In 1his series INFO RMA 11 u l"l ~ l Ku\._, J. u .

l'-L

AND SENTENCE FORM

sz MICHAEL S. ROCHEMONT and PETER W. CULICOVER: English focus
amstn.w:tion1 and the t~ry of gnimmar
S3 PHIUP CARR: Linguistic realities: an autonornist me:tath«>ry for the generative
enterprile Topic, focus, and the mental representations of
54 EVE SWEETSER: From etymology to pragmatics: metaphorical and cultural
Upecl$ of semantic structure discourse referents
~ ll£GfNA BLASS: Relevance relations in discourse: a study with special reference
to Sissala
ANDJ\EW CHESTERMAN; On definiteness; a s1udy with special nfercncc 10
"
51
English and Finnish
ALESSANDRA GIORGI and GIUSEPPE LONGOBARDI: The syntax of
noun phrascs: configuration, parameters and empty categories
58
!19
MONIK CHARE:ITE: Conditions on phonological government
M. H. KLAIMAN: Grammatical voice
KNUD LAMBRECHT
60 SARAH M. 8. FAGAN: The syntax and semantics of middle constructions: a Department of French and Italian
study with special reference to German Univer.riry cf Texas al Au.rlin
fit ANJUM P. SALEEM!: Universal Grammar and language leamabi!ity
62 STEPHEN R. ANDERSON: A·Morphous Morphology
63 LESLEY $TIRLING: Switch reference and discourse representation.
64 HENK J. VERKUYL: A theory of aspectuality: the in1eraction between
fftllporal and a1emporal slructure
65 EVE V. Cl.ARK: The lexicon in acquisition
Mi ANTