The Oxford Reference Guide to Lexical Functional Grammar

The Oxford
Reference Guide
to Lexical Functional
Grammar

MARY DALRYMPLE
JOHN J. LOWE
LOUISE MYCOCK

1
 3
Great Clarendon Street, Oxford, ox2 6dp,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Mary Dalrymple, John J. Lowe, and Louise Mycock 2019
The moral rights of the authors have been asserted
First Edition published in 2019
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2018965641
ISBN 978–0–19–873330–0
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
 Preface
This book is based on work first published by Dalrymple (2001). Some chapters
from that book appear in this one, although the entire text has been overhauled
and revised. Some of the chapters in Part II of this book are entirely new. All in all,
this is not a new edition of the 2001 book, but a new book which includes parts of
the previous one.
As usual, the LFG community has been incredibly supportive of our work
in producing this book, and we are grateful to the many people who provided
comments, feedback, and support. We begin by conveying our thanks to Luke
Carr, Jamie Findlay, and Miltiadis Kokkonidis for helpful comments on Dalrymple
(2001).
We would like to single out three heroic individuals for special praise for their
dedication and thoroughness in reading and commenting on multiple versions of
the second edition as it took shape. Bozhil Hristov, Adam Przepiórkowski, and
Amanda Thomas each read through several versions of the entire book, giving
valuable comments and feedback each time. We have been particularly impressed
by Adam’s ability to detect problems from the microscopic to the macroscopic level.
We hope they will feel that the final version of the book reflects their hard work
and the helpful comments and suggestions that they made.
We are also grateful to those of our readers who commented in detail on entire
sections of the book. For detailed and helpful comments on the chapters in Part I,
we are grateful to Ash Asudeh, Alex Biswas, Ken Kahn, Leslie Lee, Joey Lovestrand,
Agnieszka Patejuk, Liselotte Snijders, and Martin Trpovski. For comments on Part
II, we are grateful to Ash Asudeh, Alex Biswas, Leslie Lee, and Joey Lovestrand.
We are also grateful to readers who provided comments on individual chapters
of the book. In Part I: Miriam Butt for Chapter 1 (Background and theoretical
assumptions); I Wayan Arka, Agnieszka Patejuk, and Ida Toivonen for Chapter 2
(Functional structure); Ida Toivonen for Chapter 3 (Constituent structure); Oleg
Belyaev and Ron Kaplan for Chapter 5 (Describing syntactic structures); and Ron
Kaplan for particularly detailed and helpful comments on Chapter 6 (Syntactic
relations and syntactic constraints). In Part II: Miriam Butt for Chapter 7 (Beyond
c-structure and f-structure: linguistic representations and relations) and Chapter
9 (Argument structure and mapping theory); Tina Bögel, Stephen Jones, and
Aditi Lahiri for Chapter 11 (Prosodic structure); and Miriam Butt, Louisa Sadler,
and Andy Spencer for Chapter 12 (The interface to morphology). In Part III:
Doug Arnold, Ash Asudeh, and Andy Morrison for Chapter 13 (Modification);
Dag Haug for Chapter 14 (Anaphora) and Chapter 15 (Functional and anaphoric
control); Doug Arnold, Oleg Belyaev, Dag Haug, John Lamping, Louisa Sadler,
and Vijay Saraswat for Chapter 16 (Coordination); Doug Arnold, Ash Asudeh,
Ron Kaplan, and Louisa Sadler for Chapter 17 (Long-distance dependencies); Ash
Asudeh, Aaron Broadwell, Kersti Börjars, Ron Kaplan, Peter Sells, Nigel Vincent,
and Andy Way for Chapter 18 (Related research threads and new directions); and
xviii preface

Liselotte Snijders for the bibliography. This book has been long in the making, and
we apologize if we have omitted reference to anyone whose comments we have
benefited from but who is not mentioned here.
For administrative support, we gratefully acknowledge the Faculty of Linguis-
tics, Philology, and Phonetics at the University of Oxford. We thank Julia Steer and
Vicki Sunter of OUP for very helpful advice, and for their patience over the long
course of the gestation of this book.
Mary Dalrymple is grateful to the following organizations for research support:
2011–14: Ministerio de Ciencia e Innovación, Gobierno de España: “The syntax
and information structure of unbounded dependencies,” principal investigator
Prof. Alexandre Alsina, Universitat Pompeu Fabra, Barcelona); 2012: British
Academy Small Grant “Plural semantics in Austronesian”; 2012–13: Leverhulme
Research Fellowship “Plurals: Morphology and Semantics,” RF-2012-295;
2017–2018: Centre for Advanced Study at the Norwegian Academy of Science
and Letters, “SynSem: From Form to Meaning—Integrating Linguistics and
Computing,” principal investigators Prof. Dag Haug and Prof. Stephan Oepen.
She is also grateful to Ken Kahn for all kinds of support, and to her sister Matty
for being an inspiration as a writer.
John Lowe is grateful to the following organizations for research support: 2012–
15: Leverhulme Early Career Fellowship (ECF-2012-081); 2013–14: Ministerio de
Ciencia e Innovación, Gobierno de España: “The syntax and information structure
of unbounded dependencies,” principal investigator Prof. Alexandre Alsina, Uni-
versitat Pompeu Fabra, Barcelona); 2016: Jill Hart Fund for Indo-Iranian Philology,
University of Oxford. He is also grateful to his family for being a source of strength
and inspiration, especially to Helen, Henry, and Wilfred.
Louise Mycock is grateful to the following for research support: 2013–14: Min-
isterio de Ciencia e Innovación, Gobierno de España: “The syntax and informa-
tion structure of unbounded dependencies,” principal investigator Prof.Alexandre
Alsina, Universitat Pompeu Fabra, Barcelona). She is also grateful to her family
for their ongoing support: her husband Chi Lun Pang, her grandparents Iris and
Brian, her dad Dave, her sisters Kathryn and Fiona and their partners, her nephew
Charlie, and her nieces Lily and Maisie.
 List of Abbreviations
We have modified the glosses from our source materials for consistency with the
Leipzig Glossing Rules (Bickel et al. 2015), a simplified version of which we use
throughout the book. We use the following abbreviations:

1 first person
2 second person
3 third person
abs absolutive
acc accusative
active active
av active voice
aux auxiliary
b Chechen gender class B
caus causative
comp complementizer
d Chechen gender class D
dv dative voice
en Catalan en
erg ergative
excl exclusive
f feminine
fut future
fv final vowel
gen genitive
hi Catalan hi
incl inclusive
indf indefinite
inf infinitive
iprf imperfect
iv instrumental voice
link Tagalog linker
loc locative
m masculine
n neuter
nfut nonfuture
xx list of abbreviations

nom nominative
nonpl nonplural
nonsg nonsingular
npst nonpast
ov objective voice
pass passive
pfv perfective
pl plural
postneg postnegation
pot potential
prep preposition
prf perfect
prs present
pst past
q question particle
refl reflexive
scv Chechen simultaneous converb
sg singular
vm Hungarian verbal modifier
wp Chechen witnessed past
 1
Background and theoretical
assumptions

Lexical Functional Grammar (LFG) is a nontransformational theory of linguistic

structure which assumes that language is best described and modeled by parallel
structures representing different facets of linguistic organization and information,
related to one another by means of functional constraints.

1.1 Historical roots

The theory had its beginnings in the 1970s, at a time of some upheaval in the theory
of generative grammar. Early transformational grammar proposed the existence of
“kernel sentences” (Chomsky 1957), basic simple declarative clauses generated by
a simple phrase structure grammar. More complex sentences were derived from
these simple sentences by various transformations: for example, passive sentences
were derived from their active counterparts by means of a passive transformation,
described in terms of properties of the phrase structures of the input and output
sentences. The influence of the transformational view persists to the present day
in the process-oriented terminology commonly used for various grammatical
phenomena: wh-movement, passivization, dative shift, and so on.
In time, however, the lack of generality of the early transformational approach
began to be seen as problematic. It was not easy to see how the very specific trans-
formations that had been proposed could capture crosslinguistic generalizations.
For example, as discussed by Perlmutter and Postal (1977), there seemed to be
no way to give a uniform statement of transformational rules across languages
with different phrase structural descriptions for obviously similar transformations
such as Passive. It became increasingly clear that the generalizations underlying
many transformational rules depend not on phrase structure configuration, but
on traditional abstract syntactic concepts such as subject, object, and complement.
If rules could be stated in terms of these abstract concepts, a crosslinguistically
uniform statement of generalizations about such rules would emerge.
At the same time, it was noted that a large class of transformations were
“structure-preserving” (Emonds 1976: 3):

A transformational operation is structure-preserving if it moves, copies, or inserts a node C

into some position where C can be otherwise generated by the grammar.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
2 background and theoretical assumptions

The existing transformational framework would not have led to the prediction that
transformations would operate in this way. Since transformations were not con-
strained as to the output structure they produced, it was surprising that they would
produce structures like those that the basic grammar could otherwise generate.
This important finding had wide-reaching implications: the basic phrase structure
of languages is invariant, and the application of particular transformations does
not alter this basic phrase structure.
Why should so many transformations have been structure-preserving in this
sense? Bresnan (1978) made the key observation: all structure-preserving trans-
formations can be reformulated as lexical redundancy rules. According to this view,
operations on the abstract syntactic argument structure of a lexical item produce
a new syntactic argument structure, with a surface form that is realized in an
expected way by a basic phrase structure grammar. This allowed an abstract and
uniform crosslinguistic characterization of argument alternations like the active-
passive relation, while also allowing for a theory of crosslinguistic similarities and
differences in the phrasal expression of the different alternations.
With this, the need emerged for a theory allowing simultaneous expression
of both the phrasal constituency of a sentence and its more abstract functional
syntactic organization. The formal insights leading to the development of Lexi-
cal Functional Grammar arose originally from the work of Woods (1970), who
explored methods for representing the surface constituent structure of a sentence
together with more abstract syntactic information. Building on this work, Kaplan
(1975a,b, 1976) realized that placing certain constraints on the representation of
abstract syntactic structure and its relation to surface phrasal structure would
lead to a simple, formally coherent, and linguistically well-motivated grammatical
architecture. Based on these formal underpinnings, the relation of the abstract
functional syntactic structure of a sentence to its phrase structure could be fully
explored. More information about the historical development of the theory is
provided by Dalrymple et al. (1995a) and Bresnan et al. (2016).

1.2 “Lexical” and “Functional”

The name of the theory, “Lexical Functional Grammar,” encodes two important
dimensions along which LFG differs from other theories. First, the theory is lexical
and not transformational: it states relations among different verbal diatheses in
the lexicon rather than by means of syntactic transformations. In 1978, when the
theory was first proposed, this was a fairly radical idea, but in the intervening years
it has come to be much more widely accepted; it is a fundamental assumption of
Categorial Grammar (Moortgat 1988; Morrill 1994; Steedman 2001) as well as of
Head-Driven Phrase Structure Grammar (Pollard and Sag 1994; Sag et al. 2003;
Levine 2017), Construction Grammar (Kay 2002; Boas and Sag 2012), Simpler
Syntax (Culicover and Jackendoff 2005), and some works in the transformational
tradition (Grimshaw 1990).
Unlike some other theories of syntax, then, the lexicon is not merely a repos-
itory for exceptions, a place in which syntactically or semantically exceptional
 structure of the book 3

information is recorded. Since LFG is a lexical theory, regularities across classes

of lexical items are part of the organization of a richly structured lexicon, and
an articulated theory of complex lexical structure is assumed. Work on lexical
issues has been an important focus of LFG from the beginning, and this research
continues with work to be described in the following pages.
The second dimension that distinguishes Lexical Functional Grammar is that it
is functional and not configurational: abstract grammatical functions like subject
and object are not defined in terms of phrase structure configurations or of seman-
tic or argument structure relations, but are primitives of the theory. LFG shares
this view with Relational Grammar (Perlmutter and Postal 1977) and Arc Pair
Grammar (Johnson and Postal 1980), as well as with Construction Grammar (Kay
2002; Boas and Sag 2012) and Simpler Syntax (Culicover and Jackendoff 2005).
LFG assumes that functional syntactic concepts like subject and object are
relevant for the analysis of every language: that the same notions of abstract
grammatical functions are at play in the structure of all languages, no matter how
dissimilar they seem on the surface. Of course, this does not imply that there are no
syntactic differences among languages, or among sentences in different languages
that have similar meanings; indeed, the study of abstract syntactic structure in
different languages is and has always been a major focus of the theory. Just as
the phrase structure of different languages obeys the same general principles
(for example, in adherence to X-bar theory; see §3.3.2), in the same way the
abstract syntactic structure of languages obeys universal principles of functional
organization and draws from a universally available set of possibilities, but may
vary from language to language. In this sense, the functional structure of language
is said to be “universal.”
In work on the theory of linking between semantic arguments and syntactic
functions, similarities and differences among grammatical functions have been
closely analyzed, and natural classes of grammatical functions have been proposed.
To analyze these similarities, grammatical functions like subject and object are
decomposed into more basic features such as +restricted, as described in §9.4.1.
On this view, grammatical functions are no longer thought of as atomic. Even
given these decompositions, however, the grammatical functions of LFG remain
theoretical primitives, in that they are not derived or defined in terms of other
linguistic notions such as agenthood or phrasal configuration.

1.3 Structure of the book

The book consists of three parts. In the first part, Chapter 2 (Functional structure),
Chapter 3 (Constituent structure), and Chapter 4 (Syntactic correspondences)
examine the two syntactic structures of LFG, the constituent structure and the
functional structure, discussing the nature of the linguistic information they rep-
resent, the formal structures used to represent them, and the relation between
the two structures. Chapter 5 (Describing syntactic structures) and Chapter 6
(Syntactic relations and syntactic constraints) outline the formal architecture of
LFG and explain how to describe and constrain the constituent structure, the
4 background and theoretical assumptions

functional structure, and the relation between them. A clear understanding of the
concepts presented in Chapter 5 is essential for the discussion in the rest of the
book. Chapter 6 is best thought of as a compendium of relatively more advanced
formal tools and relations, and may be most profitably used as a reference in
understanding the analyses presented in the rest of the book.
The second part of the book explores nonsyntactic levels of linguistic struc-
ture and the modular architecture of LFG. Chapter 7 (Beyond c-structure and
f-structure: Linguistic representations and relations) sets the scene for our explo-
ration of other linguistic levels and their relation to constituent structure and
functional structure, presenting LFG’s projection architecture and outlining how
different grammatical levels are related to one another. Chapter 8 (Meaning and
semantic composition) introduces the LFG view of the syntax-semantics interface
and semantic representation, according to which the meaning of an utterance
is determined via logical deduction from a set of premises associated with the
subparts of the utterance. Chapter 9 (Argument structure and mapping theory)
discusses the content and representation of argument structure, its relation to
syntax, and its role in determining the syntactic functions of the arguments of a
predicate. Chapter 10 (Information structure) introduces the level of information
structure, the structuring of an utterance in context, and explores the relation of
information structure to other linguistic levels. Chapter 11 (Prosodic structure)
introduces the level of prosodic structure, which analyzes the string in parallel with
constituent structure, but with respect to prosodic units rather than phrasal units.
Chapter 12 (The interface to morphology) discusses the place of morphology in
the architecture of LFG, showing how a realizational theory of morphology can be
integrated in an LFG setting.
The third part of the book illustrates the concepts of the theory more explicitly
by presenting a series of sketches of the syntax and semantics of a range of repre-
sentative linguistic phenomena. We present the syntactic aspects of the analyses
separately from the semantic aspects, so readers who are not interested in formal
semantic analysis should still be able to profit from the syntactic discussion in these
chapters. In this part, we often leave aside analysis of the information structure,
prosody, and morphology of these phenomena, though we sometimes include an
analysis of these other aspects as well, in line with the increasing awareness of the
importance of adopting a holistic approach and taking account of the interplay
of linguistic modules in a full account of the data. Chapter 13 (Modification)
discusses the syntax and semantics of modifiers, particularly concentrating on
modification of nouns by adjectives. Chapter 14 (Anaphora) presents a theory
of the syntax and semantics of anaphoric binding, including both intrasentential
and intersentential anaphora. Chapter 15 (Functional and anaphoric control)
discusses constructions involving control, where the referent of the subject of a
subordinate clause is determined by lexical or constructional factors. Chapter 16
(Coordination) presents an analysis of aspects of the syntax and semantics of coor-
dination, and Chapter 17 (Long-distance dependencies) discusses long-distance
dependencies in topicalization, relative clause formation, and question formation.
The final chapter of the book, Chapter 18 (Related research threads and new
directions), discusses LFG-based work in areas not covered elsewhere in the book,
as well as new developments in the theory of LFG, including work in historical
 how to use the book 5

linguistics and language acquisition, computational and algorithmic research in

parsing and generation, LFG-based theories of language acquisition, and Optimal-
ity Theory-based work.
The book concludes with three indexes: an index of cited authors, a language
index, and a subject index. The language index contains information about the
linguistic family to which each cited language belongs, as well as a rough charac-
terization of where the language is spoken.
This book concentrates primarily on the theory of LFG as it has developed since
its inception in the late 1970s. The analyses we present are focused on syntactic
and nonsyntactic relations and structures within the sentence; we will have far
less to say about the structure of larger units of discourse or the relations between
sentences.

1.4 How to use the book

Most of the book should be accessible to upper-level undergraduate or graduate
students who have some background in syntax. Part I is concerned solely with
syntax and its representation by LFG’s constituent structure and functional struc-
ture. In Part II, we widen the discussion to other modules of grammar, including
semantics, argument structure, information structure, prosodic structure, and the
morphological component, and their grammatical interfaces. For those whose
primary interest is in syntax, the chapters in any of these areas in Part II can be
skipped. Part III provides syntactic and semantic analyses of a range of linguistic
phenomena; it should be possible to follow the syntactic discussion with only
the background provided in Part I, but for the semantic discussions in Part III,
familiarity with the material covered in Chapter 8 of Part II will also be necessary.
The introduction to Part II provides more information about dependencies among
the chapters in Part II and Part III.
Some of the chapters in Part II and Part III will be easier to follow for readers
with some background in the areas that are discussed.

• For the semantics chapter in Part II (Chapter 8) and the semantics sections of
the chapters in Part III, Gamut (1991a,b) and Partee et al. (1993: Chapter 7)
provide useful background.
• Chapter 10 discusses information structure, its representation, and its place
in the overall LFG architecture. There is some discussion of information
structure in Chapter 17, but it should be possible to follow almost all of the
discussion in Chapter 17 even without familiarity with the material presented
in Chapter 10. For an overview and introduction to information structure, see
Lambrecht (1994) and Erteschik-Shir (2007: Chapters 1–3).
• The content and representation of prosodic structure is discussed in
Chapter 11, but does not figure in the analyses presented in Part III. For
an introduction to the concepts discussed in Chapter 11, see Selkirk (1984),
Nespor and Vogel (2007), and Ladd (2008).
• The analyses presented in Part III also do not include morphological analysis,
and so the morphology chapter in Part II (Chapter 12) can be skipped by
6 background and theoretical assumptions

those who are not concerned with morphology and its interface with the rest
of the grammar. Spencer (2004) and Haspelmath and Sims (2011) provide a
solid introduction to morphology, and Stewart (2015) provides an overview
of contemporary morphological theories. Stump (2001: Chapter 1) is an
introduction to issues in morphological theory with a focus on the word-
and-paradigm model, providing a theoretical underpinning for the family of
realizational theories which that chapter adopts.

1.5 Other LFG overviews and introductions

Bresnan (2001c), Falk (2001b), and Kroeger (2004) continue to provide invaluable
introductions to LFG from different perspectives and for different audiences. Bres-
nan (2001c) and Falk (2001b) both came out in the same year as Dalrymple (2001),
on which much of this book is based, and each provides an excellent pedagogically-
oriented introduction to the theory, including useful exercises. Kroeger (2004)
is a lucid introduction to syntactic theory from an LFG perspective, suitable for
an introductory syntax course. Bresnan et al. (2016) is a newly revised edition
of Bresnan (2001c), updating the treatments presented in the first edition and
providing detailed discussion and insights in many new areas.
Besides these book-length introductions, a number of shorter articles provide
overviews of the theory from various perspectives. Recent works include Dalrym-
ple (2006), Butt (2008), Lødrup (2011a), Börjars (2011), Nordlinger and Bresnan
(2011), Carnie (2012a), Sells (2013), Broadwell (2014), Asudeh and Toivonen
(2015), Butt and King (2015a), and Dalrymple and Findlay (2019). The on-line
proceedings of the LFG conferences (Butt and King 1996–) are also valuable
repositories of LFG research. Kuiper and Nokes (2013), Frank (2013), and Müller
(2016) provide an overview and comparison of LFG to other grammatical frame-
works, and Schwarze and de Alencar (2016) provide a computationally oriented
introduction to LFG with a focus on French.
The foundational papers in the Bresnan (1982b) collection provide a snapshot of
LFG at the earliest stages of the theory’s development. Overviews and summaries at
various subsequent stages include Sells (1985), Wescoat and Zaenen (1991), Neidle
(1994), Kaplan (1995), Kiss (1995), Neidle (1996), Sadler (1996), Butt et al. (1999),
and Austin (2001). The section introductions in Dalrymple et al. (1995b) provide
a historical perspective (from the vantage point of the mid-1990s) in a number of
areas: Formal Architecture, Nonlocal Dependencies, Word Order, Semantics and
Translation, and Mathematical and Computational Issues.
 2
Functional structure

LFG assumes two different ways of representing syntactic structure, the constituent
structure or c-structure and the functional structure or f-structure. Within the
overall system of linguistic structures, these constitute two subsystems. Functional
structure is the abstract functional syntactic organization of the sentence, famil-
iar from traditional grammatical descriptions, representing syntactic predicate-
argument structure and functional relations like subject and object. Constituent
structure is the overt, more concrete level of linear and hierarchical organization
of words into phrases.
Section 2.1 presents motivation for the categories and information appearing
in functional structure and outlines some common characteristics of functional
structure categories. Section 2.2 demonstrates that grammatical functions are best
treated as primitive concepts, as they are in LFG, rather than defined in terms
of morphological or phrase structure concepts. Section 2.3 shows that syntac-
tic subcategorization requirements, the array of syntactic arguments required by
a predicate, are best stated in functional terms. The formal representation of
functional structure and constraints on f-structure representations are discussed
in §2.4. Finally, §2.5 provides an overview of the content and representation of
functional structure features.

2.1 Grammatical functions and functional structure

Abstract grammatical relations have been studied for thousands of years. Apol-
lonius Dyscolus, a grammarian in Alexandria in the second century ad, gave a
syntactic description of Ancient Greek that characterized the relations of nouns to
verbs and other words in the sentence, providing an early characterization of tran-
sitivity and “foreshadow[ing] the distinction of subject and object” (Robins 1967).
The role of the subject and object and the relation of syntactic predication were
fully developed in the Middle Ages by the modistae, or speculative grammarians
(Robins 1967; Covington 1984).
Subsequent work also depends on assuming an underlying abstract regularity
operating crosslinguistically. Modern work on grammatical relations and syntactic
dependencies was pioneered by Tesnière (1959) and continues in the work of
Mel’čuk (1988), Hudson (1990), and others working within the dependency-
based tradition. Typological studies are also frequently driven by reference to
grammatical relations: for instance, Greenberg (1966b) states his word order uni-
versals by reference to subject and object. Thus, LFG aligns itself with approaches

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
10 functional structure

in traditional, nontransformational grammatical work, in which these abstract

relations were assumed.

2.1.1 Distinctions among grammatical functions

It is abundantly clear that there are differences in the behavior of phrases depend-
ing on their grammatical function. For example, when a question is formed in
English, a question phrase appears in initial position. This question phrase may
be related to a grammatical function in a subordinate clause, as indicated by the
subscript i in (1c) and (1d):

(1) a. Who invited Mary?

b. Who did John invite?
c. Whoi do you think [ i invited Mary]?
d. Whoi do you think [John invited i ]?

However, the subject phrase in such a question may not appear in sentence-initial
position if the subordinate clause begins with a word like that, though this is
possible for the object phrase:1

(2) a.⃰Whoi do you think [that i invited Mary]?

b. Whoi do you think [that John invited i ]?

In fact, the subject-object distinction is only one aspect of a rich set of distinc-
tions among grammatical functions. Keenan and Comrie (1977: 66) propose a
more fine-grained analysis of abstract grammatical structure, the Keenan-Comrie
Hierarchy or Accessibility Hierarchy, stating that “the positions on the Accessibility
Hierarchy are to be understood as specifying a set of possible grammatical dis-
tinctions that a language may make.” They claim that the hierarchy is relevant
for various linguistic constructions, including relative clause formation, where
it restricts the grammatical function of the argument in the relative clause that
is interpreted as coreferent with the modified noun. The border between any
two adjacent grammatical functions in the hierarchy can represent a distinction
between acceptable and unacceptable relative clauses in a language, and different
languages can set the border at different places on the hierarchy:2

(3) Keenan-Comrie Hierarchy:

subj > do > io > obl > gen > ocomp

1 Ungrammatical examples are standardly marked with an asterisk.

2 The nomenclature that Keenan and Comrie use is slightly different from that used in this book: in
their terminology, do is the direct object, which we call obj; io is the indirect object; obl is an oblique
phrase; gen is a genitive/possessor of an argument; and ocomp is an object of comparison.
 grammatical functions and functional structure 11

In some languages, the hierarchy distinguishes subjects from all other grammatical
functions: only the subject of a relative clause can be relativized, or interpreted as
coreferent with the noun modified by the relative clause. Other languages allow rel-
ativization of subjects and objects in contrast to other grammatical functions. This
more fine-grained hierarchical structure allows further functional distinctions to
emerge.
Keenan and Comrie speculate that their hierarchy plays a role in the analysis
of other grammatical constructions besides relative clause formation, and indeed
Comrie (1974) applies the hierarchy in an analysis of grammatical functions in
causative constructions. In fact, the Keenan-Comrie hierarchy closely mirrors the
“relational hierarchy” of Relational Grammar, as given by Bell (1983), upon which
much work in Relational Grammar is based:

(4) Relational Hierarchy of Relational Grammar:

1 (subject) > 2 (object) > 3 (indirect object)

The Obliqueness Hierarchy of Head-Driven Phrase Structure Grammar (Pollard

and Sag 1994; Sag et al. 2003) also reflects a hierarchy of grammatical functions
like this one. As demonstrated by a large body of work in Relational Grammar,
HPSG, LFG, and other theories, the distinctions inherent in these hierarchies
are relevant across languages with widely differing constituent structure repre-
sentations, languages that encode grammatical functions by morphological as
well as configurational means. There is a clear and well-defined similarity across
languages at this abstract level.
LFG assumes a universally available hierarchy of grammatical functions,3 with
subject at the top of the hierarchy:

(5) Functional hierarchy, Lexical Functional Grammar:

subject > object > obj𝜃 > comp, xcomp > oblique𝜃 > adjunct, xadjunct

The labels obj𝜃 and obl𝜃 represent families of relations indexed by thematic roles,
with the 𝜃 subscript representing the thematic role associated with the argument.
For instance, objtheme is the member of the group of thematically restricted obj𝜃
functions that bears the thematic role theme, and oblsource and oblgoal are
members of the obl𝜃 group of grammatical functions filling the source and goal
thematic roles. These clearly stated relations between thematic roles and grammat-
ical functions reflect a broader generalization about grammatical functions within
the LFG framework. As Bresnan et al. (2016: 9) put it, grammatical functions are
the “relators” of phrase structure positions to thematic roles. Structural expressions
of a particular grammatical function vary crosslinguistically, but a characteristic
mapping to argument structure exists in all cases.

3 The grammatical function hierarchy given in (5) lists only grammatical functions that play a
role within the clause. We discuss the grammatical function poss, the function borne by prenominal
genitives within the English noun phrase, in §2.1.10.
12 functional structure

Grammatical functions can be cross-classified in several different ways. The

governable grammatical functions subj, obj, obj𝜃 , comp, xcomp, and obl𝜃 can be
subcategorized, or required, by a predicate; these contrast with modifying adjuncts
adj and xadj, which are not subcategorizable.
The governable grammatical functions form several natural groups. First, the
terms or core arguments (subj, obj, and the family of thematically restricted objects
obj𝜃 , discussed in §2.1.6.1) can be distinguished from the family of nonterm or
oblique functions obl𝜃 . Crosslinguistically, term functions behave differently from
nonterms in constructions involving agreement and control (Chapter 15); we
discuss these and other differences between terms and nonterms in §2.1.3.
Second, subj and the primary object function obj are the semantically unre-
stricted functions, while obl𝜃 and the secondary object function obj𝜃 are restricted
to particular thematic roles, as the 𝜃 in their name indicates. Arguments with
no semantic content, like the subject it of a sentence like It rained, can fill the
semantically unrestricted functions, while this is impossible for the semantically
restricted functions. We discuss this distinction in §2.1.4.
Finally, open grammatical functions (xcomp and xadj), whose subjects are
controlled by an argument external to the function, are distinguished from closed
functions. These are discussed in §2.1.7.
Some linguists have considered inputs and outputs of relation-changing rules
like the passive construction to be good tests for grammatical functionhood: for
example, an argument is classified as an object in an active sentence if it appears as
a subject in the corresponding passive sentence, under the assumption that the pas-
sive rule turns an object into a passive subject. However, as we discuss in Chapter 9,
grammatical function alternations like the passive construction are best viewed not
in terms of transformational rules, or even in terms of lexical rules manipulating
grammatical function assignments, but in terms of a characterization of possible
mappings between thematic roles and grammatical functions. Therefore, appeal to
these processes as viable diagnostics of grammatical functions requires a thorough
understanding of the theory of argument linking, and these diagnostics must be
used with care.
In the rest of this chapter, we present the inventory of grammatical functions
assumed in LFG theory and discuss a variety of grammatical phenomena that make
reference to these functions. Some of these phenomena are sensitive to a gram-
matical hierarchy, while others can refer either to specific grammatical functions
or to a larger class of functions. Thus, the same test (for example, relativizability)
might distinguish subjects from all other grammatical functions in one language,
but might group subjects and objects in contrast to other grammatical functions
in another language. A number of tests are also specific to particular languages
or to particular types of languages: for example, switch-reference constructions,
constructions in which a verb is inflected according to whether its subject is coref-
erential with the subject of another verb, do not constitute a test for subjecthood
in a language in which switch-reference plays no grammatical role. In a theory like
LFG, grammatical functions are theoretical primitives, not defined in phrasal or
semantic terms; therefore, we do not define grammatical functions in terms of a
particular, invariant set of syntactic behaviors. Instead, grammatical phenomena
 grammatical functions and functional structure 13

can be seen to cluster and distribute according to the grammatical organization

provided by these functions.

2.1.2 Governable grammatical functions and modifiers

A major division in grammatical functions is the distinction between arguments of
a predicate and modifiers. The arguments are the governable grammatical functions
of LFG; they are subcategorized for, or governed, by the predicate. Modifiers modify
the phrase with which they appear, but they are not governed by the predicate:

(6) Governable grammatical functions:

subj obj obj𝜃 xcomp comp obl𝜃 adj xadj
! "# $ ! "# $
governable grammatical functions modifiers

A number of identifying criteria for governable grammatical functions have been

proposed. Dowty (1982) proposes two tests to distinguish between governable
grammatical functions and modifiers: what he calls the entailment test, namely that
using a predicate entails the existence of all of its arguments, but not its modifiers;
and what he calls the subcategorization test, namely that it is possible to omit
modifiers but not arguments when a predicate is used. These tests do capture some
intuitively correct properties of the distinction between governable grammatical
functions and modifiers; however, neither test is completely successful in distin-
guishing between them.
Dowty’s first test, the entailment test, fails for some phrases that seem uncontro-
versially to be modifiers. In particular, since the use of many predicates entails that
some event occurred at some place at some time, the test implies that temporal
modifiers are arguments of those predicates. For instance, the use of the verb
yawned in a sentence like David yawned entails that there was some past time
at which David yawned; however, few linguists would conclude on this basis that
previously is an argument of yawned in a sentence like David yawned previously.
Additionally, as pointed out by Anette Frank (p.c.), the entailment test incorrectly
predicts that the object argument of an intensional verb such as deny or seek is not
a governable grammatical function, since a sentence like David is seeking a solution
to the problem does not imply that a solution exists. Further, syntactically required
but semantically empty phrases that are governed by a predicate are not classified
as syntactic arguments by this test; the existence of some entity denoted by the
subject of rained is not entailed by the sentence It rained.
Dowty’s second test is also problematic. It clearly fails in “pro-drop” languages—
languages where some or all arguments of a predicate can be omitted—but even in
English the test does not work well. The test implies that because a sentence like
David ate is possible, the object lunch in David ate lunch is not an argument but a
modifier.
Even though Dowty’s tests do not succeed in correctly differentiating arguments
and modifiers, certain valid implications can be drawn from his claims. If a phrase
is an argument, it is either obligatorily present or it is entailed by the predicate. If
14 functional structure

a phrase is a modifier, it can be omitted. Stronger conclusions do not seem to be

warranted, however.
A number of other tests have been shown to illuminate the distinction between
arguments and modifiers:4

Multiple occurrence: Modifiers can be multiply specified, but arguments

cannot, as noted by Kaplan and Bresnan (1982):

(7) a. The girl handed the baby a toy on Tuesday in the morning.
b. ⃰David saw Tony George Sally.

VP anaphora: It is possible to add modifiers to a clause containing do so, but

not arguments (Lakoff and Ross 1976):

(8) a. John bought a book last Friday and I did so today.

b. ⃰John bought a book and I did so a magazine.

Alternation: Only arguments can alternate with subjects and objects; see
Needham and Toivonen (2011):

(9) a. The patient benefited from the medicine.

b. The medicine benefited the patient.

Anaphoric binding patterns: In some languages, binding patterns are sen-

sitive to the syntactic argument structure of predicates and therefore to the argu-
ment/modifier distinction. For example, as we discuss in §14.2.1, the Norwegian
reflexive pronoun seg selv requires as its antecedent a coargument of the same
predicate. Since a modifier is not an argument of the main predicate, the reflexive
seg selv may not appear in a modifier phrase if its antecedent is an argument of the
main verb (Hellan 1988; Dalrymple 1993). Coreference between the anaphor and
its intended antecedent is indicated by coindexation, here using the subscript i:

(10) Jon forakter seg selv.

Jon despises self
‘Joni despises himselfi .’

(11) Jon fortalte meg om seg selv.

Jon told me about self
‘Joni told me about himselfi .’

(12) ⃰Hun kastet meg fra seg selv.

She threw me from self
‘Shei threw me away from herselfi .’

4 These tests do not give clear-cut results in all cases. For discussion of contentious cases, see §9.10.5.
 grammatical functions and functional structure 15

In (10) and (11), the reflexive’s antecedent is the subject Jon, which is an argument
of the same verbal predicate. By contrast, when the reflexive appears as part of a
modifier phrase, as in (12), the sentence is ungrammatical.

Order dependence: The contribution of modifiers to semantic content can

depend on their relative order, as noted by Pollard and Sag (1987: §5.6). The
meaning of a sentence may change if its modifiers are reordered: example (13a) is
true if Kim jogged reluctantly, and this happened twice a day, while example (13b)
is true if Kim jogged twice a day, but did so reluctantly (but perhaps would have
been happy to jog once a day, or three times a day):

(13) a. Kim jogged reluctantly twice a day.

b. Kim jogged twice a day reluctantly.

In contrast, reordering arguments may affect the meaning of the sentence, focusing
attention on one or another argument, but does not alter the conditions under
which the sentence is true.

Extraction patterns: A long-distance dependency cannot relate a question

phrase that appears in sentence-initial position to a position inside some modifiers,
as noted by Pollard and Sag (1987: §5.6) (see also Huang 1982; Rizzi 1990):

(14) a.⃰Which famous professor did Kim climb K2 without oxygen in order to
impress ?
b. Which famous professor did Kim attempt to impress by climbing K2
without oxygen?

This generalization is not as robust as those discussed above, since as Pollard and
Sag point out, it is possible to extract a phrase from some modifiers:

(15) Which room does Julius teach his class in ?

2.1.3 Terms and nonterms

The governable grammatical functions can be divided into terms or direct func-
tions, and nonterms or obliques. The subject and object functions are grouped
together as terms:5

(16) Terms and nonterms:

subj obj obj𝜃 obl𝜃 xcomp comp
! "# $ ! "# $
terms nonterms

A number of tests for termhood in different languages have been proposed.

5 Relational grammar (Perlmutter and Postal 1983) also recognizes this basic division of grammatical
functions into “term relations” and “oblique relations.” Terms are also sometimes referred to as “core
functions” (Andrews 2007a; Bresnan et al. 2016).
16 functional structure

Agreement: In some languages, termhood is correlated with verb agreement;

in fact, this observation is encoded in Relational Grammar as the Agreement
Law (Frantz 1981): “Only nominals bearing term relations (in some stratum)
may trigger verb agreement.” Alsina (1993), citing Rosen (1990) and Rhodes
(1990), notes that all terms, and only terms, trigger verb agreement in Ojibwa and
Southern Tiwa.

Control: Kroeger (1993) shows that in Tagalog, only a term can be the con-
trollee in the participial complement construction, and only a term can be a con-
troller in the participial adjunct construction. The sentences in (17) are examples
of the participial complement construction. The controllee can be a term (17a,b),
but never a nonterm (17c):

(17) a. In-abut-an ko siya=ng [nagbabasa

TERM
pfv-find-dv 1sg.gen 3sg.nom=comp av.ipfv.read
ng=komiks sa=eskwela].
gen=comics dat=school
‘I caught him reading a comic book in school.’
b. In-iwan-an ko siya=ng [sinususulat
TERM
pfv-leave-dv 1sg.gen 3sg.nom=comp ipfv.write.ov
ang=liham].
nom=letter
‘I left him writing the letter.’
c.⃰In-abut-an ko si=Luz na [ibinigay ni=Juan
pfv-find-dv 1sg.gen nom=Luz link iv.ipfv.give gen=Juan
ang=pera ].
OBLGOAL
nom=money
‘I caught Luz being given money by Juan.’

Alsina (1993) provides an extensive discussion of termhood in a number of

typologically very different languages, and Andrews (2007a) further discusses the
term/nonterm distinction.
Often, discussion of terms focuses exclusively on the status of nominal argu-
ments of a predicate, and does not bear on the status of verbal or sentential
arguments. The infinitive phrase to be yawning in example (18) bears the open
grammatical function xcomp:

(18) Chris seems to be yawning.

The clausal complement that Chris was yawning bears the grammatical function
comp in (19):

(19) David thought that Chris was yawning.

 grammatical functions and functional structure 17

The xcomp function differs from the comp function in not containing an overt
subj internal to its phrase; xcomp is an open function, whose subj is determined
by means of lexical specifications on the predicate that governs it, as discussed in
§2.1.7. What is the termhood status of the xcomp and comp arguments?
Zaenen and Engdahl (1994) classify xcomp as a kind of oblique in their
analysis of the linking of sentential and predicative complements, though without
providing explicit evidence in support of this classification. Oblique arguments are
nonterms, and so if Zaenen and Engdahl are correct, xcomp would be classified as
a nonterm.
Word order requirements on infinitival and finite complements in English
provide some support for this position. Sag (1987) claims that in English, term
phrases always precede obliques:

(20) a. David gave a book to Chris.

b.⃰David gave to Chris a book.

Infinitival and clausal complements bearing the grammatical functions xcomp

and comp obey different word order restrictions from term noun phrases. The
following data indicate that xcomps are obliques:

(21) a. Kim appeared to Sandy to be unhappy.

b. Kim appeared to be unhappy to Sandy.

Since the xcomp to be unhappy is not required to precede the oblique phrase to
Sandy but can appear either before or after it, Sag’s diagnostic indicates that the
xcomp must also be an oblique. Similar data indicate that comp is also an oblique
phrase:

(22) a. David complained that it was going to rain to Chris.

b. David complained to Chris that it was going to rain.

We will return to a discussion of comp and xcomp in §2.1.7.

2.1.4 Semantically restricted and unrestricted functions

The governable grammatical functions can be divided into semantically restricted
and semantically unrestricted functions (Bresnan 1982a):

(23) Semantically unrestricted and restricted functions:

subj obj obj𝜃 obl𝜃
! "# $ ! "# $
semantically unrestricted semantically restricted

Semantically unrestricted functions like subj and obj can be associated with any
thematic role, as Fillmore (1968) shows:
18 functional structure

(24) a. He hit the ball.

b. He received a gift.
c. He received a blow.
d. He loves her.
e. He has black hair.

The examples in (24) show that the subj of different verbs can be associated with
different thematic roles: agent in a sentence like He hit the ball, goal in a sentence
like He received a gift, and so on. Similar examples can be constructed for obj.
In contrast, members of the semantically restricted family of functions obj𝜃 and
obl𝜃 are associated with a particular thematic role. For example, the objtheme
function is associated only with the thematic role of theme, and the oblgoal
is associated with goal. Languages may differ in the inventory of semantically
restricted functions they allow. For example, English allows only objtheme :

(25) a. I gave her a book .

theme
b. I made her a cake .
theme
c. I asked him a question.
theme

Other thematic roles cannot be associated with the second object position, under-
lined in these examples:

(26) a.⃰I made a cake the teacher.

goal
b.⃰I cut the cake a knife.
instrument

Section 2.1.6 provides a more complete discussion of object functions, including

the double object construction and verb alternations; see also Levin (1993).
The division between semantically restricted and semantically unrestricted
arguments predicts what in Relational Grammar is called the Nuclear Dummy
Law (Frantz 1981; Perlmutter and Postal 1983): only semantically unrestricted
functions can be filled with semantically empty arguments like the subject it
of It rained. This is because the semantically restricted functions are associated
with a particular thematic role; since a semantically empty argument is
incompatible with these semantic requirements, it cannot fill a semantically
restricted role.
The functions xcomp and comp have rarely figured in discussions of semanti-
cally restricted and unrestricted arguments, and the question of how they should
be classified remains under discussion. There does seem to be some evidence
 grammatical functions and functional structure 19

for classifying comp as semantically unrestricted, since different semantic entail-

ments can attach to different uses of xcomp and comp, as shown in a pio-
neering paper by Kiparsky and Kiparsky (1970). If these different entailment
patterns are taken to indicate that xcomp and comp can be associated with
different thematic roles, then xcomp and comp should be classified as semantically
unrestricted.
Kiparsky and Kiparsky (1970) note that clausal complements bearing the comp
function may be factive or nonfactive: for factive complements, “the embed-
ded clause expresses a true proposition, and makes some assertion about that
proposition,” whereas such a presupposition is not associated with nonfactive
complements. Kiparsky and Kiparsky also distinguish emotive from nonemotive
sentential arguments; emotive complements are those to which a speaker expresses
a “subjective, emotional, or evaluative reaction”:

(27) a. Factive emotive: I am pleased that David came.

b. Factive nonemotive: I forgot that David came.
c. Nonfactive emotive: I prefer that David come.
d. Nonfactive nonemotive: I suppose that David came.

It is not clear, however, whether the semantic differences explored by Kiparsky

and Kiparsky should be taken to indicate that these arguments, which all bear the
grammatical function comp in English, bear different thematic roles.
Zaenen and Engdahl (1994) take a different position, arguing that xcomp—and
therefore, by extension, comp—is restricted in that only propositional arguments
can fill these two functions, a constraint which they claim is directly comparable to
being semantically restricted (see also Falk 2005). In their analysis of complement
clauses, Dalrymple and Lødrup (2000) also assume that comp is semantically
restricted, a feature which distinguishes comp from obj and which is central to
their proposal; see §2.1.7.
Falk (2001b: 137–41) presents an alternative view, one which does not involve
the semantically restricted/unrestricted distinction. He assumes that comp and
xcomp, as propositional arguments, are distinguished from all other grammatical
functions by having a positive value for a special feature [c]. Falk (2001b: 133)
characterizes this as a “more cautious position” than that of Zaenen and Engdahl
(1994), with which it shares some foundational assumptions. Berman’s (2007)
analysis of German represents yet another approach: she claims that comp is an
underspecified complement function, which may be either restricted or unre-
stricted depending on the context. This underspecified analysis offers a different
view of comp and the possible reasons why it has proved so difficult to categorize
with respect to the semantically restricted/unrestricted distinction.

We have explored several natural classes of grammatical functions: governable

grammatical functions and modifiers, terms and nonterms, semantically restricted
and unrestricted functions. We now turn to an examination of particular grammat-
ical functions, beginning with the subject function.
20 functional structure

2.1.5 SUBJ
The subject is the term argument that ranks the highest on the Keenan-Comrie
relativization hierarchy. As discussed in §2.1.1, Keenan and Comrie’s hierarchy is
applicable to a number of processes including relativization: if only one type of
argument can participate in certain processes for which a grammatical function
hierarchy is relevant, that argument is often the subject.
There is no lack of tests referring specifically to the subject function.

Agreement: The subject is often the argument that controls verb agreement in
languages in which verbs bear agreement morphology; indeed, Moravcsik (1978)
proposes the following language universal:

There is no language which includes sentences where the verb agrees with a constituent
distinct from the intransitive subject and which would not also include sentences where the
verb agrees with the intransitive subject. (Moravcsik 1978: 364)

English is a language that exhibits subject-verb agreement; the fullest paradigm is

found in the verb to be:

(28) I am / You are / He is

Honorification: Matsumoto (1996) calls this the most reliable subject test in
Japanese. Certain honorific forms of verbs are used to honor the referent of the
subject:

(29) sensei wa hon o o-yomi ni

teacher topic book acc honorific-read copula
narimashi-ta
become.polite-pst
‘The teacher read a book.’

The verb form o-V ni narimasu is used to honor the subject sensei ‘teacher’. It can-
not be used to honor a nonsubject, even if the argument is a “logical subject”/agent.
Like example (29), example (30) is grammatical because the honorific verb form is
used to honor the subject sensei ‘teacher’:

(30) sensei wa Jon ni o-tasuke-rare ni

teacher topic John by honorific-help-pass copula
narimashi-ta
become.polite-pst
‘The teacher was saved by John.’

A proper name is used only when the referent is not superior to the speaker in
age or social rank: example (31) is unacceptable because Jon is not an appropriate
target of honorification, and sensei is not the subject:
 grammatical functions and functional structure 21

(31) ⃰Jon wa sensei ni o-tasuke-rare ni

John topic teacher by honorific-help-pass copula
narimashi-ta
become.polite-pst
‘John was saved by the teacher.’

Subject noncoreference: Mohanan (1994) shows that the antecedent of a

pronoun in Hindi cannot be a subject in the same clause, although a nonsubject
antecedent is possible. In example (32), the possessive pronoun uskii cannot corefer
with Vijay (as indicated by the asterisk on the subscript i), though coreference with
Ravii (j) or another individual (k) is possible:

(32) Vijay ne Ravii ko uskii saikil par bit.haayaa

Vijay erg Ravi acc his bicycle loc sit.caus.prf
‘Vijayi seated Ravij on his i,j,k bike.’
⃰
Launching floated quantifiers: Kroeger (1993: 22) shows that the subject
launches floating quantifiers, quantifiers that appear outside the phrase they quan-
tify over, in Tagalog:6

(33) sinusulat lahat ng-mga-bata ang-mga-liham

iprf.write.ov all gen-pl-child nom-pl-letter
‘All the letters are written by the/some children.’
(Does not mean: ‘All the children are writing letters.’)

Bell (1983: 154 ff.) shows that the same is true in Cebuano.
This is only a sampling of the various tests for subjecthood. Many other tests could,
of course, be cited (see, for example, Andrews 2007a; Falk 2006; Li 1976; Zaenen
1982; Zaenen et al. 1985).
The question of whether every clause has a subj is controversial. The Subject
Condition7 was discussed by Bresnan and Kanerva (1989), who attribute it origi-
nally to Baker (1983) (see also Levin 1987; Butt et al. 1999):

(34) Subject Condition:

Every verbal predicate must have a subj.

Though the Subject Condition seems to hold in English, and perhaps in many
other languages as well, there are languages in which the requirement does not so
clearly hold. For example, German impersonal passives, as in (35), are traditionally
analyzed as subjectless clauses:

6 Kroeger attributes example (33) to Schachter (1976).

7 The Subject Condition is called the Final 1 Law in Relational Grammar (Frantz 1981; Perlmutter
and Postal 1983) and the Extended Projection Principle in Chomskyan generative grammar (Chomsky
1981).
22 functional structure

(35) . . . weil getanzt wird

because danced become.prs.3sg
‘because there is dancing’

However, Berman (1999, 2003) claims that clauses like (35) contain an unpro-
nounced expletive subject and thus that the Subject Condition is not violated. She
argues that in a German impersonal passive construction the verbal agreement
morphology (specifically the 3sg form) introduces a subject that has only person
and number agreement features: that is, an expletive subject without semantic
content. In this way, every finite clause in German satisfies the Subject Condition.
Kibort (2006) discusses Polish constructions which, similar to the German
construction in (35), appear to be subjectless. These clauses contain one of a small
number of inherently impersonal predicates, such as słychać ‘hear’, as in (36a), or
an intransitive predicate which has been passivized, as in (36b), which is similar to
the German example in (35). In (36a), the lack of a syntactic subject is reflected
in the form of the verb used: the verb słychać ‘hear’ has the same form as the
infinitive. Since this is a nonfinite verb form, there is no agreement morphology
to introduce subject features, in contrast to the German construction analyzed
by Berman. (36b) contains the passivized form of an intransitive predicate, było
sypane ‘has been spreading’, which bears default impersonal agreement (3sg.n).
The agent can optionally appear as an oblique argument:

(36) a. Słychać ja˛ / jakieś mruczenie.

hear her.acc some.n.acc murmuring.(n).acc
‘One can hear her/some murmuring.’
b. Czy na tej ulicy już było sypane
q on this street already was.3sg.n throw/spread.ptcp.sg.n
(przez kogokolwiek)?
by anyone
‘Has there already been spreading [of grit] on this street (by anyone)?’

On the basis of such data, Kibort (2006) argues that truly subjectless constructions
exist in Polish. It may be, then, that the Subject Condition is a language-particular
requirement imposed by some but not all languages, rather than a universal
requirement.
An alternative, more fine-grained view of subject properties and the notion of
subjecthood is explored in detail by Falk (2006). The core of Falk’s proposal is
that the grammatical function subject represents the conflation of two distinct
functions: the element that expresses the most prominent argument of the verb
(gf)
% and the element that connects its clause to other clauses in a sentence (pivot).
Prominence in the case of gf % is defined in terms of the mapping between the
thematic hierarchy (comprising thematic roles such as agent, theme, location) and
the relational hierarchy (comprising grammatical functions). Falk (2006: 39) thus
defines gf
% as being “the element with the function of expressing as a core argument
the hierarchically most prominent argument,” and argues that a subset of “subject
 grammatical functions and functional structure 23

properties”—those relating to argument hierarchies, such as the antecedent of a

reflexive or the addressee of an imperative—are in fact gf % properties, whereas
those “subject properties” relating to elements shared between clauses are pivot
properties.
In languages like English, with nominative-accusative alignment, gf % and pivot
always coincide, and the argument which we have identified as subj exhibits
both gf% and pivot properties. However, Falk claims that other alignments are
possible, and that in what he calls “mixed-subject” languages, pivot is not neces-
sarily identified with gf.
% Mixed subject languages include those which have been
described as syntactically ergative: in a syntactically ergative language, pivot is the
object if there is one, and otherwise pivot is gf.
% For example, in the syntactically
ergative language Greenlandic, gf % plays an important role in anaphoric binding:
the antecedent of a reflexive is gf,
% not pivot. This is shown in (37), where the
antecedent of the reflexive pronoun is the single argument of the intransitive verb
ingerlavoq ‘go’ or the more agentive argument of the transitive verb tatig(i-v)aa
‘trust’:8

(37) a. Aani illu-mi-nut ingerlavoq.

Anne house-refl.poss-dat go
‘Annei is going to heri house.’
b. Ataata-ni Juuna-p tatig(i-v)aa.
father-refl.poss Juuna-erg trust
‘Juunai trusts hisi father.’

However, pivot and not gf % is the target of relativization in relative clause forma-
tion: the relativized argument is the single argument of the intransitive predicate
‘angry’ or the object of ‘kill’, but not the gf
% of ‘take’.

(38) a. miiraq [kamattuq]

child angry.sg
‘the child [who is angry]’
b. nanuq [Piita-p tuqua]
polar.bear Peter-erg kill.3sg
‘a polar bear [that Peter killed ]’
c. ⃰angut [aallaat tigusimasaa]
man gun take.3sg
‘the man [who took the gun]’

So-called “Philippine-type” languages are also mixed languages: in a Philippine-

type language such as Tagalog, lexical marking on the verb determines which of
the verb’s arguments is identified with pivot. In such languages, gf % and pivot
properties diverge in the same way as in syntactically ergative languages.

8 Falk attributes the examples in (37) and (38) to Manning (1996b). We present them in simplified
form, leaving out morphological detail.
24 functional structure

Falk (2006) proposes that gf% plays a role in all languages, in the sense that
all languages have clauses including gf% as an argument. However, he holds that
the requirement for all clauses to include gf
% as an argument holds only for some
languages, similar to Kibort’s (2006) claims in relation to the Subject Condition.
In particular, Falk (2006: 170) proposes that certain intransitive Acehnese verbs
(those with agentive subjects, such as “go”) require a gf, % while certain others
(those with nonagentive subjects, such as “fall”) require an object, not gf.
% pivot
has a different status: Falk tentatively proposes that some languages, including
Choctaw/Chickasaw and Warlpiri, lack a pivot altogether. Such languages would
be expected to lack pivot-sensitive constructions such as long-distance dependen-
cies and functional control constructions.
In order to maintain consistency with the vast majority of previous work in the
LFG framework and because the distinction between gf % and pivot is not crucial
to the analyses we present, we will continue to use subj. However, this does not
mean that we reject the important distinctions involving different aspects of the
subj role that Falk explores.

2.1.6 The object functions

Grammatical phenomena in which a functional hierarchy is operative may some-
times group subject and object arguments together in distinction to other argu-
ments, and in fact a number of grammatical processes refer to subject and object
functions in distinction to other grammatical functions. Other phenomena are
describable specifically in terms of the object function; for the purposes of our
current discussion, these object tests are more interesting. Some of these are:

Agreement: As noted in §2.1.3, terms are often registered by agreement mor-

phemes on the verb. Often, the object is uniquely identified by agreement. For
example, Georgopoulos (1985) describes obj agreement in Palauan:

(39) ak-uldenges-terir a resensei er ngak

1sg.prf-honor-3pl teachers prep me
‘I respected my teachers.’

In (39), the morpheme -terir encodes third person plural agreement with the obj
a resensei ‘teachers’.

Casemarking: In some limited circumstances, objects can be distinguished

by casemarking, though this test must be used with care: in general, there is
no one-to-one relation between the morphological case that an argument bears
and its grammatical function, as we will see in §2.2.1. Mohanan (1982) discusses
casemarking in Malayalam, showing that accusatively marked noun phrases are
unambiguously objects (see also Mohanan 1994: 89–90):

(40) kut..ti aanaye nul..li

child elephant.acc ¯pinched
‘The child pinched the elephant.’
 grammatical functions and functional structure 25

However, Mohanan goes on to show that many phrases in Malayalam that are obj
are not marked with acc case. That is, every phrase in Malayalam that is acc is an
obj, but not all objs are acc.

Relativization: Givón (1997: §4.4.3) notes that only subjects and objects can
be relativized in Kinyarwanda, and only objects can be relativized with a gap;
relativization of subjects requires the use of a resumptive pronoun.

Further discussion of object tests is provided by Baker (1983) for Italian,

Dahlstrom (1986b) for Cree, and Hudson (1992) for English. Andrews (2007a)
also provides a detailed discussion of object tests in various languages.

2.1.6.1 Multiple objects In many languages it is possible to have more than one
phrase bearing an object function in a sentence. English is one such language:

(41) He gave her a book.

Zaenen et al. (1985) discuss Icelandic, another language with multiple object
functions, and note the existence of asymmetries between the two kinds of objects.
For instance, the primary object can be the antecedent of a reflexive contained in
the secondary object:

(42) Ég gaf ambáttina [konungi sínum].

I gave slave.def.acc king.dat self ’s
‘I gave the slavei to self ’si king.’
obj obj2

However, the secondary object cannot antecede a reflexive contained in the pri-
mary object:

(43) ⃰Sjórinn svipti manninum [gömlu konuna sína].

sea.def deprived man.def.dat old wife.def.acc self ’s
‘The sea deprived self ’si old wife of the mani .’
obj obj2

Dryer (1986) also presents an extensive discussion of the behavior of objects

in languages with multiple obj functions and their groupings with respect to
thematic roles.
Earlier work in LFG concentrated on languages like English and Icelandic, which
each have two object functions. In such languages, the primary object was called
the obj, and the secondary object was called the obj2, as in examples (42)–(43).
Further research has expanded our knowledge of the properties of objects, and in
later work it became evident that this simple classification is neither sufficient nor
explanatory.
In fact, languages allow a single thematically unrestricted object, the primary obj.
In addition, languages may allow one or more secondary, thematically restricted
objects. That is, the argument that was originally identified as obj2 in English
26 functional structure

is only one member of a family of semantically restricted functions, referred to

collectively as obj𝜃 (Bresnan and Kanerva 1989). This classification more clearly
reflects the status of secondary objects as restricted to particular thematic roles,
and also encompasses analyses of languages whose functional inventory includes
more than two object functions.
In English, as discussed in §2.1.4, the thematically restricted object must be a
theme; other thematic roles, such as goal or beneficiary, are not allowed:

(44) a. I made her a cake.

b. ⃰I made a cake her.

In contrast, as Bresnan and Moshi (1990) show, languages like Kichaga allow
multiple thematically restricted objects with roles other than theme:9

(45) n-a̋-l! é-kú-shí-kí-kór.-í-à

focus-1subj-pst-17obj-8obj-7obj-cook-appl-fv
‘She/he cooked it with them there.’

This example contains three object markers, representing a locative object (17obj),
an instrumental object (8obj), and a patient object (7obj). According to Bresnan
and Moshi’s analysis, in this example the instrumental obj is the unrestricted
obj; the locative and patient arguments bear thematically restricted obj functions
objloc and objpatient . Bresnan and Moshi provide much more discussion of obj𝜃
in Kichaga and other Bantu languages.
While the division of the object functions into thematically unrestricted obj and
a family of thematically restricted object functions (obj𝜃 ) represents the main-
stream approach in LFG, the notion of object and its possible subtypes remains
relatively understudied, as Börjars and Vincent (2008) point out. In a paper which
considers object as a pretheoretical construct as well as the concept obj within LFG,
Börjars and Vincent (2008) argue that the object role is not associated with specific
semantic content but should rather be viewed as a “semantically inert” grammatical
function.

2.1.6.2 “Direct” and “indirect” object In traditional grammatical descriptions,

the grammatical function borne by her in the English example in (46) has some-
times been called the “indirect object,” and a book has been called the “direct
object”:

(46) He gave her a book.

The phrase a book is also traditionally assumed to be a direct object in examples

like (47):

(47) He gave a book to her.

9 Numbers in the glosses indicate the noun class of the arguments.

 grammatical functions and functional structure 27

The classification of a book as a direct object in both (46) and (47) may have a
semantic rather than a syntactic basis: there may be a tendency to assume that
a book must bear the same grammatical function in each instance because its
thematic role does not change. The LFG view differs: in example (46), the phrase
her bears the obj function, while in example (47), the phrase a book is the obj.
Within the transformational tradition, evidence for the LFG classification of
English objects came from certain formulations of the rule of passivization, which
applies uniformly to “transform” an object into a subject:

(48) a. Active: He gave her a book.

Passive: She was given a book.
b. Active: He gave a book to her.
Passive: A book was given to her.

If the “passive transformation” is stated in terms of the indirect object/object

distinction, or its equivalent in phrase structure terms, the generalization is com-
plicated to state: the direct object becomes the passive subject only if there is no
indirect object present; otherwise, the indirect object becomes the subject. On the
other hand, the transformation is easy to state if the first noun phrase following
the verb is classified as the object and the second bears some other function.
In LFG, however, the theory of grammatical function alternations is formulated
in terms of a characterization of possible mappings between thematic roles and
grammatical functions, as described in Chapter 9, and is not transformational in
nature. Thus, we must look to other grammatical phenomena for evidence bearing
on the classification of object functions.
Dryer (1986) presents several arguments that in English, an opposition between
primary/unrestricted objects (obj) and secondary/restricted objects (obj𝜃 ), as
proposed in LFG, allows a more satisfactory explanation of the facts than the
direct/indirect object distinction. Dryer primarily discusses evidence from prepo-
sitional casemarking and word order. For example, given a distinction between
primary and secondary objects, we can succinctly describe word order within the
English VP: the primary object immediately follows the verb, with the secondary
object following it.10
In other languages, the situation is even clearer. Alsina (1993) examines the
object functions in Chicheŵa and their role in the applicative construction. In this
construction, an affix is added to the verb to signal a requirement for an additional
syntactic argument besides the arguments ordinarily required by the verb; we
focus here on the benefactive applicative construction, in which the applicative
affix signals that an obj argument bearing a beneficiary thematic role is required.
Alsina (1993) shows that the syntactic obj properties of the patient argument in
the nonapplied form are displayed by the beneficiary argument in the applied form.
The primary/nonrestricted obj is the argument that immediately follows the verb;

10 Dryer assumes a more complicated crosslinguistic typology of object functions than is generally
accepted in LFG. His richer typology turns out to be best explained in terms of different strategies for
relating thematic roles to object grammatical functions, as described in Chapter 9.
28 functional structure

this argument is the patient in the nonapplied form, and the beneficiary in the
applied form of the verb:

(49) a. nkhandwe zi-ku-mény-á njōvu

10.foxes 10subj-prs-hit-fv 9.elephant
‘The foxes are hitting the elephant.’
b. nkhandwe zi-ku-mény-ér-a aná njōvu
10.foxes 10subj-prs-hit-appl-fv 2.children 9.elephant
‘The foxes are hitting the elephant for the children.’

The patient argument alternates with the obj marker in the nonapplied form, and
the beneficiary argument alternates with the obj marker in the applied form:

(50) a. nkhandwe zi-ku-í-me̋ny-a

10.foxes 10subj-prs-9obj-hit-fv
‘The foxes are hitting it [the elephant].’
b. nkhandwe zi-ku-wá-mény-er-á njōvu
10.foxes 10subj-prs-2obj-hit-appl-fv 9.elephant
‘The foxes are hitting the elephant for them [the children].’

This and other evidence is best explained by assuming that the patient argument
in (49a) and (50a) and the beneficiary argument in (49b) and (50b) each bear the
nonrestricted/primary obj function, while the patient argument in (49b) and (50b)
bears the restricted/secondary obj𝜃 function and behaves differently. In other
words, the syntactic behavior of the arguments in examples (49) and (50) is best
explained by reference to a distinction between obj and obj𝜃 , not between direct
and indirect objects.
Although obj𝜃 can be regarded as a secondary object function, this does not
mean that obj𝜃 appears only when an obj argument is also required. Drawing on
patterns involving passivization, causativization, and raising, Çetinoğlu and Butt
(2008) present data which support an analysis of certain Turkish transitive verbs
as subcategorizing for a subj and an obj𝜃 . Similarly, Dahlstrom (2009) proposes
that certain verb stems in the Algonquian language Meskwaki subcategorize for a
subj and an obj𝜃 . Dahlstrom shows that the obj𝜃 argument of these verbs does
not share properties with obj; for instance, it cannot undergo lexical processes
such as antipassivization (which suppresses an obj argument), and it does not
trigger agreement on the verb. In both Turkish and Meskwaki, the verbs involved
are lexically specified as taking an obj𝜃 argument. As Dahlstrom (2009) points
out, these data provide evidence in favor of retaining obj and obj𝜃 as distinct
grammatical functions, contrary to the possibility raised by Börjars and Vincent
(2008) of collapsing the two. Dalrymple and Nikolaeva (2011) also rely on the
obj/obj𝜃 distinction in their work on differential object marking and information
structure. Based on data from a range of languages, Dalrymple and Nikolaeva
(2011) claim that in certain languages a monotransitive predicate can take either
a topical obj argument or a nontopical obj𝜃 ; that is, the argument is either obj
or obj𝜃 depending on the argument’s information structure role.11 Without the

11 Chapter 10 discusses information structure and the distinction between topics and non-topics.
 grammatical functions and functional structure 29

distinction between obj and obj𝜃 , this generalization would be lost. Maintaining
the distinction between these two grammatical functions therefore appears to be
justified.

2.1.7 COMP, XCOMP, and XADJ

The comp, xcomp, and xadj functions are clausal functions, differing in whether
they contain an overt subj noun phrase. The comp function is a closed function
containing an internal subj phrase. The xcomp and xadj functions are open
functions that do not contain an internal subject phrase; their subj must be
specified externally to their phrase:12

(51) Open and closed functions:

subj obj obj𝜃 comp obl𝜃 adj xcomp xadj
! "# $ ! "# $
closed open

The comp function is the function of clausal complements, familiar from tra-
ditional grammatical description. A comp can be declarative, interrogative, or
exclamatory (Quirk et al. 1985):

(52) a. David complained that Chris yawned.

b. David wondered who yawned.
c. David couldn’t believe how big the house was.

The xcomp function is an open complement function, the one borne by a phrase
like to yawn in the examples in (53). In those examples, the subj of the xcomp is
also an argument of the matrix verb; in both of the examples in (53), it is David:

(53) a. David seemed to yawn.

b. Chris expected David to yawn.

Like xcomp, the xadj function is an open function, whose subj must be specified
externally; unlike xcomp, xadj is a modifier, not a governable grammatical func-
tion. In example (54), the subj of the xadj stretching his arms is also the subj of
the matrix clause, David:

(54) Stretching his arms, David yawned.

We will return to a discussion of xcomp, xadj, and control in Chapter 15.

12 Arka and Simpson (1998, 2008) propose that some control constructions in Balinese involve an
open subj function xsubj: for instance, in the Balinese equivalent of I tried to take the medicine, the
infinitive phrase to take the medicine can bear the subj function, with its subj controlled by the term
argument I. Falk (2005) presents an alternative analysis of the relevant data based on his proposal that,
as well as xcomp, two other governable grammatical functions exist: xobj𝜃 and xobl𝜃 . We do not
explore either of these alternatives further here.
30 functional structure

There is ongoing debate concerning the status of the grammatical function

comp. Discussion centers on whether the comp function can be discarded; propo-
nents of this position argue that sentential complements can be analyzed instead
as bearing the same types of grammatical function as nominal complements, and
that therefore a specific function for clausal complements is not motivated. Three
stances can be found in the literature: all clausal complements have the grammat-
ical function comp (Kaplan and Zaenen 1989a; Bresnan et al. 2016); clausal com-
plements have the same grammatical functions as their nominal or adpositional
counterparts, and comp should be eliminated from the inventory (Alsina et al.
1996, 2005; Berman 2003; Forst 2006); and comp should be retained in order to
account for the differences in the syntactic behavior of some clausal complements
in some languages (Dalrymple and Lødrup 2000; Berman 2007). In seeking to
reanalyze some or all instances of comp, the second and third approaches must
show that clausal complements analyzed as objects or obliques exhibit syntactic
properties characteristic of nominal objects or adpositional obliques.
Dalrymple and Lødrup (2000) propose to reclassify certain comps as obj, but
argue that comp should not be eliminated from the inventory of grammatical
functions. They observe that there are phenomena that can only be explained by
assuming the existence of the grammatical function comp as distinct from obj.
First, if all sentential complements are obj and not comp, they would be classified
as terms. In this case, the evidence presented in §2.1.3, indicating that English has
clausal complements that are not terms, would remain unexplained. Second, if
English clausal complements are analyzed as objects, then we must assume that
English admits sentences with three obj functions, but only when one of the obj
functions is sentential rather than nominal:

(55) David bet [Chris] [five dollars] [that she would win].

Most importantly, there is evidence for a split in the syntactic behavior of clausal
complements in a number of languages; Dalrymple and Lødrup (2000) propose
that this evidence is best explained by analyzing some clausal complements in these
“mixed languages” as obj, and some as comp.
For example, in Swedish, the clausal complement of a verb such as antar
‘assumes’ bears the obj function; it can be pronominalized and can topicalize, as
shown in examples (56a–c) from Engdahl (1999):

(56) a. Man antar att sossarna vinner valet.

one assumes that social.democrats.def win election.def
‘One assumes that the Social Democrats will win the election.’
b. Man antar det.
one assumes that
‘One assumes that.’
c. [Att sossarna vinner valet] antar man.
that social.democrats.def win election.def assumes one
‘That the Social Democrats will win the election, one assumes.’
 grammatical functions and functional structure 31

In contrast, the complement clause of a Swedish verb such as yrkade ‘insisted’ bears
the comp function, and does not display these properties:

(57) a. Kassören yrkade att avgiften skulle höjas.

cashier.def insisted that tax.def should be.increased
‘The cashier insisted that the tax should be increased.’
b. ⃰Kassören yrkade det.
cashier.def insisted that
‘The cashier insisted it.’
c. ⃰[Att avgiften skulle höjas] yrkade kassören.
That tax.def should be.increased insisted cashier.def
‘That the tax should be increased, the cashier insisted.’

As Dalrymple and Lødrup (2000) show, other languages also show a similar split in
the behavioral properties of clausal complements, with some clausal complements
patterning with nominal obj arguments and others exhibiting behavior typical of
comp arguments. Thus, they argue, the comp grammatical function cannot be
eliminated from grammatical description, since many clausal complements must
be analyzed as bearing the comp function.
By contrast, Alsina et al. (2005) present data from Spanish, Catalan, and Malay-
alam in support of discarding comp entirely and instead reinterpreting clausal
complements as having one of the grammatical functions obj, obj𝜃 , or obl𝜃 .13
According to this approach, any difference between non-clausal and clausal com-
plements bearing the same grammatical function follows solely from their differ-
ence in phrase structure category. In support of this claim, Alsina et al. (2005)
argue that there are syntactic differences among clausal complements that are
not accounted for on Dalrymple and Lødrup’s view. For instance, in Catalan it is
necessary to distinguish two types of clausal complements which alternate with
different pronominal clitics, en or hi, as shown in examples (58) and (59):

(58) a. M’ heu de convèncer que torni a casa.

me have.2pl to convince that return.1sg to home
‘You have to convince me to return home.’
b. Me n’ heu de convèncer.
me en have.2pl to convince
‘You have to convince me of that.’

(59) a. Estàvem d’ acord que ens apugessin el sou.

were.1pl of agreement that us raised.3pl the salary
‘We agreed that they should raise our salary.’
b. Hi estàvem d’ acord.
hi were.1pl of agreement
‘We agreed on that.’

13 In their analysis of control phenomena in Balinese, Arka and Simpson (2008) make a similar
proposal in relation to xcomp.
32 functional structure

An analysis relying on the grammatical function comp would fail to capture

the similarities between these predicates when they take clausal complements
(without a preposition) and when they select a particular preposition with a
nominal complement (de or en), as shown in (60). Instead, each verb would have
to be associated with two different subcategorization frames: one for comp clausal
complements, and the other for obl𝜃 prepositional phrases:

(60) a. M’ heu de convèncer de les seves possibilitats.

me have.2pl to convince of the 3gen possibilities
‘You have to convince me of his possibilities.’
b. Estàvem d’ acord en alguns punts.
were.1pl of agreement on some points
‘We agreed on certain points.’

Alsina et al. (2005) claim that the shared properties of the complements of such
verbs are best captured by an analysis which treats them all as having the same
grammatical function, obl𝜃 . The result is a significant simplification, as each
verb has a single subcategorization frame and redundancy is reduced: convèncer
‘convince’ subcategorizes for sentential or prepositional oblde , and estar d’acord
‘agree’ subcategorizes for sentential or prepositional oblen .
The status of comp is still poorly understood. Dalrymple and Lødrup (2000)
provide data which indicate that comp and obl𝜃 behave differently in a number of
respects in German, and cannot be collapsed: comp clauses cannot be topicalized,
while obl𝜃 phrases can; comp clauses cannot appear in the middle field, while
obl𝜃 phrases can; and comp phrases cannot be coordinated with obl𝜃 phrases.
In contrast, Forst (2006) discusses data which indicate that German complement
clauses are best analyzed as bearing one of the grammatical functions obj, obl𝜃 , or
obj𝜃 , and that the comp grammatical function is not needed in German. Lødrup
(2012) provides a new argument for the relevance of comp, arguing that the comp
grammatical function is needed not only in the analysis of clausal complements,
but also for certain nominal complements in Norwegian: that is, that there are
nominal as well as clausal comps. Until convincing arguments can be made that all
comps in languages such as English, German, and Norwegian can be reanalyzed
in terms of other grammatical functions, comp cannot be abandoned on the basis
of being redundant.

2.1.8 PREDLINK
predlink is a closed grammatical function originally introduced by Butt et al.
(1999) to analyze predicative constructions that include a linking or copular verb,
such as those in (61):

(61) a. Fiona is a nurse.

b. The book is red.
c. The vase is on the table.
d. The problem is that they yawned.
 grammatical functions and functional structure 33

According to this analysis, the copular predicate in these examples selects for a subj
and a predlink. predlink has also been employed in the analysis of some verbless
sentences that express similar meanings to the examples in (61) (see Attia 2008).
The predlink analysis can be contrasted with the traditional approach, which
assumes that the copular predicate selects xcomp, an open argument whose subj is
specified externally to the phrase. The xcomp analysis is problematic in those cases
where the constituent identified as the xcomp does not otherwise appear to have an
internal subject phrase, or has a different subject from that of the copular predicate.
This is the case with noun phrases (61a), adjective phrases (61b), prepositional
phrases (61c), and especially with clauses (61d). As Butt et al. (1999: 70) point
out, if the xcomp analysis is to be maintained for all predicative constructions, it
is necessary to assume that a phrase belonging to one of the relevant categories,
such as a nurse in (61a), has two sets of subcategorization requirements depending
on whether it is being used predicatively, and thus selects a subj argument, as in
(61a), or not. Under a predlink analysis, no such issue arises because predlink is
a closed function—all of its relevant grammatical dependents are included; none
are externally specified.
However, when it can be argued that the subjects of both the copula and the non-
verbal complement are the same, for instance when the post-copular complement
exhibits agreement with the subject of the main predicate as in the French examples
in (62), an xcomp analysis has been argued to capture the relation between the two
in a more straightforward manner:14

(62) a. Elle est petite.

she.f.sg is small.f.sg
‘She is small.’
b. Il est petit.
he.m.sg is small.m.sg
‘He is small.’

The question of whether a predlink or an xcomp analysis is most appropriate for

copular constructions, both within a particular language and crosslinguistically,
has been the focus of attention in subsequent work. Those who propose a unified
predlink analysis, and thus seek to capture the functional equivalence of copular-
type constructions at the level of functional structure, include Butt et al. (1999),
Attia (2008), and Sulger (2011). In this book, we adopt a programmatic approach
that aims to capture the ways in which such constructions are diverse, following
Falk (2004), Nordlinger and Sadler (2007), and Laczkó (2012). Further discussion
of copular constructions is provided in §5.4.5.

2.1.9 Oblique arguments: OBL𝜃

Oblique arguments are those that are associated with particular thematic roles
and marked to indicate their function overtly. In languages like English, oblique

14 See Dalrymple et al. (2004a) and Attia (2008) for a different perspective.
34 functional structure

arguments are prepositional phrases, while in other languages, as discussed by

Nordlinger (1998), oblique arguments are casemarked rather than appearing as
prepositional or postpositional phrases.
LFG assumes that there are two types of oblique arguments (Bresnan 1982a).
Arguments of the first type are marked according to the thematic role of the
argument, such as the goal to-phrase of a verb such as give. This class corresponds
to the category of semantic case in the casemarking classification scheme of Butt
and King (2004c), since semantic case is governed by generalizations about the
relation between case and thematic role.
Arguments of the second type are marked idiosyncratically, and their form
is lexically specified by the governing predicate. This class corresponds to the
category of quirky case in Butt and King’s classification scheme; see §2.5.7.5.

2.1.9.1 Semantically marked obliques The phrase to Chris in example (63)

bears the oblgoal grammatical function:

(63) David gave the book to Chris.

The thematic role of the oblgoal argument is marked by the preposition to. It is
not possible for more than one oblique argument to have the same thematic role:

(64) ⃰David gave the book to Chris to Ken.

In languages like Warlpiri, an oblloc phrase such as kirri-ngka ‘large camp’

is marked with locative casemarking rather than a preposition or postposition
(Simpson 1991; Nordlinger 1998):

(65) kirri-ngka wiri-ngka-rlipa nyina-ja

large.camp-loc big-loc-1pl.incl.subj sit-pst
‘We sat in the large camp.’

Locative casemarking plays a similar role to the preposition in example (64), to

mark the thematic role of the argument.

2.1.9.2 Idiosyncratic prepositional marking An oblique phrase may also be

required to bear a particular form unrelated to the thematic role of the argument.
For such cases, Bresnan (1982a) suggests the presence of a form feature that is
specified by the predicate. For example, in (66) the form of the preposition on in the
phrase on David is stipulated by the predicate relied, which requires a semantically
restricted nonterm argument: oblon .

(66) Chris relied on/ ⃰to/ ⃰about David.

Butt et al. (1999) provide more discussion of oblique phrases with idiosyncratic
prepositional marking.
 grammatical functions and functional structure 35

2.1.10 POSS
We follow Bresnan et al. (2016) and many others in analyzing the genitive phrase
which appears in the prenominal position in an English nominal phrase as bearing
the grammatical function poss:15

(67) Chris’s book

The grammatical function poss is the most prominent grammatical function

associated with nominals, just as the subj function is the most prominent clausal
function. Similarities between poss and subj have often been noted, particularly
in the realization of arguments of finite clauses and their corresponding gerunds
and derived nominals (Chomsky 1970):

(68) a. John criticized the book.

b. John’s criticizing the book
c. John’s criticism of the book

Like the subj function, arguments with the poss function can bear any of a number
of semantic roles, including possessor (as in example 67), agent (as in examples
68b,c), and other roles. poss arguments often share other syntactic characteristics
of subjects: for example, poss and subj often behave similarly with respect to
anaphoric binding constraints. In §2.1.5, we saw that an identifying characteristic
of subjects in Hindi is their inability to serve as the antecedent of a pronoun in the
same clause. This was shown in example (32), repeated here:

(69) Vijay ne Ravii ko uskii saikil par bit.haayaa

Vijay erg Ravi acc his bicycle loc sit.caus.prf
‘Vijayi seated Ravij on his i,j,k bike.’
⃰
As observed by Ghulam Raza (p.c.), the same constraints hold of poss in the noun
phrase; in (70), coreference between the poss pronoun uskii and the poss Vijay
is disallowed, though coreference with the non-poss argument Ravii or another
individual is allowed:

(70) Vijay kaa Ravii ko uskii saikil kaa tohfaa

Vijay gen Ravi dat his bicycle gen gift
‘Vijayi ’s gift to Raviij of his i,j,k bike’
⃰
Sulger (2015) provides additional evidence showing that possessors exhibit
subject-like properties in Hindi-Urdu. For example, Mohanan (1994) shows that

15 Dalrymple (2001) analyzes prenominal genitives as spec rather than poss. We do not adopt that
analysis here, though we will make use of the feature spec as a feature of nominal f-structures: see §2.5.7
for a discussion of the functional features of nominal phrases, including def and spec.
36 functional structure

the antecedent of the Hindi possessive reflexive apnaa/apnii must be a logical or

grammatical subject, and cannot be a nonsubject:

(71) a. Ravii apnii saikil par bait.haa

Ravi self.gen bicycle loc sit.prf
‘Ravii sat on self ’si,⃰j bike.’
b. Vijay ne Ravii ko apnii saikil par bit.haayaa
Vijay erg Ravi acc self.gen bicycle loc sit.caus.prf
‘Vijayi seated Ravij on self ’si,⃰j,⃰k bike.’

As Sulger shows, possessors are also able to bind reflexives within the nominal
phrase:

(72) a. meraa apnaa makaan

my self.gen house
‘myi owni house’
b. sadr kaa ilekšan karaane kaa apnaa elaan
president gen election do.caus gen self.gen announcement
‘the announcement made by the presidenti himselfi to conduct elections’

Indeed, Chisarik and Payne (2003) and Sulger (2015) analyze possessor arguments
as subj rather than poss, arguing that this provides a more adequate and general
picture of nominal and clausal syntactic relations. Laczkó (1995) discusses similar-
ities between subj and poss in Hungarian, though he does not argue for collapsing
the distinction between the two; see Falk (2002a) for additional arguments in favor
of maintaining a distinction between subj and poss.
Bresnan et al. (2016: 242) propose that poss is a governable grammatical
function for nominals. On this view, most or all nouns are ambiguous, either
requiring (73a) or disallowing (73b) a poss phrase:

(73) a. Chris’s book

b. the book

Bresnan et al. (2016: 315–16) treat poss as a derived argument, in the sense of
Needham and Toivonen (2011); for discussion of derived arguments, see §9.10.5.2.
Though we are sympathetic with this view, we do not treat nouns as ambiguous,
and we do not represent the poss argument as one of the governable grammatical
functions of a noun.
Some languages have more than one way of expressing the possession relation
within the noun phrase. For example, the possessor in English can be expressed
either by a prenominal genitive or by a postnominal of-phrase; these are sometimes
referred to as “short” and “long” possessive phrases (Falk 2002a), or as “genitive”
and “of-oblique” possessors (Chisarik and Payne 2003):

(74) a. the linguist’s house

b. the house of the linguist
 grammatical functions and functional structure 37

In English, it is the short-form, prenominal genitive phrase, the phrase in (74a),

which bears the poss function and shares syntactic properties with subj. The
long-form postnominal phrase shares syntactic properties with clausal oblique
phrases, and we treat such phrases as a kind of obl.16 Other languages may differ
from English in the syntactic expression of possessors: see Laczkó (1995, 2004)
and Chisarik and Payne (2003) for a discussion of the syntax of possession in
Hungarian, Sadler (2000) for Welsh, Falk (2001a, 2002a) for Hebrew, and Sulger
(2015) for Hindi-Urdu.

2.1.11 Overlay functions

Each of the functions that we have discussed so far specifies the syntactic relation
of an argument to a local predicate. However, as Falk (2001b: 59) reminds us,
clauses and the syntactic elements of which they are comprised do not exist in
isolation; they are a part of larger syntactic and discourse structures to which they
may be linked in a variety of ways. Falk refers to the set of secondary functions
which a syntactic element may have as overlay functions (following Johnson and
Postal 1980); in contrast, the primary functions already introduced are nonoverlay
functions. An overlay function must be integrated into the structure of a clause via
its relation to a primary, nonoverlay function, hence the term. The one overlay
function that we assume in this book is dis, which can be thought of as standing
for dislocation or long distance dependency.17 We use this overlay function in the
analysis of long-distance dependencies and instances of dislocation.
An unbounded or long-distance dependency is the relationship that exists
between a filler and a gap in, for example, the constituent questions in (75). In each
of these sentences, the question word what appears in sentence-initial position
and is the filler, while the gap to which it is related has a grammatical function
in either the same clause (75a) or a subordinate clause (75b). The gap corresponds
to a primary argument or adjunct grammatical function. In these examples, the
relevant nonoverlay function is obj because the question word refers to the item(s)
purchased. The filler to which a gap is related has the overlay function dis:

(75) a. What did Charlie buy ?

b. What do you think [Charlie bought ]?

Details of the analysis of long-distance dependencies can be found in Chapter 17.

16 Chisarik and Payne (2003) present arguments in favor of analyzing these as a new grammatical
function ncomp; we do not adopt this analysis here.
17 In much LFG work, including Bresnan and Mchombo (1987), Dalrymple (2001), Falk (2001b),
and Bresnan et al. (2016), the syntacticized discourse functions topic and focus are used as overlay
functions. However, this approach has been rejected as inadequate in other work on displaced con-
stituents, notably by Alsina (2008), who proposes an overlay feature op (“operator”) instead of discourse
functions like topic and focus, and Asudeh (2004, 2011, 2012), who proposes an overlay feature
udf (“unbounded dependency function”). We use the feature dis rather than op or udf: the term
“unbounded dependency” generally refers to the relation between a filler and a gap, as in example (75)
but not example (76), and the term “operator” is widely used in semantics but with a different (though
related) meaning.
38 functional structure

A dislocation construction involves an anaphoric dependency between the

dislocated constituent and a primary, nonoverlay function. The sentence in (76)
exemplifies left dislocation in English:

(76) a. Beansi , Charlie likes themi .

b. Beansi , David thinks [Charlie likes themi ].

In both (76a) and (76b), there is an anaphoric dependency between the left-
dislocated constituent Beans, with the overlay function dis, and the pronoun them
with the primary function obj. Details of the analysis of anaphoric dependencies
can be found in Chapter 14.

Some LFG work (Falk 2001b, for instance) has also treated subj as being an
overlay function, the only overlay function which is also an argument. However,
this work predates Falk’s major work on subjecthood (Falk 2006), discussed in
§2.1.5, in which he proposes that subj be abandoned and replaced with two sepa-
rate functions: gf
% and pivot. gf % is a primary (nonoverlay) governable grammatical
function, while pivot is a secondary overlay function. This approach has the
advantage of eliminating from the original set of grammatical functions the only
overlay function that is an argument, subj. We use subj rather than gf % or pivot in
this book (see §2.1.5), but we follow Falk (2006) in classifying subj as a primary
nonoverlay grammatical function.
Falk (2006: 74) proposes that pivot is an overlay function which relates elements
which are “shared” between clauses, defining pivot as “the element with the
function of connecting its clause to other clauses in the sentence.” Like the overlay
function dis, a syntactic element with the function pivot must be identified with a
primary function; see Falk (2006) for details. To give an example, in Dyirbal (Dixon
1994: 15) pivot is invariably identified with either the intransitive subject or the
transitive object; in (77), the pivot obj of the verb bura ‘see’ in the first clause is
identified with the pivot gf% of the verb banaga ‘return’ in the second clause:

(77) Ny urra ŋana-na bura-n banaga-ny u

2pl.nom 1pl.acc see-nfut return-nfut
‘You saw us and (we) returned.’

Falk (2006) speculates that not all languages make use of the function pivot; as
noted in §2.1.5, Choctaw/Chickasaw and Warlpiri seem to be pivotless languages,
lacking pivot-sensitive constructions such as long-distance dependencies and
functional control constructions.
As mentioned in §2.1.5, we set aside pivot in this book. The only overlay
function we employ is dis.

2.2 The autonomy of functional organization

LFG does not assume that abstract grammatical functions are defined in terms
of their phrase structural position in the sentence or in terms of morphological
 the autonomy of functional organization 39

properties like casemarking; instead, grammatical functions are primitive concepts

of the theory. However, there is also clear evidence for structure at other levels: for
example, there is abundant evidence for morphological and phrasal organization
and structure. Given this, one might conclude that constituent structure is the
only structure with a firm linguistic basis, and that the appearance of abstract
grammatical functions is actually only an illusion. On this view, the generalizations
assumed by traditional grammarians are actually derivative of phrasal organization
and structure. We will see in the following that this view is misguided: attempts to
define functional structure in terms of morphological or phrase structure concepts
do not succeed.

2.2.1 Grammatical functions defined?: Casemarking

It is clear that arguments of predicates have certain superficial morphological
properties, and it is equally clear that it is not possible to provide a simple definition
of grammatical functions in terms of these properties. A cursory look at languages
with complex casemarking systems is enough to show that the relation between
case and grammatical function is not at all straightforward.
Examples given in §2.1.6 of this chapter show that it is possible to demonstrate
a correlation between grammatical function and casemarking in Malayalam: if
an argument is acc, it is an object. However, the overall picture is much more
complex, and Mohanan (1982) argues convincingly against defining grammatical
functions in terms of superficial properties like case. Objects in Malayalam are
marked acc if animate and nom if inanimate:

(78) a. Nominative subject and object:

kut..ti waatil at.accu
¯
child.nom door.nom closed
‘The child closed the door.’
b. Nominative subject, accusative object:
kut..ti aanaye kan..tu
child.nom elephant.acc saw
‘The child saw the elephant.’

In Malayalam, then, there is clearly no one-to-one relation between casemarking

and grammatical function, since a grammatical function like obj is not always
marked with the same case.
Similarly, arguments that can be shown to bear the subj function in Icelandic
are marked with a variety of cases, as shown by Andrews (1982). These cases also
appear on arguments filling nonsubject grammatical functions; for instance, as
examples (79a) and (79b) show, acc case can appear on both subjects and objects,
and examples (79c) and (79d) show that subjects can bear other cases as well:

(79) a. Accusative subject:

Hana dreymdi um hafið.
She.acc dreamed about sea
‘She dreamed about the sea.’
40 functional structure

b. Accusative object:
Stúlkan kyssti drengina.
girl.nom kissed boys.acc
‘The girl kissed the boys.’
c. Dative subject:
Bátnum hvolfdi.
boat.dat capsized
‘The boat capsized.’
d. Genitive subject:
Verkjanna gætir ekki.
pains.gen is.noticeable not
‘The pains are not noticeable.’

In sum, the relation between grammatical function and case is complex. Even when
there is a close relation between case and grammatical function, as discussed in
§2.1.6, a clear and explanatory description of casemarking and other morphosyn-
tactic properties is best obtained by reference to abstract functional properties.

2.2.2 Grammatical functions defined?: Constituent structure

Another visible, easily testable property of languages is their surface phrase struc-
ture. Given the necessity for this structure, one might claim that grammatical func-
tions are not universally manifest, but instead that the appearance of grammatical
functions in a language like English is due to the fact that grammatical functions
are associated with certain phrasal configurations in English syntax: in a nutshell,
English has subjects and objects because English is configurational. This claim
entails that languages that are not configurational would not be expected to exhibit
the same abstract functional relations.18
Warlpiri is a language whose phrasal syntactic structure is completely different
from that of languages like English. Warlpiri (like many Australian languages) is
known for displaying “nonconfigurational” properties, including free word order
and “discontinuous phrases.” The following Warlpiri sentences involve permuta-
tions of the same words; they are all grammatical and have more or less the same
meaning (Hale 1983: 6):

(80) a. ngarrka-ngku ka wawirri panti-rni

man-erg aux kangaroo spear-npst
‘The man is spearing the kangaroo.’
b. wawirri ka panti-rni ngarrka-ngku
kangaroo aux spear-npst man-erg
c. panti-rni ka ngarrka-ngku wawirri
spear-npst aux man-erg kangaroo

18 For in-depth discussion of configurationality in an LFG setting, see Snijders (2015).

 the autonomy of functional organization 41

It would be difficult to find a language less like English in its phrase structure
configurational properties. Thus, Warlpiri would seem to be a good candidate to
test the hypothesis that evidence for grammatical functions can be found only in
English-like configurational languages.
However, as Hale (1983) demonstrates, languages like Warlpiri do show evi-
dence of abstract grammatical relations, just as English-like configurational lan-
guages do. Hale discusses person marking, control, and interpretation of reflexive/
reciprocal constructions, showing that constraints on these constructions are not
statable in terms of surface configurational properties. Simpson and Bresnan
(1983) and Simpson (1991) provide further evidence that properties like control
in Warlpiri are best stated in terms of abstract functional syntactic relations. In
particular, Simpson and Bresnan (1983) examine the karra-construction, in which
the subject of a subordinate clause with affix karra is controlled by the subject of
the matrix clause:

(81) ngarrka ka wirnpirli-mi [karli jarnti-rninja-karra]

man.abs aux whistle-npst boomerang.abs trim-inf-comp
‘The mani is whistling while (hei is) trimming a boomerang.’

As Simpson and Bresnan show, the controller subject may be discontinuous or

absent, and it may be marked with nom, abs, or erg case. The correct generaliza-
tion about this construction involves the abstract grammatical function subj of the
controller, not its case or its surface configurational properties.
Thus, even in a language that appears to have completely different phrase
structure properties from English, and which has been analyzed as “nonconfig-
urational,” evidence for abstract functional syntactic relations is still found. The
hypothesis that functional structure is epiphenomenal of surface configurational
properties is not viable.

2.2.3 Grammatical functions defined?: Semantic composition

Dowty (1982) proposes to define grammatical functions like subj and obj in
compositional semantic terms, by reference to order of combination of a predicate
with its arguments: a predicate must combine with its arguments according to a
functional obliqueness hierarchy, with the subj defined as the last argument to
combine with the predicate. This approach is also adopted by Gazdar et al. (1985)
for Generalized Phrase Structure Grammar, and has to some extent carried over to
Head-Driven Phrase Structure Grammar (Pollard and Sag 1994; Sag et al. 2003).
There are several ways in which an approach like Dowty’s, where grammatical
functions are defined as an ordering on the arguments of a predicate, might lead to
incorrect predictions. First, if the order of semantic composition is very closely tied
to the order of composition of the surface configurational structure, this approach
would predict that the subject could not intervene between the verb and the object;
of course, this prediction is not correct, since many languages exhibit VSO word
order, with the subject immediately following the verb and preceding the object.
The theory that Dowty and most other adherents of this position advocate does
42 functional structure

not suffer from this difficulty, however, since the hypothesized relation between the
surface order of arguments in a sentence and the order of semantic composition is
more complex.
A more subtle problem does arise, however. It is not clear that such an approach
can make certain distinctions that are necessary for syntactic analysis: in particular,
it does not seem possible to distinguish between predicates that take the same
number of arguments with the same phrasal categories. For example, any two-
argument verb that requires a nominal subject and a clausal complement should
behave like any other such verb. However, it has been claimed that there are
languages in which some clausal complements bear the obj function, while others
bear the comp function, as discussed in §2.1.7. In a theory like LFG, this distinction
is reflected in a difference in the grammatical function of the complement; some
complements are obj, and others are comp. However, it is not clear how such
a distinction can be made in a theory that does not allow explicit reference to
grammatical functions.
Dowty (1982: 107) also argues against theories which, like LFG, assume that
grammatical functions are undefined primitives by claiming that in his approach
“grammatical relations play an important role in the way syntax relates to composi-
tional semantics.” This statement is a non sequitur. In LFG, grammatical functions
are primitive concepts and also play an important role in compositional semantics
and the syntax-semantics interface. Indeed, this is the topic of Chapter 8 and the
chapters in Part III (see also Bresnan 1982a: 286).
Leaving aside these difficulties, there is a strong degree of similarity between
theories that define grammatical functions in terms of abstract properties such as
order of semantic combination and theories like LFG, in which grammatical func-
tions are not definable in terms of phrasal or argument structure. For both types of
theories, grammatical functions are abstract and are analyzed independently from
phrasal and other structures.

2.3 Subcategorization
At a minimum, the idiosyncratic information that must be specified for a word is
its pronunciation and its meaning. Research has shown that the syntactic behavior
of a word can be partially predicted from its meaning; this is because a number
of regularities govern the relation between the meaning of a predicate and the
grammatical functions of its arguments, as we discuss in detail in Chapter 9. LFG
and other linguistic theories define and capitalize on this relation in their theory
of syntactic subcategorization.
LFG assumes that syntactic subcategorization requirements of predicates are
stated at the f-structure level, in functional rather than phrasal terms. Predicates
require a set of arguments bearing particular semantic roles. These roles are asso-
ciated with grammatical functions according to a theory of argument mapping,
to be discussed in Chapter 9. In turn, these grammatical functions are realized
at the level of constituent structure in a variety of ways, as required by particular
languages: in some languages, grammatical functions are associated with particular
 subcategorization 43

phrase structure positions, while in other languages, grammatical functions may

be signaled by particular kinds of morphological marking on the head or on the
argument (see §5.4).
In contrast to this view, and in line with some proposals in transformational
grammar (Chomsky 1965), some linguistic theories state subcategorization
requirements in terms of phrase structure positions rather than abstract functional
syntactic organization: in such theories, for example, a transitive verb in English is
defined as a verb that must be followed by a noun phrase object. There are many
reasons to question the viability of this position, since the bulk of phrase structure
information is never relevant to the satisfaction of subcategorization requirements.
As Grimshaw (1982b) points out, predicates never vary idiosyncratically in terms
of which phrasal position they require their arguments to be in; for example,
there are no exceptional transitive verbs in English which require their objects
to appear preverbally rather than postverbally. Subcategorization according to
constituent structure configuration rather than functional structure leads to the
incorrect expectation that such exceptional verbs should exist. In fact, however, we
can cleanly state subcategorization requirements in terms of abstract functional
structure; the claim that all phrasal and configurational information is always
relevant to subcategorization is too strong.
There is evidence that one particular type of constituent structure information
may in some cases be relevant to subcategorization requirements: cases in which
a predicate idiosyncratically requires an argument of a particular phrasal category.
Other kinds of phrasal information, such as position, never play a role in subcat-
egorization requirements. However, one must take care in identifying situations
in which such requirements seem to hold. Often, as Maling (1983) demonstrates,
apparent evidence for subcategorization for a particular phrase structure category
turns out on closer examination to be better analyzed as a requirement for an
argument of a particular semantic type, together with a strong correlation between
that type and the particular phrasal category most often used to express it. Maling
notes that predicates like seem have often been claimed to require adjective phrase
complements and to disallow prepositional phrase complements:

(82) a. Sandy seems clever.

b. ⃰Sandy seems out of town.

However, Maling shows that the true criterion at work in these examples is
not based on phrase structure category, but is semantic in nature: only gradable
predicates, those that can hold to a greater or lesser degree, are acceptable as
complements of seem (see also Bresnan et al. 2016: 295–301, and references cited
there). Many prepositional phrases do not express gradable predicates, accounting
for the unacceptability of example (82b). However, prepositional phrases that
denote gradable predicates are acceptable as complements of seem:

(83) a. That suggestion seemed completely off the wall.

b. Lee sure seems under the weather.
44 functional structure

Further, as Maling shows, adjective phrases that are not gradable predicates are
unacceptable as complements of seem. In the following examples, the adjective
irrational as a description of a mental state is gradable and can be used as the
complement of seems, while as a technical mathematical term it is not gradable
and cannot be used:

(84) a. Lee seems irrational.

b. ⃰The square root of two seems irrational.

In some cases, then, requirements that appear to depend on phrase structure

category prove on closer inspection to be functional or semantic in nature.
In other cases, however, the particular constituent structure category of the
complement is at issue, and no functional or semantic distinction is involved.
The circumstances under which these extra specifications are necessary are rare:
subcategorization for a particular phrasal category is a marked exception rather
than the general rule. In §6.10.3, we discuss these cases, showing that the phrase
structure category of a complement can be specified in these limited cases.

2.4 Functional structure representation

In LFG, functional information is formally represented by the functional structure
or f-structure. Mathematically, the f-structure can be thought of as a function19
from attributes to values, or equivalently as a set of pairs, where the first member of
the pair is an attribute and the second is its value.20 There is a simple and common
way of representing f-structures in tabular form, that is, as a table of attributes and
values:21

19 A function is a special kind of relation which assigns a unique value to its argument. For example,
the relation between a person and his or her age is a function, since every person has exactly one age.
The relation between a person and his or her children is not a function, since some people have no
children and some people have more than one child.
20 We use the term feature as a synonym of attribute, so that feature-value pair and attribute-value
pair mean the same thing. Bresnan et al. (2016) use the term feature in a way that differs from ours: for
them, a feature is an attribute-value pair in which the value is simple, such as [def +] or [vtype fin].
21 In some literature, particularly in Head-Driven Phrase Structure Grammar (see, for example,
Pollard and Sag 1994), the objects that are represented in LFG as structures like (85) are instead
represented via diagrams such as:
attribute1
• value1
•
• value2
attribute2

Attributes are labeled arcs in the diagram, and values are nodes. A sequence of attributes, a path through
the f-structure, corresponds to the traversal of several labeled arcs. A possible source of confusion
for those trained within the Head-Driven Phrase Structure Grammar framework is that the same
formal notation used to represent LFG functional structures in examples like (85) is used to represent
constraints on structures in HPSG. What is depicted in (85) is not a constraint; it is a formal object.
 functional structure representation 45
& '
(85) attribute1 value1
attribute2 value2

2.4.1 Simple f-structures

The following is a simplified f-structure for the noun phrase the man:

(86) & '

pred ‘man’
def +

This f-structure does not contain all the syntactic information that the man
contributes. We assume here and elsewhere that the full f-structure representation
for the examples we exhibit contains at least the information shown, but may also
contain other information not relevant to the particular point under discussion.
The f-structure in (86) contains two attributes: pred and def. The value of def
is +, indicating a positive value for the definiteness feature def. This is a simple
atomic value, with no internal structure.
For the sentence The man yawned, we have the following f-structure:

(87)  
pred ‘yawn!subj"’
 tense pst 
 '
y & 
 pred ‘man’ 
subj m
def +

As (87) shows, f-structures can themselves be values of attributes: here, the value of
the attribute subj is the f-structure for the subject of the sentence. We can annotate
f-structures with labels for subsequent reference; in (87), we have annotated the
subj f-structure with the label m and the f-structure for the sentence with the label
y. These labels are not a part of the f-structure, but provide a convenient way of
referring to an f-structure and its parts.

2.4.2 Semantic forms

The value of the pred attribute is special: it is a semantic form. A full discussion of
semantic forms will be presented in §5.2.2; additionally, Chapter 8 presents a more
complete discussion of the information that semantic forms represent. In example
(87), the semantic form value of the pred for the f-structure labeled m is ‘man’, and
the value of the pred attribute of y is ‘yawn!subj"’. The single quotes surrounding
a semantic form indicate uniqueness: each instance of use of the word man gives
rise to a uniquely instantiated occurrence of the semantic form ‘man’.
We use English names for semantic forms throughout. For example, we provide
the semantic form ‘man’ for the Warlpiri noun wati ‘man’. This is done for ease of
readability and to emphasize the distinction between the semantic form associated
with a word and its surface realization; uniform use of Warlpiri names instead of
46 functional structure

English ones for semantic forms would be equally satisfactory, though generally
less clear for an English-speaking audience.
The list of grammatical functions mentioned in a semantic form is called the
argument list. We discuss its role in determining wellformedness constraints on
f-structures in §2.4.6. In (87), the argument list for the verbal predicate yawn
indicates that it subcategorizes for a subject only.
It is not just verbs that have arguments, of course. A preposition which con-
tributes semantic content has its own pred value which includes an argument list.
It forms a separate embedded f-structure with its argument, as shown in (88):

(88) David fell from the chair.

 
’
 
 
 
 
  
 

 ’ 
  
  
  

+

Compare this with a preposition such as on in (89), whose presence is required to

mark the oblique argument of relied, but which makes no semantic contribution
and does not contribute a pred:

(89) David relied on Chris.

 
’
 
 
 
 
 

The f-structure representation of PPs is discussed further in §6.2.

2.4.3 Attributes with common values

It is possible for two different attributes of the same f-structure to have the same
value. When the value is an atom like + or m rather than an f-structure, we simply
repeat the value each time:
& '
(90) attribute1 v
attribute2 v

It is also possible for two different attributes to have the same f-structure as their
value. Here the situation is slightly more complex. Recall that an f-structure is a
 functional structure representation 47

set of pairs of attributes and values: f-structures, like other sets, obey the Axiom of
Extension, which states that two sets are the same if and only if they have the same
members (Partee et al. 1993: §8.5.8). Thus, two f-structures are indistinguishable
if they contain the same attribute-value pairs.22
Notationally, it is in some cases clearer to represent two identical f-structures
separately, repeating the same f-structure as the value of the two attributes:
 & '
(91) a1 v1
 attribute1 a2 v2 
 
 & '
 a1 v1 
 attribute2 
a2 v2

Care must be taken if a semantic form, the value of the attribute pred, is repeated.
Since each instance of a semantic form is unique, a repeated semantic form must
be explicitly marked with an index to indicate identity; see §5.2.2.1 for more
discussion of this point. If no such index appears, the two semantic forms are
assumed to be different.
In other cases, it may be easier and more perspicuous not to repeat the f-
structure, but to use other notational means to indicate that the same f-structure
appears as the value of two different attributes. We can accomplish this by drawing
a line from one occurrence to another, a common practice in LFG literature; this
notation conveys exactly the same information as in (91):
 & '
(92) a1 v1
 attribute1 a2 v2 
 
 
attribute2

This convention is notationally equivalent to another common way of representing

the same structure, using numbers enclosed in boxes:

(93)

22 This view of f-structures is different from the view of similar structures in Head-Driven Phrase
Structure Grammar (Pollard and Sag 1994; Sag et al. 2003; Levine 2017); the attribute-value structures
of HPSG are graphs, not set-theoretic objects. On the HPSG view, two attribute-value structures can
contain the same attributes and values and can nevertheless be different structures. To state the same
point in a different way: HPSG relies on a type-token distinction in attribute-value structures (Shieber
1986), meaning that two attribute-value structures are of the same type if they have the same set of
attributes and values, but may be different tokens of that type. In the set-theoretic view of LFG, the
Axiom of Extension precludes a type-token distinction, so two f-structures that have the same attributes
and values are not distinguished.
48 functional structure

There is no substantive difference between these two conventions; following LFG

tradition, we generally represent identical values for two attributes by drawing a
line connecting the two values, as in (92).

2.4.4 Sets
Sets are also valid structures, and may appear as values of attributes. Sets are
often used to represent structures with an unbounded number of elements.23 For
instance, there is in principle no limit to the number of modifiers that can appear
with any phrase, and so the value of the adj attribute is the set of all modifiers that
appear. Formally, sets are enclosed in curly brackets, with the elements of the set
inside the brackets:

(94) David yawned quietly.

 
’
 
 
 
 
 
 

In (94), only a single modifier appears, but other sentences may contain more
modification:

(95) David yawned quietly yesterday.

 
’
 
 
 
 
  
 
 
 

 
 
   
 

Any valid structure can be an element of a set: for example, some sets can have
atomic values as their elements. As we discuss in §2.5.7, sets of atomic values have
been used to represent the values of the pers, gend, and case attributes in analyses
of feature resolution and feature indeterminacy.

23 The f-structure remains a function from attributes to values even if the value of an attribute is a
set. A structure like (a), in which an attribute has as its value a set with three members, is a function,
while a structure like (b), in which the same attribute has three different values simultaneously, is not a
function.
5 67 87 87 8 9:
(a) Function: attribute1 a1 v1 a2 v2 a3 v3
 7 8
attribute1 a1 v1
 7 8
 a2 v2 
(b) Not a function:  attribute1 
 7 8
attribute1 a3 v3
 functional structure representation 49

2.4.5 Sets with additional properties

Since there is no limit to the number of coordinated elements in a coordinate
structure, we also use sets in their representation, with the coordinated elements
(the conjuncts) as members of the set. Thus, the f-structure of a coordinate
structure like David and Chris is a set, and the f-structure for each of the conjuncts
David and Chris is an element of the set, as shown in (97).
Sets representing coordinate structures are treated specially: coordinate struc-
tures are hybrid objects, sets that can have their own attributes and values as well
as having elements.24 Attributes which can be associated with a set are called
nondistributive features, while attributes which cannot be associated with a set
are called distributive features. We will return to a discussion of nondistributive
features and the distributive-nondistributive distinction in §2.5.7, in §6.3.2, and in
our discussion of feature resolution in §16.8.1.
As noted above, we represent sets in curly brackets that contain the element
f-structures. If a set has additional nondistributive attributes, we enclose the set
in square brackets and list the attributes and values of the set within the square
brackets. For example, if a set f has the nondistributive attribute a with value v it
looks like this:
& '
(96) a v
f { }

In the following example, the conjuncts of the coordinate subject David and Chris
are each singular (with the value sg for the num feature in the index feature
bundle) but the coordinate structure as a whole is a plural phrase. Thus, the set
bears the nondistributive attribute index with pl as the value for the num feature:25

(97) David and Chris yawn.

 
’
 
  
 
 
  
  
   
  
 
  




   
  
  

 

   

 








Notice that nondistributive attributes such as index may appear as attributes of the
conjuncts in a set as well as attributes of the set itself: in example (97), each of the

24 The proposal to treat sets representing coordinate structures as hybrid objects is originally due
to John Maxwell (p.c.), and is foreshadowed in work on coordination in LFG by Peterson (1982) and
Andrews (1983a).
25 We motivate a more complex representation of the number feature in §2.5; in this section we make
the simplifying assumption that num has atomic values like sg (singular) and pl (plural).
50 functional structure

conjuncts David and Chris has a singular index value, while the set as a whole has
a plural index. In contrast, conjuncts can have distributive attributes like pred,
but a set cannot have a distributive attribute.

2.4.6 Wellformedness conditions on f-structures

F-structures are subject to certain wellformedness conditions: Completeness,
Coherence, and Consistency (Kaplan and Bresnan 1982). The Completeness and
Coherence conditions ensure that all the arguments of a predicate are present and
that there are no additional arguments that the predicate does not require. The
Consistency condition ensures that each attribute of an f-structure has a single
value. We also discuss these requirements in §5.2.2.

2.4.6.1 Completeness The Completeness requirement tells us what is wrong

with a sentence like:

(98) ⃰David devoured.

Intuitively, some required material is missing from an f-structure that is

incomplete. The required material is specified as a part of the value of the pred
attribute, the semantic form. The pred attribute and semantic form for a verb like
devoured are:
5 :
(99) pred ‘devour!subj,obj"’

The argument list of a semantic form is a list of governable grammatical functions26

that are governed, or mentioned, by the predicate: in example (99), devour governs
the grammatical functions subj and obj. Example (98) contains a subj but no obj;
this accounts for its unacceptability according to the Completeness requirement.27
Previous LFG literature has employed a variety of notations for the argument
list. In the notation employed here, the argument list consists of a list of names
of governable grammatical functions. In other work, the argument list is some-
times depicted as a list of f-structures which are the values of the subcategorized
functions:

(100) David devoured the cake.

 
, ’
 
 
 
 
 
 
 
+

26 Recall from §2.1.2 that the governable grammatical functions are subj, obj, obj𝜃 , obl𝜃 , xcomp,
and comp.
27 Optional dependents (for example, instruments and benefactives) are discussed in §9.10.5.2.
 functional structure representation 51

It is also common for the argument list to be represented in the following way,
where (↑ subj) represents the subject f-structure and (↑ obj) represents the object
f-structure, as explained in §5.3.1:

(101) David devoured the cake.

 
(↑ ↑ ’
 
 
 
 
 
 
 
+

These notational variants are used equivalently, though technically the variant
shown in (101) is incorrect, since it contains the uninstantiated f-structure
metavariable ↑ (which we introduce in §5.3.1); here, we choose the more succinct
representation in (99) to save space and make the f-structures more readable.
There is a difference between grammatical functions that appear inside the
angled brackets and those that appear outside. In (99), the functions subj and
obj are semantic as well as syntactic arguments of devour, contributing to its
meaning as well as filling syntactic requirements. In contrast, the semantically
empty expletive subject it of a verb like rain makes no semantic contribution—it
is athematic. This intuitive difference is reflected in the formal requirement that
arguments of a predicate that appear inside angled brackets must contain a pred
attribute whose value is a semantic form; this is not required for arguments outside
angled brackets. Thus, the subj function appears outside the angled brackets of the
argument list of the semantic form of rain, since there is no semantic role associated
with it:

(102) It rained.
5 :
pred ‘rain!"subj’

Note that it is not a violation of any condition for more than one predicate to
govern an f-structure with a semantic form. In fact, this is a common situation with
“raising” verbs like seem, whose subject is also the subject of its xcomp argument
(see Chapter 15). Since the subject of seem is a syntactic but not a semantic
argument of the seem predicate, the subj in the value of the pred attribute of seem
appears outside the angled brackets:

(103) David seemed to yawn.

 
 
 
 
 
 ’ 
 
 
52 functional structure

The line connecting the f-structure for David to the subj position of seem indicates
that the same f-structure is the value of two different attributes, as discussed in
§2.4.3 of this chapter: it is both the subj of seem and the subj of yawn. Complete-
ness is satisfied for both predicates: each requires a subj, and this requirement is
satisfied for each verb.
Following Kaplan and Bresnan (1982), the Completeness requirement can be
formally defined as follows:

(104) Completeness:
An f-structure is locally complete if and only if it contains all the governable
grammatical functions that its predicate governs. An f-structure is complete
if and only if it and all its subsidiary f-structures are locally complete.

Chapter 8 provides further discussion of the place of the pred feature in the theory
of the syntax-semantics interface and its role in ensuring syntactic wellformedness.

2.4.6.2 Coherence F-structures which do not obey the Coherence requirement

include extra governable grammatical functions that are not contained in the
argument list of their semantic form:

(105) ⃰David yawned the sink.

The f-structure for this sentence is ill-formed:

(106) Ill-formed f-structure:

 
’
 
 
 
 
 
 
+

The governable grammatical function obj is present in this f-structure, though it

is not governed by the semantic form of yawn. Consequently, the f-structure is
incoherent.
Of course, the Coherence requirement applies only to governable grammatical
functions, not functions that are ungoverned, such as modifying adjuncts. The
following f-structure is perfectly coherent. Besides the single governable function
subj, it contains a modifier adj, which is not a governable function:

(107) David yawned yesterday.

 
’
 
 
 
 functional structure representation 53

An f-structure is coherent if all of its governable grammatical functions are

governed by a predicate. The following f-structure is incoherent, since there is no
pred in the larger f-structure whose argument list contains obj:

(108) Ill-formed f-structure:

5 7 8:
obj pred ‘David’

It is usual but not necessary for the argument list of a predicate to mention single
grammatical functions and not lists of functions or paths through the f-structure.
In some treatments of subcategorized oblique phrases, however, the argument list
of a predicate contains longer expressions, such as oblon obj; see, for example,
Levin (1982) and Falk (2001b):

(109) David relied on Chris.

 
’
 
 
 
 

This f-structure is coherent because the governable grammatical functions it

contains are mentioned in the argument list of rely. That is, ‘rely!subj,oblon obj"’
governs the obj of the oblique function oblon . We do not adopt this treatment of
oblique phrases here, but merely display an example to illustrate this possibility;
the representation we will use is shown in (89).
The Coherence requirement can be formally defined as follows (Kaplan and
Bresnan 1982):

(110) Coherence:
An f-structure is locally coherent if and only if all the governable gram-
matical functions that it contains are governed by a local predicate. An
f-structure is coherent if and only if it and all its subsidiary f-structures
are locally coherent.

2.4.6.3 Consistency The Consistency requirement, or Uniqueness Condition,

reflects the functional (as opposed to relational) nature of the f-structure. An
attribute of an f-structure may have only one value, not more (Kaplan and Bresnan
1982):

(111) Consistency:
In a given f-structure a particular attribute may have at most one value.

This requirement disallows f-structures satisfying incompatible constraints:

54 functional structure

(112) ⃰The boys yawns.

Ill-formed f-structure:
 
’
 
 
 

The subj noun phrase the boys is a plural (pl) phrase, but the verb yawns requires
its subject to be singular (sg). These two requirements cannot be simultaneously
met: the value of the attribute num must be either sg or pl, and it cannot have both
values at the same time.

2.5 Functional features

The f-structure attribute pred, whose value is a semantic form, is crucial in
the statement of syntactic subcategorization requirements. The appearance and
distribution of other f-structural features is constrained in terms of functional
syntactic relations, and so their presence at f-structure is essential: agreement
features such as pers and num are specified for particular grammatical functions,
such as subj in subject-verb agreement; features specifying form, such as case
and vtype (discussed in §2.5.5.1), are relevant at a functional syntactic level for
specifying the required morphological form of an argument; and “sequence of
tense” phenomena govern syntactic requirements on tense and aspect realization.
Only features that can be argued to play a role in functional syntactic constraints
are represented at f-structure; features encoding purely semantic, prosodic, or
information-structural content are represented at other, nonsyntactic levels of
representation, as we discuss in Part II of the book.

2.5.1 Simple and complex values

Features can have various kinds of values: the value of features like subj and obj
is an f-structure, the value of the pred feature is a semantic form, and the value of
a feature like def (encoding definiteness) is an atom such as +, with no internal
structure. Some features have complex, nonatomic values whose structure is moti-
vated by patterns involving feature resolution, feature indeterminacy, syntactic
compositionality, or natural classes of feature values.
For example, the value of the feature pers (person) is often represented as a
simple atomic value such as 1, 2, or 3:

(113) Simple atomic values for the feature pers: 1 (first person), 2 (second
person), 3 (third person)

However, atomic values such as these do not underpin a predictive theory of feature
resolution in coordination. For example, a theory of syntactic feature resolution
 functional features 55

should predict that coordinating a second person pronoun with a third person
pronoun produces a coordinate phrase with a second person feature in every
language (Corbett 2000), and that other resolution patterns are impossible:

(114) [you2 and he3 ]2

Similarly, the use of atomic features does not allow for a formally explicit theory
of natural classes of features and syncretism (Baerman et al. 2005). For example,
Baerman et al. (2002) present the following paradigm for present tense verbs in
Rongpo,28 exemplifying with the verb ‘go’:

(115) Present tense of the verb ‘go’, Rongpo:

sg pl
1 gyən̊
2 gyən gyən̊i
3

The same verb form gyən is used with second and third person singular subjects.
Representation of the value of the pers feature as a simple atomic value such
as 2 or 3 does not allow for the formal expression of properties (such as “non-
first-person”) that the two values have in common. As Baerman et al. (2005:
Chapter 3) point out, other languages display syncretisms involving first/third
person and first/second person, and so the complex value of the pers feature
must allow reference to these commonalities as well. We discuss proposals for the
complex representation of the value of the pers feature in §2.5.7.2.
It is also often assumed that the num (number) feature has simple atomic values
such as sg (singular), du (dual), and pl (plural):

(116) Simple atomic values for the feature num: sg (singular), du (dual), pl
(plural)…

However, this representation does not support an explanatory theory of con-

structed number, where the value of the num feature is determined by a combina-
tion of constraints associated with different words in the sentence. Sadler (2011)
discusses the data in (117) from Hopi, taken from Corbett (2000: 169), in which
pronouns distinguish singular number (sg) from nonsingular number (nonsg,
more than one individual), while verbs distinguish plural number (pl, more than
two individuals) from nonplural number (nonpl, one or two individuals). A
subject is interpreted as having dual number if a nonsingular pronoun (reference
to more than one individual) appears with nonplural verb agreement (reference to
one or two individuals):

28 Baerman et al. (2002) attribute the data in (115) to Zoller (1983).

56 functional structure

(117) a. Pam wari

that.sg run.nonpl
‘S/he ran.’
b. Puma yùutu
that.nonsg run.pl
‘They ran.’
c. Puma wari
that.nonsg run.nonpl
‘They (two) ran.’

Thus, dual number in Hopi is constructed by a combination of nonsg nomi-

nal number and nonpl verbal agreement. In §2.5.7.4, we discuss proposals for
accounting for these patterns on the basis of a complex value for the num feature.
Finally, it is often assumed that the value of the feature case is a simple atomic
value such as nom, acc, or dat:

(118) Simple atomic values for the feature case: nom (nominative), acc
(accusative), dat (dative), erg (ergative), abs (absolutive), gen (genitive),
loc (locative)…

Atomic values do not support a predictive theory of case indeterminacy (Zaenen

and Karttunen 1984; Pullum and Zwicky 1986), illustrated in (119), in which a
form can satisfy more than one case requirement at the same time. The German
verb findet ‘finds’ requires an accusative object, and the verb hilft ‘helps’ requires
a dative object. The noun phrase Papageien ‘parrots’ is indeterminately specified
for case, and can appear with a verb requiring an accusative object (119a), a verb
requiring a dative object (119b), or, crucially, as the object of both verbs at the same
time (119c):

(119) a. Er findet Papageien.

he finds parrots
obj:acc nom/acc/dat/gen
‘He finds parrots.’
b. Er hilft Papageien.
he helps parrots
obj:dat nom/acc/dat/gen
‘He helps parrots.’
c. Er findet und hilft Papageien.
he finds and helps parrots
obj:acc obj:dat nom/acc/dat/gen
‘He finds and helps parrots.’

Complex values for the case feature allow indeterminate forms to satisfy multiple
conflicting case requirements simultaneously imposed by different predicates, as
we discuss in §2.5.7.5.
 functional features 57

A number of important theoretical and empirical issues arise in the determina-

tion of the proper representation of the values of features such as pers, num, and
case. We do not always take a firm stand on the proper analyses of the relevant
cases, but it is vital to keep these issues in mind when determining the proper
representation of the value of a feature. We hope that an explicit statement of these
issues will help to guide future research on the structure and representation of
complex feature values.
The first issue relates to universality. Is the representation and structure of the
value of a feature determined on a language-by-language basis, or do all languages
make use of the same representation? Can the same feature have a simple, atomic
value in one language, and a complex value in another language? Can the same
feature have different complex values in different languages?
The second is the issue of the relation between syntax and meaning. For seman-
tically motivated features such as pers and num, do we expect the structure of a
complex f-structure feature value to partially or completely reflect details of the
meaning of the elements bearing the features? If there is a systematic relation
between the representation of a syntactic feature value at f-structure and the
meaning of phrases bearing the feature, should this be attributed to historical
factors, or do we require a synchronic theory of this relation?
The third is the issue of markedness.29 Some feature values are traditionally clas-
sified as unmarked: for example, third person is usually classified as the unmarked
value of the pers feature. It is often proposed that there is an important relation
between the markedness of a feature value and its representation: unmarked
properties should be formally represented either by a negative value for a feature
or by the absence of the feature, while marked values are represented by a positive
value or by the presence of the feature. Should our representation of feature
values conform to this generalization? We discuss technical issues that arise in
the representation of one of the values of a feature by the absence of the feature
in §2.5.2.
The fourth is the issue of consistency of representation. Do patterns of fea-
ture resolution, feature indeterminacy, syntactic compositionality, syncretism, and
markedness converge on a single representation for the value of a feature? It may
be, for example, that evidence from syncretism and natural classes of feature values
motivates a particular representation of the complex value of a feature, while
evidence from resolution motivates a different, incompatible representation. In
such cases, which phenomenon takes precedence in determining how the value
of a feature should be represented?
Besides these theoretical issues, a practical issue arises with regard to the
f-structure representations that we will provide. Even though the representation
of the case feature in (120a) does not support a theory of case indeterminacy, it is
clearer, simpler, and easier to read than the representations in (120b) and (120c):

29 For definition and discussion of markedness and the representation of feature values, see Jakobson
(1932), Greenberg (1966a), Moravcsik (1988), Blevins (2000), and references cited in those works.
58 functional structure

(120) (simple atomic representation)

b. (set-based representation)
  
−
  
  +
c. 
  
−
 (feature structure representation)
  
..
.

Often, if we are not concerned with issues of feature indeterminacy, feature

resolution, constructed values for features, or representing commonalities among
feature values, our discussion will not depend on a particular representation of
the value of the case feature or other features with complex values. In fact,
the f-structures which we display will be easier to read if we adopt a simple
representation such as the one in (120a) when possible. In this book, therefore,
we follow the convention that if the details of the representation of a complex
feature value are not important to the discussion, we represent it as a simple atomic
value. In such cases, the simple value can be thought of as an abbreviation for the
corresponding complex value.

2.5.2 The absence of a feature as a feature value

In some approaches, the absence of a feature is taken to represent one of the
possible values of the feature. Features treated in this way are sometimes called
privative features, though this term is often reserved for the particular case in
which the feature has two values, represented as the presence or absence of the
feature. In such cases, the absence of a feature is often thought of as representing
the “unmarked” value for the feature, with the “marked” value represented as the
presence of the feature.
For example, Falk (2001b: 82) proposes the following features to represent the
different forms of the English verb:

(121) present tense:

past tense:
-ing form:
-ed/-en form:

On Falk’s view, tensed forms are associated with a tense feature, participial forms
are associated with a part feature, and the infinitive is identified by the absence of
both of these features.
One issue that arises for such approaches is the treatment of feature under-
specification. In many approaches, a word form which can have any of several
values for a feature is treated by leaving the value for the feature underspecified
 functional features 59

or unspecified. For example, the verb form put is the present tense form, the past
tense form, or the infinitive form of the verb:

(122) a. David and Chris put their dishes in the dishwasher every day. [present
tense]
b. David and Chris put their dishes in the dishwasher yesterday. [past tense]
c. David and Chris don’t want to put their dishes in the dishwasher today.
[infinitive]

Given Falk’s system of feature specifications, we cannot analyze put by leaving

the value of the tense feature unspecified, since the lack of a tense feature
(along with the lack of a part feature) marks the form as infinitival, and is not
compatible with the representation of a tensed form. This means that if the absence
of a feature is treated as signaling one of the values of the feature, some other
approach to underspecification must be taken. See Payne and Börjars (2015) for
further discussion of this issue, and Kaplan (2017) for an overview discussion of
underspecification and feature indeterminacy.

We now turn to an examination of the particular f-structure features and their

values that we assume in this book.

2.5.3 FORM features

Some predicates require one or more of their arguments to contain words of a
particular form. For example, English weather verbs such as rain and snow require
a pleonastic or semantically empty subject which must have the form it:

(123) It/ ⃰There/ ⃰∅ rained.

It is, then, necessary to encode information about the form of words at f-structure,
in order to allow such requirements to be stated. Word form is encoded by means
of features such as form; the f-structure for It rained is:

(124)

Notice that the subj argument of rain appears outside the angled brackets in the
semantic form, since it is syntactically required but does not make a semantic
contribution (§2.4.6.1). The f-structure of the subject contains the attribute form
with value it, satisfying the special requirements of the verb rained.
This means that the pronoun it makes one of two different f-structure contribu-
tions on each occasion of its use: one of them, shown in (124), has a form attribute
with value it, while the other is a standard referential pronoun containing a pred
attribute with value ‘pro’, as in a sentence like David saw it. It is generally assumed
60 functional structure

that the only words which have an f-structure with a form feature are those whose
presence is syntactically required by some predicate in the language but which
do not make a semantic contribution. For example, there is no English verb that
requires the presence of an argument with the form hippopotamus, and so there is
no f-structure for the word hippopotamus which contains a form feature and no
pred feature.
Some verbs place constraints on the form of the complementizer of their
complement clause; for example, Quirk et al. (1972: §15.6) claim that the verb
justify allows the complementizer whether, but not the complementizer if, in its
comp argument:

(125) You have to justify whether/ ⃰if your journey is really necessary.

A complementizer like that or whether contributes its features to the same f-

structure as its complement clause, as we will see in Chapter 4. When more than
one word contributes information to the same f-structure, it is not possible for both
words to contribute values for the same form feature, since this causes a clash. This
means that complements must contribute a special attribute compform whose
value is the form of the complementizer. Thus, a sentence like Chris knows whether
David left has the f-structure given in (126), in which the subordinate clause comp
has a compform feature with value whether:

(126) Chris knows whether David left.

 
’
 
 
 
  
 ’ 
 
  
  
 

It is the value of the compform feature that is constrained by the verb justify,
which disallows the value if for the compform feature in its comp. We return
to a discussion of this example in §5.2.5.
In their comparative grammars of English, French, and German, Butt et al.
(1999) propose a number of additional specialized form features, including pron-
form for pronouns, spec-form for nominal specifiers like the, conj-form and
preconj-form for conjunctions like and and preconjunctions like either or both,
prt-form for particles (including German verbal particles as well as English
particles like up), and comp-form (which we call compform). In the compu-
tational setting of Butt et al.’s work, the linguistic analysis of a sentence is input
to a computer application that performs further processing for some practical
purpose. In such a setting, features such as these are often included in the analysis
in case they are useful for such applications, rather than for purely syntactic
reasons. In the analyses we present, we introduce only f-structure features which
 functional features 61

we can show to be clearly motivated by syntactic patterns and which take part in
syntactic selection requirements. It may be that further research will reveal the
need for additional form features at f-structure such as those that Butt et al. (1999)
propose.
In §11.4.1, we will introduce an fm string feature whose value is the form of
the unit of the string under analysis. We will see that the fm feature bears a close
relation to the f-structure form feature, but is relevant for relations in the string
rather than f-structural relations and constraints.

2.5.4 PFORM and PCASE

Prepositions and particles contribute information about their form by means of the
pform feature. Additionally, prepositions may be specified with information about
the grammatical function of the argument which they mark; this information is
recorded as the value of the pcase feature.
The pcase feature reflects the fact that the grammatical role of an oblique
prepositional phrase is often determined by the preposition. Kaplan and Bresnan
(1982) propose the (now non-standard) f-structure in (127) for the sentence A girl
handed a toy to the baby:

(127) Non-standard f-structure for A girl handed a toy to the baby according to
Kaplan and Bresnan (1982, (45)):
 
 
 
 
 
 
 ’ 
 
 
 
 
 
  
 
 
  
  
  
  
 

According to this early analysis, the value of the pcase attribute is to, which
is also the grammatical function of the phrase to the baby (see Butt et al. 1999
for a different but related treatment). In later work, two different attributes have
generally been assumed: the form of the preposition is instead recorded as the
value of the pform feature, as we describe below, while the value of the pcase
attribute is specified as the particular oblique grammatical function contributed
by the prepositional phrase. For example, in (128), the value of the pform
attribute is to, and the value of the pcase attribute is the grammatical function
oblgoal , which is also the grammatical function of the prepositional phrase
to the baby:
62 functional structure

(128) F-structure for A girl handed a toy to the baby, standard treatment of pcase
feature:
 
 

 ’ 

 
 
  
 
 
  
  
  
 

In §6.2, we describe the constraints which allow the value of the pcase attribute to
specify the grammatical function borne by the oblique phrase.30
F-structural specification of the form of particles or prepositions is necessary
because predicates can place idiosyncratic requirements on the form of their
prepositional complements or particles (§2.1.9.2):

(129) a. David relied on / ⃰to / ⃰about Chris.

b. David used the sugar up / ⃰down / ⃰through.

The pform feature records the value of the form of the particle or preposition so
that it can be constrained and checked by the predicate:31

(130) a. David relied on Chris.

 
’
 
 
 
 
 

b. David used the sugar up.

 
’
 
 
 
 
 

30 As we discuss in §§6.1.3 and 6.2, the pcase analysis is closely related to the constructive case
theory of Nordlinger (1998, 2000).
31 See Toivonen (2003) for an in-depth discussion of particles in Germanic languages, and an
alternative proposal for the f-structure of English particle verbs.
 functional features 63

2.5.5 Verbal and clausal features

Much previous LFG work has made use of verbal and clausal features such as
vtype and tense, but there has not always been agreement on the appropriate
values for these features. Influential proposals have been made by Butt et al. (1999),
Falk (2001b §3.5 discussed in §2.5.2 above), and Payne and Börjars (2015). mood,
aspect, and voice are also important features in verbal and clausal analysis.

2.5.5.1 vtype and tense We adopt the proposal of Payne and Börjars (2015)
for the vtype and tense features and their values. On their proposal, all verb forms
have a vtype feature, and tensed forms also have a tense feature:32

(131) Finite present non-third singular (help):

Finite present third singular (helps):

Finite past (helped):

Present participle (helping):

Past participle (helped):

help):

The vtype feature has the values listed in (132) in English; additional values may
be required for other languages, depending on the richness of the verbal paradigm:

(132) Values for the vtype feature: fin (finite), prs.ptcp (present participle),
pst.ptcp (past participle), inf (bare infinitive)

For English, the tense feature has at least the values past (pst) and present (prs).
There is some controversy over the status of English will as a future tense marker
(Huddleston 1995); we do not take a position on this issue here, though we note
that fut is a possible value for the tense feature in many languages:

(133) Values for the tense feature: pst (past), prs (present), fut (future)

Some languages make additional distinctions by subdividing past or future tense,

distinguishing, for example, recent from remote past tense or near from remote
future tense (Comrie 1985); additional values of the f-structure tense feature may

32 Payne and Börjars use the attribute vform rather than vtype for this feature. For consistency with
the Leipzig glossing rules (Bickel et al. 2015), we provide the value prs.ptcp for the present participial
form of a verb, rather than Payne and Börjars’s ger-part.
64 functional structure

be needed in the analysis of these languages, depending on the relevance of these

semantic distinctions for syntactic analysis.33
The tense feature has also been argued to have a complex, nonatomic value.
Nordlinger and Bresnan (1996) present an insightful analysis of tense marking
in Wambaya that decomposes the value of the tense feature into three primitive
binary features: +past, +future, and +uncertain. In this book, we will supply an
atomic value such as pst or prs for the feature tense: as discussed in §2.5.1, we may
think of this as an abbreviation for a complex value in languages like Wambaya,
where a decomposition of the value of the tense feature is motivated.

2.5.5.2 mood and aspect Besides the tense feature, aspect and mood con-
stitute the other two dimensions of tense-aspect-mood systems (often abbreviated
as TAM). We adopt Butt et al.’s (1999) treatment of the mood feature, and a variant
of their analysis for the aspect feature.
Butt et al. (1999) propose the values ind (indicative), sbjv (subjunctive), and
imp (imperative) for the feature mood in English, French, and German; to these,
we add the value interr (interrogative):

(134) Values for the mood feature: ind (indicative), sbjv (subjunctive), imp
(imperative), interr (interrogative)

For other languages, additional values may be required: for example, Ancient
Greek has an optative (opt) mood.
The treatment of aspect is more complex. Butt et al.’s analysis of English,
French, and German makes use of an aspect feature which they abbreviate as perf.
Butt et al. (1999) do not distinguish between perfect and perfective aspect in their
representation of the aspect feature; we follow standard practice in using prf for
perfect aspect, and pfv for perfective aspect. We also follow Butt et al. in using a
single feature, which we represent as prf, in the analysis of aspectual distinctions
in English, French, and German.
According to Butt et al.’s analysis of English, the feature prf appears with value +
for sentences with the perfect auxiliary, and does not appear in sentences without
the perfect auxiliary:
5 7 8:
(135) a. David has eaten a banana: aspect prf +

b. David ate a banana: no aspect feature

Butt et al. also propose a feature prog (progressive), since in English, a sentence
can have both perfect and progressive aspect:

33 Here and in general, it is worth keeping in mind that the f-structure is a purely syntactic
representation, and need not encode fine-grained semantic distinctions. The distinction between near
and remote past may not be syntactically significant in any language, even when it is important
semantically; whether or not the full range of semantically relevant distinctions is encoded in the
f-structure is one aspect of the general issue of the relation between syntax and meaning, discussed
in §2.5.1.
 functional features 65

+
(136) David has been eating bananas:
+

A representation like this works well for languages like English, in which two
separate dimensions of the aspect feature must be represented in the same
f-structure. We provide the representation in (137) for such languages, similar to
the representation proposed by Butt et al. (1999):

(137) Complex values for the aspect feature:

Perfect: +

Progressive: +

+
Perfect progressive: +

We occasionally use atomic symbols as abbreviations for the complex structure in

(137): for example, we can use the abbreviation prf.prog as an abbreviation for
the complex value given in (137) for perfect progressive aspect.
For languages whose aspectual syntax does not require consideration of more
than one dimension of aspect for any sentence, a simpler representation suffices,
using the attribute aspect with an atomic value:

(138) Atomic values for the aspect feature: prf (perfect), iprf (imperfect), pfv
(perfective), ipfv (imperfective), prog (progressive)

For an alternative proposal for the f-structural representation of aspect, including

a discussion of lexical aspect or aktionsart, see Glasbey (2001).
It is worth noting that tense, aspect, and mood are not exclusively clausal
categories. Nordlinger and Sadler (2004a,b) discuss TAM marking on nominals
in a number of languages, and Arka (2013) discusses nominal aspect marking in
Marori. Nordlinger and Sadler show that in some cases, TAM marking appearing
on nominals marks the TAM of the clause in which the nominal phrase appears; in
other cases, nominal TAM marking is relevant to the timespan of the referent of the
noun phrase, and may differ from the TAM of the clause. Arka shows that the same
is true of nominal aspect in Marori. As motivation for representing nominal TAM
in the f-structure, Nordlinger and Sadler show that tense is involved in syntactic
agreement relations within the noun phrase in Somali: adjectives agree in tense
and gend with the head noun, and so tense must be represented in the f-structure
of nouns in Somali and other languages with TAM agreement within the noun
phrase. Nordlinger and Sadler (2004b) and Arka (2013) present LFG analyses of
nominal tense and aspect which ensure that the TAM features are available for
syntactic agreement relations where necessary, and that the features appear in the
correct f-structure in each instance.
66 functional structure

2.5.5.3 voice Many voice systems encode a two-way distinction between active
and passive voice; to represent this distinction, Sadler and Spencer (2001) propose
a feature voice with values active (active) and pass (passive). Often, active voice
is treated as the unmarked, default value for the voice feature: if we adopt this view
in Sadler and Spencer’s setting, active voice can be represented by the absence of the
voice feature, and passive voice can be represented by the presence of the voice
feature with the value pass. Additional values for the voice feature are necessary
in the analysis of languages with antipassive voice (antip) and middle voice (mid)
(Fox and Hopper 1993).
As we discuss in Chapter 9, voice alternations are often viewed in terms of
“promotion” and “demotion” of arguments, with the active voice taken as basic,
and the passive voice involving “demotion” of the subject to oblique status, and
“promotion” of a nonsubject argument to subject status. Many languages have
voice systems which do not work in this way: in so-called “symmetric voice
systems” (Foley 1998), voice alternations involve singling out one argument as
the subj, but without demoting the other core arguments or altering their core
status. Such systems often have several voices, each named after the role of the
argument that is singled out as the subj. For example, Kroeger (1993) discusses
five voices in Tagalog: actor voice (sometimes called “active voice”: av), objective
voice (ov), dative/locative voice (dv), instrumental voice (iv), and benefactive
voice (bv). Interestingly, Arka (2003) shows that Balinese has a three-way voice
alternation, with demotional and symmetric voice alternations coexisting in the
same language: Balinese distinguishes agentive voice (similar to actor voice in
Tagalog, and also represented as the value av), a non-demotional objective voice
(ov) in which the agent remains a core argument, and a demotional passive voice
(pass) in which the agent is demoted to oblique status. Kroeger (1993), Arka
(2003), and Falk (2006) provide an in-depth discussion of these “symmetric voice”
languages, their clausal syntactic organization, and their analysis within the theory
of argument mapping, the relation between grammatical functions and semantic
roles; we discuss argument mapping in Chapter 9.
In languages with direct and inverse voice, the status of the arguments of a
verbal predicate and their position on an animacy/saliency hierarchy determines
the voice of a clause: for example, a first person agent acting on a third person
patient must be expressed with direct voice (dir), but a third person agent acting
on a first person patient must be expressed in inverse voice (inv) (see §5.2.10 for
a discussion of related patterns in Lummi). Arnold (1994) provides an analysis of
argument mapping and inverse voice in Mapudungan.
Our inventory of values for the voice feature must, then, encompass values
that are relevant in the analysis of demotional voice alternations such as the
active/pass alternation as well as voice alternations in symmetric voice languages
and languages with direct and inverse voice. Additional values may be needed
in the analysis of symmetrical voice languages with voice distinctions other than
those attested in Tagalog and Balinese:

(139) Values for the voice feature: active (active), pass (passive), antip
(antipassive), mid (middle), av (actor/agentive voice), ov (objective voice),
dv (dative voice), bv (benefactive voice), iv (instrumental voice), dir
(direct voice), inv (inverse voice)
 functional features 67

2.5.6 Negation
The presence of negation affects syntactic patterns such as case assignment and
negative concord (see Przepiórkowski and Patejuk 2015 for Polish), and so nega-
tion must be represented explicitly at f-structure. Clausal negation is represented
by a feature neg with value + by King (1995), Butt et al. (1999), and Falk (2001b),
among others:

(140) F-structure for David didn’t yawn with neg feature:

 
’
 
 
+

A related feature, neg-form, is proposed by Butt et al. (1999: 142–3) in the analysis
of bipartite clausal negation forms such as French, with the initial component
contributing a neg feature, and the postnegation component contributing the
value of neg-form:

(141) F-structure for French bipartite negation ne. . . . pas according to Butt et al.
(1999):
David n’ a pas mangé de soupe.
neg have postneg eaten of soup
‘David did not eat any soup.’
 
’
 
 
 
 
 
 
 + 

Butt et al. (1999: 143) propose that the neg feature is also used in the treatment
of constituent negation, where only a subconstituent is negated and not the entire
clause:

(142) F-structure for David found the ball not in the box [but on the floor…]
according to Butt et al. (1999):
 
’
 
 
 
 
 
 
   
 ’ 
 
  
   
 
+
68 functional structure

However, a single neg feature may not be adequate in the analysis of multiple
negation examples such as (143), from Przepiórkowski and Patejuk (2015):

(143) John doesn’t not like Mary.

An undesirable analysis of example (143) would be for each occurrence of not in

(143) simply to require the value + for the neg feature in the clausal f-structure,
since this would render the resulting f-structure indistinguishable from the
f-structure for John doesn’t like Mary, with only one negation:

(144) F-structure for John doesn’t like Mary with neg feature:
 
’
 
 
 
 
 
 
+

To ensure the proper treatment of examples with multiple negation, the ParGram
Consortium (2017) proposes to treat at least some instances of clausal negation
adverbially, with an f-structure representing clausal negation in the adj set:

(145) F-structure for John doesn’t like Mary with negation as an adjunct:
 
’
 
 
 
 
 
 
 

This representation allows multiple instances of negation to be represented sepa-

rately, each as a member of the adj set:

(146) F-structure for John doesn’t not like Mary with negation as an adjunct:
 
’
 
 
 
 
 
 
  
 
  
 
 
 

In his analysis of negation in Hungarian, Laczkó (2015) introduces an additional

polarity feature pol, and proposes that different morphosyntactic realizations of
negation are represented differently: affixal negation (which cannot be iterated) is
 functional features 69

represented by the [neg +] feature, while clausal or constituent negation (encoded

by the negative particle in Hungarian) is represented as a member of the adj set, as
in (146), and additionally by a [pol neg] feature to mark the scope of negation. In
related work, Przepiórkowski and Patejuk (2015) explore the possibility of using
only the neg and pol features, eschewing the adverbial representation in (146).
They argue that the two features neg and pol are adequate to handle doubly
negated examples such as (143) as well as the distinction between clausal negation
(which they call “eventuality negation”) and constituent negation. For more on the
morphological and syntactic representation of negation, see Laczkó (2014c, 2015),
Przepiórkowski and Patejuk (2015), Bond (2016), and the ParGram Consortium
(2017).

2.5.7 Nominal features

Nominal features are different from clausal features in that they behave specially
in coordination; an important component of the analysis of nominal features is the
distinction between distributive and nondistributive features, introduced in §2.4.5.
Another dimension of difference is the classification of nominal features governing
agreement relations into index features and concord features. In this section, we
discuss nominal features and their values, including def, spec, deixis, prontype,
pers, gend, num, and case.

2.5.7.1 Agreement features index and concord Nominal features relevant

for syntactic agreement and government relations include person (pers), number
(num), gender (gend), and case (case). Wechsler and Zlatić (2003) argue con-
vincingly that these features make up two different feature bundles, index and
concord (see also Pollard and Sag 1994; Kathol 1999; King and Dalrymple 2004;
Wechsler 2011; Hristov 2012; Belyaev et al. 2015; Haug and Nikitina 2016).34, 35
index features group together pers, num, and gend features; although index
features are syntactic features and appear in f-structure, they are closely related
to semantic features and the meaning of the phrase bearing the features. Wech-
sler (2011) proposes that index features derive historically from incorporated
pronouns, accounting for the presence of the pers feature in the index bundle.
In contrast, concord features group together case, num, and gend features.
Like index, concord features are f-structure features, but concord is more
closely related to morphological declension class; according to Wechsler (2011),
concord features derive historically from nonpronominal sources, often from
incorporated nominal classifiers, accounting for the presence of case but not pers
in the concord bundle:

34 Bresnan et al. (2016: Chapter 8) also group pers, num, and gend features together as components
of the value of the index feature, as we do here. They provide a different treatment of index and
agreement features in their discussion of anaphoric binding in their Chapter 10, where they use the
cover term agr for pers, num, and gend features, and provide a variable such as i or j for the value of
the attribute index.
35 Alsina and Arsenijević (2012a,b) argue against the index/concord distinction and in favor of
a single type of syntactic agreement feature; Wechsler and Zlatić (2012) and Hristov (2013a) provide
counterarguments to Alsina and Arsenijević’s position.
70 functional structure
  
(147) num ...
 concord  gend ...  
 
 case ... 
 
  
 pers ... 
 
 index  num ...  
 
gend ...

As shown in (147), both the index feature bundle and the concord feature bundle
contain num and gend features. Since index and concord features are specified
separately, the values of these features can differ: for example, a form can have
singular concord but plural index, or vice versa. Wechsler and Zlatić (2003:
51) provide the Bosnian/Croatian/Serbian36 example in (148) to illustrate this
possibility:

(148) Ta dobra deca dolaze.

that.f.sg good.f.sg children come.3pl
‘Those good children are coming.’

The noun deca ‘children’ controls feminine singular agreement on the determiner
and adjective, but third person plural agreement on the verb. Wechsler and Zlatić
propose that deca is unusual in having feminine singular concord features but
neuter plural index features:

(149) index and concord features for deca, Wechsler and Zlatić (2003):
  
  
 
 
 
   
 
 
   
   
   
 

Subject-verb agreement is often determined by index features, as in example (148):

the verb dolaze ‘come.3pl’ shows third person plural index agreement with its
subject. In contrast, NP-internal agreement, including adjective-noun agreement,
is often determined by concord features: here, the demonstrative determiner ta
and the adjective dobra ‘good.f.sg’ show feminine singular concord agreement
with deca ‘children’.
index and concord features have been claimed to differ in their status as
distributive or nondistributive features in coordination. Recall from §2.4.5 that

36 Due to the complexity of the political and linguistic divisions in the former Yugoslavia we use
the name Bosnian/Croatian/Serbian to refer to the primary languages of Bosnia and Herzegovina,
Croatia, Montenegro, and Serbia, without prejudice as to how many distinct languages or varieties this
encompasses and without implying primacy of one variety or language over others.
 functional features 71

nondistributive features are associated with a coordinate structure as a whole; in

contrast, distributive features are associated with each conjunct of a coordinate
phrase, but not with the coordinate structure as a whole. King and Dalrymple
(2004) argue that index is a nondistributive feature, meaning that a coordinate
structure can have its own index features that may be different from the index
features of the conjuncts, while concord is a distributive feature, and not a feature
of a coordinate structure:

(150) index as a nondistributive feature (King and Dalrymple 2004):

 
  
 
 
 
 
 

 
 
   
 
 
 
 .. 
.

The classification of index as a nondistributive feature and concord as a distribu-

tive feature is adopted by Dalrymple and Nikolaeva (2006), Vincent and Börjars
(2007), Sadler and Nordlinger (2010), Hristov (2012), and many others. The view
that concord is always a distributive feature is challenged by Belyaev et al. (2015),
who propose that concord is distributive in some languages but nondistributive
in others.
The status of index as a nondistributive feature is important in the theory of
syntactic feature resolution: the person (pers) and gender (gend) index features
of a coordinate phrase are determined in a systematic way by the pers and gend
index features of the conjuncts. The num feature is also a part of the index feature
complex, but it does not resolve in the same way as pers or gend. In our discussion
of the pers, num, and gend features, we will examine different proposals for the
complex values of these features and the role that they play in the f-structures
of coordinated phrases and the theory of feature resolution. We postpone a full
discussion of feature resolution and the features of coordinate noun phrases to
Chapter 16.

2.5.7.2 Person The pers feature is often represented with a simple atomic
value:37

(151) Simple atomic values for the feature pers: 1 (first person), 2 (second
person), 3 (third person)

37 Wechsler (2004) proposes additional atomic values for languages like French, which have a
distinction in the first person between a singular pronoun (je) and a plural or ‘authorial’ singular
pronoun (nous), and in the second person between an informal singular pronoun (tu) and a plural or
formal singular pronoun (vous). According to Wechsler’s analysis, values for the pers feature in French
are 1s (first person singular), 1a (first person associative or “authorial” first person), 2s (second person
singular informal), and 2a (second person associative or formal); third person is represented by the
absence of a first or second person feature.
72 functional structure

There are two sorts of motivations for a more complex representation of the value
of the pers feature. First, as noted in §2.5.1, the pers feature resolves in predictable
ways in coordination: for English and many other languages, the resolution rules
are as in (152). In a formal theory of syntactic feature resolution, complex values
allow resolution to be modeled in a way that accounts for these patterns:

(152) Resolution of the pers feature:

first & second = first
first & third = first
second & third = second
third & third = third

Secondly, as noted in §2.5.1, values of the pers feature pattern together in agree-
ment relations, and a complex representation allows for the definition of natural
classes and commonalities in values of the pers feature. We discuss analyses
focusing on each of these phenomena in turn.
Dalrymple and Kaplan (2000) propose a theory of feature resolution which
defines the value of the pers feature as a closed set38 containing the atoms s
(mnemonic for speaker) and h (mnemonic for hearer):

(153) Values for the pers feature according to Dalrymple and Kaplan (2000):
{s, h}: first person
{h}: second person
{ }: third person

Dalrymple and Kaplan further propose that feature resolution can be modeled as
set union:39 the pers value of a coordinate structure is determined by taking the
set union of the pers values of the conjuncts. The table in (154) represents exactly
the same pattern of person agreement as the table in (152):

(154) Resolution of the pers feature via set union (Dalrymple and Kaplan 2000):
{s, h} (first) ∪ {h} (second) = {s, h} (first)
{s, h} (first) ∪ { } (third) = {s, h} (first)
{h} (second) ∪ { } (third) = {h} (second)
{ } (third) ∪ { } (third) = { } (third)

Other languages have a richer system of personal pronouns. For example, Fula and
many other languages exhibit a distinction between inclusive and exclusive first
person pronouns: the referent of an inclusive first person pronoun includes the
hearer, while the referent of an exclusive first person pronoun excludes the hearer.
For such languages, Dalrymple and Kaplan propose the following refinement to
the feature values in (153):

38 We define and discuss closed set descriptions in §6.3.4.

39 See §6.3 for definitions of set relations such as set union (∪) and intersection (∩).
 functional features 73

(155) Values for the pers feature in languages with a first person inclusive/
exclusive distinction according to Dalrymple and Kaplan (2000):
{s, h}: first person inclusive
{s}: first person exclusive
{h}: second person
{ }: third person

In their analysis of the pers feature, Vincent and Börjars (2007) propose that
the pers feature has the inventory of feature values in (155) in all languages,
distinguishing inclusive from exclusive pronouns in the first person, whether or
not there is a morphological inclusive/exclusive distinction in the pronominal
system.
In §2.5.1, we discussed theoretical issues that bear on the determination of the
value of features like pers. The issue of universality is whether all languages share
the same representation and structure for the values of a feature like pers. For
Vincent and Börjars, universality is a high priority: they propose as a working
hypothesis that “in grammatical sub-systems with a clear referential basis like per-
son, the definitions of the persons should be constant across languages” (Vincent
and Börjars 2007: 301–2). The issue of the relation between syntax and meaning
is whether and to what extent the complex value of a syntactic feature like pers
reflects details of the meaning of the form with which it is associated. In a language
like English, the first person plural pronoun we can be used to refer to the speaker
and the hearer (an inclusive reading) or to the speaker and another person (an
exclusive reading); prioritizing a transparent relation between syntax and meaning
entails that this meaning distinction should be encoded as a syntactic ambiguity
in the f-structural representation, even if there is no morphological or syntactic
evidence for this ambiguity. On the basis of these motivations, Vincent and Börjars
(2007) argue that the more complex system in (155) is universal, and relevant for
languages like English in addition to languages like Fula.
Sadler (2011) observes that set-based analyses of complex feature values can be
directly translated to equivalent feature-based analyses, with a positive value rep-
resenting the presence of an element in the set, and a negative value representing
its absence (see §6.3.4). Otoguro (2015) proposes a feature-based analysis of the
pers feature in his analysis of pers agreement in Germanic and Romance:

(156) Values for the pers feature according to Otoguro (2015):

; & '<
1 +
First person: pers
2 −
; & '<
1 −
Second person: pers
2 +
; & '<
1 −
Third person: pers
2 −
74 functional structure

According to Otoguro’s analysis, first person is represented with two features,

[1 +] and [2 −]. This corresponds to the first-person exclusive value in (155),
with the attribute 1 corresponding to s, the attribute 2 corresponding to h, the
value + corresponding to the presence of an element in the set, and the value −
corresponding to its absence. An additional value [1 +, 2 +] could be added for
first-person inclusive. Notably, unlike a set-based analysis, a feature-based analysis
allows underspecification: a first-person pronoun in a language like English, with
no inclusive/exclusive distinction, can be represented simply as [1 +]. Similarly,
non-first-person forms (as in the Rongpo verbal paradigm in 115) require [1 −],
but are unspecified for the 2 feature or its value: this means that they are compatible
with additional specification as either second person ([1 −, 2 +]) or third person
([1 −, 2 −]).

2.5.7.3 Gender Many LFG analyses assume simple atomic values for the gend
feature:

(157) Simple atomic values for the feature gend: m (masculine), f (feminine), n
(neuter) . . .

However, analyses of syntactic gender resolution often depend on a complex

value for the gend feature. For example, gender resolution in Icelandic obeys the
following rule (Corbett 1983):

(158) Gender resolution in Icelandic:

If the conjuncts are all masculine, the masculine form is used.
If the conjuncts are all feminine, the feminine form is used.
Otherwise the neuter form is used.

This is shown in example (159), where mixed-gender coordinate phrases control

neuter agreement (Dalrymple and Kaplan 2000):

(159) a. [Drengurinn og telpan] eru þreytt.

the.boy.m and the.girl.f are tired.n
‘The boy and the girl are tired.’
b. [Maðurinn og barnið] eru þreytt.
the.man.m and the.baby.n are tired.n
‘The man and the baby are tired.’
c. Ég sá [á og lamb], bæði svört.
I saw a.ewe.f and a.lamb.n both black.n
‘I saw a ewe and a lamb, both black.’

The generalization in (158) can be restated as:

 functional features 75

(160) Icelandic:
masculine & masculine = masculine
feminine & feminine = feminine
neuter & neuter = neuter
masculine & feminine = neuter
masculine & neuter = neuter
feminine & neuter = neuter

Dalrymple and Kaplan (2000) propose that the value of the gend feature is a
closed set containing members like m and f, and that, as in their analysis of person
resolution, gender resolution is modeled as set union. For Icelandic, they propose
the gend values in (161):

(161) Values for the gend feature in Icelandic according to Dalrymple and
Kaplan (2000):
masculine {m}
feminine {f}
neuter {m, f}

Given these values, modeling gender resolution as set union produces the correct
result, as can be seen by comparing the tables in (160) and (162):

(162) Resolution of the gend feature as set union (Dalrymple and Kaplan 2000):
{m} (masculine) ∪ {m} (masculine) = {m} (masculine)
{f} (feminine) ∪ {f} (feminine) = {f} (feminine)
{m, f} (neuter) ∪ {m, f} (neuter) = {m, f} (neuter)
{m} (masculine) ∪ {f} (feminine) = {m, f} (neuter)
{m} (masculine) ∪ {m, f} (neuter) = {m, f} (neuter)
{f} (feminine) ∪ {m, f} (neuter) = {m, f} (neuter)

Vincent and Börjars (2007) also propose an analysis using sets in the representation
of the value of the gend feature, but object to the gender values that Dalrymple
and Kaplan present, and to the use of set union to model gender resolution. They
observe that Dalrymple and Kaplan’s analysis of gender violates the markedness
criterion discussed in §2.5.1, according to which the formally more marked value
of a feature should behave as the more marked member in an opposition of feature
values. In Icelandic, neuter is the unmarked gender value; however, in Dalrymple
and Kaplan’s system, neuter is formally the most marked, since it is represented as
a set with two members (m and f), while the masculine and feminine values each
contain one member and so are less formally marked.
Vincent and Börjars (2007: 297) propose as a desideratum for the formal
representation of complex feature values that “for whatever morphosyntactic sub-
system we examine, we must be able to show by independent tests that the member
76 functional structure

of that system which is assigned the representation { } acts as the unmarked form.”
They propose the values given in (163) for the gend feature in Icelandic:

(163) Values for the gend feature in Icelandic according to Vincent and Börjars
(2007):
masculine {m}
feminine {f}
neuter { }

Here the unmarked form is the neuter value, represented as the empty set { }, as
desired. As Vincent and Börjars observe, these values allow for a pleasingly simple
generalization about gender resolution in Icelandic and many other languages: for
conjuncts of different genders, agreement is with the unmarked form, represented
as { }. Crucially, however, under these assumptions it is necessary to model gender
resolution by set intersection (∩) rather than set union (∪):

(164) Resolution of the gend feature as set intersection (Vincent and Börjars
2007):
{m} (masculine) ∩ {m} (masculine) = {m} (masculine)
{f} (feminine) ∩ {f} (feminine) = {f} (feminine)
{ } (neuter) ∩ { } (neuter) = { } (neuter)
{m} (masculine) ∩ {f} (feminine) = { } (neuter)
{m} (masculine) ∩ { } (neuter) = { } (neuter)
{f} (feminine) ∩ { } (neuter) = { } (neuter)

As mentioned above and discussed in more detail in §6.3.4, Sadler (2011) demon-
strates that set-based analyses of complex feature values have a ready translation to
feature-based analyses. She provides a feature-based analysis of Icelandic gender
which conforms to Vincent and Börjars’s observations about markedness if we
assume that a negative value for a feature corresponds to the unmarked value for
that feature.

2.5.7.4 Number In much LFG work, the number feature is assumed to have
atomic values such as sg, du, paucal, and pl:

(165) Simple atomic values for the feature num: sg (singular), du (dual), paucal
(paucal), pl (plural)

However, patterns involving constructed number are often taken as evidence that
the value of the num feature is not atomic, but a complex value whose form
may be constrained in different ways by different parts of the sentence. As briefly
discussed in §2.5.1, Sadler (2011) provides an analysis of the constructed dual in
Hopi which assumes a complex value for the num feature. Example (117), repeated
here, illustrates the Hopi patterns:
 functional features 77

(166) a. Pam wari

that.sg run.nonpl
‘S/he ran.’
b. Puma yùutu
that.nonsg run.pl
‘They ran.’
c. Puma wari
that.nonsg run.nonpl
‘They (two) ran.’

Sadler proposes the following complex values for the num feature:

(167) Values for the num feature according to Sadler (2011):

; & '<
sg +
Singular: num
pl −
; & '<
sg +
Dual: num
pl +
; & '<
sg −
Plural: num
pl +

Crucially, according to Sadler’s analysis, number may be partially specified by

different components of the sentence; as long as the partial specifications are
compatible, the sentence is wellformed. For example, the demonstrative puma is
specified as nonsingular, with a [pl +] specification but no specification for the sg
feature; the verb wari is specified as nonplural, with a [sg +] specification but no
specification for the pl feature. This allows the complex value for the num feature
to be constructed by combining the num of the subject and the num of the verb;
in (168c), dual number results from the compatible partial specifications on the
subject and the verb:

(168) a. Pam wari

that.sg run.nonpl
[sg +, pl −] [pl −]
‘S/he ran.’ (singular: [sg +, pl −])
b. Puma yùutu
that.nonsg run.pl
[pl +] [sg −, pl +]
‘They ran.’ (plural: [sg −, pl +])
c. Puma wari
that.nonsg run.nonpl
[pl +] [sg +]
‘They (two) ran.’ (dual: [sg +, pl +])
78 functional structure

Arka (2011, 2012) also proposes a feature-based analysis for constructed number
in Marori, and Jones (2015a,b) proposes a feature-based analysis for the com-
plex number system of Meryam Mir. Interestingly, Nordlinger (2012) presents a
detailed examination of the number system of Murrinh-Patha, arguing that the
complex patterns of number marking in that language are not best analyzed in
terms of a feature-based treatment of the num feature such as Sadler, Arka, and
Jones propose.
We know of no set-based proposals for the value of the number feature, and
indeed it is not clear that it would be possible to formulate an analysis of the
number feature of coordinate noun phrases that could take advantage of the
set-based operations of union or intersection that have been proposed for pers
resolution (§2.5.7.2) or gend resolution (§2.5.7.3). If we assume a set-based rep-
resentation of the num feature, and we assume that the num value for a coordinate
structure is determined by set union or set intersection, as for pers and gend, we
would incorrectly expect a coordinate structure with two singular conjuncts to be
singular: no matter what set we choose to represent the singular value for the num
feature, the union of any set with itself is that set, and the intersection of any set
with itself is also that set. In fact, however, a coordinate structure such as David
and Chris, with singular conjuncts, has plural number and not singular number;
set union and intersection do not provide a straightforward foundation for a theory
of number in coordinate structures. Dalrymple (2012) provides a general overview
of issues related to the treatment of the num feature and the representation of its
value in LFG.
The Bosnian/Croatian/Serbian example discussed in §2.5.7.1 shows that the
same noun phrase can control different values for the num feature on different
agreement targets. Mittendorf and Sadler (2005) present two alternative analyses of
number mismatch in Welsh NPs: one analysis assumes a complex f-structure, and
the other relies on the index/concord distinction and the possibility for the num
feature to be specified differently for index and concord. Hristov (2012) proposes
an analysis of nouns which seem to exhibit both singular and plural properties,
such as British English “company,” which require singular determiners but can
appear with plural verbs (as in the attested example This company are superbly
managed: Hristov 2012: 174). Wechsler (2011) proposes that some agreement
mismatches can be driven by the absence of syntactic agreement features on the
controller (see also Wechsler and Hahm 2011); this is his Agreement Marking
Principle, informally stated in (169):

(169) Agreement Marking Principle (Wechsler 2011: 1009):

Agreement is driven by a syntactic feature of the controller, if the controller
has such a feature. If the controller lacks such a feature, then the target
agreement inflection is semantically interpreted as characterizing the con-
troller denotation.

In support of this view, Wechsler (2011) discusses the French examples in (170), in
which the verb êtes displays syntactic (second person plural) index agreement with
 functional features 79

the pronominal subject vous ‘you’ whether the subject is semantically singular (as
in example 170a, exemplifying the polite plural use of the pronoun) or semantically
plural (as in 170b), while the predicate adjective displays semantic agreement
(singular in 170a, plural in 170b). Wechsler proposes that the French pronoun vous
does not bear concord features, and so there are no syntactic features to control
agreement with the predicate adjective; in this situation, the Agreement Marking
Principle correctly predicts that the adjective agreement features reflect semantic
features of the controller:

(170) a. Vous êtes loyal.

you.pl be.2pl loyal.m.sg
‘You (singular, formal, male) are loyal.’
b. Vous êtes loyaux.
you.pl be.2pl loyal.pl
‘You (plural) are loyal.’

In Bosnian/Croatian/Serbian, as in French, the auxiliary verb shows second person

plural index agreement with the pronoun vi ‘you’. Unlike French, however, the
predicative adjective shows plural and not singular agreement, even when the
subject refers to a single individual:

(171) Vi ste duhovit-i.

you.pl aux.2.pl funny-m.pl
‘You (one formal addressee/multiple addressees) are funny.’

According to Wechsler (2011), this indicates that the Bosnian/Croatian/Serbian

pronoun vi differs from French vous in being specified for concord as well as
index features: the pronoun vi has plural concord, controlling plural agreement
on the predicate adjective.

2.5.7.5 Case A great deal of work in LFG has been done on the morphosyntax
of case. Butt and King (2004c) discuss the kinds of syntactic relations that can
be marked by case: semantic case and quirky case can appear on either terms or
obliques, and structural case and default case appear on terms. Otoguro (2006)
discusses the theoretical status of case, illustrating his discussion with examples
from Icelandic, Hindi-Urdu, and Japanese. For a thorough and wide-ranging
overview of case and its role in grammar, see Butt (2006); for a detailed discussion
of case in an LFG setting, see Butt (2008).
We have seen that features such as index must be treated as nondistributive,
since a coordinate structure has its own value for index features which is deter-
mined by rules referring to the value of the index features of the conjuncts. A
distinguishing characteristic of nondistributive features, including the components
of the index feature, is that for any resolving feature and any combination of
features of the conjuncts, there is a means of determining the resolved value of
the feature for the coordinate structure. As a component of the concord feature
80 functional structure

bundle, the case feature behaves differently, in that there are no resolution rules
for case: that is, there is no sense in which different values for the case feature
systematically combine to determine a resolved case value for the coordinate
structure.
In fact, coordinate structures in which the conjuncts have mismatching case
values are often unacceptable, and this is predicted if we treat case as a distributive
feature, as we discuss in §6.3.2.2. For example, if we assume that the concord
feature is distributive and that subjects must bear nominative case, we predict that
each conjunct in a coordinate subject phrase must specify nominative case in its
concord feature bundle. However, the situation is in fact more complex: Peterson
(2004a) and Hristov (2012: Chapter 5; 2013b) argue that case in English is best
treated as a nondistributive feature, in light of attested examples exhibiting case
mismatch such as . . .[him and I] have this fight, right?, from the British National
Corpus, and Patejuk (2015) provides further discussion of case mismatches in
coordinate structures in Polish. In this book, we provisionally treat concord and
its component case feature as a distributive feature, but further research may
show the need for treating it instead as a nondistributive feature, and analyzing
its behavior in coordination by means of constraints on the relation between
the concord value of the coordinate structure and the concord values of the
conjuncts.
Simple values such as those listed in (172) are often assumed for the case feature:

(172) Simple atomic values for the feature case: nom (nominative), acc
(accusative) erg (ergative), abs (absolutive), gen (genitive), loc (locative),
dat (dative), abl (ablative), ins (instrumental). . .

This simple representation does not allow for a treatment of case indeterminacy
(Groos and van Reimsdijk 1979; Zaenen and Karttunen 1984), where a form with
underspecified case can satisfy more than one case requirement at the same time.
The German nouns Männer/Männern ‘men’ and Papageien ‘parrots’ differ in the
cases that they express:

(173) a. Papageien
‘parrots.nom/acc/dat/gen’
b. Männer
‘men.nom/acc/gen’
c. Männern
‘men.dat’

As noted in §2.5.1, Papageien but not Männer or Männern can appear as the
shared object of coordinated verbs if one verb requires accusative case and the
other requires dative case (174c). Example (174) shows that the verb findet ‘find’
requires an accusative object (174a), while the verb hilft ‘helps’ requires a dative
object (174b) (Pullum and Zwicky 1986):
 functional features 81

(174) a. Er findet Papageien / Männer / ⃰Männern.

he finds parrots men men
obj:acc nom/acc/dat/gen nom/acc/gen dat
‘He finds parrots/men.’
b. Er hilft Papageien / ⃰Männer / Männern.
he helps parrots men men
obj:dat nom/acc/dat/gen nom/acc/gen dat
‘He helps parrots/men.’
c. Er findet und hilft Papageien / ⃰Männer /
he finds and helps parrots men
obj:acc obj:dat nom/acc/dat/gen nom/acc/gen
⃰Männern.
men
dat
‘He finds and helps parrots/⃰men.’

Dalrymple and Kaplan (2000) propose to represent the value of the case feature of
a nominal as a set whose members are the case values that the nominal can express.
A verb taking an accusative object requires acc to be a member of the case set of
its object, and similarly for a verb taking a dat object:40

(175) Set-based analysis of indeterminate case according to Dalrymple and

Kaplan (2000):
Er findet und hilft Papageien.
he finds and helps parrots
acc ∈ obj case dat ∈ obj case [case {nom,acc,dat,gen}]
‘He finds and helps parrots.’

This treatment is adequate for simple cases of indeterminacy, but Dalrymple et al.
(2009) show that it does not make correct predictions for some more complex
cases. They examine the transitivity problem (where a modifier restricts the pos-
sibilities for case expression of its head noun) and the second-order indeterminacy
problem (where a predicate imposes indeterminate requirements on an argument
which may itself be indeterminately specified); see Dalrymple et al. (2009) for
detailed discussion of these issues, and a demonstration that a feature-based
approach with underspecification solves both of these problems. Dalrymple et al.
propose the following representation for the determinate dative case value for a
noun like Männern:

40 The symbol ∈ is the set-membership predicate: a ∈ S means that a is a member of the set S.
See §6.3.
82 functional structure

(176) Determinate dat case for Männern according to Dalrymple et al. (2009):
  
nom −
 
 case 
  acc −  
  gen −   
dat +

Predicates specify a positive value for the case that they require: for example, a
predicate requiring dative case specifies [dat +] for its argument. Indeterminate
nouns like Männer and Papageien are underspecified; they can be assigned a value
of + for one or more case features, as long as the value of the feature is not already
specified as −. For clarity of exposition, we explicitly represent all case features in
(177) and (178), with a blank value for features with no assigned value:

(177) a. Indeterminate nom/acc/gen case for Männer according to Dalrymple

et al. (2009):
  nom 
 acc 
 case   
  gen 
 
dat −
b. Completely indeterminate case specification for Papageien according to
Dalrymple et al. (2009):
  
nom
 acc 
 case  
  gen 
 
dat

(178) Feature-based analysis of indeterminate case according to Dalrymple et al.

(2009):
Er findet und hilft
he finds and helps
obj: [case [acc +]] obj: [case [dat +]]
Papageien.
parrots
  nom 
 acc +  
 case   
  gen 
 
dat +
‘He finds and helps parrots.’

Complex values for the case feature (whether set-based or feature-based) bear a
superficial resemblance to the complex values of the pers, num, and gend features;
importantly, however, they express very different linguistic intuitions, despite their
 functional features 83

formal similarity. For indeterminate features like case, a complex value such as
{nom,acc} or [nom +, acc +] enumerates a set of alternative possibilities which
may be simultaneously realized, allowing an indeterminate form to simultaneously
fulfill conflicting case requirements. In contrast, the specification of the dual value
of the num feature as [sg +, pl +] does not entail that dual nouns are in any
sense simultaneously sg and pl, or alternately sg or pl with respect to different
predicates; instead, complete patterns of values of the pers, num, and gend
features holistically represent the values for these features.

2.5.7.6 Nominal specification and quantification The function spec has been
used in the analysis of articles such as the and a as well as quantifiers and
occasionally possessors, thus grouping together elements which “specify” rather
than modify the head in a nominal phrase. Sadler (2000) and Falk (2002a) note that
treating both possessors and articles as contributing a value of the spec attribute
makes it difficult to analyze languages which allow both an article and a possessor
in the same nominal phrase; Falk (2002a) provides example (179) to show that the
definite article is obligatory in Romanian noun phrases containing a possessor:41

(179) a. cas-a unei vecine

house-the a.gen neighbor
‘a neighbor’s (the) house’
b.⃰casă unei vecine
house a.gen neighbor
‘a neighbor’s house’

The article and the possessor cannot both contribute values for the same spec
attribute in examples like (179): either the article or the possessor must contribute
a value for an attribute other than spec. As discussed in §2.1.10, we adopt the poss
grammatical function for possessors.
Similarly, English noun phrases can contain both an article and a quantifier:

(180) a. the many dogs

b. those few cats

Therefore, we also need separate attributes for the contributions of articles and
quantifiers.
We make use of the features def, deixis, and spec for nominal f-structures. In
English, the value for the feature def is either + or −:

(181) Values for the feature def: + (definite), − (indefinite)

Demonstrative determiners like this and those are definite, contributing the value
+ for the def feature, as well as contributing a deixis feature. The English deictic

41 Falk attributes example (179) to Dobrovie-Sorin (2001).

84 functional structure

system makes only a two way proximal/distal distinction. The deictic systems of
other languages are more complex, including additional values such as medial:

(182) Values for the feature deixis: prox (proximal), medial (medial), dist
(distal) …

There is an important difference between articles like the and quantifiers like every
or all: articles cannot be modified ( ⃰almost the), but quantifiers can be (almost
every). This means that a quantifier must contribute not an atomic value for a
feature, but an f-structure with its own pred attribute and semantic form value,
in which an adj attribute can also appear:

(183) a. every:

b. almost every:

Therefore, the value for the feature spec is an f-structure whose semantic form
represents the quantifier:

(184) Values for the feature spec: f-structures with semantic form ‘all’, ‘every’,
‘both’…

2.5.7.7 prontype The prontype feature marks the type of a pronominal

f-structure. The type of a pronoun is syntactically constrained in a number of
constructions: a complex fronted phrase in a relative clause must contain a relative
pronoun with type rel; the subject of a tag question such as isn’t it? or aren’t they?
in standard English must be a personal pronoun (pers); verbs such as perjure
(oneself) require a nonthematic reflexive pronominal object (refl), and so on. To
account for these requirements, prontype must be marked in the f-structure. We
assume the following values for the feature prontype:

(185) Values for the feature prontype: rel (relative pronoun), wh (interroga-
tive pronoun), pers (personal pronoun), refl (reflexive pronoun), recp
(reciprocal pronoun)

2.5.8 Summary: Functional features

Our discussion has focused primarily on the features that play a role in the
functional syntax of English. Many of these features are also relevant for other
languages, but some languages may make more or fewer distinctions in the values
of these features, and other languages may not make any use of these features
in their functional syntax. It is important to examine the feature system of each
 functional features 85

language carefully in order to determine the inventory of functional features and

values that are appropriate for that language.
We have seen that the features in (186) have simple atomic values:

(186) F-structure features with atomic values:

Feature Value
Form features form Surface word form
(§2.5.3) pform Form of the preposition
compform Form of the complementizer: that,
whether, if…
Prepositional “case” pcase The family of grammatical
(§2.5.4) functions obl𝜃
Verbal and clausal vtype fin, prs.ptcp, pst.ptcp, inf…
features (§2.5.5) mood ind, imp, sbjv, interr…
voice pass, antip, mid, dir, inv, av, ov,
dv, iv, bv…
Nominal features def +, −
(§2.5.7) prontype rel, wh, refl, pers…
deixis prox, medial, dist…

The features in (187) have been argued to have complex values, at least in some
languages. For details of the complex value of the feature, see the indicated section.
As discussed in §2.5.1, we often use atomic abbreviations for these complex values
when the details of the representation of the feature are not important to the issues
under discussion, and it is these abbreviations which appear in (187):

(187) Abbreviations for complex f-structure values:

Feature Abbreviated Value
Verbal and clausal tense prs, pst, fut… (§2.5.5.1)
features aspect simple values: prf, prog, iprf, ipfv,
pfv; complex values are represented
by combining simple values:
prf.prog… (§2.5.5.2)
Nominal features pers 1, 2, 3 (§2.5.7.2)
gend m, f, n… (§2.5.7.3)
num sg, du, pl… (§2.5.7.4)
case nom, acc, dat, erg, abs, gen, abl,
loc, ins, part… (§2.5.7.5)

Features can also be cross-classified in several ways. Section 2.5.7.1 discussed two
kinds of nominal features, index and concord features:

(188) index features: pers, num, gend

concord features: num, gend, case
86 functional structure

The features num and gend appear in both feature bundles, and in fact they
can take on different values in each feature bundle: for example, as shown in
example (148) of this chapter, the Bosnian/Croatian/Serbian noun dobra ‘children’
has singular concord, but plural index.
In §2.4.5, we demonstrated the need to distinguish betwen distributive and
nondistributive features in coordinate structures. We assume that the only nondis-
tributive features are index (specifying the resolved pers, num, and gend features
of coordinate noun phrases), adj (specifying modifiers of coordinate structures
in examples like David yawned and sneezed simultaneously, where simultaneously
modifies the coordinated verbs yawned and sneezed), and conj and preconj
(specifying the form of the conjunction and preconjunction). We return to a
discussion of distributive and nondistributive features in §6.3.2 and Chapter 16:

(189) Nondistributive features: index, adj, conj, preconj

Distributive features: all other features, including pred and the governable
grammatical functions

Example f-structures for a range of simple English nominal phrases are given in
(190). Recall that sg and pl are abbreviations standing for complex values for the
num feature, as discussed in §2.5.7.4, and similarly for the simple abbreviatory
values presented in (190k) for the case, pers, and gend features. We omit
concord features in most of the examples:

(190) a dog:
 
 − 
 
 
 

b. the dog:
 
 + 
 
 
 

c. this dog:
 
 + 
 
 
 
 
 
 functional features 87

d. that dog:
 
 + 
 
 
 
 
 

e. every dog:
 
 
 
 
 
 

f. all dogs:
 
 
 
 
 
 

g. all the dogs:

 
 
 
 
 + 
 
 
 

h. both the dogs:

 
 
 
 
 + 
 
 
 

i. the many dogs:

 
 
 
 
 + 
 
 
 
88 functional structure

j. David’s dog:
 
 
 
 
 
 

k. his dog:
 
 
 
 
 
  
 
 
  
  
  
  
  
   

  
  
   

 

2.6 Further reading and related issues

Within LFG, there has been more discussion of grammatical functions and func-
tional structure than can be summarized in a brief space. Bresnan et al. (2016:
Chapter 6) provide an in-depth discussion of the classification of grammatical
functions. In their lucid summary of LFG theory, Asudeh and Toivonen (2015:
380) provide a useful table of grammatical functions with illustrative examples,
reprinted in Bresnan et al. (2016: 99). Alsina (1996), Przepiórkowski (2017),
and Patejuk and Przepiórkowski (2017) argue for a more restricted inventory of
grammatical functions than is assumed in these works and in this book. Duncan
(2007) discusses the argument-adjunct distinction, as does Rákosi (2006b); for
more on this issue see §9.10.5.2. Andrews (2007a) provides a good overview of
the grammatical functions of nominals. The grammatical function of incorporated
nouns is discussed in Asudeh (2007) and Duncan (2007). On double object con-
structions and the challenges that they present to LFG treatments of objecthood,
see Lam (2008) and Thomas (2012). On the obj/obj𝜃 distinction in Indonesian, see
Musgrave (2008). Allen (2001) analyzes the development of the recipient passive
as involving reanalysis of grammatical functions, from objrecipient and obj in Old
English to obj and objtheme respectively in Modern English. Butt et al. (1999)
provide a general overview of English, French, and German functional and phrasal
structure, and Dipper (2003) provides a computationally-oriented discussion of
the syntactic structure of the German nominal phrase.
 3
Constituent structure

We have seen that there is a large degree of unity in the abstract functional syntactic
structure of languages. In contrast, phrasal structure varies greatly: some languages
allow phrases with no lexical heads, and some have no such categories; some
languages have a VP constituent, and others do not; and so on. In this chapter,
we will discuss the organization of overt phrasal syntactic representation, the
constituent structure or c-structure. We will explore commonalities in constituent
structure across the world’s languages, and talk about how languages can differ
in their phrasal organization. We will also consider the relation between phrasal
syntactic structure and the surface linear order of syntactic units; we consider the
surface linear order to be a separate level of representation, the syntactic string or
s-string.
Section 3.1 begins by discussing some traditional arguments for constituent
structure representation. Many of these arguments turn out to be flawed, since
the theory of phrase structure has a different status in LFG than in theories in
which grammatical functions are defined configurationally and abstract syntac-
tic (and even semantic and information structural) relations are represented in
phrase structure terms. Some arguments for particular phrase structure theories
and configurations are based on phenomena which in LFG are better treated
in other structures, such as f-structure, and do not constitute good arguments
for constituent structure at all. For this reason, we must examine the status of
arguments for and against particular phrase structures or phrase structure theories
particularly carefully. Section 3.2 proposes some valid criteria within LFG for
phrase structure determination.
Research on constituent structure has revealed much about the universally
available set of categories and how they can combine into phrases. We adopt a view
of phrase structure that incorporates insights primarily from the work of Kroeger
(1993), King (1995), Sadler (1997), Sells (1998), Toivonen (2003), and Bresnan
et al. (2016). We assume lexical categories like N and V, as well as functional
categories like I and C. These categories appear as the heads of phrases like NP
and IP; the inventory of constituent structure categories that are crosslinguistically
available and the theory of the organization of words and categories into phrases
are explored in §3.3. The general theory of constituent structure organization is
exemplified in §3.4, where we provide more specific discussion of the constituent
structure organization of clauses and the role of the functional categories I and C in
clausal structure. In §3.5, we discuss the relation between hierarchical constituent
structure and surface linear order.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
90 constituent structure

3.1 Traditional arguments for constituent structure

The sorts of constituent structure rules and representations used in LFG and
most other theories of constituent structure are prefigured in work by Bloomfield
(1933), who assumes complex phrasal structures described in terms of “immediate
constituent analysis”: the combination of words into phrases. These structures
were originally motivated within the transformational tradition by the desire to
formulate a finite characterization of the infinite set of sentences of a language. As
Chomsky (1955: 116) says:

If there were no intervening representations between Sentence and words, the gram-
mar would have to contain a vast (in fact, infinite) number of conversions of the form
Sentence→X, where X is a permissible string of words. However, we find that it is possible
to classify strings of words into phrases in such a way that sentence structure can be stated
in terms of phrase strings, and phrase structure in terms of word strings, in a rather simple
way. Further, a phrase of a given type can be included within a phrase of the same type, so
that a finite number of conversions will generate an infinite number of strings of words.

Chomsky’s first point is that the same sequence of categories may appear in
more than one environment, and any adequate grammar must characterize this
regularity. For instance, the phrase the dachshund can appear in many positions in
a sentence:

(1) a. The dachshund is barking.

b. David petted the dachshund.
c. Matty gave a treat to the dachshund.

No matter where it appears, it can be replaced by a phrase with additional

modifiers:

(2) a. The black dachshund is barking.

b. David petted the black dachshund.
c. Matty gave a treat to the black dachshund.

Generalizations about the structure of the phrase the dachshund and the phrases
that can be substituted for it are captured by assuming that the dachshund is a
noun phrase and can appear wherever other noun phrases can appear. This is
intuitively appealing, since it captures the intuition that the dachshund (and the
black dachshund) are phrasal units in a sense that dachshund is or petted the are
not. Further, it simplifies the linguistic description: the different ways in which
a noun phrase can be formed do not have to be separately enumerated for each
environment in which a noun phrase can appear.
Chomsky’s second point is that a phrase may contain subconstituents of the
same type: a clause can contain a subordinate clause, a noun phrase can contain
other noun phrases, and so on. Our linguistic descriptions therefore need the same
kind of recursive character.
 traditional arguments for constituent structure 91

This formal motivation for a level of constituent structure analysis and repre-
sentation is buttressed by a range of diagnostics for phrase structure constituency.
However, current syntactic theories vary greatly in their criteria for determining
the validity of these diagnostics, and tests that are accepted in some theoretical
frameworks are not recognized as valid in other frameworks. For instance, Chom-
sky (1981) and work growing out of the transformationally based theory presented
there (see, for example, Webelhuth 1995; Rizzi 1997; Chomsky 2001; Adger 2003)
propose that abstract syntactic properties and relations are represented in terms of
phrase structure trees, and that relations between these trees are statable in terms of
the movement of constituents (“move-α,” Move, or Internal Merge). In fact, in such
work not only syntactic relations, but also semantic and information structural
properties and relations, are represented in phrase structure terms. Given such a
theory, the criterion for constituenthood is fairly straightforward: any unit that
plays a role in abstract syntactic structure, semantic structure, or information
structure is a phrase structure constituent, since it must be represented as a unit
and must be eligible to undergo movement.
Such criteria have no relevance in a theory like LFG. We have seen that abstract
syntactic structure is represented by functional structure, not in phrasal terms;
further, phrase structural transformations and movement rules play no role in
LFG theory. We will see in Part II (Chapter 7 to Chapter 12) that semantic and
information structural features, as well as other nonsyntactic aspects of grammar,
are likewise represented by separate structures in LFG. LFG’s constituent structure
trees represent tangible phrasal configurations, not more abstract relations. Thus,
many criteria commonly proposed within the transformational tradition to iden-
tify phrase structure constituents turn out to be irrelevant in an LFG setting.
In this light, it is interesting to carefully examine one detailed proposal for
testing constituency within a transformationally based theory. Radford (2009:
58–69) discusses a set of tests for constituency, including the following:

(3) (i) Coordination: a constituent can be coordinated with another similar

constituent.
(ii) Ellipsis: only (some) constituents can be omitted, under appropriate
discourse conditions.
(iii) Substitution: a constituent can be replaced by, or serve as the antecedent
of, a proform.
(iv) The fragment test: only (some) constituents can serve as sentence frag-
ments (that is, valid free-standing expressions that are not complete
sentences).

Test (i), coordination, proves not to be very successful. It is well known that many
strings that are shown not to be constituents by other criteria can be coordinated:

(4) a. David gave [a flower] [to Chris] and [a book] [to Pat].
b. [David] [likes], and [Chris] [dislikes], carrots.
92 constituent structure

On most theories of the constituent structure of English, the phrases a flower to

Chris and David likes are not constituents. It does seem to be true, at least in
English, that if a string is a constituent it can be coordinated, barring semantic
unacceptability, but the converse implication, that if a string can be coordinated it
is a constituent, does not hold.1
Test (ii), ellipsis, is also relatively unsuccessful. Many constituents cannot be
omitted:

(5) a. Chris told me that David yawned.

b. ⃰Chris told that David yawned.
c. ⃰Told me that David yawned.

At the same time, sequences that are not constituents can be omitted:

(6) Chris can write better plays than David (can write) novels.

Even if, as Radford argues, this example involves two separate omitted constituents,
can and write, the usefulness of this test for determining constituency is seri-
ously undermined. The same problem applies to test (iii), substitution. Many
constituents cannot be replaced by a proform:

(7) a. Chris gave a book [to David].

b. ⃰Chris gave a book him/there.

Conversely, some nonconstituents can be replaced by a proform. In example (8),

the discontinuous phrase two pies . . . eat is replaced by do it, and in example (9),
the antecedents of they are the separate phrases Chris and David:

(8) a. Chris can eat one pie, but two pies he won’t be able to eat.
b. Yes, he will be able to do it.
(9) a. Chris ran into David yesterday.
b. They decided to see a movie.

Likewise, test (iv) is not fully reliable. While sentence fragments usually involve
constituents, there are constituents that cannot appear as sentence fragments (10b),
and nonconstituents that can (10c):

(10) a. Q: What has Chris written?

A: A best-selling novel.
b. Q: What has Chris done?
A: ⃰Has written a best-selling novel.

1 See Chapter 16 for discussion of coordination in general and nonconstituent coordination in

particular.
 evidence for constituent structure 93

c. Q: Who did you see?

A: Chris yesterday, and David today.

Although none of these tests are fully successful (as noted by Radford himself),
some prove to be distinctly less useful than others. This attests to the difficulty of
formulating unequivocal criteria for determining constituent structure units.

3.2 Evidence for constituent structure

Criteria for phrase structure constituency in LFG appeal to the surface syntactic
properties of utterances, not to semantic intuitions or facts about abstract func-
tional syntactic structure. Since these surface properties vary from language to
language, the tests discussed below make reference to properties of particular
languages, and may not be applicable to every language. This is just what we would
expect, since surface form and organization vary from language to language, while
the more abstract functional structure is more uniform.

Verb-second: Certain syntactic generalizations refer specifically to phrase

structure constituents. For instance, in some languages, the position of the
verb provides a test for constituency: the verb appears in second position in
the sentence, after the first constituent. German is such a language; in German
the sequence of words directly preceding the verb in a main clause must be a
single constituent (leaving aside certain rare and clearly identifiable exceptions, cf.
Zaenen and Kaplan 2002):

(11) [Dem Mädchen] schien Hans das Buch zu geben.

the.dat girl seem.pst Hans the.acc book to give.inf
‘Hans seemed to give the girl the book.’

This argument can be repeated for many other verb-second languages, at least
for the majority of data: Kroeger (1993) discusses verb-second phenomena in
Germanic languages and in Tagalog; Zaenen and Kaplan (2002) and Cook and
Payne (2006) discuss the German data in more detail; and Simpson (1991, 2007)
discusses similar phenomena in Warlpiri.

Question formation: A similar argument for constituency comes from the

English constituent question construction. Zwicky (1990) observes that in English,
only a single displaced constituent can appear in clause-initial position:

(12) a. [Which people from California] did you introduce to Tracy?

b. ⃰[Which people from California] [to Tracy] did you introduce?
c. [To how many of your friends] did you introduce people from California?
d. ⃰[People from California] [to how many of your friends] did you introduce?
94 constituent structure

Distribution of adverbs: It is also possible to determine the presence and

distribution of phrases of a particular type. Kaplan and Zaenen (1989b), following
work by Þráinsson (1986), discuss the distribution of adverbs in Icelandic; Sells
(1998, 2001b, 2005) builds on their work, analyzing Icelandic constituent structure
in more detail. In an Icelandic sentence containing an auxiliary or modal verb such
as mun ‘will’, an adverb like sjaldan ‘seldom’ has a restricted distribution:

(13) a. Hann mun sjaldan stinga smjörinu í vasann.

he will seldom put butter.def in pocket.def
‘He will seldom put the butter in the pocket.’
b. ⃰Hann mun stinga sjaldan smjörinu í vasann.
he will put seldom butter.def in pocket.def
c. ⃰Hann mun stinga smjörinu sjaldan í vasann.
he will put butter.def seldom in pocket.def
d. Hann mun stinga smjörinu í vasann sjaldan.
he will put butter.def in pocket.def seldom

In sentences with no modal, the distribution of the adverb is more free:

(14) a. Hann stingur sjaldan smjörinu í vasann.

he puts seldom butter.def in pocket.def
‘He seldom puts the butter in the pocket.’
b. Hann stingur smjörinu sjaldan í vasann.
he puts butter.def seldom in pocket.def
c. Hann stingur smjörinu í vasann sjaldan.
he puts butter.def in pocket.def seldom

Kaplan and Zaenen (1989b) propose that the distribution of the adverb depends on
the presence or absence of a VP constituent: an adverb cannot appear as daughter
of VP. Sentences with an auxiliary verb contain a VP constituent:

(15) Hann mun [stinga smjörinu í vasann]VP .

he will put butter.def in pocket.def
‘He will put the butter in the pocket.’

The adverb can appear as a daughter of S (or I" , in Sells’s 2005 analysis), but not
as a daughter of VP, accounting for the ungrammaticality of examples (13b,c). In
contrast, sentences with no auxiliary verb have no VP, allowing for the wider range
of possible adverb positions shown in (14).

Prosody: A particularly interesting sort of evidence for phrase structure con-

stituency arises from the interaction of constituent structure and prosody. King
(1995: 129–30) notes that a Russian constituent may be signaled as in focus by
 constituent structure organization and relations 95

falling intonation on its right edge; conversely, then, focus intonation may be
viewed as indicating alignment with a right-edge phrasal boundary. Here, the
feature +F represents this right-edge focus intonation, and the subscript focus
marks the focused constituent:2

(16) kolxoz zakončil [uborku urožaja+F ].

kolxoz finished harvest crop
‘The kolxoz finished [the crop harvest]focus .’

A sentence is unambiguously a constituent, so as we would predict an entire

sentence can be put in focus in this way (Junghanns and Zybatow 1997):

(17) Q: What happened?

A: [sgorela ratuša+F ].
burn.down town.hall
‘[The town hall burned down]focus .’

See Zybatow and Mehlhorn (2000) for further discussion of focus intonation in
Russian. While this example clearly demonstrates alignment between prosodic
features such as intonation and syntactic groupings, we assume (with Lahiri and
Plank 2010) that such alignment is not necessary and, indeed, is not often found;
our approach to the formal relation between syntax and prosody will be presented
in Chapter 11.

Clitic placement: Zwicky (1990) proposes other “edge tests,” tests that pick
out the first or last word in a phrase. He argues that the distribution of the
possessive clitic in an English possessive noun phrase is best described by referring
to the right edge of the possessive phrase:

(18) [my friend from Chicago]’s crazy ideas

In this example, the possessor is my friend from Chicago, and the placement of
the clitic ’s shows that my friend from Chicago is a syntactic constituent (and not,
for example, two separate constituents my friend and from Chicago). The syntax of
the English possessive clitic is discussed in more detail by Payne (2009) and Lowe
(2016b).

3.3 Constituent structure organization and relations

LFG assumes a strict version of the Lexical Integrity Principle (§4.4): constituent
structure positions cannot be filled only by affixes or (as Kroeger 1993: 6 puts
it) “disembodied morphological features.” Every terminal node of the constituent

2 King attributes example (16) to Krylova and Khavronina (1988: 80).

96 constituent structure

structure tree corresponds to an individual word. There is no syntactic process

of word assembly, though individual words can make complex syntactic contri-
butions at the functional level (§4.4). We discuss the place of morphology in the
architecture of LFG in Chapter 12; in the following, we discuss the inventory of
constituent structure categories and the organization of constituent structure.

3.3.1 Lexical categories

We assume the following set of major lexical categories:

(19) Major lexical categories:

N(oun), P(reposition), V(erb), A(djective), Adv(erb)

Chomsky (1986) assumes that N, P, V, and A are major lexical categories; foll-
owing Jackendoff (1977), we also assume that Adv is a major lexical category.
These major lexical categories can head phrases of the corresponding category:

(20) a. NP: the boy

b. PP: on the boat
c. VP: sail the boat
d. AP: very fearful of the storm
e. AdvP: quite fearfully

In addition to these major lexical categories, some LFG work assumes a set of
more crosslinguistically variable minor lexical categories. For example, Dalrymple
(2001), Butt et al. (2002a), and Forst et al. (2010) assume the category Part, the
category of English particles such as up in examples like (21):

(21) David called Chris [up]Part .

Similarly, negative particles such as English not are sometimes assigned to a minor
category Neg (see, for example, Sells 2000a on negation in Swedish). Another minor
category is CL, which has been used to represent the category of clitics in some lan-
guages, for example in Romance languages (Grimshaw 1982a; Schwarze 2001a), in
Bosnian/Croatian/Serbian (O’Connor 2002a; Bögel et al. 2010), in Sanskrit (Lowe
2011, 2014a), and in Pashto (Lowe 2016a). Sequences of CL clitics that appear as
single unit (a “clitic cluster”) in a clause are sometimes grouped under a super-
ordinate “CCL” node (Bögel et al. 2010; Lowe 2011, 2014a; Lowe and Belyaev
2015).3
In contrast, some authors assume that there are no minor categories, and that
the inventory of c-structure categories is fixed. For example, Toivonen (2003) and

3 We use the term “clitic” here in purely syntactic terms; words that are treated as “clitics” on purely
prosodic grounds (such as “prosodic deficiency”) are not necessarily “clitics” in syntactic terms, and may
in principle be part of any major lexical or functional c-structure category. See Lowe (2014a, 2016a) for
further discussion of these issues.
 constituent structure organization and relations 97

Bresnan et al. (2016) assume only four lexical categories, V, P, N, and A (treating
adverbs as part of A). Under such assumptions, apparent minor categories must be
analyzed as subtypes of major categories. English particles of the sort seen in (21),
for instance, are all homophonous with prepositions; it is therefore possible to treat
them as belonging to the category P. The c-structure possibilities for English not
partially match those of other English adverbs like never; not may therefore be
analyzed as a member of the category Adv (as in example (47) on page 108).
In fact, it is not necessary to propose a distinct minor c-structure category in
order to account for different c-structure constraints on different subsets of a single
c-structure category (for example, constraints on particles as opposed to preposi-
tions, or the negative not as opposed to other adverbs). The theory of nonprojecting
categories (Toivonen 2003), discussed in §3.3.4, permits a distinction between
members of the same c-structure category that project full phrasal structure and
those that do not. In addition, complex categories, discussed in §3.3.5, enable us
to model differences in c-structure constraints between words of any category that
differ in respect of some morphological or other feature.
In this work, we do not make use of any minor categories; all the words that
we analyze can be assigned to one of the five lexical categories stated in (19),
or else to one of the functional categories introduced in §3.3.3 below.4 In doing
this, however, we make no claim as to the inventory of c-structure categories in
Universal Grammar, and leave open the possibility that minor lexical categories
may exist in some languages.

3.3.2 X" theory

Constituent structure organization obeys the basic principles of X" (X-bar or X̄)
theory (Jackendoff 1977; Chomsky 1986): words appear as heads of phrases and
may be associated with specifier and complement positions within the same phrase.
X" theory allows for general and abstract statements about the organization of
phrases both crosslinguistically and within a language.
Using X as a variable over categories like N, V, or P, the basic principle of X"
theory is that a lexical or functional category X is related to projections of that
phrase, often written as X" (with one “bar level”), X"" (with two “bar levels”), and
so on. Equivalently, we can speak of a lexical or functional category X projecting
the phrases X" and X"" . For example, we can speak of V" as a projection of the
lexical category V, or of V as projecting the category V" . A lexical category such
as V is sometimes written with a superscript 0 to indicate that it has zero “bar
levels,” so that V0 can be used as an alternative way of representing the lexical
category V. However, Toivonen (2003) proposes a distinction between these two
representations in her analysis of nonprojecting words (see §3.3.4), according to
which V and V0 are not equivalent. We adopt Toivonen’s distinction in this work.

4 However, we do assume a non-X" -theoretic category Cnj for conjunctions such as and; we postpone
discussion of conjunctions and coordinate structures to Chapter 16, where we provide a thorough
discussion of the syntax and semantics of coordination.
98 constituent structure

A standard assumption of X" theory is that one of the projections of a category

is a maximal phrase, and is thus usually written as XP. In other words, the category
XP is the maximal projection of the category X. We adopt a simple two-level version
of X" theory in which X"" is the maximal phrase: thus, a phrase of category XP
dominates a nonmaximal projection of category X" which, in turn, dominates a
word of category X0 , for any lexical or functional category X0 .
We further assume that a word of category X0 may be sister to a series of
complement phrases (YP . . . ) and forms a constituent of category X" whose phrasal
head is X0 . The configurations in (22) and (23) illustrate the general X" -theoretic
configuration we assume; languages vary as to where the head of the phrase appears
in relation to the other daughters:

(22) X"

X0 YP ···

The X" node may dominate any number of daughter phrases; we do not assume
that constituent structure trees must be binary branching.
The X" category may be sister to a series of specifier phrases (ZP . . . ) and forms
an XP phrasal constituent with X" as its head:

(23) XP

ZP ··· X

Some languages allow only a single specifier phrase; other languages (for example,
languages like Russian, in which more than one question phrase can appear in
sentence-initial specifier position) allow multiple specifiers.
We also assume that an XP category can directly dominate a single X0 category,
as long as that is the only node it dominates. This permits us to simplify our
representation by omitting intermediate X" nodes that have no sisters and only one
X0 daughter:

(24) XP
X0

Besides these X" -theoretic structures, other permissible phrase structure config-
urations have been proposed. It is generally assumed that it is possible to adjoin
a maximal phrase to either a maximal (XP) or a nonmaximal (X" ) projection,
so-called “Chomsky-adjunction.” In example (25a), YP is adjoined to XP, and in
example (25b), YP is adjoined to X" :

(25) a. XP b. X"

YP XP YP X"
 constituent structure organization and relations 99

Since we do not assume binary branching, it is possible to adjoin more than one
phrase at a time. In example (26), YP1 and YP2 are adjoined to XP:

(26) XP

YP1 YP2 XP

The adjoined phrase is often a modifier. Although the structures seen in examples
(25) and (26) are the most widely accepted structures that involve adjunction, they
are not the only types of adjunction that have been proposed. Another type, which
we accept in this work, is the adjunction of nonmaximal, nonprojecting categories
to other categories, which we discuss in §3.3.4.

3.3.3 Functional categories

Besides the set of lexical categories described in §3.3.1, we assume a set of “func-
tional” phrase structure categories; the terminology is standard but somewhat
confusing, given the existence of functional structure (with a different sense of the
word “functional”) as a separate syntactic level in LFG. In this work, we assume
the functional categories C, I, and D. Other functional categories have also been
proposed in work within LFG, particularly in work on the structure of noun
phrases, and we discuss these briefly in §3.3.3.4. Börjars et al. (1999) provide
further discussion of functional categories and their motivation in LFG.

3.3.3.1 The functional category I As Kroeger (1993) notes, the clausal syntax of
a number of languages that are genetically unrelated and typologically quite differ-
ent nevertheless organizes itself around a finite verbal element, either an auxiliary
or a main verb, which appears in a specified (often second) position; languages
as different as English, German, Warlpiri, and Tagalog all have a distinguished
position in which a finite main or auxiliary verb appears. The position in which
this element appears is called I (originally for INFL, i.e. inflection). The idea that
the functional category I is the head of a finite clause was originally proposed by
Falk (1984) in his LFG-based analysis of the English auxiliary system, and has since
been incorporated, in different ways, into both transformational and nontransfor-
mational analyses of clausal structure (Chomsky 1986, 1993; Pollock 1989; Kroeger
1993; King 1995; Nordlinger 1998; Cinque 1999, 2003, 2004; Bresnan et al. 2016).
Languages can differ as to which lexical categories can fill the I position. In
English, the tensed auxiliary appears in I, but nonauxiliary verbs may not appear
there (see the discussion in §3.4.1.1):

(27) David is yawning.

IP

NP I"

N I VP

David is V

yawning
100 constituent structure

In other languages, such as Russian (King 1995), all finite verbs appear in I, as
shown in example (44) on page 107.

3.3.3.2 The functional category C The functional category C (for complemen-

tizer) was first proposed as the head of CP by Fassi-Fehri (1981), following
unpublished work on English auxiliary inversion by Ken Hale. In English, as shown
in example (28), and in many other languages, the C position can be filled by a
complementizer like that.5 Other elements may also appear in C position: like the
functional category I, the C position may be filled by a verbal element, as in cases
of subject-auxiliary inversion, seen in example (54) on page 111:

(28) David knows that Chris yawned.

IP

NP I"

N VP

David V"

V CP

knows C"

C IP

that NP I"

N VP

Chris V

yawned

In other languages, the rules may differ. King (1995: Chapter 10) provides a
thorough discussion of I and C in Russian, and Sells (1998, 2001b, 2005) discusses
the structure of IP in Icelandic and Swedish.

3.3.3.3 The functional category D Following work by Brame (1982) and Abney
(1987), many researchers have assumed that at least some instances of the category
traditionally labeled “noun phrase” are more accurately treated as a determiner
phrase or DP. According to this theory, the head of a phrase like the boy is the
determiner the, as shown in example (29). This contrasts with the more traditional
structure, shown in example (30), in which the noun is the head:

5 As example (28) shows, we do not assume that a phrase like IP must dominate an I head; we discuss
phrase structure optionality in §3.3.7.
 constituent structure organization and relations 101

(29) the boy as a DP (30) the boy as an NP

DP NP

D" D N"

D NP the N

the N boy
boy
The projection DP was first assumed in an LFG context by Bresnan (1997, 2000)
and by Sadler (1997), who presents an analysis of Welsh pronominal clitics as Ds
within a more complete DP analysis; Börjars (1998) discusses the internal structure
of nominal phrases and the status of D in a range of languages.
DPs are now widely utilized in LFG work on many languages. The internal
structure of nominal phrases is complex, however, varying considerably between
languages. For example, Laczkó (2007) argues that Hungarian NPs can appear as a
complement of D, as in (29), but that (unlike English) a (possessive) DP can appear
as a specifier of NP, between the determiner and the noun. This is shown in (31),
based on Laczkó’s analysis of the Hungarian phrase az én kalapjaim ‘my hats’:6

(31) az én kalap-ja-i-m
the I hat-poss-pl-1sg
‘my hats’
DP

D"

D NP

az
DP N"
the
N
D
kalap-ja-i-m
én hat-poss-pl-1sg
I

In some languages, there is little empirical motivation for a category DP, while in
others there is much evidence. In this work, we will not make detailed proposals
concerning the internal syntactic structure of noun phrases; for simplicity, we will
use the traditional category NP for noun phrases like the boy wherever possible
(as in 30), using DP only where clearly motivated.

6 Laczkó’s (2007) proposal is actually somewhat more complicated, in that he assumes a complex
structure below the n. For simplicity, we omit the details of this structure here.
102 constituent structure

3.3.3.4 Other functional categories Some analyses of noun phrases assume

further levels of complexity and propose additional functional categories besides
DP. We do not make use of any such additional categories in this work, but we
provide here a brief overview of more complex analyses of noun phrases.
The category KP for phrases headed by a case marker K was prefigured in
the seminal work of Fillmore (1968); in an LFG setting, Butt and King (2004c)
propose that Hindi-Urdu has the category KP (see also Davison 1998; Raza and
Ahmed 2011). Butt and King provide the following c-structure for the ergatively
casemarked Hindi-Urdu phrase lar.ke ne ‘boy erg’:

(32) lar.ke ne
boy erg
KP

K"

NP K

N ne
erg
lar.ke
boy

KPs are also assumed for Persian by Nemati (2010). In contrast, Spencer (2005a)
argues that Hindi ne is a postposition, analyzed as a nonprojecting P adjoined to
NP, as shown in (39).
Another functional category sometimes assumed is QP, for phrases headed by
a quantifier Q. Guo et al. (2007) assume a QP in their analysis of Chinese noun
phrases with classifiers, and Spector (2009) assumes a QP for Hebrew nominal
phrases. Wescoat (2007) assumes QP dominating DP for quantified noun phrases
in French:

(33) tout le personnel

all the personnel
QP

Q DP

tout D
all
D NP

le N
the
personnel
personnel
 constituent structure organization and relations 103

Occasionally other functional categories are proposed in order to account for the
complex structure of noun phrases in certain languages. For example, Laczkó
(2007) assumes a NumberP in Hungarian to account for the position of number
words in nominal phrases, and Bögel et al. (2008) and Bögel and Butt (2013)
propose an “Ezafe Phrase,” EzP, to account for the Urdu ezafe-construction, in
which a particle e (the ezafe) connects a modifying noun or adjective with a
preceding head NP. See Laczkó (1995) for an extensive discussion of the functional
and phrasal syntax of noun phrases within the LFG framework.

3.3.4 Nonprojecting categories

All the rules of constituent structure organization that we have discussed so
far involve categories that project in a standard X" -theoretic way. There is also
evidence, however, that some words can be “nonprojecting.” According to Sadler
and Arnold (1994), lexical categories like A, which usually project full phrasal
categories, can also appear as “small” or X0 categories, adjoined to other X0
categories. Sadler and Arnold call these adjoined structures “small” constructions,
building on work by Poser (1992).7 They argue that English prenominal adjectives
participate in these constructions, proposing the following structure for the phrase
an extremely happy person:

(34) an extremely happy person according to Sadler and Arnold (1994):

NP

D N"

an N0

A0 N0

Adv0 A0 person
extremely happy

Treating extremely happy as a zero-level category A0 explains why example (35a)

is impossible while examples (35b) and (35c) are grammatical: English happy can
take an IP complement (such as to have finished) only when it heads an AP, which
it does in all contexts except prenominal position. That is, prenominal modifiers
cannot be full AP phrases:8

7 A “small” construction in Sadler and Arnold’s sense is different from a “small clause” (Stowell 1981),
a tenseless phrase consisting of a subject and a predicate.
8 There is a highly restricted set of exceptions to this generalization, which are not accounted for by
the structure proposed here. As noted by Arnold and Sadler (2013: 65), a restricted set of prenominal
adjectives can take phrasal complements, as long as those complements do not contain referential or
quantificational nouns. Examples include the eager to please student, a larger than average house, and a
difficult to solve problem. In some cases, these adjectival phrases have become conventionalized and
are treated as units in the syntax, but this cannot explain all such examples. A more wide-ranging
104 constituent structure

(35) a. ⃰an extremely happy to have finished person

b. She is extremely happy to have finished.
c. She came in, extremely happy to have finished.

In subsequent work, Toivonen (2003) proposes that some words may be inher-
ently nonprojecting; that is, they never project full phrasal categories. Such non-
projecting words may be of any lexical category, and they adjoin to an X0 word just
like Sadler and Arnold’s “small” categories. Toivonen refers to projecting words as
X0 in the usual way, but distinguishes nonprojecting words with the notation ! X
(following Asudeh 2002c). The plain label X is used by Toivonen as a cover term
for X0 or !
X. Toivonen (2003: 2) proposes the structure in (36) for Swedish particle
verbs (ihjäl ‘to death’ is a particle):

(36) Eric har slagit ihjäl ormen

Eric has beaten to.death snake.def
‘Eric has beaten the snake to death.’
IP

NP I"

N0 I0 VP

Eric har V"

Eric has
V0 NP

V0 P
! N0

slagit ihjäl ormen

beaten to.death snake.def

Nonprojecting words are found in a variety of unrelated languages. Sells (1994)

shows that nonprojecting words play an important role in the phrasal syntax of
Korean, and nonprojecting words are also found in Zapotec (Broadwell 2007) and
Welsh (Mittendorf and Sadler 2011). Lowe (2015d) shows that non-projecting
words can provide an insightful account of compounding forms in Sanskrit. It is
important to note that this proposal somewhat changes the nature of the category
X0 : it is not restricted purely to single words, but can also refer to phrases consisting
of more than one word, such as extremely happy in (34) or slagit ihjäl in (36). We
adopt Toivonen’s terminology and notation in this work.

exception to the generalization made here is that almost any type of phrase can be used as a prenominal
modifier, for example I’m having one of those I’m-so-sick-of-this-bloody-job-that-I-could-scream days
(Arnold and Sadler 2013: 65 ex. 93). The particular intonation required with such examples suggests
that they require a separate account. Bresnan and Mchombo (1995: 194) propose that phrases can be
innovatively lexicalized, and thus used as if single words.
 constituent structure organization and relations 105

Xs may adjoin to a single X0 , and permits

Toivonen (2003) assumes that multiple !
adjunction by a nonprojecting category only to a projecting X0 category:

(37) X0

X0 !
Y ···

Arnold and Sadler (2013) argue that nonprojecting categories must also be permit-
ted to adjoin to other nonprojecting categories, as in the following configuration:

(38) X
!

!
X !
Y

Spencer (2005a) argues that nonprojecting categories may also adjoin to full
phrasal categories; for example, he proposes that Hindi-Urdu postpositions are
nonprojecting !
Ps that adjoin to NP. His proposal implies the following analysis
of the Hindi-Urdu phrase lar.ke ne ‘boy erg’ (for a contrasting analysis, compare
example (32) on page 102):

(39) lar.ke ne
boy erg
NP

NP P
!

N ne
erg
lar.ke
boy

We do not assume that there are any necessary restrictions on what categories
nonprojecting words may adjoin to.
From now on in this book, we use X as an abbreviation for X0 ; nonprojecting
words are always referred to as !
X.

3.3.5 Complex categories

In any language, the inventory of forms that make up a particular c-structure
category includes sets of forms with differing features. For example, in English both
finite verb forms and nonfinite verb forms such as participles and infinitives are
part of the same c-structure category V. While such c-structure categories permit
us to make valid generalizations over the varied sets of forms they contain, it is
sometimes useful to be able to make c-structure generalizations over only some
members of a category. For example, some c-structure rules may apply not to all
members of the category V, but only to those that are finite.
106 constituent structure

Complex c-structure categories provide a way of stating fine-grained constraints

on categories through the use of c-structure parameters. For example, Frank
and Zaenen (2002) propose that the VP category in French is parametrized for
finiteness, and that the V category is parametrized for finiteness and verb type.
The parameters of a category are written in square brackets after the category label:
for example, VP[fin] is a VP with the parameter for finiteness set to fin, for finite.
Frank and Zaenen (2002) assume a non-standard, flat structure for the French VP,
providing the following sample configuration:

(40) Frank and Zaenen (2002: 51):

VP[fin]

V[aux,fin] (V[aux,part]) V[main,part]

In this configuration, a finite VP (VP[fin]) dominates a finite verbal auxiliary

(V[aux,fin]), an optional participial auxiliary (V[aux,part]), and a participial main
verb (V[main,part]). On this view, the category of a verb must include the parameter
specifications appropriate for its form. For example, the French finite auxiliary verb
est has the c-structure category V[aux,fin]:

(41) est: V[aux,fin]

As we show in §5.1.4, this pattern can be generalized to other values for finiteness
by allowing the parameters of the rule to be instantiated in different ways, placing
fine-grained constraints on how different subtypes of categories can be realized.
Falk (2003, 2008) builds on Frank and Zaenen’s proposals in his analysis of the
English auxiliary system. Complex categories are also discussed by Kuhn (1999)
and Crouch et al. (2008).

3.3.6 Categorial inventory

The inventory of phrasal categories may vary from language to language; we do not
assume that a phrasal category exists in a language unless there is direct evidence
for it. Criteria for determining the presence of a phrasal category differ according
to whether the phrase is a projection of a lexical or a functional category.
The existence of a projection of a lexical category in a language must in the
first instance be motivated by the presence in the language of some word of that
lexical category (King 1995). That is, for example, the existence of a VP phrase
in the constituent structure of a language implies that there are lexical items of
category V in that language. If a lexical category does not appear in a language, its
corresponding phrasal projection does not appear either.
Even when there is evidence for a lexical category in a language, the corre-
sponding phrasal category may in some cases not appear. For instance, although
some Warlpiri words are of category V, Simpson (1991) shows that there is strong
evidence against the existence of VP in Warlpiri. Any single XP constituent may
appear before the Warlpiri auxiliary:

(42) [watiya-rlu wiri-ngki] ji paka-rnu

stick-erg big-erg aux hit-pst
‘He hit me with a big stick.’
 constituent structure organization and relations 107

As Simpson points out, if the V formed a VP constituent with the object phrase,
we would expect example (43) to be grammatical; in fact, however, it is not
grammatical:

(43) ⃰[wawirri] [panti-rni] ka ngarrka-ngku

kangaroo.abs spear-npst aux man-erg
‘The man speared the kangaroo.’

Simpson (1991) presents a number of other arguments to show that Warlpiri has
no VP, although it does have a V" constituent.
In the case of functional categories, evidence is often less directly available;
here too, however, categories are only presumed to exist if they are motivated by
direct evidence. For instance, Laczkó (2016) argues that there is no evidence in
Hungarian for the existence of the category IP or I.

3.3.7 Optionality of constituent structure positions

In the theory of constituent structure proposed by Chomsky (1986), heads of
phrases are obligatorily present, complements (sisters to X categories) are present
according to predicate valence, and specifier positions (sisters to X" categories)
are optional. In contrast to this view, we do not assume that any phrase structure
position is crosslinguistically obligatory, including the head and complements. As
discussed in §2.3, subcategorization requirements are most appropriately specified
at the level of f-structure, and so there is no necessity for predicate valence to be
reflected in c-structure representation. And since heads can appear outside their
phrases (cases of so-called “head movement”: see §4.3.1 and Zaenen and Kaplan
1995, King 1995, and Nordlinger 1998), the position of the head must also be
optional.
The following Russian example illustrates the optionality of a number of con-
stituents (King 1995: 172):

(44) kogda rodilsja Lermontov?

when born Lermontov
‘When was Lermontov born?’
CP

AdvP C"

Adv IP

kogda I"
when
I VP

rodilsja NP
born
N

Lermontov
Lermontov
108 constituent structure

As the tree in (44) shows, there is no specifier position of IP if there is no topicalized

or focused non-interrogative constituent (King 1995: 172). Additionally, the tree
illustrates “headless” constructions; the VP constituent does not dominate a V
node, since the tensed verb in Russian appears in I; also, the CP does not dominate
a C. For more discussion of phrasal categories that do not contain a lexical head,
see Zaenen and Kaplan (1995).
Some LFG researchers, including Kroeger (1993), King (1995), and Bresnan
et al. (2016), make the strong claim that all constituent structure positions in
all languages are optional. However, Snijders (2012, 2015) demonstrates that,
while optionality is a strong crosslinguistic tendency, it is not without exceptions,
and certain positions are obligatory in certain languages. In particular, Snijders
discusses constraints on optionality in Latin, showing that the NP complement in
a PP is obligatory, even though all elements of an NP are themselves optional within
the NP. That is, some portion of the NP complement, not necessarily including the
head, must appear adjacent to the P.

3.4 Clausal organization

3.4.1 IP and CP
In many languages, IP corresponds to a sentence, and CP corresponds to what was
called S" in early transformationalist literature, a sentence with a complementizer
or a displaced phrase in sentence-initial position. Here we examine some basic
structures in English and draw some contrasts with constituent structures in
other languages. Of course, no conclusions for the structure of other languages
should be drawn from the organization of English: LFG does not claim that the
phrasal organization of every language is the same. Since there is a great deal of
crosslinguistic typological variation in constituent structure organization, what is
true for English cannot be assumed to be true in other languages. We must carefully
examine each language on its own terms to determine its phrasal structure and
organization.

3.4.1.1 The IP phrase In English, the tensed auxiliary verb appears in I, and the
rest of the verb complex appears inside the VP. Evidence that the English VP forms
a constituent comes from the fact that, like other constituents, it can be preposed
for information structure reasons:

(45) a. David wanted to win the prize, and [win the prize] he will.
b. . . . there is no greater deed than to die for Iran. And [dying] they are, . . .
(from Ward 1990)

It is not possible to prepose only the auxiliary and verb, since they do not form a
constituent:

(46) ⃰David wanted to win the prize, and [will win] he the prize.
 clausal organization 109

Since the sentential negation morpheme not must be preceded by a tensed auxiliary
" right-adjoined to the tensed
verb, we assume that it is a nonprojecting adverb (Adv)
verb in I:

(47) David is not yawning.

IP

NP I"

N I VP

David I "
Adv V

is not yawning

It is not possible in (modern) English for the negation marker to follow a nonaux-
iliary verb (in contrast with French; see Pollock 1989):

(48) ⃰David yawned not.

Thus, in English only a tensed auxiliary verb can appear in I, and main verbs,
whether tensed or nontensed, appear inside the VP. When there is no auxiliary
verb, the I position is not filled and the IP is “headless”:

(49) David yawned.

IP

NP I"

N VP

David V

yawned

Nontensed auxiliaries appear in V and not the I position in English:

(50) David has been yawning.

IP

NP I"

N I VP

David has V"

V VP

been V

yawning
110 constituent structure

In contrast to English, only nontensed verbs appear within the VP in Russian. King
(1995) provides several pieces of evidence supporting this claim. A VP constituent
that contains a nontensed verb can be scrambled to a higher clause in colloquial
Russian:9

(51) mne [otpustit’ Katju odnu] kažetsja, čto bylo by bezumiem.

me let.go.inf Katja alone seem that would be insane
‘It seems to me that it would be insane to allow Katja to go alone.’

A similar construction involving a tensed verb is not grammatical, since the tensed
verb is not a constituent of VP:

(52) ⃰ja [pošel v školu] skazal, (čto) on.

I went to school said that he
‘I said that he had gone to school.’

King concludes from this and other evidence, based on coordination and the
distribution of the negative marker ne, that nontensed verbs are of category V in
Russian. In contrast, all tensed verbs in Russian appear in I position (King 1995),
as shown in example (53):

(53) prislal muž den’gi.

sent husband money
‘My husband sent (me) money.’
IP

I"

I VP

prislal NP V"
sent
N NP

muž N
husband
den’gi
money

9 King attributes examples (51)–(52) to Yadroff (1992).

 clausal organization 111

3.4.1.2 The CP phrase In English questions, the auxiliary verb appears in C:

(54) Is David yawning?

CP

C"

C IP

Is NP I"

N VP

David V

yawning

Thus, a tensed English auxiliary appears in C in constructions involving subject-

auxiliary inversion, and in I otherwise. Constraints involving the f-structure of
constructions requiring or forbidding subject-auxiliary inversion ensure that the
auxiliary appears in the proper position in each instance. The question word
appears in the specifier position of CP in English, with the auxiliary verb in C
(King 1995: Chapter 10):

(55) What is David eating?

CP

NP C"

N C IP

What is NP I"

N VP

David V

eating

Many languages are unlike English in configurational structure. We now examine

some of this variability, and discuss how languages can vary within the limits
imposed by X" theory and the universally available mappings between c-structure
and f-structure which we examine in Chapter 4.

3.4.2 Exocentricity and endocentricity

We have seen that the constituent structure of English abides by X" -theoretic
principles: a head of category X can project a nonmaximal phrase X" and a
112 constituent structure

maximal phrase XP of the same category. Headed categories like XP and X" are
called endocentric (Bloomfield 1933), and the tendency for languages to make
use of endocentric categories and for grammatical functions to be determined by
configurational syntactic relations is what Bresnan et al. (2016) call the principle of
endocentricity.
In contrast, some languages allow an exocentric category, one that has no
head: the category S (Bresnan 1982a; Kroeger 1993; Austin and Bresnan 1996;
Nordlinger 1998). According to Bresnan et al. (2016), languages making use of
the category S are lexocentric, meaning that syntactic functions are determined
by features associated with words rather than by configurational structure. S is a
constituent structure category that can contain a predicate together with any or all
of its arguments and modifiers, including the subject; for this reason, Austin and
Bresnan (1996) call languages with the category S, including Tagalog, Hungarian,
Malayalam, and Warlpiri, “internal subject” languages. As an exocentric category,
the category S can dominate a series of either lexical or phrasal constituents. Some
authors permit the category S even in thoroughly endocentric languages such as
English (Bresnan et al. 2016), but we restrict its use to those languages which show
clear evidence for exocentric structure.
According to Austin and Bresnan (1996), the phrase structure of a simple
Warlpiri sentence is as follows:10

(56) kurdu-jarra-rlu kapala maliki wajilipi-nyi wita-jarra-rlu

child-du-erg prs dog.abs chase-npst small-du-erg
‘The two small children are chasing the dog.’
IP

NP I"

N I S

kurdu-jarra-rlu kapala NP V NP
child-du-erg prs
N wajilipi-nyi N
chase-npst
maliki wita-jarra-rlu
dog-abs small-du-erg

The Warlpiri I" consists of an I, in which the clitic auxiliary appears, and an
S complement. A phrasal constituent of any category appears as the specifier
daughter of IP; in (56), a noun phrase fills the specifier of IP. Importantly, there
are no syntactic constraints on the order of words in the sentence, so long as the
auxiliary appears in second position (Simpson 1991; Austin and Bresnan 1996).11

10 Legate (2002), Laughren (2002), and Simpson (2007) show that many Warlpiri sentences are in
fact more complicated than this, insofar as various projections exist above the IP, but this does not affect
the phrasal configuration below the IP.
11 The order of words is highly constrained according to other, namely information structural,
factors (Simpson 2007).
 clausal organization 113

The daughters of S can in principle appear in any order, and no element obligatorily
appears as daughter of S, since any phrasal constituent can appear in the specifier
of IP position.
In Chapter 4, we discuss the close correlation between specifier and complement
constituent structure positions as defined by X" theory and the grammatical
functions of the phrases that appear in those positions. For example, in many
languages the subject must appear in the specifier position of IP. In contrast, the
relation between constituent structure positions dominated by S and their abstract
functional role does not obey the same X" -theoretic constraints. Thus, in languages
which, like Warlpiri, make use of the exocentric category S, constituent structure
position is often not an indicator of grammatical function. Instead, grammatical
functions are often marked morphologically, by means of case endings, with a
concomitant tendency to freer word order.
The category S has also been proposed for languages that have relatively fixed
word order. Sadler (1997) proposes the following clausal structure for Welsh, a
VSO language:

(57) Gwnaeth hi weld y draig.

do.3sg.pst 3sg.f see the dragon
‘She saw the dragon.’
IP

I S

Gwnaeth NP VP
do.3sg.pst
N V NP

hi weld
y draig
3sg.f see
the dragon

On Sadler’s analysis, the tensed verb in Welsh is always of category I. The comple-
ment of I is S; S dominates the subject NP and a predicate XP, which may be a VP,
as in example (57), or a maximal phrase of different category, for example NP or
AP in copular constructions (on copular constructions, see §5.4.5). Similar clausal
structures are assumed by Kroeger (1993) and Asudeh (2012: 157–63) for Irish,
and by Falk (2001b: 51–2) for Hebrew.
A further functional category E (for Expression or External-Topic), originally
proposed by Banfield (1973), has been proposed in analyses of left-dislocated
phrases, such as external topics that are intonationally and syntactically separated
from the rest of the clause. In (58), David is an external topic (and not a vocative),
and the syntactic separation of David from the rest of the sentence is clear from
the fact that a resumptive pronoun appears and a gap is not allowed:

(58) (As for) David, what did he/ ⃰ ∅ do today?

114 constituent structure

Within LFG, the category E is used by King (1995), building on work by Aissen
(1992), in her analysis of Russian c-structure, and also by Jaeger and Gerassimova
(2002) for Bulgarian, Simpson (2007) for Warlpiri, and Lowe (2012) for Sanskrit.
However, it is often possible to account for external topics and similarly positioned
phrases by assuming adjunction to a clausal node, without recourse to an addi-
tional phrase-structure category. In example (58), for example, the NP David may
be analyzed as adjoined to CP. We have no need for E in this work, and so will make
no use of it.

3.5 The s-string

As we have seen, constituent structure represents hierarchical phrasal syntactic
relations between words. This hierarchy is related, of course, to the linear order of
those words, but is separate from it. So, in example (57), the word hi ‘3sg.f’ is more
closely related to draig ‘dragon’ than to gwnaeth ‘do.3sg.pst’ in hierarchical terms
(since both hi and draig are part of the same S constituent), even though in linear
terms hi is adjacent to gwnaeth but three words distant from draig. However, some
syntactic phenomena appear to make reference to linear relations rather than, or
in addition to, hierarchical ones.
In the architecture proposed by Kaplan (1987, 1995), constituent structure is
projected from the string, the linearly ordered sequence of words that we are famil-
iar with when we speak or write anything, one word after another. Dalrymple and
Mycock (2011) propose a more detailed model of the string, whereby it consists
of two parts or “sides”: the s-string, representing the string of syntactic units, and
the p-string, representing the string of phonological units that correspond to the s-
string. The p-string, and the architectural assumptions underlying this conception
of the string, are discussed in more detail in Chapter 11; for present purposes only
the s-string is relevant, as a level of syntactic representation encoding linear order
between syntactic units. We can therefore represent the s-string for example (57)
as follows:

(59) gwnaeth hi weld y draig

This may appear trivial, but it is important for several reasons. Lowe (2016b)
adapts the theory of “Lexical Sharing” developed by Wescoat (2002, 2005, 2007,
2009), according to which it is possible for single items in the s-string (Wescoat’s
“linear structure”) to be associated with two terminal nodes in the c-structure. In
French, for example, some sequences of preposition followed by determiner are,
unremarkably, sequences of two separate words, but certain other sequences of
preposition followed by determiner are single words:

(60) a. à la fille
to the girl
b. ⃰à le garçon
to the boy
 the s-string 115

c. au garçon
to.the boy

As Wescoat points out, forms like au cannot be derived by regular phonological

rules from two-word sequences; there are no synchronic phonological processes
in French that would produce au ([o]) from à le ([a lə]). Lexical Sharing permits
a consistent representation of all preposition-determiner sequences in French (we
use the formulation of Lowe 2016b, rather than that of Wescoat):

(61) à la fille (62) au garçon

to the girl to.the boy
‘to the girl’ ‘to the boy’
PP PP

P DP P DP

D NP D NP

N N

à la fille au garçon
to the girl to.the boy

The arrows between the s-string and the c-structure represent the fact that these
are not constituency relations, but a “projection” from one syntactic level (the
s-string, encoding linear order) to another (the c-structure, encoding hierar-
chical relations). This is compatible with Carnie’s (2012b: 93) observation that
the relation between words (s-string units) and the c-structure is different from
the relation between c-structure nodes.12 The s-string and its relation to c-structure
are discussed in more detail in Chapter 11.
Lexical Sharing is also assumed by Broadwell (2007, 2008) in his analyses of
Zapotec and Turkish, by Alsina (2010) in his analysis of Catalan, by Belyaev
(2014) in his analysis of Ossetic, and by Lowe (2015d) in his analysis of Sanskrit
compounds. Mismatches between the s-string and c-structure may also be relevant
to the phenomenon of “prosodic inversion,” examined by Halpern (1995) and, in
an LFG setting, by Kroeger (1993), Newman (1996), Austin and Bresnan (1996),
Nordlinger (1998), Bögel et al. (2010), Bögel (2010), and Lowe (2011). Lowe
(2016a), building on a suggestion by Dalrymple and Mycock (2011), argues that the
observed distinction between what Halpern calls “surface order” (the order of the
words in a sentence) and “syntactic order” (their order in the constituent structure
tree) is best captured as a difference in the order of items between the s-string and
c-structure, rather than a distinction between syntax and prosody, as is assumed

12 We do not adopt Carnie’s convention of representing this difference by omitting the line between
the s-string unit and the c-structure.
116 constituent structure

in many previous works. Other authors who investigate different aspects and uses
of the s-string include Mycock (2006) and Asudeh (2006, 2009).
Like other linguistic levels, the s-string is subject to wellformedness constraints,
which may differ on a language-by-language basis. These constraints will be
discussed in more detail in §6.11 and Chapter 11. In §11.4.1, we will see that a
considerably more refined treatment of the string enables us to analyze interactions
involving syntax, prosody, and other aspects of the linguistic structure of an utter-
ance. For certain phenomena, then, the s-string, as a separate syntactic component,
is an essential component of analysis. Nevertheless, in the vast majority of cases,
as in examples (57) and (61), the relation between the s-string and c-structure
is straightforward. In the following chapters, we explicitly represent the s-string
as separate from the c-structure only where necessary; otherwise, we assume the
separate s-string and the arrows connecting s-string units to terminal c-structure
nodes only implicitly, and we maintain the simpler representation (which does not
include arrows), as illustrated in the previous sections of this chapter.

3.6 Further reading and related issues

Bresnan et al. (2016: Chapter 6) provide an alternative theory of constituent struc-
ture categories involving a featural decomposition of category types that define
lexical and functional categories, treating natural classes of categories in terms of
a set of basic features “predicative,” “transitive,” and “functional.” Marcotte (2014)
and Dalrymple (2017) also explore the decomposition of c-structure categories
into sets of primitive features. Otoguro (2015) discusses the morphological basis
for the classification of verbs as V or I. Marcotte (2014) and Lovestrand and Lowe
(2017) explore alternative formalizations of phrase structure which preserve many
of the intuitions of X" theory (§3.3.2) but require fewer non-branching nodes.
Besides the work we have discussed so far, the following are of note on con-
stituent structure issues in different languages: Börjars et al. (2003) on Swedish;
Berman (2003) on German; Sells (2005) on Icelandic; King (2005) on the treatment
of clitics; Dipper (2005) on German quantifiers; Bresnan (1997), Seiss (2008), and
Bresnan and Mugane (2006) on the treatment of so-called “mixed categories” such
as the English gerund; Lødrup (2011b) on Norwegian possessive pronouns; and
Cavar and Seiss (2011) on Bosnian/Croatian/Serbian clitics.
 4
Syntactic correspondences

LFG hypothesizes that constituent structure and functional structure are mutually
constraining structures and that the relation between these structures is governed
by constraints associated with words and phrasal configurations. We turn now to
an investigation of the relation between these two facets of linguistic structure: in
this chapter, we explore universally valid generalizations regarding the correlation
between phrasal positions and grammatical functions.
Section 4.1 discusses the formal representation of the relation between c-
structure and f-structure. Section 4.2 explores the relation between c-structure and
f-structure: how c-structure phrases and their heads relate to f-structure, and the
c-structure/f-structure realization of arguments and modifiers. Next, we exam-
ine apparent mismatches between units at c-structure and those at f-structure;
Section 4.3 shows that these cases have a natural explanation within LFG. In
§4.4, we discuss the Lexical Integrity Principle, the concept of wordhood, and the
possibly complex contribution of words to functional structure, and in §4.5, we
discuss the principle of Economy of Expression.

4.1 Relating configurational and functional structure

As aspects of a single complex linguistic structure, c-structure and f-structure
are related to one another in a finely specified way: the pieces of the c-structure
that constitute the subject are related to the f-structure for the subj, for example.
Formally, the relation between the two structures is given by a function1 called 𝜙
that relates c-structure nodes to f-structures. Each c-structure node is related by
this function to a particular f-structure, and since 𝜙 is a function, no c-structure
node can be related to more than one f-structure. Pictorially, the 𝜙 function is
represented by an arrow, which can be labeled 𝜙, from the c-structure node to its
corresponding f-structure:

(1) V
’
yawned

As this diagram shows, the word yawned is of category V, and the V node is associ-
ated with certain functional syntactic information: the f-structure corresponding

1 Chapter 2, Footnote 19 provides a definition of the term “function.”

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
118 syntactic correspondences

to the V node has a pred with value ‘yawn!subj"’, and the attribute tense with
value pst.
Often, more than one c-structure node is related to the same f-structure.
For instance, a phrase and its head correspond to the same f-structure:

(2) VP
’
V

yawned

There are also f-structures that are not related to any c-structure node. In
Japanese (Kameyama 1985), a “pro-drop” language, the single word kowareta
‘broke’ can appear without an overt subject noun phrase. In such cases, the subj
f-structure is interpreted pronominally, and does not correspond to any node in the
c-structure:

(3) koware-ta
break-pst
‘[It/Something] broke.’
S  
’
V  
 
kowareta

4.2 Regularities in the c-structure/f-structure mapping

The mapping function from c-structure to f-structure obeys universal principles:
a phrase and its head always correspond to the same f-structure, for example, and
phrase structure complements play particular grammatical roles. In this section,
we examine these regularities and how they constrain the relation between the two
syntactic structures of LFG.

4.2.1 Heads
The functional properties and requirements of the head of a phrase are inherited
by its phrasal projections and become the functional properties and requirements
of the phrases projected by the head. This means that a constituent structure head
and the phrases it projects are mapped onto the same f-structure, as shown in (2).
This condition was originally proposed by Bresnan (1982a: 296) and discussed by
Zaenen (1983), who calls it the Head Convention.
regularities in the c-structure/f-structure mapping 119

4.2.2 Complements
4.2.2.1 Complements of functional categories Recall that complement posi-
tions are sisters to the head of a phrase. Complements of functional categories
are f-structure co-heads (Kroeger 1993; King 1995; Bresnan et al. 2016), meaning
that a functional category shares its functional properties with its complement.
For example, the English IP shares the functional syntactic properties of its VP
complement. Thus, an f-structure can be associated with two different c-structure
heads, one a functional category and the other a lexical category, as shown in the
English example in (4) and the Russian example in (5) (from King 1995: 227):

(4) David is yawning.

 
’
 
 
IP  
 
NP I

N I VP

David is V

yawning

(5) Anna budet čitat’ knigu.

Anna fut read.inf book
‘Anna will be reading a book.’
IP
 
NP I
’
 
 
N I VP  
 
 
Anna budet V  
 
Anna
V NP

čitat’ N

knigu
book

4.2.2.2 Complements of lexical categories Bresnan et al. (2016) claim that

complements of lexical categories can bear any of the governable grammatical
functions except for subj. In the following English example, the obj phrase Chris
is a sister to V:
120 syntactic correspondences

(6) David greeted Chris.

IP  
’
NP I  
 
 
N VP

David V

V NP

greeted N

Chris

Of course, more than one complement may appear; in example (7), the first
complement bears the obj role, and the second complement is objtheme :

(7) David gave Chris a book.

IP

NP I  
’
N VP
 
 
 
 
David V  
 
 
 − 
V NP NP

gave N D N

Chris a N

book

The complement of a lexical category can also correspond to the same f-structure
as the head of the phrase, as in the case of English nonfinite auxiliaries and their
complement VPs (Bresnan et al. 2016: Chapter 6):

(8) David has been yawning.

IP

NP I
 
’
N I VP
 
 
David has V 



V VP

been V

yawning

Falk (1984, 2001b), Butt et al. (1996b), Sadler (1998), and Bresnan et al. (2016)
provide further discussion of the English auxiliary system.
regularities in the c-structure/f-structure mapping 121

According to Laczkó (2014a), Hungarian is an exception to Bresnan et al.’s

(2016) claim that only nonsubject arguments can appear in complement positions.
In Hungarian, a subject can appear postverbally, and its position is not fixed; it can
appear in any order with other postverbal arguments:

(9) a. Szereti János Mari-t.

loves John.nom Mary-acc
‘John loves Mary.’
b. Szereti Mari-t János.
loves Mary-acc John.nom
‘John loves Mary.’

As É. Kiss (2002) and others have observed, the different orders of postverbal
arguments in Hungarian are essentially equivalent in terms of interpretation.
Laczkó (2014a) proposes that the subject can appear within V$ in Hungarian,
as a complement of the lexical category V, in the same position as postverbal
complements bearing other governable grammatical functions.
Since LFG does not define subcategorization properties of predicates in c-
structure terms, the arguments of a predicate may but are not required to appear in
complement positions. They may, for example, appear in the specifier position of
some projection, and are linked to their within-clause function by means of purely
functional specifications. Resolution of long-distance dependencies is discussed
briefly in the next section, and in more detail in Chapter 17.

4.2.3 Specifiers
Recall that the specifier of a constituent structure phrase is the daughter of XP
and sister to X$ . We propose that specifier positions are filled by phrases that
are prominent either syntactically or in information-structural terms (compare
Bresnan et al.’s “Clause-prominence of DFs” principle: Bresnan et al. 2016: 209).
Syntactically prominent phrases that can appear in specifier positions in the clause
are those bearing either the function subj or the overlay function dis heading
a long-distance dependency (see §2.1.11). Information-structurally prominent
phrases can also appear in specifier position; if they are not syntactically promi-
nent, they may bear any grammatical function within the local clause. The table in
(10) shows the permitted grammatical and information structure roles of phrases
appearing in specifier positions (see Snijders 2015 for further discussion):

(10) For phrases in specifier position within the clause:

Grammatical role Syntactically Prominent at informa-
prominent? tion structure?
subj in the local clause yes yes or no
any grammatical func- no yes (topic or focus)
tion in the local clause
dis, related to a poten- yes yes (topic or focus)
tially nonlocal gramm-
atical function

In this, our position differs from some other LFG work (Dalrymple 2001; Bresnan
et al. 2016), which assumes that specifier positions are filled by phrases bearing
122 syntactic correspondences

particular f-structure roles: subj, topic, or focus. Following Alsina (2008) and
Asudeh (2012), we assume that topic and focus are information structure roles,
and are not represented at f-structure. Rather, phrases that play the role of topic or
focus at information structure can appear in a specifier position with no necessary
restriction on the grammatical function that these phrases can bear, or in other
positions within the clause. See Chapter 10 for more discussion of information
structure in LFG’s projection architecture, including a historical overview of the
treatment of information structure in LFG.

4.2.3.1 Syntactically prominent specifiers A phrase in specifier position that

is syntactically prominent bears either the subj function in the same clause, or
the overlay function dis. If it bears the dis function, there are two possibilities: it
participates in an unbounded dependency and is identified with a (nonoverlay)
grammatical function in the same clause or an embedded clause, as in (11a),
or it is a dislocated constituent which is anaphorically related to a (nonoverlay)
grammatical function in the same clause or an embedded clause, as in (11b):

(11) a. Chocolate, David loves .

b. Chocolatei , David loves iti .

In examples like (11a), the within-clause function of a displaced phrase is specified

by means of functional uncertainty; see §6.1.2 for definition and discussion of
functional uncertainty, and Chapter 17 for a more detailed discussion of long-
distance dependencies. Examples like (11b) involve an anaphoric relation between
the displaced phrase and a pronoun; see Chapter 14.

SUBJ in specifier position: In English and many other languages, the specifier
position of IP is associated with the subj function:2

(12) David yawned.

IP

NP I
’
N VP

David V

yawned

King (1995) shows that only subjects can appear in the specifier position of VP
in Russian:

2 The tree in example (12) has an IP and an I$ node, but no I node; recall from our discussion in
§3.3.7 that no c-structure position is crosslinguistically obligatory, and heads are often optional.
regularities in the c-structure/f-structure mapping 123

(13) posadil ded repku.

planted old.man turnip
‘An old man planted a turnip.’
IP

I VP  
’
posadil NP V 



planted  
N NP

ded N
old.man
repku
turnip

dis in specifier position: In English, a question phrase like what appears

in the specifier position of CP and bears the function dis, as shown in (14). As
discussed in §2.1.11, dis (which stands for dislocation or long distance depen-
dency) is an overlay function, and as such it must be integrated into the internal
structure of the clause by a relation to a primary, nonoverlay function. It may head
an unbounded dependency, as it does in (14), or anaphorically bind a non-overlay
function, as in left dislocation structures such as Chocolatei , David loves iti . In (14),
dis is identified with the obj of the verb eat:

(14) What is David eating?

CP
 
NP C ’
 
N C IP  
 
 
What is NP I 



 
N VP  

David V

eating

The value of dis in (14) is a set because more than one constituent can bear the
dis function in a single f-structure. This is the case, for instance, in a version of the
sentence in (14) in which David is left dislocated (David, what is he eating?), and
the f-structures for David and what are in the dis set. However, while this analysis
is appropriate for English, the same is not necessarily true for all languages. In this
book, we assume unless we are aware of evidence to the contrary that the value of
dis is a set. However, this must be established independently, on a language-by-
language basis.
124 syntactic correspondences

The dis phrase can also bear a grammatical function in an embedded clause
(for more discussion, see §6.1.2 and Chapter 17). In (15), dis in the main clause is
identified with the obj in the subordinate clause:

(15) What does Chris think David ate?

CP

NP C ’
N C IP
What does NP I ’
N VP
Chris V IP
think NP I
N VP
David V
ate

The specifier of CP is also filled by dis in Russian, as King (1995) shows.

Question phrases in Bulgarian bear the function dis and fill the specifier
position of IP (see Rudin 1985); another dis can be included in this sentence as
well, as example (16) shows. Since more than one dis can appear in the same
clause, as discussed above and shown in (16), the value of the dis attribute is a set.
The difference between the two specifier phrases is one of information structure
status: Ivan has the information structure function topic, while kakvo ‘what’ has
the information structure function focus in this sentence:3

(16) Ivan kakvo pravi?

Ivan what does
‘What is Ivan doing?’
CP  
’
NP C   
 
 
 
  
N IP  
 
 
 
 
Ivan NP I   
 
Ivan  
N I  
 
 
kakvo pravi
what does

3 For more on the information structure functions of question words, see Mycock (2013) and
Chapter 17.
regularities in the c-structure/f-structure mapping 125

Example (17) shows that a phrase in specifier position can play a syntactic role in
a subordinate clause. Here, Ivan is in the specifier position of CP and kakvo is in
the specifier position of IP in the main clause; they bear the syntactic roles of subj
and obj in the subordinate clause:

(17) Ivan kakvo kaza, če pravi?

Ivan what say.pst.2sg comp does
‘What did you say that Ivan is doing?’
CP
 
NP C
’
   
 
 
 
N IP    
 
 
Ivan NP I
 
 
 
   
Ivan  
N I VP
 
  
 
 
kakvo kaza V 





what say 



CP  
  
C  
 
  
C IP   
  
 
če
pravi
does

Izvorski (1993) provides further discussion of Bulgarian constituent structure.

Mycock (2006) shows that in Hungarian, the specifier position of VP can be
filled by a phrase that bears some grammatical function within the same clause or
a subordinate clause, and is associated with the information structure role of focus:

(18) Q: Who does John love?

A: János Mari-t szereti.
John.nom Mary-acc loves
‘As for John, it is Mary who he loves.’
 
S ’
 
 
NP VP  
 
N
NP V
János
N V

Mari-t szereti
loves
126 syntactic correspondences

Laczkó (2014a,b) goes further, providing a detailed analysis of the sorts of phrases
that can fill the specifier of VP in Hungarian; he argues that besides hosting
phrases associated with the information structure role of focus (including question
phrases), the specifier of VP can also host verbal modifiers, a category including
verbal particles as well as parts of idiomatic expressions and certain object or
oblique arguments. He also discusses the consequences of the Hungarian data for
LFG assumptions regarding regularities in the c-structure/f-structure mapping.

4.2.3.2 Information-structurally prominent specifiers King (1995) shows that

in Russian, the phrase appearing in the specifier of IP bears a prominent role
at the level of information structure: it is either in focus, or is an internal (new
or continuing) topic. Syntactically, the phrase bears an unspecified grammatical
function within the same clause, and does not head a long-distance dependency,
as was the case in the Hungarian example in (18). We provide more discussion
of information structure roles such as topic and focus and their representation in
Chapter 10; here we simply discuss the f-structure representation, and note that a
phrase with any grammatical role in the clause can fill the specifier of IP in Russian.
In (19), the phrase in the specifier of IP ‘Evgenija Onegina’, is a topic and the
object of the clause (King 1995: 206):

(19) ‘Evgenija Onegina’ napisal Puškin.

Eugene Onegin wrote Pushkin
‘Pushkin wrote ‘Eugene Onegin’.’
IP

NP I
 
Evgenija Onegina ’
I VP  
Eugene Onegin  
napisal NP  
wrote
N

Puškin
Pushkin

The specifier of IP in Russian is information-structurally prominent, but not

syntactically prominent. It is not associated with the dis function, since the phrase
appearing in the specifier of IP does not control a long-distance dependency: it
must bear a grammatical function within the same clause, and cannot be identified
with a grammatical function in a different clause. Thus, in (19) ‘Evgenija Onegina’
is the object of the clause within which it appears: phrases appearing in the specifier
of IP in Russian do not play a syntactic role in a subordinate clause.

4.2.4 Nonprojecting words

We turn now to the f-structure role of words that do not project full phrases.
As discussed in §3.3.4, we follow Toivonen (2003) in assuming the existence of
regularities in the c-structure/f-structure mapping 127

nonprojecting words: that is, lexical and functional categories that are not the heads
of full phrases. Toivonen (2003: 68) proposes the following principle governing
the f-structural role of nonprojecting categories, according to which nonprojecting
categories cannot be adjuncts:

(20) Nonprojecting words:

Words adjoined to heads are co-heads or argument functions.

Unlike Toivonen, we allow a wider range of f-structure roles for nonproject-

ing words, including adjuncts; this allows an analysis of, for example, English
prenominal adjectives in terms of head adjunction, as proposed by Sadler and
Arnold (1994) and Arnold and Sadler (2013), and discussed in §3.3.4. In (21), the
nonprojecting adjective happy is adjoined to the head noun person, and appears as
a member of the adjunct (adj) set at f-structure:

(21) a happy person

 
 − 
 

NP

D N

a N0

A N0

happy person

4.2.5 Idiosyncratic constructions and minor categories

There is some degree of tension between the general principles relating constituent
structure and functional structure which we have been discussing and the demands
of idiosyncratic words and constructions in particular languages. Like Kay and
Fillmore (1999), LFG aims to provide analyses of idiomatic language patterns
as well as the relatively general properties of languages. The need to provide
analyses which are crosslinguistically well-motivated and supported by robust
typological generalizations must be balanced by the equally important requirement
for analyses that are faithful to the phenomenon under analysis; it is important to
avoid the danger of forcing a construction with unusual and interesting properties
into a more familiar analytic mold.
As discussed in §3.3.1, some analyses of idiosyncratic words and constructions
assume the existence of minor categories such as Part (for Particle) and CL (for
Clitic) in addition to the major lexical and functional categories. Such analyses
often assume that at least in some instances, the distribution of these minor
categories is language dependent and does not fall under the general rules we
have been discussing. If we adopt such analyses, we must examine idiomatic,
128 syntactic correspondences

language-specific, or construction-specific syntactic properties of minor categories

on a case-by-case basis to determine their properties. Other researchers have pro-
posed generalizations about the f-structural role of these categories; for example,
Zaenen (1983) proposes the Minor category convention, according to which minor
categories are f-structure co-heads:

(22) Minor category convention:

Minor categories map onto the same f-structure as the node that immedi-
ately dominates them.

We have not found motivation for assuming minor categories in this work, and so
we will not adopt any particular view about the permitted f-structure roles of such
categories.

4.2.6 The exocentric category S

The exocentric category S is not an X$ -theoretic category (see §3.4.2) and does not
obey the X$ -theoretic generalizations governing the relation between c-structure
positions and f-structure functions. Thus, in languages that are unlike English
in having the exocentric category S, such as Warlpiri, Tagalog, and Malayalam
(Austin and Bresnan 1996), phrase structure configuration is not always an unam-
biguous indicator of grammatical function: phrases with different grammatical
functions may appear in the same constituent structure position. For this reason,
as Nordlinger (1998) points out, languages with S often allow relatively free word
order and rely more heavily on morphological marking than phrase structure
configuration for the identification of grammatical functions.
S can dominate phrases with any grammatical function, including subj. In
Warlpiri, the subject ngarrka-ngku ‘man’ may appear inside S (Simpson 1991), as
shown in example (23a); in contrast, in example (23b) the subject appears outside S:

(23) a. wawirri ka ngarrka-ngku panti-rni

kangaroo prs man-erg spear-npst
‘The man is spearing the kangaroo.’
 
’
 
 
 
IP  
 
NP I

N I S

wawirri ka NP V
kangaroo
N panti-rni

ngarrka-ngku
 “movement” and discontinuity 129

b. ngaju ka-rna parnka-mi  

’
 
‘I am running.’   
 
 
IP   
  
 
NP I

N I S

ngaju ka-rna V

parnka-mi

4.3 “Movement” and discontinuity

4.3.1 Apparent head “movement”
As discussed in §3.3.7, there are no constituent structure positions that are crosslin-
guistically obligatory; in particular, the head of a phrase need not be realized.
Consider the following Russian example, in which the tensed verb appears in I
(King 1995: Chapter 10):

(24) ona pročitala knigu.

she read.pst book
‘She read the book.’
IP  
’
NP I  
  
 
N I VP  
   
  
ona pročitala NP   
   

she  
N  
 
 
knigu
book

Examples such as these need no special analysis within LFG, and the verb is
not thought of as having “moved” to the position in which it appears. Rather,
the principles we have outlined so far predict that this configuration is possible
and wellformed. In Russian, all tensed verbs have the phrase structure category I;
the verb in (24) appears in I and not within VP. It is not possible for there to
be two lexical verbs in a single Russian sentence, one in the I position and one
within the VP, since this would produce an ill-formed f-structure; each lexical verb
contributes a pred value to its f-structure, and the Consistency Principle forbids
130 syntactic correspondences

an f-structure from containing a pred attribute with two different semantic forms
as its value. Conversely, a sentence with no verb is ill-formed, since in that case the
main f-structure would contain no pred, and the Completeness Principle would be
violated. Exactly one lexical verb must appear, and it must appear in the appropriate
position for its constituent structure category. Bresnan et al. (2016: Chapter 7)
provide more discussion of such cases, which they refer to as head mobility or
variable head positioning.

4.3.2 Clitic doubling

In most dialects of Spanish, full noun phrase objects appear after the verb, while
clitic object pronouns appear preverbally. This has been analyzed in other theories
as an instance of movement (see, for example, Baltin 1982):

(25) a. Juan vió a Pedro.

Juan saw prep Pedro
‘Juan saw Pedro.’
b. Juan lo vió.
Juan 3sg.acc.clitic saw
‘Juan saw him.’

In fact, though, no special analysis is required for Spanish, and no “movement”

need be assumed. Two phrase structure positions are associated with the obj
function in Spanish, a preverbal clitic position and a postverbal phrasal position.
If a verb is transitive, an obj is required. This means that one of these positions
must be filled, and either a clitic or a full noun phrase must appear. In standard
Spanish, it is not possible to fill both phrase structure positions: this would cause
the obj function to be associated with a pred with two different values, and the
Consistency Principle would rule out this ill-formed possibility:

(26) Standard Spanish, no clitic doubling:

⃰Juan lo vió a Pedro.
Juan 3sg.acc.clitic saw prep Pedro
‘Juan saw Pedro.’
Ill-formed f-structure:
IP

NP I  
’
N I VP
 
 
 
 
Juan N I V
Juan
lo vió PP
saw
P

P NP

a N

Pedro
Pedro
 “movement” and discontinuity 131

Thus, exactly one object phrase must appear, and it must appear in the constituent
structure position appropriate for its phrase structure category.
Interestingly, however, complementary distribution between a clitic and a full
noun phrase is not always found. Andrews (1990b) discusses dialects of Spanish
which allow clitic doubling, in which it is possible for both obj positions to be filled,
as in (27):

(27) River Plate and Peruvian Spanish, clitic doubling allowed:

Juan lo vió a Pedro.
Juan 3sg.acc.clitic saw prep Pedro
‘Juan saw Pedro.’
IP

NP I  
’
N I VP
 
 
 
 
Juan N I V
Juan
lo vió PP
saw
P

P NP

a N

Pedro
Pedro

Clitic doubling is possible in the River Plate and Peruvian dialects of Spanish as
described by Jaeggli (1980) (see also Grimshaw 1982b; Estigarribia 2005; Mayer
2006; Bresnan et al. 2016; and, on Limeño, Mayer 2008). In these dialects, the object
clitic pronoun lo is undergoing reanalysis as an agreement marker, and has begun
to lose its semantic status as a pronominal. Formally, this reanalysis is reflected
as optionality of the pred value of the clitic (see §5.2.4): lo need not contribute a
pred value to the f-structure. Since its other features (pers 3, num sg, case acc)
are compatible with the full phrasal obj, both the clitic and the full noun phrase
can appear in the same sentence.

4.3.3 Category mismatches

Phrases not appearing in their canonical positions can sometimes exhibit category
mismatches, as discussed by Kaplan and Bresnan (1982), Kaplan and Zaenen
(1989b), and Bresnan et al. (2016: Chapter 2). Kaplan and Bresnan (1982) discuss
the verb think, showing that it subcategorizes either for a sentential argument as in
(28a), or an oblique phrase as in (28b):

(28) a. He didn’t think that he might be wrong.

b. He didn’t think of his mistake.
132 syntactic correspondences

Kaplan and Bresnan (1982) observe that surprising grammaticality patterns

emerge when the complement of think is displaced, appearing at the beginning of
the sentence; they provide the examples in (29) to illustrate this point:

(29) a. ⃰[That he might be wrong] he didn’t think.

b. ⃰He didn’t think of [that he might be wrong].
c. [That he might be wrong] he didn’t think of.

Example (29a) contrasts in grammaticality with (28a); further, although example

(29b) is ungrammatical, example (29c) is fully acceptable. Such examples are my-
sterious on theories that analyze displaced phrases in terms of phrase structure
movement: if we assume that an example like (29a) is derived from a source like
(28a) by movement of the complement to initial position, and that (29c) is derived
by movement from (29b), it is not clear what would account for their different
status.
As noted by Kaplan and Zaenen (1989b) and Dalrymple and Lødrup (2000),
a nontransformational account fares better in accounting for instances of appar-
ent mismatch. In LFG, the relation between a sentence-initial constituent in an
example like (29c) and its within-clause role is defined at f-structure rather than
c-structure. Example (30) shows that English allows preposition stranding, with
the sentence-initial phrase that problem filling the role of object of the stranded
preposition of:

(30) That problem, he didn’t think of.

’

IP

NP IP

D N NP I

that N N I VP

problem he didn’t V

V PP

think P

of
 “movement” and discontinuity 133

In (30), there is no sense in which the phrase that problem has been moved
to clause-initial position from a position within the clause. Rather, the relation
between the initial NP and its within-clause function is specified in functional
rather than constituent structure terms, by functional uncertainty (see §6.1.2),
and phrase structure category is not a part of the specification. Exactly the same
constraints allow for a clause-initial CP to fill the role of object of the preposition
of, as shown in (31):4

(31) That he might be wrong he didn’t think of.

’
’

IP
CP IP

C NP I
C IP N I VP

that NP I he didn’t V
N I VP V PP

he might V think P
V AP of
be A
wrong

Example (29a) is ruled out by functional constraints on extraction relations: in par-

ticular, the clausal complement of think (as opposed to think of ) is a comp, which,
at least in English, cannot participate in constructions involving displacement. In
contrast, example (29b) is ruled out as an impossible c-structure configuration: a
preposition such as of does not take a CP as its complement at c-structure.

4 The f-structure for the clause that he might be wrong has been simplified; the functional structure
of copular constructions is discussed in detail in §5.4.5.
134 syntactic correspondences

4.3.4 Constituent structural discontinuity

Simpson (1991: Chapter 5) provides a very interesting discussion of nonconfigu-
rationality and discontinuity in an LFG setting (see also Austin and Bresnan 1996;
Kuhn 2001c; Bresnan et al. 2016). Simpson shows that the subject of a Warlpiri
sentence can appear as a daughter of S (as in example 23a) as well as in the specifier
position of IP (as in example 23b). As we would expect, then, phrases in each of
these positions—or, crucially, in both positions—can contribute to the structure of
the subj of a sentence. In the “merged” reading5 of example (32), parts of the subj
appear in S while other parts appear in the specifier of IP: the subj phrase kurdu-
ngku ‘child’ appears in the specifier of IP, while a modifier of the subj, wita-ngku
‘small’, appears within the S. Since Warlpiri is a pro-drop language, the obj is not
represented in the c-structure tree (see §4.1):

(32) kurdu-ngku ka wajilipi-nyi wita-ngku

child-erg prs chase-npst small-erg
‘The small child is chasing it.’
 
’
 
 
 
 
 
 
 
 
 
 

IP

NP I

N I S

kurdu-ngku ka V NP

wajilipi-nyi N

wita-ngku

These two phrases are functionally compatible, since both bear erg case: one
represents the head of the subject phrase, and the other represents a modifier. Here
too, we need say nothing special about examples involving constituent structural

5 There is no categorical distinction between adjectives and nouns in Warlpiri, and we follow
Simpson in categorizing both kurdu-ngku ‘child’ and wita-ngku ‘small’ as nouns. Simpson shows that
there are two readings for example (32); in the merged reading, the one which is of interest here,
wita-ngku ‘small’ is interpreted as a restrictive modifier of kurdu-ngku ‘child’. The other reading is the
nonrestrictive, “unmerged” interpretation, which Simpson paraphrases as ‘The child is chasing it and
she is small’; on Simpson’s analysis, this reading has a different syntactic structure.
 the lexical integrity principle 135

discontinuity such as these; our theory predicts that they will occur, and we need
no special analysis to encompass them.

4.4 The Lexical Integrity Principle

The Lexicalist Hypothesis was first discussed within a transformational setting
by Chomsky (1970), who advanced the claim that rules of word formation are
lexical rather than transformational: words and phrases are built up from different
elements and by different means. The hypothesis exists in two forms. Its weaker
form (the “weak lexicalist hypothesis”) states that the lexicon consists of all roots
formed by derivational morphology, and that roots are not formed by transfor-
mational rules. However, on this view inflectional morphemes may be treated as
seperate syntactic units from the root, occupying separate nodes in the phrase
structure tree.
Bresnan (1978) adopts a stronger form of the hypothesis, claiming that “syntac-
tic transformations do not play a role in word formation,” either in derivational or
inflectional morphology. Lapointe (1980) builds on this claim, proposing what is
usually called the “strong lexicalist hypothesis” or the Lexical Integrity Principle:6

(33) Lexical Integrity Principle, Lapointe (1980: 8): No syntactic rule can refer to
elements of morphological structure.

Bresnan et al. (2016) state the principle in the following way:

(34) Lexical Integrity Principle, Bresnan et al. (2016: 92): Morphologically com-
plete words are leaves of the c-structure tree, and each leaf corresponds to
one and only one c-structure node.

Current work in LFG adheres to this principle: LFG does not assume the exis-
tence of processes that assemble word forms in constituent structure or reorder
subparts of word forms during syntactic composition. In this way, the LFG view
contrasts with approaches such as Distributed Morphology (Halle and Marantz
1993; Embick and Noyer 2007) and the Exo-Skeletal Model (Borer 2013), which
reject the Lexicalist Hypothesis. We will not address this controversy in detail
here; cogent defenses of the Lexicalist Hypothesis are provided by Bresnan and
Mchombo (1995), Spencer (2005b, 2016), Stewart and Stump (2007), and refer-
ences cited there. We provide a detailed discussion of the role of morphology in
the LFG architecture in Chapter 12.
Although words are not subject to processes of assembly in the constituent
structure, their syntactic contributions can be complex. A word that forms a unit
at the level of c-structure may introduce complex functional structure, structure
that in other languages might be associated with a phrase rather than a single
word. For example, Bresnan and Mchombo (1987) present evidence that subject

6 Lapointe (1980) refers to this as the “generalized lexicalist hypothesis.”

136 syntactic correspondences

and object markers on the Chicheŵa verb can function as incorporated pronouns.
These affixes form a part of the verb at c-structure, meaning that the f-structure
corresponding to this verb is complex, similar to the f-structure for a sentence in
a language like English.7 In (35), the marker zi- represents a pronominal subj of
noun class 10, and the marker -wá- represents a pronominal obj of noun class 2:
 
(35) ’
 
 
 
 
V  
 
 
zi-ná-wá-lúm-a  

‘They bit them.’

As this example shows, the notion of word is multifaceted and must be fully defined
at each syntactic level of representation; units at c-structure need not correspond
to simple units at other levels.
Much other work has been done within LFG in support of the strong lexicalist
hypothesis, demonstrating convincingly that a word that is atomic at the level of
c-structure can project the structure of a phrase at functional structure. Simpson
(1983, 1991), Ackerman (1987), O’Connor (1987), Mohanan (1994, 1995), Bres-
nan and Mchombo (1995), Nordlinger and Bresnan (1996), Matsumoto (1996),
Sadler (1997), Ackerman and LeSourd (1997), Börjars (1998), Alsina (1999),
Broadwell (2008), Asudeh et al. (2013), and Lowe (2016b) have all made valuable
contributions to our understanding of wordhood and lexical integrity.

4.5 Economy of Expression

The principle of Economy of Expression requires the choice of the simplest and
smallest c-structure tree that expresses the intended meaning while also obeying
wellformedness conditions on c-structures and corresponding to a wellformed
f-structure. Economy of Expression is formulated by Bresnan et al. (2016) as
follows:

(36) Economy of Expression:

All syntactic phrase structure nodes are optional and are not used unless
required by independent principles (Completeness, Coherence, Semantic
expressivity). (Bresnan et al. 2016: 90)

Since a node may be used only if it is required by independent principles to express

a particular meaning, speakers are required to choose the most economical way

7 See §5.4.3 and §14.1 for more discussion of the pronominal status of the subject and object markers
in Chicheŵa.
 economy of expression 137

of expressing a meaning, where economy is measured in terms of the number of

nodes in the c-structure tree. That is, if there is more than one possible c-structure
tree for an utterance, the correct tree is the one with the smallest number of nodes.
The effects of the Economy metric are dependent on the independently-
specified wellformedness conditions on c-structure trees that are assumed. For
example, according to the version of X$ theory assumed by Bresnan et al. (2016)
and in this book, a nonbranching X$ node is not required. In this setting, Economy
disallows trees with a nonbranching single-bar-level category X$ , since the X$ node
is not necessary to express the intended meaning, and the tree without the X$ node
has fewer nodes. Given a choice between the following two trees, then, Economy
mandates the choice of the tree in (37b) and not the tree in (37a):

(37) Bresnan et al. (2016):

a. Uneconomical tree (disallowed): b. Economical tree:
NP NP

N$ N

N David
David

Toivonen (2003) proposes a slightly different definition of Economy:

(38) Economy of Expression: All syntactic phrase structure nodes are optional
and are not used unless required by X$ -constraints or Completeness. (Toivo-
nen 2003: 200)

Toivonen assumes a stricter version of X$ theory, according to which the X$ node

may not be omitted. Her stricter version of X$ theory leads to the choice of a
different tree as the correct one for the NP David; on her view, the smaller tree
is not wellformed and may not be selected:

(39) Toivonen (2003):

Wellformed tree: Ill-formed tree violating X$ theory (disallowed):
NP NP

N$ N

N David
David

Both definitions of Economy proceed from the claim that every node in a
c-structure tree is optional. In fact, the issue of optionality of nodes as a general
principle of the grammar is separable from particular definitions of an Economy
principle. As discussed in §3.3.7, there are no phrase structure positions that are
obligatory in every phrase and in every language: a phrase XP need not dominate a
138 syntactic correspondences

head X in the tree, and the complements of a lexical predicate X need not appear as
sisters to X. However, Snijders (2012, 2015) shows that in some configurations in
some languages, c-structure nodes are obligatory and not optional: in particular,
Snijders shows that the NP complement of P in the Latin PP is obligatory. The
existence of obligatory nodes is not incompatible with the adoption of an Economy
principle: even if some nodes in some constructions are obligatory, an Economy
measure can still be used to choose among larger and smaller trees that otherwise
obey the rules of the grammar.
Dalrymple et al. (2015) discuss the various versions of Economy which have
been proposed, providing explicit formal definitions of three versions of the
Economy metric, and an overview of analyses which have appealed to an Economy
measure. They discuss whether Economy is best treated as a very general principle,
as on Toivonen’s and Bresnan et al.’s approaches, or if it is preferable to appeal
to more specific principles for particular subcases rather than a general Economy
principle.

4.6 Further reading and related issues

There has been a great deal of work in LFG on the relation between constituent
structure and functional structure in a typologically diverse set of languages.
Of particular note, besides the works mentioned earlier, are Johnson (1986),
Dahlstrom (1987), Kaplan and Zaenen (1989b), and Huang (1990), who discuss
discontinuity, phrasal constituency, and the c-structure/f-structure relation; Lee
(2001) and Mahowald (2011) on word order freezing; and Sells (2000b), who
discusses evidence from raising in Philippine languages and its bearing on clause
structure. Bresnan et al. (2016: Chapter 5) provides a lucid discussion of the
fragmentability of language, the fact that the c-structure and f-structure of sentence
fragments can be easily inferred based on lexically and phrasally specified con-
straints on syntactic structures. Butt et al. (1999) present a comparative overview
of English, French, and German constituent and functional structures and the
relation between them, discussing a wide range of syntactic constructions and
issues that arise in the analysis of these languages.
 5
Describing syntactic structures

Up to this point, our discussion has concentrated on the nature and represen-
tation of the two syntactic structures of LFG. We will now demonstrate how
to formulate descriptions of and constraints on c-structure, f-structure, and the
relation between them, and we will see how these constraints are important in the
statement of universal typological generalizations about linguistic structure. These
constraints are a part of the formal architecture of LFG theory.
The job of the designer of a formal linguistic framework such as LFG is to
provide a way of stating linguistic facts and generalizations clearly and precisely,
in a way that is conducive to a solid understanding of the linguistic structures that
are described and how they are related. Designing an appropriate representation
for a linguistic structure is very important in helping the linguist to understand
it clearly and in avoiding confusion about the properties of the structure and its
relation to other structures. A formal linguistic theory must provide efficient and
transparent ways of describing the facts of a language. Linguistically important
generalizations should be easy to express, and the linguistic import of a constraint
should be evident.
The job of a linguist working within a formal framework is to discover the facts
of a language or of languages and to express these facts by using the tools provided
by the designer of the framework so that the linguistic facts emerge precisely and
intuitively. Most importantly, the linguist must distinguish between constraints
that are needed and used in describing the facts of language, the linguistically
relevant constraints, and those that may be expressible within the overall formal
framework but do not play a role in linguistic description. A well-designed formal
framework aids the linguist in deciding which constraints are relevant and which
are not in the description of linguistic structures.

5.1 Constituent structure rules

5.1.1 Phrase structure rules
Most linguistic theories that represent phrasal information in terms of phrase
structure trees make use of phrase structure rules to determine the possible and
admissible phrase structure configurations in a language:

(1) S −→ NP VP

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
140 describing syntactic structures

This rule permits a node labeled S to dominate two nodes, an NP and a VP, with
the NP preceding the VP. In LFG, phrase structure rules are interpreted as node
admissibility conditions, as originally proposed by McCawley (1968): a phrase
structure tree is admitted by a set of phrase structure rules if the rules license
the tree. In other words, phrase structure rules are thought of as descriptions
of admissible trees, and the trees of the language must meet these descriptions.
McCawley’s groundbreaking work constituted an important alternative to the way
of thinking about phrase structure rules that was prevalent in the mid-1960s, which
viewed phrase structure rules as a procedural, derivational set of instructions to
perform a series of rewriting steps.
In many theories of phrase structure specification, all phrase structure rules
are of the type illustrated in (1): the right-hand side of the rule (here, NP VP)
specifies a particular unique admissible configuration. Constituent structure rules
in LFG are more expressive than this, in that the right-hand side of a phrase
structure rule consists of a regular expression (Kaplan and Bresnan 1982: 277; see
§6.1 for a full discussion and definition of regular expressions). The use of a regular
expression allows a sequence of category labels in which some categories may be
optional, some categories may be repeated any number of times, and disjunction
is permitted.
Since phrase structure rules represent constraints on phrase structure configura-
tions, it is reasonable and desirable to abbreviate a large or even an infinite number
of rules by means of a regular expression. A phrase structure rule is not thought
of as a large or infinite set of rules that can be applied individually in a series of
rewriting steps, but as a characterization of the daughters that nodes of a phrase
structure tree can dominate. This constraint-based view of linguistic descriptions
pervades the formal theory of LFG: a grammar of a language consists of a set of
constraints on linguistic structures, and these constraints admit only wellformed
analyses of utterances.
A simple LFG phrase structure rule can have the following form:

(2) IP −→ {NP | PP} I#

This rule indicates that either an NP or a PP can appear in the specifier position of
IP. The curly brackets mark a disjunction of phrase structure categories, with the
possibilities separated by a vertical bar |. This rule abbreviates the following two
rules, where the disjunction in (2) has been fully expanded into two separate rules:

(3) a. IP −→ NP I#
b. IP −→ PP I#

The abbreviation in (2) is not only more compact than the two-fold expansion
in (3), but more revealing of the linguistic facts: the specifier position of IP can
be filled either by NP or by PP without affecting any properties of the second
daughter of IP, namely I# . Stating this fact by means of two separate rules, as in
(3), makes it appear that the I# might have different properties in the two different
 constituent structure rules 141

rules, since there is no apparent relation between the I# in the first rule and the I#
in the second rule.
In (4), the parentheses around the NP indicate optionality. The Kleene star
annotation ⃰ on the PP indicates that zero or more PPs may appear in the expansion
of the rule; more discussion of the meaning and use of the Kleene star is given
in §6.1:

(4) V# −→ V (NP) PP⃰

Thus, this rule admits trees in which a V# node dominates a V, an optional NP, and
any number of PPs. The use of the Kleene star means that an infinite set of possible
rules—rules with any number of PPs—can be abbreviated with a single expression.
Thus, phrase structure generalizations can be stated once rather than separately
for each specific phrase structure configuration licensed by the rule.

5.1.2 Node labels

The rule in (1) admits a tree in which a node labeled S dominates two nodes, one
labeled NP and the other labeled VP. Generally, we refer to a node of a tree by its
label, and so we talk of the S, NP, or VP nodes in the tree. In fact, however, S, NP, and
VP are node labels, and each node of the tree is related to its label by the labeling
function λ (Kaplan 1995). To be completely explicit, then, we could represent a tree
and its node labels in the following more cumbersome way, with dots representing
the nodes themselves, and arrows representing the λ function from nodes to node
labels:
λ
(5) • S
NP
• •
VP

We generally follow standard practice in displaying trees with node labels repre-
senting the nodes which they label, meaning that we represent the tree in (5) as in
(6). The node labeling function λ will occasionally be important in our subsequent
discussion, however.

(6) S

NP VP

5.1.3 C-structure metacategories

Example (2) illustrates the use of disjunction over a set of categories in a rule. It
is also possible to introduce an abbreviation over a set of categories in a rule: for
instance, an abbreviation like XP is often used to represent a set of categories that
behave similarly in some way. In such a situation, XP is a c-structure metacategory
representing several different categories.
King (1995) uses metacategory abbreviations like the following in her analysis
of Russian:
142 describing syntactic structures

(7) CP −→ XP C#
IP −→ XP I#

The use of XP in these rules indicates that a full phrase of any category (NP, PP,
and so on) can appear as the first daughter of CP and IP. XP is defined as follows:1

(8) XP ≡ {NP | PP | VP | AP | AdvP}

The use of abbreviations like this allows for the expression of general statements
about all phrases that appear in a particular phrase structure position.
In (8), the metacategory XP stands for any one of a number of phrasal categories.
In fact, a metacategory can abbreviate a longer sequence of categories, not just
a single category (Kaplan and Maxwell 1996). This is shown in the following
putative definition of the metacategory PREDICATE:

(9) PREDICATE ≡ V NP

More generally, a metacategory can be used as an abbreviation for any regular

predicate over categories. What do such abbreviations mean, and how are they
used?
An abbreviation like PREDICATE ≡ V NP can be used to express a generalization
about where a sequence of categories like V NP can appear in the grammar without
introducing a node dominating those categories into the tree. Instead, wherever
a phrase structure rule refers to the metacategory PREDICATE, the sequence of
categories V NP is permitted to appear in the phrase structure tree. For example,
we can interpret the rule in (10) as referring to the definition of PREDICATE
given in (9):

(10) S −→ NP PREDICATE

Given this definition, the rule in (11) produces exactly the same result as the rule
in (10):

(11) S −→ NP V NP

The phrase structure rule in (10) and the definition of PREDICATE in (9) admit the
following tree:

(12) S

NP V NP

Notably, there is no PREDICATE node in this tree.

The possibility of using a metacategory to characterize a sequence of categories
in this way has an interesting impact on one of the clearest traditional motivations

1 The symbol ≡ connects two expressions that are defined to be equivalent; the expression in (8) can
be read as: “XP is defined as the disjunction {NP | PP | VP | AP | AdvP}”; that is, XP is a phrase with
category NP, PP, VP, AP, or AdvP.
 constituent structure rules 143

for phrase structure constituency, described in §3.1: generalizations governing the

distribution of sequences of categories. In many theories of phrase structure, the
fact that a phrase like the dachshund has the same syntactic distribution as a phrase
like the black dachshund is taken as evidence that both phrases are constituents that
are dominated by an NP node; on this view, generalizations about the distribution
of the category sequence Det (A) N are stated in terms of the distribution of an NP
constituent. The use of a metacategory like VP in the rule in example (11) allows
for the statement of generalizations about sequences of categories in the same
way. Importantly, however, the resulting phrase structure tree does not contain
a constituent labeled VP: the V NP sequence does not form a phrasal unit in the
constituent structure tree.
It is interesting to note that some (but not all) of the criteria for phrase struc-
ture constituenthood presented in §3.2 are based in part on the distribution of
sequences of categories. Further research may reveal more about the possibility and
desirability of capturing generalizations about category distribution by means of
metacategories defined over sequences of categories, rather than by assuming the
existence of a phrasal constituent dominating these categories in the constituent
structure tree.

5.1.4 Complex categories

Constraints on phrase structure may involve fine-grained specifications over sub-
types of standard categories like VP or V. Such patterns can be captured by the use
of complex categories involving parametrization of c-structure categories.
For example, as discussed in §3.3.5, Frank and Zaenen (2002) assume a non-
standard flat structure for the French VP in their analysis of auxiliary selection in
French. They propose that the VP category is parametrized for finiteness, and the
V category is parametrized for finiteness and verb type. Notationally, parameters of
a category are written in square brackets after the category: VP[fin] is a VP with the
feature fin(ite), and V[fin] is a finite V. According to this analysis, one acceptable
VP configuration in French is shown in (13), where a VP[fin] consists of a finite
auxiliary verb, an optional participial auxiliary, and a participial main verb:

(13) Frank and Zaenen (2002: 51):

VP[fin]

V[aux,fin] (V[aux,part]) V[main,part]

This pattern can be generalized to other values for finiteness by using parametrized
rules to allow information about these features to flow through the c-structure tree,
placing fine-grained constraints on how different subtypes of categories can be
realized. In Frank and Zaenen’s grammar, there are three possible instantiations
of the finiteness parameter: fin (finite), inf (infinitive), and part (participle). The
requirement for identical values of the finiteness parameter for the VP and the
initial V is captured by the use of variables over features. Notationally, parameter
variables are prefixed by an underscore: VP[_ftness] and V[main,_ftness] both con-
tain occurrences of the variable _ftness. Within the same configuration, identical
144 describing syntactic structures

variables must match. This means that the rule in (14a) is a succinct abbreviation
of the three rules in (14b):

(14) a. −→
b. −→
VP[inf] −→ V[main,inf]
VP[part] −→ V[main,part]

Thus, the rule in example (15) licenses the c-structure in example (13), as well
as any other c-structures in which the finiteness value of the initial V daughter
matches the finiteness value of the VP:

(15) VP[_ftness] −→ V[aux,_ftness] (V[aux,part]) V[main,part]

This analysis requires the French finite auxiliary est to have the parametrized
category V[aux,fin]:

(16) est: V[aux,fin]

We return to a discussion of parametrized categories in §12.2.2, where we will see

how the morphological component of the grammar determines the c-structure
category of a word form like est, including the parameters of a complex category.

5.1.5 Immediate Dominance and Linear Precedence rules

Immediate Dominance/Linear Precedence rules were introduced by Gazdar and
Pullum (1981), and independently developed within the LFG framework by Falk
(1983), to allow dominance relations to be stated separately from precedence
relations in phrase structure rules. Dominance relations are stated in terms of
Immediate Dominance or ID rules, and precedence constraints are stated in terms
of Linear Precedence or LP rules. These rules allow the statement of generalizations
across families of phrase structure rules: for example, that the head of a phrase of
any category precedes its complements in a head-initial language.
An ID rule expressing only dominance relations is written with commas sepa-
rating the daughter nodes in the rule:

(17) VP −→ V, NP

This rule states that a VP node dominates two other nodes in a tree, a V node and
a NP node, but does not specify the order of V and NP. Thus, it can be regarded as
an abbreviation for the two rules in (18), in which the order is fully specified:

(18) VP −→ V NP
VP −→ NP V

If we wish to specify the order, we can write a separate LP ordering constraint:

 functional constraints 145

(19) VP −→ V, NP V<NP

The ID phrase structure rule in (17), combined with the linear precedence con-
straint V<NP stating that V must precede NP, is equivalent to the standard
ordered phrase structure rule VP −→ V NP. A more complicated example is
given in (20):

(20) VP −→ V, NP, PP V<NP, V<PP

The ID phrase structure rule requires VP to dominate three nodes, V, NP, and PP.
The LP ordering constraints require V to precede both NP and PP, but do not place
any constraints on the relative order of NP and PP. Thus, the ID rule and constraint
in (20) are equivalent to the rule in (21), in which ordering is fully specified, and
NP appears either before or after PP:

(21) VP −→ V {NP PP | PP NP}

ID/LP rules are used in King’s (1995) analysis of Russian. She proposes rules of the
following form:

(22) a. CP −→ XP, C#
b. C# −→ C, IP
c. IP −→ XP, I#
d. XP<Y# , Y<XP

XP is defined as a metacategory over any maximal phrasal category, Y# is a

metacategory over any nonmaximal category of bar level one, and Y is a meta-
category over any lexical or functional category of bar level zero. The constraint
XP<Y# means that the specifier XP of a phrase of category Y appears before the
head Y# , determining the order of the daughters for rules (22a) and (22c). The
constraint Y<XP means that the head Y appears before its complement phrase XP,
determining the order for rule (22b).
In §6.1.1, we discuss more complicated constraints on c-structure configura-
tions, including the use of combinations of regular expressions in defining phrase
structure rules.

5.2 Functional constraints

Just as we have ways of talking about the set of permissible trees in a language,
we would also like to have a way of describing acceptable f-structures. In the fol-
lowing, we discuss how simple constraints on functional structure are interpreted
and when these constraints are satisfied. Chapter 6 provides more discussion of
c-structure and f-structure constraints.
146 describing syntactic structures

Formally, an f-structure is a set of attribute-value pairs, or a function from

attributes to values, as discussed in §2.4. The usual way of presenting an f-structure
is in tabular form (that is, as a table):2

(23)
+

The f-structure in (23) contains two attribute-value pairs: %pred, ‘man’& and
%def, +&. We can place various requirements on f-structures: they may be required
to contain certain attribute-value pairs or one of several possible attribute-value
pairs, or they may be required not to contain certain material. The following
sections explain how such constraints can be imposed.

5.2.1 Functional equations

The equation in (24) specifies the f-structure named m as having an attribute def
whose value is +:

(24) (m def) = +

This is a simple functional description or f-description. This f-description consists

of only a single equation, but f-descriptions can in general consist of any number
of such equations.
The f-structure labeled m in (25) satisfies the constraint in (24), since it has an
attribute def with value +:
! "
(25) m def +

An equation requiring an attribute a of an f-structure to have a certain value v

holds if (and only if) the pair consisting of the attribute a and its value v belongs
to the f-structure.
There are many other f-structures (in fact, an infinite number) that also satisfy
the constraint in (24). Any f-structure that has an attribute def with value + as
well as additional attributes and values satisfies the constraint; for example:
 
(26)
m + 

This f-structure, which represents the expression this man, contains the attribute
def with value +; it also contains additional attributes and values.

2 As explained in Chapter 2, Footnote 21 (page 44), what is depicted in example (23) is an f-structure,
a formal object, not a constraint on f-structures.
 functional constraints 147

The f-structure in (25) is special in that it is the smallest f-structure that satisfies
the constraint in (24). We call such an f-structure the minimal solution to the
f-description in (24): it satisfies all the constraints in the f-description, and it has
no additional structure that is not relevant in satisfying the constraints.
We require the f-structure for any utterance to be the smallest f-structure that
satisfies all of the constraints imposed by the f-description for the utterance, and
has no additional properties not mentioned in the f-description:

(27) The f-structure for an utterance is the minimal solution satisfying the con-
straints introduced by the words and phrase structure of the utterance.

The f-description for an utterance is given by annotations on the phrase structure

rules and the lexical entries involved in the utterance. In §5.3, we discuss how
annotated phrase structure rules and lexical entries give rise to an f-description
for a phrase.
Kaplan and Zaenen (1989b) provide the following formal characterization of
when an equation like the one in (24) holds of an f-structure:3

(28) ( f a) = v holds if and only if f is an f-structure, a is a symbol, and the pair

%a, v& ∈ f .

It is also possible for an expression to involve multiple attribute names—that is,

a longer path through the f-structure:

(29) ( f subj def) = +

For these cases, the following definition is relevant (Kaplan and Bresnan 1982;
Kaplan and Zaenen 1989b):4

(30) ( f as) ≡ (( f a) s) for a symbol a and a string of symbols s.

( f 𝜖) ≡ f , where 𝜖 is the empty string.

This definition tells us that an expression like ( f subj def) denotes the same
f-structure as (( f subj) def): that is, the f-structure that appears at the end of the
path subj def in the f-structure f is the same f-structure that appears as the value
of the attribute def in the f-structure ( f subj). Longer expressions are treated
similarly. The second part of the definition tells us that the empty path 𝜖 can be
ignored: the expression ( f 𝜖) is the same as f .

3 The symbol ∈ is the set-membership symbol: the expression x ∈ Y (“x is in (the set) Y”) means
that x is a member of the set Y. Chapter 6, §3 provides a detailed discussion of predicates and relations
involving sets.
4 The epsilon symbol 𝜖, which represents the empty string or path, is not the same as the set
membership symbol ∈.
148 describing syntactic structures

A Hindi sentence like (31) (McGregor 1972) is associated with the f-description
given in (32) and (33) (some detail has been omitted):

(31) Raam calegaa

Ram go.fut.3m.sg
‘Ram will go.’

The constraints in (32) come from the lexical entries for the proper noun Raam
and the verb calegaa ‘will go’; here, we consider only case and num requirements
imposed by the verb on its subject, ignoring tense, person, and gender features:

(32) Raam (r pred) = ‘Ram’

(r index num) = sg
(r concord case) = nom
calegaa (c pred) = ‘go%subj&’
(c subj index num) = sg
(c subj concord case) = nom

We also know that Raam is the subject of this sentence: that is, the f-structure r
for Raam is the subj of the f-structure c for the sentence. This is encoded in the
constraint in (33):

(33) (c subj) = r

Given this equality, we can substitute r for (c subj) in the constraints in (32). We are
then left with the set of constraints in (34), which are equivalent to the constraints
in (32) and (33):

(34) (r pred) = ‘Ram’

(r index num) = sg
(r concord case) = nom
(c pred) = ‘go%subj&’
(r index num) = sg
(r concord case) = nom
(c subj) = r

The minimal solution to this f-description—the f-structure that satisfies the

f-description and contains no additional attribute-value pairs—is:
 
(35) ’
   
 
 
c 

 r 
 
 functional constraints 149

Notice that some of the equations in the f-description constrain the same attribute-
value pairs. For example, the proper noun Raam requires its f-structure r to have
an index attribute specifying singular number (containing the attribute num with
value sg):

(36) (r index num) = sg

The verb calegaa ‘go’ places the same requirement on its subject:

(37) (c subj index num) = sg

We also know that c’s subject is r:

(38) (c subj) = r

Thus, the two equations in (36) and (37) both require the f-structure labeled r,
which is also called (c subj), to contain the attribute num with value sg in the
index feature bundle. It is common for an attribute and its value to be multiply
constrained in this way; here, subject-verb agreement is enforced, since the num
requirements of both the subject and the verb are satisfied.
We will not discuss particular methods or algorithms for solving systems of
equations like the f-description in (34). Kaplan and Bresnan (1982) provide a
particularly clear presentation of one algorithm for solving such sets of equations,
and we briefly discuss some work on parsing and generation algorithms in §18.3.
There is a large body of work, summarized in detail by Rounds (1997), which
explores the logic of feature structures and their descriptions. Some work in LFG
and related frameworks discusses the operation of unification, first introduced in
linguistic work by Kay (1979) (see also Shieber 1986). Unification is an operation
(represented by () that combines compatible feature structures into a new feature
structure by taking the union of all the attribute-value pairs in the original
structures. For instance, the unification of the two feature structures f and g in
(39a) is the feature structure in (39b):
) * ) *
(39) a. a b c d
f g e [f g]
c d
 
a b
b. f ( g  c d 
e [f g]

The operation of unification on feature structures is related in a clear way to

the conjunction of constraints on those structures. Consider two f-descriptions
F and G, and their minimal solutions, the feature structures f and g. If the two f-
descriptions F and G are consistent and are taken to describe the same f-structure,
then the minimal solution to the conjunction of the two f-descriptions F ∧ G is
exactly the unification of the two f-structures f and g.
150 describing syntactic structures

5.2.2 Semantic forms

5.2.2.1 Uniqueness The value of the pred attribute is called a semantic form.
The difference between a semantic form and other atomic values is represented
notationally by the presence of single quotes. Semantic forms behave in a special
way in terms of the constraints described here: as discussed in §2.4.2, the semantic
form is instantiated to a unique value each time it appears in an f-description. For
example, the name David is lexically associated with an equation like the one in
(40), specifying its semantic form:

(40) ( f pred) = ‘David’

Abstractly, a semantic form like ‘David’ appearing in a lexical entry or phrase

structure rule can be thought of as abbreviating an unbounded number of distinct
forms with different indices (Kaplan and Bresnan 1982). For each use of a word or
rule associated with the semantic form, a new and distinct indexed form is chosen.
Thus, for example, a particular instance of use of the name David gives rise to an
f-structure like the one in (41), where we have arbitrarily chosen the number 3 as
the unique index:
+ ,
(41) f pred ‘David3 ’

This instantiated semantic form ‘David3 ’ is distinct from the semantic form
‘David7 ’ or ‘David25 ’ for different instances of the use of David.
Recall the discussion in §2.4 of f-structures in which two attributes have the
same value:

(42) David seemed to yawn.

 
 
 
 3’ 
 
f 
 ’ 
 
 

The verb seem requires its subj to be the same as its xcomp’s subj by means of an
equation like the following:

(43) ( f subj) = ( f xcomp subj)

The value of the pred attribute of the subj ‘David’ is instantiated to a unique value
for this instance of its use. An equally correct but less succinct way of representing
the same f-structure is:
 functional constraints 151

(44) David seemed to yawn.

 
 
 
 3’ 
 
f
  

 ’ 
  
 3 ’ 

Crucially, the indices on the semantic forms of the subj and the xcomp subj
f-structures are the same. The f-structure in (44) is not the same as the one in (45),
in which two different semantic forms for David have distinct indices:

(45) Incorrect f-structure for David seemed to yawn:

 
 
 
 3’ 
 
  
 ’ 
 
  
 4’


Distinctness of semantic forms is important in enforcing Consistency (§2.4.6.3).

As discussed by Simpson (1991), word order in Warlpiri is very free; the subject
may appear in any position in the sentence. However, even though subjects may
appear in either sentence-initial or sentence-final position, a sentence with two
subjects (one sentence-initial and one sentence-final) is ungrammatical (Simpson
1991: 93):5

(46) ⃰wati ka parnka-mi karnta

man.abs prs run-npst woman.abs
‘The man runs the woman.’

Intuitively, the sentence is unacceptable because of the simultaneous presence of

two different subjects. In slightly more formal terms, the presence of two different
semantic forms for the subject of the sentence causes a clash, and the resulting
f-structure is ill-formed:

5 In our discussion of Warlpiri in Chapter 4, we discussed example (32) (page 134) in which two
separate phrases contributed to the subj function. In that example, unlike (46), the two phrases are
compatible and can both appear in the same utterance, since one phrase is interpreted as the head and
the other is interpreted as a modifier of the head.
152 describing syntactic structures

(47) Ill-formed f-structure:

 
’
 
 

The clitic and full pronouns of Bosnian/Croatian/Serbian, as discussed by Franks

and King (2000), provide further evidence for the behavior of semantic forms. For
pronominal objects in Bosnian/Croatian/Serbian, a clitic pronoun (here, the third
person singular feminine clitic ju) is generally used:6

(48) Mirko ju je čitao.

Mirko 3f.sg.acc.clitic aux.3sg.clitic read
‘Mirko read it.’

Bosnian/Croatian/Serbian also has full pronominal forms (here, the third singular
feminine pronoun nju), used for emphasis, which do not appear in the second-
position clitic cluster:

(49) Mirko je čitao nju.

Mirko aux.3sg.clitic read 3f.sg.acc
‘Mirko read it.’

The f-structures for these examples are fundamentally similar; (50) gives an
abbreviated f-structure for both example (48) and example (49), with the obj
f-structure labeled f :
 
(50) ’
 
 
  
 
 
   
  
 f 
   



The lexical entries for the clitic and full pronoun contain the information in (51):

6 Following Franks and King, we translate the feminine full and clitic pronouns nju and ju as ‘it’
when they appear as the object of the verb čitao ‘read’.
 functional constraints 153

(51) ju (f pred) = ‘pro’

(f index pers) = 3
(f index num) = sg
(f index gend) = f
nju (f pred) = ‘pro’
(f index pers) = 3
(f index num) = sg
(f index gend) = f

In Bosnian/Croatian/Serbian, the clitic pronoun and the full pronoun cannot be

used in the same sentence:

(52) ⃰Mirko ju je čitao nju.

Mirko 3f.sg.acc.clitic aux.3sg.clitic read 3f.sg.acc
‘Mirko read it it.’

Despite the fact that the pred value contributed by both pronominal forms is ‘pro’,
the sentence is ungrammatical; again, as above, a clash is produced by multiple
specification of semantic forms with different indices as the value of the obj pred:7

(53) Ill-formed f-structure:

 
’
 
 
 
  
 1 2’

 
   
  
 f  

  
 

In general, following standard LFG practice, we will not display numeric indices on
semantic forms unless it is necessary for clarity, and two different semantic forms
are treated as distinct even if they look the same. If we want to indicate that the same
semantic form appears in two different places in the f-structure, as in example (42),
we will draw a line connecting the two occurrences.
In some cases, the value of an attribute other than pred might be required to be
uniquely contributed; for instance, the value of the tense attribute is contributed
by only a single form, and multiple contributions are disallowed:

(54) a. Is David yawning?

b. ⃰Is David is yawning?
c. ⃰Is David yawns?

7 In contrast to Bosnian/Croatian/Serbian, clitic doubling is possible in certain dialects of Spanish,

as discussed in §4.3.2, as well as other Slavic languages. This is because the pred value of the clitic
pronoun in these languages is optional, as we will see in §5.2.4.
154 describing syntactic structures

An instantiated symbol can be used as the value of the tense attribute in such a
situation. Like a semantic form, an instantiated symbol takes on a unique value
on each occasion of its use. In general, any syntactic uniqueness requirement for
an attribute can be imposed by the use of an instantiated symbol as the value of
that attribute. Notationally, instantiated symbols are followed by an underscore;
for example, to indicate that the value for the attribute tense is the instantiated
symbol prs, we write:

(55) ( f tense) = prs

5.2.2.2 Argument lists A semantic form, unlike other values, may contain an
argument list. In example (50) of this chapter, the pred value contributed by the
verb čitao ‘read’ is the complex semantic form ‘read%subj,obj&’. As discussed in
§2.4.6, this f-structure is complete and coherent because the requirements specified
by the semantic form for čitao ‘read’ are satisfied: the f-structure has a subj and an
obj, and there are no other governable grammatical functions in the f-structure
that are not mentioned in the argument list of čitao ‘read’. Additionally, since the
subj and obj arguments of čitao ‘read’ are semantic arguments appearing inside
the angled brackets of the argument list, each must contain a pred; nonsemantic
arguments mentioned outside the angled brackets in the argument list must also
be present in the f-structure, but are not required to contain a pred.

5.2.2.3 Reference to parts of the semantic form It is possible to use a special set
of attributes to refer to parts of a semantic form. For a simple semantic form like
‘David’, the attribute fn can be used to refer to the function component:

(56) If ( f pred) = ‘David’, then:

( f pred fn) = David

Numbered attributes such as arg1 and arg2 can be used to refer to the arguments
of the semantic form. For example, given the semantic form ‘read%subj,obj&’, the
following equations hold:

(57) If ( f pred) = ‘read%subj,obj&’, then:

( f pred fn) = read
( f pred arg1) = subj
( f pred arg2) = obj

Nonthematic arguments appearing outside the angled brackets are referred to by

numbered attributes such as notarg1 and notarg2:

(58) If ( f pred) = ‘seem%xcomp&subj’, then:

( f pred fn) = seem
( f pred arg1) = xcomp
( f pred notarg1) = subj
 functional constraints 155

More generally, for some semantic form p:

• (p fn) is the function component;

• (p arg1) is the first argument inside the angled brackets in the argument list,
and (p argn) is the nth argument inside the angled brackets in the argument
list;
• (p notarg1) is the first nonthematic argument outside the angled brackets in
the argument list, and (p notargn) is the nth nonthematic argument outside
the angled brackets.

5.2.3 Disjunction
An f-description can also consist of a disjunction of two or more descriptions.
When this happens, one of the disjuncts must be satisfied for the f-description
to hold.
For instance, the form met of the English verb meet is either a past tense form
or a past participle:

(59) I met/have met him.

This is reflected in the following disjunctive f-description in the lexical entry for
met, which says that the f-structure f for met must contain either the attribute-
value pair %tense, pst& or the attribute-value pair %vtype, pst.ptcp&:

(60) met ( f pred) = ‘meet%subj,obj&’

{( f tense) = pst | ( f vtype) = pst.ptcp}

There are two minimal solutions to this f-description:

(61) ’
f

b. ’
f

Each of these minimal solutions satisfies one of the disjuncts of the description.
Formally, a disjunction over descriptions is satisfied when one of the disjuncts
is satisfied:

(62) Disjunction:
A disjunction {d1 | . . . | dn } over f-descriptions d1 . . . dn holds of an
f-structure f if and only if there is some disjunct dk , 1 ≤ k ≤ n, that holds
of f .
156 describing syntactic structures

5.2.4 Optionality
An f-description can also be optional. When this happens, the f-description may
but need not be satisfied.
Bresnan and Mchombo (1987) show that verbs in Chicheŵa optionally carry
information about their subjects; in a Chicheŵa sentence, a subject noun phrase
may be either present (63a) or absent (63b):

(63) a. njûchi zi-ná-lúm-a alenje

bees 10subj-pst-bite-ind hunters
‘The bees bit the hunters.’
b. zi-ná-lúm-a alenje
10subj-pst-bite-ind hunters
‘They bit the hunters.’

Bresnan and Mchombo propose that the verb zi-ná-lúm-a ‘bit’ optionally con-
tributes an f-description constraining the value of the pred attribute of its subject.
This optional f-description is enclosed in parentheses:

(64) zi-ná-lúm-a: (( f subj pred) = ‘pro’)

Since the equation ( f subj pred) = ‘pro’ is optional, it may but need not contribute
to the minimal solution of the f-description for the sentence. If an overt subject
noun phrase does not contribute its own pred value, the f-structure for the sen-
tence is incomplete unless this equation is satisfied, and in this case the wellformed
f-structure for the subj contains the pair %pred,‘pro’&. If an overt subject noun
phrase appears, the equation cannot be satisfied, since the pred value of the overt
subject would produce a clash; instead, the pred value for the subj is the one
specified by the subject noun phrase:

(65) a. njûchi zi-ná-lúm-a alenje

bees 10subj-pst-bite-ind hunters
‘The bees bit the hunters.’
 
’
 
 
 
 
f 
 
 
 

b. zi-ná-lúm-a alenje
10subj-pst-bite-ind hunters
‘They bit the hunters.’
 
’
 
 
 
 
f 
 
 
 
 functional constraints 157

A similar analysis is appropriate for languages that allow clitic doubling. As dis-
cussed in §4.3, the River Plate and Peruvian dialects of Spanish allow either a clitic
or a full noun phrase object to appear:

(66) a. Juan vió a Pedro.

Juan saw prep Pedro
‘Juan saw Pedro.’
b. Juan lo vió.
Juan 3sg.acc.clitic saw
‘Juan saw him.’

Unlike many other dialects of Spanish, in these dialects the clitic pronoun can co-
occur with a full noun phrase object:

(67) Juan lo vió a Pedro.

Juan 3sg.acc.clitic saw prep Pedro
‘Juan saw Pedro.’

We account for these facts by assuming that in the River Plate and Peruvian
dialects, the pred value contributed by the clitic pronoun lo is optional:

(68) Pedro ( f pred) = ‘Pedro’

lo (( f pred) = ‘pro’)

A skeletal f-structure for (66a) and (67) is:

 
(69) ’
 
 
 
 
f

When a full noun phrase object is present, the optional equation contributing the
pred value of the clitic pronoun is not satisfied; if two preds were present, the
example would not satisfy Consistency. When there is no full noun phrase, in order
to satisfy Completeness, the pred contributed by the clitic noun phrase appears.
The f-structure for example (66b) is given in (70):
 
(70) ’
 
 
 
 
f

Formally, optionality of an f-description d is treated like a disjunction between d

and the f-description true, a statement that is always true:

(71) Optionality:
An f-description d optionally holds of an f-structure f if and only if the
disjunction {d | true} holds of f .
158 describing syntactic structures

5.2.5 Negation
An f-description can be negated; when this happens, the f-description must not be
satisfied. Notationally, negation of a single equation is indicated by a diagonal line
through the equals sign: f += g means that f does not equal g.
For example, as discussed in §2.5.3, Quirk et al. (1972: §15.6) claim that it is not
possible to use the complementizer if in the clausal complement of certain verbs,
including the verb justify:

(72) a. I know whether/if David yawned.

b. You have to justify whether/ ⃰if your journey is really necessary.

We can analyze the verb justify as described by Quirk et al. differently from a
verb like know by prohibiting the value if for the attribute compform in its comp
argument:

(73) justify ( f comp compform) += if

The f-structure in (74) satisfies this constraint:

(74) Chris justified whether David deserved the prize.

 
’
 
 
  
 
 ’ 
  
  
 
f 
 
  
  
  
  
  
  
 
+

However, the f-structure in (75) does not satisfy the constraint; the offending value
is circled:

(75) ⃰Chris justified if David deserved the prize.

 
’
 
 
 
   

 ’  
  
   
  
f   
  
   
  
   
  
   
  
 
+
 functional constraints 159

This example shows that a single equation can be negated, requiring a particular
attribute-value pair not to appear. More generally, it is possible to negate not
just a single equation, but an entire f-description. The following expressions are
notationally equivalent:

(76) ( f a) += v ≡ ¬[( f a) = v]

The negation of a conjunction of descriptions holds just in case at least one of

the descriptions does not hold. For example, the base form for verbs in English
also serves as the present tense form for all person/number combinations except
third person singular: I/you/⃰he/we/they yawn. We might analyze this by means of
a constraint like the one in (77), which states that the index f-structure f for the
subject of a bare verb must not contain both %pers, 3& and %num, sg&:

(77) ¬[( f pers) = 3

( f num) = sg]

Formally, negation of f-descriptions is defined in the following way:

(78) Negation:
A negated f-description ¬d holds of an f-structure f if and only if the
description d does not hold of f .

5.2.6 Existential constraints

An f-structure may be required to contain an attribute, but its value may be uncon-
strained: this kind of constraint is called an existential constraint. The f-structural
requirement of Completeness (§2.4.6.1) is a kind of existential constraint: Com-
pleteness requires the presence of all of the governable grammatical functions
specified by a predicate, but does not place any constraints on the particular values
of these functions.
Existential constraints can be used in the analysis of relative clauses. The English
relative clause must be tensed, but no particular value for the tense attribute is
required:

(79) a. the man who yawned

b. the man who yawns
c. the man who will yawn
d. ⃰the man who yawning

The f-structure for example (79a) is shown in (80). Note that the f-structure labeled
f contains the attribute tense with value pst:
160 describing syntactic structures

(80) the man who yawned

f ’

We can enforce the requirement for relative clauses to be tensed by means of a

constraint like the following:

(81) ( f tense)

This constraint requires the f-structure f to contain the attribute tense, but it
does not constrain the value of the tense attribute; any value is acceptable. The
f-structure in (80) satisfies this constraint. However, an f-structure like the one in
(82) does not satisfy the constraint, since it has no tense attribute:

(82) ⃰the man who yawning

f ’

Formally, an existential constraint has the following interpretation:

(83) Existential constraint:

The existential constraint ( f a) holds of an f-structure f if and only if there
is some value v for which the pair %a, v& ∈ f .

5.2.7 Negative existential constraints

Just as an f-structure can be required to contain some attribute, it can be prohibited
from containing some attribute: this is a negative existential constraint. The f-
structural requirement of Coherence is a constraint of this kind (§2.4.6.2): a
grammatical function that is not mentioned in the argument list of the pred must
not appear in the f-structure.
 functional constraints 161

Another use of a negative existential constraint is in the analysis of participial

modifiers, as discussed by Bresnan (1982a). Such modifiers must not be tensed:
(84) a. Scratching his head, Chris yawned.
b. Struck on the head, Chris slumped to the floor.
c. ⃰Scratched/Scratches his head, Chris yawned.
This constraint can be expressed in the following way:

(85) ¬( f tense)

The constraint in (85) ensures that the f-structure f has no tense attribute. The
constraint is satisfied in (86):

(86) Scratching his head, Chris yawned.

 
’
 
 
 
 
 
    

   
 
 
 
  




  ’ 




 
  


  


 
   


   

 
 
f     
    
 
    
   
 
     
 
 
     
 
 
       
  

 
     
 
  
 
 
 


Formally, a negative existential constraint is interpreted in the following way:

(87) Negative existential constraint:

The negative existential constraint ¬( f a) holds of an f-structure f if and
only if there is no value v for which the pair %a, v& ∈ f .

5.2.8 Defining and constraining equations

Besides defining equations like ( f a) = v, LFG allows constraining equations,
which contribute in a different way to the solution: defining equations determine
the minimal solution, and constraining equations check that the minimal solution
is wellformed. An example will help to show the difference between the two kinds
of equations.
In English, a clausal argument need not contain the complementizer that when
it bears the grammatical function comp, but it must contain that when it is a subj:
162 describing syntactic structures

(88) a. Chris thought [that David yawned].

b. Chris thought [David yawned].
c. [That David yawned] surprised Chris.
d. ⃰[David yawned] surprised Chris.

We assume that the lexical entry for the complementizer that contributes the
following defining equation:

(89) ( f compform) = that

In (88c), the clausal argument that David yawned is the subj of the clause, and
contains the attribute-value pair %compform, that&:

(90) That David yawned surprised Chris.

 
’
 
 
  

 ’ 

  
  
 f 
  
 
 
 
 

The defining equation in (89) is satisfied, because the f-structure labeled f has
an attribute compform with value that. Hence, (90) is an acceptable minimal
solution to an f-description that includes the defining equation in (89).
In contrast, the pair %compform, that& is not required to belong to the
f-structure of a comp; example (88b) is grammatical, and its f-structure is
wellformed:

(91) Chris thought David yawned.

 
’
 
 
 
 
  
 
 ’ 
  
 f 
 

The equation in (89) is not a part of the f-description for example (88b), since the
word that does not appear; in the minimal solution to the f-description for this
sentence, f does not have an attribute compform with any value.
 functional constraints 163

Our task, then, is to require a clausal subject such as the one in (90) to contain
the attribute compform with value that, and thus to ensure the presence of the
complementizer that. To impose this requirement, we can use the constraining
equation in (92):

(92) ( f compform) =c that

Notationally, a constraining equation differs from a defining equation by the

presence of the subscript c on the equals sign. A constraining equation is not
used in determining the minimal solution to an f-description. Instead, it imposes
an additional requirement on the minimal solution obtained from the defining
equations in the f-description: it requires the pair %compform, that& to be in the
minimal solution for f , as is the case in (90). Some other defining equation must
specify this attribute-value pair for the final solution to be acceptable.
The constraining equation in (92) holds of the f-structure in (90), since the
f-structure labeled f contains the pair <compform, that>. It does not hold of
the f-structure in (91). It is only when the complementizer that contributes the
defining equation in (89) that the constraining equation in (92) can be satisfied.
Notice in particular that we cannot impose a requirement for the presence of
that via a defining equation. Introducing an additional defining equation like
( f compform) = that would just ensure that the attribute compform with its
value that is a part of the minimal solution for f , no matter whether the word that
is present in the sentence or not.
We can also use constraining equations to encode requirements on semantic
forms which would otherwise be impossible to state. Recall from §5.2.2 that each
use of a semantic form is instantiated to a unique value: we cannot require the
object of an f-structure f to be pronominal by introducing an equation like the one
in (93), since such an equation would introduce a unique semantic form value for
the pred of the obj which is incompatible with any other semantic form introduced
by another equation:

(93) Introducing the semantic form ‘pro’ as the value of the obj’s pred:
( f obj pred) = ‘pro’

Instead, we can require an object to be pronominal by requiring its pred fn to be

pro. The constraining equation in (94) holds of all semantic forms contributed by
equations like the one in (93) (see §5.2.2.3 for an explanation of the attribute fn):

(94) Requiring the obj to be pronominal:

( f obj pred fn) =c pro

We can propose a formal definition for constraining equations, following Kaplan

and Bresnan (1982):

(95) Constraining equation:

( f a) =c v holds if and only if f is an f-structure, a is a symbol, and the
pair %a, v& is in the minimal solution for the defining equations in the
f-description of f .
164 describing syntactic structures

Kaplan and Bresnan (1982) provide an interesting discussion of the formal role of
constraining equations and existential constraints in LFG (see also Johnson 1999;
Saraswat 1999). As they note, constraining equations and existential constraints
are particularly useful when the presence of a feature is associated with a marked
value for the feature, with the absence of the feature indicating the unmarked value
(§2.5.2). For instance, all marked forms of a particular paradigm may be associated
with a certain feature: assume, for instance, that only passive verbs have a value pass
for the voice feature, and active verbs have no voice feature. In this situation, if a
particular voice of a verb is required, a constraining equation must be used to check
for the presence or absence of the voice feature; a defining equation mentioning
the voice feature would be compatible with passive verb forms as well as with
active forms unspecified for the voice feature, the wrong result.
In certain limited situations, the use of a defining equation does not produce a
different result from the use of a constraining equation. A constraining equation is
used to check whether an attribute and its value are present in the minimal solution
of an f-description. If we know that a particular attribute is always present in a
certain type of f-structure, we need not use a constraining equation to check its
value: in that situation, using a defining equation produces the same result. For
instance, suppose that all noun phrases in English are marked for number, so
that they are all either singular or plural. Suppose further that the value for the
number feature is not an instantiated symbol and can be specified by more than
one defining equation. Then, in specifying number agreement with a noun phrase,
it does not matter whether the specification involves a defining equation or a
constraining equation: we know that the minimal solution to the constraints always
contains a number attribute with some value, since all noun phrases are specified
for number. We can include an additional defining specification of the feature, or
we can use a constraining equation to ensure that the feature and its required value
are present in the minimal f-structure satisfying the defining equations.

5.2.9 Feature defaults

Some features can be treated as having default values which can be overridden by
the specification of a non-default value. Let us assume, for example, that we would
like to treat the attribute case as having the default value nom; we can accomplish
this by means of the following disjunctive specification:8

(96) The
4 attribute case has the default value nom:
5
( f case) ( f case) = nom
( f case) += nom

According to this disjunctive constraint, one of two alternative descriptions must

hold of an f-structure f :

8 Disjunction is discussed in §5.2.3, negation is discussed in §5.2.5, and existential constraints are
discussed in §5.2.6.
 functional constraints 165

• The existential constraint ( f case) holds: f has a case attribute; furthermore,

the value of ( f case) is not the default value nom; or
• f has an attribute case whose value is (perhaps redundantly) specified as the
default value nom.

If this constraint is present, the f-structure f must have some value for the attribute
case. Some other component of the grammar may specify a non-default value
(a value other than nom) for case. If there is no other specification of a different
value for case elsewhere in the full f-description, then the value of the case
attribute for f is the default, nom.
Some works adopt a simpler treatment of defaults:

(97) The attribute case has the default value nom, alternative (simpler) specifi-
cation:
6 7
( f case) ( f case) = nom

According to this simple specification, there are two possibilities:

• The existential constraint ( f case) holds: a value for case is specified by some
component of the grammar; or
• the value for case is the default value, nom.

This simpler statement has the same effect as the more complex one in (96), in
that it specifies a default value nom for the attribute case. However, it is less
constrained, in that it allows more than one equivalent way of satisfying a default
specification in some circumstances. If another component of the full f-description
specifies the default value nom for the case attribute, both alternatives in (97) are
satisfied: (i) there is a value for the case attribute, and (ii) it is the default value,
nom. This means that there may be more than one solution for an f-description
with this simpler form of default specification, giving rise to the appearance of
ambiguity where no ambiguity exists. Thus, we prefer the more complex statement
of defaults as in (96), since it does not permit both alternatives to be satisfied at the
same time.
The general form of default specifications is, then, as follows:

(98) The attribute a has the default value v:

4 5
( f a) ( f a) = v
( f a) += v

It is important for default specifications to interact in the desired way with

specifications of feature values in other components of the grammar. Grammars
appealing to default specifications must use defining equations to specify a non-
default value for a feature, and constraining equations must be used to check the
value of a default feature in circumstances where a non-default value should not be
imposed (for more discussion of defining and constraining equations, see §5.2.8).
166 describing syntactic structures

5.2.10 Implication
It is often convenient and intuitive to express a linguistic generalization as an
implication. For example, implicational constraints can be used to encode effects
of the person hierarchy in Lummi transitive sentences. Building on work by Jelinek
and Demers (1983), Bresnan et al. (2001) observe that Lummi requires the person
hierarchy in (99a) to be aligned with the grammatical function hierarchy, with the
result that a first or second person argument cannot be lower than a third person
argument on the grammatical function hierarchy:

(99) a. Person hierarchy in Lummi: 1, 2 > 3

b. Grammatical function hierarchy: subj > obj > obl

This means that a third-person subject may not appear with a first or second person
object in an active Lummi sentence, and in a passive sentence, a third person
subject may not appear with a first or second person oblique argument.9 Jelinek
and Demers (1983) provide the data in (100–101) to illustrate these patterns.10 In
the active sentences in (100), a third person subject may not appear with a first or
second person object, as in (100c), but other combinations are allowed:

(100) Active sentences:

a. x.či-t=sən cə swəyʔqəʔ
know-active=1nom the man
‘I know the man’. 1subj, 3obj
b. x.či-t=sxw cə swəyʔqəʔ
know-active=2nom the man
‘You know the man.’ 2subj, 3obj
c. ⃰‘The man knows me/you.’ ⃰3subj, 1/2obj
d. x.či-t-s cə swəyʔqəʔ cə swiʔqoʔəƚ
know-active-3 the man the boy
‘The man knows the boy.’ 3subj, 3obj

In the passive sentences in (101), a third person subject may not appear with a
first or second person oblique phrase, as in (101c), but other combinations are
allowed:

(101) Passive sentences:

a. x.či-tŋ=sən cə swəyʔqəʔ
know-pass=1nom the man
‘I am known by the man.’ 1subj, 3obl
b. x.či-tŋ=sxw ə cə swəyʔqəʔ
know-pass=2nom by the man
‘You are known by the man.’ 2subj, 3obl

9 Constraints on voice and the person hierarchy in Lummi are more complicated than the data in
this section indicate; for a detailed discussion, see Jelinek and Demers (1983).
10 Jelinek and Demers do not provide Lummi equivalents for the ungrammatical patterns.
 functional constraints 167

c. ⃰‘The man is known by me/you.’ ⃰3subj, 1/2obl

d. x.či-tŋ cə swiʔqoʔəƚ ə cə swəyʔqəʔ
know-pass the boy by the man
‘The boy is known by the man.’ 3subj, 3obl

Thus, we would like a way of formally stating the generalization in (102):

(102) In Lummi, if the subject is third person, the object or oblique phrase cannot
be first or second person.

An implication is written with the double right arrow “⇒.” For any clausal
f-structure f in Lummi, the following implication holds:11

(103) If f ’s subject is third person, its object or oblique is not first or second
person:
( f subj index pers) = 3 ⇒ ( f {obj|obl} index pers) += 1
( f {obj|obl} index pers) += 2

Since negation and disjunction are permitted in f-descriptions, we can define the
implication relation by using negation and disjunction. In the general case, the
implication in (104a) can be paraphrased as in (104b):12

(104) a. d1 ⇒ d2 (“d1 implies d2” or “if d1 holds, then d2”)

b. “either d1 does not hold, or d1 holds and we can introduce the
f-description d2”

As the paraphrase in (104b) indicates, d1 in the implicational statement in (104a)

has the status of a constraining equation: we first check the minimal solution for
the rest of the f-description of an utterance to see if d1 holds, and if it does, we
augment the f-description for the utterance with d2.
We can give any f-description the status of a constraining equation by negating it
twice: ¬¬d1 (“it is not the case that d1 is false”). Twice negated, d1 must hold of the
minimal solution for the rest of the f-description, but it does not have the status of
a defining equation. We can now define the implication relation for f-descriptions:

11 For simplicity, we assume that the value of the pers feature is one of the atomic values 1, 2, or 3. The
complex person features discussed in §2.5.7.2 do not provide a means of referring to first and second
person as a natural class of feature values, to the exclusion of third person; further refinement to the
values proposed for the pers feature by Otoguro (2015) would allow us to simplify the rule in (103) by
referring to such a class.
12 Formally, this equivalence is motivated by the conditional law, which states that the implication
d1 → d2 holds if and only if either d1 fails to hold, or d2 holds:
[d1 → d2] ⇔ {¬d1 | d2}
Since either an f-description d1 or its negation ¬d1 is true of any f-structure, we can use the stronger
formulation of equality in (104).
168 describing syntactic structures

(105) Implication:
Given two f-descriptions d1 and d2, d1 ⇒ d2 (“d1 implies d2”) holds if
and only if {¬d1 | ¬¬d1, d2} (“either d1 does not hold, or it is not the case
that d1 does not hold, and d2 holds”).

Bresnan et al. (2016: 61) provide more discussion of implicational constraints.

5.3 The c-structure/f-structure correspondence

5.3.1 Annotated phrase structure rules
Chapter 4 discussed universally valid correspondences between c-structure and
f-structure: a c-structure head and the phrases it projects correspond to the same
f-structure, for example. Here we show how these correspondences are formally
stated.
Recall that the 𝜙 function defines a relation between c-structure nodes and
f-structures:

(106) V
’
V

yawned

To make the following discussion simpler, we assume that the rule expanding V# is:

(107) V# −→ V

As discussed in §4.2.1, a phrase and its head correspond to the same f-structure.
Here, V is the head of V# , and so V# and V correspond to the same f-structure.
We would like a way to express this fact.
We accomplish this by annotating the V node with an expression requiring the
f-structure of the V to be the same as the f-structure for the V# . In general, any
daughter node—that is, any node on the right-hand side of a phrase structure
rule—may be annotated with constraints on the relation between its f-structure
and the f-structure of the mother node. If the daughter node is the head, the
f-structures must be the same. If the daughter node is a nonhead, its f-structure
bears some relation (say, the obj relation) in the mother’s f-structure.
In order to do this, we need a notation for the following concepts:

(108) the current c-structure node (“self ”): ⃰

the immediately dominating node (“mother”): ˆ⃰
the c-structure to f-structure function: 𝜙

The symbol ⃰ stands for the node corresponding to the rule element on which the
constraint is written; note that this use of ⃰ is not related to the Kleene star notation
 the c-structure/f-structure correspondence 169

indicating that a category or attribute can be repeated zero or more times, as

discussed in §5.1.1. The symbol ˆ⃰ stands for the node immediately dominating the
⃰ node. In some LFG literature, the immediately dominating node ˆ⃰ is represented
by means of the mother function M, as M(⃰); the two expressions ˆ⃰ and M(⃰) are
equivalent.
The function 𝜙 applies to a c-structure node to give the f-structure correspond-
ing to that node. Thus, 𝜙(⃰) is the f-structure corresponding to the current node,
and 𝜙(ˆ⃰) is the f-structure corresponding to the mother node in a rule.
To indicate that the f-structure for the V# is the same as the f-structure for the
V in the rule given in (107), we can write:

(109) V# −→ V
𝜙(ˆ⃰) = 𝜙(⃰)
mother’s (V# ’s) f-structure = self ’s (V’s) f-structure

A convenient abbreviation is usually used for 𝜙(ˆ⃰) and 𝜙(⃰):

(110) 𝜙(ˆ⃰) (mother’s f-structure) =↑

𝜙(⃰) (self ’s f-structure) =↓

The intuition behind this notation comes from the way trees are usually repre-
sented: the up arrow ↑ points to the mother node, while ↓ points to the node
itself. Using these abbreviations, we can rewrite the rule in (109) in the following
more standard way:

(111) V# −→ V
↑=↓
mother’s f-structure = self ’s f-structure

This rule represents the following configuration:

(112) V# []

In some LFG literature, f-structure annotations are written above the node labels
of a constituent structure tree, making the intuition behind the ↑ and ↓ notation
clearer; written this way, the arrows point to the appropriate phrase structure
nodes:

(113) V#

↑=↓
V
170 describing syntactic structures

In the following, we will stick to the more common practice of writing functional
annotations beneath the node label in the phrase structure rule, as in (111).
Let us turn to a slightly more complicated rule, one that describes the following
c-structure and f-structure:
+ ,
(114) obj [ ]
V #

V NP

Here, the NP bears the obj function. Thus, the rule in (115) contains the additional
information that the f-structure for the NP daughter of V# is the value of the obj
attribute in the f-structure for the V# :

(115) V# −→ V NP
𝜙(ˆ⃰) = 𝜙(⃰) (𝜙(ˆ⃰) obj) = 𝜙(⃰)
mother’s f-structure’s obj = self ’s f-structure

or, in the abbreviated notation:

(116) V# −→ V NP
↑=↓ (↑ obj) = ↓

Some LFG work follows another abbreviatory convention according to which the
annotation ↑ = ↓ is omitted when it is the only annotation on a node. According
to this convention, a rule like (116) can be written as:

(117) V# −→ V NP
(↑ obj) = ↓

In this book, we will try to be as explicit as possible, so we will not follow this
convention of omitting equations.
Restricting ourselves to rule annotations referring only to the f-structures corre-
sponding to a daughter node and the mother node in a phrase structure rule makes
a strong claim about the local applicability of syntactic constraints. For instance,
we cannot refer to the grandmother node in a tree, or to its f-structure. This means
that nonlocal syntactic relations are statable only in functional terms, not in terms
of constituent structure configuration. Within the tree, only local phrase structure
relations can be constrained by phrase structure rules.13

13 Some analyses in LFG involve reference to the f-structures of sister nodes in the phrase structure
rule: see §6.4.
 the c-structure/f-structure correspondence 171

5.3.2 Lexical entries

We can use the same notation in writing lexical entries that we used in annotations
on phrase structure rules. For instance, assume that we would like to describe a
c-structure/f-structure pair like the following:

(118) V ’
yawned

The following lexical entry for yawned provides information about the f-structure
corresponding to V, the immediately dominating c-structure node:

(119) yawned V (↑ pred) = ‘yawn%subj&’

(↑ tense) = pst

This lexical entry asserts that the f-structure ↑ corresponding to the V node
immediately dominating yawned has an attribute pred whose value is the semantic
form ‘yawn%subj&’, and an attribute tense whose value is pst. The f-structure
displayed in (118) is the minimal solution to these constraints, the smallest
f-structure that satisfies the constraints.14

5.3.2.1 ↑ in lexical entries The use of ↑ in a lexical entry is exactly the same as
its use in a rule: ↑ refers to the c-structure node dominating the word itself. This
can be seen more easily if we recast the lexical entry in (119) in the equivalent
phrase structure rule format:

(120) V −→ yawned
(↑ pred) = ‘yawn%subj&’
(↑ tense) = pst

5.3.2.2 ↓ in lexical entries In most cases, lexical entries specify only information
about ↑, the f-structure of the immediately dominating c-structure node. Some
analyses also refer to properties of the f-structure ↓ corresponding to the word
itself and how the f-structure for the word relates to the f-structure for the
immediately dominating node. For example, Zaenen and Kaplan (1995) explore
an analysis of word order constraints in Dutch which assumes lexical entries such
as the partial entry in (121) for the verb wil ‘want’:

(121) wil V (↑ pred) = ↓

↓ = ‘want%subj,xcomp&’

14 We rely here on a simplified view of the relation between elements of the string and nodes in the
c-structure. See §3.5 for a more complete view, and Chapter 11 for full details.
172 describing syntactic structures

In this lexical entry, ↑ refers to the f-structure of the c-structure node V, and ↓
refers to the f-structure of the terminal node, the word wil. The structure associated
with wil is the semantic form ‘want%subj,xcomp&’. According to this lexical entry,
then, the value of the pred attribute of wil is the structure associated with the
terminal node wil. The entire configuration is depicted in (122):
) *
(122) V pred ‘want%subj,xcomp&’
8 9: ;
wil

Bresnan et al. (2016) also use ↓ in lexical entries of words containing incorporated
pronouns to anchor these pronouns to particular positions in the tree (for a discus-
sion of incorporated pronouns, see §14.1). For instance, Bresnan et al. (2016: 160)
propose that the object marker (abbreviated as om) that appears with Chicheŵa
transitive verbs has the following lexical entry, where agr is an abbreviatory symbol
representing the attributes pers, num, and gend:

(123) om- (↑ obj) = ↓

(↓ agr) = α
(↓ pred) = ‘pro’

According to this analysis, a Chicheŵa transitive verb with an incorporated object

pronoun prefix participates in the configuration in (124):

(124)  
V
 
agr α

5.3.3 An example
We now have the notational equipment to do a complete analysis of a sentence
like David yawned. For clarity, in the following example the rules and lexical
entries have been considerably simplified; for example, the phrase structure rules
expanding VP and NP are clearly too impoverished to account for very many
constructions in English.
We assume the following annotated phrase structure rules:

(125) IP −→ NP I#
(↑ subj) = ↓ ↑=↓

I# −→ I VP
↑=↓ ↑=↓
 the c-structure/f-structure correspondence 173

VP −→ V
↑=↓

NP −→ N
↑=↓

In English, the specifier of IP is associated with the grammatical function subj.

Other daughter nodes in the rules in (125) are heads or complements of functional
categories, and are associated with the annotation ↑ = ↓, requiring that they
correspond to the same f-structure as the mother node. Parentheses around each
annotated daughter node indicate that all of the daughter nodes are optional;
see §3.3.7 for a discussion of phrase structure optionality.
We also make use of the following simplified lexical entries:

(126) yawned V (↑ pred) = ‘yawn%subj&’

(↑ tense) = pst
David N (↑ pred) = ‘David’

These rules and lexical entries admit the following tree, with as yet uninstantiated
variables ↑ and ↓ over f-structures:

(127) David yawned.

IP

NP I#
(↑ subj) = ↓ ↑=↓

N VP
↑=↓ ↑=↓

David V
(↑ pred) = ‘David’ ↑=↓

yawned
(↑ pred) = ‘yawn%subj&’
(↑ tense) = pst

The next task is to instantiate the ↑ and ↓ metavariables to the f-structures that
they stand for in this example. It will be useful to have names for the f-structures
corresponding to each node. Taking advantage of the coincidental fact that in this
tree, every node has a different label, we will give the name fv to the f-structure
corresponding to the node labeled V, fvp to the f-structure for the node labeled VP,
and so on.
We begin with the information contributed by the lexical entry for David. The
f-structure variable ↑ in the annotation (↑ pred) = ‘David’ for David refers to fn ,
174 describing syntactic structures

the f-structure of the N node immediately dominating the leaf node David, and so
we replace ↑ in that expression with fn :

(128)
IP

NP I#
(↑ subj) = ↓ ↑=↓

N VP
↑=↓ ↑=↓

David V
( fn pred) = ‘David’ ↑=↓

yawned
(↑ pred) = ‘yawn%subj&’
(↑ tense) = pst

Let us now consider the N node. Its annotation is ↑ = ↓, meaning that the f-
structure fnp corresponding to its mother node NP is the same as fn , the f-structure
for the N node:

(129)
IP

NP I#
(↑ subj) = ↓ ↑=↓

N VP
fnp = fn ↑=↓

David V
( fn pred) = ‘David’ ↑=↓

yawned
(↑ pred) = ‘yawn%subj&’
(↑ tense) = pst

In a similar way, we replace the ↑ and ↓ nodes in the rest of the tree with the names
of the f-structures they refer to:
 the c-structure/f-structure correspondence 175

(130)
IP

NP I#
( fip subj) = fnp fip = fi#

N VP
fnp = fn fi# = fvp

David V
( fn pred) = ‘David’ fvp = fv

yawned
( fv pred) = ‘yawn%subj&’
( fv tense) = pst

We now have an instantiated f-description of the f-structure for this sentence:

(131) ( fip subj) = fnp

fnp = fn
( fn pred) = ‘David’
fip = fi#
fi# = fvp
fvp = fv
( fv pred) = ‘yawn%subj&’
( fv tense) = pst

According to these constraints, fip , fi# , fvp , and fv all name the same f-structure,
which has three attributes, pred, tense, and subj. The subj of this f-structure is fnp ,
which is also called fn . The f-structure for this sentence is the minimal solution to
these constraints, the f-structure that satisfies all of these constraints and contains
no extra structure not mentioned in the constraints:

(132)  
’
 
fip , fi , fvp , fv  
fnp , fn

We have now deduced that the sentence David yawned has the following anno-
tated c-structure and f-structure:
176 describing syntactic structures

(133)  
’
 
IP  

NP I
( fip fnp fip = fi

N VP
fnp = fn fi = fvp

David V
( fn fvp = fv

yawned
( fv ’
( fv

To summarize: an f-structure is admitted in correspondence with a particular

constituent structure tree if the annotations on the phrase structure rules and
the lexical items admit the pairing of that tree with that f-structure, and if the
f-structure is the minimal solution—the smallest f-structure—that satisfies the
constraints in the annotations and the lexical entries.

5.4 Variation in grammatical function encoding

Grammatical functions are encoded in different ways in different languages, and
languages may employ mixed or multiple strategies for grammatical function
encoding. These typological differences are reflected in the constraints associated
with lexical items and phrase structure rules.
As discussed in §2.2.2, in some languages the grammatical function of a phrase
is determined by its constituent structure position. Languages of this type make
use of configurational encoding (Bresnan 1982a), and obey the principle of endo-
centricity (Bresnan et al. 2016): specifier and complement positions at constituent
structure are associated with particular grammatical functions by means of anno-
tations on phrase structure rules.
In other languages, there may be no uniform position in which a particular
grammatical function must appear; instead, the grammatical role of a phrase is
marked morphologically or by means of an adposition or particle. This is what
Bresnan (1982a) calls nonconfigurational encoding, and Bresnan et al. (2016) call
lexocentric organization: an association between morphological marking and syn-
tactic function. Languages may tend to employ one of these types of encoding
more heavily, but there are many cases in which a single language employs both
types. For instance, in English, the object grammatical functions are encoded
configurationally; in contrast, the oblique functions are encoded by means of
 variation in grammatical function encoding 177

prepositional marking. Below, we will examine other languages making use of a

combination of these strategies.
In an important typological study, Nichols (1986) shows that some languages are
head marking and some are dependent marking: in other words, the surface indi-
cation of grammatical function can appear either on the argument of a predicate
(dependent marking) or on the predicate itself (head marking). Often, this surface
indication involves nonconfigurational encoding, with morphological marking
of grammatical function on either the head or the dependent. In fact, though,
configurational languages can be said to exhibit a type of dependent marking, since
a surface syntactic property of the dependent—its constituent structure position—
indicates its grammatical function. We will see examples of both head-marking
and dependent-marking languages below.
In the following, we illustrate the kind of variability that LFG predicts by exam-
ining simple structures in four typologically different languages: Balinese, Ancient
Greek, Chicheŵa, and Bulgarian. In the final section, we provide a more detailed
snapshot of the copular construction in several languages as an illustration of
variability with respect to a single construction. These sketches are not intended as
complete analyses of these languages or the constructions we discuss; only enough
detail is provided so that broad outlines become evident. For very interesting
discussions of nonconfigurationality, head marking, and dependent marking in a
variety of languages, see Nordlinger (1998) and Bresnan et al. (2016). The analyses
in this section make use of the following definition of the abbreviatory symbol
ggf, which ranges over governable grammatical functions; see §6.1.2 for more
discussion of the use of abbreviatory symbols such as ggf:

(134) ggf ≡ {subj | obj | obj𝜃 | comp | xcomp | obl𝜃 }

5.4.1 Balinese
In Balinese, as described by Arka (2003), the grammatical functions subj, obj, and
obj𝜃 are primarily encoded configurationally. This means that phrase structure
rules contain specifications of particular grammatical functions: just as in English
(§4.2), the specifier position of IP is filled by the subject, and the object appears
as the first nominal complement of V.15 These principles of mapping between
c-structure and f-structure configurations are reflected in the annotations on the
rules given in (135): heads of phrases bear the annotation ↑ = ↓, ensuring that a
phrase and its head correspond to the same f-structure; the specifier of the func-
tional category IP bears the annotation (↑ subj) = ↓, ensuring that it is associated
with the grammatical function subj; the S complement of the functional category I
is an f-structure co-head, bearing the annotation ↑ = ↓; and the complement of the
lexical category V bears the annotation (↑ obj) = ↓, ensuring that it is associated
with the grammatical function obj:

15 Balinese exhibits a symmetric, “Philippine-type” alternation in morphosyntactic alignment in

transitive verbs, as described in §2.5.5.3: in objective voice (ov) the theme or patient argument appears
as the subject, while in agentive voice (av) the more agentive argument appears as the subject.
178 describing syntactic structures
< = < # =
(135) IP −→ NP I
(↑ subj) = ↓ ↑=↓
< = < =
I# −→ I S
↑=↓ ↑=↓

S −→ { VP | NP }⃰
↑=↓ (↑ ggf) = ↓
< # =
VP −→ V
↑=↓
< = < =
V# −→ V NP
↑=↓ (↑ obj) = ↓

Predicates specify a list of the governable grammatical functions that they require:

(136) ngamah V (↑ pred) = ‘eat%subj,obj&’

The c-structure and f-structure for the sentence sampi ngamah padang ‘a cow eats
grass’ are given in (137), with the relation between the clausal head c-structure
nodes and the main f-structure indicated by arrows:

(137) sampi ngamah padang

cow eat.av grass
‘A cow eats grass.’
IP

NP I
(↑ ↓ ↑ =↓

N S
↑ =↓ ↑ =↓
 
’
sampi VP  
 
(↑ ↑ =↓  

V
↑ =↓

V NP
↑ =↓ (↑ ↓

ngamah N
(↑ ’ ↑ =↓

padang
(↑
 variation in grammatical function encoding 179

The requirement for the presence of a subj and an obj is lexically specified by the
verb, and the grammatical function of each argument is determined by its phrase
structure position.

5.4.2 Ancient Greek

Ancient Greek is typologically quite different from Balinese. Ancient Greek is a
nonconfigurational language; unlike in Balinese, phrase structure configuration
does not determine the grammatical function of an Ancient Greek phrase.16
Instead, grammatical function is determined by morphological casemarking on
the argument phrase:
 
(138) CP −→ XP C#
 
 (↑ dis = ↓  ↑=↓
(↑ ggf = ↓

C# −→ C S
↑=↓ ↑=↓

S −→ XP XP V XP∗
(↑ ggf = ↓ (↑ ggf = ↓ (↑ = ↓) (↑ ggf = ↓

In Ancient Greek, the specifier position of CP is associated with a phrase that fills
the dis function at f-structure. This is indicated by the annotation (↑ dis) = ↓;
unbounded dependencies are discussed further in Chapter 17. The dis function
is an overlay function and must be syntactically or semantically linked to a
primary non-overlay function: in this CP rule, the annotation (↑ ggf) = ↓ requires
this phrase to bear not only the dis role, but also a grammatical function ggf
in the sentence.17 The XP daughters of S are also annotated with the equation
(↑ ggf) = ↓, indicating that a phrase with any grammatical function can appear
there. Complementizers and some conjunctions can be analyzed as appearing in C.
In contrast to Balinese, the Ancient Greek verb specifies a great deal of informa-
tion about its arguments. The case of each argument is specified, and additionally
an optional pred value for each argument is provided. As described in §5.2.4, this
allows for the absence of overt phrasal arguments (pro-drop): an Ancient Greek
sentence may consist simply of a verb, with no overt subject or object phrases
present at c-structure. In such a case, the pred values of the arguments of the verb
come from the verb’s specifications:

16 The analysis of Ancient Greek presented here relies on the work of Dik (1995, 2007) on the relation
between word order and information structure in Ancient Greek, and on the work of Haug (2008a,b,
2011, 2012) on Ancient Greek syntax. The rules provided are, as with Balinese, necessarily simplified
for the purposes of exposition.
17 This is a simplification, since phrases from embedded clauses can appear in the specifier of CP. The
annotation (↑ ggf) = ↓ on the specifier of CP can be read as an abbreviated equivalent (in a monoclausal
sentence) for a more complex annotation involving specification of a path through the f-structure, as
we will see in Chapter 17.
 180 describing syntactic structures

(139) élaben V (↑ pred) = ‘catch%subj,obj&’

((↑ subj pred) = ‘pro’)
(↑ subj concord case) = nom
((↑ obj pred) = ‘pro’)
(↑ obj concord case) = acc

(140) strouthòn dè oudeìs é-laben

ostrich.acc but no_one.nom pst-catch.pfv
‘But no one has caught an ostrich.’
 
’
   
 
   
 
 
 
 
 
 
  
 
 
  
 
 
 
CP  

NP
C
(↑ ↓
↑=↓
(↑ ↓

N C S
↑=↓ ↑=↓ ↑=↓

strouthòn
dè
NP V
but
(↑ (↑ ↓ ↑=↓
(↑
(↑
é-laben

(↑ ’
N
((↑
↑=↓
(↑
((↑
oudeìs
(↑
(↑
(↑

The verb requires its subject to be in the nominative case; phrase structural annota-
tions allow the phrase oudeìs ‘no one’ to bear any governable grammatical function
ggf, but only the subj grammatical function is compatible with the nominative
casemarking requirements imposed by the verb and captured by the equation
(↑ concord case) = nom for oudeìs. Similarly, its accusative casemarking requires
the phrase strouthòn ‘ostrich’ to bear the obj function.
The verb also provides optional ‘pro’ values, enclosed in parentheses, for the
pred of its subject and object. These values do not appear in the final f-structure,
since the overt subject and object noun phrases oudeìs ‘no one’ and strouthòn
‘ostrich’ are present and contribute their pred values to the final f-structure. If
these phrases did not appear, the ‘pro’ value optionally provided by the verb would
appear as the value of the pred of these arguments.
 variation in grammatical function encoding 181

5.4.3 Chicheŵa
Chicheŵa is typologically different from both Balinese and Ancient Greek, and
illustrates an interesting combination of configurational and nonconfigurational
characteristics (Bresnan and Mchombo 1987; Bresnan et al. 2016). The relevant
phrase structure rules for Chicheŵa are:

(141) S −→ NP , NP , VP
(↑ subj = ↓ ↓ ∈ (↑ dis ↑=↓

VP −→ V#
↑=↓

V# −→ V NP
↑=↓ (↑ obj = ↓

These rules show that grammatical functions in Chicheŵa are specified configura-
tionally to some extent, though not in the same way as in Balinese. Like Balinese
(and Ancient Greek), Chicheŵa makes use of the exocentric category S; it differs
from Balinese, however, in that the daughters of S are the subject, a dis phrase
(which is the topic at the separate level of information structure; see Chapter 10),
and the VP, which may appear in any order (the first rule in (141) is an unordered
ID rule, as described in §5.1.5). There is a configurationally specified postverbal
object position: the complement of the lexical category V is obj, as indicated by
the equation (↑ obj) = ↓.
The lexical entry for the Chicheŵa transitive verb zi-ná-wá-lum-a ‘bit’ is given
in (142):

(142) zi-ná-wá-lum-a V (↑ pred) = ‘bite%subj,obj&’

((↑ subj pred) = ‘pro’)
(↑ subj nounclass) = 10
(↑ obj pred) = ‘pro’
(↑ obj nounclass) = 2

Unlike the Balinese verb, and like the Ancient Greek verb, this verb contains an
optional ‘pro’ value for the pred of its subject; this means that an overt subj phrase
may but need not appear. The verb also carries information about the noun class
of its arguments: Chicheŵa, like many Bantu languages, has a complex noun class
system, and the prefix zi- indicates that the subj belongs to noun class 10.
The obj is treated differently from the subj. As Bresnan and Mchombo (1987)
show, this verb contains an incorporated pronominal object wá. This means that
the equation specifying the obj pred is obligatory, not optional. The c-structure
and f-structure for the sentence njûchi zi-ná-wá-lum-a ‘The bees bit them’ are
displayed in (143):
182 describing syntactic structures

(143) njûchi zi-ná-wá-lum-a

bees 10subj-pst-2obj-bite-ind
‘The bees bit them.’
 
’
 
 
 
 
S  
 
 
NP VP
 
(↑ ↓ ↑ =↓

N
V
↑ =↓
↑ =↓
njûchi
bees
(↑ V
(↑ ↑ =↓

zi-ná-wá-lum-a

(↑ ’
((↑
(↑
(↑
(↑

When the incorporated object pronoun wá does not appear, the sentence is
incomplete unless an overt noun phrase is present. The lexical entry for the verb
zi-ná-lum-a ‘bit’, with no incorporated obj pronoun, is:

(144) zi-ná-lum-a V (↑ pred) = ‘bite%subj,obj&’

((↑ subj pred) = ‘pro’)
(↑ subj nounclass) = 10

In the following Chicheŵa sentence, there is no overt subj noun phrase. This
means that the subject noun phrase is an incorporated pronoun, as optionally
specified by the verb. An overt obj noun phrase, alenje ‘hunters’, also appears;
if there were no overt obj noun phrase, the sentence would be incomplete and
therefore ungrammatical. The grammatical function of alenje is determined by
the phrase structure configuration in which it appears:
 variation in grammatical function encoding 183

(145) zi-ná-lum-a alenje

10subj-pst-bite-ind hunters
‘They bit the hunters.’
 
’
 
 
 
 
S  
 
 
VP  
↑=↓

V
↑=↓

V NP
↑=↓ (↑ ↓

zi-ná-lum-a
N
(↑ ’
↑=↓
((↑
(↑
alenje
hunters
(↑
(↑

It is also possible for the incorporated object pronoun to be anaphorically linked

to an overt, displaced dis noun phrase that has the function of topic at the
separate level of information structure, a relationship that we do not indicate in
the functional structure. In this case, the incorporated obj pronoun wá appears,
as shown in (146) (Bresnan and Mchombo 1987: 745). Example (146) is different
from (145) in that the phrase alenje ‘hunters’ appears not in canonical obj position,
but in the c-structure position associated with a dis phrase (Bresnan and Mchombo
1987). As dis is an overlay function, this phrase must therefore also be linked to a
primary (nonoverlay) function. In this case, the dis phrase is anaphorically linked
to the incorporated obj pronoun, as indicated by the subscript i indexes in the
translation:
 184 describing syntactic structures

(146) njûchi alenje zi-ná-wá-lum-a

bees hunters 10subj-pst-2obj-bite-ind
‘The huntersi , the bees bit themi .’
 
’
 
 
 
 
 
 
 
 
 
 
 
 
 

NP NP VP
(↑ ↓ ↓ ∈ (↑ ↑=↓

N N V
↑=↓ ↑=↓ ↑=↓

zi-ná-wá-lum-a
njûchi alenje
bees hunters (↑ ’
(↑ (↑ ((↑
(↑ (↑ (↑
(↑
(↑

As these examples illustrate, the pronominal typology predicted by LFG is richer

than the one proposed by Jelinek (1984), who hypothesizes that all noncon-
figurational languages should be analyzed as pronominal-incorporating, as we
have analyzed the Chicheŵa incorporated object pronoun. Dahlstrom (1986a)
shows that this simple proposal does not provide an adequate account of the
facts in Meskwaki; Bresnan and Mchombo (1987), Austin and Bresnan (1996),
Nordlinger (1998), and Toivonen (2000) also show that an adequate analysis of the
phrasal and functional structure of many languages must allow the possibility for
a pred value to be optionally contributed. In other words, incorporated pronouns
contributing a pred (like the Chicheŵa incorporated object pronoun wá) must be
distinguished from agreement markers which do not contribute a pred as well as
from forms which, like the Chicheŵa subject marker, are ambiguous between the
two, optionally contributing a pred.

5.4.4 Bulgarian
Bulgarian is unusual in combining relatively free word order with a lack of nominal
inflection (Rudin 1985): only pronominal forms show casemarking. In some cases,
the subject can be identified as the argument that agrees with the verb; additionally,
Bulgarian allows clitic doubling, so that the case, gender, and number of clitic-
 variation in grammatical function encoding 185

doubled arguments are specified. In other cases, however, these clues do not
serve to disambiguate the grammatical functions of the argument phrases, and a
phrase may be associated with any of the grammatical functions selected by the
predicate. In such cases, only contextual information and world knowledge help in
determining the intended structure.
The relevant phrase structure rules for Bulgarian are:
 
(147) IP −→ NP I#
 
 ↓ ∈ (↑ dis)  ↑=↓
(↑ ggf) = ↓

I# −→ I S
↑=↓ ↑=↓

I −→ N I
(↑ obj) = ↓ ↑=↓
∗
S −→ { NP | V }
(↑ ggf) = ↓ ↑=↓

At this rough level of detail, these rules are very similar to the rules for languages
such as Ancient Greek, which also allow relatively free word order. An important
difference is that the I rule allows adjunction of a nonprojecting DN object clitic to I.
The third person singular past tense verb vidja ‘saw’ has this lexical entry:

(148) vidja V (↑ pred) = ‘see%subj,obj&’

((↑ subj pred) = ‘pro’)
(↑ subj index pers) = 3
(↑ subj index num) = sg
(↑ subj concord case) = nom
(↑ obj concord case) = acc

As Rudin (1985) shows, a Bulgarian clause can appear without an overt subject, as
in Ancient Greek and Chicheŵa: the verb contains an optional equation specifying
a pronominal value for the pred of its subject. If the Bulgarian verb is transitive,
either an overt object phrase or the object clitic pronoun must appear, since the
verb does not specify a pred value for its object.
Consider (149). The noun knigata ‘the book’ and the proper noun Georgi (a male
name) are unmarked for case; each of them is compatible with either nominative
or accusative case. The symbol ggf, which represents any governable grammatical
function, is arbitrarily instantiated to subj for Georgi and obj for knigata ‘the book’.
As Rudin (1985) notes, it is only world knowledge that enforces the interpretation
of knigata ‘the book’ as the object of the verb vidja ‘saw’, and Georgi as its subject.
Neither phrase structure position, casemarking, nor agreement requirements serve
to disambiguate the syntactic role of these arguments.18 However, in (149), the

18 But see King (1995) and Mahowald (2011) for discussion of ways in which word order constraints
may contribute to disambiguation in similar constructions.
 186 describing syntactic structures

subject and object are distinguished in terms of information structure role: as

discussed in §4.2.3.1, the specifier of IP is a dis position in Bulgarian, and the
phrase appearing in the specifier of IP is interpreted as bearing a prominent
information structure role. Here, the object knigata ‘the book’ is likely to be
interpreted as in focus:

(149) Knigata vidja Georgi. ’

book see.pst.3sg Georgi
‘It is the book that Georgi saw.’

IP

NP
I
↓ ∈ (↑
↑ =↓
(↑ ↓

N I S
↑ =↓ ↑ =↓ ↑ =↓

vidja
Knigata
book (↑ ’
(↑ ((↑ NP
(↑ (↑ (↑ ↓
(↑ (↑
(↑ (↑ N
(↑ ↑ =↓

Georgi
Georgi
(↑
(↑
(↑
(↑

Example (150) differs from (149) in that the dis noun phrase decata ‘the children’
is plural. Thus, since the verb shows third person singular agreement with its
 variation in grammatical function encoding 187

subj, the only available syntactic analysis is the one in which the third person
singular phrase Georgi is the subject. In this example, decata ‘the children’ appears
in the specifier of IP and can be interpreted as bearing the prominent information
structure role of focus:

(150) Decata vidja Georgi. ’

children see.pst.3sg Georgi
‘It is the children that Georgi saw.’

IP

NP
↓ ∈ (↑ I
(↑ ↓

N I S
↑=↓ ↑=↓ ↑=↓

vidja

Decata
(↑ ’
children
((↑ NP
(↑
(↑ (↑ ↓
(↑
(↑
(↑ N
(↑
(↑ ↑=↓

Georgi
Georgi
(↑
(↑
(↑
(↑

The presence of a doubled clitic object pronoun can also help to disambiguate a
potentially ambiguous sentence; example (151) is unambiguous. Here, the subject
Georgi appears in the specifier of IP, and is likely to be interpreted as a topic:
 188 describing syntactic structures

(151) Georgi ja gleda Marija.

Georgi 3sg.f.acc.clitic watch.prs.3sg Marija
‘Georgi is watching Marija.’
’

IP

NP
I
↓ ∈ (↑
↑=↓
(↑ ↓

N I S
↑=↓ ↑=↓ ↑=↓

Georgi
Georgi
(↑ N I NP
(↑ (↑ ↓ ↑=↓ (↑ ↓
(↑
(↑ gleda
ja

(↑ ’
((↑
((↑ N
(↑
(↑ ↑=↓
(↑
(↑
(↑ Marija
(↑
(↑ Marija
(↑
(↑
(↑
(↑
(↑

The lexical entry for the feminine singular accusative clitic pronoun ja is:
 variation in grammatical function encoding 189

(152) ja ((↑ pred) = ‘pro’)

(↑ index pers) = 3
(↑ index num) = sg
(↑ index gend) = f
(↑ concord case) = acc

If there is no full object noun phrase, the pred value for the obj function is given
by the object clitic pronoun phrase. Since the pred of the clitic pronoun is optional
(as in Spanish; see §5.2.4) the presence of ja is also compatible with the obj being
filled by the feminine phrase Marija (but not the masculine phrase Georgi, since
that would produce a clash in gend values).

The Balinese, Ancient Greek, Chicheŵa, and Bulgarian examples presented in

this section attest both to the diversity of expression found crosslinguistically and
to the basic underlying unity of structure at a more abstract syntactic level.

5.4.5 Copular constructions

Copular constructions represent an interesting case study with respect to many
of the issues concerning syntactic correspondences that have been discussed in
this and the previous chapter. Crosslinguistically, such constructions may vary in
significant ways, despite having very similar meanings. In a language like English,
for example, a copular verb must be used:

(153) a. Fiona is a nurse.

b. The book is red.
c. The vase is on the table.

Contrast this with comparable examples from Maori (154a), Japanese (154b), and
Russian (154c), in which no verb appears in the sentence. The following examples
are from Rosén (1996) citing Biggs (1969: 24), Dalrymple et al. (2004a: 190), and
Attia (2008) citing Avgustinova and Uszkoreit (2003), respectively:

(154) a. He taariana, te hoiho.

indf stallion def.sg horse
‘The horse is a stallion.’
b. hon wa akai
book topic red
‘The book is red.’
c. Boris na sobranii.
Boris at meeting.loc
‘Boris is at a meeting.’

In some languages, both types of constructions—with and without a verb—exist,

as these Hebrew examples from Falk (2004) show:
190 describing syntactic structures

(155) a. Pnina tinoket / b-a-bayit / nora xamuda.

Pnina baby.f in-the-house awfully cute.f
‘Pnina is a baby/in the house/awfully cute.’
b. Pnina hayta tinoket / b-a-bayit / nora xamuda.
Pnina be.pst.3f.sg baby.f in-the-house awfully cute.f
‘Pnina was a baby/in the house/awfully cute.’

Which of these two types of copular construction is used to encode the relevant
relationship may vary according to a number of different factors including tense
(as in the Hebrew examples), aspect, polarity, syntactic category, main versus
subordinate clause status, and the precise nature of the semantic relation between
the elements in the sentence. Crosslinguistically, whether or not a verbless con-
struction is available does not invariably correlate with any particular one of these
factors. In other languages, the copula may be expressed by use of a particle which
does not exhibit all of the usual properties of a verb, or by inflectional marking on
a non-verbal element in the sentence similar to that more usually associated with
a verb: for details, see Curnow (1999) and Pustet (2003). Nordlinger and Sadler
(2003, 2007) propose that the clausal pred information which in other cases would
be contributed by a verb is lexically associated with these other expressions.
In terms of c-structure (and, by extension, the correspondence between c-
structure and f-structure), there is a fundamental difference between copular
constructions with a copular particle or verb and those without. Compare the c-
structures for the copular constructions in English (156) and in Russian (157). In
English, the copular verb appears in I: it precedes negation (He is not a student),
and it inverts in question formation (Is he a student?).19 There is no copular verb
in (157); IP is headless:

19 An alternative to the English c-structure in (156) is the c-structure in (a), which includes a VP
node. The choice between the trees in (156) and (a) depends on whether a single-tier analysis is assumed
(as shown in 161), for which the tree in (156) may be preferred in order to maintain the generalization
that the complement of I is always an f-structure co-head, or a double-tier analysis (as shown in 162),
for which the tree in (a) is appropriate.
(a)
IP

NP I#
I VP
He
is V#
NP
a student
 variation in grammatical function encoding 191

(156) He is a student.
IP

NP I#

I NP
He
is
a student

(157) On student.
he student
‘He is a student.’
IP

NP I#

NP
On
he student
student

While we have seen that it is possible to have f-structures that are not related
to any c-structure node, as in cases of pro-drop (example (3), §4.1, page 118),
the examples that have been presented in this chapter and earlier chapters have
all included a verb which contributes the main pred of the clause. In a copular
construction including a verb such as (156), the verb is often assumed to introduce
the main clausal pred. In the case of verbless constructions such as (157), though,
it is not immediately clear what contributes the main pred that is required for the
f-structure to be coherent. In turn, this raises questions relating to the analysis of
copular constructions that do include a verb, and the broader issue of whether
a unified analysis of copular constructions is possible and desirable. We will see
that different analyses appear in the literature: for instance, it has been proposed
that the copular verb in a sentence such as (156) contributes only tense and
aspect information, with the main pred contributed by the copular complement.
Similarly, in the case of verbless constructions such as (157), different analyses are
possible as to what contributes the main pred at f-structure. A related issue, which
has also been the subject of debate in the literature, concerns which grammatical
functions the main pred subcategorizes for. Before considering the grammatical
functions that may be involved, we review two influential analyses of verbless
copular constructions. The first proposes that the main clausal pred is contributed
by the phrase structure configuration (§5.4.5.1); the second proposes that the main
clausal pred is contributed by the complement (§5.4.5.2).
192 describing syntactic structures

5.4.5.1 Main pred contributed by phrase structure Rosén (1996) proposes that
in the case of a verbless copular construction like the Maori example in (154a), the
phrase structure configuration rather than the lexical specifications of the words
in the construction must introduce the copular relation. Under a constructional
analysis such as this one, the pred value that licenses the grammatical functions
of non-verbal elements is introduced by the phrase structure of the copular
construction:20

(158) C-structure and f-structure for He taariana te hoiho according to Rosén

(1996):
He taariana, te hoiho.
indf stallion def.sg horse
‘The horse is a stallion.’
 
’
  
 
 
  + 
  
 
 
 
 
 
 
S
−
NP
NP
(↑ ’
(↑ ↓
(↑ ↓

D N D N

he te
N N

taariana hoiho
stallion horse

As shown in (158), Rosén (1996) associates the clausal pred with the first NP node
in the c-structure of Maori, which is interpreted as the predicative complement.21
This is one possible analysis of this construction.

20 Other analyses of verbless copular constructions have been formulated in terms of an empty node
rule (§6.10.4). On Russian, see Dalrymple et al. (2004a); on Arabic, see Attia (2008).
21 Rosén’s analysis makes use of the grammatical role ncomp, or nominal complement. This role
is no longer in general use in LFG analyses, since it violates modularity by encoding c-structure
distinctions (the phrase structure category of the complement) in the statement of f-structural subcat-
egorization requirements. The current equivalent would be predlink, discussed in §2.1.8 and below,
though the predlink role is not associated with any particular phrase structure category.
 variation in grammatical function encoding 193

5.4.5.2 Main pred contributed by the complement When a verb is not present,
it can also be argued that it is the predicative complement that contributes the
main clausal pred, selecting for arguments just as a verb would and resulting
in a fundamentally different f-structure from that shown in (158). For example,
according to Dalrymple et al.’s (2004a) analysis of Japanese sentences such as
(154b), the lexical entry for akai ‘red’ includes (↑ pred) = ‘red%subj&’:

(159) C-structure and f-structure for hon wa akai according to Dalrymple et al.
(2004a):
hon wa akai
book topic red
‘The book is red.’
’
S

NP AP
(↑ ↓ ↑ =↓

A
hon wa
book akai
red
(↑ ’

The two different approaches to the relationship between c-structure and

f-structure shown in (158) and (159) are not restricted to the analysis of verbless
copular constructions, but represent a major division in the possible analyses of
copular constructions more generally. The two approaches are referred to as the
“single-tier” analysis and the “double-tier” analysis. To illustrate the difference
between the two, Nordlinger and Sadler (2007: 141–2) present two possible
f-structure analyses of this verbless Russian copular construction:

(160) Ona vrač.

she doctor
‘She is a doctor.’

Under a single-tier analysis, a non-verbal element in the copular construction

is the syntactic head of the clause, as in the Japanese example in (159). The f-
structure of a non-verbal element is therefore identified with the f-structure of the
clause and contributes the necessary clausal pred value. For (160), there must be
a lexical entry for the predicative complement vrač ‘doctor’ which contributes the
main clausal predicate and selects a subject, as shown in (161). This lexical entry
may exist alongside a lexical entry for the same form which does not require a
194 describing syntactic structures

subject, as in a sentence like A doctor came into the room, where a doctor is not
used predicatively:
(161) Single-tier analysis of Ona vrač ‘She is a doctor’:
 
’
  
 
   
  
  
    
 

Under a double-tier approach, the relevant non-verbal elements are arguments,

one of which is the subject. The main pred that selects for these arguments can be
supplied by a copular verb (or by a non-verbal particle or inflection which serves
to establish the same kind of relationship; see, for example, Nordlinger and Sadler
2003 on Tariana), or it can be introduced by the phrase structure, as proposed by
Rosén (1996) and illustrated in (158).
Two alternative double-tier analyses of (160) are shown in (162).22 These differ
with respect to the grammatical function that the non-verbal complement in a
copular construction bears: that is, whether it is an open grammatical function
whose subj must be specified externally (xcomp, as in example 162a), or a closed
grammatical function (predlink, as in example 162b):

(162) Alternative double-tier analyses of Ona vrač ‘She is a doctor’:

a.  ’

   
 
   
  
   
     
   
 
 
 
 
 ’ 
 
 

22 A third kind of double-tier analysis with an oblique copular complement is proposed by Falk
(2004) for locative complement constructions in English and Hebrew, and by Laczkó (2012) for locative
and existential complement constructions in Hungarian (see also Bresnan et al. 2016: Chapter 12):
 
(a) ’
   
 
   
  
  
   
 
 
 
 

For other copular constructions in Hebrew, Falk argues for a predlink analysis; Laczkó (2012) argues
that a single-tier analysis is appropriate for Hungarian copular constructions involving attribution or
classification, while for constructions involving identity or possession, a double-tier predlink analysis
is motivated.
 variation in grammatical function encoding 195

b.  ’

  
 
   
  
  
   

 
 
 
 

Nordlinger and Sadler (2007: 142) point out that although most treatments of
copular constructions that include a verb assume a double-tier analysis (with
the copular verb contributing the main clausal pred), a single-tier analysis is
also possible. Under such an approach, the copular verb contributes information
about tense, aspect, mood, and other verbal features, while the non-verbal element
contributes the main clausal pred, meaning that the copular verb and the non-
verbal predicate are functional co-heads of the clause.
In summary, two main approaches to the analysis of copular constructions have
been put forward: the single-tier analysis, in which a non-verbal element supplies
the main clausal pred, and the double-tier analysis, in which the main clausal pred
is contributed by some other means (for example, by a copular verb or by the
phrase structure). Double-tier analyses can be subdivided according to whether
the complement’s grammatical function is taken to be open or closed.

5.4.5.3 A unified analysis? Determining which analysis is appropriate for cop-

ular constructions is a key issue. Some researchers adopt a unified approach
to the f-structure of copular constructions: Butt et al. (1999) reject an xcomp
analysis on the basis that the types of complement phrases that appear in copular
constructions (NPs, APs, and PPs) do not usually select for a subject, and propose
a unified double-tier analysis of copular constructions in which the complement
phrase bears the closed function predlink. Attia (2008) also argues for a unified
f-structure analysis of copular constructions, stating that the apparent variation
in copular constructions must be regarded as a matter of encoding only, and
therefore should not be taken as evidence of functional variation. For example,
Attia (2008) states that the presence or absence of a copular verb is simply a
matter of c-structure variation, the verb itself being semantically redundant, and
as such should not correspond to a difference at f-structure. Rather, the functional
equivalency of copular constructions should be captured at the level of f-structure
through a unified analysis that also reflects the fundamental ways in which copular
constructions differ from subject-verb constructions.
In those cases where a unified analysis is advanced, we must decide which of the
possible approaches to f-structure represents the default for a copular construction.
The double-tier closed complement predlink analysis, exemplified in (163), is the
most widely adopted unified analysis of copular constructions in the literature;
see Butt et al. (1999), Attia (2008), Sulger (2009, 2011), and Dione (2012). Attia
claims that a single-tier analysis may be motivated in exceptional cases, citing
those discussed in Nordlinger and Sadler (2007). Nordlinger and Sadler (2007)
themselves conclude on grounds of economy that a single-tier analysis should be
196 describing syntactic structures

regarded as the default for languages with only verbless copular constructions,
though they do not argue for a unified analysis of all copular constructions.
Other researchers have advocated an approach that allows more variation. Dal-
rymple et al. (2004a) observe that when the complement in a copular construction
has a subject that is not the same as the one in the matrix clause, a single-
tier analysis is not possible. This is because the subj of the copular complement
(David in 163) would clash with the subj of the copula (the problem in 163),
resulting in a violation of Consistency. If the main clausal pred selects for a
closed complement, however, the f-structure includes two distinct subjects and
Consistency is respected, as shown in (163):

(163) IP
NP I
I VP  
The problem ’
is V  
 
 
CP  + 
 
  
C  ’ 
 
C IP
  
  
 
that NP I
VP
David
V
yawned

Compare (163) with (164), in which the (expletive) subject of the copula and its
complement clause are the same. Dalrymple et al. (2004a) argue that in such cases
an open complement xcomp analysis is required to account for the fact that a
standard control relation appears to exist between the two subjects (see Chapter 15
for a full discussion of open complements and control):

(164)  
 
 
 
  
 
 
  
IP   
  
   
NP I   
  
   
I VP  
It
is V
AP
A
A VP
likely
to rain
 further reading and related issues 197

On the basis of examples such as (163) and (164), Dalrymple et al. (2004a) argue
that a single analysis cannot apply to all copular constructions: whether a closed
or open complement is involved must be determined on a case-by-case basis—
both crosslinguistically and within a single language—taking into account relevant
syntactic properties, including the presence or absence of a copular verb and
the presence or absence of agreement marking. On this view, the suitability of
an analysis is determined independently for each copular construction, and the
f-structures of all copular constructions need not be fundamentally the same, even
within a single language. This approach to the analysis of copular constructions
in LFG is also advocated by Falk (2004), Nordlinger and Sadler (2007), Laczkó
(2012), and Lowe (2013).
Precisely how c-structure and f-structure are related in the case of copular
constructions is the subject of ongoing debate in the LFG literature. Further
research which considers data from a range of languages is required to determine
if a unified analysis of some or all copular constructions is justified, either within
a particular language or crosslinguistically.

5.5 Further reading and related issues

In §5.2.1, we treated agreement as specification of the features of the agreement
target by the agreement controller. Haug and Nikitina (2016) present a typology
of agreement analyses that encompasses other analytic possibilities, distinguishing
asymmetric from symmetric analyses of agreement. Asymmetric analyses require
the agreement features of the target to satisfy the agreement requirements of
the controller, with the result that the controller must be fully specified for the
agreement features that are present on the target; in contrast, according to the
symmetric view, the controller and target cospecify the features involved in agree-
ment, and underspecification is allowed in both the target and controller. Haug
and Nikitina also distinguish feature sharing analyses from analyses not involving
feature sharing: on a syntactic feature sharing analysis, the controller and target
each bear a separate set of agreement features which must be compatible, while
on a non-feature-sharing analysis, the target constrains the agreement features of
the controller. The analysis we assume in this book is a symmetric analysis with
no feature sharing, as is the analysis of Bresnan et al. (2016). Haug and Nikitina
(2016) present arguments for a symmetric analysis with feature sharing, where
the controller and the target each bear agreement features which are required to
match, and Alsina and Vigo (2014) present arguments for a similar approach to
the analysis of copular agreement in Catalan. Wechsler (2015) provides a useful
overview of agreement patterns and the features that are relevant for agreement.
In the agreement patterns we examine in this book, the target of agreement
is a predicate, and the controller is an argument of the predicate: for example,
the target of agreement is often a verb, and the controller is its subject and/or
object. More complex patterns are also found: see in particular Sadler (2016) for
discussion and an LFG-based analysis of the very complex agreement patterns in
Archi. Belyaev (2013) presents an Optimality Theory-based analysis of complex
198 describing syntactic structures

cross-dialectal verb agreement patterns in Dargwa involving different controllers

of person agreement on the verb.
The formal tools and notational conventions of LFG presented in this and the
next chapter are discussed in detail by Kaplan and Maxwell (1996) and Crouch
et al. (2008).
 6
Syntactic relations and
syntactic constraints

Chapter 5 discussed ways of describing and constraining constituent structures

and functional structures. This chapter continues that thread, introducing addi-
tional relations and constraints on structures. For most readers, this chapter best
serves as a reference to be consulted for definition and discussion of concepts and
relations that are used in the analyses presented in the remainder of the book.

6.1 Regular expressions and regular languages

In formal language theory, a “language” is defined as a set of strings of symbols,
sounds, letters, words, or whatever objects are of interest, and the “alphabet” is the
set of minimal units that make up the strings. For instance, (1) shows a simple
language over the alphabet {a, b}, containing seven strings:

(1) {a, b, ab, aaab, bbbbb, ababab, bbbaaa}

A class of languages with particularly simple formal properties is the class of regular
languages, which are languages that can be characterized by regular expressions. In
an LFG setting, regular expressions play an important formal role in the expression
of constraints on both the c-structure and the f-structure. In §5.1.1, we saw that
the right-hand side of an LFG phrase structure rule is a regular expression over
c-structure categories, allowing disjunction, optionality, and repetition:

(2) V! −→ V (NP) PP⃰

As we will see in §6.1.2, functional uncertainty involves characterization of paths

through the f-structure by means of regular expressions over grammatical func-
tions. Here, we provide a brief synopsis of regular languages and regular expres-
sions and their use in formulating c-structural and f-structural constraints. For a
detailed explication, see Partee et al. (1993, Section 17.2).
Formally, a regular expression is formed in the following way:

(3) a. The symbol 𝜖 or a single symbol is a regular expression. The symbol 𝜖

denotes the empty string.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
200 syntactic relations and syntactic constraints

b. The disjunction of two regular expressions R1 and R2, written {R1 | R2},
is a regular expression. The regular language denoted by the disjunction
{R1 | R2} is the union of the language R1 and the language R2: that is, all
of the strings in R1 as well as all of the strings in R2.
c. Parentheses indicate optionality; if R is a regular expression, (R) denotes
the union of the strings in R with the empty string, which could also be
written as {R | 𝜖}.
d. The concatenation of two regular expressions R1 and R2, written R1 R2,
is a regular expression. The regular language denoted by R1 R2 is the con-
catenation of the strings in R1 with the strings in R2. (The concatenation
of a with ab is aab, and the concatenation of aa with bb is aabb.)
e. A regular expression annotated with the Kleene star operator ⃰ is a regular
expression. The regular expression R⃰ denotes the set of strings with 0 or
more repetitions of the strings in R.
f. A regular expression annotated with the Kleene plus operator + is a regular
expression. The regular expression R+ denotes the set of strings with 1 or
more repetitions of the strings in R.

Groupings of symbols can be represented by enclosing the symbols in square

brackets, which otherwise have no meaning.
Given these rules, the regular expression in (4a) is wellformed, and characterizes
the regular language in (4b) (an a, followed by an optional a, followed by either
a or b):

(4) a. Regular expression: a(a){a|b}

b. Regular language: {aa, ab, aaa, aab}

The use of the Kleene star operator allows the description of an infinite language,
since an expression such as R⃰ denotes a language in which the strings in R are
repeated any number of times (including zero). The language characterized by the
regular expression in (5a) consists of strings which begin with a, then contain any
number of sequences of bc (including none), and end with d:

(5) a. Regular expression: a[bc]⃰d

b. Regular language: {ad, abcd, abcbcd, abcbcbcd, abcbcbcbcd, . . .}

Additionally, regular languages are closed under the set-theoretic operations of

union and intersection, and regular languages over simple symbols such as phrase
structure category labels and names of grammatical functions are closed under
complementation.1 In other words, the union or intersection of two regular lan-
guages is also a regular language, and the complement of a regular language whose

1 For definitions of union, intersection, and complementation, see (46) in this chapter (page 214). It
is not currently known whether arbitrary regular languages over annotated phrase structure categories,
 regular expressions and regular languages 201

alphabet consists of constituent structure categories or grammatical functions is

also a regular language. Since this is true, regular languages can also be defined in
terms of these operators. We can also define the term complement of a collection of
symbols s as any of the symbols in the alphabet other than those in s.2 For example,
given the category labels discussed in Chapter 3, the term complement of NP is
any category label other than NP, such as P, V! , or AdvP. The relative difference
operator − is similar: A − B is the set of strings that are in A and are not in
B. Thus, we can define the term complement of a single symbol or disjunction
of single symbols s as Σ − s (all symbols in the alphabet except for those in s),
using the relative difference operator. It is common to use a special notation for
union, intersection, and complementation of regular languages, different from
the notation usually used for sets. We use the following notation for operations
involving regular languages:

(6) Union of two regular languages A and B: {A | B}

Intersection of two regular languages A and B: A&B
Relative difference between two regular languages A and B: A−B
Complement of the regular language A: ¬A
Term complement of a collection of symbols s: \s

Care must be taken when defining the term complement of an annotated phrase
structure category: the term complement of an annotated category C with anno-
tation A denotes the set of all category/f-structure pairs that do not match the
annotated category. This is the disjunction of \C (that is, any category other than
C, with any annotation) and the category C with the annotation ¬A. For example,
the term complement of the category NP with annotation (↑ subj) = ↓ is written
as in (7a), and is the same as the disjunction in (7b):
! "
NP
(7) a. \
(↑ subj) = ↓
b. {\NP | NP }
(↑ subj) &= ↓

6.1.1 Regular languages and rule descriptions

The ID/LP rule format discussed in §5.1.5 allows the decomposition of a standard
phrase structure rule into two aspects so that dominance constraints can be
specified separately from precedence constraints. For example, the ID (immediate
dominance) rule in (8a) licenses a c-structure tree in which a V! node dominates

associated with f-structure constraints such as ↑ = ↓ or (↑ subj) = ↓, are closed under complementation
(Ron Kaplan, p.c.).
2 Given an alphabet Σ, the universe (the set of all possible strings in the language) is Σ ⃰, and the
complement of a set of strings S is Σ ⃰ − S (all strings in the universe except for those in S).
202 syntactic relations and syntactic constraints

two nodes, one labeled V and one labeled NP, but without specifying an order
between V and NP. The LP (linear precedence) constraint in (8b) specifies that
the V node precedes the NP node. This is a particular instance of the general rule
in (8c), stating that a lexical (X) node must precede a phrasal (YP) node:

(8) a. V! −→ V, NP
b. V < NP
c. X < YP

This is an example of the use of a formal device, the ID/LP rule format, to express
a generalization about word order across classes of phrase structure rules: for
example, whether the phrases in a language are head-initial or head-final, or
whether the specifier of a phrase precedes or follows the head.
More generally, it is possible to write rules that combine different constraints on
phrasal structure, allowing for the succinct expression of linguistic generalizations
about the structure of phrases. Such generalizations can be expressed in terms
of separate constraints that must be simultaneously satisfied. Formally, this is
possible because we are allowed to specify any regular language as the right-hand
side of a phrase structure rule; this means that concatenation, disjunction, Kleene
star, Kleene plus, intersection, and term complement can be used in specifying
sequences of categories.
We have already seen that disjunctions over various possibilities for phrase
structure expansion can be specified:

(9) X −→ { Y1 Y2 Y3 | Z1 Z2 Z3 }

This schematic rule indicates that a phrase with category X dominates either the
series of daughters Y1 Y2 Y3 or the series Z1 Z2 Z3. As we have seen, disjunction in
a phrase structure rule corresponds to the union of two regular languages, since
the union of two regular languages encompasses all of the alternatives in each
language.
Intersection of two regular languages corresponds to the combination of two
descriptions; each description must hold of the result:

(10) X −→ X1 & X2

This schematic rule indicates that X dominates a sequence of categories that must
satisfy the description represented by X1 as well as the description represented by
X2. Formally, this corresponds to characterizing a regular language by intersecting
two regular languages, represented by X1 and X2.
To take a concrete example, we can restate ID/LP rules in this way. Consider the
following simple rule:

(11) C! −→ C, IP C < IP
 regular expressions and regular languages 203

We can think of the ID part of the rule, indicating dominance relations, as

representing one aspect of the constraints on phrasal configurations. The ID rule
in (11) can be expressed in an equivalent but much less revealing way as:

(12) C! −→ {C IP | IP C}

That is, C! dominates either C IP or IP C; the order is not determined.

The LP constraint C < IP is a requirement for C to precede IP, and this constraint
can be imposed on any ID rule involving any number of categories. In its most
general formulation, it can be written as a regular expression as follows (Ron
Kaplan, p.c.):

(13) ([\IP]⃰ C) \[C | IP]⃰ (IP [\C]⃰)

This expression allows any sequence of categories, including C, IP, and other
categories, but with constraints on the relative order of C and IP. The expression
takes advantage of the term complement operator: for example, \IP is any category
except IP. Each of the members of the set of strings in the language of this
expression consists of three parts. First, there is an optional sequence beginning
with any number of categories not including IP and ending in C; next, there is
a sequence of categories that does not contain either IP or C; finally, there is an
optional sequence of categories that begins with IP and continues with a sequence
not including C. In other words, this expression disallows IPs before the initial C,
and also disallows Cs after the final IP.
Now that we have reformulated the right-hand side of the ID rule and the LP
constraint as regular expressions, we can state the combination of the two by
intersection. The ID/LP rule in (14a) is equivalent to the expression in (14b), where
the set of strings characterized by the regular expression on the right-hand side
of the rule in (11) is intersected with the set of strings characterized by the LP
constraint in (13):

(14) a. C! −→ C, IP C < IP
b. C! −→ {C IP | IP C} & ([\IP]⃰ C) \[C | IP]⃰ (IP [\C]⃰)

This example illustrates two points. First, we can state constraints on constituent
structure configurations by defining a sequence of daughter nodes in a phrase
structure rule in terms of a regular expression. This is a powerful and general tool
for the linguist to use in the description of phrase structure.
Second, this way of stating generalizations may not be the clearest or most
revealing. Although the encoding of linguistic facts in terms of operators such as
Kleene star, union, intersection, and term complement leads to a solid understand-
ing of their underlying formal and computational properties, it may be preferable
to devise a special notation for some common operations, since (as can be seen
above, from the restatement of ID/LP rules in terms of the intersection of two
regular languages) it may not be perspicuous or revealing to state these in terms of
the standard notation of regular expressions: the ID/LP notation in (14a) is simpler
204 syntactic relations and syntactic constraints

and easier to understand than the expression in (14b), where a complex regular
expression is defined.
Kaplan and Maxwell (1996) define and discuss some operators that allow for the
compact expression of particular kinds of regular languages:

The “Ignore” operator: Using the Ignore operator, written as a forward slash
/ (Kaplan and Kay 1994), we can write a rule that allows for a category or a sequence
of categories to be interspersed with the other categories:

(15) XP −→ [X1 X2 X3] / Cat

This rule can be read as: “XP dominates X1 X2 X3, ignoring occurrences of Cat.”
In other words, XP must dominate X1 and X2 and X3 in that order, and may also
dominate any number of occurrences of Cat at any position. This rule is equivalent
to the following rule, containing a more complicated regular expression:

(16) XP −→ Cat⃰ X1 Cat⃰ X2 Cat⃰ X3 Cat⃰

The Ignore operator rule can be used to describe the appearance of parenthetical
elements, elements that can be inserted before or after any phrasal constituent
(see McCawley 1982).

The “Shuffle” operator: Using the Shuffle operator, represented as a comma,

we can specify two different sequences of nodes, each of which appears in a particu-
lar order but which may be interspersed or “shuffled” with each other. For instance:

(17) XP −→ [X1 X2 X3], [Y1 Y2 Y3]

According to this rule, XP must dominate nodes labeled X1 X2 X3 and Y1 Y2 Y3.

The relative order of the Xs and the Ys must be preserved, but no order is specified
across these sequences. The effect of this rule is similar to an ID/LP rule (§5.1.5)
where an ordering is specified among the X daughters and among the Y daughters,
but not between the Xs and the Ys.
The rule in (17) allows sequences like the following. In each case, X1 precedes
X2 and X3, and X2 precedes X3. The same holds for the Ys.

(18) a. X1 X2 X3 Y1 Y2 Y3
b. X1 X2 Y1 X3 Y2 Y3
c. X1 Y1 X2 Y2 X3 Y3
d. Y1 X1 X2 Y2 X3 Y3

Any ordering between the Xs and the Ys is allowed, as long as the order X1 X2 X3
and the order Y1 Y2 Y3 are preserved. The regular expression corresponding to this
rule is quite complex and will not be displayed.
The Shuffle operator is used to characterize constraints involving partial orders
holding among constituents in languages with otherwise fairly free word order. It
has been proposed within the Head-Driven Phrase Structure Grammar framework
 regular expressions and regular languages 205

by Reape (1994) in his analysis of word order in German; however, the phe-
nomena Reape analyzes using Shuffle are best analyzed within LFG in terms of
functional syntactic relations, as shown by Kaplan and Zaenen (1989b).

Other operators based on regular predicates can be proposed if their use simpli-
fies linguistic description. The use of these operators assists the linguist in develop-
ing firm intuitions about linguistic phenomena; as long as the operators are defin-
able in terms of regular predicates, no new formal power is added to the theory.

6.1.2 Regular languages and functional uncertainty

Regular languages are also used in stating f-structural constraints. As discussed in
§4.2.3.2, an information-structurally prominent phrase bearing some grammatical
function within the clause can appear in the specifier position of IP in Russian
(King 1995: 206):

(19) ‘Evgenija Onegina’ napisal Puškin.

Eugene Onegin wrote Pushkin
‘Pushkin wrote ‘Eugene Onegin’.’
’

IP

NP I

I S
Evgenija Onegina
Eugene Onegin napisal NP
wrote
N

Puškin
Pushkin

The specifier of IP in Russian is not associated with any particular grammatical

function. Here the phrase in the specifier of IP, Evgenija Onegina, bears the obj
function; in other examples, a phrase in the same position might be the subj or an
oblique. This functional uncertainty about the grammatical function of a phrase
in this position can be represented by defining a special abbreviatory symbol gf
representing a disjunction of all grammatical functions, similar to the definition of
ggf for governable grammatical functions:3

3 Recall the definition of ggf given in §5.4, example (134):

(a) ggf ≡ {subj | obj | obj𝜃 | comp | xcomp | obl𝜃 }
Relying on this definition of ggf, we can provide an alternative, more succinct definition of gf as a
disjunction over members of the adj and xadj set and the governable grammatical functions:
(b) gf ≡ {ggf | adj ∈ | xadj ∈}
The definitions in (b) and (20) are exactly equivalent.
206 syntactic relations and syntactic constraints

(20) gf ≡ {subj | obj | obj𝜃 | comp | xcomp | obl𝜃 | adj ∈ | xadj ∈}

This definition uses the set membership symbol ∈ as an attribute to allow refer-
ence to some member of a set, as discussed in §6.3.1.
The abbreviatory symbol gf appears in the phrase structure rule for IP in
Russian (following King 1995: 204):

(21) IP −→ XP I!
(↑ gf) = ↓ ↑=↓

An equation such as (↑ gf) = ↓ is satisfied if there is some value of gf

for which the equation is true. Here, the equation is true if the value of
gf is obj.
In this instance, the uncertainty was limited: one member of a disjunction
of grammatical functions was chosen (based, at least, on world knowledge of
the relation between an author and a book). In other cases, there might be
more uncertainty; the phrase might bear a grammatical function more deeply
embedded inside the sentence. This is true for constituent questions in English.
Example (22) shows that the question phrase what can fill the role of obj in the
complement clause comp, appearing as the value of the path comp obj in the
f-structure (for clarity, the attributes making up the path are enclosed in boxes).
In (23), which involves another complement clause, it is the value of the path
comp comp obj:

(22) What do you think Chris bought?

CP

NP C
’

N C IP

What do NP I
’
N VP

you V IP

think NP I

N VP

Chris V

bought
 regular expressions and regular languages 207

(23) What do you think Chris hoped David bought?

CP ’
NP C

N C IP ’
What do NP I

N VP

you V

V IP

think NP I

N VP

Chris V

V IP

hoped NP I

N VP

David V

bought

A simplified version of the annotated phrase structure rule for English constituent
questions is:
 
(24) CP −→ XP C!
 
↓ ∈ (↑ dis) ↑=↓
 
∗
(↑ comp gf) = ↓

The first annotation on the XP daughter of CP, ↓ ∈ (↑ dis), requires the f-structure
corresponding to the XP to be a member of the set value of the dis attribute of
the CP mother node. The second annotation on the XP node, (↑ comp⃰ gf) = ↓,
contains a new sort of expression. As discussed in §6.1, the Kleene star operator ⃰
indicates that an expression may be repeated zero or more times. In particular,
comp⃰ represents paths containing any number of comps: the empty path, comp,
comp comp, and so on. Thus, the equation (↑ comp⃰ gf) = ↓ indicates that within
208 syntactic relations and syntactic constraints

the f-structure ↑, the dis f-structure also bears some grammatical function gf
which lies at the end of some path in the set of paths comp⃰: that is, some gf that
may be embedded inside any number of comps. The constraint holds if there is
some path in the set of paths comp⃰ gf for which the equation is true. In (22), the
path is comp obj. In (23), the path is comp comp obj. In some other example, a
different path might be chosen.
More complicated paths can also be characterized. A slightly more complete
version of the rule for question formation in English is:
 
(25) CP −→ XP C!
 ↓ ∈ (↑ dis)  ↑=↓
 
∗
(↑ xcomp|comp gf) = ↓

The regular expression {xcomp | comp}⃰ denotes paths containing any number
of xcomps or comps in any order: for example, comp xcomp, comp comp, or
xcomp comp xcomp.4
Equations of this sort, involving abbreviatory symbols over grammatical
functions or more complex regular expressions denoting paths through an f-
structure, exemplify functional uncertainty. Functional uncertainty was first
introduced by Kaplan et al. (1987) and Kaplan and Zaenen (1989b) in the
treatment of long-distance dependencies such as topicalization, question formation,
and relative clause formation in English. The expression in (25) more adequately
captures constraints on question formation in English than the one in (24), but still
does not completely characterize the possible grammatical functions which can
be borne by the sentence-initial dis constituent in English questions; a detailed
discussion of the syntax and semantics of long-distance dependencies is provided
in Chapter 17.
Definition (30) in Chapter 5, repeated in (26), tells us how to interpret con-
straints involving a string as of length greater than one:

(26) (f as) ≡ ((f a) s) for a symbol a and a string of symbols s.

(f 𝜖) ≡ f , where 𝜖 is the empty string.

For instance, the f-structure ( f comp subj) is the f-structure at the end of the
path comp subj within the f-structure f ; it is the same as (( f comp) subj),
the subj of the f-structure ( f comp). In (26), the string s is the suffix of the
string as.
Building on this definition, Kaplan and Zaenen (1989b) provide the following
interpretation for constraints involving regular expressions over paths:

4 Note that this expression is not the same as the regular expression {xcomp⃰ | comp⃰} which denotes
paths containing either any number of xcomps or any number of comps: xcomp xcomp xcomp or
comp comp, but not xcomp comp.
 regular expressions and regular languages 209

(27) Functional uncertainty:

If α is a regular expression, then (f α) = v holds if and only if

(f a suff(a, α)) = v

for some symbol a, where suff(a, α) is the set of suffix strings s such that
as∈α.

The definition in (27) is stated with reference to the predicate suff, which takes
two arguments: a string a and a regular language α, a set of strings composed of
grammatical functions. suff(a, α) is the set of strings which are possible continu-
ations of a in the regular language α. For example:

(28) If α is the regular language

{comp subj, comp comp subj, comp comp comp subj, comp xcomp obj}
then suff(comp comp, α) is the set of strings that follow the string
comp comp in α:
{subj, comp subj}

According to the definition in (27), then, we interpret a constraint involving a

regular expression by making a series of left-to-right choices: that is, by choosing
a symbol a that is the first symbol in some string in α, and then choosing a
wellformed continuation of a in α, suff(a, α).
Much work has been done on the formal properties of systems using functional
uncertainty. For an overview discussion, see Dalrymple et al. (1995d). The issue of
decidability and functional uncertainty is treated in detail by Kaplan and Maxwell
(1988a), Baader et al. (1991), Backofen (1993), and Keller (1993).

6.1.3 Inside-out functional uncertainty

By using functional uncertainty, we can specify an f-structure embedded at an
arbitrary depth inside another f-structure. We can also talk about f-structures that
enclose an f-structure at an arbitrary level of distance. This is referred to as inside-
out functional uncertainty, first introduced by Kaplan (1988). The two types of
functional uncertainty are closely related, but they are used in different contexts:
“ordinary” outside-in functional uncertainty is used to define constraints on more
deeply embedded structures, while inside-out functional uncertainty is used to
define constraints on enclosing structures.
Inside-out functional uncertainty is used by Nordlinger (1998, 2000) in her
theory of constructive case. Consider the following Wambaya example (Nordlinger
1998: 65):

(29) Dawu gin-a alaji janyi-ni.

bite 3sg.m.subj-pst boy.m.acc dog.m-erg
‘The dog bit the boy.’
210 syntactic relations and syntactic constraints

The noun janyi-ni ‘dog.m-erg’ bears an ergative marker -ni, signaling its gram-
matical function: nouns marked with ergative case are subjects. According to
Nordlinger’s theory of constructive case, casemarking specifies the syntactic envi-
ronment within which the casemarked phrase must appear. The f-structure corre-
sponding to the n node dominating janyi-ni is:

(30)  
 
d




According to Nordlinger’s rules for constructive case in Wambaya, this f-structure

must appear in the following f-structure environment:

(31)   
  
 d 
b  
 

In (31), the f-structure for janyi-ni is the one labeled d; it is required to bear the
subj relation within its clause. Nordlinger enforces this requirement by means of
the following lexical entry for janyi-ni:

(32) janyi-ni (↑ pred) = ‘dog’

(↑ index gend) = m
(↑ concord case) = erg
(subj ↑)

The first three equations state that the f-structure d for janyi-ni must have a pred
with value ‘dog’, the value m for its gend attribute, and the value erg for its case
attribute, as shown in (30). The expression (subj ↑) in the fourth line of this lexical
entry is different from the other expressions in (32), in that the attribute subj
appears to the left of the f-structure designator ↑. This inside-out expression refers
to an f-structure through which the path subj leads to the f-structure ↑; this is the
outermost f-structure, labeled b, in (31). In (32), this expression is an existential
constraint (see §5.2.6) requiring such an f-structure to exist.
Formally, (a f ) is the f-structure whose value for the attribute a is f :

(33) Inside-out expression:

(a f ) = g holds if and only if g is an f-structure, a is a symbol, and the pair
)a, f * ∈ g.

In (32), the inside-out functional uncertainty path consists of just one attribute,
subj. Longer paths in an inside-out equation are interpreted incrementally, as with
outside-in expressions (see §5.2.1):
 the pcase attribute 211

(34) (𝜖 f ) = f , where 𝜖 is the empty string.

(saf ) = (s (a f )) for a symbol a and a string of symbols s.

As with outside-in functional uncertainty, it is possible to use a regular expression

to characterize a set of paths through the f-structure. This will be useful in
our analysis of anaphora, to be presented in Chapter 14. The use of regular
expressions in inside-out functional uncertainty is similar to its use in outside-
in functional uncertainty: the expression is true if there is some string in the
set of strings picked out by the regular expression for which the expression
holds:

(35) Inside-out functional uncertainty:

(α f ) ≡ g if and only if g is an f-structure, α is a set of strings, and for some
s in the set of strings α, (s f ) ≡ g.

Notice that even when the inside-out path is fixed and the expression con-
taining it appears to be determinate, it may denote any of several f-structures.
Consider the f-structure in (37) for a verb like seem, whose subject is shared
with the subject of its open infinitival complement according to the lexical entry
in (36) (see Chapter 15 for a discussion of verbs taking an open complement
xcomp):

(36) seem (↑ pred) = ‘seem)xcomp*subj’

(↑ subj) = (↑ xcomp subj)

(37) David seemed to yawn.

 
 

 d 

s


 ’ 

 y 

In (37), the f-structure for David, labeled d, is the subj of two f-structures: the
f-structure for seem, labeled s, and the f-structure for yawn, labeled y. In this case,
we can construct an inside-out expression rooted in the f-structure d, (subj d),
which refers to either of the two f-structures of which d is the subj: (subj d) is
either s or y.

6.2 The PCASE attribute

The grammatical function of an oblique PP argument is determined in English
by the preposition that is used. For example, the goal phrase to Chris in a
sentence like David gave the book to Chris bears the grammatical function
oblgoal . As discussed in §2.5.4, Kaplan and Bresnan (1982) propose that the
212 syntactic relations and syntactic constraints

constraint specifying the grammatical function of an oblique phrase is given

by the preposition. In this case, the information that to Chris is an oblgoal
is specified as the value of the pcase attribute in the lexical entry of the
preposition to:5

(38) to (↑ pcase) = oblgoal

Kaplan and Bresnan further propose that the value of the pcase attribute is also
the attribute whose value is the oblique phrase, so that oblgoal is an attribute
name as well as a value. The following annotation on the PP phrase structure node
accomplishes this:

(39) PP
(↑ (↓ pcase)) = ↓

This annotation appears as a part of the rule expanding V! . The full expansion of
the V! node is as follows:

(40) gave the book to Chris

VP

V!

PP
V NP
(↑ (↓ pcase)) = ↓

P!
gave
the book ↑=↓

P
NP
↑=↓

to
N
(↑ pcase) = oblgoal
Chris

Using mnemonic names for f-structures, such as fpp for the f-structure correspond-
ing to the PP node in the c-structure tree, the equations in (40) for the PP and the
nodes it dominates are instantiated as in (41):

5 Kaplan and Bresnan’s analysis prefigures Nordlinger’s (1998) theory of constructive case, discussed
in the previous section, according to which the grammatical function of an argument is specified by
the case morpheme with which it appears.
 talking about sets 213

(41) PP
(fv (fpp fpp

P
fpp = fp

P NP
fp = fp

to N
(fp
Chris

The relevant f-description is the following:

(42) (fv! (fpp pcase)) = fpp

fpp = fp!
fp! = fp
(fp pcase) = oblgoal

These equations tell us that the f-structure fpp corresponding to the PP node is the
same as the f-structures fp! and fp corresponding to the P! and P nodes, and that fp ’s
pcase is oblgoal . Thus, we have the following equivalences:

(43) (fpp pcase) = (fp! pcase) = (fp pcase) = oblgoal

Substituting oblgoal for (fpp pcase) in the first equation in (42), we have:

(44) (fv! oblgoal ) = fpp

The equality induced by the constraint in (44) is explicitly indicated by a line in

this f-structure:

(45)  
’
 
fv 



fpp , fp , fp

6.3 Talking about sets

Sets are used to represent several different types of objects in LFG. Primarily, sets
are used where an unbounded number of elements is in principle allowed: for
coordinate structures, for example, where there is no fixed limit to the number
214 syntactic relations and syntactic constraints

of conjuncts; or for the modifiers of a phrase, where any number of modifiers may
appear. In the following, we discuss ways of describing sets and constraining their
members. We first provide definitions for the standard symbols which are used in
talking about sets and relations between them:
(46) ∈ Set membership. a ∈ S means that a is a member of the set S.
&∈ Negated set membership. a &∈ S means that a is not a member of the set S.
⊆ Subset. A ⊆ B means that A is a subset of B: that is, all of the members of
A are also members of B. If A is a subset of B and vice versa (A ⊆ B and
B ⊆ A), then A = B.
⊂ Proper subset. A ⊂ B means that A is a proper subset of B: that is, all
of the members of A are also members of B, and B has some additional
members that are not in A.
∪ Union. A ∪ B is the set which is the union of A with B, which is the result
of combining the two sets. In other words, the union of set A with set B
contains all members of set A as well as all members of set B.
∩ Intersection. A ∩ B is the set which is the intersection of A with B, which
contains all of the members that A and B have in common.
∅ The empty set, sometimes written as { }. ∅ is a set that has no members.
A! Set complement. In defining the complement (A! ) of a set A, we must
make reference to the universe, which consists of all possible members
of the set; the complement of a set A is then the set containing all of the
elements in the universe that are not in A. For example, suppose that
the universe is all of the letters in the Roman alphabet, and A is a set
consisting of all strings containing any number of as, including none
({𝜖, a, aa, aaa, aaaa, . . . }):6 this is the set characterized by the regular
expression a⃰, as discussed in §6.1. The complement A! is, then, the set of
all strings that contain any letter other than a.

6.3.1 Open set descriptions

An open set description is given by separately specifying the individual elements of
a set. The constraints specifying the elements may be given in different parts of the
grammar, by different phrase structure rules or lexical items. For example, consider
the following rule for the English V! , in which we make the simplifying assumption
that the verb is followed by a single NP object and any number of PP adjuncts:

(47) V! −→ V NP PP∗
↑=↓ (↑ obj) = ↓ ↓ ∈ (↑ adj)

The expression PP⃰ represents a sequence of zero or more PPs. What about the
annotation ↓ ∈ (↑ adj)? This annotation means that each PP that appears is an
adjunct, and so its f-structure is a member (∈) of the adj set of the mother’s
f-structure ↑. That is, there may be zero or more occurrences of the following
annotated node:
6 Recall that 𝜖 is the empty string: see example (3) in this chapter.
 talking about sets 215

(48) PP
↓ ∈ (↑ adj)

The expression in (49) represents an alternative way of specifying set membership:

(49) (↑ adj ∈) = ↓

This expression uses the set membership symbol ∈ as an attribute and states that ↓
is a member of the set (↑ adj). Expressions such as these are sometimes useful in
stating constraints on set members, particularly in expressions involving inside-out
functional uncertainty, discussed in §6.1.3. The two expressions in (50) are equiva-
lent; each states that the f-structure ↓ is a member of the set of f-structures (↑ adj):

(50) ↓ ∈ (↑ adj)
(↑ adj ∈) = ↓

The c-structure and f-structure for a V! like yawn in class on Monday are:

(51) V

PP PP
V
↓∈ (↑ ↓ ∈ (↑

V P P

yawn P NP P NP

in N on N

class Monday
 
’
  

 
 ’ 



 
 
 
  


 


 
   
 
 ’ 
 
 
 
 
 
 


As the annotations on the rule require, the f-structure of each modifying adjunct
PP is a member of the adj set of the f-structure for the mother V! node.
Formally, an expression involving set membership is defined as we would
expect:

(52) Open set description:

g ∈ f holds if and only if f is a set and g is a member of f .

It is also possible to write a constraining expression for set membership:

216 syntactic relations and syntactic constraints

(53) Constraining statement of set membership:

g ∈c f holds if and only if f is a set and g is a member of f in the minimal
solution for the defining equations in the f-description of f .

Rounds (1988) provides more discussion of the description and representation of

sets in LFG.

6.3.2 Distributive and nondistributive features

Sets are also used in the representation of coordinate structures, but in that case
there is a difference: as explained in §2.4.5, a set representing a coordinate structure
is a hybrid object that can have its own attributes and values as well as having
elements. This captures the fact that a coordinate structure such as David and Chris
in an example like (54) has properties that the individual conjuncts do not have:

(54) David and Chris yawn/ ⃰yawns.

Although both David and Chris are singular phrases, the coordinate structure as a
whole is plural. The c-structure and f-structure for such an example are:

(55) David and Chris yawn.

 
’
  
IP  
  
  
NP I    
  
 
NP Cnj[main] NP VP
  




   
  
  

N and N V  

   

 






David Chris yawn 

We present here a simplified, preliminary phrase structure rule for NP coordina-

tion; a more detailed discussion of coordination, including the complex category
Cnj[main], can be found in Chapter 16:

(56) NP −→ NP Cnj[main] NP
↓∈↑ ↑=↓ ↓∈↑
(↑ index num) = pl

As presented in (56), the annotations on the NP daughters require the f-structure

for each conjunct NP to be a member of the f-structure for the coordinate NP.
That is, the f-structure for the NP as a whole is a set, with the NP conjuncts as its
members.
The annotation on the Cnj[main] daughter requires the coordinate structure
to have an index attribute whose value is a structure specifying plural number:
the coordinate structure is a plural phrase. In other words, besides having the
 talking about sets 217

f-structures for David and Chris as members, the set representing the coordinate
structure has the attribute index whose value contains the attribute-value pair
)num,pl*, signifying plural number. What does it mean to specify a property of
a set in this way?

6.3.2.1 Nondistributive features In specifying a property of a set, the property

may or may not distribute to the members of the set, depending on whether
the feature involved is a distributive or a nondistributive feature. We classify the
following features as nondistributive:

(57) Nondistributive features:

index, adj, conj, preconj

If a feature of a set is nondistributive, it is a property of the set as a whole.

Thus, the conj and index attributes and their values specified in the rule in
(56) are a property of the coordinate structure as a whole, not the individual
conjuncts:
(58) ( f index num) = pl (index is a nondistributive feature)
( f conj) = and (conj is a nondistributive feature)
 
 
NP  
 
   
 

Cnj[main]  
NP NP  d 
(f   
 
d∈f
(f
c∈f f
 







 
   
 
N and N 



 
  
  
 c
 
  
David Chris  
 
 

Modifiers can also modify a coordinate structure as a whole, and so the adj feature
is also a nondistributive feature. For example, the adverb alternately in the sentence
David alternately sneezed and coughed modifies the conjoined phrase sneezed and
coughed: the sentence does not mean that David alternately sneezed and that he
also alternately coughed, but that he alternated in performing the two actions. We
assume the adjunction rule in (59), which adjoins an AdvP at the VP level, as well
as the VP coordination rule in (60):

(59) VP −→ AdvP VP
↓ ∈ (↑ adj) ↑=↓

(60) VP −→ VP Cnj[main] VP
↓∈↑ ↑=↓ ↓∈↑
 218 syntactic relations and syntactic constraints

(61) a ∈ ( f adj) (adj is a nondistributive feature)

IP
NP I  
 a 
N VP  
   
 ’ 
AdvP   
David VP  s 

 
a ∈ (f )   
 




 

f 
 
VP 
Adv Cnj[main] VP   
s∈ f     
 ’ 
 

VP   
  


 
alternately V and  c  
 
c ∈ f 



 

sneezed V
coughed

We also classify the features conj and preconj as nondistributive; as we discuss

in §16.3, the form of the conjunction and preconjunction are properties of the
coordinate structure as a whole, not of the individual conjuncts. In an example
like Both Chris and David yawned, the preconjunction both and the conjunction
and contribute values for the conj and preconj features of the coordinate phrase
both Chris and David:

(62) ( f preconj = both) (preconj is a nondistributive feature)

( f conj = and) (conj is a nondistributive feature)
NP  
 
Cnj[pre] NP Cnj[main] NP  
 
(f d∈f (f c∈f f d 
 
 
both N and N  c 
David Chris

6.3.2.2 Distributive features In contrast, a distributive feature and its value are
a property of each member of the set, not of the set as a whole. We assume that all
features other than adj, index, conj, and preconj are distributive. Example (61)
shows that grammatical functions such as subj are distributive attributes, since
specifying the f-structure for David as the subj of the coordinate phrase amounts
to specifying that f-structure as the subj of each conjunct: hence the link between
the subjects in the f-structure in (61).
To take another example, suppose that concord is classified as a distributive
feature, and that we want to specify the concord case of a particular coordinate
phrase f as nom. In this situation, specifying the set f as having a particular
value for concord means that each member of the set f —here, d and c—must
be specified with that value for concord:7

7 Though concord features such as case are often analyzed as distributive features, Peterson
(2004a) and Hristov (2012: Chapter 5; 2013b) argue that case is best treated as a nondistributive feature
 talking about sets 219

(63) ( f concord case) = nom (concord is a distributive feature)

 
   
NP  
 
 
  
(f  d
 

 
  
 
 
 
  
NP Cnj NP  


d∈f (f c∈f
f 
 
    
 
 
N and N   

   

 
  
 c

 
  
David Chris  


 


When they are not features of sets, the behavior of distributive and nondistributive
features is the same.
As features of sets, distributive and nondistributive features are formally treated
in the following way (Dalrymple and Kaplan 2000):

(64) Distributive and nondistributive features:

If a is a distributive feature and s is a set of f-structures, then (s a) = v holds
if and only if (f a) = v for all f-structures f that are members of the set s.
If a is a nondistributive feature and s is a set of f-structures, then (s a) = v
holds if and only if the pair )a, v* is in the set s itself.

6.3.3 Functional uncertainty and sets

In constraints involving functional uncertainty, a regular expression representing
a path through the f-structure can be resolved differently in each conjunct of a
coordinate structure. Kaplan and Maxwell (1988b) provide the definition in (27)
of this chapter, repeated in (65):

(65) Functional uncertainty:

If α is a set of strings, then (f α) = v holds if and only if

(f a suff(a, α)) = v

for some symbol a, where suff(a, α) is the set of suffix strings s such that
as ∈ α.

in light of acceptable examples involving case mismatch. If case is distributive, case mismatch is not
allowed, since any specification of the case value for the set as a whole applies to all of the conjuncts.
See §2.5.7.5 for further discussion.
220 syntactic relations and syntactic constraints

As a consequence of this definition, if a functional uncertainty path involves a set,

the path may unwind differently in each element of the set: a valid path must be
taken within each element, but a different path may be taken for each element.
We return to this point in §17.1.1.1.

6.3.4 Closed set descriptions

A closed set description lists all of the elements of the set in a single constraint,
instead of specifying the elements of the set by means of separate constraints
mentioning each element. As discussed in §2.5.7.2, Dalrymple and Kaplan
(2000) use sets whose members are atomic symbols to represent the values
of attributes like pers; we review their proposal here to illustrate closed set
descriptions.
Dalrymple and Kaplan (2000) propose the following values for the pers feature,
where s and h are atomic symbols mnemonic for speaker and hearer:

(66) Values for the pers feature according to Dalrymple and Kaplan (2000):
{s,h}: first person
{h}: second person
{ }: third person

According to this analysis, the f-structure for a sentence such as We yawned,

containing a first-person pronominal subject, is as follows, where the value {s,h}
represents first person:

(67) We yawned.
 
’
 
  
 
 
  
  
 

The closed set description characterizing the value of the pers attribute for the
pronoun we is:

(68) we (↑ index pers) = {s,h}

This description differs from the open set description in (69), which is consistent
with the presence of other members of the set; the constraint in (68) is not:
 talking about sets 221

(69) we s ∈ (↑ index pers)

h ∈ (↑ index pers)

For example, an additional constraint o ∈ (↑ index pers) is compatible with the

constraints in (69) but not with the constraint in (68).
Notably, there is a close relation between set-based representations for the
value of a feature such as pers and feature-based representations such as the one
proposed by Otoguro (2015), illustrated in §2.5.7.2, example (156), and repeated
here:

(70) Values for the pers feature according to Otoguro (2015):

6 ! "7
1 +
First person: pers
2 −
6 ! "7
1 −
Second person: pers
2 +
6 ! "7
1 −
Third person: pers
2 −

As Sadler (2011) observes, any analysis that is expressed with set values is also
expressible using a feature-based representation, though the reverse does not hold.
In particular, it is possible to translate any set-based analysis of a complex feature
value into a feature-based analysis by representing the presence of a member of the
set with a positive (+) value for the corresponding attribute, and its absence by a
negative (−) value:

(71)
representation: representation:
+
+
−
+
−
Third person { } −

In the feature-based representation, a value of + for a feature indicates that the

feature is present in the set, while a value of − indicates that the feature is
absent. Vincent and Börjars’s (2007) more complex proposal for the representation
of the value of pers in (72) can also be restated in feature-based terms, as
shown in (73):
222 syntactic relations and syntactic constraints

(72) Values for the pers feature according to Vincent and Börjars (2007):
{s,h}: first person inclusive
{s}: first person exclusive
{h}: second person
{ }: third person

(73)
representation: representation:
+
+
+
−
−
+
−
Third person { } −

In fact, if we rename the s attribute as 1 and the h attribute as 2, it becomes clear

that Otoguro’s feature-based representation of complex pers values in (70) is very
close to the representation in (73).
The feature-based representation also differs from the set-based representation
in straightforwardly allowing for underspecification. With a closed set specifica-
tion, the members of the set must be explicitly enumerated; it is not possible for a
set to be underdetermined as to whether or not it contains a particular member.8
In contrast, in a feature-based representation a feature and its value can remain
unspecified, and this can indicate a lack of information about whether the feature
is present (with value +) or not (with value −).

6.4 Reference to sister nodes

In most work within LFG, only the up arrow ↑ and down arrow ↓ annotations are
used in rules. This embodies a strong locality claim: only the functional relation
between a daughter category and its mother can be stated, and not the relations
among the f-structures of any other nodes in the tree. In some work, however,
reference can also be made to the f-structures of sister nodes in a phrase struc-
ture rule. For example, Nordlinger (1998) presents an analysis of morphological

8 Kaplan (2017) points out that subsumption (§6.9.2) can be used in a set-based analysis to capture
many of the effects of underspecification in a feature-based setting.
 local names for f-structures 223

composition in which crucial reference is made to the immediate left sister of a

node and its f-structure.
In a phrase structure rule, the left sister of a node can be referred to by using the
symbol < ⃰ , and its f-structure is represented as 𝜙(< ⃰ ). Similarly, the symbol ⃰ > is
used to refer to the immediately adjacent right sister.9 This additional expressivity
expands the domain of locality slightly, but not beyond the local mother-daughter
configuration that is described by the c-structure rule.

6.5 Local names for f-structures

In expressing constraints on f-structures, a local name can be used in a lexical entry
or annotated phrase structure rule to refer to an f-structure (Kaplan and Maxwell
1996). The reference of a local name is restricted to the lexical item or rule element
within which it occurs: this means that if the same local name is used in more than
one daughter in a rule or more than one different lexical item, it refers to different
f-structures, not the same f-structure. A local name begins with the percent sign %.
A local name is particularly useful in expressions involving functional uncer-
tainty: it makes it possible to name a particular f-structure that participates in the
uncertainty and to place constraints on it. For example, the relative pronoun in
Russian agrees in number and gender with the head noun of the noun phrase.
Lipson (1981) discusses the following Russian example, in which the masculine
singular relative pronoun kotorogo must be used with a masculine singular noun
like park:

(74) park, okolo kotorogo ja živu

park.m.sg near which.m.sg I live
‘the park near which I live’

As (74) shows, the relative pronoun can appear as a subconstituent of a

displaced phrase such as okolo kotorogo ‘near which’. The f-structure for this
example is:10

9 Nordlinger (1998) uses the left arrow ← to refer to the f-structure of the left sister, and the right
arrow → to refer to the f-structure of the right sister. We prefer the use of 𝜙(< ⃰ ) and 𝜙(⃰ >) for the
f-structures corresponding to the left and right sisters of a node, since this avoids confusion with the
use of the ← symbol in off-path constraints: see §6.6.
10 In §17.1.2, we augment the f-structure for relative clauses with an attribute relpro, whose
value is the f-structure of the relative pronoun within the clause-initial phrase. Although the pres-
ence of the relpro attribute would allow for a considerable simplification in the analysis of this
example, we omit this attribute here in order to provide a clear illustration of how a local name is
used.
224 syntactic relations and syntactic constraints

(75) park, okolo kotorogo ja živu

park.m.sg near which.m.sg I live
 
 
 
 
 
    
   
 
  ’  
 
 
  
  
   
 
  
 
  
 
 
  
   
 
    
  
 
 
        
 
   g   
  

 
  
     
 
  
 
  
 
  
 
  
 
 
   
   
   ’  
 
   
 
     
 
   
 
   
 
     
 
     
 
   
 
   
 
   
 
 
 
 

In the analysis of (74), we would like to impose an agreement requirement that

allows us to refer to the relative pronoun, which may be embedded at an arbitrary
depth within the dis phrase, and to constrain its num and gend features. The
following phrase structure rule accomplishes this:
8 9  
(76) NP −→ N CP
 ↓ ∈ (↑ adj) 
↑=↓  
 
 (↓ dis ∈ gf⃰) = %relpro 
 
 (%relpro prontype) =c rel 
 
 (%relpro index num) = (↑ index num) 
(%relpro index gend) = (↑ index gend)

This rule states that a Russian noun phrase consists of a head noun N and a CP
relative clause. The first cp constraint, ↓ ∈ (↑ adj), requires the CP’s f-structure to
be a member of the set of modifiers of the NP (see §6.3.1 for more on specification
of set membership). This is true of the f-structure in (75).
The relative clause CP contains a relative phrase, the phrase okolo kotorogo ‘near
which’ in (74). The CP rule (not displayed here) ensures that the f-structure of this
relative phrase is a member of the dis set within the relative clause. According to
the rule in (76), this dis f-structure must contain a relative pronoun at some level
of embedding gf⃰ inside the dis. This f-structure is referred to by a name local to
this rule element as %relpro:

(77) (↓ dis ∈ gf⃰ ) = %relpro

 off-path constraints 225

The final three annotations in (76) place further constraints on the f-structure
%relpro: it must have a prontype of rel (on prontype, see 2.5.7.7), and its num
and gend must match the num and gend of the mother NP:

(78) (%relpro prontype) =c rel

(%relpro index num) = (↑ index num)
(%relpro index gend) = (↑ index gend)

These constraints are satisfied if %relpro names the f-structure labeled g in (75).
Using a local name like %relpro is essential in this instance: the use of a local name
ensures that all of the constraints in (78) refer to the same f-structure. In particular,
a set of expressions like the following are not equivalent to those in (78):

(79) (↓ dis ∈ gf⃰ prontype) =c rel

(↓ dis ∈ gf⃰ index num) = (↑ index num)
(↓ dis ∈ gf⃰ index gend) = (↑ index gend)

The equations in (79) require some f-structure in the dis set to contain an f-
structure with a prontype of rel, a possibly different f-structure to have the
same num as the full noun phrase, and a possibly different f-structure to have the
same gend as the full noun phrase; crucially, these constraints impose no require-
ment for the same f-structure to satisfy all of these constraints. It is the use of a
local name that enforces the proper requirement.

6.6 Off-path constraints

Off-path constraints (Jam 1990; Kaplan and Maxwell 1996; Crouch et al. 2008;
Bresnan et al. 2016: §4.8) allow for constraints on features of the f-structures
through which some long-distance path is defined. Let us consider the simple
functional uncertainty expression in (80), which played a role in the annotated rule
in (24). This equation states that the f-structure ↓ is the value of some grammatical
function gf that may be embedded inside one or more comps:

(80) (↑ comp⃰ gf) = ↓

This equation involves a path consisting of a grammatical function gf embedded

inside any number of comp f-structures, but places no constraints on other features
of the f-structures through which the path passes. Suppose, however, that we are
interested not in all comp f-structures, but only in those which contain some
additional feature; in other words, we are concerned with a feature that is “off
the path” traversed by the functional uncertainty expression comp⃰ gf. Off-path
constraints allow us to restrict attention to f-structures with the relevant features,
as we now show.
226 syntactic relations and syntactic constraints

6.6.1 The f-structure →

Some English verbs allow extraction from their clausal complements, while others
do not. For many (but not all) speakers, verbs of manner of speaking do not allow
extraction (Erteschik 1973):

(81) Who did Chris think/ ⃰whisper that David saw?

Verbs allowing extraction are often called bridge verbs, while those disallowing
extraction are called nonbridge verbs.
There is no reason to assume that the grammatical function of the clausal
complements of these two verbs differs. Other than this difference between them,
they behave the same syntactically; both bear the grammatical function comp.
A sentence with a bridge verb allowing extraction has an f-structure like the
following:

(82) Who did Chris think that David saw?

We propose that a nonbridge verb specifies that its complement bears a feature
that bridge verbs lack, which we will call ldd (for long-distance dependency),
with value −. The path in a long-distance dependency may not pass through an
f-structure with this feature:

(83) ⃰Who did Chris whisper that David saw?

’
−
 off-path constraints 227

In (83), the displaced question constituent is related to its within-clause function

obj by means of an equation such as the following on the phrase structure rule
dominating the question phrase:

(84) (↑ comp obj) = ↓

The attributes comp and obj do not reflect the prohibition against extraction.
Instead, this requirement must be stated “off the path” characterizing the depen-
dency, as an additional condition on the f-structures along the path. In this case,
we would like to express the following constraint:

(85) A comp in the extraction path must not contain the pair )ldd, −*.

We can use the expression → in an off-path constraint to express this requirement:

(86) (↑ comp obj) = ↓

(→ ldd) & = −

In the expression in (87), the right arrow → stands for the value of the attribute
comp:

(87) comp
(→ ldd) &= −

In (88), the comp attribute is boxed, and the f-structure denoted by → is labeled s:

(88) ⃰Who did Chris whisper that David saw?

’
−
s

The f-structure s contains the attribute ldd with value −. This is forbidden by
the negative constraint (→ ldd) &= − in (86), accounting for the ungrammaticality
of (88).
Slightly more generally, we can use an expression like the following to constrain
long-distance dependencies in English:11

11 As in §6.1.2, this provisional characterization of constraints on question formation in English

is incomplete; we provide a more complete treatment of the syntax and semantics of questions in
Chapter 17.
228 syntactic relations and syntactic constraints

(89) (↑ comp⃰ gf) = ↓

(→ ldd) & = −

This expression indicates that any number of occurrences of the annotated comp
attribute comp displayed in (87) can occur in the long-distance path; in
(→ ldd) &= −
other words, the f-structure ↓ bears some grammatical function gf embedded
inside any number of comps, as long as none of the comp f-structures on the path
contain the pair )ldd, −*.

6.6.2 The f-structure ←

We can also use the left arrow ← in off-path constraints to denote the f-structure
which contains the annotated attribute and its value. To see the difference between
the left arrow ← and the right arrow → in off-path constraints, consider the
following expression, containing a generic attribute represented by attr and a
generic value value:

(90) ( f dis) = ( f comp obj)

(← attr) = value

The occurrence of comp which is annotated with the off-path constraint is boxed
in (91), and the left arrow ← in the off-path constraint refers to the f-structure
in which the annotated attribute comp appears, the one labeled f . The off-path
constraint (← attr) = value requires f to have an attribute attr with value
value:

(91) F-structure satisfying the constraint in (90):

 ! "
pred ‘pro’
 dis 
 prontype wh 
 
f
 attr value


 : ; 
 
comp g obj

If we replace the left arrow by the right arrow, the result is as in (92): it is not f
which contains the attribute and value in question, but g, the value of the comp
attribute within f :

(92) a. ( f dis) = ( f comp obj)

(→ attr) = value
b. F-structure satisfying the constraint in (92a):
 ! " 
pred ‘pro’
 dis 
 prontype wh 
  

f 
 attr value 
 comp g   
obj
 off-path constraints 229

Przepiórkowski and Patejuk (2012) use the f-structure metavariable ← in their

analysis of case assignment in Polish. They provide the following summary of
Polish casemarking requirements:

(93) a. subjects bearing structural case are in the nominative,

b. with the exception of numeral phrase subjects, headed by so-called
governing numerals, which are in the accusative;
c. objects bearing structural case are in the accusative,
d. unless they are in the syntactic scope of sentential negation, in which case
they are in the genitive.

These requirements hold for coordinate structures as well, with the result that
the conjuncts in a coordinate structure may bear different cases. For example, a
nominative noun phrase may be coordinated with an accusative numeral phrase,
forming a coordinated subject in which each conjunct obeys the casemarking
rules in (93):

(94) [Janek i jego pie˛ć córek] głosowali

Janek.nom.sg.m and his five.acc.pl.f daughters.gen.pl.f voted.3pl.m
przeciw ACTA.
against ACTA
‘John and his five daughters voted against ACTA.’

Przepiórkowski and Patejuk propose two analyses for this pattern; the second,
which they call the “conservative solution,” involves off-path constraints and the
features sc and acm. The sc feature marks f-structures which are assigned struc-
tural case, and the acm attribute with value rec marks f-structures for numerals
governing a genitive noun. These features are involved in these implicational con-
straints on casemarking (see §5.2.10 for definition and discussion of implicational
constraints):12

(f ) =c +
(95) ⇒ (f )=
(f )=
(f ) =c +
b. ⇒ (f )=
(f ) =c

The constraint in (95a) specifies that subjects that are assigned structural case
(these have the value + for the attribute sc) and are not numeral phrases (and
so do not have the value rec for the feature acm) are assigned nominative case.
The constraint in (95b) specifies that subjects that are assigned structural case and
are numeral phrases are assigned accusative case.

12 Przepiórkowski and Patejuk’s treatment of casemarking has been simplified for the purposes of
exposition; see their paper and §2.5.7.5 for more discussion and a formal analysis of the case attribute
and its value.
230 syntactic relations and syntactic constraints

The issue, as Przepiórkowski and Patejuk observe, is that this constraint cannot
be evaluated with respect to a coordinate structure as a whole, since, as shown
in (94), some conjuncts in a coordinate phrase might be assigned nominative
case according to constraint (95a), while others are assigned accusative case
according to constraint (95b). Their solution involves passing the constraint into
the coordinate structure via an off-path constraint on a distributive feature, so that
the constraint can be satisfied independently for each conjunct.
According to Przepiórkowski and Patejuk’s analysis, casemarking for Polish
subjects is controlled by the following constraint, involving a conjunction of the
implicational constraints in (95):

(96) (↑ subj pred )

(← ) =c +
⇒ (← )=
(← )=
(← ) =c +
⇒ (← )=
(← ) =c

The entire expression is an existential constraint, requiring the subject to contain

a pred. Since subj and pred are distributive features, the off-path constraint
applies to each conjunct separately: ← refers to an f-structure with the attribute
pred, which is either the single pred of a noncoordinate subj or each pred in
a coordinate subject. This entails that the existential constraint holds of each
conjunct when the subject is a coordinate phrase, as required, and that the correct
case (though possibly a different case) is assigned in each conjunct.

Formally, we define the expressions ← and → as they are used in off-path

constraints in the following way:

(97) Off-path constraints:

In an expression like a, ← refers to the f-structure of which a is an
attribute. (← s)

In an expression like a, → refers to the value of the attribute a.

(→ s)

Using the f-structure variables ← and →, any kind of constraint can be written
as an off-path constraint: defining equations, constraining equations, existential
constraints, and other kinds of f-descriptions may be specified. We return to a
discussion of off-path constraints and long-distance dependencies in Chapter 17.

6.7 Templates
The previous examples used functional descriptions associated with lexical entries
and phrase structure rules in determining the f-structure for a particular annotated
 templates 231

c-structure tree. The lexical entries and phrase structure rules of a language are
associated with f-descriptions that characterize and constrain the f-structures
(and, as we will see, other structures) of the utterances in which they appear. In
formulating a theory of the structure of the lexicon and permissible annotations
on phrase structure rules, it is important to be able to specify commonalities in
f-descriptions associated with lexical entries and annotated phrase structure rules.
Templates are used to express these commonalities.

6.7.1 Templates in lexical descriptions

Linguistic theories have adopted different views as to how commonalities in lexical
entries should be captured. Early theories of the lexicon viewed it as “a kind of
appendix to the grammar, whose function is to list what is unpredictable and
irregular about the words of a language” (Kiparsky 1982). Such views were (and
are) common among proponents of transformational approaches to syntax, since
important linguistic generalizations are assumed to be best encoded transforma-
tionally, with the lexicon as a catch-all for linguistic facts that cannot be represented
in general transformational terms.
With the advent of constraint-based, nontransformational theories like LFG,
an alternative view of the lexicon emerged. Bresnan (1978) observed that many
structural and behavioral generalizations over words and constructions are in
fact best captured by lexical redundancy rules: for example, the active and passive
forms of a transitive verb, or the base and dative-shifted variants of a ditransitive
verb, are related by lexical rules rather than by syntactic transformations (see
Bresnan et al. 2016, Chapter 3 for further discussion of this point). On this view,
lexical information is no longer merely exceptional, idiosyncratic, and therefore
theoretically uninteresting. Instead, the lexicon and the rules relating lexical items
become a prime locus of syntactic generalizations. The architecture of a richly
structured lexicon and an articulated theory of relations among lexical forms is
a hallmark of LFG, and the relations between lexical forms and the structure of the
lexicon are an important focus of theoretical work.
One of the first proposals for explicitly representing generalizations that hold of
classes of lexical items was made by Flickinger (1987), who represents the lexicon
as a hierarchy of feature structures. Each node in the hierarchy represents some
aspect of syntactic structure which is common to a number of lexical items: the
word yawns belongs to the third-person singular present-tense class (like devours,
cooks, and so on), the intransitive class (like coughed, hiccup, and so on), and to
other classes as well. Classes of lexical items may be subclasses of other classes, or
may partition other classes along several dimensions.
Work within LFG differs from this approach in that it does not appeal to a
hierarchy of functional structures to encode linguistic generalizations. Instead,
in keeping with the constraint-based view that pervades the theory of LFG, lin-
guistic generalizations are expressed not in terms of relations between structures,
but in terms of relations between descriptions of structures. An LFG functional
description can be given a name, and this name can be used to stand for that
232 syntactic relations and syntactic constraints

functional description. These named descriptions are referred to as templates

(Dalrymple et al. 2004b).
To illustrate the use of templates, we begin with a simple uninstantiated
f-description associated with the verb yawns:

(98) yawns (↑ pred) = ‘yawn)subj*’

(↑ vtype) = fin
(↑ tense) = prs
(↑ subj index pers) = 3
(↑ subj index num) = sg

This lexical entry contributes a pred value for the verb, and also contains informa-
tion about the form of the verb and its subject agreement requirements.
Some of the elements of this f-description are shared with other verbs:
for example, all present tense verbs contribute the equations (↑ vtype) = fin
and (↑ tense) = prs, and all third person singular verbs contribute (↑ subj index
pers) = 3 and (↑ subj index num) = sg. We can define the templates present
and 3sg to encode this information, associated with the appropriate pieces of the
functional description:

(99) Template definitions for the templates present and 3sg:

a. present ≡ (↑ vtype) = fin
(↑ tense) = prs
b. 3sg ≡ (↑ subj index pers) = 3
(↑ subj index num) = sg

The template name present names the functional description consisting of the
two equations (↑ vtype) = fin and (↑ tense) = prs, and similarly for 3sg. In fact,
however, we can further subdivide the functional description present in terms of
the following template definitions:

(100) a. finite ≡ (↑ vtype) = fin

b. prestense ≡ (↑ tense) = prs

The template present can be defined in terms of these simpler templates:

(101) present ≡ @prestense

@finite

When a template is referred to in a lexical entry or in the definition of another

template, as in (101), it is preceded by an ‘at’ sign “@.” These template defi-
nitions can be arranged in a template hierarchy to indicate interdependencies.
Template hierarchies are usually represented as a graph, with the relation between
 templates 233

nodes in the graph representing relations between templates. In the case at hand,
we have the template hierarchy fragment in (102), showing that the template
present is defined by reference to the template prestense and the template
finite:

(102) present
prestense finite

Note that this template hierarchy is unlike the type hierarchies of Head-Driven
Phrase Structure Grammar (Pollard and Sag 1987, 1994; Ginzburg and Sag 2000)
in that it encodes a relation between f-descriptions rather than a relation between
structures. The template hierarchy in (102) indicates merely that the present
template is defined by reference to the prestense and finite templates. Templates
stand for f-descriptions, and f-descriptions can be conjoined, disjoined, or negated.
Thus, the f-description contributed by the daughter nodes in this hierarchy may
be related in a complex way to the f-description of the mother node, including via
negation. This is not possible in a type hierarchy, where the hierarchy represents
inheritance in a semilattice. Asudeh et al. (2008, 2013) provide further discussion
of this point.
We can also subdivide the 3sg template as follows:

(103) ≡ (↑
≡ (↑

This gives the following definition of the template 3sg:

(104) 3Sg ≡ @3PersonSubj

@SingSubj

This information can also be represented as a template hierarchy:

(105) 3sg

3PersonSubj SingSubj

Finally, we can define a template pres3sg for verb forms like yawns:

(106) pres3sg ≡ @present

@3sg

Our template hierarchy then becomes:

234 syntactic relations and syntactic constraints

(107) Pres3sg
present 3sg

prestense finite 3PersonSubj SingSubj

We can also provide a template that can be used with all intransitive verbs, by using
a parametrized template. Parametrized templates take one or more arguments,
written in parentheses after the template name. Here, we define a parametrized
template which takes a single argument providing the pred specification:

(108) intransitive(_p) ≡ (↑ pred) = ‘_p)subj*’

In the definition of the parametrized template, the argument _p is written with

an underscore. When the template is used, it must be supplied with an argument,
which is used as the pred specification for the lexical entry. For the verb yawns,
the template is called in the following way:

(109) @intransitive(yawn)

Once we have expressed all of the relevant pieces of f-description in terms of

templates, we have the following template definitions:

(110) a. intransitive(_p) ≡ (↑ pred) = ‘_p)subj*’

b. pres3sg ≡ @present
@3sg
c. present ≡ @prestense
@finite
d. 3sg ≡ @3PersonSubj
@SingSubj
e. prestense ≡ (↑ tense) = prs
f. finite ≡ (↑ vtype) = fin
g. 3PersonSubj ≡ (↑ subj index pers) = 3
h. SingSubj ≡ (↑ subj index num) = sg

Given these template definitions, the lexical entry for yawns is:

(111) yawns @intransitive(yawn)

@pres3sg

The parametrized template intransitive(_p) is shared by verbs like sneezed,

arrive, and many others. The pres3sg template is shared by verbs like appears,
goes, cooks, and many others. The template present, used in defining the pres3sg
template, is also used by verbs like bake, are, and many others. The use of templates
allows commonalities between lexical entries to be represented succinctly and
linguistic generalizations to be encoded in a theoretically motivated manner.
 templates 235

6.7.2 Templates in phrase structure rules

Since templates simply stand for pieces of functional descriptions, it is also possible
to use templates in annotations on phrase structure rules, to capture recurring gen-
eralizations in the specification of the relation between c-structure configurations
and f-structures. There is no difference in the way templates are defined or called
when they are used in phrase structure rules; the functional annotations in phrase
structure rules can simply be replaced with a template.
For example, Bresnan (1982a: 323) discusses control of depictive adjuncts,
showing that the subject of a depictive adjunct within the VP can be controlled
by the matrix subject, object, or secondary/thematically restricted object:13

(112) a. The package arrived unopened. (subj control: the package is unopened)
subj
b. John will serve the fish to you raw. (obj control: the fish is raw)
obj
c. I sent you the letter sealed. (obj𝜃 control: the letter is sealed)
obj𝜃

We assume that the adjective phrase in the VP-internal examples in (112) is

adjoined at the VP level, and that the adjective phrase bears the open adjunct role
xadj. Under these assumptions, the c-structure and f-structure for (112a) are:

(113) The package arrived unopened.

IP  
’
NP I  
 
 
D N VP  + 
 
 
the N VP AP 
 ’ 

 
package V A

arrived unopened
In this configuration, the subject of the xadj unopened is required to be the same
as the matrix subj, obj, or obj𝜃 :

(114) VP −→ VP AP⃰
↑=↓ ↓ ∈ (↑ xadj)
(↑ {subj|obj|obj𝜃 }) = (↓ subj)

For clause-initial adjuncts, the choice of the controller is more restricted;

Bresnan (1982a: 325) claims that clause-initial adjective phrase adjuncts

13 For more on control, see Chapter 15.

236 syntactic relations and syntactic constraints

permit control only by the matrix subject, and not by the object or other
arguments:

(115) a. subj control (Mary is sure of winning):

Sure of winning, Mary entered the competition yesterday.
subj
b. obj control not possible:
⃰Sure of winning, the competition excited Mary yesterday.
obj

We assume that clause-initial adjuncts are adjoined to IP according to the rule in

(116), with the matrix subj specified as controller of the xadj subj. The tree for
example (115a) is shown in (117):

(116) IP −→ AP⃰ IP
↓ ∈ (↑ xadj) ↑=↓
(↑ subj) = (↓ subj)

(117) IP

AP IP

NP I!
Sure of winning
N VP

Mary
entered the competition yesterday

The annotations on AP in the rules in (114) and (116) are similar but not the same,
and indeed, as we will see in §15.8.1, Icelandic makes use of a phrase structure rule
for depictive xadj that is identical to (114). By using a template, we can capture
the commonalities and differences in these rule annotations within and across
languages.
We can define the parametrized template xadj in the same way as for the
parametrized intransitive template in (108), with an argument for the gram-
matical function of the matrix controller:

(118) xadj(_p) ≡ ↓ ∈ (↑ xadj)

(↑ _p) = (↓ subj)

Given this definition, we can rewrite the rules in (114) and (116) in a way
that illuminates the similarities and differences between them; the similarities
are embodied in the call to the template xadj, while the difference in control
possibilities between the two constructions is reflected in the different arguments
to each of the templates:
 c-structure rule macros 237

(119) a. VP −→ VP AP⃰
↑=↓ @xadj({subj|obj|obj𝜃 })
b. IP −→ AP⃰ IP
@xadj(subj) ↑=↓

Templates have been adopted in LFG-based work on constructions: Asudeh et al.

(2008, 2013) provide an analysis of the way-construction (for example, David dug
his way out of prison) which relies crucially on templates to express commonalities
and differences in this construction and related constructions within and across
languages. Other work relying on templates includes King et al. (2005) and Asudeh
et al. (2014); Kaplan and Maxwell (1996) and Crouch et al. (2008) provide more
discussion and exemplification.

6.8 C-structure rule macros

As we have seen, the use of c-structure metacategories allows us to express gener-
alizations over natural classes of c-structure categories; as discussed in §5.1.3, the
metacategory XP can be defined as one of the phrasal categories in (120), and this
metacategory can be used in the statement of generalizations that hold across these
categories:

(120) Definition of the metacategory XP:

XP ≡ {NP | PP | VP | AP | AdvP}

As shown in the previous section, we can express generalizations in the domain

of functional descriptions by means of templates. We may also wish to encode
more complex generalizations involving c-structure categories annotated with
functional descriptions; we can do this by using a c-structure rule macro.
Consider, for example, the structure-function correlation proposed in §4.2.2.1:
complements of functional categories are f-structure co-heads. The rules in (121)
conform to this generalization, since both the head (I and C) and the complement
(VP and IP) bear the annotation ↑ = ↓:

(121) I! −→ I VP
↑=↓ ↑=↓
C! −→ C IP
↑=↓ ↑=↓

We can encode the relevant generalization by means of the rule macro in (122),
which provides a definition of coheadedness, and captures the commonalities
between these two rules. The definition of a rule macro is similar to a template
definition; in particular, rule macros can be parametrized, taking arguments. The
macro definition in (122) takes two arguments, _h1 and _h2, representing the two
cohead categories in the rule:
238 syntactic relations and syntactic constraints

(122) cohead(_h1,_h2) ≡ _h1 _h2

↑=↓ ↑=↓

This macro definition is unlike a metacategory in that its categories are not
plain (unannotated) categories, but are annotated with functional descriptions;
it is unlike a template in that the definition includes one or more annotated c-
structure categories, not only an f-description as in the definition of a template.
Of course, metacategories and templates can be used in the definition of a rule
macro, in the same way as they are normally used on the right-hand side of a phrase
structure rule.
Using the macro in (122), we can restate the two rules in (121) as in (123), by
using the cohead rule macro. The rules in (123) are exactly equivalent to the rules
in (121):

(123) I! −→ @cohead(I, VP)

C! −→ @cohead(C, IP)

As with templates, a call to a rule macro is preceded by an at sign “@.” A c-structure

rule macro can be used in place of any sequence of annotated categories in the
grammar; like a template, it allows for the expression of generalizations in the
grammar, commonalities across classes of annotated rules.

6.9 Relations between f-structures

6.9.1 F-command
F-command is a relation between f-structures analogous to the c-command rela-
tion defined on trees (Reinhart 1976). F-command was originally defined by
Bresnan (1982a) in the following way:

(124) F-command:
f f-commands g if and only if f does not contain g, and all f-structures that
contain f also contain g.

In (125a) and (125b), the f-structure labeled f f-commands the f-structure labeled
g. In (125a), but not in (125b), g also f-commands f :

(125) f f-commands g:
 
f[ ] f[ ]
a. b.  
g[ ] g[ ]

The definition of f-command given in (124) is correct for cases like (125). However,
as pointed out by Ron Kaplan (p.c.), this definition may not make the right
predictions in cases in which two attributes share the same value. Consider the
 relations between f-structures 239

f-structure in (126), where the f-structure labeled f is the subj as well as the
xcomp subj:

(126)  f[ ]

 
 g[ ] 
 
 
h

The f-structure labeled f in (126) does not f-command the f-structure labeled g
according to the definition in (124), because there is an f-structure (namely h) that
contains f but does not contain g. For the f-command relation to hold between f
and g, we can formulate a new definition of f-command using inside-out functional
uncertainty (defined in §6.1.3):

(127) F-command, definition 2:

f f-commands g if and only if (f gf⃰) &= g ( f does not contain g) and
((gf f ) gf+ ) = g (there is an f-structure (gf f ) which contains g at some
level of embedding).

F-command is important in the definition of binding relations between pronouns

and their antecedents: in many cases, the antecedent of a reflexive pronoun like
himself must f-command that reflexive pronoun. The contrast in acceptability
between (128a) and (128b) is due to the fact that in (128a), the antecedent f of
the reflexive pronoun himself f-commands the f-structure g of the pronoun, while
the f-command relation does not hold in (128b), because there is an f-structure
m that contains f but not g:

(128) a. Davidi saw himselfi .

 
’
 
 f 
 
 
 
g

b. ⃰Davidi ’s mother saw himselfi .

 
’
 
 
 
 m 
 f 
 
 
 
 g 
240 syntactic relations and syntactic constraints

Chapter 14 provides a more in-depth discussion of constraints on anaphoric bind-

ing; there, we will see that the f-command condition for antecedents of reflexive
pronouns follows as a corollary from the binding requirements for reflexives, along
the lines of the definition in (127).

6.9.2 Subsumption
Subsumption is a relation that holds between two f-structures f and g if g is
compatible with but perhaps has more structure than f . In other words, f subsumes
g if f and g are the same, or if g is the same as f except that it contains some
additional structure that does not appear in f . For example, the f-structure labeled
f in (129) subsumes the f-structure labeled g:
(129) f subsumes g:
 
’
   
’ 
  

f  g


  
  


The subsumption relation can be formally defined recursively, as follows:

(130) Subsumption:
An f-structure f subsumes an f-structure g (f 2 g) if and only if:
f = g; or
f and g are sets, and for each member f1 of f there is a member g1 of g such
that f1 2 g1 ; or
f and g are f-structures, and for each attribute-value pair )a, v1 * ∈ f , there
is a pair )a, v2 * ∈ g such that v1 2 v2 .

6.9.3 Generalization
Intuitively, the generalization of two f-structures is the structure that they have in
common. For example, in (131) the f-structure labeled f is the generalization of
the f-structures g and h:
 relations between f-structures 241

(131) f is the generalization of g and h:

’
f
+
 
  ’
’ 



   
   
g  h 
   + 
 
+  

Kaplan and Maxwell (1988b) use generalization in their analysis of coordination,

proposing that the value of an attribute of a set is the generalization of the values of
the attributes of the elements of the set. Their proposal predates the introduction of
the distinction between distributive and nondistributive features, and is no longer
generally accepted; we do not adopt it here.
Formally, the generalization f1 3 f2 of two f-structures f1 and f2 is defined
recursively as follows (see also Kaplan and Maxwell 1988b):

(132) Generalization:
An f-structure f is the generalization f1 3 f2 of two f-structures f1 and f2 if
and only if:
f1 = f2 = f ; or
f1 and f2 are f-structures, and for each pair )a, v1 * ∈ f1 , if there is a pair
)a, v2 * ∈ f2 , then )a, v1 3 v2 * ∈ f ; or
f1 and f2 are sets, and each member of f is the generalization of some
member of f1 and some member of f2 .

Unlike many previous definitions of generalization, (132) defines the generaliza-

tion of two sets. This definition has an interesting consequence: the generalization
of two sets may not be unique. For instance, consider the two sets in (133a) and
(133b):
<! "! "=
(133) a. f1 v1 f3 v3
f2 v2 f4 v4
<! "! "=
b. f1 v1 f2 v2
f3 v3 f4 v4
242 syntactic relations and syntactic constraints

According to the definition in (132), both of the following two sets constitute a
generalization of the sets in (133):
>? @? @A
(134) a. f1 v1 f4 v4
>? @? @A
b. f2 v2 f3 v3

6.9.4 Restriction
The restriction of an f-structure with respect to an attribute can be intuitively
defined as the f-structure that results from removing the attribute and its value
(Kaplan and Wedekind 1993). The f-structure labeled g |TENSE in (135) is the
restriction with respect to tense of the f-structure labeled g:
(135) g |TENSE is the restriction of g with respect to tense:
 
’
 
g 

’
g |TENSE

More formally, Kaplan and Wedekind (1993) define the restriction of an f-structure
f with respect to an attribute a as follows:14

(136) Restriction:
f |a ≡ {)s, v* ∈ f : s &= a}
If a is a set-valued attribute, restriction removes the specified member of
the set:
B
f |a if (f a) = {v}
f |)av* ≡
f |a ∪ {)a, (f a) − {v}*} otherwise

Restriction is useful if an f-structure plays two syntactic roles, but with different
syntactic requirements associated with each role. For example, let us assume that
some f-structure f is shared as the subj value of two different f-structures g1 and
g2 , but that f must take on a different case value in each structure. The equation in

14 The definition in (136) allows us to characterize a set by describing its elements, according to the
following format:
Y = {X : ...constraints on X... }
According to this schematic definition, Y is a set containing all elements X that satisfy the constraints
on X that appear after the colon. For example, in the first part of the definition in (136), f |a is a set of
all pairs )s, v* which are in f and which satisfy the constraint s & = a. In other words, the first part of the
expression, )s, v* ∈ f , directs us to consider all of the attribute-value pairs )s, v* in the f-structure f , and
the constraint s & = a restricts attention to those attribute-value pairs in f in which the attribute is not a.
 relations between f-structures 243

(137a) requires all of the attribute-value pairs of f other than case to be the same
as the attribute-value pairs of g1 ’s subj other than case, and the equation in (137b)
imposes the same requirement for g2 :

(137) a. f |CASE = (g1 subj)|CASE

b. f |CASE = (g2 subj)|CASE

We can then specify different case values for the subjects of g1 and g2 ; the
constraints in (138) are consistent with the requirements in (137):

(138) a. (g1 subj case) = nom

b. (g2 subj case) = acc

Kaplan and Wedekind (1993) use restriction in their analysis of the semantics of
modification. We do not adopt their analysis in this book; we discuss the syntax and
semantics of modification in Chapter 13. Asudeh (2011, 2012) uses the restriction
operator in his analysis of syntactically inactive resumptive pronouns (§17.2).

6.9.5 Priority union

Kaplan (1987) first proposed the operation of priority union, defined in (139):15

(139) Priority union, definition 1:

An f-structure f /g is the priority union of f with g (or “f given g”), if f /g
is the set of pairs )a, v* such that v is equal to the value of the attribute a in
f if it exists, otherwise the value of a in g.

Intuitively, the priority union of two f-structures contains all the structure that each
f-structure has, with the f-structure on the left of the slash “winning” if there is a
conflict. For example, in (140) f /g is the priority union of f with g:

(140) f /g is the priority union of f with g:

 s 
 o1
f /g 


c 
t
   
s s
f o1 g o2
c t

The priority union f /g has all the structure in f as well as all the structure in g that
does not conflict with f .

15 The forward slash notation is used both for priority union of f-structures (f /g) and as the “Ignore”
operator in c-structure rules (§6.1.1).
244 syntactic relations and syntactic constraints

Kaplan’s original definition of priority union, given in (139), was intended as a

proposal for the analysis of elliptical constructions. For example, we might assume
the following incomplete f-structures for a coordinate sentence with gapping:

(141) David saw Chris, and Matty Ken.

   

 ’  

   
g

 f



 

 

An analysis of gapping that appeals to priority union might propose that the final
f-structure f for the second conjunct Matty Ken is obtained by taking the priority
union f /g: in effect, the f-structure for Matty would replace the f-structure for
David, and similarly for the f-structures for Chris and Ken:

(142) f /g is the priority union of f with g:

 
’
 
f /g 




 
  ’
 
f  g




Priority union produces for the second conjunct an f-structure like the one that
would be associated with a sentence like Matty saw Ken.
Kaplan (1987) purposely formulated the definition of priority union to refer only
to the top-level attributes of f and g; the definition given in (139) is not recursive.
However, later work (for instance, Brun 1996b) assumes a recursive definition for
priority union like the following:

(143) Priority union, definition 2:

An f-structure f /g is the priority union of f with g (or “f given g”) if and
only if:
f is atomic, and f /g = f ; or,
f and g are sets, and f /g = f ∪ g; or,
f is an f-structure, and:
if )a, v1 * ∈ f and )a, v2 * ∈ g, then )a, v1 /v2 * ∈ f /g.
if )a, v1 * ∈ f and there is no pair )a, v2 * ∈ g, then )a, v1 * ∈ f /g.
if )a, v2 * ∈ g and there is no pair )a, v1 * ∈ f , then )a, v2 * ∈ f /g.

Future work will show which of these definitions is the most useful one.
 c-structure/f-structure constraints 245

6.10 C-structure/F-structure constraints

6.10.1 C-structure wellformedness: nonbranching dominance
Constituent structure representations are governed by a constraint on valid deriva-
tions originally proposed by Kaplan and Bresnan (1982):
(144) Nonbranching Dominance Constraint, preliminary version:
A c-structure derivation is valid if and only if no category appears twice in
a nonbranching dominance chain.

Intuitively, this requirement prevents a sentence from having an infinite number

of c-structures by preventing a c-structure node from dominating another node
with the same label in a nonbranching chain:
(145) Disallowed:
XP

..
.

XP

A tree like this one is not permitted: an XP cannot dominate another XP without
also dominating some other material as well. If this were permitted, there could be
a chain of XPs of unbounded length dominating any XP, giving rise to an infinite
number of possible constituent structure trees for that XP:
(146) Disallowed:
XP

..
.

XP

..
.

XP

Of course, a phrase may dominate a phrase of the same type if other material is
present as well:

(147) Permitted:

VP

V!

V VP
want
to go
246 syntactic relations and syntactic constraints

This constraint is also discussed by Pereira and Warren (1983), who refer to the
constraint as “off-line parsability”; this is because their formulation depends on
the application of the Nonbranching Dominance Constraint “off-line,” after the
parsing algorithm has been applied to derive a set of trees for a string. Johnson
(1988) also provides a definition of off-line parsability that is very similar to the
definition in (144).
In subsequent work, Kaplan and Maxwell (1996) revise the Nonbranching
Dominance Constraint to allow nonbranching dominance chains with nodes of
the same category if the two nodes have different functional annotations. Under
this view, the following dominance chain is ill-formed:

(148) Disallowed, revised Nonbranching Dominance Constraint:

XP
↑=↓

..
.

XP
↑=↓

However, the following configuration is permitted:

(149) Permitted, revised Nonbranching Dominance Constraint:
XP
↑=↓

..
.

XP
(↑ gf) = ↓
Under these assumptions, the final definition of the Nonbranching Dominance
Constraint is:
(150) Nonbranching Dominance Constraint, final version:
A c-structure derivation is valid if and only if there are no categories
with the same functional annotation appearing twice in a nonbranching
dominance chain.

Kaplan and Bresnan (1982) show that the Nonbranching Dominance Constraint
is important in proving that the membership problem for lexical functional gram-
mars is decidable—that it is always possible to determine whether a string is
acceptable according to a given grammar:
 c-structure/f-structure constraints 247

(151) Decidability Theorem (Kaplan and Bresnan 1982: (181)):

For any lexical functional grammar G and for any string s, it is decidable
whether s belongs to the language of G.

6.10.2 Category-function correlations

Certain lexical categories may tend to be associated only with certain grammatical
functions or requirements. For example, Bresnan and Moshi (1990) propose that
in general, only verbs and prepositions can subcategorize for the obj function (for
similar claims, see Chomsky 1970, Bresnan 1976, Jackendoff 1977, and Bresnan
and Kanerva 1989). However, exceptions to this tendency have often been noted:
for instance, Maling (1983), Mittendorf and Sadler (2008), Al Sharifi and Sadler
(2009), Vincent and Börjars (2010a), and Raza and Ahmed (2011) provide analyses
of adjectives subcategorizing for objects; Lowe (2013, 2014c) discusses transitive
nouns and adjectives in Avestan, and Lowe (2017b) presents a detailed analysis of
transitive nouns and adjectives in Sanskrit.
There are also common correlations between phrase structure positions and
functional annotations associated with those positions. For example, we have seen
that in English the specifier of IP is associated with the grammatical function subj.
English also allows sentential subjects—that is, subjects with the phrase structure
category CP; however, Bresnan (1994) presents evidence that the categories CP and
PP cannot appear in the specifier of IP position, the canonical position for subjects
(see also Bresnan et al. 2016): for instance, auxiliary inversion with a CP is not
possible (Koster 1978). Of course, as Bresnan shows, this does not prevent phrases
of those categories from bearing the subj function, even though they do not appear
in the position in which nominal subjects are generally found.

6.10.3 Inverse correspondences and the CAT predicate

In our discussion of subcategorization in §2.3, we noted that LFG defines sub-
categorization requirements in functional terms: predicates subcategorize for a
particular set of grammatical functions rather than for phrasal categories or
configurations. In many instances, as shown by Maling (1983), what appears to
be evidence for selection for a particular phrase structure category is often better
explained in semantic terms. In some cases, however, constraints on syntactic
category do seem to be at issue.
Some predicates are exceptional in that they impose a categorical requirement
on their arguments, restricting the constituent structure category of the argument
to be only a subset of the categories that may be associated with a particular
grammatical function. As pointed out by Pollard and Sag (1994), wax is a relatively
uncommon verb which requires an adjective complement, but is used mostly with
a small set of adjectives like poetical and lyrical:
248 syntactic relations and syntactic constraints

(152) a. Kim waxed poetical.

b. ⃰Kim waxed a success.
c. ⃰ Kim waxed sent more and more leaflets.
d. ⃰Kim waxed doing all the work.
e. ⃰ Kim waxed to like anchovies.

Sag et al. (1985) claim that the verb become requires either an AP or NP comple-
ment, and cannot appear with a PP or VP complement:

(153) a. Pat has become a Republican.

b. Gerry became quite conservative.
c. ⃰Connie has become of the opinion that we should get out.
d. ⃰Tracy became awarded a prize.
e. ⃰Chris will become talking to colleagues.

In a theory where constituent structure information is available as readily as

functional information in defining subcategorization requirements, the scarcity
of such verbs is surprising. Our theory of subcategorization allows for these
exceptional cases of categorical subcategorization, while reflecting the fact that in
the normal case functional information is all that is relevant.
The c-structure/f-structure correspondence for an example like Kim waxed
poetical is:

(154) Kim waxed poetical.

IP  
’
NP I
 
 
 
 
N VP  ’ 
 
 
Kim V

V AP

waxed A

poetical

The 𝜙 function relating c-structure nodes to their f-structures is indicated by

arrows. The inverse of this function, the 𝜙−1 relation, is indicated by arrows
pointing in the opposite direction in (155):
 c-structure/f-structure constraints 249

(155) Kim waxed poetical.

IP
’
NP I

N VP ’
p
Kim V

V AP
−1
waxed A

poetical

For each f-structure f , the inverse correspondence relation 𝜙−1 (f ) gives the
c-structure nodes that are associated with that f-structure; the relation between
f-structures and their corresponding c-structure nodes is therefore not a function,
because more than one c-structure node can correspond to a single f-structure.
For the f-structure labeled p in (155), 𝜙−1 (p) yields the nodes labeled AP and A.
Example (156) gives the c-structure and f-structure for the ill-formed sentence
Kim waxed a success:

(156) ⃰Kim waxed a success.

IP  
’
NP I  
 
N VP  
 
 ’ 
 
Kim V  s 
V NP

waxed D N −1

a N
success

For this example, the f-structure labeled s is associated via the inverse 𝜙
correspondence with three nodes, those labeled NP, N! , and N.
We can now state the categorical requirement imposed on the complement of
the verb wax by using the predicate CAT (Kaplan and Maxwell 1996; Crouch et al.
2008), defined in terms of the inverse 𝜙 correspondence and the λ function from
c-structure nodes to category labels (see §5.1.2). The CAT predicate allows the
250 syntactic relations and syntactic constraints

statement of constraints on the category of the c-structure nodes corresponding to

an f-structure. Formally, CAT is defined in the following way:16, 17

(157) Definition of CAT:

CAT( f , C) if and only if ∃n ∈ 𝜙−1 (f ) : λ(n) ∈ C.
CAT( f , C) is true if and only if there is some node n that corresponds to
f via the inverse 𝜙 correspondence (𝜙−1 ) whose label (λ) is in the set of
categories C.

According to this definition, the relation CAT( f , C) holds between an f-structure

f and a set of category labels C if the label of one of the nodes n in the set of
nodes corresponding to f is a member of C. In the wellformed example (155), the
following is true, since one of the nodes corresponding to f has the label AP:

(158) CAT( f , {AP})

Thus, we posit the following category requirement, lexically associated with the
verb wax:

(159) wax: CAT((↑ xcomp), {AP})

This requirement ensures that one of the c-structure nodes associated with the
xcomp has the category AP, as required.

6.10.4 Empty node rules

In her analysis of Warlpiri, Simpson (1991) discusses gaps in morphological
paradigms, showing that the Warlpiri aux does not appear in what she calls the
“null perfect aspect”:

(160) Japanangka-rlu ∅ pantu-rnu marlu

Japanangka-erg prf spear-pst kangaroo
‘Japanangka speared the kangaroo.’

According to the generalization that Simpson formulates, there are two possibilities
for phrasal expansion of I! . In (161), with a “non-null” auxiliary, the first daughter
of I! is I:

16 This definition of CAT conforms to the definition given by Kaplan and Maxwell (1996) and
Crouch et al. (2008), but differs from the definition in Dalrymple (2001).
17 The symbol ∃ is an existential quantifier: an expression like ∃n.... asserts that some individual n
exists that meets the description represented by the following formula. The first part of the expression,
∃n ∈ 𝜙−1 (f ), directs us to consider the nodes n which are related to f via the inverse 𝜙 correspondence,
and λ(n) ∈ C requires the category label λ(n) of at least one of those nodes to be in the set of node
labels C.
 c-structure/f-structure constraints 251

(161) wawirri ka ngarrka-ngku panti-rni

kangaroo prs man-erg spear-npst
‘The man is spearing the kangaroo.’
 
’
IP  
 
NP I  
 
 
N I S

wawirri ka NP V
kangaroo
N panti-rni

ngarrka-ngku

In (162), no auxiliary appears, and the sentence is interpreted as having perfect

aspect:

(162) Japanangka-rlu ∅ pantu-rnu marlu

Japanangka-erg prf spear-pst kangaroo
‘Japanangka speared the kangaroo.’

IP

NP I  
’
N S  
 
 
Japanangka-rlu V NP  
 
 
pantu-rnu N

marlu
kangaroo

In Warlpiri, then, the absence of an auxiliary requires the presence of an aspect

attribute in the f-structure with the value prf.
The rule in (163) expresses the possibilities Simpson outlines for the phrase
structure expansion of the category I! in Warlpiri:

(163) I! −→ { I | 𝜖 } S
↑=↓ (↑ aspect) = prf ↑=↓

In this rule, the symbol 𝜖 corresponds to the empty string and represents the
absence of a phrase structure constituent. Importantly, the rule does not license
252 syntactic relations and syntactic constraints

the presence of an empty category or node in the c-structure tree. Instead, it simply
constitutes an instruction to introduce some functional constraints in the absence
of some overt word or phrase. No empty node is introduced into the tree, as (162)
shows.
The rule in (163) contains a disjunction: in one case, an auxiliary appears in I,
while in the other case no auxiliary appears, and the sentence is interpreted as
having perfect aspect. In fact, the rule in (163) is exactly equivalent to:

(164) I! −→ { I S | S }
↑=↓ ↑=↓ ↑=↓
(↑ aspect) = prf

However, although the rule in (164) also represents the two possible phrase
structure expansions of I! , it is more cumbersome and fails to express Simp-
son’s generalization. By using the symbol 𝜖, the generalization can be expressed
concisely.

6.11 Linear relations and the string

6.11.1 String wellformedness conditions
Recall from §3.5 that, as proposed by Kaplan (1987, 1995), constituent structure is
projected via the function π from a string of words which constitute the terminal
nodes of the c-structure tree. Strings are governed by string wellformedness
conditions, or string constraints (Lowe and Mycock 2014), which define the valid
strings for a language; they are to strings what phrase-structure rules are to c-
structure. Only strings which conform to all of the relevant string constraints
for a language can participate in full, wellformed grammatical analyses. Here, we
briefly discuss such string wellformedness conditions and how they are formally
expressed. In Chapter 11, we will examine the string and its two aspects, the s-string
or string of syntactic units relevant for c-structure, and the p-string, the string of
prosodic units relevant for prosodic structure.
The string constraints that are relevant in a language are formally stated as a
regular expression (see §6.1) encoding universal and language-specific conditions
which all wellformed strings in the language must obey. Since more than one type
of string constraint may be relevant in a language, the regular expression encoding
the combination of all of the string constraints is likely to be fairly complex. It
is, then, preferable to define it by reference to a set of basic conditions, each of
which reflects some particular generalization about string wellformedness. Taking
advantage of the fact that regular languages are closed under the operations of
union, intersection, and complementation, we can make use of these operations to
combine a set of basic conditions into a single regular expression which imposes
each of the separate conditions.
The most useful operation for combining basic constraints is intersection, since
the intersection of two regular languages contains the strings that the two regular
 linear relations and the string 253

languages have in common: in other words, the use of intersection amounts to

imposing both of the constraints simultaneously. For example, we can impose two
basic conditions encoded by two regular expressions R1 and R2 by intersection
(&); the resulting regular expression R1&R2 picks out the set of strings that obey
the conditions imposed by both R1 and R2.18 We provide examples of basic string
constraints and their combinations in Part II of the book.
We define separate, basic string constraints as components of the string well-
formedness requirements for a language: for instance, in §11.6 we impose a con-
dition on the s-string requiring each component of a focused string to be marked
with the label “DFFoc.” Strings that do not obey this requirement are not well-
formed, and are rejected. Additional universal and language-specific wellformed-
ness conditions may also be imposed by combining them with the other string
constraints for a language; the complete collection of wellformedness conditions
are expressed as separate, basic regular expressions, and the string wellformedness
conditions for a language are defined by a combination of regular expressions
encoding each of the basic conditions.

6.11.2 String precedence

Linear order relations in the string may also play a role in defining syntactic con-
straints. For example, Asudeh (2009) provides an analysis of the Complementizer-
Adjacent Extraction (CAE) constraint on question formation in modern English,
originally observed by Perlmutter (1971), which disallows extraction of a subject
in most cases if a complementizer is present (see §17.1.3.4):

(165) Who do you think ( ⃰that) saw Kim?

Nonsubject constituent question formation is possible whether or not a comple-

mentizer is present:

(166) Who do you think (that) Kim saw ?

Extraction of a subject is also possible if the complementizer is followed by an

adverb (Bresnan 1977); this is sometimes referred to as the Adverb Effect:

(167) Who did you say that, just a minute ago, sneezed?

The pattern is reversed for relative clauses, where the complementizer is required
in subject relativization in most varieties of contemporary English (Bresnan 1977):

(168) a. This is the person [that sneezed].

b.⃰This is the person [∅ sneezed].

18 As discussed in §6.1, the symbol & that is generally used for the intersection of two regular
languages is different from the symbol ∩ used for the intersection of two sets.
254 syntactic relations and syntactic constraints

Asudeh (2009) discusses these and other patterns involving subject and nonsubject
extraction in question formation and relative clause formation in detail; for the
full account, see Asudeh (2009). He proposes the following constraint, lexically
associated with the complementizer:

(169) CAE Constraint, Asudeh (2009), informal version:

It is not the case that the string element that immediately follows the
complementizer maps to an f-structure that contains a subject that is both
phonologically realized and is the head of an unbounded dependency.

In order to provide an explicit formal treatment of this constraint, Asudeh defines

a function realized, which is true of an f-structure if it corresponds to one or
more c-structure nodes. This function is defined in terms of the inverse relation
𝜙−1 from f-structures to c-structure nodes, discussed in §6.10.3, and holds of an
f-structure f if there is some c-structure node that corresponds to f :

(170) realized( f ) holds if and only if 𝜙−1 (f ) &= ∅.

Asudeh also defines a function N (for Next), which relates a word in a string to
the next word in the string. N is defined for all words in the string except for
the last word, since there is no word immediately following the final word in the
string. The inverse of this function, N −1 , relates a word to the previous word in the
string.

(171) N(word1) = word2 if and only if word2 immediately follows word1 in the
string.
N −1 (word2) = word1 if and only if word2 immediately precedes word1 in
the string.

Given this definition of N for the next word in a sentence and the π function from
words to terminal nodes in the c-structure, we can use the familiar 𝜙 function
to refer to the f-structure corresponding to the c-structure node dominating the
next word in the string. We follow Asudeh in using the expression 5 for this
f-structure.19

(172) For a word w in the string, 5 is the f-structure corresponding to the

preterminal node dominating the next word in the string:
5 ≡ 𝜙(π(N(w)))

19 Our definition differs slightly from the one provided by Asudeh (2009) because we assume that
words in the string are related to nodes in the c-structure via the function π; see §3.5 and §11.4.1 for
more discussion of our architectural assumptions.
 linear relations and the string 255

We can define the expression ≺ in a similar way, to refer to the f-structure

corresponding to the c-structure node dominating the preceding word in the string.
The ≺ expression is not used in Asudeh’s analysis of the CAE constraint.

(173) For a word w in the string, ≺ is the f-structure corresponding to the

preterminal node dominating the previous word in the string:
≺ ≡ 𝜙(π(N −1 (w)))

Note that 5 is not defined for the last word in the string (which has no word
following it), and ≺ is not defined for the first word in the string (which has no
word preceding it).
Either 5 or ≺ may appear in a lexical entry. Since these expressions are defined
over words in a string, they may not appear in a c-structure annotation, since
their definitions rely on the relation N which holds between words in the string
rather than the c-structure precedence relation discussed in §6.11.3. Consider the
following configuration:
6 7
(174) g[ ]
f
h[ ]
C1
C2 C3 𝜙
word1
word2 word3

In (174), the terminal string is word1 word2 word3. According to the definition of
the N function, the following statements define the Next function for this string:

(175) N(word1) = word2

N(word2) = word3

The π function from words of the string to preterminal nodes is:

(176) π(word1) is the node labeled C1

π(word2) is the node labeled C2
π(word3) is the node labeled C3

We can restate the second and third line of (176), using the N function to refer to
the next word, as:

(177) π(N(word1)) is the node labeled C2

π(N(word2)) is the node labeled C3

Finally, we can refer to the f-structures corresponding to each of these nodes:

256 syntactic relations and syntactic constraints

(178) 𝜙(π(N(word1))) = g
𝜙(π(N(word2))) = h

Thus, for word1, 5 would refer to g, and for word2, 5 would refer to h.
Still following the analysis of Asudeh (2009), we can now state the CAE con-
straint in formal terms:

(179) CAE Constraint, Asudeh (2009), formal version:

¬[realized(5 subj) ∧ (dis ∈ (5 subj))]

This constraint appears in the lexical entry of the complementizer that, and rules
out the possibility that the f-structure of the word following that has a subject
(5 subj) that (i) has a c-structure realization (realized(5 subj)) and (ii) is a mem-
ber of the dis set, establishing a long-distance dependency (dis ∈ (5 subj)).20 This
constraint is satisfied in one of three circumstances. First, if the f-structure of the
word following the complementizer that has no subj, then (5 subj) does not exist
and the CAE Constraint is satisfied. Second, if the f-structure of the following word
has a subject, but it is not realized, the statement realized(5 subj) is false and the
CAE Constraint is satisfied (this is the case for relative clauses with no relative
pronoun, such as This is the person that/⃰∅ sneezed). Third, if the f-structure of the
following word has a subject that is realized but does not appear in the dis set, the
f-structure (dis ∈ (5 subj)) does not exist, and the CAE Constraint is satisfied;
this covers the case of nonsubject relatives such as Who do you think [that Kim
saw]? as well as declarative complements with no dis, such as I think [that Kim saw
David]). The analysis makes crucial use of the N relation, defined over elements of
the string.

6.11.3 C-structure precedence

The notion of c-precedence for c-structure nodes is the intuitively familiar notion
of linear precedence, definable in the following terms (see also Partee et al. 1993:
§16.3.2):

(180) A c-structure node n1 c-precedes a node n2 if and only if n1 does not

dominate n2 , n2 does not dominate n1 , and all nodes that n1 dominates
precede all nodes that n2 dominates.

6.11.4 Functional precedence

Functional precedence is a relation between two f-structures based on the
c-structure precedence relation holding between the c-structure nodes corre-
sponding to the two f-structures. Although it is based on the c-structural relation of

20 The expression (dis ∈ (5 subj)) is an existential constraint requiring that the f-structure (5 subj)
is a member (∈) of a dis set. In this expression, the set-membership symbol ∈ is used as an attribute
selecting a member of the dis set (§6.3.1).
 linear relations and the string 257

precedence, it is different in interesting ways; differences between the two relations

show up most clearly when an f-structure is related to discontinuous c-structure
elements, and when an f-structure does not correspond to any c-structure nodes.
Kameyama (1989) presents an analysis of Japanese pronominals that accounts
for the distribution of overt pronominals as well as “null” pronouns, pronouns that
appear at f-structure but not c-structure (see also Kameyama 1985). As Kameyama
shows, overt pronouns must follow their antecedents. In the unacceptable example
in (181a), the pronoun kare precedes its antecedent Taroo, while (181b), in which
Taroo precedes the pronoun, is acceptable:

(181) a. ??kare-no imooto-o Taroo-ga sewasiteiru (koto…)

his-gen sister-acc Taro-nom be.taking.care.of that
‘…(that) Taroi was taking care of hisi sister’
b. Taroo-ga kare-no imooto-o sewasiteiru (koto…)
Taro-nom his-gen sister-acc be.taking.care.of that

In contrast, there are no restrictions on the relation between the null pronominal
and the definite antecedent. Whether imooto ‘sister’ precedes or follows Taroo, we
may interpret imooto with a null possessive pronoun anteceded by Taroo:

(182) a. imooto-o Taroo-ga sewasiteiru (koto…)

[∅’s] sister-acc Taro-nom be.taking.care.of that
‘…(that) Taroi was taking care of hisi sister’
b. Taroo-ga imooto-o sewasiteiru (koto…)
Taro-nom [∅’s] sister-acc be.taking.care.of that

Simplifying Kameyama’s analysis somewhat, these and other examples show that
the antecedent of an overt pronoun must precede the pronoun, while this con-
straint does not hold for null pronouns. The facts about pronominal binding in
Japanese can be given a uniform explanation in terms of f-precedence:

(183) The antecedent of a pronoun must f-precede the pronoun.

This generalization about anaphoric binding in Japanese holds under the following
definition of f-precedence (Kaplan and Zaenen 1989b; Kameyama 1989):

(184) F-precedence, definition 1 (Kaplan and Zaenen 1989b):

f f-precedes g (f <f g) if and only if for all n1 ∈ 𝜙−1 (f ) and for all n2 ∈
𝜙−1 (g), n1 c-precedes n2 .

This definition appeals to the inverse relation 𝜙−1 , defined in §6.10.3, which
associates f-structures with the c-structure nodes they correspond to. It also relies
on the definition of c-precedence given in (180) in §6.11.3.
This definition of f-precedence states that an f-structure f f-precedes an f-
structure g if and only if all of the nodes corresponding to f c-precede all of the
258 syntactic relations and syntactic constraints

nodes corresponding to g in the c-structure. In the unacceptable example in (185),

the f-structure g of the possessive pronoun kare-no f-precedes the f-structure f
of the antecedent Taroo, since all of the nodes corresponding to g (the D node)
precede the nodes corresponding to f (the NP and N nodes):

(185) ??kare-no imooto-o Taroo-ga sewasiteiru (koto…)

his-gen sister-acc Taro-nom be.taking.care.of that
‘…(that) Taroi was taking care of hisi sister’
 
’
 
 f 
  
 
 
   
  
   
   
   

  g  

   
 

NP NP V

D N N sewasiteiru
be.taking.care.of
kare-no imooto-o Taroo-ga

What is the situation with null pronominals, pronouns that do not appear at
c-structure? Consider the c-structure and f-structure for (186), which is like (185)
except for having a null pronoun (realized only at f-structure) in place of the overt
possessive pronoun kare-no:

(186) imooto-o Taroo-ga sewasiteiru (koto…)

[∅’s] sister-acc Taro-nom be.taking.care.of that
‘…(that) Taroi was taking care of ∅i sister’
 
’
 
 f 
 
 
 
 
g
S

NP NP V

N N sewasiteiru
be.taking.care.of
imooto-o Taroo-ga
 linear relations and the string 259

Crucially, the f-structure of the null pronoun does not correspond to any
c-structure node. According to the definition of f-precedence given in (184),
null elements vacuously f-precede and are f-preceded by all other elements in
the sentence; in particular, null pronouns vacuously satisfy the constraint of
being preceded by their antecedents. Thus, a uniform statement of antecedent
requirements for Japanese null and overt pronominals, together with the definition
of f-precedence given in (184), makes the correct predictions.
The definition of f-precedence in (184) is the one originally proposed by Bresnan
(1984) and used by Kameyama (1989) and by Kaplan and Zaenen (1989b) in their
analysis of word order in Dutch and German. A different definition is proposed by
Bresnan et al. (2016) in their analysis of weak crossover:

(187) F-precedence, definition 2 (Bresnan et al. 2016: 213):

f f-precedes g if and only if the rightmost node in 𝜙−1 (f ) precedes the
rightmost node in 𝜙−1 (g).

This definition gives a different result for f-structures that do not correspond
to any c-structure nodes, since such f-structures do not have a corresponding
rightmost node. Under definition 2 in (187), f-structures that do not correspond
to any c-structure node do not f-precede any other f-structure, and they are not
f-preceded by any other f-structure.

With the formal tools that have now been introduced, we are ready to begin
a full-scale excursion into new linguistic realms. In the following chapters, we
will explore other levels of nonsyntactic linguistic structure and how they are
represented.
 7
Beyond c-structure and
f-structure: Linguistic
representations and relations

In Part I, we provided evidence for two levels of syntactic structure: an abstract

representation of functional syntactic organization, the f-structure, and a separate
level of phrasal organization, the c-structure. These linguistic structures are repre-
sented in different ways: by means of a phrase structure tree for c-structure, and by
an attribute-value structure for f-structure. Besides these two syntactic structures,
LFG research has explored other linguistic levels and their representations. As
with the relation between c-structure and f-structure, it is possible to define
and constrain these additional linguistic levels by specifying properties of their
structure and their relation to other levels of structure. In considering the overall
architecture of our theory of grammar, we must address the following questions:
How do we determine when it is necessary to postulate the existence of a new
linguistic level, distinct from those already assumed within the theory? How do we
determine the best representation for a proposed level of structure? And how can
constraints within a level, or constraints that hold across levels, be defined within
this architecture? Our theory must respect modularity and the independence of
different levels of linguistic structure, while stating the relations among levels in a
clear and easily understandable way.

7.1 Linguistic structures and modules

7.1.1 Defining a separate structure
In the analysis of new and unfamiliar linguistic phenomena, it is sometimes
tempting to assume that they constitute a new and independent linguistic system,
governed by its own rules, rather than simply hitherto unexplored manifestations
of familiar phenomena and familiar rules. Whether or not to propose a distinct
representation for a new type of linguistic information must be considered care-
fully. Sadock (1991: 214–15) suggests that:

Any postulated component should be of a kind that one might want in a full description of
the basic facts of language, regardless of the way in which the modules are related to one
another.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
264 beyond c-structure and f-structure

It is difficult to provide strict criteria for when it is appropriate to postulate the

existence of a separate level. Sadock (1991: 215) rightly cautions against assuming
“levels that are entirely abstract,” encoding relations like coindexing or traces;
of course, any new level that is postulated should deal with a coherent set of
linguistic phenomena that are found in multiple languages and are demonstrably
and cohesively related to one another.
For example, there has been some disagreement about the status of f-structure
as a grammatical level of structure, and what information should be represented
there. As we discuss in §12.2.2, Frank and Zaenen (2002) make the strong claim
that f-structure should not contain features that vary from language to language,
such as case and agreement features; instead, it should contain all and only the
features that are relevant for semantic interpretation. On their view, features that
are not semantically relevant should be represented in a separate morphosyntactic
structure; indeed, Frank and Zaenen advocate thinking of f-structure as an under-
specified semantic representation, analogous to the unscoped, purely semantic rep-
resentation Quasi-Logical Form in the Core Language Engine language processing
system (Alshawi 1992). Falk (2003) provides convincing argumentation against
this approach, however: on Falk’s view, f-structure is not a bleached semantic
structure but a syntactic structure, and it is the appropriate place to represent
morphosyntactic features that play a role in the grammar of a language (see
Dyvik 1999 for an articulate defense of a similar view). In Chapter 12, we provide
further discussion of approaches which assume a separate level of morphosyntactic
structure, but we do not assume this level of structure in this book.

7.1.2 Defining a new relation on an existing structure

With advances in our understanding of the overall grammatical architecture and
the interactions among levels, phenomena previously analyzed in terms of one
level of linguistic structure may be open to analysis in terms of a new relation
defined at another level. Compare, for example, the original treatment of long-
distance dependencies at c-structure, by means of the double-arrow annotation
relating two c-structure positions (Kaplan and Bresnan 1982), with the currently
accepted analysis of long-distance dependencies as relating f-structure roles rather
than c-structure positions by means of functional uncertainty (Kaplan et al. 1987).1
In some cases, constraints on grammatical phenomena may involve more than
one kind of linguistic information; such constraints must be stated by reference to
more than one aspect of linguistic structure. This in itself does not constitute moti-
vation for introducing a new grammatical module incorporating information from
other levels. Indeed, many phenomena that we have already examined are of this
nature. For example, the correlation between phrase structure positions and gram-
matical functions involves reference both to c-structure and to f-structure, and is
captured by annotations on c-structure rules making reference to f-structure roles.

1 For more on the history of the treatment of long-distance dependencies in LFG, see Dalrymple
et al. (1995d); we discuss long-distance dependencies in detail in Chapter 17.
 modularity 265

A fruitful strategy for the formal treatment of linguistic phenomena involving

relations and entities at more than one level of representation is to define a
new relation on an existing structure in terms of relations that are native to
another structure. An example of this is functional precedence, a relation between
f-structures that is defined in terms of the precedence relation that holds
between c-structure nodes (§6.11.4). Defining a precedence relation derivative of
c-structure properties on the f-structure, a level at which precedence is not native,
gives rise to very interesting results that are not available when the standard
c-structural precedence relation is considered. In particular, f-structures that
do not correspond to any c-structure nodes also take part in f-precedence.
Additionally, nonadjacent constituents in different parts of the phrase structure
tree may correspond to the same f-structure, and thus the c-structure nodes
corresponding to a single f-structure may be interleaved with the nodes of another;
in this case, it may be that no f-precedence relation holds between the two
f-structures, depending on the particular definition of f-precedence that is chosen.
Thus, constraining precedence relations in terms of f-precedence instead of
c-structure precedence gives rise to a richer set of predictions for how the linear
order of arguments can be constrained.
The principle of modularity is a central tenet of LFG. We explore this in the next
section.

7.2 Modularity
Analyzing an utterance requires the linguist to analyze different types of lin-
guistic structure and information; a modular approach requires different aspects
of linguistic structure to be represented as distinct, independent components
of the grammar. These separate components or grammatical modules each have
their own primitives and organizing principles, and therefore their own internal
structure and formal representation. For instance, a concept such as “noun phrase”
is relevant at the level of c-structure, but not f-structure; similarly, syllables are
important in a phonological analysis, but a semantic analysis of an utterance is
effectively blind to them.
A modular approach ensures that different types of linguistic information
are not conflated. This enables generalizations, whether universals or language-
particular constraints, to be stated for and apply only to the level of representation
for which they are relevant. Moreover, separating parts of the grammar in this way
minimizes the possibility of adverse effects should it become necessary to modify
the analysis of one component of the grammar: advances in the analysis of one
aspect of linguistic structure can be incorporated into the model without the need
for a wholesale revision of the grammatical architecture. This ensures that LFG’s
approach to the architecture of the grammar is stable yet flexible.
While the different modules of the grammar are separate in LFG, they exist in
parallel. This does not entail a one-to-one correspondence between the subparts of
structures at any of the levels, nor does it mean that the information encapsulated
within one module (for example, the precedence relation between c-structure
266 beyond c-structure and f-structure

nodes) is unavailable to another. Indeed, distinct structural levels are mutually

constraining, and analyses can be formulated which capture interactions among
syntax, semantics, prosody, information structure, and other levels. Universal and
language-particular generalizations about the interface between specific modules
of the grammar can also therefore be stated without requiring that such general-
izations apply to all modules or inter-modular relations.

7.3 The projection architecture

Kaplan (1987) originally proposed the projection architecture to define piecewise
correspondences between grammatical modules (see also Halvorsen and Kaplan
1988). The different modules are also referred to as projections, and the functions
relating the structures are referred to as correspondence functions.
We are already familiar with the basic concepts of the projection architecture
from our study of the relation between c-structure and f-structure. Section 5.3.1
presented a notation for referring to the current c-structure node and its mother:

(1) the current c-structure node (“self ”): ⃰

the immediately dominating node (“mother”): ˆ⃰

The correspondence function 𝜙 relates nodes of the c-structure to their

corresponding f-structures, as described in Chapter 4. To recap briefly, 𝜙(⃰) is
the f-structure corresponding to the node under consideration, and 𝜙(ˆ⃰) is the
f-structure corresponding to its mother node. A standard abbreviation for 𝜙(ˆ⃰) is
↑ ; a standard abbreviation for 𝜙(⃰) is ↓. Alternative notations for correspondence
functions have also been proposed; in some work, correspondence functions are
represented as a subscript, with ⃰𝜙 used as an alternative notation for 𝜙(⃰). This
notation is particularly common in discussions of structures projected from the
f-structure, and we will adopt this variant notation in Chapter 8 in our discussion
of semantic structure, and in subsequent chapters. In this chapter, for explicitness,
we represent correspondence functions in the mathematically more standard way:
for example, 𝜙(⃰) and 𝜙(ˆ⃰).
Just as the correspondence function 𝜙 relates c-structure nodes to f-structures,
other functions can be defined to relate other aspects of linguistic organization.
We can define a relation between modules by specifying a function relating parts
of any linguistic structure to parts of another. For example, a function 𝜎 relating
f-structures to semantic structures can be defined, as we will see below and
in Chapter 8. In this way, LFG models distinct aspects of linguistic structure,
including but not limited to syntactic structures, and the mappings between them.
There is a clear relation between LFG’s projection architecture and other formal
linguistic architectures. For example, Head-Driven Phrase Structure Grammar
(Pollard and Sag 1994; Sag et al. 2003) and related theories encode linguistic
information as a single structure, the sign, represented as an attribute-value struc-
ture like the f-structure. Subparts of the sign correspond to different aspects of
linguistic structure, and the various substructures are built up simultaneously.
 defining relations between structures 267

A linguistic structure that is represented as an attribute-value structure and defined

as a projection of the f-structure can be represented in an equivalent but less
revealing way as a subpart of the f-structure it is associated with: the function
defining the relation between the f-structure and the new projection can be
reinterpreted as an attribute of the f-structure, with the new structure as its
value. The significance of the projection architecture lies not in the additional
formal power that it brings, but in its expressiveness and modularity; it allows
for the relation between different linguistic components to be expressed while
also retaining the identity of these components as separate structures representing
different kinds of linguistic information. The resulting grammatical architecture
respects the principle of modularity.

7.4 Defining relations between structures

Kaplan (1995) notes that there are two ways in which relations between structures
can be defined: codescription and description by analysis. The formal difference
between these two methods of description has not yet been fully explored. Ron
Kaplan (p.c.) hypothesizes that description by analysis is the more powerful of the
two, but a complete formal analysis and proof awaits further research.

7.4.1 Description by analysis

Relations between structures can be defined by description by analysis, in which a
description of one structure is obtained by analysis of another structure. A number
of LFG proposals for semantic analysis and representation involve description by
analysis: for instance, Halvorsen (1983) defines a procedure for determining the
semantic structure for an utterance on the basis of properties of its f-structure.
Chapter 8 presents a theory of the syntax-semantics interface that differs from
Halvorsen’s proposals in a number of respects, and the theory presented there will
be used in semantic analyses of the phenomena treated in the remainder of the
book. Here we present Halvorsen’s theory of the syntax-semantics interface as a
particularly clear example of description by analysis.
Description by analysis involves the definition of properties of one structure
based on the properties of another. Informally, a rule using description by analysis
says something like: “whenever there is a structure with a certain property, it
corresponds to another structure of a certain type.”
Halvorsen presents the following partial table of correspondences between
semantic forms in the f-structure and meanings, represented as logical formulas:2

2 Halvorsen’s analysis has been considerably simplified for presentation here, leaving aside many
of the details of the meanings and rules that he proposes. Further, we have modified his analysis
by assuming that a word like every is treated as a generalized quantifier (Barwise and Cooper 1981)
specifying that the every relation holds between two properties R and S. We discuss quantification in
more detail in Chapter 8. The lambda operator λ is explained in §8.4.1.
268 beyond c-structure and f-structure

(2) Semantic form in f-structure: Meaning:

‘every’ λR.λS.every (R, S)
‘horse’ λx.horse (x)

Halvorsen also presents a set of rules for determining the meaning of an f-structure
based on its semantic forms. The following is a simplified version of his Rule III,
spec-pred configuration, which defines the meaning Mk of an f-structure fk :

(3) spec-pred configuration:

 spec v1 
 pred vn 
If fk is an f-structure of the form  , then
.. ..
. .
(Mk predicate) = λP.λQ.Q(P)
(Mk arg1) = M1 (semantic structure of the spec)
(Mk arg2) = M2 (semantic structure of the pred)

Let us see how this rule operates in the analysis of the f-structure proposed by
Halvorsen for the noun phrase every horse. For this simple example, we will leave
out most of the detail in the f-structure:

(4) F-structure for every horse according to Halvorsen (1983):

' (
spec ‘every’
f1
pred ‘horse’

We apply the rule in (3) in the following way. First, we inspect the f-structure in
(4), noting that it is of the form required by the rule in (3). Therefore, we introduce
a constraint on the structure M1 , the semantic structure corresponding to f1 : it
must contain the attribute predicate with value λP.λQ.Q(P). We also add the
additional constraint that M1 must contain the attribute arg1, and we determine
the value of that attribute by consulting the table in (2); in a similar way, we add
a constraint on the value of arg2. This yields the following set of constraints
on M1 :

(5) (M1 predicate) = λP.λQ.Q(P)

(M1 arg1) = λR.λS.every (R, S)
(M1 arg2) = λx.horse (x)

These equations describe the following semantic structure M1 for the f-structure
f1 :

(6)  
predicate λP.λQ.Q(P)
 λR.λS.every (R, S) 
M1  arg1 
arg2 λx.horse (x)
 defining relations between structures 269

We also need a set of rules for interpreting this structure: Halvorsen (1983) presents
a set of rules that produce logical formulas from these semantic structures. Among
them are rules like the following:

(7) predicate and argi

Apply the predicate to argn , where n is the polyadicity of the predicate.
Apply the result to argn−1 and so on, until n = 0.

Applying this rule to the semantic structure for every horse in (6) produces the
following result, which is just what is required for the phrase every horse:

(8) λP.λQ.[Q(P)](λx.horse (x))(λR.λS.every (R, S))

= λS.every (horse , S)

In sum, linguistic analyses formulated according to the description-by-analysis

paradigm operate by imposing constraints on one structure on the basis of an
inspection of one or more other structures.

7.4.2 Codescription
Grammatical structures can also be related by codescription, where multiple struc-
tures are simultaneously described; this is the most common way in which the
relations between grammatical modules are encoded and constrained. For exam-
ple, annotated c-structure rules simultaneously encode constraints on (at least) two
kinds of grammatical information: c-structure category and f-structure.
As discussed in §5.1.2, nodes of the c-structure tree are associated with two
different functions: the 𝜙 function relates c-structure nodes to their corresponding
f-structures, and the λ function relates nodes to their labels. We can represent
the configuration in (9a) in a more explicit but also more cumbersome way
as (9b):

(9) a. IP

NP I

b. IP •
λ
NP • •
I

The trees in (9) are licensed by the phrase structure rule in (10a), which is
equivalent to the rule in (10b) except that the λ labeling function is explicitly
shown, and 𝜙(ˆ⃰) and 𝜙(⃰) appear rather than the abbreviatory ↑ and ↓ arrows in
the former. The format in (10a) makes it clear that two different aspects of structure
are constrained by this rule. The λ labeling function constrains the phrase structure
270 beyond c-structure and f-structure

category of the mother and daughter nodes,3 and the 𝜙 function constrains the
f-structures that correspond to the mother and daughter nodes. In other words,
this rule simultaneously describes, or codescribes, two levels of structure: the 𝜙
projection from c-structure—that is, the f-structures 𝜙(ˆ⃰) and 𝜙(⃰), and the λ
projection from c-structure nodes defining the category of the node:

(10) a. • −→ • •
λ(⃰) = NP λ(ˆ⃰) = IP
(𝜙(ˆ⃰) subj) = 𝜙(⃰) λ(⃰) = I%
𝜙(ˆ⃰) = 𝜙(⃰)
b. IP −→ NP I%
(↑ subj) = ↓ ↑=↓

7.5 Defining interstructural constraints

It is often useful to refer to the relation between one representation and another:
to speak, for example, of the semantic structure that corresponds to a particular
c-structure node. In order to achieve this, we need a means of referring to the
relation between two structures related by functional projections.

7.5.1 Structural correspondence by composition

In Chapter 8, we will introduce semantic structure as a level of linguistic structure
directly related to f-structure; the correspondence function 𝜎 relates f-structures
to semantic structures in the following way:

(11) ) *
pred ‘yawn&subj'’ 𝜎 s:[]
𝜙 f
V subj [] s1 : [ ]
yawn

In the configuration depicted in (11), the correspondence function 𝜙 relates the

c-structure node labeled V to the f-structure labeled f . The 𝜎 function defines a
direct relation between f-structures and semantic structures: in (11), the semantic
structure corresponding to f is labeled s, and the semantic structure corresponding
to f ’s subject is s1. The following facts hold of the configuration in (11):4

(12) 𝜙(V) = f 𝜎(f ) = s

3 The constraint λ(ˆ⃰) = IP defining the category of the mother node could have been written on
either daughter node: here we have made the arbitrary decision to associate it with the second daughter.
4 As mentioned in §7.3, an alternative subscript notation f𝜎 , exactly equivalent to 𝜎(f ), is often used
for semantic structures.
 defining interstructural constraints 271

Given the functions 𝜙 and 𝜎, we can define a function between c-structure nodes
and semantic structures as the composition of the two correspondence functions,
𝜎 ◦ 𝜙.5 In the case at hand, we can apply the composition function 𝜎 ◦ 𝜙 to the
c-structure node V to get the semantic structure s:

(13) 𝜎 ◦ 𝜙(V) = 𝜎(𝜙(V)) = 𝜎(f ) = s

In this way, using codescription, we can define a function between c-structure

nodes and their corresponding semantic structures that is mediated by the
f-structure in terms of the function 𝜙 from c-structure nodes to f-structures
and the function 𝜎 from f-structures to semantic structures. More generally, we
can exploit the projection architecture to define relations between structures that
are not directly related via correspondence functions by defining new composite
functions to relate the two levels.
Subtle issues arise in the design of the optimal projection architecture for
linguistic description. In particular, care must be taken in arranging the various
linguistic structures properly, since distinctions that are relevant at one level but
collapsed at another cannot be reintroduced in further projections. Consider this
configuration:

(14) 𝜙
VP ' (
pred ‘yawn&subj'’
V
tense pst
yawned

At c-structure, the distinction between the nodes labeled VP and V is clearly repre-
sented: two different c-structure nodes are involved. At f-structure, the distinction
between the two nodes is collapsed, since the two nodes correspond to the same
f-structure. Thus, any level of linguistic representation that must make reference
to a distinction between the V and the VP node cannot be a projection of the
f-structure, since the distinction between VP and V is collapsed at that level and
cannot be reintroduced in projections from the f-structure.

7.5.2 Structural correspondence by inverse correspondence

This diagram depicts the familiar 𝜙 function from nodes of the c-structure to
f-structures:

(15)
IP

NP I

5 The composition of two functions is obtained by taking the result of applying one of the functions
to its argument and then applying the other function to that result (see Partee et al. 1993: 33). The
composition of two functions is also a function.
272 beyond c-structure and f-structure

In §6.10.3, we discussed inverse correspondences between structures. The inverse of

the 𝜙 function, 𝜙−1 , relates f-structures to the c-structure nodes that correspond
to them:
−1

(16) IP f s[ ]

NP I

In the configuration depicted in (16), the f-structure labeled f is related via the 𝜙−1
correspondence to two c-structure nodes, labeled IP and I% , and the subj f-structure
is related via the 𝜙−1 correspondence to the node labeled NP. The definition of the
CAT predicate (given in (157) of Chapter 6) is stated in terms of the inverse of the
𝜙 function, 𝜙−1 :

(17) Definition of CAT:

CAT(f , C) if and only if ∃n ∈ 𝜙−1 (f ) : λ(n) ∈ C.

The inverse correspondence is not in general a function, as noted in §6.10.3: there

are often several c-structure nodes that correspond to a particular f-structure. In
(16), for example, the IP and I% nodes of the c-structure tree are related to the same
f-structure f :

(18) λ(𝜙−1 (f )) = {IP, I% }

Thus, the projection architecture allows for the statement of complex relations
among linguistic structures, and for constraining these relations in appropriate
ways.

7.6 Representing linguistic structure and information

LFG shares with a number of linguistic theories the view that functional, config-
urational, and other linguistic structures reflect what Sadock (1991) calls “par-
allel organizational principles”: the various facets of linguistic organization are
copresent, and each aspect of linguistic structure is organized according to its
own set of rules and principles. LFG assumes a fine-grained relation between
representations in which subparts of one module are related to subparts of another
one; for instance, as the c-structure is related to the f-structure. Wellformedness is
defined and determined for each module: what holds for one module need not
necessarily hold for another, even if their representation is similar.
Other grammatical theories also resemble LFG in taking the view that represen-
tations of different aspects of linguistic structure need not be of the same formal
types. The formal architecture of Construction Grammar as described by Kay
(2002), in which the syntactic objects are trees whose nodes are associated with
feature structures, is fairly close to that of LFG. In Role and Reference Grammar
 representing linguistic structure and information 273

(Van Valin 2003), a semantic/logical structure resembling a formula of predicate

logic is linked to a layered, tree-like syntactic structure via a set of linking rules.
Some versions of Categorial Grammar also allow linguistic representations of
different formal types; Oehrle (1999) provides a view of LFG as labeled deduction
in categorial terms (Gabbay 1996), where the correspondences between different
structures are represented as relations between labels on formulas. The Parallel
Architecture theory of Culicover and Jackendoff (2005) and Jackendoff (2002,
2010) assumes an architecture in which phonological, syntactic, and semantic
information is organized into separate modules, each module consisting of a set
of tiers which may have different formal representations.
In adhering to the principle of modularity, LFG differs from frameworks which,
to a greater or lesser extent, seek to conflate different types of linguistic information
by using the same primitives, organizing principles, and formal representations to
model syntax and semantics (as in some versions of Chomskyan transformational
theories) or syntax and phonology/prosody (as in Head-Driven Phrase Structure
Grammar).
Does it matter how linguistic structure is represented and how the different
facets of structure are related? In fact, this issue is vitally important: the use of
inappropriate representations and relations makes it difficult to capture facts about
the linguistic structure underlying the representations, and can lead to incorrect
and obscured views of linguistic typology. In LFG, the choice of how to represent a
particular kind of information is made on the basis of how that kind of information
is best expressed.
For instance, dominance and precedence conditions and phrasal groupings
are clearly and succinctly represented in terms of a familiar phrase structure
tree, the c-structure. In the case of functional syntactic information, however, a
phrase structure tree is not the best way of perspicuously and unambiguously
representing this information, as trees carry with them certain presuppositions
about the information being represented. For instance, trees encode a linear order,
which does not make sense in the functional realm; further, nodes in a tree are
required to have a single mother, which, if used in f-structure, would imply that
a phrase can play only a single functional role. Instead, as originally proposed by
Kaplan (1975a,b), an attribute-value structure is a very good way of representing
functional information and is therefore the way in which functional structure is
represented in LFG. The functional structure does not contain information about
linear order relevant for the description of constituent structure but irrelevant to
functional organization. In addition, the fact that a single f-structure may be the
value of more than one functional structure attribute allows us to represent the fact
that a single phrase may play multiple functional roles (see §2.4).
Conversely, we might regain representational uniformity by representing phrase
structural information in the same way as the f-structure, by means of an attribute-
value structure. In fact, a tree is nothing more than a special kind of attribute-value
structure: one in which a linear ordering between nodes is imposed, so that a node
can be said to precede or follow another node; cycles are disallowed, so that a
node cannot both dominate and be dominated by another node; and values of
attributes may not be shared.
274 beyond c-structure and f-structure

However, representing phrase structure information in terms of attribute-value

structures has the potential for leading to confusion. The same formal properties
that render trees an inappropriate representation for functional information make
them a very good representation for phrase structure information. Attribute-
value structures are not inherently ordered, and they allow a node to have more
than one mother. On the other hand, the words of an utterance do appear in
a particular order, and (on most theories of phrase structure organization) a
c-structure node cannot be dominated by more than one mother node.6 The
particular characteristics of phrase structure are concisely and intuitively captured
by a phrase structure tree, not an attribute-value representation.
Other linguistic information might be best represented in terms of other kinds
of formal structures. For instance, as we will see in Chapter 8, a deductive semantic
approach meshes well with the overall LFG architecture; such an approach is well
suited to the expression of meanings in terms of formulas in a logical language like
those standardly used in formal semantic theory. On the LFG view, the represen-
tation of each type of linguistic information by means of structures that reflect
the nature of that information allows for a systematic, flexible, and principled
account of relations between modules, promotes clear intuitions about the formal
and linguistic properties of each module, and aids in developing reasonable ways
of thinking about interactions between modules.

In the remainder of Part II of this book, we will present the different linguistic
modules which the LFG projection architecture comprises, exploring their repre-
sentation and how they are related to one another by correspondence functions.

7.7 Further reading and related issues

Some early LFG work does not assume a projection architecture, but makes other
assumptions about the representation of the various aspects of linguistic structure
and how they are related. For example, Fenstad et al. (1987) outlined a proposal for
the architecture of LFG that involves codescription and that differs significantly in
its formal architecture from standard LFG approaches; Fenstad et al.’s approach
was independently developed at about the same time as the early development of
Head-Driven Phrase Structure Grammar (Pollard and Sag 1994), and it formally
resembles HPSG much more closely. Besides the c-structure tree, Fenstad et al.
propose a bipartite structure consisting of a syntactic representation like the
f-structure and a semantic representation in attribute-value format. Additionally,
they propose that other levels of structure, such as phonological structure, are also
represented as subparts of the overall structure. The following is their representa-
tion of the semantic and syntactic structure for the sentence John walks:

6 Some approaches allow phrase structure trees that violate some of these formal conditions; see,
for example, McCawley (1988). McCawley’s trees are, then, a more abstract representation of syntactic
information than the constituent structures of LFG.
 further reading and related issues 275
  
(19) walk
  
  John 
  
    
  
  
   
  
 ◦  
    
    1  
 
  
  d 
  
 
 
 
   
 
 
   
   
   
   
 
’

Andrews and Manning (1999) also present a very different view of the relations
between linguistic structures within the LFG framework, an approach that for-
mally resembles the work of Fenstad et al. (1987) as well as work in HPSG.
Andrews (2008) proposes a view of semantic structure and the syntax-semantics
interface which, unlike the codescription approach to be presented in Chapter 8,
proposes that semantic structure and semantic constraints are established on the
basis of description by analysis of the f-structure. Andrews motivates this alterna-
tive analysis by providing a set of arguments against the standard codescription
approach and in favor of his description-by-analysis alternative. Though we do
not adopt his analysis in this book, it is interesting as an alternative view of how
different aspects of linguistic structure are related.
In our discussion of linguistic modules, we will not touch on proposals for
enriching the projection architecture to enable the analysis of multimodal com-
munication: see Giorgolo and Asudeh (2011b) for a very interesting proposal to
augment the projection architecture to account for the role that gestures (specifi-
cally, spontaneous hand gestures) may play in interpretation.
 8
Meaning and semantic
composition

We now embark on an exploration of the theory of the relation between syntax

and meaning, examining how the meaning of an utterance is determined on the
basis of its syntactic structure. Early work in LFG proposed that the semantic
form value of the f-structure pred represented certain aspects of the meaning of
the f-structure. Later work assumes the existence of a separate level of semantic
structure or s-structure, related to the f-structure by a correspondence function. In
this chapter, we briefly review some previous LFG approaches to semantics and
the syntax-semantics interface. We then present the glue approach to semantic
composition. This approach, which we adopt in the rest of the book, provides a
firm theoretical foundation for the discussions and analyses that we present.

8.1 Syntax and semantic interpretation

The central problem of semantic interpretation is plain: people have no trouble
understanding the meanings of sentences in their language that they have never
heard before. Thus, it must be possible to determine the meaning of a novel
sentence on the basis of the meanings of its component parts. The idea that the
meanings of larger pieces are assembled from the meanings of the smaller pieces
that make them up is known as the Principle of Compositionality, and is generally
attributed to Gottlob Frege (though the accuracy of this attribution has been
disputed; see, for example, Janssen 1997, 2001). An adequate treatment of linguistic
meaning requires, then, a theory of the meanings of the most basic units of a
sentence, together with a theory of how these meanings are put together.
A commonly accepted version of the Principle of Compositionality is the rule-
to-rule hypothesis, which states that “a very close relation is supposed to exist
between the rules of the syntax and the rules of the semantics” (Bach 1989: 124).
This means that each syntactic rule for combining syntactic units to form a larger
syntactic unit corresponds to a semantic rule that tells how to put the meanings
of those units together to form the meaning of the larger unit. The syntactic rules
in question are often assumed to be phrase structure rules, so that instructions
for combining meanings are paired with instructions for forming constituent
structure phrases.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 syntax and semantic interpretation 277

However, this version of the rule-to-rule hypothesis is actually just one way of
enforcing an orderly theory of semantic composition, one in which the intuition
that the meaning of a whole depends on the meanings of its parts is made explicit
by defining the relevant parts as phrase structure constituents. In fact, research on
the syntax-semantics interface and semantic composition in LFG has shown that
we can remain faithful to the Principle of Compositionality without assuming that
rules for putting meanings together must depend on phrasal primitives such as
linear order and phrasal dominance.
Since the inception of semantic research in LFG, convincing arguments have
been made that semantic composition should proceed mainly by reference to
functional structure rather than constituent structure organization. As argued by
Fenstad et al. (1987: Chapter 2), the units that are primarily relevant for semantic
composition are units at f-structure and not necessarily at c-structure. For example,
as we have seen, a semantic unit may correspond to discontinuous portions of
the c-structure tree. Example (32) in Chapter 4, repeated as (1) here, shows that
the Warlpiri analog of the English phrase small child need not form a c-structure
constituent; the noun kurdu-ngku ‘child’ appears at the beginning of the sentence,
while its modifier wita-ngku ‘small’ appears at the end.

(1) kurdu-ngku ka wajilipi-nyi wita-ngku

child-erg prs chase-npst small-erg
‘The small child is chasing it.’
 
’
 
 
 
 
  
 
 
  
  
 
 
IP  
 
NP I

N I S

kurdu-ngku ka V NP

wajilipi-nyi N

wita-ngku

However, rules for semantic composition in both Warlpiri and English treat the
subject of the Warlpiri sentence kurdu-ngku ka wajilipi-nyi wita-ngku and that of
the English sentence The small child is chasing it as an f-structure constituent and as
a semantic unit; in fact, the rules for semantic composition in the two languages are
remarkably similar, considering the great differences between the two languages
at the c-structure level. Guiding semantic composition by reference to f-structure
278 meaning and semantic composition

and not c-structure relations brings out and clarifies crosslinguistic commonalities
in principles of semantic composition, commonalities that would otherwise be
obscured by properties of the more variant c-structure.
Even given the centrality of f-structure in semantic composition, however, it
must be kept in mind that semantic composition does not depend upon f-structure
relations alone. For example, as pointed out by Halvorsen (1983) and Fenstad
et al. (1987), prosody has a strong effect in determining semantic interpretation.
Prosodic information is represented at prosodic structure, a structure that is
related to but separate from the c-structure; the modeling of prosodic influence
on semantic interpretation is explored in Chapter 11. Information structure,
described in Chapter 10, is also closely connected to semantic interpretation.
In the rest of this chapter, we will not examine constraints on meaning assembly
imposed at nonsyntactic levels of representation, leaving discussion of these issues
for Chapter 10 and Chapter 11.

8.2 Semantic forms

As discussed in §2.4.1, the value of the pred feature in the f-structure is called
a semantic form. This nomenclature reveals an early LFG view of meaning and its
relation to f-structure: semantic forms were originally seen as the locus of semantic
representation. On the view presented by Kaplan and Bresnan (1982), semantic
forms represent four types of information (see also Dalrymple et al. 1993 and Kuhn
2001c):

(2) a. Specification of the semantic relation

b. Mapping of grammatical functions to semantic roles
c. Subcategorization information (the governed grammatical functions)
d. Instantiation to indicate distinctness (predicate uniqueness)

This assumes a semantic form like the one in (3) for the verb give:

(3) Semantic form for give (Kaplan and Bresnan 1982):

subj obj oblgoal
‘give! — , — , — "’
agent theme goal

This semantic form specifies that the predicate give has three arguments with roles
agent, theme, and goal; that the agent is mapped to subj, the theme is mapped
to obj, and the goal is mapped to oblgoal ; that the f-structure for a sentence with
this verb must contain a subj, an obj, and an oblgoal in order for Completeness
and Coherence conditions to be met; and that this use of the verb give is distinct
from other uses of the same verb, since each use of a semantic form is uniquely
indexed (§5.2.2.1).
More elaborated theories of several of these aspects of semantic forms have
emerged in the years since Kaplan and Bresnan’s original work. Most obviously, the
 semantic structure and meaning composition 279

mapping of grammatical functions to semantic roles has been the focus of much
theoretical attention; this is discussed in detail in Chapter 9. Further, the semantic
form is no longer assumed to represent semantic relations. Instead, the semantic
contribution of a verb like give is reflected in the semantic structure and its relation
to meaning (to be described in this chapter), as well as in argument structure
(Chapter 9). This separation leads to a more modular theory, since on this view
f-structure is a purely syntactic level of representation, not a mix of syntactic and
semantic information (on modularity, see §7.2). In addition, a more adequate view
of meaning and its relation to syntax is thereby available: the original view of the
semantic form was inadequate to represent anything but the most basic semantic
relations. Semantic forms could not represent many aspects of interpretation,
including scope of modifiers, quantification, and notions of coreference.
What, then, is the role of the semantic form value of the pred feature in the
current setting? First, the function of instantiation to indicate distinctness remains.
Many words make a syntactically unique contribution, and the fact that semantic
forms are instantiated uniquely for each instance of their use is a formal reflection
of this.
Second, semantic forms represent the array of syntactic arguments that a pred-
icate requires, making explicit the result of the application of mapping principles
to the argument structure of a predicate. As we discuss in Chapter 9, syntactic and
semantic argument structure are not the same; verbs like intransitive eat and rain
illustrate this point:

(4) a. Chris ate.

b. It rained.

Although eat denotes a two-place relation between an eater and an eaten thing,
syntactically it has an intransitive use, as shown in (4a); conversely, rain does
not take a semantic argument, but is syntactically monovalent. Semantic forms
represent the syntactic grammatical functions required by a predicate, whether
or not they make a semantic contribution. Andrews (2008) provides further
discussion of the role of semantic forms in LFG.

8.3 Semantic structure and meaning composition

Approaches to semantics and the syntax-semantics interface within the LFG
framework share a striking degree of commonality: rules for semantic composition
are formulated primarily by reference to syntactic predicate-argument structure,
the syntactic organization of f-structure; and a theory of either implicit or explicit
instructions for combining the meanings of the parts of a sentence into the
meaning of the whole—what Fenstad et al. (1987) call a “logical syntax”—is based
on these f-structure relations.
In the first comprehensive treatment of semantics and its relation to syn-
tax within LFG theory, Halvorsen (1983) proposes a semantic structure that is
280 meaning and semantic composition

obtained by analysis of the f-structure, as described in §7.4.1. Halvorsen’s semantic

structure consists of instructions on how to assemble meanings represented as
formulas of the intensional logic of Montague (1973); thus, the semantic structure
represents an explicitly stated and clearly worked out theory of semantic compo-
sition, a set of instructions for meaning assembly.
Reyle (1988) provides a different view of semantic composition, one which is in
some sense more closely tied to c-structure composition but which is interestingly
different from the standard assumptions of the rule-to-rule hypothesis. On Reyle’s
approach, the meaning contributions of the daughters in a phrase structure rule
are gathered up into a set of contributions associated with the mother node.
These contributions consist of expressions of intensional logic that are indexed
by f-structure relations like subj and obj. These contributions can combine in
different orders, and these different orders can correspond to different meanings—
for instance, to different scopes for quantifiers. This approach is similar in some
respects to the treatment of quantifier scope ambiguity described in Dalrymple
et al. (1997) and in §8.8: the order in which meanings are combined does not
necessarily mirror the order of phrasal composition, and a freer order is allowed.
Wedekind and Kaplan (1993) and Kaplan and Wedekind (1993) present a theory
of semantic interpretation that relies on the restriction operator, discussed in §6.9.4.
The restriction operator allows reference to the f-structure that results from remov-
ing an attribute and its value from another f-structure. Wedekind and Kaplan’s
analysis is primarily targeted at a treatment of the semantics of modification, which
had proven problematic in various ways in previous approaches. An interesting
and important aspect of Wedekind and Kaplan’s proposal is that it incorporates
a form of resource accounting: the semantic argument of a modifier is defined in
terms of the meaning that results from removing the modifier from the structure,
and the final meaning is obtained by applying the meaning of the modifier to
this argument. This means that each modifier is required to make exactly one
contribution to the final meaning. In the following, we will see why this property
is a particularly desirable one.
These approaches illustrate three important properties of a theory of semantic
composition. First, the theory should incorporate a systematic and explicit theory
of how meanings combine, grounded in a thorough understanding of the space
of theoretical possibilities, structures, and results. Second, it should not impose
a particular order of composition that is tied to c-structure organization. Third, it
should treat meanings as resources that are accounted for in the course of semantic
composition. Section 8.5 introduces an approach to semantic composition and the
syntax-semantics interface, the glue approach, that meets these conditions. Before
introducing the theory, we must decide on a method for representing the meaning
of an utterance and its parts; in the next section, we address the issue of meaning
representation.

8.4 Expressing meanings

In formulating a theory of the relation between syntax and meaning, one of the first
decisions to be taken is how to represent the meanings of words and phrases. In this
book, we concentrate primarily on issues related to semantic composition and the
syntax-semantics interface. Many details of semantic interpretation do not interact
 expressing meanings 281

significantly with principles of meaning assembly and semantic composition; thus,

our overall goal is to use the simplest possible meaning representations that are
adequate to represent the semantic distinctions in which we are interested.
In Part II of this book, we use standard predicate logic as a way of expressing
meanings. This formal system has several advantages: it is a simple and uncluttered
representation, and it is widely known and generally familiar. Further, meanings
represented as terms of predicate logic can often be readily translated into the
representations used in other semantic theories, so that the use of predicate logic
is not unduly limiting or confining. In fact, our predicate logic representations
might profitably be viewed as abbreviations for the full semantic representations
proposed in other semantic theories. Formally, the only requirement we impose on
our system of meaning representation is that it must permit function abstraction
and application, with a well-defined notion of variable binding, and predicate logic
meets this desideratum.
It is of course possible to work within a different, more expressive theory of
meaning representation, such as intensional logic (Montague 1973), Discourse
Representation Theory (Kamp 1981; Kamp and Reyle 1993), or Situation Seman-
tics (Barwise and Perry 1983). Importantly, these semantic theories are fully
compatible with the “glue” approach to semantic composition that we present here;
glue-based approaches to the syntax-semantics interface using intensional logic,
Discourse Representation Theory, and Situation Semantics are described in §8.4.2.
Indeed, in Part III of the book, we will make use of a version of Discourse Repre-
sentation Theory rather than predicate logic to express meanings. This is because
we require a representation of individuals relevant in the current context, as well
as a notion of dynamic variable binding, in order to treat anaphoric relations.
Additionally, in our discussion of noun phrase coordination in Chapter 16, we need
a basic theory of plurals and group formation, and certain extensions to predicate
logic will be necessary. Although we extend our theory of meaning representation
beyond simple predicate logic in Part III of the book, the glue approach to semantic
composition that we describe in this chapter remains unchanged, since it does not
depend on the particular meaning representation that is chosen.
The following few pages contain a brief introduction to some basic concepts of
predicate logic. Gamut (1991a,b) and Partee et al. (1993: Chapter 7) give a much
more complete explication of the concepts introduced, as well as a full exposition
of their formal underpinnings.

8.4.1 Predicate logic

The expression in (5) represents the meaning of the proper noun David:

(5) David

David is a constant representing the individual David. Representing the meaning

of a proper noun as an individual constant is a convenient simplification; to do
complete justice to the meaning of proper names, a fairly complex theory of indi-
vidual reference would be required. We stress that such a theory is fully compatible
with the glue theory of semantics and meaning assembly that we present, and that
282 meaning and semantic composition

the constant David can be thought of as an abbreviated representation of the fully

fleshed-out semantic contribution of the proper name David.
We use the expression in (6) to represent the meaning of the sentence David
yawned:

(6) yawn (David )

Formally, the expression in (6) indicates that the one-place function yawn is
applied to David —or, to say the same thing in a different way, the predicate yawn
holds of David . This expression means that David yawned, but does not represent
many details of the meaning of the sentence, including its tense. Again, since these
details are not immediately relevant to our discussion, we usually omit them. We
discuss how the semantics of tense and aspect can be modeled in §8.10.

8.4.1.1 Lambda expressions The expression yawn represents a function that

takes an argument like David . For greater flexibility, we would like to have a general
method for constructing functions from other expressions; this is made possible by
the use of the lambda operator λ, which allows the construction of new functions
by abstracting over variables in logical expressions:1

(7) Lambda abstraction:

λx.P represents a function from entities represented by x to entities repre-
sented by P.

Usually, the expression P contains at least one occurrence of the variable x, and we
say that these occurrences are bound by the λ lambda operator.
To avoid situations where the same variable name has accidentally been chosen
for two different variables, we might sometimes need to rename the variables that
are bound by a lambda operator. The expressions in (8a) are equivalent, even
though one contains the variable x and the other contains the variable y, and both
represent the function be a person. Similarly, the expressions in (8b) are equivalent,
representing the function admire oneself:

(8) a. λx.person (x) ≡ λy.person (y)

b. λx.admire (x, x) ≡ λy.admire (y, y)

Besides the equivalences that come from variable renaming, there are many other
equivalent ways of writing a function. We generally try to represent a function
in the clearest way possible, which is usually the simplest and shortest way. For
example, in the case of a one-place function like person , the shortest and simplest

1 A more complete discussion of the lambda operator and the lambda calculus is provided by Gamut
(1991b: Chapter 4) and Partee et al. (1993: Chapter 13). Note that this use of λ has nothing to do with
the node labeling function λ, discussed in §5.1.2, nor with the projection function λ assumed in some
argument structure work and mentioned in §9.2.
 expressing meanings 283

way is just to write the name of the function person . Alternatively, we can apply the
function person to an argument x that is bound by the λ lambda operator, and we
have constructed a one-place function λx.person (x) that is the same as the function
person . The following two expressions are equivalent:

(9) λx.person (x) ≡ person

At times it is clearer to write a function in this way; for example, writing the func-
tion as λx.person (x) shows explicitly that it is a function that takes one argument.
Another way of thinking of a function like λx.person (x) is that it picks out the
set of individuals that are people—that is, the set of individuals x for whom the
expression person (x) is true. The function λx.person (x) is called the characteristic
function of the set of people. We will sometimes refer to sets and their characteristic
functions in our discussions of meaning.

8.4.1.2 Function application As in (6), we can apply a function to its argument:

(10) Function application:

[λx.P](a)
The function λx.P is applied to the argument a.

Square brackets around the function expression have been added to make the
groupings in this expression explicit. This expression is equivalent to the expres-
sion that results from replacing all occurrences of x in P with a. For example,
the expression [λx.yawn (x)](David ) is equivalent to the expression yawn (David ),
which is the expression that results from replacing all occurrences of x in yawn (x)
with David :

(11) [λx.yawn (x)](David ) ≡ yawn (David )

There is usually at least one occurrence of x in P. If there is more than one

occurrence of x, as in (8b), each occurrence is replaced by the argument of the
function:

(12) [λx.admire (x, x)](David ) ≡ admire (David , David )

8.4.1.3 Types We assume that the expressions we are working with are typed. As
shown earlier, we propose the individual constant meaning David for the proper
name David; this meaning has type e (for entity), the type of individuals:

(13) David : e

The expression in (13) indicates that the constant David is of type e. We assume
that there are only two basic types: e is associated with individual-denoting expres-
sions, and t (for truth value) is associated with proposition-denoting expressions,
which have a truth value (that is, which are either true or false). The expression
yawn (David ) is of type t:
284 meaning and semantic composition

(14) yawn (David ) : t

Types of other expressions are built up from these basic types. For example, the
type of a one-place relation like yawn is:

(15) λx.yawn (x) : !e → t"

The function λx.yawn (x) is of type !e → t", a function from expressions of type e
(represented by x) to expressions of type t (represented by yawn (x)).
The type of a two-place relation like select is:

(16) λx.λy.select (x, y) : !e → !e → t""

This is a function from expressions of type e (represented by x) to functions from

expressions of type e (represented by y) to expressions of type t (represented by
select (x, y)).
The types we have examined so far are:

(17) expression type

David e
yawn (David ) t
λx.yawn (x) !e → t"
λx.λy.select (x, y) !e → !e → t""

As we will see, the type of an argument can be important in constraining possibil-

ities for meaning assembly.

8.4.1.4 Quantification Since the work of Montague (1973) and Barwise and
Cooper (1981), there has been a great deal of interest in the properties of quan-
tificational words like every and most. Here we present a brief discussion of
quantification; details can be found in Gamut (1991b: Chapter 7), Partee et al.
(1993: Chapter 14), Peters and Westerståhl (2006), and Keenan and Westerståhl
(2011). In §8.8, we discuss how quantifiers are treated in the glue approach adopted
in this work.
The noun phrase everyone contains a quantificational meaning corresponding
to the determiner every. A sentence like Everyone yawned has a meaning that can
be represented in the following way:2

(18) Everyone yawned.

every (x, person (x), yawn (x))

The quantifier every represents a relation between an individual (here x) and

two propositions involving that individual, the proposition person (x) and the

2 In our analysis of quantification, we use pair quantifiers—expressions like the one in (18)—
instead of standard generalized quantifiers (every (person, yawn)). There is a one-to-one correspondence
between the two types of quantifiers, as shown by Dalrymple et al. (1991).
 expressing meanings 285

proposition yawn (x). The first proposition corresponds to what is often called the
restriction of the quantifier every , and the second proposition corresponds to the
scope. The type of a quantifier like every is:

(19) every : !!e → !t, t"" → t"

This type associates an individual e with a pair of propositions !t, t" that involve
that individual. Different quantifiers place different requirements on this relation.
For example, for every (x, person (x), yawn (x)) to be true, any individual x that is a
person—for whom person (x) is true—must also yawn, satisfying yawn (x). In other
words, every individual that is a person must also be an individual that yawns.

(20) Most people yawned.

most (x, person (x), yawn (x))

The quantifier most requires that more than half of the individuals x satisfying the
proposition person (x) must also satisfy the proposition yawn (x).

(21) No person yawned.

no (x, person (x), yawn (x))

The quantifier no requires that any individual x who satisfies the proposition
person (x) must not satisfy the proposition yawn (x)—that is, there should be no
individuals that are people and that also yawn.
The restriction of a quantifier—its first propositional argument—is fixed, spec-
ified by a noun like everyone as person (x) or by a noun like everything as thing (x),
or given by the meaning of the quantified common noun (people or person in
examples 20–21) and any modifiers it might have. In contrast, the scope of a
quantifier—its second propositional argument—is chosen more freely, and seman-
tic ambiguity can result from different scope choices. For example, although (22)
is syntactically unambiguous, with a single c-structure and f-structure, it has two
different meanings, depending on the order in which the quantifiers are applied.
That is, the sentence is ambiguous because there is more than one way of putting
its component meanings together:

(22) Everyone heard something.

Reading 1: For each person, there is a thing which that person heard (but
each person may have heard a different thing).
every (x, person (x), a (y, thing (y), hear (x, y)))
Reading 2: There is something that everyone heard.
a (y, thing (y), every (x, person (x), hear (x, y)))

In Reading 1, we say that everyone has wide scope, and something has narrow scope.
The scope relations are reversed in Reading 2, where something has wide scope and
everything has narrow scope.
286 meaning and semantic composition

Similarly, as we discuss in detail in §15.2.1, (23) is syntactically unambiguous,

with only one c-structure tree and one f-structure. It is semantically ambiguous,
however, since the scope of the quantifier can vary:

(23) Someone seemed to yawn.

Reading 1: It seemed that someone yawned (although there may not be
anyone present).
seem (a (x, person (x), yawn (x)))
Reading 2: There is someone that seemed to yawn.
a (x, person (x), seem (yawn (x)))

According to Reading 1, the proposition a (x, person (x), yawn (x)) seems to hold,
but the existence of the person (and therefore the occurrence of the yawn) is not
certain—perhaps the only evidence was a noise. In contrast, Reading 2 claims that
there is some individual x that satisfies the proposition person (x) and that also
seemed to yawn, satisfying the proposition seem (yawn (x)).
Such examples show that it is not adequate to rely solely on the f-structure as a
representation of the meaning of a sentence; examples (22) and (23) have a single
f-structure corresponding to more than one meaning. Our theory of semantics and
the syntax-semantics interface allows us to deduce exactly the two meanings given
for (22) and (23), given an unambiguous syntactic structure.

This concludes our brief introduction to predicate logic. We have seen that pred-
icate logic provides a basic yet expressive way of representing linguistic meaning.
This is an advantage from our perspective, since much of our discussion focuses
on issues in meaning assembly, and our claims about the meanings of particular
constructions are fairly general.

8.4.2 Meaning representations and semantic theories

In much LFG work on meaning and semantic composition, detailed assumptions
about the nature and representation of linguistic meaning and its relation to syntac-
tic structure have been explored, and a close analysis of the semantic contributions
of particular phrases or constructions has been the main focus of concern. Work
on integrating an LFG view of semantic composition with particular semantic
theories is important and valuable, since this work not only allows for a fuller
exploration of the relation of syntactic structure to meaning, but also makes
important contributions to semantic theory.
Since the work of Montague (1970b), it has been common to use intensional
logic to express linguistic meaning. Halvorsen (1983) proposes a theory of the
association between f-structures and meanings, outlined briefly in §7.4.1, which
allows the construction of formulas of intensional logic to represent the meanings
of utterances based on their f-structures. Meanings have also been represented as
formulas of intensional logic in an LFG setting by Wedekind and Kaplan (1993).
In other work, the semantic theory of Situation Semantics (Barwise and Perry
1983) is assumed. Fenstad et al. (1987) propose that functional descriptions in
rules and lexical entries describe not only the f-structure for an utterance but also
a Situation Schema, which represents information that is relevant for semantic
 meaning assembly and logical “glue” 287

interpretation. Situation Semantics adheres to the Relational Theory of Meaning,

whereby the meaning of an utterance is a relation between the situation in which
an utterance is made—the utterance situation—and the situation described by the
utterance, the described situation. Accordingly, the situation schemata proposed by
Fenstad et al. (1987) represent a potentially underdetermined description of the
relation between an utterance situation and a described situation. Fenstad et al.
provide an extensive treatment of constraints on situation schemata as well as
an algorithm for their interpretation. Gawron and Peters (1990) also propose a
Situation-Theoretic view of anaphora, quantification, and their interactions from
an LFG perspective, and their work includes an appendix containing an LFG
grammar for the fragment of English that they treat.
Perhaps the most widely adopted theory of semantics among LFG researchers
is Discourse Representation Theory (DRT: Kamp 1981; Kamp and Reyle 1993).
Discourse Representation Theory assumes that each sentence in a discourse con-
tributes to the construction of a Discourse Representation Structure (DRS) rep-
resenting the discourse referents that are introduced as well as the conditions
they must meet. Frey and Reyle (1983) advanced one of the first proposals for
constructing Discourse Representation Structures for utterances based on their
f-structures, and this work was continued by Wada and Asher (1986) and Asher
and Wada (1988) in their proposals for LFG-based DRS construction. Muskens
(1995) also proposes an analysis involving Underspecified Discourse Representa-
tion Structures (Reyle 1993) with syntactic assumptions that are very close to LFG.
As noted above, there is no obstacle to representing linguistic meanings accord-
ing to any of these semantic theories within the glue approach. Dalrymple et al.
(1997) discuss quantification and intensionality in the glue approach, using inten-
sional logic to represent meanings. Van Genabith and Crouch (1999a) provide
a detailed and very interesting discussion of different methods for incorporating
dynamic and underspecified meaning representations, similar to the structures of
Underspecified Discourse Representation Theory, within the glue approach (see
also van Genabith and Crouch 1999b). Dalrymple et al. (1999b) briefly discuss
the construction of Discourse Representation Structures in a glue setting, where
meanings are given as expressions of Lambda DRT (Bos et al. 1994). DRT is also
used in a glue setting by Haug (2008b), Bary and Haug (2011), Haug and Nikitina
(2012), and Lowe (2012, 2015c) to model verbal tense and aspect and participial
modification, by Flouraki and Kazana (2009) to model disjunctive agreement in
Modern Greek, and by Arnold and Sadler (2012) to model affected experiencer
constructions. In Part III of this book, we will use a version of DRT, Partial
Compositional DRT (Haug 2014b), since it allows us to represent the semantics
of anaphoric relations.

8.5 Meaning assembly and logical “glue”

This section introduces the glue theory of semantic composition and presents
some basic examples of meaning assembly in the glue setting. We propose a logi-
cally based theory of semantic composition: instructions for combining meanings
are stated as premises in a logical deduction. The deduction of the meaning of an
utterance proceeds by combining these premises as the logic requires, which means
288 meaning and semantic composition

that meaning composition need not proceed according to the rules of phrasal
composition. This represents a move away from the phrase-structure based rule-
to-rule hypothesis. In addition, the logic used to state constraints on meaning
combination is a resource logic, linear logic, which treats meaning contributions
as resources that are accounted for in the meaning deduction. Thus, the theory
conforms to the desiderata introduced at the end of §8.3. The theory is often
referred to as the glue approach because of the role of linear logic in stating how the
meanings of the parts of an utterance can be “glued together” to form the meaning
of the whole utterance.

8.5.1 Meaning specifications and the projection architecture

The lexical entry for a proper name like David contains at least the syntactic
information shown in (24):

(24) David N (↑ pred) = ‘David’

We also assume the following simplified phrase structure rule for NP:

(25) NP −→ N
↑=↓

As discussed in Chapter 5, this lexical entry and phrase structure rule give rise to
the syntactic structures in (26):

(26) NP 𝜙
' (
N pred ‘David’

David

We now augment our theory with a semantic structure and its associated meaning.
As described in Chapter 7, a linguistic structure like the semantic structure
is related to other linguistic structures by means of correspondence functions.
Here, the function 𝜎 relates f-structures to semantic structures, and we say that
the semantic structure is a projection of the functional structure, related to the
functional structure via the 𝜎 (sigma) projection. This is shown pictorially in (27),
where d𝜎 is the semantic structure that is related to the f-structure labeled d by the
correspondence function 𝜎, represented as a dashed arrow:
' (
(27) NP d pred ‘David’ d𝜎 [ ]

N 𝜎
𝜙
David
 meaning assembly and logical “glue” 289

As noted in §7.1.1, there are two common and equivalent notations for the
correspondence function:

(28) d𝜎 ≡ 𝜎(d)

In the following, we use the more common subscript notation, rather than the
parenthesis notation: that is, we write d𝜎 rather than 𝜎(d) for the semantic structure
corresponding to d via the correspondence function 𝜎. Nothing of substance
depends on this notational choice; using the parenthesis notation would be equally
correct.
We propose the augmented lexical entry in (29) for the proper name David. This
lexical entry differs from the one in (24) in that the expression David : ↑𝜎 has been
added. No additions or changes to the phrase structure rule in (25) are necessary:

(29) David N (↑ pred) = ‘David’

David : ↑𝜎

The expression David : ↑𝜎 is called a meaning constructor, since it is an expression

that tells us how to construct meanings. In this simple case, there is no real meaning
construction involved, since the meaning David is complete on its own. Other cases
are more complex, as we will soon see.
Meaning constructors are pairs, with the left-hand side (the meaning side)
representing a meaning and the right-hand side (the glue side) representing a
logical formula over semantic structures corresponding to that meaning. The
expression David : ↑𝜎 says that David is the meaning associated with ↑𝜎 , the
semantic projection of the f-structure ↑. In (27), the f-structure metavariable ↑
is instantiated to the f-structure labeled d, and so the meaning constructor pairs
the meaning David with the semantic structure d𝜎 . Because meaning constructors
associate meanings with semantic structures and not nodes of the c-structure
tree, f-structural relations and not c-structure configurations are the primary
determinant of how meanings combine.
As discussed in §8.4.1.3, meaning expressions are typed; the constant David is
of type e. We assume that the basic types e and t are associated with semantic
structures, since the type of an expression is important in determining how it can
combine with other expressions. Types are written on the semantic structure as
subscripts enclosed in angled brackets:

(30) David : d𝜎 !e"

When the type of a semantic structure is clear from the context, we often omit it
to reduce notational clutter.
For brevity, we can use a label like [David] to refer to the meaning constructor
in (30). In (31), [David] is defined as a label representing the typed meaning
constructor David : d𝜎 !e" , in which d𝜎 is a semantic structure of type e and David is
an individual constant representing the individual named David:
290 meaning and semantic composition

(31) [David] David : d𝜎 !e"

Using names or labels for meaning constructors is useful for presenting deductions
in a more compact form.

8.5.2 Assembling meanings

Some words, such as verbs, must combine with other meanings to produce a
complete meaning. For example, the meaning of an intransitive verb combines with
the meaning of its subject to produce the meaning of its clause. This means that we
must provide instructions for combining the meaning of a verb with its arguments
to form the meaning of the clause as a whole. We provide these instructions in
logical terms, the “glue language” of linear logic.

8.5.2.1 Example one: Intransitive verbs The syntactic structures for the sen-
tence David yawned, together with the semantic result we desire, are:

(32) David yawned.

IP
𝜙
NP I( ) *
pred ‘yawn!subj"’
N VP ' ( yawn (David ) : [ ]
subj pred ‘David’
David V 𝜎
yawned

The semantic structure for the sentence is related to the f-structure for the sentence
by the correspondence function 𝜎, represented as a dashed arrow. We are not
concerned with the internal makeup of the semantic structure here, and so we have
represented it without internal attributes or values, as the structure [ ]. Below, we
will see cases in which the semantic structure has attributes and values which play
a crucial role in meaning deduction.
Let us see how the meaning yawn (David ) for the sentence David yawned is
obtained. We propose the following simplified lexical entry for the verb yawned:

(33) yawned V (↑ pred) = ‘yawn!subj"’

λx.yawn (x) : (↑ subj)𝜎 ⊸ ↑𝜎

The meaning constructor for yawned pairs the meaning for yawned, the one-place
predicate λx.yawn (x), with the linear logic formula (↑ subj)𝜎 ⊸ ↑𝜎 . This formula
 meaning assembly and logical “glue” 291

contains a new expression: the connective ⊸ is the linear implication symbol of

linear logic, which we discuss in more detail in §8.7. For the moment, we can
think of the symbol as expressing a meaning like if . . . then . . . : in this case, stating
that if a semantic resource (↑ subj)𝜎 representing the meaning of the subject is
available, then a semantic resource ↑𝜎 representing the meaning of the sentence
can be produced.
Additionally, the linear implication operator ⊸ carries with it a requirement for
consumption and production of semantic resources: the formula (↑ subj)𝜎 ⊸ ↑𝜎
indicates that if a semantic resource (↑ subj)𝜎 is found, it is consumed and
the semantic resource ↑𝜎 is produced. We also assume that a name like David
contributes a semantic resource, its semantic structure. In an example like David
yawned, this resource is consumed by the verb yawned, which requires a resource
for its subj to produce a resource for the sentence. This accords with the intuition
that the verb in a sentence must obtain a meaning for its arguments in order for
a meaning for the sentence to be available. Thus, in the linear logic formulas that
comprise the glue (right-hand) sides of meaning constructors, semantic structures
are treated as resources that are contributed by the words and structures of the
sentence.
In (34), we display the annotated c-structure for the sentence David yawned,
together with the f-descriptions (including meaning constructors) contributed by
the words David and yawned. In (35), we instantiate the metavariables represented
by ↑ and ↓ in this tree, using the label y for the f-structure of the entire sentence
and d for the f-structure of the subj. Only the instantiated c-structure annotations
that are important for our current discussion are displayed in (35).

(34) David yawned.

IP

NP I(
(↑ subj) = ↓ ↑=↓

N VP
↑=↓ ↑=↓

David
V
(↑ pred) = ‘David’
↑=↓
David : ↑𝜎

yawned
(↑ pred) = ‘yawn!subj"’
λx.yawn (x) : (↑ subj)𝜎 ⊸ ↑𝜎
292 meaning and semantic composition

(35) David yawned.

IP

NP
I
(y d

N VP ’
y
d
David
(d V
David : d
yawned

) ⊸y
(y ’
λx.yawn (x) : (y

The f-structure for yawn in (35) is labeled y, and the f-structure d for David is
y’s subj. Since (y subj) = d, we can replace the expression (y subj)𝜎 by d𝜎 in the
meaning constructors in (35), yielding the instantiated meaning constructors for
David (labeled [David]) and yawned (labeled [yawn]):

(36) Meaning constructors for David yawned:

[David] David : d𝜎
[yawn] λx.yawn (x) : d𝜎 ⊸ y𝜎

The meaning (left-hand) sides of the meaning constructors in (36) are familiar
from our discussion of predicate logic formulas in §8.4.1. The meaning side of
the meaning constructor labeled [David] is the proper noun meaning David ,
and the meaning side of the meaning constructor labeled [yawn] is the meaning
of the intransitive verb yawned, the one-place predicate λx.yawn (x).
The glue (right-hand) sides of these meaning constructors indicate how these
meanings are associated with the different parts of this sentence. The constant
David is associated with the semantic structure d𝜎 . The glue side of the meaning
constructor labeled [yawn] is more complex: as explained earlier, the connective
⊸ is the linear implication symbol of linear logic. In (36), we can think of this
expression as encoding a meaning like if d𝜎 , then y𝜎 . In other words, the glue side
of the meaning constructor labeled [yawn] in (36) states that if we consume a
resource associated with the semantic structure d𝜎 , then we can produce a resource
associated with the semantic structure y𝜎 .
We must also provide rules for how the glue side of each of the meaning
constructors in (36) relates to the meaning side in a meaning deduction. For
simple, nonimplicational meaning constructors like [David] in (36), the mean-
ing on the left-hand side (David ) is the meaning of the semantic structure on
the right-hand side (d𝜎 ). For implicational meaning constructors like [yawn],
which contain the linear implication operator ⊸, performing an inferential step
 meaning assembly and logical “glue” 293

on the glue side corresponds to applying a function to its argument on the

meaning side:3

(37) x : f𝜎
P : f𝜎 ⊸ g𝜎
P(x) : g𝜎

Each side of an implicational meaning constructor P : f𝜎 ⊸ g𝜎 requires a contribu-

tion: the glue side requires as its argument a semantic structure f𝜎 , and the meaning
side requires an argument for the predicate P. When an appropriate resource such
as x : f𝜎 is available to provide the appropriate contributions on both the meaning
and the glue sides, the result is a complete semantic resource on the glue side and
its corresponding meaning on the meaning side. In the case at hand, the pairing of
the linear logic formula d𝜎 ⊸ y𝜎 with the meaning term λx.yawn (x) means that we
apply the function λx.yawn (x) to the meaning David associated with d𝜎 , obtaining
the meaning constructor yawn (David ) : y𝜎 for the sentence.
With these correspondences between linear logic formulas and meanings, we
perform a series of reasoning steps like the following:

(38) David : d𝜎 This is the meaning constructor [David]. It

associates the meaning David with the subj
semantic structure d𝜎 .
λx.yawn (x) : d𝜎 ⊸ y𝜎 This is the meaning constructor [yawned].
On the glue side, if we find a semantic
resource for the subj d𝜎 , we consume that
resource and produce a semantic resource for
the full sentence y𝜎 . On the meaning side, we
apply the function λx.yawn (x) to the meaning
associated with d𝜎 .

yawn (David ) : y𝜎 By combining [David] and [yawned], we

have produced a semantic structure for the
full sentence y𝜎 , associated with the meaning
yawn (David ).

By using the rule in (37) and the meaning constructors [David] and [yawned], we
have deduced the meaning yawn (David ) for the sentence David yawned, as desired.

8.5.2.2 Example two: Transitive verbs Our next example of meaning deduction
involves a transitive verb; the example differs from the one just presented only in
that the verb takes two arguments instead of one. The c-structure and f-structure
for the sentence David selected Chris are:

3 This is the standard correspondence as defined by the Curry-Howard Isomorphism relating propo-
sitions like d𝜎 ⊸ y𝜎 to terms like λx.yawn(x); see Crouch and van Genabith (2000) for more discussion.
294 meaning and semantic composition

(39) David selected Chris.

IP

NP I
 
N VP ’
 
 
David V  
V NP

selected N

Chris

The lexical entry for the transitive verb selected is:

(40) selected V (↑ pred) = ‘select!subj,obj"’

λx.λy.select (x, y) : (↑ subj)𝜎 ⊸ [(↑ obj)𝜎 ⊸ ↑𝜎 ]

In the meaning constructor for the transitive verb selected, two arguments are
required: a resource for the subj, (↑ subj)𝜎 , and a resource for the obj, (↑ obj)𝜎 .
The square brackets in this expression are there to make the groupings in the
expression clear: selected requires a meaning for its subj, then a meaning for
its obj, to form a meaning for the sentence. In other words, this formula can
be paraphrased as: “If we find a resource for the subject and a resource for the
object, we can produce a resource for the entire sentence.” The meaning side is a
function that requires two arguments and is applied to those arguments to produce
a meaning for the sentence.
In (40), the glue side of the meaning constructor requires the verb to combine
with its arguments in a particular order—the subj first, then the obj—since this
order must respect the order of combination of meanings specified in the lambda
expression on the meaning side. Importantly, the meaning constructor shown in
(41) is exactly equivalent to the one in (40) except that the order of argument
combination on both the meaning and glue sides is reversed, so that the verb
combines with its obj first and then its subj:

(41) λy.λx.select(x, y) : (↑ obj)𝜎 ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 ]

In formal terms, the glue side of this meaning constructor is logically equivalent to
the glue side of the meaning constructor in (40). In principle, we can choose any
order of combination of premises, with no theoretical significance attached to the
choice we make.
The lexical entry for Chris is analogous to the one for David, providing a
meaning constructor with the meaning Chris :
 meaning assembly and logical “glue” 295

(42) Chris N (↑ pred) = ‘Chris’

Chris : ↑𝜎

With these lexical entries for the words in the sentence, we have the following
structures:

(43) David selected Chris.

 
’
 
s d 
 
c
IP

NP I
(↑ ↓ ↑=↓

N VP
↑=↓ ↑=↓

David
V
(↑
↑=↓
David : ↑

V NP
↑=↓ (↑ ↓

selected
N
) ⊸ [(↑ ) ⊸↑ ]
(↑ ’
↑=↓
λx.λy.select (x, y) : (↑
Chris
(↑
Chris : ↑

Instantiating the ↑ and ↓ metavariables appropriately, we have the following

meaning constructors, labeled [David], [Chris], and [select]:

(44) Meaning constructor premises for David selected Chris:

[David] David : d𝜎
[Chris] Chris : c𝜎
[select] λx.λy.select (x, y) : d𝜎 ⊸ [c𝜎 ⊸ s𝜎 ]

From these premises, we can make the following logical deduction:

296 meaning and semantic composition

(45) David : d𝜎 This is the meaning constructor [David]. It asso-

ciates the subject semantic structure d𝜎 with the
meaning David .
Chris : c𝜎 This is the meaning constructor [Chris]. It asso-
ciates the object semantic structure c𝜎 with the
meaning Chris .
λx.λy.select (x, y) : d𝜎 ⊸ [c𝜎 ⊸ s𝜎 ]
This is the meaning constructor [select]. On the
glue side, if semantic resources for the subject d𝜎
and the object c𝜎 are found, a resource for the
sentence can be produced. On the meaning side,
the two-place predicate select is applied to the
subject meaning x and then the object meaning
y to produce the meaning select (x, y) for the sen-
tence.

select (David , Chris ) : s𝜎 We have produced a semantic structure s𝜎 for

the full sentence, associated with the meaning
select (David , Chris ).

As desired, we have concluded that the meaning for the sentence David selected
Chris is select (David , Chris ).
In this book, we will present glue derivations in this informal way. Much glue
literature uses a more standard proof format, as in (46), which represents the same
proof as was presented informally in (45):

(46) David : d𝜎 λx.λy.select (x, y) : d𝜎 ⊸ [c𝜎 ⊸ s𝜎 ]

λy.select (David , y) : c𝜎 ⊸ s𝜎 Chris : c𝜎
select (David , Chris ) : s𝜎

See Asudeh (2012: Chapter 4) for an accessible overview discussion of glue proofs,
and the papers in Dalrymple (1999) for additional discussion.
We sometimes take advantage of the possibility of referring to a meaning
constructor by a label to present an abbreviated representation of a derivation from
a set of premises. For example, we can abbreviate the derivation outlined in (46) in
the following way:

(47) [David] David : d𝜎

[Chris] Chris : c𝜎
[select] λx.λy.select (x, y) : d𝜎 ⊸ [c𝜎 ⊸ s𝜎 ]

[David], [Chris], [select] ) select (David , Chris ) : s𝜎

 constructional meaning 297

The final line in (47) represents the derivation of the meaning select (David , Chris )
for the semantic structure s𝜎 from the premises labeled [David], [Chris], and
[select]. It contains a new expression ), sometimes called the turnstile, which
indicates that the conclusion on the right is derivable from the premises on the
left. Thus, the final line in (47) means that the conclusion select (David , Chris ) : s𝜎
is derivable from the premises labeled [David], [Chris], and [select].
In sum, we have used linear logic as a glue language to provide instructions on
how to glue together or assemble meanings. The use of this logical language lets
us express constraints on meaning combination in a formally coherent and flexible
way, taking advantage of the syntactic relations imposed by the f-structure.

8.6 Constructional meaning

In the examples just presented, meaning terms are associated with words and not
phrase structure rules. In a language like English, annotations on phrase structure
rules serve mainly to determine the functional syntactic roles of constituents. For
the most part, phrase structure rules play only this syntactic organizing function
and do not contribute meaning on their own. This is true for many other languages
as well.
However, this generalization is not exceptionless. There are cases in which a
meaning is associated with a phrasal construction as a whole, where the semantic
properties of the construction go beyond the semantic properties of the words
it contains. A particularly clear example of meaning associated with phrasal
configuration is provided by relative clauses with no relative pronoun, such as:

(48) the man I met

In this example, the phrase I met is a relative clause modifier of man. This
information is not associated with either the word I or the word met. Instead,
the interpretation of I met as a relative clause is due to the phrasal configuration
in which it appears. In Chapter 17, we will propose an analysis of the semantics
of relative clauses, and we will see that the phrase structure rule associated with
relative clause formation in English can be annotated with a meaning constructor,
and thus can make a contribution to meaning.
The view that meanings can be attached either to lexical items or to c-structure
configurations accords with the views of Kaplan and Bresnan (1982: fn. 2), but not
with some other proposals. In particular, Halvorsen (1983) proposes that semantic
content is introduced only in the lexicon, by words and not phrase structure rules
(see also Bresnan 1982a). In a very interesting discussion of verbless sentences,
including the topic-comment construction in Vietnamese and nominal sentences
with no copula in Maori (discussed in §5.4.5), Rosén (1996) shows that attempts to
restrict semantic content to appearing only in the lexicon are inadvisable. Phrase
structure configurations can be associated with meaning constructors, and these
constructors can make an essential contribution to meaning deduction.
298 meaning and semantic composition

8.7 The “glue” language: linear logic

We use expressions of linear logic (Girard 1987) to encode instructions on how
to assemble meanings. Here, we informally describe the properties of the small
fragment of linear logic (the multiplicative fragment) that we will use.4
Intuitively, linear logic is different from classical logic in that premises in a
linear logic deduction are treated as resources that must be kept track of; that
is, as occurrences of resources that can be introduced or consumed. In contrast,
premises in a deduction in classical logic are statements about what is or is not true.
To illustrate this difference, let us assume that we can deduce the statement You
will get wet from the premises If it is raining outside, you will get wet and It is raining
outside in classical logic:

(49) Classical logic:

If it is raining outside, you will get wet.
It is raining outside.
You will get wet.

In classical logic, if a conclusion can be deduced from a set of premises, the same
conclusion can still be deduced if additional premises are added:

(50) Classical logic:

If it is raining outside, you will get wet.
It often rains in March.
It was raining yesterday.
It is raining outside.
You will get wet.

In contrast, linear logic does not necessarily allow the same conclusion to be
deduced when additional premises are introduced. Instead, propositions in linear
logic can be thought of as resources, and an economic metaphor is sometimes used.
For instance, we can use the symbol $1 for the proposition that you have a dollar,
$1 ⊸ apple for the linear logic proposition that if you have $1, you can have an
apple, and apple for the proposition that you can have an apple. The following is
valid in linear logic:

(51) [$1 ⊸ apple ], $1 ) apple

This can be read as:

(52) If you have $1, you can have an apple.

You have $1.
You can have an apple.

4 In this section, we describe only the properties of the linear implication operator ⊸. Proof rules
for our fragment of linear logic are given in the appendix to the chapter (page 321).
 the “glue” language: linear logic 299

Just as in the real world, it is not possible to get two apples with $1, or to still have
$1 as well as the apple:

(53) INCORRECT (obtaining two apples with $1):

[$1 ⊸ apple ], $1 ) apple , apple
INCORRECT (obtaining an apple while keeping $1):
$1, [$1 ⊸ apple ] ) $1, apple

More schematically, inferences in linear logic work in the following way:

(54) INCORRECT: A ) A, A
We cannot deduce A, A from A.
A resource cannot be duplicated.
INCORRECT: A, B ) A
We cannot deduce A from A, B.
A resource cannot be discarded.
INCORRECT: A, [A ⊸ B] ) A, B
The resource A is consumed by A ⊸ B to conclude B.
A resource is consumed by an implication.
INCORRECT: A, [A ⊸ B] ) A ⊸ B, B
Both A and A⊸B are consumed in concluding B.
A linear implication is also a resource and is consumed in the deduction.
CORRECT: A, [A ⊸ B] ) B

This resource-sensitivity of linear logic allows us to model the meaning contri-

butions of words and phrases as semantic resources that must be accounted for.
The meaning of a sentence is deduced from the meanings of its component parts;
it would be incorrect to deduce the same meaning for the sentence if words or
phrases are added or subtracted. Each word or phrase makes a unique contribution
that must be reflected in the final meaning of the sentence, and meanings cannot
be arbitrarily duplicated, added, or discarded.

8.7.1 Semantic completeness and coherence

Formally, we say that a meaning derivation for an utterance is semantically complete
if the meaning derivation from the premises contributed by the meaning-bearing
items in the sentence produces a meaning for the semantic structure for the
utterance that does not contain any unsaturated expressions (that is, in which all
of the meaning contribution requirements are satisfied). If no such meaning can
be produced, some required material is missing and the utterance is semantically
incomplete.
We say that a meaning derivation for an utterance is semantically coherent if
the meaning derivation produces a meaning for the utterance with no additional
unused premises remaining. If extra resources remain, the utterance is semantically
incoherent.
300 meaning and semantic composition

Semantic completeness and coherence are related in a clear way to the syntactic
Completeness and Coherence conditions on f-structures discussed in §2.4.6. This
is as expected, since most syntactic arguments also make a semantic contribu-
tion and thus must be accounted for in a meaning derivation; indeed, our logi-
cally defined semantic completeness and coherence conditions subsume syntactic
Completeness and Coherence in all cases except for pleonastic or semantically
empty (expletive) arguments, which make no semantic contribution and are not
accounted for in a semantic derivation. The following sentence is syntactically and
semantically incomplete:

(55) ⃰Yawned.

The sentence is syntactically incomplete because the verb yawned requires a subj,
and no subject is present; the sentence is semantically incomplete because the
meaning constructor for yawned requires a semantic resource corresponding to its
subject, but none can be found. Example (56) is both syntactically and semantically
incoherent:

(56) ⃰David yawned Chris.

This example is syntactically incoherent due to the presence of an obj argument,

which yawned does not require. It is semantically incoherent because the meaning
constructor for yawned requires only a subj resource, and in a meaning deduction
from these premises the semantic resource for the obj Chris remains unused.
Semantic and syntactic completeness differ for arguments that make no seman-
tic contribution:

(57) ⃰Rained.

The verb rained requires a subj, but there is no subj in (57); therefore, the
sentence is syntactically incomplete. The semantic completeness condition is not
violated, however, because the subj of rained is not required to make a semantic
contribution—an expletive subject is syntactically required, but not semantically
necessary.5
Another difference between syntactic and semantic coherence involves mod-
ifying adjuncts: a semantic deduction in which the meaning contribution of a
modifier is not incorporated is semantically incoherent, since all meanings must
be taken into account. That is, the semantic coherence condition prevents us from
assigning an unmodified meaning to a sentence with a modifier:

(58) David ran quickly.

cannot mean: run (David )

5 This is indicated by the appearance of the subj argument outside the angled brackets in the
semantic form for rain, as shown in (102) of §2.4.6.1.
 the “glue” language: linear logic 301

The modifier quickly is not constrained by the syntactic Completeness and Coher-
ence conditions, which apply only to governable grammatical functions. Seman-
tically, however, its meaning contribution must be taken into account, and the
deduction is semantically incoherent if the modifier meaning does not appear.

8.7.2 Glue deductions and Categorial Grammar

Glue semantic deductions have an interesting property: as shown by Dalrymple
et al. (1999a), whether or not a glue deduction is possible depends only on the
linear logic glue formulas on the right-hand side of the meaning constructor, never
on the meanings involved in the deduction. This means that we can think of the
meaning deduction process purely in terms of the linear logic deduction; on the
basis of the resulting deduction, we can determine the corresponding meaning by
function abstraction and application.
For example, we can present deductions in an abbreviated form like (59), which
is the same as the deduction in (38) on page 293 except that meaning terms have
been omitted. On the basis of this deduction, we can determine the meaning
corresponding to the semantic structure y𝜎 by function application, following the
function application rule presented in (37):

(59) d𝜎 The subj semantic structure d𝜎 is present.

d𝜎 ⊸ y𝜎 If we find a resource for the subj semantic
structure d𝜎 , we can produce a resource for
the semantic structure for the full sentence y𝜎 .
y𝜎 We have produced a semantic structure for
the full sentence y𝜎 .

In fact, as discussed by Dalrymple et al. (1999a), this aspect of glue semantic

deductions is strongly similar to Categorial Grammar (Oehrle et al. 1988; Moortgat
1988, 1996; Morrill 1994, 2015, 2011; Steedman 1996, 2001; Moot and Retore
2012). Linguistic analysis in Categorial Grammar is a deductive process, in which
the syntactic structure and the meaning of a sentence are obtained by a logical
deduction from premises contributed by its words. The Lambek calculus (Lambek
1958), the logical system commonly used in syntactic analysis in categorial frame-
works, is actually a fragment of noncommutative multiplicative linear logic and so
is very close to the linear logic glue language.
Probably the most important difference between the categorial approach and
the glue approach is in the syntactic primitives that are relevant for semantic
composition. In Categorial Grammar, a predicate combines with its arguments
on the basis of relations defined on the surface string, like to-the-left-of and
to-the-right-of; in the glue approach, in contrast, semantic deductions are guided
by f-structural relations like subj, obj, and adj. This frees the glue approach from
concerns with crosslinguistically variable word order relations and allows semantic
composition to proceed according to the more abstract syntactic organization of
f-structure.
302 meaning and semantic composition

8.8 Quantification
Here we briefly outline our theory of quantification and the treatment of gener-
alized quantifiers, since an explicit theory of the syntax and semantics of noun
phrases will be important in subsequent discussion, particularly in relation to
adjectival modification and relative clauses. For a full explication of the theory of
quantification presented in this section, see Dalrymple et al. (1997).

8.8.1 Quantifier scope

As discussed in §8.4.1.4, the meaning of a sentence like Everyone yawned is:

(60) Everyone yawned.

every (x, person (x), yawn (x))

Here, every relates an arbitrary individual represented by x to two propositions

about that individual, person (x) and yawn (x). We propose the lexical entry in (61)
for the quantificational pronoun everyone:

(61) everyone N (↑ pred) = ‘everyone’

λS.every (x, person (x), S(x)) : ∀H.[↑𝜎 ⊸ H] ⊸ H

This entry has a number of new features, which we will now explain.

8.8.1.1 Quantifier scope and meaning assembly The glue side of the meaning
constructor in the second line of the lexical entry in (61) has several new aspects,
different from the meaning constructors for proper names and verbs that we have
examined thus far:

(62) ∀H.[↑𝜎 ⊸ H] ⊸ H

First, a universal quantifier ∀ binds the variable H, which ranges over semantic
structures that correspond to possible scopes of the quantifier. The universal
quantifier ∀ means something close to the English word all or every, and it binds
the variable that follows it; see Partee et al. (1993: Chapter 7) for a full explanation.
In (62), the expression [↑𝜎 ⊸ H] ⊸ H is asserted to be valid for any H: if
an implicational resource ↑𝜎 ⊸ H is provided for any H, we can consume the
implicational resource and obtain the resource H.
The second new aspect of the meaning constructor in (61) is that it contains
an embedded implication: the implication ↑𝜎 ⊸ H appears on the left side of
the main linear implication operator. We can think of the expression ↑𝜎 ⊸ H as the
argument required by the meaning constructor for everyone. As we have seen, the
arguments required by a meaning constructor appear on the left side of the main
implication operator. An intransitive verb like yawned requires as its argument the
meaning of its subject, (↑ subj)𝜎 :

(63) λz.yawn (z) : (↑ subj)𝜎 ⊸ ↑𝜎

 quantification 303

In contrast, the quantificational pronoun everyone takes a more complex argument,

an implicational meaning constructor ↑𝜎 ⊸ H, in the lexical entry in (61). That
is, everyone requires as its argument a meaning constructor that consumes a
resource for ↑𝜎 to produce some semantic structure H. An intransitive verb with
the quantificational pronoun everyone as its subject would provide such a meaning,
since it consumes a meaning for ↑𝜎 , the semantic structure for everyone, to produce
another semantic resource which we can call H. Any other meaning constructor
that consumes a meaning for ↑𝜎 to produce another semantic structure H will also
fill the bill.
As Saraswat (1999) notes, another way to think of the embedded implication
in (61) is that the quantifier must perform a test on its environment to determine
whether some implicational resource can be found which matches the required
resource ↑𝜎 ⊸ H. To perform this test, the quantifier proposes the resource ↑𝜎 . If
a resource H can then be obtained for some semantic structure H, the requirements
of the quantifier are satisfied, and the conclusion H is valid.

8.8.1.2 Quantifier scope meaning The meaning (left-hand) side of the lexical
entry for everyone in (61) is:

(64) λS.every (x, person (x), S(x))

In this expression, S(x) represents possible meanings for the scope of the quantifier.
To take a concrete example, we present the c-structure, f-structure, and
meaning constructors for the sentence Everyone yawned in (65); we have also
filled in the target meaning for the sentence, associated with the semantic
structure y𝜎 :

(65) Everyone yawned.

𝜙
IP ) *
pred ‘yawn!subj"’
NP I( y ' (
subj e pred ‘everyone’
N VP
𝜎
Everyone V
every (x, person (x), yawn (x)) : y𝜎 [ ]
yawned

[everyone] λS.every (x, person (x), S(x)) : ∀H.[e𝜎 ⊸ H] ⊸ H

[yawn] λz.yawn (z) : e𝜎 ⊸ y𝜎

The right-hand side of the meaning constructor labeled [everyone] requires as its
argument a meaning constructor of the form:

(66) e𝜎 ⊸ H

The glue side of the meaning constructor labeled [yawn] is of exactly this form,
and the derivation is successful if the variable H for the scope semantic structure
304 meaning and semantic composition

is instantiated to y𝜎 . Following the discussion in §8.7.2, we perform the glue deduc-

tion shown in (67), displaying only the glue sides of the meaning constructors:

(67) ∀H.[e𝜎 ⊸ H] ⊸ H If we are given a resource e𝜎 ⊸ H for some semantic

structure H, we can produce a resource for H.
e𝜎 ⊸ y𝜎 H can be instantiated as y𝜎 , satisfying the requirement
in the previous step.

y𝜎 We have produced a resource y𝜎 for the full sentence.

To determine the meaning that results from combining the meaning constructors
labeled [everyone] and [yawn] according to the glue deduction in (67), we follow
the function application rule presented in (37) on page 293, applying the mean-
ing of the quantificational pronoun λS.every (x, person (x), S(x)) to its argument
λz.yawn (z). The resulting meaning expression is:

(68) every (x, person (x), [λz.yawn (z)](x))

or, equivalently:

(69) every (x, person (x), yawn (x))

In sum, assuming the meaning constructors shown in (65) for everyone and
yawned, we can perform the following full glue deduction:

(70) λS.every (x, person (x), S(x)) : ∀H.[e𝜎 ⊸ H] ⊸ H

This is the meaning constructor [everyone].
On the glue side, if we are given a resource
e𝜎 ⊸ H for some semantic structure H,
we can produce a resource for H. On
the meaning side, we apply the predicate
λS.every (x, person (x), S(x)) to the meaning
corresponding to the resource e𝜎 ⊸ H.
λz.yawn (z) : e𝜎 ⊸ y𝜎
This is the meaning constructor [yawn]. If
we are given a resource e𝜎 corresponding
to the subj, we can produce a resource y𝜎
for the entire sentence. The meaning corre-
sponding to this expression is λz.yawn (z).

every (x, person (x), yawn (x)) : y𝜎

We have produced a resource y𝜎 for the
full sentence, corresponding to the mean-
ing every (x, person (x), yawn (x)), by assum-
ing that H is the semantic structure y𝜎 .

We conclude that the sentence has the meaning every (x, person (x), yawn (x)), as
desired.
 quantification 305

8.8.1.3 Determination of scope semantic structure Example (70) shows that

the variable H in the semantic constructor for the quantificational pronoun every-
one can be instantiated to the semantic structure y𝜎 . In §8.4.1.4, we saw that the
scope of a quantifier is not syntactically fixed: sentences containing quantificational
words may exhibit quantifier scope ambiguity. What are the possible semantic
structures that can be chosen as the scope of a quantifier?
First, we note that the semantic structure that is chosen as the scope of a
quantifier need not correspond to any f-structure constituent. For example, it has
long been noted that the restriction of a quantifier can serve as the scope of another
quantifier (Dalrymple et al. 1997):

(71) Every relative of a student attended.

One reading of this sentence is paraphrasable as Everyone who is a relative of some

student attended:

(72) every (x, a (y, student (y), relative.of (x, y)), attend (x))

An abbreviated f-structure for this sentence is:

(73) Every relative of a student attended.

 
’
  
 
 
  
  ’ 
  
  
  − 
 

Treating the article a as a quantificational determiner, we see that its scope is

relative.of (x, y), the proposition that y is a relative of x. This meaning corresponds
roughly to the subphrase relative of, but does not correspond to an f-structure con-
stituent. Instead, the more fine-grained semantic structure is the appropriate level
to define quantifier scoping possibilities; this will become clear in our discussion
of the meanings of determiners and common noun phrases in §8.8.2.
Second, we require the scope of the quantifier to contain the variable bound
by the quantifier. That is, the scope of the quantifier must be a function of the
argument position in which the predicate appears. As noted by Dalrymple et al.
(1997), this follows without stipulation from our logical system: the embedded
implication that the quantifier requires to determine its scope meaning is an
implication from the meaning of the quantified noun phrase to the scope meaning.
A number of other constraints on quantifier scoping have been proposed:
quantifiers may be required to find their scope inside some syntactically definable
domain, or to scope either inside or outside another quantifier. Since our focus
here is not on a complete theory of quantification, we will not discuss constraints
like these or show how they can be incorporated into the framework we propose.
For detailed discussion of quantifier scoping constraints and a proposal for how
they can be imposed in a glue setting, see Crouch and van Genabith (1999).
306 meaning and semantic composition

8.8.1.4 Abstraction The rule for function application given in (37) has been
important in the examples we have seen so far. In order to handle sentences with
multiple quantifiers, we also require a rule of abstraction that allows a hypothetical
resource to be proposed in order to create a function. The rule in (74) allows us
to temporarily posit an additional premise in the deduction, a semantic resource
f𝜎 associated with the meaning x. A semantic resource hypothesized in this way
is notationally distinguished from other premises in that it is enclosed in square
brackets: [f𝜎 ]. If we can successfully perform a deduction (represented by elliptical
.
dots .. ) from this and other meaning constructor premises, producing a semantic
resource g𝜎 with meaning P(x) as in (74), we can discharge the assumption x : [f𝜎 ],
and we are left with the meaning constructor λx.P(x) : f𝜎 ⊸ g𝜎 :

(74) x : [f𝜎 ]
..
.
P(x) : g𝜎
λx.P(x) : f𝜎 ⊸ g𝜎

Intuitively, we have shown that if we are given a resource f𝜎 , we can obtain g𝜎 ,

exactly the import of the linear logic expression f𝜎 ⊸ g𝜎 . On the meaning side,
we have shown that by providing x, we can produce the meaning P(x) — in other
words, that we have proven the existence of a function λx.P(x). This abstraction
rule will be helpful in the treatment of quantifier scope ambiguity as well as in
future discussion. The appendix to this chapter (page 321) contains the full set of
rules of deduction for our fragment of linear logic.
8.8.1.5 Quantifier scope ambiguity By allowing different choices for the scope
of quantifiers, and different orders in which multiple quantifiers scoping at the
same point can apply, we can produce multiple meanings for a sentence with
multiple quantifiers. As noted in §8.4.1.4 in our discussion of (22), the sentence
Everyone heard something is syntactically unambiguous but has two different
meanings, depending on the order in which the quantifiers are applied. This
sentence has the c-structure, f-structure, and meaning constructors in (75):

(75) Everyone heard something.

IP

NP I
 
N VP ’
 
Everyone V h e 
 
V NP s

heard N

something
[everyone] λS.every (x, person (x), S(x)) : ∀H.[e ⊸ H] ⊸ H
[something] λS.a (y, thing (y), S(y)) : ∀H.[s ⊸ H] ⊸ H
[hear] λz.λw.hear (w, z) : s ⊸ [e ⊸ h ]
 quantification 307

The meaning constructors for the quantifiers [everyone] and [something] each
require an implicational meaning constructor of the general form F𝜎 ⊸ H. They
differ in that the meaning constructor [everyone] refers to the subject semantic
structure, e𝜎 , while the meaning constructor [something] refers to the object
semantic structure, s𝜎 .
The meaning constructor [hear] is not exactly of this form, since it requires two
resources to produce a meaning for the clause: a resource corresponding to the
subject meaning e𝜎 , and a resource corresponding to the object meaning s𝜎 . In
fact, we can obtain a meaning constructor of the form that [everyone] requires as
its argument by using the abstraction rule given in (74), supplying a hypothetical
resource v : [s𝜎 ] to [hear]. We can then combine [hear] with [everyone] under
the hypothesized resource, discharge the v : [s𝜎 ] assumption, and combine the
resulting meaning constructor with [something]. The deduction is shown in (76):

(76) λz.λw.hear (w, z) : s𝜎 ⊸ [e𝜎 ⊸ h𝜎 ]

This is the meaning constructor [hear]. On
the glue side, if we are given a resource s𝜎
corresponding to the obj, we can produce
an implicational resource e𝜎 ⊸ h𝜎 . On the
meaning side, we require a meaning z for the
obj to produce the meaning λw.hear (w, z).
v : [s𝜎 ] We hypothetically assume the existence of a
resource s𝜎 with meaning v.
λw.hear (w, v) : [e𝜎 ⊸ h𝜎 ]
Taking into account the hypothesized
resource in the previous step, we have satisfied
the requirement for a resource s𝜎 , and we
obtain the implicational resource e𝜎 ⊸ h𝜎 . On
the meaning side, we apply λz.λw.hear (w, z)
to the hypothesized meaning v, obtaining
λw.hear (w, v).
λS.every (x, person (x), S(x)) : ∀H.[e𝜎 ⊸ H] ⊸ H
This is the meaning constructor [everyone].
On the glue side, it requires an implicational
meaning constructor of the form e𝜎 ⊸ H.
In the previous step, we produced a meaning
constructor that is of a suitable type, and we
can combine the two, instantiating H to h𝜎 .
every (x, person (x), hear (x, v)) : h𝜎
On the glue side, we have now obtained a
resource h𝜎 . On the meaning side, we obtain
the meaning associated with h𝜎 by applying
λS.every (x, person (x), S(x)) to λw.hear (w, v).
308 meaning and semantic composition

λv.every (x, person (x), hear (x, v)) : s𝜎 ⊸ h𝜎

We now discharge the s𝜎 assumption introduced in
the second step, producing an implication on the
glue side, s𝜎 ⊸ h𝜎 . On the meaning side, we abstract
out the meaning v. We now have a meaning con-
structor of the appropriate form to combine with
[something].
λS.a (y, thing (y), S(y)) : ∀H.[s𝜎 ⊸ H] ⊸ H
This is the meaning constructor [something]. On
the glue side, it requires an implicational meaning
constructor of the form s𝜎 ⊸ H. We have a meaning
constructor that is of a suitable type, and we again
instantiate H to h𝜎 .
a (y, thing (y), every (x, person (x), hear (x, y))) : h𝜎
On the glue side, we have obtained a resource
h𝜎 . On the meaning side, we obtain the final
meaning by applying λS.a (y, thing (y), S(y)) to
λv.every (x, person (x), hear (x, v)).
This deduction results in the reading for this sentence in which the object some-
thing has wide scope: there is a particular thing which everyone heard. The other
meaning, where something has narrow scope, proceeds similarly. Here we take
advantage of the observation made in §8.5.2.2 that a predicate may combine with
its arguments in any order; in particular, a transitive verb may combine with its
subject first and then its object, or the object first and then the subject. In formal
terms, the two meaning constructors in (77) are logically equivalent; note that the
lambda-bound variables z and w as well as the semantic structures s𝜎 and e𝜎 appear
in different orders in (77a) and (77b):

(77) a. λz.λw.hear (w, z) : s𝜎 ⊸ [e𝜎 ⊸ h𝜎 ]

b. λw.λz.hear (w, z) : e𝜎 ⊸ [s𝜎 ⊸ h𝜎 ]

Thus, we begin the deduction of the narrow-scope reading of something with the
logically equivalent version of the meaning constructor [hear] in (77b):

(78) λw.λz.hear (w, z) : e𝜎 ⊸ [s𝜎 ⊸ h𝜎 ]

This meaning constructor is a variant version
of [hear], requiring a resource e𝜎 correspond-
ing to the subj to produce an implicational
resource s𝜎 ⊸ h𝜎 .
v : [e𝜎 ] We hypothetically assume the existence of a
resource e𝜎 with meaning v.
λz.hear (v, z) : [s𝜎 ⊸ h𝜎 ]
As above, the hypothesized resource satisfies
the requirement for a resource e𝜎 .
 quantification 309

λS.a (y, thing (y), S(y)) : ∀H.[s𝜎 ⊸ H] ⊸ H

This is the meaning constructor [something].
It requires an argument of the form produced
in the previous step of the deduction.
a (y, thing (y), hear (v, y)) : h𝜎
Combining the resources in the previous two
steps, we have obtained a resource h𝜎 , still
under the hypothesized meaning v associated
with e𝜎 .
λv.a (y, thing (y), hear (v, y)) : e𝜎 ⊸ h𝜎
We now discharge the assumption introduced
in the second step, and we have a meaning
constructor of the appropriate form to com-
bine with [everyone].
λS.every (x, person (x), S(x)) : ∀H.[e𝜎 ⊸ H] ⊸ H
This is the meaning constructor [everyone].
We combine it with the meaning constructor
created in the previous step, instantiating H
to h𝜎 .
every (x, person (x), a (y, thing (y), hear (x, y))) : h𝜎
On the glue side, we have obtained
a resource h𝜎 . On the meaning side,
we obtain the final meaning by
applying λS.every (x, person (x), S(x)) to
λv.a (y, thing (y), hear (v, y)).

This illustrates the possibility of ambiguity resulting from two quantifiers which
take the same semantic structure as their scope, but apply in a different order.
Quantifier scope ambiguity can also arise when a quantifier can choose from
multiple possible scope points, as in (23). In a glue setting, these possibilities
arise naturally, given the different proofs available from the same set of meaning
constructor premises.

8.8.2 Determiners and nouns

We now turn to a sentence involving a determiner and noun, Every student yawned.
This example illustrates how the meanings of the quantificational determiner every
and the common noun student are combined. As we will see, a deduction from the
meaning constructors for every and student produces a meaning similar to the one
proposed in (61) for everyone, which can play a similar role in meaning assembly.
The c-structure, f-structure, and semantic representation for the sentence Every
student yawned are:
310 meaning and semantic composition

(79) Every student yawned.

 
’
 
y



s

IP

NP I
(↑ ↓ ↑ =↓

D N VP
(↑ ↓ ↑ =↓ ↑ =↓ every (x, student (x), yawn (x)) : y [ ]

N V
Every
↑ =↓ ↑ =↓

student yawned

We propose the following lexical entry for the determiner every:

(80) every D (↑ pred) = ‘every’

λR.λS.every (x, R(x), S(x)) :
[((spec ↑)𝜎 var) ⊸ ((spec ↑)𝜎 restr)] ⊸
[∀H.[(spec ↑)𝜎 ⊸ H] ⊸ H]

The meaning constructor for every uses inside-out functional uncertainty (§6.1.3)
to refer to the f-structure for the noun phrase that contains it. The expression
(spec ↑) in this entry refers to the f-structure in which the f-structure for every
appears as the spec value, which is the f-structure labeled s in (79).
The lexical entry for the common noun student is:

(81) student N (↑ pred) = ‘student’

λx.student (x) : (↑𝜎 var) ⊸ (↑𝜎 restr)

According to the lexical entries in (80) and (81), the semantic structure for every
student is complex and has internal structure; it contains two attributes, var and
restr, with semantic structures as their values. The attribute var represents a
variable of type e, and the attribute restr represents a restriction of type t on
that variable—in this case, that the variable must range over individuals that are
students.
These lexical entries, together with the standard English phrase structure rules,
give rise to the structures shown in (82); to save space, only the glue sides of the
meaning constructors for every and student are displayed, and the meaning sides
are omitted:
 quantification 311

(82) every student

v[ ]
s s
r[ ]

NP

D N
(↑ ↓ ↑ =↓

every

) ⊸ ((
(↑ N

⊸ [∀H.[( ↑) ⊸ H] ⊸ H]
[(( ↑) ↑) )] ↑ =↓

student

) ⊸ (↑
(↑
(↑ )

Instantiating the ↑ and ↓ variables and using the labels v and r for the semantic
structures (s𝜎 var) and (s𝜎 restr), as shown in (82), we have the meaning
constructors in (83), labeled [every] and [student]:

(83) Meaning constructor premises for every student:

[every] λR.λS.every (x, R(x), S(x)) : [v ⊸ r] ⊸ [∀H.[s𝜎 ⊸ H] ⊸ H]
[student] λx.student (x) : v ⊸ r

The meaning constructor for [every] requires two arguments: just as a transitive
verb needs two semantic contributions, one from its subject and one from its
object, a quantifier like every needs a semantic contribution from its restriction
(the meaning of the common noun and any arguments or modifiers it might have)
and its scope.
The first requirement is for a meaning for the restriction of the quantifier:

(84) v ⊸ r

This requirement exactly matches the contribution of the common noun student,
and the meaning of student becomes the restriction of the quantifier every .
The second requirement for the quantifier every is a meaning for its scope:

(85) s𝜎 ⊸ H

As described in §8.8.1 for the quantificational pronoun everyone, the quantifier

requires a contribution of the form s𝜎 ⊸ H, whose meaning corresponds to the
scope meaning S of every.
312 meaning and semantic composition

We can now deduce the meaning constructor for every student from the meaning
constructors for every and student:

(86) Combining [every] and [student]:

λR.λS.every (x, R(x), S(x)) : [v ⊸ r] ⊸ [∀H.[s𝜎 ⊸ H] ⊸ H]
This is the meaning constructor [every]. It requires
a resource v ⊸ r corresponding to its restriction
meaning R, and a resource s𝜎 ⊸ H corresponding to
its scope meaning S, to produce a resource H for its
scope semantic structure.
λx.student (x) : v ⊸ r This is the meaning constructor [student]. It provides
an implicational resource v ⊸ r corresponding to the
meaning λx.student (x).
λS.every (x, student (x), S(x)) : ∀H.[s𝜎 ⊸ H] ⊸ H
By combining [every] and [student] we get a result
that is like the meaning constructor for everyone,
except that the restriction of the quantifier every is
specified to involve students.

The resulting meaning constructor for every student is in crucial respects funda-
mentally the same as the meaning for everyone. This is desirable because in terms
of meaning construction, they behave alike; only the meanings associated with the
semantic structures differ.
Completing the deduction, we have the meaning every (x, student (x), yawn (x))
for this sentence, which is the desired result:

(87) Every student yawned.

 
’
 
y



s

[every] λR.λS.every (x, R(x), S(x)) :

[(s𝜎 var) ⊸ (s𝜎 restr)] ⊸ [∀H.[s𝜎 ⊸ H] ⊸ H]
[student] λx.student (x) : (s𝜎 var) ⊸ (s𝜎 restr)

[yawn] λx.yawn (x) : s𝜎 ⊸ y𝜎

[every], [student], [yawn] ) every (x, student (x), yawn (x)) : y𝜎

8.9 Representing semantic features

Thus far, we have represented the meaning of the proper name David simply as
David . This is sufficient for many purposes, but there are a number of properties
associated with the name David that are not clearly represented by the meaning
 tense and aspect 313

David . The name David, used in a particular context with a particular referent,
may refer to an entity that is animate, for example a human or a dog. In the former
case, the referent has the property of humanness; in the latter case, it does not. Most
of the time, the precise referent of a name like David, and the particular semantic
features associated with that referent, are not important for our purposes. In some
situations, however, we need to be able to refer to a particular semantic feature
associated with a meaning (and/or its referent). Rather than simply referring to
↑𝜎 , for example, we may need to refer to the fact that the value of the attribute
animate associated with ↑𝜎 is positive or negative. Example (88) shows the f-
structure and meaning constructor for David in a context in which it is animate and
human:

(88) F-structure and meaning constructor for David:

+ ,
' ( animate +
pred ‘David’ David :
human +

There may be many other semantic and pragmatic features associated with a name
like David in different contexts (Liao 2010, as cited by Dalrymple and Nikolaeva
2011: 78–9). As with functional structures, we usually do not explicitly represent
every attribute in the semantic structure, but only those that are relevant to the
current discussion. Semantic features may be contributed by, for example, syntactic
structure, prosody, or discourse context. Semantic features will prove to be of
particular importance to our discussion of information structure (Chapter 10).
Another semantic feature which will be relevant in Chapter 9 is rel. This feature
is often used as a label representing the meaning of the eventuality denoted by a
verb (i.e. the meaning of a verb minus any tense, aspect, or other grammatically
contributed meaning). For example:

(89) F-structure and meaning constructor for yawn:

- . ' (
pred ‘yawn!subj"’ yawn : rel yawn

Note that the value of rel is merely a label that stands for the inherent meaning
of the word concerned, and does not directly correpond either to the f-structure
semantic form or to the meaning side of a meaning constructor. This accounts for
the lack of subcategorization information in the semantic structure in (89).

8.10 Tense and aspect

In §8.4.1, the expression yawn (David ) was proposed as the meaning of the sentence
David yawned. Clearly, this analysis involves considerable simplification; in partic-
ular, it does not include a representation of the past tense of yawned. Often, such
simplification is desirable, since it permits us to focus on other aspects of semantic
composition with as little complication as possible. For this reason, such simplified
representations will be used in the rest of this book. In this section, however, we
314 meaning and semantic composition

delve more deeply into the semantics of verbs, to show how it is possible to make
distinctions relating to tense and aspect.
Early work on tense and aspect in LFG represented such information purely in
the f-structure; see, for example, Butt (2001b) and Glasbey (2001). Fry (1999b) was
one of the first to propose a model of event semantics for LFG. Haug (2008b) and
Bary and Haug (2011) present a detailed and relatively comprehensive model of the
semantics of tense and aspect in LFG in the context of participial modification (see
also Lowe 2012). The approach presented here is largely based on the proposals of
Haug (2008b), Bary and Haug (2011), and Lowe (2012), though departing from
their assumptions in certain minor ways.
Expressions such as yawn (David ) and λx.λy.select (x, y) treat the meaning of a
verb as a function that takes as arguments the meanings of the syntactic arguments
of the verb. That is, just as David is a syntactic argument of yawn in the sentence
David yawns, so the meaning David is a semantic argument of the meaning yawn
in yawn (David ). This is a convenient simplification, but it does not enable us to
explicitly distinguish the different semantic roles that syntactic arguments may
have. There is clearly a semantic difference between the roles played by the subject
arguments of the verbs yawn and select, for example: the former is a theme,
while the latter is an agent. Some models of argument structure, discussed in
Chapter 9, assume that semantic roles are also relevant for syntactic subcatego-
rization requirements and alternations. Such models represent semantic roles at
argument structure, but these representations do not imply that such relations are
not part of the semantics; rather, they tend to leave the semantic representation to
one side. Other approaches to argument structure seek to separate semantic roles
from argument structure representations as far as possible. We maintain that it is
desirable to represent the semantic relations between verbs and their arguments as
a part of the meaning representation.
In order to capture the semantic aspects of the relationship between a verb
and its syntactic arguments, we can adopt a “neo-Davidsonian” approach to the
representation of syntactic arguments, treating semantic roles as predicates of
events (Davidson 1967; Dowty 1989; Parsons 1990; Schein 1994). Thus far, our
representation has not distinguished event types from individual events. We now
use yawn to refer to an event type, that is, the set of events that can be characterized
as consisting primarily of yawning. This predicate holds of individual events.
Events are a special kind of individual, and we therefore extend our understanding
of the type e (§8.4.1.3) so that e ranges over both events and entities. The meaning
constructor in (90) represents a function from events of unspecified nature to the
proposition that the event is a yawning-type event:

(90) λx.yawn (x) : !e → t"

This meaning constructor is identical in form to that given in (15) on page 284.
It is now used, however, in an entirely different way. The variable x no longer
represents the subject of yawn, as in previous sections, but represents an event that
is characterized as an event of yawning. The simplified expression yawn (David )
can now be expressed in a rather different form as:
 tense and aspect 315

(91) λx.yawn (x) ∧ theme (x, David ) : (↑𝜎 ev) ⊸ ↑𝜎

This meaning constructor encodes the information that David is the theme of
an event x, which is an event of yawning. This expression therefore represents a
function from events in general to propositions in which David is the theme of
an event of yawning. In the glue expression, (↑𝜎 ev) refers to a semantic structure
labeled ev, which represents the event variable of the proposition:

(92) ’ yawn :

In the same way, we previously used the expression select (David , Chris ) to represent
the meaning of the sentence David selected Chris. Now, the semantic relation
between the verb select and its arguments can be more accurately represented as
in (93), where David is specified as the agent of the selecting event, and Chris as its
theme:

(93) λx.select (x) ∧ agent (x, David ) ∧ theme (x, Chris ) : (↑𝜎 ev) ⊸ ↑𝜎

This meaning expresses that David is the agent, and Chris the theme, of an event
x, which is an event of selecting.
Now that we are able to refer to a particular event (or set of events), we are able
to specify the temporal and aspectual characteristics of that event. In order to do
this, we need a theory of the representation of the semantics of tense and aspect.
Reichenbach (1947) was one of the first to develop a formal model of tense and
aspect relations by reference to a limited number of temporal variables. This model
has been further developed by Kamp and Reyle (1993), and is highly influential;
a very similar formal approach is that of Klein (1992, 1994, 1995). As discussed
by Kamp and Reyle (1993), tense and aspect can be modeled with reference to the
following four points in time relative to the discourse:6

(94) E – event time, i.e. the time during which or at which an event occurs.
R – reference time, i.e. the time referred to by the utterance.
P – perspective time, i.e. the “now” point of temporal deixis.
S – speech time, i.e. the moment of utterance.

In most simple sentences, S and P are equivalent, and even in complex sentences,
the relation of P to S can be determined purely by reference to E, R, and P; hence
S is generally omitted from the formalization of tense and aspect, and we will
do so here. Temporal and aspectual relations are indicated using logical symbols
indicating precedence, inclusion, and overlap. In line with Klein (1992), tense
is formalized as the relation between R and P, while aspect is formalized as the

6 These points may be true points, that is instants with no measurable temporal extent, or they may
be periods extending over an interval of time.
316 meaning and semantic composition

Table 8.1 Temporal and aspectual relations

Tense Aspect
Name Present Past Future Imperfective Perfective Anterior
Repr. P⊆R R≺P P≺R R⊆E E⊆R E≺R

relation between E and R. Table 8.1 illustrates the major temporal and aspectual
relations, and their logical representation.7
The symbols ≺ and . indicate a precedence relation; thus, R ≺ P (or, equiva-
lently, P . R) indicates that the time referred to by the utterance entirely precedes
the perspective time of the utterance. This is the basic meaning of relative past time.
The symbols ⊆ and ⊇ indicate an inclusion relation: E ⊆ R or R ⊇ E indicates that
E is included within or coextensive with R. This is a possible definition of perfective
aspect within this framework.8
In order to specify the tense and aspect properties of an event, we therefore need
to refer to the event time E of that event. We use the function τ (tau) to do this:
τ(x) is the temporal interval (which may or may not be a point) during which an
event x occurs (Krifka 1989). Aspect describes the relation between τ(x) and the
reference time R of an utterance. Let us take as a concrete example the sentence
David yawned. We assume for the sake of argument that the English simple past
expresses perfective aspect and past tense. The meaning of perfective aspect is
encoded in the meaning constructor in (95), labeled [perf]:

(95) [perf] λQ.λr.∃x.Q(x) ∧ τ(x) ⊆ r : [(↑𝜎 ev) ⊸ ↑𝜎 ] ⊸ [(↑𝜎 rt) ⊸ ↑𝜎 ]

The term (↑𝜎 rt) in the glue expression refers to the value of the rt (Reference
Time) attribute in the semantic structure ↑𝜎 for the verb, representing its reference
time. On the meaning side, an existential quantifier is introduced to bind the event
variable x, and an inclusion relation (⊆) is specified between τ(x), the time of the
event x, and the variable r, which represents the reference time.
The meaning of the verb yawn in combination with its subject argument
David is given by the meaning constructor in (91), repeated as (96) and labeled
[David-yawned]:

(96) [David-yawned] λx.yawn (x) ∧ theme (x, David ) : (↑𝜎 ev) ⊸ ↑𝜎

Combining the meaning constructor [David-yawned] with the meaning construc-

tor [perf] results in a meaning constructor in which the event variable is quantified,

7 The formalization of tense and aspect presented here may be too simplistic to account adequately
for the wide range of temporal and aspectual features expressed in different ways crosslinguistically. The
variety of use of the English perfect, for example, cannot be accounted for under this system without
assuming that it has two recursively applied aspect properties. Nevertheless, the system presented here
has relatively wide currency, and is adequate for the demonstration of tense-aspect semantics provided
here.
8 The alternative is proper inclusion: E ⊂ R, meaning that E is included within R but R is not included
within E. Following Kiparsky (1998) and Haug (2008b: 294, fn. 1), among others, we assume improper
inclusion here.
 tense and aspect 317

but a new variable—that of the reference time—is introduced and specified as

having a particular relation to the event time of the quantified event. This mean-
ing constructor, labeled [David-yawned-perf] in (97), represents the perfective
aspect of the meaning of David yawned, but does not include the meaning of
its tense:

(97) [David-yawned-perf]
λr.∃x.yawn (x) ∧ theme (x, David ) ∧ τ(x) ⊆ r : (↑𝜎 rt) ⊸ ↑𝜎

The meaning of tense is formally similar to the meaning of aspect. Tense quantifies
the reference time R of an event (rt) in relation to a perspective time P (pt). The
meaning of past tense, for example, can be represented as in (98); we give this
meaning constructor the label [past]:

(98) [past] λQ.λp.∃r.Q(r) ∧ r ≺ p : [(↑𝜎 rt) ⊸ ↑𝜎 ] ⊸ [(↑𝜎 pt) ⊸ ↑𝜎 ]

We can combine the meaning constructor [past] with the meaning construc-
tor [David-yawned-perf] to produce the meaning constructor in (99), labeled
[David-yawned-perf-past]:

(99) [David-yawned-perf-past]
λp.∃r.∃x.yawn (x) ∧ theme (x, David ) ∧ τ(x) ⊆ r ∧ r ≺ p : (↑𝜎 pt) ⊸ ↑𝜎

This meaning constructor represents a function from perspective times to the

proposition that an event of yawning occurred, viewed perfectively and repre-
sented as occurring prior to a reference time, and includes the information that
David is the theme of this event. The perspective time variable p is still lambda-
bound in (99). In some contexts, this variable may be anchored in relation to
another verbal meaning in the clause, for example in the case of nonfinite verb
forms like participles, whose temporal reference is usually anchored by the main
clause. The temporal reference of a finite verb, on the other hand, is independent
within its clause, depending rather on discourse and extra-linguistic context for
its interpretation. The meaning of a finite verb must therefore include a further
specification, which we can label [finiteness], which quantifies the perspective
time (pt) of the sentence:

(100) [finiteness] λQ.∃p.Q(p) : [(↑𝜎 pt) ⊸ ↑𝜎 ] ⊸ ↑𝜎

This meaning constructor can combine with a meaning constructor like

[David-yawned-perf-past] in (99). It augments the meaning by specifying that
an appropriate perspective time exists. This is a convenient simplification, since
the perspective time must be quantified in relation to some other temporal
anchor, which may be discourse internal or discourse external. This quantification
is determined by pragmatic interpretation—by reference to extra-clausal or
extra-linguistic context—and so goes beyond the limits of our semantic theory,
which is restricted to clause-internal relations.
318 meaning and semantic composition

Let us now summarize the derivation of the more complex, but more accurate,
meaning which we have obtained for the sentence David yawned:

(101) Meaning constructor premises for David yawned:

[David] David : d𝜎
[yawn] λy.λx.yawn (x) ∧ theme (x, y) : d𝜎 ⊸ [e𝜎 ⊸ y𝜎 ]
[perf] λQ.λr.∃x.Q(x) ∧ τ(x) ⊆ r : [e𝜎 ⊸ y𝜎 ] ⊸ [r𝜎 ⊸ y𝜎 ]
[past] λQ.λp.∃r.Q(r) ∧ r ≺ p : [r𝜎 ⊸ y𝜎 ] ⊸ [p𝜎 ⊸ y𝜎 ]
[finiteness] λQ.∃p.Q(p) : [p𝜎 ⊸ y𝜎 ] ⊸ y𝜎

From these premises, we can make the following logical deduction:

(102) David : d𝜎 This is the meaning constructor [David].

λy.λx.yawn (x) ∧ theme (x, y) : d𝜎 ⊸ [e𝜎 ⊸ y𝜎 ]
This is the meaning constructor [yawn]. On the glue side, if
semantic resources for the theme d𝜎 and an event e𝜎 can be
found, a resource for the sentence can be produced. On the
meaning side, the one-place predicate yawn is applied to the
event variable x, and the two-place predicate theme is applied
to the event meaning x and the subject meaning y.
λQ.λr.∃x.Q(x) ∧ τ(x) ⊆ r : [e𝜎 ⊸ y𝜎 ] ⊸ [r𝜎 ⊸ y𝜎 ]
This is the meaning constructor [perf]. On the glue side,
if a semantic resource of the form [e𝜎 ⊸ y𝜎 ] can be found,
a resource of the form [r𝜎 ⊸ y𝜎 ] can be produced. On the
meaning side, the variable x is existentially quantified (∃), and
the temporal extent of x, the event time τ(x), is defined as
being included within the temporal extent of a new variable
r representing the reference time.
λQ.λp.∃r.Q(r) ∧ r ≺ p : [r𝜎 ⊸ y𝜎 ] ⊸ [p𝜎 ⊸ y𝜎 ]
This is the meaning constructor [past]. On the glue side, if a
resource of the form [r𝜎 ⊸ y𝜎 ] can be found, a resource of the
form [p𝜎 ⊸ y𝜎 ] can be produced. On the meaning side, the
variable r is existentially quantified, and its temporal extent
is defined as wholly preceding (≺) the temporal extent of a
new variable p representing the perspective time.
λQ.∃p.Q(p) : [p𝜎 ⊸ y𝜎 ] ⊸ y𝜎
This is the meaning constructor [finiteness]. On the glue
side, if a resource of the form [p𝜎 ⊸ y𝜎 ] can be found, a
resource for the sentence can be produced. On the meaning
side, the perspective time variable p is existentially quantified,
producing a full meaning for the sentence.

∃x.∃r.∃p.yawn (x) ∧ theme (x, David ) ∧ τ(x) ⊆ r ∧ r ≺ p : y𝜎

We have produced a semantic structure y𝜎 for the sentence
David yawned, associated with the full meaning for the sen-
tence.
 further reading and related issues 319

In this way, then, it is possible to represent the semantics of tense and aspect using
the glue framework which we have presented in this chapter. The representation
is highly complex, however, even for a sentence as simple as David yawned. In
subsequent chapters, we will ignore the semantic contributions of tense and aspect
by finite verbs and other words, so as not to obscure the points made.

8.11 Further reading and related issues

This chapter has been devoted to an exploration of linguistic meaning and the
syntax-semantics interface. The intention has been to give the reader the linguistic
intuitions behind the analyses, and we have not emphasized the formal and
mathematical properties of the glue language, linear logic. The presentation of
analyses in subsequent chapters is also aimed primarily at an intuitive understand-
ing of how meaning deductions work. It is important to keep in mind, however,
that despite the informal nature of the presentation here and in the following
chapters, our theory of meaning composition is grounded in a mathematically
precise, rigorously defined logic. We will not give a more technically oriented
introduction or overview discussion of linear logic in this volume, since such
material is readily available from other sources. Dalrymple et al. (1999b) give a
more detailed introduction to linear logic in the current setting (see also Dalrymple
et al. 1995e). Linear logic originated in the work of Girard (1987); a very accessible
general overview is given by Scedrov (1993), and Crouch and van Genabith (2000)
provide an in-depth treatment with a linguistic orientation. For the logically
inclined, the appendix to this chapter presents the proof rules for our fragment of
linear logic.
Besides the work mentioned in this chapter, there are a number of papers on
linguistic issues relating to glue theory. The papers in Dalrymple (1999) provide
an overview of the theory, as well as discussions of formal aspects of the theory
and particular linguistic phenomena. Included in this volume are treatments of
quantifier scoping constraints (Crouch and van Genabith 1999), intensionality
and quantifier scope (Dalrymple et al. 1997), negative polarity (Fry 1999a), and
dynamic and underspecified semantics (van Genabith and Crouch 1999a). It is
worth noting that the meaning constructors we assume in this book are cast in
the so-called “Curry-Howard” or “new glue” format, conforming to the proposals
made by Dalrymple et al. (1999a). This format departs from most work in the
glue framework carried out in the twentieth century, including most of the papers
collected in Dalrymple (1999), in which the meaning constructor for David is
written as ↑𝜎 ↝ David (read as ‘↑𝜎 means David’). The two formats have different
expressive power, and in fact the “new glue” format adopted here and in most glue
work since 1999 is the more constrained of the two; see Dalrymple et al. (1999a)
for more discussion of the two formats and the formal differences between them.
Additional work within the glue framework includes work on ellipsis
(Crouch 1999; Asudeh and Crouch 2002b), translation within the semantic
framework of Underspecified Discourse Representation Theory (Crouch et al.
2001), the German split NP construction (Kuhn 2001c), constructional meaning
320 meaning and semantic composition

(Asudeh et al. 2008, 2013), nonrestrictive relative clauses (Arnold and Sadler
2010, 2011), adjectival constructions (Al Sharifi and Sadler 2009), the semantics of
concomitance (Haug 2009), affected experiencer constructions (Arnold and Sadler
2012), optional or ambiguous arguments (Asudeh and Giorgolo 2012; Giorgolo
and Asudeh 2012b), and distance distributivity (Przepiórkowski 2014a,b, 2015).
The logic of pronominal resumption is treated in detail by Asudeh (2012), and will
be discussed further in §17.6.4.
Emendations or extensions to the glue system which we have presented here
have been proposed by some authors, but we omit discussion of these issues.
Kokkonidis (2007) discusses the development of glue theory and proposes a “first-
order” glue system that differs somewhat from the “second-order” system used in
this book, as well as much other glue work; Bary and Haug (2011) adopt and extend
Kokkonidis’ proposals. Andrews (2007b, 2008, 2011) proposes a purely propo-
sitional glue system, and discusses the formal integration of semantic structure
into the grammatical architecture, proposing an alternative view of the relation
between semantic structure and f-structure. For Andrews, meaning constructors
are not directly associated with lexical entries in the way that we have assumed
in this chapter; rather, they are associated with special lexical entries that link
meanings to f-structures. Lowe (2014b) discusses the role and function of semantic
structures within the LFG architecture, and proposes a reformulation of meaning
constructors so that each lexical meaning is associated, via its glue term, with a
single semantic structure.
Giorgolo and Asudeh (2011a) propose a new formal treatment of the meaning
contribution of expressions such as expressives, epithets, appositives, and non-
restrictive relative clauses: such expressions are often taken to motivate a multi-
dimensional semantics, encompassing an “at-issue” dimension and a “side-issue”
dimension (Potts 2005). Giorgolo and Asudeh’s treatment of multi-dimensional
meaning extends the formal glue model to incorporate monads, a construction
from category theory; the model is explored further in their work on the argument-
adjunct distinction and argument linking (Giorgolo and Asudeh 2012b), on con-
ventional implicatures (Giorgolo and Asudeh 2012a), and on opaque contexts
(Giorgolo and Asudeh 2014). Arnold and Sadler (2010, 2011) provide an alterna-
tive treatment of multidimensional meaning which does not incorporate monads,
exploring the possibility of extending the projection architecture to separate at-
issue from side-issue meaning, and also exploring an alternative treatment which
assumes a single semantic projection associated with a meaning pair.
We will not discuss proof methods or algorithms for deduction in linear logic,
since this material is widely available for consultation by those interested in formal
and computational aspects of glue theory. Girard (1987) introduced the notion of
proof nets for proofs in the fragment of linear logic we use; for a lucid description of
the use of proof nets for deduction in the glue approach, see Fry (1999a). Efficient
proof techniques for glue semantic deductions are also explored by Gupta and
Lamping (1998).
 further reading and related issues 321

Appendix: Proof Rules for Linear Logic

The following table of proof rules for the glue fragment of linear logic is adapted
from Crouch and van Genabith (2000). In this book, the multiplicative conjunction
operator ⊗ is used only in §17.6.3; it is also used in some glue treatments of
anaphora.

[A]i
.. A A⊸B
⊸ .
B B
A⊸B
[A]i [B]j
A B ..
⊗ .
A⊗B A⊗B C
C
A ∀x.A
∀
∀x.A[x/a] A[a/x]

provided a does not occur in any

assumptions that A depends on.

In these rules, [ ]i indicates the discharge of a hypothesis labeled i.

 9
Argument structure
and mapping theory

In this chapter, we explore argument structure and its relation to syntax, particularly
concentrating on its role in determining the grammatical functions of the semantic
arguments of a predicate. We will examine different views of the representation
and content of argument structure, and outline the theory of the relation between
thematic roles and grammatical functions. The first five sections explore issues
relating to the theory of argument structure. Sections 9.6 to 9.9 focus on the
analysis of some important phenomena, and further issues relating to grammatical
functions and argument structure are considered in §9.10.

9.1 Syntax, semantics, and argument structure

As discussed in §8.2, it can be easily demonstrated that syntax and semantics are
separate structures, separately constrained, by a consideration of verbs like eat or
rain in examples such as:

(1) a. Chris ate.

b. It rained.

Semantically, the verb eat refers to a two-place relation between an agent, or

individual who eats, and a patient, or entity that is eaten. Syntactically, however,
the verb eat has an intransitive use, illustrated in (1a), where its only argument is
the agent; the patient argument is understood but is not included in the sentence.
Evidence that the verb eat has an intransitive use is given by the possibility of out-
prefixation, as discussed by Bresnan (1982d). Bresnan shows that only intransitive
verbs can participate in out-prefixation:

(2) a. The lamp shines./The lamp outshines the candle.

b. The Brownies found the treasure./ ⃰The Brownies outfound the Girl Scouts
in the treasure hunt.

The verb eat can participate in out-prefixation, indicating that it has an intransitive
use:

(3) Chris outate David.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 syntax, semantics, and argument structure 323

Thus, the verb eat can be syntactically monovalent, requiring only a subj argu-
ment, whereas it is semantically bivalent, denoting a relation between two entities.
Conversely, a verb like rain in (1b) requires a syntactic subj argument, but
semantically denotes a predicate that does not take an argument; it does not make
sense to ask ⃰Who/What rained?, or to replace the subj argument of rain by any
other argument:

(4) ⃰He/David/Something rained.

Here too, syntactic and semantic valence are different: rain requires a syntactic
argument that does not play a semantic role. These simple examples make it
clear that syntactic and semantic argument structures are different and must be
represented separately.
The influence of semantics on syntactic form is shown by the predictable
nature of the syntactic realization of arguments of newly coined verbs. Alsina
(1996: Chapter 1) illustrates this point by considering the nonce word obliquate,
which he defines as meaning “to build or place in an oblique orientation.” As he
notes, the possibilities for syntactic expression of the arguments of this verb are
limited:

(5) a. Jim obliquated the door of the closet.

b. ⃰The door of the closet obliquated Jim.

Because obliquate is a made-up verb, anyone encountering these sentences cannot

have heard sentences like them before. Thus, any constraints on the way the
arguments of this verb are realized cannot be due to any syntactic or morphological
priming. Instead, the pattern of acceptability in (5) must be ascribed to constraints
imposed by the meaning of the verb.
Which aspects of meaning are relevant for determining syntactic roles? Pinker
(1989) outlines two main hypotheses (see also Mohanan 1994): on the first
hypothesis, which Pinker calls “Unrestricted Conceptual Representation,” any
kind of semantic or culturally salient distinction can be reflected in syntax and
can constrain syntactic form and syntactic relations. The second hypothesis, the
“Grammatically Relevant Subsystem” hypothesis, is more restricted: only certain
semantic features, those represented at what is generally termed argument struc-
ture, are relevant for syntax. As Pinker notes, the second hypothesis is more
satisfying. It allows not only for a more precise characterization of argument
structure, but also for the prospect of an explanatory theory of the relation between
argument structure and syntax, as well as a realistic theory of language learning.
Work in LFG adheres to the “Grammatically Relevant Subsystem” paradigm, with
different researchers adopting different views of argument structure and the subset
of semantic information it contains.
In this chapter, we will explore some influential theories of argument structure
and its representation in LFG, and we will examine some proposed constraints on
the relation between argument structure and syntax. We will concentrate primarily
324 argument structure and mapping theory

on the theory of argument structure and its role in mapping between argument
roles and grammatical functions.1

9.2 Content and representation of argument structure

It is generally agreed that argument structure contains some amount of semantic
information, but researchers do not agree on how much. Some claim that there is
very little semantic information in argument structure; others hold that argument
structure is semantically richer, individuating thematic roles like agent, theme,
or goal, and even drawing aspectual and other semantic distinctions.
A number of different views of argument structure have been explored since the
early days of LFG. As discussed in §8.2, Kaplan and Bresnan (1982) propose that
the semantic form value of the pred attribute encodes the relation between the-
matic roles and syntactic functions (see also Bresnan 1982a,c), as in the following
semantic form for the verb give:

(6) Semantic form for give (Kaplan and Bresnan 1982):

subj obj oblgoal
‘give! — , — , — "’
agent theme goal

This expression indicates that give has three argument “slots,” which we refer to
as argument positions. These argument positions are associated with the roles of
agent, theme, and goal; such roles are often referred to as thematic roles. The
three argument positions are also associated with syntactic functions: the agent
with subj, the theme with obj, and the goal with oblgoal . As Kaplan and Bresnan
(1982) note, the angled brackets are supposed to remind us of the parentheses
commonly used in logical expressions, so that the semantic form is thought of as
a kind of logical formula encoding some aspects of the meaning of the sentence as
well as the relation between thematic roles and their grammatical functions.
Since these first steps, researchers in LFG have sought to provide an explicit
view of the content and representation of argument structure and its place in the
overall grammar. Proposals for representing argument structure are often based
on one of two rather different sources: Zaenen (1993), Alsina (1996), Rákosi
(2008), Ackerman and Moore (2013), and others adopt and modify the Proto-
Role argument classification proposals of Dowty (1991), while Butt (1995, 1998),
Broadwell (1998), and others present analyses based on the Conceptual Semantics
framework of Jackendoff (1983, 1990).

1 In this book, we will not provide much discussion of the interaction of argument structure with
other grammatical processes. See Manning (1996a,b) for a very interesting theory of the syntactic
role of argument structure and its interactions with functional structure. Manning proposes that what
he calls construal processes such as binding, determination of imperative addressee, and control in
adverbial and complement clauses are sensitive to relations at argument structure rather than a syntactic
representation like functional structure. For a critical evaluation of Manning’s proposals, see Falk
(2006), particularly §7.4.
 content and representation of argument structure 325

Alsina (1993, 1996) assumes that argument structure is represented via a refine-
ment of the traditional notion of semantic form, appearing in the f-structure as
the value of the pred feature. In line with much research on argument structure
and thematic roles, Alsina relies on an ordering of the arguments of a predicate
according to a semantically motivated hierarchy of thematic roles (to be described
in §9.4.2): for example, arguments bearing the agent role are higher on the
hierarchy than patients. Alsina claims, however, that the differences among
particular thematic roles like agent and patient are best represented semantically
rather than at argument structure; in his theory, such distinctions do not play a
direct role in determining the syntactic functions of arguments.
Besides the hierarchical ordering imposed on the arguments of a predicate,
Alsina builds on work by Dowty (1991) on Proto-Role argument classification,
distinguishing between proto-agent and proto-patient properties of arguments.
Unlike Dowty, however, Alsina assumes a set of criterial definitions for proto-agent
and proto-patient arguments; for example, he proposes that a causer argument with
volitional involvement is necessarily classified as a proto-agent, while an “incre-
mental theme” argument is a proto-patient. Arguments that do not meet these
criteria are not assigned proto-role status. Alsina (1996) provides the following
argument structure representations for the verbs come and give:

(7) Argument
! structures for come
" and give (Alsina 1996):
pred ‘come![P-P] [ ]"’
! "
pred ‘give![P-A] [(P-P)] [P-P]"’

According to these representations, the verb come has two arguments: the individ-
ual who comes is the proto-patient (P-P), and the destination argument bears no
proto-role status and so is not labeled as either P-A or P-P. The verb give has three
arguments: a proto-agent (P-A) or giver; a recipient argument that is optionally
classified as a proto-patient (indicated by parentheses around the P-P label),
depending on whether or not it is causally affected; and a second proto-patient,
the entity that is given. On Alsina’s view of argument structure, no other semantic
information appears in argument structure besides the proto-role status of each
argument and the ordering of arguments according to the thematic hierarchy.
Zaenen (1993) also builds on Dowty’s proposals to define argument structure in
terms of semantically based properties of a predicate. Unlike Dowty, Zaenen does
not assume that these properties relate to semantic entailments of particular uses
of the predicate. Instead, Zaenen proposes that predicates have lexically specified,
semantically definable characteristics that she terms dimensions; for instance, a
predicate has a volitional dimension if it can be used with an adverb like on purpose.
The existence of a volitional dimension in the argument structure of a verb does
not entail that every use of the verb denotes a volitional act; rather, the verb denotes
an act that can be volitional. Based on these semantic dimensions, Zaenen (1993)
proposes a set of role-defining properties that the arguments of a verb can bear,
326 argument structure and mapping theory

and the assignment of a grammatical function to an argument is made on the basis

of these properties.
In contrast with Proto-Role-theoretic accounts, Butt (1995) proposes a seman-
tically richer theory of argument structure, based on the Conceptual Semantics
approach of Jackendoff (1990). On her view, the argument structure for a verb like
give is:

Lexical Conceptual Structure for give (Butt 1995):

(8)  
give
   

 CS([α], GOPoss ([ ], TO[β])) 

  + α  
  AFF ([ ] ,)  
 
ASP( , , )
event

This argument structure representation is tripartite. The first line is the Thematic
Tier, representing causation, motion, and location: here, that the “cause” relation
CS holds between an actor α and an event GOPoss in which some entity comes
to be possessed by a beneficiary β. The second line is the Action Tier, which
encodes information about agency and affectedness. The function aff (affect)
in the Action Tier indicates actor/patient/beneficiary relations; aff+ is a relation
between an actor and a beneficiary (aff− would indicate that the arguments are
an actor and a patient). In this example, the second line indicates that α is the first
argument of the aff+ predicate and is therefore the agent. As Butt points out,
the Action Tier represents roughly the same subset of information that is taken
to be relevant for the Proto-Role theory of Dowty (1991). An innovation in Butt’s
approach is the postulation of an Aspectual Tier, represented on the third line,
which indicates whether the beginning, middle, and end of the event are lexically
specified; in the case of the verb give, neither the inception, duration, nor end point
of the event is intrinsically specified. Compare this with the Hindi-Urdu verb bh uul
‘forget’ shown in (9), which is negatively specified (with the integer 0) for inception
(Butt 1995: 151). Butt shows that aspectual information is important in character-
izing the mapping relation for complex predicates in Urdu:

Lexical Conceptual Structure for bhuul ‘forget’ (Butt

(9)   1995):
bhuul ‘forget’
   

 CS([α], GOInfo ([ ], FROM[α])) 

  AFF([ ]α ,)  
   
 
ASP(0, , )
event

Butt et al. (1997) assume a more traditional view of argument structure in which
the arguments of a predicate bear particular thematic roles like agent, goal, and
theme. Unlike many other approaches, they provide an explicit characterization
of the relation between argument structure and other grammatical structures:
 content and representation of argument structure 327

(10) C-structure, argument structure, and f-structure for cut (Butt et al. 1997):
   
λ
α    
V    

cut

=λ◦α

On this view, argument structure bears a direct relation to c-structure, defined by

the correspondence function α from c-structure nodes to argument structures.
In turn, argument structure is related to f-structure by a function λ.2 Thus, the
familiar 𝜙 mapping between c-structure nodes and f-structures, described in §4.1
and represented by the dashed arrow in (10), is redefined as the composition λ ◦ α
of the α and λ functions.
Butt et al. propose that the argument roles required by a predicate at argument
structure are specified in the meaning constructor (Chapter 8) which represents
the lexical meaning of the predicate. On this view, the verb cut has the following
meaning constructor:
(11) λx.λy.cut (x, y) : (ˆ⃰α agent)λ𝜎 ⊸ (ˆ⃰α theme)λ𝜎 ⊸ ˆ⃰αλ𝜎
In the glue term in (11), ˆ⃰α refers to the argument structure projected via the α
function from the immediately dominating (mother) c-structure node (ˆ⃰), and
(ˆ⃰α agent) refers to the agent feature in the a-structure for this word. (ˆ⃰α agent)λ𝜎
then refers to the semantic structure projected from the f-structure projected from
this a-structure feature. Crucially, no direct reference is made to the grammatical
function label. Based on these relations between structures, Butt et al. (1997)
propose a theory of mapping between argument roles and grammatical functions
in which candidate mappings defined by the λ function are evaluated and the
highest-ranking candidate mapping is selected. This model of argument structure
is also utilized by Asudeh et al. (2008, 2013); for further details, see §9.10.5.2.
A rather different, though not incompatible, approach to argument structure—
one which has significant consequences for the LFG architecture—has been devel-
oped by Asudeh and Giorgolo (2012). In contrast with much LFG work, Asudeh
and Giorgolo propose an approach in which no separate level of argument struc-
ture exists; like Butt et al.’s proposal, the approach includes a glue semantics
component. Asudeh and Giorgolo’s approach shares some of the key goals of Butt
et al.’s model, but solves some problems with Butt et al.’s approach by positing a
connected semantic structure.
In developing their analysis, Asudeh and Giorgolo make a number of new
proposals regarding the representation of f-structure and argument structure
in LFG. Firstly, they argue that the semantic form does not encode syntactic
subcategorization requirements. For example, the lexical entry for the verb select
specifies the verb’s semantic form as in example (12a) and does not, as more usually
assumed, include an argument list, as in (12b):

2 Note that this use of λ is different from the λ projection from c-structure nodes to node labels,
discussed in §5.1.2. It is also unrelated to the λ operator used in semantic representation and introduced
in §8.4.1.1.
328 argument structure and mapping theory

(12) a. (↑ pred) = ‘select’ (Asudeh and Giorgolo 2012)

b. (↑ pred) = ‘select!subj,obj"’ (standard representation)

The f-structure principles of Completeness and Coherence, which in a standard

LFG approach depend upon the subcategorization specifications in f-structure
pred values, are enforced in Asudeh and Giorgolo’s analysis by the resource
sensitivity of the glue semantics (see §8.7.1) for all non-expletive arguments.
Asudeh and Giorgolo (2012) propose that grammatical functions at f-structure
are directly associated with s-structure attributes in the lexical entries of predicates.
For example, under this approach, the lexical entry for English selected contains at
least the following information:

(13) selected V (↑
(↑

(↑
(↑ = (↑ 1)
(↑ = (↑ 2)

2 ) ⊸(↑ 1 ) ⊸(↑ ⊸↑
λy.λx.λe.select (e) ∧ agent (e) = x ∧ theme (e) = y :
(↑

The fourth and fifth lines of the lexical entry in (13) state that the s-structure
of the subj is the value of the s-structure attribute arg1 , and the s-structure of
the obj is the value of the s-structure attribute arg2 . The final two lines contain
the meaning constructor encoding the relevant semantic information: Asudeh and
Giorgolo adopt a neo-Davidsonian event semantics for their meaning language
(§8.10) encoding information about thematic roles, treating them as functions
from events to individuals. At semantic structure, Asudeh and Giorgolo distin-
guish the arguments of a predicate by means of attributes like arg1 and arg2
rather than the names of specific thematic roles such as agent and patient for
the relevant s-structure attributes, since the meaning constructor in the final line
includes an encoding of these roles in the meaning language.
Given the lexical entry in (13), Asudeh and Giorgolo’s f-structure and s-
structure for David selected Chris is:

(14)    
   
   
   
   1 [] 
 
2 []
 grammatical function alternations 329

According to Asudeh and Giorgolo’s proposal, then, the relation between gram-
matical functions and semantic arguments is specified directly in the lexical
entries of predicates, and there is no separate level of argument structure. Using
this model, Asudeh and Giorgolo (2012) develop a detailed analysis of optional
arguments, which we describe in §9.10.5.
A number of other theories of the nature and content of argument structure have
also been proposed; Alsina (1996: Chapter 1) and Butt (2006: Chapter 5) provide
a useful overview and summary. Spencer (2013: Chapter 7) provides a detailed
discussion of argument structure as a level of representation distinct from syntax
and semantics, containing only those aspects of the semantic representation that
are relevant to the morphosyntax, and only those aspects of the functional syntax
that are relevant to the semantics.

9.3 Grammatical function alternations

A primary focus of LFG theory since its inception has been the relation between
different syntactic realizations of a predicate and its arguments, and the character-
ization of grammatical function alternations such as the active/passive alternation
and the dative alternation. LFG adheres to the Principle of Direct Syntactic Encoding
(Kaplan and Bresnan 1982; Bresnan et al. 2016: §5.2), which states that syntac-
tic operations may not alter the subcategorization requirements specified by a
predicate:3

(15) Principle of Direct Syntactic Encoding:

No rule of syntax may replace one function name by another. (Kaplan and
Bresnan 1982: 180)
All grammatical relation changes are lexical. (Bresnan et al. 2016: 77)

Within the constraints imposed by the Principle of Direct Syntactic Encoding,

there has been a steady evolution in the LFG view. In early formulations, alter-
nations like the active/passive relation were characterized by lexical rules relating
lexical entries corresponding to different diatheses of a verbal form. As Bresnan
(1990) notes, however, the original theory of lexical rules did not provide a com-
pletely general picture of grammatical function alternations and their interactions
or of the linking between grammatical functions and thematic roles, and the
need for a more general theory subsequently became apparent. Since these early
formulations, grammatical function alternations have been investigated in the
context of mapping or linking theory, a theory of the mapping between thematic
roles and grammatical functions. A number of different versions of mapping
theory have been proposed. We discuss some of the more influential views in the
following sections; Section 9.11 of this chapter contains a brief overview of other
work on the subject.

3 Complex predicates raise a potential challenge for the Principle of Direct Syntactic Encoding;
see §9.9.
330 argument structure and mapping theory

9.3.1 Lexical rules

Kaplan and Bresnan (1982) represent the relation between grammatical functions
and thematic roles as a part of the semantic form value of the pred feature, as
discussed in §9.2. They treat different assignments of grammatical functions to the
same thematic roles (as in, for example, the active and passive versions of a verb) by
means of lexical rules, rules encoding a systematic relation between lexical entries
for different forms of a predicate.
Early LFG treatments of lexical rules had a distinctly transformational character.
Bresnan (1982c) proposes the following lexical rule for relating passive verbs to
their active counterparts:

subj obj oblagent /∅ subj

(16) ‘L! — — "’ → ‘L! — — "’
agent theme agent theme

or in abbreviated form:

(17) subj → oblagent

obj → subj

This lexical rule was thought of as producing a lexical entry for the passive form of
a predicate (with the agent argument realized as oblagent and the theme as subj)
on the basis of the lexical entry for the corresponding active form (with the agent
argument realized as subj and the theme as obj). The fact that relation-changing
lexical rules were formulated as operations on one lexical entry to produce another
is telling of the transformationally oriented way in which these rules tended to
be viewed in the early days of LFG: though the rules were formulated to apply to
structures pairing thematic roles with grammatical functions, thematic roles did
not usually play much of a role in regulating or constraining the rules. Instead,
research focused primarily on the morphological and syntactic characteristics of
grammatical function alternations rather than on a theory of the alignment of
thematic roles with grammatical functions.

9.3.2 A theory of argument-function mapping

That there are regularities in the mapping between argument structure roles
and grammatical functions has been clear since the pioneering work of Gruber
(1965), Fillmore (1968), and especially Chomsky (1970); an early attempt to
characterize these regularities was made by Ostler (1979), who proposed that the
relation between argument structure roles and grammatical functions is given by
a set of linking rules. Within LFG, Zaenen and Maling (1983) were among the
first to propose a set of Association Principles relating grammatical functions to
thematic roles. For example, they give the following Association Principles for
Icelandic:
 argument classification 331

(18) Icelandic Association Principles (Zaenen and Maling 1983):

1. Agents are linked to subj. (universal)
2. Casemarked Themes are assigned to the lowest available grammatical
function. (specific to Icelandic)
3. If there is only one thematic role, it is assigned to subj; if there are
two, they are assigned to subj and obj; if there are three, they are
assigned to subj, obj, and the secondary object function obj2. This
principle applies after principle 2 and after the assignment of restricted
grammatical functions.

Zaenen et al. (1985) provide further discussion of association principles for Ice-
landic, and also give a similar set of association principles for German; see also
Rappaport (1983).
From this beginning, a vibrant body of research has developed. We only have
room here to scratch the surface of the complex issues that are involved.

9.4 Argument classification

In subsequent research on argument mapping, it was found that grammatical
functions like subj and obj can be grouped into natural classes, and generalizations
about argument mapping can be expressed in terms of these classes rather than
specific grammatical functions. This allows for thematic roles to be specified as
associated with classes of grammatical functions. Levin (1986) was the first to
propose such restrictions on the mapping between thematic roles and grammatical
functions as:

(19) theme is unrestricted.

As discussed in §2.1.4, the (semantically) unrestricted functions are subj and

obj; these are the functions that can be filled by an argument with any thematic
role, by an expletive or semantically empty argument, or by an argument of a
raising verb (as discussed in Chapter 15). Levin’s rule requires themes to be
associated with either subj or obj; more generally, on Levin’s view, arguments with
a particular thematic role can be linked to a class of grammatical functions such as
the unrestricted functions, as in the case of her theme rule.
Continuing the line of research forged by Zaenen and Maling (1983), Levin
(1986), and others, Bresnan and Kanerva (1989) present a comprehensive theory
of mapping from thematic roles to grammatical functions, referred to originally
as lexical mapping theory (LMT), but later as simply mapping theory. Bresnan
and Kanerva’s theory has been further developed and refined by a number of
authors, including Bresnan and Zaenen (1990), Zaenen (1993), Kibort (2001, 2004,
2006, 2007), and Bresnan et al. (2016). In the following, we base our presentation
332 argument structure and mapping theory

on Bresnan and Kanerva’s influential proposal, together with some subsequent

revisions and alternative proposals.4

9.4.1 Cross-classification of grammatical features

Bresnan and Kanerva (1989) propose that the syntactic functions subj, obj,
obj𝜃 , and obl𝜃 are decomposable into the feature specifications +r(estricted) and
+o(bjective) in the following way:

(20) −r: semantically unrestricted functions subj and obj; arguments with any
thematic role (or with no thematic role) can fill these functions (see §2.1.4).
+r: semantically restricted functions obj𝜃 and obl𝜃 : only arguments bear-
ing particular thematic roles can fill these functions. For example, only an
argument with the role of agent can appear as an oblagent .
−o: nonobjective (non-object-like) functions subj and obl𝜃 .
+o: objective functions obj and obj𝜃 .

These features cross-classify the grammatical functions in this way:

(21)
−r +r
−o subj obl𝜃
+o obj obj𝜃

9.4.2 Intrinsic and default classification of thematic roles

Bresnan and Kanerva (1989) assume that at argument structure, the arguments of
a predicate are associated with thematic roles such as agent, goal, and theme.
The relation between argument roles and grammatical functions is expressed by
associating the features +r and +o with thematic roles, given a theory of intrinsic
and default classification principles. For example, Bresnan and Kanerva propose
that arguments bearing the agent role are intrinsically classified thus:

(22) agent encoding: agent

[−o]

This rule states that the agent argument must be a −o function: that is, the agent
may not be associated with an objective or object-like function. Therefore, the
agent role must be associated either with the subj or the oblagent function (one
of the obl𝜃 family of functions).

4 Some authors, such as Alsina (1996), have developed approaches which diverge in more significant
ways from Bresnan and Kanerva’s original formulation; see §9.11.
 argument classification 333

Turning next to the theme or patient thematic role, Bresnan and Kanerva
propose the following classification, reminiscent of the theme-linking rule in (19):

(23) theme encoding: theme/patient

[−r]

According to this classification, a theme argument must be realized as an unre-

stricted function, either subj or obj.
Bresnan and Kanerva classify locative arguments in the same way as agents:

(24) locative encoding: locative

[−o]

A locative argument must therefore be linked to a nonobjective function, either

subj or oblloc .
In addition to the intrinsic classifications, Bresnan and Kanerva (1989) propose
a set of default assignments depending on the relative ranking of thematic roles on
the thematic hierarchy. The thematic hierarchy which they assume is:

(25) Thematic hierarchy (Bresnan and Kanerva 1989):

agent > benefactive > recipient/experiencer
> instrument > theme/patient > locative

The thematic role of a predicate that is highest on the thematic hierarchy, also
ˆ By default, it is classified
called the logical subject, is notationally represented as 𝜃.
as unrestricted:

(26) Default encoding: 𝜃ˆ

[−r]

Other roles are classified as restricted:

(27) Default encoding: 𝜃

[+r]

These default classifications apply in every instance except when a conflict with
intrinsic specifications would result.

9.4.3 Wellformedness principles

Intrinsic and default classifications delimit the possible mappings between
thematic roles and grammatical functions. Principles of wellformedness of the
334 argument structure and mapping theory

mapping relation further specify and constrain these associations, determining

which particular grammatical function is associated with each role.
One such wellformedness principle is the Subject Condition (§2.1.5), which
holds in at least some languages; this principle requires each verbal predicate to
have an argument associated with the subj function. Another is the principle
of Function-Argument Biuniqueness, originally proposed by Bresnan (1982d) (see
also Bresnan and Kanerva 1989). In the following definition, g1 . . . gn is a list of
grammatical functions, and P(1 . . . m) is a semantic form with a list of arguments
1 . . . m:

(28) Function-Argument Biuniqueness (Bresnan 1982d: 163):

G = g1 . . . gn is a possible grammatical function assignment to P(1 . . . m)
if and only if the mapping from 1 . . . m to G defined by i )→ gi is injective
(one-to-one and into).

This principle rules out a situation where the same grammatical function is
associated with more than one thematic role, or where the same role is associated
with more than one grammatical function.5

9.4.4 Other proposals for intrinsic classification

Subsequent crosslinguistic research building on these foundations has explored
alternative intrinsic classification principles applying in particular languages or
groups of languages. Bresnan and Zaenen (1990) present an analysis of intransitive
and resultative verbs that makes use of the standard +r, +o argument classifica-
tions, but eschews the use of default principles in argument mapping. On their
view, thematic roles like agent and patient are intrinsically associated with the
+r, +o argument-classifying features, and further restrictions are imposed on
the basis of how “patient-like” the argument role is. This provides an account of
the unergative/unaccusative distinction (Perlmutter and Postal 1984): the subject
of an unergative intransitive verb is agent-like, while the subject of an unaccusative
intransitive verb is patient-like. Typical unergative verbs are bark and run, with
agentive subjects; typical unaccusative verbs are freeze and die, with non-agentive,
patient-like subjects. This difference in semantic role has syntactic consequences,
in that unaccusative and unergative verbs exhibit different syntactic behavior in
certain constructions: for example, unaccusative and unergative verbs in English

5 The Function-Argument Biuniqueness principle has been challenged by Alsina (1995, 1996) in his
analysis of reflexive and reciprocal constructions, where a single grammatical argument appears to fill
two distinct thematic roles in relation to the predicate. Alsina’s proposals are adopted by Miliĉević
(2009). Reinhart and Siloni (2005) and Rákosi (2008) present alternative analyses of reflexive and
reciprocal constructions which preserve the principle of Function-Argument Biuniqueness (see also
Aranovich 2013). See Asudeh (2012: Chapter 5) for a discussion of Function-Argument Biuniqueness
and its relation to the Theta Criterion (Chomsky 1981) and other theoretical principles governing the
relation between thematic roles and grammatical functions. Argument Fusion in complex predicate
formation also raises potential issues for Function-Argument Biuniqueness; see §9.9.
 argument classification 335

differ in their behavior in constructions with resultative phrases. An unaccusative

verb can appear with a resultative phrase, while an unergative verb cannot (The
river froze solid/ ⃰The dog barked hoarse); however, when a “fake reflexive” is added,
unergative verbs but not unaccusative verbs can appear (The dog barked itself
hoarse/ ⃰The river froze itself solid). Section 9.10.1 provides additional discussion
of the unergative/unaccusative distinction.
Zaenen (1993) presents an approach to argument mapping based on her theory
of argument structure, discussed in §9.2, according to which arguments of a
predicate are associated with certain semantically defined proto-agent and proto-
patient properties. Zaenen proposes that the assignment of intrinsic feature clas-
sifications to an argument is based on whether the argument has a preponderance
of proto-agent or proto-patient properties. Her analysis successfully accounts for
the syntactic and semantic differences between two classes of unaccusative verbs
in Dutch.
Her (2013), building on Her (2003), proposes that in accusative languages the
only intrinsic classification of thematic roles is that patient/theme is classified as
[−r], as in (23); there are no other intrinsic classifications. Her (2013) also argues
that different intrinsic classifications are required to account for ergative languages,
and Her and Deng (2012) claim that in Yami and other ergative languages there are
no intrinsic classifications at all.
Bresnan et al. (2016) also assume a distinction between patient-like roles and
other roles, compatible with a Proto-Role approach. They propose the following
intrinsic classifications:6

(29) patient-like roles: 𝜃

[−r]
secondary patient-like roles: 𝜃

[+o]
other thematic roles: 𝜃

[−o]

Secondary patient-like roles include, for example, the second object argument in a
ditransitive construction, though Bresnan et al. observe that the inventory of roles
belonging to this category seems to be subject to crosslinguistic variation.
Most theories assuming intrinsic argument classifications analyze the agent
argument as filling a nonobjective −o role, as shown in (22). However, Kroeger
(1993) argues that in Tagalog, the agent can be syntactically realized as a non-
subject term, or object. Manning (1996b) makes a similar argument for Green-
landic and other ergative languages. Arka (1998, 2003) and Arka and Simpson

6 For an approach which makes a finer distinction between different sorts of patient-like roles, see
Butt (1998).
336 argument structure and mapping theory

(2008) present analyses of the “objective voice” in Balinese which involve agent
objects. Lødrup (1999b) discusses the Norwegian presentational focus construc-
tion, which also allows agent objects. He gives an Optimality-Theoretic analysis
of Norwegian argument mapping that accounts for the differences between Nor-
wegian and languages like English, which do not allow this possibility. Lødrup
(2000) also analyzes Norwegian existential constructions as permitting agent
objects.

9.4.5 The markedness hierarchy and mapping principles

Markedness principles underlying the association of thematic roles with gram-
matical functions have often been argued to play an important role in argument
mapping. Bresnan and Moshi (1990) develop a theory of markedness principles
which assumes that positive values for the r and o attributes are marked, meaning
that the four major grammatical functions can be partially ordered in a markedness
hierarchy (Bresnan and Moshi 1990: 167):

(30) Markedness Hierarchy of Grammatical Functions (Bresnan and Moshi

1990):
subj > obj, obl𝜃 > obj𝜃

subj is the least marked grammatical function, since it has negative values for
both r and o, while obj𝜃 is the most marked. Notably, this markedness hierarchy
generalizes a segment of LFG’s grammatical function hierarchy, given in (5) of
Chapter 2. The recognition of distinct hierarchies for thematic roles and gram-
matical functions is important, as Falk (2006: 39–44) points out: while the same
argument can be the most prominent with respect to both hierarchies, it is also
possible for each hierarchy to have a different most prominent argument, resulting
in a mismatch. Such a mismatch is exactly what is found, for example, in the
passive construction (to be discussed in §9.6), where the argument ranked highest
on the thematic hierarchy does not map to the most prominent argument on the
grammatical function hierarchy.
The Markedness Hierarchy in (30) motivates a set of Mapping Principles for
associating thematic roles with grammatical functions. The Mapping Principles in
(31), from Bresnan et al. (2016: 334), were originally proposed by Bresnan (2001c):

(31) a. Subject roles:

i. 𝜃ˆ is mapped onto subj when initial in the a-structure;
[−o]
ii. otherwise, 𝜃 is mapped onto subj.
[−r]
b. Other roles are mapped onto the lowest compatible function on the
Markedness Hierarchy (30).
 argument classification 337

According to rule (31a.i), if the highest thematic role 𝜃ˆ is associated with the feature
[−o], it is mapped onto subj;7 if this rule does not apply, rule (31a.ii) states that any
argument associated with the feature [−r] is mapped to subj. Rule (31b) defines
mapping rules for the remaining thematic roles to grammatical functions other
than subj.
Both the Markedness Hierarchy and the Mapping Principles have been subject
to revision as mapping theory continues to develop. Kibort (2001, 2004, 2006,
2007) argues for a simplified reformulation of the Mapping Principles in (31) into
a single Mapping Principle. (What we refer to as argument positions are termed
ordered arguments by Kibort.)

(32) Mapping Principle (Kibort 2001, 2004, 2006, 2007):

The ordered arguments are mapped onto the highest (i.e. least marked)
compatible function on the Markedness Hierarchy.

Kibort argues that this reformulation achieves the correct mappings for various
classes of predicates discussed in the literature, but avoids redundancy by depend-
ing upon the partial ordering of the Markedness Hierarchy. According to this
formulation (and in contrast to Bresnan et al.’s formulation), the Subject Condition
(§2.1.5) is not necessary: this is because the requirement for most predicates to
have a subject is ensured by the Mapping Principle, while impersonal predicates in
many languages do not require subjects.
Her (2013) proposes a revision of both the Markedness Hierarchy and the
Mapping Principles (see also Her 2009). Her proposes that [−r] is less marked
than [−o], with the result that the Markedness Hierarchy is fully ordered:

(33) Markedness Hierarchy of Syntactic Functions (Her 2013):

subj > obj > obl𝜃 > obj𝜃

Her (2013) also proposes a “Unified Mapping Principle” (see also Her 1998, 2003,
2010; Her and Deng 2012) which is very similar to the Mapping Principle of Kibort:

(34) Unified Mapping Principle (Her 2013):

Map each a-structure role that is available onto the highest function that is
compatible and available.
• A role 𝜃 is available for mapping if all roles to the left of 𝜃 are mapped;
a function F is available for mapping to 𝜃 if F is not fully specified for by
another role and also not linked to a role to the left of 𝜃.
• A function is compatible if it contains no conflicting feature.

7 Following Zaenen and Engdahl (1994), Bresnan et al. (2016) assume that semantically empty
arguments of verbs like rain or seem also appear in the argument structure and are eligible for linking
to a grammatical function. Semantically empty arguments can appear to the left of the highest thematic
ˆ this accounts for the additional restriction in (31a.i), which applies only when the highest
role 𝜃;
thematic role 𝜃ˆ is initial in the a-structure. If a semantically empty argument appears to the left of
ˆ the linking rule in (31a.i) does not apply.
𝜃,
338 argument structure and mapping theory

Under this definition, as with Kibort’s proposal, the Function-Argument Biunique-

ness constraint and the Subject Condition are no longer required as independent
constraints, since they are effectively incorporated into the definition of the Map-
ping Principle. Due to Her’s views regarding intrinsic classifications (mentioned
in the previous section), the details of how this Unified Mapping Principle applies
are slightly different from the application of Kibort’s Mapping Principle, but both
appear equally capable of accounting for attested patterns of linking between
thematic roles and grammatical functions.

9.5 Selection and classification: syntactic or semantic?

As described in §9.1, the earliest representations of argument structure in LFG
implicitly or explicitly dissociate argument positions and thematic roles (as in (6)
on page 324). In contrast, the mapping theory developed by Bresnan and Kanerva
(1989) and Bresnan and Zaenen (1990) effectively involves a fusion of argument
positions with thematic roles. Kibort (2007) discusses this development, providing
the following quote from Bresnan and Zaenen (1990: 48):

The grammatically significant participant-role relations in the structure of events are

represented by a-structures. An a-structure consists of a predicator with its argument roles,
an ordering that represents the relative prominence of the roles, and a syntactic classification
of each role indicated by a feature.

On this view, the a-structure for a verb like English select is represented as in (35).
The verb subcategorizes for two arguments, labeled according to their thematic
role. These roles each have an intrinsic classification, as discussed in §9.4.2, and
these classifications, together with default and construction-specific classifications,
constrain the grammatical functions of these roles at f-structure:

(35) select ! agent patient "

intrinsic: [−o] [−r]

In this model, the arguments of a predicate are defined by thematic role, and
argument position plays no role. However, a number of authors including Joshi
(1993), Mohanan (1994: Chapter 3), Alsina (1996: 37), and Ackerman and Moore
(2001: 40–59) have argued for the earlier view, that it is necessary to separate
argument positions from thematic roles.
Kibort (2007) observes that some predicates show variation in the assignment
of thematic roles to grammatical functions, and argues that these patterns are
best analyzed by appeal to argument position rather than thematic role. Verbs
exhibiting the spray/load alternation are examples of such predicates: in terms of
argument positions, both variants are three-place predicates selecting for a subj,
obj, and obl𝜃 , but the two non-agent thematic roles can associate with either the
second or third argument position.8

8 The spray/load alternation is discussed by Ackerman (1990, 1992), Levin (1993: Chapter 2), and
Ackerman and Moore (2001). Similar alternations including verbs such as ‘present’ in Hungarian are
analyzed in mapping theory terms by Ackerman (1992) and Laczkó (2013).
selection and classification: syntactic or semantic? 339

(36) a. David sprayed (the) paint onto the wall.

subj obj obl𝜃
b. David sprayed the wall with paint.
subj obj obl𝜃

Kibort also observes that morphosyntactic operations can make reference to argu-
ment structure positions independent of thematic roles. For example, passivization
and locative inversion can be described as lexical operations affecting the first
argument position of a verb, but there is no way to generalize over the particular
thematic role which that first argument may have. On the basis of such evidence,
Kibort (2007) concludes that argument positions must be separated from thematic
roles, and represents the former using labels such as arg1 , arg2 , and so on.9
Kibort proposes that a single “valency template” is available to non-derived
predicates in all languages. Following Zaenen (1993), Ackerman and Moore
(2001), and others, Kibort proposes that a fixed total ordering exists within this
valency template:10

(37) ! arg1 arg2 arg3 arg4 … argn "

[−o/−r] [−r] [+o] [−o] [−o]

Every predicate selects for one or more arguments from this template. The slots
which a particular predicate selects are lexically specified, and it is possible for
one or more of the argument positions to be skipped. Notice that while this
template broadly corresponds to the hierarchy of syntactic functions (subj >
obj > obj𝜃 > obl𝜃 ), it is defined in terms of the features [±o] and [±r]. The
thematic roles entailed by a particular predicate are associated with argument
positions partly on the basis of intrinsic associations between roles and [±o]/[±r]
classifications (such as the association of agentive roles with [−o] in many lan-
guages), but there is no sense in which a particular thematic role is inherently
associated with a particular argument position. Kibort (2007) also recognizes the
influence of a hierarchical ordering of thematic roles but, following Levin and
Rappaport Hovav (2005: Chapter 6), points out that no thematic hierarchy can be
defined that captures all generalizations over the association of argument positions
with thematic roles. Thematic roles may therefore be understood to be ordered,
but this ordering is only default, and different actual orderings are possible: under
certain conditions roles may map to slots in different ways, as for example in
the spray/load alternation shown in (36). We agree with Kibort that a separation
should be maintained between thematic roles and argument positions, and adopt
her approach to argument structure in the following four sections. We examine
some of the central phenomena which have been the focus of attention in mapping
theory, providing an argument structure analysis based on Kibort’s proposals.

9 Note that Kibort’s use of labels like arg1 , arg2 , etc. is not the same as in Asudeh and Giorgolo’s
model, described in §9.2, where arg1 and arg2 represent s-structure attributes whose values are
s-structures. We distinguish between the two typographically by using small capital letters for Asudeh
and Giorgolo’s s-structure attributes, and lower case letters for Kibort’s argument structure labels.
10 Kibort’s analysis of arg1 as either −o or −r is designed to account for the distinction between
unergative and unaccusative verbs; see §9.10.1.
340 argument structure and mapping theory

9.6 The active/passive alternation

As a first illustration of the theory, we consider the active and passive versions of a
verb like select in examples such as:

(38) a. David selected Chris.

b. Chris was selected by David.

The intrinsic meaning of the verb select involves a selector and a selectee, or
an agent and theme argument. As a canonical transitive verb in English, its
subcategorization frame includes arg1 and arg2 , the first two argument positions
in the universal valency template presented in (37); the agent is associated with
arg1 , and the theme is associated with arg2 .

(39) agent theme

select ! arg1 arg2 "

These argument positions are intrinsically associated with mapping feature speci-
fications:

(40) select ! arg1 arg2 "

[−o] [−r]

We now apply the Mapping Principle (32). The highest, least marked function
on the Markedness Hierarchy is subj. Since arg1 is specified as [−o], assignment
to subj is compatible with its specifications, and it is mapped to subj. Of the
remaining core functions, only obj is compatible with the [−r] specification of
arg2 , so arg2 is associated with obj. We therefore have the following associations,
where M.P. stands for Mapping Principle:

(41) agent theme

select ! arg1 arg2 "

[−o] [−r]
M.P. subj obj

Thus, the theory correctly captures the fact that in the active voice, the agent
argument of select is the subj, and the patient is the obj:

(42) David selected Chris.

agent patient
 
’
 
 
 
 the active/passive alternation 341

We now examine the passive counterpart of this active sentence. Most work
within LFG, following Bresnan and Moshi (1990), assumes an account of the
passive in which the thematically highest argument is suppressed and, therefore,
unavailable for linking. For example, under this approach, the passive of select has
this argument structure:

(43) Suppression-based account of the passive (Bresnan and Moshi 1990):

agent theme

select ! arg1 arg2 "

[−o] [−r]

Under these assumptions, only the second argument slot is available for linking to
a grammatical function. Since [−r] is compatible with the highest function on the
Markedness Hierarchy, arg2 maps to subj. This can be thought of as a promotion
of arg2 from obj, as in (41), to subj:

(44) Suppression-based account of the passive:

agent theme

select ! arg1 arg2 "

[−o] [−r]

∅
M.P. subj

By contrast, Kibort (2007) argues for a rather different account of the passive.
In her view, morphosyntactic operations such as voice alternations do not affect
the selection of argument positions by a predicate, nor the alignment of argu-
ment positions with thematic roles; passivization and similar morphosyntactic
operations affect only the mapping between grammatical functions and argument
positions. Given the Markedness Hierarchy and the Mapping Principle, there is
only one way in which this mapping can be changed: by means of a “mechanism
of increasing markedness,” augmenting (and thus further restricting) the classi-
fication of argument positions prior to the application of the Mapping Principle.
Only the addition of positive specifications, i.e. [+r] or [+o], counts as increasing
markedness. There are a limited number of logical possibilities for the addition of
further specifications to argument positions: a [−r] position can be additionally
specified as [+o], a [−o] position can be additionally specified as [+r], or a [+o]
position can be additionally specified as [+r]. It is not possible for a [−r] position
to be specified as [+r], or for a [−o] position to be specified as [+o], since an
argument position may have only one value for each of these features; in addition,
there is no argument position prespecified as [+r].
342 argument structure and mapping theory

Under Kibort’s analysis, then, passive voice is the result of adding a [+r]
specification to the first argument position of a predicate pre-specified as [−o].
This means that the first argument position of a passive predicate (arg1 ) can
associate only with the oblique grammatical function obl𝜃 , while the second
argument position (arg2 ), if present, is necessarily associated with the subj func-
tion. We therefore have the following argument structure for select in the passive
voice:

(45) Demotion-based account of the passive (Kibort 2007):

agent theme

select ! arg1 arg2 "

[−o] [−r]
passive: [+r]
M.P. obl𝜃 subj

This produces the correct result, namely that the theme argument of passive select
is the subj, as in this f-structure:

(46) Chris was selected by David.

 
’
 
 
 

Kibort (2001; 2004, in particular pages 360–3) provides detailed argumentation

for this demotional approach to the passive. One significant difference between
this and the suppression-based, promotional account of Bresnan and Moshi (1990)
is in the status of the agent argument. Under Bresnan and Moshi’s approach to
passivization, the agent may be expressed only as a modifying adjunct, not as
an oblique argument of the verb, since its suppression leaves it unavailable for
linking to obl𝜃 . Under Kibort’s analysis, the converse is true: the agent argument
is explicitly mapped to obl𝜃 , and so cannot be expressed as an adjunct.11
Kibort (2012) draws out an interesting contrast between the passive and an
apparently similar operation, the anticausative, exemplified in the following two
pairs of English and Polish sentences:

(47) a. I spilled the soup.

b. The soup spilled.

11 Within the context of their own approach to argument structure (as described in §9.5 and
discussed further in §9.10.5), Asudeh and Giorgolo (2012) treat the oblique agent as an obl𝜃 , but state
that their model could easily be reformulated to treat it as an adj.
 impersonal predication 343

(48) a. Tomek wylał zupe˛.

Tomek.nom spilled.3sg.m soup.acc
‘Tomek spilled the/some soup.’
b. Zupa wylała sie˛.
soup.nom spilled.3sg.f refl
‘The soup spilled.’

Kibort argues that voice alternations are meaning preserving, while the anti-
causative is not meaning preserving. In the anticausative, an event affecting a
patient/theme is expressed without the cause of the event being specified. Mor-
phosyntactic operations are necessarily meaning preserving; therefore, the anti-
causative alternation cannot be morphosyntactic, but must be accounted for in
lexical terms. Kibort (2012) consequently argues that two distinct lexical entries
must be assumed for verbs that display the anticausative alternation:

(49) agent patient/theme

(trans) spill ! arg1 arg2 "

[−o] [−r]
M.P. subj obj

(50) agent patient/theme

(anticaus) spill ! arg2 "

[−r]
M.P. subj

In both cases, the verb entails a spiller agent and a spillee patient/theme, but
in the anticausative variant there is no argument position with which the agent
can be associated. If it is expressed, it must be expressed as an adjunct. Thus,
even though the anticausative alternation may initially appear to be very similar
to the passive, Kibort (2012) analyzes the two as being quite different in argument
structure terms.

9.7 Impersonal predication

In English, the passive alternation is restricted to predicates with more than
one argument. In some languages, however, intransitive predicates may undergo
passivization: this may result in an impersonal construction which violates the
Subject Condition, as discussed in §2.1.5. The account of passivization proposed by
Kibort accounts equally well for such impersonal passives. Kibort (2006: example
55) provides the following sentence, which illustrates a passive use of the Polish
intransitive verb palić ‘smoke’.
344 argument structure and mapping theory

(51) Wchodzisz i czujesz, że było palone.

come.in.2sg and smell.2sg that was.3sg.n smoke.ptcp.sg.n
‘You come in and you can smell that there has been smoking (here).’

Active forms of this verb have the following argument structure, with the agent
of the verb realized as its subj:

(52) agent

palić ‘smoke’ ! arg1 "

[−o]
M.P. subj

In the passive, Kibort analyzes the first argument position as being additionally
classified [+r]; it must therefore map to obl𝜃 . Since there is only one argument
position, and this position maps to obl𝜃 , the passive of an intransitive verb does
not have a subj argument, in violation of the Subject Condition:12

(53) agent

palić ‘smoke’ ! arg1 "

[−o]
passive: [+r]
M.P. obl𝜃

This analysis can be contrasted with the analysis of inherently impersonal predi-
cates like English rain under this model:

(54) rain ! arg1 "

[−r]
M.P. subj

This gives the following f-structure:

(55) It rained.

The difference between an inherently impersonal predicate like rain and other
intransitive verbs, such as the one shown in (52), is that rain does not entail any
participants, as (55) shows, i.e. it is not associated with any thematic role.13 Thus,

12 Impersonal constructions are not necessarily uniform in this respect; see, for example, German
impersonal passives as analyzed by Berman (2003), outlined in §2.1.5.
13 For more on intransitive verbs and argument structure, see §9.10.1.
 locative inversion 345

arg1 appears in the subcategorization frame for rain, and maps to a grammatical
function, but there is no thematic role with which it is associated.
Other work on impersonal constructions and expletives within the LFG frame-
work includes Berman (2003: Chapters 4 and 8) on expletive subjects in Ger-
man; Maling and Sigurjónsdóttir (2013) on impersonal passive constructions
in Icelandic; Kelling (2006) on personal and impersonal passive constructions
in Spanish; Rubio Vallejo (2011), also on impersonal constructions in Spanish;
Lødrup (2000) on existential constructions in Norwegian; and Kibort and Maling
(2015) on the distinction between active and passive impersonal constructions.

9.8 Locative inversion

The locative inversion construction is analyzed in detail by Bresnan and Kanerva
(1989) (see also Bresnan 1994) in an exploration of the effect of other levels of
linguistic structure on mapping principles. The locative inversion construction
is a particularly fruitful arena for such study, since locative inversion involves
the interaction of a particular information structural property—presentational
focus—with mapping principles.
Bresnan and Kanerva (1989) examine argument structure alternations with
verbs like Chicheŵa -im- ‘stand’. Example (56) illustrates the uninverted construc-
tion, with a theme subject and a locative oblique argument:14

(56) Nkhandwe y-a-im-a pa-m-chenga.

9.fox 9subj-prf-stand-ind 16-3-sand
‘The fox is standing on the sand.’

Example (57), with a preposed locative and a focused theme argument, illus-
trates the locative inversion construction:

(57) Pa-m-chenga p-a-im-a nkhandwe.

16-3-sand 16subj-prf-stand-ind 9.fox
‘On the sand is standing the fox.’

In the locative inversion construction, as Bresnan and Kanerva (1989) demon-

strate, the noun phrase nkhandwe ‘fox’ bears the obj relation, while the locative
phrase pa-m-chenga ‘on the sand’ bears the subj relation. Evidence for this comes
from patterns of verb agreement: in (56), the verb agrees in noun class with the
subject nkhandwe ‘fox’, which is class 9, whereas in (57), the verb shows class
16 agreement with the locative subject pa-m-chenga ‘on the sand’. Additionally,
Bresnan and Kanerva describe syntactic restrictions on nonfinite verb phrase
modifiers, which may not appear with an overt subject. In verb phrase modifiers
involving locative inversion, it is the locative subject that must be nonovert:

14 Numbers in the glosses indicate the noun class of the arguments.

346 argument structure and mapping theory

(58) m-nkhalangó [m-ó-khál-á mi-ângo]VP

18-9-forest 18subj-inf-live-ind 4-lion
‘in the forest where there live lions’

Constructions involving subject raising in Chicheŵa also provide evidence that the
locative phrase is the subject and the theme phrase is the object in examples like
(57); see Bresnan and Kanerva (1989) for discussion.
Under Kibort’s (2007) analysis, uninverted examples such as (56) can be ana-
lyzed as involving verbs which select for arg1 and arg4 , that is:

(59) theme locative

-im- ‘stand’ ! arg1 arg4 "

[−r] [−o]
M.P. subj obl𝜃

Once again, the mapping from argument positions to grammatical functions

derives without difficulty from the Mapping Principle. In thematic terms, this
verb entails the existence of a stander and a location of standing, or a theme
and a locative, and these associate with the argument positions as shown, in
accordance with the thematic hierarchy.
Kibort treats locative inversion as a morphosyntactic operation which involves
adding the classification [+o] to the first argument of a verb pre-specified as [−r],
restricting it further so that arg1 must map to obj. According to the Mapping
Principle, arg4 , which is intrinsically associated with the classification [−o], maps
to subj in the inverted construction.15 The inverted construction is therefore
analyzed in the following way:

(60) theme locative

-im- ‘stand’ ! arg1 arg4 "

[−r] [−o]
loc. inv.: [+o]
M.P. obj subj

Morimoto (1999) proposes an alternative, Optimality-Theoretic account of argu-

ment mapping in locative inversion and related constructions.

9.9 Complex predicates and argument linking

In early LFG work on argument structure, it was often assumed that argument link-
ing is a process applying to individual words in the lexicon. Butt (1995) and Alsina
(1996) challenge this view by providing evidence that syntactic structures which

15 This explanation for argument linking in locative inversion is distinct from the earlier proposal
of Bresnan and Kanerva (1989), although the intrinsic classifications are the same: on Bresnan and
Kanerva’s analysis, the locative argument is further classified [−r], and so must map to subj, leaving
the theme argument to map to obj.
 complex predicates and argument linking 347

are monoclausal at f-structure can be associated with constructions consisting of

more than one (potentially nonadjacent) word.
Butt (1995) discusses the permissive construction in Urdu, illustrated in (61):

(61) Anjum-ne Saddaf-ko cit..thii likh-ne d-ii

Anjum-erg Saddaf-dat note.nom.f.sg write.inf.obl let-prf.f.sg
‘Anjum let Saddaf write a note.’

The permissive construction consists of two verbs, the main verb (here likhne
‘write’) and the light, auxiliary-like verb de (here dii) ‘let’. Butt shows that at the
level of c-structure, the permissive construction involves either a V* constituent
containing the main and light verbs (62a), or a VP constituent containing the main
verb and the obj argument (62b,c):

(62) a. Anjum-ne Saddaf-ko cit..thii [likh-ne d-ii]V*

Anjum-erg Saddaf-dat note.nom.f.sg write-inf.obl let-prf.f.sg
b. Anjum-ne d-ii Saddaf-ko [cit..thii likh-ne]VP
Anjum-erg let-prf.f.sg Saddaf-dat note.nom.f.sg write-inf.obl
c. Anjum-ne [cit..thii likh-ne]VP Saddaf-ko d-ii
Anjum-erg note.nom.f.sg write-inf.obl Saddaf-dat let-prf.f.sg
‘Anjum let Saddaf write a note.’

Although the main verb/light verb combination need not form a c-structure con-
stituent, and in fact the verbs need not appear adjacent to each other, Butt (1995)
shows that the permissive construction is monoclausal at f-structure, involving
a single complex predicate constructed from the two verbal forms. Evidence for
this comes from verb agreement patterns in Urdu. The Urdu verb agrees with
the nominative argument that is highest on the grammatical function hierarchy;
if none of its arguments are nominative, the verb shows default (third person
masculine singular) agreement. In (61), we see that the light verb dii ‘let’ agrees
with the nominative feminine singular noun cit..thii ‘note’. This shows that cit..thi
‘note’ is a syntactic argument of the main predicate—one which must include
dii—since agreement is possible only between a predicate and one of its syntactic
arguments. Butt provides further evidence from constructions involving control
and anaphora that points to the same conclusion: the verbs likhne ‘write’ and dii
‘let’ combine to form a single syntactic predicate at f-structure, taking a single array
of syntactic arguments.
As Butt’s work makes clear, complex predicate formation—and, therefore, argu-
ment linking—cannot be defined by reference to the argument roles of a single
word or even a single phrase structure constituent; data discussed by Alsina (1996)
point to a similar conclusion. As mentioned above, the theory of argument linking
developed by Bresnan and Kanerva (1989) was originally referred to as lexical
mapping theory, but the work of Butt and Alsina, together with other research,
led to the relabeling of the theory as functional mapping theory or simply mapping
348 argument structure and mapping theory

theory, in acknowledgement of the fact that the theory cannot apply exclusively to
individual words.

9.9.1 Case study: Complex predicates in Urdu

The analysis of complex predicates in Urdu has been the topic of continuing
research in LFG.16 In the rest of this section, we will illustrate the theory of
argument structure and mapping theory in relation to complex predicates using
examples from Urdu, before moving on to consider work on complex predicates
in other languages in §9.9.2.
Based on work by Alsina and Joshi (1991) and Butt (1998), Butt (2014a)
distinguishes two types of causative complex predicates, depending on the type
of argument merger that takes place between the light verb and the lexical verb.
She discusses Urdu permissive constructions, distinguishing the type introduced
in the previous section, which she labels the ‘allow-to-do’ permissive (example 61,
repeated as 63), from the ‘allow-to-happen’ permissive (example 64):

(63) Anjum-ne Saddaf-ko cit..thii likh-ne d-ii

Anjum-erg Saddaf-dat note.nom.f.sg write-inf.obl let-prf.f.sg
‘Anjum let Saddaf write a note.’

(64) kacce lamhe-ko ʃaak=par pak-ne

unripe.m.obl moment.m.sg.obl-acc branch.m.sg=on ripen-inf.obl
d-o
give-imp
‘Let the tender moment ripen on the bough.’

In the ‘allow-to-do’ permissive in (63), argument merger between the two parts
of the complex predicate involves what Butt (2014a) calls “Argument Fusion”: an
argument of the lexical predicate likh ‘write’ (here the participial form likhne) is
coindexed with an argument of the light verb predicate de. In this construction, the
highest argument of the lexical predicate is coindexed with the lowest argument of
the light verb predicate.
The argument structure for the verb likh ‘write’ is shown in (65); (66) illustrates
the use of this verb on its own, as the only verb in the clause:

(65) agent theme

likh ‘write’ ! arg1 arg2 "

[−o] [−r]
M.P. subj obj

16 In addition to the works cited, complex predicates in Urdu have been discussed by Butt and Geuder
(2001), Butt and Ramchand (2005), Ahmed and Butt (2011), Raza (2011), Ahmed et al. (2012), Butt
et al. (2012), Sulger (2012), and Lowe (2015a).
 complex predicates and argument linking 349

(66) Saddaf-ne cit..thii likh-ii

Saddaf-erg note.nom.f.sg write.prf.f.sg
‘Saddaf wrote a note.’

The subcategorization frame for the ‘allow-to-do’ permissive verb de ‘let’ (dii in 63)
can be represented as:

(67) ‘allow-to-do’ de: ‘let!arg1 , arg3 , %pred"’

Here ‘%pred’ in the argument structure for de indicates that this verb, in its
use as a permissive light verb, is incomplete in terms of its subcategorization
frame: it requires another predicate’s argument structure with which to merge. The
argument structure merger of these predicates involves the coindexation of the
lowest argument of the matrix predicate de ‘let’, in this case arg3 , with the highest
argument of the embedded predicate likh ‘write’, in this case (and usually) arg1 .
This results in the following merged argument structure:

(68) de ‘let’ ! arg1 arg3 ‘write’ ! arg1 arg2 ""

[−o] [+o] ([−o]) [−r]
M.P. subj obj𝜃 obj

The coindexed arguments are treated as a single argument for the purposes of
mapping theory: they must be linked to a single grammatical function. Mapping
proceeds entirely in accordance with the Mapping Principle, but it must proceed
according to the hierarchy of arguments as denoted by their indices, and not simply
left-to-right. In this example, the arg2 of the merged predicate is ‘to the right’ of the
arg3 , but according to the hierarchy represented by their indices, it is arg2 , and so
must be mapped first. If arg3 were mapped before arg2 , it would be mapped to
the highest available [+o] argument, which would be obj. But with subj already
linked to arg1 , there would then be no [−r] argument available for arg2 to map
to. If arg2 is mapped first, it is mapped to obj, and subsequently arg3 is mapped to
obj𝜃 . Note that the properties of the fused argument (arg3 of ‘let’ fused with arg1
of ‘write’) derive entirely from the arg3 of the matrix predicate, and not from the
arg1 of the embedded predicate: the fused argument maps to obj𝜃 , even though
this is not compatible with the [−o] classification of the embedded arg1 (hence it
is presented in parentheses).
Butt (2014a) contrasts Argument Fusion with the distinct Argument Raising
construction, which is found with the ‘allow-to-happen’ permissive in Urdu. Butt
argues that the ‘allow-to-happen’ construction, shown in (64), involves a different
subcategorization frame for the light verb de ‘let’:

(69) ‘allow-to-happen’ de: ‘let!arg1 , %pred"’

The two distinct subcategorization frames of de ‘let’ are associated with different
kinds of argument merger: the subcategorization frame in (67) is associated
350 argument structure and mapping theory

with Argument Fusion, while the subcategorization frame in (69) is associated

with Argument Raising. Argument Raising involves the merger of two argument
domains without any coindexation. Thus, the merged argument structure for the
‘allow-to-happen’ complex predicate in (64) is:

(70) de ‘let’ ! arg1 ‘ripen’ ! arg1 ""

[−o] [−r]
M.P. subj obj

In this type of argument merger, the argument structure for the embedded predi-
cate, here pak ‘ripen’, is simply inserted into the slot of the light verb’s argument
structure, with no coindexation between (or fusion of) the arguments of the
two predicates. Mapping then proceeds entirely in accordance with the Mapping
Principle; since both arguments have the same index (i.e. both are arg1 ), the order
of mapping proceeds by left-to-right precedence. That is, the leftmost arg1 must
map first, followed by the embedded arg1 .
As shown in detail by Butt et al. (2010), complex predicates in Urdu can be
recursively embedded, and when this happens we find recursive argument merger.
There are many different light verbs in Urdu besides de ‘let’ which form complex
predicates. One of these is le ‘take’; this is an aspectual light verb, which adds
perfectivity, or a sense of completion, to the event described by the predicate:17

(71) Saddaf-ne cit..thii likh l-ii

Saddaf-erg note.nom.f.sg write take.prf.f.sg
‘Saddaf wrote a note (completely).’

The subcategorization frame for the light verb le ‘take’ is shown in (72); the
argument structure for the complex predicate likh lii ‘wrote (completely)’, which
involves Argument Fusion, is shown in (73):

(72) le: ‘take’ ! arg1 , %pred "

(73) ‘take’ ! arg1 ‘write’ ! arg1 arg2 ""

[−o] ([−o]) [−r]
subj obj

This aspectual complex predicate can be embedded as the predicate argument of

the permissive ‘allow-to-do’ light verb de shown in (67):

(74) Anjum-ne Saddaf-ko cit..thii likh le-ne d-ii

Anjum-erg Saddaf-dat note.nom.f.sg write take-inf.obl let-prf.f.sg
‘Anjum let Saddaf write a note (completely).’

17 The precise semantic contribution of the light verb and how this might be accounted for is
discussed in Butt and Tantos (2004). Butt and Tantos consider the use of Petri Nets (Peterson 1981)
for the representation of lexical meaning in LFG with particular reference to complex predicates.
 complex predicates and argument linking 351

As with complex predicates that show only one level of embedding, this predicate is
a true complex predicate: it is monoclausal at f-structure. The argument structure
for this complex predicate, which involves multiple instances of Argument Fusion,
combining (65) with (67) and (72), is:

(75) ‘let’ ! arg1 arg3 ‘take’ ! arg1 ‘write’ ! arg1 arg2 """
[−o] [+o] ([−o]) ([−o]) [−r]
subj obj𝜃 obj

As one would expect, this complex predicate can be further embedded under
other light verbs; Butt et al. (2010) provide detailed discussion of this recursive
embedding.

An alternative line of analysis regarding complex predicates involves the use of

the restriction operator (§6.9.4). Kaplan and Wedekind (1993) propose a very dif-
ferent analysis of complex predicates in Urdu from the one we have just presented;
their analysis makes use of the restriction operator to provide an alternative lexical
entry for verbs which, like likh ‘write’ in (63), combine with light verbs like de
‘let’ to form a complex predicate. The restriction operator is used in a lexical rule
that provides a new lexical entry for the main verb when it is used as a part of a
complex predicate: the subj of the original lexical entry of a verb like likh appears
as a thematically restricted obj of a new lexical entry. This new lexical item must
be used with light verbs that introduce a new subject, as de ‘let’ does.
Under Kaplan and Wedekind’s proposals, complex predicate formation in Urdu
occurs in the lexicon, not in the syntax. Butt et al. (2003) build on the proposals
of Kaplan and Wedekind (1993), but show that the restriction operator can
be used in f-descriptions in phrase structure rules, enabling complex predicate
formation to be treated syntactically, in line with the argument structure analyses
presented above. The use of the restriction operator is further discussed by Butt
and King (2006c), who show that it can apply both in the syntax and within
the morphological component of grammar, to model both syntactic and lexical
complex predicate formation. The use of the restriction operator for complex
predicates is also discussed by Lowe (2015a: 420–3).

9.9.2 Complex predicates crosslinguistically

The crosslinguistic analysis of complex predicates has been a focus of LFG research
since its earliest days. One of the earliest treatments of complex predicates in an
LFG setting was proposed by Ishikawa (1985), who discusses Japanese morpho-
logical causatives, passives, potentials, and desideratives, and proposes to treat
them as involving a type of raising that can involve the obj as well as the subj
of the complement verb. Ackerman and Webelhuth (1996, 1998) present a theory
of complex predicates in which a single lexical item can be expressed by more than
one morphological word. Matsumoto (1996) also explores the notion of word at
different levels of linguistic representation, providing an illustrative examination
of complex predicates in Japanese; Matsumoto (1998, 2000) examines Japanese
352 argument structure and mapping theory

causatives and proposes a parameter of semantic variation in the typology of

causatives. Ackerman and Moore (2001) discuss argument linking in causatives
and other structures in Finnic. Çetinoğlu et al. (2009) discuss the monoclausal
status of causatives in Turkish, presenting an LFG-based analysis involving com-
plex predication and argument structure merger. Spencer (2013: 286–94) discusses
causative constructions in the context of his own model of argument structure.
Frank (1996) addresses the issue of complex predicate formation in French
and Italian, noting that the assumption that complex predicates are formed in a
similar manner in the two languages does not allow for a revealing analysis of the
differences between them. She proposes a new sort of lexical rule that combines
two verbs to produce a new lexical specification for complex predicates. On this
view, complex predicate formation is lexically specified, though involving possibly
discontinuous expressions consisting of more than one word. By contrast, Falk
(2001b: 114–19) provides a rather more traditional, and influential, analysis of
Romance causatives which captures commonalities in their structure.
Other phenomena involving complex predication which have received LFG
analyses include: serial verbs (Bodomo 1996, 1997; Andrews and Manning 199918),
noun incorporation (Ball 2004; Asudeh and Ball 2005; Asudeh 2007; Duncan
2007; Nordlinger and Sadler 2008; Baker et al. 2010), directional complex pred-
ication in Choctaw (Broadwell 2000a), and a category of complex verbs involving
an “associated path” in some Australian languages (Simpson 2001). Seiss and
Nordlinger (2010) discuss complex predicates in Murrinh-Patha and the interac-
tions of complex predicate formation, applicativization, and reflexivization. Dras
et al. (2012) provide an analysis of complex predicates in Arrernte which relies on
glue semantics (Chapter 8). In a slightly different vein, Arka et al. (2009) provide
a detailed analysis of the applicative construction in Indonesian that makes use of
the restriction operator, and integrates this with an argument structure analysis
involving predicate fusion in the morphology.
The identification of complex predicates, and the distinction between light verbs
and auxiliaries, has also been a topic of investigation. Butt and Ramchand (2005)
discuss the distinction between light verbs and auxiliaries in detail, concentrat-
ing on those that contribute aspectual semantics. Seiss (2009) proposes criteria
distinguishing light verb constructions, serial verb constructions, and auxiliary
constructions. Butt (2010) provides a comprehensive discussion of the identifi-
cation of light verbs and complex predicates crosslinguistically. Butt and Lahiri
(2013) discuss the diachrony of light verbs and complex predicates, with particular
reference to Urdu and Bengali. They draw a clear distinction in diachronic terms
between light verbs and auxiliaries, showing that light verbs are diachronically
relatively stable in comparison to auxiliaries, and are more closely tied to their
corresponding full verb.

18 The approach of Andrews and Manning (1999) is LFG-based but architecturally quite different,
involving structures resembling those used in Head-Driven Phrase Structure Grammar much more
closely.
 grammatical functions 353

9.10 Grammatical functions

In this section, we examine the issue of mapping from the standpoint of gram-
matical functions. This provides us with the opportunity to explore grammatical
features and cross-classification from a slightly different perspective. We also
consider the important issue of gradient distinctions and optionality.

9.10.1 SUBJ
The grammatical function subj is classified as [−o,−r], making it nonobjective
and semantically unrestricted. It is interesting to consider intransitive verbs and
some of the ways in which their subjects may differ, specifically in relation to the
unergative/unaccusative distinction (§9.4.4): the subject of an unergative intran-
sitive verb is agent-like, while the subject of an unaccusative intransitive verb is
patient-like.
Falk (2001b: 111) observes that if a separate level of a-structure is assumed, it
is possible to provide a straightforward account of the unergative/unaccusative
distinction without the need for additional assumptions. Falk illustrates the differ-
ence between unergative and unaccusative verbs in English using the resultative
construction.19 As discussed in §9.4.4, when the verb in a resultative construction
is transitive, the resultative predicate applies to the more patient-like argument and
not the more agent-like argument. This is true in both the active and the passive
voice, whether the more patient-like argument is the obj or the subj:

(76) a. They wiped the table clean.

b. The table was wiped clean.
c. ⃰They wiped the table tired.

Given this orientation to the patient-like argument, it is unsurprising to find that

the subject of an unaccusative verb, which is patient-like, can control a resultative
(77a), but the subject of an unergative verb, which is agent-like, cannot (77b):

(77) a. My son grew tall.

b. ⃰John screamed hoarse.

Kibort (2007) proposes the disjunctive classification [−o] or [−r] for arg1 in her
version of mapping theory in order to account for the unergative/unaccusative
distinction (§9.5). An unaccusative verb requires a [−r] argument, and shares this
classification with the obj of a transitive verb like wipe, whereas an unergative verb
requires a [−o] argument, and shares this classification with the subj of a transitive
verb like wipe. In this way, the grammaticality facts exemplified in (76) and (77) are
captured. Bresnan et al. (2016: Chapter 14) provide further discussion of argument
linking and unaccusativity.

19 On resultative constructions in Norwegian, see Lødrup (2000). Markantonatou and Sadler (1995)
provide an analysis of English resultatives which departs in significant respects from standard assump-
tions of LFG and lexical mapping theory. Bresnan et al. (2016: Chapter 14) provide a brief discussion
of resultatives, unergative intransitive verbs, and “fake” reflexives.
354 argument structure and mapping theory

9.10.2 OBJ
Börjars and Vincent (2008) discuss the grammatical function obj in detail, and
its relation to the more general concept of “object” and the thematic role theme.
They identify many problems in the definition and distinction of obj as a distinct
grammatical category, and argue that current versions of mapping theory which
make use of binary feature decompositions of grammatical functions are inade-
quate when it comes to the specific properties of obj.

9.10.3 OBJ𝜃 and OBL𝜃

A number of authors have addressed the mapping of restricted [+r] arguments
and the crosslinguistically common alternation between oblique arguments and
restricted objects, for instance in the dative alternation and the applicative con-
struction.
To illustrate first with an English example, the dative alternation involves three
argument positions and three thematic roles (agent, theme, benefactive/
recipient). The benefactive/recipient can be expressed either as a PP oblique
argument (78a) or as an object NP (78b):

(78) a. John gave flowers to Mary.

b. John gave Mary flowers.

Within the LFG framework, the dative alternation is discussed by Allen (2001)
and Kibort (2008); Her (1999) examines the dative alternation in Chinese in the
context of a new set of mapping proposals. Under the approach that we have
adopted, following Kibort (2008), the dative alternation involves a pre-syntactic
realignment of two of the participants in the event. In the construction illustrated
in (78a), including the PP oblique argument which we identify as oblgoal , the
argument position associated with the benefactive/recipient argument is arg4 ;
this is based on the set of semantic entailments invoked by the predicate in relation
to this participant in the event. According to Kibort’s theory, arg4 is intrinsically
associated with [−o]. Kibort proposes that the dative alternation involves the
“remapping” of the benefactive/recipient to primary object argument position
(arg2 ) plus the “downgrading” of the theme to secondary object argument position
(arg3 ), as shown in examples (79) and (80). In the following examples, b represents
the participant with the most proto-beneficiary/recipient properties:

(79) Ditransitive verb with oblique argument benefactive: John gave flowers to
Mary.
x y b

give1 ! arg1 arg2 arg4 "

[−o] [−r] [−o]
M.P. subj obj oblgoal
 grammatical functions 355

(80) Ditransitive verb with shifted dative benefactive: John gave Mary flowers.
x b y

give2 ! arg1 arg2 arg3 "

[−o] [−r] [+o]
M.P. subj obj objtheme

The alternation between obliques and restricted objects is also discussed by

Ackerman and Moore (2011), while the dative alternation has been the subject
of much work by Joan Bresnan and colleagues; see, for example, Bresnan et al.
(2007a), Bresnan (2007), Bresnan and Hay (2008), Bresnan and Ford (2010),
Bresnan and Nikitina (2010), Kendall et al. (2011), de Marneffe et al. (2012), and
Ford and Bresnan (2013).
Applicative constructions, which involve the addition of an argument to a
predicate’s subcategorization frame, are discussed by Alsina and Mchombo (1990),
Bresnan and Moshi (1990), Kibort (2007, 2008), Arka et al. (2009), and Ackerman
and Moore (2011, 2013). The benefactive applicative construction in Chicheŵa is
illustrated in (81b). In each of the sentences in (81), the argument that immediately
follows the verb is the primary (nonrestricted) object:

(81) a. Non-applied transitive construction:

nkhandwe zi-ku-mény-á njōvu
10.foxes 10subj-prs-hit-fv 9.elephant
‘The foxes are hitting the elephant.’
b. Benefactive applicative construction:
nkhandwe zi-ku-mény-ér-a aná njōvu
10.foxes 10subj-prs-hit-appl-fv 2.children 9.elephant
‘The foxes are hitting the elephant for the children.’

Kibort (2008) analyzes the benefactive applicative construction as involving the

addition of an argument pre-specified as [+o] to the valency frame of the predicate
in question, resulting in an increase in valence. Note that while (83) and the
analysis provided for the construction that includes the English shifted dative
benefactive in (80) are the same, it is only the applicative construction that is
morphologically marked:

(82) Non-applied transitive construction:

x y

hit ! arg1 arg2 "

[−o] [−r]
M.P. subj obj
356 argument structure and mapping theory

(83) Benefactive applicative construction:

x b y

(applicative) hit ! arg1 arg2 arg3 "

[−o] [−r] [+o]
M.P. subj obj objtheme

In the context of the applicative construction, a number of authors have discussed

the phenomenon of “object asymmetries” and the distinction between symmetrical
and asymmetrical object languages. Bresnan and Moshi (1990) show in detail
that there is a distinction between “asymmetrical object type” languages, in which
only one of a verb’s complements can exhibit primary object properties like
passivizability, and “symmetrical object type” languages, in which more than one
of the verb’s complements is potentially able to display such properties. In some
symmetrical languages, certain subclasses of objects can show asymmetric patterns
due to factors such as person and animacy. Bresnan and Moshi (1990), building on
an analysis proposed by Alsina and Mchombo (1993) for Chicheŵa, argue that a
single parameter can account for the difference between symmetrical and asym-
metrical languages: the “Asymmetrical Object Parameter” (AOP), which holds in
asymmetrical languages such as Chicheŵa. The AOP is a constraint on intrinsic
classifications, such that only one argument position may be intrinsically classified
as unrestricted. This means, for example, that at most one argument of a predicate
is available to become the subject of a passive construction. The AOP is also
discussed by Falk (2001b: 111–14). Kibort (2008) claims that object asymmetries
are easily handled in her model of argument structure, since according to the
universal valency template in (37), only one argument position is classified as [−r].

9.10.4 COMP and XCOMP

There has been comparatively little work on mapping of the grammatical functions
comp and xcomp. Zaenen and Engdahl (1994) were among the first to propose a
detailed theory of mapping to comp and xcomp, assuming that these arguments
bear the thematic role proposition and are intrinsically associated with the [−o]
feature.20 Butt (1995) and Alsina (1996) also discuss the application of linking
theory to xcomp, comp, and raising verbs.
It is worth noting that clausal arguments need not map onto comp and xcomp.
As discussed in §2.1.7, some clausal arguments have been analyzed as obj, obj𝜃 ,
or obl𝜃 . Arka and Simpson (2008) argue on the basis of control phenomena in
Balinese that complex clausal arguments can map to subj, obj, or obl𝜃 . Rákosi
and Laczkó (2005) make the same point for clausal arguments in Hungarian.

20 Zaenen and Engdahl (1994) also discuss argument mapping and nonthematic or semantically
empty arguments, which they claim are present at argument structure with the intrinsic feature [−r];
see Bresnan et al. (2016: 332) for additional discussion.
 grammatical functions 357

9.10.5 Gradient distinctions and optionality

Early models of argument structure within LFG were based on the assumption that
the distinctions between argument and adjunct, on the one hand, and between
core (subj, obj, obj𝜃 ) and oblique (obl𝜃 ) arguments, on the other, are absolute
distinctions which can be made unambiguously in any particular context. How-
ever, a growing body of work within LFG has sought to address phenomena which
point instead toward a gradient distinction between arguments and adjuncts, and
between core and oblique arguments.

9.10.5.1 The core-oblique distinction Arka (2005) discusses the nature of what
he labels the “core-oblique” distinction in Austronesian languages spoken in
Indonesia, and particularly the obj/obl𝜃 distinction. Arka argues in detail that
there is no absolute distinction between obj and obl𝜃 in some languages. He
proposes a “core index,” based on a set of language-specific syntactic properties
characteristic of core arguments. The index ranges from 1.00 (“highly core”) to
0.00 (“highly oblique”). Arka shows that in Indonesian and Balinese the core index
of core arguments can vary, but it is always over 0.60, while the core index of
typical obliques is between 0.00 and 0.10. However, certain arguments of verbs
in Indonesian have a core index around 0.50. This is the case, for example, with
the stimulus NP argument of Indonesian suka ‘like’, which has a core index of
0.54. The puzzle of how to classify this stimulus argument is also discussed by
Musgrave (2001, 2008), who argues that this argument cannot be classified as
either subj or obj. The best analysis within the inventory of LFG’s grammatical
functions is, Musgrave argues, obj𝜃 . Arka (2005) rejects this proposal on the
basis that the stimulus argument does not show sufficient core properties. Both
Arka (2005) and Musgrave (2008) consider the possibility of adding to LFG’s
inventory of grammatical functions a ‘semi-obj’ function, but they both dismiss
this as a viable option given that such a class would be only negatively defined
and not necessarily coherent. Arka (2005) argues that the best solution may be to
reanalyze the distinctions between grammatical functions so that only subject and
complement are differentiated.

9.10.5.2 Optional dependents: Derived arguments, thematic adjuncts Another

distinction which appears to be problematic when conceived of as binary
and discrete is the argument/adjunct distinction. Key to this issue are those
constituents whose properties indicate that they do not belong unambiguously
to either category. For instance, Rákosi (2006a,b) discusses the status of so-called
“circumstantial” dependents, such as comitatives, instrumentals, and benefactives,
which are realized as PPs in English. Within LFG, such dependents are usually
treated as arguments. However, they are different from prototypical arguments
in a number of ways, most obviously in the fact that they are optional. In the
sentences in (84), the subject and object arguments cannot be omitted, whereas in
each case the PP dependent is optional:

(84) a. David selected Chris (for the first team).

b. The masked man assassinated the president (with a knife).
358 argument structure and mapping theory

The standard LFG approach to such dependents, following Bresnan (1982d), is

to treat them as arguments, and to treat their optionality by means of a lexical
rule which creates new verb forms by the addition of an argument to the lexical
representation. The lexical rule Instrumentalization, for example, derives the verb
in example (85b) from the one in (85a) by adding an instrumental argument:

(85) a. assassinate ! agent, patient "

b. assassinate ! agent, patient, instrument "

The version of (84b) which lacks the PP contains the verb in (85a), while the version
of (84b) which includes the PP contains the related but distinct verb in (85b). The
optionality of such dependents is therefore not due to any essential difference in
status between them and the subj or obj arguments of the same predicates: when
the PP is present it is as much an argument, and no less obligatory, than the subject
and object arguments. The PP’s optionality is only apparent, due to the existence
of two distinct lexical entries for the governing verb, one of which selects for an
obligatory instrument argument, the other of which does not.
However, Rákosi (2006a,b) shows that a more fine-grained approach to optional
dependents is required in order to capture the full range of variation in the status
of such dependents. In particular, some such optional dependents are, in semantic
terms, entailed by the meaning of the predicate, and so must form part of the lexical
conceptual structure of the verb. Rákosi (2006b) contrasts English assassinate with
a verb like drill.21 Superficially, the instrumental dependent of drill appears to be
just like the instrumental dependent of assassinate, since it is optional:

(86) a. David drilled the hole (with the hand drill).

b. The masked man assassinated the president (with the knife).

However, there is a difference between the two: assassinate does not necessarily
entail the existence of an instrument, whereas drill does. Rákosi (2006b) claims
that “one can in principle assassinate the president simply by jumping on him/her,”
while it is not possible to drill something without an instrument. This difference
is reflected grammatically since, as noted by Reinhart (2002), there exists a variant
of verbs like drill in which the instrument functions as the subject, whereas there
exists no such variant for verbs like assassinate:

(87) a. The hand drill drilled the hole.

b. ⃰The knife assassinated the president.

On the basis of such observations, Rákosi argues that the optional dependents of
verbs like assassinate are intermediate, in descriptive terms, between arguments

21 In fact, Rákosi (2006b) bases his discussion on the verb peel, following Reinhart (2002) in
classifying this verb as entailing the existence of an instrument. However, it is possible to peel an orange
or tangerine without the aid of a peeling instrument; in fact, peel patterns with verbs like eat in allowing
but not requiring the presence of an instrument.
 grammatical functions 359

proper and adjuncts proper. In formal terms, however, Rákosi seeks to maintain
the absolute distinction between arguments and adjuncts which is fundamental to
the treatment of grammatical functions in LFG. In order to resolve this paradox,
Rákosi argues for a class of thematic adjuncts (adj𝜃 ), which are adjuncts in syntactic
terms but which bear a thematic relation to the predicate, unlike “pure” adjuncts.22
A particular subset of the constituents which Rákosi (2006a,b) treats as thematic
adjuncts, namely optional to-PPs with verbs like seem, receives a different analysis
from Asudeh and Toivonen (2007, 2012). Asudeh and Toivonen refer to these PP
experiencers as instantiating a semantic role of pgoal (goal of perception). They
argue that such dependents should be treated as “pure” adjuncts, but ones which
bear a semantic relation to the predicate without bearing a relation in syntactic or
argument structure terms.
As mentioned above, the standard LFG analysis of the type of optional depen-
dents that Rákosi (2006a,b) analyzes as thematic adjuncts is, following Bresnan
(1982d), to treat them as arguments. This approach is explored further by Need-
ham and Toivonen (2011), who discuss a variety of optional dependents whose
status is, they argue, intermediate between true argument and true adjunct. They
present a number of argumenthood tests, showing that several types of optional
dependents display inconsistent properties. These include the constituents in
parentheses in the following examples: the demoted subject in a passive (88a),
possessive phrases with event nominals (88b), benefactives (88c), “displaced”
themes (88d), instruments (88e), experiencers (88f), and directionals (88g):

(88) a. The letter was written (by David).

b. David’s writing of the letter / The writing of the letter
c. David cooked a meal (for Chris).
d. David loaded the wagon (with hay).
e. David cut up the apple (with a knife).
f. It looks (to me) like it’s going to rain.
g. David arrived (from London) yesterday.

In clarifying their concept of argumenthood, Needham and Toivonen (2011)

address the problem of semantic entailment, i.e. whether the existence of a par-
ticular argument role is entailed by a predicate. They point out (as also observed
in Chapter 2) that it is not, in fact, helpful to think of arguments as “entailed” or
“semantically necessary,” because the existence of a time and place at which an
event occurs is necessarily entailed by every verb, yet such concepts are almost uni-
versally expressed using adjunct phrases. What matters instead is verb specificity:
arguments are “semantically distinctive,” that is, they are associated with particular
classes of verbs, and thus serve to distinguish those classes from other classes. Time
and place are therefore not likely to be expressed as arguments, because they are

22 For a similar treatment motivated primarily by implementational concerns, see Zaenen and
Crouch (2009).
360 argument structure and mapping theory

relevant to the majority of verbs and so do not generally distinguish between verb
classes.
With this definition of argument in place, Needham and Toivonen (2011) pursue
a mapping-theoretic analysis of optional dependents which is similar in spirit to
the proposals of Bresnan (1982d) in treating them as derived arguments. In the
case of instrumentals, for example, they provide a lexical rule which is roughly
equivalent to Bresnan’s Instrumentalization, exemplified in (85). Reformulated in
terms of the approach that we have been using, following Kibort’s work, the rule
would look like this:

(89) Optionally add the following argument to verbs whose first argument is an
agent:
y

arg4
[−o]
M.P. oblinstr

Similar optional lexical rules are proposed for the other optional dependents
exemplified in (88).
The argument-adjunct distinction continues to be a focus of research within
LFG; see, for example, Rákosi (2013), Toivonen (2013), and Przepiórkowski
(2017). To date, such research has overwhelmingly focused on data from English.
As Needham and Toivonen (2011) observe, a fuller understanding of this issue is
likely to be gained in light of data from a wider range of languages.
Building on the derived argument approach of Needham and Toivonen (2011),
Asudeh and Giorgolo (2012) propose an integrated syntactic and semantic analysis
of the argument-adjunct distinction.23 They argue that the standard LFG approach,
involving distinct lexical entries for verbs that have optional dependents, fails to
capture the intuition that fundamentally the same verb is involved in both of the
following sentences, for example:

(90) a. David ate.

b. David ate cheese.

Asudeh and Giorgolo (2012) propose to capture such patterns by using optional
f-descriptions within a single lexical entry. In presenting their analysis, Asudeh
and Giorgolo contrast their model specifically with that of Butt et al. (1997),
presented in §9.2, according to which argument structure is interpolated between
c-structure and f-structure. Asudeh and Giorgolo argue that such an approach
could deal with optional dependents only if a further correspondence function
were assumed between a-structure and s-structure. That is because the patient
argument of a verb like eat, for example, must be present at argument structure and

23 For an alternative implementation of the same proposal using monads, building on Giorgolo and
Asudeh (2011a), see Giorgolo and Asudeh (2012b).
 grammatical functions 361

semantic structure even if it is not realized in the c-structure and does not appear
in the f-structure. Therefore, alongside the decomposition of the 𝜙 function into
a function α (from c-structure to a-structure) and a function λ (from a-structure
to f-structure), it would also be necessary to assume a function, which Asudeh
and Giorgolo (2012) label 𝜃, from a-structure to s-structure. On this view, 𝜃 is the
composition of λ with 𝜎, as shown in example (91):

(91) Necessary adjustment to Butt et al.’s proposal, considered and rejected by

Asudeh and Giorgolo (2012):
IP

(↑ subj) = ↓ ↑=↓
NP I*

↑ =↓
Kim VP

↑ =↓
V*

α ↑ =↓ ↓ ∈ (↑ adj)
V* PP

α ↑ =↓
V at noon

 ate  𝜎 ! "
 rel eat  pred ‘eat’
 subj ) * e event ev[ ]
 pred ‘Kim’   𝜎
 agent [ ]  λ   k[ ]
patient [ ]  adj { [ “at noon” ] } 
tense pst p[ ]
𝜃
𝜃
As discussed in §9.5, Asudeh and Giorgolo (2012) propose an alternative, archi-
tecturally simpler model in which no distinct a-structure and no additional corre-
spondence functions are required. Under this approach, grammatical functions at
f-structure are directly associated with s-structure attributes in the lexical entries
of predicates. Lexical entries also include an obligatory meaning constructor that
represents the essential meaning of a predicate; additional optional meaning con-
structors can introduce or existentially quantify optional arguments. For example,
Asudeh and Giorgolo (2012) provide the following lexical entry for English ate,
modified in accordance with the version of mapping theory proposed by Findlay
(2014, 2016) (see also Asudeh et al. 2014):
362 argument structure and mapping theory

(92) ate V (↑ pred) = ‘eat’

(↑ tense) = pst

(↑ subj)𝜎 = (↑𝜎 arg1 )

{(↑ obj)𝜎 = (↑𝜎 arg2 ) | (↑𝜎 arg2 )𝜎−1 = ∅}

λx.λy.λe.eat (e) ∧ agent (e) = x ∧ patient (e) = y :

(↑𝜎 arg1 ) ⊸ [(↑𝜎 arg2 ) ⊸ [(↑𝜎 event) ⊸ ↑𝜎 ]]
+ ,
λP.λe.∃x.P(e, x) :
[(↑𝜎 arg2 ) ⊸ (↑𝜎 event) ⊸ ↑𝜎 ] ⊸ [(↑𝜎 event) ⊸ ↑𝜎 ]

Line 3 of the lexical entry in (92) associates the s-structure of the subj with the
s-structure attribute arg1 . Line 4 ensures that arg2 is mapped appropriately unless
it is unrealized. The rest of the lexical entry comprises meaning constructors.
The first meaning constructor introduces the meaning eat of the verb and its
agent and patient, and is obligatory. The second meaning constructor (enclosed
in parentheses) is optional, and existentially quantifies over the patient argument,
allowing the verb to appear with only the agent expressed. These f-descriptions
license the analyses shown in (93) and (94), which capture the two different uses
of the optionally transitive verb eat:

(93) F-structure and s-structure for David ate cheese, Asudeh and Giorgolo
(2012):
   
   
   
   
   1 [] 
 
2 []

(94) F-structure and s-structure for David ate, Asudeh and Giorgolo (2012):
 
 
 
   
   
 1 [] 

2 []

Under these proposals, the verb eat in English has two arguments at s-structure,
regardless of whether the object argument is realized at c-structure and f-structure;
it is realized syntactically in (93), but not in (94). Importantly, we are dealing with
only one lexical entry—only one verb eat—whether or not the optional object is
present.
Asudeh and Giorgolo (2012) also provide an analysis for instrumental with-
phrases, based on the derived argument approach of Needham and Toivonen
(2011). They propose that argument addition is best captured by associating the
 further reading and related issues 363

relevant lexical information with the entry of the instrumental preposition with,
as in (95), rather than the verb:

(95) with P (↑ pred) = ‘with’

(↑ obj)𝜎 = ((obl ↑)𝜎 instrument)

λy.λP.λx.λe.[P(x)(e) ∧ animate (x) ∧ instrument (e) = y] :

(↑ obj)𝜎 ⊸
[((obl ↑) subj)𝜎 ⊸ ((obl ↑)𝜎 event) ⊸ (obl ↑)𝜎 ] ⊸
((obl ↑) subj)𝜎 ⊸ ((obl ↑)𝜎 event) ⊸ (obl ↑)𝜎

This lexical entry not only includes the information that the derived argument is an
instrument, but also requires the subj to be animate, capturing the generalization
that instrumental with-phrases are only possible with “agent verbs” (Reinhart
2002). Example (96) shows the f-structure and s-structure for the sentence David
peeled the apple with the knife (example 86), based on Asudeh and Giorgolo’s (2012)
proposal:

(96)  
   
 
 
 +   
   
   
   
    
   
  
 + 
  

In addition to optional transitives and instrumental with-phrases, Asudeh and

Giorgolo (2012) provide an analysis of passive constructions with and without a
by-phrase, and consider the wider implications of this approach for the argument-
adjunct distinction and the LFG architecture as a whole.

9.11 Further reading and related issues

The discussion in this chapter cannot do justice to the large volume of literature
on argument structure and its place in the LFG architecture. For more on the
nature of and motivation for argument structure, see Simpson (1991: Chapter 1),
who presents an interesting discussion of thematic roles, syntactic functions, and
the regularities that hold between syntax, semantics, and argument structure. Falk
(2001b) provides a useful overview, including a very interesting discussion of link-
ing in ditransitive verbs. Theories of argument linking and argument structure and
their development within LFG are discussed in detail by Butt (2006: Chapter 5).
364 argument structure and mapping theory

Bresnan et al. (2016: Chapter 14) also provide an overview discussion of argument
structure and mapping theory, discussing a range of alternative approaches and
providing pointers to related literature.
The concept of argument structure, and the specific proposals of mapping
theory, have been subject to criticism, for example by Davis and Koenig (2000) and
Levin and Rappaport Hovav (2005); their arguments are addressed and countered
by, among others, Ackerman and Moore (2013) and Spencer (2013: Chapter 7).
As mentioned earlier, a number of alternative proposals have been made regard-
ing the theory of argument mapping and argument classification. Alsina (1996)
argues for a very different feature decomposition of grammatical functions, mak-
ing use of the features +subj, indicating whether or not the argument is “subject-
like,” and +obl, indicating whether the argument is a direct (−obl) or an oblique
argument. Alsina also eliminates the Function-Argument Biuniqueness condition,
arguing that its elimination allows the correct treatment of, among other construc-
tions, reflexive clitics in some Romance languages.
Rákosi (2006a,b) proposes a different feature decomposition of grammatical
functions, based on Reinhart’s (2002) Theta System decomposition of thematic
roles. Rákosi makes use of Reinhart’s binary features +c(ause) and +m(entally
involved). In this system, absence of a feature differs from a minus value for that
feature. For example, the [+c,+m] role corresponds to the thematic role of agent,
the [+c,−m] role corresponds to the thematic role of instrument, and the [+c]
role corresponds to a thematic role cause.
Linking in nominal predicates has been a subject of interest since relatively early
in the theory’s development, and continues to be a focus of research. Iida (1987)
and Saiki (1987) discuss deverbal nominals in Japanese, the realization of gram-
matical functions in nominals, and the role of argument structure; Markantonatou
(1992, 1995) presents an analysis of linking in deverbal nominals in Modern Greek.
Laczkó (1995, 2000) presents a detailed analysis of the syntax of Hungarian noun
phrases, including a theory of linking of nominal arguments. An alternative theory
of linking for nominal arguments is provided by Kelling (2003).
Another proposal for linking nominal arguments is made by Chisarik and Payne
(2003). They argue for a third binary feature [±d] “discourse-related,” alongside
[±r] and [±o]. This feature serves to distinguish two [−o], [−r] functions: subj
and “adnom.” Both of these functions can be governed by nominals, but adnom
cannot be governed by verbs.
A topic of debate in the literature concerns whether nominals can select for
unrestricted [−r] functions, or only restricted [+r] functions. Rappaport (1983)
and Kelling (2003) argue that nominals can govern only restricted functions.
Markantonatou (1995) assumes that nominals can select for both restricted and
unrestricted functions. According to Laczkó (2000), deverbal nominals can select
for both restricted and unrestricted functions, with poss being an unrestricted
function. Laczkó’s research also includes work on event nominalizations (Laczkó
2003, 2013). The selectional possibilities for adjectives are discussed by Vincent
and Börjars (2010a), while subcategorization by both nouns and adjectives is
discussed by Lowe (2013, 2017b). Spencer (2013: Chapter 7) develops a the-
ory of argument structure representation which is primarily designed to model
 further reading and related issues 365

attributive modification; he also briefly discusses adverbial and prepositional

argument structure.
A variety of other phenomena with interesting argument structure properties
have been discussed within the context of mapping theory. That some two-
argument verbs can select for a subj and an obj𝜃 , rather than the usual subj
and obj, is discussed by Çetinoğlu and Butt (2008), Dahlstrom (2009, 2013),
and Dalrymple and Nikolaeva (2011). Valency alternations in Balinese, which
exhibits a partially symmetrical voice system, are discussed by Arka (2003) and
Arka and Simpson (2008). Arnold (1994) provides an analysis of inverse voice in
Mapudungan.
In his work on argument alternations, Findlay (2014, 2016) develops an account
within the framework of Asudeh and Giorgolo (2012) which utilizes the argument
structure model proposed by Kibort (2001, 2004, 2006, 2007, 2008), drawing
together these two approaches to argument mapping and subcategorization.
There is a small body of work within LFG’s mapping theory that has been
conducted within an Optimality-Theoretic framework (Prince and Smolensky
2004): Lødrup (1999b) provides an analysis of the Norwegian presentational focus
construction; Asudeh (2001) proposes an account of argument linking in Marathi,
noting some consequences for Dowty’s (1991) theory of proto-roles; Morimoto
(1999) presents an analysis of locative inversion and argument reversal, and
Morimoto (2000) continues this work. Optimality-Theoretic approaches to LFG
more generally are briefly discussed in §18.1.3.
 10
Information structure

We now consider how information is organized, or structured, within an utterance.

As with semantic information, early work in LFG represented certain aspects
of information packaging by means of f-structure attributes. Subsequent work
assumes a separate level of information structure, related to other structures via
correspondence functions. In this chapter, we begin by discussing how the infor-
mation conveyed by an utterance is structured or packaged to facilitate communi-
cation, and we explore the nature of the units that are relevant to the structuring
of information. We then review some early LFG approaches to the representation
of information structural features, followed by an overview of the model of
information structure that we adopt in the rest of this book.

10.1 Structuring information

The main function of language is the exchange of information between par-
ticipants. Information structure is the level of sentence organization that repre-
sents how sentences are structured in a particular context in order to facilitate
information exchange.1 All languages make it possible to facilitate information
exchange by, for example, highlighting certain pieces of information as particularly
relevant or important, or by distinguishing pieces of information that are entirely
new in the discourse from pieces of information that are “old,” corresponding
to information already present in the relevant context or discourse. As a result,
the same proposition can be expressed by sentences with different structures.
For instance, the Russian sentences in (1) (Comrie 1987: 95) encode the same
proposition and are truth-conditionally equivalent, regardless of the difference
in word order. In both cases, the individual called Maxim defends the individual
called Victor:

1 There are different uses of the term “information structure” in the literature. O’Connor (2006)
uses i(nformation)-structure as a term for relations between multiple levels, including prosody and
semantics, and refers to the level at which information is organized within an utterance as d(iscourse)-
structure. Here, we use the more standard term “information structure” (or i-structure) for this level,
and the term “discourse structure” rather for the relations between successive utterances in a discourse,
following King and Zaenen (2004). We will not have anything detailed to say on discourse structure in
this book.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 structuring information 367

(1) a. Viktor-a zaščiščaet Maksim.

Victor-acc defends Maxim.nom
‘Maxim defends Victor.’
b. Maksim zaščiščaet Viktor-a.
Maxim.nom defends Victor-acc
‘Maxim defends Victor.’

However, these sentences are not interchangeable: their acceptability is dependent

on the precise discourse context. In the context provided in (2a), (1a) is the most
natural way to answer a question about the subject of the sentence (the defender);
in the context provided in (2b), (1b) is the most natural way to answer a question
about the object of the sentence:

(2) a. Q: Kto zaščiščaet Viktor-a?

who.nom defends Victor-acc
‘Who defends Victor?’
A: Viktor-a zaščiščaet Maksim.
Victor-acc defends Maxim.nom
‘Maxim defends Victor.’
b. Q: Kogo zaščiščaet Maksim?
who.acc defends Maxim.nom
‘Who does Maxim defend?’
A: Maksim zaščiščaet Viktor-a.
Maxim.nom defends Victor-acc
‘Maxim defends Victor.’

Russian has a high degree of word order flexibility, meaning that constituents
can appear in different orders for the purpose of encoding information structure
distinctions. Another possibility is for a language to use specific constructions
which can be used to “package” information. For instance, in English a cleft
construction can be used as an answer to a constituent question when the clefted
element is the “answer” constituent; in (3) A is a discourse felicitous response, and
A! is infelicitous in the context provided.2 (A! would, on the other hand, be an
acceptable response to the question Who hates beans? while A would not be.)

(3) Q: What does David hate?

A: It’s beans that David/he hates.
A : #It’s David that hates beans.
!

Of course, syntax is not the only means by which information structure can be
encoded crosslinguistically. Morphology and prosody may also be used to package

2 The symbol # indicates semantic or pragmatic unacceptability, in the same way as the asterisk ⃰
indicates syntactic ill-formedness.
368 information structure

information, either on their own or in combination. An example of how morphol-

ogy can encode information structure status in some languages is differential object
marking; Dalrymple and Nikolaeva (2011) provide an LFG analysis which makes
an explicit link between object marking and information structure in a number of
languages. With respect to prosody, the location of the main stress in an English
sentence (indicated by bold face in the following example) in part depends on
information structure, as the infelicity of A! in response to this particular question
illustrates:

(4) Q: Who bought flowers?

A: Lily bought flowers.
A! : #Lily bought flowers.

We provide a detailed discussion of the relationship between prosody and informa-

tion structure in §11.6. In this chapter, we assume “neutral” intonation and confine
ourselves to the syntactic encoding of information structure.
Crosslinguistically, languages differ in the extent to which syntax encodes infor-
mation structural distinctions. At one end of the spectrum are nonconfigurational
languages. These are languages whose word order appears flexible if one considers
only grammatical functions such as subject and object, but where word order is in
fact often determined on the basis of information structure relations. For example,
as discussed in §3.4.2, Warlpiri is a nonconfigurational language: beyond requiring
that the auxiliary verb occupy a particular syntactic position, no purely syntactic
generalization can adequately account for word ordering in Warlpiri. A configu-
rational language like English is fundamentally different in this respect: phrase
structure positions in English tend to be associated with particular grammatical
functions, and word order is thus highly restricted in this respect. For instance, the
grammatical subject in an English sentence appears in the specifier of IP, regardless
of its information structure status. The answer A in (5), with David in the specifier
of IP, is an acceptable response to either Q or Q! :

(5) Q: Who selected Chris?

Q! : Who did David select?
A: David selected Chris.

Another possibility attested in many languages, particularly those with rich

morphological agreement systems, is discourse configurationality. In a discourse-
configurational language, phrase structure and word order are determined not by
grammatical function but by information structure status. As in a configurational
language (or more accurately a grammatical function configurational language),
therefore, word order is highly restricted. The key difference between these two
types of configurational language is whether word order is determined on the
basis of information structure categories such as topic and focus (which we will
define shortly) or grammatical functions such as subject and object. Snijders
(2015) claims that word order in all languages is constrained either syntactically
 the categories of information structure 369

(in terms of grammatical functions) or by information structure relations; that is,

all languages which have been classified as nonconfigurational, such as Warlpiri,
are in fact discourse-configurational.
Hungarian is a discourse-configurational language. In Hungarian, the con-
stituent which supplies the missing information elicited by a question (the focus
of the answer sentence) occupies the immediately preverbal position, regardless of
its grammatical function:

(6) a. Q: János ki-t hívott fel?

John.nom who-acc called vm
‘Who did John call?’
A: János [Mari-t]focus hívta fel.
John.nom Mary-acc called vm
‘John called Mary.’
b. Q: Mari-t ki hívta fel?
Mary-acc who.nom called vm
‘Who called Mary?’
A: Mari-t [János]focus hívta fel.
Mary-acc John.nom called vm
‘John called Mary.’

A language may be discourse configurational to a greater or lesser degree, making

it difficult to establish a strict division between configurational languages and
discourse-configurational languages. For example, Catalan (Vallduví 1992) is a
language which is discourse configurational only with respect to certain types
of topic; focus, by contrast, is not associated with a particular phrase structure
position (Erteschik-Shir 2007: 85).
Before we consider how the relationship between syntax and information struc-
ture has been analyzed within the LFG framework, it is necessary to negotiate
the “terminological minefield” (Vallduví and Vilkuna 1998: 80) that is a feature of
research on pragmatics and information structure. In the next section, we define
the key concepts that underlie the definitions of information structure units that
have generally been adopted in LFG analyses.

10.2 The categories of information structure

There are many different theories of the ways that information is structured in
language, with studies of the relation between syntax and information structure
stretching back at least to the eleventh-century Arabic grammarian Al-Jurjani
(see Owens 1986, among others). Perhaps the most widely known information
structure distinction, due to Weil (1844), is between “Topic” or “Theme”—what
the sentence is about, the “point of departure”—and “Comment” or “Rheme,” what
is stated about the topic. Though the precise definitions of these terms may vary,
this fundamental bifurcation in one way or another underlies much subsequent
work on the categories of information structure, including important work within
370 information structure

the Prague School tradition (for example Firbas 1964; Sgall 1967; Daneš 1974;
Mathesius 1983; Hajičová et al. 1998) and beyond (for example, Bolinger 1965;
Halliday 1967; Dahl 1969; Kuno 1972; Gundel 1974).3 To give an example, the
answer to the question translated as Who did John call? in the Hungarian dialogue
in (6a), repeated here as (7), can be divided into Topic and Comment as shown:

(7) János Mari-t hívta fel.

John.nom Mary-acc called vm
! "# $ ! "# $
topic comment
‘John called Mary.’

Another key binary distinction made in the information structure literature is

between “Background” or “Presupposition” and “Focus” (for example, Chomsky
1971; Jackendoff 1972; Prince 1981; Krifka 1991; Steedman 1991; Lambrecht
1994). This distinction relates to whether the information in question can be
assumed to be shared by the interlocutors (Background) or not (Focus). Thus, as
with the Topic/Comment distinction, the fundamental insight is that information
is structured in such a way as to relate the utterance to the wider discourse context
and to satisfy the communicative requirements of the discourse participants.
The information expressed in the Hungarian example (7) can be divided on the
basis of this distinction as well, resulting in a different division to the previous
Topic/Comment one:

(8) János Mari-t hívta fel.

John.nom Mary-acc called vm
! "# $ ! "# $ ! "# $
background focus background
‘John called Mary.’

The categories of topic and focus have been of particular importance in the analysis
of information structure in the LFG literature. Based on much previous research,
including Gundel (1974), Reinhart (1981), and Lambrecht (1994), we understand
topic to be the entity or entities that the proposition is about. We understand
focus to be the informationally unpredictable part of the proposition, that which
is informative or contrary to expectation (Vallduví 1992; Vallduví and Engdahl
1996), or in Lambrechtian terms, the semantic component of a pragmatically
structured proposition whereby the assertion differs from the presupposition
(Lambrecht 1994). For more on the definitions of topic and focus in an LFG setting,
see Dalrymple and Nikolaeva (2011).
One approach to information structure and its primitives which has been
particularly influential within LFG is that of Vallduví (1992). Vallduví proposes
a trinomial articulation, integrating key insights relating to the two bifurcations

3 A useful summary of information structure terminologies and their dependencies in twentieth-

century research is provided by Kruijff-Korbayová and Steedman (2003); see also Erteschik-Shir (2007)
for an overview.
 the categories of information structure 371

described and exemplified above, which he uses to analyze the syntax-information

structure interface in Catalan. He distinguishes between ground and focus, parallel
to the Background/Focus distinction in other works on information structure.
He additionally proposes that ground information can be further divided into
link and tail. In relating Vallduví’s (1992) approach to more traditional informa-
tion structure categories, Choi (1999: 75–6) states that “we can interpret link as
topic . . . or theme . . . , which is, roughly speaking, what the sentence is ‘about’, and
tail as ‘other’ given information or the rest of the ground information, which is less
conspicuous in the sentence.” Under such an analysis, the information structure of
our Hungarian example would be classified like this:

(9) János Mari-t hívta fel.

John.nom Mary-acc called vm
! "# $ ! "# $ ! "# $
topic focus tail
‘John called Mary.’

Working within the LFG framework, Choi (1999) proposes an analysis of the
relationship between syntax and information structure in German and Korean
which takes Vallduví’s approach as its starting point.4 Choi acknowledges the
importance of the categories which Vallduví proposes, but also points out that
even a trinomial system is not sufficiently fine-grained to capture the data whose
analysis is her main concern. She therefore proposes a feature-based approach to
information structure, in which a term like “topic” is a label for certain values of
a pair of binary-valued features. The advantage of such a system is that it allows
reference to linguistically significant natural classes of roles sharing a common
feature. Choi proposes two such features in her system to encode information
structural distinctions: +new and +prom(inent). Based on these features, Choi
(1999: 92) defines the four discourse functions shown in Table 10.1: topic, tail,
contrastive focus, and completive focus.

Table 10.1 Information features and discourse functions

according to Choi (1999: 92)

+prom −prom
+new contrastive focus completive focus
−new topic tail

The feature +new expresses “discourse-newness”: that is, information is classi-

fied according to how novel it is in the relevant discourse context. Choi analyzes
“focused” arguments as being discourse-new (+new), while “topic” and “tail”
(in the sense of Vallduví 1992) are discourse-old (−new). The second feature,
+prom, refers to the relative prominence that a speaker accords to a particular
element in the discourse; it picks out what is “important” or “urgent” in the

4 In fact, Choi (1999) proposes an Optimality Theory (OT)-LFG analysis. For a brief overview of
OT-LFG, see §18.1.3.
372 information structure

sentence (Choi 2001: 21). Under Choi’s classification, topic and contrastive focus
are prominent (+prom), while tail and completive focus are non-prominent
(−prom). The analysis in (11) of the answer sentence in (10) exemplifies Choi’s
feature system:

(10) Q: What about John? What does he drink?

A: John drinks beer.

(11) John drinks beer.

topic tail completive focus
[−new, +prom] [−new, −prom] [+new, −prom]

Choi’s (1999) approach has the significant advantage of capturing, for example,
what topic and tail have in common: that is, what it means for information to be
classed as ground. Under this approach, ground can be defined as that information
which has a negative value for the feature new. Similarly, topic and contrastive
focus form a natural class because they have in common the specification +prom.
This is a crucial advantage over Vallduví’s system for Choi because in the German
and Korean data which she analyzes, topic and contrastive focus behave alike.
Under Vallduví’s “atomic” approach to information structure categories, by con-
trast, the fact that they pattern together like this is surprising: there is no reason to
expect that they should have more in common than any two of the other categories.
At the same time, under Choi’s feature system these two categories are not expected
to be identical because they have different values for the other feature new: topic
is −new, while any kind of focus is +new.
The binary features originally proposed by Choi in her thesis (published as Choi
1999) were adopted by Butt and King (1996b), who also use them to distinguish
four information structure roles, though not the same ones as Choi; compare Table
10.1 and Table 10.2. Butt and King (2000b) provide the following descriptions of
the four discourse functions which they identify as being associated with different
phrase structure positions in Hindi-Urdu:

• topic is old or known information that is relevant in the current context.

In Hindi-Urdu, the topic appears in clause-initial position, in the specifier
position of IP.
• focus is new and prominent information. It appears in preverbal position
in Hindi-Urdu if there is only one focused element; additionally, a phrase

Table 10.2 Information features and discourse

functions according to Butt and King (1996b)

+prom −prom
+new focus completive information
−new topic background information
 the categories of information structure 373

may be intonationally marked as focus when it appears in its canonical

position.
• background information is like topic in consisting of old or known
information; it provides information as to how new information relates to
old information in an utterance. It appears postverbally in Hindi-Urdu.
• completive information is new information that is not prominent in the
discourse. It is not associated with a particular Hindi-Urdu phrase structure
position, but occurs preverbally.

In Butt and King’s system, completive information is, like focus, new to the
addressee, but it differs from focus in terms of its relative significance and hence
is classified −prom. For example, the phrase in the kitchen in (12A) is, by their
classification, completive information:

(12) Q: What is Anna reading?

A: She is reading War and Peace in the kitchen.
topic background focus completive
[−new, +prom] [−new, −prom] [+new, +prom] [+new, −prom]

Background information is comparable to tail in Choi’s system. Like completive

information, background information is not of great relative significance, but may
be a necessary part of the sentence either for syntactic reasons or to further clarify
the relation between what is already known and the discourse-new material in the
sentence. Topic is classified in the same way by both Choi and Butt and King.
The four-way distinction in information structure roles proposed by Butt and
King (1996b, 2000b) has been adopted by a number of authors in LFG, including
Dalrymple and Nikolaeva (2011), Mycock (2013), and Mycock and Lowe (2013).
While this four-way distinction suffices for our purposes here, it can be seen to
be inadequate given that, for instance, the existence of different types of Topic
with different properties (such as Switch Topic and Continuing Topic) is regularly
acknowledged in the literature (for example, Choi 1999). One could deal with this
by simply proposing an additional information structure feature in order to define
a greater variety of information structure roles. However, this requires careful
thought: it is not clear where such augmentation of the inventory of information
structure features should stop, given that there are a number of subtle information
structural distinctions which even an eight-way distinction does not capture.
While we do not adopt a more fine-grained approach to information structure
roles in this book, we acknowledge that such an approach is ultimately required
if we are to account for the full range of information structural distinctions
that are attested crosslinguistically. Lowe and Mycock (2014) and Mycock and
Lowe (2014) propose a set of four features encoding aboutness, informativeness,
discourse newness, and hearer newness. They demonstrate that these features
allow principled decompositions of the notions of relative salience and newness,
concepts which are often vaguely defined in the wider literature on information
structure but which are fundamental to its analysis.
374 information structure

10.3 Representing information structure: early approaches

In this section, we examine previous LFG approaches to the representation of infor-
mation structure. The earliest approaches utilized the f-structure attributes topic
and focus to represent information structure status. Subsequent approaches have
represented information structure, or i-structure, as a separate level of grammatical
representation.

10.3.1 Grammaticized discourse functions TOPIC and FOCUS

As highlighted in the previous section, the information structure categories of
topic and focus have been a central concern in the LFG literature on information
structure. Bresnan and Mchombo (1987) discuss the f-structural properties of
phrases appearing in sentence-initial position in interrogative and relative clauses
in English and other languages (what in (13a), which in (13b)) with respect to topic
and focus roles. They propose that such phrases bear a grammaticized discourse
function in the f-structure: question phrases in interrogative clauses bear the focus
function, and relativized constituents in relative clauses bear the topic function.
Grammaticized discourse functions have also been called overlay functions; see
§2.1.11.

(13) a. I know [what you want].

focus
b. The car [which you don’t want] is a Renault.
topic

Bresnan and Mchombo note that in a cleft construction, the clefted constituent
bears both functions: it is the focus in the matrix clause and the topic in the
embedded clause, as shown in (14a). They propose that the same constituent
cannot bear both the topic and focus functions in the same clause, accounting
for the grammaticality judgment assigned to (14b):5

(14) a. It is my car [that you want].

focus topic
b. ?? I bought the car [that it was [that you want]].
topic focus

10.3.2 Information structure represented in the f-structure

The grammaticized discourse functions topic and focus proposed by Bresnan
and Mchombo have clear syntactic roles and as such are appropriately represented

5 Dalrymple (2010) discusses examples from Malayalam that appear to violate this constraint.
 representing information structure 375

syntactically, in the f-structure. Building on Bresnan and Mchombo’s proposals,

King (1995), in a key work on information structure, assumes f-structure repre-
sentations like the following:

(15) F-structure for Inna, John claimed that he saw adapted from King (1995:
199):
 
 
 ’ 
 
 
 
 
  
 
 ’ 
  
  
   
  
   
   
   
    

   
  
  
  
 

In (15), the phrase Inna is displaced from its expected position inside the clausal
complement to the beginning of the sentence. At f-structure, Inna is both the
topic of the clause and the obj of the verb saw. This sentence therefore exhibits
a long-distance syntactic dependency, and is governed by the Extended Coherence
Condition, first proposed by Zaenen (1980) and discussed in detail by Fassi-Fehri
(1988). The Extended Coherence Condition requires the information structural
functions topic and focus to be associated with some syntactic function in the
f-structure; Bresnan and Mchombo (1987) state the condition as follows:

(16) Extended Coherence Condition:

focus and topic must be linked to the semantic predicate argument
structure of the sentence in which they occur, either by functionally or by
anaphorically binding an argument.

Used in this way, the f-structure features topic and focus are grammaticized dis-
course functions which have a purely syntactic role as overlay functions, consistent
with Bresnan and Mchombo’s original proposal.
However, the use of the f-structure labels topic and focus was extended in
some literature to include not only grammaticized syntactic functions, but also
the information structure roles to which those functions are related. For example,
King (1995) discusses the Russian example (17), in which the object den’gi ‘money’
bears the role of focus in information structure terms but is not involved in a long-
distance dependency and is not syntactically distinguished from other objects.
Since at the time no other way of representing information structure roles like
topic and focus existed, it was perhaps natural to co-opt the related but different
376 information structure

f-structure features topic and focus to represent these information structure

roles. King proposes the f-structure in (17) for this sentence:

(17) F-structure for prislal muž den’gi according to King (1995: 212):
prislal muž den’gi
sent husband money
‘My husband sent (me) the money.’
 
 

 ’ 

 
 
 

When topic and focus are used in this way, f-structure no longer represents only
syntactic information; it now includes information structural roles as well. This
undermines the principle of modularity that underpins the grammatical architec-
ture of LFG, discussed in §7.2.6 This approach to information structure representa-
tion was subsequently adopted in other LFG work, predominantly for expediency,
it seems, as its inadequacies have been acknowledged in the literature. For example,
Butt and King (1996b), who discuss information structure and its relation to
word order patterns in Urdu and Turkish, use topic and focus to represent
information structure topic and focus in f-structure, but note that a separate level
of representation for information structure features would be preferable. Likewise,
Choi (1999) represents the information structure features +new and +prom at
f-structure, but expresses similar dissatisfaction.
As noted in §2.1.11, the inclusion of the attributes topic and focus in
f-structure is problematic for other reasons: Alsina (2008) and Asudeh (2004,
2011, 2012) criticize their use in f-structure even for the purely syntactic
representation of displaced constituents in long-distance dependencies. Both
authors point out that the labels topic and focus have some justification owing
to the crosslinguistically widespread connection between displaced phrases in
long-distance dependencies and information structural topic or focus status.
These labels are, in fact, used in a deliberately ambiguous way in such cases
in order to capture a syntactic relation and an information structure relation
simultaneously. However, it is not necessarily the case that a particular type of
long-distance dependency always involves the same information structure role,
and thus collapsing the distinction between the two in this way is undesirable. As
an alternative, Alsina proposes the neutral term op, standing for “operator,” in place
of both topic and focus, as the f-structure attribute for all displaced elements.

6 There is also the question of the other constituents’ information structure status, for example of
muž ‘husband’ and the verb prislal ‘sent’ in (17), which remains unrepresented under such an f-structure
approach.
 representing information structure 377

Similarly, Asudeh (2004) proposes the single attribute udf (for “unbounded
dependency function”). These terms permit syntactic analyses of those syntactic
features that are common to all long-distance dependency constructions without
introducing (or apparently introducing) reference to information structure
categories. This is a desirable outcome, since there is no necessary one-to-one
relation between a particular information structure category and a particular
syntactic construction involving a long-distance dependency. For this reason, we
do not use topic and focus as f-structure attributes, but instead reserve these
terms for use solely in relation to information structure. We use the term dis to
represent long-distance dependencies in f-structure in this book. We prefer the use
of dis rather than udf because the term “unbounded dependency” generally refers
to the relation between a filler and a gap, but the feature dis is also used for the
initial constituent in constructions involving left-dislocation and no gap (Beans,
David likes them). We prefer dis to op because the term “operator” is widely used
in semantics with a related but different meaning, and we wish to emphasize the
syntactic rather than semantic nature of dis as a syntactic overlay function.

10.3.3 Information structure as a separate structure

In line with the reservations expressed by Butt and King (1996b) and Choi (1999)
about the use of f-structure to encode information structure relations, King (1997)
provides a series of arguments against such practice. A key issue highlighted by
King is the potential for mismatch between f-structure and information structure
units, where a particular information structural constituent, such as a topic or
focus, does not match a single f-structure constituent, but may match, for example,
only part of an f-structure constituent. This is exemplified in the following:

(18) a. Was it the ex-convict with the red shirt that he was warned to look out for?
b. No, it was an ex-convict with a red tie that he was warned to look out for.
(King 1997: 8, citing Jackendoff 1972: 232)

(19) F-structure for ex-convict with a red tie according to King (1997: 8):
 
   
 
 ’ 

    
   
   
 
  
 

In this particular exchange, the (contrastive) focus in (18b) is tie, which is only
part of the value of the obj attribute in the f-structure in (19); there is no
f-structure consisting only of the focused material which can serve as the value of
a focus attribute in the f-structure. To solve this problem, King argues for treating
378 information structure

information structure as a separate level of the grammar, distinct from f-structure,7

and makes one of the first explicit formal proposals for such a level. King (1997)
proposes the following arrangement of linguistic levels:

(20) c-structure

f-structure i-structure

s-structure

semantics

King provides the abbreviated partial i(nformation)-structure in (21) for (18b).

(Note that the value of background is a set because more than one element can
have the status of background information.)

(21) I-structure for example (18b) King (1997: 9):

 focus tie 
 

  ex-convict   
 background with 
 
red

Under King’s (1997) proposal, i-structure is represented as an attribute-value

structure, just like f-structure. Because it is a separate level of representation,
however, its attributes are distinct from those found in f-structure—they are true
primitives of information structure rather than a means of capturing facts about
the syntax-information structure interface. As a result, the “grouping” of elements
in the i-structure may differ considerably from the grouping of the corresponding
elements in the f-structure.
Building on King’s proposals, and on work by Butt and King (1996b), Butt and
King (2000b) also discuss the position of i(nformation)-structure as a distinct
level of representation within LFG’s projection architecture. They propose that
i-structure is projected from c-structure independently from the projection of
f-structure, as shown in (20). The projection of i-structure from c-structure is
defined by the function ι (iota) from c-structure nodes to i-structures, parallel to
the function 𝜙 from c-structure nodes to f-structures. They assume c-structure
rules such as the following for Hindi-Urdu:

(22) IP −→ XP I!
(ι(ˆ⃰) topic) = ι(⃰) ι(ˆ⃰) = ι(⃰)
(↑ gf) = ↓ ↑=↓

7 This view is somewhat similar to the proposal of Engdahl and Vallduví (1996) in the Head-Driven
Phrase Structure Grammar framework, whereby the features focus and ground are represented
within a distinct structure called info-struct. O’Connor (2006) and Mycock (2006) provide detailed
discussions of these issues.
 representing information structure 379

This rule states that in Hindi-Urdu an IP dominates a phrase of any category (XP)
and an I! constituent. The XP has some syntactic role at f-structure (gf), and at the
same time has the information structure role topic at i-structure. As discussed in
§4.2.1, the IP corresponds to the same f-structure and i-structure as its head, I! . This
rule implies the configuration of c-structure, f-structure, and i-structure shown in
(23), with the ι function from c-structure nodes to i-structures represented by a
dashed arrow:
2 3
(23) gf [ ] (f-structure)
𝜙

IP

XP I!
ι 2 3
topic [ ] (i-structure)

Butt and King (2000b) illustrate their proposal by reference to (24).8 Although
projected independently from the c-structure, i-structure functions and
f-structure functions are not entirely dissociated from each other. The structures
in (23) respect the Extended Coherence Condition given in (16), since the
i-structure topic Naadyaa has the grammatical function subj at f-structure, and
the i-structure focus bazaar-mẽ is a member of the adj set at f-structure.
Butt and King (2000b), like King (1997), assume that i-structure is a projection
from c-structure. A similar proposal is made by Mycock (2006), who argues that
information structure is directly related to both c-structure and prosodic structure.
The shared conception to these proposals is that there is a direct connection
between i-structure and c-structure, and a correspondingly indirect connection
between i-structure and f-structure. The primary motivation for this view is
King’s (1997) observation, discussed above, that f-structure constituents often do
not correspond to information structure constituents, since f-structures are often
either too small or too large to define information structure roles. This is labeled
the granularity problem by Dalrymple and Nikolaeva (2011: 92). The analysis of
constituent questions proposed by Mycock (2006) raises a similar issue, insofar
as it requires reference to units that do not match f-structure constituents, and so
cannot be defined in f-structure terms.
The solution to the granularity problem which King (1997) proposed, and which
was adopted by Butt and King (2000b) and Mycock (2006), is that i-structure
is projected directly from c-structure. This overcomes the granularity problem
because c-structure is more fine-grained than f-structure. For example, although
the f-structure of a modifier is contained within the f-structure of the head, the
c-structure nodes of the modifier may map to a different i-structure from that of
the head. A direct projection from c-structure to i-structure therefore appears to
permit more fine-grained distinctions.

8 The attribute comp.inf in (24) represents completive information.

380 information structure

(24) I-structure according to Butt and King (2000b):

Naadyaa abhii .tofii bazaar-mẽ xarid rahii thii
Nadya now toffee market-loc buy stative.f.sg be.pst.f.sg
‘Nadya was buying toffee at the market just now.’
’

IP

NP I

N S

Naadyaa AdvP NP VP
Nadya
abhii N PP V
now ι
.t bazaar-mẽ xarid rahii thii
buy

This model of information structure is also adopted by Andréasson (2007),

Sulger (2009), and Dione (2012), among others. It is unclear how the theory
of s-structure presented in Chapter 8 would be integrated into this approach.
S-structure is projected via the 𝜎 function from f-structure, but s-structure is also
clearly related to i-structure; connecting the two in this model is not straightfor-
ward. The model that we propose, described in the next section, overcomes these
difficulties by projecting i-structure directly from s-structure, rather than from
c-structure. Moreover, this approach is not susceptible to the granularity problem,
which was the motivation for assuming a direct connection between c-structure
and i-structure in the first place.
 modeling information structure 381

10.4 Modeling information structure

In this section, we present the formal model of information structure and its
relation to other components of the grammar which we assume in the subsequent
chapters of this book. The model of information structural representation used in
this book is essentially the one presented by Dalrymple and Nikolaeva (2011).

10.4.1 Overview
Dalrymple and Nikolaeva (2011) adopt King’s (1997) view, shared by Butt and
King (2000b) and many other LFG researchers, that information structure should
be treated as a separate level of representation, independent from f-structure.
Dalrymple and Nikolaeva propose that information structure contains four
attributes, topic, focus, background, and completive, representing the four
information structure roles originally proposed by Butt and King (1996b), as
discussed in §10.2 above. These information structure roles are determined
by discourse context, and, as discussed in §10.1, can be signaled by various
different means, such as word order and phrase-structure position.9 An important
feature of Dalrymple and Nikolaeva’s proposal is that the elements categorized at
information structure, on the basis of these four features, are meaning constructors
(see Chapter 8), following Mycock’s (2009) insight that information structure and
s-structure are closely related.

10.4.2 Information structure and its relation to semantics

Information structure represents the structuring of the propositional content of
an utterance based on the speaker’s assumptions about the addressee’s state of
knowledge at the time of utterance. As it is concerned with the meaning of an
utterance, it is therefore crucial that a formal theory of information structure
should represent the structuring of meanings, and the assignment of information
structure roles to meanings, and not, for example, to syntactic elements.
There is no consensus on the relation between truth-conditional semantics and
information structure. Some researchers in formal semantics and information
structure argue that information structure should be understood and represented
as an independent module, separate from the module in which truth-conditional
semantics is represented. Others argue that information structure is really a means
of partitioning truth-conditional meaning. The approach described here adopts
this second view: information structure partitions sentence meaning into informa-
tion structure categories, as described in §10.4.4. In certain respects, this approach
resembles “structured meaning” approaches (von Stechow 1982; Krifka 2001), and
correlates with Lambrecht’s (1994) understanding of information structure as the
pragmatic structuring of a proposition.

9 More discussion of these issues is provided by Mycock (2006), Erteschik-Shir (2007), Féry and
Krifka (2008), and references cited in those works.
382 information structure

An early proposal for information structure as means of partitioning or “struc-

turing” meanings was made by von Stechow (1982). Von Stechow represents the
individual meanings in an utterance as a list, the first element of which represents
the topic, and the remaining elements of which represent foci. Krifka (2001)
represents utterance meaning as a pair, with background as the first member and
focus as the second. Krifka (2006) develops this idea further, proposing a three-
part structured meaning for a VP such as ‘introduced bill to Sue’ (where ‘bill’ has
focus status):

(25) Representation of “introduced bill to Sue” (Krifka 2006: ex. 2):

&Bill , A , λx.introduce (Sue , x)'

In (25), the structured meaning is a triple: the first member represents the focus
Bill , the second member contains a set of alternatives to the focus A , and the third
member contains the background meaning ‘introduced to Sue’. The set of alter-
natives A represents all the relevant individuals that might have been introduced
to Sue, including Bill. For example, in the context under consideration A might be
{Bill, David, Chris, …} . In a similar way, our approach also assumes that meanings of
the parts of an utterance are separated and classified according to their information
structure roles in context, though the representation used is somewhat different.
Dalrymple and Nikolaeva (2011) criticize certain assumptions commonly associ-
ated with structured meaning approaches to information structure; for example,
the transformation of structured meanings into non-structured (“standard”)
meanings in the presence of “focus-sensitive” operators such as only depends
on a considerably more restricted view of the structuring of meanings, and plays
no part in the analysis that Dalrymple and Nikolaeva propose. Their approach
assumes that the information structuring of utterance meaning is relevant for all
utterances, regardless of the presence or absence of operators like only; in addition,
it gives no preferential role to the information structure category of focus, but
considers all information structure categories to be relevant to the structuring of
meaning.

10.4.3 Information structure categories: the role of meaning constructors

Most treatments of semantic composition in the glue language assume that the
meaning constructors appearing in a semantic derivation are undifferentiated
and unordered (see Chapter 8). In contrast, Dalrymple and Nikolaeva (2011)
argue that meaning constructors are grouped according to information structure
role, as in the structured meaning approaches described in the previous section.
Thus, meaning constructors that constitute the focus of an utterance are grouped
separately from those that constitute the topic, and likewise those that constitute
background or completive information are grouped separately. This does not alter
in any way the semantic completeness and coherence requirements that are central
to the glue framework. That is, an utterance meaning must still be derivable from
the complete set of premises contributed by the different parts of the utterance, the
utterance meaning should not contain unsaturated expressions, and all contributed
 modeling information structure 383

meaning constructors must be used in the derivation of the utterance meaning

(see §8.7.1).
In Chapter 8, we discussed in detail the following f-structure and meaning
constructors for the sentence David selected Chris:
 
(26) ’
λy.λx.select (x, y) : c ⊸[d ⊸s ]
  David : d
s d 
 
c Chris : c

As in the previous chapter, these meaning constructors refer to s𝜎 , d𝜎 , and c𝜎 ,

respectively the semantic structures related by the 𝜎 function to the f-structures
s, d, and c.
We are now ready to add information structure to the representation. We rep-
resent the syntactic, semantic, and information structural aspects of the sentence
David selected Chris as in (27), given the context Q: David constitutes the topic and
selected Chris constitutes the focus. As more than one meaning constructor may
have topic or focus status, the value of each corresponding i-structure attribute is
a set:

(27) What did David do?

A: David selected Chris.

IP
 
NP I ’
 
s d 
N VP  
c
N V
 
David V NP
λy.λx.select (x, y) : c ⊸ [d ⊸ s ] 
{ David : d }
 
selected N s ι
 
Chris : c
N

Chris

Here, in addition to the familiar c-structure and f-structure, we see the i-structure,
labeled s𝜎ι , which represents the organization of the meaning constructors from
(26) according to their status as topic or focus in the context provided.
As noted above, Butt and King (2000b) defined a function ι from c-structure
nodes to information structures. In contrast, Dalrymple and Nikolaeva (2011)
define ι as a function from semantic structures to information structure. In this
way, i-structure is projected directly from s-structure, and essentially organizes
meaning constructors, the building blocks of s-structure, according to i-structure
status. An expression like s𝜎ι , which labels the i-structure shown in (27), is
384 information structure

defined in terms of the composition of the 𝜎 function and the ι function, ι ◦ 𝜎

(recall our discussion of function composition in §7.5.1). So, s𝜎ι , which is the
i-structure corresponding to s, is obtained by the application of the function 𝜎 to an
f-structure s, and the subsequent application of the function ι to the s-structure s𝜎 .
In other words, s𝜎ι can be understood as the i-structure that is projected from the
s-structure s𝜎 by the ι projection function, or, equivalently, as the i-structure which
is projected from the f-structure s by the composite projection function ι ◦ 𝜎. The
resulting arrangement of levels is distinct from the arrangement in (20), as this
diagram shows:
𝜙 𝜎 ι
(28) c-structure f-structure s-structure i-structure
ι◦𝜎

Using the labels in (29) to represent meaning constructors in a compact and

readable way, as we did in Chapter 8, we can represent the semantic proof for
sentence (27b) as in (30):

(29) [David] David : d𝜎

[Chris] Chris : c𝜎
[selected] λy.λx.select (x, y) : c𝜎 ⊸ [d𝜎 ⊸ s𝜎 ]
[selected-Chris] λx.select (x, Chris ) : d𝜎 ⊸ s𝜎

(30) [selected] [Chris]

[selected-Chris] [David]
select (David , Chris ) : s𝜎

We can also use the labels in (29) to recast the full representation given in (27) in
a more reader-friendly format:

(31) David selected Chris.

IP
 
NP I
’
s
 d


N VP  
c
N V

David V NP  
{ [David] }
 [selected] 
selected N s ι 
N [Chris]

Chris

As shown in (30), the meaning constructor [selected-Chris] can be deduced via

linear logic proof from the two meaning constructors [selected] and [Chris]. We
can therefore represent this configuration equally well as:
 modeling information structure 385

(32) David selected Chris.

IP

NP I
 
’
 
N VP s d 
 
N V c

David V NP
{ [David] }
selected N s ι
{ [selected-Chris] }
N

Chris
In this representation, meaning constructors are categorized according to their
information structure role by virtue of being included in one of the set of values for
an s-structure attribute (topic, focus, background, or completive). In (32), the
meaning associated with the phrase David fills the topic role, while the meaning
associated with the phrase selected Chris fills the focus role. Meaning constructors
contributed by all parts of an utterance are categorized in this way according to
their information structure contribution, and appear in the relevant category at
i-structure.

10.4.4 Formal specification of information structure categories

We now consider how the information structure category of a particular meaning
can be specified in formal terms. Dalrymple and Nikolaeva (2011) propose that this
specification can be achieved through the use of a special feature at s-structure.

10.4.4.1 Semantic structure features Liao (2010) proposes the use of features
to represent the activation and accessibility of discourse referents in context.
The features she uses are adopted from Lambrecht (1994): status, with values
identifiable and unidentifiable; actv (“activation”), whose values include
active, accessible, and inactive; and the binary feature anchored. Liao uses
these features to analyze the distribution of overt and null anaphora in Mandarin
Chinese and to determine information structure roles. Example (33) repeats and
augments (88) from Chapter 8 in line with the view that s-structure is the proper
level for the representation of these features. In (33), we see the f-structure and
meaning constructor for David in a context where David is identifiable and active
in the discourse:
386 information structure

(33) F-structure and meaning constructor for David:

 
+
 + 
David :




The meaning constructor for David pairs the meaning David with a semantic
structure in which an array of semantic and pragmatic features of that meaning are
represented; other features and values might be relevant in other contexts (§8.9).

10.4.4.2 The feature df The model of information structure presented here

depends crucially on an additional s-structure feature, df (for Discourse
Function), whose value is specified by the linguistic context as topic, focus,
background, or completive. In a context in which David is topical, the
associated meaning constructor can be represented as follows:

(34) Meaning constructor for David, with specification of information structure

role:
 
+
 + 
 
David :



 

The value of the df feature is not specified in the lexical entry for David (nor any
similar lexical entry, with the possible exception of question words; see §17.7) since
it is not an intrinsic lexical property of David that it plays a particular information
structure role. Rather, the information structure role of David depends, on each
occasion of its use, on the linguistic context in which it appears. As discussed
in §10.1, information structure roles are determined by the context of utterance;
they can be signaled linguistically in a variety of ways, such as by agreement
or casemarking, phrasal position, or prosody. It is also possible for information
structure roles to be associated by default with particular grammatical functions.
For example, in many languages, including English, the subject is the default topic
(Dalrymple and Nikolaeva 2011; Bresnan et al. 2016). In this way, any of a number
of components of the grammar may determine the df value of a particular meaning
in an utterance.
The relation between the df feature at s-structure and the categorization of a
meaning at i-structure depends on special constraints appearing in lexical entries
in Dalrymple and Nikolaeva’s model. The lexical entry for David, for example,
contains at least the following information:
 modeling information structure 387

(35) David N (↑ pred) = ‘David’

[David] ∈ (↑𝜎ι (↑𝜎 df))

The first line of this lexical entry is entirely familiar. It is the specification in
the second line that is crucial for determining the correct information structural
configuration. This specification involves the meaning constructor David : d𝜎 ,
abbreviated with the label [David]. The functional description specifies that this
meaning constructor is required to be a member of the set of meaning constructors
specified by the value of the s-structure attribute df. The expression can be
paraphrased as follows:

(36) The meaning constructor David : ↑𝜎 , abbreviated as [David], is a member

of the set of values, within the information structure ↑𝜎ι , of the discourse
function attribute signified by (↑𝜎 df).

For example, if the value of the semantic feature df in the s-structure correspond-
ing to ↑𝜎 is topic, then the meaning constructor must be a member of the topic set
at i-structure; see, for example, (32). If the value of df is specified as focus, then the
meaning constructor is a member of the focus set at i-structure, and likewise for
background and completive. Thus for any meaning constructor, when a value is
specified for the s-structure feature df, the information structure categorization of
that meaning constructor, whether topic, focus, background, or completive,
is concurrently specified.10 If no value is specified for the df feature, the meaning
constructor is not integrated into the information structure of the sentence, and the
resulting meaning for the sentence is semantically incomplete, incoherent, or both.

10.4.4.3 Positional specification of df When an information structural role

is associated with a particular syntactic position, this information is included in
annotated phrase structure rules like the following for English:

(37) IP −→ NP I!
(↑ subj) = ↓ ↑=↓
↑𝜎ι = ↓𝜎ι
((↓𝜎 df) = topic)

In the sentence David selected Chris, David is the subject, appearing in the specifier
of IP. As mentioned previously, there is a well-known default relationship between
the grammatical function subject and the information structural role topic. This
is captured by the phrase structure rule in (37) combined with a lexical entry
like (35). Taking the two together, we obtain the following partial configuration,
representing here only the c-structure and f-structure:

10 The way in which this is specified is similar to the use of the pcase feature in the specification of
the grammatical role of a prepositional phrase; see §6.2.
388 information structure

(38) IP

NP
(s subj) = d
I!
s𝜎ι = d𝜎ι 2 4 53
((d𝜎 df) = topic) s subj d pred ‘David’
N

David
(d pred) = ‘David’
[David] ∈ (d𝜎ι (d𝜎 df))

In the annotations under the phrase structure nodes and in the lexical entry in
(38), the ↑ and ↓ metavariables have been instantiated to the f-structure names s
and d. The ↑ = ↓ annotation on N, which ensures that the NP and its head David
correspond to the same functional structure, has been left implicit. The arrows
represent the familiar 𝜙 function from c-structure nodes to f-structures. In the
following exposition, we omit the c-structure in the interests of readability, retain-
ing only the functional descriptions harvested from the annotated c-structure. The
full f-description obtained from the c-structure in (38) is:

(39) (s subj) = d
s𝜎ι = d𝜎ι 2 4 53
((d𝜎 df) = topic) s subj d pred ‘David’
(d pred) = ‘David’
[David] ∈ (d𝜎ι (d𝜎 df))

The first line of (39) requires the f-structure d to be the subject of s, which is true
for the f-structure shown. The second line is crucial: it requires the information
structure corresponding to s and d to be the same. All nonhead daughters in the
phrase structure rules of a language bear such a specification, ensuring that all
members of a clause share the same information structure.11
The third line provides the optional, default discourse function topic associated
with the subject. If its discourse function is not otherwise specified, and as long
as compatible specifications are provided by the linguistic context (as discussed
below), and the prosodic and discourse prominence features of David are con-
sistent with the topic role, then the subject is associated with the i-structure role
topic.
The fourth line is contributed by the lexical entry for David, and requires that the
subject f-structure, d, have a feature pred with value ‘David’. Again, this is true for
the f-structure in (39). The fifth line, also contributed by the lexical entry, specifies

11 Heads need not be explicitly marked with this specification. The head of a phrase corresponds to
the same f-structure as the mother node according to the Head Convention (§4.2.1). Therefore, since
s-structure is projected from f-structure, and i-structure is projected from s-structure, a phrase and its
head correspond to the same f-structure, s-structure, and i-structure.
 modeling information structure 389

that the meaning constructor [David] must bear the role specified by (d𝜎 df) at
i-structure, as discussed above.
If we assume that the default equation ((d𝜎 df) = topic) holds, we can simplify
the final line of these constraints as follows, according to the equality s𝜎ι = d𝜎ι :
2 4 53
(40)
s subj d pred ‘David’
(s subj) = d
(d𝜎 df) = topic 4 5
d𝜎 df topic
(d pred) = ‘David’
[David] ∈ (s𝜎ι topic) 2 3
s𝜎ι topic { [David] }

The equations produce the configuration shown:

• at f-structure, the subject of s is d, and d’s pred is ‘David’;

• the s-structure d𝜎 corresponding to d has the feature df with value topic;
• the value of the attribute df for d𝜎 specifies the topic attribute in the
i-structure for the clause, s𝜎ι .

It is in this way that meaning constructors are specified as bearing a particular

information structure role in the particular linguistic context in which they appear.

10.4.4.4 An example A more complete set of phrase structure rules for a simple
English clause is:

(41) IP −→ NP I!
(↑ subi) = ↓ ↑=↓
↑ ι =↓ ι
((↑ df) = topic)
I! −→ I VP
↑=↓ ↑=↓

VP −→ V
↑=↓
 
V −→ V NP
 
↑ = ↓ (↑ obj = ↓
↑ ι =↓ ι

We are now able, using these rules, to model the information structure of a sentence
such as David selected Chris in (27). The annotations on these rules should be
familiar and unremarkable, except for the annotation ↑𝜎ι = ↓𝜎ι on the NP daughter
of V! , which specifies that the object’s information structure is the same as the
information structure for the entire utterance. We augment the lexical entries for
selected and Chris with the requirement for their meaning constructors to bear
an information structure role: they are required to be a member of some set of
meaning constructors at i-structure, the set determined by the value of the attribute
df at s-structure:
390 information structure

(42) selected V (↑ pred) = ‘selected&subj,obj'’

[selected] ∈ (↑𝜎ι (↑𝜎 df))
Chris N (↑ pred) = ‘Chris’
[Chris] ∈ (↑𝜎ι (↑𝜎 df))

In addition, we require the information structure roles of David, selected, and

Chris to be specified. This can be achieved in a number of ways, through syntactic
encoding or as a result of the discourse context, for example. For this particular
sentence, we assume that the discourse context identifies David as topical, rein-
forcing the default specification in the phrase structure rule for IP that the subject
is the topic. We also assume that the phrase selected Chris has focus status, based on
its context as the answer to the question What did David do?.12 The c-structure and
f-structure for the sentence David selected Chris, including annotations relevant for
the formation of information structure, is given in (43):

(43) IP
 
’
NP
 
(s d s d 
I  
s ι=d ι
c
((d

N VP

N V

David NP
(d V (s c
[David] ∈ (d ι (d s ι=c ι
selected
(s ’ N
[selected] ∈ (s ι (s
N

Chris
(c
[Chris] ∈ (c ι (c

Contribution from linguistic and pragmatic context: (d

(s
(c

12 The discourse prominence features and prosodic contour must reinforce these assignments, or at
least must not conflict with the assignment of these roles. The contribution of prosody is explored in
Chapter 11. As yet there exists no widely agreed LFG model of how discourse context contributes to
the determination of information structure roles.
 modeling information structure 391

Again, we harvest the functional description from the annotated c-structure in

(43). In (44), the constraints are reordered and categorized for ease of reference,
separated into those that refer only to the f-structure, those that refer to s-structure,
and those that are relevant for i-structure. The f-structure is specified by the
constraints under (A); the constraints under (B) specify df values at s-structure;
and the constraints under (C) define the i-structure for this utterance on the basis
of the df specifications. (The equation specifying David as topic, which is both
optionally specified on the phrase structure rule and reinforced by the context, has
not been repeated.)

(44) s ’  
’
(s d  
(d s d 
 
(s c c
(c
(B) (d d :
(s s :
(c c :
(C) [David] ∈ (d ι (d
[selected] ∈ (s ι (s
[Chris] ∈ (c ι (c
s ι=d ι
s ι=c ι

We simplify the equations in (C), using the equalities in (B) and in the last two lines
of (C), to produce the compact description in (45) of the information structure for
this utterance:

(45) (C) [David] ∈ (s𝜎ι topic)

[selected] ∈ (s𝜎ι focus)
[Chris] ∈ (s𝜎ι focus)
 
{ [David] }
 
s ι [selected] 
 
[Chris]

Since [selected-Chris] can be derived by linear logic proof from [selected] and
[Chris] (as shown in the proof in 30), we can, once again, simplify this further to:

(46) { [David] }
s ι
{ [selected-Chris] }
392 information structure

Example (47) is an augmented version of the diagram in (43), showing the full
configuration and relationships between structures. The dashed arrows represent
the 𝜎 function from f-structure to s-structure, while the solid arrows represent the
ι function from s-structure to i-structure:

(47) IP

NP
(s d
I
s ι=d ι
((d

N VP

N V

David NP
(d V (s c
s ι=c ι
selected
N
(s ’
N

Chris
(c
 
’
 
f-structure: s d 
 
c

David : d

λy.λx.select (x, y) : d ⊸ (c ⊸ s
s-structure: Chris : c
)

{ David : d }
Chris : c
λy.λx.select (x, y) : d ⊸ [c ⊸ s ]
i-structure:

We see in this diagram that this clause has a single i-structure in which meaning
constructors are classified according to their information structure role. This
i-structure is formed on the basis of information in the s-structures corresponding
to individual meaning constructors. For instance, the discourse function value
topic in the s-structure corresponding to David ensures that the relevant meaning
 modeling information structure 393

constructor is a member of the topic set at i-structure (in this case, the only
member).

10.4.4.5 Granularity and i-structure role Dalrymple and Nikolaeva’s (2011)

approach to information structure, which we have explored in this chapter, is not
susceptible to the granularity problem as it was presented in §10.3.3. Approaches
that use f-structure units to specify information structure roles fail due to the
granularity problem, since reference to the contents of an f-structure necessarily
includes reference to any arguments and modifiers that appear in that f-structure
(King 1997). In the approach of Dalrymple and Nikolaeva (2011), on the other
hand, the specification of a particular information structure role for a head does
not entail that everything within the same f-structure is associated with that role.
Rather, only the meaning constructors contributed by the head or heads of the
f-structure are associated with the specified information structure role. For
instance, in (47), the meaning constructor associated with the verb appears in
the focus set at information structure. This does not entail, however, that the
arguments of the verb must also appear in the focus set. Indeed, in (47) the
subject of the verb is not (part of) the focus but is the topic, and accordingly
appears in the topic set at i-structure.
One consequence of Dalrymple and Nikolaeva’s approach is that every meaning
constructor associated with a clause must bear a particular role at i-structure,
that is, each one must appear in the set value of one of the information structure
categories. In the Dalrymple and Nikolaeva model, this is enforced by requiring
that every meaning constructor is associated with an equation of the following
form:

(48) [meaning-constructor] ∈ (↑𝜎ι (↑𝜎 df))

This equation states that the meaning constructor must appear in the i-structure
set which is specified by the value of (↑𝜎 df). Very often, this value is not specified
grammatically, but is determined by the linguistic and discourse context. Even in
cases where the value of (↑𝜎 df) is not grammatically specified, the equation in
(48) requires a value to be found. In this way, all meaning constructors play a
role at i-structure. This contrasts with claims that some elements can be entirely
excluded from information structure. For example, Kwon and Zribi-Hertz (2008)
assume that certain unmarked subjects and objects in Korean have no information
structure role. In Dalrymple and Nikolaeva’s model, such an analysis is impossible:
all parts of an utterance meaning are categorized at information structure. Infor-
mation structure does not simply pick out certain parts of a sentence and specify
particular roles for them, therefore, but partitions the entire sentence meaning
according to the relevant criteria.
394 information structure

10.5 Further reading and related issues

Hong (1991), King (1997), Lowe and Mycock (2014), and Mycock and Lowe (2014)
discuss information structure and its place in the overall architecture of LFG. King
and Zaenen (2004) and O’Connor (2006) provide a useful overview of information
structure within LFG. More specific works on information structural issues in
different languages include Dahlstrom (2003) on focus constructions in Meskwaki,
Cook, and Payne (2006) on the relevance of information structure to distributive
scope in German, Andréasson (2007) on word order in Swedish, Mayer (2008)
on differential object marking in non-standard Limeño Spanish contact varieties,
Sulger (2009) on cleft constructions in Irish, Dione (2012) on cleft constructions
in Wolof, and Zymla et al. (2015), who develop a computationally grounded model
of the Common Ground in their treatment of German modal particles.
As noted above, in this chapter we have concentrated exclusively on the orga-
nization of information within clauses. A more complete model would, of course,
also account for the structuring and organization of information across clauses
and the effect that this can have on other aspects of linguistic structure. Little work
currently exists in LFG on discourse structure (that is, cross-clausal information
structuring); King and Zaenen (2004) and Gazdik (2011) are exceptions.
 11
Prosodic structure

In this chapter, we investigate the relationship between the phonological or

prosodic structure of a spoken utterance and its syntactic, semantic, and
information structural analysis. A full theory of the form-meaning correspondence
must account for the effect of prosodic features such as intonation patterns
on interpretation. Work in LFG that is concerned with the contribution made
by phonology or prosody to grammatical structure and interpretation usually
assumes the existence of a separate prosodic structure (sometimes called
phonological structure) within the overall grammatical architecture. We will review
previous LFG approaches to prosody and the place of prosodic structure within
the grammar, before presenting and exemplifying the approach that we adopt.

11.1 Prosody and grammar

Prosody refers to the patterns of rhythm and intonation in spoken language.
Prosodic (also referred to as suprasegmental) features of speech are acoustic char-
acteristics associated with a particular phonological domain. The relevant features
are generally assumed to be pitch, length, and loudness; their physical correlates are
fundamental frequency (F0, measured in Hz), duration, and intensity (measured
in dB), respectively. Our concern here is with prosody at the sentence level; we are
interested in word-level prosody (for example, the lexically determined location
of accent or tonal pattern) insofar as it interacts with sentence-level prosody, such
as intonational contours that distinguish declaratives from questions. Reflecting
the vast majority of research in LFG, the data we consider in this chapter will
overwhelmingly focus on patterns of intonation and the analysis of F0; for LFG
work that considers duration and pitch as well, see Bögel (2015: Chapter 3).
The approach that we adopt in this book, in line with the majority of work in LFG
on the interface between intonation and the other aspects of linguistic structure
discussed in Part II, is a phonological model of intonation designed to capture
generalizations about this system and its meaningful contrastive elements.1 Under
such a model, contours are analyzed as melodies, abstracting away from their
precise shape and the specific Hz values with which they are produced. This kind
of model is inherently flexible with respect to the relationship between prosody
and other aspects of linguistic structure. This is a necessary condition if the model

1 Bögel’s (2012, 2013, 2014, 2015) p-diagram approach, grounded in details of the speech signal,
represents an alternative approach within the LFG framework; this is briefly discussed in §11.7.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
396 prosodic structure

is to account for the attested variability in intonation that can relate to factors such
as speech rate, speaker, style, and dialect, inter alia (see, for example, Arvaniti and
Ladd 2009; Arvaniti 2016). For a general introduction to phrasing and intonation,
which also includes data referred to in this chapter, see Hayes and Lahiri (1991).
The central concerns of a theory of the relation between prosody and other parts
of the grammar are the understanding and analysis of the relation between the
phonological features of a word or utterance and its abstract grammatical prop-
erties. It is clear enough, for example, that some connection must exist between
the phonological realization of a word and the c-structural, f-structural, and s-
structural features and structures that correspond to that phonological realization.
However, the precise nature of that connection and how it is constrained are less
clear. The problem is particularly acute above the word level: how is a particular
phonological or prosodic domain related to a particular syntactic or semantic unit?
This is not merely a matter of incorporating an additional, ancillary aspect of
linguistic analysis into an otherwise complete model of grammar. Prosodic features
such as pitch and duration can, and very often do, make crucial contributions to the
interpretation of an utterance. For example, in Japanese a question is distinguished
from a declarative sentence in writing by the presence of a sentence-final particle
such as ka. In spoken Japanese, the inclusion of this particle is optional (Hinds
1986); what is critical is the intonation pattern that is used. If the version of (1)
without ka is spoken with a final fall in intonation, the utterance is interpreted as a
declarative; if there is a final rise in intonation, it is interpreted as a yes-no question.
In the latter case, the contribution to meaning that prosody makes effectively
“overrides” the apparently unambiguous declarative syntax of the sentence. This
means that the pitch movement associated with the end of the utterance is crucial
to the interpretation of (1) when it does not include the question particle. Such
data illustrate how prosody can make an independent contribution to meaning:

(1) Norio-ga Mayumi-ni omocha-o erabimashita (ka)

Norio-nom Mayumi-dat toy-acc choose.pst q
‘Did Norio choose a toy for Mayumi?’

We represent intonational contours, which may also be referred to as melodies or

tunes, using two abstract tone specifications: High (H) and Low (L). For example,
our representation of the final falling declarative intonational contour in Japanese
is shown in (2a) as a High tone preceding a Low tone, while the final rising
intonational contour associated with the interrogative reading of (1) is represented
in (2b) as the sequence Low–High:

(2) a. Norio-ga Mayumi-ni omocha-o erabimashita.

HL
b. Norio-ga Mayumi-ni omocha-o erabimashita?
LH

Prosody may also be used to resolve ambiguity. For example, it has been claimed
that differences in prosodic phrasing can be used to distinguish between candidate
 prosody: an independent level of structure? 397

syntactic structures, though speakers in general do not do this consistently (see,

for instance, Allbritton et al. 1996). Consider the following examples, taken from
Price et al. (1991), in which square brackets delimit the subordinate clause, and
parentheses indicate the prosodic constituency that can be associated with each
reading. Note how the boundaries of the two types of constituents coincide:

(3) When you learn gradually you worry more.

a. Reading 1: When you learn gradually, you worry more than when you
learn quickly.
= ([When you learn gradually]) (you worry more)
b. Reading 2: When you learn, you gradually begin to worry more.
= ([When you learn]) (gradually you worry more)

Which of the two syntactic analyses is preferred may depend upon the location
of cues such as pauses which signify boundaries between prosodic constituents:
when a pause occurs after gradually, the interpretation is the one given in (3a), i.e.
gradually modifies the verb learn in the subordinate clause; when a pause occurs
after learn, the interpretation is the one given in (3b), i.e. gradually is part of the
main clause. Such data indicate that a close relationship can exist between prosodic
phrasing and syntactic structure (though one should be cautious about claiming
that disambiguation is the result of a one-to-one relationship between the two).
Furthermore, they demonstrate the importance of integrating prosody into any
account of the form-meaning relation.
Debate in the literature on prosody has centered on one fundamental issue in
particular, which we too must address before presenting the approach that we
adopt, namely whether the phonological/prosodic component is an independent
module with its own internal structure.

11.2 Prosody: an independent level of structure?

At the heart of the debate on the place of prosody in the grammar is the issue of its
relation to syntactic structure. Specifically, the question is whether phonological
processes make direct reference to syntactic information or indirect reference via an
interface relation, the nature of which must be identified and defined.

11.2.1 Direct versus indirect reference

Support for the direct reference approach comes from data which indicate that
phonological output is defined on the basis of syntactic structure, as one might
argue is the case in (3). This information is generally assumed to concern syntactic
constituency (for example Kaisse 1985; Odden 1995) but could in principle refer
to other types of syntactic relation, as Pak (2008) points out. Such an approach to
the syntax-phonology interface is problematic on a view which regards syntax and
phonology as fundamentally distinct. For instance, Scheer (2011: 347) holds that
398 prosodic structure

syntax and phonology are separate aspects of linguistic structure and as such are
each unable to interpret units, features, and structures specific to the other.
Even setting aside this incompatibility, the direct reference approach has been
criticized in the light of data which show that syntactic and prosodic struc-
ture are not, as one would predict under such a model, necessarily isomorphic;
for discussion of relevant data, see Chen (1987), Selkirk and Shen (1990), and
Bošković (2001). Lahiri and Plank (2010) seek to emphasize, based on observations
about Germanic languages that date back at least to Steele (1775/1779), that
the prosody-syntax relation is in fact characterized by extensive misalignment;
a lack of isomorphism between the two is more the rule than the exception (see
example 12).
The alternative to direct reference is an indirect reference approach, according
to which an interface relation serves to connect the syntax and phonology macro-
modules. An indirect reference approach copes well with data which demonstrate
a lack of isomorphism between syntactic and prosodic structure. While the two
are systematically related to one another via a mapping algorithm, under such an
approach there is no expectation that isomorphism is the result. With respect to
modularity and domain specificity, the distinction between syntax and prosody can
be maintained by assuming indirect reference: the objects and structures native to
one module cannot be interpreted or manipulated by the other module.
An indirect reference approach to the syntax-prosody interface is not by def-
inition restricted to one particular theoretical framework, but it clearly fits well
with the parallel grammatical architecture and general commitment to modular
specificity of LFG (§7.2). Of course, such an approach requires an independent
level of representation with its own primitives and organizing principles, which
provides the domains of application relevant to postlexical phonological processes.
The issue of precisely how to define and represent prosodic structure continues
to be the subject of research and debate in the wider literature. In the following
section, we review the ideas which have been most influential in the modeling of
prosody and its interfaces within the LFG framework.

11.2.2 Units, constraints, and internal structure

Early work on the representation of prosodic features in generative phonology,
for example by Chomsky and Halle (1968), assumed binary features such as
[±stress]. Such features could be assigned to particular syllables by rules applied
in particular phonological contexts, but no cohesive theory existed of the prosodic
structure of whole phrases or utterances. Building on earlier work by Liberman
(1975) and Liberman and Prince (1977), which explored hierarchical structure
in phonology at the level of the word and below, Selkirk (1978, 1980a,b, 1981)
first proposed that the phonological or prosodic features of phrases and utterances
could be described by reference to a structured prosodic representation. Under this
“Prosodic Phonology” approach, developed further by Nespor and Vogel (1986),
an utterance can be analyzed not only in terms of hierarchical syntactic structure,
i.e. constituent structure, but also in terms of hierarchical prosodic structure.
 prosody: an independent level of structure? 399

According to Prosodic Phonology, the categories of prosodic structure are, like

the categories of constituent structure, finite in number and available crosslin-
guistically. In contrast to constituent structure, prosodic phrases are not projected
by a head, but are related to one another according to the Prosodic Hierarchy.
The precise inventory of categories which form the Prosodic Hierarchy varies
in the literature (as does the inventory of constituent structure categories, as
discussed in Chapter 3); we assume the inventory and set of hierarchical rela-
tions shown in (4), taken (with modified labels) from Selkirk (1995), with these
units being defined independently of units at any other level such as c-structure.
Much research has investigated the units of the Prosodic Hierarchy, focusing on
prosodic/phonological phenomena that have these units as their domain, and the
associated prosodic features which can be taken as cues for their detection:2

(4) The Prosodic Hierarchy:

Utterance (Utt)

Intonational Phrase (IntP)

Phonological Phrase (PhP)

Prosodic Word (PW)

Foot (Ft)

Syllable (S)

The syllable is the smallest unit of the Prosodic Hierarchy that is of concern to us.3
In many languages, syllables are the smallest units with which certain phonological
features, such as stress, are associated.4 A foot is a prosodic constituent consisting
of one or more syllables. Patterns in the alternation of stressed and unstressed
syllables in languages such as English are often stated by reference to rules of
foot formation. One or more feet make up a prosodic word. Feet are crucial in
the formation of prosodic words, but beyond this the foot is not relevant for the
purposes of modeling prosody’s interfaces with other modules of the grammar.
(For this reason, and in the interests of clear presentation, the foot level is omitted
in the analyses of prosodic structure which we present in this book.) Prosodic
words are usually assumed to be the domain within which lexical phonological
processes, and often other processes such as cliticization, apply. One or more
prosodic words make up a phonological phrase, the unit within which a single
prosodic contour may apply. The intonational phrase consists of one or more

2 For an interesting critical review of the Prosodic Hierarchy and Prosodic Phonology in general, see
Scheer (2011).
3 Syllables consist of one or more morae, the smallest prosodic units, but we will not make reference
to morae here. Morae themselves are analyzed as consisting of smaller phonological units, segments and
features, units which are generally considered to be sub-prosodic.
4 As it is not our intention to provide a comprehensive introduction to prosody and phonology, we
set aside the issue of precisely how stress should be defined. For an overview, see Fox (2000).
400 prosodic structure

prosodic phrases. This is the unit within which the smaller prosodic contours of
prosodic phrases are subject to larger overarching processes, such as the gradual
declination of median pitch; the boundaries of intonational phrases are therefore
generally characterized by pitch reset. Finally, the utterance is the largest prosodic
constituent, comprising any single unit of continuous speech consisting of one or
more intonational phrases.5
As with constituent structure, it is possible to state rules or constraints on the
formation of valid prosodic structures. These constraints (see Selkirk 1984 for
details) are usually subsumed under the “Strict Layer Hypothesis” (Nespor and
Vogel 1986: 7; Selkirk 2011: 347), which constitutes a general wellformedness
condition on prosodic structure. The Strict Layer Hypothesis requires that an
utterance be parsed exhaustively into non-recursive prosodic constituents, which
form “layers” that correspond to categories of the Prosodic Hierarchy. Thus, an
Intonational Phrase (IntP) can immediately dominate only Phonological Phrases
(PhPs), and a PhP can immediately dominate only Prosodic Words (PWs).6 For
example:

(5) Utt

IntP

PhP PhP

PW PW PW

teachers maintain standards

The basic approach to prosodic structure illustrated by (5) is widely assumed in

LFG approaches to postlexical phonological processes, though the hierarchical
structuring of prosodic constituents has not always been represented as a tree
structure, as we will see.

11.3 Representing prosodic structure in LFG: early approaches

Early work in LFG made no attempt to integrate a theory of prosody or its relation
to the syntactic component of grammar. The first significant steps in this direction
were made by Butt and King (1998a), who made an explicit proposal regarding the
LFG representation of prosodic features relevant to syntax. Butt and King (1998a)

5 There are many alternative definitions of these prosodic constituents. See, for example, Selkirk
(1978), Nespor and Vogel (1986), Levelt (1989), Wheeldon (2000), and Frota (2012). As stated
previously, we seek to define these constituents in purely prosodic terms, whereas many of these authors
define them in a combination of syntactic and prosodic terms.
6 This strict view of the constraints on prosodic domination has been called into question by a
number of authors, including Inkelas (1989) and Itô and Mester (2003). In particular, it has been
noted that exhaustivity and the bar on recursivity may be best viewed as tendencies, given that they
can be violated in a number of languages. For a recasting of the Strict Layer Hypothesis in Optimality-
Theoretic terms that takes this into account, see Selkirk (1995). Bögel (2015) provides an alternative
approach within LFG which explicitly avoids the prosodic hierarchy as a basis for prosodic structure.
 representing prosodic structure in lfg 401

propose a p(honological)-structure, projected from c-structure and parallel to but

separate from f-structure. Butt and King’s phonological structure encodes only
phonological information that is relevant to the syntax and can contribute to the
full interpretation of an utterance. Butt and King (1998a) illustrate their proposals
with reference to ambiguous verb phrases in Bengali such as this sentence from
Hayes and Lahiri (1991):

(6) ami bh ut dekh -l-am.

I ghost see-pst-1sg
a. ‘I was startled.’ (idiomatic reading)
b. ‘I saw a ghost.’ (literal reading)

When spoken, the two readings of the sentence in (6) are associated with different
phrasings: the idiomatic reading ‘was startled’ is licensed when the VP corresponds
to a single PhP (7a), while the literal reading ‘saw a ghost’ emerges when the verb
form and bh ut ‘ghost’ constitute two distinct PhPs (7b), as indicated by the labeled
brackets:

(7) a. (ami)PhP (bh ut dekh lam)PhP = (6a)

b. (ami)PhP (bh ut)PhP (dekh lam)PhP = (6b)

Butt and King (1998a) represent p-structure as a feature structure, similar to f-

structure. The p-structure they propose for the idiomatic reading composed of
two PhPs is given in (8), and the p-structure for the literal reading composed of
three PhPs is given in (9). The p-structure feature dom (prosodic domain) defines
the level of constituent within the Prosodic Hierarchy that a particular feature
structure represents, while the feature p-form represents a word’s phonological
form:7

(8) ami bh ut dekh -l-am.

I ghost see-pst-1sg
‘I was startled.’ (idiomatic reading)
 dom intonational phrase 
  
  dom p-phrase  
  () * +  
  
  dom p-word


 
    
 
   p-form ami 
 
   
 
   
  



 

   dom p-phrase 
 
    ) dom p-word *  

       
   
  
  
  p-form bh
ut 
 
 
 
   
  
 
 ) * 
 
 
    
 dom p-word 
 

  
  
 
 
 p-form dekh lam 

7 Note that the p-structure feature p-form is not the same as the f-structure feature pform encoding
the form of a preposition (§2.5.4).
402 prosodic structure

(9) ami bh ut dekh -l-am.

I ghost see-pst-1sg
‘I saw a ghost.’ (literal reading)
 
dom intonational phrase
    
  dom p-phrase  
  
 ( ) * + 
 
 dom p-word  
    
 
   p-form ami

 
   
  



 
   
  
   

 
  dom p-phrase 

  
 ( ) * + 
 

  dom p-word 


  p-form b ut h  
  


 
   
  
  


 
  dom p-phrase
 
  
 ( ) * +


 
   
  


 dom p-word 
 
 
   

 p-form dekh lam 

Butt and King (1998a) propose that p-structure is projected from c-structure just
as f-structure is, and that prosodic constituents such as IntPs and PhPs are derived
from syntactic constituents by means of mapping processes. The relevant map-
ping processes permit not only isomorphism, but also some degree of difference
between the two structures (for example, by increasing or decreasing the levels of
embedding in p-structure relative to the c-structure input). With respect to the
ambiguity of (6), Butt and King argue that the prosodic distinction between the
two readings corresponds to a difference in c-structure. In the case of the idiomatic
reading, Butt and King analyze the N bh ut ‘ghost’ and the V dekh lam ‘saw’ as sisters
at c-structure; together they constitute a V! , as shown in (10). This single bar-
level syntactic constituent corresponds to a single PhP in the relevant prosodic
representation (8). By contrast, in the literal reading the NP bh ut ‘ghost’ occupies
specifier position in the VP, as shown in (11). The syntactic relationship between
the NP and the verb is not as close as in the case of the idiomatic reading, and
this is also the case in the prosodic representation (9): the two relevant syntactic
constituents correspond to two separate PhPs rather than forming one PhP.
(10) ami bh ut dekh -l-am. (11) ami bh ut dekh -l-am.
I ghost see-pst-1sg I ghost see-pst-1sg
‘I was startled.’ ‘I saw a ghost.’
S S

NP VP NP VP

Pron V! Pron NP V!

ami ami
N V N V
I I
bh ut dekh lam bh ut dekh lam
ghost saw ghost saw
 representing prosodic structure in lfg 403

Butt and King (1998a) state that their feature-structure representation of p-

structure is equivalent to a tree structure representation of the type shown in
(5), but they acknowledge a deficiency in the representation: the feature structure’s
contents are unordered. They therefore find it necessary to rely on a notion of pro-
jection precedence, assuming that information about ordering is indirectly avail-
able from the c-structure, as for the f-precedence relation on f-structures (§6.11.4).
Another issue with feature structures as a means of representing prosodic
structure, highlighted by O’Connor (2006: 151–2), is that an enriched theory
of phonological structure is required in order to associate text (i.e. the words in an
utterance) with a tune (i.e. an intonational contour) because tonal information at
the phrasal level cannot be directly associated with the relevant text in the feature-
structure approach. In subsequent work, the original feature-structure approach of
Butt and King (1998a) has been rejected in favor of other means of representation.
On Butt and King’s view, p-structure is fundamentally derivative of the c-
structure, even though it is a distinct level of linguistic structure. While the
overall approach is one of indirect reference, by conceiving of p-structure as a
projection from c-structure Butt and King (1998a) ultimately base their definitions
of p-structure units on syntactic constituency. On this view, isomorphism is the
default; when p-structure differs from c-structure, it is as a result of additional
processes applying. As noted in §11.2.1, this assumption about the nature of the
relationship between syntax and prosody has been challenged.
The next significant work on prosodic structure within LFG was on English and
Bosnian/Croatian/Serbian by O’Connor (2005a,b, 2006) and by Mycock (2006)
in the context of the typology of constituent question formation. They assume,
with Butt and King (1998a), that p-structure is a separate level within the parallel
architecture, but both abandon the feature-structure representation in favor of tree
structures. In this respect, O’Connor (2006) and Mycock (2006) share a good deal
of common ground. They propose “tune structure” and “contour” rules respec-
tively to capture facts about intonational contours that apply to specific prosodic
domains. O’Connor adopts the Autosegmental-Metrical/Tones and Break Indices
(AM/ToBI) framework for the expression of tune structure rules.8 Mycock (in her
PhD thesis and in later work) stops short of doing the same, but the p-structure
rules with which she works are broadly compatible with an AM/ToBI approach. In
terms of empirical focus, O’Connor (2006) and Mycock (2006) are both concerned
with how prosody interfaces with other parts of the grammar. In particular, they
seek to capture the important role that prosody can play in signaling information
structure status. Mycock (2006) also analyzes scope marking in constituent ques-
tions with reference to p-structure, offering a way to model the prosody-semantics
interface in LFG. Where these two approaches differ significantly is in the location
of p-structure in the architecture. Mycock (2006) follows Butt and King (1998a)
in assuming that p-structure is projected from c-structure (and thus inherits the
issues with such an approach discussed above), but augments this with the proposal
that p-structure maps to s-structure (and i-structure) independently of syntax in

8 For an introduction to AM/ToBI see Beckman et al. (2005) and Ladd (2008); for ToBI analyses of
data from a range of languages, see the papers in Jun (2005, 2014).
404 prosodic structure

order to account for the type of phenomena exemplified in (2), which show that
prosody can make a crucial contribution to meaning. O’Connor (2006) takes a
different view, according to which p-structure has a direct relationship only with a
level which he refers to as discourse structure. The architecture which O’Connor
proposes is non-standard and has not been adopted beyond his work; we do not
consider it further here.
Perhaps surprisingly, given LFG’s co-description architecture and the general
commitment to modularity, the next major proposal concerning the analysis of
prosody, put forward by Bögel et al. (2009), did not include a separate level of
prosodic representation. As a result, this work has less in common with the indirect
reference approaches outlined in §11.2.1 than the LFG analyses which preceded
it. Bögel et al. (2009) proposed a “pipeline architecture” as a way of capturing
facts about the (mis)alignment of prosodic and syntactic constituents. Rather than
treating prosodic structure as projected from c-structure, Bögel et al. argue that
prosodic information feeds into c-structure analysis. In their model, the input to
c-structure is a prosodically bracketed string. Phrase structure rules then apply
as normal to produce a valid c-structure for the string, but with one significant
augmentation: phrase structure rules are formulated so as to make reference not
only to syntactic categories but also to prosodic brackets. These rules are used in
Bögel et al. (2009) to capture the misalignment which characterizes the syntax-
prosody relation in Germanic languages that, as Lahiri and Plank (2010) discuss, is
often due to the differing behavior of function words in the formation of syntactic
and prosodic constituents. Lahiri and Plank exemplify using an old advertising
slogan:

(12) a. Syntactic Phrasing: [Drink [a [pint [of milk]]] [a day]]

b. Prosodic Phrasing: (Drink a) (pint of) (milk a) (day)

The bracketing in (12) shows that the English function words a and of group with
a constituent to their right in the syntax, but to their left in the prosody. (In fact,
the slogan appeared in writing as Drinka pinta milka day, reflecting the prosodic
bracketing shown.) In their analysis of this example, Bögel et al. (2009) propose that
the phonological string undergoes prosodic parsing, which introduces the relevant
prosodic boundaries into the string. This “prosodically annotated string” is then
the input to the syntactic component:

(13) Phonological String Drink a pint a milk a day

Prosodic Parser |
Prosodically Bracketed (Drink a) (pint a) (milk a) (day)
Syntactic Parser |
Syntactic Analysis [(Drink [a) [(pint [of) (milk]]] [a) (day)]]

A key feature of this pipeline architecture is that syntax can interpret the prosodic
boundaries that are inserted into the string. Though this is not relevant for the
 representing prosodic structure in lfg 405

example in (13), it is crucial to Bögel et al.’s (2009) account of the role of prosody
in resolving syntactic ambiguity. For example, the phrase in (14) is ambiguous,
depending on what exactly the AdjP old modifies, and is compatible with either of
the syntactic analyses indicated:

(14) old men and women

a. [[old men] and [women]]
b. [old [men and women]]

The ambiguity is resolved when prosodic phrasing is available. (15a) is compatible

with the syntactic analysis in (14a), while (15b) pairs with (14b). That is, the
preference is for alignment of syntactic and prosodic boundaries:

(15) old men and women

a. (old men) (and women) = (14a)
b. (old) (men and women) = (14b)

Bögel et al. capture this fact about alignment with their Principle of Prosodic
Preference: “syntactic structures with constituent boundaries that do not coincide
with prosodic boundaries are dispreferred.” While this approach successfully
accounts for the role that prosody can play in resolving syntactic ambiguity
and captures cases of mismatch between syntax and prosody, it does so at the
cost of introducing prosodic information into the syntax macromodule. In an
important respect, this violates the grammatical principle of modularity (§7.2):
the syntactic module should not be able to interpret nonsyntactic objects. This
is circumvented by proposing that a prosodic boundary has an additional role
as one of two syntactic categories (left boundary, LB, or right boundary, RB),
with the result that prosodic objects are effectively “disguised” as being syntactic:
under this approach, LB and RB are terminal nodes in the c-structure tree. This
does not represent a strictly modular approach to the issue. Moreover, the rather
simplified prosodic representation involving LB and RB that Bögel et al. (2009)
adopt does not allow easy reference to the different types of units that belong
to the Prosodic Hierarchy, something which is essential in the analysis of many
phenomena. Presumably, reference to distinct prosodic units could be made via a
separate, additional prosodic component introducing a large number of differently
labeled prosodic brackets into the string, but this adds further complexities. Bögel
et al. (2009) do not explore the details or the wider implications of including
such a component. Subsequent work on prosody and its interfaces within LFG has
moved away from the “pipeline” architecture proposed by Bögel et al. (2009) and
returned to an approach in which prosodic structure represents a distinct level
in the parallel architecture, a position more compatible with strict modularity. In
later work such as Bögel (2015), Bögel rejects the prosodic hierarchy as the basis for
p-structure.
406 prosodic structure

11.4 Modeling the prosody-syntax interface

We now present the model of the prosody-syntax interface that we adopt in
this book. Espousing a strong commitment to modular specificity and the indi-
rect reference approach, Dalrymple and Mycock (2011) propose a new model
for integrating prosody with the wider LFG grammar. This model conforms to
three important criteria, addressing issues raised in the two previous sections.
Firstly, it enables the contribution of prosody to different components of the
grammar to be modeled with relative ease. Secondly, it does not assume any sort
of close correspondence between the syntactic and prosodic phrase structures of
an utterance, though it does permit any such correspondence to be accounted
for. Thirdly, it respects the principle of modularity of the grammar, maintaining
a strict separation between syntactic and prosodic information. In Dalrymple
and Mycock’s model, although prosody may contribute to syntax, semantics, and
information structure, no structure is able to interpret prosodic information except
the prosodic component, and the prosodic component itself is unable to interpret
anything but prosodic information.
Dalrymple and Mycock’s (2011) model of the prosody-syntax interface was
further developed and streamlined by Mycock and Lowe (2013). In the following
sections, the model that we present is essentially the Mycock and Lowe version.

11.4.1 The p-string and the s-string

In §3.5, we adopted the proposal of Kaplan (1987, 1995) that the string of words
usually treated as constituting the terminal nodes of the c-structure tree should be
thought of as a separate structure from the c-structure. Under this proposal, the
string is related to the c-structure by a projection function, labeled π, from string
elements to terminal c-structure nodes; terminal c-structure nodes are therefore
not words, but X0 or 2 X level categories. This is illustrated in (16) for the sentence
David arrived:

(16) David arrived.

IP

NP I!

N VP

π V
π

David arrived
 modeling the prosody-syntax interface 407

The string can be thought of as a grammatical signal parsed into minimal units.
The majority of analyses assume this view of the string: it is regarded as compris-
ing minimal syntactic units which are then subject to phrase-structural analysis
in the c-structure, via the π projection. However, this ignores the fact that a
string can also be parsed into minimal phonological/prosodic units. The string,
as Asudeh (2009: 107) observes, thus represents part of the syntax-phonology
interface.
Dalrymple and Mycock (2011) were the first to make a detailed proposal con-
cerning this dual nature and function of the string. They argue that the string must
be analyzed as having two distinct aspects: one syntactic, the s-string, the other
phonological/prosodic, the p-string. Any linguistic signal can be parsed in two
ways: the s-string represents the parsing of a signal into minimal syntactic units,
and so corresponds to the conception of the string in most earlier works, while
the p-string represents the parsing of a signal into minimal phonological/prosodic
units. Just as the s-string is the basis of the syntactic phrase-structural analysis of
an utterance, so the p-string is the basis of the prosodic phrase-structural analysis.
To give an example, in Germanic languages units are formed according to the
types of rhythmic principles highlighted by Lahiri and Plank (2010), along with the
structure of the Prosodic Hierarchy; see §11.2. The units of p(rosodic)-structure
are native to the phonology macromodule, and the projection of p-structure is
based on phonological and prosodic features; there is no sense in which p-structure
is derived from syntactic phrase structure, and no assumption of a necessarily
close correlation between syntactic and prosodic phrase structure. The two are,
of course, related to one another, but it is the string which represents the point of
interface; thus, modularity is respected. Dalrymple and Mycock (2011) illustrate
their proposals using the sentence Anna was studying at the university, providing
a “double-tree” analysis of the two aspects of the string. The s-string, p-string,
c-structure, and p-structure which they propose—with some emendations to bring
it in line with proposals by Mycock and Lowe (2013)—are shown in (17). The
units of the s-string are each related to a terminal node of the c-structure via the π
projection. The prosodic units related to these syntactic units form the p-string.
Parallel to the syntactic analysis of this utterance, the units of the p-string are
each related to a terminal node of the p-structure via the β projection, with “S”
standing for syllable. Thus, the two aspects of the string receive equivalent but
distinct analyses:9

9 In (17), each PW is also a separate PhP, but it could be the case that, for example, the PWs are
grouped into two PhPs, giving (æ nə wəz stʌ di ɪŋ ət ðə ju nə)PhP (vɜ: sə ti)PhP . Our approach is flexible
in order to account for the attested variability in intonation mentioned in §11.1. While for our purposes,
nothing in (17) hinges on the placement of PhP boundaries, their inclusion is important because they
can be associated with other aspects of prosody. For instance, voicing assimilation in Bengali is bounded
within PhPs (Hayes and Lahiri 1991). Thus, even if a PhP comprises a single PW, this does not mean
that the two can or should be conflated.
408 prosodic structure

(17) IP

I!
VP
V!
PP
P!
!
NP V NP
N!
N I V P D N
π π π π π π
s-string Anna was studying at the university

p-string æ nə wəz stʌ di ɪŋ ət ðə ju nə vɜ: sə ti

β β β β β β β β β β β β β
S S S S S S S S S S S S S

PW PW PW

PhP PhP PhP

IntP

Utt

The relation between s-string and p-string units or, to put it another way, between
the syntactic and prosodic parses of an utterance, is represented in (17) using
dotted lines. Dalrymple and Mycock (2011) propose that the s-string–p-string
relation is defined by information stored in lexical entries. Underpinning their
proposal is a more fine-grained understanding of the nature and content of lexical
entries, according to which a lexical entry specifies both a s(yntactic)-form and a
p(honological)-form, as well as the c-structure category and f-description. The co-
occurrence of s-forms and p-forms in lexical entries constrains the analysis of a
string by requiring that any lexical item identified as a syntactic unit necessarily
corresponds to a phonological/prosodic unit or units, in the same relative position
in the string. The role of the lexical entry in mediating between the s-string and
the p-string is illustrated in (18). The diagram in (19) exemplifies the portion of
the grammatical architecture for the noun university which appears in (17):
 modeling the prosody-syntax interface 409

(18) C-structure
Lexical entry
π
s-form s-string
p-form p-string
β
P-structure

(19) Lexical entry, s-string, and p-string: Dalrymple and Mycock (2011)

C-structure
Lexical entry
π
university university
/ju.nɪ.vɜ:.sɪ.ti/ /ju nə vɜ: sə ti/
β
P-structure

Dalrymple and Mycock (2011) therefore use the p-string, the s-string, and the
lexicon to model the interface between the syntax and phonology macromodules.
Under this analysis, information of various kinds can be associated with s-string
and p-string units. Mycock and Lowe (2013) develop Dalrymple and Mycock’s
(2011) approach to lexical entries by proposing that s-string and p-string units are
not atomic, but in fact are bundles of information which they represent as feature
structures.
A key component of any feature structure representing a unit of the s-string or
p-string is the feature fm (form), whose value is related to the s-form or p-form
of a lexical entry. For example, the lexical entry for university includes:

(20) (• fm) = university

The symbol • is used to refer to s-string units (Mycock and Lowe 2013). In the case
of an s-string fm feature such as that shown in (20), the relation with the s-form in
the relevant lexical entry (i.e. university) is one of identity. That is, the s-form of a
word as it appears in a lexical entry is identical to the value of the fm feature in any
s-string instantiation of that word. In the case of the p-string, on the other hand,
this is not the case. For instance, the p-form of university includes five syllables
410 prosodic structure

(shown separated by periods), of which the second and fourth syllables contain
the vowel /ɪ/:10

(21) /ju.nɪ.vɜ:.sɪ.ti/

Notice that in the p-string in (17) the vowels in those same syllables are realized
as /ə/. Phonological processes related to speech tempo and other contextual
factors may apply, rendering the relation between p-forms and p-string fm fea-
tures opaque. (It is important that the underlying vowels are retrievable though,
as speakers can clearly access and produce them in instances of slow, careful
production.) In informal terms, it is not difficult to see how the application of
such phonological processes affects the derivation of p-string fm features from
p-forms, and likewise how the corresponding undoing of those same rules affects
the reconstruction of p-forms from p-string fm features. For the present purposes,
however, we make no proposals regarding the precise formalization of this aspect
of phonology.11
The other respect in which fm features differ from the s-forms and p-forms
of lexical entries is the possibility of many-to-one correspondences. Again, this is
clearly seen by comparison of (21) with (17). The single p-form of the lexical entry
in (21) corresponds to five distinct units in the p-string representation. Under a
feature-structure approach, each of these p-string units is represented as a distinct
p-string feature structure with its own p-string fm feature; that is, there is one
p-string unit, represented as a separate feature structure, for each syllable. We
assume that the specification of p-string units is part of the lexical information
stored in lexical entries.12
Of course, fm features and their values are not the only components of a lexical
entry. Each lexical entry also includes a c-structure category and f-description. At
this point, we must note an inconsistency in the formal details of the presentation
so far. Following long-established tradition in LFG, we have been using the
variable ↑ in f-descriptions in lexical entries. Thus, for example, the f-structure
number feature for university is specified by the f-description (↑ index num) = sg.
However, as discussed in §5.3.1, the variable ↑ is an abbreviation for 𝜙(ˆ⃰), that
is, the f-structure projected, via the 𝜙 function, from the mother of the current
c-structure node. This variable was appropriate when a syntactic element such
as university was conceived of as a terminal node of the c-structure tree, with
the preterminal node N as its mother. In that case, ↑ refers to the f-structure
corresponding to the N node, as required. However, as discussed in §3.5 and at
the start of the current section, most major approaches to the LFG architecture
follow Kaplan (1987, 1995) in understanding the (s-)string to be distinct from the

10 This is a simplification. As Bögel (2015) observes, following Levelt et al. (1999), segments and the
metrical frame are most likely stored separately in the lexicon. See Bögel (2015) for more on the precise
contents of the p-form portion of a lexical entry.
11 For a formalization of some postlexical phonological processes in an LFG setting, see Bögel (2015).
12 This issue affects not only the p-form–p-string relation, but also the s-form–s-string relation, since
single lexical entries can specify multiple s-string elements; see Lowe (2016a).
 modeling the prosody-syntax interface 411

c-structure, related to it by the projection function π. Since ↑ is defined in terms

of the mother function M on c-structure nodes, our use of ↑ in a lexical entry is
therefore technically incorrect. What should rather be used is the distinct function
𝜙(π(•)), which represents the f-structure projected, via the 𝜙 function, from the
c-structure terminal node projected, via the π function, from the s-string unit with
which the lexical entry is associated, as shown here:

(22)

The lexical entry for university (with its parts labeled) is therefore:13

(23) s-form (• fm) = university

c-structure category λ(π(•)) = N
f-description (𝜙(π(•)) pred) = ‘university’
(𝜙(π(•)) index num) = sg
university ∈ (ι(𝜎(𝜙(π(•)))) (𝜎(𝜙(π(•))) df))
p-form /ju.nɪ.vɜ:.sɪ.ti/

While a lexical entry of this type is technically more accurate, the use of ↑ is a
well-established tradition in LFG, and is certainly more readable. We therefore
retain this traditional notation, on the understanding that ↑ appearing in a lexical
entry refers to a different function from ↑ when it appears in phrase structure
annotations. In a lexical entry, ↑ is defined as 𝜙(π(•)), where the π function maps
from units of the s-string to terminal nodes of the c-structure. In a phrase structure
rule, ↑ is defined as 𝜙(ˆ⃰) (or, to be completely explicit, as 𝜙(M(⃰)), in terms of the
mother function M on c-structure nodes).

11.4.2 Edge features

Besides the fm feature, whose value is specified in each lexical entry, a range of
other features can appear in string feature structures. Some of these are lexically
specified; for example, a lexically stressed syllable is specified in the lexicon as
having the feature syllstress p. Other features represent properties that string
elements may possess in a particular context. The most important of these contex-
tually specified properties for our purposes is the association of string elements
with information about the location of the beginning or end of syntactic or
prosodic constituents. Reliance on such edge information is an acknowledged
characteristic of prosody and its interaction with other aspects of linguistic struc-
ture, and as such has been integral to models of prosody and its interfaces over

13 On the function λ, which relates a node of the tree to its label, see §5.1.2.
412 prosodic structure

the years, for instance the end-based (also known as the edge-based) approach
of Selkirk (1986) and the Align family of constraints in Optimality-Theoretic
approaches including Selkirk (1995). Following Dalrymple and Mycock (2011),
Mycock and Lowe (2013) assume that string units appearing in particular contexts
are associated with sets of labels. These labels represent the left and right edges of
the constituents in the associated structure that belongs to the relevant module
of the grammar (i.e. c-structure or p-structure). So, in (17), the s-string element
‘Anna’ is associated with the left edge of the IP in the c-structure, and also with
both the left and right edges of an NP and N. This is represented in the relevant
string unit by means of features l (left edge) and r (right edge) whose values are
sets comprising information concerning which constituents this particular s-string
unit represents the left or right edge of. Thus, because the N Anna is the left edge
of the IP, NP, and N constituents and the right edge of the NP and N constituents,
the representation of the s-string unit for Anna in (17) is:
 
(24) fm Anna
 l { IP, NP, N } 
 
r { NP, N }

There are two p-string elements corresponding to the s-string unit for Anna: ‘æ’
and ‘nə’. The first of these is associated with the left edge of a PW, a PhP, the IntP,
and the Utt in the p-structure. As is the case for edge information related to s-string
units, this edge information about p-string units appears as values of the features
l and r within p-string attribute structures, as shown in (25).14
The features fm, l, and r are not the only features that have been proposed for
s-string and p-string attribute structures. Mycock and Lowe (2013) also include
prosodic information associated with particular p-string units in p-string feature
structures. For example, they assume the feature syllstress with value p in order
to represent the location of primary stress. Thus, the full representation of the two
p-string units for Anna in (17) is:

(25)  fm æ 
 fmnə 
 syllstress p 
  l { } 
 l { Utt, IntP, PhP, PW } 
r { }
r { }

The mechanism by which c-structure and p-structure category edge information

is specified as appearing in s-string and p-string features is defined by Mycock and
Lowe (2013) in terms of the relation between a c-structure or p-structure node and
its rightmost or leftmost daughter. They define a relation D which, when applied
to a c-structure or p-structure node, finds the set of immediate daughter nodes

14 For a different approach to capturing prosodic phrasing and its encoding in LFG, see Bögel’s work
on the p-diagram. Bögel (2015) provides full details; see also §11.7.
 modeling the prosody-syntax interface 413

of the node to which it applies; it is therefore the inverse of the mother relation
M discussed in §5.3.1. The leftmost immediate daughter of the node in question
can then be defined as the member of the set of daughter nodes which is not
preceded, in linear terms, by any other member of the same set. Likewise, the
rightmost immediate daughter of the node can be defined as the member of the
set of daughter nodes which is not followed, in linear terms, by any other member
of the same set. These functions, from nodes to leftmost and rightmost immediate
daughters, are labeled Dl and Dr respectively, and are defined, in relation to c-
structure nodes, as follows:

(26) a. Dl ( ⃰ ) ≡ node n, where n ∈ D( ⃰ ) ∧ ¬∃x.x ∈ D( ⃰ ) ∧ x ≺ n.

b. Dr ( ⃰ ) ≡ node n, where n ∈ D( ⃰ ) ∧ ¬∃x.x ∈ D( ⃰ ) ∧ x ( n.

This is shown graphically in (27), where the mother node is represented as c1 , the
leftmost daughter is Dl (c1 ), and the rightmost daughter is Dr (c1 ):

(27) c1

Dl (c1 ) ... Dr (c1 )

The leftmost and rightmost terminal nodes dominated by a node can be defined
by the recursive application of these functions until a node is reached that has no
daughters. Mycock and Lowe (2013) label the functions that find these terminal
nodes as Tl and Tr respectively. The definitions they provide are given in (28).
These rules can be informally read as follows: the leftmost/rightmost terminal
node from the current node is the current node if the current node has no daugh-
ters; otherwise, it is the leftmost/rightmost terminal node of the leftmost/rightmost
daughter (respectively) of the current node. The rule applies recursively to find the
appropriate terminal descendant of any node:
(
(28) a. Tl (⃰) ≡ ⃰ if D(⃰) = ∅
else Tl (Dl (⃰))
(
if D(⃰) = ∅
b. Tr (⃰) ≡ ⃰
else Tr (Dr (⃰))

In (29), we display a sample tree in which the current node (⃰ in the definitions
above) is represented as c1 , the leftmost terminal daughter is Tl (c1 ), and the
rightmost terminal daughter is Tr (c1 ):

(29) Sample configuration: Tl (c1 ) is the leftmost terminal daughter of c1 , and

Tr (c1 ) is the rightmost terminal daughter
c1

Tr (c1 )
Tl (c1 )
414 prosodic structure

Finally, the s-string elements corresponding to these terminal nodes are straight-
forwardly obtained by applying the inverse function π−1 from terminal nodes
of the c-structure to elements of the s-string. In (30), the leftmost s-string ele-
ment of the node labeled c1 is π−1 (Tl (c1 )), and the rightmost s-string element is
π−1 (Tr (c1 )):

(30) Sample configuration: π−1 (Tl (c1 )) is the leftmost s-string element of c1 , and
π−1 (Tr (c1 )) is the rightmost s-string element
c1

Tr (c1 )
Tl (c1 )
π−1
π−1
π−1 (Tl (c1 )) π−1 (Tr (c1 ))

Mycock and Lowe (2013) propose to use arrows as abbreviations for the paths
from c-structure nodes to leftmost and rightmost string elements. The arrow ⇙
abbreviates the function from a c-structure node to the leftmost s-string element
dominated by it, and the arrow ⇘ abbreviates the function from a c-structure node
to the rightmost s-string element dominated by it:

(31) a. ⇙ ≡ π−1 (Tl ( ⃰ ))

b. ⇘ ≡ π−1 (Tr ( ⃰ ))

The tree in (32) provides an example of the use of ⇙ and ⇘:

(32) Example: ⇙ and ⇘

c1
(⇙fm) = sform1
(⇘fm) = sform2

Tr (c1 )
Tl (c1 )
π−1
π−1
3 4 3 4
fm sform1 fm sform2
 modeling the prosody-syntax interface 415

In a parallel manner, the arrows # and $ abbreviate the functions from a

p-structure node to its leftmost and rightmost p-string elements. The definitions
for these arrows are given in (33), where + refers to any p-structure node (i.e. this
is the equivalent of ⃰ at c-structure). Recall that β is the mapping from elements of
the p-string to terminal nodes of the p-structure; β−1 is the inverse function:

(33) a. # ≡ β−1 (Tl (+))

b. $ ≡ β−1 (Tr (+))

Use of the “syntax arrows” ⇙ and ⇘ and the “phonology/prosody arrows” #

and $ therefore permits direct reference from any c-structure or p-structure node
to the s-string or p-string elements corresponding to the leftmost and rightmost
terminal nodes which are descendants of the node in question. It is thus possible to
express simple generalizations about the passing of category information to string
elements, with annotation principles like those given in (34); these generalizations
are encoded in all phrase structure rules in a language with phrase structure
categories NP and N, and similar generalizations hold of other categories:

(34) a. For any NP, NP ∈ (⇙l) and NP ∈ (⇘r)

b. For any N, N ∈ (⇙l) and N ∈ (⇘r)

The tree in (35) explicitly shows these specifications on all c-structure and p-
structure nodes. Notice that the labels appearing in the sets which are the values
of each string unit’s l and r features correspond exactly to the specifications made
by the arrows in the c- and p-structures.
As mentioned previously, in addition to the features fm, l, and r, Mycock
and Lowe (2013) include prosodic information associated with particular p-string
units in p-string feature structures. The feature syllstress with value p, which
appears in (35), represents the location of primary stress:15

15 Lowe (2016a) also proposes a feature clitic, which can appear in both s-string and p-string
units, to distinguish syntactic or prosodic clitics from non-clitics at the level of the string, and an s-
string or p-string feature field, which contains labels referring to all syntactic categories dominating a
particular element in the c-structure or to all the prosodic categories dominating a particular element
in the p-structure.
416 prosodic structure

(35) Anna hit Norman.

IP
IP ∈ (⇙L), IP ∈ (⇘R)

NP I!
NP ∈ (⇙L), NP ∈ (⇘R)

VP
VP ∈ (⇙L), VP ∈ (⇘R)

N V!
N ∈ (⇙L), N ∈ (⇘R)

V NP
V ∈ (⇙L), V ∈ (⇘R) NP ∈ (⇙L), NP ∈ (⇘R)
π N
π N ∈ (⇙L), N ∈ (⇘R)
π

     
fm Anna fm hit fm Norman
 l { IP, NP, N }   5 ! 6  l { NP, N } 
s-string   l I , VP, V! , V   
   5 6
r { NP, N } r {V} r IP, I! , VP, V! , NP, N

 fm     fm 
æ  fm nə  fm hıt nɔː  fm 
 syllstress p  mən
 syllstress p    syllstress p 
 5 6 l { }   5 6  
 l 5 PhP, PW 6 
l { } 
p-string  l
 IntP, PhP, PW 

 5 6 l PhP, PW   
 5 6
r PhP, PW   r IntP, PhP, PW
5 6
r { } r PhP, PW r { }

β β β β β

S S S S S

PW PW PW
PW ∈ (#L), PW ∈ ($R) PW ∈ (#L), PW ∈ ($R) PW ∈ (#L), PW ∈ ($R)

PhP PhP PhP

PhP ∈ (#L), PhP ∈ ($R) PhP ∈ (#L), PhP ∈ ($R) PhP ∈ (#L), PhP ∈ ($R)

IntP
IntP ∈ (#L), IntP ∈ ($R)

The approach to prosody and its interfaces developed by Dalrymple and Mycock
(2011) and illustrated in (35) makes crucial reference to edge information asso-
ciated with string elements. Dalrymple and Mycock (2011) themselves note that
“passing information about the edges of all major constituents into the p-string
and s-string . . . may seem excessive,” but also point out that in at least some cases,
for instance when speech is slower and more careful, a greater degree of alignment
exists between syntactic and prosodic constituents, meaning that “information
 modeling the prosody-syntax interface 417

about constituent boundaries must be available at the interface in order that

the relevant alignment principles can apply to give phrasing which reflects the
appropriate degree of increased isomorphism.” However, Dalrymple and Mycock
also state that “this does not mean that we necessarily expect all of this information
to emerge as being relevant at the interface. In fact, it is an important feature of this
architecture that it enables us to explore the question of precisely which aspects
of prosodic and syntactic structure are important at the interface and to investi-
gate why this should be, both crosslinguistically and on a language-by-language
basis.”
One example of a case where alignment between syntactic and prosodic con-
stituents is key is the type of disambiguation analyzed by Butt and King (1998a) and
presented in §11.3. To recap briefly, two readings are associated with the Bengali
sentence in (36), with the ambiguity being resolved by prosodic phrasing:

(36) ami bh ut dekh -l-am.

I ghost see-pst-1sg
Reading 1: (ami)PhP (bh ut dekh lam)PhP ‘I was startled.’
Reading 2: (ami)PhP (bh ut)PhP (dekh lam)PhP ‘I saw a ghost.’

Crucial to disambiguation of the sentence in (36) is the alignment of phrase-level

p-structure and c-structure boundaries: the difference between Reading 1 and
Reading 2 is whether bh ut (‘ghost’) is a phrase in its own right (i.e. a separate
NP in the c-structure tree, and a separate PhP at prosodic structure). If bh ut is
simultaneously the left and the right edge of a PhP in the p-structure, then it is
also the left and right edge of a NP in the c-structure and the appropriate reading
is Reading 2; otherwise, it is Reading 1.
In Dalrymple and Mycock’s model (see also Mycock and Lowe 2013), a principle
of “Interface Harmony” is important to the analysis of this and other instances
of prosodic disambiguation, as well as to the increasing alignment which occurs
as speech rate decreases. Interface Harmony requires that certain information
introduced at c-structure (such as the location of a phrase’s left or right edge, for
example), which has its presence recorded via a label in the s-string (by virtue
of being a member of the l or r set in such a unit’s feature structure), must
be matched with a label in the corresponding p-string unit; both labels should
be values of the same feature (i.e. l or r) in the relevant string units. That is,
at the interface between s-string and p-string, a principle of Interface Harmony
applies which can require that two labels co-occur—one in an s-string unit and
one in a p-string unit. Note that these are labels: they are associated with but do
not in themselves constitute specific syntactic or prosodic information. Prosodic
disambiguation relies on the alignment of c-structure and p-structure boundaries
and therefore can be understood in terms of Interface Harmony. The Bengali
example provided by Butt and King (1998a) illustrates this well. In (37), we see how
the boundaries of phrase-level prosodic constituents (parentheses) and syntactic
constituents (square brackets) align in the respective readings:
418 prosodic structure

(37) a. Reading 1: ‘I was startled.’

[(ami)PhP ]NP [(bh ut dekh lam)PhP ]VP
S

NP VP

Pron V!

N V

ami bh ut dekh lam

a mi bh ut dekh lam
S S S S S

PW PW PW

PhP PhP

IntP
b. Reading 2: ‘I saw a ghost.’
[(ami)PhP ]NP [[(bh ut)PhP ]NP (dekh lam)PhP ]VP
S

NP VP

Pron NP V!

N V

ami bh ut dekh lam

a mi bh ut dekh lam
S S S S S

PW PW PW

PhP PhP PhP

IntP

Each phrasal boundary in the syntax (NP or VP) has a corresponding PhP boundary
in the prosodic analysis.16 (There is no “double marking” or “double edge effect” if
an edge of the VP also represents an edge of an NP.) The two readings are therefore
associated with distinct c-structure and p-structure pairs. As proposed by Dal-
rymple and Mycock (2011), information about the edges of all major constituents
is recorded as labels in l or r feature-value sets, by means of requirements like
those given in (34). The string representations for the Bengali utterances, based
on the analysis presented by Butt and King (1998a) (which itself is based on that
of Hayes and Lahiri 1991), are as shown in (38) and (39). Under neutral focus, a

16 These boundaries are crucial in accounting for different voicing assimilation possibilities in these
two examples; see Hayes and Lahiri (1991).
 modeling the prosody-syntax interface 419

High tone, represented in the p-string feature structure as tone h, is associated

with the leftmost PW of the rightmost PhP in Bengali (Hayes and Lahiri 1991).
This serves as an indicator of prosodic constituency, as examples (38) and (39)
show. The labels that align and upon which the disambiguation relies are given in
bold in both examples; the first word ami ‘I’ is omitted.

(38) Reading 1: ‘I was startled.’

   
fm bh ut fm dekh lam
   
s-string  l { VP, V , N } 
!
l { V } 
r {N} r { S, VP, V , V }
!

   
fm bh ut fm dekh  
 syllstress p   syllstress p  fm lam
    l { } 
     
p-string  tone h   l { PW } 
  r { IntP, PhP, PW }
 l { PhP, PW }  r { }
r { PW }

(39) Reading 2: ‘I saw a ghost.’

   
fm bh ut fm dekh lam
   
s-string  l { VP, NP, N }  l { V , V }
!

r { NP, N } r { S, VP, V , V }
!

   
fm bh ut fm dekh  
 syllstress p   syllstress p  fm lam
   
p-string     l { } 
 l { PhP, PW }   tone h   
 
r { PhP, PW }  l { PhP, PW }  r { IntP, PhP, PW }
r { }

In examples (38) and (39), the principle of Interface Harmony, coupled with
the rules of alignment in Bengali, mean that XP labels in an s-string feature
structure are matched with phrase labels in the values of the l or r attributes of the
corresponding p-string feature structure. This means that, ordinarily at least, the
s-string in (38) is not paired with the p-string in (39), for example. Such constraints
on wellformedness at the interface between the syntax and phonology macro-
modules, situated in the string, account for the relationship between c-structure
and p-structure in cases of prosodic disambiguation. They also account for the
ambiguity that remains when only the written form of the sentence is available,
because without p-structure constituent edges to match with, in principle either of
the two c-string/s-string analyses is possible.
When it comes to Interface Harmony, determining exactly what prosodic infor-
mation (for example, boundary type) is important to the syntax, and vice versa, is a
key area for future research. For instance, in the Bengali examples we see that each
NP forms a separate PhP comprising the relevant PWs, with the attendant result that
in the literal reading (Reading 2), the V dekh lam forms a PhP on its own. However,
as we will see in the following sections, information about the location of syntactic
and prosodic constituent boundaries is not the only sort of information that may
be associated with string units and may therefore appear as values of the l and r
420 prosodic structure

features; importantly, any grammatical or prosodic feature may contribute a label

to the feature structure of the s-string or p-string unit at its left or right edge. In
the next two sections, we show how this approach to p-structure can be used to
analyze phenomena in which prosody plays an important role.

11.5 Declarative questions

The first phenomenon which we will explore with respect to prosody and its
interfaces is that of English declarative questions. Similar to the Japanese example
at the start of this chapter, in declarative questions prosody makes a crucial
and unique contribution to the semantic and pragmatic interpretation of the
utterance. Specifically, the intonational contour (or tune) is the only means by
which a declarative question is marked as a type of question with interrogative
semantics, rather than simply an ordinary declarative statement.17 The relevant
tune is characterized by a L(ow) tone preceding a final H(igh) tone; each tone is
associated with a particular syllable, as shown:

(40) Declarative question: declarative syntax, final rise in intonation

Anna was studying at the university?
ˈæ.nə.wəz.ˈstʌ.di.ıŋ.ət.ðə.ju.nə.ˈvɜː.sə.ti
L H

Following a considerable body of work on intonational phonology (for an

overview, see Ladd 2008), we assume that a tune such as the one associated
with interrogative semantics in (40) minimally consists of a nuclear tone and a
boundary tone. The nuclear tone is the pitch target which bears the main stress in
an IntP. Boundary tones appear at one edge or both edges of an IntP. A left boundary
tone is associated with the first syllable in an IntP, while a right boundary tone is
associated with the final syllable in an IntP.
The English interrogative tune in (40) can be characterized as consisting of a
l nuclear tone and a h right boundary tone. In the English interrogative tune,
the nuclear tone is associated with the stressed syllable of the first PW in the last
PhP in the IntP; see, for example, Hayes and Lahiri (1991). A tone is rendered
simultaneously with the associated syllable. In order to model this, we must refer
to the rightmost and leftmost p-string units in a prosodic projection that are
marked for primary stress; that is, the leftmost and rightmost syllables within a
projection that are specified as being the location of primary stress (represented
as the p-string feature syllstress p). Recall that Tl (+) is defined as β−1 (Tl (+)),
the leftmost p-string element of the prosodic structure node +; similarly, Tr (+) is

17 In what follows, we assume that a declarative question has the same interpretation as an equivalent
polar interrogative that is syntactically marked as such by subject-auxiliary inversion, for example Was
Anna studying at the university? This is a simplification, as declarative questions exhibit contextual
restrictions which polar interrogatives do not; on this, see, for example, Gunlogson (2003). Our aim
here is to illustrate the mechanism whereby prosody can have an effect on meaning within the LFG
framework. The full details of an LFG analysis of declarative questions await further research.
 declarative questions 421

defined as β−1 (Tr (+)), the rightmost p-string element. β is the function from p-
string elements to terminal nodes of the prosodic structure, and β−1 is the inverse
of β. An example configuration is:

(41) β−1 (Tl (+)) is the leftmost p-string element of the prosodic constituent +,
and β−1 (Tr (+)) is the rightmost s-string element
+

Tr (+)
Tl (+)
β−1
β−1
β−1 (Tl (+)) β−1 (Tr (+))

Thus, we define the relations Tls and Trs , which pick out the stressed syllable closest
to the left and right edge of a constituent +, as follows:18

(42) Tl ( ) if (β−1 (Tl ( )) ) =c

a. Tls ( ) ≡
else Tls (β(N(β−1 (Tl ( )))))
Tr ( ) if (β−1 (Tr ( )) ) =c
b. Trs ( ) ≡
else Trs (β(N −1 (β−1 (Tr ( )))))

According to the first line of this definition, Tls (+) is the leftmost syllable Tl (+) in
the prosodic phrase if the p-string element corresponding to Tl (+) contains the
feature syllstress with value p. A sample configuration meeting these require-
ments is given in (43):

(43) (β−1 (Tl (+)) syllstress) = p

+

Tr (+)
Tl (+)

β−1
3 4
β−1 (Tl (+)): syllstress p

The second line of the definition of Tls applies in case the leftmost syllable does not
correspond to a stressed element of the p-string. In that case, we check the next
p-string element to the right, and continue rightwards until we encounter a
p-string element containing the feature syllstress with value p.
As usual, we can introduce a convenient abbreviation for these concepts, using
arrows superscripted with s:

18 The function N, which appears in (42), finds the next element in linear order when applied to
string elements, as discussed in §6.11.2; N −1 finds the preceding element.
422 prosodic structure

(44) a. #s ≡ β−1 (Tls (+))

b. $s ≡ β−1 (Trs (+))

In sum, the definitions of the functions #s and $s are similar to the definitions
of # and $ in (33). The only difference is the presence of an additional specifica-
tion: these functions do not necessarily find the absolute leftmost and rightmost
syllables, but rather the leftmost and rightmost syllables that have the feature-value
pair which we use to represent primary stress, i.e. syllstress p. These two pairs of
arrows are all that we require to analyze declarative questions and prosodic focus
marking, which is the subject of §11.6.
For the prosodic part of the analysis of a declarative question, the interrogative
tune is a property associated with an IntP. The specification of this tune is intro-
duced by a prosodic phrase structure rule, parallel to the syntactic phrase structure
rules with which we are familiar:

(45) IntP −→ PhP⃰ PhP

((#s n_tone) = l)
($ rb_tone) = h

Line 1 of the annotation below the final PhP in (45) means that the value of n_tone
is l for the leftmost stressed syllable (i.e. the leftmost p-string unit including the
feature-value pair syllstress p) in this PhP. Line 2 requires the value of the
rightmost p-string unit in this PhP to have an rb_tone (right boundary tone)
feature whose value is h. (45) captures the intonational contour we are concerned
with, but what it does not do is link it to the semantic contribution associated with
this interrogative tune. To capture this, we must turn now to the syntactic part of
the analysis.
As discussed in Chapter 8, s(emantic)-structure is projected from f-structure,
that is, from the syntax. As a consequence, it is necessary to associate the meaning
constructor for interrogative semantics with a syntactic unit representing the
clause as a whole. We assume a constructional meaning, introduced in the IP
phrase structure rule, which contributes the relevant meaning constructor (§8.6).
For present purposes, we define the semantics of polar interrogativity as follows
(which represents a simplification, but suffices for the purposes of illustration in
this case):

(46) Polar interrogative meaning constructor:

λP.Ques(P) : ↑𝜎 ⊸ ↑𝜎

We abbreviate this meaning constructor as [PolarInt]. This interrogative meaning

contribution is associated with the right edge of the root node, IP. The phrase
structure rule for English declarative questions is therefore:

(47) IP −→ XP I!
↑=↓
PolarIntSem ∈ (⇘ r)
[PolarInt]
 declarative questions 423

This rule ensures that a label, “PolarIntSem,” appears as a member of the set value
of r in the rightmost s-string unit corresponding to the root IP. This specifica-
tion always appears together with the polar interrogative meaning constructor
[PolarInt], as illustrated in (52).
Returning now to the prosodic part of our analysis, just as on the syntactic
side, the interrogative tune specification is always accompanied by a further
specification which defines a label PolarInt. This label appears as a member of the
set value of r in the rightmost p-string unit in the IntP:19

(48) PolarInt ∈ ($ r)

Putting together (48) and (45), we have a specification of the interrogative tune
which combines its tonal features with a label relating to polar interrogativity:20

(49) IntP −→ PhP⃰ PhP

((#s n_tone) = l)
($ rb_tone) = h
PolarInt ∈ ($ r)

Following the analysis of Lahiri and Plank (2010), adopted here, the last three
syllables of university form a prosodic word (PW). A partial representation of
the p-structure and p-string for these three syllables illustrates the way this rule
determines certain feature values:
 
(50) fm vɜː  
 syllstress p   fm sə  fm ti
   rb_tone h 
 n_tone l  l { }   
p-string   l { } 
 l { PhP, PW }   
  r { }
r { Utt, IntP, PhP, PW, PolarInt }
r { }
β β
β
S S S

PW

We are now in a position to put together the syntactic and prosodic aspects of our
analysis of declarative questions. Crucial to this analysis are the feature values in the
rightmost string feature structures, specifically “PolarIntSem” in the final s-string
unit (see 47) and PolarInt in the final p-string unit (see (49)), which are shown in
this partial representation:

19 We follow Mycock and Lowe (2013) in distinguishing syntactic labels like “PolarIntSem” from
prosodic labels like PolarInt by putting the latter in italics. Note that these are simply labels, and do
not imply the presence of semantic properties in either the c-structure or p-structure: PolarInt could
equally well be labeled XYZ, or “PolarIntSem” could be “XYZ.”
20 Note that we do not assume a one-to-one relationship between a tune and meaning.
424 prosodic structure

(51) NP
N!
N
π
 
fm university
l { N } 
s-string  
r { IP, VP, PP, NP, N, PolarIntSem }

 
fm ti
 rb_tone h 
 
p-string l { } 
 
r { Utt, IntP, PhP, PW, PolarInt }
β
S

PW

But what is the connection between the labels “PolarIntSem” and PolarInt? Both
are clearly intended to link with polar interrogativity, which is characteristic of
this construction, but the labels and the syntactic/prosodic structures with which
they are associated are not directly dependent on one another: neither is derived
from the other and their sphere of application is entirely separate, in line with
strict modularity. And yet it is clear that the two must be related if the analysis
is to achieve its aim of pairing the interrogative tune with the appropriate meaning
constructor. Once again, the key assumption is the principle of Interface Harmony.
Interface Harmony requires that information about a meaning constructor
introduced at c-structure (for example [PolarInt]), which has its presence recorded
via a label in the s-string (i.e. “PolarIntSem”), must be matched with an equivalent
label in the corresponding p-string unit (i.e. PolarInt). As in other cases, both
labels should be associated with the same feature (i.e. l or r) in the relevant
string units. Because the labels “PolarIntSem” and PolarInt are intrinsically related
to one of the two separate phrase structure rules which introduce either the
relevant meaning constructor or the relevant tune, the appropriate semantics and
intonational contour must also co-occur. In the case of a declarative question,
then, the principle of Interface Harmony applies so that, if “PolarIntSem” appears
as a member of the set value of r in any s-string unit, then the equivalent label
PolarInt must appear in the corresponding p-string unit as a member of the set
value of r (along with the feature values associated with the rest of the interrogative
tune phrase structure rule, as shown in (49) and (50)). In the case of (52), the
relevant labels co-occur and correspond to one another; if they did not, Interface
Harmony would be violated and the resulting structures would be ungrammatical.
The phrases to which these labels are related (IP on the syntactic side, IntP on the
prosodic side) are also wellformed according to (47) and (49). This means that the
structure shown in (52) is grammatical. In this way, the contribution of prosody
to syntax and semantics can be modeled in a framework that does not require a
direct relation between syntactic and prosodic structure, and does not blur the
distinction between these two modules of grammar:
426 prosodic structure

In this section, we have seen how to model p-structure and account for the ways
in which it interacts with syntax and semantics. In the next section, we turn to
another aspect of interpretation that can have a close relationship with prosody:
information structure.

11.6 Prosodic focus marking

In this section, we explore how prosody can contribute to the interpretation of an
utterance as a result of its relationship to information structure. In the following,
we analyze data first discussed by Mycock and Lowe (2013); we broadly follow the
proposals of Lowe and Mycock (2014), but alter their account slightly in order
to maintain consistency with the model of information structure described in
Chapter 10.
In the following English examples, the focused element can be identified as the
element which bears the main or nuclear stress in the utterance (see, for instance,
Ladd 2008). We identify this element as bearing the Nuclear Tone. The Nuclear
Tone is the final pitch accent within the relevant IntP domain. This pitch accent
is perceived as being the most prominent within the intonational contour under
consideration. In a declarative statement such as the answer in (53), the Nuclear
Tone is a h(igh) tone:

(53) Q: Who did Anna hit?

A: Anna hit [Norman]focus .
h

In this example, Norman in the answer corresponds to the queried constituent

in the question, and thus functions as the focus in the reply sentence. Contrast
this with the following example, where Anna corresponds to the questioned
constituent, and hence is the focus. In this case, Anna bears the Nuclear Tone:

(54) Prosodic focus marking

Q: Who hit Norman?
A: [Anna]focus hit Norman.
h

The answer sentences in (53) and (54) exemplify narrow focus: the word that
bears prosodic focus marking is also the only element that bears focus status
at information structure. However, it is also possible for a single word to bear
prosodic focus marking when more than one word has focus status at information
structure. This is known as broad focus, and is exemplified in (55). In (55a), the
whole VP in the answer sentence has focus status, but the Nuclear Tone is associated
with Norman (specifically, with the initial stressed syllable of Norman).21 In (55b),

21 Here, we set aside the issue of an event, but not the type of that event, being presupposed in the
context of What happened? or What did X do (to Y)? type questions. When such a question is asked, it
 prosodic focus marking 427

the subject NP some old woman is the focus of the answer sentence, but the Nuclear
Tone is associated with the initial stressed syllable of woman; cf. (54):

(55) a. Q: What did Anna do?

A: Anna [hit Norman]focus .
h
b. Q: Who hit Norman?
A: [Some old woman]focus hit Norman.
h

Although the distinction between broad and narrow focus is generally accepted in
the literature on focus marking, in an important sense the analysis that we adopt
here transcends the distinction between broad and narrow focus. This distinction
depends on different correlations between focus in the syntax and information
structure on the one hand, and the prosodic marking of focus status on the other. In
the modular grammatical architecture which we assume, these different aspects of
focus and its encoding are specified separately. Important to understanding these
inextricably linked dimensions of the notion “focus” are two concepts which we
refer to as Extent of Focus (Foc-Extent) and Exponent of Focus (Foc-Exponent).
Foc-Extent (also known as the Focus Domain) refers to the portion of a sentence
which can be said to have focus status in information structure terms, while Foc-
Exponent is the indication at some level of representation, for example prosody
(p-structure), of the focus status of part or all of a sentence.22 In the examples in
(55), square brackets enclose the syntactic elements that constitute the Foc-Extent,
while the h Nuclear Tone annotation indicates the word (given in bold) which
bears the focus marking (the main stress in the sentence).
Of course, the precise definition of Foc-Extent depends on the general approach
taken to information structure and its relation to other aspects of linguistic
structure within the grammar. We define the Foc-Extent as the set of meaning
constructors corresponding to syntactic elements that are associated with focus.
In the approach introduced in Chapter 10, elements of a sentence’s meaning—i.e.
meaning constructors, which appear in bold in (56)—are categorized according
to their discourse function (df) at s(emantic)-structure, and consequently belong
to the relevant set (for example topic, focus) at the level of i(nformation)-
structure.

is in fact only the type of the event that is in focus: the occurrence of some event is presupposed. On
this, see Mycock (2006). For the purpose of illustrating the analysis of prosodic focus marking in LFG,
we make the simplifying assumption that a single meaning constructor is associated with a verb, and
thus that the verb is associated with a single discourse function at information structure. If one were to
adopt a neo-Davidsonian approach to verb meaning instead, the distinction between event and event
type could be captured using the s-structure attributes rel and event (see §9.10.5.2), and these two
aspects of verb meaning could be analyzed as belonging to different information structure categories.
22 Of course, information structure may also be indicated at other levels of structure, for example
c-structure, but our concern here is with prosodic marking.
428 prosodic structure

In (56), Anna is the subject of a clause, and is also the topic at i-structure by virtue
of appearing in the specifier of IP. Key to this analysis is the attribute-value pair df
topic included in the s-structure for ‘Anna’, a𝜎 . This information, combined with
the annotations on the terminal node and the specifier of IP node in the c-structure,
serves to categorize the relevant meaning constructor as belonging to the topic set
in the clause’s i-structure, f𝜎 ι :

(56) IP

NP
(f a
I
f ι =a ι
((a f : a:
N
a :
Anna
(a f ι: [Anna] }
[Anna] ∈ (a ι (a

Given the approach to i-structure exemplified in (56), Foc-Extent is equivalent

to the set of meaning constructors which are the value of the attribute focus at
i-structure. These meaning constructors are semantic units and thus correspond—
sometimes imperfectly—to units at other levels of representation, such as syntactic
constituents.
It is possible to provide a relatively straightforward informal generalization
concerning the relationship between Foc-Extent and Foc-Exponent in English:
the Foc-Exponent is associated with the Prosodic Word which corresponds to
the rightmost syntactic word that is the syntactic realization of the Foc-Extent.
However, a full formal analysis represents a significant challenge, in large part
because of the extensive misalignment that is a feature of the correspondences
between units belonging to different levels of representation.
When we consider Foc-Exponent and Foc-Extent separately, the difference
between narrow and broad focus collapses to an extent. Based purely on their
syntax, these two types of focus are fundamentally the same: in all of the examples
in (53), (54) and (55), the Foc-Extent is a single syntactic constituent. Similarly, the
basic facts about the Foc-Exponent are the same in all cases: the rightmost Prosodic
Word of the focus constituent bears the Nuclear Tone.
In order to provide an analysis of prosodic focus marking, we employ the
mechanisms introduced previously, including the arrows defined in §11.4.1.
As with declarative questions, we represent the prosodic contour of a phrase
with focus marking by means of prosodic phrase structure rules. Focus marking
of the kind under discussion involves sentences that are declarative statements. We
assume this prosodic phrase structure rule for declarative statements:23

23 For the use of the ampersand here, see §6.1.1.

 prosodic focus marking 429

(57) IntP → [PhP⃰ PhP PhP⃰]

($s n_tone) = h
& [PhP⃰ PhP ]
($ rb_tone) = l

In addition to this, we require another rule to capture the focus-specific elements of

the prosody of such clauses. It is possible to formulate a single p-structure rule (58)
to account for nearly all the types of intonational focus marking in which we are
interested. This rule states that any Prosodic Word (PW) in a Phonological Phrase
may be marked for n_tone = h, and if so, the label DFFoc appears as a member of
the r value of the corresponding p-string element that carries the Nuclear Tone.24
The p-string element in which this feature appears is, specifically, the rightmost
syllable that bears primary stress ($s ) in the relevant PW.25

(58) PhP → PW⃰ PW PW⃰

($s n_tone) = h ⇒
DFFoc ∈ ($s r)

The Foc-Exponent, captured by the rule in (58), serves to delimit the Foc-Extent.
Foc-Extent is expressed via separate c-structure rules which, together with the
p-structure rule in (58) and the principle of Interface Harmony, play an equally
important role in the analysis of prosodic focus marking as an interface phe-
nomenon.
In the case of the sentence in (53), the focus constituent in the answer sentence
(Norman) could be identified with either the NP, the N! , or the N which dominates
the word Norman in the c-structure. Following Mycock and Lowe (2013), we
assume here that the phrase in question is the NP, since focused constituents often
correspond to XP categories, but it makes no difference to the analysis of the present
example.
We propose the following node annotation principle for Extent of Focus in
English:

(59) Any c-structure node may optionally be annotated with the following two
specifications:
DFFoc ∈ (⇙l)
DFFoc ∈ (⇘r)

These constraints specify that the label “DFFoc” appears as a member of the
set value of the features l in the leftmost string element (⇙l) and r in the

24 As with PolarInt above, note that DFFoc is simply a label; it does not situate discourse features in
p-structure.
25 We specify and require harmony for only the r features in the s-string and p-string. This reflects
the fact that focus marking in English is fundamentally right-edge based. We make no claims as to
whether or not the same is true for other languages.
430 prosodic structure

rightmost string element (⇘r) corresponding to the c-structure node on which

the annotations appear. Thus this label serves to effectively delimit the focused
constituent. In the present example, the annotated node is an NP consisting of only
a single word (Norman), so the leftmost and rightmost string elements dominated
by the NP are the same. This is not the case when the phrase in focus consists of
more than one word, as, for example, in instances of “broad” focus (see (55)), and
it is for this reason that separate specifications are required for the left edge and
right edge of the phrase.
The principle in (59) specifies only the presence of a label, “DFFoc,” as a member
of the l and r sets in particular string elements. On its own, it does not say anything
about the information structure status of the words in the focused phrase. In order
to specify this, we must introduce a wellformedness constraint on strings, or a
string constraint (see Lowe and Mycock 2014 and §6.11.1). String constraints define
the valid s-strings or p-strings for a language. A string constraint consists of a
regular expression encoding the constraint, and a label which specifies the aspect
of the string referred to by the constraint (here s-string, the other possibility being
p-string). We make use of the symbol Σ, which represents a string element of any
form.
The labels specified in (59) serve to delimit a span in the s-string: the leftmost
s-string element in the span is marked by having the label “DFFoc” as a member of
the l value set, while the rightmost one is marked by having the label “DFFoc” in
its r value set. Our aim is to define a string constraint which requires every word in
the relevant span to be marked in such a way that its meaning constructor appears
in the focus set at i-structure. For clarity, we define the following abbreviatory
templates:

(60) a. inclLeft(_α) ≡ _α ∈c (• l)
b. inclRight(_α) ≡ _α ∈c (• r)
c. sdf(_α) ≡ (↑𝜎 df) = _α
d. nsdf(_α) ≡ (↑𝜎 df) /= _α

The template inclLeft ensures that its argument _α is a member of the l set of
the current string element, and similarly for the template inclRight. The template
sdf specifies that its argument _α is the value of the df attribute of the s-structure
↑𝜎 , and nsdf specifies that the df value of ↑𝜎 is different from _α. These basic
templates are used in the definition of the templates lfocus and rfocus:

(61) a. lfocus ≡ @incLeft(DFFoc)

@sdf(focus)
b. rfocus ≡ @incRight(DFFoc)
@sdf(focus)

The f-description lfocus holds of a string element if “DFFoc” is a member of its

l set, and the value of the df attribute in its mother’s semantic structure is focus.
 prosodic focus marking 431

Recall that we use the up arrow ↑ in lexical entries to refer to the c-structure node
accessible from an s-string element via the π−1 relation (§11.4.1):

NP
(62)
DFFoc ∈ (⇙l)

3 4 3 4
N pred ... df focus
𝜙 𝜎
π−1
53 46
l DFFoc

Similarly, rfocus holds of a string element if “DFFoc” is a member of its r set, and
the df value of its mother’s semantic structure is focus:

NP
(63)
DFFoc ∈ (⇘r)

3 4 3 4
N pred ... df focus
𝜙 𝜎
π−1
53 46
r DFFoc

We then define the metacategory in (64), representing the Foc-Extent in terms of

s-string items:

(64) ΣFoc ≡ { Σ Σ⃰ Σ | Σ }
@lfocus @sdf(focus) @rfocus @lfocus
@rfocus

The definition in (64) is a disjunction over two possible sequences: in both

sequences, the leftmost element must be associated with the feature df focus at
s-structure and must have the label “DFFoc” as a member of its l value set (in
both cases, as a result of application of the template @lfocus); likewise in both
sequences, the rightmost element must be associated with df focus at s-structure
and must have the label “DFFoc” as a member of the r value set (by the application
of the template @rfocus). The only difference between the two possibilities is
whether the Foc-Extent is more than one s-string unit (that is, whether the s-string
element with l {DFFoc} is different from the element with r {DFFoc}), or only one
s-string unit with “DFFoc” as a member of both its l and r value sets. If there are
more than two elements in the string, the elements that are neither leftmost nor
rightmost must be associated with df focus at s-structure. This is achieved by
application of the template @sdf(focus).
We can use these definitions in formulating the s-string constraint in (65),
which we call focus. It allows for the appearance of several focused constituents
in a string, each preceded and followed by string elements that are not in focus
432 prosodic structure

(marked with @nsdf(focus)). The regular expression focus string constraint

constitutes one component of the s-string constraints that are relevant for English.
It is intersected with all of the other s-string constraints for English to obtain the
fully specified s-string wellformedness condition, as discussed in §6.11.1.
7 8
(65) s-string focus: ( Σ⃰ ΣFoc Σ⃰ )+
@nsdf(focus) @nsdf(focus)

The rule in (65) permits one or more sequences consisting of any number of
string elements that are not associated with focus at s-structure (by the template
@nsdf(focus)), optionally followed by a sequence fitting the definition of ΣFoc ,
followed by any number of string elements that are not associated with focus
at s-structure. That is, for any string in English the rule in (65) licenses one or
more substrings of which every element is associated with df focus at s-structure,
and of which the leftmost and rightmost elements are marked by the appearance
of “DFFoc” in their l or r value sets respectively. Following the approach to
information structure presented in Chapter 10, the association with df focus at
s-structure has the effect that the meaning constructors corresponding to the s-
string unit(s) in this sub-string appear in the focus set at i-structure.
The p-structure rules in (57) and (58), the c-structure constraint in (59), and
the s-string constraint in (65) are the basis for the analysis shown in (66) for the
example given in (53).
In this utterance, the stressed syllable of the word ‘Norman’ /nɔː/ is associated
with the Nuclear Tone (n_tone h). The prosodic phrase structure rule in (58)
may thus apply to the Prosodic Word dominating that syllable, in this case the
final Prosodic Word in the p-structure pronounced /nɔː.mən/. As a result of
(58) applying, the label DFFoc appears as a member of the set value of r in the
syllable which has primary stress within this Prosodic Word, i.e. the first syllable of
‘Norman’. On the syntactic side, the phrase-structure rule in (59) applies to the NP
that corresponds to the s-string element ‘Norman’. By this rule, the label “DFFoc”
must appear in the l and r sets in the corresponding leftmost and rightmost
s-string elements respectively, which in this case are the same element. By the
string constraint in (65), the meaning constructor corresponding to this s-string
element, that is the meaning constructor introduced in the lexical entry for the
name Norman, appears in the focus set at information structure. The principle of
Interface Harmony requires that an s-string element which has a feature “DFFoc”
in its r value set must be associated with a p-string element that has the feature
DFFoc in its r value set. In (66), this requirement is satisfied, and the analysis
succeeds. If, say, the c-structure rule in (59) were to apply not to the final NP in
the c-structure but to the first one (meaning that Anna was in focus at information
structure), the analysis would fail because “DFFoc” would appear in the r set in
the s-string element Anna but would not be associated with a DFFoc feature in a
corresponding p-string element. That is, prosodic and syntactic specifications are
effectively required to match up, such that whichever prosodic element bears the
Nuclear Tone necessarily corresponds to the syntactic element associated with the
meaning constructor(s) that represent the focus at information structure.
 prosodic focus marking 433

(66) Anna hit [Norman]focus .

IP h: a:
n:

n :
NP I

ι
 
VP Norman : n
 
h ι : 
.. ..
. .
N V

V NP
DFFoc ∈ (⇙L)
π
DFFoc ∈ (⇘R)
π

N
   π 
 
    { NP, N, DFFoc } 
s-string  { IP, NP, N } 
  I , VP, V , V   
   
{ NP, N } {V} IP, I , VP, V , NP, N, DFFoc

 ɔː 
   
 ə     ən 
     
   { }       { } 
p-string  IntP, PhP, PW    


PhP, PW 



  
     PhP, PW 
PhP, PW   IntP, PhP, PW
{ } PhP, PW
{ DFFoc }

β β β β β

S S S S S

PW PW PW
( s ⇒
DFFoc ∈ ( s

PhP PhP PhP

IntP

Precisely the same set of rules can equally well account for instances of broad
focus, where the Foc-Extent in the c-structure and s-string spans more than one
word. Example (67) shows the analysis for (55a), an example of broad focus in
which the Foc-Extent is hit Norman. The only difference from the structure in (66)
is that the c-structure rule in (59) applies at the VP level, meaning that the label
“DFFoc” appears in the l set in the s-string element corresponding to the word hit.
This in turn means that the meaning constructor for hit appears in the focus set
at i-structure along with that for Norman, whose r set in the s-string contains the
label “DFFoc” due to the annotation DFFoc ∈ (⇘r) on the VP node.
434 prosodic structure

(67) Anna [hit Norman]focus .

IP h: a:
n:
ι

n :

NP I h :

 ι 
Norman : n
VP 



h ι : hit : h 
DFFoc ∈ (⇙L) 
.


.. ..
DFFoc ∈ (⇘R) .
N

π
V NP

π N
   π 
 
    { NP, N } 
s-string  { IP, NP, N } 
  I , VP, V , V, DFFoc   
   
{ NP, N } {V} IP, I , VP, V , NP, N, DFFoc

 ɔː 
   
 ə     ən 
     
   { }       { } 
p-string  IntP, PhP, PW    


PhP, PW 



  
     PhP, PW 
PhP, PW   IntP, PhP, PW
{ } PhP, PW
{ DFFoc }

β β β β β

S S S S S

PW PW PW
( s ⇒
DFFoc ∈ ( s

PhP PhP PhP

IntP

In this section, we have provided an analysis of prosodic focus marking in

English. The approach which we have presented can also be used to analyze
prosodic marking of this and other information structure categories crosslinguis-
tically.
 further reading and related issues 435

11.7 Further reading and related issues

Bögel (2012, 2013, 2014, 2015) has developed a distinct approach to prosody
within the LFG framework whose formulation makes crucial reference to actual
values of the acoustic characteristics of the speech signal. Bögel’s proposals are
in part influenced by the proposals of Dalrymple and Mycock (2011), but in
certain respects depart significantly from other approaches to prosody within
LFG. Bögel (2015) argues against a dependence on the prosodic hierarchy as a
means of modeling p-structure, and proposes a model based on the “p-diagram.”
The “p-diagram” represents speech as a series of feature vectors, each of which
encodes the phonetic and phonological properties of a single syllable. Bögel’s
prosodic analysis is therefore fundamentally syllable-based, although in principle
it could be made more fine-grained. The phonetic and phonological properties
of individual syllables are accessible by other components of the grammar via a
projection function. Bögel uses this approach in her analysis of prosodic resolution
of syntactic ambiguities (see Bögel 2015: 70–7), as well as other phenomena.
Clitic positioning and the treatment of syntax-prosody mismatches are dis-
cussed by O’Connor (2002a,b, 2005a), Lowe (2011, 2016a), and Bögel (2014, 2015).
Mycock (2006, 2007) presents an LFG approach to constituent question formation
crosslinguistically which includes description and analysis of question intonation,
and Mycock (2010) discusses the modeling of prosody and its interfaces, exempli-
fying the analysis with data from Hungarian. It must be noted that in these works
the authors do not necessarily assume the version of LFG’s parallel architecture
that is adopted here.
 12
The interface to morphology

In this chapter, we examine the place of morphology in the LFG architecture. Work
on the morphology-syntax interface in LFG (notably by Sadler and Spencer 2001
and Kaplan and Butt 2002) assumes a modular view of the morphological compo-
nent, in line with the overall modular architecture of LFG: the morphological com-
ponent has its own internal structure and obeys universal and language-particular
constraints on word formation that need not be shared by other levels of structure.
As Sadler and Spencer (2001) point out, this view accords well with Beard’s
Separation Hypothesis (Beard 1988, 1995), which argues for a separation between
the representation of the phonological form of a word and the features encoding
the morphosyntactic and morphosemantic contribution that the word makes.
Following Sadler and Spencer (2001), Kaplan and Butt (2002), Spencer (2006,
2013), and many others, we assume that the morphological component associates
a word form with a set of features representing the morphological structure and
grammatical contribution of the word, and that these features are interpreted at the
interface of morphology with the rest of the grammar, producing the lexical entry
for the word form. This view fits well with a realizational theory of morphology1
as proposed by, among others, Anderson (1992), Stump (2001, 2002, 2006, 2012),
and Brown and Hippisley (2012), though, like Sadler and Spencer (2001), Kaplan
and Butt (2002), and Andrews (2005), our proposals are compatible not only
with explicitly paradigm-based models, but with any morphological theory which
relates words to feature sets encoding their grammatical properties and structure,
including finite state theories of morphology (Kaplan and Kay 1994; Beesley and
Karttunen 2003). Our work builds on and extends work on morphology and LFG
by Butt et al. (1996a), Butt et al. (1996b), Frank and Zaenen (2002), Sadler and
Nordlinger (2004), Andrews (2005), Otoguro (2006), Bond (2016), and especially
Sadler and Spencer (2001) and Kaplan and Butt (2002).

12.1 The morphological component and the lexicon

In the overall architecture of the grammar of a language, the morphological
component serves to populate the lexicon with lexical entries. We first review

1 Stump (2001) presents a taxonomy of morphological theories, contrasting realizational with

incremental theories, and lexical with inferential theories. In terms of this taxonomy, we adopt an
inferential/realizational approach.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 the morphological component and the lexicon 437

our definition of the lexical entry, and we then show how the morphological
component, together with a theory of the interface between morphology and the
rest of the grammar, produces the complete set of lexical entries for a language.

12.1.1 The lexical entry

The full lexical entry for the plural noun dogs is given in (1), following the
discussion in §11.4.1. As elsewhere in the book, we use the familiar up and
down arrows to refer to f-structures rather than the technically correct but more
complex notation given in (23) of Chapter 11; recall that the s-form is the lexical
representation specifying the s(yntactic)-string associated with a word, and the
p-form is the lexical representation specifying the p(rosodic)-string contributed
by that word:

(1) Full lexical entry for dogs:

s-form (• fm) = dogs
c-structure category λ(π(•)) = N
f-description (↑ pred) = ‘dog’
(↑ index num) = pl
[dog] ∈ (↑𝜎ι (↑𝜎 df))
[pl] ∈ (↑𝜎ι (↑𝜎 df))
p-form /dɔgz/

A note about terminology is in order. Morphologists often use the term “lexical
entry” to refer to information associated with a lexeme rather than a word form,
under the assumption that the contribution of a word form depends on the
contribution of the lexeme, specified in the lexical entry for the lexeme, together
with the features of the particular word form. We follow common LFG practice
in using the term “lexical entry” to refer instead to a word form (here, the plural
noun dogs) and its associated grammatical information, as in (1). We use the term
lexemic entry to refer to a lexeme and its associated grammatical information. The
aim of this chapter is to relate the lexical entry for a word form to the morphological
feature set associated with the word form by the morphological component, given
its lexemic entry.2
For ease of presentation, and since our focus in this chapter is not on the details
of syntactic representation or semantic composition, we will assume simplified
f-descriptions containing only certain f-structure features, and omitting many
details of f-structural constraints as well as semantic and other non-f-structural

2 In this chapter, we describe the task of producing the lexical entry for a word form, including its
f-description, on the basis of its morphological features. Importantly, this is only for ease of exposition,
and there is no directionality intrinsic to the theory; our overall aim is simply to define and constrain the
relation between lexical entries and lexemic entries for a language in light of a theory of morphological
realization for the language. Indeed, the opposite order of exposition is often adopted in the setting of
Paradigm Function Morphology (Stump 2001), where a mapping from f-descriptions to morphological
features is described.
438 the interface to morphology

information. Thus, we will work with a simplified entry such as the one in (2),
which we will use as a stand-in for the full entry in (1):

(2) Simplified entry, including only basic syntactic constraints:

s-form (• fm) = dogs
c-structure category λ(π(•)) = N
f-description (↑ pred) = ‘dog’
(↑ index num) = pl
p-form /dɔgz/

The lexical entry in (2) encodes the same syntactic information as the traditional-
style LFG lexical entry in (3), but also contains information about the word’s
phonological form p-form.

(3) Traditional lexical entry (§5.3.2):

dogs N (↑ pred) = ‘dogs’
(↑ index num) = pl

Kaplan and Butt (2002)3 propose an alternative way of representing lexical entries,
by means of a relation which they call M; to avoid confusion with the function
M from c-structure nodes to their mother nodes in the tree, we call this relation
L. For Kaplan and Butt, L is a three-place relation involving a word form, a
c-structure category, and an f-description: they propose the representation in (4),
incorporating the same information as the lexical entry in (3).

(4) (3) as an L relation according to Kaplan and Butt (2002):

L<dogs, N, {(↑ pred) = ‘dogs’, (↑ index num) = pl}>

We extend L to relate the different components of the enhanced lexical entry in

(2): an s-form, a p-form, a possibly complex c-structure category (for definition
and discussion of complex c-structure categories, see §3.3.5), and an f-description.
For us, then, L is a four-place relation:

(5) L<s-form, p-form, category, f-description>

Rewriting the lexical entry in (2) in this format, the L relation for the plural noun
form dogs is:
(6) The lexical entry in (2) as an L relation:
L<dogs,
/dɔgz/,
N,
{(↑ pred) = ‘dog’, (↑ index num) = pl}>

3 We follow the presentation in the unpublished Kaplan and Butt (2002) draft paper, which differs
slightly from the version in the handout for the talk presented at the LFG02 conference.
 the morphological component and the lexicon 439

The representation in (6) encodes exactly the same information as the representa-
tion in (2), but facilitates the discussion of the definition of L in the following. Our
task, then, is to provide a way of defining the set of lexical entries L for a language,
given an independently specified set of lexemic entries and an independently
specified theory of morphological realization for the language.
In sum, the aim of this chapter is to elucidate the interface between the morpho-
logical component and the rest of the grammar, and to demonstrate how the set of
lexical entries for a language is produced, given a set of lexemic entries and a theory
of morphological realization for the language. We will provide a general framework
which allows exploration of alternative theories of morphological realization and
its interface with the rest of the grammar in an LFG setting.

12.1.2 Lexemic entries and the Lexemic Index

We first provide some information about our expectations governing the structure
of lexemic entries and how they are related. A lexemic entry contains information
of the kind which we would expect to find in an ordinary dictionary. Spencer
(2013) divides this information into the following four components:

(7) Lexemic entries according to Spencer (2013):

FORM: the form of the root and any non-predictable stem forms
SYN: syntactic information and requirements
SEM: a representation of the meaning of the lexeme
LI: a Lexemic Index, an arbitrary label identifying the lexeme

Our approach is broadly compatible with Spencer’s view, though it is clear that
word forms may be associated with information of other types as well: for example,
the presence of certain affixes might mark a particular information structure
role (Dalrymple and Nikolaeva 2011). Although the difference between the SYN
and SEM components of the lexemic entry is important to the theory of lexical
relatedness proposed by Spencer (2013), our primary concern in this chapter is to
demonstrate how a realizational theory of morphology can be integrated into the
architecture of LFG. Hence, in our discussion of the morphological component
and its place in the grammar, we do not enforce a distinction in the lexemic entry
between syntactic, semantic, and other kinds of information.
We follow Spencer (2013) in assuming that each lexemic root is associated
with a unique identifier, its Lexemic Index (LI).4 The LI cannot be equated
with the semantic form, at least if we assume that the semantic form defines
subcategorization requirements, since morphologically related forms with the
same LI may have different subcategorization frames. Even if we assume that
the LI corresponds to a subpart of the semantic form, the fn (see §5.2.2.3), it
is unclear whether identity of lexemic index for two different word forms has
any consequences for compositional syntax or semantics, and thus whether the
morphological derivational history of a lexeme should be encoded in the semantic
form. For clarity, we provide mnemonic LIs such as dog1 for the lexeme related

4 The Lexemic Index is similar to the LexID proposed by Stump (2001).

440 the interface to morphology

to the noun forms dog/dogs, but we assume that the LI plays a role only within
the morphological component, and is not visible to any other component of the
grammar.
A lexemic entry is, then, a three-place relation LE involving (i) the form of the
root and any non-predictable stem forms, together with morphological features
encoding information about inflectional class, category, and other unpredictable
features; (ii) an f-description that encodes syntactic, semantic, and other informa-
tion associated with all word forms of the lexeme, filling the role of Spencer’s SYN
and SEM; and (iii) the Lexemic Index, which we abbreviate as LI:5

(8) General form of lexemic entry:

LE <root/idiosyncratic stem forms/features, f-description, LI>

For example, we assume that the lexemic entry that is relevant in the analysis of
the plural noun dogs can be represented as follows, using the LI dog1, and leaving
the morphological features unspecified for the moment:6

(9) Lexemic entry for the lexeme dog1 with morphological features unspecified:
LE <morphological features, {(↑ pred) = ‘dog’}, dog1>

The f-description that is associated with all forms of the lexeme dog1 is
{(↑ pred) = ‘dog’}: this is the portion of the f-description that the word forms
dog and dogs of this lexeme have in common. We now turn to a discussion of
the morphological features which we expect to find as the first argument of this
lexemic entry.

12.1.3 Morphological features

For the lexeme dog1, we propose the following full lexemic entry, including the
appropriate set of morphological features as the first argument:

(10) Full lexemic entry for the lexeme dog1:

LE <{m-root:dog, m-iclass:1, m-cat:noun}, {(↑ pred) = ‘dog’}, dog1>

Following Sadler and Spencer (2001) and Spencer (2006), we use the term m-
feature for morphological features, and we use the m-prefix for particular m-
features.
There are several types of m-features. Features that are relevant only within
the morphological component are morphomic features. A standard example of a
morphomic feature is inflectional class. Inflectional class is specified as a part of

5 The lexemic entry relation LE is similar to what Andrews (2005) refers to as the lookup function,
for which he uses the symbol L.
6 Recall that for simplicity, we use the pred feature and its value to stand for the full f-description
associated with a root. In a full treatment, the f-description in the lexemic entry would include more
information, including (at least) the semantic content of the root: [dog] ∈ (↑𝜎ι (↑𝜎 df)), with an
appropriate definition of the meaning constructor abbreviated as [dog].
 the morphological component and the lexicon 441

the lexemic entry, and we will see that it can play a role in determining the lexical
entry for a word form.
Besides morphomic features, some m-features are relevant in determining the
syntactic contribution of a word form: these are morphosyntactic m-features,
relevant in defining c-structure or f-structure relations and constraints. Others are
associated with morphological signaling of meaning: these are morphosemantic
m-features. Still others are relevant at other levels of structure: for example, a word
form may bear morphological marking to indicate its information structure role.
A single m-feature can also play multiple roles, determining the contribution of a
word form at more than one level of structure.
Thus, the lexemic entry in (10) associates the lexeme dog1 with an m-feature
m-root which specifies the root form of the lexeme dog1, an m-feature m-
iclass which specifies inflectional class, and an m-feature m-cat which specifies
the lexical category of the lexeme. We do not rely on a particular theory of
morphological realization or inflectional classes, and so we use the arbitrary integer
1 to represent the inflection class (m-iclass) of the lexeme dog1.
The lexemic entry also includes all of the unpredictable stems for a lexeme. For
example, the lexemic entries for the verb roots print and drink are given in (11);
drink differs from print in that its past tense form drank and past participle form
drunk are irregular, and must be lexically specified. Again, we provide arbitrary
inflectional class information for these lexemes:7

(11) a. Lexemic entry for print:

LE <{m-root:print, m-iclass:2, m-cat:verb},
{(↑ pred fn) = print},
print1>
b. Lexemic entry for drink:
LE <{m-root:drink, m-stem1:drank, m-stem2:drunk, m-iclass:3,
m-cat:verb},
{(↑ pred fn) = drink},
drink1>

Depending on the theory of morphology that is assumed, different kinds of mor-

phological information specified by different sets of m-features may be required.
Again, since our analysis is compatible with a range of realizational morphological

7 Unlike the lexemic entry for dog in (10), only the pred fn value for the verbs print and drink is
specified in (11). This is because the lexemic entry for a verb like print or drink produces both active and
passive lexical entries, with different semantic forms, and the correlation between passive morphology
and passive argument structure must be maintained: the semantic form for the active version of print
is ‘print#subj,obj$’, and the passive version is ‘print#oblagent ,subj$’. The f-description for a verbal
lexical entry, including the full semantic form linking thematic roles with grammatical functions, is
specified as required by mapping theory (Chapter 9) on the basis of its functional features, including
voice features. For an explanation of the structure of semantic forms and reference to parts of a
semantic form by features such as fn, see §5.2.2.3.
442 the interface to morphology

theories, we are agnostic about the details of the content and representation of m-
features encoding purely morphological information.
To summarize, every language specifies a set of lexemic entries LE which asso-
ciate a Lexemic Index with a set of morphological m-features and an f-description.
The detailed content of the m-features is dependent on the particular theory of
morphological realization that is assumed, and we place no further constraints on
the form of individual lexemic entries and the relations between them.

12.1.4 The morphological realization relation R

The second relation relevant to defining L is the morphological realization relation
R, which encodes a relation between a word form and its associated morphological
features. Our modular theory of the interface between the morphological compo-
nent and the rest of the grammar makes no assumptions about the precise nature
of R or the internal details of the morphological component; in the current context,
R is simply a means of associating m-features with p-forms and s-forms relative to
a lexemic root.
Specifically, R is a set of four-place relations which we call m-entries, associating
a Lexemic Index, an s-form, and a p-form with a set of m-features:

(12) General form of m-entry:

R <LI, s-form, p-form, m-features>

In the setting of Stump’s (2001) version of Paradigm Function Morphology, R is the

Paradigm Function. In more recent versions of the theory (Ackerman and Stump
2004; Stewart and Stump 2007; Stump 2016), m-features specify a cell in the form
paradigm, and the s-form and p-form occupy a cell in the realized paradigm, related
to the form paradigm by paradigm linkage principles.
For example, the m-entry for the plural noun dogs relates the Lexemic Index
dog1, the s-form ‘dogs’, the p-form /dɔgz/, and the set of m-features associated
with plural nouns, {m-cat:noun, m-num:pl}:

(13) M-entry for dogs:

R <dog1, dogs, /dɔgz/, {m-cat:noun, m-num:pl}>

To define the lexical entries for the word forms of a language, we require access
to the complete set of m-entries as defined by the morphological realization
component R.8 The realization relation R for a language accounts for all aspects
of the morphological realization of word forms in the language, encompassing a

8 We do not take a position on whether some or all lexical entries for morphologically invariant
forms are stored separately, rather than being defined on the basis of m-entries for those forms: for
example, whether there is an m-entry for the English determiner the from which the lexical entry for
the is derived. It may be that such forms are in fact stored as underived lexical entries, and do not play
a role in the morphological component.
 the morphological component and the lexicon 443

theory of derivational and inflectional morphology, and encoding generalizations

about affix ordering, stress placement, and other morphological patterns for the
language. The proper theory of R and its internal structure are of vital interest
to morphological theory, but from the point of view of the interface between
morphology and the rest of the grammar, we require only a theory which provides
an empirically adequate and explanatory set of m-entries for a language, and we
will not be concerned with the internal structure and definition of R.

12.1.5 The functional description function D

We now turn to the main focus of this chapter: providing a means of interpreting
the m-features which are associated via the realization relation R with a word form,
producing the c-structure category and an f-description reflecting morphologically
encoded information for the word form.9 We use the abbreviation f-descriptionM ,
with subscript M, for the portion of the f-description for a word form that is
morphologically encoded. Our third function D maps a set of m-features provided
by R to the appropriate c-structure category and f-descriptionM for a word form:

(14) General form of the description function D:

D <Lexemic Index, m-features, category, f-descriptionM >

D is what Kaplan and Butt (2002) call the “description function,” and what Sadler
and Nordlinger (2004) call a “lexical transducer,” relating m-features and their val-
ues to grammatical specifications. In a Paradigm Function Morphology setting, D
corresponds to what Stump (2002) calls rules of paradigm linkage, or (roughly) the
correspondence relation between the form paradigm (pairing a lexeme with a set
of m-features) and the content paradigm (pairing a lexeme with an f-description)
in work by Stump (2015).
In §12.1.4, we saw that the realization relation R produces an m-entry which
associates the set of m-features {m-cat:noun, m-num:pl} with the s-form dogs,
the p-form /dɔgz/, and Lexemic Index dog1, as shown in (13). For the plural noun
dogs, D interprets the m-features {m-cat:noun, m-num:pl}, mapping them to the
c-structure category N and the f-description {(↑ index num) = pl}:

(15) D <dog1,
{m-cat:noun, m-num:pl},
N,
{(↑ index num) = pl}>

9 Marcotte and Kent (2010) explore the possibility of eschewing the D mapping and requiring the
realizational component R to produce a full f-description rather than a set of m-features. We side
with Sadler and Spencer (2001), Kaplan and Butt (2002), and many others in recognizing the need to
distinguish between the m-features relevant within the morphological component and the f-description
that is relevant outside it.
444 the interface to morphology

In this example, there is a simple mapping from the m-feature m-cat:noun to

the c-structure category N, and from the m-feature m-num:pl to the f-structure
constraint (↑ index num) = pl. We postpone a full discussion of the description
function D to §12.3, where we will see that D can be considerably more complicated
than this simple example indicates.

12.1.6 L defined in terms of D, LE, and R

Still building on work by Kaplan and Butt (2002), we can now provide an explicit
definition of the lexicon L in terms of D, LE, and R:10

(16) L is the set of lexical entries for a language:

L = {<s-form, p-form, category, f-descriptionL ∪ f-descriptionM >:
LE <m-features, f-descriptionL , LI> and
R <LI, s-form, p-form, m-features> and
D <LI, m-features, category, f-descriptionM >}

Legitimate members of L are those which meet the conditions imposed by LE, R,
and D. First, a lexemic entry LE must provide a partial f-descriptionL associated
with a Lexemic Index LI and a set of m-features. Second, the morphological
realization relation R must include an m-entry relating the LI and the m-features
to an s-form and a p-form. Third, the description function D must map the m-
features to a category and f-descriptionM . The f-description for the lexical entry
is, then, obtained by combining (via set union, ∪) the f-descriptionL intrinsic to
the lexeme and the f-descriptionM obtained from the m-features for the particular
word form in question. We can represent the requirements in (16) graphically, as
in (17):

(17) Lexemic entries LE

<[morphological features and forms], f-descriptionL , LI>

Morphological realization R Description function D

<LI, s-form, p-form, m-feats> <LI, m-feats, category, f-descriptionM >

L<s-form, p-form, category, f-descriptionL ∪ f-descriptionM >

10 The definition in (16) characterizes a set by describing its elements, as discussed in Foot-
note 14 of Chapter 6. In (16), L is a set whose elements have the form <s-form, p-form,
category, f-descriptionL ∪ f-description M >; elements of that form are included in L if they satisfy the
description in the next three lines.
 the morphological component and the lexicon 445

For the word form dogs, we have the following, as desired:

(18) Lexemic entry LE

< ↑ >

Morphological realization R Description function D

< ɔ > < ↑ >

L<dogs, /dɔgz/, N, {(↑ )= (↑ )= >

As advocated by Kaplan and Butt (2002), m-features and the Lexemic Index are
crucial in defining the description function D, but neither the Lexemic Index nor
any of the m-features appear in the lexical entries L which are accessible to the rest
of the grammatical system. It is only the information in the lexical entries in L—
the s-form, p-form, c-structure category, and f-description—that is visible outside
the morphological component. This maintains a clean separation between the
morphological component and the other components of the grammar. In this way,
our proposal aligns itself with the Principle of Morphology-Free Syntax (Pullum
and Zwicky 1988; Zwicky 1992), and contrasts with proposals that reject the
Lexical Integrity Principle, including Distributed Morphology (Halle and Marantz
1993; Embick and Noyer 2007) and the Exo-Skeletal Model (Borer 2013).

In sum, the following components are required for morphological analysis:

• A set of lexemic entries LE for a language, associating Lexemic Indices

(LIs) with morphological roots, stems, and their idiosyncratic morphological
features, and an f-description containing idiosyncratic syntactic, semantic,
and other information and constraints for the lexeme.
• A morphological realization relation R for the language which relates full
word forms and their associated m-features to a Lexemic Index.
• A description function D which interprets the m-features provided by the
realization component R, producing fully specified lexical entries L for the
word forms of the language.

The set of lexemic entries LE and morphological realization relation R are indepen-
dently specified for each language, in conformance to a suitable realizational theory
of morphology. In the remainder of the chapter, our focus is on the definition of the
description function D, which constitutes the relation between morphology (sets
of m-features) and the information and constraints that are relevant for the rest of
the grammar.
446 the interface to morphology

12.2 Morphological features and morphological classes

Our aim is to provide a framework which allows the linguist to define the proper
D-mapping for a language in a clear and intuitive way, given a particular set of
lexemic entries and a theory of the realization relation R for the language. In
this section, we show how the D description function works by presenting some
examples of mappings which are established by D between a set of m-features
and a (possibly complex) c-structure category and f-description. The examples
in this section are intended as simple illustrations of the range and types of D-
mappings that may be required, under various alternative assumptions about the
best way of treating a particular grammatical construction or the grammatical
consequences of a particular morphological alternation. Of course, there may be
alternative lexemic entries, alternative realization relations R, or alternative D-
mappings which could be argued to provide a more satisfying analysis of the
phenomena under discussion: the purpose of the discussion in this section is not to
advocate these particular analyses over alternatives, but to illustrate and exemplify
various kinds of D-mappings in a simple and intuitive way. A focus of future
research is to examine the mappings presented here in more detail; in fact, it may
be possible and desirable to adopt a more constrained theory of the D-mapping,
according to which some of these types of mappings are dispreferred or disallowed.
In line with the overall architecture of LFG, our approach is modular in the
sense that any particular theory of the realization relation R for a language is likely
to be compatible with several different possibilities for syntactic and semantic
analysis of the language. An example of this is presented in §12.2.2, where two
alternative syntactic analyses of the English “affix hopping” pattern are considered
in the context of the same theory of morphological realization R for English.
These approaches differ from ours in appealing to a separate morphosyntactic
structure; further, one analysis also assumes complex c-structure categories. We
point out how the D mapping is defined differently in these two settings. We
also provide examples of mappings involving lexical exceptions, taking the Lexical
Index as well as m-features into account; mappings which take into account more
than one m-feature; default mappings involving privative features, in which an
f-description is introduced in the absence of an m-feature; and mappings involving
complex m-feature specifications. Our theory of the D-mapping as the interface
between the morphological component and the rest of the grammar must be
flexible enough to allow expression of alternative grammatical analyses on the basis
of the same morphological realization relation R, and also to encompass alternative
assumptions about the nature of R.
In this section, rather than representing all four components of D, we abbreviate
D
by informally indicating an association represented as ⇒ between an m-feature or
set of m-features and the corresponding c-structure category or f-description. We
provide a precise formal definition of D in §12.3.

12.2.1 Simple D-mappings

Many m-features specify information that is relevant for syntax or semantics.
Often, for example, the morphological form of a noun or pronoun transparently
morphological features and morphological classes 447

signals its f-structure case; morphologically accusative nouns contribute the m-

feature m-case:acc, and this corresponds to a specification of accusative case
at f-structure.11 This means that the description function D must establish the
correspondence given in (19) between the m-feature m-case:acc and an equation
specifying the value acc for the case attribute at f-structure:

(19) Typical D-mappings:

D
m-case:acc ⇒ (↑ concord case) = acc
D
m-num:pl ⇒ (↑ index num) = pl

12.2.2 C-structurally relevant m-features

As has been known since Chomsky’s (1957) formulation of the Affix Hopping
transformation, English auxiliaries select a particular form of their verbal com-
plements: progressive be selects the present participial form of the following verb,
perfect have selects the past participial form of the following verb, and so on.

(20) a. David is swimming.

b. David has swum.
c. David has been swimming.
d. David must be swimming.
e. ⃰David have is swum. / ⃰David has swimming.

In their analysis of auxiliary selection, Butt et al. (1996a) and Butt et al. (1996b)
propose a new projection from c-structure, morphosyntactic structure. Building
on this work, but adopting a slightly different projection architecture, Frank and
Zaenen (2002) propose an alternative analysis in which morphosyntactic structure
is projected from f-structure rather than c-structure, and which also requires
the use of complex c-structure categories. As noted in §7.1.1, we do not adopt
morphosyntactic structure as a level of representation elsewhere in this book;
we present these two analyses here simply as illustrations of how the D-mapping
works, and in particular to show how different definitions of the D-mapping from
the same m-features give rise to different lexical entries, depending on the syntactic
analysis that is adopted.
Butt et al. (1996a) and Butt et al. (1996b) argue that Affix Hopping patterns
should not be treated in terms of complementation or embedding at functional
structure. Instead, they propose to encode verbal dependencies at morphosyntactic
structure, which on their approach is related to c-structure through a projection
function μ, represented in (21) as a dashed arrow:

11 §2.5 provides more discussion of the structure and representation of the values of features such as
case and num; here we provide simplified atomic values such as acc and pl.
448 the interface to morphology

(21) be swimming according to Butt et al. (1996a) and Butt et al. (1996b):
! $
vtype inf
" # (Morphosyntactic structure)
dep vtype prs.ptcp
μ
V'

V VP
% &
ˆ⃰μ = ⃰μ (ˆ⃰μ dep) = ⃰μ pred ‘swim#subj$’
↑=↓ ↑=↓ (F-structure)
aspect prog

be V

swimming

On Butt et al.’s analysis, be swimming contributes a monoclausal f-structure: the

V and VP daughters of V' are f-structure coheads, each annotated with ↑ = ↓. The
requirement for the complement of be to appear in present participial form holds
at morphosyntactic structure.
Butt et al.’s lexical entries for be and swimming are:

(22) Lexical entries according to Butt et al. (1996a):

L<be, /bi/, V, {(ˆ⃰μ vtype) = inf, (ˆ⃰μ dep vtype) = c prs.ptcp}>
L<swimming, /swɪmɪŋ/, V, {(↑ aspect) = prog, (ˆ⃰μ vtype) = prs.ptcp}>

According to Butt et al., morphosyntactic structure is accessible via the μ projection

function from the nodes in the c-structure, as shown in (21). As usual, ⃰ refers to
the current phrase structure node, and ˆ⃰ refers to the mother of the current node
(as explained in §5.3.1); thus, in the lexical entries in (22) and in the annotations
on the tree in (21), ˆ⃰μ refers to the μ-projection of the mother node, i.e. the
morphosyntactic structure corresponding to the mother node, and ⃰μ refers to the
μ-projection of the current node. According to these lexical entries and phrase
structure annotations, then, progressive be requires its verbal dependent dep to
be a present participle at morphosyntactic structure, with the value prs.ptcp for
the vtype attribute. Swimming is specified with the value prs.ptcp at morphosyn-
tactic structure, and the (ˆ⃰μ dep) = ⃰μ constraint on the VP node ensures that the
morphosyntactic structure of swimming appears as the value for the dep feature.
This approach provides a means of handling morphosyntactic dependencies by
a mechanism similar to complementation at f-structure, but without assuming
embedding at f-structure.
In this setting, we assume the m-entry for swimming in (23), and consider
the D-mapping which produces the lexical entry for swimming in (22) from this
m-entry:
morphological features and morphological classes 449

(23) M-entry for swimming:

R <swim1, swimming, /swɪmɪŋ/, {m-cat:verb, m-vtype:prs.ptcp}>

For clarity, we have provided the same value prs.ptcp for the m-feature m-vtype
in (23) and the vtype feature at morphosyntactic structure in (21). Impor-
tantly, these features and their values have a very different status: the m-feature
m-vtype:prs.ptcp in (23) is visible and relevant only within the morphological
component, and plays a crucial role in defining the lexical entry for the word form
swimming, while the vtype feature in (21) is relevant in defining and constraining
syntactic dependencies involving verbs.
Given the m-entry in (23), Butt et al.’s analysis requires the morphosyntactic
m-vtype feature that is contributed by swimming to correspond to a vtype
feature at morphosyntactic structure via the μ mapping, and an aspect feature at
f-structure; this allows the required dependencies to be established:

(24) D mapping for f-description according to Butt et al. (1996a,b):

D
m-vtype:prs.ptcp ⇒ {(ˆ⃰μ vtype) = prs.ptcp, (↑ aspect) = prog}

Butt et al.’s analysis also requires the m-cat feature for verbal forms to map to the
phrase structure category V:

(25) D mapping for phrase structure category according to Butt et al. (1996a,b):
D
m-cat:verb ⇒ V

In this way, Butt et al.’s analysis enforces “affix hopping” requirements at mor-
phosyntactic structure while maintaining a simple, monoclausal f-structure with
no embedding.
It is possible for the same theory of morphological realization, and in particular
the same m-entry in (23), to underpin an analysis of the English auxiliary system
that makes very different syntactic assumptions. Frank and Zaenen (2002) provide
an alternative analysis of auxiliary selection, illustrated primarily for French, which
assumes a different relation between morphosyntactic structure and f-structure,
and which also appeals to complex c-structure categories; see § 3.3.5 and §5.1.4 for
definition and discussion of complex categories, including parametrized complex
categories. Frank and Zaenen’s proposal differs from Butt et al.’s in the inventory of
features that appear in morphosyntactic structure: according to Frank and Zaenen,
morphosyntactic structure contains a number of additional features not treated
by Butt et al., including person, number, gender, and case. Constraints on these
features are generally stated in terms of f-structure relations rather than c-structure
configuration: verbs show agreement in person, number, and/or gender with their
subjects or objects, for example, and case is often an indication of f-structure
role. According to Frank and Zaenen (2002), then, morphosyntactic structure is
450 the interface to morphology

best treated as directly related to f-structure rather than c-structure, since this
allows easy reference to f-structural relations in characterizing morphosyntactic
dependencies such as agreement and case assignment. For a discussion of the
grammatical architecture that Frank and Zaenen assume, see § 7.1.1.
Frank and Zaenen (2002) redefine μ as a function from f-structures (rather than
nodes of the c-structure) to morphosyntactic structures:

(26) The μ projection function:

Butt et al. (1996a), Butt et al. (1996b): ˆ⃰μ (= μ(ˆ⃰))
Frank and Zaenen (2002): ↑μ (= μ(𝜙(ˆ⃰)))

Adapting Frank and Zaenen’s analysis of French auxiliaries to conform more

closely to Butt et al.’s assumptions about the structure of the English verb phrase,
we obtain this configuration:

(27) be swimming, following Frank and Zaenen (2002):

% &
pred ‘swim#subj$’
𝜙 (F-structure)
aspect prog
V' [fin]

V[fin,be] VP[nonfin,prs.ptcp]
μ
be V[prs.ptcp]

swimming
(Morphosyntactic
! structure) $
vtype inf
" #
dep vtype prs.ptcp

Since Frank and Zaenen assume that morphosyntactic structure is directly related
to f-structure and only indirectly related to c-structure, their approach cannot
take advantage of c-structural embedding relations to control auxiliary sequencing.
Instead, Frank and Zaenen propose a combination of constraints involving mor-
phosyntactic features and complex c-structure categories to constrain the position
and morphological form of verbs in the verbal complex.
As discussed in §3.3.5 and §5.1.4, complex categories allow certain features to
be realized as a component of a complex c-structure category, and parametrized
rules allow for these features to be passed through the tree. For example, the
parametrized phrase structure rule in (28) states that a V' with a finiteness value
represented as _ftness dominates a V with form ‘be’ and the same finiteness value
_ftness as the V' , followed by a VP headed by a nonfinite present participial
verb. This rule ensures that progressive be is followed by a present participial
complement, as in be swimming:
morphological features and morphological classes 451

(28) V' phrase structure rule according to Frank and Zaenen (2002):
V' [_ftness] −→ V[_ftness,be] VP[nonfin,prs.ptcp]

On Frank and Zaenen’s approach, then, be and swimming have the lexical entries
in (29), containing specifications of vtype for be and dep vtype for swimming at
morphosyntactic structure, as well as complex c-structure categories:

(29) Lexical entries according to Frank and Zaenen (2002):

L<be, /bi/, V[fin,be], {(↑μ vtype) = inf, (↑μ dep vtype) = c prs.ptcp}>
L<swimming, /swɪmɪŋ/, V[prs.ptcp], {((↑ aspect) = prog,
(↑μ dep vtype) = prs.ptcp}>

To derive these lexical entries, we again assume the m-entry for swimming given in
(23). However, Frank and Zaenen’s approach requires a different D-mapping from
m-features to f-descriptions. On their approach, the description function D must
establish the following correspondence between the set of m-features for swimming
and a complex c-structure category:

(30) D-mapping to complex c-structure category according to Frank and Zaenen

(2002):
D
{m-cat:verb, m-vtype:prs.ptcp} ⇒ V[prs.ptcp]

As this example illustrates, more than one m-feature may be involved in a sin-
gle mapping relation: here, the m-features {m-cat:verb, m-vtype:prs.ptcp} are
involved in a single D-mapping specification to the complex category V[prs.ptcp].
Frank and Zaenen’s analysis also involves constraints on morphosyntactic struc-
ture and f-structure. This means that the same m-vtype feature as in (30) maps to
the following f-description:

(31) D-mapping to f-description according to Frank and Zaenen (2002):

D
m-vtype:prs.ptcp ⇒ {(↑ aspect) = prog, (↑μ dep vtype) = prs.ptcp}

Note, then, that the same m-features can be relevant in two different aspects of the
D-mapping: here, the same m-feature, m-vtype, is involved in determining both
a complex c-structure category (30) and an f-description (31).

12.2.3 Mixed categories

Spencer (2013, pages 122–3) discusses the Russian noun stolovaja ‘dining room’,
which historically derives from an adjective, and is synchronically a noun which
inflects as an adjective (see also Baerman et al. 2005). This can be seen by
452 the interface to morphology

inspecting the paradigms for the noun stolovaja ‘dining room’, the adjective bol' šoj
‘big’, and the noun lampa ‘lamp’:

(32) Noun Adjective Noun

‘dining room’ ‘big’ ‘lamp’
nom stolov-aja bol' š-aja lamp-a
acc stolov-uju bol' š-uju lamp-u
gen stolov-oj bol' š-oj lamp-i
dat stolov-oj bol' š-oj lamp-e
ins stolov-oj bol' š-oj lamp-oj
prepositional stolov-oj bol' š-oj lamp-e

As Spencer shows, stolovaja belongs to a class of words that are syntactically

nouns, but morphologically adjectives; he refers to such forms as “mixed lexical
categories.”12 We assume the m-entries in (33) for the nominative forms. For bol' š-
aja (with m-cat:adj) and lamp-a (with m-cat:noun), the m-class feature has
value regular; in contrast, stolovaja patterns morphologically as an adjective,
with the m-feature m-cat:adj, but is marked as a mixed category with the syntactic
behavior of a noun by the feature m-class:mixed-a-n:

(33) M-entries for Russian nominative word forms:

R <stolov1, stolovaja, /stɐlovəjə/,
{m-cat:adj, m-class:mixed-a-n, m-case:nom}>
R <bolš1, bol' šaja, /bɐlj ʂajə/,
{m-cat:adj, m-class:regular, m-case:nom}>
R <lamp1, lampa, /lampə/,
{m-cat:noun, m-class:regular, m-case:nom}>

Given these m-entries, the D-mapping rule in (34) produces the correct category
for these verb forms; notably, for the mixed category stolovaja with m-feature
m-class:mixed-a-n, the m-cat:adj m-feature maps to the c-structure category
N:13

(34) a. C-structure category N for a Russian mixed category inflecting as an

adjective (for example, stolovaja):
D
m-cat:adj ⇒ N in the presence of the m-feature m-class:mixed-a-n.

12 The term “mixed category” is also used in a different sense, to refer to categories such as the English
gerund, which exhibit both nominal and verbal syntactic characteristics simultaneously (Bresnan 1997;
Bresnan and Mugane 2006; Seiss 2008; Lowe 2017a). “Mixed lexical categories” in Spencer’s sense are
categories which have morphological characteristics of one category, but syntactic characteristics of
another category.
13 If we assume a morphological theory which makes use of privative features and defaults, the
mapping for the regular nouns bol' šaja and lampa could be treated as a default mapping (§12.2.6).
morphological features and morphological classes 453

b. C-structure category A for a regular category inflecting as an adjective

(for example, bol' šaja):
D
m-cat:adj ⇒ A in the presence of the m-feature m-class:regular.
c. C-structure category N for a regular category inflecting as a noun (for
example, lampa):
D
m-cat:noun ⇒ N in the presence of the m-feature m-class:regular.

Lowe (2017a) proposes a parallel D-mapping rule which provides an account of

participles, a category of word which is syntactically verbal but morphologically
adjectival.

12.2.4 Deponency
Börjars et al. (1997) and Sadler and Spencer (2001) discuss the Latin verb loquor
‘I speak’, which has the morphological form of a passive verb but is syntactically
active, as can be seen by comparing the forms of loquor, with active voice, to the
active and passive forms of lego ‘I read’. This verb illustrates deponency (Baerman
2007), a mismatch in which a word has a form implying a particular category (here,
passive voice) but realizing a different category (active voice):

(35) deponent non-deponent non-deponent

active active passive
present loquitur legit legitur
‘he speaks’ ‘he reads’ ‘it is read’
imperfect loquebatur legebat legebatur
‘he spoke’ ‘he read’ ‘it was read’
future loquetur leget legetur
‘he will speak’ ‘he will read’ ‘it will be read’

We assume that the realization relation R contributes an m-feature m-voice encod-

ing morphological voice. We must, then, define the D-mapping in such a way that
for deponent verbs (with m-feature m-class:deponent-voice), the m-feature
for passive morphological form maps to an f-description specifying active voice.
For non-deponent verbs (with m-feature m-class:regular), morphological voice
matches syntactic voice. According to this analysis, the third-person singular
present tense forms of these verbs have the following m-entries:

(36) M-entries for Latin verb forms:

R <loq1, loquitur, /lokw ɪtur/,
{m-cat:verb, m-class:deponent-voice, m-voice:pass, m-pers:3,
m-num:sg, m-tense:pres}>
R <leg1, legitur, /legitur/,
{m-cat:verb, m-class:regular, m-voice:pass, m-pers:3, m-num:sg,
m-tense:pres}>
454 the interface to morphology

The D-mapping rules given in (37) produce the correct voice specification for these
word forms: active voice for the deponent verb loquitur, and passive voice for the
regular verb legitur:

(37) a. Passive form, active voice for deponent verbs (for example, loquitur):
D
m-voice:pass ⇒ (↑ voice) = active in the presence of the m-feature
m-class:deponent-voice.
b. Passive form, passive voice for regular verbs (for example, legitur):
D
m-voice:pass ⇒ (↑ voice) = pass in the presence of the m-feature
m-class:regular.

12.2.5 Lexical exceptions: f-descriptionM dependent on Lexemic Index

We include the Lexemic Index as a component of the description function D in
order to allow for the possibility that the interpretation of an m-feature or a set
of m-features varies idiosyncratically for different lexemes, and that this is best
analyzed by encoding a dependence in the D-mapping on the Lexemic Index.
For example, Acquaviva (2008: 19) proposes that the word form measles is
exceptional in being morphologically plural (involving suffixation of plural -s
to a base, and so carrying the m-feature m-num:pl) but syntactically singular,
and so bearing the f-description (↑ index num) = sg. That measles is syntactically
singular is shown by its requirement for singular verb agreement:

(38) Measles is/ ⃰are a terrible disease.

On this view, measles is a deponent noun, morphologically plural but syntactically

singular. Support for the view that measles is morphologically plural, consisting
of the root measle followed by plural -s, is provided by attested examples of the
uninflected form measle as the first member of a noun-noun compound:

(39) a. New needle-free measle vaccine ‘could save thousands of children’s lives’
(headline in The Telegraph, 17 Aug 2009)
b. . . . it is reasonable because though we have never found a measle
germ associated with measle-symptoms we have in cases with like
symptoms found, not indeed measle germs, but things of the same sort . . .
(Wisdom 1968)

One way to capture Acquaviva’s analysis in our setting is by assuming that the m-
feature m-num:pl maps to the syntactic feature (↑ index num) = pl in most cases,
but with particular lexically specified exceptions involving reference to particular
lexemic entries: if the LI is measle1 or a handful of other syntactically singular
nouns with plural morphology, m-num:pl maps to (↑ index num) = sg. Such an
analysis requires the following m-entries and D-mapping rule:
morphological features and morphological classes 455

(40) M-entries for English noun forms:

R <measle1, measles, /mi:zlz/, {m-cat:noun, m-num:pl}>
ˈ
R <dog1, dogs, /dɔgz/, {m-cat:noun, m-num:pl}>

(41) Measles as a lexical exception:

D
m-num:pl ⇒ (↑ index num) = sg if LI = measle1,
D
otherwise m-num:pl ⇒ (↑ index num) = pl

This analytic possibility may not arise, depending on the particular theory of
morphological realization and the morphology-syntax-semantics interface that
is adopted. There are two alternative analytic possibilities that do not involve
reference to the LI. First, it may be preferable to analyze measles as a root that
belongs to a particular (very small) inflectional class, consisting of deponent nouns
that are morphologically plural but syntactically singular. Under this analysis, the
m-feature m-class for this class, rather than the LI for the lexemic entry, would
be important in the D-mapping for measles. Second, if the realization relation R
specifies measles as having m-num:sg even though it bears plural morphology, the
definition of D need not take the LI into account. If alternative accounts which do
not involve reference to the LI are shown to be available and preferable for all word
forms, we need not include the LI as a component of the description function D.
On the other hand, if the interpretation of a set of m-features is best analyzed as
varying according to the LI, we must take the LI into account in our definition of
D. Future work will show which hypothesis is viable, and whether the LI must be
taken into account in at least some cases of the D-mapping.

12.2.6 Feature defaults

Some morphological theories use privative features: that is, the absence of a feature
is interpreted as indicating the presence of some grammatical property. To allow for
this possibility, the D-mapping must be formulated so as to permit the introduction
of a particular f-description if no m-feature of a certain type is present.14 For
example, we might assume that the realization relation R provides a privative m-
feature m-num:pl for plural nouns, but no m-feature for singular nouns: nouns
are assumed to be syntactically singular if the plural m-feature does not appear. In
other words, according to this view plural nouns such as dogs are associated with
the m-feature m-num:pl, and singular nouns such as dog do not have a m-num
feature. This analysis assumes m-entries like the following:

14 Note that we neither advocate nor reject the use of privative features or defaults in defining the
D-mapping relation; whether or not such rules are needed depends on the theory of the morphological
realization relation R that is adopted and the syntactic rules and constraints for the language. Privative
features are not used in Paradigm Function Morphology (Stump 2001), and so feature default rules of
this type are not relevant in a PFM setting.
456 the interface to morphology

(42) M-entries for English noun forms:

R <dog1, dog, /dɔg/, {m-cat:noun}>
R <dog1, dogs, /dɔgz/, {m-cat:noun, m-num:pl}>

Under these assumptions, we require a default rule which applies when the m-num
feature is absent:

(43) Singular number as a morphological default for nouns:

Introduce the f-description {(↑ index num) = sg} if there is an m-feature
m-cat:noun but no m-num m-feature in the m-description.

In §12.3.2.3, we will see how to express default rules within the formal framework
that we adopt.

12.2.7 Complex m-feature specifications

According to some theories of the morphological realization relation R, m-features
can have a complex structure involving reference not to the f-structure denoted by
↑, but to a different f-structure related in a specified way to ↑. Our definition of
the D mapping must, then, allow for the proper treatment of such complex m-
feature specifications. For example, Stump (2001) proposes m-features such as the
following:

(44) Example: Complex m-feature, Stump (2001)

agr(su):{pers:1, num:pl}

Here, a set of m-features is relevant for subject agreement for inflected verbs,
encoding first person plural subject agreement. Thus, the D mapping must map
a complex m-feature of this form to the following f-description:

(45) D-mapping for a complex m-feature:

D
agr(su):{pers:1, num:pl} ⇒ {(↑ subj index pers) = 1,
(↑ subj index num) = pl}

12.3 The description function D

We have seen that the set of lexical entries L is determined for a language on the
basis of a set of lexemic entries LE, a set of m-entries defined by the morphological
realization relation R, and the description function D mapping from m-features
to a c-structure category and f-description. The realization relation R is provided
by a separately specified theory of morphological realization. Lexemic entries LE
 the description function d 457

include f-descriptions which are structured according to the theory of syntax,

semantics, information structure, and prosody that we discuss elsewhere in this
book. The description function D constitutes the interface between the morpho-
logical component and the rest of the grammar, since it defines the relation between
the m-features for a word form, which are invisible outside the morphological
module, and the phrase structure category and f-description that appear in its
lexical entry.

12.3.1 Previous definitions of D

Previous approaches have assumed that the f-description corresponding to a set
of m-features can be constructed by examining one m-feature at a time, mapping
each m-feature to a partial f-description independent of the presence or absence
of other m-features. This is, for example, how Kaplan and Butt (2002) define D,
and how Andrews (2005) defines his version of D for features not involved in case
stacking (§12.3.3.6). Simple definitions such as these do not allow for dependencies
among m-features, nor for defaults: they do not allow the introduction of an f-
description in the absence of an m-feature, for example.
Kaplan and Butt (2002) analyze the ambiguous German noun Kätzchen ‘kitten/
kittens’ by means of two different m-entries, one associated with singular number
(with syncretic nominative/dative/accusative case) and the other associated with
plural number (with syncretic nominative/genitive/dative/accusative case). Trans-
lating their analysis into the current setting, the s-form and p-form for the two
m-entries are the same, but the m-features differ in terms of the morphological
specification of number (m-num) and case (m-case):

(46) Kätzchen (Kaplan and Butt 2002):

R <Kätzchen1, Kätzchen, /kɛtsçən/,
{m-root:Katze, m-cat:noun, m-gend:neut, m-num:sg, m-case:nom/
dat/acc}>
R <Kätzchen1, Kätzchen, /kɛtsçən/,
{m-root:Katze, m-cat:noun, m-gend:neut, m-num:pl, m-case:nom/gen/
dat/acc}>

Kaplan and Butt propose that the description function D maps from individual
m-features to component parts of the f-description for a word form:

(47) Kaplan and Butt’s (2002) description function D for each m-feature:
D(m-root:Katze) = {(↑ pred) = ‘Katze’}
D(m-cat:noun) = {N, (↑ ntype) = count}
D(m-gend:neut) = {(↑ concord gend) = n}
D(m-num:sg) = {(↑ index num) = sg}
D(m-num:pl) = {(↑ index num) = pl}
D(m-case:nom/dat/acc) = {(↑ concord case) ∈ {nom,dat,acc}}
D(m-case:nom/gen/dat/acc) = {(↑ concord case) ∈ {nom,gen,dat,acc}}
458 the interface to morphology

They then define the description function D for the set of m-features as the union
(∪) of the result of applying D to each feature independently:

(48) Kaplan and Butt’s (2002) description function D for a set of m-features:
D({d1 , d2 , . . . , dn }) = D(d1 ) ∪ D(d2 ) ∪ …D(dn )

Andrews (2005) proposes a similar definition for his version of D, which he

calls F : each m-feature is placed in correspondence to a particular f-description,
and the full f-description for a word form is obtained by collecting together the
f-descriptions for each m-feature.15
This simple approach is adequate for many cases, but is not adequate for all of
the analytical possibilities that may arise. We have seen the need to consider certain
m-features in the context of other m-features, to specify an f-description in the
absence of an m-feature, and to provide an analysis of complex m-features. We
have also seen that some analyses may require reference to the Lexemic Index in
the definition of D.

12.3.2 Definition of D
Recall that the D-mapping is a relation involving the following terms:

(49) General form of the description function D:

D <Lexemic Index, m-features, category, f-descriptionM>

The D-mapping for the plural noun dogs given in (15) is repeated in (50). It relates
the m-features {m-cat:noun, m-num:pl} to the c-structure category N and the
f-description {(↑ index num) = pl}:

(50) D <dog1,
{m-cat:noun, m-num:pl},
N,
{(↑ index num) = pl}>

We propose the following definition of D:

(51) D <LI, m-features, category, f-descriptionM > if and only if

Dcat <LI, m-features, category> and
Dfeats <LI, m-features, ↑, f-descriptionM >.

The c-structure category for a word form is specified by the Dcat mapping on
the basis of the L(exemic) I(ndex) and the m-features, and the f-descriptionM is
determined by the mapping defined by Dfeats . These mappings are defined on a
language-by-language basis, though there is likely to be a great deal of commonality

15 Andrews (2005) also extends his analysis to case stacking, discussed in §12.3.3.6.
 the description function d 459

in their definitions across languages; this is an important topic of research on the

interface between morphology and other components of the grammar.
A complicating factor in the definition of Dfeats is that in some cases, in
particular for complex m-features as described in §12.2.7 and for case stacking
as described in §12.3.3.6, the D-mapping might require reference not to the f-
structure ↑, but to other f-structures related to ↑. This means that we must
explicitly specify the f-structure that is relevant for the definition of Dfeats in the
definition of D. For this reason, in the definition in (51), ↑ is specified as the third
argument of Dfeats .

12.3.2.1 Dcat : defining the c-structure category If we assume that English does
not make use of complex c-structure categories, and furthermore that there are
no “mixed” categories in English with differing morphological and syntactic
categories, the definition of Dcat for English is:

(52) Dcat for English:

Dcat <LI, m-features, N> if and only if m-cat:noun ∈ m-features.
Dcat <LI, m-features, V> if and only if m-cat:verb ∈ m-features.
Dcat <LI, m-features, A> if and only if m-cat:adj ∈ m-features.
..
.

An additional m-feature may be necessary to specify the category as projecting or

nonprojecting (§3.3.4).
For analyses assuming complex or mixed categories, additional m-features must
be considered, as discussed in §12.2.2. We can restate the informal complex
category rule in (30) in the current format:16

(53) Dcat <LI, m-features, V[prs.ptcp]>

if and only if {m-cat:verb, m-vtype:prs.ptcp} ⊆ m-features.

An informal rule was given in (34a) for a complex category which is morphologi-
cally an adjective but syntactically a noun. That rule can be restated in the current
format as in (54):

(54) Dcat <LI, m-features, N>

if and only if
{m-cat:adj, m-class:mixed-a-n} ⊆ m-features, or
{m-cat:noun, m-class:regular} ⊆ m-features.

According to the definition of D in (51), Dcat is required to apply in order to

determine the c-structure category of a word form. It does not do any “feature
accounting,” however; whether the definition of Dcat appeals to one m-feature or

16 The subset relation ⊆ is defined and discussed in §6.3.

460 the interface to morphology

more than one, all of the m-features are passed on to Dfeats for determination of
the f-description of the word form.

12.3.2.2 Dfeats : f-descriptions corresponding to m-features The language-

specific Dfeats function determines the f-description of the f-structure f , obtained
by considering each member of a set of m-features:

(55) Mapping m-features to f-descriptions, schematic definition:

Dfeats <LI, {m1 , m2 , . . . mn }, f , d0 ∪ d1 ∪ d2 ∪ . . . dn > if and only if
Ddefault <LI, {m1 , m2 , . . . mn }, f , d0 > and
Dfeat <LI, m1 , {m1 , m2 , . . . mn }, f , d1 > and
Dfeat <LI, m2 , {m1 , m2 , . . . mn }, f , d2 > and
..
.
Dfeat <LI, mn , {m1 , m2 , . . . mn }, f , dn >.

In this definition, Dfeats (with four arguments) relates a L(exemic) I(ndex), a set of
m-features {m1 , m2 , . . . mn }, an f-structure f , and an f-description d1 ∪ d2 ∪ . . . dn
corresponding to those m-features. Dfeats is defined in terms of two other func-
tions: Ddefault defines a set of default features d0 which appear when certain m-
features are absent, and the Dfeat functions map individual m-features m1 . . . mn to
f-descriptions d1 . . . dn . That is, Dfeat (with five arguments) specifies the mapping
relation for a particular m-feature (for example, m1 ) in the context of the LI, the
complete set of m-features {m1 , m2 , . . . mn }, and a target f-structure f , mapping
m1 to the f-description d1 . According to this definition, then, each m-feature
is considered in the context of all of the other m-features, and the appropriate
f-description of the f-structure f is added to the full f-description. This requires
a definition of Ddefault for the language and of Dfeat for each m-feature, as we now
show.
The definition of Dfeats is formulated so as to allow the D-mapping to specify
constraints on any relevant f-structure. Usually, the relevant f-structure is ↑, as
specified in (51); we discuss the D-mapping for complex m-features and case
stacking in §§12.3.3.5 and 12.3.3.6, where the f-description may refer not to ↑,
but to another f-structure related to ↑.

12.3.2.3 Ddefault : privative m-feature and defaults As noted in §12.2.6 above,

some but not all morphological theories assume the existence of privative features
and feature defaults.17 In a theoretical setting in which it is relevant, the Ddefault
function introduces an f-description in the absence of a particular m-feature or m-
features. If we assume that the D-mapping function for a language includes some
number n of default rules D1default . . . Dndefault , the definition of Ddefault is the

17 Privative features are not assumed in Paradigm Function Morphology (Stump 2001). When
such theories of morphological realization are adopted, the Ddefault function is not relevant, and the
definition of Ddefault which we provide in this section can be ignored.
 the description function d 461

following, where the default f-description (d0 in the definition in (55)) results from
applying each of the n rules in turn:

(56) Default mappings Ddefault :

Ddefault <LI, m-features, f , d1 ∪ d2 ∪ . . . dn > if and only if
D1default <LI, m-features, f , d1 > and
D2default <LI, m-features, f , d2 > and
..
.
Dndefault <LI, m-features, f , dn >.

Notice that we have specified f as the relevant f-structure in the third argument
position of the definition of Ddefault and, in turn, of the individual default rules
D1default . . . Dndefault , to allow the D-mapping to impose constraints on the rele-
vant f-structure, whether it is ↑ or another f-structure related to ↑. In most cases,
the relevant f-structure is ↑, as specified in the definition in (51), but for complex
m-features or case stacking another f-structure might be relevant, as we show in
§§12.3.3.5 and 12.3.3.6.
Schematically, individual Ddefault rules are stated as in (57); in this definition,
the absence of the m-feature m1 licenses the introduction of the f-description d1
as a constraint on the f-structure f . In other words, if the privative feature m1 is
absent, d1 appears in the default f-description of f ; if m1 is present, d1 does not
appear, and the empty f-description ∅ is introduced:18

(57) Schematic default rule:

D1default <LI, m-features, f , d1 >
if and only if m1 -∈ m-features (and possibly other conditions as well),
otherwise D1default <LI, m-features, f , ∅>.
Recall the example given in §12.2.6, in which we explore the possibility that
plural nouns have a privative m-feature m-num:pl, while singular nouns have no
m-num feature. Given these assumptions, the default rule introducing the con-
straint (f index num) = sg in the absence of a plural m-feature is:

(58) Example: Default mapping to a singular f-description for nouns without a

plural m-feature
D1default <LI, m-features, f , {(f index num) = sg}>
if and only if m-cat:noun ∈ m-features and m-num:pl -∈ m-features,
otherwise D1default <LI, m-features, f , ∅>.

As with the Dcat rules, the Ddefault rules are not involved in “feature accounting”:
once the default rules have applied to a set of m-features, that set is passed
unchanged to the Dfeats rule.

18 The negated set membership relation - ∈ is defined and discussed in §6.3.

462 the interface to morphology

12.3.3 Dfeat : examples

12.3.3.1 Simple D-mappings from m-features to f-descriptions For a language
with a very straightforward mapping between morphological and syntactic case,
we have D-mappings such as the following, producing an f-description which the
f-structure f must satisfy:19

(59) Example: Simple mapping from m-case to syntactic case

Dfeat <LI, m-case:nom, m-features, f , {(f concord case) = nom}>.
Dfeat <LI, m-case:acc, m-features, f , {(f concord case) = acc}>.
Dfeat <LI, m-case:dat, m-features, f , {(f concord case) = dat}>.
Dfeat <LI, m-case:gen, m-features, f , {(f concord case) = gen}>.

This rule does not require or disallow the presence of other m-features in defin-
ing the mapping between m-case and syntactic case; m-case:nom maps to the
f-description {(f concord case) = nom} for the f-structure f independent of
the presence or absence of other m-features, and similarly for the other case
possibilities.
If the mapping between morphological and syntactic case is completely straight-
forward, we can make use of a convenient abbreviation capturing the effects of each
of the four separate D-rules in (59), borrowing the underscore notation for the
argument of a parametrized template (§6.7) to indicate that morphological case
always matches syntactic case:

(60) Example: General mapping from any m-case to the corresponding f-

structure case specification, abbreviating the set of rules in (59)
Dfeat <LI, m-case:_case, m-features, f , {(f concord case) = _case}>.

12.3.3.2 Context-sensitive D-mappings The Latin deponent verb loquor ‘I

speak’ is morphologically passive but syntactically active. According to the analysis
of this pattern presented in §12.2.4, deponent verbs are associated with the m-
feature m-class:deponent-voice, and in the presence of this feature, verbs
with the m-feature m-voice:pass are associated with the f-structure constraint
(↑ voice) = active. Non-deponent verbs have the m-feature m-class:regular,
with a straightforward relation between morphological and syntactic voice. This
pattern can be stated by means of the following set of Dfeat rules, according to
which a morphologically passive form is syntactically active if it has the features
m-voice:pass and m-class:deponent-voice; if the form is morphologically
regular, with the feature m-class:regular, morphological voice matches syntactic
voice. The second rule uses the underscore notation introduced in §12.3.3.1,
requiring matching values for the two features:

19 As we have done so far, we use simple values like nom and acc as the value of the case attribute.
Section 2.5.7.5 provides a full discussion of the representation of the values of the case feature.
 the description function d 463

(61) Example: D-mapping dependent on the m-class m-feature

Dfeat <LI, m-voice:pass, m-features, f , {(f voice) = active}>
if and only if m-class:deponent-voice ∈ m-features.
Dfeat <LI, m-voice:_voice, m-features, f , {(f voice) = _voice}>
if and only if m-class:regular ∈ m-features.

12.3.3.3 Vacuous D-mappings Depending on the particular theory of the real-

ization relation R and the mapping D from m-features to f-descriptions, certain m-
features may be important in providing context for other D-mappings, but do not
themselves map to an f-description. For these features, we must specify a mapping
to the empty f-description:

(62) D-mapping to the empty f-description for an inert m-feature m1

Dfeat <LI, m1 , m-features, f , ∅>.

For example, there may be no mapping to an f-description for the m-iclass feature
representing inflectional class, though that feature may be important in providing
the proper context for other D-mapping rules, as outlined in §12.2.4 and §12.3.3.2:

(63) Example: D-mapping to the empty f-description for the m-iclass:1 and
m-iclass:2 m-features
Dfeat <LI, m-iclass:1, m-features, f , ∅>.
Dfeat <LI, m-iclass:2, m-features, f , ∅>.

To take another example, there may be no mapping to an f-description for the

m-cat feature, though that feature is important in determining the c-structure
category through the mapping Dcat . For cases like this, where all values for an
m-feature map to the same f-description (here, the empty f-description ∅), we
can make use of an abbreviatory convention according to which mentioning
an m-feature without specifying its value is interpreted as specifying the same
D-mapping for all possible values for the feature:

(64) Example: D-mapping to the empty f-description for the m-cat m-feature
with any value, using an abbreviatory convention mentioning only the
m-cat feature
Dfeat <LI, m-cat, m-features, f , ∅>.

The D-mapping rule in (64) maps the m-cat attribute with any value to the empty
f-description ∅.

12.3.3.4 D-mapping dependent on the LI In §12.2.5, we saw that according to

some analyses, lexical exceptions are analyzed by specifying a D-mapping which
refers to the Lexemic Index. A schematic mapping of this type is:
464 the interface to morphology

(65) Schematic mapping for a lexical exception dependent on the Lexemic Index:
Dfeat <LI, m1 , m-features, f , d1 >
if and only if LI = l1 ,
otherwise Dfeat <LI, m1 , m-features, f , d2 >.

For example, if we assume that the noun measles is specified as morphologically

plural (with m-num:pl) but syntactically singular (with the value sg for the num
feature at f-structure), it may be necessary for the D-mapping to take the Lexemic
Index into account in the D-mapping for number. The rule in (66) states that the
m-feature m-num:pl maps to a plural f-description in all cases except when the
Lexemic Index is measle1:

(66) Example: Mapping to syntactically singular f-description for the morpho-

logically plural noun measles, and to plural f-description for all other nouns
Dfeat <LI, m-num:pl, m-features, f , {(f index num) = sg}>
if and only if LI = measle1,
otherwise Dfeat <LI, m-num:pl, m-features, f , {(f index num) = pl}>.

12.3.3.5 D-mapping for complex m-feature specifications As outlined in

§12.2.7, some theories of the morphological realization relation R assume
m-features with a complex structure, requiring an f-structure related to ↑ to be
constrained. For example, Stump (2001) proposes the complex subject agreement
m-feature in (67), encoding first person plural subject agreement:

(67) Example: Complex m-feature, Stump (2001)

agr(su):{pers:1, num:pl}

Our goal is to map the m-feature in (67) to this f-description:

(68) Example: F-description corresponding to the m-feature in (67)

(↑ subj index pers) = 1
(↑ subj index num) = pl

To align more closely with the m-feature representations we have been assuming,
we reformulate the complex m-feature in (67) in the following equivalent way:

(69) Variant representation of complex m-feature in (67):

m-agr:<su, {m-pers:1, m-num:pl}>

In this reformulation, the value of the m-agr m-feature is a pair in which the first
argument su is a morphological marker providing information about the syntactic
role of the relevant f-structure for the complex feature specification, and the second
 the description function d 465

argument is a set of m-features. To interpret this complex constraint, we require a

rule which maps the morphological marker su to the grammatical function subj;
given that mapping, we can identify the f-structure (↑ subj) which is the relevant
f-structure for the set of constraints in the second argument.
We assume that the correspondence between the morphological marker su and
the grammatical function subj is also specified as a part of the D-mapping, and
thus that we have D-mapping rules of the following form:

(70) Example: Mapping between the morphological marker su and the gram-
matical function subj
Dfeat <LI, su, m-features, subj, ∅>

The rule in (70) states a correspondence between su and the grammatical function
subj in the presence of a set of m-features, but does not contribute an f-description.
We can make use of the mapping in (70) in the following D-mapping rule:

(71) Example: Mapping for a complex m-feature

Dfeat <LI, m-agr:<su, {m-pers:1, m-num:pl}>, m-features, f , d>
if Dfeat <LI, su, m-features, subj, ∅> and
and Dfeats <LI, {m-pers:1, m-num:pl}, (f subj), d>.

In (71), the relevant f-structure is f , and the value of the complex m-arg m-
feature is a pair whose first argument is the morphological marker su. The relation
between su and the grammatical function subj is specified in the D-mapping rule
defined in (70). The second argument is a set of m-features, {m-pers:1, m-num:pl},
which map to the f-description constraining the relevant f-structure. The rule in
(71) states that the set of m-features specified as the second member of the pair
should be interpreted as constraining the Dfeats mapping with respect to the subj
attribute in the f-structure f . That is, the relevant f-structure is (f subj), and the
Dfeats mapping requires the f-description to hold of (f subj).
This rule is a specific instance of the following general pattern:

(72) Mapping for complex m-features:

Dfeat <LI, m-feat:<attr, m-features>, M, f , d>
if Dfeat <LI, attr, m-features, _gf, ∅>
and Dfeats <LI, m-features, (f _gf), d>

According to this rule, the value of a complex m-feature is a pair in which the first
argument specifies a morphological marker attr and the second argument is a set
of m-features. The Dfeat mapping for a complex feature with respect to f-structure f
is defined as the Dfeats mapping for the set of m-features, where the morphological
marker attr maps to a grammatical function _gf, and the relevant f-structure for
the constraints specified by the set of m-features is (f _gf).
466 the interface to morphology

12.3.3.6 Case stacking and constructive case In §6.1.3, we discussed

Nordlinger’s (1998) theory of constructive case, according to which the case of
an argument determines its syntactic role. Nordlinger (1998) also extends her
theory to case stacking, in which an argument can bear more than one case feature
determining the environment in which it must appear (see also Plank 1995). An
example is the Martuthunira word form in (73) with three case endings, originally
discussed by Dench (1995):

(73) thara-ngka-marta-a
pouch-loc-prop-acc

The locative casemarking closest to the root thara ‘pouch’ marks the word as
the head of a locative phrase (‘in (its) pouch’). The proprietive and accusative
casemarking further constrains the syntactic environment within which the phrase
can appear: thara-ngka ‘in its pouch’ must modify a proprietive phrase (‘with X
in its pouch’) which itself modifies an accusatively-marked phrase. According to
Sadler and Nordlinger (2004), the f-structure for the form thara-ngka-marta-a
is the one labeled p in (74), and it must appear in the f-structure environment
shown:20

(74) Functional structure p for thara-ngka-marta-a and the environment in

which it is required to appear (Sadler and Nordlinger 2004):
  * + 
case acc

  concord 
      
  * + 
  
 concord case prop 
 
 obj  
  
 
 


      
 
  adj  
  pred ‘pouch’   
  adj * +   
  
  p
case loc  
 
  

  concord  


Given these constraints, thara-ngka-marta-a can felicitously appear in a clause

such as:

(75) Ngayu nhawu-lha ngurnu tharta-a mirtily-marta-a

I saw-pst that.acc euro-acc joey-prop-acc
thara-ngka-marta-a.
pouch-loc-prop-acc
‘I saw the euro with a joey in (its) pouch.’21 (Dench 1995)

20 The representation in (74) differs from the proposals of Nordlinger (1998) and Sadler and
Nordlinger (2004) in using the standard set-valued attribute adj for adjuncts, rather than assuming
different adjunct functions indexed by semantic role.
21 A euro is a type of wallaroo, a kangaroo-like animal.
 the description function d 467

The f-structure for (75) is given in (76), which meets the requirements given in
(74) for the environment in which thara-ngka-marta-a must appear:

(76)  
’
   
 
   
 
   
 
 
 
  
 
 
  
  
  
     
  
  
 
 
  
 
  
  
  
 
   

      
  

        
  
  
 
  
  p   
  
 
    
   

Sadler and Nordlinger (2004, 2006b) provide an analysis of case stacking within the
theory of Paradigm Function Morphology (Stump 2001) which, like the complex
m-features discussed in the previous section, appeals to the possibility of an
enriched structure for m-features: the order of the case features in case stacking is
captured by allowing the value of a complex m-feature to consist of an embedded
set of m-case m-features.
The table in (77), based on Sadler and Nordlinger (2004), states the relation
between morphological case and the corresponding f-description. The C and
M subscripts represent a distinction between Core cases, used adnominally and
relationally, and Modal cases, encoding temporal contributions as well as comple-
mentizing and associating functions; for more discussion of this distinction, see
Sadler and Nordlinger (2004, 2006b):

(77) m-feature f-description

CaseC :Loc (↑ concord case) = loc, (adj ∈ ↑)
CaseC :Abl (↑ concord case) = abl, (adj ∈ ↑)
CaseC :Prop (↑ concord case) = prop, (adj ∈ ↑)
CaseC :Erg (↑ concord case) = erg, (subj ↑)
CaseC :Nom (↑ concord case) = nom, (subj ↑)
CaseM :Abl (↑ tense) = pst

In line with the theory of constructive case originally proposed by Nordlinger

(1998) and described in §6.1.3, Sadler and Nordlinger (2004, 2006b) propose the
468 the interface to morphology

f-description for thara-ngka-marta-a given in (78), which describes the f-structure

in (74):22

(78) (↑ pred) = ‘pouch’

(↑ case) = loc
((adj ∈ ↑) concord case) = prop
((adj ∈ adj ∈ ↑) concord case) = acc
(obj adj ∈ adj ∈ ↑)

Our goal, then, is to produce the f-description in (78) on the basis of the lexemic
entry for thara, the m-entry for thara-ngka-marta-a, and a D-mapping given the
m-entry.
We propose the following lexemic entry for thara:

(79) Lexemic entry for the root thara1:

LE <{<m-root:thara>}, {(↑ pred) = ‘pouch’}, thara1>

The f-description in this lexemic entry provides the first line of the f-description
in (78).
Next, we require an m-entry for the word form thara-ngka-marta-a as pro-
vided by the realizational component R for Martuthunira. We adopt Sadler and
Nordlinger’s insight that the value of the m-case feature in case stacking is best
represented as a recursively embedded set reflecting the order in which the case
affixes appear in the word form:

(80) M-entry for thara-ngka-marta-a:

<thara1, thara-ngka-marta-a, /yɐɻɐŋkɐmɐʈɐ:/,
{ m-case:{m-caseC :loc, {m-caseC :prop, {m-caseC :acc}}} }>

The task is now to extend the definition of D in such a way that this complex
m-case specification maps to the following f-description, as desired:

D
(81) m-case:{m-caseC :loc, {m-caseC :prop, {m-caseC :acc}}} ⇒
{(↑ case) = loc,
((adj ∈ ↑) concord case) = prop,
((adj ∈ adj ∈ ↑) concord case) = acc,
(obj adj ∈ adj ∈ ↑)}

Unlike the analysis of Sadler and Nordlinger (2004), we do not construct a tree
as a part of the D-mapping; rather, we follow Andrews (2005) in proposing a
recursive definition of D for set-valued features which produces the appropriate
f-description.

22 The f-description in (78) involves inside-out functional uncertainty (§6.1.3) and uses the set
membership symbol ∈ as an attribute (§6.3.1).
 the description function d 469

It is helpful to reformulate the f-description in (81) to indicate which part of

the m-case feature is responsible for which subpart of the f-description. Following
Andrews (2005), we augment the f-description in (81) with separately specified
existential constraints for each case specification. For example, in (82) we include
the existential constraint (adj ∈ ↑), which requires the f-structure ↑ to be a
member of an adjunct set; this is part of the contribution of locative casemarking.
This constraint is implied by the constraint (obj adj ∈ adj ∈ ↑) in (81), which
requires ↑ to be a member of an adjunct set which is itself in an adjunct set of
an object. Including this simpler constraint and the other existential constraints in
(82) produces a slightly expanded f-description which is equivalent to the one in
(81), but where the commonalities in the different components of the f-description
are more clearly revealed:
6
(↑ case) = loc
(82) m-caseC :loc
(adj ∈ ↑)
6
((adj ∈ ↑) concord case) = prop
{m-caseC :prop}
(adj ∈ adj ∈ ↑)
6
((adj ∈ adj ∈ ↑) concord case) = acc
{{m-caseC :acc}}
(obj adj ∈ adj ∈ ↑)

We can now give a slightly lengthier but more revealing restatement of the relevant
f-description, closely following proposals by Andrews (2005) and using local
names for f-structures (prefixed by a percent sign %: see §6.5):

(83) ↑ = %f0

 (%f0 concord case) = loc
m-caseC :loc (adj ∈ %f0 ) = %f1

%f1

 (%f1 concord case) = prop
{m-caseC :prop} (adj ∈ %f1 ) = %f2

%f2

 (%f2 concord case) = acc
{{m-caseC :acc}} (obj %f2 ) = %f3

%f3

As this restatement makes clear, the f-structure which is constrained at each

level of embedding is the one which is existentially quantified at the previous
level of embedding. For the m-case feature, the following mappings are relevant.
The definition in (84a) applies when the set value of the m-case attribute has
two elements: a particular case value, and an embedded set containing further
specification of the f-structure environment. The definition in (84b) applies when
the set value of the m-case attribute has only one element. The definitions in (84c)
specify the inside-out path _path required for each of the first two definitions:
470 the interface to morphology

(84) D-mapping for constructive m-case:

a. Dfeat <LI, m-case:{_case1, _cases}, m-features, f ,
{(f concord case) = _case1, (_path f ) = f1 , f1 } ∪ d>
if Dfeat <LI, m-case:_cases, m-features, f1 , d>.
b. Dfeat <LI, m-case:{_case1}, m-features, f ,
{(f concord case) = _case1, (_path f ) = f1 , f1 }>.
c. If _case1 is loc, abl, or prop, _path is adj ∈.
If _case1 is erg or nom, _path is subj.
If _case1 is acc, _path is obj.

The definitions in (84a) and (84b) are relevant for all languages with case stacking.
The definitions in (84c) are particular to Martuthunira, and are specified on a
language-by-language basis.

12.4 Further reading and related issues

We have not provided a detailed history of LFG work on morphology and the
interface between the morphological component and the rest of the grammar.
The seminal Workshop on Morphology organized by Louisa Sadler and Andy
Spencer at the LFG2000 conference marked the beginning of serious discussion
of morphology in an LFG setting and the role of the morphological component
in the LFG architecture; Sadler and Spencer (2000) present an overview of the
workshop and provide pointers to some of the papers and presentations. Sadler and
Spencer (2004) collects subsequent LFG-based work on morphology, and Sadler
and Nordlinger (to appear) provide a very useful overview and comparison of
work on morphology in the LFG and HPSG frameworks, including a discussion of
morpheme-based approaches assuming annotated sublexical trees (Simpson 1991;
Nordlinger 1998; Bresnan et al. 2016); see also Bonami and Crysmann (2016).
Morphological periphrasis has been a focus of interest since the foundational
work of Sadler and Spencer (2001); Spencer (2013) provides a discussion of the
relevant issues, and Biswas (2017) proposes a treatment of some of the central cases
which combines LFG and Paradigm Function Morphology.
In the foregoing, we have assumed that f-structures and lexical entries are fully
specified, and we have not appealed to any markedness or blocking principles that
would lead us to choose the least marked form or the most specific compatible form
in any given instance. This issue was discussed in detail by Andrews (1990b), who
proposes the Morphological Blocking Principle, requiring that the most specific
compatible lexical item must be chosen. Much LFG research discusses and relies on
the Morphological Blocking Principle, and Bresnan (2001b) proposes a recasting of
the principle in Optimality-Theoretic terms; Optimality-Theoretic LFG analyses
are briefly discussed in §18.1.3.
 13
Modification

This chapter explores issues in the syntax and semantics of modification. Since
there is in principle no limit to the number of modifiers that a phrase can have,
we represent modifiers at functional structure as members of a set of modifying
adjuncts adj (§2.4.4). Functional annotations on c-structure rules ensure that at
f-structure, each modifier appears as a member of the adjunct set of the phrase it
modifies.
In this chapter, we concentrate in particular on adjectival modification, since
the syntax and semantics of adjectives is fairly complex and illustrates many of
the issues of interest to us. §13.1 provides an overview of the syntax of adjectival
modification, §13.2 discusses three semantic classes of adjectives and how their
meanings are represented, and §13.3 discusses adjectival modification at the
syntax-semantics interface within the glue approach.
Defining the semantic contribution of a modifier brings up a set of tricky
problems, as first noticed by Kasper (1997). In §13.4, we address these issues and
show that they have a straightforward solution within our framework.
The chapter concludes with a brief examination of the syntax and semantics
of adverbial modification: §13.5 discusses the syntax and semantics of manner
adverbs like skillfully, as well as sentential adverbs like clearly.

13.1 Syntax of adjectival modification

13.1.1 Modification at functional structure
As discussed in §2.1.2, modifiers are different from arguments in that an f-structure
can contain any number of modifiers, including none; the clause in (1a) has two
modifiers, while (1b) shows that there cannot be multiple obj arguments in a
clause:

(1) a. The girl handed the baby a toy on Tuesday in the morning.
b. ⃰David saw Tony Mr. Gilroy my next-door neighbor.

At f-structure, each modifier is a member of the adj set. In (2), the adjectival
modifier Swedish is treated as a member of the adj set of modifiers of the noun
man:

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
474 modification

(2) Swedish man

! &
pred ‘man’
"# $%
adj pred ‘Swedish’

In phrases with more than one modifier, the f-structure for each modifier appears
as a member of the adj set:

(3) tall Swedish man

 
  
  
 
 
 

The lexical entries for the attributive adjectives tall and Swedish and the noun
man contain at least the following syntactic information, including specification of
syntactic form (fm), phrase structure category (the result of applying the labeling
function λ to the terminal c-structure node π(•)), and the value of the pred
feature:1

(4) Lexical entries for tall, Swedish, and man:

a. tall A
3 (↑ pred) = ‘tall’
b. Swedish A
3 (↑ pred) = ‘Swedish’
c. man N (↑ pred) = ‘man’

13.1.2 Constituent structure constraints

At constituent structure, modifiers are often adjoined to the phrases they modify
(§3.3.4). The c-structure and f-structure for the English noun phrase Swedish man
is shown in (5). The prenominal modifier Swedish is a non-projecting adjective 3
A,
adjoined at the N level (see §3.3.4, Toivonen 2003, Arnold and Sadler 2013, and
references cited there):2

1 Since we do not provide morphological or prosodic analyses in Part III, we present lexical entries
in the traditional format introduced in §5.3.2, and not the fully-specified format, including the p-form,
which was introduced in §11.4.1.
2 For discussion of postnominal AP modifiers, see Arnold and Sadler (2013).
 syntax of adjectival modification 475

(5) Swedish man

NP

N
↑ =↓

N
↑ =↓

A N
↓ ∈ (↑ ↑ =↓

Swedish man

We propose the following adjunction rule for prenominal adjective modifiers in

English:

(6) N −→ A
3 N
↓ ∈ (↑ adj) ↑=↓

This rule allows nonprojecting adjectives to be adjoined at the N level. At

f-structure, each adjective is a member of the adjunct set adj. In (7), two adjectives
have been adjoined:

(7) tall Swedish man

N
↑ =↓

A N
↓ ∈ (↑ ↑ =↓

tall A N
(↑ ↓ ∈ (↑ ↑ =↓

Swedish man
(↑ (↑
476 modification

13.2 Semantic classes of adjectives

Foundational work on the semantics of adjectives was carried out by Montague
(1970a) and Kamp (1975), who focused primarily on the three types of adjectives
to be examined in this section. Their basic view of the semantics of adjectival
modification has been widely adopted.
As discussed in §8.8.2, we define a common noun’s meaning thus:

(8) Meaning of man:

λx.man (x)

This meaning is of type &e → t'. It picks out the set of men—that is, the set of
entities x for whom the proposition man (x) is true. When a meaning like the one
in (8) is modified, the result is a meaning which is of the same type but which
reflects a modified meaning rather than the original unmodified meaning. In this
section, we describe how noun meanings are modified in different ways by different
semantic classes of adjectives.

13.2.1 Intersective adjectives

The meaning of Swedish man can be represented as in (9), in which the conjunction
operator ∧ conjoins the two expressions Swedish (x) and man (x):

(9) Meaning of Swedish man:

λx.Swedish (x) ∧ man (x)

The type of this meaning is &e → t', just like the unmodified meaning man; the
difference in meaning is that this expression picks out the set of individuals x that
satisfy both the predicate Swedish (x) and the predicate man (x)—the individuals
that are both Swedish and men. Adjectives like Swedish are called intersective, since
the individuals who are Swedish men are those that are in the intersection3 of the
set of individuals that are Swedish with the set of individuals that are men. As Partee
(1995) points out, for intersective adjectives the conclusion in (10c) follows from
the premises in (10a) and (10b):

(10) a. David is a Swedish man. [David is in the intersection of the set of Swedish
individuals and the set of men.]
b. David is a football player. [David is in the set of football players.]
c. Therefore, David is a Swedish football player. [David is in the intersection
of the set of Swedish individuals and the set of football players; that is,
David is both Swedish and a football player.]

3 The intersection of two sets contains all of the members that the two sets have in common: see §6.3.
 semantic classes of adjectives 477

13.2.2 Subsective adjectives

Not all adjectives are intersective. For example, Partee (1995) notes that we cannot
perform the same kind of reasoning as in (10) with an adjective like big:

(11) a. Mortimer is a big mouse.

b. Mortimer is an animal.
c. [does not follow:] Therefore, Mortimer is a big animal.

Such adjectives are called subsective: the set of big mice is a subset of the set of
mice, but a big mouse is not in the intersection of the set of things that are big in
general with the set of mice. Indeed, it is not clear whether it makes sense to talk
about the set of things that are “big in general”; to the extent that it does, the set
of things that are “big in general” would contain solar systems and galaxies, and
possibly mountains and oceans, but certainly not mice.
There are several kinds of subsective adjectives.4 One type is gradable adjectives
like big or tall, which, as noted by Montague (1970a), Kamp (1975), Siegel (1976),
Kennedy (1997), and many others, must be interpreted relative to some relevant
standard. For example, a particular mouse might count as a big mouse, even though
the same mouse is probably not a big animal or even a big rodent. Similarly, a seven-
year-old boy can be correctly characterized as a tall seven-year-old even if he is not
tall compared to an adult.
We propose the following simplified meaning for big mouse (see Kennedy 1997
for a full discussion of the semantics of gradability and comparison):

(12) Meaning of big mouse:

λx.big (x, P ) ∧ mouse (x)

The argument P of big represents the contextually salient property that determines
the relevant standard of measurement. If the property P of the individual is that it
is a mouse, modification by the adjective big requires the individual to exceed some
standard of size that is determined by reference to mousehood. In other words, in
order to claim that something is big relative to the property of being a mouse, we
need to know the range of sizes that are usual for mice.
In a neutral setting, the contextually relevant property is often the property
denoted by the modified noun; for example, the contextually salient property P in
an example like big mouse is generally resolved to the property of being a mouse.
However, as pointed out by McConnell-Ginet (1979) and Pollard and Sag (1994),

4 One well-studied type of subsective adjective is exemplified by beautiful in one reading of beautiful
dancer, which can refer either to an individual who is both beautiful and a dancer, or (in the non-
intersective, subsective reading) to an individual who dances beautifully (Siegel 1976). Larson (1998)
proposes to treat this ambiguity by allowing beautiful to specify either the individual variable for dancer,
giving rise to the meaning ‘beautiful and a dancer’, or the event variable associated with the dancing
event, giving rise to the meaning ‘dance beautifully’. Though we will not provide an explicit analysis of
these examples here, such an analysis can be easily formulated in a glue setting by adopting a semantic
representation including event variables, as introduced in §8.10.
478 modification

in certain contexts other interpretations are also possible. Pollard and Sag provide
the following example:

(13) The Linguistics Department has an important volleyball game coming up

against the Philosophy Department. I see the Phils have recruited Julius to
play with them, which means we are in real trouble unless we can find a good
linguist to add to our team in time for the game.

Here the property P relevant to the interpretation of the adjective good is being a
volleyball player, since in this example good linguist means, more or less, linguist
who is good at playing volleyball. Examples such as these show that the property
P need not correspond to the property denoted by the modified noun, but can be
determined contextually.
Of course, modified phrases can undergo further modification. The meaning of
the doubly modified phrase tall Swedish man is:

(14) Meaning of tall Swedish man:

λx.tall (x, P ) ∧ Swedish (x) ∧ man (x)

Even in a neutral context, the contextually relevant property P involved in the

interpretation of the adjective tall can be resolved in more than one way. It can
refer to someone who is Swedish, a man, and tall for a man, in which case the
contextually relevant property P is the property of being a man. It can also refer
to someone who is Swedish, a man, and tall for a Swedish man, in which case the
contextually relevant property P is the property of being a Swedish man.

13.2.3 Nonsubsective adjectives

Nonsubsective adjectives are a third class of modifying adjectives, including adjec-
tives like imaginary, former, fake, and alleged. The statements in (15b) and (16b)
do not follow from the statements in (15a) and (16a) respectively:

(15) a. David is a former student.

b. [does not follow:] David is a student.

(16) a. This is a fake gun.

b. [does not follow:] This is a gun.

These adjectives, studied by Kamp (1975) and in more detail by Siegel (1976),
are different from those discussed in the previous section in an important way:
a Swedish man is a man, and a big mouse is a mouse, but a fake gun is not a
gun; instead, it may actually be a toy or a piece of soap. Thus, the meaning of a
phrase with a nonsubsective adjective depends on the meaning of the unmodified
phrase, but the resulting property may hold of an individual even if the unmodified
meaning does not. In other words, the set of individuals referred to by the
modified noun is in general not a subset of the set of individuals referred to by
the unmodified noun: this is why these adjectives are called nonsubsective.
 modifiers and semantic composition 479

Like other adjectives, a nonsubsective adjective operates on the description it

modifies and produces a new description of the same type:5

(17) Meaning of former student:

λx.former (x, student )

A former student is one who at some previous time was a student, but who is
no longer a student; the meaning of student is important in understanding the
meaning of former student, but the individuals represented by x in the meaning
given in (17) for former student are not required to be students. Thus, former in (17)
denotes a relation between the property of being a student and some individual
who formerly had that property. Similarly, a fake gun is an entity that is not a gun,
but which has some properties in common (for example, appearance) with entities
that are actually guns; again, although a fake gun is not a gun, the meaning of gun
is important in determining the meaning of fake gun:

(18) Meaning of fake gun:

λx.fake (x, gun )

Importantly, the resulting meaning still has the same type as the unmodified noun;
nonsubsective adjectives, like intersective adjectives and gradable adjectives, turn
an unmodified &e → t' meaning into a modified &e → t' meaning.6 In fact, the
type of an unmodified meaning is always the same as the type of the modified
meaning, and this is important in our analysis of modification and meaning
composition.

13.3 Modifiers and semantic composition

As Montague (1970a) and Kamp (1975) point out, adjectival modifiers are func-
tions that take a property of type &e → t' (such as the property of being a man)
and produce a new property (such as the property of being a Swedish man). This
intuition is reflected in the glue semantic premises contributed by modifiers.

5 Partee (2010) argues that adjectives that have traditionally been classified as nonsubsective are
actually subsective, inducing coercion of the meaning of the modified noun. In Partee’s example (a),
for instance, the denotation of the first underlined occurrence of fur is expanded to include not only
fur, but material resembling fur in some respect, while the second underlined occurrence of fur has its
normal meaning, and refers only to actual fur:
(a) I don’t care whether that fur is fake fur or real fur.
Adopting this view and eliminating the category of nonsubsective adjectives would allow a simplifica-
tion of the compositional semantic treatment that we present in this chapter; for the sake of clarity of
analysis and consistency with much of the literature on adjective semantics, we maintain the traditional
three-way distinction between intersective, (non-intersective) subsective, and nonsubsective adjectives
in our discussion.
6 We ignore the role of intensionality in the interpretation of subsective and nonsubsective adjectives;
for discussion of the semantics of adjectives and intensionality, see Kamp and Partee (1995) and Partee
(2010).
480 modification

13.3.1 Adjectival modification

As shown in §8.8.2, a common noun like man is associated with the syntactic and
semantic structures and meaning constructor given in (19), where the semantic
structures v and r are the values of the attributes var and restr in the semantic
structure m𝜎 :

(19) F-structure, s-structure, and meaning constructor for man:

v[ ]
m m
r[ ]
λx.man (x) : v ⊸ r

A modified noun like Swedish man is associated with a meaning constructor whose
right-hand (glue) side is exactly the same as the meaning constructor for man, but
whose left-hand (meaning) side is associated with a modified meaning rather than
an unmodified one:

(20) F-structure, s-structure, and meaning constructor for Swedish man:

v[ ]
m m
r[ ]
λx.Swedish (x) ∧ man (x) : v ⊸ r

In this section, we show how a meaning constructor like the one in (20) is derived
from the meaning constructors for Swedish and man.
Augmented lexical entries for Swedish and man, including meaning construc-
tors, are as follows:7

(21) Lexical entries for man and Swedish, including meaning constructors:
a. man N (↑ pred) = ‘man’
λx.man (x) : (↑𝜎 var) ⊸ (↑𝜎 restr)
b. Swedish 3
A (↑ pred) = ‘Swedish’
λP.λx.Swedish (x) ∧ P(x) :
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]

The meaning constructor for man is familiar from our discussion of common
nouns in §8.8.2. The meaning constructor for Swedish uses inside-out functional
uncertainty (§6.1.3) to refer to the semantic structure of the phrase it modifies. The
expression (adj ∈ ↑) refers to the f-structure in which ↑ appears as a member of
the modifier set;8 in (20), this f-structure is labeled m. The expression (adj ∈ ↑)𝜎

7 In §13.4, we will refine this analysis in our discussion of recursive modification (modification of
modifiers).
8 The use of the set membership symbol ∈ as an attribute is discussed in §6.3.1.
 modifiers and semantic composition 481

refers to the semantic structure corresponding to that f-structure (so m𝜎 in 20),

and the expression ((adj ∈ ↑)𝜎 var) refers to the value of the attribute var in that
semantic structure, labeled v in (20). Similarly, the expression ((adj ∈ ↑)𝜎 restr)
refers to the value of the restr attribute, labeled r in (20).
Instantiating the meaning constructors in (21) according to the labels on the
structures displayed in (20), we have the following instantiated meaning construc-
tors for Swedish and man:

(22) Meaning constructor premises for Swedish man:

[man] λx.man (x) : v ⊸ r
[Swedish] λP.λx.Swedish (x) ∧ P(x) : [v ⊸ r] ⊸ [v ⊸ r]

The right-hand side of the meaning constructor for Swedish illustrates the charac-
teristic glue contribution of a modifier: it requires a resource of the form v ⊸ r as
its argument and produces a resource of exactly the same form. The general form
for modifiers is given in (23), where M is the meaning of the modifier and S is the
glue contribution of the phrase it modifies:

(23) General form of modifier meaning constructor:

M:S⊸S

Modifiers consume a meaning resource S and produce an identical new meaning

resource S for the phrases they modify.
Given the premises [Swedish] and [man], we can perform a deduction that
produces the meaning constructor for Swedish man given in (20):

(24) λx.man (x) : v ⊸ r

This is the meaning constructor [man]. It associates
the meaning λx.man (x) with the implicational con-
tribution v ⊸ r.
λP.λx.Swedish (x) ∧ P(x) : [v ⊸ r] ⊸ [v ⊸ r]
This is the meaning constructor [Swedish]. On the
glue side, it consumes the noun contribution v ⊸ r
and produces a new modified meaning which is also
associated with v ⊸ r. On the meaning side, we apply
the function λP.λx.Swedish (x) ∧ P(x) to the unmod-
ified meaning contributed by man, λx.man (x).

λx.Swedish (x) ∧ man (x) : v ⊸ r

We have produced a modified meaning
λx.Swedish (x) ∧ man (x) associated with the impli-
cational contribution v ⊸ r. Note that the glue side
of this meaning constructor is exactly the same as
the glue side of the unmodified meaning constructor
[man]; only the meaning is different.
482 modification

We can also represent this deduction in abbreviated form, as shown in Chapter 8,

using the labels in (22):

(25) [Swedish], [man] ) λx.Swedish (x) ∧ man (x) : v ⊸ r

13.3.2 Gradable adjectives

We have seen that gradable adjectives like big differ from intersective adjectives like
Swedish in introducing a contextually salient property P in their interpretation:

(26) big mouse

v[ ]
m m
r[ ]
λx.big (x, P ) ∧ mouse (x) : v ⊸ r

The meaning contribution of big mouse given in (26) refers to a mouse that exceeds
the size of individuals that are described by the contextually determined property
P . Since the property P is determined by contextual factors, not by semantic
composition on the basis of syntactic relations, we will not specify a means for
determining P but instead will leave it uninstantiated.
Although the meaning contribution of a gradable adjective like big is not the
same as that of an intersective adjective like Swedish, the right-hand sides of the
two meaning constructors are the same, since the two kinds of adjective play a
similar role in meaning assembly. The lexical entry for big in (27) again includes
a somewhat simplified meaning constructor, pending our discussion of recursive
modification in §13.4.

(27) Lexical entry for big:

big A
3 (↑ pred) = ‘big’
λR.λx.big (x, P ) ∧ R(x) :
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]

Like the entry for Swedish given in (21), this entry uses inside-out functional
uncertainty to refer to the f-structure of the phrase it modifies. The lexical entry
for mouse is exactly analogous to the one for man, and will not be displayed.
Instantiating the lexical entries for big and mouse according to the labels in (26),
we obtain the following instantiated meaning constructors:

(28) Meaning constructor premises for big mouse:

[mouse] λx.mouse (x) : v ⊸ r
[big] λR.λx.big (x, P ) ∧ R(x) : [v ⊸ r] ⊸ [v ⊸ r]
 modifiers and semantic composition 483

The meaning constructor for big requires a meaning resource of the form v ⊸ r,
and mouse provides such a resource. The resulting meaning is obtained by applying
the expression λR.λx.big (x, P ) ∧ R(x) to its argument λx.mouse (x). The result is
as desired—from the meaning constructors labeled [big] and [mouse] in (28), we
derive the following meaning constructor for big mouse:

(29) [big], [mouse] ) λx.big (x, P ) ∧ mouse (x) : v ⊸ r

13.3.3 Nonsubsective adjectives

The syntactic and semantic structures and meaning constructor for the phrase
former student are:

(30) former student

v[ ]
s s
r[ ]
λx.former (x, student ) : v ⊸ r

The (simplified) lexical entry of former is as follows:

(31) Lexical entry for former:

former A
3 (↑ pred) = ‘former’
λP.λx.former (x, P) :
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]

As shown earlier, the meaning contribution of an nonsubsective adjective like

former is different from Swedish and big. Nevertheless, it contributes a meaning
resource of the same form: it consumes a resource corresponding to the phrase it
modifies and produces a new resource of the same form. The instantiated meaning
constructors for former and student are as follows:

(32) Meaning constructor premises for former student:

[student] λx.student (x) : v ⊸ r
[former] λP.λx.former (x, P) : [v ⊸ r] ⊸ [v ⊸ r]

As desired, these meaning constructors combine to produce the meaning construc-

tor for former student given in (30):

(33) [former], [student] ) λx.former (x, student ) : v ⊸ r

Although each type of modifier makes a different kind of contribution to meaning,

their roles in meaning assembly are similar; this is reflected in the meaning
484 modification

resources on the right-hand sides of the meaning constructors for the modifying
adjectives we have examined.

13.4 Recursive modification

In the previous section, we assumed that the function of a modifier is to specify
the result that is obtained when it combines with the phrase it modifies—in other
words, that the meaning of an adjective is defined in terms of its effect on the
element that it modifies. This common assumption is challenged in an important
paper by Kasper (1997), who discusses evidence from recursive modification:
cases in which a modifier is itself modified. In this section, we review Kasper’s
observations and show how they are accounted for in the glue approach.
Consider a modifier like Swedish, which we assumed in the previous section to
have a meaning constructor like this:

(34) Swedish
v[ ]
r[ ]
λP.λx.Swedish (x) ∧ P(x) : [v ⊸ r] ⊸ [v ⊸ r]

The meaning constructor for Swedish given in (34) provides information about
how to determine the meaning of the phrase it modifies. It does not provide a
representation for the meaning of Swedish independent of its modifying effect;
instead, it represents only the conjunctive meaning that results from combining
Swedish with the phrase it modifies.
Kasper (1997) shows that this view is inadequate by considering examples like
the following:

(35) obviously Swedish

 
v[ ]
 
r[ ]

λP.λx.obviously (Swedish (x)) ∧ P(x) : [v ⊸ r] ⊸ [v ⊸ r]

In this example, the modifier Swedish is itself modified by the adverb obviously. The
effect of modification by obviously is to modify the proposition that x is Swedish,
Swedish (x), to produce a new proposition obviously (Swedish (x)). However, the
proposition Swedish (x) is not by itself associated with the meaning of the adjective
Swedish, and in fact there is no clear way to disentangle the meaning Swedish (x)
from the rest of the meaning contribution for Swedish in (34).
For a meaning like Swedish (x) to be available, we require an independent,
modifiable characterization of the intrinsic meaning of Swedish, which we assume
 recursive modification 485

to be the &e → t' type meaning λx.Swedish (x). We also require a theory of how
this meaning combines with the meaning of the modified noun. Kasper (1997)
provides an analysis of examples like (35) within the framework of Head-Driven
Phrase Structure Grammar; though it is stated in different formal terms, our
analysis has a clear basis in Kasper’s proposal.

13.4.1 Meaning constructors for modifiers

In light of these observations, we must refine our assumptions about the meaning
contributions of adjectives. We first note that adjectives like Swedish can also be
used predicatively:9

(36) Sven is Swedish.

For this case as well, the appropriate meaning for Swedish is λx.Swedish (x), the
intrinsic, modifiable meaning of the adjective.
There is an independent difference in English between adjectives used predica-
tively, as in (36), and adjectives used as prenominal attributive modifiers: as shown
in (5), English prenominal attributive adjectives are nonprojecting ( 3A ) categories,
while other, nonprenominal uses of adjectives involve the projecting category A as
the head of AP. In fact, some English adjectives are only nonprojecting, and can
be used only prenominally (late with the meaning ‘deceased’: the late president,
but not The president is late, which means The president is tardy and not The
president is deceased), while other adjectives do not have a nonprojecting form and
cannot be used prenominally (afraid: The man is afraid/the man afraid of spiders,
but not ⃰the afraid man). Thus, the different meaning contributions of attributive
and predicative adjectives in English correlate with their status as projecting or
nonprojecting c-structure categories. Predicative adjectives are associated only
with a type &e → t' meaning, while attributive adjectives contribute an additional
meaning constructor which combines the intrinsic adjective meaning with the
meaning of the noun.
We now assume that the semantic structures of adjectives are internally struc-
tured, containing at least the attribute var. In (37), the f-structure f corresponds
to a semantic structure f𝜎 with the attributes var and restr; as shown earlier,
the values of these attributes are labeled v and r. The f-structure s of the adjective
Swedish also has an attribute var, whose value we have labeled sv:

(37) Swedish
v[ ]
f s f
r[ ]

s sv[ ]

9 See Lowe (2013) for discussion of commonalities and differences in meaning constructors for
attributive and predicative uses of adjectives, and §5.4.5 for discussion of copular constructions.
486 modification

The intrinsic meaning of the adjective Swedish is of type &e → t'. Since we assume
that the basic types e and t are associated with semantic structures, we assign the
type e to sv and the type t to s𝜎 . Thus, the meaning constructor in (38), which
represents the core meaning of the word Swedish, is common to both projecting
and nonprojecting forms of the word:10

(38) Meaning constructor for the adjective Swedish:

λx.Swedish (x) : (↑𝜎 var) ⊸ ↑𝜎

The meaning constructor in (39) for the nonprojecting, attributive use of Swedish
contains an additional meaning constructor, whose function is to provide instruc-
tions for combining the first meaning constructor with the noun it modifies:

A Swedish (final):
(39) Lexical entry for nonprojecting 3
Swedish 3
A (↑ pred) = ‘Swedish’
λx.Swedish (x) : (↑𝜎 var) ⊸ ↑𝜎
λR.λQ.λx.R(x) ∧ Q(x) :
[(↑𝜎 var) ⊸ ↑𝜎 ] ⊸
[[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]]

Instantiating the two meaning constructors in (39) according to the labels given in
(37) makes them much easier to read; we label the first meaning constructor in the
lexical entry in (39) as [Swedish-basic] and the second as [attr-adj]:

(40) Meaning constructor premises for Swedish, used attributively:

[Swedish-basic] λx.Swedish (x) : [sv ⊸ s𝜎 ]
[attr-adj] λR.λQ.λx.R(x) ∧ Q(x) : [sv ⊸ s𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]

Importantly, we can deduce the meaning constructor for Swedish given in

(22) from the two meaning constructors in (40). The meaning constructor
[Swedish-basic] provides the semantic resource sv ⊸ s𝜎 that is required by
[attr-adj], and the resulting meaning is obtained by function application: the
meaning contribution of [attr-adj], λR.λQ.λx.R(x) ∧ Q(x), is applied to the
meaning contribution of [Swedish-basic], λx.Swedish (x).

(41) [Swedish-basic], [attr-adj] ) [Swedish]

10 We assume that in an example like (36) the meaning constructor for the copula manages the
relation between the value of the var attribute in the predicative adjective’s semantic structure and
the subj of the copular construction.
 recursive modification 487

Therefore, the two meaning constructors [Swedish-basic] and [attr-adj] can play
exactly the same role in meaning assembly as the simple meaning constructor
[Swedish] discussed in §13.3.1. In particular, from the premises [Swedish-basic],
[attr-adj], and [man], we correctly derive the meaning constructor for Swedish
man given in (24):

(42) [Swedish-basic], [attr-adj], [man] ) λx.Swedish (x) ∧ man (x) : v ⊸ r

We treat other adjectival modifiers similarly: the basic meaning constructor for an
adjective combines with a meaning constructor that reflects the function of the
adjective, such as attributive modification, in the syntactic context of its use.
More generally, the example just presented illustrates that the simple and intu-
itive assumptions we make about meanings and how they combine often turn out
to be largely correct, but may be in need of refinement to account for more compli-
cated examples. In logical terms, the intuitively motivated meaning constructors
often correspond to conclusions resulting from a deduction from a more refined set
of basic meaning constructor premises. It is often easier to work with the simpler
and more intuitive constructors; this is legitimate and theoretically sound, as long
as it can be shown that they follow as a logical consequence from the more basic
premises.

13.4.2 Modification of modifiers

We now show the derivation of the meaning for obviously Swedish man, an example
in which the modifier Swedish is itself modified. As above, we introduce a var
attribute with value ov in the semantic structure o𝜎 corresponding to obviously:

(43) [obviously Swedish] man

v[ ]
f
r[ ]
 
 
f  s sv[ ]
 s 
o

o ov[ ]
λx.obviously (Swedish (x)) ∧ man (x) : v ⊸ r

Like adjectives, the adverb obviously can be modified:

(44) [[quite obviously] Swedish] man

As with the meaning constructors for Swedish given in (40), which we labeled
[Swedish-basic] and [attr-adj], the meaning contribution of obviously is twofold.
The lexical entry for obviously is therefore:
488 modification

(45) Lexical entry for obviously:

obviously Adv
4 (↑ pred) = ‘obviously’

λP.obviously (P) : [(↑𝜎 var) ⊸ ↑𝜎 ]

λQ.λR.λx.Q(R(x)) :
[(↑𝜎 var) ⊸ ↑𝜎 ] ⊸
[[((adj ∈ ↑)𝜎 var) ⊸ (adj ∈ ↑)𝜎 ] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ (adj ∈ ↑)𝜎 ]]

Again, readability increases when the meaning constructors are instantiated

according to the labels in (43):

(46) Meaning constructor premises for obviously:

[obviously-basic] λP.obviously (P) : [ov ⊸ o𝜎 ]
[adv-adj] λQ.λR.λx.Q(R(x)) : [ov ⊸ o𝜎 ] ⊸ [[sv ⊸ s𝜎 ] ⊸ [sv ⊸ s𝜎 ]]

The meaning constructor labeled [obviously-basic] in (46) specifies the intrinsic,

modifiable meaning of obviously, and the meaning constructor labeled [adv-adj]
combines the intrinsic meaning of the adverb with the meaning of the adjective
which it modifies.
The two meaning constructors [obviously-basic] and [adv-adj] combine to
produce this meaning constructor:

(47) [obviously] λR.λx.obviously (R(x)) : [sv ⊸ s𝜎 ] ⊸ [sv ⊸ s𝜎 ]

This meaning constructor consumes a meaning resource of the form sv ⊸ s𝜎 ,

producing a new meaning resource of the same form but corresponding to a
modified meaning.
We can now combine the meaning constructor [obviously] with the meaning
constructor [Swedish-basic] in (40) to yield this meaning constructor:

(48) [obviously-Swedish] λx.obviously (Swedish (x)) : [sv ⊸ s𝜎 ]

Notably, the right-hand side of this meaning constructor is the same as the right-
hand side of the unmodified meaning constructor [Swedish-basic], repeated here,
and plays the same role in meaning composition:

(49) [Swedish-basic] λx.Swedish (x) : [sv ⊸ s𝜎 ]

Next, we combine the meaning constructors [obviously-Swedish], [attr-adj], and

[man] to produce the meaning constructor given in (43) for obviously Swedish
man:

(50) [obviously-Swedish], [attr-adj], [man] )

λx.obviously (Swedish (x)) ∧ man (x) : v ⊸ r
 recursive modification 489

Thus, our refined theory of the semantic contribution of modifiers not only enables
the clean and intuitive treatment of modification presented in §13.3 but also allows
an analysis of recursive modification, which, as Kasper (1997) shows, has proven
problematic in many other approaches.

13.4.3 Modified gradable adjectives

For modified gradable adjectives, the derivation proceeds in exactly the same
way as for intersective adjectives. We propose the following lexical entry for the
adjective big:

A big (final):
(51) Lexical entry for nonprojecting 3
big A
3 (↑ pred) = ‘big’
λx.big (x, P ) : (↑𝜎 var) ⊸ ↑𝜎
λR.λQ.λx.R(x) ∧ Q(x) :
[(↑𝜎 var) ⊸ ↑𝜎 ] ⊸
[[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]]

The f-structure and semantic structures for obviously big mouse are:

(52) [[obviously big] mouse]

v[ ]
m
r[ ]
 
 
m  b bv[ ]
 b 
o

o ov[ ]
λx.obviously (big (x, P )) ∧ mouse (x) : v ⊸ r

Again, instantiating the meaning constructor premises for big in (51) according to
the labels in (52) makes them easier to read:

(53) Meaning constructor premises for big, used attributively:

[big-basic] λx.big (x, P ) : [bv ⊸ b𝜎 ]
[attr-adj] λR.λQ.λx.R(x) ∧ Q(x) : [bv ⊸ b𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]

Notably, the meaning constructor labeled [attr-adj], combining the intrinsic

meaning λx.big (x, P ) of the adjective big with the noun meaning, is exactly the
same as for the adjective Swedish. We assume the same meaning for obviously as
in (47), where the two lexical premises in (46) are combined:

(54) [obviously] λR.λx.obviously (R(x)) : [bv ⊸ b𝜎 ] ⊸ [bv ⊸ b𝜎 ]

490 modification

Given these premises, we can perform the following deduction:

(55) λR.λx.obviously (R(x)) : [bv ⊸ b𝜎 ] ⊸ [bv ⊸ b𝜎 ]

This is the meaning constructor [obviously]. It asso-
ciates the meaning λR.λx.obviously (R(x)) with the
implicational contribution [bv ⊸ b𝜎 ] ⊸ [bv ⊸ b𝜎 ],
consuming and then producing a resource of exactly
the form contributed by [big-basic].
λx.big (x, P ) : [bv ⊸ b𝜎 ]
This is the meaning constructor contributing the
intrinsic meaning of the adjective big, [big-basic]. On
the glue side, it contributes an implication bv ⊸ b𝜎 ,
and on the meaning side, it contributes the meaning
λx.big (x, P ), which holds of an individual x if x is big
relative to some contextually-supplied standard P .
λx.obviously (big (x, P )) : [bv ⊸ b𝜎 ]
By combining the two previous meaning construc-
tors [obviously] and [big-basic], we have produced
a modified meaning λx.obviously (big (x, P )) associ-
ated with the implicational contribution bv ⊸ b𝜎 .
Note that the glue side of this meaning constructor
is exactly the same as the glue side of the unmodified
meaning constructor [big-basic].
λR.λQ.λx.R(x) ∧ Q(x) : [bv ⊸ b𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]
This is the meaning constructor [attr-adj] (with the
semantic structure variables instantiated as in (52)),
whose purpose is to combine the meaning in the
previous step with the meaning constructor for the
noun mouse.
λQ.λx.obviously (big (x, P )) ∧ Q(x) : [v ⊸ r] ⊸ [v ⊸ r]
By combining the meaning constructors in the pre-
vious two steps, we have produced a meaning for
the attributive modifier obviously big, which can now
combine with the meaning constructor for mouse.
λx.mouse (x) : [v ⊸ r]
This is the meaning constructor [mouse]. On the glue
side, it is of the appropriate form to be consumed
by the modifier meaning constructor in the previous
step.
λx.obviously (big (x, P )) ∧ mouse (x) : [v ⊸ r]
By combining the meaning constructors in the previ-
ous two steps, we have produced the desired meaning
for obviously big mouse.
 recursive modification 491

13.4.4 Modified nonsubsective adjectives

We now turn to the slightly more complex question of modification of a nonsub-
sective adjective such as fake. Syntactically, there is nothing exceptional about fake,
and the f-structure and semantic structures for obviously fake gun are similar to the
structures for obviously Swedish man and obviously big mouse:

(56) [[obviously fake] gun]

v[ ]
g
r[ ]
 
 
g  f fv[ ]
 f 
o

o ov[ ]
λx.obviously (fake (x, gun )) : v ⊸ r

We assume the same meaning constructor for obviously as in the previous exam-
ples, but with the semantic structure variables instantiated as in (56):

(57) [obviously] λR.λx.obviously (R(x)) : [fv ⊸ f𝜎 ] ⊸ [fv ⊸ f𝜎 ]

However, as a nonsubsective adjective, fake makes a different meaning contribu-

tion from adjectives like Swedish and big. These differences are reflected in its
lexical entry:

A fake (final):
(58) Lexical entry for nonprojecting 3
fake A
3 (↑ pred) = ‘fake’
λP.λx.fake (x, P) : [((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]
⊸ [(↑𝜎 var) ⊸ ↑𝜎 ]
λQ.λx.Q(x) : [(↑𝜎 var) ⊸ ↑𝜎 ] ⊸ [((adj ∈ ↑)𝜎 var) ⊸
((adj ∈ ↑)𝜎 restr)] :

Instantiating the meaning constructors in (58) according to the semantic structure

labels in (56), we have:

(59) Meaning constructor premises for fake:

[fake-basic] λP.λx.fake (x, P) : [v ⊸ r] ⊸ [fv ⊸ f𝜎 ]
[nonsubsective-adj] λQ.λx.Q(x) : [fv ⊸ f𝜎 ] ⊸ [v ⊸ r]

The most important difference distinguishing fake from the other adjectives we
have examined is that fake requires the meaning of the noun it modifies as an
argument: on the glue side, it requires a meaning contribution of the form [v ⊸ r]
to obtain an intrinsic meaning of the form [fv ⊸ f𝜎 ], which is modifiable by
492 modification

obviously. Although these meaning constructors are quite different from those
contributed by intersective and subsective adjectives, no special versions of the
meaning constructors for adjectival modifiers like obviously or nouns like man,
mouse, or gun are required. Given the meaning constructors for obviously, fake,
and gun, we can perform the following deduction:

(60) λx.gun (x) : [v ⊸ r]

This is the meaning constructor [gun].
λP.λx.fake (x, P) : [v ⊸ r] ⊸ [fv ⊸ f𝜎 ]
This is the meaning constructor [fake-basic].
λx.fake (x, gun ) : [fv ⊸ f𝜎 ]
This meaning constructor results from combining
[fake-basic] with [gun].
λR.λx.obviously (R(x)) : [fv ⊸ f𝜎 ] ⊸ [fv ⊸ f𝜎 ]
This is the meaning constructor [obviously].
λx.obviously (fake (x, gun )) : [fv ⊸ f𝜎 ]
Combining [obviously] with the meaning construc-
tor resulting from the step above it, we have produced
a modified meaning associated with the implicational
contribution fv ⊸ f𝜎 .
λQ.λx.Q(x) : [fv ⊸ f𝜎 ] ⊸ [v ⊸ r]
This is the meaning constructor [nonsubsective-adj].
Its purpose is to assign a modified meaning to the
phrase, incorporating the meaning of the noun gun
and the meaning of obviously fake.
λx.obviously (fake (x, gun )) : [v ⊸ r]
By combining the meaning constructors in the previ-
ous two steps, we have produced the desired meaning
for obviously fake gun.

Again, the glue side of the meaning constructor for obviously fake gun is [v ⊸ r],
exactly the same as the unmodified noun gun, as with all cases of modification.

13.5 Adverbial modification

We now turn to an examination of the syntax and semantics of adverbial modifica-
tion. The treatment provided here is brief, and we concentrate primarily on aspects
of meaning composition; Butt et al. (1999: Chapter 7) provide more discussion of
the syntax of adverbial modifiers from an LFG perspective.
 adverbial modification 493

13.5.1 Adverbs at c-structure and f-structure

In English, adverbs such as clearly and skillfully are adjoined to the phrases they
modify. Like other modifiers, their f-structures appear as members of the set of
adj modifiers. In (61), the sentential adverb clearly is adjoined to IP:

(61) Clearly David fell.

IP

AdvP
IP  
↓ ∈ (↑ ’
 
Adv NP I  
 
Clearly N VP

David V

fell

A manner adverb like skillfully can be adjoined to VP, as in (62). Evidence that the
adverb skillfully is adjoined to VP in this example comes from the VP preposing
construction, discussed in §3.4 and illustrated in (63), where a VP appears in
fronted position. If the VP includes an adverb, the latter is also preposed, showing
that the adverb forms a constituent with the VP:

(62) David played skillfully.

IP

NP I
 
N VP ’
 
 
AdvP  
David VP
↓∈ (↑

V Adv

played skillfully

(63) David wants to play skillfully, and [play skillfully] he will.

494 modification

13.5.2 Adverbs and semantic composition

13.5.2.1 Adverb meaning The semantic contribution of adverbs has long been a
focus of generative linguistic research. Heny (1973) gives a cogent overview of the
state of research on adverb meaning in the early 1970s, when much research on
adverb meaning was done; though it was conducted on the basis of very different
syntactic assumptions, this work nevertheless forms the foundation upon which
much current work on the semantics of adverbs is based. We will examine two
semantically different kinds of adverbs, illustrated in the previous section by the
sentential adverb clearly and the manner adverb skillfully.
Within the LFG semantic tradition, Halvorsen (1983) discusses sentential
adverbs like clearly and necessarily and proposes to treat them in the standard
way, as proposition modifiers. The meaning of the sentence Clearly David fell is
therefore:

(64) Clearly David fell.

clearly (fall (David ))

The predicate clearly takes as its argument the proposition David fell, and the
meaning represented in (64) is roughly paraphrasable as It is clear that David fell.
Heny (1973), writing at the time at which Nixon was the president of the United
States, considers the following pair of sentences:

(65) a. The U.S. president is necessarily a citizen of the United States.

b. Nixon is necessarily a citizen of the United States.

As Heny notes, sentence (65a) is true under the rules of the constitution of the
United States, while sentence (65b) is not true by necessity. In other words, though
it turns out to be true that Nixon is a citizen of the United States, this is not
necessarily the case, since Nixon could have decided to become a citizen of another
country. On the other hand, it is necessarily the case that the U.S. president must be
a U.S. citizen under the laws of the United States. The sentences in (66), containing
the sentential adverb clearly, differ from one another in a similar way:

(66) a. Clearly the running back fell.

b. Clearly David fell.

Even in a situation where the running back fell and David is the running back, it
may not be clear that David fell, since the identity of the running back may not be
certain. Adverbs like clearly and necessarily induce opacity in the subject position
of the sentence they modify, since different ways of referring to the same individual
can affect the truth or falsity of the sentence (Quine 1953).
This aspect of the meaning of sentential adverbs is different from manner
adverbs. Intuitively, a manner adverb like skillfully modifies the action that is
performed, producing a new action that is performed skillfully:
 adverbial modification 495

(67) David played skillfully.

skillfully (David , λx.play (x))

In (67), skillfully is a two-place predicate: its arguments are the person that per-
formed the action (here, David) and the action that is performed skillfully (here,
playing). Unlike the situation with sentential adverbs, the following two sentences
are both true if David is the running back and he played skillfully. Manner adverbs
like skillfully do not induce opacity in subject position, so that if David is the
running back, the sentences in (68) are true in the same circumstances:

(68) a. David played skillfully.

b. The running back played skillfully.

In general, a manner adverb like skillfully takes two arguments, one corresponding
to the subject of the sentence and the other roughly corresponding to the English
verb phrase—the action that is performed. For this reason, such adverbs are
sometimes called VP or verb phrase adverbs, and as shown in (62), such adverbs
can be adjoined to VP in English. However, meaning combination with adverbs
like skillfully in fact depends on f-structural relations like subj, not c-structure
constituency relations. This is clearly shown by the existence of manner adverbs
in languages without a VP constituent; Broadwell (2005) provides a detailed
examination of the phrase structure of Zapotec, showing that Zapotec lacks a VP
constituent, and that manner adverbs appear adjoined to S:

(69) Dìáp g-ú’ld Màrìì.

strongly pot-sing Maria
‘Maria will sing strongly/loudly.’
S
↑ =↓

Adv S
↓ ∈ (↑ adj) ↑ =↓

Dìáp V NP
strongly ↑ =↓ (↑ subj) = ↓

g-ú’ld Màrìì
pot-sing Maria

13.5.2.2 Adverbs and meaning assembly We assume the syntactic and semantic
structures and meaning constructor in (70) for the sentence Clearly David fell.
The f-structure for clearly is labeled c, and its semantic structure c𝜎 contains the
attribute var whose value we have labeled cv:
496 modification

(70) Clearly David fell.

 
’
 
f d  c cv[ ]
 
c
clearly (fall (David )) : f

From now on, we will simplify our representations by displaying only semantic
structures whose internal structure is of interest in the constructions we are consid-
ering. Therefore, we do not display the semantic structures d𝜎 or f𝜎 corresponding
to the subject f-structure d and the sentence f-structure f .
We propose this lexical entry for the sentential adverb clearly:

(71) Lexical entry for clearly:

clearly Adv (↑ pred) = ‘clearly’
λP.clearly (P) : (↑𝜎 var) ⊸ ↑𝜎
λP.λQ.P(Q) :
[(↑𝜎 var) ⊸ ↑𝜎 ] ⊸ [(adj ∈ ↑)𝜎 ⊸ (adj ∈ ↑)𝜎 ]

As in the previous sections, the lexical entry in (71) uses inside-out functional
uncertainty to refer to the f-structure of the phrase it modifies. The expression
(adj ∈ ↑) refers to the f-structure modified by clearly, which in (70) is f . The
first meaning constructor in (71) contributes the intrinsic, modifiable meaning
of the adverb, and the second meaning constructor combines with the first one
to produce a meaning capable of modifying a proposition. We assume that both
of these meaning constructors are part of the lexical entry of the adverb, as in
our analysis of attributive adjectives. Unlike adjectives, there is no independent
predicative use of the intrinsic meaning of an adverb, as (72) illustrates, and so
there is no other version of the adverb lexical entry which contains only the
intrinsic meaning of the adverb and not the second meaning constructor:

(72) a. David is happy.

b.⃰David is clearly. / ⃰The situation is clearly.

The instantiated meaning constructors for the sentence Clearly David fell are given
in (73): the meaning constructors contributed by clearly are labeled [clearly1] and
[clearly2], and the meaning constructors [David] and [fall] follow the proposals
for proper names and intransitive verbs given in Chapter 8:
 adverbial modification 497

(73) Meaning constructor premises for Clearly David fell:

[David] David : d𝜎
[fall] λx.fall (x) : d𝜎 ⊸ f𝜎
[clearly1] λP.clearly (P) : cv ⊸ c𝜎
[clearly2] λP.λQ.P(Q) : [cv ⊸ c𝜎 ] ⊸ [f𝜎 ⊸ f𝜎 ]

Since the modifying adverb clearly is not itself modified, we first combine the two
meaning constructor premises [clearly1] and [clearly2] to obtain this meaning
constructor:

(74) [clearly] λQ.clearly (Q) : [f𝜎 ⊸ f𝜎 ]

As described in §8.5.2.1, we can combine the premises labeled [David] and [fall]
to obtain this meaning constructor:

(75) [David-fall] fall (David ) : f𝜎

Finally, we combine the meaning constructors [David-fall] and [clearly] to obtain

the desired result, that is, the meaning of the sentence is clearly (fall (David )):

(76) [David-fall], [clearly] ) clearly (fall (David )) : f𝜎

The derivation is semantically complete and coherent: we have obtained a well-

formed, nonimplicational meaning constructor for the sentence, with no premises
left unused.
The meaning deduction of a sentence with the manner adverb skillfully pro-
ceeds somewhat differently. The syntactic and semantic structures and meaning
constructor for the sentence David played skillfully are given in (77), where the
semantic structure s𝜎 corresponding to the adverb f-structure has the attribute var
with value sv and prop with value sp:

(77) David played skillfully.

 
’
  sv[ ]
p d  s
  sp[ ]
s
skillfully (David , λx.play (x)) : f

Again, we assume a bipartite semantic contribution for the adverb skillfully. The
lexical entry for skillfully is given in (78), and the instantiated meaning constructor
premises for this sentence are given in (79):
498 modification

(78) skillfully Adv (↑ pred) = ‘skillfully’

λQ.λy.skillfully (y, Q) :
[(↑𝜎 var) ⊸ (↑𝜎 prop)] ⊸ [(↑𝜎 var) ⊸ ↑𝜎 ]
λP.λQ.λx.P(Q)(x) :
[[(↑𝜎 var) ⊸ (↑𝜎 prop)] ⊸ [(↑𝜎 var) ⊸ ↑𝜎 ]] ⊸
[((adj ∈ ↑) subj)𝜎 ⊸ (adj ∈ ↑)𝜎 ] ⊸
[((adj ∈ ↑) subj)𝜎 ⊸ (adj ∈ ↑)𝜎 ]

(79) Meaning constructor premises for David played skillfully:

[David] David : d𝜎
[play] λx.play (x) : d𝜎 ⊸ p𝜎
[skillfully1] λQ.λy.skillfully (y, Q) : [sv ⊸ sp] ⊸ [sv ⊸ s𝜎 ]
[skillfully2] λP.λQ.λx.P(Q)(x) :
[[sv ⊸ sp] ⊸ [sv ⊸ s𝜎 ]] ⊸ [[d𝜎 ⊸ p𝜎 ] ⊸ [d𝜎 ⊸ p𝜎 ]]

We begin the derivation by combining the premises [skillfully1] and [skillfully2]

to obtain this meaning constructor:

(80) [skillfully] λQ.λy.skillfully (y, Q) : [d𝜎 ⊸ p𝜎 ] ⊸ [d𝜎 ⊸ p𝜎 ]

The right-hand side of the meaning contribution of the intransitive verb play,
d𝜎 ⊸ p𝜎 exactly matches the requirements of [skillfully]. We combine [skillfully]
and [play], and obtain this meaning constructor:

(81) [skillfully-play] λy.skillfully (y, λx.play (x)) : d𝜎 ⊸ p𝜎

Finally, we combine [skillfully-play] and [David] to obtain a wellformed, seman-

tically complete and coherent meaning constructor for the sentence:

(82) [skillfully-play], [David] ) skillfully (David , λx.play (x)) : p𝜎

13.6 Further reading and related issues

There has been much work on modification within LFG, particularly on the
syntax of modifiers and adjunction, that has not been discussed in this chapter. In
particular, Butt et al. (1999) discuss the syntax of adjectives and adverbs in English,
French, and German, and Colban (1987) provides a syntactic and semantic analysis
of prepositional phrases as verbal arguments and modifiers. Mittendorf and Sadler
(2008) discuss adjectival constructions in Welsh, and Al Sharifi and Sadler (2009)
discuss the adjectival construct in Arabic. Vincent and Börjars (2010a), Raza and
Ahmed (2011), and Lowe (2013, 2017b) discuss complementation within noun
and adjective phrases. Arnold and Sadler (2013) discuss displaced dependent
 further reading and related issues 499

constructions such as too complex for anyone to understand. A general overview

of the syntax and semantics of coordination is given in Chapter 16, but we leave an
analysis of coordinated modifiers (as in tall and thin man) for future work.
We have also omitted definition and discussion of the important issue of scoping
relations between modifiers. As noted by Andrews (1983b), Andrews and Manning
(1993), Pollard and Sag (1994), and many others, the contribution of modifiers
to the meaning of an utterance can depend on the order in which they appear.
Andrews (2018) discusses the contrast in (83), in which the modifiers unscrupulous
and former appear in different orders: (83a) refers to someone who is unscrupulous
and was a property developer in the past, while (83b) refers to someone who
was an unscrupulous property developer in the past but may currently be neither
unscrupulous nor a property developer:

(83) a. He is an unscrupulous former property developer.

b. He is a former unscrupulous property developer.

Various approaches to constraining modifier scope have been explored in an LFG

setting. Some approaches define syntactic scoping relations between modifiers in
terms of f-precedence (§6.11.4); on such approaches, semantic scope relations are
constrained by the syntactic scope relations determined by f-precedence. Andrews
(2018) develops an alternative approach which introduces set-based nesting rela-
tions into the f-structure, resulting in a close match between f-structure embedding
and the c-structure nesting relations that correlate with modifier scope. Meaning
assembly can then take advantage of these finer-grained relations at f-structure to
constrain modifier scope relations appropriately. Morrison (2017) builds on this
approach in an analysis of displaced adjectival focus, including constraints on the
position of relational adjectives.
 14
Anaphora

A variety of types of anaphora are attested crosslinguistically. Section 14.1 of this

chapter shows that incorporated pronominal elements behave differently from
elements that alternate with agreement markers, and these differ from morpholog-
ically independent pronouns in interesting ways. Anaphoric relations and binding
patterns have been fairly well studied in LFG; §14.2 discusses constraints on
anaphoric binding stated in terms of structural relations holding at f-structure,
and §14.3 discusses prominence relations which hold between the anaphor and
its potential antecedents stated at f-structure as well as other linguistic levels.
Our glue-theoretic treatment of the semantics of anaphoric binding is presented
in §14.4.1 This semantic treatment will be drawn upon in subsequent chapters,
particularly in our discussion of anaphoric control in Chapter 15.

14.1 Incorporated pronouns and “pro-drop”

As discussed in §5.4.3, a predicate may specify information about how its argu-
ments are interpreted when no overt argument phrases are present. In Chicheŵa,
for example, a verb like zi-ná-wá-lum-a ‘bite’ provides such information about its
subject and its object. The obj affix -wá- is unambiguously an incorporated obj
pronoun, so that a better gloss for this form might be ‘bite them’. This incorporated
pronominal obj may be anaphorically linked to a topic phrase, as in the English
example Those students, the bees bit them.
In contrast, the subj marker zi- behaves either as an agreement marker or as
an incorporated pronoun. In the presence of an overt subj phrase, zi- simply
marks agreement with the subject. When no overt subj phrase appears, zi- is an
incorporated pronoun like the obj marker. Since the subject marker may behave as
an incorporated pronoun, it can be anaphorically linked to a topic phrase, like the
incorporated obj pronoun -wá-; the important difference between the two markers
is that the subj marker has an alternate use as an agreement marker in addition to
its use as an incorporated pronominal.
As Bresnan and Mchombo (1987) show, there is a great deal of evidence for
the different status of the subj and obj markers in Chicheŵa. For example, since
the subject marker can simply mark agreement, it can appear with an idiomatic
subject. In contrast, the object marker is always an incorporated pronoun and not
an object agreement marker, and it cannot appear with an idiomatic object, since
an idiomatic object cannot be interpreted as an information structure topic and

1 The analyses presented in §§14.3.4 and 14.4 are the result of joint work with Dag Haug.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 incorporated pronouns and “pro-drop” 501

cannot bear an anaphoric relation to an incorporated pronominal object. Further,

the subject can be questioned when the subject marker is present, but the object
cannot be questioned when the object marker is present: since the question word
bears the information structure focus function, it is compatible with the subject
agreement marker, but not with the incorporated pronominal obj, which must bear
an anaphoric relation to an information structure topic and not a focus. Bresnan
and Mchombo (1987) enumerate additional ways in which the subj and obj
markers behave differently, showing that all of these differences can be explained
on the basis of the different status of the two markers.
The syntactic difference between the subj and obj affixes is formally reflected in
the following lexical entry:

(1) Lexical entry for the Chicheŵa verb zi-ná-wá-lum-a:

zináwáluma V (↑ pred) = ‘bite"subj,obj#’
((↑ subj pred) = ‘pro’)
(↑ subj nounclass) = 10
(↑ obj pred) = ‘pro’
(↑ obj nounclass) = 2

In this lexical entry, the pred value of the obj of zi-ná-wá-lum-a ‘bite’ is
unambiguously specified by the equation (↑ obj pred) = ‘pro’: the object of this
verb is pronominal. In contrast, the pred value of the subj is optionally specified,
as indicated by the parentheses. In the presence of an overt subj phrase, the subject
marker specifies only agreement information (here, information about noun class),
and the pred value of the subj is provided by the overt subject phrase, as shown
in (2):

(2) njûchi zi-ná-wá-lum-a

bees subj-pst-obj-bite-ind
‘The bees bit them.’
 
’
 
S  
 
 
 
NP VP  
 
(↑ ↓ ↑=↓  

N V
↑=↓ ↑=↓

zi-ná-wá-lum-a
njûchi
bees (↑ ’
(↑ ((↑
(↑ (↑
(↑
(↑
502 anaphora

In contrast, (3) shows that when there is no subject phrase, the specifications
associated with the verb provide the pred value for the subj:

(3) zi-ná-wá-lum-a
subj-pst-obj-bite-ind
‘They bit them.’
S  
’
VP  
 
↑=↓  
 
 
 
V  
 
↑=↓

zi-ná-wá-lum-a

(↑ ’
((↑
(↑
(↑
(↑

As discussed in §5.4, the typology of agreement and pronominal incorporation that

is reflected in these different specifications is richer than is assumed in some other
theories. In her analysis of nonconfigurationality, Jelinek (1984) proposes that all
nonconfigurational languages should be analyzed as pronominal-incorporating,
as we have analyzed the Chicheŵa incorporated object pronoun. Bresnan and
Mchombo (1987) conclusively demonstrate that this inflexible approach is incor-
rect and that a wider range of distinctions is necessary. The object marking on the
Chicheŵa verb must be analyzed as an incorporated pronoun with an obligatory
pronominal pred value supplied by the verb, while the subject marking represents
an optional pronominal pred, behaving as an agreement marker in the presence of
an overt subject phrase. Further evidence for the richer typology assumed in LFG
is provided by Austin and Bresnan (1996) in their analysis of Warlpiri, Toivonen
(2001) in her analysis of Finnish, Butt (2007) in her analysis of Punjabi, and
Coppock and Wechsler (2012) in their analysis of Hungarian.
Bresnan et al. (2016) provide much more discussion of the typology of pronom-
inal elements in LFG, including a detailed discussion of the differences between
overt pronouns and “null” pronominals like the Chicheŵa incorporated obj pro-
noun. In their Chapter 8, Bresnan et al. (2016) explore the role of null pronominals
in introducing and referring to sentential and discourse topics, and in focus
constructions.

14.2 Binding relations

Anaphoric binding relations are semantic in nature, having to do with coreference
between a pronoun and its antecedent: this is the topic of §14.4. Nonsemantic levels
 binding relations 503

of linguistic structure also play a role in anaphoric binding, as we discuss in §14.3,

since they are often important in constraining possible binding relations. In this
section, we discuss constraints on the binding relation which are defined in terms
of f-structure domains and relations. We assume the phrase structure rules and
lexical entries that we have discussed so far, and we display only the f-structures
for the examples under discussion.
Within LFG, Bresnan et al. (1985a) were the first to propose that a theory of
syntactic constraints on binding relations can be stated in terms of f-structural
properties such as coargumenthood or the presence of a subj function. Continuing
this work, Dalrymple (1993) proposed a universally available and lexically speci-
fied inventory of binding constraints, and also provided a formal specification of
these constraints. This work has been extended by many others, including Strand
(1992) in work on Norwegian anaphora, Sung (1996) in work on Korean, Lapata
(1998) in work on Modern Greek, Henadeerage (1998) in work on Sinhala, Lødrup
(2008a) in work on Norwegian, Strahan (2009, 2011) in work on Icelandic and
other Scandinavian languages, Rákosi (2009, 2010) in work on Hungarian, Snijders
(2014) in work on Warlpiri, and Dalrymple (2015) in work on Ya̧g Dii. Bresnan
et al. (2016: Chapters 10–11) provide a detailed discussion of binding constraints,
extending the theory to cover coreference relations between nonpronominal ele-
ments as well. In the following, we give a brief overview of the theory.

14.2.1 Positive binding constraints

Some anaphoric elements, such as the English reflexive pronoun himself, must
appear in a particular syntactic relation to their antecedent. We say that elements
like himself obey positive constraints, that is, constraints which state the syntactic
relation that an anaphor must bear to its antecedent.
Example (4) is potentially ambiguous since the antecedent of himself can be
either the subj David or the obj Chris. Here we focus on the reading on which
the antecedent of himself is David, as indicated by the subscript i annotation
coindexing David and himself :

(4) Davidi compared Chris to himselfi .

pred compare subj,obj,obl ’
subj ‘David’
obj ‘Chris’

c pro

obl

In the f-structure in (4), the antecedent David of the reflexive pronoun himself
is the subj of the f-structure labeled c, and the reflexive pronoun is the oblto
of the same f-structure. The semantic antecedency relation, which establishes
504 anaphora

coreference between the anaphor and its antecedent, is defined and discussed
in §14.4.3; here we concentrate on syntactic factors that constrain the anaphor-
antecedent relation.
The antecedent of himself need not be an argument of the same predicate. In
(5), David is an acceptable antecedent of himself, even though himself is the object
of the preposition around, and David is the subject of the verb wrapped:

(5) Davidi wrapped the blanket around himselfi .

 
’
 
 
 
 
 
 
 + 
 
  
 ’ 
 
   
  
  
    
      
    
   
    
 
   
 

However, it is not possible for himself to appear in a sentence with no antecedent,

or with an antecedent in a syntactically unacceptable relation to it. For example,
himself may not be separated from its antecedent by a finite clause boundary:

(6) ⃰Davidi thought that Susan had seen himselfi .

 
’
 
 
  
 

 ’ 
  
  
  
    
  
  
    
    
      
    

   

    
  


As demonstrated by Bresnan et al. (1985a), the English reflexive pronoun himself

obeys the following positive constraint, governing the syntactic relation between
himself and its antecedent:

(7) The antecedent of the English reflexive pronoun himself must appear in the
Minimal Complete Nucleus containing the pronoun.
 binding relations 505

The Minimal Complete Nucleus is defined by reference to the presence of a subj

function (Bresnan et al. 1985a; Dalrymple 1993; Bresnan et al. 2016):

(8) Minimal Complete Nucleus containing an f-structure p:

The smallest f-structure that contains p and a subj function.

According to this definition, the antecedent of the anaphor himself must appear
in the smallest f-structure that contains both the anaphor and a subj. We call the
domain in which the antecedent of the anaphor must appear the binding domain
of the anaphor, and we say that the binding domain of himself is the Minimal
Complete Nucleus.
Languages with multiple anaphors provide evidence for expanding the range
of constraints that anaphors can obey and also demonstrate that constraints
on anaphoric binding must be specified lexically, not universally or on a per-
language basis: different anaphoric elements in the same language may obey
different anaphoric binding constraints. The pronominal system of Norwegian is
particularly rich.2
Hellan (1988) shows that although both the Norwegian reflexive anaphor seg
selv and the reciprocal hverandre must be locally bound, the binding domain for
the reciprocal hverandre is larger than the domain for seg selv. The reflexive seg selv
must be bound to a coargument, an argument governed by the same pred as the
reflexive. In contrast, the reciprocal hverandre, like the English reflexive pronoun
himself, must be bound in the Minimal Complete Nucleus.
In (9), the verb fortalte ‘told’ has three arguments: subj, obj, and an oblabout
argument marked by the preposition om. The antecedent of the oblabout seg selv
is the subj, Jon (Hellan 1988: 67):

(9) Jon fortalte meg om seg selv.

Jon told me about self
‘Joni told me about selfi .’
 
’
 
 
 
 
  
 
 
  
  
 
 
 
   
 
 
   
   
 

2 Hellan (1988), Strand (1992), and Dalrymple (1993) provide further discussion of binding con-
straints on Norwegian anaphors. In a very interesting series of articles, Lødrup (1999a, 2007, 2008a)
challenges some of Hellan’s generalizations and provides alternative analyses of binding constraints for
the anaphoric elements in Norwegian, while maintaining the generalization that binding constraints
can differ for different anaphoric elements.
506 anaphora

In (9), the antecedent of seg selv is a coargument, as required: both seg selv and Jon
are arguments of the verb fortalte ‘tell’. In contrast, (10) is ungrammatical:

(10) ⃰Hun kastet meg fra seg selv.

she threw me from self
‘Shei threw me away from selfi .’
 
’
   
 
   
   
 
 
 
 
   
 
   
 
   
 
 
 
   
  ’ 
  
 

   
 

  
   
 
    
 
 
  

 


In (10), the verb kastet ‘threw’ takes only two arguments, subj and obj; the phrase
fra seg selv is an adjunct. The reflexive pronoun seg selv is the obj of the preposition
fra ‘from’, and the intended antecedent is the subj of the verb kastet ‘throw’. The
reflexive and its intended antecedent are not coarguments of the same pred, and
the sentence is unacceptable.
Example (11) contains the Norwegian reciprocal pronoun hverandre, whose
antecedent is the coargument subj de ‘they’ (Hellan 1988: 67). The f-structure for
(11) is roughly the same as the one displayed in (9) for the reflexive seg selv:

(11) De fortalte meg om hverandre.

they told me about each.other
‘Theyi told me about each otheri .’

The antecedent of the reciprocal hverandre is not required to be a coargument.

The structure of (12) is similar to (10), but (12) is completely acceptable with the
reciprocal hverandre, since the reciprocal appears in the Minimal Complete
Nucleus with respect to its antecedent, as required (Hellan 1988: 69).

(12) De kastet meg til og fra hverandre.

they threw me to and from each.other
‘Theyi threw me to and from each otheri .’
 binding relations 507

These examples show that different anaphors in the same language may be required
to obey different binding constraints. Thus, these constraints must be lexically
specified for each anaphoric element.
Crosslinguistic examination of anaphoric binding patterns reveals four domains
relevant for anaphoric binding, defined by the f-structure properties pred, subj,
and tense, and by the entire sentence:

(13) Coargument Domain: minimal domain defined by a pred and

the grammatical functions it governs
Minimal Complete Nucleus: minimal domain with a subj function
Minimal Finite Domain: minimal domain with a tense attribute
Root Domain: f-structure of the entire sentence

Interestingly, all of these domains denote some syntactically or semantically com-

plete entity: the Coargument Domain corresponds to a syntactically saturated
argument structure, the Minimal Complete Nucleus corresponds to a predication
involving some property and a subject, the Minimal Finite Domain represents an
event that has been spatiotemporally anchored, and the Root Domain represents
a complete sentence. The binding conditions defined in (13) are illustrated by the
binding requirements for the Norwegian anaphors seg selv, sin, and seg (Hellan
1988) and the Chinese anaphor ziji (Tang 1989); see Dalrymple (1993) for more
discussion:

(14) Positive binding domains:

Coargument Minimal complete Minimal finite Root

domain nucleus domain domain
seg selv sin seg ziji

These binding requirements are specified as part of the lexical entry of each
anaphoric element (Dalrymple 1993). They limit the possibilities for pronoun
antecedency, which, as we show in §§14.3.2 and 14.3.3, may be further constrained
by other levels of structure.
Formally, we can define the syntactic domain in which an anaphor must find its
antecedent by means of expressions involving inside-out functional uncertainty. As
discussed in §6.1.3, inside-out functional uncertainty allows reference to enclosing
structures, those in which a particular f-structure is contained. In the case at hand,
we can use inside-out functional uncertainty to define the binding domain of an
anaphor, that is, the f-structure domain within which the anaphor is required to
find its antecedent.
Assuming that the f-structure for the pronoun is p, we can define each of the
anaphoric binding domains as follows:
508 anaphora

(15) Coargument Domain: ( gf⃰ gfpro p)

¬(→ pred)

Minimal Complete Nucleus: ( gf⃰ gfpro p)

¬(→ subj)

Minimal Finite Domain: ( gf⃰ gfpro p)

¬(→ tense)

Root Domain: (gf⃰ gfpro p)

These expressions constrain the f-structure domain within which the antecedent
of the anaphor can appear. Recall that gf is an abbreviation for any grammatical
function (§6.1.2). In the expressions in (15), the path leading to the anaphor
through the binding domain is as follows:

(16) Path defining the binding domain:

gf⃰ gfpro

For clarity, we use the abbreviation gfpro for the grammatical function borne by
the anaphoric pronoun, which can be any grammatical function. Constraints on
the domain within which the anaphor must be bound are stated by means of
off-path constraints on the path to the anaphor.3 In the case of the Coargument
Domain definition, for example, the path leading through the binding domain to
the anaphor is as follows:

(17) Coargument Domain path:

gf⃰ gfpro
¬(→ pred)

This expression is defined in terms of a series of attributes gf, each of which must
obey the off-path constraint ¬(→ pred). In this expression, the symbol → refers to
the f-structure value of the attribute gf, and the expression ¬(→ pred) represents
a negative existential constraint preventing that f-structure from containing the
attribute pred. The effect of this constraint is that the path through the binding
domain leading to the anaphor may not pass through an f-structure containing
the attribute pred. If the path were to pass through such an f-structure, the binding
domain would incorrectly extend beyond the Coargument Domain.
The other binding domains are similarly defined by off-path constraints that pre-
vent the path from passing through an f-structure of a certain type. For anaphors
subject to the Minimal Complete Nucleus constraint, the path from the pronoun
to its antecedent may not pass through an f-structure with a subj attribute, and for
the Minimal Finite Domain constraint, an f-structure with a tense attribute may
not be crossed. For the Root Domain, the path is unconstrained and may be of
arbitrary length. In this way, anaphors that obey positive constraints are required
to find an antecedent within a certain f-structural domain.

3 Off-path constraints are defined and discussed in §6.6.

 binding relations 509

Although the antecedent of the anaphor must appear within its binding domain,
antecedents that are too deeply embedded within the binding domain are not
acceptable:

(18) ⃰Davidi ’s mother nominated himselfi .

 
’
 
 
 
 
 
 
   
 
 
   
  
    
   
  
    
 

As noted in §6.9.1, the antecedent of an anaphor is generally required to f-command

the anaphor.4 In (18), the intended antecedent David of the reflexive himself
appears in the proper binding domain, the Minimal Complete Nucleus; however,
the f-command condition does not hold, and the sentence is ill-formed.
The syntactic relation between the anaphor and its antecedent is given by a
constraint of the following form, where gfante is the grammatical function of the
antecedent:

(19) F-commanding f-structures:

((gf⃰ gfpro p) gfante )

In this expression, as above, p is the f-structure for the anaphor, and (gf⃰ gfpro p)
defines the binding domain containing the anaphor and its antecedent. The
expression in (19) refers to some f-structure within the binding domain which
bears the unspecified grammatical function gfante . The f-command requirement
follows from the form of this expression, since the expression in (19) picks out all
and only the f-structures that f-command the anaphor within the binding domain.
In the expression in (19), the grammatical function of the antecedent gfante
is unconstrained, and any grammatical function may be chosen. In some cases,
however, the grammatical function of the antecedent may also be constrained in
that the antecedent of the anaphor may be required to bear either the function
subj or the function poss within the binding domain. Hellan (1988) shows that
the antecedent of the Norwegian possessive reflexive sin must be a subject:

4 See Culy (1991) for discussion of the Fula pronominal ɗum, which is unusual in not obeying a
command condition.
510 anaphora

(20) a. Jon ble arrestert i sin kjøkkenhave.

Jon was arrested in self ’s kitchen.garden
‘Joni was arrested in hisi kitchen garden.’
b. ⃰Vi arresterte Jon i sin kjøkkenhave.
we arrested Jon in self ’s kitchen.garden
‘We arrested Joni in hisi kitchen garden.’

In contrast, nonsubjects are acceptable antecedents for the Norwegian possessive

pronoun hans:

(21) Vi fant Jon under sengen hans.

we found Jon under bed his
‘We found Joni under hisi bed.’

Thus, it is necessary in some cases, including Norwegian sin, to constrain the

grammatical function of the antecedent. The following expression picks out all
subjects within the Minimal Finite Domain relative to the pronoun f-structure p:

(22) F-commanding subjects in the Minimal Finite Domain:

(( gf⃰ gfpro p) subj)
¬(→ tense)

F-structures with the subj function in other domains are picked out similarly; see
Dalrymple (1993) and Bresnan et al. (2016) for details.
Other conditions on the antecedent must also be met in some cases. Culy
(1996) discusses the pronominal systems of several varieties of Fula in which
certain pronouns place very specific syntactic requirements on their antecedents:
for instance, that the antecedent must be a pronoun. Culy analyzes this as a type of
agreement between the pronoun and its antecedent.

14.2.2 Negative binding constraints

Just as some anaphoric elements require their antecedent to appear within some
syntactic domain, some elements can require noncoreference with every element
within some domain. We call such noncoreference constraints negative constraints.
For instance, the pronoun him may not corefer with its coargument Chris in this
example:

(23) ⃰Chrisi nominated himi .

 
’
 
 
  
 
 
   
  
  
   

 
 binding relations 511

The constraint obeyed by him can be stated in the following way:

(24) The antecedent of the English pronoun him must not appear in the Coargu-
ment Domain containing the pronoun.

We formalize this in §14.4.3.

Different anaphoric elements in the same language may obey different neg-
ative constraints. Thus, as with positive constraints, negative constraints must
be lexically associated with the relevant anaphoric elements. Interestingly, the
same domains that are relevant for defining positive constraints, described in
§14.2.1, are also relevant for negative constraints, as Dalrymple (1993) shows for
the Norwegian pronouns ham selv and seg (Hellan 1988), the Hindi possessive
pronoun uskaa (Mohanan 1994), and the Yoruba pronoun ó (Pulleyblank 1986):

(25) Negative binding domains:

Coargument Minimal complete Minimal finite Root

domain nucleus domain domain
seg ham selv uskaa ó

In the case of positive constraints, the anaphor is required to corefer with some
element picked out by the constraint; in the case of negative constraints, the
anaphor must be noncoreferent with all elements picked out by the constraint.

14.2.3 Positive and negative binding constraints

Since positive and negative constraints are lexically associated with individual
anaphors, we would expect to find anaphors that simultaneously obey both kinds of
constraints. The Norwegian anaphoric element ham selv exemplifies this situation
(Hellan 1988; Dalrymple 1993): ham selv must be bound to an argument in the
Minimal Complete Nucleus (a positive constraint), but it is also required to be
noncoreferent with a coargument subj (a negative constraint).
In (26a), the obj Jon and the oblique phrase ham selv are coarguments. There-
fore, ham selv is coreferent with a coargument in (26a); (26b) shows that corefer-
ence with noncoarguments is also permitted:

(26) a. Vi fortalte Jon om ham selv.

we told Jon about self
‘We told Joni about selfi .’
b. Jeg ga Jon en bok om ham selv.
I gave Jon a book about self
‘I gave Joni a book about selfi .’

The antecedent of ham selv must appear in the Minimal Complete Nucleus
containing it. This accounts for the unacceptability of (27), since the intended
antecedent Jon does not appear in the Minimal Complete Nucleus relative to
ham selv:
512 anaphora

(27) ⃰Jeg lovet Jon [å snakke om ham selv].

I promised Jon to talk about self
‘I promised Joni to talk about selfi .’

However, ham selv may not corefer with a coargument subject, even one that is in
its Minimal Complete Nucleus:

(28) ⃰Jon snakker om ham selv.

Jon talks about self
‘Joni talks about selfi .’

Thus, the Norwegian pronoun ham selv obeys two binding conditions, one nega-
tive and one positive: it must be noncoreferent with a subj coargument, and it must
be coreferent with an argument in the Minimal Complete Nucleus. Both of these
requirements must be satisfied. This is accomplished by including both a positive
constraint and a negative constraint in the lexical entry for ham selv.

14.3 Binding and prominence

As discussed by Dalrymple (1993:5) and Bresnan et al. (2016), not all f-commanding
elements in the binding domain are relevant for the application of binding
constraints, but only those which are more prominent than the anaphor. In this
section, we discuss prominence relations defined in terms of the grammatical
function hierarchy (§14.3.1), f-precedence (§14.3.2), and argument structure
(§14.3.3). Bresnan et al. (2016) show that different anaphors impose different
combinations of prominence constraints in the statement of binding constraints.
Evidence of binding constraints defined at multiple levels of structure has also
been explored by Dalrymple (1993) for English and Norwegian and by Arka
and Wechsler (1996) for Balinese. Arka and Wechsler show that constraints on
Balinese binding relations depend on linear order, thematic prominence, and the
term/oblique distinction. We provide a formal statement of prominence conditions
in §14.3.4.

14.3.1 Prominence and the grammatical function hierarchy

As we have seen, the English pronoun him obeys the negative Coargument Condi-
tion, and may not corefer with a coargument. However, (29) is wellformed despite
him being coreferent with a coargument, himself:

(29) I compared himi with himselfi .

Similarly, in (30) the pronoun him is the object of believe and also the subject of like,
and himself is the object of like. This means that him and himself are coarguments
of like, but this does not lead to ungrammaticality:
 binding and prominence 513

(30) I believed himi to like himselfi .

As these examples show, coreference with certain coarguments does not violate
the negative binding constraints which him obeys. In contrast, him in (31a) cannot
corefer with the object of compare, and him as the object of like cannot corefer with
the subject of like, David (31b):

(31) a. ⃰I compared Davidi with himi .

b. ⃰I believed Davidi to like himi .

Only arguments which are superior on the grammatical function hierarchy are
relevant for the binding constraints associated with him. Bresnan et al. (2016:
230) refer to this requirement as the syntactic rank condition, and require binding
constraints to be defined in terms of syntactic rank:

(32) Syntactic rank: A locally outranks B if A and B belong to the same f-

structure and A is more prominent than B on the grammatical function
hierarchy. A outranks B if A locally outranks some C which contains B.

14.3.2 Prominence and f-precedence

In addition to constraints on anaphoric binding defined purely in terms of
f-structure properties, some pronouns obey constraints that are defined in terms
of f-precedence relations holding between the anaphor and its antecedent. For
example, Mohanan (1983) shows that overt pronouns in Malayalam cannot
precede their antecedents:

(33) [kut..tiyute ammaye] awan nul..li

child.gen mother.acc he ¯pinched
‘Hei pinched the childi ’s mother.’

(34) ⃰[awante ammaye] kut..ti nul..li

his mother.acc child ¯pinched
‘The childi pinched hisi mother.’

In (35), an overt pronoun is not acceptable if coreference with the matrix subject
is intended:

(35) [awan aanaye nul..liyatinə s̄ees.am] kut..ti uraŋŋi

he elephant.acc ¯pinched ¯ after child slept
‘The childi slept, after he⃰i,j pinched the elephant.’

However, pronominals that are not overtly realized at c-structure do not obey such
ordering restrictions. The subordinate clause in (36) contains a null pronoun that
may corefer with the matrix subject kut..ti ‘child’:
514 anaphora

(36) [∅ aanaye nul..liyatinə s̄ees.am] kut..ti uraŋŋi

elephant.acc ¯pinched ¯ after child slept
‘The childi slept, after hei,j pinched the elephant.’

Kameyama (1985) examines similar data from Japanese, discussed in §6.11.4, and
proposes the following generalization, valid for pronouns in Japanese, Malayalam,
and many other languages:

(37) The antecedent of a pronoun must f-precede the pronoun.

As noted in §6.11.4, constraining binding relations by f-precedence makes exactly

the right predictions concerning overt and null elements.
Intuitively, an f-structure f f-precedes an f-structure g if the c-structure nodes
corresponding to f precede the c-structure nodes corresponding to g. The formal
definition of f-precedence provided by Kaplan and Zaenen (1989b) is given in
(184) of §6.11.4, and repeated here:

(38) F-precedence:
f f-precedes g ( f <f g) if and only if for all n1 ∈ 𝜙−1 (f ) and for all n2 ∈
𝜙−1 (g), n1 c-precedes n2 .

This definition, together with the generalization in (37), predicts the patterns
of acceptability for the Japanese and Malayalam examples examined above: null
pronouns f-precede and are f-preceded by every element in the sentence, so
no matter where the antecedent of a null pronoun appears in the sentence, the
condition in (37) is satisfied. In contrast, overt pronouns are not permitted to
f-precede their antecedents, accounting for the unacceptability of coreference in
examples (34) and (35).
Bresnan (1995, 1998) and Bresnan et al. (2016) also discuss linear precedence
conditions on anaphoric binding with particular attention to weak crossover vio-
lations in extraction, providing a different definition of f-precedence. We discuss
these proposals in §17.4.2.

14.3.3 Prominence and argument structure

As discussed in §2.1.3, binding may also be constrained by argument structure
relations: the antecedent of an anaphor may be required to outrank the anaphor or
a phrase containing it on the thematic hierarchy. The thematic hierarchy presented
in (25) of §9.4.2 is repeated here:

(39) Thematic hierarchy:

agent > benefactive > recipient/experiencer
> instrument > theme/patient > locative
 binding and prominence 515

Sells (1988) shows that reference to the thematic hierarchy is necessary in an

account of binding conditions in Albanian. A term argument in Albanian can
antecede a term or oblique reflexive, while an oblique can only antecede another
oblique. Among the term arguments, binding relations are constrained by the the-
matic hierarchy: if the antecedent and the anaphor are both terms, the antecedent
must be higher on the thematic hierarchy than the anaphor.
Hellan (1988), Dalrymple and Zaenen (1991), and Dalrymple (1993) discuss
Norwegian data that point to a similar conclusion. Hellan (1988) shows that some
Norwegian verbs have two possibilities for passivization:

(40) a. Vi overlot Jon pengene.

we gave Jon money
goal theme
‘We gave Jon the money.’
b. Jon ble overlatt pengene.
Jon was given money
goal theme
‘Jon was given the money.’
c. Pengene ble overlatt Jon.
money was given Jon
theme goal
‘The money was given to Jon.’

However, when the object contains a possessive reflexive whose antecedent is the
subject of the passive verb, only one reading is possible (Hellan 1988: 162):

(41) Barnet ble fratatt sine foreldre.

child was taken self ’s parents
‘The childi was deprived of self ’si parents.’
malefactive theme
NOT: ‘The childi was taken away from self ’si parents.’
theme malefactive

We might expect that either argument in (41) could be interpreted as the theme
and either as the malefactive, in the same way that there is flexibility in (40).
In fact, however, the only possible construal of this sentence is one where the
subject barnet is the malefactive argument, and the object sine foreldre is the theme.
Assuming that the goal and malefactive arguments occupy the same position on
the thematic hierarchy, the malefactive argument outranks the theme argument.
In the permissible reading of (41), then, the antecedent barnet outranks the phrase
containing the pronoun sine foreldre on the thematic hierarchy. The other reading,
where the phrase containing the pronoun thematically outranks the antecedent, is
not available.
516 anaphora

14.3.4 Prominence constraints

Formally, we impose prominence conditions by constraining the relation between
the potential antecedent and the f-structure containing the anaphor which appears
in the same f-structure as the antecedent, just as in the syntactic rank condition
defined in (32). Recall that the f-structures of the potential antecedents of an
anaphor are those which f-command the anaphor, as shown in (19) and repeated
here:

(42) F-commanding f-structures:

((gf⃰ gfpro p) gfante )

We can restrict application of binding constraints to f-structures which are more

prominent than the anaphor by imposing constraints on gfante of the form in
(43), using the name prominent as a cover term for prominence relations defined
in terms of the grammatical function hierarchy, f-precedence, or the thematic
hierarchy. As a part of the expression in (42), this constraint is lexically associated
with the anaphor:

(43) Off-path constraints on gfante encoding the prominence condition (gfante is

as shown in 42):
gfante
%coarg = (← gf)
(%coarg gf⃰) = ↑
prominent(→, %coarg)

In (42) and (43), gfante is the grammatical function whose value is the potential
antecedent of the anaphor, labeled f in (44) and referred to as → in (43). The
f-structure labeled c in (44) is a coargument of gfante , and is given the name
%coarg in the first constraint in (43). According to the second constraint, there is
a path5 gf⃰ through c leading to the f-structure p of the anaphor; in other words, c is
the coargument of f which contains the anaphor. We can now require the potential
antecedent f to be more prominent than its coargument c by specifying a particular
definition for the relation prominent:
. /
(44) gfante f[ ]
gf c[ ... p[ ]]

For example, we can impose a prominence constraint based on the grammat-

ical function hierarchy by replacing prominent in (43) with the relation gf-
outranks, which holds between arguments of the same predicate:

5 If the anaphor and its potential antecedent are coarguments, the path gf⃰ is empty, and c and p are
the same f-structure.
 anaphora and semantic composition 517

(45) gf-outranks(f1 , f2 ) if and only if f1 outranks f2 on the grammatical function

hierarchy.

We can then define the template gf-prominent as encoding exactly this set of
constraints:

(46) Definition of the template gf-prominent:

gf-prominent ≡ %coarg = (← gf)
(%coarg gf⃰) = ↑
gf-outranks(→, %coarg)

And we can state the binding conditions on him by means of the expression in (47),
which uses this template:

(47) Prominent elements in the coargument domain:

(( gf⃰ gfpro ↑ ) gfante )
¬(→ pred) @gf-prominent

The expression in (47) ranges over elements in the coargument domain which
are higher on the grammatical function hierarchy than the f-structure of the
anaphor: these are the f-structures which are relevant for the application of binding
constraints. Argument structure and f-precedence prominence constraints are
encoded similarly, by altering the definition of prominent in (43).

14.4 Anaphora and semantic composition

We turn now to the central issue in the interpretation of anaphora: how the seman-
tic relation is established between an anaphor and its antecedent. Within the glue
semantics approach, several proposals have been made for the semantic treatment
of anaphoric binding. The first proposal was made by Dalrymple et al. (1997),
whose approach correctly handles the interactions of anaphora and quantification
within the sentence; Asudeh (2004, 2012) adopts this approach, extending it to
the treatment of resumptive pronouns (see §17.6.4). However, that proposal only
peripherally addresses issues that arise in analyzing intersentential anaphora—the
interpretation of anaphors whose antecedents are not in the same sentence.
Subsequently, Crouch and van Genabith (1999) observed that the glue approach
can also be harnessed to handle the context-changing potential of sentences. They
propose that the linear logic glue language not only manages the dynamics of
meaning composition within a sentence, but also manages context resources and
context update. In their analysis, Crouch and van Genabith adopt an e-type treat-
ment of anaphoric binding (Evans 1980), where descriptions of entities relevant
in the discourse are constructed in the course of the derivation of the meaning of
an utterance and are used in anaphora resolution both within the sentence and in
subsequent discourse. Dalrymple (2001) follows Crouch and van Genabith (1999)
518 anaphora

in assuming that the glue approach not only accounts for resource-sensitive aspects
of meaning assembly within sentences, but also manages contextual contributions
in discourse interpretation. In contrast to their approach, however, Dalrymple
(2001) provides a theory of coreference and anaphoric binding that is closer to
Discourse Representation Theory (DRT: Kamp 1981; Kamp and Reyle 1993) and
Dynamic Predicate Logic (Groenendijk and Stokhof 1991).
An alternative approach to using the glue language itself to manage contextual
contributions across discourse is to combine glue with a meaning language which
is more dynamic than standard predicate logic, and thus better able to handle the
variability in coreference involved in anaphora. Kokkonidis (2005) proposes to
pair glue expressions with meanings couched in a compositional version of DRT
which is capable of dealing with anaphoric resolution. Haug (2014b) discusses
anaphoric resolution in compositional DRT in detail, and develops a Partial version
of Compositional DRT (PCDRT) which provides a clear separation of the semantic
and pragmatic aspects of anaphora. Haug’s proposals are further developed within
the glue approach by Haug (2013), Belyaev and Haug (2014), and Dalrymple et al.
(2018). This is the approach that we adopt and build on in this book.

14.4.1 Anaphora in context

It has long been clear that phrases like David, someone, or a man introduce new
individuals into the context, and that these individuals can be referred to in later
discourse. Karttunen (1976) was among the first to propose a formal theory of
the ability of noun phrases to introduce a discourse referent representing some
individual into the discourse context; discourse referents persist across the dis-
course and allow for the pragmatic resolution of anaphors in subsequent sentences
to a discourse referent introduced at an earlier point. Karttunen’s basic idea was
elaborated and refined in the work of Heim (1982) on File Change Semantics and
by Kamp (1981) in his work on Discourse Representation Theory (DRT). In both
of these theories, certain kinds of phrases introduce discourse referents into the
discourse context. Pronouns in subsequent discourse can take these phrases as
antecedents and are resolved to the discourse referents that correspond to them.
For instance, part of the semantic contribution of sentences like David arrived,
I met David yesterday, or I am going to give the book to David is to introduce a
new discourse referent representing the individual David into the context. In the
Discourse Representation Structure (DRS) for the sentence in (48), the proper
name David introduces the discourse referent x1 :

(48) David arrived.

x1
David (x1 )
arrive (x1 )

In a DRS, discourse referents are listed in the “universe,” the upper portion of the
“box”; the discourse conditions, that is the truth-conditional requirements placed
 anaphora and semantic composition 519

on the discourse referents, are represented in the lower portion of the box. In (48),
there is only one discourse referent in the universe: x1 , introduced by David.
In our analysis, we assume the Partial Compositional DRT (PCDRT) approach
of Haug (2014b), which differs from earlier versions of DRT in its treatment of
anaphora. The standard DRT assumption is that a sentence containing a pronoun
cannot be interpreted until the pronoun is resolved. This is problematic, as Haug
points out: if a sentence with an unresolved pronoun is uninterpretable, the content
of the sentence cannot play a role in determining how the pronoun is resolved.
On the PCDRT view, anaphoric resolution is assumed to be part of the non-
monotonic, pragmatic content of discourse. This also allows for the fact that as
the hearer acquires additional information, it may become clear that the initial
resolution for the pronoun was incorrect and must be revised.
In PCDRT, as in standard DRT, non-anaphoric phrases introduce discourse
referents into the discourse context, as shown in (48).6 Pronouns also introduce
discourse referents, but unlike standard DRT, these are not directly resolved to
their antecedents in the truth-conditional meaning representation. In the PCDRT-
based DRS for the two sentences in (49), both the proper name and the pronoun
introduce discourse referents, x1 and x2 respectively:

(49) David arrived. He yawned.

x1 x2
David (x1 )
arrive (x1 ) , A(x2 ) = x1
∂(male (x2 ))
yawn (x2 )

In (49), there are two discourse referents in the universe: x1 , introduced by the
proper noun David, and x2 , introduced by the pronoun he. The referent introduced
by the pronoun is marked with an overbar (x2 ), representing the fact that x2 is a
discourse referent whose reference must be resolved. The descriptive content of
the anaphoric pronoun he, in this case simply the information that the discourse
referent is male, is embedded under the presupposition operator, ∂ (Beaver 1992),
representing the fact that this aspect of the meaning constrains the possibilities for
anaphoric resolution. The presupposition connective ∂ maps ∂(𝜙) to true if 𝜙 is
true and to the undefined truth value otherwise. In this example, the only available
antecedent for x2 is x1 , but in more complex examples there may be other ways to
resolve the reference of x2 .
Anaphoric resolution is represented as a function, A, from anaphoric discourse
referents to their antecedents: in (49), the relation A holds between x2 and x1 . All
anaphoric discourse referents must obey the requirement in (50) for coreference

6 Of course, discourse referents can also be introduced by various other kinds of phrases and can
correspond to events, to plural individuals formed from the individuals relevant in the context, and so
on; see Kamp and Reyle (1993) for discussion of these issues. Haug’s (2014b) approach can also manage
these aspects of context update; see Dalrymple et al. (2018) for discussion of these issues.
520 anaphora

with an accessible antecedent; otherwise, there is a truth value gap. In (50), s is a

DRS state, x is an anaphoric discourse referent, and ν is an interpretation function
which assigns an individual to each discourse referent in every state. Notice that
x and its antecedent must both be defined in the same state s; this yields the usual
restrictions on anaphoric accessibility, as in DRT. We often omit the subscript s
on A, but with the understanding that the anaphoric relation A is defined only
between discourse referents in the same state s. As above, ∂ is the presupposition
connective:

(50) Constraint on anaphoric discourse referents:

∂(ν(s)(x) = ν(s)(As (x)))
Every anaphoric discourse referent x must be identical to its antecedent
As (x) in every state s, thus yielding coreference or, if As (x) is itself bound
by an operator, covariation.

Haug (2014b: 493–7) defines A as a composite function, depending on a function

I from discourse referents to syntactically accessible objects called “indices”: as
shown in (51), A is the composition of the function from discourse referents to
their indices, a function R from anaphor indices to antecedent indices, and the
inverse function I −1 from indices to discourse referents:

(51) Definition of A (see Haug 2014b: 497, ex. 69):

As (x) ≡ Is−1 (Rs (Is (x)))

The core of pragmatic anaphora resolution is then the function R, which maps
indices to antecedent indices; crucially, as we will see, indices are accessible to syn-
tactic representations and constraints. This allows us to keep the simple idea under-
lying the coindexation approach, namely that anaphoric relations are just relations
between syntactic tokens, but without presupposing that the resolution actually
takes place in the syntax. We often omit the subscript s on R, as we do with A.
We thus have the following set-up: indices, which are syntactically accessible,
introduce discourse referents; by mapping from the discourse referent of an
anaphor to its index, then from the anaphor index to the index of its antecedent,
and finally from the antecedent index to the antecedent discourse referent, we
obtain a mapping between the discourse referent of an anaphor and its antecedent
discourse referent (in a particular state s; as stated above, we often omit specifica-
tion of s to avoid clutter):

(52) R as a relation between indices:

A = I −1 ◦ R ◦ I
antecedent disc. ref. anaphor disc. ref.

I −1 I

index1 index2
R
 anaphora and semantic composition 521

Since A is uniquely determined by R, we can define constraints on R to capture

the constraints of binding theory.
In the remainder of this chapter, we present a theory of semantic composition
which produces representations like the one in (49) for a sequence of sentences in
a discourse, and also incorporates a theory of anaphora resolution which allows
for the expression of the anaphoric binding constraints discussed in §14.2.

14.4.2 Meaning and context update

As in standard DRT, PCDRT assumes that certain phrases introduce discourse
referents into the context, and thus that the DRS is a representation of the context
and the discourse referents it contains. As shown in (48), the DRS for the sentence
David arrived includes the conditions David (x1 ) and arrive (x1 ), which are required
to hold of the discourse referent x1 introduced by the name David. The same
observations can be made for indefinite noun phrases like a man. The sentence
A man arrived introduces a new discourse referent into the context, which can
be used to resolve a pronoun in subsequent discourse. In §8.4.1.4, we proposed
the meaning a (x, man (x), arrive (x)) for the sentence A man arrived. The compo-
nents of this meaning man (x) and arrive (x) also play a role in the DRS for the
sentence, which asserts the existence of some individual x1 who is a man and who
arrived:

(53) A man arrived.

x1
man (x1 )
arrive (x1 )

If the discourse starting with the sentence in (53) continues with the sentence He
yawned, we have Haug’s (2014b) PCDRT representation in (54). The discourse
referent x2 for the pronoun he is anaphoric, as indicated by the overbar. We assume
that it takes as its antecedent the discourse referent x1 for the phrase a man, as
represented by the A relation:

(54) A man arrived. He yawned.

x1 x2
man (x1 )
arrive (x1 ) , A(x2 ) = x1
∂(male (x2 ))
yawn (x2 )

As in standard DRT, quantifiers like nobody and everyone are different from
proper names and indefinites in their effect on the context. They do not introduce
new discourse referents into the wider discourse; instead, they effectively “trap”
discourse referents and do not allow them to contribute to the global context.
522 anaphora

When a sentence like Nobody arrived or Everyone arrived is uttered, no discourse

referents are made contextually available outside these structures:7

(55) Nobody arrived.

x1
¬ person (x1 )
arrive (x1 )

(56) Everyone arrived.

x1
⇒
person (x1 ) arrive (x1 )

In these examples, the universe of the main discourse structure is empty. This
explains the infelicity of the sentences in (57), as there are no discourse referents
available in the main discourse structure to resolve pronouns in subsequent
discourse:

(57) a. Nobody arrived. #He yawned.

b. Everyone arrived. #He yawned.

In fact, even a referent introduced by an indefinite in the scope of a quantifier like

nobody or everyone or in the scope of negation is not assumed to persist (Karttunen
1976; Groenendijk and Stokhof 1991; Kamp and Reyle 1993). As shown in (58), the
reading of a sentence like Nobody saw a mouse in which a mouse has narrow scope
relative to the subject Nobody does not allow reference to a mouse in subsequent
discourse:8

(58) Nobody saw a mouse. #It squeaked.

This observation is reflected in the fact that the universe of the main discourse
structure for Nobody saw a mouse does not contain discourse markers that repre-
sent either a mouse or a person:

7 The symbol ¬ in front of the sub-DRS in (55) is the negation symbol. On the treatment of negative
quantifiers such as nobody in DRT, see Corblin (1996).
8 For a brief discussion of quantifier scope, including narrow and wide scope readings, see §8.4.1.4.
 anaphora and semantic composition 523

(59) Nobody saw a mouse.

x1 x2
person (x1 )
¬
see (x1 , x2 )
mouse (x2 )

14.4.3 Anaphora and meaning assembly

Section 14.2 showed that constraints on anaphora resolution are defined in terms
of the syntactic and semantic role of the antecedent relative to the anaphor. In
this section, following Dalrymple et al. (2018), we show how the proposals of
Haug (2014b) can be used in a treatment of both intrasentential and intersentential
anaphora.
As discussed in §14.4.1, Haug (2014b) proposes to represent the anaphor-
antecedent relation by the function A, mediated via the function R from anaphor
indices to antecedent indices. We assign indices as the value of the index attribute
in semantic structure.9 To make it clear which index corresponds to which dis-
course referent, we represent indices as numbers, and we use the same number
(for instance, the number 1 in example 60) for an index and the discourse referent
to which it corresponds. For example, in (60) we assume that the word David
introduces the index 1, which appears as the value of the index attribute in the
semantic structure projected from the f-structure headed by David, and is related
via the function I to the discourse referent x1 :

I
(60) 2 3
0 1 x1
d pred ‘David’ 𝜎 d𝜎 index 1
David (x1 )

Similarly, the index introduced by a pronoun appears as the value of the index
attribute in the s-structure of the pronoun, and is related via the function I to the
discourse referent of the pronoun. This is important in allowing the statement of
syntactic constraints on anaphor-antecedent relations, since it allows the semantic
relation between anaphors and their potential antecedents to be syntactically
mediated and constrained.

14.4.3.1 Positive binding constraints With the indices introduced by a pro-

noun and its potential antecedents represented as the value of the index attribute

9 The index attributes at f-structure and semantic structure have a very different status: the value of
the f-structure index feature is a syntactic agreement feature bundle specifying pers, num, and gend
features (§2.5.7.1), while the value of the semantic index feature is a semantic index.
524 anaphora

at semantic structure, we can take advantage of the relation between semantic

structure and f-structure to state the syntactic constraints on binding relations
between anaphors and their antecedents discussed in §14.2. We make crucial use
of the following familiar relations:

(61) If ↑ is the f-structure of the anaphor, then:

↑𝜎 is the semantic structure of the anaphor;
the value of (↑𝜎 index) is the index contributed by the anaphor;
((gf⃰ gfpro ↑ ) gf) represents the potential antecedents of the anaphor; and
(((gf⃰ gfpro ↑ ) gf)𝜎 index) represents the indices contributed by the
potential antecedents of the anaphor.

This means that we can constrain anaphoric resolution possibilities for an anaphor
by referring to the index values of the anaphor and of its potential antecedents. We
assume that the appropriate prominence constraints are also imposed, as described
in §14.3, though for the sake of simplicity of explanation we do not explicitly
specify prominence constraints in this section.
As discussed in §14.2.1, positive binding constraints define the permissible
syntactic relations between a pronoun and its antecedent. For instance, we have
seen that the antecedent of the English reflexive pronoun himself must appear
in the Minimal Complete Nucleus relative to the pronoun, that is, the minimal
f-structure containing the pronoun and a subj function. In the sentence David
compared Chris to himself, there are two potential antecedents for himself in the
Minimal Complete Nucleus: David and Chris. From a syntactic perspective, either
of these antecedents may be chosen; here we assume a context in which David
is chosen as the antecedent for himself. Example (62) illustrates a syntactically
and semantically wellformed binding relation between himself and David. For
clarity, we represent the relation R by an arrow from the pronoun to its potential
antecedent:

(62) Davidi compared Chris to himselfi .

’
d

d
R
p
p

If we assume that p is the f-structure for himself, as in (62), the expression in (63)
represents the set of permissible f-structure antecedents of himself, including David
(the chosen antecedent) as well as Chris (a syntactically permitted antecedent, but
 anaphora and semantic composition 525

not the one chosen in this context). The expression in (63) reflects the definition
of the Minimal Complete Nucleus binding condition given in (15), according to
which the f-structure path gf⃰ delimiting the binding domain is not allowed to
pass through an f-structure with a subj function:

(63) F-commanding f-structures appearing in the Minimal Complete Nucleus

relative to p:
(( gf⃰ gfpro p) gf)
¬(→ subj)

We can now give a complete and explicit formulation of the positive Minimal
Complete Nucleus binding constraint as it applies to the reflexive pronoun himself.
The specification in (64) is a part of the lexical entry of himself, and constrains the
R relation between the index of himself and the index of its antecedent:

(64) Positive Minimal Complete Nucleus binding constraint:

R(↑𝜎 index) = ((( gf⃰ gfpro ↑ ) gf)𝜎 index)
¬(→ subj)

The equation in (64) requires the index of the antecedent of the pronoun,
R(↑𝜎 index), to appear as the value of index in the s-structure of a syntactically
permitted antecedent for the pronoun. In (62), this condition is met, and the
binding relation is permitted.
Example (65) illustrates an impermissible binding relation:

(65) ⃰Davidi thought that Chris had seen himselfi .

Ill-formed binding relation violating (64):
’
d
’

d
R
p
c
p

The index for David, 1, cannot be chosen as the antecedent of the reflexive pronoun
himself, and so cannot be the value of R(2), because the f-structure for David
(labeled d) does not stand in a syntactically permissible antecedent relation to the
pronoun f-structure p: the Minimal Complete Nucleus for the pronoun p is the
526 anaphora

comp f-structure, labeled c in (65). In other words, the binding requirement in

(64) is not satisfied when the antecedent David is chosen for the reflexive pronoun
himself in (65).

14.4.3.2 Negative binding constraints Section 14.2.2 discusses negative con-

straints, which rule out particular antecedents for a pronoun. In terms of the
antecedency relation R, negative constraints prevent syntactically impermissible
antecedents from being chosen as the value of R for a pronoun.
For example, the pronoun him in (66) obeys the negative Coargument Condi-
tion: it may not corefer with its coargument Chris, but it may take David as its
antecedent, since David is not a coargument, as the arrows representing the R
relation indicate:

(66) ×
Davidi thought that Chrisj had seen himi,⃰j .
Negative binding constraints are more complex to state than positive constraints, in
that negative constraints cannot be stated simply as constraints on the antecedency
relation R. To see this, consider the following sentence:

(67) David thought that he had seen him.

The antecedent of him may not be he, since him and he are coarguments. However,
as in (66), the antecedent of him may be David:

(68)
Davidi thought that he had seen himi .

Independently, David can serve as antecedent for he since, again, David does not
stand in a coargument relation to he:

(69)
Davidi thought that hei had seen him.
However, as observed by Wasow (1972), Lasnik (1976), and Higginbotham (1983),
if he takes David as its antecedent, then David cannot also serve as antecedent for
him, since in that case he and him would corefer, in violation of the requirement
for a pronoun like him not to corefer with a coargument.

(70) ×

Davidi thought that hei had seen him⃰i .

It is not possible to disallow this unacceptable pattern simply by forbidding
the anaphor-antecedent (R) relation to hold between coarguments. Rather, we
require noncoreference with coarguments: a pronoun which obeys the negative
Coargument Condition is prevented from having as its antecedent not just its
 anaphora and semantic composition 527

coarguments, but also any index which is related to one of its coarguments via
an antecedent relation.
In order to impose the proper constraints, we define a function R⃰, a recursive
version of R. R⃰ follows the R path back from index to antecedent index, stopping
only when it finds a index which does not have an antecedent:

(71) Definitionof R⃰:

 x if x does not have an antecedent as defined by the
R⃰(x) = R relation

R⃰(R(x)) otherwise

We can now formalize the negative constraint on him in the following way:

(72) Negative coargument condition:

R⃰(↑𝜎 index) .= R⃰((( gf⃰ gfpro ↑ ) gf)𝜎 index)
¬(→ pred)

This complex constraint appears in the lexical entry of the pronoun him. Notice
that the negation scopes over the disjunction over grammatical functions in the
coargument domain, giving universal force: R⃰(↑𝜎 index) may not be equal to R⃰
applied to any of the superior coarguments of ↑ . We examine each part of this
complex constraint in turn.
The constraint in (72) is repeated in (73), with the framed portion delimiting
the noncoreference domain of the anaphor. The off-path constraint ¬(→ pred)
defines the noncoreference domain as the Coargument Domain (compare exam-
ple 17): the minimal f-structure containing a pred.

(73) Coargument domain:

R⃰(↑𝜎 index) .= R⃰(( ( gf⃰ gfpro ↑ ) gf)𝜎 index)
¬(→ pred)

The framed portion of the constraint in (74) picks out the semantic structures of
the coarguments of the anaphor:10

(74) Semantic structures of the coarguments of the anaphor:

R⃰(↑𝜎 index) .= R⃰( (( gf⃰ gfpro ↑ ) gf)𝜎 index)
¬(→ pred)

Since indices appear as the value of the index attribute in semantic structure, the
framed portion of the constraint in (75) refers to the indices of the coarguments of
the anaphor; these indices and any indices related to them via the R relation may
not corefer with the anaphor or any of its antecedents:

10 Recall that we leave the prominence constraint implicit (see §14.3): the fully specified constraint
would refer only to more prominent coarguments of the anaphor, not to all coarguments.
528 anaphora

(75) Indices of the coarguments of the anaphor:

R⃰(↑𝜎 index) .= R⃰ ((( gf⃰ gfpro ↑ ) gf)𝜎 index)
¬(→ pred)

Thus, the constraint in (72) states that the discourse referent which is the
antecedent (of the antecedent (of the antecedent . . . )) of the pronoun may not be
identical with any coargument of the pronoun, or an antecedent (of an antecedent
(of an antecedent . . . )) of any coargument of the pronoun. In the case of (70), this
constraint produces the desired result: David is not an appropriate antecedent for
him because David is the antecedent of he, and he is a coargument of him.
In the case of (76) and (77), this constraint prevents him from coreferring with
Chris (76), but does not prevent him from coreferring with David (77), since David
is neither a coargument of him, nor the antecedent of a coargument of him.

(76) ⃰David thought that Chrisi had seen himi .

Ill-formed binding relation:
’
d
’ d
c
c
R
s p
p

(77) Davidi thought that Chris had seen himi .

’
d
’ d
c
c R
s p
p

14.4.3.3 Intersentential anaphora There is no syntactic relation between f-

structures that represent distinct clauses not related by coordination or subordi-
nation. That is, for the discourse

(78) David arrived. He yawned.

 anaphora and semantic composition 529

there is no syntactic relation between the f-structure for David arrived and
the f-structure for He yawned, and therefore no syntactic relation between the
f-structure for the anaphor he and the f-structure for its antecedent David. In such
cases, the function R is not constrained by syntax. We assume that speakers have
access to s-structures (and indices) from the preceding discourse, enabling them
to make pragmatic connections between discourse referents belonging to different
sentences and utterances. We can therefore use R for intersentential anaphora in
exactly the same way as for intrasentential anaphora.
The first sentence of the discourse in (78) is David arrived, with the following
f-structure, s-structure, and DRS:

(79) David arrived.

’ x1
a d David (x1 )
d
arrive (x1 )

The discourse continues with the sentence He yawned; the f-structure and s-
structure for He yawned and the DRS for (78) are:

(80) He yawned.
’
x 1 x2
David (x1 )
y h arrive (x1 )
h
∂(male (x2 ))
yawn (x2 )

Although there is no syntactic relation between the f-structure for he and the f-
structure for its antecedent David, we assume that the discourse representation
built up so far is available to the conversational participants, and that it is possible
for the function R to relate the index 2 of the pronoun to the index 1 of its
antecedent. Importantly, the negative constraint on pronouns given in (72) allows
this, because the antecedent David is not coreferent with a more prominent
syntactic coargument of the pronoun he. Example (81), repeated from (49), shows
the resolved meaning representation for the discourse under discussion, with
the anaphoric resolution function R relating the index for he to the index for
David:

(81) David arrived. He yawned.

x1 x2
David (x1 )
arrive (x1 ) , R(2) = 1
∂(male (x2 ))
yawn (x2 )
530 anaphora

We now show how the DRS in (81) is produced from its component parts.

14.4.3.4 Discourse Representation Structures and meaning assembly We

assume this f-structure for the first sentence, David arrived:

(82) David arrived.

. /
pred ‘arrive"subj#’
a 0 1
subj d pred ‘David’

The lexical entries for David and arrived are given in (83); we type-raise David,
treating it as a quantifier:

(83) David N (↑ pred) = ‘David’

x1
λP. ⊕ P(x1 ) : ∀H.(↑𝜎 ⊸ H) ⊸ H
David (x1 )
arrived V (↑ pred) = ‘arrive"subj#’
λx. : (↑ subj)𝜎 ⊸ ↑𝜎
arrive (x)

The symbol ⊕ represents the merge operation for combining DRSs by taking the
union of the discourse referents in the universe and the union of the discourse
conditions on the discourse referents. Instantiating the meaning constructors in
(83) according to the labels in the f-structure in (82), we get:

x1
(84) [David] λP. ⊕ P(x1 ) : ∀H.(d𝜎 ⊸ H) ⊸ H
David (x1 )

[arrive] λx. : d𝜎 ⊸ a𝜎
arrive (x)

These meaning constructors combine as follows:

 anaphora and semantic composition 531

x1
(85) λP. ⊕ P(x1 ) : ∀H.(d𝜎 ⊸ H) ⊸ H
David (x1 )
This is the meaning constructor [David]. On the
glue side, if we find a semantic resource of the
form d𝜎 ⊸ H for some semantic structure H, we
consume that resource and produce a semantic
resource for H. On the meaning side, we apply the
x1
function λP. ⊕ P(x1 ) to the meaning
David (x1 )
associated with d𝜎 ⊸ H.

λx. : d𝜎 ⊸ a𝜎
arrive (x)
This is the meaning constructor [arrive]. On the
glue side, it contributes an implicational resource
d𝜎 ⊸ a𝜎 , of exactly the form required by [David].
On the meaning side, it contributes a function
from individuals x to a constraint requiring x to
be an individual who arrived.
x1
[λP. ⊕ P(x1 )](λx. ) : a𝜎
David (x1 ) arrive (x)
By applying the meaning of [David] to the mean-
ing of [arrived], we have derived a meaning for
the semantic structure a𝜎 . In the next step, we
simplify this result by replacing all occurrences of
P in the meaning contributed by [David] with the
meaning contributed by [arrive].
x1
⊕ λx. (x1 ) : a𝜎
David (x1 ) arrive (x)
We can now simplify again, by replacing all occur-
rences of x in the meaning of [arrive] with the
discourse referent x1 .
x1
⊕ : a𝜎
David (x1 ) arrive (x1 )
The final simplification consists of merging the
two DRSs connected by ⊕ into a single DRS, as
required by the ⊕ merge operator.
x1
David (x1 ) : a𝜎
arrive (x1 )
We have produced the desired meaning for David
arrived.

14.4.3.5 Context update In DRT terms, sentences are understood as context

modifers: functions from a discourse context to a new, updated discourse context.
Therefore, we must provide a way for the meaning of a sentence to update the
532 anaphora

context in which it appears. To accomplish this, we provide root sentences with

an s-structure feature context, whose value is associated with the DRS for the
context in which the sentence appears. To model context update, we introduce a
meaning constructor which consumes the input context for a root sentence and
produces a new context which is relevant for subsequent discourse. We make the
following assumptions:

(86) a. For a sequence of sentences s1 s2, where s1 immediately precedes s2, the
value of the context attribute of s1 is assigned as the initial value of the
context attribute of s2. If a sentence is the first sentence in a discourse
and there is no preceding context, its context value is initially associated
with the null context.11
b. Each sentence in a discourse contributes a context update meaning con-
structor which updates its initial context to produce a new context
which can be further updated in subsequent discourse.

We begin our discussion of context update with an analysis of (78), in which the
sentence David arrived is the first sentence in the discourse. In this circumstance,
we associate the value of the context attribute of David arrived with the null
context, represented as a DRS with an empty universe and no conditions. This is
shown in (87) for the s-structure a𝜎 for the sentence David arrived:

(87) : (a𝜎 context)

Since sentences in DRT are treated as context modifiers, we associate root clauses
with a meaning constructor enforcing modification of the current context. This is
shown in (88):

(88) Context update meaning constructor:

λP.λQ.P ⊕ Q : ↑𝜎 ⊸ (↑𝜎 context) ⊸ (↑𝜎 context)

This meaning constructor consumes the meaning resource ↑𝜎 of the root sentence
to produce a modifier of its context. For the sentence David arrived, the up arrow
↑ in (88) is instantiated to a, as shown in (89a). Combining the context update
meaning constructor in (89a) and the meaning of David arrived given in the last
line of (85), repeated in (89b), produces the meaning constructor in (89c):

(89) a. λP.λQ.P ⊕ Q : a𝜎 ⊸ (a𝜎 context) ⊸ (a𝜎 context)

x1
b. David (x1 ) : a𝜎
arrive (x1 )

11 This is a simplification; we set aside representation of aspects of the context such as world
knowledge, location and time of the speech situation, and the identities of the interlocutors.
 anaphora and semantic composition 533

x1
c. λQ. David (x1 ) ⊕ Q : (a𝜎 context) ⊸ (a𝜎 context)
arrive (x1 )

Note that the meaning constructor in (89c) has the form of a context modifier
(Chapter 13), consuming the meaning of the context of a sentence and producing
an updated context meaning.
We now combine the meaning constructor in (89c)—which updates the input
context with the meaning of the current sentence David arrived—with the input
context in (87) representing the null context, since David arrived is the first
sentence in the discourse. This produces the updated discourse context in (90),
incorporating the effect of the sentence David arrived on the null context:

(90) David arrived.

’ x1
a David (x1 ) : (a )
arrive (x1 )

This provides the input context for any subsequent sentence in the discourse.

14.4.3.6 Anaphora: combining clauses In the discourse David arrived. He

yawned., the meaning derived in (90) for David arrived provides the context
in which the sentence He yawned is evaluated:

(91) He yawned.
’
y
h

The meaning constructors for he and yawned are:

x1
(92) [he] λP. ⊕ P(x1 ) : ∀H.(h𝜎 ⊸ H) ⊸ H
∂(male (x1 ))

[yawn] λx. : h𝜎 ⊸ y𝜎
yawn (x)

The meaning constructor for he differs from the one for David in only two ways.
First, the discourse referent x1 is marked as an anaphoric discourse referent, as
indicated by the overbar. This means that the discourse in which this referent
appears can only be evaluated for truth in the context of an anaphoric resolution
function R. Secondly, the condition requiring x1 to be male is in the scope of the
presupposition operator ∂. The meaning constructor for yawn differs from the one
for arrive only in referring to an act of yawning rather than to an act of arriving.
534 anaphora

Combining these meaning constructors along the same lines as in (85) produces
the meaning in (93) for the sentence He yawned:

(93) He yawned.
x1
∂(male (x1 )) : y𝜎
yawn (x1 )

Since He yawned is a root sentence like David arrived, it also contributes the
meaning constructor in (88), reflecting its role as a context modifier. Combining
the meaning constructor in (93) with the meaning constructor in (88) produces the
meaning constructor in (94), which requires the sentence to modify the context in
which it appears:

x1
(94) λQ. ∂(male (x1 )) ⊕ Q : (y𝜎 context) ⊸ (y𝜎 context)
yawn (x1 )

The discourse context for the sentence He yawned is the context inherited from the
previous sentence David arrived, shown in (90); that is, the initial context value
of the sentence He yawned is the same as the context value of David arrived. Thus,
(y𝜎 context) is initially associated with the same context as in (90):

x1
(95) David (x1 ) : (y𝜎 context)
arrive (x1 )

The meaning constructor in (95) must combine with the meaning constructor in
(94) to update this context. When DRSs are combined, their discourse referents
may have to be renumbered to prevent a clash; see Haug (2014b) for discussion
of this point. Linear precedence determines which discourse referents are renum-
bered: referents introduced earlier in the discourse retain their subscript. In this
example, then, x1 in the meaning for He yawned must be renumbered as x2 when
its DRS is composed with the DRS for David arrived, since a discourse referent x1
is already present. The result is:

x1 x2
David (x1 )
(96) arrive (x1 ) : (y𝜎 context)
∂(male (x2 ))
yawn (x2 )
 anaphora and semantic composition 535

This meaning constructor is semantically complete, but it is uninterpretable in the

absence of an anaphoric resolution for x2 . In this case, there is only one possible
antecedent, and therefore there is only one possible resolution:

x1 x2
David (x1 )
(97) arrive (x1 ) : (y𝜎 context), R(2) = 1
∂(male (x2 ))
yawn (x2 )

Note that anaphoric resolution does not remove discourse referents from the
context, or restrict the extent to which a particular discourse referent may function
as an antecedent to further anaphors. For example, if the discourse above is
continued by the sentence He fell asleep, the fact that x1 serves as the antecedent
to x2 (by virtue of the resolution R(2) = 1) does not prevent either x1 or x2
from serving as the antecedent to the discourse referent x3 introduced by the new
pronoun.

14.4.3.7 Indefinites in context Like a proper name, an indefinite noun phrase

like a man or someone introduces a discourse referent into the context. The
f-structure and DRS for the sentence Someone arrived are:

(98) Someone arrived.

. / x1
pred ‘arrive"subj#’
a 0 1 person (x1 ) : a𝜎
subj s pred ‘someone’
arrive (x1 )

After this sentence is uttered, the context (the universe of the DRS) contains the
discourse referent x1 representing an individual who is a person and who arrived.
To derive this meaning, we assume this lexical entry for someone:

(99) someone N (↑ pred) = ‘someone’

x1
λP. ⊕ P(x1 ) : ∀H.(↑𝜎 ⊸ H) ⊸ H
person (x1 )

This is similar in form to the lexical entry for the type-raised proper name David
in (83).
In the derivation of the meaning of Someone arrived, the premises in (100) are
relevant:
536 anaphora

(100) Meaning constructor premises for Someone arrived:

[arrive] λx. : s𝜎 ⊸ a𝜎
arrive (x)
x1
[someone] λP. ⊕ P(x1 ) : (s𝜎 ⊸ a𝜎 ) ⊸ a𝜎
person (x1 )

These meaning constructors combine in the way shown in (85) to produce the
complete and coherent meaning in (101), as desired:

x1
(101) [someone], [arrive] 1 person (x1 ) : a𝜎
arrive (x1 )

As a context modifier, this meaning constructor combines with the context-update

meaning constructor in (88) to modify its context. In our example, this sentence
is uttered at the beginning of a discourse, and so it combines with the meaning
constructor [null-context] given in (87) just as described for David arrived in the
previous section. The result is an updated context for a𝜎 , which provides the input
context for the next sentence in the discourse:

x1
(102) person (x1 ) : (a𝜎 context)
arrive (x1 )

14.4.3.8 Indefinites as antecedents The context made available by the sentence

Someone arrived contains the discourse referent x1 , representing a person who
arrived:

(103) Someone arrived.

x1
person (x1 )
arrive (x1 )

Interpretation of the sentence He yawned in this context proceeds as described

in §14.4.3.6, since the input context in the two situations is very similar. The
f-structure, semantic structure, and meaning constructor for He yawned are:

(104) He yawned.

’ x1
y ∂(male (x1 )) : y
h
yawn (x1 )

The meaning constructor premises in (105) are involved in the derivation of the
meaning of the discourse Someone arrived. He yawned. as follows:
 anaphora and semantic composition 537

(105) Meaning constructor premises for Someone arrived. He yawned.

[null-context] : (a )

x1
[someone-arrive] person (x1 ) : a

λP.λQ.P ⊕ Q : a ⊸ (a ) ⊸ (a
arrive (x1 )
)

x1
[he-yawn] ∂(male (x1 )) : y
yawn (x1 )

λP.λQ.P ⊕ Q : y ⊸ (y ) ⊸ (y )

We begin by combining the meaning constructor [someone-arrive] with the

meaning constructor [context-mod1] to produce a context modifier. Since this
sentence is the first sentence in the discourse, the resulting meaning constructor
combines with [null-context] to produce a context for Someone arrived:

x1
(106) person (x1 ) : (a𝜎 context)
arrive (x1 )

We likewise combine the meaning constructor [he-yawn] with the meaning con-
structor [context-mod2] to produce the following context modifier:

x1
(107) λQ. ∂(male (x1 )) ⊕ Q : (y𝜎 context) ⊸ (y𝜎 context)
yawn (x1 )

As with our previous examples, the context to be modified, (y𝜎 context), is inher-
ited from the previous sentence, here Someone arrived. In other words, the context
for He yawned is the one given in (106), and the initial value for (y𝜎 context) is
inherited from (a𝜎 context). Combining these meaning constructors produces
the complete, updated context in (108) for He yawned, and the meaning for the
discourse as a whole:

(108) [null-context], [he-yawn], [context-mod1], [someone-arrive],

[context-mod2] 1
x1 x2
person (x1 )
arrive (x1 ) : (y𝜎 context)
∂(male (x2 ))
yawn (x2 )
538 anaphora

Since the subject of each sentence introduces its own discourse referent, there must
be two distinct discourse referents in the combined meaning for the discourse.
Thus the referent labeled x1 in the meaning [he-yawn] must be renumbered as x2
in the DRS in (108) to avoid identity with the referent introduced by the meaning
[someone-arrive].
As with our previous example, since this meaning contains anaphoric discourse
referents, it is uninterpretable in the absence of a resolution function. Since there
is only one possible antecedent for the pronoun, the result is as follows:

x1 x2
person (x1 )
(109) arrive (x1 ) : (y𝜎 context), R(2) = 1
∂(male (x2 ))
yawn (x2 )

14.4.4 Context and quantifiers

Indefinite phrases like someone or a man are treated differently from quantified
noun phrases like nobody or every woman in Discourse Representation Theory, as
they are in Dynamic Predicate Logic (Groenendijk and Stokhof 1991): as discussed
in §14.4.1, quantified noun phrases do not introduce a discourse referent that is
available in subsequent discourse. The f-structure and meaning constructor for
the sentence Nobody arrived are:

(110) Nobody arrived. (#He yawned.)

. /
pred ‘arrive"subj#’ x1
a 0 1 : a𝜎
subj n pred ‘nobody’ ¬ person (x1 )
arrive (x1 )

No discourse referent is available outside the scope of the negation after this
sentence is uttered; this explains the infelicity of a continuation like He yawned,
which requires an accessible antecedent for interpretation of its pronoun subject.
We assume the following lexical entry for the negative quantifier nobody:

(111) nobody N (↑ pred) = ‘nobody’

λP. x1 : ∀H.(↑𝜎 ⊸ H) ⊸ H
¬[ ⊕ P(x1 )]
person (x1 )
 anaphora and semantic composition 539

Instantiating the meaning constructor in (111) according to the f-structure labels

in (110), we have the instantiated meaning constructor in (112), labeled [nobody]:

(112) [nobody] λP. x1 : (n𝜎 ⊸ a𝜎 ) ⊸ a𝜎

¬[ ⊕ P(x1 )]
person (x1 )

The crucial difference between this meaning constructor and the one for someone
is that the discourse referent introduced and the condition placed upon it are
embedded within a larger DRS, under the scope of the negation operator. This
means that the contextual contribution of nobody is to leave the wider context
(that is, the context outside the scope of the negative) unchanged. Thus, like
the earlier proposal of Dalrymple et al. (1997), this analysis has the desirable
property of correctly characterizing interactions between quantifier scope and
bound anaphora: any pronouns bound by a quantifier like everyone or nobody must
appear within the scope of the quantifier, since the discourse referent introduced
by the quantifier is only available in contexts within its scope, not outside it.
For the sentence Nobody arrived, we assume these meaning constructor
premises:

(113) Meaning constructor premises for Nobody arrived:

[arrive] λx. : n𝜎 ⊸ a𝜎
arrive (x)

[nobody] λP. x1 : (n𝜎 ⊸ a𝜎 ) ⊸ a𝜎

¬[ ⊕ P(x1 )]
person (x1 )

These premises combine to produce:

x1
(114) [nobody], [arrive] 1 : a𝜎
¬ person (x1 )
arrive (x1 )

Next, we briefly examine the derivation of the meaning of the sentence Nobody saw
a mouse. On the most accessible reading of this sentence, the indefinite a mouse
appears inside the scope of the quantifier nobody, and does not contribute to the
540 anaphora

overall context. This accounts for the infelicity of a sentence like It squeaked in
subsequent discourse:

(115) Nobody saw a mouse. (#It squeaked.)

 
’
  x1 x 2
 n 
s


 person (x1 ) :s
 −  ¬
m mouse (x2 )
see (x1 , x2 )

The derivation of this reading proceeds on the basis of the meaning constructors
in (116), which have been instantiated according to the f-structure labels in (115):

(116) Meaning constructor premises for Nobody saw a mouse:

[see] λy.λz. : n𝜎 ⊸ (m𝜎 ⊸ s𝜎 )

see (y, z)

[nobody] λP. x1 : ∀H.(n𝜎 ⊸ H) ⊸ H

¬[ ⊕ P(x1 )]
person (x1 )

x2
[a] λP.λQ. ⊕ P(x2 ) ⊕ Q(x2 ) : ∀H.((m𝜎 index) ⊸ (m𝜎 restr))
⊸ (m𝜎 ⊸ H) ⊸ H

[mouse] λx. : (m𝜎 index) ⊸ (m𝜎 restr)

mouse (x)

The premises [a] and [mouse] combine to produce the meaning constructor in
(117), labeled [a-mouse]:12

x2
(117) [a-mouse] λQ. ⊕ Q(x2 ) : ∀H.(m𝜎 ⊸ H) ⊸ H
mouse (x2 )

Combining the premises labeled [a-mouse] and [see], we have the meaning
constructor labeled [see-a-mouse]:13

12 For a full discussion of how the meaning of a determiner like a or every combines with the meaning
of a common noun like mouse, see §8.8.2.
13 For the sake of space, we omit the introduction and discharging of a hypothetical subject. For
details, see §8.8.1.5.
 anaphora and semantic composition 541

x2
(118) [see-a-mouse] λy. mouse (x2 ) : n𝜎 ⊸ s𝜎
see (y, x2 )

We combine this meaning constructor with [nobody], producing the meaning

constructor labeled [nobody-see-a-mouse]:

x1 x2
(119) [nobody-see-a-mouse] person (x1 ) : s𝜎
¬
mouse (x2 )
see (x1 , x2 )

This reading of the sentence makes no discourse referents available for anaphora
resolution in subsequent discourse. Outside of a special context, it is difficult to
get a wide scope reading for the indefinite, according to which there is a particular
mouse that no one saw.
The quantifier everyone works in an entirely parallel way. We assume the
following meaning constructor for everyone:

(120) λP. x1 : ∀H.(↑𝜎 ⊸ H) ⊸ H

⇒ P(x1 )
person (x1 )

For the sentence Everyone saw a mouse, two readings are readily available: the
indefinite can take narrow scope, meaning that each person saw a (possibly
different) mouse, or wide scope, meaning that there is a particular mouse which
everybody saw.14 Under the second reading, a discourse referent corresponding to
a mouse is available for pronoun resolution in subsequent discourse. We assume
the following f-structure:

(121) Everyone saw a mouse.

 
’
 
 e 
s



 − 
m

14 Meaning constructors and deductions of narrow and wide scope readings for quantifiers are
discussed in §8.8.1.5.
542 anaphora

The derivation of the narrow-scope indefinite reading of this sentence proceeds

on the basis of the following meaning constructors, which have been instantiated
according to the f-structure labels in (121):

(122) Meaning constructor premises for Everyone saw a mouse:

[see] λx.λy. : e𝜎 ⊸ (m𝜎 ⊸ s𝜎 )

see (x, y)

[everyone] λP. x1 : ∀H.(e𝜎 ⊸ H) ⊸ H

⇒ P(x1 )
person (x1 )

x2
[a-mouse] λP. ⊕ P(x2 ) : ∀H.(m𝜎 ⊸ H) ⊸ H
mouse (x2 )

As before, we combine the premises labeled [a-mouse] and [see] to produce the
meaning constructor [see-a-mouse]:

x2
(123) [see-a-mouse] λy. mouse (x2 ) : e𝜎 ⊸ s𝜎
see (y, x2 )

We combine this meaning constructor with [everyone], producing the meaning

constructor labeled [everyone-see-a-mouse]:

x2
(124) [everyone-see-a-mouse] x1 : s𝜎
⇒ mouse (x2 )
person (x1 )
see (x1 , x2 )

As with the narrow scope reading of nobody saw a mouse, this makes no discourse
referents available for anaphora resolution in subsequent discourse.
When the indefinite takes wide scope, a discourse referent corresponding to
a mouse is available. Beginning with the premises in (122), we combine the
premises labeled [everyone] and [see], producing the meaning constructor labeled
[everyone-see]:

(125) [everyone-see] λy. x1 : m𝜎 ⊸ s𝜎

⇒
person (x1 ) see (x1 , y)
 further reading and related issues 543

Combining this meaning constructor with [a-mouse] produces the meaning

constructor labeled [everyone-see-a-mouse2]:

x2
mouse (x2 )
(126) [everyone-see-a-mouse2] x1 : s𝜎
⇒
person (x1 ) see (x1 , x2 )

We have shown that the glue approach is valuable not only in accounting for mean-
ing contributions in the derivation of the meaning of a sentence, but also in man-
aging contextual contributions across sentences in a discourse. The representation
of context and the DRT-based glue-theoretic treatment of anaphora resolution is of
crucial importance in our treatment of anaphoric control in Chapter 15. Elsewhere,
however, when context and anaphora resolution are not relevant to the semantic
issues we examine, we employ simpler representations using predicate logic. No
loss of generality results from this simplification.

14.5 Further reading and related issues

LFG-based treatments of anaphoric binding often assume that the anaphor-
antecedent relation is explicitly represented in the semantic structure of pronouns
by means of an attribute antecedent whose value is the semantic structure of the
antecedent of the pronoun (Dalrymple 2001; Asudeh 2012). This attribute plays a
role in Asudeh’s semantic treatment of resumptive pronouns, discussed in §17.6.4.
Research on the syntax of anaphoric binding has revealed correlations between
the morphological form of a pronoun and its binding properties (Faltz 1977; Pica
1987): monomorphemic pronouns tend to have different binding properties from
polymorphemic pronouns, for example. Bresnan et al. (2016: Chapter 11) provide
a discussion and analysis of this issue from an LFG perspective; see also Lødrup
(2007, 2008a) and Rákosi (2009).
Bresnan (2001a) addresses issues of markedness and asymmetries in pronom-
inal systems: for instance, many languages have both free pronouns and null or
incorporated pronominals, but there are no languages that have incorporated pro-
nouns but no free pronominal forms. Bresnan proposes an Optimality-Theoretic
analysis explaining these and other asymmetries. On the basis of an inspection of a
large sample of languages, Siewierska (2003) verifies and refines some of Bresnan’s
claims.
The semantics of a predicate can play a role in the acceptability and distribu-
tion of anaphoric arguments of the predicate. Park (2012) argues that reflexive
requirements can be lexically specified for distinct groups of predicates, and that
these specifications can narrow down the choice of anaphor in cases of overlapping
syntactic binding domains.
Analyses of anaphoric binding often concentrate primarily on reflexives and
non-reflexive pronouns, and less on reciprocals and reciprocal binding; an
544 anaphora

important exception is the analysis of reciprocals in Icelandic, Malagasy, and

Swahili by Hurst (2010, 2012).
It is sometimes controversial or unclear whether a particular form is a reflexive
pronoun with its own semantic form or a marker indicating detransitivization: that
is, that one of the arguments of a predicate has been suppressed. Grimshaw (1982a)
discusses the French reflexive clitic, arguing that it is not in fact a reflexive pronoun,
but a marker indicating that the verb has been detransitivized. Alsina (1996)
provides similar arguments for Romance clitics, and Patejuk and Przepiórkowski
(2015) for the Polish word sie˛. De Alencar and Kelling (2005) argue for the opposite
view for German and Romance: that reflexive clitics are in fact reflexive pronouns
which fill an argument position of the predicate. Sells et al. (1987) provide a
very useful discussion of diagnostics for syntactic and semantic transitivity and
detransitivization.
In this chapter, our discussion has covered pronouns whose antecedency con-
ditions are syntactically defined, that is, statable in terms of f-structural properties
such as the presence of a pred or subj. However, not all pronouns obey purely
syntactic binding conditions. Antecedency conditions for some pronouns depend
on information structural properties of the sentence or discourse; Culy (2000)
discusses topic anaphora, distinguishing among anaphors that refer to the topic
of the sentence, paragraph, or story, and Strahan (2009) discusses discourse
constraints on anaphoric binding in Icelandic. Other conditions on pronoun
antecedency are also found: a logophoric pronoun is one which is used to refer to
“the author of a discourse or to a participant whose thoughts are reported” (Hagége
1974, translation by Stirling 1988). Culy et al. (1995) examine the pronominal
systems of three Dogon languages and trace their evolution from a common
ancestor, showing how logophoric pronouns can evolve in the course of language
change. Bresnan et al. (2016: Chapter 11) discuss logophoricity and constraints
on logophoric pronouns, and Haug (2013) discusses the role of logophoricity in
constructions involving partial control (see §15.6).
Related to systems of pronominal binding, coreference, and logophoricity are
systems of obviation and switch reference, where certain antecedents are allowed
or disallowed for anaphors in certain syntactic positions. Simpson and Bresnan
(1983) analyze control and obviation in Warlpiri; their analysis is reviewed by Dal-
rymple (1993), who provides a reanalysis using inside-out functional uncertainty.
Our semantic analysis depends heavily on work by Dag Haug and colleagues
on Partial Compositional Discourse Representation Theory (Haug 2013, 2014a,b;
Belyaev and Haug 2014). Haug (2014b) provides a full overview of the theory,
Belyaev and Haug (2014) discuss the treatment of correlatives (on which see also
§17.1.2.4), and Haug (2014a) discusses anaphoric control and bridging.
 15
Functional and anaphoric
control

This chapter explores the syntax and semantics of functional and anaphoric
control, constructions in which either syntactic or lexical constraints require
coreference between an argument of the matrix clause and an argument of a
subordinate or modifying adjunct clause. In English, such cases include the classes
of equi and raising verbs. Crosslinguistically, descriptions of such constructions
involve reference to functional syntactic relations such as subj, obj, and so on;
therefore, the syntactic discussion in this chapter is primarily centered around the
f-structures of functional and anaphoric control constructions.
The open grammatical functions xcomp and xadj and the closed function
comp were first introduced by Bresnan (1982a) in a pioneering study of clausal
relations and complementation. Since then, there has been a wealth of work in LFG
building on these proposals. In the following sections, we will review the major
proposals within LFG for the syntactic and semantic treatment of functional and
anaphoric control constructions.

15.1 Open complements and functional control

As an illustration of functional control, we first examine raising verbs,1 verbs like
seem:

(1) David seemed to yawn.

Raising verbs are distinguished by the fact that the “raised” argument, the subj
David in (1), is not a semantic argument of the raising verb. In other words, raising
verbs impose no semantic constraints on the raised argument. Notationally, this is
indicated by the position of the raised argument outside of the angled brackets in
the semantic form of the raising verb, as discussed in §2.4.6.1:

1 Raising verbs are so called because of their analysis in transformational grammar (Kiparsky and
Kiparsky 1970; Postal 1974), in which the subject phrase of the subordinate clause was assumed to have
raised, or moved up, from the subordinate clause to its final position in the matrix clause.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
546 functional and anaphoric control

(2) David seemed to yawn.

 
 
 
 
 
 ’ 
 
 

(3) David believed Chris to know the answer.

 
 
 
 
 
 
 
  
 ’ 
 
  
  
  
  
  + 
 

Raising verbs in English and other languages exemplify functional control. Func-
tional control verbs require as an argument an open complement xcomp (§2.1.7).
In (2), the subj of the raising verb functionally controls the subj of the subordinate
xcomp. This means that the subj of the verb seemed is required to be the same
f-structure as the subj of the subordinate xcomp, as indicated by the line con-
necting the two subject positions in (2); see §2.4.3. Other raising verbs exhibit
functional control by an obj, as shown in (3), where the obj of the matrix verb
is the same as the subj of the xcomp.

15.1.1 Evidence for functional control

In functional control analyses as seen in (2) and (3), in which the same argument
is both an argument of the matrix verb and the subj of the subordinate xcomp, any
syntactic restrictions that are imposed on the subj in the subordinate clause must
also hold for the “raised” argument, since the same f-structure is an argument in
both the matrix and the subordinate clause.
As discussed in §2.5.3, some English predicates require a semantically empty
subject—an expletive subject—with a particular syntactic form. For example, the
verb rain requires its subject to have the form it, not there:

(4) a. It is raining.
b. ⃰There is raining.
 open complements and functional control 547

The f-structure for (4a) is:

(5)

In contrast, the verb be in the existential there-construction requires its subject to

have the form there:

(6) There is a problem.

Since raising verbs impose no semantic constraints on their “raised” argument,

we expect expletive arguments—arguments with no semantic content—to appear
felicitously in raising constructions (Postal 1974; Bresnan 1982a,c). Syntactic
requirements on the form of the “raised” argument must also be met in the
functional control construction, since both matrix and subordinate clause require-
ments must be satisfied:

(7) a. It seems to be raining.

b. There seems to be a problem.

(8) a. David believed it to be raining.

b. David believed there to be a problem.

In (7) and (8), the “raised” argument/controller is not a semantic argument of

either the matrix or the subordinate predicate; syntactic requirements imposed by
the subordinate clause verb are satisfied, no semantic requirements are violated,
and the examples are wellformed. The f-structures for (7a) and (8a) are:2

(9) a. It seems to be raining.

 
 
 
 
 
 
 
 

b. David believed it to be raining.

 
 
 
 
 
 
 
 
 
 
 

2 On the form features in (9), see §2.5.3.

548 functional and anaphoric control

In English and many other languages, raising constructions like these are not
limited to verbal predicates, but are also found with adjectival predicates. For
example, the adjectival predicate be likely displays exactly the same behavior as
subject-raising verbs like seem. Such predicates have a syntactically selected subj
argument which is not semantically selected, and is filled by the “raised” argument
of the embedded complement:

(10) a. It is likely to be raining.

b. David believed it to be likely to be raining.

(11) a. It is likely to be raining.

 
 
 
 
 
 
 
 

b. David believed it to be likely to be raining.

 
 
 
 
 
 
  
 
 
  
  
  
  
  
  
  
 

Similarly, in languages in which predicates place requirements on the case of

their arguments, any such requirements imposed by the subordinate predicate
must likewise hold for the “raised” argument of the matrix predicate. Andrews
(1982) discusses raising and functional control in Icelandic, exploring the behavior
of “quirky case” verbs, verbs that lexically specify a particular case for their
arguments. As discussed in §2.2.1, Icelandic subjects of quirky case verbs can
be marked with one of several cases, depending on requirements imposed by
the verb:

(12) a. Accusative subject:

Drengina vantar mat.
boys.def.acc lacks food.acc
‘The boys lack food.’
 open complements and functional control 549

b. Dative subject:
Barninu batnaði veikin.
child.def.dat recovered.from disease.def.nom
‘The child recovered from the disease.’
c. Genitive subject:
Verkjanna gætir ekki.
pains.def.gen is.noticeable not
‘The pains are not noticeable.’

Andrews shows that in a functional control construction, the case of a “raised” obj
depends on the casemarking requirements on the subj of the lower clause:

(13) a. Accusative “raised” object:

Hann telur mig (í barnaskap sínum) vanta peninga.
he believes me.acc (in foolishness his) to.lack money.acc
‘He believes me (in his foolishness) to lack money.’
b. Dative “raised” object:
Hann telur barninu (í barnaskap sínum) hafa
he believes child.def.dat (in foolishness his) to.have
batnað veikin.
recovered.from disease.def.nom
‘He believes the child (in his foolishness) to have recovered from the
disease.’
c. Genitive “raised” object:
Hann telur verkjanna (í barnaskap sínum) ekki
he believes pains.def.gen (in foolishness his) not
gæta.
noticeable
‘He believes the pains (in his foolishness) not to be noticeable.’

The position of the parenthesized adjunct í barnaskap sínum ‘in his foolishness’,
which is a matrix clause modifier, shows that the “raised” constituent does indeed
appear as the obj of the matrix clause and not in the subordinate clause. Since
this argument is also the subj of the subordinate xcomp, the constraints on
casemarking that the subordinate xcomp verb imposes on this argument must be
met for the sentence to be wellformed. Andrews (1990a) and Zaenen and Maling
(1990) provide more discussion of default case, quirky case, and raising verbs in
Icelandic.
Jacobson (1990, 1992b) discusses some tests which may be taken to demonstrate
the syntactic characteristics of the open complement xcomp in English. First, VP
complement drop is never possible for the open complement xcomp:

(14) a. [Did David really yawn?]

'
to (yawn).
He seemed
⃰∅
550 functional and anaphoric control

b. [Did Chris really know the answer?]

'
to (know the answer).
David believed him
⃰∅ (wrong meaning)

Second, xcomp is not among the syntactic categories that can appear in sentence-
initial position in a long-distance dependency in English:

(15) a. ⃰To yawn, David seemed.

b. ⃰To know the answer, David believed Chris.

As we discuss further in §15.3.2, the closed complement comp may behave

differently in each of these respects.

15.1.2 Constituent structure and functional constraints

We propose the following annotated phrase structure rule for functional control
constructions in English (only details relevant to this construction are displayed):3

(16) V& −→ V NP VP
↑ =↓ (↑ obj) = ↓ (↑ xcomp) = ↓

This rule constrains the constituent structure of functional control constructions

and the functional syntactic role of each constituent. Notice that the phrase struc-
ture rule does not specify the control relation between the matrix argument and
the subj of the xcomp: the difference between verbs like seemed, whose subj is also
the subj of its xcomp, and believed, whose obj is the subj of its xcomp, is lexically
specified by the raising verb, not given by constituent structure requirements.
The lexical entries for the English raising verbs seemed and believed contain the
following syntactic information:

(17) seemed V (↑ pred) = ‘seem'xcomp(subj’

(↑ subj) = (↑ xcomp subj)

(18) believed V (↑ pred) = ‘believe'subj,xcomp(obj’

(↑ obj) = (↑ xcomp subj)

3 With Bresnan et al. (2016), we classify the infinitival to complement as a VP, and we assume that the
word to is head of the verbal projection. Falk (2001b) assumes a different c-structure for to-infinitives,
arguing that to appears in C, as the head of CP.
 open complements and functional control 551

These lexical entries contain a control equation specifying the relation between a
particular argument of the matrix clause and the subj of the subordinate xcomp.
The control equations for the above verbs are:

(19) seemed (↑ subj) = (↑ xcomp subj)

believed (↑ obj) = (↑ xcomp subj)

These lexical entries and phrase structure rules give rise to the c-structures and
f-structures shown in (20) and (21) for examples (2)–(3). The theory of controller
selection and how the control equation is determined will be discussed in §15.7.

(20) David seemed to yawn.

IP

NP I  
(↑ ↓ ↑=↓  
 
 
N VP  
 ’ 
↑=↓ ↑=↓  
 

David V
(↑ ↑=↓

V VP
↑=↓ (↑ ↓

seemed
V
(↑
↑=↓
(↑ ↑
V VP
↑=↓ ↑=↓

V
to
↑=↓

yawn
(↑ ’
552 functional and anaphoric control

(21) David believed Chris to know the answer.

 
 
 
 
 
 
 
  
 ’ 
IP  
  
  
NP I   
  
  + 
N VP  
David V

V NP VP

believed N V

Chris V VP

to V

V NP

know D N

the N
answer

Raising predicates with to-infinitive complements are not the only structures in
English which can be analyzed using xcomp. Bresnan et al. (2016: §§12.1 and 12.2)
discuss predicate complements such as the following:

(22) a. Mary didn’t sound ashamed of herself.

b. Jogging keeps Susan in a bad mood.
c. Linda will have your brother working again.

In (22a), the underlined phrase is an AdjP, in (22b) the underlined phrase is a PP,
and in (22c) the underlined phrase is a participial VP. All of the underlined phrases
in (22) can be analyzed as xcomp complements of their respective matrix verbs,
as shown by Bresnan et al. (2016). Bresnan et al. also discuss interactions between
raising and anaphoric binding in constructions such as these.
Falk (2005) notes that the vast majority of predicates which take xcomp com-
plements require those complements to be headed by a verbal predicate, and
do not permit nominal, adjectival, or prepositional complements. Falk argues
that this is unexpected if all open complements have the same function xcomp,
since subcategorization is primarily functional, not categorial (§2.3). Falk therefore
argues for the differentiation of two additional open argument functions, xobj𝜃
and xobl𝜃 , alongside xcomp. Under this proposal, adjectival or nominal open
 open complements and functional control 553

complements are xobj𝜃 , prepositional open complements are xobl𝜃 , and xcomp is
restricted to verbal open complements. As discussed in §2.1.7, the status of clausal
complement functions, in particular comp, is controversial; here we follow the
traditional analysis of xcomp as the grammatical function for all types of open
complement.

15.1.3 Backward control and subsumption

Functional descriptions such as (↑ subj) = (↑ xcomp subj), as in (19), specify
a relation of equality between the values of two f-structures. In this case, the
f-structure which serves as the value of (↑ subj) is the same f-structure which
serves as the value of (↑ xcomp subj). In contrast, Zaenen and Kaplan (2002)
propose that some functional control relations are better analyzed in terms of sub-
sumption (§6.9.2), and this proposal is extended by Sells (2006b) to all functional
control relations.
Zaenen and Kaplan (2002) contrast the ungrammaticality of the German sen-
tence in (23a) with the grammaticality of (23b):

(23) a. ⃰[Ein Aussenseiter zu gewinnen] versuchte hier noch nie.

An outsider to win tried here still never
‘An outsider never tried to win here.’
b. [Ein Aussenseiter zu gewinnen] schien hier eigentlich nie.
An outsider to win seemed here actually never
‘An outsider never actually seemed to win here.’

In both of these sentences, the four words preceding the finite verbs (versuchte
and schien respectively) constitute a single phrase, meaning that the phrase Ein
Aussenseiter, which is the subject of both verbs, appears within the embedded
clause, rather than at the level of the matrix verb. This phenomenon is called
backward control: rather than an argument of the matrix clause controlling an
argument of the embedded clause, the control relation works “backwards,” an
argument from the embedded clause controlling the missing argument position in
the matrix clause (Polinsky and Potsdam 2002a,b, 2006). The f-structure assumed
for (23b) is:
 
(24) ’
 
 
  
 
 ’ 
  
  
  
    
  
  
 
 
  
 
 
  
  
    
    
 
 

554 functional and anaphoric control

In (24), we represent the “backward control” relation by depicting the f-structure

for ein Aussenseiter ‘an outsider’ as the value of the xcomp subj, with a line con-
necting it to the matrix subj. Importantly, this is done only for ease of exposition,
and as an explicit indication that this example involves backward control. Zaenen
and Kaplan (2002) analyze the verb scheinen ‘seem’ as requiring its subj to be equal
to its xcomp’s subj; since the same f-structure appears as the value of both the subj
and the xcomp subj, there is no formal difference between this representation and
one in which the ‘outsider’ f-structure is depicted as the value of the matrix subj,
with a line connecting it to the xcomp subj. For further discussion of this point,
see §2.4.3.
The question is why backward control is acceptable for the verb scheinen ‘seem’,
but not for versuchen ‘try’. Unlike the raising verb seem, the verb try is an equi verb.
The analysis of equi verbs will be presented in more detail in §15.3, but at this point
it is sufficient to note that Zaenen and Kaplan (2002) assume a functional control
analysis for both raising and equi verbs in German. Crucially, Zaenen and Kaplan
propose that equi verbs differ from raising verbs in that the control relation for
raising verbs is stated in terms of equality, while the control relation for equi verbs
is stated in terms of subsumption. They propose the following control relation for
German versuchen:

(25) versuchen (↑ subj) ) (↑ xcomp subj)

The difference between equality and subsumption in this context can be thought
of in terms of information flow. In the case of equality, information flow is bidi-
rectional: any information specified for the embedded argument is automatically
shared with the matrix argument, and vice versa, since they are equal. This allows
the backward control relation observed for scheinen ‘seem’ in (23b), since a phrase
filling the subordinate xcomp subj role is also required to fill the role of the
main clause subj. In the case of the subsumption relation required by the verb
versuchen ‘try’, however, information flow is monodirectional: if a ) b, then any
information specified for a must be shared with b, but information specified for b is
not necessarily shared with a. With versuchen, then, if the shared argument appears
in the matrix clause as the value of (↑ subj), as desired, its pred also appears as the
value of (↑ xcomp subj), in much the same way as it would if the relation were one
of equality. But if the shared argument appears in the embedded clause, as the value
of (↑ xcomp subj), its pred does not appear as the value of (↑ subj), rendering the
matrix f-structure incomplete. This models the fact that versuchen does not permit
backward control: if the controller appears in the embedded clause, the matrix subj
position is empty, leading to an incomplete f-structure.
Sells (2006b) extends Zaenen and Kaplan’s proposals in an analysis distinguish-
ing forward and backward control, proposing that all functional control relations
can be analyzed in terms of subsumption, and arguing that this provides a good
model of the typology of control constructions. Sells argues that in German both
raising and equi verbs specify the control relation in terms of subsumption; so, the
functional description specifying the control relation for the raising verb scheinen
‘seem’ is exactly the same as that for versuchen ‘try’:
 open complements and functional control 555

(26) scheinen (↑ subj) ) (↑ xcomp subj)

Why, then, is (23b) grammatical? Sells follows Berman (2003) in assuming that in
German, subjects do not need to contain a pred feature, but may consist solely of
agreement features if and only if they are non-thematic. The controller argument
of raising predicates is, as we have seen above, not semantically selected by the
raising predicate. Thus in the case of (23b), the lack of a semantically contentful
(pred-less) subj at the matrix level does not lead to an incomplete f-structure.
On the other hand, as we discuss in §15.4, the controller argument of an equi
predicate is semantically selected, and so the f-structure is ill-formed if the matrix
subj lacks a pred feature. This means that if control relations are stated in terms
of subsumption, backward control is possible with raising predicates, but not with
equi predicates.
In some languages, there are predicates which allow only backward control.
Following work by Polinsky and Potsdam (2002b), Sells analyzes the Malagasy
examples in (27) as “forward control,” with the controller overtly realized in the
main clause, and the examples in (28) as “backward control,” with the controller
overtly realized in the subordinate clause:

(27) Forward control:

a. m-an-andrana [m-i-tondra ny fiara] Rabe
prs-active-try [prs-active-drive the car] Rabe
‘Rabe is trying to drive the car.’
b. m-an-andrana Rabe [m-i-tondra ny fiara]
prs-active-try Rabe [prs-active-drive the car]
‘Rabe is trying to drive the car.’

(28) Backward control:

a. m-an-omboka [m-i-tondra ny fiara Rabe]
prs-active-begin prs-active-drive the car Rabe
‘Rabe is beginning to drive the car.’
b. ⃰m-an-omboka Rabe [m-i-tondra ny fiara]
prs-active-begin Rabe [prs-active-drive the car]
‘Rabe is beginning to drive the car.’

The ungrammaticality of (28b) is explained, following Polinsky and Potsdam

(2002b), on the assumption that verbs such as omboka ‘begin’ in Malagasy display
backward control: since the shared argument appears in the embedded clause, it
cannot precede the embedded verb, unlike the shared argument in (27), which can
either precede or follow the embedded verb since it is a part of the matrix clause.
For constructions like these, in which only backward control is possible, Sells
(2006b) proposes that the subsumption relation is stated the other way around:

(29) omboka (↑ xcomp subj) ) (↑ subj)

556 functional and anaphoric control

Since all of the constraints on the embedded subject also hold for the matrix
subject, and since there is nothing else which can provide a subj value for the
xcomp, this equation ensures that the controller must appear within the embedded
clause.
Whether a comprehensive account of all functional control relations can be
stated in terms of subsumption rather than equality is not clear. Haug (2011)
discusses backward control phenomena in Ancient Greek, and concludes that
equality, rather than subsumption, provides the most satisfying account (see also
Haug 2017). Backward control in Indonesian is also discussed by Arka (2014).

15.2 Raising verbs and semantic composition

15.2.1 Semantics of raising verbs
The semantic contribution of raising verbs like seem, appear, and tend has been
widely studied. We propose the representation in (30) for the sentence David
seemed to yawn:4

(30) David seemed to yawn.

seem (yawn (David ))

In this example, the predicate seem holds of the proposition yawn (David ), and the
sentence has more or less the same meaning as the sentence It seemed that David
yawned.
As noted in §8.4.1.4, scope ambiguities are available with the subj argument of
a subject raising verb like seem or appear and with the obj argument of an object
raising verb like believe (May 1977; Williams 1983; Jacobson 1990; Halvorsen
1983): a raising verb allows a narrow scope reading for its “raised” argument. For
example, a sentence like Someone appeared to yawn has two readings, a narrow
scope reading paraphrasable as It appeared that someone yawned and a wide scope
reading on which there is some person who appeared to yawn, as shown in (32):

(31) Someone appeared to yawn.

narrow scope interpretation = It appeared that someone yawned (but perhaps
no one is present).
wide scope interpretation = There is some person who appeared to yawn (but
that person may not have yawned).

This ambiguity depends on the scope of someone, whether it is inside or outside

appear.

4 Since the analysis of raising verbs does not depend on our theory of anaphoric binding, we
provide simple meaning constructors rather than the PCDRT-based meaning constructors presented
in Chapter 14.
 raising verbs and semantic composition 557

(32) Someone appeared to yawn.

Narrow scope: appear (a (x, person (x), yawn (x)))
Wide scope: a (x, person (x), appear (yawn (x)))

In the next section, we will see how both readings for this sentence are derived.

15.2.2 Raising verbs and meaning assembly

The sentence David seemed to yawn is associated with the following f-structure
and meaning constructor:

(33) David seemed to yawn.

 
 

 d 

s


 ’ 

 y 

seem (yawn (David )) : s

The presence of the subj argument outside the angled brackets in the matrix pred’s
semantic form in (33) indicates that the subj of seem is not a semantic argument
of the verb and that the sole semantic argument is the xcomp. In keeping with this
intuition, we follow Asudeh (2000, 2002a) in proposing the following lexical entry
for the verb seemed:

(34) seemed V (↑ pred) = ‘seem'xcomp(subj’

(↑ subj) = (↑ xcomp subj)
λP.seem (P) : (↑ xcomp)𝜎 ⊸ ↑𝜎

The meaning constructor in the last line of this lexical entry requires an xcomp
argument. A meaning contribution corresponding to the subj is not required, since
the subj meaning is not a semantic argument of seem . If the subj contributes a
meaning resource, it must be consumed by the xcomp predicate for the sentence
to be semantically complete and coherent.
We now instantiate the meaning constructor given in (34) according to the
f-structure labels in (33). We also provide instantiated meaning constructors for
the proper name David and the intransitive verb yawn:

(35) Meaning constructor premises for David seemed to yawn:

[seem] λP.seem (P) : y𝜎 ⊸ s𝜎
[David] David : d𝜎
[yawn] λx.yawn (x) : d𝜎 ⊸ y𝜎
558 functional and anaphoric control

The meaning constructor labeled [yawn] requires a meaning for its subject, d𝜎 ,
to produce a meaning for y𝜎 . Thus, we first combine the meaning constructors
[David] and [yawn] to produce the meaning constructor [David-yawn] for y𝜎 .

(36) [David-yawn] yawn (David ) : y𝜎

A meaning for y𝜎 is exactly what the meaning constructor [seem] requires, and a
semantically complete and coherent meaning constructor for the sentence results.

(37) [David-yawn], [seem] * seem (yawn (David )) : s𝜎

Meaning deduction from the premises contributed by the sentence Someone

appeared to yawn yields two different conclusions, since the example is ambiguous,
with both a narrow and a wide scope reading, as shown in (31):

(38) Someone appeared to yawn.

 
 

 s 

a


 ’ 

 y 

Narrow scope: appear (a (x, person (x), yawn (x))) : a

Wide scope: a (x, person (x), appear (yawn (x))) : a

The meaning constructor premises in (39) are relevant for this example. Again,
since we are examining the semantics of raising verbs and are not concentrating
on issues related to context update and anaphoric binding, we provide the simple
meaning constructor for the indefinite quantifier someone which was discussed in
§8.8, rather than the PCDRT constructor discussed in §14.4.3.7:

(39) Meaning constructor premises for Someone appeared to yawn:

[appear] λP.appear (P) : y𝜎 ⊸ a𝜎
[someone] λS.a (x, person (x), S(x)) : ∀H.[s𝜎 ⊸ H] ⊸ H
[yawn] λx.yawn (x) : s𝜎 ⊸ y𝜎

For the narrow scope reading, we note that the quantifier meaning constructor
[someone] requires as its argument a resource of the form s𝜎 ⊸ H for some
semantic structure H. The meaning constructor [yawn] provides such a resource.
Combining [someone] and [yawn], we have the meaning constructor labeled
[someone-yawn]:

(40) [someone-yawn] a (x, person (x), yawn (x)) : y𝜎

The meaning constructor in (40) provides a meaning resource y𝜎 , exactly

what the meaning constructor labeled [appear] in (39) requires. Combining
 raising verbs and semantic composition 559

[someone-yawn] and [appear], we obtain the semantically complete and coherent

meaning constructor in (41), yielding the narrow scope reading:

(41) [someone-yawn], [appear] * appear (a (x, person (x), yawn (x))) : a𝜎

To derive the wide scope reading for the example, we make use of the abstraction
rule given as (74) in Chapter 8. Recall that this rule permits the introduction of
a hypothetical premise on the glue side, which is discharged at a later point in
the deduction; on the meaning side, hypothetical premise discharge corresponds
to abstracting over the variable introduced by the premise. For this example, we
hypothesize the premise x : [s𝜎 ] in the first line:

(42) x : [s𝜎 ] [yawn]

yawn (x) : y𝜎 [appear]

appear (yawn (x)) : a𝜎

[appear-yawn] λx.appear (yawn (x)) : s𝜎 ⊸ a𝜎

As shown in (42), we combine the hypothesized premise x : [s𝜎 ] with the premise
[yawn], producing the meaning constructor yawn (x) : y𝜎 . This meaning construc-
tor provides the semantic resource y𝜎 required by [appear]. Combining these two
meaning constructors, we produce the meaning constructor appear (yawn (x)) : a𝜎 .
Finally, the hypothesized premise x : [s𝜎 ] is discharged in the last line of (42),
producing the meaning constructor labeled [appear-yawn]: the variable x is
abstracted over on the left-hand side, producing the function λx.appear (yawn (x)),
and the implicational meaning constructor s𝜎 ⊸ a𝜎 is produced on the right-hand
side.
The meaning constructor [appear-yawn] provides a resource of the form
s𝜎 ⊸ H, which is what the quantifier [someone] requires; for this reading, the
semantic structure a𝜎 is chosen to provide the scope meaning of the quantifier.
Combining the meaning constructors [someone] and [appear-yawn], we obtain
the semantically complete and coherent meaning constructor in (43), which
provides the wide scope reading for this example:

(43) [appear-yawn], [someone] * a (x, person (x), appear (yawn (x))) : a𝜎

15.2.3 Copy raising

Copy raising is closely related to the “standard” instances of raising discussed
above, but differs from standard raising in interesting ways. Extensive investigation
of the syntax and semantics of copy raising has been undertaken by Asudeh (2002b,
2004, 2005, 2012) and Asudeh and Toivonen (2006, 2007, 2012). The following
examples of copy raising are taken from Asudeh and Toivonen (2006):

(44) a. They seem like they’ve missed the bus.

b. John appears as if he is tired.
 560 functional and anaphoric control

Such sentences are parallel in many ways to sentences involving raising from
infinitival complements: (44a) is similar to They seem to have missed the bus, and
(44b) is parallel to John appears to be tired. The crucial similarity between the
copy raising sentences and the standard raising sentences is that the subject of the
matrix verb is not semantically selected, as is evident from sentences like There
seems like there’s a problem; the crucial difference is that there is a pronoun in the
embedded clause corresponding to the controlled argument of the standard raising
sentences (the “copy”). Thus in contrast to standard raising, in copy raising there
are two different noun phrases corresponding to a single semantically selected
argument position. This leads to a semantic difficulty involving resource surplus:
the pronoun subject of the embedded predication makes a contribution to the set
of meaning resources available for semantic composition, but that contribution is
neither required nor usable, because its role is already satisfied by the subject of
the matrix predicate.
In terms of the syntactic analysis of copy raising, Asudeh and Toivonen propose
that the apparent complementizers in English copy raising, e.g. like in (44a) and as
if in (44b), are not in fact complementizers but rather prepositions which select
for two arguments, a subj and a closed complement comp. The subject of the
preposition is controlled by the subject of the matrix raising verb, just as in the
standard raising examples, meaning that we can use exactly the same lexical entry
and functional descriptions for the seem that occurs in copy raising as for the seem
that occurs in standard raising:

(45) They seem like they’ve missed the bus.

IP

NP I
(↑ ↓ ↑=↓ ’

N VP
↑=↓ ↑=↓ ’

V
(↑ ↑=↓

V PP
↑=↓ (↑ ↓

seem
P
(↑
↑=↓
(↑ ↑
P IP
↑=↓ (↑ ↓

like
(↑ ’ they’ve missed the bus
 closed complements and anaphoric control 561

In terms of semantic analysis, the main difficulty is the apparent duplication of

argument resources, since in (44a) there are two instances of they corresponding
to a meaning seem (miss (they , bus )), in which there is only one instance of the
meaning of they. To account for this, Asudeh (2004) introduces a manager resource,
lexically associated with copy raising verbs, which effectively serves to remove
the superfluous pronoun meaning from the semantic composition. For reasons of
space we do not provide an illustration here; see §17.6.4 for discussion of manager
resources, and Asudeh (2004, 2005) for details of the analysis of copy raising.

15.3 Closed complements and anaphoric control

Anaphoric control contrasts with functional control in several interlinked ways.
The subordinate complement in an anaphoric control construction is the closed
function comp, not the open function xcomp. Some constraints on the controller
in an anaphoric control construction are similar to those in functional control,
but the nature of the control is different: the relation in anaphoric control is
semantically much closer to an anaphoric binding relation, and does not involve
syntactic identity.
Anaphoric control constructions are of two types: obligatory anaphoric control
and arbitrary anaphoric control (Bresnan 1982a; Zec 1987; Bresnan 2001c). In
an obligatory anaphoric control construction, coreference is required between an
argument of the matrix clause and the controlled argument in the subordinate
clause. In contrast, in an arbitrary anaphoric control construction, no coreference
constraints are imposed by the control verb. Instead, the controlled argument
in the subordinate clause finds its referent in a way very similar to an ordinary
pronoun, and split antecedents and syntactically remote controllers are possible,
as we will see.

15.3.1 Obligatory anaphoric control

Obligatory anaphoric control constructions were first examined in LFG by Bresnan
(1982a), who showed that an anaphor in an anaphoric control construction may be
assigned an antecedent by the rules of sentence grammar. Further exploration of
anaphoric control in Bosnian/Croatian/Serbian was carried out by Zec (1987). In
the following, we explore the syntax of obligatory anaphoric control; the semantics
of these constructions will be discussed in §15.4.
In this work, we adopt an anaphoric control analysis of English equi verbs.5
Under this analysis, English equi verbs exhibit obligatory anaphoric control of the
subj of a closed complement comp:

5 Equi verbs are so called because of their participation in the “Equi-NP Deletion” transformation
proposed for them by Postal (1970) and others in a transformational framework, in which an NP in
subject position of the subordinate clause of an equi verb was assumed to be deleted under conditions
of identity with an NP in the matrix clause.
562 functional and anaphoric control

(46) David tried to leave.

 
’
 
 
 
 
 ’ 
 
 

In (46), the subj of the obligatory anaphoric control verb tried anaphorically
controls the subj of the comp, and the sentence is interpreted as meaning that
David tried to bring about a situation where David leaves. The controller in an
anaphoric control construction can also be a matrix clause object if an appropriate
matrix verb appears:

(47) David convinced Chris to leave.

 
’
 
 
 
 
 
 
 
 ’ 
 
 

Here, the controller of the subj of leave is the obj of the matrix verb, Chris, and the
sentence means that David convinced Chris that Chris should leave.
It is not necessarily the case that if the matrix control verb has an obj, that
argument must be the controller: with verbs such as promise it is possible for the
matrix subj to control the subj of its comp, even though there is also a matrix obj:

(48) David promised Chris to leave.

 
’
 
 
 
 
 
 
 
 ’ 
 
 
 closed complements and anaphoric control 563

Here, the controller of the subj of leave is the subj of the matrix verb, David, and
the sentence means that David promised Chris that David would leave.6 We discuss
syntactic and semantic constraints on the controller in §15.7.

15.3.2 Anaphoric versus functional control

As we have already noted in passing in §15.1.3, obligatory anaphoric control is not
the only possible way of analyzing English equi verbs; in fact, many authors assume
a functional control analysis of equi verbs in English. It is clear that equi verbs differ
from raising verbs in a number of ways, but to what extent those differences are
reflected in the syntactic control structures assumed, and whether the differences
are better explained in semantic terms, is less clear.
Perhaps the clearest evidence for obligatory anaphoric control as contrasted with
functional control can be found in languages which show case mismatches between
the controller and controlled position in equi constructions. In an anaphoric
control construction, the anaphorically controlled subj of the subordinate comp
is syntactically independent from the matrix controller, although the two are
semantically related by an anaphoric binding relation. Thus, unlike the situation
with functional control, we do not expect syntactic restrictions imposed on the
controlled subj of the comp to be relevant for the matrix clause controller. As
Andrews (1982) shows for anaphoric control in Icelandic, the case restrictions
found in Icelandic functional control constructions are not found in constructions
involving anaphoric control.
As the sentences in (12) show, subjects in Icelandic can bear a case that is
idiosyncratically assigned by the verb. In constructions involving functional con-
trol, the case requirements of the subordinate clause verb must be satisfied, as
shown in (13). In contrast, in anaphoric control constructions in Icelandic, case
requirements of the subordinate clause verb do not apply to the matrix controller.
As discussed in §15.1.1, the Icelandic verb vanta ‘lack’ requires an accusative subj;
in (49b), the subject of the verb vanta ‘lack’ is interpreted as coreferent with the
subject of the equi verb vonast ‘hope’. The subject of the anaphoric control verb
vonast ‘hope’ is marked with nom case, not acc, even though the subordinate verb
vanta ‘lack’ requires its subject to be marked with acc case:

(49) a. Drengina vantar mat.

boys.def.acc lacks food.acc
‘The boys lack food.’
b. Ég vonast til að vanta ekki efni í ritgerðina.
I.nom hope to to lack not material for thesis.def
‘I hope to not lack material for the thesis.’

6 This use of the verb promise appears to be dialect-specific, and is not available for all speakers.
564 functional and anaphoric control

This contrasts with the situation with functional control, illustrated in (13), where
the case specified by the subordinate clause verb must appear on the matrix subject.
Since English lacks such case marking, the evidence for a functional distinction
between raising and equi verbs is less clear.7 As discussed in §15.1.1, Jacobson
(1990, 1992b) investigates certain syntactic differences between English raising
and equi verbs. One such difference is VP complement drop: the VP complement of
many equi predicates may be dropped, but this is impossible with raising predicates
(see also (14)):

(50) Equi:
a. [Did David really leave?] He tried.
b. [Will Chris leave?] If David can convince him.

(51) Raising:
'
to (leave).
a. [Did David really leave?] He seemed
⃰∅
'
to (vote).
b. [Does Chris vote?] David believes him
⃰∅

This is a lexically governed option, which is not possible for all equi verbs:
'
to (leave).
(52) a. [Did David really leave?] He wanted
⃰∅

One way of accounting for this difference is by reference to grammatical function:

closed VP comp arguments may be dropped where lexically licensed, but open VP
xcomp arguments may not be dropped. Alternatively, it may be possible to account
for the difference by reference to the major semantic difference between raising
and equi constructions: the controller in a raising construction is not semantically
selected by the raising verb, whereas the controller in an equi construction is
semantically selected by the equi verb. If the entire VP complement is dropped in a
raising construction, there is no overt signal of a predicate which can semantically
select for the controller. Thus in (51b), the only possible interpretation of David
believes him involves the transitive, non-raising version of believe, in which the obj
is semantically selected by the verb.
Evidence from topicalization is also unclear. Jacobson (1990, 1992b) claims that
while the complement of a raising verb cannot appear at the beginning of the
sentence, as shown in (15), examples in which the infinitival complement of an
equi verb appears in initial position are marginally acceptable:

7 Zaenen and Kaplan (2002) even suggest that case mismatches between controller and controllee
do not necessitate an anaphoric control analysis, since sharing of the case feature can be managed by
use of the restriction operator (§6.9.4). Such an approach is technically possible, but does not result in
a linguistically appealing analysis.
 closed complements and anaphoric control 565

(53) a. ?To leave, David (at least) tried.

b. ?To leave, David (at least) convinced Chris.

In fact, these examples are not acceptable for all speakers. Nevertheless, there does
seem to be a contrast between the examples with sentence-initial infinitives in (53)
and those in (15). The relatively low acceptability of the examples in (53) may in
part be due to the unsuitability of the infinitive in the pragmatic/information struc-
ture role associated with the sentence-initial phrase (Tracy Holloway King, p.c.).
In principle, then, equi verbs can be analyzed in terms of either functional or
anaphoric control, and it is sometimes assumed that equi constructions as well
as raising constructions in English involve functional control (see, for example,
Bresnan 1982a, 2001c; Asudeh 2000, 2002a; Falk 2001b; Bresnan et al. 2016).
However, the evidence presented above is consistent with the proposal that at
least some English equi verbs in fact participate in obligatory anaphoric control.
In the absence of unequivocal evidence in favor of the functional control analysis
for English equi verbs, we adopt the obligatory anaphoric control analysis in this
work. Falk (2001b) provides an illuminating discussion of the difference between
anaphoric and functional control in equi constructions, though the conclusions he
draws are different from those adopted here.
In other languages, more variation is found: some languages have two clearly
distinct types of equi verb, some specifying anaphoric control and some specifying
functional control. Kroeger (1993: Chapter 4) shows that Tagalog has two different
types of equi constructions. The first type involves functional control of a subject
argument in the complement clause:

(54) nagpilit si-Maria-ng bigy-an ng-pera ni-Ben

prf.av.insist.on nom-Maria-comp give-dat gen-money gen-Ben
‘Maria insisted on being given money by Ben.’
 
’
 
 
 
  

 ’ 

   
  
   
  
   
  
 

The second type involves anaphoric control of a term (possibly nonsubject) argu-
ment in the complement clause. In (55), the matrix subject Maria anaphorically
controls the objagent argument in the subordinate clause:

(55) nagpilit si-Maria-ng bigy-an ng-pera si-Ben

prf.av.insist.on nom-Maria-comp give-dat gen-money nom-Ben
‘Maria insisted on giving money to Ben.’
566 functional and anaphoric control
 
’
 
 
 
  

 ’ 

  
  
  
  
  
  
 

Thus, equi verbs that can specify either functional or anaphoric control can be
found, even within a single language.

15.3.3 Constituent structure and functional constraints

To analyze anaphoric control, we augment the phrase structure rule in (16) to allow
for VP complement daughters bearing the comp function:

(56) V& −→ V NP VP
↑ =↓ (↑ obj) = ↓ (↑ xcomp|comp ) = ↓

The rule in (56) differs from the one in (16) in allowing the VP daughter of V&
to bear either the xcomp or the comp function. There is no constituent structure
distinction between VP complements that are functionally controlled, bearing the
xcomp function, and those that are anaphorically controlled and bear the comp
function; xcomp and anaphorically controlled comp appear in the same position
relative to adverbs and direct objects, for example:

(57) a. The students seem clearly to be intelligent. (xcomp)

b. The students tried hard to be on time. (comp)

(58) a. The students believed David to have left. (xcomp)

b. The students convinced David to leave. (comp)

Thus, it is only the functional annotations on the rule that distinguish the two cases,
and not the c-structure configuration.
We propose the following syntactic lexical entries for the English equi verbs tried
and convinced:
 closed complements and anaphoric control 567

(59) tried V (↑ pred) = ‘try'subj,comp(’

(↑ comp subj pred) = ‘pro’

(60) convinced V (↑ pred) = ‘convince'subj,obj,comp(’

(↑ comp subj pred) = ‘pro’

The last line in each of these lexical entries specifies a pronominal subj for the
subordinate comp. In §15.4, we discuss how the anaphoric control relation is
established between the matrix controller and the subordinate clause subj.
These lexical entries, together with the annotated phrase structure rule in (56),
give rise to the anaphoric control structures in (61) and (62):

(61) David tried to leave.

IP

NP I  
’
(↑ ↓ ↑=↓  
 
 
N VP  
 ’ 
↑=↓ ↑=↓  
 

David V
(↑ ↑=↓

V VP
↑=↓ (↑ ↓

tried
V
(↑ ’
↑=↓
(↑
V VP
↑=↓ ↑=↓

V
to
↑=↓

leave
(↑ ’
568 functional and anaphoric control

(62) David convinced Chris to leave.

 
’
 
 
IP  
 
 
NP I  
 
 ’ 
N VP  
 
David V

V NP VP

convinced N V

Chris V VP

to V

leave

15.4 Equi verbs and semantic composition

As discussed in §15.3.2, we adopt the position that English equi verbs exemplify
anaphoric control, in contrast with English raising verbs, which exhibit functional
control. Equi and raising verbs differ in other ways as well: unlike raising verbs, the
controller in an equi construction is semantically as well as syntactically selected
by the verb. Notationally, this is reflected in the fact that the subj of an equi verb
such as try appears inside rather than outside the angled brackets in the semantic
form (see §2.4.6.1):

(63)  ’

 
 
 
 
 ’ 
 
 

Since the controller is semantically selected, equi verbs cannot appear with a
subordinate comp verb that selects an expletive or semantically empty subject,
since expletive subjects contribute no semantic content and therefore cannot
satisfy the semantic requirements of the equi predicate or enter into an anaphoric
dependency:

(64) ⃰There tried to be a problem.

 equi verbs and semantic composition 569

15.4.1 Semantics of equi verbs

Since equi verbs semantically as well as syntactically select for their controller argu-
ment, we treat an equi verb like try as denoting a two-place predicate, following
Halvorsen (1983), Reyle (1988), and many others. The meaning for the sentence
David tried to leave is therefore:

(65) David tried to leave.

try (David , leave (David ))

In (65), try denotes a relation between an individual David and the proposition
leave (David ) that David is trying to bring about. The equi verb tried requires
the pronominal subject of the subordinate verb leave to be coreferent with the
controller subject of tried, David .
Asudeh (2000, 2002a) proposes an analysis of English equi verbs within the
glue framework that differs from the analysis presented here in its assumptions
about the syntax and meaning of the complement of the equi verb. Asudeh analyzes
equi verbs as taking xcomp complements, not comp; further, following Chierchia
(1984a,b, 1985), Asudeh claims that the xcomp of an equi verb denotes a property
rather than a proposition. On this view, the meaning of a sentence like David tried
to leave is represented as:

(66) David tried to leave.

try (David , λx.leave (x)) (Chierchia 1984a, 1985; Asudeh 2000, 2002a)

In (66), try denotes a relation between an individual and the property that the
individual aspires to have. Chierchia (1984a) argues for the property theory of
control on the basis of entailments like this one:

(67) Nando tries anything/whatever Ezio tries.

Ezio tries to jog at sunrise.
Nando tries to jog at sunrise.

On Chierchia’s view, the validity of this argument is reflected in the following

logical paraphrases:

(68) ∀x.try (Ezio , x) → try (Nando , x)

try (Ezio , λx.jog.at.sunrise (x))
try (Nando , λx.jog.at.sunrise (x))

Chierchia argues that if the complement of the verb try is treated as a proposition
roughly corresponding to the meaning of a sentence like Ezio jogs at sunrise, an
unwanted entailment seems to follow, namely that Nando tries for Ezio to jog at
sunrise:
570 functional and anaphoric control

(69) ∀x.try (Ezio , x) → try (Nando , x)

try (Ezio , jog.at.sunrise (Ezio ))
try (Nando , jog.at.sunrise (Ezio ))

However, there are several difficulties with the property theory of control advo-
cated by Chierchia and Asudeh.8
First, Chierchia’s proposal is based on the lack of availability of a strict reading
for the conclusion of the argument in (67): the conclusion means that Nando tries
for Nando (the sloppy reading) and not Ezio (the strict reading) to jog at sunrise.9
However, other cases of sloppy-only readings for arguments similar to (67) are
not amenable to a solution that, like Chierchia’s, relies on the property theory of
control. For example, the argument in (70) is also valid:

(70) Nando does anything/whatever Ezio does.

Ezioi broke hisi arm playing football.
Nandoj broke hisj arm playing football.

However, there is no obvious way in which the canonical representation of the

meaning of the sentence Ezio broke his arm playing football can be adjusted to
predict the validity of the entailment in (70). Thus, whatever means accounts for
the sloppy-reading entailment in (70) can presumably account for the validity of
the argument in (67) without assuming the property analysis of control.
Additionally, Higginbotham (1992) points out that in at least some cases, argu-
ments analogous to the one in (67) do not provide evidence either for or against
Chierchia’s position. Higginbotham argues that the entailment in (71) should not
be taken as a linguistic fact about the verb practice:

(71) Nando does anything/whatever Ezio does.

Ezio practices playing the piano.
Nando practices playing the piano.

Coreference between the subject of the verb practice and the understood subject of
the gerund playing is enforced by real-world constraints on situations of practicing
the piano. Regardless of whether the gerund complement of practice is taken to be
a property or a proposition, it makes no sense to talk about Nando practicing Ezio’s
playing the piano. Therefore, Higginbotham argues, examples like this do not shed
light on the issue of the type of the complement of equi verbs.

8 Other authors to have defended the property theory of control include Chierchia and Jacobson
(1985), Dowty (1985), and Jacobson (1990).
9 The distinction between strict and sloppy readings is best known from analyses of ellipsis (see
Dalrymple et al. 1991 and references cited there); in the following, (b) paraphrases the sloppy reading
of (a), and (c) paraphrases the strict reading:
(a) David rode his bike, and Chris did too.
(b) Chris rode Chris’s bike. (sloppy)
(c) Chris rode David’s bike. (strict)
 equi verbs and semantic composition 571

Higginbotham (1992) also discusses examples such as:

(72) They expected to sit next to each other.

As Higginbotham points out, this sentence has two readings, paraphrasable in the
following way:

(73) a. Each of them expects to sit next to the other one.

b. They expect: they will sit next to each other.

On the proposition theory of control adopted here, both of these readings are
readily available (we use the notation of Dalrymple et al. 1998 to represent
reciprocal meaning):

(74) a. Each of them expects to sit next to the other one:

RECIP (they , λx.λy.expect (x, sit.next.to (x, y)))
b. They expect that they will sit next to each other:
expect (they , RECIP (they , λx.λy.sit.next.to (x, y)))

The reading paraphrased in (73a) is also readily available on the property theory:

(75) Each of them expects to sit next to the other one:

RECIP (they , λx.λy.expect (x, λz.sit.next.to (z, y)))

However, the reading paraphrased in (73b) is difficult to account for on the

property theory. It might be thought that the representation in (76) corresponds
to the desired interpretation:

(76) They expect that they will sit next to each other:
expect (they , λx.RECIP (x, λz.λy.sit.next.to (z, y)))

The problem with this representation is that a predicate like expect denotes a
relation between an individual and the property that the individual expects to have.
However, an individual cannot enter into a relationship involving a RECIP predicate,
which must hold of a group and not an individual. These conflicting requirements
make a coherent account of this reading difficult or impossible to obtain under the
assumptions of the property theory.
In addition to these arguments, Higginbotham (1992) presents a number of
other arguments supporting the proposition theory of control; we conclude with
Higginbotham that the proposition theory has clear advantages over the property
theory of control, and we adopt the proposition theory here.10

10 For further arguments in favor of the proposition theory of control, see Sag and Pollard (1991)
and Pollard and Sag (1994).
572 functional and anaphoric control

15.4.2 Equi and obligatory anaphoric control

Since we treat the subject of the comp argument of an equi verb like tried as
an unexpressed pronoun whose antecedent is the controller, we reintroduce the
PCDRT-based meaning constructors used in our analysis of anaphoric binding and
the syntax-semantics interface in Chapter 14. We assume the following f-structure
and meaning for the sentence David tried to leave:

(77) David tried to leave.

 
’
 

 d 

t



 ’ 
 l 
p

x1
David (x1 )
x2 , A(x2 ) = x1
try (x1 , )
leave (x2 )

As in Chapter 14, we assume that both the matrix subject David and the unex-
pressed pronominal subject of the complement verb leave contribute a discourse
referent. The unexpressed pronoun contributes an anaphoric discourse referent,
whose antecedent is constrained to be the subject of try. The contribution of a
discourse referent for the subject of the complement clause allows the resolution
of bound anaphors in examples like (78), where the antecedent of the reflexive
pronoun himself is the unexpressed pronominal subject of vote:

(78) David tried to vote for himself.

We propose this lexical entry for the equi verb tried.

(79) tried V (↑ pred) = ‘try'subj,comp(’

(↑ comp subj pred) = ‘pro’
R((↑ comp subj)𝜎 index) = ((↑ subj)𝜎 index)

λx.λP. : (↑ subj)𝜎 ⊸ [(↑ comp)𝜎 ⊸ ↑𝜎 ]

try (x, P)

x1
λP. ⊕ P(x1 ) : ∀H.((↑ comp subj)𝜎 ⊸ H) ⊸ H
 equi verbs and semantic composition 573

The first and second lines of this lexical entry are familiar from our syntactic
discussion in §15.3.3. The third line establishes the antecedency of the pronominal
subject of the comp argument: semantically, the antecedent of the comp subj is the
matrix clause controller, the subj of tried. As discussed in §14.4.1, this is modeled
in terms of the antecedency relation R between the values of index in the semantic
structures projected from the respective f-structures. In (80), we instantiate this
constraint according to the f-structure labels in (77):

(80) R(p𝜎 index) = (d𝜎 index)

The meaning constructor in the fourth line of the lexical entry, which is labeled
[try] in (81), provides the main predicate try . This PCDRT-based meaning con-
structor is parallel to that for arrive, discussed in §14.4.3.4, except that in addition
to the requirement for the meaning of the subj, it also requires a meaning for the
comp argument, which corresponds to the proposition P on the meaning side.
The final line of the lexical entry is the pronominal meaning constructor
supplied by the equi verb for the interpretation of the pronominal subj of its
complement clause, labeled [pro] in (81). This meaning constructor is similar to
the meaning constructor for he discussed in §14.4.3.6, except that, since it is asso-
ciated with a null pronoun, it imposes no person, number, or gender requirements
on the antecedent: in particular, it lacks the requirement for the antecedent to
be male.
The instantiated meaning constructors in (81) are relevant in the analysis of (77):

(81) Meaning constructor premises for David tried to leave:

[try] λx.λP. : d𝜎 ⊸ [l𝜎 ⊸ t𝜎 ]

try (x, P)
x1
[David] λP. ⊕ P(x1 ) : ∀H. (d𝜎 ⊸ H) ⊸ H
David (x1 )

[leave] λx. : p𝜎 ⊸ l𝜎
leave (x)
x1
[pro] λP. ⊕ P(x1 ) : ∀H. (p𝜎 ⊸ H) ⊸ H

We first combine [pro] with [leave], producing the meaning constructor

[pro-leave]:

x1
(82) [pro-leave] : l𝜎
leave (x1 )
574 functional and anaphoric control

We also combine [David] with the meaning constructor [try] to produce the
meaning constructor [David-try]:11

x1
(83) [David-try] λP. David (x1 ) : l𝜎 ⊸ t𝜎
try (x1 , P)

The meaning constructor [David-try] in (83) requires a meaning resource for l𝜎 ,

which is provided by [pro-leave]. Combining [David-try] and [pro-leave], we
have the meaning constructor labeled [David-try-pro-leave]:12

x1
David (x1 )
(84) [David-try-pro-leave] x2 : t𝜎
try (x1 , )
leave (x2 )

We have now produced a complete meaning for (77), but the anaphoric discourse
referent x2 still requires an antecedent. According to the constraint in the third
line of (79), the function R applied to the s-structure index feature corresponding
to the pronoun finds the index feature corresponding to David . Thus, we can
augment the meaning constructor in (84) with the correct antecedency relation:

x1
David (x1 )
(85) [David-try-pro-leave] x2 : t𝜎 , R(2) = 1
try (x1 , )
leave (x2 )

15.4.3 Equi and functional control

In §15.3.2, we saw that some equi verbs in Tagalog involve functional rather than
anaphoric control. Example (54), repeated here, exemplifies functional control:

11 As defined in §8.8.1.1, the variable H ranges over semantic structures. In order to apply
the meaning [David] to the meaning [try], it is therefore necessary first to hypothesize (§8.8.1.4)
meanings for the subj and comp of try, and then to discharge the hypothesized subject meaning,
resulting in a meaning: [try-P] λx. try (x, P) : d𝜎 ⊸ t𝜎 . The meaning [David] is then applied to
the meaning [try-P], following which the hypothesized meaning P is discharged, resulting in the
meaning [David-try]. Here, and elsewhere in this chapter, we pass over the details of such abstractions,
describing meanings like [David] as directly applying to meanings like [try].
12 Notice that the discourse referents in (82) have been renumbered to avoid a clash with those in
(83); see §14.4.3.6 for discussion.
 equi verbs and semantic composition 575

(86) nagpilit si-Maria-ng bigy-an ng-pera ni-Ben

prf.av.insist.on nom-Maria-comp give-dat gen-money gen-Ben
‘Maria insisted on being given money by Ben.’
’
m
’
i
g
f
b

insist (Maria , give (Ben , money , Maria )) : i

Since this example involves functional rather than anaphoric control, anaphora
resolution is not a central issue in the derivation of its meaning, and we use simple
meaning constructors rather than PCDRT constructors.
We assume that the xcomp complement of the equi verb nagpilit ‘insist on’
patterns with other equi verbs in denoting a proposition, though we have not been
able to verify the semantic patterns presented in §15.4.1 with a native speaker of
Tagalog. On this assumption, we propose the following lexical entry for nagpilit
‘insist on’:

(87) nagpilit V (↑ pred) = ‘insist'subj,xcomp(’

(↑ subj) = (↑ xcomp subj)
λP.λx.insist (x, P(x)) :
[(↑ subj)𝜎 ⊸ (↑ xcomp)𝜎 ] ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 ]

The first and second lines of this lexical entry enforce the syntactic constraints
appropriate for a subject raising verb. The last line gives the meaning con-
structor for this verb, which requires two arguments: an implicational resource
[(↑ subj)𝜎 ⊸ (↑ xcomp)𝜎 ] corresponding to its xcomp argument and a resource
(↑ subj)𝜎 corresponding to its subject. The xcomp resource takes the form of an
implication because it is an argument that is “missing” its subject (which is also
the subject of the main verb). The subject’s meaning is represented by x on the
left-hand side of the meaning constructor in (87); the xcomp’s meaning is the
property P, which is applied to the subject meaning x to produce the proposition
P(x) filling the second argument position of insist .
The meaning constructor in the lexical entry in (87) is labeled [insist] in (88),
instantiated with the f-structure labels given in (86); we also provide the standard
meaning constructors for the proper names Maria and Ben and for the ditransitive
verb bigy-an ‘give’. Since the internal structure of the noun phrase ng-pera ‘money’
is not at issue here, we make the simplifying assumption that this noun phrase
makes a contribution like that of a proper name and has the meaning money :
576 functional and anaphoric control

(88) Meaning constructor premises for nagpilit si-Maria-ng bigy-an ng-pera

ni-Ben:
[insist] λP.λx.insist (x, P(x)) : [m𝜎 ⊸ g𝜎 ] ⊸ [m𝜎 ⊸ i𝜎 ]
[Maria] Maria : m𝜎
[give] λx.λy.λz.give (x, y, z) : b𝜎 ⊸ [f𝜎 ⊸ [m𝜎 ⊸ g𝜎 ]]
[money] money : f𝜎
[Ben] Ben : b𝜎

We first combine the premises labeled [Ben], [money], and [give] to produce the
meaning constructor labeled [give-Ben-money]:

(89) [give-Ben-money] λz.give (Ben , money , z) : m𝜎 ⊸ g𝜎

This meaning constructor provides the implicational resource m𝜎 ⊸ g𝜎 required

by the meaning constructor labeled [insist]. On the left-hand side, we obtain
the corresponding meaning by applying the function λP.λx.insist (x, P(x)) to its
argument λz.give (Ben , money , z). Combining [give-Ben-money] and [insist], we
have the meaning constructor labeled [insist-give-Ben-money] in (90):

(90) [insist-give-Ben-money] λx.insist (x, give (Ben , money , x)) : m𝜎 ⊸ i𝜎

Finally, we combine this meaning constructor with the subject meaning construc-
tor, labeled [Maria]. The resulting meaning constructor, displayed in (91), provides
a semantically complete and coherent meaning for this example:

(91) [insist-give-Ben-money], [Maria] *

insist (Maria , give (Ben , money , Maria )) : i𝜎

15.5 Arbitrary anaphoric control

15.5.1 Syntax of arbitrary control
As we have argued above, English equi verbs such as try or persuade participate
in obligatory anaphoric control, where the referent of the controlled subordinate
clause subj is determined by constraints associated with the matrix verb. This
situation contrasts with constructions involving arbitrary control, in which the
reference of the pronominal element in the subordinate clause is not syntactically
determined. In an arbitrary control construction, the reference of the controlled
argument in the subordinate clause is resolved much like an overt non-reflexive
pronoun, as described by Bresnan (1982a) and Bresnan et al. (2016). Example
(92) involves arbitrary anaphoric control and means that David gestured for some
unspecified individual or individuals to follow Chris:
 arbitrary anaphoric control 577

(92) David gestured to follow Chris.

 
’
 
 
 
  

 ’ 

  
  
  
 

There are a number of differences between arbitrary control and functional or

obligatory anaphoric control. Bresnan (1982a) and Bresnan et al. (2016) show that
for arbitrary anaphoric control, though not for functional or obligatory anaphoric
control, a split antecedent (an antecedent that does not form a syntactic unit) is
possible. In (93), the subject of follow can be interpreted as the group consisting
of Chris and Matty, even though there is no single constituent representing this
group in the matrix clause:

(93) Chris told Matty that David had gestured to follow Ken.
[possible interpretation: Chris and Matty follow Ken]

A syntactically remote antecedent is also possible. In (94), the subject of follow can
be interpreted as coreferent with Chris, although the noun phrase Chris is not an
argument of the immediately higher clause:

(94) Chris thought that it was clear that David had gestured to follow Ken.
[possible interpretation: Chris follows Ken]

Additionally, there may be no expressed antecedent in the same sentence at all, as

in (92).
These semantic differences will emerge more clearly in the next section when
we discuss the interpretation of arbitrary anaphoric control constructions. Syn-
tactically, obligatory and arbitrary control constructions do not differ; the same
phrase structure rule, given in (56), covers both, and the syntactic portions of the
lexical entries are similar. We propose the following syntactic lexical entry for a
verb like gesture:

(95) gesture V (↑ pred) = ‘gesture'subj,comp(’

(↑ comp subj pred) = ‘pro’

Example (96) shows the c-structure and f-structure for the example David gestured
to follow Chris:
578 functional and anaphoric control

(96) David gestured to follow Chris.

 
’
 
 
 
  

 ’ 

  
  
  
 
IP

NP I
(↑ ↓ ↑ =↓

N VP
↑ =↓ ↑ =↓

David V
(↑ ↑ =↓

V VP
↑ =↓ ↑ =↓

gestured
V
(↑ ’
↑ =↓
(↑
V VP
↑ =↓ (↑ ↓

V
to
↑ =↓

V NP
↑ =↓ ↑ =↓

follow N
(↑ ’ ↑ =↓

Chris
(↑
 arbitrary anaphoric control 579

15.5.2 Semantics of arbitrary control

Consider this short discourse:

(97) Chris yawned. David gestured to leave.

The most likely interpretation of the sentence David gestured to leave in (97) is that
David gestured for Chris to leave:

(98) Chris yawned. David gestured to leave.

yawn (Chris ), gesture (David , leave (Chris ))

In other contexts, other antecedents are possible. Unlike the situation with obliga-
tory anaphoric control, lexical or syntactic constraints do not determine the refer-
ent of the complement subject. Instead, the pronominal subject of the complement
to leave obeys the same constraints on pronoun resolution as an overt pronoun.
However, lexical constraints can play an important role in constraining the range
of possible referents for the controller in an arbitrary control construction. In
(98), the matrix subject in the second sentence cannot corefer with the subject
of the comp: that is, a sentence like David gestured to leave cannot mean that
David gestured for himself to leave. This is a negative constraint on anaphoric
reference, which in this case is imposed by the verb gesture; negative constraints
were discussed in §§14.2.2 and 14.4.3.

15.5.3 Arbitrary control and meaning assembly

Since the unexpressed subject of the comp of the verb gestured is interpreted as
a pronoun whose antecedent must be resolved, we require a representation of the
context in the derivation of the meanings of the sentences in the discourse in (97),
and so we provide PCDRT-based meaning constructors in our semantic analysis.
The f-structure and meaning constructor for the first sentence are:

(99) Chris yawned.

/ 2
pred ‘yawn'subj(’
y 0 1
subj c pred ‘Chris’

x1
Chris (x1 ) : y𝜎
yawn (x1 )
580 functional and anaphoric control

We assume the premises in (100) for the analysis of this sentence:

(100) Meaning constructor premises for Chris yawned:

[context] : (y𝜎 context)

[context-mod] λP.λQ.P ⊕ Q : y𝜎 ⊸ (y𝜎 context) ⊸ (y𝜎 context)

[yawn] λx. : c𝜎 ⊸ y𝜎
yawn (x)
x1
[Chris] λP. ⊕ P(x1 ) : ∀H.(c𝜎 ⊸ H) ⊸ H
Chris (x1 )

The first two premises in (100) are identical to those presented in (87) and (88) of
Chapter 14 in the analysis of the sentence David arrived at the start of a discourse.
The derivation of the meaning of this sentence proceeds analogously. Combining
the premises in (100) produces the following semantically complete and coherent
meaning constructor, as desired:

x1
(101) [context], [context-mod], [yawn], [Chris] * Chris (x1 ) : (y𝜎 context)
yawn (x1 )

The second sentence in (97), David gestured to leave, has the following f-structure
and meaning constructor:

(102) David gestured to leave.

 
’
 

 d 

g



 ’ 
 l 
p

x1
David (x1 )
x2 :g
gesture (x1 , )
leave (x2 )

The lexical entry for the arbitrary control verb gestured is as follows:
 arbitrary anaphoric control 581

(103) gestured V (↑ pred) = ‘gesture'subj,comp(’

(↑ comp subj pred) = ‘pro’

R⃰((↑ comp subj)𝜎 index) -= R⃰((↑ subj)𝜎 index)

λx.λP. : (↑ subj)𝜎 ⊸ [(↑ comp)𝜎 ⊸ ↑𝜎 ]

gesture (x, P)
x1
λP. ⊕ P(x1 ) : ∀H.((↑ comp subj)𝜎 ⊸ H) ⊸ H

The first line of this lexical entry provides the pred value for the verb gestured.
The second line of the lexical entry specifies a ‘pro’ value for the pred attribute
of the complement clause’s subject. The third line introduces a constraint on its
antecedent:

(104) R⃰((↑ comp subj)𝜎 index) -= R⃰((↑ subj)𝜎 index)

As noted earlier, the subj of gesture cannot corefer with its comp subj: the sentence
David gestured to leave cannot mean that David gestured for himself to leave.
The constraint in (104) enforces this requirement by preventing the complement’s
subject’s semantic structure index ((↑ comp subj)𝜎 index) from being related to
the index of (↑ subj)𝜎 via the R⃰ relation.13
The fourth line in (103) is the meaning constructor that introduces the predicate
gesture , labeled [gesture] in (105); it is analogous to the meaning for try given
in (79).
The final line of the lexical entry in (103) provides a pronominal meaning
constructor for the comp subj of gesture, labeled [pro] in (105). Again, this
meaning constructor is identical to that found in the lexical entry of the equi verb
try in (79), and is similar to the meaning constructor for he discussed in §14.4.3.6,
except that it lacks the requirement that the antecedent be male.
With the context [context] produced by uttering the first sentence in the
discourse, Chris yawned, the premises in (105) are relevant for the deduction of
the meaning of David gestured to leave:

13 For discussion of negative binding constraints and the R⃰ relation, see §14.4.3.2.
582 functional and anaphoric control

(105) Meaning constructor premises for David gestured to leave:

x1
[context] Chris (x1 ) : (g𝜎 context)
yawn (x1 )
[context-mod] λP.λQ.P ⊕ Q : g𝜎 ⊸ (g𝜎 context) ⊸ (g𝜎 context)

[gesture] λx.λP. : d𝜎 ⊸ [l𝜎 ⊸ g𝜎 ]

gesture (x, P)
x1
[David] λP. ⊕ P(x1 ) : ∀H.(d𝜎 ⊸ H) ⊸ H
David (x1 )
x1
[pro] λP. ⊕ P(x1 ) : ∀H.(p𝜎 ⊸ H) ⊸ H

[leave] λx. : p𝜎 ⊸ l𝜎
leave (x)

We first combine [David] with [gesture] to produce this meaning constructor.

x1
(106) [David-gesture] λP. David (x1 ) : l𝜎 ⊸ g𝜎
gesture (x1 , P)

We also combine the meaning constructor [pro] with [leave] to produce

[pro-leave].

x1
(107) [pro-leave] : l𝜎
leave (x1 )

We can now combine [David-gesture] with [pro-leave], producing the meaning

constructor labeled [David-gesture-pro-leave]:14

x1
David (x1 )
(108) [David-gesture-pro-leave] x2 : g𝜎
gesture (x1 , )
leave (x2 )

Since David gestured to leave is a root sentence like Chris yawned, we must also
make use of the meaning constructor labeled [context-mod] in (105), reflecting
its role as a context modifier. Combining [context-mod] with the meaning con-
structor in (108) produces the context-modifying meaning constructor in (109):

14 As in (84), the discourse referents in (107) have been renumbered to prevent a clash with the
discourse referents in (106).
 partial (“quasi-obligatory”) control 583

(109) x1
David (x1 )
[David-gesture-pro-leave-mod] x2 : (g𝜎 context) ⊸ (g𝜎 context)
gesture (x1 , )
leave (x2 )

The discourse context for this sentence is the meaning constructor labeled
[context] in (105), which was derived in (101). The meaning constructor in (109)
functions as a modifier on this context, producing an updated context which
includes the contribution of the sentence David gestured to leave. Combining the
meaning [David-gesture-pro-leave-mod] with [context], and rewriting indices
as needed, we arrive at the following meaning for the discourse:

x1 , x2
Chris (x1 )
yawn (x1 )
(110) David (x2 ) : (g𝜎 context)
x3
gesture (x2 , )
leave (x3 )

Finally, we find a licit antecedent for the anaphoric discourse referent x3 . By the
negative constraint in the lexical entry for gesture, the discourse referent x2 is
not a licit antecedent for x3 . The discourse referent x1 is available to serve as the
antecedent of x3 , so the pragmatic resolution function R may apply to the index
corresponding to x3 to establish an anaphoric relation to the index of x1 :

x1 , x2
Chris (x1 )
yawn (x1 )
(111) David (x2 ) : (g𝜎 context), R(3) = 1
x3
gesture (x2 , )
leave (x3 )

15.6 Partial (“quasi-obligatory”) control

In the preceding sections, we have drawn a contrast between obligatory anaphoric
control, as exemplified by English equi verbs such as try, and arbitrary anaphoric
control, as exemplified by the verb gesture. However, this binary distinction does
not account for the full variety of anaphoric control possibilities, even in English.
As argued in detail by Haug (2013), the phenomenon of “partial control” is best
analyzed as a kind of “quasi-obligatory” anaphoric control, in which the con-
straints on the resolution of the controlled pronoun are looser than the constraints
584 functional and anaphoric control

involved in obligatory anaphoric control, but stronger than those involved in arbi-
trary anaphoric control. Partial control is exemplified in the following sentences
(Haug 2013):

(112) a. The Chairi preferred to gatheri+ at six.

b. Johni told Maryj that hei preferred to meeti+ at six today.
c. Mary wants to bomb Hanoi.

In (112a), there is an inclusion relation between the controlled argument and its
controller in the matrix clause: the subject of gather is a group, for example a
committee, which includes the Chair of that committee. The same is true in (112b):
the subject of meet must be a group of at least two individuals, which may or
may not include Mary, but must include John. In (112c), the situation is slightly
different: this sentence is felicitous in a situation where Mary herself does not want
to be directly involved in the bombing, but wants some salient and related referent
to bomb Hanoi, for example her country’s airforce.
Partial control has been investigated by a number of authors, including Landau
(2000, 2004) and Grano (2015). Within LFG, the first author to address this
phenomenon was Asudeh (2005), who proposed an analysis involving func-
tional control. Haug (2013) shows that a functional control analysis is not viable,
and provides a semantically-based anaphoric control analysis within the PCDRT
framework adopted in this work. Haug (2013, 2014a) proposes to account for
partial control by enriching the pragmatic resolution function A in such a way as to
license relations other than identity between anaphor and antecedent, for example
a relation of inclusion. Specifically, Haug proposes that the validity of the relation A
between anaphor and antecedent depends on an additional relation B : the relation
A is valid if and only if an appropriate B relation holds between the antecedent and
the anaphor. For example, we can represent the meaning and control resolution of
example (112a) as follows:

x1
chair (x1 )
(113) x2 , A(x2 ) = x1 ; B = λx.λy.chair(x, y)
prefer (x1 , )
gather.at.six (x2 )

Here, the antecedent relation holds between x2 and x1 , because the relation
λx.λy.chair(x, y) holds between x1 and x2 . In all of the examples discussed in the
previous sections, the A relation indicates coreference, and this can be modeled as a
default interpretation of B as λx.λy.x = y. The possibilities of anaphoric resolution,
which are significantly increased by the extension of B beyond coreference, are
constrained in different ways by the specific semantic properties of different
control verbs; for example, a constraint associated with prefer ensures that the
controlled argument in (112a) includes the logophoric center of the control verb,
in this case the subject.
the controller in anaphoric or functional control 585

Thus, the model of anaphoric control presented above is capable of capturing a

wide variety of constraints on the anaphoric resolution of the controlled argument,
from strict syntactic constraints in obligatory anaphoric control, to more flexible
semantic constraints in partial control, and relative freedom of resolution in the
case of arbitrary control.

15.7 The controller in anaphoric or functional control

Determination of the controller in functional and obligatory anaphoric control
is constrained by both syntactic and semantic factors. In both kinds of control
constructions, the controller must be a term (§2.1.3). Further, as discussed in detail
by Sag and Pollard (1991), the choice of controller in equi verbs is semantically
constrained; equi verbs can be divided into semantic classes, and the semantic role
of the controller can be predicted from these classes.

15.7.1 Syntactic requirements

Bresnan (1982a) presents the following constraints on determination of the con-
troller in control constructions:

(114) a. The controller must be a term (subj, obj, or obj𝜃 ).

b. By default, the controller is the lowest available argument on the gram-
matical function hierarchy subj > obj > obj𝜃 .

The first requirement is that the controller is required to be a term: either subj,
obj, or obj𝜃 . As demonstrated by Bresnan (1982a), this makes several strikingly
correct predictions.
First, it accounts for what is known as Visser’s Generalization (Visser 1963–1973;
Bresnan 1982a), according to which there is no passive version of a verb involving
subject control:15

(115) a. John promised Mary to be on time.

b. ⃰Mary was promised by John to be on time.

15 As discussed by van Urk (2013), impersonal passives such as the following constitute an exception
to this generalization:
(a) It was decided to leave.
(b) Es wurde versucht, Eichhörnchen zu fangen.
it was tried squirrels to catch
‘It was tried to catch squirrels.’
Such examples constitute an exception to Visser’s Generalization, but do not violate the constraint in
(114a): there is no explicit controller, so whatever the correct analysis of such sentences, they cannot
involve functional or obligatory anaphoric control.
586 functional and anaphoric control

(116) a. He strikes his friends as pompous.

b. ⃰His friends are struck by him as pompous.

Visser’s Generalization follows from the constraint in (114a) since an oblique or

adjunct phrase such as by John is not a term and therefore cannot participate as a
controller in a control construction.
Second, the constraint in (114a) accounts for what is known as Bach’s Gen-
eralization (Bach 1980): there is no detransitivized version of a verb involving
object control. Bresnan (1982a) presents the following illustrative examples, which
involve both functional and anaphoric control:

(117) Anaphoric control by subject:

a. Louise promised Tom to be on time.
b. Louise promised to be on time.

(118) Anaphoric control by object:

a. Louise taught Tom to smoke.
b. ⃰Louise taught to smoke.

(119) Functional control by object:

a. Louise believed Tom to smoke.
b. ⃰Louise believed to smoke.

Bach’s Generalization follows from the constraint in (114a), since the controller
must be present as a term argument in a control construction.16 Of course, Bach’s
Generalization does not apply to arbitrary anaphoric control constructions, since
there are no syntactic constraints on the determination of the controller of the
subject of the subordinate comp in an arbitrary anaphoric control construction:

(120) a. Louise signaled Tom to follow.

b. Louise signaled to follow.

The requirement in (114b) involves a violable syntactic hierarchy of default con-

trollers in control constructions: Bresnan (1982a) claims that the unmarked choice

16 Many verbs which show object control entirely lack a detransitivized version, even when they have
no controlled complement:
(a) Louise believed ⃰(Tom) regarding his accident.
(b) Louise convinced ⃰(Tom) of her innocence.
Nevertheless, some such verbs do permit detransitivized versions outside control contexts, supporting
the validity of Bach’s Generalization:
(c) Louise taught (them) about anaphoric binding.
 control in adjuncts 587

for a controller is obj𝜃 if there is one, otherwise obj if there is one, and otherwise
subj. The following control constructions obey this rule:

(121) a. David tried to leave.

b. David persuaded Chris to leave.

Verbs like promise constitute exceptions to this default, since the subj and not the
obj is the controller:

(122) Chris promised David to leave.

Both of the generalizations in (114) are best thought of as constraining the deter-
mination of grammatical functions for verbs involving functional or anaphoric
control; for discussion of the theory of mapping between semantic roles and
grammatical functions, see Chapter 9.

15.7.2 Semantic requirements

In addition to syntactic constraints on how the controller is realized, there are also
semantic generalizations about controller choice in constructions involving equi
verbs (which, as shown above, can involve either functional or anaphoric control).
Sag and Pollard (1991) and Pollard and Sag (1994) provide a detailed exploration
of different classes of equi verbs and propose semantic principles for determination
of the controller in equi constructions. For example, they propose that verbs such
as order, persuade, and bid are members of what they call the order/permit class of
equi verbs. Verbs of this class refer to situations where a participant is influenced
by another participant to perform a certain action, and the controller is always
the obj argument of the active equi verb. This class contrasts with the promise
class, containing verbs like promise, agree, and demand, in which the controller
is always the subj of the active equi verb. Thus, generalizations about linking in
equi verbs—how the semantic arguments of these verbs are linked to the syntactic
functions subj, obj, obj𝜃 , and comp, as described in Chapter 9—are formulated
with reference to these larger classes, not separately specified for each verb. Within
LFG, the shared properties of such classes are captured by the use of templates
(§6.7). Sag and Pollard’s useful classifications are used in analyses of control in
Balinese by Arka (1998) and Arka and Simpson (1998). As discussed in §15.6,
semantic constraints are also relevant to the analysis of partial control, as shown
by Haug (2013, 2014a).

15.8 Control in adjuncts

Thus far, we have examined functional and anaphoric control involving the closed
grammatical function comp and the open function xcomp. Modifiers may also
participate in either functional or anaphoric control. The open function xadj
588 functional and anaphoric control

contains an open position which is functionally controlled by an argument of

the matrix clause, and some modifying adjuncts with the closed function adj
participate in anaphoric control. Mohanan (1983) discusses control in modifier
phrases, showing that control relations must be defined in functional and not
phrasal terms: English and Malayalam are quite different in constituent structure,
but obey similar functional constraints in control constructions.

15.8.1 Functional control and XADJ

The open adjunct function xadj is similar to the open function xcomp in partici-
pating in functional control. The subj of the xadj is identified with an argument of
the matrix clause, and the same f-structure fills both functions. Andrews (1990a)
discusses the following Icelandic examples involving control of an xadj:17

(123) Njósnaranum var kastað einum út úr þyrlunni.

spy.def.dat was thrown alone.dat out from helicopter.def
‘The spy was thrown out of the helicopter alone.’

(124) Ég mœtti Sveini drukknum.

I met Svein.dat drunk.dat
‘I met Svein drunk.’

In (123) and (124), the dative form of the open xadj adjuncts einum ‘alone’ and
drukknum ‘drunk’ appears. Andrews (1990a) shows that the subjects of these open
adjuncts are functionally controlled by a term argument of the matrix verb. In
particular, in (124) the object Sveini ‘Svein’ appears in dative case, as the verb mœtti
‘met’ requires, and the subj of the dative adjunct drukknum ‘drunk’ is controlled
by the matrix dative obj Sveini.
We propose the rule in (125) for xadj in Icelandic, which allows for the xadj
phrase drukknum ‘drunk’ to be adjoined to a VP:18

(125) VP −→ VP AP ⃰
↑=↓ ↓ ∈ (↑
(↑ | | }) = (↓

According to this rule, the subj of the open adjunct xadj is identified with a term
argument of the matrix clause: a subj, obj, or obj𝜃 . We also propose the lexical
entry in (126) for the adjective drukknum ‘drunk’, which requires a dative subject:

17 Andrews (1990a) attributes these examples to Rögnvaldsson (1984).

18 It is unclear whether the xadj is adjoined to VP in Icelandic or to some higher c-structure position.
In either case, the relevant functional annotations are as shown in (125).
 control in adjuncts 589

(126) drukknum A (↑ pred) = ‘drunk'subj(’

(↑ subj case) = dat

The case constraint imposed by drukknum is satisfied by the dative noun phrase
Sveini:

(127) Sveini N (↑ pred) = ‘Svein’

(↑ case) = dat

The c-structure and f-structure for (124) are:

(128) Ég mœtti Sveini drukknum

I met Svein.dat drunk.dat
‘I met Svein drunk.’
 
’
   
 
   
   
 
 
 
IP  
 
 
NP I  
 
 
N I S  
 ’ 
 
 
Ég mœtti VP
I met
AP
VP ↓ ∈ (↑
(↑ | | }) = (↓

NP A

N drukknum

Sveini

15.8.2 Open adjuncts and semantic composition

Examples of functional control in constructions involving xadj are also
found in English. We follow Bresnan (1982a) in assuming that, in a sen-
tence like Walking the dog, Chris saw Tracy, the subj of the participial
adjunct walking the dog is functionally controlled by the subj of the matrix
clause Chris. We propose the following representation of the meaning of this
example:
590 functional and anaphoric control

(129) Walking the dog, Chris saw Tracy.

during (walk (Chris , dog ), see (Chris , Tracy ))

Like other adjuncts, an open adjunct xadj such as walking the dog combines with
the clause it modifies, producing a new modified meaning of the same semantic
type. In (129), the predicate during relates the interval at which the subordinate
clause event occurs to the interval at which the main clause event occurs, and the
sentence means that during the interval at which Chris was walking the dog, the
event of Chris seeing Tracy occurred.19 The f-structure and meaning constructor
for (129) are:

(130) Walking the dog, Chris saw Tracy.

 
’
 

 c 

 

 t 

s
    
 
 ’ 
    
 w  
    
 
 

d

during (walk (Chris , dog ), see (Chris , Tracy )) : s

What linguistic element contributes the during predicate in this example? As

discussed in §8.6, meanings can be contributed not only by the words of a sentence,
but also by certain syntactic configurations. In this case, the information that the
event of Chris seeing Tracy occurred during the time at which Chris was walking
the dog is not contributed by any of the words in either the subordinate or the main
clause. Instead, the meaning is contributed by the phrase structure configuration
associated with this construction: the meaning constructor associated with the
during predicate appears on the c-structure rule associated with functionally
controlled xadj adjuncts. In other languages, this meaning might be expressed
constructionally (as it is in English) or by morphological or lexical means.

19 This is a simplification of what is in reality a more complex temporal-aspectual relation between

the event of the adjunct participle clause and that of the matrix clause. In particular, the interpretation
of the relation as ‘during’ holds only when the adjunct’s predicate is atelic: contrast the sentence in (129)
with e.g. Opening the book, Chris read the title page, where the reading event follows the opening event.
In this section, we consider only atelic adjuncts. The semantics of tense and aspect is discussed in more
detail in §8.10; for an event-semantic approach to participial adjuncts which can more precisely capture
the temporal-aspectual relations involved, see Haug (2008b), Bary and Haug (2011), and Lowe (2015c).
 control in adjuncts 591

The annotated c-structure tree for (129) is:

(131) Walking the dog, Chris saw Tracy.

IP

VP
↓ ∈ (↑ xadj)
IP
(↑ subj) = (↓ subj)
[during-xadj]

V& NP I&

V NP N VP

Walking Chris V&

the dog
V NP

saw N

Tracy

The rule that gives rise to this c-structure is:

  9 :
(132) IP −→ VP IP
 ↓ ∈ (↑ xadj) 
  ↑=↓
 
(↑ subj) = (↓ subj)
[during-xadj]

The VP in the rule in (132) has three annotations that are crucial for our current
discussion. The set-membership expression ↓ ∈ (↑ xadj) requires the f-structure
for the VP to appear as a member of the xadj set of the mother IP. The equation
(↑ subj) = (↓ subj) means that the subj of the xadj phrase walking the dog is
the same as the subj of the matrix clause. The third annotation, [during-xadj],
abbreviates the following meaning constructor:

(133) [during-xadj] λP.λQ.λx.during (P(x), Q(x)) :

[(↓ subj)𝜎 ⊸ ↓𝜎 ] ⊸ [[(↑ subj)𝜎 ⊸ ↑𝜎 ] ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 ]]

Instantiating the variables in this meaning constructor according to the f-structure

labels in (130), we have the instantiated meaning constructor for which we reuse
the label [during-xadj]:

(134) [during-xadj] λP.λQ.λx.during (P(x), Q(x)) :

[c𝜎 ⊸ w𝜎 ] ⊸ [[c𝜎 ⊸ s𝜎 ] ⊸ [c𝜎 ⊸ s𝜎 ]]
592 functional and anaphoric control

The right-hand side of this meaning constructor requires two implicational

resources, [c𝜎 ⊸ w𝜎 ] and [c𝜎 ⊸ s𝜎 ], to produce a meaning resource [c𝜎 ⊸ s𝜎 ].
The meaning resource [c𝜎 ⊸ w𝜎 ] represents a resource w𝜎 for the xadj that has
not yet combined with its subject c𝜎 . In other words, [c𝜎 ⊸ w𝜎 ] is a function from
the subject meaning c𝜎 to the xadj meaning w𝜎 . Similarly, [c𝜎 ⊸ s𝜎 ] represents a
main clause meaning “missing” its subject c𝜎 . When both of these resources are
found, a new resource [c𝜎 ⊸ s𝜎 ] is produced, reflecting a matrix clause meaning
that is modified by the participial adjunct walking the dog. On the left-hand side,
two arguments P and Q are required; each of these arguments is applied to the
subject meaning x to produce a modified meaning during (P(x), Q(x)) for the
entire sentence.
The meaning constructor premises in (135) are relevant in the derivation of
the meaning of this sentence; we have simplified the meaning contribution of the
phrase the dog, representing it simply as the constant dog .

(135) Meaning constructor premises for Walking the dog, Chris saw Tracy:
[walk] λx.λy.walk (y, x) : d𝜎 ⊸ [c𝜎 ⊸ w𝜎 ]
[dog] dog : d𝜎
[see] λx.λy.see (y, x) : t𝜎 ⊸ [c𝜎 ⊸ s𝜎 ]
[Chris] Chris : c𝜎
[Tracy] Tracy : t𝜎
[during-xadj] λP.λQ.λx.during (P(x), Q(x)) : [c𝜎 ⊸ w𝜎 ] ⊸ [[c𝜎 ⊸ s𝜎 ] ⊸ [c𝜎 ⊸ s𝜎 ]]

We begin by combining the premises labeled [walk] and [dog], producing the
meaning constructor labeled [walk-dog]:

(136) [walk-dog] λy.walk (y, dog ) : c𝜎 ⊸ w𝜎

This meaning constructor provides the resource c𝜎 ⊸ w𝜎 required by the meaning

constructor [during-xadj]. Combining [during-xadj] and [walk-dog], we have
the meaning constructor labeled [xadj-walk-dog]:

(137) [xadj-walk-dog]
λQ.λx.during (walk (x, dog ), Q(x)) : [[c𝜎 ⊸ s𝜎 ] ⊸ [c𝜎 ⊸ s𝜎 ]]

Next, we combine the premises labeled [see] and [Tracy], producing the meaning
constructor [see-Tracy]:
 control in adjuncts 593

(138) [see-Tracy] λy.see (y, Tracy ) : c𝜎 ⊸ s𝜎

This meaning constructor provides the resource needed by [xadj-walk-dog]. We

combine [xadj-walk-dog] with [see-Tracy] to produce the meaning constructor
labeled [xadj-walk-dog-see-Tracy]:

(139) [xadj-walk-dog-see-Tracy]
λx.during (walk (x, dog ), see (x, Tracy )) : c𝜎 ⊸ s𝜎

Finally, we combine the meaning constructor in (139) with the remaining meaning
constructor [Chris], producing the following semantically complete and coherent
meaning constructor, as desired:

(140) [xadj-walk-dog-see-Tracy], [Chris] *

during (walk (Chris , dog ), see (Chris , Tracy )) : s𝜎

15.8.3 Anaphoric control and ADJ

Some adjuncts participate in obligatory anaphoric control, where an unexpressed
pronominal argument of a clausal adjunct is anaphorically controlled by an argu-
ment of the matrix clause. Bresnan et al. (2016: Chapter 13) contrast participial VPs
in English, which they analyze via functional control, with gerundive VPs, which
are superficially similar but instead exemplify anaphoric control. Here we discuss
the Warlpiri kurra and karra constructions, both of which exemplify anaphoric
control. In both constructions, the subj of a subordinate adjunct clause with
complementizer kurra or karra is controlled by an argument of the matrix clause:
the matrix subj is the controller in the karra construction, and the matrix obj is
the controller in the kurra construction.
The sentence in (141) exemplifies the kurra construction, in which the obj of
the matrix clause anaphorically controls the subj of the adjunct clause:

(141) karnta ka-rla wangka-mi ngarrka-ku [ngurra-ngka-rlu

woman.abs prs-dat speak-npst man-dat camp-loc-erg
jarnti-rninja-kurra-(ku)]
trim-inf-comp-(dat)
‘The woman is speaking to the man (while he is) trimming it in camp.’
594 functional and anaphoric control

IP

NP I

N I S

karnta ka-rla V NP S

wangka-mi N NP V

ngarrka-ku N jarnti-rninja-kurra-(ku)

ngurra-ngka-rlu

Simpson and Bresnan (1983) provide evidence from case agreement to demon-
strate that the kurra construction involves anaphoric rather than functional control
and therefore that the modifying adjunct phrase ngurra-ngka-rlu jarnti-rninja-
kurra-(ku) ‘trimming it in camp’ is an adj and not an xadj. As they point out, (141)
contains an adjunct phrase ngurra-ngka-rlu ‘in camp’ which has erg casemarking
-rlu. Such adjunct phrases agree in case with the subj of the clause they modify.
Here, the phrase ngurra-ngka-rlu ‘in camp’ modifies the subordinate adjunct
clause jarnti-rninja-kurra-(ku) ‘trimming it’. Therefore, the erg casemarking on
the modifier ngurra-ngka-rlu ‘in camp’ shows that the subj of the subordinate
clause jarnti-rninja-kurra-(ku) ‘trimming it’ is also erg.
 control in adjuncts 595

However, the matrix obj phrase ngarrka-ku ‘man’, which anaphorically controls
the subordinate subj, is dat and not erg, in accordance with the case requirements
imposed by the matrix verb wangka-mi ‘speak’. This difference in case require-
ments between the matrix controller and the subordinate clause controllee shows
that the f-structures of the controller and the controllee are different, and that
anaphoric and not functional control is involved.

15.8.4 Controlled adjuncts and semantic composition

Like the kurra construction, the Warlpiri karra construction involves anaphoric
control of the subordinate clause subj by an argument of the matrix clause. How-
ever, Simpson and Bresnan (1983) show that the karra construction contrasts with
the kurra construction in its requirements on the controller-controllee relation: in
the karra construction, the subj of the adjunct clause is anaphorically controlled
by the matrix subj, not the matrix obj:

(142) ngarrka ka wirnpirli-mi [karli jarnti-rninja-karra]

man.abs prs whistle-npst boomerang.abs trim-inf-comp
‘The man is whistling while trimming a boomerang.’

In (142), the subj of the adjunct phrase is anaphorically controlled by the matrix
subj ngarrka ‘man’, as required by the affix karra on the subordinate clause verb
jarnti- ‘trim’: the example means that the man is whistling while he is trimming
a boomerang. Since this analysis involves anaphoric control, we represent the
meaning of (142) in DRS form, as follows:

x1 , x2 , x3

(143) , A(x3 ) = x1
during ( boomerang (x2 ) , man (x1 ) )
trim (x3 , x2 ) whistle (x1 )

The treatment of the predicate during in (143) is the same as in §15.8.2. It is the ver-
bal affix karra which contributes the meaning during , specifying (approximately)
that the interval corresponding to the event of trimming the boomerang occurs
during the interval corresponding to the whistling event. We assume a slightly
simplified representation of the noun phrase man, treating it as an indefinite (like
boomerang).
596 functional and anaphoric control

The f-structure for (142) is:

(144) ngarrka ka wirnpirli-mi [karli jarnti-rninja-karra]

man.abs prs whistle-npst boomerang.abs trim-inf-comp
‘The man is whistling while trimming a boomerang.’
 
’
 
 
 m 
 
 
     
w
  ’  

 

 
  
 
 

 t p   
  

 
  
 


 

b

The lexical entry for jarnti-rninja-karra ‘trim’ is as follows:20

(145) jarnti-rninja-karra V (↑ pred) = ‘trim'subj,obj(’

λy.λx. : (↑ obj)𝜎 ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 ]
trim (x, y)

λP.λQ. : ↑𝜎 ⊸ [(adj ∈ ↑)𝜎 ⊸ (adj ∈ ↑)𝜎 ]

during (P, Q)
(↑ subj pred) = ‘pro’
R((↑ subj)𝜎 index) = (((adj ∈ ↑) subj)𝜎 index)
[pro]

The first line of this lexical entry specifies the pred value for the verb jarnti-rninja-
karra ‘trim’, and the second line provides the trim predicate: like any transitive verb,
this verb requires a meaning for its obj and a meaning for its subj in order to
produce a meaning for the entire sentence.
The third line of this lexical entry represents part of the meaning contribution
of the affix karra, the during predicate:

(146) λP.λQ. : ↑𝜎 ⊸ [(adj ∈ ↑)𝜎 ⊸ (adj ∈ ↑)𝜎 ]

during (P, Q)

On the right-hand side, this meaning constructor requires a meaning resource ↑𝜎

(to be provided by the meaning constructor in the second line of the lexical entry
when its other requirements are satisfied) and a meaning (adj ∈ ↑)𝜎 for the clause
it modifies. When these are provided, a resource (adj ∈ ↑)𝜎 results, corresponding
to the modified main clause meaning. On the left-hand side, P corresponds to the

20 The expression (adj ∈ ↑) exemplifies inside-out functional uncertainty, discussed in §6.1.3; the
use of the set membership symbol ∈ as an attribute is discussed in §6.3.1.
 control in adjuncts 597

meaning of the adj clause, and Q corresponds to the unmodified meaning of the
main clause.
The fourth line of this lexical entry provides the f-structure ‘pro’ value
for the pred of the subj. The fifth line establishes the antecedency relation
between the s-structure index value of the subj of the adjunct clause,
((↑ subj)𝜎 index), and the s-structure index value of the subj of the matrix
clause, (((adj ∈ ↑)subj)𝜎 index):

(147) R((↑ subj)𝜎 index) = (((adj ∈ ↑) subj)𝜎 index)

Finally, the last line represents a pronominal meaning constructor [pro] for the
subj of jarnti-rninja-karra ‘trim’, defined as follows:

x1
(148) [pro] λP. ⊕ P(x1 ) : ∀H.((↑ subj)𝜎 ⊸ H) ⊸ H

This is identical to the meaning [pro] used in the analyses of obligatory and
arbitrary anaphoric control provided previously. Instantiating this meaning con-
structor according to the f-structure labels in (144), we obtain the meaning
constructor in (149), for which we reuse the label [pro]:

x1
(149) [pro] λP. ⊕ P(x1 ) : ∀H.(p𝜎 ⊸ H) ⊸ H

The meaning constructors in (150) are relevant in the analysis of example (142):

(150) Meaning constructor premises for ngarrka ka wirnpirli-mi karli

jarnti-rninja-karra:

[whistle] λx. : m𝜎 ⊸ w𝜎
whistle (x)
x1
[man] λP. ⊕ P(x1 ) : ∀H.(m𝜎 ⊸ H) ⊸ H
man (x1 )

[trim] λy.λx. : b𝜎 ⊸ [p𝜎 ⊸ t𝜎 ]

trim (x, y)

[during] λP.λQ. : t𝜎 ⊸ [w𝜎 ⊸ w𝜎 ]

during (P, Q)
x1
[pro] λP. ⊕ P(x1 ) : ∀H.(p𝜎 ⊸ H) ⊸ H

x1
[boomerang] λP. ⊕ P(x1 ) : ∀H.(b𝜎 ⊸ H) ⊸ H
boomerang (x1 )
598 functional and anaphoric control

In order to obtain the correct reading for this sentence, it is necessary first to intro-
duce hypothetical meanings for the arguments of the main verb (see §8.8.1.4).21
We therefore make use of the hypothetical meaning constructors a : [p𝜎 ] and
b : [b𝜎 ] to obtain a complete meaning for the predicate trim, labeled [trim-ab]:

(151) [trim-ab] : t𝜎
trim (a, b)

We likewise hypothesize the meaning constructor c : [m𝜎 ] for the subject of the
verb whistle, to produce the meaning constructor [whistle-c]:

(152) [whistle-c] : w𝜎
whistle (c)

The predicate during can then apply to these two propositions, to yield the meaning
constructor labeled [whistle-trim]:22

(153) [whistle-trim] : w𝜎
during ( , )
trim (a, b) whistle (c)

We now discharge the hypothesized meaning for m𝜎 introduced in (152), yielding

the meaning constructor labeled [whistle-trim-c]:

(154) [whistle-trim-c] λx. : m𝜎 ⊸ w𝜎

during ( , )
trim (a, b) whistle (x)

This meaning constructor is of the correct form for the meaning constructor [man]
to apply to it; combining these produces the meaning constructor [man-w-t]:

x1
man (x1 )
(155) [man-w-t] : w𝜎
during ( , )
trim (a, b) whistle (x1 )

21 This is necessary in order to ensure that the predicate during does not scope over the discourse
referents x1 , x2 , and x3 .
22 We simplify here by treating during as a relation between DRSs, which incorrectly allows for the
possibility for discourse referents to scope inside the DRS arguments of during . The problem would
disappear if we adopted a more sophisticated analysis in which during is a relation between event
variables (as introduced in §8.10), since there would then be no subordinate DRSs as arguments to
during .
 control in adjuncts 599

Next, we discharge the hypothesized meaning for b𝜎 introduced in (151), yielding

the meaning constructor labeled [man-w-t-b]:

x1
man (x1 )
(156) [man-w-t-b] λy. : b𝜎 ⊸ w𝜎
during ( , )
trim (a, y) whistle (x1 )

Now the meaning constructor [boomerang] can be applied to the meaning con-
structor [man-w-t-b], producing the meaning constructor [man-bmr]:

x1 , x2
man (x1 )
boomerang (x2 )
(157) [man-bmr] : w𝜎
during ( , )
trim (a, x2 ) whistle (x1 )

We now discharge the hypothesized meaning for p𝜎 introduced in (151), yielding

the meaning constructor labeled [man-bmr-a]:

x1 , x2
man (x1 )

: p𝜎 ⊸ w𝜎
boomerang (x2 )
(158) [man-bmr-a] λz.
during ( , )
trim (z, x2 ) whistle (x1 )

The resulting meaning constructor is of the correct form for the meaning construc-
tor [pro] to be applied to it, producing a complete and coherent meaning for the
clause:

x1 , x2 , x3
man (x1 )
boomerang (x2 )
(159) [man-bmr-w-t-a] : w𝜎
during ( , )
trim (x3 , x2 ) whistle (x1 )

Finally, the constraint in (147) enforces an antecedency relation between the

subject of the adjunct clause and the subject of the matrix clause. We therefore
augment the meaning constructor in (159) with a specification of A:
600 functional and anaphoric control

x1 , x2 , x3
man (x1 )
boomerang (x2 )
(160) : w𝜎 , A(x3 ) = x1
during ( , )
trim (x3 , x2 ) whistle (x1 )

In this way, constraints on the antecedent of the unexpressed pronominal subject

of the adj clause are appropriately enforced.

15.9 Further reading and related issues

The syntax of functional and anaphoric control has been a central topic of LFG
research from the inception of the theory; besides the work discussed in this
chapter, important early work exploring the nature of control crosslinguistically
was done by Andrews (1982), Neidle (1982), and Mohanan (1983).
Traditionally, research in LFG assumes that the open complement in a func-
tional control construction bears the grammatical function xcomp and that no
other governable grammatical function is an open function. Arka and Simp-
son (1998, 2008) provide evidence that this assumption cannot be universally
maintained. In English and many other languages, the controller in an equi or
raising construction can bear the subj role, while the controlled clause can be an
xcomp; alternatively, in some languages but not in English, obj controllers are
possible with open subj complements. In the Balinese equivalent of a sentence
like Chris wants to leave, then, the controller Chris can bear the obj function,
with to leave bearing the subj function. This situation counterexemplifies previous
claims about the syntactic role of the controller and the controlled clause: Jacobson
(1990) claims that the controller in a raising construction must be higher on
the grammatical function hierarchy than the controllee, precluding an analysis of
the open complement in a Balinese control construction as a subj. As Arka and
Simpson (1998, 2008) show, control constructions in Balinese do not obey this
generalization.
This is not the only proposal to extend the inventory of open grammatical
functions. As discussed in §15.1.2, Falk (2005) argues that alongside xcomp it is
necessary to distinguish xobj𝜃 and xobl𝜃 .
One type of control construction not discussed in this chapter is so-called
“possessor raising”:

(161) She kissed him on the cheek.

In this example, the obj of the verb, him, also functions as the possessor of the
obl obj the cheek. Lødrup (2009) discusses possessor raising constructions in
Norwegian, analyzing them in terms of functional control. Further discussion of
control structures in Norwegian is provided by Lødrup (2001, 2002a,b, 2008b,
2017a,b), among others.
 further reading and related issues 601

The analysis of predicative participial and adjectival adjuncts in terms of func-

tional control, as presented in §15.8.1, has been widely adopted. Whether attribu-
tively used participial adjuncts should be analyzed in terms of functional or
anaphoric control is more controversial. Authors who argue for the former position
include Haug and Nikitina (2012, 2016) and Spencer (2015); Lowe (2015c, 2017a)
argues for the latter.
The analysis of auxiliary verb constructions in English and other languages is
also a matter for debate. Falk (2003, 2008) discusses the English auxiliary system,
and argues that a uniform analysis is not possible: some auxiliaries in English
must be analyzed as raising verbs, while others do not contribute a pred but
merely contribute features to the f-structure headed by the lexical verb. Patejuk
and Przepiórkowski (2014) take up this question with regard to the Polish passive,
arguing that it should be treated in terms of raising. Butt and Lahiri (2013)
contrast the properties of auxiliary constructions with those of complex predicate
constructions in a diachronic perspective.
The diachronic development of control constructions has been investigated by
Barron (1997, 2001), Lowe (2015c), and Lødrup (2017a), among others.
 16
Coordination

In this chapter, we examine constructions involving coordination. In §16.1, we

examine the syntax of simple clausal coordination. Section 16.2 discusses coordi-
nate structures involving verbs and other argument-taking predicates. Section 16.3
discusses f-structure features and their behavior in coordinate structures, with a
particular focus on the distinction between distributive and nondistributive features
introduced in §6.3.2. Section 16.4 presents the theory of nonconstituent coordina-
tion, and §16.5 discusses unlike coordination. Section 16.6 discusses coordination
patterns crosslinguistically, and introduces rule macros for coordinate structures.
In §16.7, we turn to an examination of the semantics of clausal, predicate, and
verb coordination. In some cases, such as clausal coordination, simple conjunction
of the meanings of the conjuncts is involved. In other cases involving sharing of
dependents, the resource-sensitive nature of our glue language becomes an issue,
and a theory of resource sharing is required.
Section 16.8 examines noun phrase coordination and the syntactic and semantic
properties of coordinated noun phrases, which may include person, number, and
gender features. Semantically, noun phrase coordination differs from clausal or
predicate coordination in that it involves group formation, and so we also briefly
discuss the semantics of plurals, including conjoined quantifiers.

16.1 Clausal coordination

Coordination was first examined in an LFG setting by Bresnan et al. (1985b),
and the formal properties of coordination were explored in detail by Kaplan
and Maxwell (1988b). We begin our discussion with the simple case of clausal
coordination, as described by Kaplan and Maxwell (1988b). Kaplan and Maxwell
propose a constituent structure like (1) for coordinated clauses like Chris yawned
and David sneezed:

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 clausal coordination 603

(1) Chris yawned and David sneezed.

IP

IP Cnj[main] IP

NP I! and NP I!

N VP N VP

Chris V David V

yawned sneezed

As this c-structure shows, we do not treat coordinate structures as X! -theoretic

phrases of category CnjP, headed by Cnj, as proposed by Johannessen (1998) and
others. Rather, a coordinate structure is a non-X! -theoretic structure in which the
mother node inherits the category of its conjuncts, so that a coordinate structure
with IP conjuncts is a phrase of category IP. Borsley (2005) and Hristov (2012)
present convincing arguments in favor of the view adopted here, and against the
X! -theoretic CnjP view.
There is no limit to the number of conjuncts in a coordinate structure; therefore,
as discussed in §2.4.4, a coordinate f-structure is represented as a set whose
members are the f-structures of the individual conjuncts. The c-structure in (1)
corresponds to this f-structure:

(2) Chris yawned and David sneezed.

 

 ’ 


 


 

 

 ’ 


 


 

 

The c-structure and f-structure in (1) and (2) are constrained by the rule in
(3) for IP coordination. We adopt the complex c-structure category Cnj[main]
(with parameter main) for conjunctions like and; below, we introduce the category
Cnj[pre] for preconjunctions such as both in both Chris and David. The preliminary
rule in (3) does not account for certain aspects of the f-structure of coordinate
structures, including features contributed by the conjunction and; in §16.3, we pro-
vide a coordination rule which incorporates this aspect of the syntax of coordinate
structures:

(3) Preliminary rule for IP coordination, to be amended:

IP −→ IP+ Cnj[main] IP
↓∈↑ ↓∈↑

This rule makes use of the Kleene plus operator +, which licenses one or more
occurrences of IP. Thus, this rule allows a coordinate IP to consist of one or more
IPs, followed by a conjunction, followed by the final IP conjunct.
604 coordination

The functional annotations on this rule use the set-membership relation symbol
∈ to specify that each f-structure corresponding to an IP conjunct is a member
of the set of f-structures corresponding to the mother IP; set descriptions are
discussed in more detail in §6.3. The diagram in (4) shows the relation between
the c-structure and the f-structure of the example under discussion:

(4) Chris yawned and David sneezed.

 

 ’ 


 


 

 
IP  


 ’ 



 

IP IP  
Cnj[main]
↓∈↑ ↓∈↑

NP I and NP I

N VP N VP

Chris V David V

yawned sneezed

16.2 Predicate coordination

When unsaturated predicates are coordinated, the situation is more complex:
coordinated verbs often share some arguments, and Completeness and Coherence
requirements must be satisfied for each verb. In (5), the verbs selected and hired
are transitive, and so to meet Completeness and Coherence requirements each
must have a subj and an obj. In the c-structure and f-structure shown in (5), this
requirement is met; for both verbs, Chris is the subj, and David is the obj:

(5) Chris selected and hired David.

IP  
 ’ 

 

NP I 
  

  


  


N VP 
 


 

Chris V NP
   


 ’ 


V N 
  


   


   

V Cnj[main] V David 
 

 
selected and hired

We propose this preliminary phrase structure rule for verb coordination:

(6) Preliminary rule for V coordination, to be amended:

V −→ V+ Cnj[main] V
↓∈↑ ↓∈↑
 predicate coordination 605

Given the rule in (6), the c-structure and f-structure for the coordinated verbs
selected and hired are:

(7) selected and hired . /

 pred ‘select'subj,obj(’ 
 
V  . / 
 pred ‘hire'subj,obj(’ 
V V
Cnj[main]
↓∈↑ ↓∈↑

selected and hired

Both of these verbs are transitive, requiring a subj and an obj. In the example under
discussion, the coordinate V is the head of V! , and the f-structure corresponding
to V! (with obj structure not yet included) is the set labeled c, as the annotations
in (8) indicate:

(8) selected and hired David

F-structure not including structure contributed by David:
. /
 pred ‘select'subj,obj(’ 
 
c . / 
V! 
 pred ‘hire'subj,obj(’ 
V NP
↑=↓ (↑ obj) = ↓

V V
Cnj[main] N
↓∈↑ ↓∈↑

selected and hired David

In (8), the equation on the NP node dominating David refers to the f-structure
corresponding to the V! node—the set c—and requires the f-structure for David
to be the obj of that set. These requirements are summarized in (9), where the
f-structure for David is labeled d, and the annotation on the NP node is instantiated
to (c obj) = d:

(9) selected and hired David

. /
 pred ‘select'subj,obj(’ 
 
c . /

 pred ‘hire'subj,obj(’ 

0 1
d pred ‘David’

(c obj) = d

Bresnan et al. (1985b) and Dalrymple and Kaplan (2000) provide a definition of
function application to sets that allows us to interpret an equation like (c obj) = d
when c is a set: in such a situation, d is required to be the obj of each member of
606 coordination

the set c. This is because governable grammatical functions like obj are distributive
features, as defined in (189) of §2.5.8 and (64) of §6.3.2, repeated here:1

(10) Nondistributive features: index, adj, conj, preconj

Distributive features: all other features, including pred and the governable
grammatical functions

(11) If a is a distributive feature and s is a set of f-structures, then (s a) = v holds

if and only if (f a) = v for all f-structures f that are members of the set s.

Thus, the constraints in (9) entail that d is the obj of each f-structure in c:

(12) selected and hired David

V  

 ’ 

V NP 
 

 
↑=↓ (↑ ↓ 
 


 ’ 

V V 
 

Cnj[main] N 
 

↓∈↑ ↓∈↑  

selected and hired David

By the same reasoning, Chris is the subj of the set c in (5) and is thus the subj of
each member of c, as shown in (5).
Peterson (2004a) observes that the treatment of grammatical functions as dis-
tributive features nicely accounts for this pattern:

(13) a. Kate faxed and emailed the results to Paul.

b. ⃰Kate faxed and disliked the results to Paul.
c. ⃰Kate faxed the results to Paul and Kate disliked the results to Paul.

The verbs fax and email require a subj (Kate), an obj (the results), and an oblgoal
(to Paul), and these arguments are shared in the coordinate structure in (13a),
as shown in (14a). In contrast, (13b) is ungrammatical for the same reason that
(13c) is ungrammatical: dislike requires an obj but not an oblgoal . The incoherent
f-structure for (13b) is shown in (14b):

1 Peterson (2004a: 655) provides a definition similar to (11), but restricted to situations in which a
is a grammatical function: in other words, for Peterson, the only distributive features are grammatical
functions.
 predicate coordination 607

(14) a. Kate faxed and emailed the results to Paul.

Complete and coherent f-structure:
   

 


   


   


   


   


   


   


  +  


   


   


 


  


 


 


  


  


  


  


  


  


  


 


b. ⃰Kate faxed and disliked the results to Paul.

Incoherent f-structure (dislike does not govern oblgoal ):
  
 ’ 

  


  


  


  


  


  


  


  + 


  


 

 

   


 ’ 


 


   


   


   


   


   


   


   


 


As Frank (2006) shows, the classification of grammatical functions as distributive

features also underpins a straightforward account of the German asymmetric
coordination construction, also known as SGF coordination (Subject Gap in
Finite/Fronted constructions; Wunderlich 1988). In these constructions, one of the
dependents (subj or a member of the dis set) appears in the initial conjunct, but is
shared across all of the conjuncts of the coordinate structure. This is shown in (15),
where the subject der Jäger ‘the hunter’ appears in the first conjunct, but is inter-
preted as the subject of both the first and the second conjunct (Wunderlich 1988:
289). Other arguments are not shared in this way: for example, the locative phrase
in den Wald is associated only with the initial conjunct.

(15) [In den Wald ging der Jäger] und [fing einen Hasen].
into the forest went the hunter and caught a rabbit
‘The hunter went into the forest and caught a rabbit.’
608 coordination

Frank (2006) analyzes these constructions by means of asymmetric projection of

either subj or a member of the dis set from the first conjunct. She defines the
abbreviatory symbol gdf as a disjunction including subj and dis, and adds an
optional constraint to associate the gdf in the first clause with the coordinate
structure as a whole. Since both subj and dis are distributive, the result is that
the gdf distributes to each conjunct:

(16) gdf ≡ {subj|dis ∈}

CP −→ CP Cnj CP
↓∈↑ ↓∈↑
((↓ gdf) = (↑ gdf))

The annotated phrase structure and f-structure for (15) are shown in (17), with the
subject der Jäger ‘the hunter’ distributed over both conjuncts:
(17) [In den Wald ging der Jäger] und [fing einen Hasen].
into the forest went the hunter and caught a rabbit
‘The hunter went into the forest and caught a rabbit.’
CP

CP
CP
↓∈ ↑ Cnj[main]
↓ ∈↑
(↓ ↑

PP
↓∈ (↑ C und C
(↑ ↓ and

In den Wald C VP C VP
into the forest
NP NP
ging
(↑ ↓ (↑ ↓
went caught
einen Hasen
der Jäger
a rabbit
the hunter
’

−
 syntactic properties of coordinate structures 609

This allows for the correct c-structural and f-structural analysis of such examples,
with only a minimal addition to the annotation on the standard c-structure rule
for CP coordination.
Frank (2006) also discusses the discourse-functional properties of this construc-
tion, showing that the special properties of this construction are related to general
discourse subordination effects also found in modal subordination (Frank 1997);
this provides an explanation of some of the other syntactic, semantic, and discourse
properties of the construction.

16.3 Syntactic properties of coordinate structures

Coordinate structures are special in that the coordinate structure as a whole often
has its own properties, distinct from the properties of its elements. The syntactic
features that a set of f-structures can have are nondistributive features. The behavior
of nondistributive features is defined in (64) of §6.3.2, repeated here:

(18) If a is a nondistributive feature, then (f a) = v holds if and only if the pair

'a, v( ∈ f .

We now expand our treatment of conjunction to allow preconjunctions such as

both, either, and neither. Notice that the preconjunction both is allowed only for
some categories: V, NP, PP, and AP, for example, but not N:2

(19) a. Chris both [made] and [ate] a sandwich.

b. Both [the man] and [the woman] sneezed.
c. I looked both [under the table] and [in the box].
d. David is both [tall] and [thin].
e. ⃰The both [sandwich] and [soup] are cold.

We propose the following lexical entries for both and and:

(20) both Cnj[pre] (↑ preconj) = both

(↑ conj) =c and
and Cnj[main] (↑ conj) = and

2 Huddleston and Pullum (2002: Chapter 15) provide more discussion of preconjunctions and
what they call correlative coordination: both . . . and, either . . . or, neither . . . nor. Constraints on this
construction are complex: for example, although both is not permitted in IP coordination, either is
permitted:
(a) ⃰Both [it is raining] and [the sun is shining].
(b) Either [it is raining] or [the sun is shining].
We leave a full treatment of the syntax of correlative coordination for future work.
610 coordination

The features preconj and conj are classified as nondistributive features. Treating
the conj feature as a nondistributive feature of the coordinate structure also
ensures that different conjunctions do not appear in the same coordinate structure:
for example, Chris and David or Tracy may not be treated as a single coordinate
structure with three conjuncts, but must be analyzed as nested coordination, either
Chris and [David or Tracy] or [Chris and David] or Tracy. The second line of the
lexical entry for both contains a constraining equation ensuring that both does
not appear with conjunctions other than and, allowing both tall and thin and
disallowing ⃰both tall or thin.
The rule in (21) permits the analysis of coordinated verbs with a preconjunction
(Chris both yawned and sneezed), and associates the information contributed by the
conjunction and any preconjunctions like both or either with the f-structure for the
coordinate phrase:

(21) Rule for V coordination including preconjunction:

V −→ Cnj[pre] V+ Cnj[main] V
↑=↓ ↓ ∈↑ ↑=↓ ↓ ∈↑

Given the lexical entries in (20) and the rule in (21), the c-structure and f-structure
for the phrase both selected and hired are:

(22) both selected and hired

 preconj both 
 conj and 
 . / 
 
 pred ‘select'subj,obj(’ 
 
 
 . /  
 pred ‘hire'subj,obj(’ 

Cnj[pre] V Cnj[main] V
↑=↓ ↓∈↑ ↑=↓ ↓∈↑

both selected and hired

In (22), the preconj and conj features are attributes of the coordinate structure,
as required.
 syntactic properties of coordinate structures 611

The coordination rule in (21) also permits more complex nested coordinate
structures like either [selected and hired] or [interviewed and rejected], which has
the following c-structure and f-structure:

(23) either selected and hired or interviewed and rejected

 preconj either 
 conj or 
    

 conj and 




   . /  
 
 

   pred ‘select'subj,obj(’ 
 




   
 
  . /   
 
    
 

 pred ‘hire'subj,obj(’ 
 
 

 
   
 conj and  

 . /


 
   
 
  pred ‘interview'subj,obj(’ 

 
    
   
  . /  
 
    
 
  pred ‘reject'subj,obj(’
   

Cnj[pre] V Cnj[main] V
↑ =↓ ↓∈↑ ↑ =↓ ↓∈↑

V Cnj[main] V V Cnj[main] V
either or
↓∈↑ ↑ =↓ ↓∈↑ ↓∈↑ ↑ =↓ ↓∈↑

selected and hired interviewed and rejected

As we have seen, grammatical functions are distributive; thus, a verbal dependent

appearing outside this complex coordinate structure distributes to the innermost
conjuncts, as shown in (24), where David is the obj of each verb:
612 coordination

(24) either selected and hired or interviewed and rejected David

V NP
↑=↓ (↑ ↓

Cnj[pre] V Cnj[main] V
N
↑=↓ ↓∈↑ ↑=↓ ↓∈↑

V Cnj[main] V V Cnj[main] V
either or David
↓∈↑ ↑=↓ ↓∈↑ ↓∈↑ ↑=↓ ↓∈↑

selected and hired interviewed and rejected

16.4 Nonconstituent coordination

Coordination has often been considered a reliable indicator of constituenthood;
for example, as discussed in §3.1, Radford (2009) proposes that a string of words is
a constituent if “it can be coordinated with another similar constituent.” However,
constructions involving nonconstituent coordination show that the situation is
more complex. Strings that are clearly not phrase structure constituents can be
coordinated:

(25) David introduced [[Chris] [to Tracy]] and [[Matty] [to Ken]].
 nonconstituent coordination 613

In (25), the sequence Chris to Tracy is not a phrase structure constituent; nev-
ertheless, it can be coordinated with the sequence Matty to Ken, which is not a
constituent either.
Maxwell and Manning (1996) propose a theory of nonconstituent coordination
that accounts for the grammaticality of examples like (25). Their account has its
basis in Wasow’s Generalization:

(26) Wasow’s Generalization (Pullum and Zwicky 1986):

If a coordinate structure occurs in some position in a syntactic represen-
tation, each of its conjuncts must have syntactic feature values that would
allow it individually to occur in that position.

According to Wasow’s Generalization, (25) is acceptable because both Chris to

Tracy and Matty to Ken constitute valid completions of a VP constituent beginning
with introduced. Maxwell and Manning’s theory captures this generalization by
allowing phrase structure rules to be split, and rules of coordination to refer to
the partial constituents that are described by these partial rules. Building on the
approach of Maxwell and Manning (1996), Brun (1996a,b) develops a detailed
analysis of coordination in French, and Forst and Rohrer (2009) and Kuhn et al.
(2010) explore nonconstituent coordination and right node raising in German.

16.4.1 Constituent structure constraints

To illustrate Maxwell and Manning’s approach, we assume the following simplified
phrase structure rule for the English VP:

(27) VP −→ V NP PP

Maxwell and Manning propose that we can think of the right side of this rule as
being divided into two portions, which we will call VP-x for the first part and x-VP
for the second part:

(28) VP −→ VP-x x-VP

To analyze (25), we assume that the first half of the VP rule analyzes V, and the
second half analyzes the sequence NP PP; for other constructions, a different split
of the rule might be necessary:

(29) VP-x −→ V
x-VP −→ NP PP

Crucially, rules of coordination can refer to the partial phrase structure con-
stituents that result from splitting rules in this way. This allows the following
c-structure analysis of (25):
614 coordination

(30) David introduced Chris to Tracy and Matty to Ken.

IP

NP I!

N VP

David VP-x x-VP

V x-VP Cnj[main] x-VP

introduced NP PP and NP PP

Chris to Tracy Matty to Ken

On Maxwell and Manning’s theory, any phrase structure rule can be broken up
into two or more parts in this way. However, the only rules that can refer to
these partial phrase structure constituents are rules of coordination. Therefore, the
partial constituents that result from rule splitting play no other role in the grammar
besides their role in the analysis of nonconstituent coordination.
In analyzing more complex examples, a c-structure rule may be broken into
more than two pieces. For the purposes of this example, we assume this VP rule:

(31) VP −→ V NP NP CP

We then break the VP into three parts. The first part of the VP, VP-x, analyzes V;
the second part, x-VP-y, analyzes the sequence NP NP; and the third part, y-VP,
analyzes CP:

(32) VP −→ VP-x x-VP-y y-VP

VP-x −→ V
x-VP-y −→ NP NP
y-VP −→ CP

These rules allow an analysis of the following example, in which there is shared
material both before and after the coordinate structure:

(33) David bet Chris five dollars and Matty ten dollars that it would rain.
IP

NP I!

N VP

David VP-x x-VP-y y-VP

V x-VP-y Cnj[main] x-VP-y CP

bet NP NP and NP NP
that it would rain
Chris five dollars Matty ten dollars
 nonconstituent coordination 615

Other examples show greater degrees of complexity. Examples (31) and (33)
involve partial constituents that are all dominated by a single mother node;
Maxwell and Manning also discuss examples in which more than one rule is
involved in the split, treating these examples by the use of a stack of partial
constituents that must be combined in a compatible way.
In formal terms, Maxwell and Manning (1996) state their proposal by reference
to the state of the finite-state automaton that corresponds to the regular expression
on the right-hand side of a phrase structure rule.3 In constructions involving
nonconstituent coordination, the automaton can stop in a particular state in a
phrase structure rule and can then continue from that state to analyze each
conjunct of a coordinate phrase.4 In other words, each conjunct in a coordinate
structure must constitute a valid expansion of the mother category.
Maxwell and Manning note that their analysis allows a natural treatment of
cases where each conjunct contains a different number of constituents. The only
constituent structure requirement in an example like (34) is that each conjunct
must constitute a valid completion of the VP rule, and different numbers of phrases
are allowed:

(34) You can call me [directly] or [[after three p.m.] [through my secretary]].

As pointed out by Milward (1994), such cases are problematic for some other
approaches to coordination, particularly the “3-D” approaches of Goodall (1987)
and Moltmann (1992).

16.4.2 Functional annotations

A desirable property of Maxwell and Manning’s analysis is that no special stipula-
tions are required concerning the functional structures of constructions involving
nonconstituent coordination; the rules we have outlined so far give the desired
result for all of the examples of nonconstituent coordination that we have dis-
cussed. We assume the standard functional annotations on phrase structure rules
that we have discussed so far, and we also impose the intuitively reasonable
requirement that the f-structures of the subconstituent parts of a split constituent
are the same as the f-structure for the full constituent, as the annotations in the
following rule indicate:

(35) VP −→ VP-x x-VP

↑=↓ ↑=↓

3 A finite-state automaton is an abstract “machine” that advances through a string, moving from
state to state as the string is traversed. If the string is a member of the language of the regular expression
corresponding to the automaton, the automaton is in a final state when the end of the string is reached.
An automaton corresponding to the right-hand side of an LFG phrase structure rule advances through
the daughter categories it encounters, moving from state to state as the daughter phrases are analyzed.
4 As Maxwell and Manning (1996) point out, another way of thinking of the theory is in terms of the
regular expression that appears on the right-hand side of a phrase structure rule; on this view, the partial
phrase structure constituents that are involved in nonconstituent coordination must be members of
the suffix language of the regular expression representing the right-hand side of a phrase structure rule,
where the prefix consists of the phrase structure categories that precede the coordinated subconstituent.
616 coordination

Under these assumptions, the annotated c-structure for (25) is:

(36) David introduced Chris to Tracy and Matty to Ken.

IP

NP I!
(↑ subj) = ↓ ↑ =↓

VP
David ↑ =↓

VP-x x-VP
↑ =↓ ↑ =↓

V x-VP x-VP
Cnj[main]
↑ =↓ ↓∈↑ ↓∈↑

NP PP NP PP
introduced and
(↑ obj) = ↓ (↑ oblgoal ) = ↓ (↑ obj) = ↓ (↑ oblgoal ) = ↓

Chris to Tracy Matty to Ken

These annotations give us the following f-structure, as desired:

(37) David introduced Chris to Tracy and Matty to Ken.

 
   
’ 
 
 
 

 

 

 
 







  





 

 
   
 
  
 
 
  
 
   
 
   
 
   
 
   
 








 
 
 

 unlike category coordination 617

16.5 Unlike category coordination

16.5.1 Unlike category coordination as nonconstituent coordination
Maxwell and Manning’s approach to nonconstituent coordination also addresses
one aspect of another long-standing problem in the syntax of coordination. Sag
(2002) discusses coordination of constituents of different categories, providing
examples like:

(38) We walked [slowly]AdvP and [with great care]PP .

In this example, each conjunct of the coordinate phrase is an acceptable con-

tinuation of the VP in which it appears. Maxwell and Manning’s treatment of
nonconstituent coordination extends unproblematically to such cases:

(39) We walked slowly and with great care.

IP

NP I

N VP

We VP-x x-VP

V x-VP Cnj[main] x-VP

walked AdvP and PP

Adv
with great care
slowly

’
618 coordination

Notice that as an instance of nonconstituent coordination, the resulting f-structure

has the status of a coordinated clausal structure, resembling the f-structure for the
full clausal coordination We walked slowly, and we walked with great care.5 The
multiclausal analysis in (39) is suitable for examples like (38), and necessary in
the analysis of examples in which the conjuncts are composed of more than one
phrasal constituent, as in (25) and (34).

16.5.2 Unlike category coordination as true coordination

Some examples of unlike category coordination cannot be analyzed in this way,
however, and must be treated as standard coordination of phrases of different c-
structure categories. Peterson (2004a) provides the following examples of unlike
category coordination, which differ crucially from the examples discussed in the
previous section:

(40) a. [A plumber] and [making a fortune] though Bill may be, he’s not going to
be invited to my party. [NP and VP]
b. [Stupid] and [a liar] Paul undoubtedly is, but he is still my friend.
[AP and NP]
c. [In town] and [itching for a fight] is the scourge of the West, Zitty Zeke.
[PP and VP]

There are several reasons why these examples cannot be analyzed as noncon-
stituent coordination. First, the coordinate structures in these examples are not
partial constituents which are completed by other constituents in the immediate
context, and so they cannot be analyzed in terms of the rule-splitting analysis
proposed by Maxwell and Manning (1996). Second, it is not possible in general
for partial constituents in nonconstituent coordination to appear in contexts like
those in (40):

(41) a. David introduced [[Chris] [to Tracy]] and [[Matty] [to Ken]].
b. ⃰[[Chris] [to Tracy]] and [[Matty] [to Ken]] though David introduced, ...

Examples such as those in (40) show the need for a rule which allows constituents
of different categories to be coordinated to form a single coordinated constituent.
What is the c-structure category of a coordinate structure with conjuncts of
different categories? Valuable clues come from predicates whose arguments must
be phrases of a particular category. As discussed in §6.10.3, Pollard and Sag (1994)
show that the verb wax requires an adjectival complement, while the verb become
requires either a nominal or an adjectival complement:

5 See Patejuk (2015: Chapter 5) for discussion of multiclausal versus monoclausal analyses of lexico-
semantic coordination in Polish, which resembles these structures in certain respects.
 unlike category coordination 619

(42) a. Fred waxed [poetical]AP /[lyrical]AP .

b. ⃰ Fred waxed [a success]NP .
c. ⃰ Fred waxed [in a good mood]PP .
d. ⃰ Fred waxed [liking oranges]VP .
e. ⃰ Fred waxed [quickly]AdvP .

(43) a. Gerry became [quite conservative]AP .

b. Pat has become [a Republican]NP .
c. ⃰ Connie has become [of the opinion that we should get out]PP .
d. ⃰ Tracy became [awarded a prize]VP .
e. ⃰ Chris will become [talking to colleagues]VP .

The CAT predicate, defined in (157) of Chapter 6 (§6.10.3) allows for the imposi-
tion of these constraints; the xcomp complement of wax must be of category AP,
and the complement of become must be either NP or AP:

(44) a. wax: CAT((↑ xcomp), {AP})

b. become: CAT((↑ xcomp), {AP, NP})

Notably, unlike-category coordinated complements containing one NP comple-

ment and one AP complement are acceptable with become, but other unlike
category coordinations are disallowed:

(45) a. Pat has become [a Republican]NP and [quite conservative]AP .

b. ⃰Pat has become [a Republican]NP and [awarded a prize.]VP
c. ⃰Pat has become [quite conservative]AP and [talking to colleagues.]VP

These and other examples show that each conjunct in an unlike category coordi-
nation structure must satisfy the category requirements imposed by the predicate.
To impose the appropriate constraints in coordinate structures, Kaplan and
Maxwell (1996) and Crouch et al. (2008) propose that the CAT predicate is
distributive, and must hold of each conjunct in a coordinate structure. Under this
assumption, the requirements in (44) produce the correct result for constructions
involving these verbs: the complement of wax must be a phrase of category AP
or a coordinated phrase in which all conjuncts are of category AP, while the
complement of become must be a phrase of category NP or AP, or a coordinated
phrase in which each conjunct is either NP or AP.
Treating CAT as distributive allows for checking of category requirements
imposed by predicates like wax and become, but still leaves open the question
of the c-structure category of an unlike category coordinate structure such as a
Republican and quite conservative in (45a). Patejuk (2015) proposes the rule in (46),
in which YP and ZP are metacategories ranging over full phrasal categories, while
620 coordination

UP is a category label standing for Unlike Phrase for all instances of unlike category
coordination:

(46) C-structure rule for unlike category coordination according to Patejuk

(2015):
UP −→ YP Cnj[main] ZP
↓∈↑ ↓∈↑

On this view, all unlike category coordinate structures have the same c-structure
category: here, UP. This approach is problematic in that it complicates the grammar
by requiring introduction of the category UP wherever unlike category coordina-
tion structures are allowed, in addition to standard categories such as NP or AP.
Peterson (2004a: 652) proposes that the category of an unlike category coordi-
nate structure is the same as the category of the first conjunct:

(47) C-structure rule for unlike category coordination according to Peterson

(2004a):
X −→ X Cnj[main] Y
↓∈↑ ↓∈↑

In this rule, X and Y are metacategories ranging over all categories; X is used as the
category label of the mother node as well as the category label of the first conjunct.
As noted by Patejuk (2015), this leads to the expectation that the distribution of
a coordinate phrase mirrors the distribution of its first conjunct. This proposal
runs into problems originally noted by Sag et al. (1985: 141), where a coordinate
structure does not have the same distribution as its first conjunct. The coordinate
structure longwinded and a bully is acceptable as a copular complement:

(48) The man was longwinded and a bully.

However, this coordinate structure does not have the distribution of a normal AP,
and cannot appear in other places in which AP can appear:

(49) a. The longwinded man was my brother.

b. ⃰The [longwinded] and [a bully] man was my brother.

The problem for each of these proposals is that the category of the coordinate
phrase does not reflect the categories of any of the conjuncts (Patejuk 2015)
or the categories of the non-initial conjuncts (Peterson 2004a). As discussed by
Dalrymple (2017), this makes it difficult to enforce category-function correlations
and to control the distribution of phrases of different categories. For example, if
we assume a standard annotated phrase structure rule that associates the comp
grammatical function with a CP verbal complement, the Patejuk and Peterson
proposals allow conjuncts of categories other than CP in an unlike category
 coordination patterns crosslinguistically 621

coordinate structure bearing the grammatical function comp. We can prevent this
by introducing CAT constraints throughout the grammar to control the category
of the conjuncts in unlike category coordination, but the result is a considerable
complication of the grammar.
Dalrymple (2017) provides an analysis of unlike category coordination which
assumes that the CAT predicate is not distributive. Instead, the phrase structure
category of an unlike category coordinate structure reflects properties of the cate-
gories of each of the conjuncts, and the CAT predicate constrains the category of
the coordinate structure as a whole. Similar to the view of category labels advocated
by Sag et al. (1985), Marcotte (2014), Bresnan et al. (2016), and Lovestrand and
Lowe (2017), this proposal assumes that category labels are feature structures with
features encoding phrasal category as well as bar level, the status of the category as
functional or lexical, and whether the category is a projecting or a nonprojecting
category. The category label in a coordinate structure is determined on the basis
of the category features of each of the conjuncts: for example, a noun phrase bears
the feature n with value +; an adjective phrase bears the feature adj with value
+; and a coordinate structure containing a noun phrase conjunct and an adjective
phrase conjunct bears both features, n + and adj +. Predicates such as wax and
become place requirements on the feature composition of their complement by use
of the CAT predicate: wax requires an adjectival complement, with value − for all
features except for the feature adj, and become requires an adjectival or nominal
complement, with value − for all features except for the features adj and n. See
Dalrymple (2017) for further discussion and exemplification.
As observed by Bayer (1996), the successful treatment of category features in
unlike category coordination is very much dependent on the feature inventory
that is assumed: Bayer discusses difficulties in the analysis of predicates which
require either an NP or S complement under Sag et al.’s assumptions, where NP
has the features [+N, −V], and S has the features [−N, +V], with no features
in common. In fact, the same issues in the definition and representation of
f-structure features discussed in §2.5.1 arise for c-structure category features. For
example, Bresnan et al. (2016: chapter 6) propose a two-dimensional classification
of c-structure features as “predicative” (verbs and adjectives) and “non-predicative”
(prepositions and nouns), and as “transitive” (verbs and prepositions) and “non-
transitive” (nouns and adjectives). These features successfully capture the phrase-
internal properties of these categories: verbs and prepositions can take objects,
but nouns and adjectives generally do not, for example. However, they do not
straightforwardly allow for the statement of selection requirements when these
phrases appear in coordinate structures.

16.6 Coordination patterns crosslinguistically

We now propose a set of coordination macros, allowing us to incorporate the pat-
terns and constraints discussed in the preceding sections of this chapter. Haspel-
math (2004b) points out that there are four types of coordination patterns attested
622 coordination

crosslinguistically, depending on the number of conjunctions in the coordinate

structure:6

(50) a. Asyndetic coordination (no conjunction): X Y Z

[mox c’iiza b-yylira]; [darc hwovziira].
wind howl.inf b-start.wp blizzard turn.around.wp
‘The wind started to howl, and the blizzard turned around.’
(Chechen: Jeschull 2004: 252)
b. One conjunction in the coordinate structure: X Y and Z
[David got up], [Chris sat down], and [Matty left the room].
c. One fewer conjunction than the number of conjuncts: X and Y and Z
[David got up] and [Chris sat down] and [Matty left the room].
d. One conjunction for each conjunct: X and Y and Z and
hun eer d-u caara, [shain k’ant-a
what say.fut d-be.prs they.erg themselves.gen boy-erg
moliila] ’a, [i lovzush v-yyla] ’a xi’cha.
drink.subj and he play.scv V-be.subj and find.out.when
‘What will they say if they find out that their son drinks and he plays?’
(Chechen: Jeschull 2004, page 254)7

In (51a), we define a rule macro8 cnj(_c) for a sequence consisting of a conjunction

such as and or or and a conjunct of category _c. This macro is useful for English,
where the conjunction precedes the conjunct; in a language where the conjunction
follows the conjunct, as in the Chechen example in (50d), an alternative macro
encoding the other order would be required. The basic macro in (51a) allows us
to define additional general macros for each of these construction types, as shown
in (51b–e), using the Kleene plus operator to require one or more occurrences of
a phrase. These macros will also be useful in our discussion of the semantics of
coordinate phrases in §16.8.3:

(51) a. Definition of the rule macro cnj(_c):

cnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
b. Asyndetic coordination:
asyndetic-coord(_c) ≡ _c+
↓∈↑

6 The terminology used by Haspelmath (2004b) for these four types is somewhat confusing: he
uses the term monosyndetic coordination for what we call “one fewer conjunction than the number
of conjuncts,” treating the one-conjunction type as monosyndetic coordination with ellipsis of all but
one conjunction, and he uses the term bisyndetic coordination for what we call “one conjunction for
each conjunct.”
7 Placement of the conjunction ’a in Chechen is governed by complex rules; see Jeschull (2004) for
a full discussion.
8 See §6.8 for definition and discussion of rule macros.
 coordination patterns crosslinguistically 623

c. One conjunction in the coordinate structure:

one-conj(_c) ≡ _c+ @cnj(_c)
↓∈↑
d. One fewer conjunction than the number of conjuncts:
one-fewer-conj(_c) ≡ _c @cnj(_c)+
↓∈↑
e. One conjunction for each conjunct:
one-each-conj(_c) ≡ @cnj(_c)+

For English, the structures represented by the rule macros one-conj and one-
fewer-conj are both used; we thus provide a single macro english-coord which
allows either the “one conjunction in the coordinate structure” pattern or the “one
fewer conjunction than the number of conjuncts” pattern, which we will build on in
our discussion of the semantics of coordination. This macro requires a coordinate
structure to consist of an optional preconjunction,9 an initial conjunct, any number
of medial conjuncts, and a final conjunction and conjunct phrase. The medial
conjuncts must either all have conjunctions (as defined by the rule macro @cnj(_c)
or all lack conjunctions (the simple category _c):

(52) Rule macro for English coordination:

english-coord(_c) ≡
2 3 4 5
Cnj[pre] _c _c⃰ @cnj(_c)⃰ @cnj(_c)
↑ =↓ ↓∈↑ ↓∈↑

This macro allows the statement of coordination rules for a range of c-structure
categories as in (52), exemplifying with V and IP:

(53) Coordination rules using the rule macro in (52):

V −→ @english-coord(V)
IP −→ @english-coord(IP)

Constraints lexically associated with the preconjunctions both and either prevent
their appearance with certain phrase structure categories. For instance, as shown
in (19), both cannot appear with coordinated nouns (⃰The both [sandwich] and
[soup] are cold), though it can appear with coordinated noun phrases (Both [the
man] and [the woman] sneezed); this constraint is reflected in the final line of this
lexical entry for both, which disallows coordinate structures with category N:

9 The preconjunction both requires the coordinate phrase to have only two conjuncts; in light of
noncoordinate examples involving the related determiner both in examples like both (the) girls, which
refers to two girls, we take this to be a semantic fact, and we do not encode this constraint in the
c-structure rule. Prescriptively, the preconjunction either is sometimes also claimed to require only
two conjuncts, but examples with three or more conjuncts are commonly found: I’ll either [call out] or
[bang on the door] or [blow my whistle] (Huddleston and Pullum 2002, 2005).
624 coordination

(54) both Cnj[pre] (↑ preconj) = both

(↑ conj) =c and
N * ∈ 𝜙−1 (↑)

The rule macro in (52) is sufficient to account for the syntactic patterns which
we have examined so far. We will augment this rule macro in two ways in the
remainder of this chapter. First, in our discussion of the compositional semantics of
coordination, we designate one of the conjuncts as the “seed” and treat it specially
in semantic composition; this requires a separate macro encoding the special
semantic role of the “seed” conjunct. Second, in our discussion of feature resolution
patterns in nominal coordination, we propose additional constraints to ensure that
the appropriate resolved features are assigned to a nominal coordinate structure.

16.7 Coordination and semantic composition

16.7.1 Clausal coordination
The meaning of a coordinated sentence like Chris yawned and David sneezed is
usually represented simply as the conjunction of the meanings of the conjuncts:10

(55) Chris yawned and David sneezed.

yawn (Chris ) ∧ sneeze (David )

For the sentence Chris yawned and David sneezed to be true, it must be the
case both that Chris yawned and that David sneezed. We assume the following
f-structure and meaning constructor for this example:

(56) Chris yawned and David sneezed.

 
 ’

  
y 

 c

 
 

a 
 
 ’  
 
 


 s




d

yawn (Chris ) ∧ sneeze (David ) : a

To obtain this result, we adopt the compositional analysis of Asudeh and Crouch
(2002a),11 which can be intuitively described as follows. One of the conjuncts in a

10 Since we are not concerned with anaphoric relations in this chapter, we do not need the expressive
power of PCDRT, and we use predicate logic as our meaning language.
11 Dalrymple (2001) follows Kehler et al. (1995) in proposing an analysis that uses the of course
operator of linear logic, written as an exclamation point, to turn off resource accounting in the analysis
 coordination and semantic composition 625

coordinate structure is the “seed,” or the initial resource contributed to the meaning
of a coordinate structure. A special meaning constructor assigns the meaning of
this conjunct to be the initial meaning of the coordinate structure as a whole.
The other conjuncts are associated with meaning constructors which add their
meanings to the “seed” one by one, with the result that the meanings of all of the
conjuncts are combined to produce the final, complete, and coherent meaning of
the coordinate structure.
We begin our discussion by inspecting the meaning constructor in (57), which
is associated with the seed conjunct in the phrase structure rule for clausal coor-
dination. Its effect is to consume the resource associated with the seed conjunct
daughter (↓𝜎 , of type t) and reassociate its meaning with the coordinate structure
as a whole (↑𝜎 , also of type t). Note that this meaning constructor is specified as
applying only to conjuncts of type t; we extend this treatment to coordination
of conjoinable types (types ‘ending in t’) in §16.7.2, and we discuss nominal
coordination and group formation in §16.8.2.

(57) Preliminary meaning constructor for seed conjunct, to be revised

λP.P : ↓𝜎 't( ⊸ ↑𝜎 't(

We also require a specification of the meaning of the conjunction. Following

Asudeh and Crouch (2002a), we assume that the meaning of the conjunction is
associated with the coord-rel attribute at semantic structure (∧ for and, ∨ for
or).12 The type of this meaning is 't → 't → t((; to save space, we do not explicitly
represent this in the meaning constructors which we present:

of coordinate structures. Asudeh and Crouch (2002a) observe that this analysis goes against the spirit
of the glue approach, which has a strong commitment to the preservation of resource sensitivity in
meaning composition, and so we do not adopt that analysis here.
12 This raises a resource issue. In §16.6, we noted the “one conjunction for each conjunct” pattern
of coordination, where multiple conjunctions can appear in a single coordinate structure. If each
conjunction contributes a meaning constructor to the same coordinate structure, but only one meaning
constructor is needed for interpretation of the coordinate structure, resource surplus results. We can
address this issue by providing a more complex lexical entry such as (a), which uses an instantiated
symbol (§5.2.2.1) as the value for the conj feature; recall that an instantiated symbol is like a semantic
form in that it takes on a unique value for each instance of use. The lexical entry in (a) contains a
disjunction: in the first case, the conj feature is contributed with an instantiated (unique) value, and
the meaning constructor is also contributed; in the second case, a constraining equation checks that
another instance of the conjunction has contributed a unique value for the conj feature. This ensures
that only one instance of the conjunction contributes a meaning constructor.
(a) and Cnj[main] { (↑ conj) = and
∧ : (↑𝜎 coord-rel)
| (↑ conj) =c and }
In the following discussion, for simplicity, we do not discuss the “one conjunction for each conjunct”
pattern, and we use the less complicated lexical entry in (58).
626 coordination

(58) Lexical entries for and and or, including meaning constructors
and Cnj[main] (↑ conj) = and
∧ : (↑𝜎 coord-rel)
or Cnj[main] (↑ conj) = or
∨ : (↑𝜎 coord-rel)

We also require meaning constructors which have the effect of adding the meaning
of the non-seed conjuncts to the meaning of the seed to produce the meaning
for the full coordinate structure.13 The meaning constructor for each non-seed
conjunct consumes the meaning resource of the conjunct (associated with ↓𝜎 ,
which is of type t) and combines the meaning of the conjunct with the meaning of
the coordinate structure derived so far (associated with ↑𝜎 ), using the meaning of
the conjunction (associated with (↑𝜎 coord-rel)) to combine them.14 This means
that we must also manage the meaning resource contributed by the conjunction,
which may need to be reused several times in a coordinate structure with more
than two conjuncts: for this reason, the coord-rel resource is both consumed and
produced by this meaning constructor, so that it can be consumed and produced
again by another conjunct:

(59) Meaning constructor for non-seed clausal conjuncts:

[conj0] λP.λQ.λC.C(P, Q(C)) :
↓𝜎 't( ⊸ [(↑𝜎 coord-rel) ⊸ ↑𝜎 't( ] ⊸ [(↑𝜎 coord-rel) ⊸ ↑𝜎 't( ]

To produce a complete and coherent meaning for the sentence, the meaning
constructor in (58) for the conjunction must be consumed in the course of
the derivation, like other meaning resources. According to Asudeh and Crouch
(2002a), this is accomplished by the meaning constructor for the seed conjunct,
which appears in its final form in (60). In fact, the meaning of the seed conjunct
consumes the resource (↑𝜎 coord-rel) contributed by the conjunction, but does
not incorporate the meaning C of the conjunction; the resulting meaning reflects
only the meaning P of the seed conjunct:

13 Winter (2007) provides an illuminating discussion of the complex issues raised by the distinction
between binary coordination (as assumed here and by Asudeh and Crouch 2002a), where a coordinate
structure with three conjuncts A, B, C is given a binary analysis and (A, and (B, C)), as opposed to a “flat”
treatment like and ({A, B, C}). Winter provides evidence indicating that the binary analysis is superior,
though it may be that some of the advantages he adduces for binary structure are better explained by
assuming syntactic ambiguity rather than ambiguity of semantic composition. There is no obstacle to
the formulation of a “flat” treatment of coordination in a glue setting, though we do not provide the
details here; indeed, a “flat” treatment would be simpler in some respects, in that it would not require
multiple copies of the conjunction in assembling the meaning of the coordinate structure.
14 The meaning constructor in (59) produces the meaning ∧(yawn (Chris ), sneeze (David )) for exam-
ple (1), with ∧ as a predicate taking the meanings of the two conjuncts as its arguments. This is a
notational variant of the expression yawn (Chris ) ∧ sneeze (David ), and we will use the two variants
interchangeably.
 coordination and semantic composition 627

(60) Meaning constructor for seed clausal conjunct:

[seed0] λP.λC.P : ↓𝜎 't( ⊸ [(↑𝜎 coord-rel) ⊸ ↑𝜎 't( ]

We now associate the [seed0] and [conj0] meaning constructors with the phrase
structure rule for coordination, exemplifying with a modified version of the rule
macro for English coordination given in (52). Following Asudeh and Crouch
(2002a), we treat the final conjunct as the seed, though nothing rests on this
choice; the first conjunct or any other conjunct could be chosen instead. Given
this choice, the final conjunct is associated with a different meaning constructor
from the nonfinal conjuncts, and so we define a new macro final-conj for the
final conjunct, associated with the [seed0] meaning constructor; the non-final
conjuncts are associated with the [conj0] meaning constructor:

(61) a. Definition of the rule macro cnj(_c):

cnj(_c) ≡ Cnj[main] _c
↑ =↓ ↓∈↑
[conj0]
b. Definition of the rule macro final-cnj(_c):
final-cnj(_c) ≡ Cnj[main] _c
↑ =↓ ↓∈↑
[seed0]
c. Revised rule macro for English clausal coordination:
english-coord(_c) ≡
2 3  
Cnj[pre] _c  _c⃰ @cnj(_c)⃰  @final-cnj
↑ =↓ ↓∈↑ ↓∈↑
[conj0]  
[conj0]

We now go through the glue deduction for the clausal coordination Chris yawned
and David sneezed:

(62) Chris yawned and David sneezed.

 
 

 ’ 
 
 y 
 
 

 
a


 a c[ ]
  
 ’  


 s



 
 

yawn (Chris ) ∧ sneeze (David ) : a

628 coordination

(63) yawn (Chris ) : y𝜎 This is the meaning constructor

[Chris-yawn]. It associates the meaning
yawn (Chris ) with the semantic structure y𝜎 .
λP.λQ.λC.C(P, Q(C)) : y𝜎 ⊸ [[c ⊸ a𝜎 ] ⊸ [c ⊸ a𝜎 ]]
This is the meaning constructor [conj0] asso-
ciated with the first conjunct of the phrase
structure rule for clausal coordination.
λQ.λC.C(yawn (Chris ), Q(C)) : [[c ⊸ a𝜎 ] ⊸ [c ⊸ a𝜎 ]]
Combining [conj0] with [Chris-yawn], we
have this meaning constructor, which we call
[Chris-yawn-conj].
sneeze (David ) : s𝜎 This is the meaning constructor
[David-sneeze]. It associates the meaning
sneeze (David ) with the semantic structure s𝜎 .
λP.λC.P : s𝜎 ⊸ [c ⊸ a𝜎 ] This is the meaning constructor [seed0] asso-
ciated with the final conjunct of the phrase
structure rule for clausal coordination.
λC.sneeze (David ) : [c ⊸ a𝜎 ]
Combining [seed0] with [David-sneeze], we
have this meaning constructor, which we call
[David-sneeze-seed]. Notice that this mean-
ing involves vacuous abstraction over the
meaning of the conjunction C.
λC.C(yawn (Chris ), sneeze (David )) : [c ⊸ a𝜎 ]
On the glue side, [David-sneeze-seed] is
associated with the resource [c ⊸ a𝜎 ], which
is of exactly the form required by Chris-
yawn-conj. Combining David-sneeze-seed
and [Chris-yawn-conj], we obtain this mean-
ing constructor, which requires a resource for
the conjunction C.
∧:c This is the meaning constructor [and]. It pro-
vides the meaning ∧ contributed by the con-
junction and.
yawn (Chris ) ∧ sneeze (David ) : a𝜎
Combining the meaning constructors in the
previous two steps, we have the meaning con-
structor for Chris yawned and David sneezed,
as desired.

16.7.2 Subsentential coordination

As with clausal coordination, coordination of subsentential units involves con-
junction of the meanings of the conjunct phrases. The f-structure and meaning
constructor for the sentence Tracy sang and danced are:
 coordination and semantic composition 629

(64) Tracy sang and danced.

 
 ’

 

 s  

 t


 
 
a 
 
 ’  
 
 

d






sing (Tracy ) ∧ dance (Tracy ) : a

In this example, the subject Tracy of the conjoined verbs sang and danced is shared
across the conjuncts in the coordinate structure, as described in §16.2. Argument
sharing in subsentential coordination presents a special challenge for a resource-
based account of the syntax-semantics interface.
In (64), the subject Tracy is shared by the verbs sang and danced, and each
verb requires a meaning contribution from its subject. Our theory of meaning
assembly and the syntax-semantics interface relies crucially on the assumption that
the meaning constructor for Tracy makes a single, unique meaning contribution.
Clearly, however, the acceptability of (64) entails that this single meaning contri-
bution can satisfy the requirements imposed by each of the verbs in the coordinate
structure sang and danced. Therefore, in the analysis of examples like (64), we
require a theory of resource management in argument sharing that accounts for
the grammaticality of examples like Tracy sang and danced while maintaining the
desirable properties of our linear-logic-based glue approach.
In fact, reliance on a theory of resource management in the analysis of examples
like (64) is of paramount importance: the acceptability of (64) does not indicate
that resource accounting is completely abandoned for the shared argument Tracy.
If resource accounting were switched off entirely for shared dependents—for
example, by prefixing the meaning constructor for Tracy with the linear logic of
course operator, allowing it to be used any number of times—we would have no
way of accounting for the unacceptability of examples like:

(65) ⃰Chris selected and arrived Tracy.

This example is syntactically and semantically incoherent. Syntactic requirements

on verb coordination, discussed in §16.2, entail that Tracy must appear as the
obj argument of the verbs selected and arrived. However, the intransitive verb
arrived does not require an obj resource, and so if an obj resource is provided,
it is not consumed in the meaning derivation and remains unused at the end of the
deduction, leading to semantic incoherence. If resource accounting were switched
off for Tracy, it could contribute one semantic resource to this sentence, and not
the expected two. This would satisfy the requirements of the verb selected, violate
no semantic requirements imposed by the verb arrived, and incorrectly result in
a semantically coherent deduction. This example shows that resource accounting
630 coordination

must in fact be enforced for (65), as it is in all other cases. We require, then, a
complete and explicit theory of resource accounting, argument sharing, and their
interactions.
Particular care must be taken in the treatment of certain kinds of arguments
shared across conjunctions. As noted by Partee (1970), a sentence like Someone
sang and danced has the following meaning:

(66) Someone sang and danced.

a (x, person (x), sing (x) ∧ dance (x))

Here, the single quantifier someone scopes over the coordinate structure, and the
variable x bound by the quantifier appears as an argument of both sing and dance .
This meaning is not the same as the one for the sentence Someone sang and someone
danced:

(67) Someone sang and someone danced.

a (x, person (x), sing (x)) ∧ a (y, person (y), dance (y))

In (67), different people might be involved in each activity, while (66) requires that
there is a single person who both sang and danced. This fact must also be captured
by our theory of argument sharing and semantic composition.
Kehler et al. (1995) were the first to propose a treatment of argument sharing
and resource management within the glue approach, and a similar approach
was adopted by Dalrymple (2001). That proposal appeals to the geometry
of f-structures in constructions involving argument sharing: intuitively, their
approach focuses on occurrences of f-structures, where an f-structure occurs more
than once if there is more than one path leading to it. Semantic resources are
associated with paths through the f-structure and thus with occurrences of f-
structures; in essence, Kehler et al. (1995) provide a means for making one copy
of a semantic resource for each f-structure path leading to its corresponding f-
structure.
For instance, in the analysis of (64), there are two paths leading to the f-structure
for Tracy, since the f-structure for Tracy appears as the value of two different
subj attributes. On Kehler et al.’s analysis, each verb requires a semantic resource
associated with the path leading to its subj. Since the f-structure for Tracy appears
at the end of each of these two paths, two copies of the meaning constructor
for Tracy are made available, and a semantically complete and coherent meaning
deduction results.
This approach successfully accounts for the acceptability of examples involving
argument sharing by allowing the creation of as many semantic resources as are
needed to satisfy the requirements of each predicate involved in the sharing of a
single argument. However, a major problem with this approach is that it also allows
resource duplication in cases where such duplication is unwarranted. For example,
constructions with raising verbs also exhibit argument sharing: the subject of a
verb like seem is also the subject of its xcomp (§15.1). As pointed out by Asudeh
(2000), we do not wish to enforce resource duplication in this case; as discussed
 coordination and semantic composition 631

in §15.2.2, the derivation of the meaning seem (yawn (David )) of a sentence like
David seemed to yawn requires exactly one occurrence of the meaning resource
contributed by David, not two. Similarly, resource duplication is not warranted in
constructions in which an f-structure is involved in a long-distance dependency
and appears as a member of the dis set as well as bearing an argument function.
In these cases, relying simply on the geometry of the f-structure to license feature
duplication leads to the wrong result; a more constrained theory is needed.
Instead, we adopt a variant of the approach to resource management proposed
by Asudeh and Crouch (2002a). This approach provides special rules to combine
the semantic requirements imposed by each predicate that shares an argument into
a requirement for a single semantic resource.
In order to introduce the approach, we propose special-purpose templates
[conj1] and [seed1] for coordinated predicates which share one argument, as in
(68); these templates are special cases of the general approach to be presented later
in this section. These resemble the [seed0] and [conj0] templates in (59) and (60),
except that the meaning constructors for the conjoined phrases each require a subj
argument, and the coordinate structure as a whole requires a subj argument which
is shared across the conjunct daughters. The underlined portion of the glue side
of the meaning constructors in (68) is different from the meaning constructors in
(59) and (60) in that the resources involved are implicational, requiring a subj;
other aspects of their structure are the same. Thus, the meaning constructor in
(68a) combines two conjuncts which are each missing a subject, and produces
a resource for the coordinate structure which is still missing a subject. A single
resource satisfies the missing-subject requirement of the coordinate structure:

(68) a. Meaning constructor for non-seed conjuncts sharing subj, provisional:

[conj1] λP.λQ.λC.λx.C(P(x), Q(C, x)) :
[(↓subj)𝜎 ⊸ ↓𝜎 't( ] ⊸ [(↑𝜎 coord-rel) ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 't( ]] ⊸ [(↑𝜎 coord-rel)
⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 't( ]]
b. Meaning constructor for seed conjunct sharing subj, provisional:
[seed1] λP.λC.λx.P(x) : [(↓subj)𝜎 ⊸ ↓𝜎 't( ] ⊸ [(↑𝜎 coord-rel) ⊸ [(↑ subj)𝜎 ⊸ ↑𝜎 't( ]]

We now associate the components of the English coordination rule macro with
these templates. Below, we will provide a general definition that covers coordina-
tion at various clause levels.

(69) a. Alternative definition of the rule macro cnj(_c), provisional:

cnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[conj1]
b. Alternative definition of the rule macro final-cnj(_c), provisional:
632 coordination

final-cnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[seed1]
c. Alternative rule macro for English coordination, provisional:
english-coord(_c) ≡
2 3  
Cnj[pre] _c  _c⃰ @cnj(_c)⃰  @final-cnj
↑=↓ ↓ ∈↑ ↓∈↑
[conj1]  
[conj1]

Using these modified templates and this rule, we can derive the desired meaning
for Tracy sang and danced, where the requirement for each verb to combine with a
subj is satisfied by a single resource associated with the subject Tracy:

(70) Tracy sang and danced.

 
 

 ’ 
 
 s 
 
 t 

 
a


 a c[ ]
  
 ’  


 d



 
 

sing (Tracy ) ∧ dance (Tracy ) : a

(71) λx.sing (x) : t𝜎 ⊸ s𝜎 This is the meaning constructor [sing].

It associates the meaning λx.sing (x) with
the implicational resource t𝜎 ⊸ s𝜎 , which
requires a subject resource t𝜎 .
λP.λQ.λC.λx.C(P(x), Q(C, x)) : [t𝜎 ⊸ s𝜎 ] ⊸ [[c ⊸ [t𝜎 ⊸ a𝜎 ]] ⊸ [c ⊸ [t𝜎 ⊸ a𝜎 ]]]
This is the meaning constructor [conj1]
associated with the first conjunct of
the phrase structure rule for clausal
coordination.
λQ.λC.λx.C(sing (x), Q(C, x)) : [[c ⊸ [t𝜎 ⊸ a𝜎 ]] ⊸ [c ⊸ [t𝜎 ⊸ a𝜎 ]]]
Combining [conj1] with [sing], we have
this meaning constructor, which we call
[sing-conj].
λx.dance (x) : t𝜎 ⊸ d𝜎 This is the meaning constructor [dance]. It
associates the meaning λx.dance (x) with the
implicational resource t𝜎 ⊸ d𝜎 .
λP.λC.λx.P(x) : [t𝜎 ⊸ d𝜎 ] ⊸ [c ⊸ [t𝜎 ⊸ a𝜎 ]]
This is the meaning constructor [seed1]
associated with the final conjunct of
the phrase structure rule for clausal
coordination.
 coordination and semantic composition 633

λC.λx.dance (x) : [c ⊸ [t𝜎 ⊸ a𝜎 ]]

Combining [seed1] with [dance], we have
this meaning constructor, which we call
[dance-seed]. Notice that this meaning
involves vacuous abstraction over the mean-
ing of the conjunction C.
λC.λx.C(sing (x), dance (x)) : [c ⊸ [t𝜎 ⊸ a𝜎 ]]
On the glue side, [dance-seed] is associated
with the resource [c ⊸ [t𝜎 ⊸ a𝜎 ]], which is
of exactly the form required by [sing-conj].
Combining [dance-seed] and [sing-conj],
we obtain this meaning constructor, which
we call [sing-dance]; it requires a meaning
C for the conjunction c and a meaning x for
the subject.
∧:c This is the meaning constructor [and]. It
provides the meaning ∧ contributed by the
conjunction and.
λx.sing (x) ∧ dance (x) : t𝜎 ⊸ a𝜎
Combining [sing-dance] with [and], we
have the meaning constructor correspond-
ing to the conjunction sang and danced,
which requires a single resource t𝜎 for the
subject.
Tracy : t𝜎 This is the subj meaning constructor
[Tracy].
sing (Tracy ) ∧ dance (Tracy ) : a𝜎
Combining the meaning constructors in the
previous two steps, we have the meaning
constructor for Tracy sang and danced, as
desired.
This derivation shows that it is possible to provide meaning constructors of very
similar shape—[seed0] and [seed1], [coord0] and [coord1]—for conjunction
of what Partee and Rooth (1983) call conjoinable types, sometimes called “types
ending in t”: meanings of type t, 'e → t(, 'e → 'e → t((, and so on:

(72) Definition of Conjoinable Type (Partee and Rooth 1983: 363):

a. t is a conjoinable type
b. if b is a conjoinable type, then for all a, 'a, b( is a conjoinable type.

What is needed, then, is a single rule schema which can be used for coordination
of conjuncts of any conjoinable type: clausal coordination with conjuncts of type t
producing a coordinate structure of type t, predicate coordination with conjuncts
634 coordination

of type 'e → t( producing a coordinate predicate of type 'e → t(, and so on. As
Asudeh and Crouch (2002a) point out, similar proposals have been made in many
theoretical settings (Gazdar 1980; Keenan and Faltz 1985; Emms 1990), though the
approach is most commonly associated with Categorial Grammar treatments of
coordination (Steedman 1985, 2001; Morrill 2011): Steedman (2001: 39) refers to
the “ancient intuition that coordination is an operation that maps two constituents
of like type onto a constituent of the same type.” Building on observations by
Asudeh and Crouch (2002a), we propose the following schematic definitions
of [conj] and [seed], which differ from [conj0]/[conj1] and [seed0]/[seed1]
in the underlined portion; these template schemas allow coordinate structures
of all conjoinable types (resources of type t and any resources that depend
on them):

(73) a. Meaning constructor for non-seed conjuncts, final:

! ! !
[conj] λP.λQ.λC.λ x .C(P( x ), Q(C, x )) :
[' α(n ⊸ ↓𝜎 't( ] ⊸ [(↑𝜎 coord-rel) ⊸ [' α(n ⊸ ↑𝜎 't( ]] ⊸ [(↑𝜎 coord-rel)
! !

⊸ [' α(n ⊸ ↑𝜎 't( ]]

!

b. Meaning constructor for seed conjunct, final:

[seed] λP.λC.λ x .P( x ) : [' α(n ⊸ ↓𝜎 't( ] ⊸ [(↑𝜎 coord-rel) ⊸ [' α(n ⊸ ↑𝜎 't( ]]
! ! ! !

!
In (73), the expression ' α(n represents a sequence of n (zero or more) dependents
!
that are shared across the conjuncts of the coordinate structure; x represents the
corresponding meanings. Importantly, the number of required dependents for the
seed conjunct is inherited by the coordinate structure as a whole, and the fact
that the [conj] macro consumes a resource requiring n dependents and produces
a resource which also requires n dependents ensures that valency requirements
for all conjuncts are shared. Thus, semantic completeness and coherence (§8.7.1)
ensures that requirements for each conjunct are met, with no surplus resources
remaining. For instance, the ungrammaticality of (65) (Chris selected and arrived
Tracy) is due to a mismatch of valency requirements. The seed conjunct arrived
requires only a subj argument, and the schema is instantiated so that the seed
conjunct shares the requirement for a subj argument (and no other arguments)
with the coordinate structure as a whole; the non-seed conjuncts must match this
requirement. The result is that the two verbs selected and arrived share a subj but
not an obj: the resource for Tracy is left unused, violating semantic coherence,
and the requirement imposed by selected for an obj resource is not met, violating
semantic completeness.
As before, these templates are used in the final definitions of the rule macros for
coordinate structures with conjoinable types:
 noun phrase coordination 635

(74) a. Final definition of the rule macro cnj(_c):

cnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[conj]
b. Final definition of the rule macro final-cnj(_c):
final-cnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[seed]
c. Final rule macro for English coordination:
english-coord(_c) ≡
2 3  
Cnj[pre] _c  _c⃰ @cnj(_c)⃰  @final-cnj
↑=↓ ↓∈↑ ↓∈↑
[conj]  [conj] 

This macro allows any dependent that is shared across a coordinate structure to
satisfy the requirements of each conjunct.
Asudeh and Crouch (2002a) discuss a very similar general coordination schema
to the templates in (73), and present two objections to the resulting analysis. First,
they point out that not all coordination involves conjoinable types; in particular,
it has been argued that nominal coordination works differently, and cannot be
captured by this rule schema. We are in agreement with this observation, and in
§16.8.3 we propose a different set of templates for nominal conjunction; however,
this does not constitute motivation for abandoning the analysis of conjoinable
types which we have presented in this section. Second, they propose that more
control over the types involved in coordination is available if a coordination rule
is separately specified for each c-structure category. However, there is no one-to-
one relation between c-structure categories and semantic types: for example, in V
coordination, the type of the coordinate structure depends on the type of the verb
('e → t( for intransitive verbs, ''e → t( → t( for transitive verbs, and so on). This
means that a rule schema like (74) is needed even if separate rules are proposed for
each c-structure category, and we do not see any advantage to proposing a separate
schema for each c-structure category over the general schema in (74). Thus, we do
not consider either of these objections as counting against this approach.

16.8 Noun phrase coordination

Syntactically, noun phrase coordination is similar to clausal coordination in that
it involves set formation at f-structure, with the f-structures of the conjuncts as
members of the set. It differs from clausal coordination in several respects. First,
as discussed in §2.5.7.1, a coordinate noun phrase has its own index features,
which may be different from the index features of its conjuncts. We discuss the
features of coordinated nominal phrases in §16.8.1, building on our discussion of
f-structure features in §2.5. Like simple noun phrases, a coordinate noun phrase
636 coordination

can be the controller of agreement: in some cases, agreement depends on the

resolved (§16.8.1.1) features of the coordinate noun phrase, while in other cases
one of the conjuncts in the coordinate structure (the “distinguished conjunct”)
controls agreement (§16.8.1.2).
Semantically, noun phrase coordination differs from clausal coordination in
involving group formation. We discuss the semantics of noun phrase coordination
in §16.8.2, and provide a glue-based treatment in §16.8.3.

16.8.1 Nominal features in coordination

As discussed in §6.3.2, a coordinate noun phrase has features that may differ from
the features of the individual conjunct phrases: these are nondistributive features.
Besides possibly having both the features preconj and conj contributed by the
preconjunction and conjunction, a coordinated noun phrase has an index feature
encoding the resolved person, number, and gender features of the coordinate
phrase.
Agreement relations may depend on either the resolved syntactic index features
of a coordinate phrase or the features of a “distinguished” conjunct; see Hristov
(2012) for an overview discussion. Additionally, agreement with a coordinate
noun phrase is sometimes partially or completely semantically based, in which
case it does not depend on syntactic features of the coordinate structure or a
distinguished conjunct alone or possibly at all. Sadler (2006) proposes a feature-
based analysis of complex gender resolution patterns in Romanian, and Wechsler
(2008) proposes a set-based analysis of gender resolution in French; both authors
show that a distinction between syntactic and semantic resolution is vital to an
adequate analysis of gend resolution in these languages. In the following, we focus
on the index feature bundle and syntactic feature resolution.

16.8.1.1 Feature resolution in coordination: pers and gend In §2.5.7, we dis-

cussed nominal features and theories of representation and resolution for person
and gender features. Resolution patterns for the pers feature are:

(75) Resolution of the pers feature

first & second = first
first & third = first
second & third = second
third & third = third

Our theory of the representation of the values of the pers feature must account for
this pattern. As discussed in §2.5.7.2, Dalrymple and Kaplan (2000) and Vincent
and Börjars (2007) adopt a set-based representation of the pers feature, as shown
in (76). Sadler (2011) demonstrates that any set-based analysis of complex feature
values can be directly translated to an equivalent feature-based analysis, with a
positive value representing the presence of an element in the set, and a negative
value representing its absence:
 noun phrase coordination 637

(76) Values of the pers feature:

Set-based value Feature-based equivalent
6 7 89
1 +
1: {s,h} pers
2 +
6 7 89
1 −
2: {h} pers
2 +
6 7 89
1 −
3: { } pers
2 −

If we assume a set-based theory of the values of the pers feature, treating resolution
as set union of the values of the pers features of the conjuncts produces the same
pattern as in (75), as desired:

(77) Resolution of the pers feature via set union:

{s,h} (first) ∪ {h} (second) = {s,h} (first)
{s,h} (first) ∪ { } (third) = {s,h} (first)
{h} (second) ∪ { } (third) = {h} (second)
{ } (third) ∪ { } (third) = { } (third)

In the following, we show how to define resolution rules for the pers feature in
a feature-based setting, preserving the essential intuitions underlying the union
operation while retaining the advantages of a feature-based analysis, including
underspecification.

“Union” in a feature-based analysis: Given a feature-based representation

of the value of the pers feature, resolution rules must have the same effect as the
union analysis; if the value of the attributes 1 or 2 is + for any conjunct daughter,
the coordinate structure as a whole has the value + for that attribute:

(78) Resolution of the pers feature in a feature-based approach:

6 7 89 6 7 89 6 7 89
1 + 1 − 1 +
pers (first) & pers (second) = pers (first)
2 + 2 + 2 +
6 7 89 6 7 89 6 7 89
1 + 1 − 1 +
pers (first) & pers (third) = pers (first)
2 + 2 − 2 +
6 7 89 6 7 89 6 7 89
1 − 1 − 1 −
pers (second) & pers (third) = pers (second)
2 + 2 − 2 +
6 7 89 6 7 89 6 7 89
1 − 1 − 1 −
pers (third) & pers (third) = pers (third)
2 − 2 − 2 −
638 coordination

In (78), the resolved value in the third column has a + value for the attribute 1 or
2 whenever one or both of the conjuncts has the value + for that attribute.
In order to impose the proper constraints, we define the following basic tem-
plate, which we call plus:15

(79) plus(_set, _path) ≡ (_set ∈ _path) = + ⇒ (_set _path) = +

(_set ∈ _path) *= + ⇒ (_set _path) = −

In this definition, _set is a set of f-structures and _path is a sequence of f-structure

attributes. The template requires that if a member (∈) of the set _set has a positive
value (+) for _path, then the set itself has a positive value (+) for _path. If no
member has the value + for _path, the set has the negative value (−) for _path.
Given this definition, the template call in (80a) is exactly equivalent to the fully
specified set of constraints in (80b), requiring that if any member of the set of
f-structures ↑ has the value + for the 1 attribute in the index pers feature bundle,
then the set ↑ has the value + for index pers 1, and otherwise the set ↑ has
the value − for index pers 1. This produces the correct result for the resolution
patterns in (78):

(80) a. @plus(↑, index pers 1)

b. (↑ ∈ index pers 1) = + ⇒ (↑ index pers 1) = +
(↑ ∈ index pers 1) *= + ⇒ (↑ index pers 1) = −

We can now define a template @pers-res which ensures the correct pers resolu-
tion patterns: if any conjunct has a positive value for the 1 or 2 attributes of the
pers feature, the set as a whole has a positive value for that attribute. Otherwise,
the set has a negative value for that attribute:

(81) pers-res(_set) ≡ @plus(_set, index pers 1)

@plus(_set, index pers 2)

In this definition, _set is a set of f-structures representing a coordinate structure,

with the conjuncts as members of the set. The template requires that the value of
index pers 1 for the coordinate structure is + if any of the conjuncts has the value
+ for index pers 1, and − otherwise. The same pattern holds for index pers 2.
By the use of these templates, we can preserve the intuitions and advantages
of the union analysis of resolution for the pers feature while also retaining the
advantages of the feature-based approach.
We specify the pers-res template as an additional constraint in the phrase
structure rule for nominal coordination. In (82), we define a new rule macro
english-ncoord for nominal coordination, which we also use in our discussion
of the semantics of nominal coordination in §16.8.2. For explicitness, we associate

15 See §5.2.10 for discussion of implicational constraints using the ⇒ operator; see §6.3.1 for an
explanation of the symbol ∈ used as an attribute.
 noun phrase coordination 639

the pers-res template with the final conjunct, but since the template refers only to
the f-structure ↑, it could appear on any of the daughters:

(82) Rule macro for English nominal coordination including person resolution
constraints:
english-ncoord(_c) ≡
2 3 4 5
Cnj[pre] _c _c⃰ @cnj(_c)⃰ @cnj(_c)
↑=↓ ↓∈↑ ↓∈↑ @pers-res(↑)

“Intersection” in a feature-based analysis: As discussed in §2.5.7.3, set

values have also been proposed for the gend feature, with resolution modeled as
either union (Dalrymple and Kaplan 2000) or intersection (Vincent and Börjars
2007). Corbett (1983) proposes the following generalization for gender resolution
in Icelandic:

(83) Gender resolution in Icelandic:

If the conjuncts are all masculine, the masculine form is used.
If the conjuncts are all feminine, the feminine form is used.
Otherwise the neuter form is used.

Vincent and Börjars (2007) propose a set-based treatment of gender resolution,

relying on the values in (84) for the gend feature:

(84) Values for the gend feature in Icelandic according to Vincent and Börjars
(2007):
masculine: {m}
feminine: {f}
neuter: {}

Given these set values, gender resolution can be modeled as set intersection:

(85) Resolution of the gend feature as set intersection (Vincent and Börjars
2007):
{m} (masculine) ∩ {m} (masculine) = {m} (masculine)
{f} (feminine) ∩ {f} (feminine) = {f} (feminine)
{ } (neuter) ∩ { } (neuter) = { } (neuter)
{m} (masculine) ∩ {f} (feminine) = { } (neuter)
{m} (masculine) ∩ { } (neuter) = { } (neuter)
{f} (feminine) ∩ { } (neuter) = { } (neuter)

In a feature-based setting, the corresponding values for masculine, feminine, and

neuter gender are as in (86), with a positive (+) value representing the presence
of a set element, and a negative (−) value representing its absence; as discussed
by Dalrymple and Kaplan (2000) and Vincent and Börjars (2007), fewer or more
640 coordination

features may be necessary when analyzing simpler or more complex gender

systems:

(86) Feature-based values for the gend feature in Icelandic:

7 8
m +
masculine:
f −
7 8
m −
feminine:
f +
7 8
m −
neuter:
f −

In order to produce the same effect as set intersection in a feature-based setting,

we first define a macro which we call minus:

(87) minus(_set, _path) ≡ (_set ∈ _path) *= − ⇒ (_set _path) = +

(_set ∈ _path) = − ⇒ (_set _path) = −

In this definition, _set is a set of f-structures and _path is a sequence of f-structure

attributes. The template specifies that if no member of the set _set has a negative
value − for _path (in this case, all of the members of the set have the value +
for _path), then the set itself has the value + for _path. If any member has the
negative value − for _path, the set itself has the negative value − for _path.
We can make use of the template minus to define a gender resolution template
gend-res for Icelandic; additional constraints may be needed for languages with
more complex gender systems:

(88) gend-res(_set) ≡ @minus(_set, index m)

@minus(_set, index f)

In this definition, _set is a set of f-structures. The template requires that the value
of index m for the set is + if all of the conjuncts have the value + for index m, and
− otherwise. The same pattern holds for index f.
The gender resolution template in (88) is associated with one of the conjuncts
in the Icelandic nominal coordination rule; because the template refers only to the
mother f-structure ↑, it can be associated with any of the conjuncts.

16.8.1.2 Distinguished conjunct agreement Agreement with coordinate noun

phrases often depends on the syntactic or semantic features of the coordinate
phrase; however, other patterns are also found. Sadler (1999, 2003) was among
the first to propose an LFG-based analysis of distinguished conjunct agreement,
constructions in which the controller of agreement is one of the conjuncts of
a coordinate phrase. The distinguished conjunct is either the initial or the final
conjunct: often (but not always), it is the conjunct that is closest to the agreement
target (the initial conjunct if the agreement target precedes the coordinate phrase,
 noun phrase coordination 641

and the final conjunct if it follows the coordinate phrase), and so this agreement
pattern is sometimes called “closest conjunct agreement.”
Sadler (1999, 2003) examines agreement patterns in Welsh, a VSO language in
which only pronominals control agreement: the verb agrees with the pronominal
subject in (89a), and shows default (third person singular) agreement in (89b).

(89) a. Daethan (nhw).

came.3pl (they)
‘They came.’
b. Daeth y dynion.
came.3sg the men
‘The men came.’

With coordinated subjects, verb agreement does not reflect the resolved features
of the coordinate phrase; instead, the verb agrees with the initial conjunct if it
is pronominal (examples 90a,b) and shows default agreement otherwise (exam-
ple 90c):

(90) a. Daethost ti a minnau/Siôn.

came.2sg you and I/Siôn
‘You and I/Siôn came.’
b. Roeddwn i a Mair i briodi.
was.1sg 1sg and Mair to marry
‘I and Mair were to marry.’
c. Roedd Mair a fi i briodi.
was.3sg Mair and 1sg to marry
‘Mair and I were to marry.’

A potential (but incorrect) analysis of these patterns is that verb agreement does in
fact reflect the features of the coordinate structure as a whole, but that in examples
like (90), the coordinate structure is unexpectedly associated with the features of
just one of its conjuncts, rather than the resolved features. For example, on this
view the coordinate phrase i a Mair ‘I and Mair’ in (90b) would have the features
of a first person singular phrase, matching the features of the initial conjunct.
Sadler (2003) demonstrates convincingly that such an analysis is not viable,
showing that distinguished conjunct agreement patterns as in (90) can coexist in
the same clause with resolved agreement patterns. In (91), the passive auxiliary
chaffodd shows third person singular (distinguished conjunct) agreement with the
initial conjunct e ‘he’, while the nonfinite main verb is preceded by an anaphoric
form eu showing third person plural (resolved) agreement with the coordinate
phrase e a’i milwyr ‘he and his soldiers’:

(91) Ni chaffodd e a’i milwyr eu lladd yma.

neg got.3sg he and-3sg.m soldiers 3pl kill there
‘He and his soldiers were not killed there.’
642 coordination

Similarly, in (92) the verb gwelais shows first person singular (distinguished
conjunct) agreement with the initial conjunct i ‘I’, and the reflexive anaphor ein
hunain shows first person plural (resolved) agreement with the coordinate phrase
i a’m brawd ‘I and my brother’:

(92) Gwelais i a’m brawd ein hunain.

saw.1sg I and.1sg brother 1pl self
‘I and my brother saw ourselves.’

Thus, the coordinate phrase must be analyzed as having the expected resolved
features rather than the features of one of the conjuncts, since the resolved features
are referenced in auxiliary agreement in (91) and anaphoric agreement in (92).
Arnold et al. (2007) provide an in-depth discussion of Brazilian Portuguese data
that show even more complex patterns. Brazilian Portuguese allows resolved agree-
ment with the features of a coordinate structure as well as distinguished conjunct
agreement with the initial conjunct or the final conjunct. Arnold et al. provide
(93) to show that distinguished conjunct agreement with different conjuncts can
coexist in the same structure; a prenominal determiner can exhibit distinguished
conjunct agreement with the initial conjunct, while a postnominal adjective can
exhibit distinguished conjunct agreement with the final conjunct:

(93) os corações e mentes brasileiras

the.m.pl hearts.m.pl and minds.f.pl Brazilian.f.pl
‘the Brazilian hearts and minds’

Thus, our theory must allow reference to the syntactic and semantic features of
a coordinated noun phrase (in resolved agreement), as well as the syntactic and
semantic features of the initial and final conjuncts (in distinguished conjunct
agreement).
Kuhn and Sadler (2007) review earlier proposals for the treatment of distin-
guished conjunct agreement, dividing them into representation-based approaches
and description-based approaches. Representation-based approaches augment the
f-structure of coordinate nominal phrases with additional features to encode
agreement information for the distinguished conjunct. Kuhn and Sadler (2007)
discuss several problems with representation-based approaches: these additional
features are purely for bookkeeping, and introduce unneeded complexity into
the f-structure representation; further, features for both the initial conjunct and
the final conjunct would be needed to handle examples like the Brazilian Por-
tuguese example in (93), where material on the left agrees with the initial con-
junct, and material on the right agrees with the final conjunct. Kuhn and Sadler
(2007) propose a description-based approach which introduces a new kind of
structure, a local f-structure sequence, in the representation of coordinate struc-
tures; they also propose an expanded classification of f-structure features as
left-peripheral, right-peripheral, or proximity-based, in addition to the standard
distributive/nondistributive classification. Dalrymple and Hristov (2010) build on
Kuhn and Sadler’s insights in a revised proposal that does not depend on an
 noun phrase coordination 643

enriched feature classification, but allows constructional specification of agree-

ment requirements as referring to the initial or final conjunct.

16.8.2 Semantics of noun phrase coordination

Much work has been done on the semantics of coordinated noun phrases. A
clearly presented and useful overview is given by Winter (2001); as he notes,
approaches to nominal coordination generally fall into two categories. The first
type of analysis assumes that conjunctions such as and are ambiguous, mak-
ing different semantic contributions in coordinating conjoinable types (“ending
in t”) and in nominal coordination, while the second type of analysis (advocated
by Winter 2001) assumes that there is only one semantic contribution for and
regardless of whether it coordinates clausal or nominal phrases. The prospect of
a uniform analysis of coordination is explored in very interesting work by Winter
(1998, 2001), Champollion (2016), and others. However, these analyses postulate
a variety of unpronounced operators to shift and augment the meaning of the
nominal conjuncts in various ways, and it is not yet clear whether any or all of
these unpronounced operators are necessary or desirable in a glue setting. In the
following, we explore a constructional analysis that treats some cases of nominal
coordination as “group-forming,” combining individuals of type e to produce a new
e-type individual, while also preserving the standard meanings of and and or (see
also Partee and Rooth 1983, Appendix B; Hoeksema 1988).16
In fact, treating coordination of “conjoinable types” (types “ending in t”) differ-
ently from at least some cases of nominal coordination is a natural move in the
context of the glue approach to the syntax-semantics interface and the treatment
of quantifiers discussed in Chapter 8, since it allows us to treat coordination
involving quantifiers in the same way as coordination of non-quantifiers, without
any necessary modification to the treatment of quantification and scope. Recall the
glue treatment of quantifiers such as everyone or every student, illustrated in (65),
§8.8.1.2, repeated here with the types explicitly indicated:

(94) Everyone yawned.

𝜙
IP 6 9
pred ‘yawn'subj(’
NP I! y 0 1
subj e pred ‘everyone’
N VP
𝜎
Everyone V
every (x, person (x), yawn (x)) : y𝜎 [ ]
yawned

[everyone] λS.every (x, person (x), S(x)) : ∀H 't( .[e𝜎 'e( ⊸ H 't( ] ⊸ H 't(
[yawn] λz.yawn (z) : e𝜎 'e( ⊸ y𝜎 't(

16 Not all cases of nominal coordination are best treated as group-forming; a standard example which
must be treated as conjunction of predicates is the president and CEO, which refers to a single individual
who is both the president and the CEO (the (x.president (x) ∧ CEO (x))). For discussion of such examples,
see Bergmann (1982), Dowty (1988), King and Dalrymple (2004), and Heycock and Zamparelli (2005),
among others.
644 coordination

Notably, the semantic structure e𝜎 for the NP everyone corresponds to the e-type
variable bound by the quantifier: it is a syntactically accessible reflex of the e-type
component of a quantifier meaning. When a quantifier appears as a conjunct in
nominal coordination, it is the e-type component of its meaning that corresponds
to the syntactic structure involved in coordination. This means that our proposal
does not require the complicated machinery for meaning assembly in quantifier
coordination which is required by other analyses, including Hoeksema (1988).
We adopt a variant of Asudeh and Crouch’s (2002a) proposed semantics for
e-type coordination:

(95) a. David and Tracy met.

∃X.David ≤ X ∧ Tracy ≤ X ∧ meet (X)
b. David or Tracy yawned.
∃X.David ≤ X ∨ Tracy ≤ X ∧ yawn (X)

Here, the coordinate structure introduces a possibly complex individual X of type

e. In (95a), X is composed of the atomic e-type individuals David and Tracy , while
in (95b), X is either the individual David or the individual Tracy . We do not take a
position on the precise interpretation of the “part-of ” relation ≤ in (95); according
to Link’s (1983) lattice-theoretic treatment, it can be interpreted as the “individual
part” relation, so that in (95a), the group X is the plural individual David ⊗ Tracy .
Landman (1989) and Schwarzschild (1994) provide more discussion of this issue
and an alternative view.
As we will see, this proposal extends straightforwardly to cases in which indefi-
nite singulars are coordinated. We propose this meaning for the sentence A student
and a professor met:

(96) A student and a professor met.

a (x, student (x), a (y, professor (y), ∃X.x ≤ X ∧ y ≤ X ∧ meet (X)))

This sentence means that a group consisting of the individuals x and y met, where
x is a student and y is a professor. We can also coordinate a proper name with an
indefinite:

(97) David and a professor met.

a (y, professor (y), ∃X.David ≤ X ∧ y ≤ X ∧ meet (X))

Here, the group that met consists of y, a professor, and the individual David .
This approach works well for these simple cases. As noted by Schwarzschild
(1994: 22), complications arise in the treatment of other cases; in particular, the
quantifier no is problematic:

(98) No soldier and no officer met.

Does not mean:
no (x, soldier (x), no (y, officer (y), ∃X.x ≤ X ∧ y ≤ X ∧ meet (X)))
 noun phrase coordination 645

The meaning representation given in (98) can be paraphrased as ‘There is no

soldier that met no officer’ (that is, ‘Every soldier met some officer’), clearly not
a possible meaning for this sentence. In fact, the scoping problem posed by this
example can be circumvented if (98) is analyzed as an instance of branching
quantification (Barwise 1979), in which neither quantifier takes scope over the
other; many examples that have been taken to exemplify branching quantification
involve coordinated noun phrases. We will not attempt to provide a full analysis of
these cases here, but will restrict ourselves to providing an analysis of the simpler
cases, in the belief that getting a firm handle on simpler cases of noun phrase
coordination will ultimately enable a more revealing analysis of those cases that
are more complex. For discussion of the semantics of noun phrase coordination
and plurality, see Krifka (1990), Schwarzschild (1994), Lasersohn (1995), Winter
(1998, 2007), Champollion (2016), and references cited in those works.

16.8.3 Noun phrase coordination and meaning assembly

The f-structure and meaning constructor for the sentence David and Tracy
met are:

(99) David and Tracy met.

 
’
  
 
  
   c[ ]
 
m  
 a
 a
 d  cr[ ]
  
  

t 

∃X.David ≤ X ∧ Tracy ≤ X ∧ meet (X) : m

As with our analysis of clausal coordination, we assume that one of the conjuncts is
the seed; again, we treat the final conjunct as the seed, though nothing depends on
this choice. The meanings of the non-seed conjuncts are then added to the meaning
of the coordinate structure one by one. Partee and Rooth (1983) note that it is
not generally assumed that there are “e-conjoinable” types (types “ending in e”)
other than e itself (in particular, Montague (1973) does not propose any additional
“e-conjoinable” types), so we provide only the definition for e-type conjunction
here, and do not allow for complex types “ending in e.”
The meaning for (99) has a structure similar to a generalized quantifier, with
constraints on the composition of X analogous to the restriction of the quantifier,
and the predication on X as its scope (§8.4.1.4). Thus, for nominal coordination
we introduce a semantic structure attribute coord-restr of type t, associated with
constraints on the composition of the group X (for example, that David and Tracy
are individual parts of X). There are two meaning constructors associated with the
seed, which we call [nseed1] and [nseed2]. The meaning constructor [nseed1]
requires the meaning of the seed conjunct (here, Tracy) to be a component of
646 coordination

the coordinate structure meaning X. The meaning constructor [nconj] is asso-

ciated with the non-seed conjuncts (here, there is only one non-seed conjunct,
David) and asserts that the meanings of each of the non-seed conjuncts are also
components of the meaning of the coordinate structure. Finally, the meaning
constructor [nseed2] consumes the coord-restr meaning and its dependency
on the coordinate structure ↑𝜎 , producing a quantifier meaning ([↑𝜎 'e( ⊸ S] ⊸ S)
which combines the coordinate structure with its predicate (here, meet):

(100) a. Meaning constructor for non-seed nominal conjuncts:

[nconj] λx.λQ.λC.λX.C(x ≤ X, Q(C, X)) :
↓𝜎 'e( ⊸ [(↑𝜎 coord-rel) ⊸ [↑𝜎 'e( ⊸ (↑𝜎 coord-restr)]] ⊸ [(↑𝜎 coord-rel)
⊸ [↑𝜎 'e( ⊸ (↑𝜎 coord-restr)]]
b. First meaning constructor for seed nominal conjunct:
[nseed1] λx.λC.λX.x ≤ X : ↓𝜎 'e( ⊸ [(↑𝜎 coord-rel) ⊸ [↑𝜎 'e( ⊸ (↑𝜎 coord-restr)]]
c. Second meaning constructor for seed nominal conjunct:
[nseed2] λP.λQ.∃X.P(X) ∧ Q(X) : ∀S.[↑𝜎 'e( ⊸ (↑𝜎 coord-restr)] ⊸ [↑𝜎 'e( ⊸ S] ⊸ S
d. Definition of the rule macro ncnj(_c):
ncnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[nconj]
e. Definition of the rule macro final-ncnj(_c), including person and
gender resolution macros (§16.8.1.1):
final-ncnj(_c) ≡ Cnj[main] _c
↑=↓ ↓∈↑
[nseed1]
[nseed2]
@pers-res(↑)
@gend-res(↑)
f. Full rule macro for English nominal coordination:
english-ncoord(_c) ≡
2 3  
Cnj[pre] _c  _c⃰ @ncnj(_c)⃰  @final-ncnj(_c)
↑=↓ ↓∈↑ ↓∈↑
[nconj]  
[nconj]

Thus, we have the following deduction of the meaning of David and Tracy met,
using the f-structure labels introduced in (99):

(101) David : d𝜎 This is the meaning constructor [David].

It associates the meaning David with the
semantic structure d𝜎 .
 noun phrase coordination 647

λx.λQ.λC.λX.C(x ≤ X, Q(C, X)) : d𝜎 ⊸ [[c ⊸ [a𝜎 ⊸ cr]] ⊸ [c ⊸ [a𝜎 ⊸ cr]]]

This is the meaning constructor [nconj].
It consumes the resource for the con-
junct David and produces a modifier of
the coord-restr meaning cr, which itself
depends on the meaning of the conjunction c
and the meaning of the coordinate structure
a𝜎 .
λQ.λC.λX.C(David ≤ X, Q(C, X)) : [[c ⊸ [a𝜎 ⊸ cr]] ⊸ [c ⊸ [a𝜎 ⊸ cr]]]
Combining [David] and [nconj], we have
this meaning constructor, which we call
[David-conj].
Tracy : t𝜎 This is the meaning constructor [Tracy].
It associates the meaning Tracy with the
semantic structure t𝜎 .
λx.λC.λX.x ≤ X : t𝜎 ⊸ [c ⊸ [a𝜎 ⊸ cr]]
This is the meaning constructor [nseed1].
It consumes the resource for the seed con-
junct Tracy to produce a meaning for the
coord-restr meaning cr which depends on
c and a𝜎 .
λC.λX.Tracy ≤ X : [c ⊸ [a𝜎 ⊸ cr]]
Combining [Tracy] and [nseed1], we have
this meaning constructor, which we call
[Tracy-conj].
λC.λX.C(David ≤ X, Tracy ≤ X) : c ⊸ [a𝜎 ⊸ cr]
Combining [David-conj] and [Tracy-conj],
we have this meaning constructor, which we
call [David-Tracy-conj].
∧:c This is the meaning constructor [and]. It
provides the meaning ∧ contributed by the
conjunction and.
λX.David ≤ X ∧ Tracy ≤ X : a𝜎 ⊸ cr
Combining [David-Tracy-conj] and [and],
we have this meaning constructor, which we
call [David-and-Tracy].
λP.λQ.∃X.P(X) ∧ Q(X) : ∀S.[a𝜎 ⊸ cr] ⊸ [a𝜎 ⊸ S] ⊸ S
This is the meaning constructor [nseed2].
λQ.∃X.David ≤ X ∧ Tracy ≤ X ∧ Q(X) : ∀S.[a𝜎 ⊸ S] ⊸ S
Combining [David-and-Tracy] and
[nseed2], we have this meaning constructor,
which we call [David-and-Tracy-quant].
648 coordination

λx.meet (x) : a𝜎 ⊸ m𝜎 This is the meaning constructor [meet].

∃X.David ≤ X ∧ Tracy ≤ X ∧ meet (X) : m𝜎
Combining [David-and-Tracy-quant] and
[meet], we have produced the desired mean-
ing for the sentence.
Next we consider (102), which differs from the example just discussed in that a
quantifier is coordinated with a proper name:

(102) David and a professor met.

 
’
  
 
 
  
  
   
m  d 
 a  
 
  
  
  − 
 p 


a (y, professor (y), ∃X.David ≤ X ∧ y ≤ X ∧ meet (X)) : m

The meaning constructors for this example combine as shown in (103). The
general form of the proof is very close to (101), and in fact the first three steps are
exactly the same, since in both examples the non-seed conjunct is David. We do not
show the details of the combination of the determiner a with the noun professor;
see §8.8.2 for discussion and illustration of how the meaning constructors for
determiners and nouns are combined:

(103) David : d𝜎 This is the meaning constructor [David].

It associates the meaning David with the
semantic structure d𝜎 .
λx.λQ.λC.λX.C(x ≤ X, Q(C, X)) : d𝜎 ⊸ [[c ⊸ [a𝜎 ⊸ cr]] ⊸ [c ⊸ [a𝜎 ⊸ cr]]]
This is the meaning constructor [nconj].
It consumes the resource for the con-
junct David and produces a modifier of
the coord-restr meaning cr, which itself
depends on the meaning of the conjunction
c and the meaning of a𝜎 .
λQ.λC.λX.C(David ≤ X, Q(C, X)) : [[c ⊸ [a𝜎 ⊸ cr]] ⊸ [c ⊸ [a𝜎 ⊸ cr]]]
Combining [David] and [nconj], we have
this meaning constructor, which we call
[David-conj].
h : [p𝜎 ] We temporarily introduce the hypothetical
meaning h corresponding to p𝜎 for the seed
conjunct (see §8.8.1.4). The glue side of a
meaning constructor is marked as a hypo-
thetical premise by square brackets. We must
discharge this assumption later in the proof.
 noun phrase coordination 649

λx.λC.λX.x ≤ X : p𝜎 ⊸ [c ⊸ [a𝜎 ⊸ cr]]

This is the meaning constructor [nseed1].
It consumes the meaning resource of the
seed conjunct to produce a meaning for the
coord-restr meaning cr which depends on
c and a𝜎 .
λC.λX.h ≤ X : [c ⊸ [a𝜎 ⊸ cr]]
Combining the meaning constructors in the
previous two steps, we have this meaning
constructor, which we call [h-conj].
λC.λX.C(David ≤ X, h ≤ X) : c ⊸ [a𝜎 ⊸ cr]
Combining [David-conj] and [h-conj], we
have this meaning constructor, which we call
[David-h-conj].
∧:c This is the meaning constructor [and]. It
provides the meaning ∧ contributed by the
conjunction and.
λX.David ≤ X ∧ h ≤ X : a𝜎 ⊸ cr
Combining [David-h-conj] and [and], we
have this meaning constructor, which we call
[David-and-h].
λP.λQ.∃X.P(X) ∧ Q(X) : ∀S.[a𝜎 ⊸ cr] ⊸ [a𝜎 ⊸ S] ⊸ S
This is the meaning constructor [nseed2].
λQ.∃X.David ≤ X ∧ h ≤ X ∧ Q(X) : ∀S.[a𝜎 ⊸ S] ⊸ S
Combining [David-and-h] and [nseed2],
we have this meaning constructor, which we
call [David-and-h-quant].
λx.meet (x) : a𝜎 ⊸ m𝜎 This is the meaning constructor [meet].
∃X.David ≤ X ∧ h ≤ X ∧ meet (X) : m𝜎
Combining [David-and-h-quant] and
[meet], we have produced a meaning for
the sentence containing the hypothesized
meaning h. We must now discharge this
premise, which we introduced earlier.
λh.∃X.David ≤ X ∧ h ≤ X ∧ meet (X) : p𝜎 ⊸ m𝜎
This is the result of discharging the premise
introduced in the fourth step of the proof.
λS.a (y, professor (y), S(y)) : ∀H.[p𝜎 ⊸ H] ⊸ H
This is the meaning constructor
[a-professor], the result of combining
the meaning constructors for a and professor
(see §8.8.2).
a (y, professor (y), ∃X.David ≤ X ∧ y ≤ X ∧ meet (X)) : m𝜎
Combining the meaning constructors in
the previous two steps gives us the desired
meaning for the sentence.
650 coordination

The meaning deduction for an example such as (96), in which two quantifiers are
coordinated, proceeds similarly; we do not provide the proof here.

16.9 Further reading and related issues

Besides the seminal work of Bresnan et al. (1985b), Andrews (1983a) contributed
early and influential (although unfortunately unpublished) work on coordination
in LFG. The 2004 LFG conference hosted a workshop “Coordination and Agree-
ment,” organized by Peter Peterson, at which a number of issues related to the
syntax and semantics of coordination were discussed; some of the workshop papers
are available online (Butt and King 2004a), including the overview presentation by
Peterson (2004b).
Belyaev (2015) explores clausal coordination and subordination constructions
in Ossetic, and shows that these phenomena provide a lens on the overall archi-
tecture of grammar: the Ossetic data are best explained by appeal to a three-
way distinction in grammatical levels, distinguishing c-structure, f-structure, and
semantic structure, as assumed in LFG. Belyaev provides a close examination of
constructions which are coordinate at c-structure but subordinate at f-structure
and semantically, as well as constructions which are coordinate at both c-structure
and f-structure but subordinate semantically.
Although coordinate structures are often distinguished by the presence of a
conjunction, this is not the only possibility. Brown and Dryer (2008) describe
so-called “and-verbs” in Walman, which are morphologically verbs but are used
to coordinate noun phrases. Lawyer (2010) presents an LFG-based analysis of
Walman and-verbs and serial verb constructions.
A particularly challenging type of coordinate construction, which Mel’čuk
(1988) calls lexico-semantic coordination, involves coordination of phrases with
different grammatical roles; this construction is attested in Hungarian and Polish,
as shown in (104), as well as Russian and other languages. For example, in (104a)
the first conjunct in the clause-initial coordinate phrase bears the subj role, and
the second conjunct bears the obj role:

(104) a. Ki és mit talált fel?

who and what invent.pst vm
‘Who invented something and what was it?’
(Hungarian: Gazdik 2011: 43)
b. Kogo i komu przedstawił?
who.acc and who.dat introduced
‘Who did he introduce to whom?’
(Polish: Patejuk 2015: 80, citing Kallas 1993)
 further reading and related issues 651

c. czy komukolwiek, kiedykolwiek i do czegokolwiek

particle anybody.dat anytime and for anything
przydał sie˛ poradnik
come.in.handy guide
‘Has a(ny) guide ever come in handy to anybody for anything?’
(Polish: Patejuk 2015: 80, citing Kallas 1993)

The challenge in the analysis of such structures is to ensure that each conjunct bears
the appropriate f-structure role. For in-depth discussion and analysis of lexico-
semantic coordination from an LFG perspective, see Gazdik (2011) and Patejuk
(2015). Patejuk (2015) goes beyond lexico-semantic coordination to provide a
more general discussion and analysis of examples in which the conjuncts differ in
some respect, including not only conjuncts with different grammatical functions,
but also conjuncts with the same grammatical function but different cases or
different c-structure categories; see also Przepiórkowski and Patejuk (2012) and
Patejuk and Przepiórkowski (2012).
The relation between apposition and coordination is explored in detail by Sadler
and Nordlinger (2006a), who illustrate their discussion with examples from a
range of Australian languages. Building on this foundation, Nordlinger and Sadler
(2008) and Sadler and Nordlinger (2010) explore the use of sets as an appropriate
representation not only of coordinate structures, but also of apposition, generic-
specific constructions, inclusory constructions, and other related constructions.
Some early LFG work on coordination is based on syntactic assumptions that
have since been abandoned. Kaplan and Maxwell (1988b) proposed a theory of
function application in coordination that does not rely on a distinction between
distributive and nondistributive features or on the definition presented in (11) for
application of a function involving a distributive feature to a set; instead, they pro-
pose that the properties of a set are defined as the generalization of the properties
of its elements (see §6.9.3 for definition and discussion of generalization). Many
of the predictions of the Kaplan and Maxwell (1988b) theory are indistinguishable
from the theory presented in this chapter. However, Kaplan and Maxwell’s theory
makes unwanted predictions involving constraining properties of sets, and also has
difficulty in cases where a set has a property that is different from its conjuncts; see
Dalrymple and Kaplan (2000) for more discussion.
 17
Long-distance dependencies

Topicalization and left- or right-dislocation constructions, relative clauses, and

constituent questions in English and many other languages exemplify long-distance
dependencies, constructions in which a constituent appears in a position other than
the one with which its syntactic function is usually associated. In this sense, it is
displaced. For example, in an English topicalization construction like Chris, David
likes, the initial constituent Chris plays two roles: it simultaneously has the status
of a displaced phrase (dis) and is the object of the verb likes. Since many syntactic
constraints in long-distance dependency constructions are definable in terms of
the grammatical function of the displaced phrase and the f-structural path involved
in the long-distance dependency, f-structural constraints on the relation between
a displaced constituent and its within-clause functional role will feature heavily
in our syntactic discussion. Section 17.1 discusses the syntax of long-distance
dependencies, showing how the syntactic relation is established between the
fronted phrase and its within-clause grammatical function. Section 17.2 discusses
constructions in which the displaced phrase is related not to a “gap” within the
clause, but to a resumptive pronoun. Section 17.3 discusses cases in which a long-
distance dependency is signaled by special morphological marking: in particular,
we discuss and analyze long-distance dependencies in Kikuyu, where sentences
with long-distance dependencies exhibit a special tonal change, and in Irish, where
the form of complementizers is dependent on the exact type of long-distance
dependency involved.
In the earliest LFG work on long-distance dependencies (for example, Kaplan
and Bresnan 1982), it was assumed that the relation between a displaced con-
stituent and its corresponding within-clause “gap” was definable in constituent
structure terms, using the double arrow notation ⇑ and ⇓ to relate the displaced
constituent to the position of the “gap.” Kaplan and Zaenen (1989b) showed that
this treatment made it difficult to account for functional constraints on long-
distance dependencies, and the original analysis based on c-structure relations
and defined in terms of the ⇑ and ⇓ notation was subsequently abandoned;
see Dalrymple et al. (1995d) for more discussion of the history of the analy-
sis of long-distance dependencies in LFG. However, even though the primary
constraints on long-distance dependencies are formulated in f-structure terms,
some LFG analyses assume the existence of traces, phonologically null c-structure
elements corresponding to the within-clause position of the displaced constituent.
In §17.4, we discuss evidence for and against traces, with particular attention to
the phenomenon of weak crossover. The final syntactic construction we examine

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 syntax of long-distance dependencies 653

in §17.5 involves multiple gaps corresponding to a single filler, including “across-

the-board” extraction and parasitic gaps.
We then turn to a discussion of the semantics of constructions involving long-
distance dependencies. Section 17.6 discusses the semantics of relative clauses
and meaning composition, and §17.7 discusses issues that arise in the semantic
treatment of constituent questions.

17.1 Syntax of long-distance dependencies

17.1.1 Topicalization and dislocation
17.1.1.1 Topicalization We begin our discussion of long-distance dependencies
with the English topicalization construction, in which a displaced constituent
appears at the beginning of the sentence. The term “topicalization” is something
of a misnomer: while the initial constituent can be the topic of the sentence, it
is often the case that it has focus status (Prince 1981). We retain the standard
term “topicalization” for this construction, but without the implication that the
displaced constituent bears any particular information structure role.
Following early transformational analyses, the displaced constituent is some-
times spoken of as having been “fronted” or “extracted,” and we will also use
this terminology in our discussion. The fronted phrase also plays a grammatical
role within the clause, according to the Extended Coherence Condition (§10.3.3),
originally proposed by Zaenen (1980) (see also Fassi-Fehri 1988):

(1) Extended Coherence Condition:

focus and topic must be linked to the semantic predicate argument
structure of the sentence in which they occur, either by functionally or
by anaphorically binding an argument.

We do not include the features topic and focus at f-structure, instead representing
them at the separate level of i-structure (Chapter 10). Rather, we use the overlay
function dis to represent long-distance dependencies in f-structure in this book;
see §2.1.11. We therefore assume that the Extended Coherence Condition applies
to dis rather than to topic and focus: dis must be linked to the semantic predicate
argument structure of the sentence in which it occurs, either by functionally or by
anaphorically binding an argument.
In this section, we will examine the syntax of the topicalization construction in
English. Examination of this construction reveals c-structural constraints on the
permitted constituent structure categories of the fronted constituent as well as f-
structural constraints on the path relating the fronted constituent to its within-
clause grammatical function. Constraints on the topicalization construction in
other languages may differ: as we will see, other languages may allow a different
654 long-distance dependencies

set of phrasal categories to appear in the relevant position or may place different
constraints on the f-structural relation between the displaced constituent and its
within-clause role.

Category of the displaced phrase: The fronted phrase in the English

topicalization construction may be one of several phrase structure categories:

(2) a. NP: Chris, I like.

b. PP: To Chris, I gave a book.
c. AP: Happy, Chris will never be.
d. CP: That Chris was a movie star, I never would have guessed.
e. VP: ?To leave, we convinced Chris. (acceptable for some speakers; see
§15.3.2)

The sentences in (2) exemplify the permitted range of phrase structure categories
in this construction:

(3) Phrasal category of displaced phrases in the English topicalization construc-

tion: NP, PP, AP, CP, VP

Grammatical function of the displaced phrase: The within-clause

grammatical function of the fronted phrase is also constrained: some functions
can be related to the displaced phrase, whereas others cannot. For instance, the
fronted constituent can fill the role of obj:

(4) Chris, we like.

’

In this sentence, Chris is the obj of the verb like, but this phrase is displaced: it does
not appear in the usual immediately postverbal position associated with objects in
English. The f-structure for Chris is a member of the dis set, and there is also a path
through the f-structure defining its within-clause grammatical function; in (4), the
path is obj. In (5), the displaced phrase is the obj in the subordinate clause, comp,
so that the path leading to the within-clause function of the displaced phrase is
comp obj:
 syntax of long-distance dependencies 655

(5) Chris, we think that David saw.

’

Not all within-clause functions can be related to a member of the dis set, however.
As discussed in §6.6, for some speakers it is possible for a displaced phrase to be
related to a position within the comp of a so-called “bridge verb” like think, but
not to a position within the comp of a nonbridge verb like whisper. The distinction
between bridge and nonbridge verbs is not reflected in the grammatical function
of the clausal complement: in both cases, the path to the within-clause argument
is comp obj. Instead, this requirement is stated as an additional condition on the
f-structures in the extraction domain which applies in the case of those speakers
who identify a difference in grammaticality that is related to the bridge/nonbridge
verb distinction.
Where bridge and nonbridge verbs are judged to be distinct in this respect, we
use the f-structure attribute ldd (for “long-distance dependency”) to distinguish
between the two. The value − for the feature ldd is lexically specified by a
nonbridge verb like whisper as appearing in its comp. Such f-structures cannot
participate in a long-distance dependency. This captures the unacceptability of (6)
for the relevant set of speakers, since the comp f-structure contains the attribute-
value pair $ldd, −%:

(6) ⃰Chris, we whispered that David saw.

F-structure violating Bridge Verb Constraint:

−
656 long-distance dependencies

In (6), the path relating the f-structure for Chris to its within-clause function goes
through the comp f-structure of the verb whisper, which has the value − for the
attribute ldd; this is disallowed. We demonstrate later in this section how, for the
relevant dialects, this constraint can be captured in formal terms.
A number of other constraints on long-distance dependencies were originally
explored by Ross (1967) and have since been the subject of intense scrutiny and
debate. Ross identified a number of constraints involving “islands”: areas of a
sentence from which extraction of an element to form a long-distance depen-
dency is prohibited. Among these constraints is the Sentential Subject Constraint,
according to which a long-distance dependency cannot involve a position inside a
sentential subject—that is, subjects are islands for extraction.

(7) ⃰Chris, [that David saw ] surprised me.

(cf. That David saw Chris surprised me.)
F-structure violating Sentential Subject Constraint:
’

This constraint is simply stated: the path to the within-clause function of dis
may not include subj. Other constraints on long-distance dependencies can be
characterized similarly, either as constraints on grammatical functions permitted
on the path or as constraints on attributes and values in f-structures through which
the path passes.
There is little consensus on the proper characterization of long-distance depen-
dencies involving modifying adjuncts. The situation appears to be similar to the
one we have reported for bridge and nonbridge verbs when it comes to variability
in speakers’ judgments. For example, Williams (1992: 18) claims that examples
like (8), which involve a relation between an initial question phrase and a position
inside an adverbial modifier, are “marginal though possible”:

(8) ?Who did John go to New York [after talking to]?

However, Cinque (1990) and Hornstein and Weinberg (1995) count as ungram-
matical examples that are very similar to (8):

(9) a. ⃰To whom did you leave [without speaking]?

b. ⃰What did John drink cognac [after singing]?
 syntax of long-distance dependencies 657

Constraints on long-distance dependencies involving modifying adjuncts are dif-

ficult to characterize and judgments, as we have stated, vary. However, we believe
that some basic conclusions can be drawn. First, some dependencies involving
modifying adjuncts are generally acceptable:

(10) This room, Julius teaches his class in.

Other examples involving a dependency between a member of the dis set and a
position inside an adj are not acceptable. For example, the unacceptability of (11)
shows that extraction from a tensed sentential modifier is not permitted:

(11) ⃰Chris, we think that David laughed after we selected.

(cf. We think that David laughed after we selected Chris.)
658 long-distance dependencies

F-structure violating Tensed Adjunct Constraint:

’

We propose the following general characterization of the possible within-clause

grammatical roles of the displaced phrase in the English topicalization construc-
tion, which takes into account relevant island constraints:

(12) In the English topicalization construction, the displaced phrase can be

related to a grammatical function that is embedded inside any number of
xcomp, comp (for some speakers, the comp function must be governed
by a bridge verb), or tensed sentential obj functions, or to a grammatical
function inside a possibly embedded nontensed adj function.

Other languages place different constraints on the topicalization path. For example,
Kroeger (1993: Chapter 7) shows that in Tagalog, the topicalization path must con-
sist exclusively of subjects: only a subj may be extracted, and the only function out
of which extraction is permitted is subj. This is true not only for the topicalization
construction but also for other long-distance dependencies in Tagalog; the path
in the constituent question construction and the relative clause construction must
also contain only subj attributes.1

1 A formal definition of the topicalization path in Tagalog is provided in (21).

 syntax of long-distance dependencies 659

Constituent structure and functional constraints: In §4.2.3.2, we

discussed the relation between constituent structure specifier positions and dis-
course functions: in many languages, specifier positions must be occupied by
phrases bearing the discourse functions topic or focus. King (1995) analyzes the
configurational encoding of topic and focus in Russian, showing that the Russian
topic appears in the specifier position of IP, and Kroeger (1993) shows that the
same is true in Tagalog.
Interestingly, however, in English a fronted (dis) phrase which can bear the
discourse function topic does not appear in a specifier position. The specifier
position of IP in English is reserved for the subject, not a displaced phrase
(§4.2.3.1); evidence that the displaced phrase does not appear in the specifier of
CP comes from examples involving topicalization in a subordinate clause:

(13) Matty thinks that [Chris, we like].

If the displaced phrase Chris appeared in the specifier of CP, it would be expected
to appear before the complementizer that, not after it. Bresnan et al. (2016: 9) show
that topicalized phrases are adjoined to IP in English:

(14) Chris, we like.

’

IP

NP IP

N NP I

Chris N VP

we V

like

To analyze examples like (14), we propose this phrase structure rule:

 
(15) IP −→ TpczP IP
 
 ↓∈ (↑ dis)  ↑ =↓
(↑ TpczPath) = ↓

Here we use the metacategory abbreviation TpczP for the phrase structure category
of the fronted phrase, where Tpcz stands for topicalization. We also use the
functional abbreviation TpczPath for the path through the f-structure to the
within-clause function of the fronted phrase.
660 long-distance dependencies

The set of phrasal categories that can participate in the topicalization construc-
tion in English is given in (3) of this chapter. On this basis, we define TpczP for
English in rule (15) as the following disjunction of categories:

(16) TpczP [English] ≡ {NP | PP | VP | AP | CP}

We must also properly constrain the path through the f-structure that defines
the within-clause grammatical function of the displaced phrase. Formally, this
relation involves functional uncertainty, discussed in §6.1.2. In (12) above, a set of
constraints on the long-distance path for topicalization in English were outlined.
We can formally characterize these constraints as follows.
We start with the Sentential Subject constraint, exemplified in (7) and defined
as SSC:

(17) Sentential Subject Constraint

SSC ≡ [gf − subj]⃰ gf

The initial portion of the path defined in (17) may not include subj,2 though it
may end in subj or any other grammatical function gf. The possibility for deep
embedding is represented by the Kleene star operator ⃰ permitting any number of
attributes defined as members of the metacategory gf other than subj on the path.
The Bridge Verb Constraint, which seems to apply for a subset of speakers and
was exemplified in (6), is captured via the expression defined as BVC in (18). The
definition relies on an off-path constraint (§6.6) referring to a feature of any f-
structures on the path, (→ ldd) += −:

(18) Bridge Verb Constraint

BVC ≡ gf⃰ gf
(→ ldd) += −

The off-path constraint in (18) permits any path relating the displaced phrase
to its within-clause grammatical function as long as it does not pass through an
f-structure bearing the attribute ldd with value −.
Finally, according to the Tensed Adjunct Constraint exemplified in (11), the path
defining the within-clause grammatical function of a displaced phrase may include
a nontensed adj function, but not a tensed adj. The following definition of the path
TAC accomplishes this:3

2 The relative-difference operator for regular expressions, written as a minus sign (−), is defined and
discussed in §6.1. The expression gf − subj represents any grammatical function (gf) except subj, and
[gf − subj]⃰ represents any sequence of grammatical functions that does not contain subj.
3 In §6.3.1, we discussed the use of expressions in which the set membership symbol ∈ is used as an
attribute to allow reference to some member of a set.
 syntax of long-distance dependencies 661

(19) Tensed Adjunct Constraint

TAC ≡ ¬[gf⃰ adj ∈ gf⃰]
(→ tense)

The off-path constraint (→ tense) requires a member of the adj set to contain the
attribute tense. Thus, the path within the square brackets ([gf⃰ adj ∈ gf⃰])
(→ tense)
picks out all paths which pass through a tensed adjunct. The complement operator
for regular expressions, written as ¬, picks out all paths which do not obey these
constraints: in other words, paths which do not go through a tensed adjunct. (See
§6.1 for discussion of regular expressions and complementation.)
Together, the constraints in (17), (18), and (19) express restrictions which apply
to the path through the f-structure that defines the within-clause grammatical
function of the displaced phrase. Putting these three constraints together via
intersection (§6.1), we are able to define TpczPath in English thus:

(20) TpczPath [English] ≡ SSC & BVC & TAC

TpczPath is therefore restricted by a set of off-path constraints in addition to

constraints on the grammatical functions that can appear on the path between
filler and gap. The within-clause grammatical function of the displaced phrase
is permitted to be arbitrarily deeply embedded inside any number of properly
constrained xcomp, comp, or obj functions, and optionally to appear within a
nontensed member of an adj set.
As noted earlier, different languages impose different constraints on the paths
involved in long-distance dependencies: Kroeger (1993: Chapter 7) shows that only
subjects can participate in long-distance dependencies (LDDs) in Tagalog.

(21) TpczPath [Tagalog] ≡ subj+

This expression uses the Kleene plus operator + to indicate that the relevant path in
Tagalog consists of at least one occurrence of subj: only a subj may be involved in
establishing a long-distance dependency, and only a subj may contain a displaced
phrase. LFG research has not yet established a complete typology of possible paths
in long-distance dependencies: future research will no doubt reveal more about
the possible range of crosslinguistic and cross-constructional variation in long-
distance paths, as well as universal generalizations concerning constraints on long-
distance paths.
662 long-distance dependencies

Topicalization in coordinate structures: It has long been noted that

coordinate structures obey the across-the-board constraint (Ross 1967; Williams
1990), according to which a topicalized phrase bears a grammatical function
inside one conjunct of a coordinate structure only if it also bears a grammatical
function inside the other conjunct(s). Example (22) is ungrammatical because the
topicalized phrase David bears the obj function only in the first conjunct of the
coordinate sentence:

(22) ⃰David, [Chris hates and Matty likes Ursula].

The across-the-board constraint requires the topicalized phrase to bear a gram-

matical function “across the board,” that is, within each conjunct in a coordinate
construction:

(23) David, [Chris hates and Matty likes ].

In (23), the topicalized phrase is the obj of both hates and likes. The across-the-
board constraint falls out as a consequence of our theory of coordination and
long-distance dependencies. Since grammatical functions are distributive features
(§6.3.2), asserting that an f-structure is the obj of a set means that it is the
obj of each member of the set. That is, resolving the path to the within-clause
grammatical function of the topic correctly entails that the topicalized phrase must
bear a grammatical function within each conjunct of the coordinate phrase.
Kaplan and Maxwell (1988b) discuss examples like (24), in which the topicalized
phrase David is the obj of the first conjunct and the objtheme of the second
conjunct:
 syntax of long-distance dependencies 663

(24) This book, [David bought and Chris gave Matty ].

These examples show that the fronted phrase can bear different grammatical
functions within each conjunct.4 To analyze these examples, Kaplan and Maxwell
(1988b) provide the definition of functional uncertainty given in (27) of §6.1.2,
repeated here:

(25) Functional uncertainty:

If η is a regular expression, then (f η) = v holds if and only if

(f a suff(a, η)) = v

for some symbol a, where suff(a, η) is the set of suffix strings s such that
as ∈ η.

4 Kaplan and Zaenen (1989b) discuss constraints on the long-distance path in coordinate structures
in Japanese, based on work by Saiki (1985): the grammatical function of the fronted constituent must
either be subj in all conjuncts, or a nonsubject function in all conjuncts. Similar constraints hold in
English.
(a) ⃰Who did Chris think [David met ] and [ saw Matty]?
Kaplan and Zaenen show how the relevant path can be defined so as to predict these facts. Falk (2000)
also discusses these patterns and provides an alternative analysis.
664 long-distance dependencies

The effect of this definition is to allow a regular expression representing a path

through the f-structure to be resolved differently in each conjunct of a coor-
dinate structure. When a long-distance dependency involves elements inside a
set representing a coordinate structure, the definition in (25) allows the as yet
unexplored suffix of the regular expression representing the path to be expanded
independently within each element of the set, as is required in the analysis of (24).

17.1.1.2 Left and right dislocation Another type of displacement is dislocation.

Dislocation constructions differ from topicalization constructions in having a pro-
noun rather than a gap within the clause. This means that without the dislocated
constituent, the clause would still be grammatical—it is syntactically and semanti-
cally complete in its own right. Dislocation can be to the right or left of the clause,
though the information structure roles associated with these syntactic positions
are distinct. The following English examples illustrate left dislocation used to
introduce a topic and right dislocation with a clarifying function, respectively:

(26) Q: What’s the matter?

A: [This little girl]i , shei ’s lost heri parents. She doesn’t know where they are.

(27) Theyi filed a complaint last Friday, [Karen and Dave]i .

Whether leftwards or rightwards, dislocation involves an anaphoric dependency

between the dislocated constituent, which is a member of the dis set at f-structure,
and a within-clause grammatical function.
In at least some dialects of British English, pronouns can be right dislocated, as
illustrated in (28).5 These right-dislocated pronouns (ProTags; Mycock 2017) bear
default accusative case (on which, see Hristov 2013b) and therefore case mismatch
regularly occurs. Thus an anaphoric control analysis is justified:

(28) a. Ii love sushi, mei .

b. I can never remember hisi name, himi .
c. Theyi ’re good, themi .

Of course, at f-structure the direction of the dislocation is not specified. (29)

provides an analysis of a sentence involving both left and right dislocation in a
dialect of English that permits personal ProTags; the f-structures of both dislocated
phrases appear as members of the dis set:

5 Unlike right-dislocated full noun phrases, ProTags do not have a clarifying function. See Mycock
(2017) for details.
 syntax of long-distance dependencies 665

(29) Maryi , Ij saw heri yesterday, mej .

 
’
  
   
 
 
 
 
  
 
  
 
 
 

   
   
    
 
 
 
  
 
 
   
 
 

 
 
   
 
 
   
   
   
   
   
 
 
 
   
 
 
   
   
     
   
   
     
   
 
 
 
 

How are different types of dislocation distinguished then, if not at f-structure?

As mentioned previously, the information structure roles of dislocations differ
depending on which clausal periphery—left or right—they occupy. This is an issue
which has not received much attention in LFG work (though see Berman 2003 and
Szűcs 2014), but it is clear that relevant syntactic positions at the level of c-structure
can be annotated with the necessary information to capture the differences in
question (similar to the way in which discourse configurationality is analyzed in
LFG; see §10.4.4.3). This possibility remains to be fully explored.

17.1.2 Relative clauses

Relative clauses in English and many other languages also involve long-distance
dependencies. Unlike the situation with topicalization, two long-distance depen-
dencies are involved in a relative clause construction.
The first dependency holds between the fronted phrase and the within-clause
grammatical function it fills. In early LFG work, Bresnan and Mchombo (1987)
propose that the fronted phrase in a relative clause bears the syntacticized topic
function. By the Extended Coherence Condition, given in (1), the topic must be
linked to a grammatical function within the clause. As discussed in §2.1.11 (see
666 long-distance dependencies

also §4.2.3), we do not adopt this approach, but use instead the overlay function
dis in the analysis of long-distance dependencies and instances of dislocation. As
an overlay function, members of the dis set must be integrated into the structure
of a clause via an association with a primary, nonoverlay function, consistent with
the basic principle underlying the Extended Coherence Condition.
The second dependency involves the relative pronoun and its position, possibly
embedded, within the fronted phrase. We follow Kaplan and Bresnan (1982) in
representing this syntactic dependency at f-structure; the f-structure of the relative
pronoun appears as the value of the attribute relpro within the relative clause.
Similar representations have also been adopted by Butt et al. (1999) and Falk
(2001b).
The c-structure and f-structure for the phrase a man who Chris saw are:

(30) a man who Chris saw

’
NP

D N

a N CP

N NP C

man N IP

who NP I

N VP

Chris V
saw

In (30), the relative pronoun appears in initial position in the relative clause, and
its f-structure is both a member of the set that is the value of dis and the relpro
of the relative clause.
Example (31) shows that the relative pronoun can also appear as a subcon-
stituent of the initial phrase. Here the relative pronoun whose is a subconstituent
of the fronted phrase whose book:
 syntax of long-distance dependencies 667

(31) a man whose book Chris read

NP

D N

a N CP

N NP C

man D N IP

whose N NP I

book N VP

Chris V

read

In (31), the value of the dis attribute is a set with one member, the f-structure of the
fronted phrase whose book, and the value of the relpro attribute is the f-structure
of the relative pronoun whose. We examine syntactic constraints on both of these
dependencies below.
We propose the following phrase structure rules for the analysis of these
examples:

(32) N, −→ N, CP∗
↑ =↓ ↓ ∈ (↑ adj)





(33) CP −→ RelP C,






↓ ∈ (↑ dis)  ↑=↓



(↑ RTopicPath) = ↓ 



(↑ relpro) = (↓ RelPath) 
(↑ relpro prontype) =c rel
668 long-distance dependencies

The first two annotations on the RelP daughter in (33) are similar to the annota-
tions on the rule for English topicalization constructions in (15). The constraint
↓ ∈ (↑ dis) requires the f-structure corresponding to the RelP node to appear as
a member of the dis set in the f-structure. The constraint (↑ RTopicPath) = ↓
ensures that the f-structure of the initial phrase also fills a grammatical function
within the clause, constrained by the long-distance path RTopicPath; we define
RTopicPath below.
The third and fourth annotations constrain the value of the relpro attribute
in the relative clause f-structure. The constraint in the third line, (↑ relpro) =
(↓ RelPath), requires the relpro f-structure to appear at the end of the path
RelPath within the f-structure of the RelP. Below, we provide a definition of
RelPath that properly constrains the role of the relative pronoun within the
fronted phrase. Finally, the constraint (↑ relpro prontype) =c rel is a con-
straining equation (§5.2.8) requiring the value of the relpro attribute to have a
prontype attribute with value rel: the value of the relpro attribute must be a
relative pronoun.

17.1.2.1 Category of the displaced phrase The constituent structure metacat-

egory RelP in (33) represents the phrase structure categories that can appear in
initial position in a CP relative clause. The phrases in (34) exemplify the possible
instantiations of RelP in English:6

(34) a. NP: a man who I selected

b. PP: a man to whom I gave a book
c. AP: the kind of person proud of whom I could never be
d. AdvP: the city where I live

Therefore, we define RelP for English in the rule in (33) as the following disjunction
of categories:

(35) RelP [English] ≡ {NP | PP | AP | AdvP}

17.1.2.2 Grammatical function of the displaced phrase Next, we define

RTopicPath, the path relating the fronted constituent in a relative clause to its
within-clause grammatical function. Constraints on RTopicPath are very similar
to the constraints on TpczPath discussed in §17.1.1.1, as the following examples
show:

(36) a. Chris, we like.

b. a man who we like

(37) a. Chris, we think that David saw.

b. a man who you think that David saw

6 Example (34c) is due to Webelhuth (1992).

 syntax of long-distance dependencies 669

(38) Bridge Verb Constraint violation (ungrammatical for some but not all
speakers)
a. ⃰Chris, we whispered that David saw.
b. ⃰a man who you whispered that David saw

(39) Sentential Subject Constraint violation

a. ⃰Chris, [that David saw ] surprised me.
b. ⃰a man who [that David saw ] surprised me

(40) Tensed Adjunct Constraint violation

a. ⃰Chris, we think that David laughed when we selected.
b. ⃰a man who we think that David laughed when we selected

We therefore propose that the same constraints restrict both the English
RTopicPath and the long-distance path in topicalization constructions. The
expressions in (20) and (41) are exactly the same:

(41) RTopicPath [English] ≡ SSC & BVC & TAC

Examination of other languages reveals different constraints on RTopicPath. As

noted earlier, Kroeger (1993: Chapter 7) shows that in Tagalog, RTopicPath is
subj+ , that is, paths consisting only of subj. Saiki (1985) discusses the definition
of RTopicPath in Japanese, exploring constraints on RTopicPath in the causative
and passive constructions.

17.1.2.3 Grammatical function of the relative pronoun The definition of

RelPath must appropriately constrain the grammatical function of the relative
pronoun within the fronted dis f-structure. As originally noted by Ross (1967)
and explored in detail by Bresnan (1976), Webelhuth (1992), Falk (2001b), and
many others, the relative pronoun may be embedded inside the fronted phrase.
Ross (1967) provides this example of a deeply embedded relative pronoun:

(42) [Reports [[the height of the lettering on the cover of which] the government
prescribes ]] should be abolished.

Ross (1967) originally used the term pied piping in the transformational analysis
of these constructions: in moving to the front of the sentence, the relative pronoun
lures some additional material along with it, like the Pied Piper of Hamelin lured
rats and children along with him as he left Hamelin.
670 long-distance dependencies

Research on pied piping has revealed a range of constraints on the long-distance

path RelPath to the relative pronoun in the fronted phrase:

(43) a. the man [who] I met

b. the man [whose book] I read
c. the man [whose brother’s book] I read
d. the report [the cover of which] I designed
e. the man [faster than whom] I can run
f. the kind of person [proud of whom] I could never be
g. the report [the height of the lettering on the cover of which] the government
prescribes
h. ⃰the man [a friend of whose brother] I met
i. the room [in which] I teach
j. ⃰the man [the woman next to whom] I met

In all of these examples, the phrase structure category of the fronted phrase is one
of the categories defined by RelP, and no constraints on RTopicPath are violated.
Example (43a) shows that the relative pronoun can itself appear in fronted position;
in such a case, RelPath is the empty path. Examples (43b–c) indicate that the
relative pronoun can appear as a possessor phrase, filling the poss role in the dis
f-structure, or as the possessor of a possessor. It can also appear as the object of an
oblique argument, as in (43d–f), or embedded inside an oblique argument, as in
(43g), though it may not fill the poss role inside an oblique phrase (43h).7 It can
appear as the object of a fronted adjunct phrase (43i), though it may not appear as
an adjunct inside the fronted phrase (43j).
Given these facts, we propose the following definition of RelPath in English:

(44) RelPath [English] ≡ {poss⃰ | [(obl𝜃 ) obj]⃰}

In other languages, the definition of RelPath differs. Webelhuth (1992) provides

a thorough discussion of pied piping in Germanic, showing that constraints on
pied piping in English relative clauses are different from the constraints that hold
in German, Dutch, Swedish, and other Germanic languages.

17.1.2.4 Correlatives Another type of relative clause construction which has

received attention in an LFG setting is the correlative construction, in which a
relative nominal or pronominal expression appears in a subordinate clause and

7 Louisa Sadler (p.c.) notes that some speakers judge constructions such as (43h) to be acceptable.
 syntax of long-distance dependencies 671

is resumed by a phrase in the matrix clause. Butt et al. (2007) provide an analysis
of Urdu correlatives, illustrated in (45a), while Belyaev and Haug (2014) consider
correlatives in Ossetic (a language in which direct objects bear either nom or gen
case marking), illustrated in (45b):

(45) a. [jo khar.-ii hai] [vah lar.kii lambii

which stand-prf.f.sg be.prs.3sg that girl.f.sg.nom tall.f.sg
hai]
be.prs.3sg
‘The girl who is standing is tall.’ [‘Who is standing, that girl is tall.’]
b. [didinʓ̌-ətɜ sə čəžg-ɜn ba-lɜvar kotː-aj], [wəz
flower-pl what girl-dat pfv-present do-pst.2sg that[gen]
fetː-on]
see.pfv-pst.1sg
‘I saw the girl who you gave flowers to.’

The syntactic analysis of correlatives is fundamentally the same for these exam-
ples. The correlative clause appears initially8 and bears the dis function, while
simultaneously having the appropriate within-clause grammatical function. (At
i-structure, the correlative clause bears the discourse function topic.) Belyaev and
Haug (2014) provide the following f-structure analysis of the Ossetic correlative
construction in (45b):9

(46) didinʓ̌-ətɜ sə čəžg-ɜn ba-lɜvar kotː-aj, wəz

flower-pl what girl-dat pfv-present do-pst.2sg that[gen]
fetː-on
see.pfv-pst.1sg
‘I saw the girl who you gave flowers to.’

8 Butt et al. (2007) note that in Urdu, the subordinate clause containing the relative expression
can appear either before or after the matrix clause; following Srivastav (1991), they classify as true
correlatives only structures in which the subordinate clause appears initially.
9 Belyaev and Haug (2014) use the attribute stype with value correl to mark correlative clauses.
672 long-distance dependencies

’
’

17.1.3 Constituent questions

In §4.2.3, we noted that the question phrase in an English constituent question
appears in initial position in the sentence, in the specifier position of CP:

(47) Who does David like?

CP

NP C

N C IP

Who does NP I ’
N VP

David V

like
 syntax of long-distance dependencies 673

To analyze constructions like (47), the following simplified rule was proposed in
§6.1.2:
 
(48) CP −→ XP C,
 ↓ ∈ (↑ dis) ↑=↓
 
∗
(↑ comp gf) = ↓

This rule ensures that the phrase in the specifier position of CP bears the dis
function and also fills a grammatical function within the utterance. We now refine
this rule to take into account constraints on the phrase structure category of
the fronted phrase and to give a more complete characterization of the path to
its within-clause function. We also introduce the q attribute, whose value is the
f-structure of the possibly embedded interrogative pronoun within the initial dis
phrase, analogous to the relpro attribute for relative clauses: for example, in the
analysis of the question Whose book did you read?, the f-structure for the fronted
phrase Whose book is a member of the dis set, and the question word whose
provides the value for the q attribute. See Kaplan and Bresnan (1982) for more
discussion of the q attribute.
We use the constituent structure metacategory QuesP and the functional abbre-
viations QDisPath and WhPath in the following reformulation of rule (48):
 
(49) CP −→ QuesP C,
 ↓ ∈ (↑ dis)  ↑=↓
 
 
 (↑ QDisPath) = ↓ 
 
 (↑ q) = (↓ WhPath) 
(↑ q prontype) =c wh

The annotations on the QuesP node in rule (49) are similar to those on the relative
clause rule in (33). The first two annotations require the f-structure corresponding
to the QuesP node to fill the dis role and also to bear some grammatical function
defined by the long-distance path QDisPath. The third annotation requires the
value of the q attribute to appear at the end of the long-distance path WhPath
within the dis f-structure; we discuss constraints on WhPath below. The fourth
annotation requires the prontype attribute of the q f-structure to bear the value
wh, ensuring that an interrogative pronoun plays the q role.

17.1.3.1 Category of the displaced phrase The first issue is the correct defi-
nition of QuesP: which phrasal categories can appear as a dis constituent in the
specifier of CP? These examples are wellformed:

(50) a. NP: Who do you like?

b. PP: To whom did you give a book?
c. AdvP: When did you yawn?
d. AP: How tall is Chris?
674 long-distance dependencies

Thus, we define QuesP in (49) as the following disjunction of categories:

(51) QuesP [English] ≡ {NP | PP | AdvP | AP}

17.1.3.2 Grammatical function of the displaced phrase Next, we define

QDisPath, the long-distance path involved in question formation. Constraints
on QDisPath appear to be the same as those defined for TpczPath in (20):

(52) a. Chris, we like.

b. Who do you like?

(53) a. Chris, we think that David saw.

b. Who do you think that David saw?

(54) a. ⃰Chris, we whispered that David saw.

b. ⃰Who did you whisper that David saw?

(55) a. ⃰Chris, [that David saw ] surprised me.

b. ⃰Who did [that David saw ] surprise you?

(56) a. This hammer, we smashed the vase with.

b. What did you smash the vase with?

(57) a. ⃰Chris, we think that David laughed when we selected.

b. ⃰Who did you think that David laughed when we selected?

Therefore, we can assume the same definition for QDisPath in English as we did
for TpczPath in (20):10

(58) QDisPath [English] ≡ SSC & BVC & TAC

17.1.3.3 Grammatical function of the interrogative element In (44), we pro-

vided a constraint on RelPath, the path to the relative pronoun within the initial
dis phrase in a relative clause. Similarly, we must define WhPath, the path to the
interrogative element within the initial dis phrase in a constituent question. Like
the relative pronoun, an interrogative pronoun may be embedded inside the initial
phrase, appearing as a possessor (59a) or the possessor of a possessor (59b), or as
the obj of the fronted argument (59c):

10 The path definitions assumed here are sufficient for present purposes, but Postal (1998) discusses
certain differences between these two types of path which may require further complications and
refinements.
 syntax of long-distance dependencies 675

(59) a. [Whose book] did you read?

b. [Whose brother’s book] did you read?
c. [In which room] do you teach?

However, WhPath is more constrained than RelPath:

(60) a. ⃰[The cover of which report] did you design?

(cf. Which report did you design the cover of?)
b. the report [the cover of which] I designed

(61) a. ⃰[The height of the lettering on the cover of which report] does the govern-
ment prescribe?
b. the report [the height of the lettering on the cover of which] the government
prescribes

(62) a. ⃰[Faster than whom] can you run?

b. the man [faster than whom] I can run

(63) a. ⃰[Proud of whom] are you?

b. the kind of person [proud of whom] I could never be

The following definition of WhPath for English captures these judgments:

(64) English WhPath:

{poss⃰ | obj}

Webelhuth (1992) provides more discussion of constraints on pied piping in

Germanic languages, showing that pied piping constraints in English constituent
questions are the same as in other Germanic languages.

17.1.3.4 The Complementizer-Adjacent Extraction constraint As discussed in

§6.11.2, question formation is restricted by the Complementizer-Adjacent Extrac-
tion (CAE) constraint (sometimes referred to as the that-trace or comp-trace
effect). This constraint blocks subject extraction in long-distance dependencies
when a complementizer is present; notice the difference in grammaticality that
arises in the case of extraction of a subj (65a) versus a non-subject (65b):

(65) a. Who do you think (⃰that) saw Kim?

b. Who do you think (that) Kim saw ?
676 long-distance dependencies

The presence of an adverb has an ameliorating effect, permitting subj extraction

even in the presence of the complementizer (Bresnan 1977):

(66) Who did you say that, just a minute ago, sneezed?

One approach to the CAE constraint is proposed by Falk (2002b, 2006), framed
within his pivot/gf 4 approach to subjecthood (see §2.1.5). Recall that Falk (2006)
defines pivot as the grammatical function that connects its own clause to other
clauses. In (65a), for instance, the pivot is who. To account for the CAE constraint,
Falk proposes the following generalization:

(67) CAE generalization according to Falk (2006):

If 𝜙−1 (↑ pivot) exists,11 one of the nodes in 𝜙−1 (↑) must immediately
dominate one of the nodes in 𝜙−1 (↑ pivot).

Falk’s intention is for the immediate dominance relation referred to in (67) to be

“f-structure aware,” and to this end he defines the “f-ID relation:”

(68) Definition of the f-ID relation according to Falk (2006):

For any f-structures f1 and f2 , f1 f-IDs f2 (f1 →f f2 ) if and only if there exists
a node n1 in 𝜙−1 (f1 ) and a node n2 in 𝜙−1 (f2 ) such that n1 immediately
dominates n2 .

Falk’s final formalization of the CAE constraint is as follows:

(69) CAE constraint according to Falk (2006: 133), associated with the comple-
mentizer that:
𝜙−1 (↑ pivot) ⇒ ↑ →f (↑ pivot)

As a constraint included in the lexical entry of complementizer that in English, this

allows (65b) but rules out (65a). The guiding intuition behind the constraint is that
a clause containing the complementizer that must contain its own pivot (which
will map to the gf
4 argument in a nominative-accusative language like English; see
§2.1.5) and this pivot must not be displaced; the pivot of a that-clause must be
immediately dominated by some node (here, the IP node) that maps to the same
f-structure as the complementizer.
One issue with Falk’s analysis is that it does not account for the Adverb Effect—
the ameliorating effect of an adverbial expression, illustrated in (66)—because an

11 𝜙−1 is the relation that maps f-structures to c-structure nodes: see §6.10.3. 𝜙−1 (↑ pivot) is an
existential constraint that holds if there is some c-structure node corresponding to (↑ pivot) via the
𝜙−1 relation.
 syntax of long-distance dependencies 677

intervening adverb does not change the f-structure relations which hold between
the pivot and the complementizer.
In §6.11.2, we presented Asudeh’s (2009) analysis of the CAE constraint: briefly,
Asudeh proposes a constraint, to be included in the lexical entry of the comple-
mentizer that, which rules out the ungrammatical sequences of words on the basis
of string precedence. Asudeh (2009) cites the Adverb Effect as a reason why the
CAE constraint must be treated as a precedence-based phenomenon rather than
a dominance-based one, as in Falk’s analysis. Asudeh’s (2009) own analysis of the
CAE constraint makes crucial reference to linear order relations and does account
for the Adverb Effect. His formulation of the CAE constraint is given in (179) of
§6.11.2, repeated here:

(70) CAE Constraint, Asudeh (2009), formal version:

¬[realized(. subj) ∧ (dis ∈ (. subj))]

Asudeh’s CAE constraint rules out a situation in which the word following the
complementizer has a subject which is both phonologically realized and the head
of an unbounded dependency. When an adverb follows the complementizer that,
as in (66), the constraint in (70) is satisfied.

17.1.3.5 Multiple questions Thus far, we have considered English constituent

questions that include a single initial question phrase. For languages allowing mul-
tiple question phrases occurring together in a single sentence there is typological
variation: one question phrase may appear initially (so-called simple fronting), as
in English (71a); all question phrases may be displaced (multiple fronting), as in
Russian (King 1995; 71b); or all question phrases may occupy the same position as
they would in a declarative sentence (in situ), as in Japanese (71c).

(71) a. What did Charlie put where?

b. Kogda kto udaril Borisa?
when who hit Boris
‘Who hit Boris when?’
c. Mari-ga doko-de ojoosan-ni nani-o eranda ka
Mari-nom where-loc daughter-dat what-acc choose.pst q
‘What did Mari choose for her daughter where?’

Mycock (2006) provides an LFG approach to the typology of questions. Key to

her analysis is the proposal that question phrases must be focused. Thus, a more
complete analysis of a simple constituent question in English such as (47) (Who
does David like?) would include the i-structure in (72), which captures the focus
status of the question phrase. In English, this focus status is marked syntactically:
678 long-distance dependencies

the question word bearing the grammatical function obj is displaced to initial
position, and hence also bears the function dis.12
 
(72) { [David] }
 { [who] } 
 
{ [like] }

In some languages, such as Russian, the focus status of all question phrases is
signaled syntactically; that is, they occupy a syntactic position associated with the
discourse function focus and hence all appear initially, as in (71b). Of course,
focus status need not be marked by syntactic means only. Mycock (2006) argues
that the focus status of question phrases in Japanese, which appear in situ (71c), is
marked prosodically. Constituent questions in this language do not involve a long-
distance dependency and question phrases do not bear the function dis, but at
i-structure—like their Russian counterparts—all question words are members of
the focus set, according to Mycock (2006). What of the English-type formation,
which involves a single question phrase bearing the function dis appearing in ini-
tial position and other question phrases in situ, as in (71a)? Mycock claims that in-
situ question words in English are prosodically marked for focus status, like those
in Japanese. Therefore, while the c-structures for English, Russian, and Japanese
multiple constituent questions differ according to whether or not question phrases
bear the function dis, in each case the i-structures are fundamentally the same:
they contain a focus set which includes all of the question words in the sentence.
Whether all in-situ languages should be analyzed in the same way is an interest-
ing question. In his LFG analysis of Chinese constituent questions, Huang (1993)
shows that although these constructions do not involve long-distance dependen-
cies, the same sorts of scope phenomena found in English constituent questions
can also be found in Chinese. Huang uses inside-out functional uncertainty
(§6.1.3) in his analysis to capture the similarities between questions in English
and Chinese. Mycock (2006) considers an apparent island effect in Japanese, but
it emerges that this may be an issue of prosody rather than syntax (see Deguchi
and Kitagawa 2002). It remains to be determined beyond doubt whether in-situ
languages may be subject to exactly the same kinds of syntactic constraints as
languages which involve displacement of question phrases in multiple constituent
questions.
We assume that only question words that indisputably participate in a long-
distance dependency appear in the dis set at f-structure and provide the value of
the q attribute. This distinguishes displaced question words from those that appear
in situ under our approach. This difference is illustrated by the following analysis
of the English multiple constituent question in (71a), which includes one displaced
question word (what) and one in-situ question word (where):

12 On the representation of information structure in LFG, see Chapter 10.

 syntax of long-distance dependencies 679

(73) What did Charlie put where?

f-structure:

CP

NP C ’
N C IP

What did NP I

N VP

Charlie V

V AdvP
i-structure:
put Adv
{ [Charlie] }
where [what]
[where]
[put]

The kind of “regular” constituent questions that we have considered so far are the
most well studied. LFG research has also investigated a variety of other construc-
tions that include question phrases. Mycock (2004, 2006) offers an LFG analysis
of the q-expletive (also referred to as the wh-expletive or wh-scope marking)
construction, in which a question phrase takes scope over a clause higher than the
one in which it appears. The question word’s extended scope is marked in some
languages, including Hungarian, by the presence of another element, represented
in the gloss as what:

(74) István mi-t gondol, [hogy Mari mi-t

Stephen.nom what-acc think.prs.3sg that Mary.nom what-acc
mond-ott, [hogy János ki-t hív-ott fel]]?
say-pst.3sg that John.nom who-acc call-pst.3sg vm
‘Who does Stephen think that Mary said that John called?’

Mycock adopts the position that only the final question word (in example (74), kit)
is the “true” question word; any higher ones are expletives associated with a “true”
question expression. The q-expletive is semantically empty, but it is syntactically
and/or prosodically prominent. For instance, in (74) the q-expletive (mit) occupies
immediately preverbal Focus position, just as the “true” question word kit does
in the most deeply embedded clause. The q-expletive’s presence thus serves to
focus “by proxy” the content of the phrase with which it is associated (which
680 long-distance dependencies

Mycock claims is the entire clause within which the “true” question words appear
in Hungarian). In this way, the q-expletive stands in for the relevant expression
in a higher clause. See Mycock (2006) for full details of an LFG analysis of scope-
marking constructions, including the q-expletive construction.
Other work on questions in LFG has considered the information structure status
of question words and phrases and how this interacts with syntactic structure.
We have seen that in “regular” information-seeking constituent questions there
is evidence that the question expression has focus status at i-structure, but there
are other possibilities. On the basis of question formation data from 15 languages,
Mycock (2013) proposes that question words can bear other discourse functions
as well (background, completive, topic), just like non-question words. For
instance, following Butt (2012), Mycock proposes to analyze question words in
echo questions as background rather than focus. This provides an explanation
for syntactic differences between these two types of constituent questions. Butt
(2014b) also deals with non-canonical constituent questions and the role that
information structure plays in the analysis of Hindi-Urdu. Butt draws on the work
of Mycock (2006) on questions and Krifka (2008) on information structure and the
concept of Common Ground Management to give an analysis of the immediately
postverbal position in Hindi-Urdu as a second focus position with particular
interpretational characteristics. Butt (2014b) also presents an analysis of polar
questions including the question marker kya in Hindi-Urdu, which provides a
similarly fine-grained approach to their distinct information structural properties.

17.1.4 The ‘tough’ construction

In certain constructions including predicates such as tough (“tough-movement”
constructions), the complement of the tough predicate is a nonfinite subordinate
clause with a missing object, and the subject of the tough predicate is interpreted
as controlling the missing object:

(75) This race is tough [to finish ].

(cf. It is tough [to finish this race].)

As Dalrymple and King (2000) show, the missing object can be embedded arbi-
trarily deeply in the complement of a tough construction.

(76) This book is tough [to get her [to avoid [reading ]]].

An early LFG-based analysis of tough constructions was proposed by Kaplan and

Bresnan (1982), according to which the subject of tough functionally controls a
topic as well as the obj in the subordinate clause. In contrast, Dalrymple and King
(2000) analyze tough constructions as involving a long-distance dependency in
which the gap object is anaphorically controlled by the matrix subject. A functional
control analysis, requiring identity between the matrix subject (the controller) and
the complement object (the controllee), is rejected because, for example, there is
case mismatch between the filler and the gap:
 syntax of long-distance dependencies 681

(77) a. It is tough to find him.

b. ⃰Him is tough to find .
c. He is tough to find .

Under anaphoric control, there is no expectation of case preservation because the

f-structure for the subject of tough is not the same as the f-structure for the object in
the complement clause; rather, two different f-structures are anaphorically related.
Following a body of work by, among others, Lasnik and Fiengo (1989), Jacobson
(1992a), Pollard and Sag (1994), Kim (1995), and Clark (2000), Dalrymple and
King (2000) claim that the matrix subject is assigned a thematic role. Evidence for
this is that, for example, the subject cannot be replaced by an expletive, and the
construction is incompatible with true idioms (Nunberg et al. 1994). According to
Dalrymple and King’s (2000) analysis, then, tough predicates select for a subj and
a comp, with the subj anaphorically controlling an unpronounced dis element in
the subordinate clause. The usual identity relation holds between the dis element
and an obj within the comp. In this way, the matrix subj and the subordinate obj
are related to one another via both anaphoric control (via the relation between
the matrix subj and the subordinate dis) and functional control (via the relation
between the subordinate dis and the subordinate obj). The reference of the
subordinate subj (in (78), the subject of kill) is determined by arbitrary anaphoric
control (§15.5).

(78) Moths are tough to kill.

’

This biclausal analysis is supported by, for instance, the interaction of the tough
predicate and a for-phrase, as shown in (79). Notice that the object in the for-phrase
can act as an overt controller of the comp subj:

(79) This race is tough for me to finish.

Dalrymple and King (2000) also discuss syntactic constraints on the path between
filler and gap in tough constructions. Notably, it is not possible for the gap in this
682 long-distance dependencies

construction to bear any grammatical function other than obj. Dalrymple and
King (2000) show that this obj can be inside an adjunct or oblique phrase, for
instance:

(80) This violin is easy to play the sonata on .

Revising Dalrymple and King’s original rule slightly in order to bring it in line with
our assumptions about f-structure and the dis function, the relevant constraints
are captured in (81), which appears in the lexical entry of the tough predicate:

(81) (↑ comp xcomp⃰ ({obl𝜃 | adj}) obj) ∈ (↑ comp dis)

According to this constraint, the complement (comp) of a tough-predicate has

a dis attribute, one of whose members controls an unexpressed obj which may
be embedded inside any number of xcomps, and may appear inside an oblique
or adjunct phrase. This constraint rules out a tensed complement clause on the
path: for many speakers, only xcomp can appear on the permissible path, and in
English xcomp must be nonfinite. A different path specification is required for
some speakers: Kaplan and Bresnan (1982) allow finite complements on the path,
classifying (82) as grammatical.

(82) %Mary is tough for me to believe that John would ever marry .

Note that the long-distance path specification in (81) is very different from
the path that is relevant for, for example, English topicalization (20); thus, tough
constructions neatly illustrate that long-distance dependencies and the constraints
upon them may vary not only across languages, but also within a single language.
Dalrymple and King (2000) observe that another set of predicates, including
need and require, also involve a “missing object,” but display distinct properties.
Their interaction with a for-phrase, for one, is different:

(83) ⃰The car needs/requires for him washing .

Further, the dependency involved is not arbitrarily long but bounded in the case
of need-type predicates:

(84) a. This house needs/requires renovating .

b. ⃰This house needs/requires trying [to renovate ].

In line with work by Barron (1999), Dalrymple and King (2000) analyze need-type
predicates as complex predicates that appear in a monoclausal structure; no long-
distance dependency is established. The f-structures of need-type and tough-type
predicates are therefore distinct; compare (78) with (85):
 resumptive pronouns 683

(85) The car needs washing.

 
’
  
 
  
  + 
 

17.2 Resumptive pronouns

Up to this point, we have considered only constructions in which the within-clause
function of a displaced phrase corresponds to the gap in a filler-gap dependency.
This is, of course, not the only possibility: many languages require a resumptive
pronoun, as in example (86a) from Swedish (Doron 1982) and example (86b) from
Irish (McCloskey 2002):

(86) a. Kalle letar efter en bok som han inte vet hur den slutar.
Kalle looks for a book that he not knows how it ends
‘Kalle is looking for a book that he does not know how (it) ends.’
b. an ghirseach a-r ghoid na síogaí í
the girl comp-pst stole the fairies her
‘the girl that the fairies stole (her) away’

The challenge for any LFG analysis of resumptives is to deal with the apparent
violation of wellformedness which arises. If the resumptive pronoun is the same as
any other pronoun, then both it and the dis phrase that binds it contribute pred
values to the same f-structure, in violation of Consistency (§2.4.6.3).
Falk (2002b) proposes that a pronoun in a language with resumptive pronouns
makes one of two different f-structure contributions: in its use as a standard
pronoun it contributes a pred attribute with value ‘pro’ to the f-structure, while
in its resumptive use, it does not contribute a pred value, but its f-structure is
required to be identical with a member of the dis set. Under the second option,
the same kind of long-distance dependency is established as in the filler-gap
structures that we have examined so far. This dual nature of pronouns in languages
with resumptive pronouns accounts for the differences in syntactic constraints
restricting “ordinary” pronouns compared to those restricting resumptive pro-
nouns: according to Falk the type of dependencies involved are fundamentally
different, and therefore it is unsurprising that their properties are not necessarily
the same.
What this analysis fails to capture, Asudeh (2004, 2012) argues, is McCloskey’s
Generalization (McCloskey 2002: 192): resumptive pronouns are, in fact, ordinary
pronouns, and do not take on special forms in their use as resumptives. Asudeh
observes that under Falk’s analysis, the similarities between ordinary and resump-
tive pronouns are a matter of coincidence. Asudeh’s own analysis differs in treating
684 long-distance dependencies

resumptive pronouns as no different from ordinary pronouns in terms of their

syntactic specification.
This does not mean that all resumptive pronoun constructions are identical
at f-structure, however. Asudeh (2012), building on work by McCloskey (2002),
shows that an important distinction can be made between syntactically active
resumptives (SARs) and syntactically inactive resumptives (SIRs). SIRs display
gap-like properties, and involve the kind of long-distance dependency that we
have exemplified for English in §17.1. Importantly, SIRs are sensitive to constraints
such as the across-the-board constraint (§17.1.1.1). By contrast, SARs exhibit
no such gap-like properties: an across-the-board violation would not result in
ungrammaticality, for instance. Asudeh (2012) identifies resumptive pronouns
in Swedish (86a) and Vata as SIRs, and resumptive pronouns in Irish (86b) and
Hebrew as SARs. In other LFG work, Camilleri and Sadler (2011) argue that
Maltese too has SARs.
Asudeh (2012) proposes that because SIRs display gap-like properties, they
should be analyzed in the same way as long-distance dependencies with a gap;
that is, that constructions including SIRs involve functional control. However, this
gives rise to an apparent Consistency violation, as noted above: if the resumptive
pronoun in a SIR construction is an ordinary pronoun, as Asudeh argues, then
both it and the dis phrase contribute pred values to the same f-structure, in vio-
lation of Consistency (§2.4.6.3)—an issue that Falk (2002b) addresses by allowing
resumptive pronouns to lack a pred value, in contrast to their ordinary pronoun
counterparts. Asudeh’s approach is to employ the restriction operator (§6.9.4) in
the specification of the relation between the dis f-structure and its within-clause
function, as outlined in general terms in (87) (adapted from Asudeh 2012: 147,
and similar to RTopicPath in §17.1.2):13

(87) (↑ dis)|pred = (↑ gf⃰ gf )|pred

((→ pred) = (↑ dis pred))

This constraint ensures that the features of the dis f-structure other than pred
are identical with the features of the f-structure of its corresponding within-clause
function. At the same time, the optional off-path constraint in (87) permits, but
does not require, identification between the pred value of dis and its within-clause
function. If the optional off-path constraint applies, the effect of the restriction of
pred features in (87) is neutralized, resulting in simple identity of f-structures.
This is exactly what is required for long-distance dependencies with a gap. But
in a long-distance dependency involving a resumptive pronoun, the f-structure
of the within-clause grammatical function has its own pred, meaning that the
optional off-path constraint must not apply. Thus a single path specification, that in
(87), handles both ordinary long-distance dependencies with a gap, and those with
a SIR.

13 Following Asudeh (2012), we assume that the value of the dis attribute (which Asudeh calls udf)
in Swedish is a single f-structure, not a set.
 resumptive pronouns 685

The resulting f-structure for a Swedish subject resumptive example is:

(88) Vemi trodde Maria att hani skulle fuska?

who thought Maria that he would cheat
‘Who did Maria think that (he) would cheat?’
 
’
   
 
   
   
   
     
   
   
    
  
 
 
 
 
 
 
 
 
 
   
 ’ 
 
  
   
   
   
  
    
    
      
    
    
    
 
    
 

The f-structures for dis and comp subj are not identical, since they have distinct
preds; yet they share all other features, which is why the features for gend and
case, which are not contributed by vem ‘who’ but by the resumptive pronoun,
appear in the dis f-structure. Compare this with the f-structure for the equivalent
sentence without a subject resumptive, shown in (89): since there is no pronoun to
contribute a pred value for the comp subj, the optional off-path constraint given in
(87) applies, with the result that the f-structures for the dis and the comp subj are
identical; at the same time, the features gend and case are absent in (89), because
the only pronoun present does not supply these features.

(89) Vemi trodde Maria i skulle fuska?

who thought Maria would cheat
‘Who did Maria think would cheat?’
686 long-distance dependencies
 
’
   
 
   
   
   
 
   
 
 
 
 
 
 
 
 
 
 
 ’ 
 
 

In this way, Asudeh addresses the issue of Consistency while preserving

McCloskey’s Generalization (McCloskey 2002: 192): resumptive pronouns in
SIR structures are just ordinary pronouns that can participate in a long-distance
dependency when licensed by the rule in (87).
By contrast, the fact that SARs do not exhibit gap-like properties indicates that
they are anaphorically bound pronouns, and thus Asudeh proposes a different
analysis for SARs, as this Hebrew example demonstrates:

(90) ʔet ha-yeledi she rina ʔohevet ʔoto

acc the-boy comp Rina loves him
‘the boy that Rina loves (him)’
 
 + 
 
 
 
   
 
 
 
 
 
   
 
   
 
   
 
   
 
  ’  
 
   
 
   
 
  

   
   
   
 
    
 
   
 
 
     
 
    
    
 
 



 p   
 
 
 
    
 
 
 
 

Here, the f-structure for the pronoun is labeled p: it is a standard pronominal

f-structure, contributing a pred attribute with value ‘pro’. Asudeh (2012) ana-
 morphological marking of ldd paths 687

lyzes the binding relationship between the SAR resumptive pronoun and the
dis f-structure as being established by the complementizer system in Hebrew.
Crosslinguistically, this is subject to variation: Asudeh argues that in Irish the
relationship is established by a particular complementizer, as we discuss in §17.3.2.

17.3 Morphological marking of LDD paths

Some languages signal long-distance dependency constructions by means of spe-
cial morphological or phonological forms, as noted by Clements and Ford (1979)
for Kikuyu, McCloskey (1979) for Irish, Chung (1982) for Chamorro, and Geor-
gopoulos (1985) for Palauan. These constructions were first analyzed in LFG by
Zaenen (1983) in work on Kikuyu, based on the early LFG treatment of long-
distance dependencies as relations between c-structure positions (Kaplan and
Bresnan 1982). Here we review the Kikuyu data that Zaenen (1983) originally
treated, presenting a distinct analysis that focuses on f-structure rather than
c-structure relations.14 We then turn to a discussion of the Irish complementizer
system, which signals the presence of a long-distance dependency by the use of
different complementizers.

17.3.1 Kikuyu
As discussed by Zaenen (1983) and in more detail by Clements and Ford (1979),
in affirmative declarative Kikuyu sentences the verb is associated with a downstep
tone.15 This downstep tone affects the phonology of the words following the verb:
the downstep turns a sequence of low tones into high tones in the immediately
following phrasal category. In (91), downstep is associated with the two verbs ɛ́:! rírɛ́
‘tell’ and átɛ́mírɛ́ ‘cut’:

(91) Kamaú ɛ́:! rírɛ́ Ka:nákɛ́ áté Káriók! í átɛ́mírɛ́ mótě

Kamau subj.tell.pst Kanake that Kariũki subj.cut.pst tree
‘Kamau told Kanake that Kariũki cut the tree.’

The downstep associated with ɛ́:! rírɛ́ ‘tell’ affects the words áté ‘that’ and Káriók! í
‘Kariũki’, whose citation forms are:

(92) a. ate
b. Kariokı̆

As Zaenen (1983) shows, the explanation for these differences can be ascribed to
the tonal shift imposed by the verb ɛ́:! rírɛ́ ‘tell’.

14 This analysis is based on unpublished joint work by Ron Kaplan, John Maxwell, Annie Zaenen,
and Mary Dalrymple.
15 In the following examples, high tones are marked with an acute accent, downstep tones are marked
with !, and low tones are unmarked.
688 long-distance dependencies

Interestingly, however, this tone change does not appear within the domain of a
long-distance dependency:

(93) nóo Kámaú ɛ́:! rírɛ́ Ka:nákɛ́ áte otɛ́mírɛ́ mote

who Kamau subj.tell.pst Kanake that subj.cut.pst tree
‘Who did Kamau tell Kanake that cut the tree?’

The question word nóo ‘who’ bears the dis function and also fills the role of the
subj of the verb otɛmírɛ́ ‘cut’ in the subordinate comp. The downstep that would
be expected to appear on both the matrix and subordinate clause verb does not
occur, and the effect that this downstep would have on the following words is
not present.16 As Clements and Ford (1979) and Zaenen (1983) show, the absence
of the downstep tone marks the domain of extraction of the fronted interrogative
pronoun. Here, the domain of extraction is the f-structure for the entire sentence.
We propose that every f-structure in the domain of extraction in Kikuyu is
marked with the attribute-value pair $ldd, +%:

(94) Each attribute on the path relating a member of the dis set to its within-
clause grammatical function must be an attribute of an f-structure that also
contains the pair $ldd, +%.

In (95), this constraint is satisfied, since the f-structures labeled t and c both contain
the attribute-value pair $ldd, +%:

(95) nóo Kámaú ɛ́:! rírɛ́ Ka:nákɛ́ áte otɛmírɛ́ mote

who Kamau subj.tell.pst Kanake that subj.cut.pst tree
‘Who did Kamau tell Kanake that cut the tree?’

’
+

c +

16 The high tone on the initial syllable of ate ‘that’ in (93) is spread from the final high tone of the
preceding word Ka:nákɛ́ ‘Kanake’ by an independent rule.
 morphological marking of ldd paths 689

Within the domain marked by the attribute-value pair $ldd, +%, spreading of the
downstep tone does not occur.
To constrain the path defining a long-distance dependency in Kikuyu, ensuring
that the path passes only through f-structures that are appropriately marked with
the ldd feature, we use off-path constraints, defined and discussed in §6.6 and in
§17.1. We assume a simplified definition of QDisPath for Kikuyu, representing it
as comp⃰ gf:17

(96) (↑ comp⃰ gf ) ∈ (↑ dis)

(←ldd) = + (←ldd) = +

In (96), the expression comp⃰ refers to zero or more occurrences of the

(←ldd) = +
constrained attribute comp . The metavariable ← in this off-path constraint
(←ldd) = +
refers to the f-structure that contains the attribute on which the constraint is
written. Thus, the off-path constraint (←ldd) = + ensures that each f-structure
that has a comp attribute on the path also has the attribute ldd with value +. The
equation in (96) associates the constraint (←ldd) = + with every attribute on
the path, and marks the entire Kikuyu domain of extraction with the ldd feature,
as desired.

17.3.2 Irish
Asudeh (2004, 2012) explores the marking of long-distance dependencies in
Irish, in which morphophonology varies according to the precise type of long-
distance dependency construction: the form of the complementizer in the extrac-
tion domain is dependent on whether the path terminates in a gap or a resumptive
pronoun. In the non-past, two of the complementizers have the same form, but
they are distinct in triggering different phonological mutations on the following
word. The complementizer used with a filler-gap construction triggers lenition
mutation (and is therefore often glossed as aL), while the one that co-occurs with a
resumptive pronoun triggers nasalization mutation (and therefore is often glossed
as aN). Asudeh (2012) cites the following examples from McCloskey (1979); notice
the effect of lenition on the forms of the verbs glossed ‘praise’:

(97) a. an scríbhneoir a mholann na mic léinn

the writer aL praise the students
‘the writer whom the students praise’
b. an scríbhneoiri a molann na mic léinn éi
the writer aN praise the students him
‘the writer whom the students praise (him)’

17 It is likely that grammatical functions other than comp are allowed in the extraction domain
in Kikuyu. In a full treatment of Kikuyu long-distance dependencies, the equation in (96) should be
enriched appropriately to characterize properly the full range of permitted relations between the initial
question word and its grammatical function within the clause, as was done for English in (58).
690 long-distance dependencies

Irish relative clauses lack relative pronouns. This means that the specifier of CP
position is unoccupied in an Irish relative clause, though not in Irish constituent
questions and cleft constructions. The phrase structure rule in (98) allows for the
analysis of relative clauses as well as constituent questions and clefts with a phrase
(XP in 98) filling the specifier position. When the specifier position is unfilled, the
CP is a relative clause: it must be a member of the adj set of the phrase in which it
is contained, and it has a dis attribute containing the f-structure we would expect
for a relative pronoun.18

(98) CP −→ { XP | 𝜖 } C,
↓ ∈ (↑ dis) (adj ∈ ↑) ↑=↓
%rel ∈ (↑ dis)
(%rel pred) = ‘pro’
(%rel prontype) = rel

In the rule in (98), a local name %rel is used. When the specifier of CP is not filled
by an XP at c-structure (hence 𝜖 for the empty string), the dis set includes a member
%rel whose feature pred has the value ‘pro’, and whose feature prontype has
value rel.
The two Irish long-distance dependency constructions that Asudeh analyses
differ in another significant respect. When a long-distance dependency involves
multiple clauses and a gap, aL appears in every clause on the path between filler and
gap. By contrast, when a long-distance dependency involves multiple clauses and
a resumptive pronoun, a second pattern emerges: aN appears in the “top” clause
in the dependency (the furthest clause from the resumptive pronoun, but the one
closest to its binder), and all intervening clauses contain a third complementizer
go instead.19 The full long-distance dependency path is therefore marked in Irish,
similar to the situation in Kikuyu examined earlier, but only in the case of a filler-
gap dependency and not a resumptive pronoun, as these data from McCloskey
(1979, 2002) demonstrate:

(99) a. an t-úrscéal [a mheas mé [a thuig mé ]]

the novel aL thought I aL understood I
‘the novel that I thought I understood’
b. firi [ar shíl Aturnae an Stáit [go rabh siadi díleas do’n
men aN thought Attorney the State go were they loyal to.the
Rí]]
King
‘men that the Attorney General thought (they) were loyal to the King’

Thus, the complementizer aL appears to have two functions. First, it relates a

member of the dis set to the within-clause grammatical function associated with

18 For a similar analysis of the pronoun-less relative clause construction in English, see §17.6.
19 McCloskey (2002) identifies three other multi-clause complementizer patterns which he refers to
as mixed chains; see McCloskey (2002) and Asudeh (2012) for discussion of these patterns.
 morphological marking of ldd paths 691

the gap, the defining relation in a filler-gap dependency. Second, in its “marking”
role in higher clauses, it establishes a link between the dis set in its clause and
the one in its comp. The latter applies in each intervening clause, resulting in the
multiple marking that we see in (99a).
Asudeh (2012) captures these features of long-distance dependency construc-
tions in Irish by means of a lexical entry like the one in (100) for the complemen-
tizer aL, which he analyzes as a nonprojecting word:20

(100) aL C
4 (↑ compform) = aL
(↑ comp⃰ gf) ∈ (↑ dis)
(→ dis) = (↑ dis)

The first constraint in the lexical entry in (100) ensures that aL specifies the
value aL for the attribute compform. The second constraint establishes the long-
distance dependency between a member of the dis set and some within-clause
function gf. The off-path constraint annotated on the comp attribute ensures that
if the within-clause function gf is embedded in one or more comps, the dis value of
each intervening comp f-structure is the same as the top-level dis value. Asudeh’s
analysis of Irish long-distance dependency path-marking thus crucially involves a
comp-to-comp relation established by the complementizer aL. As Asudeh notes,
the analysis has commonalities with the successive cyclic movement analyses of
Irish proposed by McCloskey (1990, 2002). On Asudeh’s analysis, however, there
is no movement involved; rather, a successive cyclic dependency is introduced.21
Combined with the phrase structure rule in (98), the lexical entry in (100)
enables us to formulate an analysis of a simple filler-gap dependency including
the complementizer aL in Irish:

(101) an scríbhneoir a mholann na mic léinn

the writer aL praise the students
‘the writer whom the students praise’

20 For discussion and motivation for the treatment of aL as a head-adjoined nonprojecting word C 4,
see Asudeh (2012: 157–63).
21 In fact, the analysis we present in this section differs from Asudeh’s analysis in assuming that the
value of the dis attribute in Irish is a set rather than a single f-structure: the constraint in (100) identifies
the entire top-level dis set with the dis set in the subordinate comps. This does not quite express the
intuition behind Asudeh’s analysis, which is stated in terms of a successive cyclic dependency involving
only one dis f-structure, not a set of f-structures. To express his intuition more faithfully, we could
instead propose the lexical entry in (a) for aL, which requires only that one member of the dis set is
shared between the top-level f-structure and the subordinate comps:
(a) aL C
4 (↑ compform) = aL
%d ∈ (↑ dis)
%d = (↑ comp⃰ gf)
%d ∈ (→ dis)
Since we are not concerned with cases in which the dis set has more than one member (and we do not
know of situations in which such a configuration would arise), we stick with the simpler formulation
in (100).
692 long-distance dependencies

DP

D0 NP

an NP CP
the
scríbhneoir IP
writer
I0 S

C I0
na mic léinn
a the students
aL mholann
praise

The specifier position of CP is unfilled in (101), and the dis set member pred ‘pro’
is introduced by the phrase structure rule in (98). The dependency established
between this member of dis and a within-clause function (in this case, obj) is
consistent with the lexical entry in (100).
The off-path constraint in (100) further requires the dis set in each intervening
comp to be identified with the dis set in the top-level f-structure.22 This takes care
of the “path-marking” function of aL in intervening clauses:

(102) an t-úrscéal [a mheas mé [a thuig mé ]]

the novel aL thought I aL understood I
‘the novel that I thought I understood’

22 In order to ensure successive cyclic marking, the possibility of the complementizer go appearing
in any intervening comp has to be ruled out by including ¬ (↑ dis) in the lexical entry for go (Asudeh
2012: 173).
 morphological marking of ldd paths 693

DP

D0 NP

an NP CP
the
t-úrscéal IP
novel
I0 S

C I0 DP VP

a mheas mé CP
aL thought I
IP

I0 S

C I0 DP

a thuig
mé
aL understood
I
694 long-distance dependencies

The final piece of the puzzle is semantic and concerns how the head is understood
to be modified by the relative clause(s). We discuss the semantic analysis of relative
clauses in §17.6, including the treatment of resumptive pronouns.

17.4 Traces and empty categories

The theory of the constituent structure properties of long-distance dependencies
in LFG has undergone a major revision since the inception of the theory. Kaplan
and Bresnan (1982) originally proposed to treat long-distance dependencies in
terms of a relation between two positions in the constituent structure tree, one
corresponding to the displaced constituent and the other to an unpronounced gap
or trace within the clause. On this view, the within-clause f-structure role of a dis-
placed constituent in a long-distance dependency is determined by the c-structure
position of its corresponding gap. This theory was further developed in work by
Zaenen (1980, 1983). Subsequently, however, Kaplan et al. (1987) and Kaplan
and Zaenen (1989b) showed that constraints on long-distance dependencies are
in fact primarily functional in nature and are best characterized in f-structure,
not c-structure, terms. Kaplan et al. (1987) and Kaplan and Zaenen (1989b) first
proposed functional uncertainty as a way of stating constraints on long-distance
dependencies, and we have adopted this analysis in the foregoing discussion.

17.4.1 Outside-in or inside-out?

Any analysis of long-distance dependencies must capture the duality of the dis-
placed phrase’s functional role in the sentence: not only is it displaced, it also has
a within-clause function. We refer to the top of the long-distance dependency (the
displaced phrase in English) as the filler and its bottom as the gap. Functional
uncertainty, as we have seen, enables us to state a relation between the filler and
the gap in terms of a path through f-structure which may be subject to certain
constraints. The standard approach is to use outside-in functional uncertainty to
define a path from filler to gap, which can be characterized in general terms
by annotations on the filler, as in our characterization in the foregoing. The
constraints in (103) define a relation between the filler and its within-clause gram-
matical role gf in terms of an appropriately-constrained long-distance path Path:

(103) Outside-in specification of the filler-gap dependency:

{[]}
CP

XP
↓ ∈ (↑ C
(↑ ↓
 traces and empty categories 695

In fact, however, an outside-in expression annotated on the filler is not the only
possible way in which a filler-gap dependency can be licensed. While in this book
we consistently adopt an outside-in functional uncertainty analysis (in line with the
vast majority of work on long-distance dependencies in LFG), it is also possible
to use inside-out functional uncertainty (§6.1.3) in constraining the filler-gap
relationship.
One of the issues with adopting an inside-out functional uncertainty approach
to long-distance dependencies is the question of what precisely the inside-out
expression should be associated with. One possibility is to assume the existence
of an unpronounced c-structure element, a trace or gap (represented as t in (104)),
with which the inside-out constraint can be associated:

(104) Inside-out specification of the filler-gap relation:

{[]}
CP

XP
C
↓ ∈ (↑
..
.

XP
(↑ ↓

t
↑∈ ↑

An alternative possibility, following analyses such as Ginzburg and Sag (2000) and
Bouma et al. (2001) in the HPSG framework, is to treat the constraint as an optional
lexical specification associated with all predicates, allowing any of the dependents
of the predicate to participate in a long-distance dependency. (Recall that gf ranges
over arguments and adjuncts of a predicate.)

(105) Optional inside-out constraint specifying a potential filler-gap dependency

involving a dependent of the predicate:
((↑ gf) ∈ ((Path ↑ ) dis))

This approach faces complications in languages in which more than one dependent
of the same predicate can participate in a long-distance dependency: Russian
examples such as (71b), repeated in (106), would require multiple inside-out
constraints associated with each predicate.

(106) Kogda kto udaril Borisa?

when who hit Boris
‘Who hit Boris when?’
696 long-distance dependencies

Falk (2006) argues against assuming a uniform approach to the type of functional
uncertainty involved in long-distance dependencies. Falk proposes that long-
distance dependencies involving subj as the within-clause grammatical function
of the dependency involve an outside-in path specification, while long-distance
dependencies involving non-subj functions require inside-out specifications.23 As
discussed in §17.4.2, Falk (2006) assumes that in the case of non-subj dependen-
cies, the inside-out specification is associated not with the predicate governing the
gap position, but with the gap itself, and he therefore assumes the existence of a
trace, or empty category, in the gap position at c-structure.
Most LFG analyses which use inside-out functional uncertainty to constrain
filler-gap dependencies assume the existence of traces. As discussed in §17.4.2, we
do not accept the existence of traces in this work, and adopt a uniform outside-in
analysis of long-distance dependencies.

17.4.2 Evidence for traces

Under a functional uncertainty approach, the role of constituent structure in
constraining long-distance dependencies is considerably diminished, and it is rea-
sonable to reevaluate the role of constituent structure and to reexamine evidence
for traces. Indeed, Kaplan and Zaenen (1989b) present a set of arguments against
the existence of traces (see also Sag and Fodor 1994), showing that many arguments
that have been made in support of traces are flawed. However, these works do
not address arguments for the presence of traces based on weak crossover, nested
dependencies, and wanna contraction.

17.4.2.1 Weak crossover Weak crossover, originally discussed by Wasow

(1979), refers to constructions such as the following:

(107) ⃰Whoi does hisi mother like ?

(cf. Whosei mother likes himi ?)

The name crossover comes from early transformational analyses of examples like
(107), which assumed that a violation ensues when a question phrase like who
“crosses over” a coindexed pronoun in moving to the front of the sentence: in
(107), who crosses over the coindexed pronoun his. Bresnan (1995, 1998, 2001c)
and Bresnan et al. (2016) propose a theory of prominence for anaphoric binding
that accounts for examples like (107) by assuming the presence of a trace in the
object position of like.
As Bresnan demonstrates, languages vary in imposing different types of promi-
nence requirements on the binding relation between a question phrase and the
pronouns it binds. Three prominence dimensions are relevant: the functional

23 Falk’s (2006) proposal is framed within his theory of subjecthood, according to which the function
subj is to be understood as the conflation of two distinct functions, pivot and gf.
4 As such, Falk’s claims
are stated relative to pivot vs. non-pivot functions, rather than subj vs. non-subj functions. In other
work, as here, Falk refers to the distinction in terms of subj vs. non-subj, for the sake of simplicity.
 traces and empty categories 697

hierarchy (on which, for example, subj outranks obj); the thematic hierarchy (on
which agent outranks theme); and linear order (f-preceding elements outrank
the elements they f-precede). Bresnan et al. (2016) propose that in English, the
relation between the question phrase and the pronoun is governed by either the
grammatical function hierarchy (Syntactic Prominence) or the linear precedence
hierarchy (Linear Prominence), so that satisfaction of either type of prominence
requirement is sufficient. The two relevant dimensions of prominence are summa-
rized by Dalrymple et al. (2007) as follows, where the “operator” is the displaced
question phrase:

(108) Prominence requirements (based on Bresnan et al. 2016):

Syntactic Prominence (A unit containing) the pronoun may not outrank
(a unit containing) the operator on the grammatical function hierarchy:
subj > obj > comp > ...
Linear Prominence The pronoun must not f-precede the operator.

In (107), the question phrase who fills the obj role, and the pronoun his is
contained in the subj. Thus, the pronoun outranks the question phrase in Syntactic
Prominence, violating the first prominence requirement in (108).
In examining whether the Linear Prominence condition in (108) is satisfied,
we must determine whether the pronoun his f-precedes the question phrase who.
Bresnan et al. assume that constituent question formation in English involves the
presence of traces, so that the f-structure for who in (107) corresponds to two c-
structure positions: the initial position in which who appears, and the gap position
following like. They rely on the definition of f-precedence given in (187) of §6.11.4,
repeated here:

(109) F-precedence, alternate definition (Bresnan et al. 2016: 213):

f f-precedes g if the rightmost node in 𝜙−1 (f ) precedes the rightmost node
in 𝜙−1 (g).

In (107), the rightmost node corresponding to the f-structure for who is the trace,
which follows the pronoun; therefore, the pronoun f-precedes the question phrase
according to the definition given in (109), and the Linear Prominence requirement
is not satisfied. Since neither the Syntactic Prominence condition nor the Linear
Prominence condition is met, (107) is correctly predicted to be ungrammatical.
Bresnan (1995, 1998), Berman (2003), and Bresnan et al. (2016) examine
weak crossover violations in German, proposing that German imposes the same
prominence conditions on binding by question phrases as English, and ascribing
differences between the two languages to variation in the presence of traces in the
two languages. The German equivalent of (107) is fully grammatical:

(110) Wen mag seine Mutter?

who.acc likes his mother.nom
‘Whoi does hisi mother like?’
698 long-distance dependencies

Berman (2003) and Bresnan et al. (2016) argue that no trace is present in German
examples like (110), since the grammatical function of the fronted argument is
determined locally, by reference to its casemarking, rather than by its position.
Since no trace appears in this clause, the pronoun does not f-precede the question
phrase; the Linear Prominence condition is thereby satisfied, accounting for the
acceptability of the example.
In contrast, they argue that traces must be assumed in extractions from sub-
ordinate clauses in German, since case information is insufficient to identify the
grammatical function of the extracted element in the embedded clause. Bresnan
et al. (2016) do not explicitly represent the trace in (111), since its position is
uncertain, though it appears somewhere within the subordinate clause:

(111) ⃰Wen meinte seine Mutter, [habe sie getrösted]?

who.acc said his mother has she consoled
‘Whoi did hisi mother say that she consoled?’

Here, the Linear Prominence condition is not satisfied because the f-structure for
wen ‘who’ also corresponds to the position of the trace in the subordinate clause.
The subordinate clause trace position is rightmost, and is therefore the one that
is relevant in evaluating f-precedence requirements. The matrix clause pronoun
f-precedes the trace in the subordinate clause, and so the Linear Prominence
condition is not met. Since the Syntactic Prominence condition is also not met, the
sentence is correctly predicted to be ungrammatical with the indicated indexing.
Thus, according to Bresnan’s proposal, weak crossover phenomena provide a new
view of the availability and distribution of empty c-structure categories in German,
Hindi, Malayalam, and Palauan.
However, not all accounts of weak crossover appeal to the existence of traces.
Dalrymple et al. (2001, 2007) explore a different definition of the Linear Promi-
nence condition, based on unpublished work by Sag (1998). In this approach, the
Linear Prominence condition is evaluated by considering the f-precedence relation
between the pronoun and an f-structure that contains the question phrase.
The approach proposed by Dalrymple et al. (2001) is defined in terms of co-
dependenthood; co-dependents are defined as in (112):24

(112) Co-Dependents:
the arguments and adjuncts of a single predicate

Dalrymple et al. provide revised prominence requirements which make reference

to co-dependents and, crucially, define Linear Prominence in terms of the rela-
tionship between the pronoun and CodepOp (the f-structure which contains the
operator rather than just the f-structure of the operator):

(113) Let CodepOp and CodepPro be co-dependents such that CodepOp con-
tains O and CodepPro contains P. Then:

24 Dalrymple et al. (2001) use the term “coargument” for this concept, but since the term ranges over
adjuncts as well as arguments, we use the term “co-dependent.”
 traces and empty categories 699

Syntactic Prominence An operator O is more prominent than a pronoun

P if and only if CodepOp is at least as high as CodepPro on the functional
hierarchy.
Linear Prominence An operator O is more prominent than a pronoun P
if and only if CodepOp f-precedes P.

According to these requirements, it is the head of the overt phrase containing the
gap (for instance, the preposition in the case of an oblique) which determines
Linear Prominence under Dalrymple et al.’s analysis. In (114), CodepOp (the
f-structure which contains the question phrase operator) is the comp f-structure,
as labeled:25

(114) ⃰Wen meinte seine Mutter, [habe sie getrösted]?

who.acc said his mother has she consoled
‘Whoi did hisi mother say that she consoled?’
’

Pron
CodepPro

CodepOp

CP

NP C
N C VP

Wen meinte NP VP
said
D N CP

seine Mutter
his mother habe sie getrösted
has she consoled

25 For clarity, in these examples displaced f-structures are depicted as the value of their within-clause
grammatical function at f-structure rather than as a member of the dis set. See §2.4.3 for a discussion
of the representation of f-structures which are the value of more than one different attribute.
 700 long-distance dependencies

According to either definition of f-precedence given in §6.11.4, the CodepOp’s

f-structure does not f-precede the pronoun’s f-structure, and the example is ruled
out. In contrast, in (115) CodepOp is the f-structure for wen ‘who’:

(115) Wen mag seine Mutter?

who.acc likes his mother-nom
‘Whoi does hisi mother like?’
’

CP

NP C Pron
CodepPro
N C VP

Wen mag NP
likes
D N CodepOp

seine Mutter
his mother

The CodepOp’s f-structure f-precedes the pronoun’s f-structure, and the Linear
Prominence condition is satisfied. The ungrammaticality of the corresponding
English example in (107) is accounted for by the Syntactic Prominence condition,
as described above; thus, this alternative account of weak crossover violations
differs from accounts presented by Berman (2003) and Bresnan et al. (2016)
in assuming that in English, though not in German, the Syntactic Prominence
condition must always be met. On this approach, weak crossover violations are
accounted for without assuming traces.
Nadathur (2013) proposes a further refinement of the definition of Linear
Prominence in a new LFG analysis of weak crossover, based on Pickering and
Barry’s (1991) Direct Association Hypothesis, which also eschews traces. Pickering
and Barry provide psycholinguistic evidence that the relevant link in a long-
distance dependency is not between the filler and the gap, but between the filler
and the predicate (verb, preposition, or other category) that selects for the gap,
which Nadathur refers to as its anchor. She provides the following definition of
Linear Prominence:

(116) Linear Prominence (adapted from Nadathur 2013):

The anchor (of the operator) must f-precede the pronoun.

This differs from Dalrymple et al.’s (2001) approach in treating the element which
subcategorizes for the operator, rather than a phrase containing the operator, as
participating in the relevant structural relationship.
 traces and empty categories 701

The primary contribution of this refinement to Dalrymple et al.’s (2001)

approach to weak crossover is identified by Nadathur as a lesser reliance
on Syntactic Prominence generally. For instance, on the definition of Linear
Prominence given in (113), there is no difference between subj and obj f-structures
in examples such as:

(117) a. ⃰[Whoi ]Op did [hisi ]Pronoun mother [greet]Anchor ?

b. [Whoi ]Op [greeted]Anchor [hisi ]Pronoun mother?

(117a) is ruled out because it violates Syntactic Prominence under Dalrymple

et al.’s (2001) analysis. By contrast, the definition of Linear Prominence provided in
(116) is by itself sufficient to account for the difference in grammaticality between
(117a) and (117b). This represents a considerable simplification. Nadathur (2013)
shows that the anchor approach to Linear Prominence also provides a promising
account of some Hindi and German weak crossover data.

17.4.2.2 Nested dependencies In many but not all languages, multiple long-
distance dependencies must not cross, but rather must be nested (Kuno and
Robinson 1972: 77):

(118) a. Nested dependency:

?This is the knife that this salami is easy to cut with .

b. Crossed dependency:
⃰This is the salami that my knife is easy to cut with .

These facts about crossed dependencies have been taken as evidence for traces,
the argument being that it is the position of the traces that is responsible for the
attested ungrammaticality. However, non-trace analyses of this phenomenon are
also possible: Dalrymple and King (2013) offer a non-trace approach.
Dalrymple and King (2013) reject an f-precedence analysis on the basis that no
definition of f-precedence makes it possible to single out unacceptable crossing
dependencies from other types of configurations involving multiple long-distance
dependencies. Instead, they offer an analysis framed in terms of operator superi-
ority, where each filler is identified as an operator. Their definition of superiority
incorporates a particular version of f-command (§6.9.1), dis-command, which
holds between operators each of which is in a dis set:

(119) F-command between operators (“dis-command”):

Operator f dis-commands operator g if and only if:
¬(f gf⃰) = g (f does not contain g) and
g ∈ ((dis ∈ f ) gf⃰ dis) (an f-structure whose value for the attribute dis
contains f also contains an f-structure whose value for the attribute dis
contains g).
702 long-distance dependencies

For example, f dis-commands g in the following f-structure:

 5 6 
(120) dis f[ ]
 7 5 68
 
... dis g[ ]

Dalrymple and King provide an initial formulation of the constraint which disal-
lows crossed dependencies:

(121) If two dependencies f-command each other, then if operator O1 dis-

commands operator O2, then T2 must be superior to T1.

The rule in (121) appeals to the definition of dis-command given in (119). What
remains to be defined is the superiority relation that holds between gaps (T1, T2).
For this, Dalrymple and King (2013) rely on the Direct Association Hypothesis
(Pickering and Barry 1991; Pickering 1993) and the notion of anchor discussed in
the previous section. On this view, the relevant superiority relation is not between
traces T1 and T2, but between Anchor1 (the predicate which selects for T1) and
Anchor2 (the predicate which selects for T2). Dalrymple and King (2013) define
the final version of the constraint on nested dependencies as follows:

(122) If two dependencies f-command each other, then if operator O1 dis-

commands operator O2, then the anchor for O1 (Anchor1) must not f-
precede the anchor for O2 (Anchor2).

The difference between nested and crossed dependencies can therefore be

accounted for without positing traces; in general, work relying on the Direct
Association Hypothesis appeals to the position of anchors rather than the position
of traces in precedence-related constraints on filler-gap dependencies.
(123) satisfies the constraint in (122): O1 dis-commands O2 and Anchor1 does
not f-precede Anchor2:

(123) Nested dependencies:

Which violin is this sonata tough to play on?
O1 O2 Anchor2 Anchor1

By contrast, (124) does not satisfy the constraint in (122) because Anchor1 f-
precedes Anchor2:

(124) Crossed dependencies:

⃰ hich sonata is this violin
W tough to play on?
O1 O2 Anchor1 Anchor2
 traces and empty categories 703

Dalrymple and King (2013) also review previous work which shows that, while
crossing dependencies result in ungrammaticality in English, for example, this is
not true crosslinguistically. Crossed dependencies are permitted in Norwegian and
merely dispreferred in Swedish (Maling and Zaenen 1982). Thus, the constraint in
(122) does not apply across all languages. Dalrymple and King (2013) consider how
this variation can be accounted for within the LFG framework without recourse to
traces.

17.4.2.3 “Wanna” contraction Another construction which has been argued to

support the existence of traces involves contraction. The claim is that the presence
of a trace or empty category blocks the possibility of contraction of, for example,
want to to wanna (Lakoff 1970; Pullum 1997):

(125) a. Who do you want to/wanna see ?

b. Who do you want to/⃰wanna see Maisie?

Falk (2007) presents a contraction-based account of this pattern which is framed

within a “last resort” approach to empty categories, which Bresnan et al. (2016)
view as a consequence of the Principle of Economy of Expression (§4.5). On Falk’s
view, “empty categories have no semantic content, so their only possible role is in
licensing grammatical f-structures.” Only when no alternative exists for licensing
an f-structure is an empty category’s presence justified. Falk argues that this is
the case in non-subj long-distance dependencies such as (125b), which, on Falk’s
analysis, are defined in terms of inside-out functional uncertainty (§17.4.1). Falk
argues that in the case of an inside-out path specification, there must be a node
in the c-structure for the inside-out equation to be associated with, and this node,
being devoid of lexical content, must be an empty category. The fact that empty
categories are required only in the case of non-subj long-distance dependencies
apparently permits an account of why wanna contraction is blocked in only these
contexts. However, it is unproblematic to introduce functional descriptions, such
as inside-out path specifications, without requiring the existence of a correspond-
ing node in the c-structure: empty node rules, introduced in §6.10.4, have precisely
this function. It therefore does not follow that an inside-out path specification
necessitates a “last resort” empty category, since an alternative is available.
Various alternatives to the trace-based analysis of wanna contraction have been
proposed, for example by Postal and Pullum (1978) and Bouchard (1982). Perhaps
the most promising alternative is that advocated by Pullum (1997), according to
which wanna is a distinct lexical item, restricted to subject-raising contexts, and
there is therefore no contraction in wanna sentences. This position is criticized by
Falk (2007), but Lorenz (2012, 2013) provides a detailed synchronic and diachronic
analysis of wanna and similar contractions, showing that they have been gram-
maticized as independent “semi-modal” verb forms and are not synchronically
mere phonological contractions of want to, etc.
704 long-distance dependencies

In sum, we find no incontrovertible evidence for the existence of traces or empty

categories, and in the absence of such evidence, a simpler and more parsimonious
theory of long-distance dependencies results if traces are avoided.

17.5 Multiple-gap constructions

In multiple-gap constructions, a single filler is related to multiple gaps:

(126) a. [Which articles]i did John file i without reading i?

b. Whoi did you tell i that you would visit i?

Multiple-gap constructions are distinct from “across-the-board” constructions; in

the former (127a) but not the latter (127b), a pronoun can appear in the position
of one of the gaps:

(127) a. [Which articles]i did John file i without reading i /themi ? (multiple
gap construction)
b. These are [the articles]i that [John filed i ] and [everyone else read
i /⃰themi ]. (‘across-the-board’ construction)

Multiple-gap constructions have been referred to in the literature as involving

“parasitic gaps,” the idea being that there is one “main” gap, and that additional gaps
are licensed by virtue of being parasitic on the “main” gap. However, Falk (2011)
points out that the term “parasitic gap” in fact is more appropriately applied to
only a subset of multiple-gap constructions. For example, the second gap in (127a)
is parasitic on the first gap: as shown in (128), a sentence with a full noun phrase in
the object position of reading is grammatical (128b), but a similar sentence with a
full noun phrase in the object position of file is not grammatical (128a). In contrast,
neither gap is parasitic on the other in the multiple-gap construction in (126b),
since either of the gaps can alternate with a pronoun or full noun phrase, as shown
in (129):

(128) a. ⃰[What/Which articles]i did John file the book without reading i?

b. [What/Which articles] did John file i without reading the book?

(129) a. [Who/Which friend]i did you tell i that you would visit your brother?
b. [Who/Which friend]i did you tell your brother that you would visit i?

For this reason, Falk advocates the term “multiple-gap construction” to refer to
both types of example, and we adopt Falk’s terminology here. Falk (2011) provides
the following f-structure for a multiple-gap construction:
 multiple-gap constructions 705

(130) Whoi did you tell i that you would visit i?

Two main issues are the subject of continuing investigation in LFG analyses of
such multiple-gap constructions. First, there is the question of how the relation
is established between a single filler and multiple gaps. As depicted in (103),
it is often assumed that filler-gap dependencies are constrained via an outside-in
specification on the filler:

(131) Annotation on the filler specifying a filler-gap dependency (see (103)):

(↑ Path gf) = ↓

As Falk (2011) notes, this outside-in annotation establishes a dependency between

a filler and a single gap. Additional annotations of a similar form would be needed
to establish a similar dependency involving multiple gaps. This issue would not
arise if such dependencies were established via inside-out specifications, as in
(104), since there is no obstacle to a single filler being the target of multiple filler-
gap specifications by multiple gaps. As discussed in §17.4.1, we do not adopt
the inside-out approach to long-distance dependencies advocated by Falk (2006),
and we assume that in languages and constructions which permit multiple gaps,
it is optionally possible for more than one instance of any outside-in filler-gap
dependency specification to apply.
A second challenge is accounting for the constraints which apply to multiple-gap
constructions. Early work, including Engdahl (1983), assumed that the relevant
constraints were best formulated in terms of c-command: specifically, that in a
multiple-gap construction one gap cannot c-command the other. This explains the
ungrammaticality of (132), in which the second gap is c-commanded by the first:
706 long-distance dependencies

(132) ⃰Whoi did you say i convinced you [ i should pass the course]?
(cf. You said Davidi convinced you (that) [hei should pass the course].)

However, Engdahl (1983) points out that the c-command requirement does not
account for all of the data: there are sentences, in whose non-gap equivalent a
reflexive can appear, that do not support multiple gaps, regardless of the fact that
no c-command relation holds between the two gaps.

(133) ⃰Whoi did you talk to i about i?

(cf. Whoi did you talk to i about himselfi ?)

Furthermore, grammaticality patterns differ when the higher gap is a c-

commanding non-subject (Chomsky 1986), indicating that the constraint is
properly formulated in terms of grammatical functions rather than c-command:

(134) Whoi did you tell i [that you would visit i ]?

Based on these patterns, Falk (2011) proposes that the appropriate constraint
is best expressed in terms of f-structure relations: a non-subject gap cannot
participate in a multiple gap construction with an f-commanding subject.

17.6 Relative clauses and semantic composition

17.6.1 Semantics of relative clauses
Like the adjectival modifiers studied in Chapter 13, a relative clause is a noun
modifier, producing a modified meaning of type $e → t% when combined with
a noun meaning of the same type. Recall that a common noun like man has
a meaning like the following, which we can think of as picking out the set of
individuals who are men:

(135) man
λx.man (x)

The meaning of a noun modified by a relative clause, like man who Chris saw,
is of the same type. It represents the set of individuals who are men (the meaning
contribution of man), who are people (the meaning contribution of who), and who
were seen by Chris (the meaning contribution of Chris saw):

(136) man who Chris saw

λx.person (x) ∧ see (Chris , x) ∧ man (x)

The meaning contribution of the relative pronoun who is redundant here, since the
fact that an individual is a man entails that he is a person. In some cases, however,
 relative clauses and semantic composition 707

the information contributed by the relative pronoun is not redundant: consider an

example like pitcher who/which Chris saw.
Relative clauses with possessive relative pronouns are interpreted similarly. We
assign the interpretation in (137) to the phrase man whose pen Chris used. We treat
the possessive determiner as definite, following Partee and Borschev (1998), and
we use the predicate name poss to represent the generalized “possession” relation
holding between the possessor and the possessed entity, which can encompass
temporary or permanent ownership as well as a range of other relations, as
discussed in §2.1.10.

(137) man whose pen Chris used

λx.the (y, poss (x, y) ∧ pen (y), use (Chris , y)) ∧ man (x)

17.6.2 Relative clauses and meaning assembly

We assume the following f-structure, semantic structure, and meaning constructor
for the phrase man who Chris saw:

(138) man who Chris saw

v[ ]
m m
s r[ ]
’
c

λx.person (x) ∧ see (Chris , x) ∧ man (x) : v ⊸ r

As described in Chapter 13, noun modifiers such as the relative clause who Chris
saw combine with the meaning of the noun that they modify to produce a new,
modified meaning of the same type. Thus, the implicational meaning constructor
in (138) is similar to the meaning constructor for a common noun like man, but
with a modified meaning reflecting the meaning of the relative clause:

(139) man
v[ ]
m m
r[ ]
λx.man (x) : v ⊸ r
708 long-distance dependencies

In the analysis of (138), we assume the relative clause rule given in (33) of this
chapter, augmented with the meaning constructor labeled [rel]:
   
(140) CP −→ RelP C,
 ↓ ∈ (↑ dis)   
  ↑ = ↓
 


(↑ RTopicPath) = ↓ 

[rel]
 (↑ relpro) = (↓ RelPath) 
(↑ relpro prontype) =c rel

As discussed in §8.6, there are cases in which meaning contributions are associated
with phrase structure configurations rather than lexical items. Relative clause
formation in English is one such case: the meaning constructor [rel], which
combines the relative clause meaning with the meaning of the modified noun, is
associated with the relative clause CP rule. The definition of [rel] is given in (141):

(141) [rel] λP.λQ.λx.P(x) ∧ Q(x) :

[(↑ relpro)𝜎 ⊸ ↑𝜎 ] ⊸
[[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)] ⊸
[((adj ∈ ↑)𝜎 var) ⊸ ((adj ∈ ↑)𝜎 restr)]]

Instantiating the ↑ metavariables in (141) according to the f-structure labels in

(138), we obtain this meaning constructor premise:

(142) [rel] λP.λQ.λx.P(x) ∧ Q(x) : [w𝜎 ⊸ s𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]

This meaning constructor consumes the meaning resource of the relative clause
w𝜎 ⊸ s𝜎 and the meaning resource of the modified noun v ⊸ r to produce a new
noun resource, also associated with v ⊸ r but reflecting a modified meaning.
We must also provide a meaning for the relative pronoun who. We propose that
who augments the relative clause meaning by providing the additional meaning
that the entity involved is a person. This is the lexical entry for who:

(143) who (↑ pred) = ‘pro’

(↑ prontype) = rel
λS.λx.person (x) ∧ S(x) :
[↑𝜎 ⊸ (relpro ↑)𝜎 ] ⊸ [↑𝜎 ⊸ (relpro ↑)𝜎 ]

Instantiating the ↑ metavariables in the meaning constructor in (143), we have the

meaning constructor labeled [who] in (144). We also assume the standard meaning
constructor for the name Chris as given in Chapter 8. Since the example we discuss
does not involve anaphoric binding, we provide simple noncontextual meaning
constructors rather than the PCDRT-based meaning constructors of Chapters 14
and 15. Thus, the following are the meaning constructor premises that are relevant
in the analysis of man who Chris saw:
 relative clauses and semantic composition 709

(144) Meaning constructor premises for man who Chris saw

[man] λx.man (x) : v ⊸ r
[who] λS.λx.person (x) ∧ S(x) : [w𝜎 ⊸ s𝜎 ] ⊸ [w𝜎 ⊸ s𝜎 ]
[Chris] Chris : c𝜎
[see] λx.λy.see (x, y) : c𝜎 ⊸ [w𝜎 ⊸ s𝜎 ]
[rel] λP.λQ.λx.P(x) ∧ Q(x) : [w𝜎 ⊸ s𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]

We begin by combining the premises labeled [Chris] and [see] to obtain the
meaning constructor labeled [Chris-see]:

(145) [Chris-see] λy.see (Chris , y) : w𝜎 ⊸ s𝜎

This provides the meaning resource required by the premise [who]. Combining
[Chris-see] and [who], we have the meaning constructor labeled [who-Chris-see]:

(146) [who-Chris-see] λx.person (x) ∧ see (Chris , x) : w𝜎 ⊸ s𝜎

Next, we combine the premises [who-Chris-see] and [rel], producing the premise
[who-Chris-see-rel]:

(147) [who-Chris-see-rel] λQ.λx.person (x) ∧ see (Chris , x) ∧ Q(x) :

[v ⊸ r] ⊸ [v ⊸ r]

As discussed in Chapter 13, this meaning constructor has the characteristic form
of a modifier: it consumes the resource [v ⊸ r] and produces a new resource of the
same form but with a modified meaning. Combining [who-Chris-see-rel] with
[man], we have, as desired:

(148) [man], [who-Chris-see-rel] 1

λx.person (x) ∧ see (Chris , x) ∧ man (x) : v ⊸ r

We now expand our treatment of English relative clauses to encompass relative

clauses with no relative pronoun, as in the phrase the man Chris saw. An advantage
of our analysis is that no additions or changes must be made to the lexical entries
or meaning constructors provided so far. All that is necessary is to augment the
c-structure rule given in (140) to provide the proper syntactic constraints when a
relative pronoun is not present, using the 𝜖 notation introduced in §6.10.4, as was
done in the rule for Irish relative clauses in (98):
 
(149) CP −→ { RelP | 𝜖 } C,
↓ ∈ (↑ dis) (↑ relpro) ∈ (↑ dis)  
↑ = ↓
(↑ RTopicPath) = ↓ (↑ RTopicPath) = (↑ relpro) [rel]
(↑ relpro) = (↓ RelPath) (↑ relpro pred) = ‘pro’
(↑ relpro prontype) =c rel
710 long-distance dependencies

According to this rule, when no RelP phrase is present, the equations under 𝜖
must be satisfied. The annotations on the 𝜖 node provide a dis attribute whose
value contains the relpro f-structure; this f-structure also plays a role inside the
clause defined by RelPath, and its value for the attribute pred is ‘pro’. With these
assumptions, the phrase man Chris saw has this f-structure, semantic structure,
and meaning constructor:

(150) man Chris saw

v[ ]
m ’ m
s r[ ]
c

λx.see (Chris , x) ∧ man (x) : v ⊸ r

The meaning derivation proceeds straightforwardly from the premises in (151),

which are contributed by the lexical items and the CP phrase structure rule figuring
in the analysis of this phrase:

(151) Meaning constructor premises for man Chris saw

[man] λx.man (x) : v ⊸ r
[see] λx.λy.see (x, y) : c𝜎 ⊸ [w𝜎 ⊸ s𝜎 ]
[Chris] Chris : c𝜎
[rel] λP.λQ.λx.P(x) ∧ Q(x) : [w𝜎 ⊸ s𝜎 ] ⊸ [[v ⊸ r] ⊸ [v ⊸ r]]

As above, we combine the premises labeled [Chris] and [see], obtaining the
premise labeled [Chris-see]:

(152) [Chris-see] λy.see (Chris , y) : w𝜎 ⊸ s𝜎

We combine [Chris-see] with [rel] to obtain [Chris-see-rel]:

(153) [Chris-see-rel] λQ.λx.see (Chris , x) ∧ Q(x) : [v ⊸ r] ⊸ [v ⊸ r]

Combining [Chris-see-rel] with [man], we obtain the desired result:

(154) [man], [Chris-see-rel] 1 λx.see (Chris , x) ∧ man (x) : v ⊸ r

 relative clauses and semantic composition 711

Our final task is to examine relative clauses with possessive relative pronouns. The
following are the f-structure, semantic structures, and meaning constructor for
man whose pen Chris used:

(155) man whose pen Chris used

pv[ ]
p
pr[ ]

w
p

m
u

’
c

mv[ ]
m
mr[ ]
λx.the (y, poss (x, y) ∧ pen (y), use (Chris , y)) ∧ man (x) : mv ⊸ mr

We assume this lexical entry for the possessive relative pronoun whose:

(156) whose (↑ pred) = ‘pro’

(↑ prontype) = rel
(↑ concord case) = gen
λP.λQ.λx.the (y, poss (x, y) ∧ P(y), Q(y)) :
[((poss ↑)𝜎 var) ⊸ ((poss ↑)𝜎 restr)] ⊸
[[(poss ↑)𝜎 ⊸ (relpro ↑)𝜎 ] ⊸ [↑𝜎 ⊸ (relpro ↑)𝜎 ]]

In (156), we instantiate this meaning constructor according to the f-structure labels

in (155), obtaining the meaning constructor labeled [whose] in (157). We also
provide the standard meaning constructor for the common noun pen, labeled
[pen]:

(157) Meaning constructor premises for whose and pen

[pen] λx.pen (x) : pv ⊸ pr
[whose] λP.λQ.λx.the (y, poss (x, y) ∧ P(y), Q(y)) :
[pv ⊸ pr] ⊸ [[p𝜎 ⊸ u𝜎 ] ⊸ [w𝜎 ⊸ u𝜎 ]]
712 long-distance dependencies

The meaning constructor of whose is similar to the meaning constructor for

every, discussed in Chapter 8: every requires a meaning for its restriction and a
meaning for its scope to produce a meaning for the sentence in which it appears.
The possessive determiner whose also requires a meaning for its restriction; this
requirement is represented on the right-hand side of the meaning constructor
labeled [whose] by a requirement for a resource pv ⊸ pr, corresponding to the
meaning P, exactly as for a determiner like every. This requirement is satisfied
by the meaning constructor [pen]. Combining [whose] and [pen], we have the
meaning constructor [wh-pen]:

(158) [wh-pen] λQ.λx.the (y, poss (x, y) ∧ pen (y), Q(y)) :

[p𝜎 ⊸ u𝜎 ] ⊸ [w𝜎 ⊸ u𝜎 ]

Besides a meaning for its restriction, a determiner like every requires a meaning
resource for its scope. When an appropriate scope meaning resource is found, it is
consumed and a new meaning resource for the scope is provided, incorporating
the semantics of the quantifier. Analogously, the possessive relative determiner
whose requires a meaning resource p𝜎 ⊸ u𝜎 , specifying its scope.26 When the
resource p𝜎 ⊸ u𝜎 is consumed, a meaning resource w𝜎 ⊸ u𝜎 for the relative clause
is provided.
In the derivation of the meaning of man whose pen Chris used, the meaning
constructors in (159) are relevant:

(159) Meaning constructors for man whose pen Chris used

[man] λx.man (x) : mv ⊸ mr
[wh-pen] λQ.λx.the (y, poss (x, y) ∧ pen (y), Q(y)) :
[p𝜎 ⊸ u𝜎 ] ⊸ [w𝜎 ⊸ u𝜎 ]
[use] λx.λy.use (x, y) : c𝜎 ⊸ [p𝜎 ⊸ u𝜎 ]
[Chris] Chris : c𝜎
[rel] λP.λQ.λx.P(x) ∧ Q(x) :
[w𝜎 ⊸ u𝜎 ] ⊸ [[mv ⊸ mr] ⊸ [mv ⊸ mr]]

We begin by combining [Chris] and [use] to produce [Chris-use]:

(160) [Chris-use] λy.use (Chris , y) : p𝜎 ⊸ u𝜎

This meaning constructor provides the scope resource required by [wh-pen].

Combining [Chris-use] and [wh-pen], we have the meaning constructor labeled
[wh-pen-Chris-use]:

26 Note that we treat the scope of the relative determiner whose as syntactically fixed (p𝜎 ⊸ u𝜎 ),
rather than unspecified as for a quantificational determiner like every. For discussion of scope ambiguity
for quantifiers, see §8.8.1.1.
 relative clauses and semantic composition 713

(161) [wh-pen-Chris-use]
λx.the (y, poss (x, y) ∧ pen (y), use (Chris , y)) : w𝜎 ⊸ u𝜎

We can now combine [rel] with [wh-pen-Chris-use], obtaining the meaning

constructor labeled [wh-pen-Chris-use-rel]:

(162) [wh-pen-Chris-see-rel]
λQ.λx.the (y, poss (x, y) ∧ pen (y), use (Chris , y)) ∧ Q(x) :
[mv ⊸ mr] ⊸ [mv ⊸ mr]

Combining this meaning with the meaning constructor [man], we have the desired
result for man whose pen Chris used:

(163) [wh-pen-Chris-see-rel], [man] 1

λx.the (y, poss (x, y) ∧ pen (y), use (Chris , y)) ∧ man (x) : mv ⊸ mr

17.6.3 Nonrestrictive relative clauses

Thus far, we have examined restrictive relative clauses. These can be distinguished
from nonrestrictive (or appositive) relative clauses: Arnold and Sadler (2010, 2011)
review the differences between the two. Semantically, restrictive relative clauses
(164) restrict the denotation of the noun which they modify, unlike nonrestrictive
relative clauses (165), which simply supply additional information. In English,
finite nonrestrictive relative clauses must contain a relative pronoun:

(164) Restrictive relative clause:

a. A person [who Sandy invited] will arrive later.
b. A person [that Sandy invited] will arrive later.
c. A person [Sandy invited] will arrive later.

(165) Nonrestrictive relative clause:

a. Kim, who Sandy invited, will arrive later.
b.⃰Kim, that Sandy invited, will arrive later.
c.⃰Kim, Sandy invited, will arrive later.

Nonrestrictive relative clauses are also associated with “wide scope” effects. For
example, Arnold and Sadler note that (166a), with a restrictive relative clause,
entails that Kim has a belief only about linguists who use the IPA. In contrast,
(166b), with a nonrestrictive relative, entails that Kim has a belief about linguists
in general, and the claim that linguists use the IPA is not a part of Kim’s beliefs:

(166) a. Kim believes that linguists who use the IPA are clever. [restrictive]
b. Kim believes that linguists, who use the IPA, are clever. [nonrestrictive]
714 long-distance dependencies

Arnold and Sadler (2010, 2011) assume that restrictive and nonrestrictive relative
clauses are syntactically integrated in the same way: they are both CPs adjoined
at the single-bar level, forming a phrase with the head that they modify. The
difference between them is rooted in their semantics. We assume the following
f-structure and semantic structure for the phrase Kim, who Sam dislikes, which
includes an appositive relative clause (ARC):

(167) Kim, who Sam dislikes

v[ ]
k k
d r[ ]
’
s

Arnold and Sadler (2010, 2011) base their semantic analysis of ARCs on the
approach to the semantics of supplemental constructions developed by Potts
(2005). As well as truth-conditional or at-issue content, Potts proposes another
dimension comprising Gricean conventional implicatures, ci content. Thus, each
expression is associated with a pair of meanings and a pair of types: one of the usual
at-issue type, and the other of the additional ci type, distinguished notationally
by a superscript c on the type specification. Potts proposes that parenthetical
and supplemental/appositive constructions, including ARCs, contribute to the
ci dimension of meaning. He analyzes constructions like ARCs as involving a
function comma which moves semantic content from the at-issue dimension to
the ci-dimension: formally, it takes an argument of type $e, t% and returns a result
of type $e, t c %, where t c is a ci rather than an at-issue type. In the case of an
ARC, the argument whose type is changed to ci content by comma is the relative
clause.
Arnold and Sadler (2010, 2011) combine Potts’s approach with the analysis of
relative clause semantics discussed in §17.6.2; their analysis derives an at-issue
meaning for the relative clause, which is reclassified by the [comma] meaning
constructor as ci meaning if the relative clause is nonrestrictive. We can begin by
instantiating the metavariables in the [rel] meaning constructor (141) according
to the f-structure labels in (167), resulting in the premise [rel] given in (168):27

(168) [rel] λP.λQ.λx.P(x) ∧ Q(x) : [w𝜎 $e% ⊸ d𝜎 $t% ] ⊸ [[v$e% ⊸ r$t% ] ⊸ [v$e% ⊸ r$t% ]]

27 Arnold and Sadler (2010, 2011) provide a slightly different formulation of [rel] from that given
in (141), because they associate [rel] with the CP node, rather than C, . The result of instantiation is
identical, so we ignore this detail of their analysis here.
 relative clauses and semantic composition 715

For clarity, we explicitly annotate semantic structures with their types, since the
distinction between at-issue and ci types is important. Note that all of the types in
the [rel] meaning constructor are at-issue types.
The meaning of the relative clause who Sam dislikes (whether used restrictively
or nonrestrictively) is derived from the following premises:

(169) Meaning constructors for the relative clause who Sam dislikes
[dislike] λx.λy.dislike (x, y) : s𝜎 $e% ⊸ [w𝜎 $e% ⊸ d𝜎 $t% ]
[Sam] Sam : s𝜎 $e%
[who] λS.λx.person (x) ∧ S(x) : [w𝜎 $e% ⊸ d𝜎 $t% ] ⊸ [w𝜎 $e% ⊸ d𝜎 $t% ]
[rel] λP.λQ.λx.P(x) ∧ Q(x) :
[w𝜎 $e% ⊸ d𝜎 $t% ] ⊸ [[v$e% ⊸ r$t% ] ⊸ [v$e% ⊸ r$t% ]]

We also assume the following meaning for the noun Kim:

(170) [Kim] Kim : k𝜎 $e%

Combining the premises for the relative clause in the usual way, we have the
meaning constructor in (171) for the relative clause:

(171) [dislike], [Sam], [who], [rel] 1

[who-Sam-dislike] λQ.λx.person (x) ∧ dislike (Sam , x) ∧ Q(x) :
[v$e% ⊸ r$t% ] ⊸ [v$e% ⊸ r$t% ]

According to Potts, in order to be interpreted as nonrestrictive, a relative clause

must combine with the comma function, by which the meaning of the relative
clause is reclassified as ci content. For their LFG analysis, Arnold and Sadler (2011)
define the meaning constructor [comma] in (172), with s-structure variables
instantiated according to (167):

(172) [comma] λP.λy.y × [P(λz.true )(y)] :

[[v$e% ⊸ r$t% ] ⊸ [v$e% ⊸ r$t% ]] ⊸ [k𝜎 $e% ⊸ [k𝜎 $e% ⊗ k𝜎 $tc % ]]

This meaning constructor contains some elements we have not previously encoun-
tered. On the glue side, the meaning constructor requires a resource of the type
[v ⊸ r] ⊸ [v ⊸ r], which corresponds to the meaning of the relative clause who
Sam dislikes. Corresponding to this on the meaning side, the meaning P of the rela-
tive clause is applied to the vacuous $e → t%-type meaning λz.true. This satisfies the
variable Q in the meaning of the relative clause in (171). The result of combining
the meaning constructors [who-Sam-dislike] and [comma] is as follows:

(173) [who-sam-dislike], [comma] 1

[who-Sam-dislike-comma] λx.x × [person (x) ∧ dislike (Sam , x) ∧ true] :
k𝜎 $e% ⊸ [k𝜎 $e% ⊗ k𝜎 $tc % ]
716 long-distance dependencies

On the glue side, this meaning constructor requires a resource corresponding to

k𝜎 , and returns the resource k𝜎 conjoined with a ci-type resource associated with k𝜎
and corresponding to the meaning of the relative clause. This is represented in the
glue expression by the multiplicative conjunction operator ⊗, introduced in the
Appendix of Chapter 8 (page 321). Corresponding to this on the meaning side, if a
meaning for the variable x is found, the result is a conjunction of that meaning
with the meaning of the relative clause. The result of combining the meaning
constructors [who-sam-dislike-comma] and [Kim] is therefore:

(174) Kim × [person (Kim ) ∧ dislikes (Sam , Kim ) ∧ true ] : k𝜎 $e% ⊗ k𝜎 $tc %

This meaning constructor can be used in a larger phrase or clause in exactly the
same way as the meaning constructor [Kim]. How the ci meaning content interacts
with the at-issue content in larger phrases and clauses is discussed by Arnold and
Sadler (2011). Giorgolo and Asudeh (2011a, 2012a) present an alternative analysis
of supplemental, non-at-issue, content which appeals to monads, a construction
from category theory.

17.6.4 Resumptive pronouns and semantic composition

Resumptive pronouns represent a challenge to a resource-sensitive approach to
semantic composition. This is an issue explored in detail in Asudeh’s (2004,
2012) work on pronominal resumption in LFG. In short, resumptives appear
to represent a surplus resource. As Asudeh (2012: 126) states, “if a resumptive
pronoun is a surplus resource then the pronoun constitutes a barrier to the
basic compositional requirements of the sentence in which it occurs. This would
normally lead to ungrammaticality.” This is the case in English, where the following
is ungrammatical:

(175) ⃰The man who Chris saw him laughed.

(cf. The man who Chris saw laughed.)

However, as discussed in §17.2, there are languages where the equivalent of

(175) is grammatical. If a resumptive pronoun is, as Asudeh proposes following
McCloskey (2002), an ordinary pronoun that makes the usual syntactic and
semantic contribution, what consumes the “surplus” semantic resource provided
by the resumptive pronoun? Asudeh (2012) proposes that languages in which
resumptive pronouns are grammatical have as part of their grammar manager
resources which consume the surplus pronominal resource. It is the availability
of manager resources, contributed by complementizers and possibly other lexical
items, that determines whether resumptive pronouns are licensed in a language.
As noted in §14.4, Asudeh’s analysis of resumptive pronouns is based on the
semantic treatment of pronouns proposed by Dalrymple et al. (1997). Though we
do not adopt that proposal in this book, we describe it briefly here as a prelude to
our discussion of Asudeh’s theory of pronominal resumption.
 relative clauses and semantic composition 717

Asudeh (2012) assumes that the anaphor-antecedent relation is explicitly speci-

fied by the anaphor’s antecedent attribute at semantic structure. For example, in
the sentence Davidi thought that Chris had seen himi , where David is the antecedent
of him, the semantic structure for David is the value of the antecedent attribute
in the semantic structure of the pronoun him:

(176) F-structure and s-structure for Davidi thought that Chris had seen himi
according to Asudeh (2012), with the antecedent relation explicitly
represented at s-structure:
’
d
’

p d []

Following Dalrymple et al. (1997), Asudeh (2012) proposes that the meaning
constructor for a pronoun consumes the meaning of the pronoun’s antecedent and
produces it again, in the process assigning the same meaning to the pronoun. This
is accomplished by the following meaning constructor:

(177) Pronoun meaning constructor according to Asudeh (2012):

him (↑ pred) = ‘pro’
λx.x × x : (↑𝜎 antecedent) ⊸ [(↑𝜎 antecedent) ⊗ ↑𝜎 ]

On this view, the instantiated meaning constructors for the pronoun him and its
antecedent David in (176) are:

(178) [David] d𝜎 : David

[him] λx.x × x : d𝜎 ⊸ [d𝜎 ⊗ p𝜎 ]

The meaning constructor for the pronoun consumes the meaning of its antecedent
and produces a pair of resources, one for the pronoun and one for the antecedent,
with the same meaning assigned to both resources. Combining [David] and [him],
we have the meaning constructor in (179), where the pronoun p𝜎 and its antecedent
d𝜎 are both associated with the meaning David :

(179) David × David : [d𝜎 ⊗ p𝜎 ]

718 long-distance dependencies

According to Asudeh (2012), languages with resumptive pronouns have manager

resources to consume a surplus pronominal resource of the form in (177). Manager
resources have the following basic form, with A standing for antecedent and P
standing for (resumptive) pronoun meaning:

(180) Manager resources (Asudeh 2012):

[(A ⊸ A ⊗ P)] ⊸ (A ⊸ A)

As shown in (177), the pronoun’s meaning constructor can be represented as

A ⊸ (A ⊗ P), consuming the antecedent meaning A to produce a pair of resources
A ⊗ P for the antecedent and the pronoun. The manager resource in (180) con-
sumes the pronoun meaning constructor A ⊸ (A ⊗ P), and produces a resource
which does not include P, only A ⊸ A. This new resource combines with the
meaning resource of the antecedent A, leaving only the meaning of the antecedent
A, as desired. In this way, the manager resource removes the meaning resource of
the pronoun, with the result that there is no surplus resource.
Asudeh (2012) bases his discussion of the resumptive pronoun construction in
Irish on this example from McCloskey (2002: 190):

(181) firi ar shíl Aturnae an Stáit go rabh siadi díleas do’n

men aN thought Attorney the State go were they loyal to.the
Rí
King
‘men that the Attorney General thought (they) were loyal to the King’

’
g
m p a []
t ’

l p

In (181), the resumptive pronoun is labeled p, and its antecedent is the dis f-
structure labeled a.
We now provide a more explicit formulation of the manager resource which
consumes the surplus pronominal meaning constructor. Notice that the position
of the resumptive pronoun is not syntactically fixed; Asudeh uses a local name %rp
 relative clauses and semantic composition 719

to refer to the f-structure of the resumptive pronoun, which appears somewhere

within the relative clause:

(182) %rp = (↑ gf+ )

Given this definition of %rp, the manager resource consuming the pronoun mean-
ing constructor can be formulated as in (183) (similar in form to the schematic
resource in (180)):

(183) [(%rp𝜎 antecedent) ⊸ [(%rp𝜎 antecedent) ⊗ %rp𝜎 ]]

⊸ [(%rp𝜎 antecedent) ⊸ (%rp𝜎 antecedent)]

Asudeh assumes that the antecedent of the resumptive pronoun is a member of

the dis set, as shown in (181); in other words, (%rp𝜎 antecedent) is the semantic
structure a𝜎 . Thus, the expression in (184) is equivalent to the expression in (183):

(184) [a𝜎 ⊸ [a𝜎 ⊗ %rp𝜎 ]] ⊸ [a𝜎 ⊸ a𝜎 ]

There is another bookkeeping issue that must be addressed. As shown in (181),

Asudeh’s analysis of Irish does not assume a relpro attribute. Instead, his analysis
assumes that dis in Irish plays the same role as relpro in English, and thus that
relative clause meanings in Irish involve a dependency on a member of the dis
set rather than relpro. However, when the meaning of the resumptive pronoun is
consumed by the manager resource, the resulting relative clause meaning involves
a dependency on the role of the resumptive pronoun, not dis. As a result, his
analysis requires the introduction of a “relabeling” meaning constructor which
consumes a dependency involving the resumptive pronoun and produces the
required dependency involving dis.

(185) “Relabeling” meaning constructor according to Asudeh (2012):

[%rp𝜎 ⊸ ↑𝜎 ] ⊸ [(↑ dis ∈)𝜎 ⊸ ↑𝜎 ]

Instantiating the ↑ metavariables in (184) and (185) according to the f-structure

labels in (181), and adding the corresponding meanings, we obtain the meaning
constructors in (186), with the manager resource labeled [mr]:

(186) [mr] λP.λx.x : [a𝜎 ⊸ (a𝜎 ⊗ p𝜎 )] ⊸ [(a𝜎 ⊸ a𝜎 )]

[relabel] λP.P : [p𝜎 ⊸ t𝜎 ] ⊸ [a𝜎 ⊸ t𝜎 ]

The lexical entry for the complementizer aN discussed in §17.3 is revised to include
these meaning constructors.
The full set of meaning constructors that are relevant in the analysis of (181)
is given in (187). We assume that the semantic structure corresponding to m has
a var attribute which we call v, and a restr attribute which we call r. Asudeh
(2012) represents the denotation of plural fir ‘men’ as λx.man⃰ (x), where the
720 long-distance dependencies

asterisk represents pluralization. He also makes the simplifying assumption that

díleas translates as loyal-to, and that do’n Rí ‘to the King’ translates as ιk.king (k) of
type e.28

(187) Meaning constructor premises for (181)

[man⃰] λx.man⃰ (x) : v ⊸ r
[rel] λP.λQ.λz.P(z) ∧ Q(z) : (a𝜎 ⊸ t𝜎 ) ⊸ [(v ⊸ r) ⊸ (v ⊸ r)]
[mr] λP.λx.x : [a𝜎 ⊸ (a𝜎 ⊗ p𝜎 )] ⊸ [(a𝜎 ⊸ a𝜎 )]
[they] λz.z × z : a𝜎 ⊸ [a𝜎 ⊗ p𝜎 ]
[relabel] λP.P : [p𝜎 ⊸ t𝜎 ] ⊸ [a𝜎 ⊸ t𝜎 ]
[AG] atty-gen : g𝜎
[think] λx.λy.think (x, y) : g𝜎 ⊸ l𝜎 ⊸ t𝜎
[loyalto] λy.λx.loyal-to (x, y) : k𝜎 ⊸ p𝜎 ⊸ l𝜎
[theKing] ιk.king (k) : k𝜎

We begin by combining the premises labeled [loyalto] and [theKing] to obtain the
meaning constructor labeled [loyalto-theKing]:

(188) [loyalto-theKing] λx.loyal-to (x, ιk.king (k)) : p𝜎 ⊸ l𝜎

Next, we combine the premises labeled [AG] and [think] to obtain the meaning
constructor labeled [AG-think]:

(189) [AG-think] λy.think (atty-gen , y) : l ⊸ t

Combining [AG-think] and [loyalto-theKing],29 we obtain the meaning con-

structor labeled [AG-think-loyalto-theKing]:

(190) [AG-think-loyalto-theKing]
λx.think (atty-gen , loyal-to (x, ιk.king (k)))) : p ⊸ t

This meaning is of the form required by the [relabel] meaning constructor, and so
we can combine [AG-think-loyalto-theKing] and [relabel], producing the mean-
ing constructor [relabel-AG-think-loyalto-theKing]. The resulting meaning is
unchanged, but the dependency on the glue side is now on a𝜎 rather than p𝜎 :

28 The iota operator ι is something like the definite determiner the: ιk.king (k) is the individual k
satisfying the description king (k). For discussion, see Gamut (1991a: Chapter 5).
29 This step involves assumption of a hypothetical resource p to produce a resource l from
[loyalto-theKing], combining the result with [AG-think], and subsequently discharging the assump-
tion p. We do not represent these steps explicitly; for discussion of assumption and discharge of
hypothetical resources, see §8.8.1.4.
 constituent questions and semantic composition 721

(191) [relabel-AG-think-loyalto-theKing]
λx.think (atty-gen , loyal-to (x, ιk.king (k)))) : a𝜎 ⊸ t𝜎

We now use the manager resource [mr] to consume the meaning of the pronoun
[they]. The result of combining [mr] and [they] is the meaning constructor labeled
[mr-they], in which the surplus pronoun meaning has been consumed:

(192) [mr-they] λx.x : a𝜎 ⊸ a𝜎

We can now combine [mr-they] with [relabel-AG-think-loyalto-theKing], leav-

ing both the meaning side and the glue side unchanged; to indicate that the
meaning constructor [mr-they] has been consumed in the derivation, we provide
the new label [mr-they-relabel-AG-think-loyalto-theKing]:

(193) [mr-they-relabel-AG-think-loyalto-theKing]
λx.think (atty-gen , loyal-to (x, ιk.king (k))) : a𝜎 ⊸ t𝜎

This meaning constructor is of the form required by [rel]. Combining the meaning
constructor in (193) with [rel], we obtain the meaning constructor in (194):

(194) [rel-mr-they-relabel-AG-think-loyalto-theKing]
λP.P(x) ∧ think (atty-gen , loyal-to (x, ιk.king (k))) : (v ⊸ r) ⊸ (v ⊸ r)

Combining this meaning constructor with [man⃰], we have, as desired:

(195) λx.man⃰ (x) ∧ think (atty-gen , loyal-to (x, ιk.king (k))) : v ⊸ r

Belyaev and Haug (2014) provide an analysis of the syntax and semantics of
correlatives in Ossetic (discussed earlier in §17.1.2) which also deals with the issue
of resumption, though by its nature this differs from the kind of resumption with
which Asudeh (2012) is concerned. In a correlative, a subordinate clause contains a
relative pronoun or phrase that is in some sense resumed by a full nominal phrase
in the matrix clause. Belyaev and Haug (2014) analyze as an instance of pronominal
anaphora the relationship between the relative pronoun/phrase and its correlate
which appears in the matrix clause. Their LFG analysis of the semantics of Ossetic
correlatives is framed in terms of glue semantics as well as Partial Compositional
Discourse Representation Theory (see Chapter 14).

17.7 Constituent questions and semantic composition

Much current work on the semantics of questions has its roots in the work of Ham-
blin (1958, 1976), who shows that there is an intimate relationship between the
meaning of a question and the meanings of its possible complete answers. Hamblin
(1958) outlines the following postulates for the interpretation of questions:
722 long-distance dependencies

(196) (i) An answer to a question is a statement.

(ii) Knowing what counts as an answer is equivalent to knowing the
question.
(iii) The possible answers to a question are an exhaustive set of mutually
exclusive possibilities.

Subsequent work by Karttunen (1977), Groenendijk and Stokhof (1984), Ginzburg

and Sag (2000), Ginzburg (2001), Krifka (2001), Nelken and Francez (2002), and
many others has expanded and refined our view of the semantics of questions.
Useful overviews are presented by Ginzburg (1996), Higginbotham (1996), Groe-
nendijk and Stokhof (1997), and Ginzburg (2011).
Many complications arise in the semantic analysis of questions, and a complete
treatment of question semantics in a glue setting must await future research. Here
we discuss some of the issues that must be addressed.
The first issue is to determine an appropriate representation of question mean-
ing. The meaning of a question is inherently intensional, and so intensional logic
provides a more appropriate way of representing question meanings than the
predicate logic representations that we have assumed in previous chapters. Here
as always, however, our primary focus is on semantic composition, not the details
of semantic interpretation. In fact, we believe that the same issues in semantic
composition arise in considering an appropriately simplified question meaning as
would arise in a more complete treatment, so that a simple predicate logic meaning
representation is sufficient for our present discussion. Following Ginzburg (1996),
we provide the meaning in (197) for the question Who does David like?

(197) Who does David like?

λP.[∃x.(person (x) ∧ P = like (David , x))]

In this expression, P represents members of the set of propositions that constitute

answers to the question of who David likes: in other words, this expression picks
out the set of propositions of the form like (David , x) for each person x whom David
likes:

(198) λP.[∃x.(person (x) ∧ P = like (David , x))] =

{like (David , Ken ), like (David , Mary ), like (David , Matty ) . . .}

This simple treatment is compatible with the Hamblin postulates, since it identifies
the meaning of the question with the set of propositions that constitute its complete
answer.
Next, we examine the issue of semantic composition and identification of the
meaning constructors that are involved in the derivation of a question meaning.
The sentence Who does David like? has the following f-structure:
 constituent questions and semantic composition 723

(199) Who does David like?

f
’

A major difference between the meaning of a declarative sentence like David likes
Chris and the meaning of the question Who does David like? is the type of the
expressions representing their meaning. The expression in (197) is of type $t → t%:
it represents a set of possible answer meanings such as like (David , Chris ), each of
which has the type t of a declarative sentence. Up to now, we have associated only
the basic types e and t with semantic structures, not more complex types like
$t → t%. We can continue to associate only basic types with semantic structures
if the meaning of a question is associated not with a single semantic structure
but with an implicational meaning constructor, one whose right-hand side has
the form A ⊸ B. However, we have argued in Chapter 8 that a semantically
complete and coherent glue derivation results in a meaning constructor that is not
implicational, so that some refinement to the definition of semantic completeness
and coherence would need to be provided.
An alternative approach is to permit higher types such as $t → t% to be associ-
ated with semantic structures. For example, the meaning representation in (197)
is of type $t → t%, and it would be associated with the semantic structure f𝜎 in a
full analysis of the question in (199). In that case, we must provide a theory that
accounts for the discrepancy between the question type $t → t% and the basic type
t that is associated with a declarative sentence whose head is the verb likes. The
possibility of associating higher types with semantic structures is argued for by
Lowe (2014b), for reasons independent of the semantics of questions.
Finally, a desirable characteristic of any analysis of constituent questions is that
it should extend unproblematically to multiple constituent questions: questions that
contain more than one interrogative pronoun. The meaning constructors that are
relevant in the analysis of the question Who does David like? should be reusable
straightforwardly and without augmentation to produce the following meaning
representation:

(200) Who likes who?

λP.[∃x.∃y.(person (x) ∧ person (y) ∧ P = like (x, y))]

The most complete discussion and analysis of the semantics of constituent ques-
tions in an LFG setting is by Mycock (2006). Mycock adopts Ginzburg and Sag’s
(2000) propositional abstract approach to the semantics of questions, according
724 long-distance dependencies

to which questions are “open propositions” that qualify as complete semantic

objects in their own right. According to this unified, non-quantificational theory
of the semantics of questions, each question word contributes a parameter and a
place-holder. A parameter is a restriction-bearing element that links an abstracted
argument to an argument role. Parameters are members of a set which takes scope
over a proposition containing at least one place-holder, giving a propositional
abstract. A place-holder indicates the “gap” that a question word represents: it
does not fill a semantic argument role, rather the relevant abstraction substitutes
in a place-holder for a regular “role-filler.” Under this analysis place-holders,
like propositional abstracts, are themselves structured semantic objects defined
in terms of other ontological entities. Mycock proposes meaning constructors
for the two-fold semantic contribution of a question word (a parameter and a
place-holder), along with a meaning constructor for the parameter set (of which
parameters contributed by question words are members) and a meaning con-
structor that defines the parameter set’s scope (a proposition containing at least
one place-holder). Ginzburg and Sag’s (2000) semantics of questions, upon which
Mycock’s (2006) LFG analysis is based, offers a situation semantics approach to
the meaning of questions. While situation semantics has been used to represent
linguistic meaning by researchers working within LFG (for instance, Fenstad et al.
1987; Gawron and Peters 1990), a full glue approach has yet to be proposed.
In sum, the best analysis of question semantics and semantic composition in
a glue setting is obtained by assuming that the meaning constructor for an inter-
rogative pronoun like who combines seamlessly with the independently motivated
semantic contributions of the other meaning-bearing items in the sentence to
produce the desired meaning constructor, just as the meaning constructors for
the relative pronoun and the relative clause combine with the other meaning con-
structors in the relative clause to produce an appropriate relative clause meaning
constructor. It may be that some basic assumptions about meaning representations,
semantic types, or other aspects of the glue approach will require some degree of
modification to give an adequate account of the meanings of questions and other
nondeclaratives; future research will reveal more about how these issues should be
resolved.

17.8 Further reading and related issues

The standard LFG analysis presented in this chapter assumes that there is a
semantic relation between the relative pronoun and the modified head noun, as
detailed in §17.6.1, but that the modified head noun in the matrix clause does not
bear a syntactic relation within the relative clause. Falk (2010) provides a number of
arguments in favor of an alternative analysis which establishes a syntactic relation
within the relative clause involving the modified head noun.
Section 17.1.1.1 discussed across-the-board effects in long-distance dependen-
cies, providing a syntactic analysis; Asudeh and Crouch (2002a) revisit these
effects, recasting them as a constraint on proof parallelism in a glue semantics
setting.
 further reading and related issues 725

Category mismatches or movement paradoxes, in which the category of a dis-

placed constituent differs from its in-situ equivalent, were discussed briefly in
§4.3.3. Such constructions are not problematic for an LFG analysis, as mismatches
between different levels of structure are to be expected. Bresnan et al. (2016)
analyze movement paradoxes in English, and Broadwell (2010) discusses two
movement paradoxes in Zapotec.
Excluding certain grammatical functions from the path between filler and gap
on a language-by-language basis to capture island constraints, as we have done
in our definitions of the extraction paths for English and Tagalog, is challenged
by Falk (2009). Falk observes that such an approach assumes that constraints
on long-distance dependencies are “essentially arbitrary and can display infinite
variation.” He proposes a different approach to the analysis of island constraints
crosslinguistically, appealing to a set of features marking extraction paths, and
defining islands in terms of off-path constraints. His proposal is designed to
capture the extra-syntactic as well as the syntactic dimensions of island constraints
in a more uniform way.
Work on constructions in other languages similar to the English tough construc-
tion includes Huang (1997) on Chinese, and Saiki (1991) and Yamamoto (1996)
on Japanese.
Belyaev and Haug (2014) present a semantic analysis of correlative constructions
in Ossetic, formulated within a PCDRT (Partial Compositional Discourse Repre-
sentation Theory) approach to semantics. Butt et al. (2007) draw parallels between
free relatives in English and German and Urdu correlatives, suggesting that the two
constructions could be open to the same analysis. In earlier work, Butt et al. (1999)
provide a different analysis of free relatives in English and German.
Gazdik (2011) analyzes multiple questions in French and Hungarian using a
level of structure that has received relatively little attention in LFG: discourse
structure.
Alsina (2008) discusses multiple-gap constructions within the LFG framework,
offering a different approach to the one described in §17.5. Falk (2011) considers
the interaction between multiple-gap constructions and various types of islands,
concluding that the former do not represent a uniform phenomenon.
Chisarik (2002) shows that split partitive noun phrases in Hungarian display
properties of long-distance dependencies, and provides an analysis.
Besides the syntactic and semantic analysis of resumptive pronouns described in
this chapter, Asudeh (2012) presents a processing model which defines the concept
of local wellformedness in LFG-specific terms, and discusses how this captures the
facts about variation in resumption.
 18
Related research threads
and new directions

This book has presented the theory of Lexical Functional Grammar. Part I pre-
sented the two syntactic structures of LFG, constituent structure and functional
structure, and showed how they are represented, described, and constrained.
Part II discussed the representation of semantic structure, information structure,
prosodic structure, and the morphological module within the modular architec-
ture of LFG, and showed how these different aspects of linguistic structure are
related to one another. Part III provided syntactic and semantic analyses of a variety
of linguistic phenomena. Much work remains to be done on the constructions we
examined, as well as on topics that have unfortunately gone unexamined in this
book. In this final chapter, we turn to a brief discussion of some LFG work that has
not been covered in earlier chapters.

18.1 Psychological reality: processing and acquisition

One of the original aims of LFG was to produce a psychologically realistic linguistic
theory, one that would not only account for observed patterns of linguistic behavior
but would also provide insight into the mental representation of language. Bresnan
and Kaplan (1982) enumerate a list of constraints on the psycholinguistic problem
of linguistic parsing, or how a speaker determines the structure of a string of words.
Creativity: The theory must account for the fact that the speaker can under-
stand and produce entirely novel sentences.
Finite Capacity: The theory must be capable of producing an unbounded
number of possible sentences from a finite number of words and rules.
Reliability: The theory must provide a nonarbitrary, reliably computable, and
consistent method for deciding on the structure of a sentence.
Order-Free Composition: The theory must explain how a speaker can pro-
duce coherent analyses for arbitrary or incomplete fragments of sentences.
Universality: The theory must incorporate a universal, effective means for
determining the grammar of the language that the speaker encounters.

The Oxford Reference Guide to Lexical Functional Grammar. First edition. Mary Dalrymple, John J. Lowe, and Louise
Mycock. © Mary Dalrymple, John J. Lowe, and Louise Mycock 2019. First published 2019 by Oxford University Press.
 psychological reality: processing and acquisition 727

The theory of LFG meets these desiderata. Work on psycholinguistic processing in

LFG was pioneered by Ford et al. (1982), and Pinker (1982, 1989) studied issues of
language acquisition in an LFG setting.

18.1.1 Data-Oriented Parsing

The framework of Data-Oriented Parsing or DOP (Bod 1998; Bod et al. 2003)
presents an innovative view of language processing and acquisition. DOP views
language acquisition as the analysis of a pool of linguistic structures that are
presented to the language learner. The learner breaks up these structures into
their component pieces, and new utterances are assembled from these pieces. The
likelihood of assigning a particular analysis to a new sentence depends on the
frequency of occurrence of its component parts in the original pool of structures.
LFG-DOP (Bod and Kaplan 1998; Cormons 1999) specializes the general DOP
theory to LFG assumptions about linguistic structures and the relations between
them. LFG-DOP assumes that the body of linguistic evidence that a language
learner is presented with consists of wellformed c-structure/f-structure pairs, and
that language acquisition consists in determining the relevant component parts
of these structures and then combining these parts to produce new c-structure/
f-structure pairs for novel sentences. Bod and Kaplan (2003) provide an overview
and describe a parser for LFG-DOP linguistic analysis; Finn et al. (2006) and Finn
(2007) describe a variant approach, GF-DOP (Grammatical Feature DOP). The
theory is also applied in work on translation by Way (1999, 2001, 2003), Hearne
(2005), and Hearne and Way (2006).

18.1.2 Second language acquisition: Processability Theory

In combination with Levelt’s model of language production (Levelt 1989; Bock
and Levelt 1994; Levelt et al. 1999), LFG has been adopted as the core theo-
retical basis of Processability Theory (PT), a psycholinguistic theory and model
of second language acquisition (SLA) first proposed by Pienemann (1998) and
subsequently developed in works including Pienemann et al. (2005) and Bettoni
and Di Biase (2015b). PT accounts for the acquisition of an L2 grammar in terms
of developmental stages relating to universal processing procedures that form an
implicational hierarchy, the processability hierarchy. The PT approach to SLA is
thus a general one: it applies no matter what the speaker’s first language (L1)
and second language (L2) might be. LFG was selected to represent linguistic
knowledge in PT because it is “a theory of grammar which is typologically and
psychologically plausible” (Pienemann 2011: 37–8). Furthermore, LFG is, like
Levelt’s model, a lexicalist theory. For Pienemann, this commonality is crucial
given that the acquisition process is most plausibly characterized as being lexically
driven (Pienemann 2005b: 18).
The PT literature covers a wide range of issues in SLA based on data from
a number of different L1s and L2s, and combinations thereof, including Arabic
(Mansouri 2005), Chinese (Zhang 2005), French (Ågren 2009), German (Jansen
and Di Biase 2015), Italian (Di Biase and Bettoni 2015), Japanese (Kawaguchi
728 related research threads and new directions

2015), Russian (Artoni and Magnani 2015), and Swedish (Glahn et al. 2001).
Pienemann (2005a) and Bettoni and Di Biase (2015a) are significant collections
of PT work. Pienemann and Keßler (2012) provides an overview of the theory;
Pienemann and Keßler (2011) is an introductory textbook on PT.

18.1.3 LFG and Optimality Theory

Much research in phonology, morphology, syntax, and semantics has been con-
ducted in the framework of Optimality Theory or OT. OT-based analyses assume
that the grammar of a language consists of a generator component which proposes
candidate linguistic structures for an input, and an evaluation component which
selects the optimal structure from these candidates by reference to a ranked set of
universally valid constraints on linguistic structures.
In OT-LFG, the input is taken to be an underspecified f-structure, and the
generator component is a “universal” LFG grammar that generates all wellformed
c-structure/f-structure pairs that are compatible with the input. The evaluation
component determines the optimal candidate in a particular language from this
set. Bresnan (2001b) was one of the first to combine LFG’s grammatical represen-
tations with an OT-style evaluation component in an analysis of morphosyntactic
blocking; Bresnan (2000) and Sells (2006a) give a useful overview of the theory.
Besides the papers incorporating OT-LFG analyses which have already been
mentioned in earlier chapters, and the work on diachronic syntax from an OT-LFG
perspective to be described in §18.2, work by Broadwell (2000b, 2002, 2003, 2006),
Morimoto (2000, 2001), Sells (2001b), Lee (2003, 2006), Mohanan and Mohanan
(2003), Belyaev (2013), Lowe and Belyaev (2015), Lowe (2016a), and the papers
collected in Sells (2001a) are of particular note.
Important work on the formal properties of OT-LFG has also been done.
Johnson (2002) discusses formal and algorithmic issues in OT-LFG parsing.
Responding to issues raised by Johnson, Kuhn (2001a,b) proves that parsing in OT-
LFG is decidable under certain reasonable assumptions, and resolves a number of
other difficult formal issues with the theory. A clear understanding of the formal
properties of generation is crucial in analyzing the formal properties of OT-LFG;
the results outlined in §18.3.2 are particularly important.

18.2 Diachrony and change

The grammatical theory that we have presented in this book is, like other formal
grammatical theories, fundamentally oriented toward synchronic grammatical
analysis: it provides a framework within which a specific synchronic grammar
can be modeled and analyzed, without consideration of its history or subsequent
development. From a diachronic perspective, however, synchronic grammars are
not isolated: the output of the grammar of one generation serves as the input
to the development of the grammar of the next generation so that, at least in
an approximative sense, a particular synchronic grammar can be said to have
developed out of another synchronic grammar, the origins of which lie slightly
 diachrony and change 729

further back in time, and which itself developed out of yet another, chronologically
prior, synchronic grammar.1
The relations between different but diachronically related synchronic grammars
is the realm of historical or diachronic linguistics. Diachronic changes in grammar
are well studied from a descriptive perspective, but they raise certain questions
for formal grammatical theories, which are primarily oriented towards synchronic
analysis. The most basic question is the extent to which it is even possible to
provide a descriptively adequate account of diachronic syntactic developments
in a synchronically oriented formalism like LFG. Although the majority of work
undertaken in LFG is exclusively synchronic in its aims, a number of authors have
explored diachronic linguistic developments from an LFG perspective, and have
shown that LFG is indeed adequate for representing diachronic developments.
A set of important early papers are collected in Butt and King (2001b). The loss
of case marking and its effect on complementation and grammatical relations
in Germanic and Romance languages has been explored by Allen (1995, 2008)
and Vincent and Börjars (2010a); the development of verbal periphrasis, and
the grammaticalization of lexical verbs into auxiliaries, is discussed by Allen
(2001), Schwarze (2001b), Butt and Lahiri (2013), Lowe (2015c), and Börjars
and Vincent (2017); the shift from accusative to split-ergative system in Indo-
Aryan is discussed by Butt (2001a); Nikitina (2008) discusses the development of
deverbal nominalizations into “mixed,” i.e. partially verbal, nominalizations, and
into infinitives; Börjars and Vincent (2017) discuss the development of pronouns
into copulas, an instance of “lateral grammaticalization.”
A more significant question is whether or to what extent LFG can provide an
explanatory model of syntactic change. For example, in grammatical theories that
rely on the concept of “movement” of words and constituents, the (supposed)
synchronic process of movement from lexical to functional head is often invoked
as an explanation for the well-attested diachronic development of lexical items into
grammatical items. As discussed by Börjars (2013) and Börjars and Vincent (2017),
there is no equivalent process within the LFG framework which can be so obviously
invoked to explain grammatical change. However, some works have advocated the
value of LFG over other grammatical theories in advancing our understanding of
diachronic grammatical change, in particular Vincent (1999, 2001b), Vincent and
Börjars (2010b), and Börjars and Vincent (2017). For example, Vincent (1999),
Börjars et al. (2016), and Börjars and Vincent (2017) show that there is diachronic
evidence for the gradual evolution of c-structure categories: a category such as P or
D may develop in a language, but there may not be evidence for PP or DP; phrasal
structure such as complement positions and specifier positions may develop at
a later period. Such developments can be modeled in LFG, where c-structural
representation is relatively free and nonprojecting categories are admitted, but it
challenges theories with more fixed approaches to phrase structure, where if a
category P exists in a language then a full PP, including specifier and complement

1 For a detailed discussion of the nature of diachronic relations between languages at different stages,
as well as the relations between the grammars of those languages, see Hale (2007).
730 related research threads and new directions

positions, necessarily also exists. In a similar vein, Lowe (2015b) argues that the
theory of Lexical Sharing in LFG provides a means of modeling the gradual process
of development between morpheme and clitic, or vice versa, which is otherwise
problematic for theories of syntax that assume an absolute distinction between
word and morpheme.
A number of authors have shown that the loss of a pred feature is a key
characteristic of grammaticalization processes. Vincent (1999) shows that once a
pred feature has been lost, constraints on the synchronic system explain associated
diachronic developments. That is, synchronic constraints on linguistic systems
modeled in LFG by reference to the pred feature provide an explanatory model of
certain diachronic changes. The loss of pred features in grammaticalization is also
discussed by Bresnan and Mchombo (1987) and Börjars et al. (1997). Coppock
and Wechsler (2010) show that the loss of person and number features is also
implicated in such processes.
On some level, of course, the idea of a fixed and invariant synchronic grammar
is a convenient fiction. To quote from Ohala’s (1989) work on sound change,
grammatical change is (at least in part) “drawn from a pool of synchronic vari-
ation.” Some work in LFG has sought to model synchronic variation, which may
serve as the basis of diachronic change. Bresnan et al. (2007b) make use of OT-
LFG to analyze individual speaker variation and dialectal variation with respect
to auxiliary-negative contractions in English. Vincent (1999, 2001a) and Vincent
and Börjars (2010b) also investigate the value of combining OT with LFG to model
synchronic variation as well as variation over time.

18.3 Computational issues: parsing, generation, and implementation

In an LFG computational linguistic setting, parsing traditionally means providing
all possible c-structure/f-structure pairs for a given string of words, and generation
means finding the strings of words that correspond to a given f-structure. As
discussed in Part II, much work in LFG explores interrelated facets of linguistic
structure, termed projections, including semantic structure, information structure,
and prosodic structure. Under these assumptions, the result of parsing a string
is the set of syntactic and nonsyntactic structures for a given input string and
the correspondence functions that relate them. In generation, the input need
not be a syntactic representation like the f-structure; semantic input can also be
analyzed, and in this case generation involves determining the syntactic structures
corresponding to the semantic input as well as the string that expresses that
meaning.

18.3.1 Parsing
Much important work has been done on the theory of parsing with LFG grammars
as well as on efficient parsing algorithms. A summary of work up to the mid-1990s
is given by Dalrymple et al. (1995c). Significant breakthroughs have been made on
several fronts.
 computational issues 731

Maxwell and Kaplan (1991) examine the problem of processing disjunctive

specifications of constraints making up an f-description. Although solving nondis-
junctive f-descriptions can be very efficient, processing disjunctive f-descriptions
can be exponentially difficult when all of the disjunctions are multiplied out against
each other. In other words, the time needed to process all of the disjunctive
possibilities can increase exponentially as the number of disjunctions increases.2
However, this worst-case scenario assumes that every disjunctive constraint can
interact significantly with every other constraint. In linguistic processing, such
interactions are found only very rarely; in fact, interactions of disjunctive con-
straints are almost always locally confined. For example, an ambiguity in the
syntactic properties of the subj of a sentence rarely correlates with ambiguities
in the obj or other arguments. This insight is the basis of Maxwell and Kaplan’s
algorithm, which works by turning a set of disjunctively specified constraints
into a set of contexted, conjunctively specified constraints, where the context of a
constraint indicates how the constraint fits into the configuration of disjunctions.
Solving these contexted constraints turns out to be very efficient for linguistically
motivated sets of constraints, where only local interactions among disjunctions
tend to occur.
The second breakthrough was made by Maxwell and Kaplan (1993), who
explore the issue of c-structure processing and its relation to solving f-structural
constraints. As observed in §7.6, phrase structure trees provide a clear and per-
spicuous representation of dominance and precedence conditions and phrasal
structure and groupings, while functional structures are an appropriate repre-
sentation for grammatical functions and features. However, it has long been
known that constituent structure parsing—determining the phrase structure trees
for a given sentence—is very fast in comparison to solving the disjunctions of
equations that determine the f-structure for the sentence: parsing with a context-
free phrase structure grammar3 to produce a c-structure can be accomplished in
cubic time,4 in contrast to the potentially exponential problem of determining

2 An exponential increase can be thought of in terms of some constant number raised to the nth
power, where n is the number of constraints. To give a rough idea of the rate of increase in complexity
as the number of constraints grows, if solving one constraint takes 21 = 2 seconds, solving three
constraints could take 23 = 8 seconds, and solving fifteen constraints could take 215 = 32,768 seconds
or about 9 hours.
3 The formal properties of context-free languages and their grammars, context-free grammars,
are described in Partee et al. (1993: Chapter 16). The added expressivity gained by allowing regular
expressions on the right-hand side of phrase structure rules, including the Kleene star, does not take us
beyond the power of context-free languages, as demonstrated by Woods (1970).
4 Parsing in cubic time is a particular instance of a problem that can be solved in polynomial time. If
a problem can be solved in polynomial time, the time it takes to solve a problem of size n is n raised to
some constant power; in the case of cubic time parsing, that constant is 3. Thus, if parsing complexity
depends on the length of the sentence we are parsing and a sentence of one word can be parsed in 13 = 1
second, a sentence consisting of three words would be parsed in 33 = 27 seconds, and a sentence with
fifteen words would be parsed in 153 = 3,375 seconds, or about 56 minutes. Of course, actual parsing
times are much faster than this for both short and long sentences: these figures are merely intended to
illustrate the rate of growth of a polynomial problem as opposed to an exponential problem. Parsing
in cubic time is much quicker and more efficient than the exponential growth associated with solving
arbitrary sets of functional constraints.
732 related research threads and new directions

the corresponding f-structures. For this reason, an important task in designing

algorithms for linguistic processing of structures of formally different kinds, like
the c-structure and the f-structure, is to optimize the interactions between these
computationally very different tasks. Previous research often assumed that the
most efficient approach would be to interleave the construction of the phrase
structure tree with the solution of f-structure constraints. Maxwell and Kaplan
(1993) explore and compare a number of different methods for combining phrase
structure processing with constraint solving; they show that in certain situations,
interleaving the two processes can actually give very bad results.
The third breakthrough built on these findings: Maxwell and Kaplan (1996)
showed that if phrase structure parsing and f-structural constraint solving are
combined in the right way, parsing can be very fast in many cases. Although
the worst case scenario involving complex interactions among constraints is still
exponential, the linguistically more common situation is for interactions among
constraints to be limited. In such cases, if the grammar that results from combining
phrase structure and functional constraints happens to be context-free equivalent,
the algorithm for computing the c-structure and f-structure operates in cubic time,
the same as for pure phrase structure parsing.
In ongoing unpublished work, Ron Kaplan and Jürgen Wedekind explore the
question of whether the linguistically-relevant subset of the LFG formalism—the
constructs that seem to be necessary and exploited in real grammars—may be
translatable into a weaker formal system, multiple context-free grammars (Pollard
1984; Seki et al. 1991). The interest of this possibility lies in the fact that grammars
in this class can be parsed in polynomial time, guaranteeing the best-case polyno-
mial result of Maxwell and Kaplan (1996) for all grammars. Further research will
show whether these interesting speculations will bear fruit.

18.3.2 Generation
Work on generation in LFG generally assumes that the generation task is to
determine the strings of a language that correspond to a specified f-structure, given
a particular grammar. Based on these assumptions, several interesting theoretical
results have been attained; Wedekind (2006) provides a useful overview, and
Butt et al. (2006: Part I) includes several papers describing LFG-based work on
generation and translation.
Wedekind (1995, 1999, 2014) addresses the issue of the decidability of genera-
tion from f-structures: the problem of determining whether there is any sentence
that corresponds to a given f-structure according to a given grammar. Wedekind
(1995) demonstrates that the problem is decidable for fully specified, acyclic5
f-structures: if we assume that the f-structure we are given is complete, has
no cycles, and no additional features can be added, we can always determine
whether or not there is a sentence that corresponds to that f-structure. Wedekind

5 An acyclic f-structure is one in which no f-structure contains a path leading back to itself. Kaplan
and Bresnan (1982) suggest that acyclic structures are the only f-structures that are motivated for
linguistic analysis.
 computational issues 733

(1999) shows that the corresponding problem for underspecified f-structures is not
decidable: it is not always possible to determine whether there is a sentence that
corresponds to a given f-structure if we are allowed to add additional attributes and
values to the f-structure. Wedekind (2014) demonstrates that the corresponding
problem for cyclic f-structures is not decidable, even when the f-structure is fully
specified: it is not always possible to determine whether there is a sentence that
corresponds to a given f-structure if it contains one or more cycles.
In further work on the formal properties of generation from f-structures, Kaplan
and Wedekind (2000) show that if we are given an LFG grammar and an acyclic
f-structure, the set of strings that corresponds to that f-structure according to
the grammar is a context-free language. Building on and extending that work,
Wedekind and Kaplan (2012) provide a method for constructing the context-free
grammar for that set of strings by a process of specialization of the full grammar
that we are given. This result leads to a new way of thinking about generation,
opens the way to new and more efficient generation algorithms, and clarifies a
number of formal and mathematical issues relating to LFG parsing and generation.
Wedekind and Kaplan (1996) explore issues in ambiguity-preserving generation,
where a set of f-structures rather than a single f-structure is considered, and the
sentences of interest are those that correspond to all of the f-structures under
consideration; Shemtov (1997) also explores issues in ambiguity management
and ambiguity preservation in generation from sets of f-structures. The potential
practical advantages of ambiguity-preserving generation are clear: consider, for
example, a scenario involving translation from English to German. We first parse
the input English sentence, producing several f-structures if the English sentence is
ambiguous. For instance, the English sentence Hans saw the man with the telescope
is ambiguous: it means either that the man had the telescope or that Hans used the
telescope to see the man. The best translation for this sentence would be a German
sentence that is ambiguous in exactly the same way as the English sentence, if such
a German sentence exists; in the case at hand, we would like to produce the German
sentence Hans sah den Mann mit dem Fernrohr, which has exactly the same two
meanings as the English input. To do this, we map the English f-structures for the
input sentence to the set of corresponding German f-structures; our goal is then
to generate the German sentence Hans sah den Mann mit dem Fernrohr, which
expresses each of these f-structures. Though this approach is appealing, Wedekind
and Kaplan (1996) show that determining whether there is a single sentence that
corresponds to each member of a set of f-structures is in general undecidable for
an arbitrary (possibly linguistically unreasonable) LFG grammar: there are LFG
grammars and sets of f-structures for which it is impossible to determine whether
there is any sentence that corresponds to those f-structures. This result is important
in understanding the limits of ambiguity-preserving generation.

18.3.3 LFG-based grammar development platforms

The important algorithmic results on parsing and generation described above have
enabled the development of computational implementations of LFG grammars
and grammar development platforms supporting analysis and generation with
734 related research threads and new directions

LFG grammars. One of the earliest LFG implementations was the Grammar
Writer’s Workbench (Kaplan and Maxwell 1996), originally implemented in the
early 1980s in the Xerox Interlisp-D environment, and under development through
the mid-1990s at the Xerox Palo Alto Research Center (PARC). In 1993, the PARC
team undertook a new implementation in C; this became the Xerox Linguistic
Environment (XLE: Crouch et al. 2008). The XLE underpins a number of large-
scale grammar implementations within the PARGRAM project (§18.3.4). More
recently, Lionel Clément and his colleagues have developed xLFG, a web-based
LFG parsing platform (Clément and Kinyon 2001; Clément 2016), and Damir
Cavar is leading a team to produce a freely available grammar development plat-
form for LFG and related grammar formalisms, the Free Linguistic Environment
(https://gorilla.linguistlist.org/fle).
The availability of computational implementations of LFG, and in particular the
XLE, has enabled research on combining LFG grammars with other computational
tools to increase parsing efficiency, improve parsing results, or produce more useful
language-based applications. Kaplan and King (2003) experiment with tools which
augment the input to the parsing process with various kinds of additional informa-
tion: using a finite-state parser to bracket the input string to indicate the presence of
phrasal boundaries, using a tagger to add part-of-speech annotations to the words
in the input, and using a gazetteer or onomasticon to mark word sequences as the
name of a person, place, or company. For example, the unannotated string in (1a)
might be annotated as in (1b) for input to XLE processing:

(1) a. New York City is a nice place to live.

b. <city>New York City</city> is [NP a niceA placeN to liveV ].

Such information is useful in reducing ambiguity and increasing efficiency when

it is completely accurate, but this is very dependent on the tools which are used
to add this information; such tools are often prone to error. Kaplan and King
found that adding information about known sequences such as New York City and
information about phrasal boundaries tended to be helpful, but that the addition
of part-of-speech information tended not to be very useful. Of course, these results
are entirely dependent on the particular tools that provide these annotations and
how compatible they are with the LFG grammar that is used. Kaplan et al. (2004)
provide further discussion of such annotations, and also provide an overview of the
use of finite state transducers for morphological analysis in parsing and generation
(Kaplan and Kay 1994; Beesley and Karttunen 2003) and their incorporation into
the XLE environment. In other work, Riezler et al. (2002) show that a stochastic
disambiguation model can be combined with a broad-coverage LFG grammar to
improve accuracy in choosing among different parses for the same sentence.
Implementations have also been developed which add a semantic component
to LFG systems or permit compositional semantic analyses in an LFG setting. The
XLE system includes a transfer component, XFR (Crouch 2005), which allows f-
structures to be rewritten into other representations by means of a set of transfer
rules. The XFR component was originally intended for use as part of a machine
 computational issues 735

translation system, but has also been used to rewrite f-structures into semantic
structures, as described by Crouch and King (2006). Theorem provers for glue
semantics have been developed by Giorgolo (2017), including a standalone prover
as well as the “glue-xle” system, which can be used together with XLE.

18.3.4 The PARGRAM Project

The Parallel Grammar Project (PARGRAM: Butt et al. 1999; Butt et al. 2002b)
was founded in 1994 as a research consortium to write large-scale, parallel LFG
grammars using the XLE grammar development platform. The grammars are
“parallel” in the sense that they are developed under a common set of grammatical
assumptions, using a commonly agreed-on set of grammatical features; differences
among the grammars are due to differences among the grammars of the languages
of the project, not different theoretical stances or arbitrary analytical decisions on
the part of the grammar writers. In 1994, the project included only grammars
of English (developed at PARC), French (developed at the Rank Xerox Research
Centre in Grenoble), and German (developed at the Institute for Natural Language
Processing, University of Stuttgart), but the project later grew to include, at various
times, grammars of Norwegian, Japanese, Urdu, Turkish, Hungarian, Georgian,
Tigrinya, Wolof, Indonesian, Welsh, Malagasy, Chinese, Arabic, Vietnamese, Pol-
ish, and Northern Soto. Besides an in-depth understanding of these languages and
their grammars, the project has produced practical resources that have been used
in various computational applications, including automatic machine translation,
summarization, and question answering. The English grammar formed the core
of the natural language search engine built by Powerset, a company started in 2006
which was acquired by Microsoft in 2008, and has influenced the development of
the Bing search engine.
 References
The web addresses in this bibliography were valid as of December 2018. Since the
web changes rapidly, however, some of these addresses may not remain perma-
nently valid.
Abney, Stephen P. (1987). The English Noun Phrase in its Sentential Aspect. Ph.D.
thesis, Massachusetts Institute of Technology.
Ackerman, Farrell (1987). Miscreant Morphemes: Phrasal Predicates in Ugric. Ph.D.
thesis, University of California, Berkeley.
Ackerman, Farrell (1990). ‘Locative Alternation vs. Locative Inversion.’ In Aaron
Halpern (ed.), WCCFL 9: Proceedings of the Ninth West Coast Conference on
Formal Linguistics. Stanford: CSLI Publications, 1–13.
Ackerman, Farrell (1992). ‘Complex Predicates and Morpholexical Relatedness:
Locative Alternation in Hungarian.’ In Sag and Szabolcsi (1992), 55–83.
Ackerman, Farrell and Philip LeSourd (1997). ‘Toward a Lexical Representation of
Phrasal Predicates.’ In Alex Alsina, Joan Bresnan, and Peter Sells (eds.), Complex
Predicates, Stanford: CSLI Publications, 67–106.
Ackerman, Farrell and John Moore (2001). Proto-Properties and Grammatical
Encoding: A Correspondence Theory of Argument Selection. Stanford: CSLI
Publications.
Ackerman, Farrell and John Moore (2011). ‘The Object⃰ Parameter and the Func-
tional Expression Continuum: Evidence from Moro.’ Unpublished manuscript,
University of California, San Diego.
Ackerman, Farrell and John Moore (2013). ‘Proto-Properties in a Comprehensive
Theory of Argument Realization.’ In King and de Paiva (2013), 9–20.
Ackerman, Farrell and Gregory T. Stump (2004). ‘Paradigms and Periphrastic
Expression: A Study in Realization-Based Lexicalism.’ In Sadler and Spencer
(2004), 111–57.
Ackerman, Farrell and Gert Webelhuth (1996). ‘The Construct PREDICATE:
Empirical Arguments and Theoretical Status.’ In Butt and King (1996a). URL
http://csli-publications.stanford.edu/LFG/LFG1-1996/.
Ackerman, Farrell and Gert Webelhuth (1998). A Theory of Predicates. Stanford:
CSLI Publications.
Acquaviva, Paolo (2008). Lexical Plurals: A Morphosemantic Approach. Oxford:
Oxford University Press.
Adger, David (2003). Core Syntax: A Minimalist Approach. Oxford: Oxford Uni-
versity Press.
Ågren, Malin (2009). ‘Morphological Development in Written L2 French: A
Processability Perspective.’ In Jörg-U. Keßler, and Dagmar Keatinge (eds.),
Research in Second Language Acquisition: Empirical Evidence Across Languages,
Newcastle, UK: Cambridge Scholars Publishing, 121–151.
Ahmed, Tafseer and Miriam Butt (2011). ‘Discovering Semantic Classes for Urdu
N-V Complex Predicates.’ In Proceedings of the Ninth International Conference
738 references

on Computational Semantics (IWCS 2011), Stroudsburg, PA: Association for

Computational Linguistics, 305–309.
Ahmed, Tafseer, Miriam Butt, Annette Hautli, and Sebastian Sulger (2012). ‘A
Reference Dependency Bank for Analyzing Complex Predicates.’ In Nicoletta
Calzolari, Khalid Choukri, Thierry Declerck, Mehmet Uğur Doğan, Bente
Maegaard, Joseph Mariani, Jan Odijk, and Stelios Piperidis (eds.), Proceedings
of the Eighth International Conference on Language Resources and Evaluation
(LREC’12). European Language Resources Association (ELRA).
Aissen, Judith (1992). ‘Topic and Focus in Mayan.’ Language 68(1): 43–80.
Al Sharifi, Budour and Louisa Sadler (2009). ‘The Adjectival Construct in Arabic.’
In Butt and King (2009). URL http://csli-publications.stanford.edu/LFG/14/
lfg09.html.
Allbritton, David W., Gail McKoon, and Roger Ratcliff (1996). ‘Reliability of
Prosodic Cues for Resolving Syntactic Ambiguity.’ Journal of Experimental
Psychology: Learning, Memory, and Cognition 22(3): 714–735.
Allen, Cynthia (1995). Case Marking and Reanalysis: Grammatical Relations From
Old to Early Modern English. Oxford: Clarendon Press.
Allen, Cynthia (2001). ‘The Development of a New Passive in English.’ In Butt and
King (2001b), 43–72.
Allen, Cynthia (2008). Genitives in Early English: Typology and Evidence. Oxford:
Oxford University Press.
Alshawi, Hiyan (ed.) (1992). The Core Language Engine. Cambridge, MA: The MIT
Press.
Alsina, Alex (1993). Predicate Composition: A Theory of Syntactic Function Alter-
nations. Ph.D. thesis, Stanford University.
Alsina, Alex (1995). ‘The Fall of Function-Argument Biuniqueness.’ In Glyn
Morrill, and Richard Oehrle (eds.), FG 1995: Proceedings of the First Conference
on Formal Grammar. Barcelona: ESSLLI, 1–16.
Alsina, Alex (1996). The Role of Argument Structure in Grammar: Evidence From
Romance. Stanford: CSLI Publications.
Alsina, Alex (1999). ‘Where’s the Mirror Principle?’ Linguistic Review 16(1):
1–42.
Alsina, Alex (2008). ‘A Theory of Structure-Sharing: Focusing on Long-Distance
Dependencies and Parasitic Gaps.’ In Butt and King (2008). URL http://csli-
publications.stanford.edu/LFG/13/lfg08.html.
Alsina, Alex (2010). ‘The Catalan Definite Article as Lexical Sharing.’ In Butt and
King (2010), 5–25. URL http://csli-publications.stanford.edu/LFG/15/lfg10.
html.
Alsina, Alex and Boban Arsenijević (2012a). ‘There Is No Third Face of Agree-
ment.’ Language 88(2): 388–389.
Alsina, Alex and Boban Arsenijević (2012b). ‘The Two Faces of Agreement.’
Language 88(2): 369–379.
Alsina, Alex and Smita Joshi (1991). ‘Parameters in Causative Constructions.’ In
Lise M. Dobrin, Lynn Nichols, and Rosa M. Rodriguez (eds.), Papers From
the 27th Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago
Linguistic Society, 31–46.
 references 739

Alsina, Alex and Sam A. Mchombo (1990). ‘The Syntax of Applicatives in

Chicheŵa: Problems for a Theta Theoretic Asymmetry.’ Natural Language and
Linguistic Theory 8(4): 493–506.
Alsina, Alex and Sam A. Mchombo (1993). ‘Object Asymmetries and the Chicheŵa
Applicative Construction.’ In Sam A. Mchombo (ed.), Theoretical Aspects of
Bantu Grammar, Stanford: CSLI Publications, number 38 in CSLI Lecture
Notes, 17–45.
Alsina, Alex, Tara Mohanan, and K. P. Mohanan (1996). ‘Untitled Submis-
sion to the LFG List.’ URL http://listserv.linguistlist.org/pipermail/lfg/1996-
September/000108.html. 3 September 1996.
Alsina, Alex, Tara Mohanan, and K. P. Mohanan (2005). ‘How to Get Rid of the
COMP.’ In Butt and King (2005). URL http://csli-publications.stanford.edu/
LFG/10/lfg05.html.
Alsina, Alex and Eugenio M. Vigo (2014). ‘Copular Inversion and Non-Subject
Agreement.’ In Butt and King (2014). URL http://csli-publications.stanford.
edu/LFG/19/lfg14.html.
Anderson, Stephen R. (1992). A-Morphous Morphology. Cambridge, UK: Cam-
bridge University Press.
Andréasson, Maia (2007). ‘The Architecture of I-Structure.’ In Butt and King
(2007), 26–43. URL http://csli-publications.stanford.edu/LFG/12/lfg07.html.
Andrews, Avery D., III (1982). ‘The Representation of Case in Modern Icelandic.’
In Bresnan (1982b), 427–503.
Andrews, Avery D., III (1983a). ‘Constituent Coordination in LFG.’ URL http://
www.essex.ac.uk/linguistics/external/lfg/www-lfg.stanford.edu/pubs/papers/
andrews/andrews-1983-0700.pdf. Unpublished manuscript, Australian
National University.
Andrews, Avery D., III (1983b). ‘A Note on the Constituent Structure of Modifiers.’
Linguistic Inquiry 14(4): 695–697.
Andrews, Avery D., III (1990a). ‘Case Structures and Control in Modern Icelandic.’
In Maling and Zaenen (1990), 187–234.
Andrews, Avery D., III (1990b). ‘Unification and Morphological Blocking.’ Natural
Language and Linguistic Theory 8(4): 507–557.
Andrews, Avery D., III (2005). ‘F-Structural Spellout in LFG Morphol-
ogy.’ URL http://members.iinet.net.au/~ada/AveryAndrews/Papers/fspell.pdf.
Unpublished manuscript, Australian National University.
Andrews, Avery D., III (2007a). ‘The Major Functions of the Noun Phrase.’ In Tim-
othy Shopen (ed.), Language Typology and Syntactic Description, Cambridge,
UK: Cambridge University Press, volume I: Clause structure, 62–154. Second
edition.
Andrews, Avery D., III (2007b). ‘Projections and Glue for Clause Union Com-
plex Predicates.’ In Butt and King (2007), 44–65. URL http://csli-publications.
stanford.edu/LFG/12/lfg07.html.
Andrews, Avery D., III (2008). ‘The Role of pred in LFG+Glue.’ In Butt and King
(2008), 46–67. URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Andrews, Avery D., III (2011). ‘Propositional Glue and the Projection Architecture
of LFG.’ Linguistics and Philosophy 33(3).
740 references

Andrews, Avery D., III (2018). ‘Sets, Heads and Spreading in LFG.’ Journal of Lan-
guage Modelling 6(1). URL http://jlm.ipipan.waw.pl/index.php/JLM/article/
view/175.
Andrews, Avery D., III and Christopher D. Manning (1993). ‘Information Spread-
ing and Levels of Representation in LFG.’ Technical Report CSLI-93-176, CSLI
Publications, Stanford.
Andrews, Avery D., III and Christopher D. Manning (1999). Complex Predicates
and Information Spreading in LFG. Stanford: CSLI Publications.
Aranovich, Raúl (2013). ‘Mismatched Spanish Unaccusativity Tests.’ In King and
de Paiva (2013), 33–45.
Arka, I Wayan (1998). From Morphosyntax to Pragmatics in Balinese. Ph.D. thesis,
University of Sydney.
Arka, I Wayan (2003). Balinese Morphosyntax: A Lexical-Functional Approach.
Canberra: Pacific Linguistics.
Arka, I Wayan (2005). ‘The Core-Oblique Distinction and Core Index in
Some Austronesian Languages of Indonesia.’ URL http://www.academia.edu/
6659164/Arka_I_Wayan._2005._The_core-oblique_distinction_and_core_
index_in_some_Austronesian_languages_of_Indonesia._Keynote_paper_
presented_at_International_ALT_VI_Association_of_Linguistic_Typology_
conference_Padang_Indonesia_July_2005. Presented at International
Association of Linguistic Typology conference, Padang, Indonesia, July 2005.
Arka, I Wayan (2011). ‘Constructive Number Systems in Marori and Beyond.’ In
Butt and King (2011). URL http://csli-publications.stanford.edu/LFG/16/lfg11.
html.
Arka, I Wayan (2012). ‘Verbal Number, Argument Number and Plural Events in
Marori.’ In Butt and King (2012). URL http://csli-publications.stanford.edu/
LFG/17/lfg12.html.
Arka, I Wayan (2013). ‘Nominal Aspect in Marori.’ In Butt and King (2013). URL
http://csli-publications.stanford.edu/LFG/18/lfg13.html.
Arka, I Wayan (2014). ‘Double and Backward Control in Indonesian: An LFG
Analysis.’ In Butt and King (2014). URL http://csli-publications.stanford.edu/
LFG/19/lfg14.html.
Arka, I Wayan, Mary Dalrymple, Meladel Mistica, Suriel Mofu, Avery D. Andrews,
III, and Jane Simpson (2009). ‘A Computational Morphosyntactic Analysis for
the Applicative -I in Indonesian.’ In Butt and King (2009). URL http://csli-
publications.stanford.edu/LFG/14/lfg09.html.
Arka, I Wayan and Jane Simpson (1998). ‘Control and Complex Arguments in
Balinese.’ In Butt and King (1998b). URL http://csli-publications.stanford.edu/
LFG/LFG3-1998/.
Arka, I Wayan and Jane Simpson (2008). ‘Objective Voice and Control Into Subject
Clauses in Balinese.’ In Austin and Musgrave (2008), 83–114.
Arka, I Wayan and Stephen Wechsler (1996). ‘Argument Structure and Lin-
ear Order in Balinese Binding.’ In Butt and King (1996a). URL http://csli-
publications.stanford.edu/LFG/LFG1-1996/.
Arnold, Doug, Miriam Butt, Berthold Crysmann, Tracy Holloway King, and Stefan
Müller (eds.) (2017). Proceedings of the Joint 2016 Conference on Head-driven
 references 741

Phrase Structure Grammar and Lexical Functional Grammar. Stanford: CSLI

Publications. URL http://cslipublications.stanford.edu/HPSG/2016/.
Arnold, Doug and Louisa Sadler (2010). ‘Pottsian LFG.’ In Butt and King (2010),
43–63. URL http://csli-publications.stanford.edu/LFG/15/lfg10.html.
Arnold, Doug and Louisa Sadler (2011). ‘Resource Splitting and Reintegration with
Supplementals.’ In Butt and King (2011), 26–46. URL http://csli-publications.
stanford.edu/LFG/16/lfg11.html.
Arnold, Doug and Louisa Sadler (2012). ‘Affected Experiencers and Mixed Seman-
tics in LFG/Glue.’ In Butt and King (2012), 44–63. URL http://csli-publications.
stanford.edu/LFG/17/lfg12.html.
Arnold, Doug and Louisa Sadler (2013). ‘Displaced Dependent Constructions.’ In
Butt and King (2013). URL http://csli-publications.stanford.edu/LFG/18/lfg13.
html.
Arnold, Doug, Louisa Sadler, and Aline Villavicencio (2007). ‘Portuguese: Cor-
pora, Coordination and Agreement.’ In Featherston and Sternefeld (2007),
9–28.
Arnold, Jennifer (1994). ‘Inverse Voice Marking in Mapudungun.’ In S. Gahl,
A. Dolbey, and C. Johnson (eds.), Proceedings of the 20th Annual Meeting of
the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, 28–41.
Artoni, Daniele and Marco Magnani (2015). ‘Acquiring Case Marking in Russian
as a Second Language: An Exploratory Study on Subject and Object.’ In Bettoni
and Di Biase (2015a), 177–193.
Arvaniti, Amalia (2016). ‘Analytical Decisions in Intonation Research and The
Role of Representations: Lessons from Romani.’ Laboratory Phonology: Journal
of the Association for Laboratory Phonology 1(7): 1–43.
Arvaniti, Amalia and D. Robert Ladd (2009). ‘Greek Wh-Questions and the
Phonology of Intonation.’ Phonology 1(26): 43–74.
Asher, Nicholas and Hajime Wada (1988). ‘A Computational Account of Syntactic,
Semantic and Discourse Principles for Anaphora Resolution.’ Journal of Seman-
tics 6(3/4): 309–344.
Asudeh, Ash (2000). ‘Functional Identity and Resource Sensitivity in Control.’ In
Butt and King (2000c). URL http://csli-publications.stanford.edu/LFG/LFG5-
2000/.
Asudeh, Ash (2001). ‘Linking, Optionality, and Ambiguity in Marathi: An Opti-
mality Theory Analysis.’ In Sells (2001a), 257–312.
Asudeh, Ash (2002a). ‘A Resource-Sensitive Semantics for Equi and Raising.’ In
David I. Beaver, Stefan Kaufmann, Brady Clark, and Luis Casillas (eds.), The
Construction of Meaning, Stanford: CSLI Publications, 1–22.
Asudeh, Ash (2002b). ‘Richard III.’ In Mary Andronis, Erin Debenport, Anne
Pycha, and Keiko Yoshimura (eds.), Papers From the 38th Regional Meeting of
the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 31–46.
Asudeh, Ash (2002c). ‘The Syntax of Preverbal Particles and Adjunction in Irish.’ In
Butt and King (2002). URL http://csli-publications.stanford.edu/LFG/7/lfg02.
html.
Asudeh, Ash (2004). Resumption as Resource Management. Ph.D. thesis, Stanford
University.
742 references

Asudeh, Ash (2005). ‘Control and Semantic Resource Sensitivity.’ Journal of Lin-
guistics 41(3): 465–511.
Asudeh, Ash (2006). ‘Direct Compositionality and the Architecture of LFG.’ In Butt
et al. (2006), 363–387.
Asudeh, Ash (2007). ‘Some Notes on Pseudo Noun Incorporation in Niuean.’ URL
http://users.ox.ac.uk/~cpgl0036/pdf/niuean.pdf. Unpublished manuscript,
Carleton University.
Asudeh, Ash (2009). ‘Adjacency and Locality: A Constraint-Based Analysis of
Complementizer-Adjacent Extraction.’ In Butt and King (2009). URL http://
csli-publications.stanford.edu/LFG/14/lfg09.html.
Asudeh, Ash (2011). ‘Towards a Unified Theory of Resumption.’ In Alain Rouveret
(ed.), Resumptive Pronouns at the Interfaces, Amsterdam: John Benjamins,
121–188.
Asudeh, Ash (2012). The Logic of Pronominal Resumption. Oxford: Oxford Univer-
sity Press.
Asudeh, Ash and Douglas Ball (2005). ‘Niuean Incorporated Nominals as
Non-Projecting Nouns.’ URL http://users.ox.ac.uk/~cpgl0036/handouts/lsa05-
niuean-ho.pdf. Presented at the 2005 Annual Meeting of the Linguistic Society
of America, Oakland, CA.
Asudeh, Ash and Richard Crouch (2002a). ‘Coordination and Parallelism in Glue
Semantics: Integrating Discourse Cohesion and the Element Constraint.’ In Butt
and King (2002). URL http://csli-publications.stanford.edu/LFG/7/lfg02.html.
Asudeh, Ash and Richard Crouch (2002b). ‘Derivational Parallelism and Ellipsis
Parallelism.’ In WCCFL 21: Proceedings of the 21st West Coast Conference on
Formal Linguistics. Medford, MA: Cascadilla Press.
Asudeh, Ash, Mary Dalrymple, and Ida Toivonen (2008). ‘Constructions With
Lexical Integrity: Templates as the Lexicon-Syntax Interface.’ In Butt and King
(2008). URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Asudeh, Ash, Mary Dalrymple, and Ida Toivonen (2013). ‘Constructions With
Lexical Integrity.’ Journal of Language Modelling 1(1): 77–130. URL http://jlm.
ipipan.waw.pl/index.php/JLM/article/view/56.
Asudeh, Ash and Gianluca Giorgolo (2012). ‘Flexible Composition for Optional
and Derived Arguments.’ In Butt and King (2012), 64–84. URL http://csli-
publications.stanford.edu/LFG/17/lfg12.html.
Asudeh, Ash, Gianluca Giorgolo, and Ida Toivonen (2014). ‘Meaning and Valency.’
In Butt and King (2014). URL http://csli-publications.stanford.edu/LFG/19/
lfg14.html.
Asudeh, Ash and Ida Toivonen (2006). ‘Expletives and the Syntax and Semantics of
Copy Raising.’ In Butt and King (2006b). URL http://csli-publications.stanford.
edu/LFG/11/lfg06.html.
Asudeh, Ash and Ida Toivonen (2007). ‘Copy Raising and Its Consequences for
Perception Reports.’ In Zaenen et al. (2007), 49–67.
Asudeh, Ash and Ida Toivonen (2012). ‘Copy Raising and Perception.’ Natural
Language and Linguistic Theory 30(2): 321–380.
Asudeh, Ash and Ida Toivonen (2015). ‘Lexical-Functional Grammar.’ In Heine
and Narrog (2015).
 references 743

Attia, Mohammed (2008). ‘A Unified Analysis of Copula Constructions in LFG.’ In

Butt and King (2008), 89–108. URL http://csli-publications.stanford.edu/LFG/
13/lfg08.html.
Austin, Peter (2001). ‘Lexical Functional Grammar.’ In N. J. Smelser, and P. Baltes
(eds.), International Encyclopedia of the Social and Behavioural Sciences,
Elsevier, 8748–8754.
Austin, Peter and Joan Bresnan (1996). ‘Non-Configurationality in Australian
Aboriginal Languages.’ Natural Language and Linguistic Theory 14(2): 215–268.
Austin, Peter and Simon Musgrave (eds.) (2008). Voice and Grammatical Relations
in Austronesian Languages. Stanford: CSLI Publications.
Avgustinova, Tania and Hans Uszkoreit (2003). ‘Reconsidering the Relations in
Constructions With Non-Verbal Predicates.’ In Peter Kosta, Joanna Blaszczak,
Jens Frasek, Ljudmila Geist, and Marzena Zygis (eds.), Investigations Into
Formal Slavic Linguistics. Contributions of the Fourth European Conference on
Formal Description of Slavic Languages (FDSL IV). Peter Lang, volume Part II,
483–498.
Baader, Franz, Hans-Jürgen Bürckert, Bernhard Nebel, Werner Nutt, and Gert
Smolka (1991). ‘On the Expressivity of Feature Logics With Negation, Func-
tional Uncertainty and Sort Equations.’ Technical Report RR-91-01, Deutsches
Forschungszentrum für Künstliche Intelligenz, Saarbrücken.
Bach, Emmon (1980). ‘In Defense of Passive.’ Linguistics and Philosophy 3:
297–341.
Bach, Emmon (1989). Informal Lectures on Formal Semantics. SUNY Series in
Linguistics. Albany: State University of New York Press.
Backofen, Rolf (1993). ‘On the Decidability of Functional Uncertainty.’ In Proceed-
ings of the 31st Annual Meeting of the ACL. Columbus, OH: Association for
Computational Linguistics, 201–208.
Baerman, Matthew (2007). ‘Morphological Typology of Deponency.’ In Matthew
Baerman, Greville G. Corbett, Dunstan Brown, and Andrew Hippisley (eds.),
Deponency and Morphological Mismatches, Oxford: The British Academy &
Oxford University Press, number 145 in Proceedings of the British Academy,
1–19.
Baerman, Matthew, Dunstan Brown, and Greville G. Corbett (2002). ‘Surrey
Syncretisms Database.’ URL http://dx.doi.org/10.15126/SMG.10/1.
Baerman, Matthew, Dunstan Brown, and Greville G. Corbett (2005). The Syntax-
Morphology Interface: A Study of Syncretism. Cambridge, UK: Cambridge
University Press.
Baker, Brett, Kate Horrack, Rachel Nordlinger, and Louisa Sadler (2010). ‘Putting
it All Together: Agreement, Incorporation, Coordination and External Posses-
sion in Wubuy (Australia).’ In Butt and King (2010), 64–84. URL http://csli-
publications.stanford.edu/LFG/15/lfg10.html.
Baker, Mark C. (1983). ‘Objects, Themes and Lexical Rules in Italian.’ In Levin et al.
(1983), 1–45. Reprinted in Butt and King (2006a, 11–58).
Ball, Douglas (2004). ‘Pseudo Noun Incorporation and Argument Structure in
Niuean.’ In Butt and King (2004b), 1–20. URL http://csli-publications.stanford.
edu/LFG/9/lfg04.html.
744 references

Baltin, Mark (1982). ‘A Landing Site Theory of Movement Rules.’ Linguistic Inquiry
13(1): 1–38.
Banfield, Ann (1973). ‘Narrative Style and the Grammar of Direct and Indirect
Speech.’ Foundations of Language 10: 1–39.
Barlow, Michael and Charles A. Ferguson (eds.) (1988). Agreement in Natural
Language: Approaches, Theories, Descriptions. Stanford: CSLI Publications.
Barron, Julia (1997). ‘LFG and the History of Raising Verbs.’ In Butt and King
(1997). URL http://csli-publications.stanford.edu/LFG/LFG2-1997/.
Barron, Julia (1999). Perception, Volition, and Reduced Clausal Complementation.
Ph.D. thesis, University of Manchester.
Barron, Julia (2001). ‘Perception and Raising Verbs: Synchronic and Diachronic
Relationships.’ In Butt and King (2001b), 73–103.
Barwise, Jon (1979). ‘On Branching Quantifiers in English.’ Journal of Philosophical
Logic 8: 47–80.
Barwise, Jon and Robin Cooper (1981). ‘Generalized Quantifiers and Natural
Language.’ Linguistics and Philosophy 4(2): 159–219.
Barwise, Jon and John Perry (1983). Situations and Attitudes. Cambridge, MA: The
MIT Press.
Bary, Corien and Dag T. T. Haug (2011). ‘Temporal Anaphora Across and Inside
Sentences: The Function of Participles.’ Semantics and Pragmatics 4, article 8:
1–56.
Bäuerle, Rainer, Christoph Schwarze, and Arnim von Stechow (eds.) (1983).
Meaning, Use, and Interpretation of Language. Berlin: Walter de Gruyter.
Bayer, Samuel (1996). ‘The Coordination of Unlike Categories.’ Language 72(3):
579–616.
Beard, Robert (1988). ‘On the Separation of Derivation from Morphology: Toward
a Lexeme-Morpheme Based Morphology.’ Quaderni di Semantica 9: 3–59.
Beard, Robert (1995). Lexeme-Morpheme Base Morphology: A General Theory of
Inflection and Word Formation. Stony Brook: SUNY Press.
Beaver, David I. (1992). ‘The Kinematics of Presupposition.’ In Paul Dekker,
and Martin Stokhof (eds.), Proceedings of the Eighth Amsterdam Colloquium.
Amsterdam: Institute for Logic, Language, and Information, University of
Amsterdam, 17–36.
Beckman, Mary E., Julia Hirschberg, and Stefanie Shattuck-Hufnagel (2005). ‘The
Original ToBI System and the Evolution of the ToBI Framework.’ In Jun (2005),
9–54.
Beesley, Kenneth R. and Lauri Karttunen (2003). Finite-State Morphology.
Stanford: CSLI Publications.
Bell, Sara (1983). ‘Advancements and Ascensions in Cebuano.’ In Perlmutter
(1983), 143–218.
Belyaev, Oleg I. (2013). ‘Optimal Agreement at M-Structure: Person in Dargwa.’ In
Butt and King (2013), 90–110. URL http://csli-publications.stanford.edu/LFG/
18/lfg13.html.
Belyaev, Oleg I. (2014). ‘Osetinskij kak jazyk s dvuhpadežnoj sistemoj: gruppovaja
fleksija i drugie paradoksy padežnogo markirovanija [Ossetic as a Two-Case
Language: Suspended Affixation and Other Case Marking Paradoxes].’ Voprosy
Jazykoznanija 6: 74–108.
 references 745

Belyaev, Oleg I. (2015). ‘Systematic Mismatches: Coordination and Subordination

at Three Levels of Grammar.’ Journal of Linguistics 51(2): 267–326.
Belyaev, Oleg I., Mary Dalrymple, and John J. Lowe (2015). ‘Number Mismatches
in Coordination: An LFG Analysis.’ In Butt and King (2015b). URL http://csli-
publications.stanford.edu/LFG/20/lfg15.html.
Belyaev, Oleg I. and Dag T. T. Haug (2014). ‘Pronominal Coreference in Ossetic
Correlatives and the Syntax-Semantics Interface.’ In Butt and King (2014),
89–109. URL http://csli-publications.stanford.edu/LFG/19/lfg14.html.
Bergmann, Merrie (1982). ‘Cross-Categorial Semantics for Conjoined Common
Nouns.’ Linguistics and Philosophy 5: 399–401.
Berman, Judith (1999). ‘Does German Satisfy the Subject Condition?’ In Butt and
King (1999). URL http://csli-publications.stanford.edu/LFG/LFG4-1999/.
Berman, Judith (2003). Clausal Syntax of German. Stanford: CSLI Publications.
Berman, Judith (2007). ‘Functional Identification of Complement Clauses in
German and the Specification of comp.’ In Zaenen et al. (2007), 69–83.
Bettoni, Camilla and Bruno Di Biase (eds.) (2015a). Grammatical Development
in Second Languages: Exploring the Boundaries of Processability Theory. Italy:
European Second Language Association.
Bettoni, Camilla and Bruno Di Biase (2015b). ‘Processability Theory: Theo-
retical Bases and Universal Schedules.’ In Bettoni and Di Biase (2015a),
19–79.
Bickel, Balthasar, Bernard Comrie, and Martin Haspelmath (2015). ‘Leipzig Gloss-
ing Rules: Conventions for Interlinear Morpheme-By-Morpheme Glosses.’ URL
http://www.eva.mpg.de/lingua/resources/glossing-rules.php.
Biggs, Bruce (1969). Let’s Learn Maori. Wellington: Reed Education.
Biswas, Alexander (2017). Inflectional Periphrasis in LFG. M.Phil. dissertation,
University of Oxford. URL https://www.academia.edu/33800486/Inflectional_
Periphrasis_in_LFG.
Blevins, James P. (2000). ‘Markedness and Agreement.’ Transactions of the Philo-
logical Society 98(2): 233–262.
Bloomfield, Leonard (1933). Language. London: Allen and Unwin.
Boas, Hans Christian and Ivan A. Sag (eds.) (2012). Sign-Based Construction
Grammar. Stanford: CSLI Publications.
Bock, Kathryn and Willem J. M. Levelt (1994). ‘Language Production: Grammati-
cal Encoding.’ In Morton Ann Gernsbacher (ed.), Handbook of Psycholinguistics,
New York: Academic Press, 945–984.
Bod, Rens (1998). Beyond Grammar: An Experience-Based Theory. Stanford: CSLI
Publications.
Bod, Rens and Ronald M. Kaplan (1998). ‘A Probabilistic Corpus-Driven Model
for Lexical-Functional Analysis.’ In COLING/ACL98: Joint Meeting of the 36th
Annual Meeting of the ACL and the 17th International Conference on Com-
putational Linguistics. Montréal: International Committee on Computational
Linguistics, 145–151.
Bod, Rens and Ronald M. Kaplan (2003). ‘A Data-Oriented Parsing Model for
Lexical-Functional Grammar.’ In Bod et al. (2003).
Bod, Rens, Remko Scha, and Khalil Sima’an (eds.) (2003). Data-Oriented Parsing.
Stanford, CA: CSLI Publications.
746 references

Bodomo, Adams B. (1996). ‘Complex Verbal Predicates: The Case of Serial Verbs
in Dagaare and Akan.’ In Butt and King (1996a). URL http://csli-publications.
stanford.edu/LFG/LFG1-1996/.
Bodomo, Adams B. (1997). Paths and Pathfinders: Exploring the Syntax and Seman-
tics of Complex Verbal Predicates in Dagaare and Other Languages. Ph.D. thesis,
Norwegian University of Science and Technology.
Bögel, Tina (2010). ‘Pashto (Endo-)Clitics in a Parallel Architecture.’ In Butt and
King (2010), 85–105. URL http://csli-publications.stanford.edu/LFG/15/lfg10.
html.
Bögel, Tina (2012). ‘The P-Diagram – A Syllable-Based Approach to P-Structure.’
In Butt and King (2012), 99–117. URL http://csli-publications.stanford.edu/
LFG/17/lfg12.html.
Bögel, Tina (2013). ‘A Prosodic Resolution of German Case Ambiguities.’ In Butt
and King (2013), 131–151. URL http://csli-publications.stanford.edu/LFG/18/
lfg13.html.
Bögel, Tina (2014). ‘Degema and the String Interface.’ In Butt and King (2014).
URL http://csli-publications.stanford.edu/LFG/19/lfg14.html.
Bögel, Tina (2015). The Syntax-Prosody Interface in Lexical Functional Grammar.
Ph.D. thesis, Universität Konstanz. URL http://ling.uni-konstanz.de/pages/
home/boegel/Dateien/Boegel_thesis2015.pdf.
Bögel, Tina and Miriam Butt (2013). ‘Possessive Clitics and Ezafe in Urdu.’ In Kersti
Börjars, David Denison, and Alan Scott (eds.), Morphosyntactic Categories and
the Expression of Possession, Amsterdam: John Benjamins, 291–322.
Bögel, Tina, Miriam Butt, Ronald M. Kaplan, Tracy Holloway King, and John T.
Maxwell, III (2009). ‘Prosodic Phonology in LFG: A New Proposal.’ In Butt and
King (2009). URL http://csli-publications.stanford.edu/LFG/14/lfg09.html.
Bögel, Tina, Miriam Butt, Ronald M. Kaplan, Tracy Holloway King, and John T.
Maxwell, III (2010). ‘Second Position and the Prosody-Syntax Interface.’ In
Butt and King (2010). URL http://csli-publications.stanford.edu/LFG/15/lfg10.
html.
Bögel, Tina, Miriam Butt, and Sebastian Sulger (2008). ‘Urdu Ezafe and the
Morphology-Syntax Interface.’ In Butt and King (2008). URL http://csli-
publications.stanford.edu/LFG/13/lfg08.html.
Bolinger, Dwight L. (1965). Forms of English. Cambridge, MA: Harvard University
Press.
Bonami, Olivier and Berthold Crysmann (2016). ‘The Role of Morphology in
Constraint-Based Lexicalist Grammars.’ In Andrew Hippisley, and Gregory T.
Stump (eds.), Cambridge Handbook of Morphology, Cambridge, UK: Cambridge
University Press.
Bond, Oliver (2016). ‘Negation Through Reduplication and Tone: Implications for
the Lexical Functional Grammar/Paradigm Function Morphology Interface.’
Journal of Linguistics 52(2): 277–310. doi:10.1017/S0022226715000134.
Borer, Hagit (2013). Structuring Sense, Volume III: Taking Form. Oxford: Oxford
University Press.
Börjars, Kersti (1998). ‘Clitics, Affixes, and Parallel Correspondences.’ In Butt and
King (1998b). URL http://csli-publications.stanford.edu/LFG/LFG3-1998/.
 references 747

Börjars, Kersti (2011). ‘Lexical Functional Grammar.’ In Patrick Colm Hogan

(ed.), The Cambridge Encyclopedia of the Language Sciences, Cambridge, UK:
Cambridge University Press.
Börjars, Kersti (2013). ‘Modelling Linguistic Change: Grammaticalization, Grad-
ualness and Persistence.’ Keynote lecture, 21st International Conference on
Historical Linguistics, University of Oslo, 6 August 2013.
Börjars, Kersti, Erika Chisarik, and John Payne (1999). ‘On the Justification
for Functional Categories in LFG.’ In Butt and King (1999). URL http://csli-
publications.stanford.edu/LFG/LFG4-1999/.
Börjars, Kersti, Elisabet Engdahl, and Maia Andréasson (2003). ‘Subject and
Object Positions in Swedish.’ In Butt and King (2003b), 43–58. URL http://csli-
publications.stanford.edu/LFG/8/lfg03.html.
Börjars, Kersti, Pauline Harries, and Nigel Vincent (2016). ‘Growing Syntax: The
Development of a DP in Northern Germanic.’ Language 92(1): e1–e37.
Börjars, Kersti and Nigel Vincent (2008). ‘Objects and OBJ.’ In Butt and King
(2008). URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Börjars, Kersti and Nigel Vincent (2017). ‘Lexical-Functional Grammar.’ In Adam
Ledgeway, and Ian Roberts (eds.), The Cambridge Handbook of Historical
Syntax, Cambridge, UK: Cambridge University Press.
Börjars, Kersti, Nigel Vincent, and Carol Chapman (1997). ‘Paradigms, Periphrases
and Pronominal Inflection: A Feature-Based Account.’ In Geert Booij, and Jaap
van Marle (eds.), Yearbook of Morphology, Dordrecht: Kluwer, 155–180.
Borsley, Robert D. (2005). ‘Against ConjP.’ Lingua 115(4): 461–482.
Borsley, Robert D. and Kersti Börjars (eds.) (2011). Non-Transformational Syntax:
Formal and Explicit Models of Grammar. Oxford: Blackwell.
Bos, Johan, Elsbeth Mastenbroek, Scott McGlashan, Sebastian Millies, and
Manfred Pinkal (1994). ‘A Compositional DRS-Based Formalism for NLP-
Applications.’ In Proceedings of the International Workshop On Computational
Semantics. Tilburg, 21–31. Also published as Verbmobil Report 59, Universität
des Saarlandes, Saarbrücken, Germany.
Bošković, Željko (2001). On the Nature of the Syntax-Phonology Interface: Cliticiza-
tion and Related Phenomena. Oxford: North-Holland.
Bouchard, Denis (1982). On the Content of Empty Categories. Ph.D. thesis, Mas-
sachusetts Institute of Technology.
Bouma, Gosse, Robert Malouf, and Ivan A. Sag (2001). ‘Satisfying Constraints
on Extraction and Adjunction.’ Natural Language and Linguistic Theory 19(1):
1–65.
Brame, Michael (1982). ‘The Head-Selector Theory of Lexical Specifications and
the Nonexistence of Coarse Categories.’ Linguistic Analysis 10(4): 321–325.
Bresnan, Joan (1976). ‘On the Form and Functioning of Transformations.’ Linguis-
tic Inquiry 7(1): 3–40.
Bresnan, Joan (1977). ‘Transformations and Categories in Syntax.’ In Robert E.
Butts, and Jaakko Hintikka (eds.), Basic Problems in Methodology and Lin-
guistics: Part Three of the Proceedings of the Fifth International Congress of
Logic, Methodology and Philosophy of Science, London, Ontario, Canada, 1975.
Dordrecht: D. Reidel, 261–282.
748 references

Bresnan, Joan (1978). ‘A Realistic Transformational Grammar.’ In Morris Halle,

Joan Bresnan, and George A. Miller (eds.), Linguistic Theory and Psychological
Reality, Cambridge, MA: The MIT Press, 1–59.
Bresnan, Joan (1980). ‘Polyadicity.’ In Teun Hoekstra, Harry van der Hulst, and
Michael Moortgat (eds.), Lexical Grammar, Dordrecht: Foris, 97–121.
Bresnan, Joan (1982a). ‘Control and Complementation.’ Linguistic Inquiry 13(3):
343–434. Reprinted in Bresnan (1982b, 282–390).
Bresnan, Joan (ed.) (1982b). The Mental Representation of Grammatical Relations.
Cambridge, MA: The MIT Press.
Bresnan, Joan (1982c). ‘The Passive in Lexical Theory.’ In Bresnan (1982b), 3–86.
Bresnan, Joan (1982d). ‘Polyadicity.’ In Bresnan (1982b), 149–172. Originally
published as Bresnan (1980).
Bresnan, Joan (1984). ‘Bound Anaphora on Functional Structures.’ Presented at the
Tenth Annual Meeting of the Berkeley Linguistics Society, February 17, 1984.
Bresnan, Joan (1990). ‘Monotonicity and the Theory of Relation Changes in LFG.’
Language Research 26(4): 637–652.
Bresnan, Joan (1994). ‘Locative Inversion and the Architecture of Universal Gram-
mar.’ Language 70(1): 72–131.
Bresnan, Joan (1995). ‘Linear Order, Syntactic Rank, and Empty Categories: On
Weak Crossover.’ In Dalrymple et al. (1995b), 241–274.
Bresnan, Joan (1997). ‘Mixed Categories as Head Sharing Constructions.’ In Butt
and King (1997). URL http://csli-publications.stanford.edu/LFG/LFG2-1997/.
Bresnan, Joan (1998). ‘Morphology Competes With Syntax: Explaining Typo-
logical Variation in Weak Crossover Effects.’ In Pilar Barbosa, Danny Fox,
Paul Hagstrom, Martha McGinnis, and David Pesetsky (eds.), Is the Best
Good Enough? Optimality and Competition in Syntax, Cambridge, MA: The
MIT Press and MIT Working Papers in Linguistics, 59–92. URL http://www.
stanford.edu/~bresnan/final-pub-rev.ps. Also presented at the MIT Workshop
on Optimality in Syntax, May 1995.
Bresnan, Joan (2000). ‘Optimal Syntax.’ In Joost Dekkers, Frank van der Leeuw, and
Jeroen van de Weijer (eds.), Optimality Theory: Phonology, Syntax and Acqui-
sition, Oxford: Oxford University Press, 334–385. URL http://www.stanford.
edu/~bresnan/pt3.pdf.
Bresnan, Joan (2001a). ‘The Emergence of the Unmarked Pronoun II.’ In Géraldine
Legendre, Jane Grimshaw, and Sten Vikner (eds.), Optimality-Theoretic Syntax,
Cambridge, MA: The MIT Press, 113–142.
Bresnan, Joan (2001b). ‘Explaining Morphosyntactic Competition.’ In Mark Baltin,
and Chris Collins (eds.), Handbook of Contemporary Syntactic Theory, Oxford:
Blackwell, 11–44.
Bresnan, Joan (2001c). Lexical-Functional Syntax. Oxford: Blackwell.
Bresnan, Joan (2007). ‘Is Syntactic Knowledge Probabilistic? Experiments with the
English Dative Alternation.’ In Featherston and Sternefeld (2007), 77–96.
Bresnan, Joan, Ash Asudeh, Ida Toivonen, and Stephen Wechsler (2016). Lexical-
Functional Syntax. Oxford: Wiley Blackwell, second edition.
Bresnan, Joan, Anna Cueni, Tatiana Nikitina, and Harald Baayen (2007a). ‘Predict-
ing the Dative Alternation.’ In Gerlof Bouma, Irene Krämer, and Joost Zwarts
(eds.), Cognitive Foundations of Interpretation, Amsterdam: Royal Netherlands
 references 749

Academy of Science, 69–94. URL http://www.ehu.eus/seg/_media/morf/5/5/6/

sak/predicting_the_dative_alternation.pdf.
Bresnan, Joan, Ashwini Deo, and Devyani Sharma (2007b). ‘Typology in Variation:
A Probabilistic Approach to be and n’t in the Survey of English Dialects.’ English
Language and Linguistics 11(2): 301–346.
Bresnan, Joan, Shipra Dingare, and Christopher D. Manning (2001). ‘Soft Con-
straints Mirror Hard Constraints: Voice and Person in English and Lummi.’ In
Butt and King (2001a). URL http://csli-publications.stanford.edu/LFG/6/lfg01.
html.
Bresnan, Joan and Marilyn Ford (2010). ‘Predicting Syntax: Processing Dative
Constructions in American and Australian Varieties of English.’ Language
86(1): 168–213.
Bresnan, Joan, Per-Kristian Halvorsen, and Joan Maling (1985a). ‘Logophoricity
and Bound Anaphors.’ Unpublished manuscript, Department of Linguistics,
Stanford University.
Bresnan, Joan and Jennifer Hay (2008). ‘Gradient Grammar: An Effect of Animacy
on the Syntax of ‘Give’ in New Zealand and American English.’ Lingua 118(2):
245–259.
Bresnan, Joan and Jonni M. Kanerva (1989). ‘Locative Inversion in Chicheŵa:
A Case Study of Factorization in Grammar.’ Linguistic Inquiry 20(1): 1–50.
Reprinted in Stowell and Wehrli (1992).
Bresnan, Joan and Ronald M. Kaplan (1982). ‘Introduction: Grammars as Mental
Representations of Language.’ In Bresnan (1982b), xvii–lii.
Bresnan, Joan, Ronald M. Kaplan, and Peter G. Peterson (1985b). ‘Coordina-
tion and the Flow of Information Through Phrase Structure.’ Unpublished
manuscript, Xerox PARC.
Bresnan, Joan and Sam A. Mchombo (1987). ‘Topic, Pronoun, and Agreement in
Chicheŵa.’ Language 63(4): 741–782.
Bresnan, Joan and Sam A. Mchombo (1995). ‘The Lexical Integrity Princi-
ple: Evidence From Bantu.’ Natural Language and Linguistic Theory 13(2):
181–254.
Bresnan, Joan and Lioba Moshi (1990). ‘Object Asymmetries in Comparative
Bantu Syntax.’ Linguistic Inquiry 21(2): 147–185.
Bresnan, Joan and John Mugane (2006). ‘Agentive Nominalizations in Gı̃kũyũ and
the Theory of Mixed Categories.’ In Butt et al. (2006), 201–234.
Bresnan, Joan and Tatiana Nikitina (2010). ‘The Gradience of the Dative Alterna-
tion.’ In Uyechi and Wee (2010), 161–184.
Bresnan, Joan and Annie Zaenen (1990). ‘Deep Unaccusativity in LFG.’ In Dzi-
wirek et al. (1990), 45–57.
Broadwell, George Aaron (1998). ‘Directionals as Complex Predicates in Choctaw.’
In Butt and King (1998b). URL http://csli-publications.stanford.edu/LFG/
LFG3-1998/.
Broadwell, George Aaron (2000a). ‘Choctaw Directionals and the Syntax of Com-
plex Predication.’ In Butt and King (2000a).
Broadwell, George Aaron (2000b). ‘Word Order and Markedness in Kaqchikel.’ In
Butt and King (2000c). URL http://csli-publications.stanford.edu/LFG/LFG5-
2000/.
750 references

Broadwell, George Aaron (2002). ‘Constraint Symmetry in Optimality Theoretic

Syntax.’ In Butt and King (2002). URL http://csli-publications.stanford.edu/
LFG/7/lfg02.html.
Broadwell, George Aaron (2003). ‘Optimality, Complex Predication, and Parallel
Structures in Zapotec.’ In Butt and King (2003b). URL http://csli-publications.
stanford.edu/LFG/8/lfg03.html.
Broadwell, George Aaron (2005). ‘It Ain’t Necessarily S(V)O: Two Kinds of VSO
Languages.’ In Butt and King (2005). URL http://csli-publications.stanford.edu/
LFG/10/lfg05.html.
Broadwell, George Aaron (2006). ‘Alignment, Precedence, and the Typology
of Pied-Piping With Inversion.’ In Butt and King (2006b). URL http://csli-
publications.stanford.edu/LFG/11/lfg06.html.
Broadwell, George Aaron (2007). ‘Lexical Sharing and Non-Projecting Words: The
Syntax of Zapotec Adjectives.’ In Butt and King (2007), 87–106. URL http://csli-
publications.stanford.edu/LFG/12/lfg07.html.
Broadwell, George Aaron (2008). ‘Turkish Suspended Affixation is Lexical Shar-
ing.’ In Butt and King (2008), 198–213. URL http://csli-publications.stanford.
edu/LFG/13/lfg08.html.
Broadwell, George Aaron (2010). ‘Two Movement Paradoxes in Zapotec.’ In Butt
and King (2010), 127–143. URL http://csli-publications.stanford.edu/LFG/15/
lfg10.html.
Broadwell, George Aaron (2014). ‘Lexical-Functional Grammar.’ In The Routledge
Handbook of Syntax, London: Routledge, 556–578.
Brown, Dunstan and Andrew Hippisley (2012). Network Morphology: A Defaults-
Based Theory of Word Structure. Cambridge, UK: Cambridge University Press.
Brown, Keith (ed.) (2006). Encyclopedia of Language and Linguistics. Amsterdam:
Elsevier, second edition.
Brown, Keith and Jim Miller (eds.) (1996). Concise Encyclopedia of Syntactic
Theories. Oxford: Elsevier Science.
Brown, Lea and Matthew S. Dryer (2008). ‘The Verbs for ‘And’ in Walman, A
Torricelli Language of Papua New Guinea.’ Language 84(3): 528–565.
Brun, Caroline (1996a). ‘La coordination dans le cadre d’une grammaire LFG
du français.’ In Proceedings of the First Student Conference In Computational
Linguistics in Montreal. Montréal, 192–199.
Brun, Caroline (1996b). ‘Using Priority Union for Non-Constituent Coordination
in LFG.’ Presented at the LFG96 Conference, Grenoble, France, August 1996.
Butt, Miriam (1995). The Structure of Complex Predicates in Urdu. Dissertations
in Linguistics. Stanford: CSLI Publications. Revised version of 1993 Stanford
University dissertation.
Butt, Miriam (1998). ‘Constraining Argument Merger Through Aspect.’ In Erhard
Hinrichs, Andreas Kathol, and Tsuneko Nakazawa (eds.), Complex Predicates in
Nonderivational Syntax, San Diego: Academic Press, volume 30 of Syntax and
Semantics, 73–113.
Butt, Miriam (2001a). ‘A Reexamination of the Accusative to Ergative Shift.’ In Butt
and King (2001b).
Butt, Miriam (2001b). ‘The Treatment of Tense (Overview).’ In Butt and King
(2001a). URL http://csli-publications.stanford.edu/LFG/6/lfg01.html.
 references 751

Butt, Miriam (2006). Theories of Case. Cambridge, UK: Cambridge University

Press.
Butt, Miriam (2007). ‘The Role of Pronominal Suffixes in Punjabi.’ In Zaenen et al.
(2007).
Butt, Miriam (2008). ‘Case in Lexical-Functional Grammar.’ In Andrej Malchukov,
and Andrew Spencer (eds.), The Handbook of Case, Oxford: Oxford University
Press, 59–71.
Butt, Miriam (2010). ‘The Light Verb Jungle: Still Hacking Away.’ In Mengistu
Amberber, Brett Baker, and Mark Harvey (eds.), Complex Predicates: Cross-
Linguistic Perspectives on Event Structure, Cambridge, UK: Cambridge Univer-
sity Press, 48–78.
Butt, Miriam (2012). ‘Questions in Urdu/Hindi: Moving Beyond Movement.’
URL http://web.stanford.edu/group/cslipublications/cslipublications/LFG/17/
abstracts/lfg12abs-butt.html. Presented at LFG12, Udayana University, Bali,
Indonesia, July 2009. Abstract in On-Line Proceedings of the LFG2012 Confer-
ence, ed. Miriam Butt and Tracy Holloway King.
Butt, Miriam (2014a). ‘Control vs. Complex Predication: Identifying Non-Finite
Complements.’ Natural Language and Linguistic Theory 32(1): 165–190.
Butt, Miriam (2014b). ‘Questions and Information Structure in Urdu/Hindi.’ In
Butt and King (2014). URL http://csli-publications.stanford.edu/LFG/19/lfg14.
html.
Butt, Miriam, Tina Bögel, Annette Hautli, Sebastian Sulger, and Tafseer Ahmed
(2012). ‘Identifying Urdu Complex Predication via Bigram Extraction.’ In
Proceedings of the 24th International Conference on Computational Linguis-
tics (COLING12). International Committee on Computational Linguistics,
409–424.
Butt, Miriam, Mary Dalrymple, and Anette Frank (1997). ‘An Architecture for
Linking Theory in LFG.’ In Butt and King (1997). URL http://csli-publications.
stanford.edu/LFG/LFG2-1997/.
Butt, Miriam, Mary Dalrymple, and Tracy Holloway King (eds.) (2006). Intelligent
Linguistic Architectures: Variations on Themes by Ronald M. Kaplan. Stanford:
CSLI Publications.
Butt, Miriam, Helge Dyvik, Tracy Holloway King, Hiroshi Masuichi, and Christian
Rohrer (2002a). ‘The Parallel Grammar Project.’ In Proceedings of the Inter-
national Conference on Computational Linguistics (COLING2002) Workshop
on Grammar Engineering and Evaluation. Taipei: International Committee on
Computational Linguistics, 1–7.
Butt, Miriam, Christian Fortmann, and Christian Rohrer (1996a). ‘Syntactic Anal-
yses for Parallel Grammars: Auxiliaries and Genitive NPs.’ In Proceedings of the
International Conference on Computational Linguistics (COLING96). Copen-
hagen: International Committee on Computational Linguistics, 182–187.
Butt, Miriam and Wilhelm Geuder (2001). ‘On the (Semi)lexical Status of Light
Verbs.’ In Norbert Corver, and Henk van Riemsdijk (eds.), Semi-Lexical Cate-
gories: On the Content of Function Words and the Function of Content Words,
Berlin: Mouton de Gruyter, 323–370.
Butt, Miriam and Tracy Holloway King (eds.) (1996–). LFG Online Proceedings.
Stanford: CSLI Publications. URL http://csli-publications.stanford.edu/LFG/.
752 references

Butt, Miriam and Tracy Holloway King (eds.) (1996a). On-Line Proceedings of the
LFG96 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/LFG1-1996/.
Butt, Miriam and Tracy Holloway King (1996b). ‘Structural Topic and Focus
Without Movement.’ In Butt and King (1996a). URL http://csli-publications.
stanford.edu/LFG/LFG1-1996/.
Butt, Miriam and Tracy Holloway King (eds.) (1997). On-Line Proceedings of the
LFG97 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/LFG2-1997/.
Butt, Miriam and Tracy Holloway King (1998a). ‘Interfacing Phonology With
LFG.’ In Butt and King (1998b). URL http://csli-publications.stanford.edu/
LFG/LFG3-1998/.
Butt, Miriam and Tracy Holloway King (eds.) (1998b). On-Line Proceedings of the
LFG98 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/LFG3-1998/.
Butt, Miriam and Tracy Holloway King (eds.) (1999). On-Line Proceedings of the
LFG99 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/LFG4-1999/.
Butt, Miriam and Tracy Holloway King (eds.) (2000a). Argument Realization.
Stanford: CSLI Publications.
Butt, Miriam and Tracy Holloway King (2000b). ‘Null Elements in Discourse
Structure.’ In K. V. Subbarao (ed.), Papers From the NULLS Seminar. Delhi:
Motilal Banarsidass.
Butt, Miriam and Tracy Holloway King (eds.) (2000c). On-Line Proceedings of the
LFG2000 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/LFG5-2000/.
Butt, Miriam and Tracy Holloway King (eds.) (2001a). On-Line Proceedings of the
LFG2001 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/6/lfg01.html.
Butt, Miriam and Tracy Holloway King (eds.) (2001b). Time Over Matter:
Diachronic Perspectives on Morphosyntax. Stanford: CSLI Publications.
Butt, Miriam and Tracy Holloway King (eds.) (2002). On-Line Proceedings of the
LFG2002 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/7/lfg02.html.
Butt, Miriam and Tracy Holloway King (eds.) (2003a). Nominals: Inside and Out.
Stanford: CSLI Publications.
Butt, Miriam and Tracy Holloway King (eds.) (2003b). On-Line Proceedings of the
LFG2003 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/8/lfg03.html.
Butt, Miriam and Tracy Holloway King (eds.) (2004a). On-Line Proceedings of the
LFG2004 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/9/lfg04.html.
Butt, Miriam and Tracy Holloway King (eds.) (2004b). On-Line Proceedings of the
LFG2004 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/9/lfg04.html.
 references 753

Butt, Miriam and Tracy Holloway King (2004c). ‘The Status of Case.’ In Veneeta
Dayal, and Anoop Mahajan (eds.), Clause Structure in South Asian Languages,
Dordrecht: Springer, 153–198.
Butt, Miriam and Tracy Holloway King (eds.) (2005). On-Line Proceedings of the
LFG2005 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/10/lfg05.html.
Butt, Miriam and Tracy Holloway King (eds.) (2006a). Lexical Semantics in LFG.
Stanford: CSLI Publications.
Butt, Miriam and Tracy Holloway King (eds.) (2006b). On-Line Proceedings of the
LFG2006 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/11/lfg06.html.
Butt, Miriam and Tracy Holloway King (2006c). ‘Restriction for Morphological
Valency Alternations: The Urdu Causative.’ In Butt et al. (2006), 235–358.
Butt, Miriam and Tracy Holloway King (eds.) (2007). On-Line Proceedings of the
LFG2007 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/12/lfg07.html.
Butt, Miriam and Tracy Holloway King (eds.) (2008). On-Line Proceedings of the
LFG2008 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/13/lfg08.html.
Butt, Miriam and Tracy Holloway King (eds.) (2009). On-Line Proceedings of the
LFG2009 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/14/lfg09.html.
Butt, Miriam and Tracy Holloway King (eds.) (2010). On-Line Proceedings of the
LFG2010 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/15/lfg10.html.
Butt, Miriam and Tracy Holloway King (eds.) (2011). On-Line Proceedings of the
LFG2011 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/16/lfg11.html.
Butt, Miriam and Tracy Holloway King (eds.) (2012). On-Line Proceedings of the
LFG2012 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/17/lfg12.html.
Butt, Miriam and Tracy Holloway King (eds.) (2013). On-Line Proceedings of the
LFG2013 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/18/lfg13.html.
Butt, Miriam and Tracy Holloway King (eds.) (2014). On-Line Proceedings of the
LFG2014 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/19/lfg14.html.
Butt, Miriam and Tracy Holloway King (2015a). ‘Lexical-Functional Grammar.’ In
Kiss and Alexiadou (2015), 839–874.
Butt, Miriam and Tracy Holloway King (eds.) (2015b). On-Line Proceedings of the
LFG2015 Conference. Stanford: CSLI Publications. URL http://csli-publications.
stanford.edu/LFG/20/lfg15.html.
Butt, Miriam and Tracy Holloway King (eds.) (2017). On-Line Proceedings of the
LFG2017 Conference. Stanford: CSLI Publications. URL http://web.stanford.
edu/group/cslipublications/cslipublications/LFG/LFG-2017/index.shtml.
754 references

Butt, Miriam, Tracy Holloway King, Hiroshi Masuichi, and Christian Rohrer
(2002b). ‘The Parallel Grammar Project.’ In J. Carroll, N.Oostijk, and R. Sutcliffe
(eds.), Proceedings of the Workshop on Grammar Engineering and Evaluation.
1–7. COLING02.
Butt, Miriam, Tracy Holloway King, and John T. Maxwell, III (2003). ‘Com-
plex Predicates via Restriction.’ In Butt and King (2003b). URL http://csli-
publications.stanford.edu/LFG/8/lfg03.html.
Butt, Miriam, Tracy Holloway King, María-Eugenia Niño, and Frédérique Segond
(1999). A Grammar Writer’s Cookbook. Stanford: CSLI Publications.
Butt, Miriam, Tracy Holloway King, and Gillian Ramchand (2010). ‘Complex
Predication: How Did the Child Pinch the Elephant?’ In Uyechi and Wee (2010),
231–256.
Butt, Miriam, Tracy Holloway King, and Sebastian Roth (2007). ‘Urdu Correl-
atives: Theoretical and Implementational Issues.’ In Butt and King (2007),
107–127. URL http://csli-publications.stanford.edu/LFG/12/lfg07.html.
Butt, Miriam and Aditi Lahiri (2013). ‘Diachronic Pertinacity of Light Verbs.’
Lingua 135: 7–29.
Butt, Miriam, María-Eugenia Niño, and Frédérique Segond (1996b). ‘Multilingual
Processing of Auxiliaries Within LFG.’ In Dafydd Gibbon (ed.), KONVENS
’96: Natural Language Processing and Speech Technology: Results of the 3rd
KONVENS Conference. Mouton de Gruyter, 111–122. Reprinted in Sadler and
Spencer (2004, 11–22).
Butt, Miriam and Gillian Ramchand (2005). ‘Complex Aspectual Structure in
Hindi/Urdu.’ In Nomi Erteschik-Shir, and Tova Rapoport (eds.), The Syntax of
Aspect, Oxford: Oxford University Press, 117–153.
Butt, Miriam and Alexandros Tantos (2004). ‘Verbal Semantics via Petri Nets.’ In
Butt and King (2004b). URL http://csli-publications.stanford.edu/LFG/9/lfg04.
html.
Camilleri, Maris and Louisa Sadler (2011). ‘Restrictive Relative Clauses in Maltese.’
In Butt and King (2011). URL http://csli-publications.stanford.edu/LFG/16/
lfg11.html.
Carnie, Andrew (2012a). ‘Lexical-Functional Grammar.’ In Carnie (2012b). URL
http://bcs.wiley.com/he-bcs/Books?action=resource&bcsId=7378&itemId=
0470655313&resourceId=28724.
Carnie, Andrew (2012b). Syntax: A Generative Introduction. Oxford: Blackwell,
third edition.
Cavar, Damir and Melanie Seiss (2011). ‘Clitic Placement, Syntactic Discontinuity
and Information Structure.’ In Butt and King (2011), 131–151. URL http://csli-
publications.stanford.edu/LFG/16/lfg11.html.
Çetinoğlu, Özlem and Miriam Butt (2008). ‘Turkish Non-Canonical Objects.’ In
Butt and King (2008). URL http://csli-publications.stanford.edu/LFG/13/lfg08.
html.
Çetinoğlu, Özlem, Miriam Butt, and Kemal Oflazer (2009). ‘Mono/Bi-Clausality
of Turkish Causative.’ In Essays on Turkish Linguistics: Proceedings of the 14th
International Conference on Turkish Linguistics, August 6–8, 2008. Wiesbaden:
Harrassowitz, 43–52.
 references 755

Champollion, Lucas (2016). ‘Ten Men and Women Got Married Today: Noun
Coordination and the Intersective Theory of Conjunction.’ Journal of Semantics
33(3): 561–622.
Chen, Matthew Y. (1987). ‘The Syntax of Xiamen Tone Sandhi.’ Phonology Year-
book 4: 109–149.
Chierchia, Gennaro (1984a). ‘Anaphoric Properties of Infinitives and Gerunds.’
In Mark Cobler, Susannah MacKaye, and Michael Wescoat (eds.), WCCFL 3:
Proceedings of the Third West Coast Conference on Formal Linguistics. Stanford:
CSLI Publications, 28–39.
Chierchia, Gennaro (1984b). Topics in the Semantics of Infinitives and Gerunds.
Ph.D. thesis, University of Massachusetts, Amherst.
Chierchia, Gennaro (1985). ‘Formal Semantics and the Grammar of Predication.’
Linguistic Inquiry 16(3): 417–443.
Chierchia, Gennaro and Pauline Jacobson (1985). ‘Local and Long Distance
Control.’ In Stephen Berman, Jae-Woong Choe, and Joyce McDonough (eds.),
Proceedings of NELS 16. Graduate Linguistic Student Association, University of
Massachusetts, Amherst.
Chisarik, Erika (2002). ‘Partitive Noun Phrases in Hungarian.’ In Butt and King
(2002). URL http://csli-publications.stanford.edu/LFG/7/lfg02.html.
Chisarik, Erika and John Payne (2003). ‘Modelling Possessor Constructions in
LFG: English and Hungarian.’ In Butt and King (2003a), 181–199.
Choi, Hye-Won (1999). Optimizing Structure in Context: Scrambling and Informa-
tion Structure. Dissertations in Linguistics. Stanford: CSLI Publications. Revised
version of 1996 Stanford University dissertation.
Choi, Hye-Won (2001). ‘Phrase Structure, Information Structure, and Resolution
of Mismatch.’ In Sells (2001a).
Chomsky, Noam (1955). The Logical Structure of Linguistic Theory. New York:
Plenum Press. Published in 1975.
Chomsky, Noam (1957). Syntactic Structures. Mouton de Gruyter.
Chomsky, Noam (1965). Aspects of the Theory of Syntax. Cambridge, MA: The MIT
Press.
Chomsky, Noam (1970). ‘Remarks on Nominalization.’ In Roderick A. Jacobs,
and Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar,
Waltham, MA: Ginn, 184–221.
Chomsky, Noam (1971). ‘Deep Structure, Surface Structure, and Semantic Inter-
pretation.’ In Danny D. Steinberg, and Leon A. Jakobovits (eds.), Semantics: An
Interdisciplinary Reader In Philosophy, Linguistics and Psychology, Cambridge,
UK: Cambridge University Press, 183–216.
Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam (1986). Barriers. Cambridge, MA: The MIT Press.
Chomsky, Noam (1993). ‘A Minimalist Program for Linguistic Theory.’ In Hale and
Keyser (1993), 1–52. Reprinted as chapter 3 of Chomsky (1995).
Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: The MIT
Press.
Chomsky, Noam (2001). ‘Beyond Explanatory Adequacy.’ Technical report,
Department of Linguistics and Philosophy, MIT, Cambridge, MA. MIT Occa-
sional Papers in Linguistics 20.
756 references

Chomsky, Noam and Morris Halle (1968). The Sound Pattern of English. New York:
Harper & Row.
Chung, Sandra (1982). ‘Unbounded Dependencies in Chamorro Grammar.’ Lin-
guistic Inquiry 13(1): 39–78.
Cinque, Guglielmo (1990). Types of Ā-Dependencies. Linguistic Inquiry Mono-
graphs. Cambridge, MA: The MIT Press.
Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-Linguistic
Perspective. Oxford: Oxford University Press.
Cinque, Guglielmo (2003). ‘The Interaction of Passive, Causative and ‘Restruc-
turing’ in Romance.’ In Christina Tortora (ed.), The Syntax of Italian Dialects,
Oxford: Oxford University Press, 50–66.
Cinque, Guglielmo (2004). ‘Restructuring and Functional Structure.’ In Adri-
ana Belletti (ed.), Structures and Beyond, Oxford: Oxford University Press,
132–191.
Clark, Brady Zack (2000). ‘Some Things are Not Susceptible to Thinking
About: The Historical Development of Tough Complementation.’ Unpublished
manuscript, Department of Linguistics, Stanford University.
Clément, Lionel (2016). ‘XLFG Documentation.’ URL https://hal.inria.fr/hal-
01277648/document.
Clément, Lionel and Alexandra Kinyon (2001). ‘XLFG – An LFG Parsing Scheme
for French.’ In Butt and King (2001a). URL http://csli-publications.stanford.
edu/LFG/6/lfg01.html.
Clements, George N. and Kevin C. Ford (1979). ‘Kikuyu Tone Shift and its
Synchronic Consequences.’ Linguistic Inquiry 10(2): 179–210.
Colban, Erik (1987). ‘Prepositional Phrases in Situation Schemata.’ In Fenstad et al.
(1987).
Comrie, Bernard (1974). ‘Causatives and Universal Grammar.’ Transactions of the
Philological Society 73(1): 1–32.
Comrie, Bernard (1985). Tense. Cambridge Textbooks in Linguistics. Cambridge,
UK: Cambridge University Press.
Comrie, Bernard (1987). ‘Russian.’ In Timothy Shopen (ed.), Languages and their
Status, Philadelphia: University of Pennsylvania, 91–151.
Cook, Phillipa and John Payne (2006). ‘Information Structure and Scope in
German.’ In Butt and King (2006b). URL http://csli-publications.stanford.edu/
LFG/11/lfg06.html.
Coppock, Elizabeth and Stephen Wechsler (2010). ‘Less-Travelled Paths From
Pronoun to Agreement: The Case of the Uralic Objective Conjugations.’ In
Butt and King (2010). URL http://csli-publications.stanford.edu/LFG/15/lfg10.
html.
Coppock, Elizabeth and Stephen Wechsler (2012). ‘The Objective Conjugation in
Hungarian: Agreement Without Phi-Features.’ Natural Language and Linguistic
Theory 30(3): 699–740.
Corbett, Greville G. (1983). ‘Resolution Rules: Agreement in Person, Number, and
Gender.’ In Gerald Gazdar, Ewan Klein, and Geoffrey K. Pullum (eds.), Order,
Concord and Constituency, Dordrecht: Foris, 175–206.
Corbett, Greville G. (2000). Number. Cambridge, UK: Cambridge University Press.
Corblin, Francis (1996). ‘Multiple Negation Processing in Natural Language.’
Theoria 62(3): 214–259.
 references 757

Cormons, Boris (1999). Analyse et désambiguïsation: une approche à base de corpus

(Data-Oriented Parsing) pour les représentations lexicales fonctionnelles. Ph.D.
thesis, Université de Rennes, France.
Covington, Michael A. (1984). Syntactic Theory in the High Middle Ages. Cam-
bridge, UK: Cambridge University Press.
Crouch, Dick, Mary Dalrymple, Ronald M. Kaplan, Tracy Holloway King, John T.
Maxwell, III, and Paula S. Newman (2008). XLE Documentation. Palo Alto
Research Center, Palo Alto, CA. URL http://www2.parc.com/isl/groups/nltt/
xle/doc/xle_toc.html.
Crouch, Dick and Tracy Holloway King (2006). ‘Semantics Via F-Structure Rewrit-
ing.’ In Butt and King (2006b). URL http://csli-publications.stanford.edu/LFG/
11/lfg06.html.
Crouch, Richard (1999). ‘Ellipsis and Glue Languages.’ In Shalom Lappin, and
Elabbas Benmamoun (eds.), Fragments: Studies in Ellipsis and Gapping, Oxford:
Oxford University Press, 32–67.
Crouch, Richard (2005). ‘Packed rewriting for mapping semantics to KR.’ In
Proceedings of the 6th International Workshop on Computational Semantics,
103–14.
Crouch, Richard, Anette Frank, and Josef van Genabith (2001). ‘Glue, Under-
specification and Translation.’ In Harry Bunt, Reinhard Muskens, and Elias
Thijsse (eds.), Computing Meaning, Dordrecht: Kluwer, volume 2 of Studies in
Linguistics and Philosophy, 165–184.
Crouch, Richard and Josef van Genabith (1999). ‘Context Change, Underspecifi-
cation and the Structure of Glue Language Derivations.’ In Dalrymple (1999).
Crouch, Richard and Josef van Genabith (2000). ‘Linear Logic for Linguists:
ESSLLI-2000 Course Notes.’ URL http://www.coli.uni-saarland.de/courses/
logical-grammar/contents/crouch-genabith.pdf. Unpublished manuscript,
PARC and Dublin City University.
Culicover, Peter and Paul M. Postal (eds.) (2001). Parasitic Gaps. Cambridge, MA:
The MIT Press.
Culicover, Peter W. and Ray Jackendoff (2005). Simpler Syntax. Oxford: Oxford
University Press.
Culy, Christopher D. (1991). ‘Command and Fula ɗum Pronominals.’ In Laurel A.
Sutton, Christopher Johnson, and Ruth Shields (eds.), Proceedings of the 17th
Annual Meeting of the Berkeley Linguistics Society: Special Session on African
Linguistics. Berkeley: Berkeley Linguistics Society, 65–73.
Culy, Christopher D. (1996). ‘Agreement and Fula Pronouns.’ Studies in African
Linguistics 25(1): 1–27.
Culy, Christopher D. (2000). ‘An Incorporated Topic Marker in Takelma.’ In
Butt and King (2000c). URL http://csli-publications.stanford.edu/LFG/LFG5-
2000/.
Culy, Christopher D., Koungarma Kodio, and Patrice Togo (1995). ‘Dogon
Pronominal Systems: Their Nature and Evolution.’ Studies in African Linguistics
23(3): 315–344.
Curnow, Timothy Jowan (1999). ‘Towards a Cross-Linguistic Typology of Copula
Constructions.’ In Proceedings of the 1999 Conference of the Australian Lin-
guistic Society. Australian Linguistic Society, 1–9. URL http://www.als.asn.au/
proceedings/als1999/curnow.pdf.
758 references

Dahl, Östen (1969). Topic and Focus: A Study in Russian and General Transforma-
tional Grammar. Göteborg: Elandres Botryckeri.
Dahlstrom, Amy (1986a). ‘Nominal Arguments and Pronominal Inflection in Fox.’
Presented at the 1986 Annual Meeting of the Linguistic Society of America,
New York.
Dahlstrom, Amy (1986b). Plains Cree Morphosyntax. Ph.D. thesis, University of
California, Berkeley. Reprinted by Garland Press, New York, 1991.
Dahlstrom, Amy (1987). ‘Discontinuous Constituents in Fox.’ In Robert E. Moore,
and Paul D. Kroeber (eds.), Native American Languages and Grammatical
Typology, Bloomington: Indiana University Linguistics Club, 53–73.
Dahlstrom, Amy (2003). ‘Focus Constructions in Meskwaki (Fox).’ In Butt and
King (2003b). URL http://csli-publications.stanford.edu/LFG/8/lfg03.html.
Dahlstrom, Amy (2009). ‘OBJ𝜃 Without OBJ: A Typology of Meskwaki Objects.’ In
Butt and King (2009). URL http://csli-publications.stanford.edu/LFG/14/lfg09.
html.
Dahlstrom, Amy (2013). ‘Argument Structure of Quirky Algonquian Verbs.’ In
King and de Paiva (2013), 61–71.
Dalrymple, Mary (1993). The Syntax of Anaphoric Binding. Number 36 in CSLI
Lecture Notes. Stanford: CSLI Publications.
Dalrymple, Mary (ed.) (1999). Semantics and Syntax in Lexical Functional Gram-
mar: The Resource Logic Approach. Cambridge, MA: The MIT Press.
Dalrymple, Mary (2001). Lexical Functional Grammar, volume 34 of Syntax and
Semantics. New York: Academic Press.
Dalrymple, Mary (2006). ‘Lexical Functional Grammar.’ In Brown (2006), 82–94.
URL http://users.ox.ac.uk/~cpgl0015/lfg.pdf.
Dalrymple, Mary (2010). ‘Constraints on Operator Binding in Malayalam.’ In
Uyechi and Wee (2010), 257–270.
Dalrymple, Mary (2012). ‘Number Marking: An LFG Overview.’ In Butt and King
(2012). URL http://csli-publications.stanford.edu/LFG/17/lfg12.html.
Dalrymple, Mary (2015). ‘Obligatory Nonlocal Binding: An Exclusively Long
Distance Anaphor in Ya̧g Dii.’ Natural Language and Linguistic Theory 33(4):
1089–1120.
Dalrymple, Mary (2017). ‘Unlike Phrase Structure Category Coordination.’ In The
Very Model of a Modern Linguist – In Honor of Helge Dyvik, Bergen: Bergen
Language and Linguistics Studies (BeLLS), volume 8. URL http://dx.doi.org/
10.15845/bells.v8i1.
Dalrymple, Mary, Helge Dyvik, and Tracy Holloway King (2004a). ‘Copular
Complements: Closed or Open?’ In Butt and King (2004b). URL http://csli-
publications.stanford.edu/LFG/9/lfg04.html.
Dalrymple, Mary and Jamie Y. Findlay (2019). ‘Lexical Functional Grammar.’ In
András Kertész, Edith Moravcsik, and Csilla Rákosi (eds.), Current Approaches
to Syntax: A Comparative Handbook, De Gruyter Mouton.
Dalrymple, Mary, Vineet Gupta, John Lamping, and Vijay A. Saraswat (1999a).
‘Relating Resource-Based Semantics to Categorial Semantics.’ In Dalrymple
(1999).
 references 759

Dalrymple, Mary, Dag T. T. Haug, and John J. Lowe (2018). ‘Integrating LFG’s
Binding Theory with PCDRT.’ Journal of Language Modelling 6(1). URL http://
jlm.ipipan.waw.pl/index.php/JLM/article/view/204.
Dalrymple, Mary, Angie Hinrichs, John Lamping, and Vijay A. Saraswat (1993).
‘The Resource Logic of Complex Predicate Interpretation.’ In Keh-Jiann Chen,
and Chu-Ren Huang (eds.), ROCLING VI: Proceedings of the Computational
Linguistics Conference. Hsitou National Park, Taiwan: Computational Linguis-
tics Society of Republic of China, 3–21. Also published as Xerox Technical
Report ISTL-NLTT-1993-08-03.
Dalrymple, Mary and Bozhil Hristov (2010). ‘Agreement Patterns and Coordina-
tion in Lexical Functional Grammar.’ In Butt and King (2010). URL http://csli-
publications.stanford.edu/LFG/15/lfg10.html.
Dalrymple, Mary, Makoto Kanazawa, Yookyung Kim, Sam A. Mchombo, and
Stanley Peters (1998). ‘Reciprocal Expressions and the Concept of Reciprocity.’
Linguistics and Philosophy 21(2): 159–210.
Dalrymple, Mary and Ronald M. Kaplan (2000). ‘Feature Indeterminacy and
Feature Resolution.’ Language 76(4): 759–798.
Dalrymple, Mary, Ronald M. Kaplan, and Tracy Holloway King (2001). ‘Weak
Crossover and the Absence of Traces.’ In Butt and King (2001a). URL http://
csli-publications.stanford.edu/LFG/6/lfg01.html.
Dalrymple, Mary, Ronald M. Kaplan, and Tracy Holloway King (2004b). ‘Linguis-
tic Generalizations Over Descriptions.’ In Butt and King (2004b). URL http://
csli-publications.stanford.edu/LFG/9/lfg04.html.
Dalrymple, Mary, Ronald M. Kaplan, and Tracy Holloway King (2007). ‘The
Absence of Traces: Evidence From Weak Crossover.’ In Zaenen et al. (2007).
Dalrymple, Mary, Ronald M. Kaplan, and Tracy Holloway King (2015). ‘Economy
of Expression as a Principle of Syntax.’ Journal of Language Modelling 2(3):
377–412. URL http://jlm.ipipan.waw.pl/index.php/JLM/article/view/82.
Dalrymple, Mary, Ronald M. Kaplan, John T. Maxwell, III, and Annie Zaenen
(1995a). ‘Formal Architecture.’ In Dalrymple et al. (1995b), 1–5.
Dalrymple, Mary, Ronald M. Kaplan, John T. Maxwell, III, and Annie Zaenen
(eds.) (1995b). Formal Issues in Lexical-Functional Grammar. Stanford: CSLI
Publications.
Dalrymple, Mary, Ronald M. Kaplan, John T. Maxwell, III, and Annie Zaenen
(1995c). ‘Mathematical and Computational Issues.’ In Dalrymple et al. (1995b),
331–338.
Dalrymple, Mary, Ronald M. Kaplan, John T. Maxwell, III, and Annie Zaenen
(1995d). ‘Nonlocal Dependencies.’ In Dalrymple et al. (1995b), 131–135.
Dalrymple, Mary and Tracy Holloway King (2000). ‘Missing-Object Construc-
tions: Lexical and Constructional Variation.’ In Butt and King (2000c). URL
http://csli-publications.stanford.edu/LFG/LFG5-2000/.
Dalrymple, Mary and Tracy Holloway King (2013). ‘Nested and Crossed Depen-
dencies and the Existence of Traces.’ In King and de Paiva (2013), 139–152.
Dalrymple, Mary, Tracy Holloway King, and Louisa Sadler (2009). ‘Indeterminacy
by Underspecification.’ Journal of Linguistics 45: 31–68.
760 references

Dalrymple, Mary, John Lamping, Fernando C. N. Pereira, and Vijay A.

Saraswat (1995e). ‘Linear Logic for Meaning Assembly.’ In Suresh Manandhar,
Gabriel Pereira Lopes, and Werner Nutt (eds.), Proceedings of Computational
Logic For Natural Language Processing. Edinburgh, 75–93. URL http://arxiv.org/
pdf/cmp-lg/9504012.pdf.
Dalrymple, Mary, John Lamping, Fernando C. N. Pereira, and Vijay A. Saraswat
(1997). ‘Quantifiers, Anaphora, and Intensionality.’ Journal of Logic, Language
and Information 6(3): 219–273. Reprinted in Dalrymple (1999).
Dalrymple, Mary, John Lamping, Fernando C. N. Pereira, and Vijay A. Saraswat
(1999b). ‘Overview and Introduction.’ In Dalrymple (1999).
Dalrymple, Mary and Helge Lødrup (2000). ‘The Grammatical Functions of
Complement Clauses.’ In Butt and King (2000c). URL http://csli-publications.
stanford.edu/LFG/LFG5-2000/.
Dalrymple, Mary and Louise Mycock (2011). ‘The Prosody-Syntax Interface.’ In
Butt and King (2011). URL http://csli-publications.stanford.edu/LFG/16/lfg11.
html.
Dalrymple, Mary and Irina Nikolaeva (2006). ‘Syntax of Natural and Accidental
Coordination: Evidence From Agreement.’ Language 82(4): 824–849.
Dalrymple, Mary and Irina Nikolaeva (2011). Objects and Information Structure.
Cambridge, UK: Cambridge University Press.
Dalrymple, Mary, Stuart M. Shieber, and Fernando C. N. Pereira (1991). ‘Ellipsis
and Higher-Order Unification.’ Linguistics and Philosophy 14(4): 399–452.
Dalrymple, Mary and Annie Zaenen (1991). ‘Modeling Anaphoric Superiority.’ In
Proceedings of the International Conference on Current Issues in Computational
Linguistics. Penang, Malaysia, 235–247.
Daneš, František (1974). ‘Functional Sentence Perspective and the Organization
of the Text.’ In František Daneš (ed.), Papers on Functional Sentence Perspective,
Prague: Academia, 106–128.
Davidson, Donald (1967). ‘The Logical Form of Action Sentences.’ In Nicholas
Rescher (ed.), The Logic of Decision and Action, Pittsburgh: University of
Pittsburgh Press. Reprinted in Davidson (1980, Chapter 6).
Davidson, Donald (1980). Essays on Actions and Events. Oxford: Oxford University
Press.
Davis, Anthony and Jean-Pierre Koenig (2000). ‘Linking as Constraints on Word
Classes in a Hierarchical Lexicon.’ Language 76: 56–91.
Davison, Alice (1998). ‘Structural Case, Lexical Case and Complex Predicates.’
South Asian Language Review VIII.
de Alencar, Leonel Figueiredo and Carmen Kelling (2005). ‘Are Reflexive Con-
structions Transitive Or Intransitive? Evidence From German and Romance.’ In
Butt and King (2005). URL http://csli-publications.stanford.edu/LFG/10/lfg05.
html.
de Marneffe, Marie-Catherine, Scott Grimm, Inbal Arnon, Susannah Kirby, and
Joan Bresnan (2012). ‘A Statistical Model of the Grammatical Choices in Child
Production of Dative Sentences.’ Language and Cognitive Processes 27(1): 25–61.
Deguchi, Masanori and Yoshihisa Kitagawa (2002). ‘Prosody and Wh-questions.’
In Masako Hirotani (ed.), Proceedings of NELS 32. Graduate Linguistic Student
Association, University of Massachusetts, Amherst, 73–92.
 references 761

Dench, Alan Charles (1995). Martuthunira: A Language of the Pilbara Region of

Western Australia. Canberra: Pacific Linguistics.
Di Biase, Bruno and Camilla Bettoni (2015). ‘The Development of Italian as a
Second Language.’ In Bettoni and Di Biase (2015a), 117–147.
Dik, Helma (1995). Word Order in Ancient Greek: A Pragmatic Account of Word
Order Variation in Herodotus. Amsterdam: J. C. Gieben.
Dik, Helma (2007). Word Order in Greek Tragic Dialogue. Oxford: Oxford Univer-
sity Press.
Dione, Cheikh Bamba (2012). ‘An LFG Approach To Wolof Cleft Constructions.’
In Butt and King (2012), 157–176. URL http://csli-publications.stanford.edu/
LFG/17/lfg12.html.
Dipper, Stefanie (2003). Implementing and Documenting Large-Scale Grammars:
German LFG. Ph.D. thesis, Universität Stuttgart.
Dipper, Stefanie (2005). ‘German Quantifiers: Determiners Or Adjectives?’ In
Butt and King (2005). URL http://csli-publications.stanford.edu/LFG/10/lfg05.
html.
Dixon, Robert M. W. (1994). Ergativity. Cambridge, UK: Cambridge University
Press.
Dobrovie-Sorin, Carmen (2001). ‘From DPs to NPs.’ Presented at the Antwerp
Conference From NPs to DPs, February 2001.
Doron, Edit (1982). ‘On the Syntax and Semantics of Resumptive Pronouns.’ In
Texas Linguistic Forum. Austin: Department of Linguistics, University of Texas
at Austin, volume 19, 1–48.
Dowty, David R. (1982). ‘Grammatical Relations and Montague Grammar.’ In
Jacobson and Pullum (1982), 79–130.
Dowty, David R. (1985). ‘On Recent Analyses of the Semantics of Control.’ Linguis-
tics and Philosophy 8: 291–231.
Dowty, David R. (1988). ‘Type-Raising, Functional Composition, and Non-
Constituent Conjunction.’ In Oehrle et al. (1988), 153–197.
Dowty, David R. (1989). ‘On the Semantic Content of the Notion of ‘The-
matic Role’.’ In Gennaro Chierchia, Barbara H. Partee, and Raymond
Turner (eds.), Properties, Types, and Meanings, Dordrecht: Kluwer, volume 1,
69–130.
Dowty, David R. (1991). ‘Thematic Proto Roles and Argument Selection.’ Language
67(3): 547–619.
Dras, Mark, François Lareau, Benjamin Börschinger, Robert Dale, Yasaman
Motazedi, Owen Rambow, Myfany Turpin, and Morgan Ulinski (2012). ‘Com-
plex Predicates in Arrernte.’ In Butt and King (2012), 157–176. URL http://csli-
publications.stanford.edu/LFG/17/lfg12.html.
Dryer, Matthew S. (1986). ‘Primary Objects, Secondary Objects, and Antidative.’
Language 62(4): 808–845.
Duncan, Lachlan (2007). ‘Analytic Noun Incorporation in Chuj and K’ichee’
Mayan.’ In Butt and King (2007), 163–183. URL http://csli-publications.
stanford.edu/LFG/12/lfg07.html.
Dyvik, Helge (1999). ‘The Universality of F-Structure: Discovery or Stipulation?
The Case of Modals.’ In Butt and King (1999). URL http://csli-publications.
stanford.edu/LFG/LFG4-1999/.
762 references

Dziwirek, Katarzyna, Patrick Farrell, and Errapel Mejías-Bikandi (eds.) (1990).

Grammatical Relations. A Cross-Theoretical Perspective. Stanford: CSLI Publi-
cations.
Embick, David and Rolf Noyer (2007). ‘Distributed Morphology and the Syntax-
Morphology Interface.’ In Ramchand and Reiss (2007), 289–324.
Emms, Martin (1990). ‘Polymorphic Quantifiers.’ In Guy Barry, and Glyn Morrill
(eds.), Studies in Categorial Grammar. Centre for Cognitive Science, University
of Edinburgh, 5, 65–111. Edinburgh Working Papers in Cognitive Science.
Emonds, Joseph (1976). A Transformational Approach to English Syntax: Root,
Structure-Preserving, and Local Transformations. New York: Academic Press.
Engdahl, Elisabet (1983). ‘Parasitic Gaps.’ Linguistics and Philosophy 6(1): 5–34.
Reprinted in Culicover and Postal (2001, 69–97).
Engdahl, Elisabet (1999). ‘The Choice Between Bli-Passive and S-Passive in
Danish, Norwegian and Swedish.’ URL http://www.science.gu.se/digitalAssets/
1311/1311028_engdahl-nordsem-passivechoice1999.pdf. Unpublished manu-
script, Goteborg University.
Engdahl, Elisabet and Enric Vallduví (1996). ‘Information Packaging in HPSG.’
In Claire Grover, and Enric Vallduví (eds.), Studies in HPSG. Edinburgh:
University of Edinburgh, volume 12 of Edinburgh Working Papers in Cognitive
Science, 1–32.
Erteschik, Nomi (1973). On the Nature of Island Constraints. Ph.D. thesis, Mas-
sachusetts Institute of Technology.
Erteschik-Shir, Nomi (2007). Information Structure: The Syntax-Discourse Inter-
face. Oxford: Oxford University Press.
Estigarribia, Bruno (2005). ‘Direct Object Clitic Doubling in OT-LFG: A New Look
at Rioplatense Spanish.’ In Butt and King (2005). URL http://csli-publications.
stanford.edu/LFG/10/lfg05.html.
Evans, Gareth (1980). ‘Pronouns.’ Linguistic Inquiry 11(2): 337–362. Reprinted in
Evans (1985, 214–248).
Evans, Gareth (1985). Collected Papers. Oxford: Clarendon Press. Edited by Anto-
nia Phillips.
Falk, Yehuda N. (1983). ‘Constituency, Word Order, and Phrase Structure Rules.’
Linguistic Analysis 11: 331–360.
Falk, Yehuda N. (1984). ‘The English Auxiliary System: A Lexical-Functional
Analysis.’ Language 60(3): 483–509.
Falk, Yehuda N. (2000). ‘Pivots and the Theory of Grammatical Functions.’ In
Butt and King (2000c). URL http://csli-publications.stanford.edu/LFG/LFG5-
2000/.
Falk, Yehuda N. (2001a). ‘Constituent Structure and Grammatical Functions in
the Hebrew Action Nominal.’ In Butt and King (2001a). URL http://csli-
publications.stanford.edu/LFG/6/lfg01.html.
Falk, Yehuda N. (2001b). Lexical-Functional Grammar: An Introduction to Parallel
Constraint-Based Syntax. Stanford: CSLI Publications.
Falk, Yehuda N. (2002a). ‘The Place of Possessors in a Theory of Grammatical
Functions.’ URL http://pluto.mscc.huji.ac.il/~msyfalk/POSS.pdf. Unpublished
manuscript, Department of English, The Hebrew University of Jerusalem.
 references 763

Falk, Yehuda N. (2002b). ‘Resumptive Pronouns in LFG.’ In Butt and King (2002).
URL http://csli-publications.stanford.edu/LFG/7/lfg02.html.
Falk, Yehuda N. (2003). ‘The English Auxiliary System Revisited.’ In Butt
and King (2003b), 184–204. URL http://csli-publications.stanford.edu/LFG/8/
lfg03.html.
Falk, Yehuda N. (2004). ‘The Hebrew Present Tense Copula as a Mixed Category.’
In Butt and King (2004b). URL http://csli-publications.stanford.edu/LFG/9/
lfg04.html.
Falk, Yehuda N. (2005). ‘Open Argument Functions.’ In Butt and King (2005). URL
http://csli-publications.stanford.edu/LFG/10/lfg05.html.
Falk, Yehuda N. (2006). Subjects and Universal Grammar: An Explanatory Theory.
Cambridge, UK: Cambridge University Press.
Falk, Yehuda N. (2007). ‘Do We Wanna (Or Hafta) Have Empty Categories?’ In
Butt and King (2007). URL http://csli-publications.stanford.edu/LFG/12/lfg07.
html.
Falk, Yehuda N. (2008). ‘Functional Relations in the English Auxiliary System.’
Linguistics 46: 861–889.
Falk, Yehuda N. (2009). ‘Islands: A Mixed Analysis.’ In Butt and King (2009),
261–281. URL http://csli-publications.stanford.edu/LFG/14/lfg09.html.
Falk, Yehuda N. (2010). ‘An Unmediated Analysis of Relative Clauses.’ In Butt
and King (2010), 207–227. URL http://csli-publications.stanford.edu/LFG/15/
lfg10.html.
Falk, Yehuda N. (2011). ‘Multiple-Gap Constructions.’ In Butt and King (2011).
URL http://csli-publications.stanford.edu/LFG/16/lfg11.html.
Faltz, Leonard M. (1977). Reflexivization: A Study in Universal Syntax. Ph.D. thesis,
University of California, Berkeley. Reprinted by Garland Press, New York,
1985.
Fassi-Fehri, Abdulkader (1981). Complémentation et Anaphore en Arabe Moderne:
Une Approche Lexicale Fonctionelle. Ph.D. thesis, Université Paris 3.
Fassi-Fehri, Abdulkader (1988). ‘Agreement in Arabic, Binding, and Coherence.’
In Barlow and Ferguson (1988), 107–158.
Featherston, Sam and Wolfgang Sternefeld (eds.) (2007). Roots: Linguistics in
Search of its Evidential Base. Berlin: Mouton de Gruyter.
Fenstad, Jens Erik, Per-Kristian Halvorsen, Tore Langholm, and Johan van Ben-
them (eds.) (1987). Situations, Language and Logic. Dordrecht: D. Reidel.
Féry, Caroline and Manfred Krifka (2008). ‘Information Structure: Notional Dis-
tinctions, Ways of Expression.’ In Piet van Sterkenburg (ed.), Unity and Diversity
of Languages, Amsterdam: John Benjamins, 123–136.
Fillmore, Charles J. (1968). ‘The Case for Case.’ In Emmon Bach, and Robert T.
Harms (eds.), Universals in Linguistic Theory, New York: Holt, Rinehart &
Winston, 1–210.
Findlay, Jamie Y. (2014). The Prepositional Passive: a Lexical Functional Account.
M.Phil. dissertation, University of Oxford.
Findlay, Jamie Y. (2016). ‘Mapping Theory Without Argument Structure.’ Journal
of Language Modelling 4(2): 293–338. URL http://jlm.ipipan.waw.pl/index.php/
JLM/article/view/171/0.
764 references

Finn, Regina (2007). Grammatical Feature Data-Oriented Parsing. Master’s thesis,

Dublin City University, School of Computing. URL http://doras.dcu.ie/17009/.
Finn, Ríona, Mary Hearne, Andy Way, and Josef van Genabith (2006). ‘GF-DOP:
Grammatical Feature Data-Oriented Parsing.’ In Butt and King (2006b). URL
http://csli-publications.stanford.edu/LFG/11/lfg06.html.
Firbas, Jan (1964). ‘On Defining the Theme in Functional Sentence Perspective.’
Travaux linguistiques de Prague 1: 267–280.
Flickinger, Daniel P. (1987). Lexical Rules in the Hierarchical Lexicon. Ph.D. thesis,
Stanford University.
Flouraki, Maria and Despina Kazana (2009). ‘Constraining Disjunctive Construc-
tions in Modern Greek.’ In Butt and King (2009), 282–296. URL http://csli-
publications.stanford.edu/LFG/14/lfg09.html.
Foley, William A. (1998). ‘Symmetrical Voice Systems and Precategoriality
in Philippine Languages.’ URL http://citeseerx.ist.psu.edu/viewdoc/summary?
doi=10.1.1.7.7620. Paper presented at the workshop on Voice and Grammatical
Functions in Austronesian Languages.
Ford, Marilyn and Joan Bresnan (2013). ‘ “They Whispered Me the Answer” in
Australia and the US: A Comparative Experimental Study.’ In King and de Paiva
(2013), 95–107.
Ford, Marilyn, Joan Bresnan, and Ronald M. Kaplan (1982). ‘A Competence-Based
Theory of Syntactic Closure.’ In Bresnan (1982b), 727–796.
Forst, Martin (2006). ‘COMP in (Parallel) Grammar Writing.’ In Butt and King
(2006b). URL http://csli-publications.stanford.edu/LFG/11/lfg06.html.
Forst, Martin, Tracy Holloway King, and Tibor Laczkó (2010). ‘Particle Verbs in
Computational LFGs: Issues From English, German, and Hungarian.’ In Butt
and King (2010), 228–248. URL http://csli-publications.stanford.edu/LFG/15/
lfg10.html.
Forst, Martin and Christian Rohrer (2009). ‘Problems of German VP Coordina-
tion.’ In Butt and King (2009), 297–316. URL http://csli-publications.stanford.
edu/LFG/14/lfg09.html.
Fox, Anthony (2000). Prosodic Features and Prosodic Structure: The Phonology of
Suprasegmentals. Oxford: Oxford University Press.
Fox, Barbara and Paul Hopper (eds.) (1993). Voice: Form and Function. Amster-
dam: John Benjamins.
Frank, Anette (1996). ‘A Note on Complex Predicate Formation: Evidence From
Auxiliary Selection, Reflexivization, and Past Participle Agreement in French
and Italian.’ In Butt and King (1996a). URL http://csli-publications.stanford.
edu/LFG/LFG1-1996/.
Frank, Anette (1997). Context Dependence in Modal Constructions. Ph.D. the-
sis, Universität Stuttgart. Published in: Arbeitspapiere des Sonderforschungs-
bereichs 340, Sprachtheoretische Grundlagen für die Computerlinguistik,
Nr. 91.
Frank, Anette (2006). ‘(Discourse-) Functional Analysis of Asymmetric Coordi-
nation.’ In Butt et al. (2006), 259–285.
Frank, Anette (2013). ‘A Tour of Grammatical Formalisms.’ In King and de Paiva
(2013), 75–94.
 references 765

Frank, Anette and Annie Zaenen (2002). ‘Tense in LFG: Syntax and Morphology.’
In Hans Kamp, and Uwe Reyle (eds.), How We Say WHEN It Happens: Con-
tributions to the Theory of Temporal Reference in Natural Language, Tübingen:
Niemeyer, 17–51. Reprinted in Sadler and Spencer (2004, 23–66).
Franks, Steven and Tracy Holloway King (2000). A Handbook of Slavic Clitics.
Oxford: Oxford University Press.
Frantz, Donald G. (1981). Grammatical Relations in Universal Grammar. Bloom-
ington: Indiana University Linguistics Club.
Frey, Werner and Uwe Reyle (1983). ‘A PROLOG Implementation of Lexical-
Functional Grammar.’ In Proceedings of the Eighth International Joint Conference
on Artificial Intelligence. Karlsruhe, 727–729.
Frota, Sónia (2012). ‘Prosodic Structure, Constituents, and Their Implementation.’
In Abigail C. Cohn, Cécile Fougeron, and Marie K. Huffman (eds.), The Oxford
Handbook of Laboratory Phonology, Oxford: Oxford University Press, 255–265.
Fry, John (1999a). ‘Proof Nets and Negative Polarity Licensing.’ In Dalrymple
(1999), 91–116.
Fry, John (1999b). ‘Resource-Logical Event Semantics for LFG.’ URL http://
citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.84.6503. Presented at
LFG99, University of Manchester.
Gabbay, Dov M. (1996). Labelled Deductive Systems. Oxford: Clarendon Press.
Gamut, L. T. F. (1991a). Logic, Language, and Meaning, volume 1: Introduction to
Logic. Chicago: University of Chicago Press.
Gamut, L. T. F. (1991b). Logic, Language, and Meaning, volume 2: Intensional Logic
and Logical Grammar. Chicago: University of Chicago Press.
Gawron, Jean Mark and Stanley Peters (1990). Anaphora and Quantification in
Situation Semantics. Number 19 in CSLI Lecture Notes. Stanford: CSLI Pub-
lications.
Gazdar, Gerald (1980). ‘A Cross-Categorial Semantics for Coordination.’ Linguis-
tics and Philosophy 3: 407–409.
Gazdar, Gerald, Ewan Klein, Geoffrey K. Pullum, and Ivan A. Sag (1985). Gener-
alized Phrase Structure Grammar. Cambridge, MA: Harvard University Press.
Gazdar, Gerald and Geoffrey K. Pullum (1981). ‘Subcategorization, Constituent
Order, and the Notion ‘Head’.’ In Michael Moortgat, Harry van der Hulst, and
Teun Hoekstra (eds.), The Scope of Lexical Rules, Dordrecht: Foris, 107–123.
Gazdik, Anna (2011). Multiple Questions in French and Hungarian. A Lexical-
Functional Analysis with Special Emphasis on the Syntax-Discourse Interface.
Ph.D. thesis, Université Paris Diderot.
Georgopoulos, Carol (1985). ‘Variables in Palauan Syntax.’ Natural Language and
Linguistic Theory 3(1): 59–94.
Ginzburg, Jonathan (1996). ‘Interrogatives: Questions, Facts, and Dialogue.’ In
Lappin (1996), 385–422.
Ginzburg, Jonathan (2001). Questions, Queries, and Facts: A Semantics and Prag-
matics for Interrogatives. Dissertations in Linguistics. Stanford: CSLI Publica-
tions. Reprinted version of 1992 Stanford dissertation.
Ginzburg, Jonathan (2011). ‘Questions: Logic and Interactions.’ In van Benthem
and ter Meulen (2011), 1133–1146.
766 references

Ginzburg, Jonathan and Ivan A. Sag (2000). Interrogative Investigations: The Form,
Meaning and Use of English Interrogatives. Stanford: CSLI Publications.
Giorgolo, Gianluca (2017). ‘Glue Repository.’ URL https://github.com/
gianlucagiorgolo. GitHub Repositories.
Giorgolo, Gianluca and Ash Asudeh (2011a). ‘Multidimensional Semantics With
Unidimensional Glue Logic.’ In Butt and King (2011), 236–256. URL http://csli-
publications.stanford.edu/LFG/16/lfg11.html.
Giorgolo, Gianluca and Ash Asudeh (2011b). ‘Multimodal Communication in
LFG: Gestures and the Correspondence Architecture.’ In Butt and King (2011),
257–277. URL http://csli-publications.stanford.edu/LFG/16/lfg11.html.
Giorgolo, Gianluca and Ash Asudeh (2012a). ‘!M, η, ⃰" Monads for Conven-
tional Implicatures.’ In Ana Aguilar-Guevara, Anna Chernilovskaya, and Rick
Nouwen (eds.), Sinn und Bedeutung 16. MIT Working Papers in Linguistics.
Giorgolo, Gianluca and Ash Asudeh (2012b). ‘Missing Resources in a Resource-
Sensitive Semantics.’ In Butt and King (2012), 219–239. URL http://csli-
publications.stanford.edu/LFG/17/lfg12.html.
Giorgolo, Gianluca and Ash Asudeh (2014). ‘Monads as a Solution for Generalized
Opacity.’ In Proceedings of the EACL 2014 Workshop on Type Theory and Natural
Language Semantics (TTNLS). 19–27. URL https://www.aclweb.org/anthology/
W/W14/W14-1400.pdf.
Girard, Jean-Yves (1987). ‘Linear Logic.’ Theoretical Computer Science 50(1):
1–102.
Givón, Talmy (1997). ‘Grammatical Relations: An Introduction.’ In Talmy Givón
(ed.), Grammatical Relations: A Functional Perspective, Amsterdam: John
Benjamins, 1–84.
Glahn, Esther, Gisela Håkansson, Björn Hammarberg, Anne Holmen, Anne
Hvenekilde, and Karen Lund (2001). ‘Processability in Scandinavian Second
Language Acquisition.’ Studies in Second Language Acquisition 23: 389–416.
Glasbey, Sheila (2001). ‘Tense, Aspect and the Temporal Structure of Discourse:
Towards an LFG Account.’ In Butt and King (2001a). URL http://csli-
publications.stanford.edu/LFG/6/lfg01.html.
Goodall, Grant (1987). Parallel Structures in Syntax. Cambridge, UK: Cambridge
University Press.
Grano, Thomas (2015). Control and Restructuring. Oxford: Oxford University
Press.
Greenberg, Joseph H. (1966a). Language Universals: With Special Reference to
Feature Hierarchies. The Hague: Mouton de Gruyter.
Greenberg, Joseph H. (1966b). ‘Some Universals of Grammar With Particular Ref-
erence to the Order of Meaningful Elements.’ In Joseph H. Greenberg (ed.), Uni-
versals of Language, Cambridge, MA: The MIT Press, 73–113. Second edition.
Grimshaw, Jane (1982a). ‘On the Lexical Representation of Romance Reflexive
Clitics.’ In Bresnan (1982b), 87–148.
Grimshaw, Jane (1982b). ‘Subcategorization and Grammatical Relations.’ In
Zaenen (1982), 35–56.
Grimshaw, Jane (1990). Argument Structure. Cambridge, MA: The MIT Press.
Groenendijk, Jeroen and Martin Stokhof (1984). Studies on the Semantics of Ques-
tions and the Pragmatics of Answers. Ph.D. thesis, University of Amsterdam.
 references 767

Groenendijk, Jeroen and Martin Stokhof (1991). ‘Dynamic Predicate Logic.’

Linguistics and Philosophy 14(1): 39–100.
Groenendijk, Jeroen and Martin Stokhof (1997). ‘Questions.’ In van Benthem
and ter Meulen (1997), 1055–1124. Co-published with Elsevier Science B.V.,
Amsterdam.
Groos, Anneke and Henk van Reimsdijk (1979). ‘Matching Effects in Free
Relatives: A Parameter of Core Grammar.’ In Adriana Belletti, Luciana
Brandi, and Luigi Rizzi (eds.), Theory of Markedness in Generative Grammar:
Proceedings of the 1979 GLOW Conference. GLOW, Pisa: Scuola Normale
Superiore di Pisa, 171–216.
Gruber, Jeffrey S. (1965). Studies in Lexical Relations. Ph.D. thesis, Massachusetts
Institute of Technology.
Gundel, Jeanette K. (1974). The Role of Topic and Comment in Linguistic Theory.
Ph.D. thesis, University of Texas at Austin. Reprinted by Garland Press, New
York, 1988.
Gunlogson, Christine (2003). True to Form: Rising and Falling Declaratives as
Questions in English. New York: Routledge.
Guo, Yuqing, Josef van Genabith, and Haifeng Wang (2007). ‘Treebank-Based
Acquisition of LFG Resources for Chinese.’ In Butt and King (2007), 214–232.
URL http://csli-publications.stanford.edu/LFG/12/lfg07.html.
Gupta, Vineet and John Lamping (1998). ‘Efficient Linear Logic Meaning
Assembly.’ In COLING/ACL98: Joint Meeting of the 36th Annual Meeting
of the ACL and the 17th International Conference on Computational
Linguistics. Montréal: International Committee on Computational Linguistics,
464–470.
Hagége, Claude (1974). ‘Les Pronoms Logophoriques.’ Bulletin de la Société de
Linguistique de Paris 69: 287–310.
Hajičová, Eva, Barbara H. Partee, and Petr Sgall (1998). Topic-Focus Articulation,
Tripartite Structures, and Semantic Content. Dordrecht: Kluwer.
Hale, Kenneth (1983). ‘Warlpiri and the Grammar of Non-Configurational
Languages.’ Natural Language and Linguistic Theory 1(1): 5–47.
Hale, Kenneth and Samuel Jay Keyser (eds.) (1993). The View From Building 20:
Essays in Honor of Sylvain Bromberger. Cambridge, MA: The MIT Press.
Hale, Mark (2007). Historical Linguistics: Theory and Method. Oxford: Blackwell.
Halle, Morris and Alec Marantz (1993). ‘Distributed Morphology and the Pieces
of Inflection.’ In Hale and Keyser (1993), 111–176.
Halliday, M. A. K. (1967). ‘Notes on Transitivity and Theme in English.’ Journal of
Linguistics 3: 199–244.
Halpern, Aaron (1995). On the Placement and Morphology of Clitics. Dissertations
in Linguistics. Stanford: CSLI Publications. Revised version of 1992 Stanford
University dissertation.
Halvorsen, Per-Kristian (1983). ‘Semantics for Lexical Functional Grammar.’
Linguistic Inquiry 14(4): 567–615.
Halvorsen, Per-Kristian and Ronald M. Kaplan (1988). ‘Projections and Semantic
Description in Lexical-Functional Grammar.’ In Proceedings of the International
Conference on Fifth Generation Computer Systems (FGCS-88). Tokyo: ICOT,
1116–1122. Reprinted in Dalrymple et al. (1995b, 279–292).
768 references

Hamblin, C. L. (1958). ‘Questions.’ Australasian Journal of Philosophy 36(3):

159–168.
Hamblin, C. L. (1976). ‘Questions in Montague English.’ In Barbara H. Partee
(ed.), Montague Grammar, New York: Academic Press, 247–259. Originally
published in Foundations of Language 10, 1973.
Haspelmath, Martin (ed.) (2004a). Coordinating Constructions. Amsterdam: John
Benjamins.
Haspelmath, Martin (2004b). ‘Coordinating Constructions: An Overview.’ In
Haspelmath (2004a), 3–40.
Haspelmath, Martin, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.)
(2008). The World Atlas of Language Structures Online. Munich: Max Planck
Digital Library. http://wals.info.
Haspelmath, Martin and Andrea D. Sims (2011). Understanding Morphology.
London: Hodder, second edition.
Haug, Dag T. T. (2008a). ‘Detecting Word Order Freezing in Old Indo-European
Languages.’ URL http://www.hf.uio.no/ifikk/english/research/projects/proiel/
Activities/proiel/publications/NRGpresentation.pdf. Presented at NRG 2008
(New Reflections on Grammaticalization 4), Leuven, 16th July 2008.
Haug, Dag T. T. (2008b). ‘Tense and Aspect for Glue Semantics: The Case of
Participial XADJ’s.’ In Butt and King (2008). URL http://csli-publications.
stanford.edu/LFG/13/lfg08.html.
Haug, Dag T. T. (2009). ‘Med – The Syntax and Semantics of Concomitance in
Norwegian.’ In Butt and King (2009), 338–356. URL http://csli-publications.
stanford.edu/LFG/14/lfg09.html.
Haug, Dag T. T. (2011). ‘Backward Control in Ancient Greek: Subsumption or
Linearization?’ In Butt and King (2011). URL http://csli-publications.stanford.
edu/LFG/16/lfg11.html.
Haug, Dag T. T. (2012). ‘Open Verb-Based Adjuncts in New Testament Greek
and the Latin of the Vulgate.’ In Cathrine Fabricius-Hansen, and Dag T. T.
Haug (eds.), Big Events, Small Clauses: The Grammar of Elaboration, Berlin: de
Gruyter, 287–321.
Haug, Dag T. T. (2013). ‘Partial Control and Anaphoric Control in LFG.’ In Butt and
King (2013). URL http://csli-publications.stanford.edu/LFG/18/lfg13.html.
Haug, Dag T. T. (2014a). ‘The Anaphoric Semantics of Partial Control.’ In Todd
Snider, Sarah D’Antonio, and Mia Weigand (eds.), SALT 24: Proceedings of the
24th Semantics and Linguistic Theory Conference. 213–233.
Haug, Dag T. T. (2014b). ‘Partial Dynamic Semantics for Anaphora:
Compositionality Without Syntactic Coindexation.’ Journal of Semantics
31(4): 457–511. doi:10.1093/jos/fft008.
Haug, Dag T. T. (2017). ‘Backward Control in Ancient Greek and Latin Participial
Adjuncts.’ Natural Language and Linguistic Theory 35(1): 99–159.
Haug, Dag T. T. and Tatiana Nikitina (2012). ‘The Many Cases of Non-Finite
Subjects: The Challenge of “Dominant” Participles.’ In Butt and King (2012),
292–311. URL http://csli-publications.stanford.edu/LFG/17/lfg12.html.
Haug, Dag T. T. and Tatiana Nikitina (2016). ‘Feature Sharing in Agreement.’
Natural Language and Linguistic Theory 34(3): 865–910.
 references 769

Hayes, Bruce and Aditi Lahiri (1991). ‘Bengali Intonational Phonology.’ Natural
Language and Linguistic Theory 9: 47–96.
Hearne, Mary (2005). Data-Oriented Models of Parsing and Translation. Ph.D.
thesis, Dublin City University, School of Computing.
Hearne, Mary and Andy Way (2006). ‘Disambiguation Strategies for Data-
Oriented Translation.’ In Proceedings of the 11th Conference of the European
Association for Machine Translation. Oslo, 59–68.
Heim, Irene (1982). The Semantics of Definite and Indefinite Noun Phrases. Ph.D.
thesis, University of Massachusetts, Amherst.
Heine, Bernd and Heiko Narrog (eds.) (2015). The Oxford Handbook of Linguistic
Analysis. Oxford: Oxford University Press, second edition.
Hellan, Lars (1988). Anaphora in Norwegian and the Theory of Grammar.
Dordrecht: Foris.
Henadeerage, Kumara (1998). ‘Anaphoric Binding in Colloquial Sinhala.’ In
Butt and King (1998b). URL http://csli-publications.stanford.edu/LFG/LFG3-
1998/.
Heny, Frank (1973). ‘Sentence and Predicate Modifiers in English.’ In John P.
Kimball (ed.), Syntax and Semantics, New York: Academic Press, volume 4 of
Syntax and Semantics, 217–245.
Her, One-Soon (1998). ‘Lexical Mapping in Chinese Inversion Constructions.’ In
Butt and King (1998b). URL http://csli-publications.stanford.edu/LFG/LFG3-
1998/.
Her, One-Soon (1999). ‘Interaction of Thematic Structure and Syntactic Struc-
tures: On Mandarin Dative Alternations.’ In Chinese Languages and Linguistics
V: Interaction of Form and Function, Taipei: Academia Sinica, 373–412.
Her, One-Soon (2003). ‘Chinese Inversion Constructions Within a Simplified
LMT.’ Journal of Chinese Linguistics: Lexical-Functional Grammar Analysis of
Chinese 1–31. Edited by Adams Bodomo and Kang Kwong Luke.
Her, One-Soon (2009). ‘Apparent Subject-Object Inversion in Chinese.’ Linguistics
47(5): 1143–1181.
Her, One-Soon (2010). Interaction and Variation in the Chinese VO Construction.
Taipei: Crane Publishing. Revised edition.
Her, One-Soon (2013). ‘Lexical Mapping Theory Revisited.’ In King and de Paiva
(2013), 47–59.
Her, One-Soon and Dun-Hong Deng (2012). ‘Lexical Mapping in Yami Verbs.’
In Butt and King (2012). URL http://csli-publications.stanford.edu/LFG/17/
lfg12.html.
Heycock, Caroline and Roberto Zamparelli (2005). ‘Friends and Colleagues:
Plurality, Coordination, and the Structure of DP.’ Natural Language Semantics
13: 201–270.
Higginbotham, James (1983). ‘Logical Form, Binding, and Nominals.’ Linguistic
Inquiry 14(3): 395–420.
Higginbotham, James (1992). ‘Reference and Control.’ In Larson et al. (1992),
79–108.
Higginbotham, James (1996). ‘The Semantics of Questions.’ In Lappin (1996),
361–383.
Hinds, John (1986). Japanese. London: Croom Helm.
770 references

Hoeksema, Jacob (1988). ‘The Semantics of Non-Boolean ‘And’.’ Journal of

Semantics 6: 19–40.
Hong, Ki-Sun (1991). Argument Selection and Case Marking in Korean. Ph.D.
thesis, Stanford University.
Hornstein, Norbert and Amy Weinberg (1995). ‘The Empty Category Principle.’
In Webelhuth (1995).
Hristov, Bozhil (2012). Agreement, Case Assignment, and Nominal Coordination.
D.Phil. thesis, University of Oxford. URL https://ora.ox.ac.uk/objects/uuid:
36b5e5b9-884e-445a-b0cd-f7dfde9cebbe.
Hristov, Bozhil (2013a). ‘Defacing Agreement.’ In Butt and King (2013). URL
http://csli-publications.stanford.edu/LFG/18/lfg13.html.
Hristov, Bozhil (2013b). ‘Pronominal Case Assignment in English.’ Journal of
Linguistics 49(3): 567–611.
Huang, C.-T. James (1982). Logical Relations in Chinese and the Theory of
Grammar. Ph.D. thesis, Massachusetts Institute of Technology.
Huang, Chu-Ren (1990). ‘A Unification-Based LFG Analysis of Lexical
Discontinuity.’ Linguistics 28(2): 263–307.
Huang, Chu-Ren (1993). ‘Reverse Long-Distance Dependency and Functional
Uncertainty: The Interpretation of Mandarin Questions.’ In Chungmin Lee,
and Boem-Mo Kang (eds.), Language, Information, and Computing, Seoul:
Thaehaksa, 111–120.
Huang, Chu-Ren (1997). ‘Morphological Transparency and Autonomous
Morphology: A Comparative Study of Tough Constructions and
Nominalization.’ In Chinese Languages and Linguistics III: Morphology
and Lexicon, Taipei: Academia Sinica, 369–399.
Huddleston, Rodney (1995). ‘The Case Against a Future Tense in English.’ Studies
in Language 19(2): 399–446.
Huddleston, Rodney and Geoffrey K. Pullum (2002). The Cambridge Grammar of
the English Language. Cambridge, UK: Cambridge University Press.
Huddleston, Rodney and Geoffrey K. Pullum (2005). A Student’s Introduction to
English Grammar. Cambridge, UK: Cambridge University Press.
Hudson, Richard A. (1990). English Word Grammar. Oxford: Blackwell.
Hudson, Richard A. (1992). ‘So-Called ‘Double Objects’ and Grammatical
Relations.’ Language 68(2): 251–276.
Hurst, Peter (2010). ‘The Syntax of Lexical Reciprocal Constructions.’ In Butt
and King (2010), 290–310. URL http://csli-publications.stanford.edu/LFG/15/
lfg10.html.
Hurst, Peter (2012). Reciprocation Strategies and the Syntax of Reciprocal
Constructions. Ph.D. thesis, University of Melbourne.
Iida, Masayo (1987). ‘Case-Assignment by Nominals in Japanese.’ In Iida et al.
(1987), 93–138.
Iida, Masayo, Stephen Wechsler, and Draga Zec (eds.) (1987). Working Papers
in Grammatical Theory and Discourse Structure, Volume 1: Interactions of
Morphology, Syntax, and Discourse. Number 11 in CSLI Lecture Notes.
Stanford: CSLI Publications.
Inkelas, Sharon (1989). Prosodic Constituency in the Lexicon. Ph.D. thesis, Stanford
University.
 references 771

Inkelas, Sharon and Draga Zec (eds.) (1990). The Phonology-Syntax Connection.
Chicago & Stanford: University of Chicago Press and CSLI.
Ishikawa, Akira (1985). Complex Predicates and Lexical Operations in Japanese.
Ph.D. thesis, Stanford University.
Itô, Junko and Armin Mester (2003). ‘Weak Layering and Word Binarity.’ In
Takeru Honma, Masao Okazaki, Toshiyuki Tabata, and Shin-ichi Tanaka (eds.),
A New Century of Phonology and Phonological Theory. A Festschrift for Pro-
fessor Shosuke Haraguchi on the Occasion of His Sixtieth Birthday, Tokyo:
Kaitakusha, 26–65. Slightly revised version of unpublished working paper, Lin-
guistic Research Center, LRC-92-09, University of California, Santa Cruz, 1992.
Izvorski, Roumyana (1993). ‘On WH-Movement and Focus Movement in
Bulgarian.’ In Regine Eckhardt, and Veerle van Geenhoven (eds.), Proceedings
of ConSoLE II. Amsterdam: Sole Publications.
Jackendoff, Ray (1972). Semantic Interpretation in Generative Grammar.
Cambridge, MA: The MIT Press.
Jackendoff, Ray (1977). X̄ Syntax: A Study of Phrase Structure. Cambridge, MA:
The MIT Press.
Jackendoff, Ray (1983). Semantics and Cognition. Cambridge, MA: The MIT Press.
Jackendoff, Ray (1990). Semantic Structures. Cambridge, MA: The MIT Press.
Jackendoff, Ray (2002). Foundations of Language: Brain, Meaning, Grammar,
Evolution. Oxford: Oxford University Press.
Jackendoff, Ray (2010). Meaning and the Lexicon: The Parallel Architecture
1975–2010. Oxford: Oxford University Press.
Jacobs, Joachim (ed.) (1992). Informationsstruktur und Grammatik. Opladen:
Westdeutscher Verlag.
Jacobson, Pauline (1990). ‘Raising as Function Composition.’ Linguistics and
Philosophy 13: 423–475.
Jacobson, Pauline (1992a). ‘The Lexical Entailment Theory of Control and the
Tough Construction.’ In Sag and Szabolcsi (1992), 269–297.
Jacobson, Pauline (1992b). ‘Raising Without Movement.’ In Larson et al. (1992),
149–194.
Jacobson, Pauline and Geoffrey K. Pullum (eds.) (1982). The Nature of Syntactic
Representation. Dordrecht: D. Reidel.
Jaeger, T. Florian and Veronica A. Gerassimova (2002). ‘Bulgarian Word Order
and the Role of the Direct Object Clitic in LFG.’ In Butt and King (2002). URL
http://csli-publications.stanford.edu/LFG/7/lfg02.html.
Jaeggli, Osvaldo (1980). On Some Phonologically Null Elements in Syntax. Ph.D.
thesis, Massachusetts Institute of Technology.
Jakobson, Roman (1932). ‘Structure of the Russian Verb.’ In Linda R. Waugh, and
Morris Halle (eds.), Roman Jakobson: Russian and Slavic Grammar. Studies
1931–1981, Berlin: Mouton, 1–14. Published in 1984.
Jam, R. (1990). ‘A Conservative Extension to the Functional Uncertainty
Notation.’ Unpublished manuscript, Xerox PARC, with authors Ron Kaplan,
John Maxwell, Annie Zaenen, and Mary Dalrymple.
Jansen, Louise and Bruno Di Biase (2015). ‘Acquiring V2 in Declarative Sentences
and Constituent Questions in German as a Second Language.’ In Bettoni and
Di Biase (2015a), 259–273.
772 references

Janssen, Theo M. V. (1997). ‘Compositionality.’ In van Benthem and ter Meulen

(1997), 419–473. Co-published with Elsevier Science B.V., Amsterdam.
Janssen, Theo M. V. (2001). ‘Frege, Contextuality and Compositionality.’ Journal
of Logic, Language and Information 10(1): 115–136.
Jelinek, Eloise (1984). ‘Empty Categories, Case, and Configurationality.’ Natural
Language and Linguistic Theory 2(1): 39–76.
Jelinek, Eloise and Richard A. Demers (1983). ‘The Agent Hierarchy and Voice
in Some Coast Salish Languages.’ International Journal of American Linguistics
49(2): 167–185.
Jeschull, Liane (2004). ‘Coordination in Chechen.’ In Haspelmath (2004a),
241–265.
Johannessen, Janne Bondi (1998). Coordination. Oxford: Oxford University Press.
Johnson, David E. and Paul M. Postal (1980). Arc Pair Grammar. Princeton:
Princeton University Press.
Johnson, Mark (1986). ‘The LFG Treatment of Discontinuity and the Double
Infinitive Construction in Dutch.’ In Mary Dalrymple, Jeffrey Goldberg, Kristin
Hanson, Michael Inman, Chris Piñon, and Stephen Wechsler (eds.), WCCFL 5:
Proceedings of the Fifth West Coast Conference on Formal Linguistics. Stanford:
CSLI Publications, 102–118.
Johnson, Mark (1988). Attribute-Value Logic and the Theory of Grammar.
Number 16 in CSLI Lecture Notes. Stanford: CSLI Publications.
Johnson, Mark (1999). ‘Type-Driven Semantic Interpretation and Feature
Dependencies in R-LFG.’ In Dalrymple (1999).
Johnson, Mark (2002). ‘Optimality-Theoretic Lexical Functional Grammar.’ In
Paola Merlo, and Suzanne Stevenson (eds.), The Lexical Basis of Sentence
Processing: Formal, Computational and Experimental Issues, John Benjamins,
volume 4, 59–73.
Jones, Stephen M. (2015a). Number in Meryam Mir. M.Phil. dissertation,
University of Oxford. URL https://ora.ox.ac.uk/objects/uuid:02b4a4de-3f33-
43b2-943c-e5bf1724e4ce.
Jones, Stephen M. (2015b). ‘Number in Meryam Mir.’ In Butt and King (2015b).
URL http://csli-publications.stanford.edu/LFG/20/lfg15.html.
Joshi, Smita (1993). Selection of Grammatical and Logical Functions in Marathi.
Ph.D. thesis, Stanford University.
Jun, Sun-Ah (ed.) (2005). Prosodic Typology: The Phonology of Intonation and
Phrasing. Oxford: Oxford University Press.
Jun, Sun-Ah (ed.) (2014). Prosodic Typology II: The Phonology of Intonation and
Phrasing. Oxford: Oxford University Press.
Junghanns, Uwe and Gerhild Zybatow (1997). ‘Syntax and Information Structure
of Russian Clauses.’ In Wayles Browne, Ewa Dornisch, Natasha Kondrashova,
and Draga Zec (eds.), Formal Approaches To Slavic Linguistics: The Cornell
Meeting, 1995. FASL, Ann Arbor: Michigan Slavic Association, 289–319.
Kaisse, Ellen M. (1985). Connected Speech: The Interaction of Syntax and
Phonology. Orlando: Academic Press.
Kallas, Krystyna (1993). Składnia współczesnych polskich konstrukcji
współrze˛dnych [The Syntax of Coordinate Constructions in Contemporary
Polish]. Toruń: Wydawnictwo Uniwersytetu Mikołaja Kopernika.
 references 773

Kameyama, Megumi (1985). Zero Anaphora: The Case of Japanese. Ph.D. thesis,
Stanford University.
Kameyama, Megumi (1989). ‘Functional Precedence Conditions on Overt and
Zero Pronominals.’ Technical report, MCC.
Kamp, Hans (1975). ‘Two Theories About Adjectives.’ In Edward L. Keenan (ed.),
Formal Semantics of Natural Language, Cambridge, UK: Cambridge University
Press, 123–155.
Kamp, Hans (1981). ‘A Theory of Truth and Semantic Representation.’ In Jeroen
Groenendijk, Theo M. V. Janssen, and Martin Stokhof (eds.), Formal Methods
in the Study of Language, Amsterdam: Mathematische Centrum, volume I,
277–321. Reprinted in Portner and Partee (2002, 189–222).
Kamp, Hans and Barbara H. Partee (1995). ‘Prototype Theory and
Compositionality.’ Cognition 57(2): 129–191.
Kamp, Hans and Uwe Reyle (1993). From Discourse to Logic: An Introduction
to Modeltheoretic Semantics of Natural Language, Formal Logic and Discourse
Representation Theory. Dordrecht: Kluwer.
Kaplan, Ronald M. (1975a). ‘On Process Models for Sentence Analysis.’ In
Donald A. Norman, and David E. Rumelhart (eds.), Explorations in Cognition,
San Francisco: W. H. Freeman, 117–135.
Kaplan, Ronald M. (1975b). Transient Processing Load in Relative Clauses. Ph.D.
thesis, Harvard University.
Kaplan, Ronald M. (1976). ‘Models of Comprehension Based on Augmented
Transition Networks.’ Presented at the ATT-MIT Convocation on
Communications, March 1976.
Kaplan, Ronald M. (1987). ‘Three Seductions of Computational Psycholinguistics.’
In Peter Whitelock, Mary McGee Wood, Harold L. Somers, Rod Johnson, and
Paul Bennett (eds.), Linguistic Theory and Computer Applications, London:
Academic Press, 149–188. Also published as CCL/UMIST Report No. 86.2:
Alvey/ICL Workshop on Linguistic Theory and Computer Applications: Tran-
scripts of Presentations and Discussions. Center for Computational Linguistics,
University of Manchester. Reprinted in Dalrymple et al. (1995b, 337–367).
Kaplan, Ronald M. (1988). ‘Correspondences and Their Inverses.’ Presented
at the Titisee Workshop on Unification Formalisms: Syntax, Semantics and
Implementation, Titisee, Germany.
Kaplan, Ronald M. (1995). ‘The Formal Architecture of Lexical-Functional
Grammar.’ In Dalrymple et al. (1995b), 7–27. Previously published in
Proceedings of ROCLING II, ed. C.-R. Huang and K.-J. Chen, Tapei 1989,
pp. 1–18, and in Journal of Information Science and Engineering 5 (1989),
pp. 305–322.
Kaplan, Ronald M. (2017). ‘Formal Aspects of Underspecified Features.’ In Cleo
Condoravdi, and Tracy Holloway King (eds.), Tokens of Meaning: Papers in
Honor of Lauri Karttunen, Stanford: CSLI Publications.
Kaplan, Ronald M. and Joan Bresnan (1982). ‘Lexical-Functional Grammar: A
Formal System for Grammatical Representation.’ In Bresnan (1982b), 173–281.
Reprinted in Dalrymple et al. (1995b, 29–130).
Kaplan, Ronald M. and Miriam Butt (2002). ‘The Morphology-Syntax Interface
in LFG.’ URL http://cslipublications.stanford.edu/LFG/7/lfg02.html. Abstract
774 references

in On-Line Proceedings of the LFG2002 Conference, ed. Miriam Butt and Tracy
Holloway King.
Kaplan, Ronald M. and Martin Kay (1994). ‘Regular Models of Phonological Rule
Systems.’ Computational Linguistics 20(3): 331–378.
Kaplan, Ronald M. and Tracy Holloway King (2003). ‘Low-Level Markup and
Large-Scale LFG Grammar Processing.’ In Butt and King (2003b). URL http://
csli-publications.stanford.edu/LFG/8/lfg03.html.
Kaplan, Ronald M. and John T. Maxwell, III (1988a). ‘An Algorithm for Functional
Uncertainty.’ In Proceedings of the International Conference on Computational
Linguistics (COLING88). International Committee on Computational
Linguistics, 297–302. Reprinted in Dalrymple et al. (1995b, 177–197).
Kaplan, Ronald M. and John T. Maxwell, III (1988b). ‘Constituent Coordination in
Lexical-Functional Grammar.’ In Proceedings of the International Conference on
Computational Linguistics (COLING88). International Committee on Compu-
tational Linguistics, 303–305. Reprinted in Dalrymple et al. (1995b, 199–210).
Kaplan, Ronald M. and John T. Maxwell, III (1996). ‘LFG Grammar Writer’s
Workbench.’ Technical report, Xerox Palo Alto Research Center, Palo Alto, CA.
URL http://www2.parc.com/isl/groups/nltt/medley/.
Kaplan, Ronald M., John T. Maxwell, III, Tracy Holloway King, and Richard
Crouch (2004). ‘Integrating Finite-State Technology with Deep LFG
Grammars.’ In Proceedings of the ESSLLI Workshop ‘Combining Shallow
and Deep Processing for NLP’. ESSLLI, 168–173.
Kaplan, Ronald M., John T. Maxwell, III, and Annie Zaenen (1987). ‘Functional
Uncertainty.’ CSLI Monthly Newsletter Stanford University.
Kaplan, Ronald M. and Jürgen Wedekind (1993). ‘Restriction and
Correspondence-Based Translation.’ In Proceedings of the 6th Meeting of the
EACL. Utrecht: European Association for Computational Linguistics, 193–202.
Kaplan, Ronald M. and Jürgen Wedekind (2000). ‘LFG Generation Produces
Context-Free Languages.’ In Proceedings of the International Conference
on Computational Linguistics (COLING2000). International Committee on
Computational Linguistics, 425–431.
Kaplan, Ronald M. and Annie Zaenen (1989a). ‘Functional Precedence and
Constituent Structure.’ In Chu-Ren Huang, and Keh-Jiann Chen (eds.),
ROCLING II: Proceedings of the Computational Linguistics Conference. Taipei,
19–40.
Kaplan, Ronald M. and Annie Zaenen (1989b). ‘Long-Distance Dependencies,
Constituent Structure, and Functional Uncertainty.’ In Mark R. Baltin, and
Anthony S. Kroch (eds.), Alternative Conceptions of Phrase Structure, Chicago:
University of Chicago Press, 17–42. Reprinted in Dalrymple et al. (1995b,
137–165).
Karttunen, Lauri (1976). ‘Discourse Referents.’ In McCawley (1976), 363–386.
Karttunen, Lauri (1977). ‘The Syntax and Semantics of Questions.’ Linguistics and
Philosophy 1: 3–44.
Kasper, Robert (1997). ‘The Semantics of Recursive Modification.’ URL http://
www.essex.ac.uk/linguistics/external/clmt/papers/hpsg/modification.ps.
Unpublished manuscript, The Ohio State University.
 references 775

Kathol, Andreas (1999). ‘Agreement and the Syntax-Morphology Interface

in HPSG.’ In Robert D. Levine, and Georgia M. Green (eds.), Studies in
Contemporary Phrase Structure Grammar, Cambridge, UK: Cambridge
University Press, 223–274.
Kawaguchi, Satomi (2015). ‘The Development of Japanese as a Second Language.’
In Bettoni and Di Biase (2015a), 149–172.
Kay, Martin (1979). ‘Functional Grammar.’ In Christine Chiarello, John Kingston,
Eve E. Sweetser, James Collins, Haruko Kawasaki, John Manley-Buser,
Dorothy W. Marschak, Catherine O’Connor, David Shaul, Marta Tobey, Henry
Thompson, and Katherine Turner (eds.), Proceedings of the 5th Annual Meeting
of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society,
142–158.
Kay, Paul (2002). ‘An Informal Sketch of a Formal Architecture for Construction
Grammar.’ Grammars 5(1): 1–19.
Kay, Paul and Charles J. Fillmore (1999). ‘Grammatical Constructions and Linguis-
tic Generalizations: The What’s X Doing Y? Construction.’ Language 75: 1–33.
Keenan, Edward L. and Bernard Comrie (1977). ‘Noun Phrase Accessibility and
Universal Grammar.’ Linguistic Inquiry 8(1): 63–99.
Keenan, Edward L. and Leonard M. Faltz (1985). Boolean Semantics for Natural
Language. Dordrecht: D. Reidel.
Keenan, Edward L. and Dag Westerståhl (2011). ‘Generalized Quantifiers in
Linguistics and Logic.’ In van Benthem and ter Meulen (2011), 859–910.
Kehler, Andrew, Mary Dalrymple, John Lamping, and Vijay A. Saraswat (1995).
‘The Semantics of Resource-Sharing in Lexical-Functional Grammar.’ In
Proceedings of the 7th Conference of the European ACL. Dublin: European
Association for Computational Linguistics, 31–38. Reprinted in Dalrymple
(1999, 191–208).
Keller, Bill (1993). Feature Logics, Infinitary Descriptions and Grammar. Number 44
in CSLI Lecture Notes. Stanford: CSLI Publications.
Kelling, Carmen (2003). ‘French Psych Verbs and Derived Nouns.’ In Butt and
King (2003a), 151–179.
Kelling, Carmen (2006). ‘Spanish se-Constructions: The Passive and the
Impersonal Construction.’ In Butt and King (2006b). URL http://csli-
publications.stanford.edu/LFG/11/lfg06.html.
Kendall, Tyler, Joan Bresnan, and Gerard Van Herk (2011). ‘The Dative Alternation
in African American English: Researching Syntactic Variation and Change
Across Sociolinguistic Datasets.’ Corpus Linguistics and Linguistic Theory 7(2):
229–244.
Kennedy, Christopher (1997). Projecting the Adjective: The Syntax and Semantics
of Gradability and Comparison. Ph.D. thesis, University of California, Santa
Cruz. Reprinted by Garland Press, New York, 1999.
Kibort, Anna (2001). ‘The Polish Passive and Impersonal in Lexical Mapping
Theory.’ In Butt and King (2001a). URL http://csli-publications.stanford.edu/
LFG/6/lfg01.html.
Kibort, Anna (2004). Passive and Passive-Like Constructions in English and Polish.
Ph.D. thesis, University of Cambridge.
776 references

Kibort, Anna (2006). ‘On Three Different Types of Subjectlessness and How to
Model Them in LFG.’ In Butt and King (2006b). URL http://csli-publications.
stanford.edu/LFG/11/lfg06.html.
Kibort, Anna (2007). ‘Extending the Applicability of Lexical Mapping Theory.’
In Butt and King (2007), 250–270. URL http://csli-publications.stanford.edu/
LFG/12/lfg07.html.
Kibort, Anna (2008). ‘On the Syntax of Ditransitive Constructions.’ In Butt and
King (2008). URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Kibort, Anna (2012). ‘Participles, Adjectives, and the Role of Argument Structure.’
In Butt and King (2012). URL http://csli-publications.stanford.edu/LFG/17/
lfg12.html.
Kibort, Anna and Joan Maling (2015). ‘Modelling the Syntactic Ambiguity of the
Active vs. Passive Impersonal in LFG.’ In Butt and King (2015b). URL http://
csli-publications.stanford.edu/LFG/20/lfg15.html.
Kim, Boomee (1995). ‘Predication in Tough Constructions.’ In José Camacho,
Lina Choueri, and Maki Watanabe (eds.), WCCFL 14: Proceedings of the 14th
West Coast Conference on Formal Linguistics. Stanford: CSLI Publications,
271–285.
King, Tracy Holloway (1995). Configuring Topic and Focus in Russian. Disser-
tations in Linguistics. Stanford: CSLI Publications. Revised version of 1993
Stanford University dissertation.
King, Tracy Holloway (1997). ‘Focus Domains and Information-Structure.’ In Butt
and King (1997). URL http://csli-publications.stanford.edu/LFG/LFG2-1997/.
King, Tracy Holloway (2005). ‘Cliticizing LFG.’ In Butt and King (2005). URL
http://csli-publications.stanford.edu/LFG/10/lfg05.html.
King, Tracy Holloway and Mary Dalrymple (2004). ‘Determiner Agreement and
Noun Conjunction.’ Journal of Linguistics 40(1): 69–104.
King, Tracy Holloway and Valeria de Paiva (eds.) (2013). From Quirky Case to Rep-
resenting Space: Papers in Honor of Annie Zaenen. Stanford: CSLI Publications.
King, Tracy Holloway, Martin Forst, Jonas Kuhn, and Miriam Butt (2005).
‘The Feature Space in Parallel Grammar Writing.’ Research on Language and
Computation 3(2-3): 139–163.
King, Tracy Holloway and Annie Zaenen (2004). ‘F-Structures, Information
Structure, and Discourse Structure.’ URL http://cslipublications.stanford.edu/
LFG/9/abstracts/lfg04kingzaenen-abs.html. Presented at the Winter School in
LFG and Computational Linguistics, University of Canterbury, Christchurch,
New Zealand.
Kiparsky, Paul (1982). ‘Word-Formation and the Lexicon.’ In F. Ingemann (ed.),
Proceedings of the 1982 Mid-America Linguistics Conference.
Kiparsky, Paul (1998). ‘Partitive Case and Aspect.’ In Miriam Butt, and Wilhelm
Geuder (eds.), The Projection of Arguments: Lexical and Compositional Factors,
Stanford: CSLI Publications, 265–307.
Kiparsky, Paul and Carol Kiparsky (1970). ‘Fact.’ In Manfred Bierwisch, and
Karl E. Heidolph (eds.), Progress in Linguistics, The Hague: Mouton de Gruyter,
143–173.
Kiss, Tibor (1995). ‘Lexical-Functional Grammar.’ In Joachim Jacobs, Arnim
von Stechow, Wolfgang Sternefeld, and Theo Vennemann (eds.), Syntax:
 references 777

Ein Internationales Handbuch Zeitgenössischer Forschung [An International

Handbook Of Contemporary Research], Berlin: de Gruyter, volume Band 9.1 of
Handbücher zur Sprach- und Kommunikationswissenschaft, 581–601.
Kiss, Tibor and Artemis Alexiadou (eds.) (2015). Syntax – Theory and Analysis:
An International Handbook, volume 42.1 of Handbooks of Linguistics and
Communication Science. Berlin: Mouton de Gruyter.
É. Kiss, Katalin (2002). The Syntax of Hungarian. Cambridge, UK: Cambridge
University Press.
Klein, Wolfgang (1992). ‘The Present Perfect Puzzle.’ Language 68(3): 525–552.
Klein, Wolfgang (1994). Time in Language. London: Routledge.
Klein, Wolfgang (1995). ‘A Time-Relational Analysis of Russian Aspect.’ Language
71(4): 669–695.
Kokkonidis, Miltiadis (2005). ‘Why Glue a Donkey to an F-Structure, When You
Can Constrain and Bind It Instead?’ In Butt and King (2005). URL http://csli-
publications.stanford.edu/LFG/10/lfg05.html.
Kokkonidis, Miltiadis (2007). ‘First-Order Glue.’ Journal of Logic, Language and
Information 17: 43–68.
Koster, Jan (1978). ‘Why Subject Sentences Don’t Exist.’ In S. Jay Keyser (ed.),
Recent Transformational Studies in European Languages, Cambridge, MA: The
MIT Press, 53–64.
Krifka, Manfred (1989). ‘Nominal Reference, Temporal Constitution and
Quantification in Event Semantics.’ In Renate Bartsch, Johan van Benthem, and
Peter van Emde Boas (eds.), Semantics and Contextual Expression, Dordrecht:
Foris, 75–115.
Krifka, Manfred (1990). ‘Boolean and Non-Boolean ‘And’.’ In Lázló Kálmán, and
Lázló Pólos (eds.), Papers From the Second Symposium on Logic and Language.
Budapest: Akadémiai Kiadó, 161–188.
Krifka, Manfred (1991). ‘A Compositional Semantics for Multiple Focus
Constructions.’ In Steven Moore, and Adam Zachary Wyner (eds.), SALT 1:
Proceedings of the First Semantics and Linguistic Theory Conference. Cornell
University: Cornell University Working Papers in Linguistics, 127–158.
Reprinted in Jacobs (1992).
Krifka, Manfred (2001). ‘A Framework for Focus-Sensitive Quantification.’ In
Caroline Fery, and Wolfgang Sternefeld (eds.), Audiatur Vox Sapientiae. A
Festschrift for Arnim von Stechow, Berlin: Akademie Verlag, 287–319.
Krifka, Manfred (2006). ‘Association With Focus Phrases.’ In Valerie Molnar, and
Susanne Winkler (eds.), The Architecture of Focus, Berlin: Mouton de Gruyter,
105–136.
Krifka, Manfred (2008). ‘Basic Notions of Information Structure.’ Acta Linguistica
Hungarica 55(3–4): 243–276.
Kroeger, Paul (1993). Phrase Structure and Grammatical Relations in Tagalog.
Dissertations in Linguistics. Stanford: CSLI Publications. Revised version of
1991 Stanford University dissertation.
Kroeger, Paul (2004). Analyzing Syntax: A Lexical-Functional Approach.
Cambridge, UK: Cambridge University Press.
Kruijff-Korbayová, Ivana and Mark J. Steedman (2003). ‘Discourse and
Information Structure.’ Journal of Logic, Language and Information 12: 249–259.
778 references

Krylova, O. and S. Khavronina (1988). Word Order in Russian. Russky Yazyk

Publishers.
Kuhn, Jonas (1999). ‘Towards a Simple Architecture for the Structure-Function
Mapping.’ In Butt and King (1999). URL http://csli-publications.stanford.edu/
LFG/LFG4-1999/.
Kuhn, Jonas (2001a). Formal and Computational Aspects of Optimality-Theoretic
Syntax. Ph.D. thesis, Universität Stuttgart.
Kuhn, Jonas (2001b). ‘Generation and Parsing in Optimality Theoretic Syntax:
Issues in the Formalization of OT-LFG.’ In Sells (2001a), 313–366.
Kuhn, Jonas (2001c). ‘Resource Sensitivity in the Syntax-Semantics Interface
and the German Split NP Construction.’ In W. Detmar Meurers, and Tibor
Kiss (eds.), Constraint-Based Approaches to Germanic Syntax, Stanford: CSLI
Publications, number 7 in Studies in Constraint-Based Lexicalism, 177–215.
Kuhn, Jonas, Christian Rohrer, and Sina Zarrieß (2010). ‘Right Node Raising in
Parsing and Generation.’ In Butt and King (2010). URL http://csli-publications.
stanford.edu/LFG/15/lfg10.html.
Kuhn, Jonas and Louisa Sadler (2007). ‘Single Conjunct Agreement and the
Formal Treatment of Coordination in LFG.’ In Butt and King (2007). URL
http://csli-publications.stanford.edu/LFG/12/lfg07.html.
Kuiper, Koenraad and Jacqui Nokes (2013). Theories of Syntax: Concepts and Case
Studies. Basingstoke: Palgrave Macmillan.
Kuno, Susumu (1972). ‘Functional Sentence Perspective: A Case Study from
Japanese and English.’ Linguistic Inquiry 3: 269–320.
Kuno, Susumu and Jane J. Robinson (1972). ‘Multiple WH-Questions.’ Linguistic
Inquiry 3(4): 463–487.
Kwon, Song-Nim and Anne Zribi-Hertz (2008). ‘Differential Function Marking,
Case and Information Structure: Evidence From Korean.’ Language 84(2):
218–257.
Laczkó, Tibor (1995). The Syntax of Hungarian Noun Phrases: A Lexical-Functional
Approach. Frankfurt: Peter Lang.
Laczkó, Tibor (2000). ‘Derived Nominals, Possessors, and Lexical Mapping
Theory.’ In Butt and King (2000a).
Laczkó, Tibor (2003). ‘On Oblique Arguments and Adjuncts of Hungarian Event
Nominals.’ In Butt and King (2003a), 201–234.
Laczkó, Tibor (2004). ‘Grammatical Functions, LMT, and Control in the
Hungarian DP Revisited.’ In Butt and King (2004b). URL http://csli-
publications.stanford.edu/LFG/9/lfg04.html.
Laczkó, Tibor (2007). ‘Revisiting Possessors in Hungarian DPs: A New
Perspective.’ In Butt and King (2007), 343–362. URL http://csli-publications.
stanford.edu/LFG/12/lfg07.html.
Laczkó, Tibor (2012). ‘On the (Un)bearable Lightness of Being an LFG-Style
Copula in Hungarian.’ In Butt and King (2012), 341–361. URL http://csli-
publications.stanford.edu/LFG/17/lfg12.html.
Laczkó, Tibor (2013). ‘On Presenting Something in English and Hungarian.’ In
King and de Paiva (2013), 181–193.
 references 779

Laczkó, Tibor (2014a). ‘Essentials of an LFG Analysis of Hungarian Finite

Sentences.’ In Butt and King (2014). URL http://csli-publications.stanford.edu/
LFG/19/lfg14.html.
Laczkó, Tibor (2014b). ‘Focus and Verbal Modifiers in Hungarian from an LFG
Perspective.’ In Butt and King (2014). URL http://csli-publications.stanford.
edu/LFG/19/lfg14.html.
Laczkó, Tibor (2014c). ‘Outlines of an LFG-XLE Account of Negation in
Hungarian Sentences.’ In Butt and King (2014). URL http://csli-publications.
stanford.edu/LFG/19/lfg14.html.
Laczkó, Tibor (2015). ‘On Negative Particles and Negative Polarity in Hungarian.’
In Butt and King (2015b). URL http://csli-publications.stanford.edu/LFG/20/
lfg15.html.
Laczkó, Tibor (2016). Aspects of Hungarian Syntax from a Lexicalist Perspective.
Doctor of the Hungarian Acadademy of Sciences thesis, University of
Debrecen. URL http://real-d.mtak.hu/960/.
Ladd, D. Robert (2008). Intonational Phonology. Cambridge: Cambridge
University Press, second edition.
Lahiri, Aditi and Frans Plank (2010). ‘Phonological Phrasing in Germanic: The
Judgement of History, Confirmed Through Experiment.’ Transactions of the
Philological Society 108(3): 370–398.
Lakoff, George (1970). ‘Global Rules.’ Language 46(3): 627–639.
Lakoff, George and John Robert Ross (1976). ‘Why You Can’t Do So Into the Sink.’
In McCawley (1976), 101–111.
Lam, Olivia S.-C. (2008). Object Functions and the Syntax of Double Object Con-
structions in Lexical Functional Grammar. D.Phil. thesis, University of Oxford.
Lambek, Joachim (1958). ‘The Mathematics of Sentence Structure.’ American
Mathematical Monthly 65: 154–170.
Lambrecht, Knud (1994). Information Structure and Sentence Form: Topic,
Focus, and the Mental Representation of Discourse Referents. Cambridge, UK:
Cambridge University Press.
Landau, Idan (2000). Elements of Control: Structure and Meaning in Infinitival
Constructions. Dordrecht: Kluwer.
Landau, Idan (2004). ‘The Scale of Finiteness and the Calculus of Control.’ Natural
Language and Linguistic Theory 22(4): 811–877.
Landman, Fred (1989). ‘Groups.’ Linguistics and Philosophy 12: 559–605, 723–744.
Lapata, Maria (1998). ‘Anaphoric Binding in Modern Greek.’ In Butt and King
(1998b). URL http://csli-publications.stanford.edu/LFG/LFG3-1998/.
Lapointe, Steven (1980). A Theory of Grammatical Agreement. Ph.D. thesis,
University of Massachusetts, Amherst.
Lappin, Shalom (ed.) (1996). The Handbook of Contemporary Semantic Theory.
Oxford: Blackwell.
Larson, Richard K. (1998). ‘Events and Modification in Nominals.’ In Devon
Strolovitch, and Aaron Lawson (eds.), SALT 8: Proceedings of the Eighth
Semantics and Linguistic Theory Conference.
Larson, Richard K., Sabine Iatridou, Utpal Lahiri, and James Higginbotham (eds.)
(1992). Control and Grammar. Dordrecht: Kluwer.
780 references

Lasersohn, Peter (1995). Plurality, Conjunction and Events. Dordrecht: Kluwer.

Lasnik, Howard (1976). ‘Remarks on Coreference.’ Linguistic Analysis 2: 1–22.
Lasnik, Howard and Robert Fiengo (1989). ‘Complement Object Deletion.’ In
Essays on Anaphora, Dordrecht: Kluwer. Originally in Linguistic Inquiry 5
(1974), 535–571.
Laughren, Mary (2002). ‘Syntactic Constraints in a “Free Word Order” Language.’
In Mengistu Amberber, and Peter Collins (eds.), Language Universals and
Variation, Westport, CO: Praeger, 83–130.
Lawyer, Lewis (2010). ‘Walman And-Verbs and the Nature of Walman
Serialization.’ In Butt and King (2010), 332–352. URL http://csli-publications.
stanford.edu/LFG/15/lfg10.html.
Lee, Hanjung (2001). ‘Markedness and Word Order Freezing.’ In Sells (2001a).
Lee, Hanjung (2003). ‘Parallel Optimization in Case Systems.’ In Butt and King
(2003a), 15–58.
Lee, Hanjung (2006). ‘Parallel Optimization in Case Systems: Evidence From Case
Ellipsis in Korean.’ Journal of East Asian Linguistics 15(1): 69–96.
Legate, Julie Anne (2002). Warlpiri: Theoretical Implications. Ph.D. thesis,
Massachusetts Institute of Technology.
Levelt, Willem J. M. (1989). Speaking: From Intention to Articulation. Cambridge,
MA: The MIT Press.
Levelt, Willem J. M., Ardi Roelofs, and Antje S. Meyer (1999). ‘A Theory of Lexical
Access in Speech Production.’ Behavioral and Brain Sciences 22: 1–75.
Levin, Beth (1993). English Verb Classes and Alternations: A Preliminary
Investigation. Chicago: University of Chicago Press.
Levin, Beth and Malka Rappaport Hovav (2005). Argument Realization.
Cambridge, UK: Cambridge University Press.
Levin, Lori S. (1982). ‘Sluicing.’ In Bresnan (1982b), 590–654.
Levin, Lori S. (1986). Operations on Lexical Forms: Unaccusative Rules in Germanic
Languages. Ph.D. thesis, Massachusetts Institute of Technology. Reprinted by
Garland Press, New York, 1988.
Levin, Lori S. (1987). ‘Toward a Linking Theory of Relation-Changing Rules in
LFG.’ Technical Report CSLI-87-115, CSLI Publications, Stanford.
Levin, Lori S., Malka Rappaport, and Annie Zaenen (eds.) (1983). Papers in
Lexical-Functional Grammar. Bloomington: Indiana University Linguistics
Club.
Levine, Robert D. (2017). Syntactic Analysis: An HPSG-Based Approach.
Cambridge, UK: Cambridge University Press.
Li, Charles N. (ed.) (1976). Subject and Topic. New York: Academic Press.
Liao, Gwen W.-T. (2010). An LFG Account of Empty Pronouns in Mandarin
Chinese. D.Phil. thesis, University of Oxford.
Liberman, Mark (1975). The Intonational System of English. Ph.D. thesis,
Massachusetts Institute of Technology.
Liberman, Mark and Alan S. Prince (1977). ‘On Stress and Linguistic Rhythm.’
Linguistic Inquiry 8: 249–336.
Link, Godehard (1983). ‘The Logical Analysis of Plurals and Mass Terms: A
Lattice-Theoretical Approach.’ In Bäuerle et al. (1983), 302–323. Reprinted in
Link (1998, 11–34).
 references 781

Link, Godehard (1998). Algebraic Semantics in Language and Philosophy. Stanford:

CSLI Publications.
Lipson, Alexander (1981). A Russian Course. Columbus, OH: Slavica.
Lødrup, Helge (1999a). ‘Inalienables in Norwegian and Binding Theory.’
Linguistics 37(3): 365–388.
Lødrup, Helge (1999b). ‘Linking and Optimality in the Norwegian Presentational
Focus Construction.’ Nordic Journal of Linguistics 22(2): 205–229.
Lødrup, Helge (2000). ‘Underspecification in Lexical Mapping Theory: The Case
of Norwegian Existentials and Resultatives.’ In Butt and King (2000a).
Lødrup, Helge (2001). ‘Pseudocoodinations in Norwegian and Control Theory.’
In Butt and King (2001a). URL http://csli-publications.stanford.edu/LFG/6/
lfg01.html.
Lødrup, Helge (2002a). ‘Infinitival Complements and the Form-Function
Relation.’ In Butt and King (2002). URL http://csli-publications.stanford.edu/
LFG/7/lfg02.html.
Lødrup, Helge (2002b). ‘The Syntactic Structures of Norwegian Pseudo-
coordinations.’ Studia Linguistica 56(2): 121–143.
Lødrup, Helge (2007). ‘A New Account of Simple and Complex Reflexives.’ Journal
of Comparative Germanic Linguistics 10(3): 183–201.
Lødrup, Helge (2008a). ‘Local Binding Without Coargumenthood: Norwegian
Noun Phrases.’ In Butt and King (2008), 333–351. URL http://csli-publications.
stanford.edu/LFG/13/lfg08.html.
Lødrup, Helge (2008b). ‘Raising to Object in Norwegian and the Derived Object
Constraint.’ Studia Linguistica 62(2): 155–181.
Lødrup, Helge (2009). ‘Looking Possessor Raising in the Mouth: Norwegian
Possessor Raising With Unergatives.’ In Butt and King (2009). URL http://csli-
publications.stanford.edu/LFG/14/lfg09.html.
Lødrup, Helge (2011a). ‘Lexical-Functional Grammar: Functional Structure.’ In
Borsley and Börjars (2011), chapter 4.
Lødrup, Helge (2011b). ‘Norwegian Possessive Pronouns: Phrases, Words or
Suffixes?’ In Butt and King (2011), 339–359. URL http://csli-publications.
stanford.edu/LFG/16/lfg11.html.
Lødrup, Helge (2012). ‘In Search of a Nominal COMP.’ In Butt and King (2012),
341–361. URL http://csli-publications.stanford.edu/LFG/17/lfg12.html.
Lødrup, Helge (2017a). ‘Norwegian Pseudocoordinations with the Verb drive
‘Carry On’: Control, Raising, Grammaticalization.’ In Butt and King (2017).
URL http://web.stanford.edu/group/cslipublications/cslipublications/LFG/
LFG-2017/index.shtml.
Lødrup, Helge (2017b). ‘Verbal Present Participles in Norwegian: Controlled
Complements or Parts of Complex Predicates.’ In Arnold et al. (2017), 380–400.
URL http://cslipublications.stanford.edu/HPSG/2016/.
Lorenz, David (2012). Contractions of English Semi-Modals: The Emancipating
Effect of Frequency. Ph.D. thesis, Albert-Ludwigs-Universität, Freiburg.
Lorenz, David (2013). ‘From Reduction to Emancipation: Is Gonna a Word?’
In Hilde Hasselgård, Jarle Ebeling, and Signe Oksefjell Ebeling (eds.),
Corpus Perspectives on Patterns of Lexis, Amsterdam: John Benjamins,
133–152.
782 references

Lovestrand, Joseph and John J. Lowe (2017). ‘Minimal C-Structure: Rethinking

Projection in Phrase Structure.’ In Butt and King (2017). URL http://web.
stanford.edu/group/cslipublications/cslipublications/LFG/LFG-2017/index.
shtml.
Lowe, John J. (2011). ‘R.gvedic Clitics and ‘Prosodic Movement’.’ In Butt and King
(2011). URL http://csli-publications.stanford.edu/LFG/16/lfg11.html.
Lowe, John J. (2012). The Syntax and Semantics of Tense-Aspect Stem Participles
in Early R.gvedic Sanskrit. D.Phil. thesis, University of Oxford. URL https://ora.
ox.ac.uk/objects/uuid:45516bfd-cffb-457a-904c-100695cbd938.
Lowe, John J. (2013). ‘(De)selecting Arguments for Transitive and Predicated
Nominals.’ In Butt and King (2013). URL http://csli-publications.stanford.edu/
LFG/18/lfg13.html.
Lowe, John J. (2014a). ‘Accented Clitics in the R.gveda.’ Transactions of the
Philological Society 112(1): 5–43.
Lowe, John J. (2014b). ‘Gluing Meanings and Semantic Structures.’ In Butt and
King (2014). URL http://csli-publications.stanford.edu/LFG/19/lfg14.html.
Lowe, John J. (2014c). ‘Transitive Nominals in Old Avestan.’ Journal of the
American Oriental Society 134(4): 553–578.
Lowe, John J. (2015a). ‘Complex Predicates: An LFG+Glue Analysis.’ Journal of
Language Modelling 2(3): 413–462. URL http://jlm.ipipan.waw.pl/index.php/
JLM/article/view/125.
Lowe, John J. (2015b). ‘Degrees of Degrammaticalization: A Lexical Sharing
Approach to the English Possessive.’ In Butt and King (2015b). URL http://csli-
publications.stanford.edu/LFG/20/lfg15.html.
Lowe, John J. (2015c). Participles in the Rigveda: The Syntax and Semantics of
Adjectival Verb Forms. Oxford: Oxford University Press.
Lowe, John J. (2015d). ‘The Syntax of Sanskrit Compounds.’ Language 91(3):
e71–e115.
Lowe, John J. (2016a). ‘Clitics: Separating Syntax and Prosody.’ Journal of
Linguistics 52(2): 375–419.
Lowe, John J. (2016b). ‘English Possessive ’s: Clitic and Affix.’ Natural Language
and Linguistic Theory 34(1): 157–195.
Lowe, John J. (2017a). ‘Participles, Gerunds and Syntactic Categories.’ In Arnold
et al. (2017), 401–421. URL http://cslipublications.stanford.edu/HPSG/2016/.
Lowe, John J. (2017b). Transitive Nouns and Adjectives: Evidence from Early
Indo-Aryan. Oxford Studies in Diachronic and Historical Linguistics. Oxford:
Oxford University Press.
Lowe, John J. and Oleg I. Belyaev (2015). ‘Clitic Positioning in Ossetic.’ In Butt and
King (2015b). URL http://csli-publications.stanford.edu/LFG/20/lfg15.html.
Lowe, John J. and Louise Mycock (2014). ‘Issues in the Representation of Informa-
tion Structure.’ Presented at the workshop The Syntax and Information Structure
of Unbounded Dependencies, University of Michigan, Ann Arbor, USA.
Mahowald, Kyle (2011). ‘An LFG Approach to Word Order Freezing.’ In Butt and
King (2011). URL http://csli-publications.stanford.edu/LFG/16/lfg11.html.
Maling, Joan (1983). ‘Transitive Adjectives: A Case of Categorial Reanalysis.’ In
Frank Heny, and Barry Richards (eds.), Linguistic Categories: Auxiliaries and
Related Puzzles, Dordrecht: D. Reidel, volume 1, 253–289.
 references 783

Maling, Joan and Sigríður Sigurjónsdóttir (2013). ‘Nothing Personal? A System-

internal Syntactic Change in Icelandic.’ In King and de Paiva (2013), 109–126.
Maling, Joan and Annie Zaenen (1982). ‘A Phrase Structure Account of Scandi-
navian Extraction Phenomena.’ In Jacobson and Pullum (1982), 229–282.
Maling, Joan and Annie Zaenen (eds.) (1990). Modern Icelandic Syntax, volume 24
of Syntax and Semantics. San Diego: Academic Press.
Manning, Christopher D. (1996a). ‘Argument Structure as a Locus for Binding
Theory.’ In Butt and King (1996a). URL http://csli-publications.stanford.edu/
LFG/LFG1-1996/.
Manning, Christopher D. (1996b). Ergativity: Argument Structure and
Grammatical Relations. Dissertations in Linguistics. Stanford: CSLI Publi-
cations. Revised version of 1994 Stanford University dissertation.
Mansouri, Fethi (2005). ‘Agreement Morphology in Arabic as a Second Language:
Typological Features and Their Processing Implications.’ In Pienemann
(2005a), 117–153.
Marcotte, Jean-Philippe (2014). ‘Syntactic Categories in the Correspondence
Architecture.’ In Butt and King (2014). URL http://csli-publications.stanford.
edu/LFG/19/lfg14.html.
Marcotte, Jean-Philippe and Kateryna Kent (2010). ‘Russian Verbal Affixes and
the Projection Architecture.’ In Butt and King (2010), 353–373. URL http://
csli-publications.stanford.edu/LFG/15/lfg10.html.
Markantonatou, Stella (1992). The Syntax of the Modern Greek Noun Phrases With
a Deverbal Nominal Head. Ph.D. thesis, University of Essex.
Markantonatou, Stella (1995). ‘Modern Greek Deverbal Nominals: An LMT
Approach.’ Journal of Linguistics 31: 267–299.
Markantonatou, Stella and Louisa Sadler (1995). ‘Linking Indirect Arguments.’
Technical Report 49, University of Essex.
Mathesius, Vilém (1983). ‘Functional Linguistics.’ In Josef Vachek, and Libuše
Dušková (eds.), Praguiana: Some Basic and Less Known Aspects of the Prague
Linguistic School, Amsterdam: John Benjamins, 121–142. First published in
1929.
Matsumoto, Yo (1996). Complex Predicates in Japanese: A Syntactic and Semantic
Study of the Notion ‘Word’. Studies in Japanese Linguistics. Stanford & Tokyo:
CSLI and Kuroiso Publishers. Revised version of 1992 Stanford University
dissertation, On the Wordhood of Complex Predicates in Japanese.
Matsumoto, Yo (1998). ‘A Reexamination of the Cross-Linguistic Parameterization
of Causative Predicates: Japanese Perspectives.’ In Butt and King (1998b). URL
http://csli-publications.stanford.edu/LFG/LFG3-1998/.
Matsumoto, Yo (2000). ‘Crosslinguistic Parameterization of Causative Predicates.’
In Butt and King (2000a).
Maxwell, John T., III and Ronald M. Kaplan (1991). ‘A Method for Disjunctive
Constraint Satisfaction.’ In Masaru Tomita (ed.), Current Issues in Parsing
Technology, Dordrecht: Kluwer, 173–190. Reprinted in Dalrymple et al. (1995b,
381–401).
Maxwell, John T., III and Ronald M. Kaplan (1993). ‘The Interface Between
Phrasal and Functional Constraints.’ Computational Linguistics 19(4):
571–590.
784 references

Maxwell, John T., III and Ronald M Kaplan (1996). ‘Unification-Based Parsers
That Automatically Take Advantage of Context Freeness.’ In Butt and King
(1996a). URL http://csli-publications.stanford.edu/LFG/LFG1-1996/.
Maxwell, John T., III and Christopher D. Manning (1996). ‘A Theory of Non-
Constituent Coordination Based on Finite-State Rules.’ In Butt and King
(1996a). URL http://csli-publications.stanford.edu/LFG/LFG1-1996/. Also
presented at the Symposium on Coordination, 1996 European Summer School
on Logic, Language, and Information, Prague.
May, Robert (1977). The Grammar of Quantification. Ph.D. thesis, Massachusetts
Institute of Technology.
Mayer, Elisabeth (2006). ‘Optional Direct Object Clitic Doubling in Limeño
Spanish.’ In Butt and King (2006b). URL http://csli-publications.stanford.edu/
LFG/11/lfg06.html.
Mayer, Elisabeth (2008). ‘Clitics on the Move: From Dependent Marking to Split
Marking.’ In Butt and King (2008). URL http://csli-publications.stanford.edu/
LFG/13/lfg08.html.
McCawley, James D. (1968). ‘Concerning the Base Component of a
Transformational Grammar.’ Foundations of Language 4: 243–269. Reprinted
in McCawley (1973).
McCawley, James D. (1973). Grammar and Meaning: Papers on Syntactic and
Semantic Topics. Tokyo: Taishukan.
McCawley, James D. (ed.) (1976). Notes From the Linguistic Underground, volume 7
of Syntax and Semantics. New York: Academic Press.
McCawley, James D. (1982). ‘Parentheticals and Discontinuous Constituent
Structure.’ Linguistic Inquiry 13(1): 91–106.
McCawley, James D. (1988). The Syntactic Phenomena of English. Chicago:
University of Chicago Press.
McCloskey, James (1979). Transformational Syntax and Model Theoretic Semantics:
A Case Study in Modern Irish. Dordrecht: D. Reidel.
McCloskey, James (1990). ‘Resumptive Pronouns, A# Binding, and Levels of
Representation in Irish.’ In Randall Hendrick (ed.), The Syntax of the Modern
Celtic Languages, New York: Academic Press, volume 23 of Syntax and
Semantics, 199–256.
McCloskey, James (2002). ‘Resumption, Successive Cyclicity, and the Locality of
Operations.’ In Samuel David Epstein, and T. Daniel Seely (eds.), Derivation
and Explanation in the Minimalist Program, Malden, MA: Blackwell,
184–226.
McConnell-Ginet, Sally (1979). ‘On the Deep (And Surface) Adjective ‘Good’.’ In
Frans Van Coetsem, and Linda R. Waugh (eds.), Contributions to Grammatical
Studies: Semantics and Syntax, Leiden: Brill, 132–150.
McGregor, R. S. (1972). Outline of Hindi Grammar. Oxford & Delhi: Oxford
University Press.
Mel’čuk, Igor A. (1988). Dependency Syntax: Theory and Practice. Albany: State
University of New York Press.
Miliĉević, Maja (2009). ‘On the Status of Clitic Reflexives and Reciprocals in
Italian and Serbian.’ In Butt and King (2009), 441–458. URL http://csli-
publications.stanford.edu/LFG/14/lfg09.html.
 references 785

Milward, David (1994). ‘Non-Constituent Coordination: Theory and Practice.’ In

Proceedings of the International Conference on Computational Linguistics (COL-
ING94). International Committee on Computational Linguistics, 935–941.
Mittendorf, Ingo and Louisa Sadler (2005). ‘Numerals, Nouns and Number in
Welsh NPs.’ In Butt and King (2005). URL http://csli-publications.stanford.
edu/LFG/10/lfg05.html.
Mittendorf, Ingo and Louisa Sadler (2008). ‘NP Would Like to Meet GF: A Welsh
Adjectival Construction.’ In Butt and King (2008), 373–393. URL http://csli-
publications.stanford.edu/LFG/13/lfg08.html.
Mittendorf, Ingo and Louisa Sadler (2011). ‘Welsh Prenominals: at the Syntax-
Morphology Interface.’ In Olivier Bonami, and Patricia Cabredo Hofherr (eds.),
Empirical Issues in Syntax and Semantics 8. Colloque de Syntaxe et Sémantique
à Paris, 383–408. URL http://www.cssp.cnrs.fr/eiss8.
Mohanan, K. P. (1982). ‘Grammatical Relations and Clause Structure in
Malayalam.’ In Bresnan (1982b), 504–589.
Mohanan, K. P. (1983). ‘Functional and Anaphoric Control.’ Linguistic Inquiry
14(4): 641–674.
Mohanan, K. P. and Tara Mohanan (2003). ‘Universal and Language Particular
Constraints in OT-LFG.’ In Butt and King (2003b). URL http://csli-
publications.stanford.edu/LFG/8/lfg03.html.
Mohanan, Tara (1994). Argument Structure in Hindi. Dissertations in Linguistics.
Stanford: CSLI Publications. Reprinted version of 1990 Stanford University
dissertation.
Mohanan, Tara (1995). ‘Wordhood and Lexicality: Noun Incorporation in Hindi.’
Natural Language and Linguistic Theory 13(1): 75–134.
Moltmann, Friederike (1992). Coordination and Comparatives. Ph.D. thesis,
Massachusetts Institute of Technology.
Montague, Richard (1970a). ‘English as a Formal Language.’ In Bruno Visentini,
et al. (eds.), Linguaggi Nella Societá e Nella Tecnica, Milan: Edizioni di
Communità, 189–224. Reprinted in Thomason (1974, 188–221).
Montague, Richard (1970b). ‘Universal Grammar.’ Theoria 36: 373–398.
Montague, Richard (1973). ‘The Proper Treatment of Quantification in Ordinary
English.’ In Jaakko Hintikka, Julius Moravcsik, and Patrick Suppes (eds.),
Approaches to Natural Language, Dordrecht: D. Reidel, 221–242. Reprinted in
Thomason (1974, 247–270).
Moortgat, Michael (1988). Categorial Investigations: Logical and Linguistic Aspects
of the Lambek Calculus. Ph.D. thesis, University of Amsterdam.
Moortgat, Michael (1996). ‘Generalized Quantifiers and Discontinuous Type
Constructors.’ In Harry Bunt, and Arthur van Horck (eds.), Discontinuous
Constituency, Berlin & New York: Mouton de Gruyter, number 6 in Natural
Language Processing, 181–207.
Moot, Richard and Christian Retore (2012). The Logic of Categorial Grammars:
A Deductive Account of Natural Language Syntax and Semantics. Dordrecht:
Springer.
Moravcsik, Edith A. (1978). ‘Agreement.’ In Joseph H. Greenberg, Charles A.
Ferguson, and Edith A. Moravcsik (eds.), Universals of Human Language:
Syntax, Stanford: Stanford University Press, volume 4, 331–374.
786 references

Moravcsik, Edith A. (1988). ‘Agreement and Markedness.’ In Barlow and Ferguson

(1988), 89–106.
Morimoto, Yukiko (1999). ‘An Optimality Account of Argument Reversal.’ In Butt
and King (1999). URL http://csli-publications.stanford.edu/LFG/LFG4-1999/.
Morimoto, Yukiko (2000). Discourse Configurationality in Bantu Morphosyntax.
Ph.D. thesis, Stanford University.
Morimoto, Yukiko (2001). ‘Deriving the Directionality Parameter in OT-LFG.’ In
Butt and King (2001a). URL http://csli-publications.stanford.edu/LFG/6/lfg01.
html.
Morrill, Glyn V. (1994). Type Logical Grammar: Categorial Logic of Signs.
Dordrecht: Kluwer.
Morrill, Glyn V. (2011). Categorial Grammar: Logical Syntax, Semantics, and
Processing. Oxford: Oxford University Press.
Morrill, Glyn V. (2015). ‘Categorial Grammar.’ In Heine and Narrog (2015).
Morrison, Andrew (2017). ‘Adjectival Focus as a Granularity Problem for
LFG-Glue.’ Unpublished manuscript, University of Oxford.
Müller, Stefan (2016). Grammatical Theory: From Transformational Grammar
To Constraint-Based Approaches. Bonn: Language Science Press. URL http://
langsci-press.org/catalog/book/25.
Musgrave, Simon (2001). Non-Subject Arguments in Indonesian. Ph.D. thesis,
University of Melbourne.
Musgrave, Simon (2008). ‘The Grammatical Function OBJ in Indonesian.’ In
Austin and Musgrave (2008), 115–138.
Muskens, Reinhard (1995). ‘Order-Independence and Underspecification.’
In Jeroen Groenendijk (ed.), Ellipsis, Underspecification, Events and More
In Dynamic Semantics, University of Amsterdam: The DYANA-2 Project
Administrator, ILLC, 17–34. DYANA-2 Deliverable R2.2.C.
Mycock, Louise (2004). ‘The Wh-Expletive Construction.’ In Butt and King
(2004b), 370–390. URL http://csli-publications.stanford.edu/LFG/9/lfg04.
html.
Mycock, Louise (2006). The Typology of Constituent Questions: A Lexical
Functional Grammar Analysis of ‘Wh’-Questions. Ph.D. thesis, University of
Manchester.
Mycock, Louise (2007). ‘Constituent Question Formation and Focus: A New
Typological Perspective.’ Transactions of the Philological Society 105(2):
192–251.
Mycock, Louise (2009). ‘ “What Do You Do?” Variation in Interrogative Predicates.’
URL http://csli-publications.stanford.edu/LFG/14/abstracts/lfg09abs-mycock.
html. Presented at the workshop “Blurring Component Boundaries: Levels of
Analysis or Growth of Information?”, LFG09, Cambridge, July 2009. Abstract
in On-Line Proceedings of the LFG2009 Conference, ed. Miriam Butt and Tracy
Holloway King.
Mycock, Louise (2010). ‘Prominence in Hungarian: The Prosody-Syntax
Connection.’ Transactions of the Philological Society 108(3): 265–297.
Mycock, Louise (2013). ‘Discourse Functions of Question Words.’ In Butt and
King (2013). URL http://csli-publications.stanford.edu/LFG/18/lfg13.html.
 references 787

Mycock, Louise (2017). ‘Right-Dislocated Pronouns in British English: The Form

and Functions of ProTag Constructions.’ English Language and Linguistics
1–23. doi:10.1017/S1360674317000399.
Mycock, Louise and John J. Lowe (2013). ‘The Prosodic Marking of Discourse
Functions.’ In Butt and King (2013). URL http://csli-publications.stanford.edu/
LFG/18/lfg13.html.
Mycock, Louise and John J. Lowe (2014). ‘S-Structure Features for Information
Structure Analysis.’ Presented at the LFG14 conference, Ann Arbor, July 2014.
Nadathur, Prerna (2013). ‘Weak Crossover and the Direct Association Hypothesis.’
In Butt and King (2013). URL http://csli-publications.stanford.edu/LFG/18/
lfg13.html.
Needham, Stephanie and Ida Toivonen (2011). ‘Derived Arguments.’ In Butt and
King (2011), 401–421. URL http://csli-publications.stanford.edu/LFG/16/
lfg11.html.
Neidle, Carol (1982). ‘Case Agreement in Russian.’ In Bresnan (1982b), 391–426.
Neidle, Carol (1994). ‘Lexical Functional Grammar.’ In Encyclopedia of Language
and Linguistics, New York: Pergamon Press, 2147–2153.
Neidle, Carol (1996). ‘Lexical Functional Grammar.’ In Brown and Miller (1996).
Nelken, Rani and Nissim Francez (2002). ‘Bilattices and the Semantics of Natural
Language Questions.’ Linguistics and Philosophy 25: 37–64.
Nemati, Fatemeh (2010). ‘Incorporation and Complex Predication in Persian.’
In Butt and King (2010), 374–394. URL http://csli-publications.stanford.edu/
LFG/15/lfg10.html.
Nespor, Marina and Irene Vogel (1986). Prosodic Phonology. Dordrecht: Foris.
Nespor, Marina and Irene Vogel (2007). Prosodic Phonology. Berlin: Mouton de
Gruyter, second edition.
Newman, Paula S. (1996). ‘Representing 2P Clitic Placement in LFG.’ In Butt and
King (1996a). URL http://csli-publications.stanford.edu/LFG/LFG1-1996/.
Nichols, Johanna (1986). ‘Head-Marking and Dependent-Marking Grammar.’
Language 62(1): 56–119.
Nikitina, Tatiana (2008). The Mixing of Syntactic Properties and Language Change.
Ph.D. thesis, Stanford University.
Nordlinger, Rachel (1998). Constructive Case: Evidence From Australian
Languages. Dissertations in Linguistics. Stanford: CSLI Publications. Revised
version of 1997 Stanford University dissertation.
Nordlinger, Rachel (2000). ‘Australian Case Systems: Towards a Constructive
Solution.’ In Butt and King (2000a).
Nordlinger, Rachel (2012). ‘Number Marking in the Daly Languages (Australia).’
In Butt and King (2012). URL http://csli-publications.stanford.edu/LFG/17/
lfg12.html.
Nordlinger, Rachel and Joan Bresnan (1996). ‘Nonconfigurational Tense in
Wambaya.’ In Butt and King (1996a). URL http://csli-publications.stanford.
edu/LFG/LFG1-1996/.
Nordlinger, Rachel and Joan Bresnan (2011). ‘Lexical-Functional Grammar:
Interactions Between Morphology and Syntax.’ In Borsley and Börjars (2011),
112–140.
788 references

Nordlinger, Rachel and Louisa Sadler (2003). ‘The Syntax and Semantics of
Tensed Nominals.’ In Butt and King (2003b). URL http://csli-publications.
stanford.edu/LFG/8/lfg03.html.
Nordlinger, Rachel and Louisa Sadler (2004a). ‘Nominal Tense in Crosslinguistic
Perspective.’ Language 80(4): 776–806.
Nordlinger, Rachel and Louisa Sadler (2004b). ‘Tense Beyond the Verb: Encoding
Clausal Tense/Aspect/Mood on Nominal Dependents.’ Natural Language and
Linguistic Theory 22(3): 597–641.
Nordlinger, Rachel and Louisa Sadler (2007). ‘Verbless Clauses: Revealing the
Structure Within.’ In Zaenen et al. (2007), 139–160.
Nordlinger, Rachel and Louisa Sadler (2008). ‘From Juxtaposition to Incorpora-
tion: An Approach to Generic-Specific Constructions.’ In Butt and King (2008),
394–412. URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Nunberg, Geoffrey, Ivan A. Sag, and Thomas Wasow (1994). ‘Idioms.’ Language
70(3): 491–538.
O’Connor, Mary Catherine (1987). Topics in Northern Pomo Grammar. Ph.D.
thesis, University of California, Berkeley. Reprinted by Garland Press, New
York, 1992.
O’Connor, Rob (2002a). ‘Clitics and Phrasal Affixation in Constructive
Morphology.’ In Butt and King (2002). URL http://csli-publications.stanford.
edu/LFG/7/lfg02.html.
O’Connor, Rob (2002b). ‘The Placement of Enclitics in Bosnian, Croatian and
Serbian.’ URL http://roa.rutgers.edu/files/521-0502/521-0502-OCONNOR-0-
0.PDF.gz. Unpublished manuscript, University of Manchester.
O’Connor, Rob (2005a). ‘Clitics in LFG: Prosodic Structure and Phrasal Affixation.’
URL http://csli-publications.stanford.edu/LFG/10/lfg05.html. Abstract in On-
Line Proceedings of the LFG2005 Conference, ed. Miriam Butt and Tracy
Holloway King.
O’Connor, Rob (2005b). ‘Information Structure in Lexical-Functional Grammar:
The Discourse-Prosody Correspondence in English and Serbo-Croatian.’ URL
http://csli-publications.stanford.edu/LFG/10/lfg05.html. Abstract in On-Line
Proceedings of the LFG2005 Conference, ed. Miriam Butt and Tracy Holloway
King.
O’Connor, Robert (2006). Information Structure in Lexical Functional Grammar:
The Discourse-Prosody Correspondence. Ph.D. thesis, University of Manchester.
Odden, David (1995). ‘Phonology At The Phrasal Level In Bantu.’ In Francis
Katamba (ed.), Bantu Phonology and Morphology, Munich: LINCOM, 40–68.
Oehrle, Dick (1999). ‘LFG as Labeled Deduction.’ In Dalrymple (1999).
Oehrle, Richard T., Emmon Bach, and Deirdre Wheeler (eds.) (1988). Categorial
Grammars and Natural Language Structures. Dordrecht: D. Reidel.
Ohala, John J. (1989). ‘Sound Change Is Drawn From A Pool Of Synchronic
Variation.’ In Leiv Egil Breivik, and Ernst Håkon Jahr (eds.), Language Change:
Contributions to the Study of its Causes, Berlin: Mouton de Gruyter, 173–198.
Ostler, Nicholas D. M. (1979). Case-Linking: A Theory of Case and Verb
Diathesis Applied to Classical Sanskrit. Ph.D. thesis, Massachusetts Institute of
Technology.
 references 789

Otoguro, Ryo (2006). Morphosyntax of Case: A Theoretical Inverstigation of the

Concept. Ph.D. thesis, University of Essex.
Otoguro, Ryo (2015). ‘Generalising Functional Categories in LFG.’ In Butt and
King (2015b). URL http://csli-publications.stanford.edu/LFG/20/lfg15.html.
Owens, Jonathan (1986). ‘The Comparative Study of Medieval Arabic
Grammatical Theory.’ In Joshua A. Fishman, Andrée Tabouret-Keller, Michael
Clyne, Bh. Krishnamurti, and Mohamed Abdulaziz (eds.), The Fergusonian
Impact: In Honor of Charles A. Ferguson, Berlin: Mouton de Gruyter, volume 1:
From Phonology to Society, 223–238.
Pak, Marjorie (2008). The Postsyntactic Derivation And Its Phonological Reflexes.
Ph.D. thesis, University of Pennsylvania.
ParGram Consortium (2017). ‘Negation.’ URL http://typo.uni-konstanz.de/
redmine/projects/pargram/wiki/Negation. PARGRAM Wiki Page, University
of Konstanz.
Park, Karen (2012). The Selective Properties of Verbs in Reflexive Constructions.
D.Phil. thesis, University of Oxford. URL https://ora.ox.ac.uk/objects/uuid:
3154fd5f-a82c-4454-9679-cd3c5c7b0fb0.
Parsons, Terence (1990). Events in the Semantics of English: A Study in Subatomic
Semantics. Number 19 in Current Studies in Linguistics. Cambridge, MA: The
MIT Press.
Partee, Barbara H. (1970). ‘Negation, Conjunction, and Quantifiers: Syntax vs.
Semantics.’ Foundations of Language 6(2): 153–165.
Partee, Barbara H. (1995). ‘Lexical Semantics and Compositionality.’ In Lila R.
Gleitman, and Mark Liberman (eds.), An Invitation to Cognitive Science:
Language, Cambridge, MA: The MIT Press, 311–360. Second edition.
Partee, Barbara H. (2010). ‘Privative Adjectives: Subsective Plus Coercion.’ In
Rainer Bauerle, Uwe Reyle, and Thomas Zimmermann (eds.), Presuppositions
and Discourse: Essays Offered to Hans Kamp, Brill, 273–285.
Partee, Barbara H. and Vladimir Borschev (1998). ‘Integrating Lexical and Formal
Semantics: Genitives, Relational Nouns, and Type-Shifting.’ In Robin Cooper,
and Thomas Gamkrelidze (eds.), Proceedings of the Second Tbilisi Symposium
On Language, Logic, and Computation. Tbilisi: Center on Language, Logic,
Speech, Tbilisi State University, 229–241. URL http://www-unix.oit.umass.
edu/~partee/docs/tbilisi17.pdf.
Partee, Barbara H. and Mats Rooth (1983). ‘Generalized Conjunction and Type
Ambiguity.’ In Bäuerle et al. (1983), 361–383. Reprinted in Portner and Partee
(2002, 334–356).
Partee, Barbara H., Alice ter Meulen, and Robert E. Wall (1993). Mathematical
Methods in Linguistics. Dordrecht: Kluwer. Corrected second printing of the
first edition.
Patejuk, Agnieszka (2015). Unlike Coordination in Polish: An LFG Account. Ph.D.
thesis, Polish Academy of Sciences. URL http://nlp.ipipan.waw.pl/Bib/pat:15.
pdf.
Patejuk, Agnieszka and Adam Przepiórkowski (2012). ‘Lexico-Semantic
Coordination in Polish.’ In Butt and King (2012). URL http://csli-publications.
stanford.edu/LFG/17/lfg12.html.
790 references

Patejuk, Agnieszka and Adam Przepiórkowski (2014). ‘In Favour of the Raising
Analysis of Passivisation.’ In Butt and King (2014). URL http://csli-publications.
stanford.edu/LFG/19/lfg14.html.
Patejuk, Agnieszka and Adam Przepiórkowski (2015). ‘An LFG Analysis of the
So-Called Reflexive Marker in Polish.’ In Butt and King (2015b). URL http://
csli-publications.stanford.edu/LFG/20/lfg15.html.
Patejuk, Agnieszka and Adam Przepiórkowski (2017). ‘Reducing Grammatical
Functions in LFG.’ In Arnold et al. (2017). URL http://cslipublications.stanford.
edu/HPSG/2016/.
Payne, John (2009). ‘The English Genitive and Double Case.’ Transactions of the
Philological Society 107(3): 322–357.
Payne, John and Kersti Börjars (2015). ‘Features and Selection in LFG: The
English VP.’ In Butt and King (2015b). URL http://csli-publications.stanford.
edu/LFG/20/lfg15.html.
Pereira, Fernando C. N. and David H. D. Warren (1983). ‘Parsing as Deduction.’ In
Proceedings of the 21st Annual Meeting of the ACL. Cambridge, MA: Association
for Computational Linguistics, 137–144.
Perlmutter, David (ed.) (1983). Studies in Relational Grammar 1. Chicago:
University of Chicago Press.
Perlmutter, David M. (1971). Deep and Surface Structure Constraints in Syntax.
New York: Holt, Rinehart & Winston.
Perlmutter, David M. and Paul M. Postal (1977). ‘Toward a Universal
Characterization of Passivization.’ In Proceedings of the Third Annual Meeting
of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society.
Reprinted in Perlmutter (1983, 3–29).
Perlmutter, David M. and Paul M. Postal (1983). ‘Some Proposed Laws of Basic
Clause Structure.’ In Perlmutter (1983), 81–128.
Perlmutter, David M. and Paul M. Postal (1984). ‘The 1-Advancement
Exclusiveness Law.’ In David Perlmutter, and Carol Rosen (eds.), Studies
in Relational Grammar 2, Chicago: University of Chicago Press.
Peters, Stanley and Dag Westerståhl (2006). Quantifiers in Language and Logic.
Oxford: Oxford University Press.
Peterson, John L. (1981). Petri Net Theory and the Modeling of Systems. London:
Prentice Hall.
Peterson, Peter G. (1982). ‘Conjunction in LFG.’ Unpublished manuscript,
Department of Linguistics, MIT and University of Newcastle, New South
Wales, Australia.
Peterson, Peter G. (2004a). ‘Coordination: Consequences of a Lexical-Functional
Account.’ Natural Language and Linguistic Theory 22(3): 643–679.
Peterson, Peter G. (2004b). ‘Introduction to the Workshop on Coordination and
Agreement.’ In Butt and King (2004b). URL http://csli-publications.stanford.
edu/LFG/9/lfg04.html.
Pica, Pierre (1987). ‘On the Nature of the Reflexivization Cycle.’ In Joyce
McDonough, and Bernadette Plunkett (eds.), Proceedings of NELS 17. Graduate
Linguistic Student Association, University of Massachusetts, Amherst,
483–500.
 references 791

Pickering, Martin (1993). ‘Direct Association and Sentence Processing: A Reply

to Gorrell and to Gibson and Hancock.’ Language and Cognitive Processes 8(2):
163–196.
Pickering, Martin and Guy Barry (1991). ‘Sentence Processing Without Empty
Categories.’ Language and Cognitive Processes 6(3): 229–259.
Pienemann, Manfred (1998). Language Processing and Second Language
Development: Processability Theory. Amsterdam: John Benjamins.
Pienemann, Manfred (ed.) (2005a). Cross-Linguistic Aspects of Processability
Theory, volume 30 of Studies in Bilingualism. Amsterdam: John Benjamins.
Pienemann, Manfred (2005b). ‘An Introduction to Processability Theory.’ In
Pienemann (2005a), 1–60.
Pienemann, Manfred (2011). ‘The Psycholinguistic Basis of PT.’ In Pienemann
and Keßler (2011), 27–49.
Pienemann, Manfred, Bruno Di Biase, and Satomi Kawaguchi (2005). ‘Extending
Processability Theory.’ In Pienemann (2005a), 199–251.
Pienemann, Manfred and Jörg-U. Keßler (eds.) (2011). Studying Processability
Theory: An Introductory Textbook. Amsterdam: John Benjamins.
Pienemann, Manfred and Jörg-U. Keßler (2012). ‘Processability Theory.’ In
Susan M. Gass, and Alison Mackey (eds.), The Routledge Handbook of Second
Language Acquisition, London: Routledge, 228–246.
Pinker, Steven (1982). ‘A Theory of the Acquisition of Lexical-Interpretive
Grammars.’ In Bresnan (1982b), 655–726.
Pinker, Steven (1989). Learnability and Cognition: The Acquisition of Argument
Structure. Cambridge, MA: The MIT Press. Reprinted by The MIT Press in
2013 with a new preface.
Plank, Frans (ed.) (1995). Double Case: Agreement by Suffixaufnahme. Oxford:
Oxford University Press.
Polinsky, Maria and Eric Potsdam (2002a). ‘Backward Control.’ Linguistic Inquiry
33(2): 245–282.
Polinsky, Maria and Eric Potsdam (2002b). ‘Backward Control: Evidence from
Malagasy.’ In Andrea Rackowski, and Norvin Richards (eds.), Proceedings of
AFLA VIII. Cambridge, MA: Department of Linguistics and Philosophy, MIT,
volume 44 of MIT Working Papers in Linguistics, 257–272.
Polinsky, Maria and Eric Potsdam (2006). ‘Expanding the Scope of Control and
Raising.’ Syntax 9(2): 171–192.
Pollard, Carl (1984). Generalized Phrase Structure Grammars, Head Grammars,
and Natural Language. Ph.D. thesis, Stanford University.
Pollard, Carl and Ivan A. Sag (1987). Information-Based Syntax and Semantics,
Volume I. Number 13 in CSLI Lecture Notes. Stanford: CSLI Publications.
Pollard, Carl and Ivan A. Sag (1994). Head-Driven Phrase Structure Grammar.
Chicago: University of Chicago Press.
Pollock, Jean-Yves (1989). ‘Verb Movement, Universal Grammar, and the
Structure of IP.’ Linguistic Inquiry 20(3).
Portner, Paul and Barbara H. Partee (eds.) (2002). Formal Semantics: The Essential
Readings. Oxford: Blackwell.
792 references

Poser, William J. (1992). ‘Blocking of Phrasal Constructions by Lexical Items.’ In

Sag and Szabolcsi (1992), 111–130.
Postal, Paul M. (1970). ‘On Coreferential Complement Subject Deletion.’ Linguistic
Inquiry 1(4).
Postal, Paul M. (1974). On Raising: One Rule of English Grammar and its
Theoretical Implications. Cambridge, MA: The MIT Press.
Postal, Paul M. (1998). Three Investigations of Extraction. Cambridge, MA: The
MIT Press.
Postal, Paul M. and Geoffrey K. Pullum (1978). ‘Traces and the Description of
English Complementizer Contraction.’ Linguistic Inquiry 9: 1–29.
Potts, Christopher (2005). The Logic of Conventional Implicatures. Oxford: Oxford
University Press.
Price, Patti, Mari Ostendorf, Stefanie Shattuck-Hufnagel, and Cynthia Fong
(1991). ‘The Use of Prosody in Syntactic Disambiguation.’ Journal of the
Acoustical Society of America 90: 2956–2970.
Prince, Alan S. and Paul Smolensky (2004). Optimality Theory: Constraint
Interaction in Generative Grammar. Oxford: Blackwell.
Prince, Ellen (1981). ‘Topicalization, Focus-Movement, and Yiddish-Movement:
A Pragmatic Differentiation.’ In Proceedings of the 7th Annual Meeting of the
Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, 249–264.
Przepiórkowski, Adam (2014a). ‘Distance Distributivity in Polish: Towards a Glue
Semantics Approach.’ In Christopher Piñón (ed.), Empirical Issues in Syntax
and Semantics 10. Colloque de Syntaxe et Sémantique à Paris, 107–124. URL
http://www.cssp.cnrs.fr/eiss10/.
Przepiórkowski, Adam (2014b). ‘Locality Constraints in Distance Distributivity:
A Propositional Glue Approach.’ In Butt and King (2014). URL http://csli-
publications.stanford.edu/LFG/19/lfg14.html.
Przepiórkowski, Adam (2015). ‘A Weakly Compositional Analysis of Distance
Distributivity in Polish.’ In Małgorzata Szajbel-Keck, Roslyn Burns, and Darya
Kavitskaya (eds.), Annual Workshop on Formal Approaches to Slavic Linguistics:
The First Berkeley Meeting 2014. Ann Arbor, MI, 262–281. URL http://nlp.
ipipan.waw.pl/Bib/prze:14:fasl.pdf.
Przepiórkowski, Adam (2017). ‘How Not To Distinguish Arguments from
Adjuncts in LFG.’ In Arnold et al. (2017). URL http://cslipublications.stanford.
edu/HPSG/2016/.
Przepiórkowski, Adam and Agnieszka Patejuk (2012). ‘On Case Assignment and
the Coordination of Unlikes: The Limits of Distributive Features.’ In Butt and
King (2012). URL http://csli-publications.stanford.edu/LFG/17/lfg12.html.
Przepiórkowski, Adam and Agnieszka Patejuk (2015). ‘Two Representations of
Negation in LFG: Evidence From Polish.’ In Butt and King (2015b). URL http://
csli-publications.stanford.edu/LFG/20/lfg15.html.
Pulleyblank, Douglas (1986). ‘Clitics in Yoruba.’ In Hagit Borer (ed.), The Syntax
of Pronominal Clitics, New York: Academic Press, volume 19 of Syntax and
Semantics, 43–64.
Pullum, Geoffrey K. (1997). ‘The Morpholexical Nature of English To-
Contraction.’ Language 73: 79–102.
 references 793

Pullum, Geoffrey K. and Arnold M. Zwicky (1986). ‘Phonological Resolution of

Syntactic Feature Conflict.’ Language 62(4): 751–773.
Pullum, Geoffrey K. and Arnold M. Zwicky (1988). ‘The Syntax-Phonology
Interface.’ In Frederick J. Newmeyer (ed.), Linguistics: The Cambridge Survey,
Cambridge, UK: Cambridge University Press.
Pustet, Regina (2003). Copulas: Universals in the Categorization of the Lexicon.
Oxford: Oxford University Press.
Quine, Willard Van Orman (1953). ‘Reference and Modality.’ In From A Logical
Point of View, Cambridge, MA: Harvard University Press, 139–159.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (1972). A
Grammar of Contemporary English. New York: Seminar Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (1985). A
Comprehensive Grammar of the English Language. New York: Longman.
Radford, Andrew (2009). Analysing English Sentences: A Minimalist Approach.
Cambridge, UK: Cambridge University Press.
Rákosi, György (2006a). Dative Experiencer Predicates in Hungarian. Ph.D. thesis,
University of Utrecht.
Rákosi, György (2006b). ‘On the Need for a More Refined Approach to
the Argument-Adjunct Distinction: The Case of Dative Experiencers in
Hungarian.’ In Butt and King (2006b). URL http://csli-publications.stanford.
edu/LFG/11/lfg06.html.
Rákosi, György (2008). ‘The Inherently Reflexive and the Inherently Reciprocal
Predicate in Hungarian: Each to Their Own Argument Structure.’ In Ekkehard
König, and Volker Gast (eds.), Reciprocals and Reflexives: Theoretical and
Typological Explorations, Berlin: Mouton de Gruyter, 411–450.
Rákosi, György (2009). ‘Beyond Identity: The Case of a Complex Hungarian
Reflexive.’ In Butt and King (2009), 459–479. URL http://csli-publications.
stanford.edu/LFG/14/lfg09.html.
Rákosi, György (2010). ‘On Snakes and Locative Binding in Hungarian.’ In Butt
and King (2010), 395–415. URL http://csli-publications.stanford.edu/LFG/15/
lfg10.html.
Rákosi, György (2013). ‘Down with Obliques?’ In King and de Paiva (2013),
127–138.
Rákosi, György and Tibor Laczkó (2005). ‘Verbal Category and Nominal Function:
Evidence From Hungarian Subject Clauses.’ In Butt and King (2005). URL
http://csli-publications.stanford.edu/LFG/10/lfg05.html.
Ramchand, Gillian and Charles Reiss (eds.) (2007). The Oxford Handbook of
Linguistic Interfaces. Oxford: Oxford University Press.
Rappaport, Malka (1983). ‘On the Nature of Derived Nominals.’ In Levin et al.
(1983), 113–142.
Raza, Ghulam (2011). Subcategorization Acquisition and Classes of Predication in
Urdu. Ph.D. thesis, University of Konstanz.
Raza, Ghulam and Tafseer Ahmed (2011). ‘Argument Scrambling Within Urdu
NPs.’ In Butt and King (2011). URL http://csli-publications.stanford.edu/LFG/
16/lfg11.html.
794 references

Reape, Mike (1994). ‘Domain Union and Word Order Variation in German.’ In
John Nerbonne, Klaus Netter, and Carl Pollard (eds.), German in Head-Driven
Phrase Structure Grammar, Stanford: CSLI Publications, number 46 in CSLI
Lecture Notes, 151–197.
Reichenbach, Hans (1947). Elements of Symbolic Logic. London: Macmillan.
Reinhart, Tanya (1976). The Syntactic Domain of Anaphora. Ph.D. thesis,
Massachusetts Institute of Technology.
Reinhart, Tanya (1981). ‘Pragmatics and Linguistics: An Analysis of Sentence
Topics.’ Philosophica 27: 53–94.
Reinhart, Tanya (2002). ‘The Theta System — An Overview.’ Theoretical Linguistics
28: 229–290.
Reinhart, Tanya and Tal Siloni (2005). ‘The Lexicon-Syntax Parameter:
Reflexivization and Other Arity Operations.’ Linguistic Inquiry 36(3): 389–436.
Reyle, Uwe (1988). ‘Compositional Semantics for LFG.’ In Reyle and Rohrer
(1988), 448–474.
Reyle, Uwe (1993). ‘Dealing With Ambiguities by Underspecification: Construc-
tion, Representation and Deduction.’ Journal of Semantics 10: 123–179.
Reyle, Uwe and Christian Rohrer (eds.) (1988). Natural Language Parsing and
Linguistic Theories. Dordrecht: D. Reidel.
Rhodes, Richard (1990). ‘Ojibwa Secondary Objects.’ In Dziwirek et al. (1990),
401–414.
Riezler, Stefan, Tracy Holloway King, Ronald M. Kaplan, Richard Crouch, John T.
Maxwell, III, and Mark Johnson (2002). ‘Parsing the Wall Street Journal Using
a Lexical-Functional Grammar and Discriminative Estimation Techniques.’ In
Proceedings of the 40th Annual Meeting of the ACL. Philadelphia: Association
for Computational Linguistics.
Rizzi, Luigi (1990). Relativized Minimality. Cambridge, MA: The MIT Press.
Rizzi, Luigi (1997). ‘The Fine Structure of the Left Periphery.’ In Liliane M. V.
Haegeman (ed.), Elements of Grammar: Handbook of Generative Syntax,
Dordrecht & Boston: Kluwer, 281–337.
Robins, R. H. (1967). A Short History of Linguistics. Bloomington: Indiana
University Press.
Rögnvaldsson, Eiríkur (1984). ‘Af Lýsingarorðsviðurlögum.’ Íslenskt mál 6: 57–80.
Rosen, Carol (1990). ‘Rethinking Southern Tiwa: The Geometry of a Triple-
Agreement Language.’ Language 66(4): 669–713.
Rosén, Victoria (1996). ‘The LFG Architecture and “Verbless” Syntactic
Constructions.’ In Butt and King (1996a). URL http://csli-publications.
stanford.edu/LFG/LFG1-1996/.
Ross, John Robert (1967). Constraints on Variables in Syntax. Ph.D. thesis,
Massachusetts Institute of Technology.
Rounds, William C. (1988). ‘Set Values for Unification-Based Grammar
Formalisms and Logic Programming.’ Technical Report CSLI-88-129,
CSLI Publications, Stanford.
Rounds, William C. (1997). ‘Feature Logics.’ In van Benthem and ter Meulen
(1997), 475–533. Co-published with Elsevier Science B.V., Amsterdam.
Rubio Vallejo, David (2011). Impersonal Constructions in Spanish: An LFG
Approach Based On Gradience. M.Phil. dissertation, University of Oxford.
 references 795

Rudin, Catherine (1985). Aspects of Bulgarian Syntax: Complementizers and WH

Constructions. Columbus, OH: Slavica.
Sadler, Louisa (1996). ‘New Developments in LFG.’ In Brown and Miller (1996).
Sadler, Louisa (1997). ‘Clitics and the Structure-Function Mapping.’ In Butt and
King (1997). URL http://csli-publications.stanford.edu/LFG/LFG2-1997/.
Sadler, Louisa (1998). ‘English Auxiliaries as Tense Inflections.’ In Special Issue of
Essex Research Reports Produced On The Occasion of the Retirement of Keith
Brown. Department of Languages and Linguistics, University of Essex.
Sadler, Louisa (1999). ‘Non-Distributive Features and Coordination in Welsh.’ In
Butt and King (1999). URL http://csli-publications.stanford.edu/LFG/LFG4-
1999/.
Sadler, Louisa (2000). ‘Noun Phrase Structure in Welsh.’ In Butt and King (2000a),
73–110.
Sadler, Louisa (2003). ‘Coordination and Asymmetric Agreement in Welsh.’ In
Butt and King (2003a), 85–118.
Sadler, Louisa (2006). ‘Gender Resolution in Rumanian.’ In Butt et al. (2006),
301–322.
Sadler, Louisa (2011). ‘Indeterminacy, Complex Features and Underspecification.’
Morphology 21: 379–417.
Sadler, Louisa (2016). ‘Agreement in Archi: An LFG Perspective.’ In Oliver Bond,
Greville G. Corbett, Marina Chumakina, and Dunstan Brown (eds.), Archi:
Complexities of Agreement in Cross-Theoretical Perspective, Oxford: Oxford
University Press.
Sadler, Louisa and Doug Arnold (1994). ‘Prenominal Adjectives and the
Phrasal/Lexical Distinction.’ Journal of Linguistics 30: 187–226.
Sadler, Louisa and Rachel Nordlinger (2004). ‘Relating Morphology to Syntax.’ In
Sadler and Spencer (2004), 159–186.
Sadler, Louisa and Rachel Nordlinger (2006a). ‘Apposition As Coordination:
Evidence From Australian Languages.’ In Butt and King (2006b). URL http://
csli-publications.stanford.edu/LFG/11/lfg06.html.
Sadler, Louisa and Rachel Nordlinger (2006b). ‘Case Stacking in Realizational
Morphology.’ Linguistics 44(3): 459–487.
Sadler, Louisa and Rachel Nordlinger (2010). ‘Nominal Juxtaposition in Australian
Languages: An LFG Analysis.’ Journal of Linguistics 46(2): 415–452.
Sadler, Louisa and Rachel Nordlinger (2018). ‘Morphology in LFG and HPSG.’ In
Jenny Audring, and Francesca Masini (eds.), Oxford Handbook of Morphological
Theory, Oxford: Oxford University Press.
Sadler, Louisa and Andrew Spencer (2000). ‘Workshop on Morphology and LFG.’
In Butt and King (2000c). URL http://csli-publications.stanford.edu/LFG/
LFG5-2000/papers/lfg00morphwrk.html.
Sadler, Louisa and Andrew Spencer (2001). ‘Syntax as an Exponent of
Morphological Features.’ In Geert Booij, and Jaap van Marle (eds.), Yearbook of
Morphology, Dordrecht: Kluwer, 71–96.
Sadler, Louisa and Andrew Spencer (eds.) (2004). Projecting Morphology. Stanford:
CSLI Publications.
Sadock, Jerrold M. (1991). Autolexical Syntax: A Theory of Parallel Grammatical
Representations. Chicago: University of Chicago Press.
796 references

Sag, Ivan A. (1987). ‘Grammatical Hierarchy and Linear Precedence.’ In Geoffrey

Huck, and Almerindo Ojeda (eds.), Discontinuous Constituency, San Diego:
Academic Press, volume 20 of Syntax and Semantics, 303–340. Also published
as Report CSLI-88-129. Stanford: Center for the Study of Language and
Information.
Sag, Ivan A. (1998). ‘Without a Trace.’ URL http://citeseer.ist.psu.edu/viewdoc/
summary?doi=10.1.1.24.4018. Unpublished manuscript, Department of
Linguistics, Stanford University.
Sag, Ivan A. (2002). ‘Coordination and Underspecification.’ In Jong-Bok Kim,
and Stephen Wechsler (eds.), Proceedings of the 9th International Conference
on Head-Driven Phrase Structure Grammar. Stanford: CSLI Publications,
267–291. URL http://cslipublications.stanford.edu/HPSG/3/.
Sag, Ivan A. and Janet D. Fodor (1994). ‘Extraction Without Traces.’ In Raul
Aranovich, William Byrne, Susanne Preuss, and Martha Senturia (eds.),
WCCFL 13: Proceedings of the 13th West Coast Conference on Formal
Linguistics. Stanford: CSLI Publications, 365–384.
Sag, Ivan A., Gerald Gazdar, Thomas Wasow, and Steven Weisler (1985).
‘Coordination and How to Distinguish Categories.’ Natural Language and
Linguistic Theory 3(2): 117–171.
Sag, Ivan A. and Carl Pollard (1991). ‘An Integrated Theory of Complement
Control.’ Language 67(1): 63–113. URL http://lingo.stanford.edu/sag/papers/
sag-pollard-91.pdf.
Sag, Ivan A. and Anna Szabolcsi (eds.) (1992). Lexical Matters. Stanford: CSLI
Publications.
Sag, Ivan A., Thomas Wasow, and Emily Bender (2003). Syntactic Theory: A
Formal Introduction. Stanford: CSLI Publications.
Saiki, Mariko (1985). ‘On the Coordination of Gapped Constituents in Japanese.’
In William H. Eilfort, Paul D. Kroeber, and Karen L. Peterson (eds.), Papers
From the 21st Regional Meeting of the Chicago Linguistic Society. Chicago:
Chicago Linguistic Society, 371–387.
Saiki, Mariko (1987). On the Manifestations of Grammatical Functions in the
Syntax of Japanese Nominals. Ph.D. thesis, Stanford University.
Saiki, Mariko (1991). ‘The Tough Construction in Japanese.’ Studies in Humanities
28(2): 91–115.
Saraswat, Vijay A. (1999). ‘LFG as Concurrent Constraint Programming.’ In
Dalrymple (1999).
Scedrov, Andre (1993). ‘A Brief Guide to Linear Logic.’ In G. Rozenberg, and
A. Salomaa (eds.), Current Trends in Theoretical Computer Science, World
Scientific Publishing Co., 377–394. Revised and expanded version of an article
originally appearing in Bulletin of the European Association for Theoretical
Computer Science 41, 1990.
Schachter, Paul (1976). ‘The Subject in Philippine Languages: Topic, Actor,
Actor-Topic, or None of the Above?’ In Li (1976).
Scheer, Tobias (2011). A Guide to Morphosyntax-Phonology Interface Theories:
How Extra-Phonological Information is Treated in Phonology Since Trubetskoy’s
Grenzsignale. Berlin: Mouton de Gruyter.
Schein, Barry (1994). Plurals and Events. Cambridge, MA: The MIT Press.
 references 797

Schwarze, Christoph (2001a). ‘On the Representation of French and Italian Clitics.’
In Butt and King (2001a), 280–304. URL http://csli-publications.stanford.edu/
LFG/6/lfg01.html.
Schwarze, Christoph (2001b). ‘Representation and Variation: On the Development
of Romance Auxiliary Syntax.’ In Butt and King (2001b), 143–172.
Schwarze, Christoph and Leonel Figueiredo de Alencar (2016). Lexikalisch-
funktionale Grammatik : Eine Einführung am Beispiel des Französischen mit
computerlinguistischer Implementierung. Tübingen: Stauffenburg. URL http://
nbn-resolving.de/urn:nbn:de:bsz:352-0-318774.
Schwarzschild, Roger (1994). ‘Plurals, Presuppositions and the Sources of
Distributivity.’ Natural Language Semantics 2(3): 201–248.
Seiss, Melanie (2008). ‘The English -ing Form.’ In Butt and King (2008), 454–472.
URL http://csli-publications.stanford.edu/LFG/13/lfg08.html.
Seiss, Melanie (2009). ‘On the Difference Between Auxiliaries, Serial Verbs and
Light Verbs.’ In Butt and King (2009), 501–519. URL http://csli-publications.
stanford.edu/LFG/14/lfg09.html.
Seiss, Melanie and Rachel Nordlinger (2010). ‘Applicativizing Complex Predicates:
A Case Study From Murrinh-Patha.’ In Butt and King (2010), 416–436. URL
http://csli-publications.stanford.edu/LFG/15/lfg10.html.
Seki, Hiroyuki, Takashi Matsumura, Mamoru Fujii, and Tadao Kasami (1991).
‘On Multiple Context-Free Grammars.’ Theoretical Computer Science 88(2):
191–229.
Selkirk, Elisabeth O. (1978). ‘On Prosodic Structure and Its Relation to Syntactic
Structure.’ In T. Fretheim (ed.), Nordic Prosody II, Trondheim: TAPIR.
Selkirk, Elisabeth O. (1980a). ‘Prosodic Domains in Phonology: Sanskrit
Revisited.’ In Mark Aronoff, and Mary-Louise Kean (eds.), Juncture, Saratoga,
CA: Anma Libri.
Selkirk, Elisabeth O. (1980b). ‘The Role of Prosodic Categories in English Word
Stress.’ Linguistic Inquiry 11: 563–605.
Selkirk, Elisabeth O. (1981). ‘On the Nature of Phonological Representation.’
In Terry Myers, John Laver, and John Anderson (eds.), The Cognitive
Representation of Speech, Amsterdam: North Holland, 379–388.
Selkirk, Elisabeth O. (1984). Phonology and Syntax: The Relation Between Sound
and Structure. Cambridge, MA: The MIT Press.
Selkirk, Elisabeth O. (1986). ‘On Derived Domains in Sentence Phonology.’
Phonology Yearbook 3: 371–405.
Selkirk, Elisabeth O. (1995). ‘The Prosodic Structure of Function Words.’ In
Jill Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk (eds.), Papers in
Optimality Theory. Amherst, MA: Graduate Linguistic Student Association,
University of Massachusetts, Amherst, volume 18 of University of Massachusetts
Occasional Papers in Linguistics, 439–469.
Selkirk, Elisabeth O. (2011). ‘The Syntax-Phonology Interface.’ In John A.
Goldsmith, Jason Riggle, and Alan C. L. Yu (eds.), The Handbook of Phonological
Theory, Oxford: Wiley-Blackwell, 435–484. Second edition.
Selkirk, Elisabeth O. and Tong Shen (1990). ‘Prosodic Domains in Shanghai
Chinese.’ In Inkelas and Zec (1990), 313–337.
798 references

Sells, Peter (1985). Lectures on Contemporary Syntactic Theories: An Introduction

to Government-Binding Theory, Generalized Phrase Structure Grammar, and
Lexical-Functional Grammar. Number 3 in CSLI Lecture Notes. Stanford: CSLI
Publications.
Sells, Peter (1988). ‘Thematic and Grammatical Hierarchies: Albanian
Reflexivization.’ In Hagit Borer (ed.), WCCFL 7: Proceedings of the Seventh
West Coast Conference on Formal Linguistics. Stanford: CSLI Publications,
293–303.
Sells, Peter (1994). ‘Subphrasal Syntax in Korean.’ Language Research 30(2):
351–386.
Sells, Peter (1998). ‘Scandinavian Clause Structure and Object Shift.’ In Butt and
King (1998b). URL http://csli-publications.stanford.edu/LFG/LFG3-1998/.
Sells, Peter (2000a). ‘Negation in Swedish: Where It’s Not At.’ In Butt and King
(2000c). URL http://csli-publications.stanford.edu/LFG/LFG5-2000/.
Sells, Peter (2000b). ‘Raising and the Order of Clausal Constituents in the
Philippine Languages.’ In Ileana Paul, Vivianne Phillips, and Lisa Travis (eds.),
Formal Issues in Austronesian Linguistics, Dordrecht: Kluwer, 117–143.
Sells, Peter (ed.) (2001a). Formal and Empirical Issues in Optimality Theoretic
Syntax. Stanford: CSLI Publications.
Sells, Peter (2001b). Structure, Alignment and Optimality in Swedish. Stanford:
CSLI Publications.
Sells, Peter (2005). ‘The Peripherality of the Icelandic Expletive.’ In Butt and King
(2005). URL http://csli-publications.stanford.edu/LFG/10/lfg05.html.
Sells, Peter (2006a). ‘Optimality-Theoretic Lexical-Functional Grammar.’ In
Brown (2006), 60–68.
Sells, Peter (2006b). ‘Using Subsumption Rather Than Equality in Functional
Control.’ In Butt and King (2006b). URL http://csli-publications.stanford.edu/
LFG/11/lfg06.html.
Sells, Peter (2013). ‘Lexical Functional Grammar.’ In Marcel Den Dikken (ed.),
The Cambridge Handbook of Generative Syntax, Cambridge, UK: Cambridge
University Press, 162–201.
Sells, Peter, Annie Zaenen, and Draga Zec (1987). ‘Reflexivization Variation:
Relations Between Syntax, Semantics, and Lexical Structure.’ In Iida et al.
(1987), 169–238.
Sgall, Petr (1967). ‘Functional Sentence Perspective in a Generative Description.’
Prague Studies in Mathematical Linguistics 2: 203–225.
Shemtov, Hadar (1997). Ambiguity Management in Natural Language Generation.
Ph.D. thesis, Stanford University.
Shieber, Stuart M. (1986). An Introduction to Unification-Based Approaches to
Grammar. Number 4 in CSLI Lecture Notes. Stanford: CSLI Publications.
Siegel, Muffy E. A. (1976). Capturing the Adjective. Ph.D. thesis, University of
Massachusetts, Amherst.
Siewierska, Anna (2003). ‘Reduced Pronominals and Argument Prominence.’ In
Butt and King (2003a), 119–150.
Simpson, Jane (1983). ‘Discontinuous Verbs and the Interaction of Morphology
and Syntax.’ In Michael Barlow, Daniel P. Flickinger, and Michael T. Wescoat
 references 799

(eds.), WCCFL 2: Proceedings of the Second West Coast Conference on Formal

Linguistics. Stanford: CSLI Publications, 275–286.
Simpson, Jane (1991). Warlpiri Morphology and Syntax: A Lexicalist Approach.
Dordrecht: Kluwer.
Simpson, Jane (2001). ‘Preferred Word Order and the Grammaticalization of
Path.’ In Butt and King (2001b).
Simpson, Jane (2007). ‘Expressing Pragmatic Constraints on Word Order in
Warlpiri.’ In Zaenen et al. (2007), 403–427.
Simpson, Jane and Joan Bresnan (1983). ‘Control and Obviation in Warlpiri.’
Natural Language and Linguistic Theory 1(1): 49–64.
Snijders, Liselotte (2012). ‘Issues Concerning Constraints on Discontinuous NPs
in Latin.’ In Butt and King (2012). URL http://csli-publications.stanford.edu/
LFG/17/lfg12.html.
Snijders, Liselotte (2014). ‘Non-Reflexive Binding in Warlpiri.’ In Butt and King
(2014). URL http://csli-publications.stanford.edu/LFG/19/lfg14.html.
Snijders, Liselotte (2015). The Nature of Configurationality in LFG. D.Phil. thesis,
University of Oxford. URL https://ora.ox.ac.uk/objects/uuid:1310f160-283e-
411e-a8d7-20ab4b3380c2.
Spector, Ilona (2009). ‘Hebrew Floating Quantifiers.’ In Butt and King (2009),
520–540. URL http://csli-publications.stanford.edu/LFG/14/lfg09.html.
Spencer, Andrew (2004). ‘Morphology: An Overview of Central Concepts.’ In
Sadler and Spencer (2004).
Spencer, Andrew (2005a). ‘Case in Hindi.’ In Butt and King (2005). URL http://
csli-publications.stanford.edu/LFG/10/lfg05.html.
Spencer, Andrew (2005b). ‘Word-Formation And Syntax.’ In Pavol Štekauer, and
Rochelle Lieber (eds.), Handbook of Word-Formation, Springer, 73–97.
Spencer, Andrew (2006). ‘Syntactic vs. Morphological Case: Implications for
Morphosyntax.’ In Leonid Kulikov, Andrej Malchukov, and Peter de Swart
(eds.), Case, Valency and Transitivity, Amsterdam: John Benjamins, 3–22.
Spencer, Andrew (2013). Lexical Relatedness: A Paradigm-Based Model. Oxford:
Oxford University Press.
Spencer, Andrew (2015). ‘Participial Relatives in LFG.’ In Butt and King (2015b).
URL http://csli-publications.stanford.edu/LFG/20/lfg15.html.
Spencer, Andrew (2016). ‘How To Individuate Lexemes.’ In Daniel Siddiqi, and
Heidi Harley (eds.), Morphological Metatheory, Amsterdam: John Benjamins.
Srivastav, Veneeta (1991). ‘The Syntax and Semantics of Correlatives.’ Natural
Language and Linguistic Theory 9(4): 637–686.
Steedman, Mark J. (1985). ‘Dependency and Coordination in the Grammar of
Dutch and English.’ Language 61: 523–568.
Steedman, Mark J. (1991). ‘Structure and Intonation.’ Language 67(2): 260–296.
Steedman, Mark J. (1996). Surface Structure and Interpretation. Cambridge, MA:
The MIT Press.
Steedman, Mark J. (2001). The Syntactic Process. Cambridge, MA: The MIT Press.
Steele, Joshua (1775/1779). An Essay Towards Establishing the Melody and Measure
of Speech. London: J. Almon. 2nd edn: Prosodia rationalis, 1779. Reprint of 1st
edn, Menston: Scolar Press, 1969.
800 references

Stewart, Thomas W. (2015). Contemporary Morphological Theories: A User’s Guide.

Edinburgh: Edinburgh University Press, first edition.
Stewart, Thomas W. and Gregory T. Stump (2007). ‘Paradigm Function Morphol-
ogy and the Morphology-Syntax Interface.’ In Ramchand and Reiss (2007).
Stirling, Lesley (1988). Switch-Reference and Logophoricity in Discourse
Representation Theory. Ph.D. thesis, University of Edinburgh.
Stowell, Timothy (1981). The Origins of Phrase Structure. Ph.D. thesis,
Massachusetts Institute of Technology.
Stowell, Timothy and Eric Wehrli (eds.) (1992). Syntax and the Lexicon, volume 26
of Syntax and Semantics. San Diego: Academic Press.
Strahan, Tania E. (2009). ‘Outside-In Binding of Reflexives in Insular
Scandinavian.’ In Butt and King (2009). URL http://csli-publications.stanford.
edu/LFG/14/lfg09.html.
Strahan, Tania E. (2011). ‘A Typology of Non-Local Reflexives in the Scandinavian
Languages.’ Nordic Journal of Linguistics 34: 157–178.
Strand, Kjetil (1992). Indeksering av Nomenfraser i et textforstående System
[Indexing of Nominal Phrases in a Text Understanding System]. Master’s thesis,
University of Oslo.
Stump, Gregory T. (2001). Inflectional Morphology: A Theory of Paradigm
Structure. Cambridge, UK: Cambridge University Press.
Stump, Gregory T. (2002). ‘Morphological and Syntactic Paradigms: Arguments
for a Theory of Paradigm Linkage.’ In Geert Booij, and Jaap van Marle (eds.),
Yearbook of Morphology 2001, Dordrecht: Kluwer, 147–180.
Stump, Gregory T. (2006). ‘Heteroclisis and Paradigm Linkage.’ Language 82(2):
279–322.
Stump, Gregory T. (2012). ‘The Formal and Functional Architecture of Inflectional
Morphology.’ In Angela Ralli, Geert Booij, Sergio Scalise, and Athanasios
Karasimos (eds.), On-Line Proceedings of the Eighth Mediterranean Morphology
Meeting (MMM8) Cagliari, Italy, 14–17 September 2011. Patras: University of
Patras, 254–268.
Stump, Gregory T. (2015). ‘The Interface of Semantic Interpretation and
Inflectional Realization.’ In Laurie Bauer, Lívia Körtvélyessy, and Pavol
Štekauer (eds.), Semantics of Complex Words, Heidelberg: Springer, 27–45.
Stump, Gregory T. (2016). Inflectional Paradigms: Content and Form at the
Syntax-Morphology Interface. Cambridge, UK: Cambridge University Press.
Sulger, Sebastian (2009). ‘Irish Clefting and Information-Structure.’ In Butt and
King (2009), 562–582. URL http://csli-publications.stanford.edu/LFG/14/
lfg09.html.
Sulger, Sebastian (2011). ‘A Parallel Analysis of Have-Type Copular Constructions
in Two Have-less Indo-European Languages.’ In Butt and King (2011),
503–523. URL http://csli-publications.stanford.edu/LFG/16/lfg11.html.
Sulger, Sebastian (2012). ‘Nominal Argument Structure and the Stage-/Individual-
Level Contrast in Hindi-Urdu.’ In Butt and King (2012), 582–602. URL http://
csli-publications.stanford.edu/LFG/17/lfg12.html.
Sulger, Sebastian (2015). Modeling Nominal Predications in Hindi/Urdu. Ph.D.
thesis, Universität Konstanz.
 references 801

Sung, Won-Kyung (1996). Etude des pronoms coréens [A Study of Korean

Pronouns]. Ph.D. thesis, Université Paris 7.
Szűcs, Peter (2014). ‘Information Structure and the English Left Periphery.’ In
Butt and King (2014). URL http://csli-publications.stanford.edu/LFG/19/lfg14.
html.
Tang, Chih-Chen Jane (1989). ‘Chinese Reflexives.’ Natural Language and
Linguistic Theory 7(1): 93–121.
Tesnière, Lucien (1959). Élements de Syntaxe Structurale. Paris: Klinksieck.
Thomas, Victoria (2012). Double Object Constructions and ‘Bill’ Verbs in English.
M.Phil. dissertation, University of Oxford.
Thomason, Richmond (ed.) (1974). Formal Philosophy: Selected Papers of Richard
Montague. New Haven: Yale University Press.
Þráinsson, Höskuldur (1986). ‘On Auxiliaries, AUX and VP’s in Icelandic.’ In Lars
Hellan, and Kirsti Koch Christensen (eds.), Topics in Scandinavian Syntax,
Dordrecht: D. Reidel, 235–265.
Toivonen, Ida (2000). ‘The Morphosyntax of Finnish Possessives.’ Natural
Language and Linguistic Theory 18(3): 579–609.
Toivonen, Ida (2001). ‘Language Change, Lexical Features and Finnish Possessors.’
In Butt and King (2001b).
Toivonen, Ida (2003). Non-Projecting Words: A Case Study of Swedish Verbal
Particles. Dordrecht: Kluwer.
Toivonen, Ida (2013). ‘English Benefactive NPs in LFG.’ In Butt and King (2013).
URL http://csli-publications.stanford.edu/LFG/18/lfg13.html.
Uyechi, Linda and Lian Hee Wee (eds.) (2010). Reality Exploration and Discovery:
Pattern Interaction in Language & Life. A Festschrift for K. P. Mohanan.
Stanford: CSLI Publications.
Vallduví, Enric (1992). The Informational Component. Outstanding Dissertations
in Linguistics. New York: Garland.
Vallduví, Enric and Elisabet Engdahl (1996). ‘The Linguistic Realization of
Information Packaging.’ Linguistics 34: 459–519.
Vallduví, Enric and Maria Vilkuna (1998). ‘On Rheme and Contrast.’ In Peter
Culicover, and Louise McNally (eds.), The Limits of Syntax, San Diego:
Academic Press, volume 29 of Syntax and Semantics.
van Benthem, Johan and Alice ter Meulen (eds.) (1997). Handbook of Logic
and Language. Cambridge, MA: The MIT Press. Co-published with Elsevier
Science B.V., Amsterdam.
van Benthem, Johan and Alice ter Meulen (eds.) (2011). Handbook of Logic and
Language. Amsterdam: Elsevier, second edition.
van Genabith, Josef and Richard Crouch (1999a). ‘Dynamic and Underspecified
Semantics for LFG.’ In Dalrymple (1999).
van Genabith, Josef and Richard Crouch (1999b). ‘How to Glue a Donkey to an
F-Structure: Porting a Dynamic Meaning Representation Into LFG’s Linear
Logic Based Glue-Language Semantics.’ In Harry Bunt, and Reinhard Muskens
(eds.), Computing Meaning, Dordrecht: Kluwer, volume 1, 129–148.
van Urk, Coppe (2013). ‘Visser’s Generalization: The Syntax of Control and the
Passive.’ Linguistic Inquiry 44(1): 168–178.
802 references

Van Valin, Robert D. (2003). Exploring the Syntax-Semantics Interface. Cambridge,

UK: Cambridge University Press.
Vincent, Nigel (1999). ‘The Evolution of C-Structure: Prepositions and PPs From
Indo-European to Romance.’ Linguistics 37(6): 1111–1153.
Vincent, Nigel (2001a). ‘Competition and Correspondence in Syntactic Change:
Null Arguments in Latin and Romance.’ In Susan Pintzuk, George Tsoulas, and
Anthony Warner (eds.), Diachronic Syntax: Models and Mechanisms, Oxford:
Oxford University Press, 25–50.
Vincent, Nigel (2001b). ‘LFG as a Model of Syntactic Change.’ In Butt and King
(2001b), 1–42.
Vincent, Nigel and Kersti Börjars (2007). ‘Feature Resolution and the Content of
Features.’ In Zaenen et al. (2007), 293–315.
Vincent, Nigel and Kersti Börjars (2010a). ‘Complements of Adjectives: A
Diachronic Approach.’ In Butt and King (2010), 458–478. URL http://csli-
publications.stanford.edu/LFG/15/lfg10.html.
Vincent, Nigel and Kersti Börjars (2010b). ‘Grammaticalization and Models
of Language.’ In Elizabeth Closs Traugott, and Graeme Trousdale (eds.),
Gradience, Gradualness and Grammaticalization, Amsterdam: John Benjamins,
279–299.
Visser, Fredericus T. (1963–1973). An Historical Syntax of the English Language.
Leiden: Brill.
von Stechow, Arnim (1982). ‘Structured Propositions.’ Technical report,
Universität Konstanz SFB. URL http://www.sfs.uni-tuebingen.de/~astechow/
Aufsaetze/Structured%20Prop%201.pdf.
Wada, Hajime and Nicholas Asher (1986). ‘BUILDRS: An Implementation
of DR Theory and LFG.’ In Proceedings of the International Conference on
Computational Linguistics (COLING86). Bonn: International Committee on
Computational Linguistics, 540–545.
Ward, Gregory L. (1990). ‘Discourse Functions of VP Preposing.’ Language 66(4):
742–763.
Wasow, Thomas (1972). Anaphoric Relations in English. Ph.D. thesis, Mass-
achusetts Institute of Technology.
Wasow, Thomas (1979). Anaphora in Generative Grammar. Ghent: E. Story-
Scientia.
Way, Andy (1999). ‘A Hybrid Architecture for Robust MT Using LFG-DOP.’
Journal of Experimental and Theoretical Artificial Intelligence 11(4). Special
Issue on Memory-Based Language Processing.
Way, Andy (2001). LFG-DOT: A Hybrid Architecture for Robust MT. Ph.D. thesis,
University of Essex.
Way, Andy (2003). ‘Translating with Examples: The LFG-DOT Models of
Translation.’ In Michael Carl, and Andy Way (eds.), Recent Advances in
Example-Based Machine Translation, Dordrecht: Kluwer, 443–472.
Webelhuth, Gert (1992). Principles and Parameters of Syntactic Saturation. Oxford:
Oxford University Press.
Webelhuth, Gert (ed.) (1995). Government and Binding Theory and the Minimalist
Program. Oxford: Oxford University Press.
 references 803

Wechsler, Stephen (2004). ‘Number as Person.’ In Olivier Bonami, and Patricia

Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics 5. Paris:
Colloque de Syntaxe et Sémantique à Paris, 255–274. URL http://www.cssp.
cnrs.fr/eiss5/.
Wechsler, Stephen (2008). “Elsewhere’ in Gender Resolution.’ In Kristin Hanson,
and Sharon Inkelas (eds.), The Nature of the Word: Essays in Honor of Paul
Kiparsky, Cambridge, MA: The MIT Press, 567–586.
Wechsler, Stephen (2011). ‘Mixed Agreement, The Person Feature, and the
Index/Concord Distinction.’ Natural Language and Linguistic Theory 29(4):
999–1031.
Wechsler, Stephen (2015). ‘The Syntactic Role of Agreement.’ In Kiss and
Alexiadou (2015), 309–342.
Wechsler, Stephen and Hyun-Jong Hahm (2011). ‘Polite Plurals and Adjective
Agreement.’ Morphology 21(2): 247–281. doi:10.1007/s11525-010-9168-z.
Wechsler, Stephen and Larisa Zlatić (2003). The Many Faces of Agreement.
Stanford: CSLI Publications.
Wechsler, Stephen and Larisa Zlatić (2012). ‘The Wrong Two Faces.’ Language
88(2): 380–387.
Wedekind, Jürgen (1995). ‘Some Remarks on the Decidability of the Generation
Problem in LFG- and PATR-Style Unification Grammars.’ In Proceedings
of the 33rd Annual Meeting of the ACL. Cambridge, MA: Association for
Computational Linguistics, 45–52.
Wedekind, Jürgen (1999). ‘Semantic-Driven Generation With LFG- and PATR-
Style Grammars.’ Computational Linguistics 25(2): 277–281.
Wedekind, Jürgen (2006). ‘On Some Formal Properties of LFG Generation.’ In
Butt et al. (2006), 53–72.
Wedekind, Jürgen (2014). ‘On the Universal Generation Problem for Unification
Grammars.’ Computational Linguistics 40(3): 533–538.
Wedekind, Jürgen and Ronald M. Kaplan (1993). ‘Type-Driven Semantic
Interpretation of F-Structures.’ In Proceedings of the 6th Meeting of the EACL.
Utrecht: European Association for Computational Linguistics, 404–411.
Wedekind, Jürgen and Ronald M. Kaplan (1996). ‘Ambiguity-Preserving
Generation With LFG- and PATR-Style Grammars.’ Computational Linguistics
22(4): 555–568.
Wedekind, Jürgen and Ronald M. Kaplan (2012). ‘LFG Generation by Grammar
Specialization.’ Computational Linguistics 38(4): 867–915.
Weil, Henri (1844). De l’ordre des mots dans les langues anciennes comparées aux
langues modernes: question de grammaire général. Paris: A. Franck. English
translation by Charles W. Super, published by Ginn & Company, Boston, 1887,
with the title The Order of Words in the Ancient Languages Compared with That
of the Modern Languages.
Wescoat, Michael T. (2002). On Lexical Sharing. Ph.D. thesis, Stanford University.
Wescoat, Michael T. (2005). ‘English Nonsyllabic Auxiliary Contractions: An
Analysis in LFG With Lexical Sharing.’ In Butt and King (2005). URL http://
csli-publications.stanford.edu/LFG/10/lfg05.html.
804 references

Wescoat, Michael T. (2007). ‘Preposition-Determiner Contractions: An Analysis

in Optimality-Theoretic Lexical-Functional Grammar With Lexical Sharing.’
In Butt and King (2007). URL http://csli-publications.stanford.edu/LFG/12/
lfg07.html.
Wescoat, Michael T. (2009). ‘Udi Person Markers and Lexical Integrity.’ In Butt
and King (2009), 604–622. URL http://csli-publications.stanford.edu/LFG/14/
lfg09.html.
Wescoat, Michael T. and Annie Zaenen (1991). ‘Lexical Functional Grammar.’ In
Flip G. Droste, and John E. Joseph (eds.), Linguistic Theory and Grammatical
Description, Amsterdam: John Benjamins, volume 75 of Current Issues in
Linguistic Theory, 103–136.
Wheeldon, Linda (2000). ‘Generating Prosodic Structure.’ In Linda Wheeldon
(ed.), Aspects of Language Production, Philadelphia: Psychology Press, 249–274.
Williams, Edwin (1983). ‘Against Small Clauses.’ Linguistic Inquiry 14: 237–308.
Williams, Edwin (1990). ‘The ATB-Theory of Parasitic Gaps.’ Linguistic Review 6:
265–279.
Williams, Edwin (1992). ‘Adjunct Control.’ In Larson et al. (1992), 297–322.
Winter, Yoad (1998). Flexible Boolean Semantics. Ph.D. thesis, Utrecht University.
Winter, Yoad (2001). Flexibility Principles in Boolean Semantics. Cambridge, MA:
The MIT Press.
Winter, Yoad (2007). ‘Multiple Coordination: Meaning Composition vs. the
Syntax-Semantics Interface.’ URL http://www.phil.uu.nl/~yoad/papers/
WinterMultipleCoordination.pdf. Unpublished manuscript, Technion.
Wisdom, John (1968). Other Minds. Berkeley: University of California Press.
Woods, William A. (1970). ‘Transition Network Grammars for Natural Language
Analysis.’ Communications of the ACM 13(10): 591–606.
Wunderlich, Dieter (1988). ‘Some Problems of Coordination in German.’ In Reyle
and Rohrer (1988), 289–316.
Yadroff, Mikhail (1992). ‘The Syntactic Properties of Adjunction (Scrambling in
Russian).’ Unpublished manuscript, Indiana University.
Yamamoto, Kazuyuki (1996). ‘The Tough Constructions in English and Japanese:
A Lexical-Functional Grammar Approach.’ In Akira Ikeya (ed.), Linguistic
Workshop Series 3: The Tough Constructions in English and Japanese. Tokyo:
Kuroiso, 43–61.
Zaenen, Annie (1980). Extraction Rules in Icelandic. Ph.D. thesis, Harvard
University. Reprinted by Garland Press, New York, 1985.
Zaenen, Annie (ed.) (1982). Subjects and Other Subjects: Proceedings of the Harvard
Conference on the Representation of Grammatical Relations. Bloomington:
Indiana University Linguistics Club.
Zaenen, Annie (1983). ‘On Syntactic Binding.’ Linguistic Inquiry 14(3): 469–504.
Zaenen, Annie (1993). ‘Unaccusativity in Dutch: Integrating Syntax and Lexical
Semantics.’ In James Pustejovsky (ed.), Semantics and the Lexicon, Dordrecht:
Kluwer, 129–161.
Zaenen, Annie and Dick Crouch (2009). ‘OBLs Hobble Computations.’ In Butt
and King (2009), 644–654. URL http://csli-publications.stanford.edu/LFG/14/
lfg09.html.
 references 805

Zaenen, Annie and Elisabet Engdahl (1994). ‘Descriptive and Theoretical Syntax in
the Lexicon.’ In Beryl T. S. Atkins, and Antonio Zampolli (eds.), Computational
Approaches to the Lexicon, Oxford: Oxford University Press, 181–212.
Zaenen, Annie and Ronald M. Kaplan (1995). ‘Formal Devices for Linguistic
Generalizations: West Germanic Word Order in LFG.’ In Jennifer S. Cole, Geor-
gia M. Green, and Jerry L. Morgan (eds.), Linguistics and Computation, Stan-
ford: CSLI Publications, 3–27. Reprinted in Dalrymple et al. (1995b, 215–239).
Zaenen, Annie and Ronald M. Kaplan (2002). ‘Subsumption and Equality:
German Partial Fronting in LFG.’ In Butt and King (2002). URL http://csli-
publications.stanford.edu/LFG/7/lfg02.html.
Zaenen, Annie and Lauri Karttunen (1984). ‘Morphological Non-Distinctiveness
and Coordination.’ In Mark Cobler, Susannah MacKaye, and Michael T.
Wescoat (eds.), Proceedings of the Eastern States Conference on Linguistics
(ESCOL ’84). 309–320.
Zaenen, Annie and Joan Maling (1983). ‘Passive and Oblique Case.’ In Levin et al.
(1983), 159–191.
Zaenen, Annie and Joan Maling (1990). ‘Unaccusative, Passive, and Quirky Case.’
In Maling and Zaenen (1990), 187–234.
Zaenen, Annie, Joan Maling, and Höskuldur Þráinsson (1985). ‘Case and
Grammatical Functions: The Icelandic Passive.’ Natural Language and Linguistic
Theory 3(4): 441–483. Reprinted in Maling and Zaenen (1990, 95–164).
Zaenen, Annie, Jane Simpson, Tracy Holloway King, Jane Grimshaw, Joan Maling,
and Christopher Manning (eds.) (2007). Architectures, Rules, and Preferences:
Variations on Themes by Joan W. Bresnan. Stanford: CSLI Publications.
Zec, Draga (1987). ‘On Obligatory Control in Clausal Complements.’ In Iida et al.
(1987), 139–168.
Zhang, Yanyin (2005). ‘Processing and Formal Instruction in the L2 Acquisition
of Five Chinese Grammatical Morphemes.’ In Pienemann (2005a), 155–177.
Zoller, Claus Peter (1983). Die Sprache der Rang Pas von Garhwal (Ran Pɔ Bhāsa):
Grammatik, Texte, Wörterbuch [The Language of the Rang Pas of Garhwal:
Grammar, Texts, Dictionary]. Wiesbaden: Harrassowitz.
Zwicky, Arnold M. (1990). ‘Syntactic Representations and Phonological Shapes.’
In Inkelas and Zec (1990), 379–397.
Zwicky, Arnold M. (1992). ‘Some Choices in the Theory of Morphology.’ In
Robert D. Levine (ed.), Formal Grammar: Theory and Implementation, Oxford:
Oxford University Press, 327–371.
Zybatow, Gerhild and Grit Mehlhorn (2000). ‘Experimental Evidence for Focus
Structure in Russian.’ In Tracy Holloway King, and Irina Sekerina (eds.),
Formal Approaches To Slavic Linguistics: The Philadelphia Meeting 1999. FASL,
Ann Arbor: Michigan Slavic Publications, 414–434.
Zymla, Mark-Matthias, Maike Müller, and Miriam Butt (2015). ‘Modelling the
Common Ground for Discourse Particles.’ In Butt and King (2015b). URL
http://csli-publications.stanford.edu/LFG/20/lfg15.html.
 Author Index

Abney, Stephen P. 100 Attia, Mohammed 33, 189, 192, 195

Ackerman, Farrell 136, 324, 338, 339, 351, Austin, Peter 6, 112, 115, 128, 134,
352, 355, 364, 442 184, 502
Acquaviva, Paolo 454 Avgustinova, Tania 189
Adger, David 91
Ågren, Malin 727 Baader, Franz 209
Ahmed, Tafseer 102, 247, 348, 498 Baayen, Harald 355
Aissen, Judith 114 Bach, Emmon 276, 586
Al Sharifi, Budour 247, 320, 498 Backofen, Rolf 209
Allbritton, David W. 397 Baerman, Matthew 55, 451, 453
Allen, Cynthia 88, 354, 729 Baker, Brett 352
Alsina, Alex 16, 27, 30–2, 37, 69, 88, 115, Baker, Mark C. 21, 25
122, 136, 197, 323–5, 329, 332, 334, 338, Ball, Douglas 352
346–8, 355, 356, 364, 376, 544, 725 Baltin, Mark 130
Anderson, Stephen R. 436 Banfield, Ann 113
Andréasson, Maia 380, 394 Barron, Julia 601, 682
Andrews, Avery D., III 15, 16, 21, 25, 39, Barry, Guy 700, 702
49, 88, 131, 275, 279, 320, 352, 355, 436, Barwise, Jon 268, 281, 284, 286, 645
440, 457, 458, 468–70, 499, 548, 549, Bary, Corien 287, 314, 320, 590
563, 588, 600, 650 Bayer, Samuel 621
Aranovich, Raúl 334 Beard, Robert 436
Arka, I Wayan 29, 31, 65, 66, 78, 177, Beaver, David I. 519
335, 336, 352, 355–7, 365, 512, 556, Beckman, Mary E. 403
587, 600 Beesley, Kenneth R. 436, 734
Arnold, Doug 103–5, 127, 287, 320, 474, Bell, Sara 11, 21
498, 642, 713–16 Belyaev, Oleg I. 69, 71, 96, 115, 197, 518,
Arnold, Jennifer 66, 365 544, 650, 671, 721, 725, 728
Arnon, Inbal 355 Bender, Emily 2, 11, 41, 47, 267
Arsenijević, Boban 69 Bergmann, Merrie 643
Artoni, Daniele 728 Berman, Judith 19, 22, 30, 116, 344, 345,
Arvaniti, Amalia 396 555, 665, 697, 698, 700
Asher, Nicholas 287 Bettoni, Camilla 727, 728
Asudeh, Ash 2, 6, 11, 15, 30, 35–7, 43, 44, Bickel, Balthasar xix, 63
69, 88, 89, 97, 99, 104, 108, 112, 113, Biggs, Bruce 189
116, 119–22, 130, 131, 134–8, 168, 172, Biswas, Alexander 470
176, 177, 181, 194, 197, 225, 231, 233, Blevins, James P. 57
237, 243, 247, 253–6, 259, 275, 296, 319, Bloomfield, Leonard 90, 112
320, 327–9, 331, 334–7, 339, 342, 352, Bock, Kathryn 727
353, 356, 359–65, 376, 377, 386, 407, Bod, Rens 727
470, 471, 502, 503, 505, 510, 512–14, Bodomo, Adams B. 352
517, 543, 544, 550, 552, 557, 559–61, Bögel, Tina 96, 103, 115, 348, 395, 400,
565, 569, 570, 576, 577, 584, 593, 621, 404, 405, 410, 412, 435
624–7, 630, 631, 634, 635, 644, 659, 677, Bolinger, Dwight L. 370
683, 684, 686, 687, 689–92, 696–8, 700, Bonami, Olivier 470
703, 716–19, 721, 724, 725 Bond, Oliver 69, 436
808 author index

Borer, Hagit 135, 445 Chisarik, Erika 36, 37, 99, 364, 725
Börjars, Kersti 6, 26, 28, 59, 63, 71, 73, 75, Choi, Hye-Won 371–3, 376, 377
76, 99, 101, 116, 136, 221, 222, 247, 354, Chomsky, Noam 1, 21, 35, 43, 90, 91, 96,
364, 453, 498, 636, 639, 729, 730 97, 99, 107, 135, 247, 330, 334, 370, 398,
Borschev, Vladimir 707 447, 706
Börschinger, Benjamin 352 Chung, Sandra 687
Borsley, Robert D. 603 Cinque, Guglielmo 99, 656
Bos, Johan 287 Clark, Brady Zack 681
Bošković, Željko 398 Clément, Lionel 734
Bouchard, Denis 703 Clements, George N. 687, 688
Bouma, Gosse 695 Colban, Erik 498
Brame, Michael 100 Comrie, Bernard xix, 10, 11, 20, 63,
Bresnan, Joan 2, 6, 11, 14, 15, 17, 21, 26, 366, 816
30, 34–7, 41–4, 50, 52, 53, 61, 64, 69, 88, Cook, Phillipa 93
89, 97, 99, 101, 104, 108, 112, 115, 116, Cooper, Robin 268, 284
118–21, 128, 130–2, 134–8, 140, 147, Coppock, Elizabeth 502, 730
149, 150, 156, 161, 163, 164, 166, 168, Corbett, Greville G. 55, 74, 451, 639
172, 176, 177, 181, 183, 184, 194, 197, Corblin, Francis 522
211, 212, 225, 231, 235, 238, 245–7, 253, Cormons, Boris 727
259, 264, 278, 297, 322, 324, 329–38, Covington, Michael A. 9
341, 342, 345–7, 353, 355, 356, 358–60, Crouch, Richard 106, 198, 225, 237, 249,
364, 374, 375, 386, 452, 470, 471, 500–5, 250, 287, 293, 305, 319, 321, 359, 517,
510, 512–14, 543–5, 547, 550, 552, 561, 619, 624–7, 631, 634, 635, 644, 724,
565, 576, 577, 585, 586, 589, 593–5, 602, 734, 735
605, 621, 650, 652, 659, 665, 666, 669, Crysmann, Berthold 470
673, 676, 680, 682, 687, 694, 696–8, 700, Cueni, Anna 355
703, 725–8, 730, 732 Culicover, Peter W. 3, 273
Broadwell, George Aaron 6, 104, 115, 136, Culy, Christopher D. 509, 510, 544
324, 352, 495, 725, 728 Curnow, Timothy Jowan 190
Brown, Dunstan 55, 436, 451
Brown, Lea 650 Dahl, Östen 370
Brun, Caroline 244, 613 Dahlstrom, Amy 25, 28, 138, 184, 365, 394
Bürckert, Hans-Jürgen 209 Dale, Robert 352
Butt, Miriam 6, 21, 28, 32–4, 60, 61, 63–5, Dalrymple, Mary 2, 6, 14, 19, 28, 30–3, 35,
67, 79, 88, 96, 102, 103, 115, 120, 138, 37, 69, 71–5, 78, 81, 82, 96, 106, 114–16,
195, 237, 314, 324, 326, 327, 329, 335, 121, 132, 136, 138, 189, 192, 193, 196–8,
346–52, 356, 360, 361, 363, 365, 372, 209, 219, 220, 225, 232, 233, 237, 249,
373, 376–81, 383, 394, 400–5, 417, 418, 250, 264, 278, 280, 284, 287, 301, 302,
436, 438, 443–5, 447–50, 457, 458, 492, 305, 313, 319, 320, 326, 327, 352, 355,
498, 502, 601, 666, 671, 680, 725, 360, 361, 365, 368, 370, 373, 374, 379,
729, 735 381–3, 385, 386, 393, 406–9, 412,
416–18, 435, 439, 503, 505, 507,
Camilleri, Maris 684 510–12, 515, 517–19, 523, 539, 543,
Carnie, Andrew 6, 115, 754 544, 570, 571, 605, 619–21, 624, 630,
Cavar, Damir 116, 734 636, 639, 642, 643, 651, 652, 680–2, 687,
Çetinoğlu, Özlem 28, 352, 365 697–703, 716, 717, 730, 734
Champollion, Lucas 643, 645 Daneš, František 370
Chapman, Carol 453, 730 Davidson, Donald 314
Chen, Matthew Y. 398 Davis, Anthony 364
Chierchia, Gennaro 569, 570 Davison, Alice 102
 author index 809

de Alencar, Leonel Figueiredo 6, 544 Foley, William A. 66

de Marneffe, Marie-Catherine 355 Fong, Cynthia 397
Deguchi, Masanori 678 Ford, Kevin C. 687, 688
Demers, Richard A. 166 Ford, Marilyn 355, 727
Dench, Alan Charles 466 Forst, Martin 30, 32, 96, 237, 613
Deng, Dun-Hong 335, 337 Fortmann, Christian 436, 447–50
Deo, Ashwini 730 Fox, Anthony 399
Di Biase, Bruno 727 Francez, Nissim 722
Dik, Helma 179 Frank, Anette 6, 13, 106, 143, 264, 319,
Dingare, Shipra 166 326, 327, 352, 360, 361, 436, 447,
Dione, Cheikh Bamba 195, 380, 394 449–51, 607–9
Dipper, Stefanie 88, 116 Franks, Steven 152
Dixon, Robert M. W. 38 Frantz, Donald G. 16, 18, 21
Dobrovie-Sorin, Carmen 83 Frege, Gottlob 276
Doron, Edit 683 Frey, Werner 287
Dowty, David R. 13, 41, 42, 314, 324–6, Frota, Sónia 400
365, 570, 643 Fry, John 314, 319, 320
Dras, Mark 352 Fujii, Mamoru 732
Dryer, Matthew S. 25, 27, 650, 816
Duncan, Lachlan 88, 352 Gabbay, Dov M. 273
Dyvik, Helge 33, 96, 189, 192, 193, 196, Gamut, L. T. F. 5, 281, 282, 284, 720
197, 264 Gawron, Jean Mark 287, 724
Gazdar, Gerald 41, 144, 248, 620, 621, 634
Embick, David 135, 445 Gazdik, Anna 394, 650, 651, 725
Emms, Martin 634 Georgopoulos, Carol 24, 687
Emonds, Joseph 1 Gerassimova, Veronica A. 114
Engdahl, Elisabet 17, 19, 30, 116, 337, 356, Geuder, Wilhelm 348
370, 378, 705, 706 Ginzburg, Jonathan 233, 695, 722–4
Erteschik-Shir, Nomi 5, 226, 369, 370, 381 Giorgolo, Gianluca 237, 275, 320, 327–9,
Estigarribia, Bruno 131 339, 342, 360–3, 365, 716, 735
Evans, Gareth 517 Girard, Jean-Yves 298, 319, 320
Givón, Talmy 25
Falk, Yehuda N. 6, 19, 21–4, 29, 33, 36–8, Glahn, Esther 728
53, 58, 59, 63, 66, 67, 83, 99, 106, 113, Glasbey, Sheila 65, 314
120, 144, 189, 194, 197, 264, 324, 336, Goodall, Grant 615
352, 353, 356, 363, 471, 550, 552, 565, Grano, Thomas 584
600, 601, 663, 666, 669, 676, 677, 683, Greenbaum, Sidney 29, 60, 158
684, 696, 703–6, 724, 725 Greenberg, Joseph H. 9, 57
Faltz, Leonard M. 543, 634 Grimm, Scott 355
Fassi-Fehri, Abdulkader 100, 375, 653 Grimshaw, Jane 3, 43, 96, 131, 544
Féry, Caroline 381 Groenendijk, Jeroen 518, 522, 538, 722
Fiengo, Robert 681 Groos, Anneke 80
Fillmore, Charles J. 17, 102, 127, 330 Gruber, Jeffrey S. 330
Findlay, Jamie Y. 6, 361, 365 Gundel, Jeanette K. 370
Finn, Regina 727 Gunlogson, Christine 420
Finn, Ríona 727 Guo, Yuqing 102
Firbas, Jan 370 Gupta, Vineet 301, 319, 320
Flickinger, Daniel P. 231
Flouraki, Maria 287 Hagége, Claude 544
Fodor, Janet D. 696 Hahm, Hyun-Jong 78
810 author index

Hajičová, Eva 370 Jackendoff, Ray 3, 96, 97, 247, 273, 324,
Håkansson, Gisela 728 326, 370, 377
Hale, Kenneth 40, 41, 100 Jacobson, Pauline 549, 556, 564, 570, 600,
Hale, Mark 729 681
Halle, Morris 135, 398, 445 Jaeger, T. Florian 114
Halliday, M. A. K. 370 Jaeggli, Osvaldo 131
Halpern, Aaron 115 Jakobson, Roman 57
Halvorsen, Per-Kristian 266–9, 278–80, Jam, R. 225
286, 297, 494, 503–5, 556, 569 Jansen, Louise 727
Hamblin, C. L. 721, 722 Janssen, Theo M. V. 276
Hammarberg, Björn 728 Jelinek, Eloise 166, 184, 502
Harries, Pauline 729 Jeschull, Liane 622
Haspelmath, Martin 6, 63, 621, 622, 816 Johannessen, Janne Bondi 603
Haug, Dag T. T. 69, 179, 197, 287, 314, 316, Johnson, David E. 3, 37
320, 500, 518–21, 523, 534, 544, 556, Johnson, Mark 138, 164, 246, 728, 734
583, 584, 587, 590, 601, 671, 721, 725 Jones, Stephen M. 78
Hautli, Annette 348 Joshi, Smita 338, 348
Hay, Jennifer 355 Junghanns, Uwe 95
Hayes, Bruce 396, 401, 407, 418–20
Hearne, Mary 727 Kaisse, Ellen M. 397
Heim, Irene 518 Kallas, Krystyna 650, 651
Hellan, Lars 14, 505–7, 509, 511, 515 Kameyama, Megumi 118, 257, 259, 514
Henadeerage, Kumara 503 Kamp, Hans 281, 287, 315, 476–9, 518,
Heny, Frank 494 519, 522
Her, One-Soon 335, 337, 338, 354 Kanazawa, Makoto 571
Heycock, Caroline 643 Kanerva, Jonni M. 21, 26, 247, 331–4, 338,
Higginbotham, James 526, 570, 571, 722 345–7
Hinds, John 396 Kaplan, Ronald M. 2, 6, 14, 30, 50, 52, 53,
Hinrichs, Angie 278 59, 61, 72–5, 81, 93, 94, 96, 106–8, 114,
Hippisley, Andrew 436 115, 131, 132, 138, 140–2, 147, 149, 150,
Hirschberg, Julia 403 163, 164, 171, 198, 201, 203–5, 208, 209,
Hoeksema, Jacob 643, 644 211, 212, 219, 220, 222, 223, 225, 232,
Holmen, Anne 728 237, 238, 241–7, 249, 250, 252, 257, 259,
Hong, Ki-Sun 394 264, 266, 267, 273, 278, 280, 286, 297,
Hornstein, Norbert 656 324, 329, 330, 351, 404–6, 410, 436, 438,
Horrack, Kate 352 443–5, 457, 458, 471, 514, 553, 554, 564,
Hristov, Bozhil 69, 71, 78, 80, 218, 603, 602, 605, 619, 636, 639, 650–2, 662, 663,
636, 642, 664 666, 673, 680, 682, 687, 694, 696–701,
Huang, C.-T. James 15 726, 727, 730–4
Huang, Chu-Ren 138, 678, 725 Karttunen, Lauri 56, 80, 436, 518, 522,
Huddleston, Rodney 63, 609, 623 722, 734
Hudson, Richard A. 9, 25 Kasami, Tadao 732
Hurst, Peter 544 Kasper, Robert 473, 484, 485, 489
Hvenekilde, Anne 728 Kathol, Andreas 69
Kawaguchi, Satomi 727
Iida, Masayo 364 Kay, Martin 149, 204, 436, 734
Inkelas, Sharon 400 Kay, Paul 2, 3, 127, 272
Ishikawa, Akira 351 Kazana, Despina 287
Itô, Junko 400 Keenan, Edward L. 10, 11, 20, 284, 634
Izvorski, Roumyana 125 Kehler, Andrew 624, 630
 author index 811

Keller, Bill 209 Ladd, D. Robert 6, 396, 403, 420, 426

Kelling, Carmen 345, 364, 544 Lahiri, Aditi 95, 352, 396, 398, 401, 404,
Kendall, Tyler 355 407, 418–20, 423, 601, 729
Kennedy, Christopher 477 Lakoff, George 14, 703
Kent, Kateryna 443 Lam, Olivia S.-C. 88
Keßler, Jörg-U. 728 Lambek, Joachim 301
Khavronina, S. 95 Lambrecht, Knud 5, 370, 381, 385
Kibort, Anna 22, 24, 331, 337–9, 341–6, Lamping, John 278, 280, 287, 301, 302,
353–6, 360, 365 305, 319, 320, 517, 539, 624, 630,
Kim, Boomee 681 716, 717
Kim, Yookyung 571 Landau, Idan 584
King, Tracy Holloway 6, 21, 32–4, 60, 61, Landman, Fred 644
63–5, 67, 69, 71, 79, 81, 82, 88, 89, 94, Lapata, Maria 503
96, 99, 100, 102, 106–8, 110, 111, Lapointe, Steven 135
114–16, 119, 122, 124, 126, 129, 138, Lareau, François 352
141, 145, 152, 185, 189, 192, 193, 195–8, Larson, Richard K. 477
205, 206, 225, 232, 237, 249, 250, 350, Lasersohn, Peter 645
351, 366, 372, 373, 375–81, 383, 393, Lasnik, Howard 526, 681
394, 400–5, 417, 418, 492, 498, 565, 619, Laughren, Mary 112
643, 659, 666, 671, 677, 680–2, Lawyer, Lewis 650
697–703, 725, 734, 735 Lee, Hanjung 138, 728
Kinyon, Alexandra 734 Leech, Geoffrey 29, 60, 158
Kiparsky, Carol 19, 545 Legate, Julie Anne 112
Kiparsky, Paul 19, 231, 316, 545 LeSourd, Philip 136
Kirby, Susannah 355 Levelt, Willem J. M. 400, 410, 727
Kiss, Tibor 6 Levin, Beth 18, 338, 339, 364
É. Kiss, Katalin 121 Levin, Lori S. 21, 53, 331
Kitagawa, Yoshihisa 678 Levine, Robert D. 2, 47
Klein, Ewan 41 Liao, Gwen W.-T. 313, 385
Klein, Wolfgang 315 Liberman, Mark 398
Kodio, Koungarma 544 Link, Godehard 644
Koenig, Jean-Pierre 364 Lipson, Alexander 223
Kokkonidis, Miltiadis 320, 518 Lødrup, Helge 6, 19, 30–2, 116, 132,
Koster, Jan 247 336, 345, 353, 365, 503, 505, 543,
Krifka, Manfred 316, 370, 381, 382, 645, 600, 601
680, 722 Lorenz, David 703
Kroeger, Paul 6, 16, 21, 66, 89, 93, 95, 99, Lovestrand, Joseph 116, 621
108, 112, 113, 115, 119, 335, 565, 658, Lowe, John J. 69, 71, 95, 96, 104, 114–16,
659, 661, 669 136, 197, 247, 252, 287, 314, 320, 348,
Kruijff-Korbayová, Ivana 370 351, 364, 373, 394, 406, 407, 409, 410,
Krylova, O. 95 412–15, 417, 423, 426, 429, 430, 435,
Kuhn, Jonas 106, 134, 237, 278, 319, 613, 452, 453, 485, 498, 518, 519, 523, 590,
642, 728 601, 621, 723, 728–30
Kuiper, Koenraad 6 Lund, Karen 728
Kuno, Susumu 370, 701
Kwon, Song-Nim 393 Magnani, Marco 728
Mahowald, Kyle 138, 185
Laczkó, Tibor 33, 36, 37, 68, 69, 96, Maling, Joan 21, 25, 43, 44, 247, 330, 331,
101, 103, 107, 121, 126, 194, 197, 345, 503–5, 549, 703
338, 364 Malouf, Robert 695
812 author index

Manning, Christopher D. 23, 166, Moot, Richard 301

275, 324, 335, 352, 499, 613–15, Moravcsik, Edith A. 20, 57
617, 618 Morimoto, Yukiko 346, 365, 728
Mansouri, Fethi 727 Morrill, Glyn V. 2, 301, 634
Marantz, Alec 135, 445 Morrison, Andrew 499
Marcotte, Jean-Philippe 116, 443, 621 Moshi, Lioba 26, 247, 336, 341, 342,
Markantonatou, Stella 353, 364 355, 356
Mastenbroek, Elsbeth 287 Motazedi, Yasaman 352
Masuichi, Hiroshi 96, 735 Mugane, John 116, 452
Mathesius, Vilém 370 Müller, Maike 394
Matsumoto, Yo 20, 136, 351 Müller, Stefan 6
Matsumura, Takashi 732 Musgrave, Simon 88, 357
Maxwell, John T., III 2, 49, 96, 106, 115, Muskens, Reinhard 287
142, 198, 204, 208, 209, 219, 223, 225, Mycock, Louise 114–16, 124, 125, 252,
237, 241, 246, 249, 250, 264, 351, 404, 373, 378, 379, 381, 394, 403, 406–9,
405, 602, 613–15, 617–19, 651, 652, 412–18, 423, 426, 427, 429, 430, 435,
662, 663, 687, 694, 730–2, 734 664, 677–80, 723, 724
May, Robert 556
Mayer, Elisabeth 131, 394 Nadathur, Prerna 700, 701
McCawley, James D. 140, 204, Nebel, Bernhard 209
274 Needham, Stephanie 14, 36, 359, 360, 362
McCloskey, James 683, 684, 686, 687, Neidle, Carol 6, 600
689–91, 716, 718 Nelken, Rani 722
McConnell-Ginet, Sally 477 Nemati, Fatemeh 102
McGlashan, Scott 287 Nespor, Marina 6, 398, 400
McGregor, R. S. 148 Newman, Paula S. 106, 115, 198, 225, 237,
Mchombo, Sam A. 37, 104, 135, 136, 156, 249, 250, 619, 734
181, 183, 184, 355, 356, 374, 375, 500–2, Nichols, Johanna 177
571, 665, 730 Nikitina, Tatiana 69, 197, 287, 355,
McKoon, Gail 397 601, 729
Mehlhorn, Grit 95 Nikolaeva, Irina 28, 71, 313, 365, 368, 370,
Mel’čuk, Igor A. 9, 650 373, 379, 381–3, 385, 386, 393, 439
Mester, Armin 400 Niño, María-Eugenia 120, 436, 447–50
Meyer, Antje S. 410, 727 Nokes, Jacqui 6
Miliĉević, Maja 334 Nordlinger, Rachel 6, 33, 34, 62, 64, 65, 71,
Millies, Sebastian 287 78, 99, 107, 112, 115, 128, 136, 177, 184,
Milward, David 615 190, 193–5, 197, 209, 210, 212, 222, 223,
Mistica, Meladel 352, 355 352, 436, 443, 466–8, 470, 651
Mittendorf, Ingo 78, 104, 247, 498 Noyer, Rolf 135, 445
Mofu, Suriel 352, 355 Nunberg, Geoffrey 681
Mohanan, K. P. 24, 25, 30–2, 39, 513, 588, Nutt, Werner 209
600, 728
Mohanan, Tara 21, 24, 30–2, 35, 136, 323, O’Connor, Mary Catherine 136
338, 511, 728 O’Connor, Robert 96, 366, 378, 394, 403,
Moltmann, Friederike 615 404, 435
Montague, Richard 280, 281, 284, 286, Odden, David 397
476, 477, 479, 645 Oehrle, Dick 273
Moore, John 324, 338, 339, 352, Oflazer, Kemal 352
355, 364 Ohala, John J. 730
Moortgat, Michael 2, 301 Ostendorf, Mari 397
 author index 813

Ostler, Nicholas D. M. 330 Quine, Willard Van Orman 494

Otoguro, Ryo 73, 74, 79, 116, 167, 221, Quirk, Randolph 29, 60, 158
222, 436
Owens, Jonathan 369 Radford, Andrew 91, 93, 612
Rákosi, György 88, 324, 334, 356–60, 364,
Pak, Marjorie 397 503, 543
ParGram Consortium 68, 69 Rambow, Owen 352
Park, Karen 543 Ramchand, Gillian 348, 350–2
Parsons, Terence 314 Rappaport Hovav, Malka 331,
Partee, Barbara H. 5, 47, 199, 256, 271, 339, 364
281, 282, 284, 302, 476, 477, 479, 630, Ratcliff, Roger 397
633, 643, 645, 707, 731 Raza, Ghulam 35, 102, 247,
Patejuk, Agnieszka 67–9, 80, 88, 229, 230, 348, 498
544, 601, 618–20, 650, 651 Reape, Mike 205
Payne, John 36, 37, 59, 63, 93, 95, 99, 364 Reichenbach, Hans 315
Pereira, Fernando C. N. 246, 280, 284, Reinhart, Tanya 238, 334, 358, 363,
287, 302, 305, 319, 517, 539, 570, 364, 370
716, 717 Retore, Christian 301
Perlmutter, David M. 1, 3, 15, 18, 21, Reyle, Uwe 280, 281, 287, 315, 518, 519,
253, 334 522, 569
Perry, John 281, 286 Rhodes, Richard 16
Peters, Stanley 284, 287, 571, 724 Riezler, Stefan 734
Peterson, John L. 350 Rizzi, Luigi 15, 91
Peterson, Peter G. 49, 80, 218, 602, 605, Robins, R. H. 9
606, 618, 620, 650 Robinson, Jane J. 701
Pica, Pierre 543 Roelofs, Ardi 410, 727
Pickering, Martin 700, 702 Rögnvaldsson, Eiríkur 588
Pienemann, Manfred 727, 728 Rohrer, Christian 96, 436, 447–50,
Pinkal, Manfred 287 613, 735
Pinker, Steven 323, 727 Rooth, Mats 633, 643, 645
Plank, Frans 95, 398, 404, 407, 423 Rosen, Carol 16
Polinsky, Maria 553, 555 Rosén, Victoria 189, 192, 194, 297
Pollard, Carl 2, 11, 15, 41, 44, 47, 69, 233, Ross, John Robert 14, 656, 662, 669
247, 267, 274, 477, 478, 499, 571, 585, Roth, Sebastian 671, 725
587, 618, 681, 732 Rounds, William C. 149, 216
Pollock, Jean-Yves 99, 109 Rubio Vallejo, David 345
Poser, William J. 103 Rudin, Catherine 124, 184, 185
Postal, Paul M. 1, 3, 15, 18, 21, 37, 334,
545, 547, 561, 674, 703 Sadler, Louisa 6, 33, 37, 55, 65, 66, 71, 73,
Potsdam, Eric 553, 555 76–8, 81–3, 89, 101, 103–5, 113, 120,
Potts, Christopher 320, 714, 715 127, 136, 190, 193–5, 197, 221, 247, 287,
Price, Patti 397 320, 352, 353, 436, 440, 443, 453, 466–8,
Prince, Alan S. 365, 398 470, 474, 498, 636, 640–2, 651, 670, 684,
Prince, Ellen 370, 653 713–16
Przepiórkowski, Adam 67–9, 88, 229, 230, Sadock, Jerrold M. 263, 264, 272
320, 360, 544, 601, 651 Sag, Ivan A. 2, 11, 15, 17, 41, 44, 47, 69,
Pulleyblank, Douglas 511 233, 247, 248, 267, 274, 477, 478, 499,
Pullum, Geoffrey K. 41, 56, 80, 144, 445, 571, 585, 587, 617, 618, 620, 621, 681,
609, 613, 623, 703 695, 696, 698, 722–4
Pustet, Regina 190 Saiki, Mariko 364, 663, 669
814 author index

Saraswat, Vijay A. 164, 278, 280, 287, Sulger, Sebastian 33, 35–7, 103, 195, 348,
301–3, 305, 319, 517, 539, 624, 630, 380, 394
716, 717 Sung, Won-Kyung 503
Scedrov, Andre 319 Svartvik, Jan 29, 60, 158
Schachter, Paul 21 Szűcs, Peter 665
Scheer, Tobias 397, 399
Schein, Barry 314 Tang, Chih-Chen Jane 507
Schwarze, Christoph 6, 96, 729 Tantos, Alexandros 350
Schwarzschild, Roger 644, 645 ter Meulen, Alice 5, 47, 199, 256, 271, 281,
Segond, Frédérique 120, 436, 282, 284, 302, 731
447–50 Tesnière, Lucien 9
Seiss, Melanie 116, 352, 452 Thomas, Victoria 88
Seki, Hiroyuki 732 Þráinsson, Höskuldur 21, 25, 94, 331
Selkirk, Elisabeth O. 6, 398–400, 412 Togo, Patrice 544
Sells, Peter 6, 89, 94, 96, 100, 104, 116, 138, Toivonen, Ida 2, 6, 11, 14, 15, 30, 35–7, 43,
515, 544, 553–5, 728 44, 62, 69, 88, 89, 96, 97, 99, 104, 105,
Sgall, Petr 370 108, 112, 116, 119–21, 126, 127, 130,
Sharma, Devyani 730 131, 134–8, 168, 172, 176, 177, 181, 184,
Shattuck-Hufnagel, Stefanie 397, 403 194, 197, 225, 231, 233, 237, 247, 259,
Shemtov, Hadar 733 320, 327, 329, 331, 335–7, 353, 356,
Shen, Tong 398 359–62, 364, 386, 470, 471, 474, 502,
Shieber, Stuart M. 47, 149, 284, 570 503, 505, 510, 512–14, 543, 544, 550,
Siegel, Muffy E. A. 477, 478 552, 559, 560, 565, 576, 577, 593, 621,
Siewierska, Anna 543 659, 696–8, 700, 703, 725
Sigurjónsdóttir, Sigríður 345 Turpin, Myfany 352
Siloni, Tal 334
Simpson, Jane 29, 31, 34, 41, 93, 106, 107, Ulinski, Morgan 352
112, 114, 128, 134, 136, 151, 250–2, 336, Uszkoreit, Hans 189
352, 355, 356, 363, 365, 470, 544, 587,
594, 595, 600 Vallduví, Enric 369–72, 378
Sims, Andrea D. 6 van Genabith, Josef 102, 287, 293, 305,
Smolensky, Paul 365 319, 321, 517, 727
Smolka, Gert 209 Van Herk, Gerard 355
Snijders, Liselotte 40, 108, 121, 138, van Reimsdijk, Henk 80
368, 503 van Urk, Coppe 585
Spector, Ilona 102 Van Valin, Robert D. 273
Spencer, Andrew 6, 66, 102, 105, 135, Vigo, Eugenio M. 197
329, 352, 364, 436, 439, 440, 443, 451–3, Vilkuna, Maria 369
470, 601 Villavicencio, Aline 642
Srivastav, Veneeta 671 Vincent, Nigel 26, 28, 71, 73, 75, 76, 221,
Steedman, Mark J. 2, 301, 370, 634 222, 247, 354, 364, 453, 498, 636, 639,
Steele, Joshua 398 729, 730
Stewart, Thomas W. 6, 135, 442 Visser, Fredericus T. 585
Stirling, Lesley 544 Vogel, Irene 6, 398, 400
Stokhof, Martin 518, 522, 538, 722 von Stechow, Arnim 381, 382
Stowell, Timothy 103
Strahan, Tania E. 503, 544 Wada, Hajime 287
Strand, Kjetil 503, 505 Wall, Robert E. 5, 47, 199, 256, 271, 281,
Stump, Gregory T. 6, 135, 436, 437, 439, 282, 284, 302, 731
442, 443, 455, 456, 460, 464, 467 Wang, Haifeng 102
 author index 815

Ward, Gregory L. 108 Wisdom, John 454

Warren, David H. D. 246 Woods, William A. 2, 731
Wasow, Thomas 2, 11, 41, 47, 248, 267, Wunderlich, Dieter 607
526, 620, 621, 681, 696
Way, Andy 727 Yadroff, Mikhail 110
Webelhuth, Gert 351, 668–70, 675 Yamamoto, Kazuyuki 725
Wechsler, Stephen 2, 6, 11, 15, 30, 35–7,
43, 44, 69–71, 78, 79, 88, 89, 97, 99, 108, Zaenen, Annie 2, 6, 17, 19, 21, 25, 30, 56,
112, 116, 119–21, 130, 131, 134–8, 168, 80, 93, 94, 106–8, 118, 128, 131, 132,
172, 176, 177, 181, 194, 197, 225, 231, 138, 143, 147, 171, 205, 208, 209, 257,
247, 259, 329, 331, 335–7, 353, 356, 364, 259, 264, 324, 325, 330, 331, 334, 335,
386, 470, 471, 502, 503, 505, 510, 337–9, 356, 359, 366, 375, 394, 436, 447,
512–14, 543, 544, 550, 552, 565, 576, 449–51, 514, 515, 544, 549, 553, 554,
577, 593, 621, 636, 659, 696–8, 700, 703, 564, 652, 653, 663, 687, 688, 694, 696,
725, 730 703, 730
Wedekind, Jürgen 242, 243, 280, 286, 351, Zamparelli, Roberto 643
732, 733 Zarrieß, Sina 613
Weil, Henri 369 Zec, Draga 544, 561
Weinberg, Amy 656 Zhang, Yanyin 727
Weisler, Steven 248, 620, 621 Zlatić, Larisa 69, 70
Wescoat, Michael T. 6, 102, 114, 115 Zoller, Claus Peter 55
Westerståhl, Dag 284 Zribi-Hertz, Anne 393
Wheeldon, Linda 400 Zwicky, Arnold M. 56, 80, 93, 95, 445, 613
Williams, Edwin 556, 656, 662 Zybatow, Gerhild 95
Winter, Yoad 626, 643, 645 Zymla, Mark-Matthias 394
 Language Index
In most cases, language families are listed according to the classifications given in the
World Atlas of Linguistic Structures (WALS: Haspelmath et al. 2008). Information on where
languages are spoken is often only approximate; see the WALS database for full details.

A Chickasaw (Muskogean; United States)

24, 38
Acehnese (Malayo-Sumbawan Chinese (Sino-Tibetan; China) 102, 354,
(Austronesian); Indonesia) 24 385, 507, 678, 725, 727, 735
Albanian (Albanian (Indo-European); Choctaw (Muskogean; United States) 24,
Albania) 515 38, 352
Arabic (Semitic (Afro-Asiatic); Middle Cree, Plains (Algonquian (Algic);
East) 192, 498, 727, 735 Canada) 25
Archi (Lezgik (Nakh-Dagestanian); Croatian, see Bosnian/Croatian/Serbian
Russia) 197
Arrernte (Pama-Nyungan; Australia) D
352
Avestan (Iranian (Indo-European); Dargwa (Lak-Dargwa
extinct) 247 (Nakh-Dagestanian); Russia) 198
Dogon (Dogon; West Africa) 544
B Dutch (Germanic; Netherlands) 171, 259,
335, 670
Balinese (Malayo-Sumbawan Dyirbal (Pama-Nyungan; Australia)
(Austronesian); Indonesia) 29, 31, 66, 38
177–9, 181, 189, 336, 356, 357, 365,
512, 587, 600 E
Bengali (Indic (Indo-European); South
Asia) 352, 401, 402, 407, 417–19 English (Germanic; Europe, North
Bosnian/Croatian/Serbian (Slavic; America, Australia) 10, 11, 13, 17–21,
Europe) 70, 78, 79, 86, 96, 116, 152, 23, 25–7, 30, 32, 33, 35–8, 40, 41, 43,
153, 403, 561 45, 46, 58–60, 62–5, 72–4, 78, 80, 83,
Bulgarian (Slavic; Bulgaria) 114, 124, 125, 84, 86, 88, 92, 93, 95–7, 99–101, 103,
177, 184–9 105, 106, 108–12, 116, 119, 120, 122,
123, 127, 128, 132, 133, 136, 138, 155,
C 159, 161, 164, 172, 173, 176, 177, 189,
190, 194, 206–8, 211, 214, 226, 227,
Catalan (Romance; Spain) 31, 32, 115, 247, 253, 277, 287, 297, 302, 310, 316,
197, 369, 371 328, 334, 336, 338, 340, 342–4, 353–5,
Cebuano (Greater Central Philippine 357, 358, 360–2, 367, 368, 374, 386,
(Austronesian); Philippines) 21 387, 389, 399, 403, 404, 420, 422,
Chamorro (Greater Central Philippine 426, 428, 429, 432, 434, 442, 446, 447,
(Austronesian); Guam) 687 449, 450, 452, 455, 456, 459, 474,
Chechen (Nakh (Nakh-Dagestanian); 475, 485, 493, 495, 498, 500, 503–5,
Russia) 622 511, 512, 524, 545, 546, 548–50,
Chicheŵa (Bantoid (Niger-Congo); 552, 560, 561, 563–6, 568, 569,
Malawi) 27, 28, 136, 156, 172, 177, 576, 583, 588–90, 593,
181–5, 189, 345, 346, 355, 356, 500–2 600, 601,
 language index 817

English (Germanic; Europe, North I

America, Australia) (cont.) 613, 622,
623, 627, 631, 632, 635, 639, 646, Icelandic (Germanic; Iceland) 25, 39, 40,
652–4, 658–61, 663–5, 668–70, 672, 74–6, 79, 94, 100, 116, 236, 330, 331,
674–8, 682, 684, 689, 690, 694, 697, 345, 503, 544, 548, 549, 563, 588, 589,
700, 703, 708, 709, 716, 719, 725, 730, 639, 640
733, 735 Indonesian (Malayo-Sumbawan
(Austronesian); Indonesia) 88, 352,
F 357, 556, 735
Irish (Celtic (Indo-European);
Finnish (Finnic (Uralic); Finland) 502 Ireland) 113, 394, 652, 683, 684, 687,
Fox, see Meskwaki 689–92, 709, 718, 719
French (Romance; France) 6, 33, 60, 64, Italian (Romance; Italy) 25, 352, 727
67, 71, 78, 79, 88, 102, 106, 109, 114,
115, 138, 143, 144, 352, 449, 450, 498, J
544, 613, 636, 725, 727, 735
Fula (Northern Atlantic (Niger-Congo); Japanese (Japanese; Japan) 20, 21, 79, 118,
West Africa) 72, 73, 509, 510 189, 193, 257–9, 351, 364, 396, 420,
514, 663, 669, 677, 678, 725, 727, 735
G
K
Georgian (Kartvelian; Georgia) 735
Kichaga (Bantoid (Niger-Congo);
German (Germanic; Central Europe) 19,
Tanzania) 26
21, 22, 32, 56, 60, 64, 80–2, 88, 93, 99,
Kikuyu (Bantoid (Niger-Congo);
116, 138, 205, 259, 319, 331, 344, 345,
Kenya) 652, 687–90
371, 372, 394, 457, 498, 544, 553–5,
Kinyarwanda (Bantoid (Niger-Congo);
585, 607, 608, 613, 670, 697–701, 725,
Rwanda) 25
727, 733, 735
Korean (Korean; Korea) 104, 371, 372,
Greek (Greek (Indo-European); Greece)
393, 503
Ancient, 9, 64, 177, 179–81, 185,
189, 556
L
Modern, 287, 364, 503
Greenlandic (Eskimo-Aleut; Greenland, Latin (Italic (Indo-European);
North America) 23, 335 extinct) 108, 138, 453, 462
Lummi (Salishan; United States) 66, 166,
H 167

Hebrew (Semitic (Afro-Asiatic); Israel) 37, M

102, 113, 189, 190, 194, 684, 686, 687
Hindi-Urdu (Indic (Indo-European); South Malagasy (Barito (Austronesian);
Asia) 21, 35–7, 79, 102, 103, 105, 148, Madagascar) 544, 555, 735
326, 347–52, 372, 373, 376, 378–80, Malayalam (Dravidian; India) 24, 25, 31,
511, 671, 680, 698, 701, 725, 735 39, 112, 128, 374, 513, 514, 588, 698
Hopi (Uto-Aztecan; United States) 55, 56, Maltese (Semitic (Afro-Asiatic);
76, 77 Malta) 684
Hungarian (Ugric (Uralic); Hungary) Maori (Oceanic (Austronesian); New
36, 37, 68, 69, 101, 103, 107, 112, Zealand) 189, 192, 297
121, 125, 126, 194, 338, 356, 364, Mapudungan (Araucanian; Chile) 66, 365
369–71, 435, 502, 503, 650, 679, 680, Marathi (Indic (Indo-European);
725, 735 India) 365
818 language index

Marori (Moraori; Southern New S

Guinea) 65, 78
Martuthunira (Pama-Nyungan; Sanskrit (Indic (Indo-European);
Australia) 466, 468, 470 India) 96, 104, 114, 115, 247
Meryam Mir (Western Fly; Australia) 78 Serbian, see Bosnian/Croatian/Serbian
Meskwaki [Fox] (Algonquian (Algic); Sinhala (Indic (Indo-European); Sri
United States) 28, 184, 394 Lanka) 503
Murrinh-Patha (Southern Daly; Somali (Cushitic (Afro-Asiatic);
Australia) 78, 352 Somalia) 65
Sotho, Northern (Bantoid (Niger-Congo);
N South Africa) 735
Spanish (Romance; Spain, Americas) 31,
Norwegian (Germanic; Norway) 14, 32, 130, 131, 153, 157, 189,
116, 336, 345, 353, 365, 503, 505–7, 345, 394
509–12, 515, 600, 703, 735 Swahili (Bantoid (Niger-Congo); East
Africa) 544
O Swedish (Germanic; Sweden) 30, 31, 96,
100, 104, 116, 394, 670, 683–5,
Ojibwa (Algonquian (Algic); North 703, 728
America) 16
Ossetic (Iranian (Indo-European); T
Georgia and Russia) 115, 650, 671,
721, 725 Tagalog (Greater Central Philippine
(Austronesian); Philippines) 16, 21,
P 23, 66, 93, 99, 112, 128, 335, 565, 566,
574, 575, 658, 659, 661, 669, 725
Palauan (Palauan (Austronesian); Tariana (Arawakan; Brazil) 194
Palau) 24, 687, 698 Tigrinya (Semitic (Afro-Asiatic); Eritrea
Pashto (Iranian (Indo-European); Pakistan and Ethiopia) 735
and Afghanistan) 96 Tiwa, Southern (Kiowa-Tanoan; United
Persian (Iranian (Indo-European); Iran) States) 16
102 Turkish (Altaic; Turkey) 28, 115, 352,
Polish (Slavic; Poland) 22, 67, 80, 229, 230, 376, 735
342–4, 544, 601, 618, 650,
651, 735 V
Portuguese, Brazilian (Romance;
Brazil) 642 Vata (Kru (Niger-Congo); Côte
Punjabi (Indic (Indo-European); South d’Ivoire) 684
Asia) 502 Vietnamese (Viet-Muong (Austro-Asiatic);
Vietnam) 297, 735
R
W
Romanian (Romance; Romania)
83, 636 Walman (Torricelli; Papua New Guinea) 650
Rongpo (Tibeto-Burman; Nepal) 55, 74 Wambaya (Mirndi; Australia) 64, 209, 210
Russian (Slavic; Russia) 94, 95, 98, 100, Warlpiri (Pama-Nyungan; Australia) 24,
107, 108, 110, 114, 119, 122–4, 126, 34, 38, 40, 41, 45, 93, 99, 106, 107,
129, 141, 145, 189–93, 205, 206, 223, 112–14, 128, 129, 134, 151, 250, 251,
224, 366, 367, 375, 376, 451, 452, 650, 277, 368, 369, 502, 503, 544, 593,
659, 677, 678, 695, 728 595, 596
 language index 819

Welsh (Celtic (Indo-European); Wales) 37, Yami (Batanic (Austronesian);

78, 101, 104, 113, 114, 498, 641, Taiwan) 335
642, 735 Yoruba (Defoid (Niger-Congo);
Wolof (Northern Atlantic (Niger-Congo); Nigeria) 511
Senegal) 394, 735
Z
Y
Zapotec (Zapotecan (Oto-Manguean);
Ya̧g Dii (Adamawa-Ubangi (Niger-Congo); Mexico) 104, 115, 495, 725
Cameroon) 503
 Subject Index

Symbols ∈ see Set, membership

0∈ see Set, membership, negated
↑ see under Functional description ⊆ see Subset; see under Temporal
↓ see under Functional description relations
≺ see under Functional description; ⊂ see Subset, proper; see under
Temporal relations Temporal relations
$ see under Functional description; ∩ see Intersection
Temporal relations ∪ see Union
5 see Set, complement
see Constituent structure, current
⃰ node; Kleene star/Kleene plus 6 see Subsumption
operator 7 see Generalization
ˆ⃰ see Constituent structure, mother 8 see Unification
node / see Phrase structure rule, “Ignore”
< see Constituent structure, left sister operator; Priority union
⃰ node \ see Regular language, term
> see Constituent structure, right sister complement
⃰ node | see Functional description,
⇙ (leftmost s-string element) disjunction; Regular language,
see under Functional description union; Restriction
⇘ (rightmost s-string element) ⇒ see Functional description,
see under Functional description implication
# (leftmost p-string element) ¬ see Functional description,
see under Functional description negation; Regular language,
$ (rightmost p-string element) complement
see under Functional description & see Regular language, intersection
#s (leftmost stressed p-string element) ∂ see Presupposition operator
see under Functional description α see Correspondence function, α
$s (rightmost stressed p-string element) β see Correspondence function, β
see under Functional description ι see Correspondence function, ι
𝜖 see Functional structure, path λ see Correspondence function, λ;
through, empty path (𝜖); Phrase Lambda operator
structure rule, empty node μ see Correspondence function, μ
← see under Off-path constraint π see Correspondence function, π
→ see under Off-path constraint 𝜎 see Correspondence function, 𝜎
=c see Functional description, τ see under Temporal relations
constraining equation 𝜙 see Correspondence function, 𝜙
≡ see Equivalency
A
∀ see Universal quantifier
∃ see Existential quantifier A see Binding, semantics, antecedency
* see Logical derivability/ relation between discourse
“Turnstile” referents
⊸ see Linear implication A-structure, see Argument structure
⊗ see Multiplicative conjunction Across-the-board constraint, see
⊕ see Discourse Representation Coordination, long-distance
Theory, merge, operator dependency and
◦ see Function, mathematical, actv (feature) 385
composition adj (grammatical function), see
∅ see Set, empty Modification
 subject index 821

Adjective 473–92, see also Copular Marking principle 78

construction; Modification mismatches 78–9
agreement 70 pronominal 510
at c-structure 103, 127, 474–5 relative pronoun 223–5
gradable, see Adjective, subsective verb 22, 149, 186, 345, 347,
intersective 476, 480–2, 486–9 500–2
lexical entry 480, 486 vs. incorporated pronoun, see Pronoun,
meaning constructor 481, 486 incorporated
nonprojecting 103–5, 127 Anaphor, see Binding; Pronoun
nonsubsective 478–9, 483–4, 491–2 Anaphoric control, see under Control
lexical entry 483, 491 anchored (feature) 385
meaning constructor 483, 491 Animacy/saliency hierarchy 66
predicative 485–6 animate (feature) 313
lexical entry 486 Annotated phrase structure rule, see under
meaning constructor 486 Phrase structure rule
raising 548 Applicative construction 27–8, 352, 354–6
subcategorization 247, 364–5 Arc Pair Grammar 3
subsective 477–8, 482–3, 489–90 Argument classification, see Mapping
lexical entry 482, 489 theory
meaning constructor 482, 489 Argument structure 314–19, 322–65,
Adjunction (phrase structure) 98–9, 114, see also Binding; Control; Mapping
474–5, 493 theory; Thematic roles
nonprojecting words 105, 109, 127 Conceptual Semantics 324, 326
Adverb 492–8, see also Modification Proto-Role argument
adjectival 487–489 classification 324–6, 335, 365
lexical entry 488 valency template 339
meaning constructor 488 Argument, syntactic, see Grammatical
at c-structure 493, 495 functions, governable; Semantic
manner 494–5, 497–8 form, argument list
lexical entry 497 Aspect 313–19, see also Temporal relations
meaning constructor 498 aspect (feature) 63–5, 85
placement, as test for constituency 94, aspectual light verb 350–2
549, 566 at argument structure 326
sentential 494–7 at f-structure 54, 64–5
lexical entry 496 Association principles, see Mapping theory
meaning constructor 496 Asymmetrical Object Parameter, see under
Adverb Effect, see Object grammatical functions
Complementizer-Adjacent Axiom of Extension, see under Set
Extraction Constraint
Agreement, see also Functional structure, B
features
adjective 70 Bach’s Generalization 586
and information structure role 386 Background information (information
as obj test 24, 28 structure), 370–3
as subj test 20 background (i-structure role) 373, 378
as test for termhood 16 background (value of df
(a)symmetric 197–8 feature) 386–93
determiner 70 Binding, 324n, 502–43, see also Control,
distinguished conjunct 640–3 anaphoric, obligatory; Pronoun
feature sharing 197–8 antecedent constraints 509–10
grammatical function and 54 as test for argumenthood 14–15
822 subject index

domain constraints morphological 446–7, 462, 466–70

Coargument Domain 507, 527–8 quirky 34, 79, 548–9, 563–4
Minimal Complete Nucleus 505, 507, semantic 34, 79
525–6 stacking, see Case, constructive
Minimal Finite Domain 507 structural 79, 229–30
Root Domain 507 CAT predicate, see under Functional
f-command 239–40, 509 description
f-precedence 257–9, 513–14 Categorial Grammar 2, 273, 301, see also
grammatical function hierarchy 512–13 Glue, and Categorial Grammar
intersentential 528–43 Category mismatch, see under
negative constraints 510–12, Long-distance dependency
526–8 Clausal complement 16–19, 29–32,
positive constraints 503–12, 523–6 see also comp; Object grammatical
prominence constraints 512–17 functions, clausal objects
definition 516–17 extraction from, see Long-distance
semantics 517–43, see also Discourse dependency, extraction from
Representation Theory, PCDRT clausal complement
anaphoric discourse referent (x) 519 extraction of, see comp (grammatical
antecedency relation between indices function), in long-distance
(R) 520–1 dependency
antecedency relation between indices Cleft construction 367, 374, 394
(R⃰) 527–8 Clitic 116, 415, 730, see also under Minor
antecedency relation between phrase structure category
discourse referents (A) 519–21 auxiliary 112
index 520–1, 523 clitic (feature) 415
index-discourse referent relation cluster 96
(I ) 520–1 doubling 130–1, 153n, 157, 184–9
thematic hierarchy 514–15 placement 435
Bridge verb, see Long-distance dependency, as test for constituency 95
extraction from clausal complement pronominal 101, 152–3, 184–9
reflexive 364, 544
C Closed grammatical functions 12, 29, 194,
545, see also comp, predlink
c(ause) (mapping feature) 364 Closed set description, see under
C-precedence, see under Constituent Functional description
structure Coargument Domain, see under Binding,
C-structure, see Constituent structure domain constraints
CAE Constraint, see Codescription 267, 269–71, 274–5
Complementizer-Adjacent Coherence 52–3, 154, 160, 328
Extraction Constraint definition 53
Case 179–80 in coordination 604–6
and information structure role 386 semantic 299–301, 723
case (feature) 56, 79–83 Comment (information structure) 369–70
constructive, 62n, 209–10, 212n, 466–70 comp (grammatical function) 29–32, 545,
default 79, 164–5 561–8
grammatical function and 24–5, 39–40, as oblique grammatical function 16–17
54 as semantically restricted/unrestricted
indeterminacy 48, 80–3 function, 18–19
loss 729 in long-distance dependency 131–3, 654
marking oblique argument 34 in mapping theory 356
 subject index 823

compform (feature) 60, 85 Constituent structure 89–116, see also

Complement (phrase structure Endocentricity; Exocentricity;
position) 97–8, 113 Phrase structure rule; X-bar theory
mapping to f-structure 119–21, 177–8 c-precedence 256
Complementizer 100, 108, 158, 593, category inventory 106–7
see also compform (feature); CP, category labels, see Correspondence
head function, λ
Complementizer-Adjacent Extraction complex category 97, 105–6, 143–4,
Constraint, 253–6, 675–7 450–1
Completeness 50–2, 130, 154, 157, 159, parameter 106, 143, 450
328 current node ( ) 168, 266
definition 52 ⃰
discontinuity 134–5, 138
in coordination 604–6 left sister node (< ) 223
semantic 299–301, 723 “mixed” category
⃰
Completive information (information morphologically mixed 451–3
structure) 372–3 syntactically mixed 116, 729
completive (i-structure role) 373 mother function (M) 169
completive (value of df mother node ( ˆ ) 168, 266
feature) 386–93 ⃰
optionality 100n, 107–8, 122n, 129–30,
Complex category, see under Constituent 137–8, see also Phrase structure
structure rule, optionality
Complex predicate 346–52 parsing 731
Composition, function, see Function,
right sister node ( >) 223
mathematical, composition ⃰
rule, see Phrase structure rule
Compositionality, Principle of
tree, definition 273
276–8
wellformedness condition, see
Computational complexity 730–2,
Nonbranching Dominance
see also Constituent structure,
Constraint
parsing; Functional description,
Constraining equation, see under
solution algorithm
Functional description
exponentially difficult problem 731n
Construction Grammar 2, 3, 272
parsing in polynomial time 731n
context (feature) 532–3
Conceptual Semantics, see under Argument
Control 324n, 347, 545–601, see also Open
structure
grammatical functions; Raising
concord (feature) 69–71, 85
anaphoric 561–8, 593–600
as distributive feature 70–1, 79–80,
arbitrary 576–83
218–19
meaning constructor 572, 581
Configurationality 40, 176, 368–9,
obligatory 561–74
see also Discourse
backward 553–6
configurationality;
controller choice 585–7
Nonconfigurationality
equation 551
configurational encoding 176
equi verb 553–6
conj (feature) 609–12
anaphoric vs. functional
as nondistributive feature 86, 217–18,
control 563–6, 574–6
610–12
semantic type of complement
Conjoinable type 633
569–71
Conjunction, see under Coordination
functional 545–53, 574–6, 587–93, 630
Consistency 53–4, 129, 130, 151,
meaning constructor 557, 576, 591
157, 196
partial 583–5
definition 53
quasi-obligatory, see Control, partial
824 subject index

Coordination 97n, 602–51, see also notation 266, 289

Agreement, distinguished conjunct; 𝜙 (c-structure to f-structure) 117–18,
Distributive/ nondistributive 168, 248–50, 266, 269–70
feature; Functional structure, set of; inverse (𝜙−1 ) 248–50, 272
Gender, resolution; Person, π (s-string to c-structure) 115, 252,
resolution 254–6, 406–7, 414
as test for constituency 91–2 𝜎 (f-structure to s-structure) 270–1,
asymmetric 607–9 288–90
c-structure 602–24 CP (phrase structure category) 100,
clausal 602–4, 624–8 111
conjunction 603, see also conj, as subj 161–3, 247
preconj complement 119
meaning constructor 626 head 100
f-structure 213, 241, 604–24 specifier 123–5
long-distance dependency and 662–4, Curry-Howard isomorphism
684, 704, 724 293n
nominal 635–50
non-seed conjunct 626–8 D
meaning constructor 626, 631, 634,
646 +d (mapping feature) 364
nonconstituent 612–18 D (description function), see under
preconjunction 609–12 Morphology
predicate 604–9 Dcat (description function), see under
“seed” conjunct 624–5 Morphology
meaning constructor 625, 627, 631, Ddefault (description function), see under
634, 646 Morphology
unlike category 617–21 Dfeat (description function), see under
verb, see Coordination, predicate Morphology
Copular construction 32–3, 189–97, Dfeats (description function), see under
485–6, see also Adjective, Morphology
predicative; predlink Dl (leftmost daughter), see under
single or double tier 193–7 Functional description
without verb 189–97 Dr (rightmost daughter), see under
Copy raising, see under Raising Functional description
Core arguments, see Terms Data-Oriented Parsing 727
Correspondence function 266–7 Dative alternation 231, 329, 354–5
α (c-structure nodes to argument Decidability theorem for LFG 247
structure) 326–7 def (feature) 83–8
β (p-string to p-structure) 407–9 deixis (feature) 83–8
inverse 247–50, 271–2 dep (feature) 448
ι (c-structure to i-structure) 378–80 Dependent marking language 177
ι (s-structure to i-structure) 383–5 Deponency, see under Morphology
λ (argument structure to Derived arguments, see under Grammatical
f-structure) 326–7 functions
λ (c-structure nodes to node labels) 141, Description by analysis 267–9, 275
249–50, 269–70 df (feature) 386–93
μ (c-structure to morphosyntactic Diachronic change 728–30
structure) 447–9 Direct grammatical functions, see Terms
μ (f-structure to morphosyntactic Direct object, see under Object
structure) 449–51 grammatical functions
 subject index 825

Direct Syntactic Encoding, Principle Equivalency (≡) 142n

of 329, see also Lexical Ergativity 23–24, 335
(redundancy) rules Existential constraint, see under Functional
dis, see under Overlay function description
Discontinuity, see under Constituent Existential quantifier (∃) 250n
structure Exo-Skeletal Model 135, 445
Discourse configurationality 368–9 Exocentricity 111–14, see also S
Discourse Representation Theory 281, (phrase structure category);
287, 319, 518–43 X-bar theory
context update 521–3, 531–43 Extended Coherence Condition 375, 379,
meaning constructor 532 653
indefinite noun phrase 535–8 definition 375
meaning constructor 536 Extended Projection Principle 21n
merge 530–43 EzP (phrase structure category) 103
operator (⊕) 530
PCDRT 518–43, see also Binding, F
semantics
quantifier 521–3, 538–43 F-command 238–40, see also under
meaning constructor 539, 541 Binding
Discourse structure 366n, 394 F-description, see Functional description
Disjunction F-precedence 256–9, 265, 403, 697,
in functional description, see under see also under Binding
Functional description definition 257, 259, 697
in phrase structure rule, see under F-structure, see Functional structure
Phrase structure rule Feature, see Functional structure, features;
Dislocation 664–5, see also Long-distance Morphology, m-features; Prosodic
dependency structure, features; Semantic
Distributed Morphology 135, 445 structure, features
Distributive/nondistributive field (feature) 415
feature 49–50, 69, 216–19, 606–13 Finite-state automaton, see under Regular
definition 219 language
inventory 86, 606 fm (feature) 409–11, see also form;
dom (feature) 401 P-form; S-form
DOP, see Data-Oriented Parsing fn, see Semantic form, reference to parts of
DP (phrase structure category) 100–1 Focus (information structure) 369–80
head 100–1 as grammaticized discourse function, see
Dynamic Predicate Logic 518, 538 Overlay function, status of focus
broad 426–34
E completive 371–2
contrastive 371–2
E (phrase structure category) exponent 427–34
113–14 extent 427–34
Economy of Expression 136–8 focus (i-structure role) 372, 377–93
Ellipsis focus (value of df feature) 386–93
as test for constituency 91–2 marking, prosodic 426–34
strict and sloppy readings 570n narrow 426–34
Endocentricity 111–14, see also X-bar presentational 336, 365
theory form (feature) 34, 59–61, 85, see also
Principle of 112, 176 compform; fm; P-form; pform;
Equi verb, see under Control S-form
826 subject index

Fragmentability of language 138 open set description 214–16, 220

Ft (prosodic structure category) 399–400 constraining 216
Function optionality 156–7
mathematical, see also Lambda operator solution algorithm 149, 730–2
abstraction 306 Tl (leftmost terminal) 413–14
application 283 Tr (rightmost terminal) 413–14
composition (◦) 270–1 Functional precedence, see F-precedence
definition 44n Functional structure 9–88, 264, see also
inverse 248, 271–2 Functional description;
syntactic, see Grammatical functions Grammatical functions
Function-Argument Biuniqueness 334, acyclic 732n
364 as set-theoretic object 44, 47n, 146
Functional category 89, 99–103, see also constraint, see Functional description
CP; DP; E; IP; KP; NumberP; QP features 54–85
complement 119 complex value 54–8
Functional control, see under Control indeterminacy 56
Functional description 146–9, see also markedness 57, 58, 75–6, 164
Phrase structure rule, macro; natural classes 55
Template privative 58–9, 164
↑ 168–72 relation to meaning 57, 73
in lexical entry 171–2 underspecification 58–9
↓ 168–72 universality 57, 73
in lexical entry 171–2 path through 44n, 147, see also
$ (f-structure of next word) 254 Functional uncertainty
≺ (f-structure of previous word) 255 empty path (𝜖) 147, 199
⇙ (leftmost s-string element) 414 set of 48–50, see also
# (leftmost p-string element) 415 Distributive/nondistributive
#s (leftmost stressed p-string feature; Functional description,
element) 421–2 closed/open set description; Set
⇘ (rightmost s-string element) 414 wellformedness condition, see
$ (rightmost p-string element) 415 Completeness; Coherence;
$s (rightmost stressed p-string Consistency
element) 421–2 Functional uncertainty 122, 205–11, 264,
CAT predicate 247–50 see also Off-path constraint
closed set description 75, 220–2 decidability 209
constraining equation (=c ) 161–4 definition 208
Dl (leftmost daughter) 413–14 inside-out 209–11, 215, 310
Dr (rightmost daughter) 413–14 definition 210, 211
default specification 164–5
disjunction (|) 155 G
processing 730–2
existential constraint 159–60, 164 Gapping 244
functional equation 147 Gender
implication (⇒) 166–8 gend (feature) 69–71, 74, 85
instantiated symbol 153–4 resolution 48, 74–6,
local name 223–5 636–40
minimal solution 146–8 Generalization (7) 240–2
relation to unification 149 Generalized Phrase Structure
negation (¬) 158–9 Grammar 41
negative existential constraint 160–1 Gesture, see Multimodal communication
 subject index 827

gf
! (grammatical function) 22–4, 38 Head marking language 177
gf (metavariable over grammatical Head-Driven Phrase Structure
functions) 206 Grammar 2, 11, 41, 44n, 47n, 205,
gf-outranks, see Grammatical functions, 266, 273, 275, 352n, 378n
hierarchy human (feature) 313
ggf (metavariable over governable
grammatical functions) 177, 205n I
Glue 287–320, see also Linear logic;
Syntax-semantics interface I see Binding, semantics, index-discourse
and Categorial Grammar 301 referent relation
computational implementation 735 I-structure, see Information structure
first-order 320 ID/LP rule, see under Phrase structure rule
propositional 320 “Ignore” operator (/), see under Phrase
Grammar Writer’s Workbench 734 structure rule
Grammatical functions 3, 11–38, see also Impersonal predicate 343–5, see also
Closed grammatical functions; Subject, condition
Modification; Oblique grammatical Implication, see under Functional
functions; Open grammatical description (⇒); see Linear
functions; Restricted grammatical implication (⊸)
functions; Terms; Unrestricted Indefinite noun phrase 285–6, 558–9, 644,
grammatical functions see also def; see also under
as theoretical primitive 3, 38–42 Discourse Representation Theory
derived arguments 357–63 meaning constructor 558
encoding 176–89, see also index (f-structure feature) 69–71, 85, 523n
Configurationality; as nondistributive feature 70–1, 86,
Dependent marking language; 217–18
Head marking index (semantic structure feature) 523,
language; Lexocentric see also Binding, semantics
organization Indirect object, see under Object
governable 12–38, 50 grammatical functions
and coherence 52–3 Infinitival complement, see comp; xcomp
status of poss 36 phrase structure category 550n
gradient distinctions 357 Information structure 5, 278, 366–94,
hierarchy 10–12, 20, 22, 24, 41, 166, see also Background information;
336, 339, 347, 697, see also Mapping Completive information;
theory, markedness hierarchy; Correspondence function, ι; Focus;
see also under Binding Tail; Topic
inventory 13 features 371–3, 377–93, see also df;
tests for 13–15, 322–3 new; prom
Grammatical relations, see Grammatical Inside-out functional uncertainty, see under
functions Functional uncertainty
Ground (information structure), see Instantiated symbol, see under Functional
Background information description
Intensional logic 280, 281, 286
H Interface Harmony, see under Prosodic
structure
Head (phrase structure position) 98 Interrogative pronoun, see under Pronoun
Convention 118, 388n Intersection (∩) 214, see also under
mapping to f-structure 118, 168–70, 177 Regular language
“movement” 129–130 in feature resolution 639–40
828 subject index

Intonation, see Prosody Lexical (redundancy) rules 2, 12, 231,

Intonational Phrase, see IntP 329–30, 358
IntP (prosodic structure Lexical Sharing 114–16
category) 399–400 Lexicalist Hypothesis, see Lexical Integrity
Inverse correspondence, see under Principle
Correspondence function Lexocentric organization 112, 176
IP (phrase structure category) 99–100, LFG-DOP, see Data-Oriented Parsing
108–10 LI, see Lexemic Index
complement 113, 119 Light verb, see Complex predicate; Aspect,
head 99–100 aspectual light verb
specifier 122 Linear implication (⊸) 291
Island constraints, see under Long-distance Linear logic 288, 290–9, 301, 319,
dependency see also Linear implication;
Monads; Multiplicative conjunction
K proof nets 320
proof rules 321
Keenan-Comrie hierarchy 10–11, 20, Linking rules, see Mapping theory
see also Grammatical functions, Local name, see under Functional
hierarchy description
Kleene star/Kleene plus operator 141, 200, Locative inversion 345–6, 365
207 Logical derivability/“Turnstile” (*)
definition 200 297
KP (phrase structure category) 102 Long-distance dependency 37–8, 652–725,
see also Dislocation; Extended
L Coherence Condition; Functional
uncertainty; Nested dependency;
l (feature) 412 Overlay function; Pronoun,
L see Lexical entry, relation L interrogative; Pronoun, resumptive;
Lambda operator (λ) 282–3, see also Question, constituent; Relative
Function, mathematical clause; Topicalization; “Tough”
LE see Lexemic entry movement; Weak crossover
Lexemic entry (morphology) 437, 439–42, across-the-board constraint, see
444–5, see also Lexemic Index (LI) Coordination, long-distance
definition 440 dependency and
Lexemic Index (LI) 439–40 as relation between c-structure
Lexical category 89, 96–7, 106, see also positions 264, 652
Minor phrase structure category as test for argumenthood 15
complement 119–21 category mismatch 131–3, 725
inventory 96 empty category in 675–7, 694–704,
Lexical entry 171–2, 437–9 see also “Wanna” Contraction
↑ see under Functional description extraction from clausal
↓ see under Functional description complement 226–8, 655–6, 658,
relation L 438–9, 444–5 660, 669
definition 444 island constraints 656–8, 660–1, 725
simplified 438 morphological marking 687–94
vs. lexemic entry 437 multiple gap construction 704–6,
Lexical Integrity Principle 95–6, 135–6, 725
445 “that-trace” effects, see
Lexical Mapping Theory, see Mapping Complementizer-Adjacent
theory Extraction Constraint
 subject index 829

M “Mixed” category, see under Constituent

structure
M see Constituent structure, mother Modification 12, 13, 473–99, 587–600,
function see also Adjective; Adjunction
m(entally involved) (mapping feature) 364 (phrase structure); Adverb; Relative
M-entry, see Morphology, realization clause; xadj
relation R adj, as nondistributive feature 86,
M-features, see under Morphology 217–18
Macro, see under Phrase structure rule anaphorically controlled 593–600
Manager resource 561, 716–21, see also meaning constructor 596
Pronoun, resumptive f-structure 214
Mapping theory 66, 278–9, 330–65, 441n, functionally controlled, see xadj
see also Argument structure; +o; +r recursive modification 484–92
argument demotion 341–2 scope 499
argument suppression 341 Modularity 192, 263–75, 278–9, 375–6,
default classification 333–4, 364 398, 404–7, 424, 427, 436
intrinsic classification 332–6, 364 Monads 320, 360n
Mapping Principle(s) 336–8, 340–1, mood (feature) 63–5, 85
346, 349 Morphological blocking 470, 728
markedness hierarchy 336–8, 340–1, Morphology 5–6, 436–70, see also Lexemic
see also Grammatical functions, entry; Lexemic Index (LI)
hierarchy D (description function) 443–70
wellformedness condition, see Subject, definition 458–9
Condition; Function-Argument Dcat 459–60
Biuniqueness Ddefault 460–1
Maximal projection, see under X-bar theory Dfeats 460
Meaning constructor 289–90, see also Dfeat 462–70
Linear logic; see also under deponency 453–5, 462–4
Adjective; Adverb; Control; finite-state 436
Coordination; Discourse m-features 440–2
Representation Theory; Indefinite morphomic features 440–1
noun phrase; Modification; Noun; morpheme-based 470
Oblique grammatical functions; realization relation R 442–5
Pronoun; Quantifier; Relative realizational 436, 442–3
clause; Verb; xadj Morphosyntactic structure 264, 446–51,
at information structure 381–93 see also Correspondence function, μ
Curry-Howard format 319 Movement paradox, see Long-distance
reference to, by label 289–90, 296–7 dependency, category mismatch
typed 289–90 Multimodal communication 275
Metacategory, see under Phrase structure Multiplicative conjunction (⊗) 321, 716
rule
Minimal Complete Nucleus, see under N
Binding, domain constraints
Minimal Finite Domain, see under Binding, N see String, N (“Next” function)
domain constraints n_tone (feature) 422
Minor phrase structure category 96–7, Negation
127–8 in functional description, see under
clitic 96–7, 127 Functional description
mapping to f-structure 127 neg (feature) 67–9, see also pol
Minor Category Convention 128 Negative existential constraint, see under
particle 96–7, 127 Functional description
830 subject index

Nested dependency 701–3 objloc 26

new (feature) 371–3 objpatient 26
Nonprojecting word objtheme 11, 18, 26, 662
mapping to f-structure 126–7 in mapping theory 354–6
Nonbranching Dominance objagent 565
Constraint 245–6 without obj 28–9, 365
Nonconfigurationality 40–1, 107, 112–13, primary object 25–9
128, 134–5, 176–7, 368–9, 502, secondary object 25–9
see also Discourse obj𝜃 , see under Object grammatical
configurationality functions
Nonconfigurational encoding 176 +obl (mapping feature) 364
Nondistributive feature, see Oblique grammatical functions (obl𝜃 ) 11,
Distributive/nondistributive feature 15–19, 33–4, 357, see also comp;
Nonmaximal projection, see under X-bar xcomp
theory as open grammatical function 29n, 552,
Nonprojecting word 103–5 600
Nonterms, see Oblique grammatical copular complement 194n
functions idiosyncratically marked 34
Noun in mapping theory
common 354–6
lexical entry 480 oblagent 332
meaning constructor 310, 480 oblgoal 11, 18, 34
proper name oblinstr 362–3
meaning constructor 289, 530 meaning constructor 363
subcategorization 247, 364–5 oblloc 34
Null pronoun, see Pronoun, incorporated oblsource 11
Number semantically marked 34
constructed 55–6, 76–8 Obliqueness Hierarchy (HPSG) 11
num (feature) 69–71, 85 Obviation 21, 544
NumberP (phrase structure category) 103 Off-line parsability, see Nonbranching
Dominance Constraint
O Off-path constraint 225–30
→ 226–30
+o (mapping feature) 332–9, 341, 364–5, definition 230
see also Object grammatical ← 228–30
functions definition 230
obj, see under Object grammatical Open grammatical functions 12, 29, 545,
functions see also xadj; xcomp
Object grammatical functions 24–9, obj as 29n, 552, 600
see also Dative alternation; +o obl𝜃 as 29n, 552, 600
Asymmetrical Object Parameter 356 subj as 29n, 600
clausal objects 30–2 Open set description, see under Functional
direct object 25–9 description
indirect object 10n, 25–9 Optimality Theory 336, 470, 728
obj 17–19 decidability of OT-LFG parsing 728
as open grammatical function 29n, Optionality
552, 600 in functional description, see under
in mapping theory 354 Functional description
obj2 25–6 in phrase structure rule, see under
obj𝜃 11, 17–19, 26 Phrase structure rule
objinst 26 OT-LFG, see Optimality Theory
 subject index 831

Overlay function 37–8 ↑ see under Functional description

dis 37–8, 121–2, 654–724 ↓ see under Functional description
in long-distance dependency 37–8, as node admissibility condition 139–40
654–724 disjunction 140–1, 199–202
phrase structure position 123–6, 659 empty node (𝜖) 250–2
pivot 38 ID/LP rule 144–5, 201
status of focus 37n, 122, 374–80 “Ignore” operator (/) 204
status of subj 38, 122 macro 237–8, see also Template
status of topic 37n, 122, 374–80 reference to 238
metacategory 141–3, 431, 659
P optionality 140–1, 173, 199–201
parametrized 143–4, 450, see also
P-diagram 395n, 412n, 435 Constituent structure, complex
P-form 409–10, 437–9, 442–3, category
see also fm (feature); Prosodic regular expression 140–1, 201–4, 615
structure; String, p-string “Shuffle” operator 204–5
p-form (feature) 401 Pied piping, see Question, consistuent, q
P-string, see under String attribute; Relative clause, relpro
P-structure, see Prosodic structure attribute
Paradigm Function Morphology 437n, Pipeline architecture, see under
442, 443, 455n, 460n, 467, see also Syntax-prosody interface
Morphology, realizational pivot (grammatical function) 22–4, 38
Parasitic gap, see Long-distance pol (feature) 68, see also Negation, feature
dependency, multiple gap (neg)
construction poss (grammatical function) 35–7,
Parenthetical 204 83, 364
Partial Compositional Discourse Precedence, see Constituent structure,
Representation Theory (PCDRT), c-precedence; F-precedence; String,
see under Discourse Representation precedence
Theory preconj (feature) 609–12
Particle 62, see also under Minor phrase as nondistributive feature 86, 217–18,
structure category 610–12
copular 190, 194 pred, see Semantic form
negative 69 Predicate logic 281–6
verbal 126 typed 283–4, 289, see also Meaning
Passive, see under Voice constructor, typed
pcase (feature) 61–2, 85, 211–13, 387n predlink (grammatical function) 32–3,
PCDRT, see under Discourse 192n, 194–7
Representation Theory Preposition, see also pform, pcase
Person marking oblique argument 34
hierarchy 166–7, see also subcategorization 133, 247
Animacy/saliency hierarchy Presupposition (information structure), see
pers (feature) 54, 69–74, 85, 220–2 Background information
resolution 48, 54–5, 72–4, 636–40 Presupposition operator (∂) 519
pform (feature) 61–62, 85 Primary object, see under Object
Phonological Phrase, see PhP grammatical functions
PhP (prosodic structure Priority union (/) 243–4
category) 399–403 Pro-drop, see Pronoun, incorporated;
Phrase structure rule 139–45 Semantic form, optional
annotated 168–70 Processability Theory 727–8
832 subject index

Projection architecture 263–74, 326–7, prosodic hierarchy 399–400, see also

377–80, 383–5, 400–11, 471, Prosodic Phonology
see also Codescription; ProTags, see Pronoun, right-dislocated
Correspondence function; Proto-Role argument classification,
Description by analysis; see under Argument structure
Information structure; Prosodic PW (prosodic structure category)
structure; Semantic structure 399–400
prom (feature) 371–3
Pronoun, see also Binding Q
incorporated 13, 135–6, 156, 179–85,
500–2, 543 QP (phrase structure category) 102
interrogative 673, see also Question, Quantifier, see also Indefinite noun phrase;
constituent see also under Discourse
logophoric 544 Representation Theory
null, see Pronoun, incorporated float, as subj test 21
personal 510–12 generalized 267n, 284–6, 302–12
meaning constructor 533, 717 meaning constructor 302, 310
reciprocal 334, 505–7, pair 284n
544, 571 scope 285–6, 302–5
reflexive 334, 503–10 ambiguity 285–6, 306–9
relative 666–7, see also Relative clause Question
absence 253–6, 297 constituent 10, 672–80, 721–4,
meaning constructor 708 see also Focus; Long-distance
possessive 711 dependency
resumptive 683–7, see also Manager in-situ 678
resource; Relative clause multiple 677–80, 725
right-dislocated (ProTags) 664–5 q attribute 673–5
prontype (feature) 84–5 declarative 420–6
prop (feature) 497 polar
Prosodic inversion 115, see also Clitic, meaning constructor 422
placement prosody 396, 420–6
Prosodic Phonology 398–400
Strict Layer Hypothesis 400 R
Prosodic structure 5, 395, 397–435, see
also P-diagram; P-form; Prosody; r (feature) 412
String, p-string; Syntax-prosody +r (mapping feature) 3, 332–9, 341, 354,
interface 364–5, see also Restricted
category, see Ft; IntP; PW; PhP; Utt grammatical functions;
features 401–3, 411–20, see also Clitic, Unrestricted grammatical
clitic; dom; field; fm; l; n_tone; functions
P-form, p-form; r; rb_tone; R see Morphology, realization relation R
syllstress; tone R see Binding, semantics, antecedency
Interface Harmony 417–20, 424, 432 relation between indices
Prosodic word, see PW R⃰ see Binding, semantics, antecedency
Prosody 395–435 relation between indices
and information structure 368 Raising 51, 150, 331, 545–59, see also
and semantic interpretation 278, 396, Control, functional; Open
420–34 grammatical functions
as test for constituency 94–5, 396–7, and backward control 553–6
400–5, 417–20 and scope 286, 556–9
 subject index 833

Raising (cont.) Restriction (|) 242–3, 280, 351

copy 559–61 Resumptive pronoun, see under Pronoun
possessor 600 Rheme (information structure), see
rb_tone (feature) 422 Comment
realized (f-structure predicate) Role and Reference Grammar 272
254, 256 Root Domain, see under Binding, domain
Regular expression 140, 199–211, 615 constraints
as constraint on string, see String, Rule-to-rule hypothesis 276
wellformedness conditions
definition 199–200 S
in functional description, see Functional
uncertainty S (phrase structure category) 112–13
in phrase structure rule, see under mapping to f-structure 128
Phrase structure rule S-form 409–10, 437–9, 442–3,
Regular language 199–211, see also see also fm; String, s-string
Regular expression S-string, see under String
closed under union, intersection, and S-structure, see Semantic structure
complementation 200 Secondary object, see under Object
complement (¬) 200 grammatical functions
finite-state automaton 615n Semantic form 45–7, 150–5, 276, 278–9,
intersection (&) 200 324
term complement (\) 201 argument list 46, 50–2, 154–5, 327–8
union (|) 200 function component (fn) 154–5
rel (feature) 313 nonthematic argument 51–2, 154, 155,
Relational Grammar 3, 11, 15n 545, 547
Agreement Law 16 optional 118, 131, 179–84, 500–2
Final 1 Law 21n reference to parts of 154–5
Nuclear Dummy Law 18 representation 50–1
relational hierarchy 11, see also semantic argument 51, 154
Grammatical functions, uniqueness 45, 150–3, 163, 278
hierarchy Semantic structure 266–71, 276, 288–90,
Relative clause 10–11, 159–60, 665–72, 327–9, 383–6, see also
706–21, see also Correspondence function, 𝜎
Complementizer-Adjacent features 312–13, 385–6, see also actv;
Extraction Constraint; anchored; animate; context;
Long-distance dependency; df; human; prop; rel; restr;
Modification; Pronoun, relative; status; var
Pronoun, resumptive typed 289
correlative 670–2, 725 Set
meaning constructor 708 as hybrid object 49–50, see also
nonrestrictive 713–16 Distributive/nondistributive
relpro attribute 666–70 feature
Relative pronoun, see Relative clause, Axiom of Extension 47
relpro attribute; see under closed 72
Pronoun description, see under Functional
Resolution, feature, see under Gender; description
Person complement (5 ) 214, see also under
restr (feature) 310, 481 Regular language
Restricted grammatical functions (obj𝜃 , empty (∅) 214
obl𝜃 ) 12, 17–19, see also +r formal representation of 48
834 subject index

intersection, see Intersection as open grammatical function

membership (∈) 147n, 214 29n, 600
negated (0∈) 214 Condition 21–2, 334, 337, 338,
symbol (∈), as f-structure 343, 344
attribute 215 in mapping theory 353
of f-structures, see under Functional Subset (⊆) 214
structure proper (⊂) 214
open, description, see under Functional Subsumption (6) 240
description syllstress (feature) 415
subset, see Subset Syntax-prosody interface 397–435,
union, see Union see also Prosodic structure;
“Shuffle” operator, see under Phrase Prosody
structure rule direct/indirect reference 397–8, 403,
Situation Semantics 281, 286–7 406
spec (feature) 83–8 pipeline architecture 404–5
Specifier (phrase structure position) 97–8, Syntax-semantics interface 279–80,
113, see also under CP; IP; VP 287–320, see also Correspondence
grammatical role 121 function, 𝜎; Meaning constructor;
mapping to f-structure 121–6, 177–9, Semantic structure
186
mapping to information structure 121 T
status (feature) 385
Strict Layer Hypothesis, see under Prosodic Tl (leftmost terminal), see under Functional
Phonology description
String Tr (rightmost terminal), see under
constraints, see String, wellformedness Functional description
conditions Tail (information structure) 370–3
N (“Next” function) 254 Template 230–7, see also Phrase structure
p-string 114, 252, 406–11, 437, see also rule, macro
Correspondence function, β; in lexical entry 231–4
P-form in phrase structure rule 235–7
as feature structure 409–20 parametrized 234
precedence 253–6 reference to 232
s-string 114–16, 252, 406–11, 437, Temporal relations 315–19, see also
see also S-form Aspect; Tense
as feature structure 409–20 ≺ (precedes) 316
relation to c-structure 115–16, $ (follows) 316
see also Correspondence function, π τ (temporal interval) 316
wellformedness conditions 116, 252–3, proper temporal inclusion (⊂) 316n
430–4 temporal inclusion (⊆) 316
Subcategorization 9, 12, 42–4, 247, 278, Tense 313–19, see also Binding,
338–9, see also Functional domain constraints, Minimal
description, CAT predicate; Finite Domain; Temporal
Grammatical functions, governable; relations
Semantic form, argument list at f-structure 54
constituent structure and 107, 121 tense (feature) 63–5, 85
+subj (mapping feature) 364 Term complement, see under Regular
Subject (subj) 20–4, see also Binding, language
domain constraints, Minimal Terms 12, 15–17, 357, see also Subject;
Complete Nucleus Object grammatical functions
 subject index 835

Thematic roles 11–12, 314–15, 324–8, 338, subcategorization 247

441n, see also Argument structure; transitive, meaning constructor 294,
Mapping theory 327–8, 361–2
hierarchy 22, 325, 333, 336, 339, 346, unaccusative 334–5, 353
697, see also under Binding unergative 334–5, 353
Theme (information structure), see Topic Visser’s Generalization, see Voice, passive,
Theta Criterion 334n control and
ToBI 403 Voice, see also Direct Syntactic Encoding,
tone (feature) 419 Principle of
Topic (information structure) 369–80 inverse 66, 365
as grammaticized discourse function, see passive 1, 164, 231, 340–5
Overlay function, status of topic and anticausative 342–3
topic (i-structure role) 372, 377–93 as test for grammatical function 12,
topic (value of df feature) 386–93 27
Topicalization 653–64, see also control and 585–6
Information structure; impersonal 21–2
Long-distance dependency lexical rule 329–30
“Tough” movement 680–2, see also “Philippine-type”, see Voice, symmetric
Long-distance dependency symmetric 66, 365
Trace, see Long-distance dependency, voice (feature) 63, 66, 85
empty category in VP (phrase structure category)
Transformational grammar 1–2 specifier 122–6
tests for grammatical functions 12 vtype (feature) 54, 63–4, 85
“Turnstile”, see Logical derivability
W
U
“Wanna” contraction 703–4
Unaccusativity, see Verb, unaccusative Weak crossover 696–701
Unbounded dependency, see Long-distance Word order freezing 138
dependency
Unification (8) 149, see also Functional X
description, solution algorithm
Union (∪) 214, see also under Regular X, see Non-projecting word
!
language X-bar theory 3, 97–9, see also
in feature resolution 636–9 Complement; Endocentricity;
Unique symbol, see Functional description, Exocentricity; Head;
instantiated symbol Non-projecting word; Specifier
Uniqueness Condition, see Consistency maximal projection 98–9, 112, 145
Universal quantifier (∀) 302 nonmaximal projection 98–9,
Unrestricted grammatical functions (subj, 111, 145
obj) 12, 17–19, 331, see also +r xadj (grammatical function) 12, 29–32,
Utt (prosodic structure category) 399–400 545, 587–93, see also Modification
meaning constructor 591
V xcomp (grammatical function) 12, 29–33,
51, 545–61
var (feature) 310, 481, 485, 487, 495, 497 as oblique grammatical function 16–17
Verb, see also Control; Raising as semantically restricted/unrestricted
auxiliary 99–100, 108–9, 120, 250–2, function 18–19
447–51, 601 in mapping theory 356
intransitive, meaning constructor 290, XFR (transfer component of XLE) 734
530 XLE 734–5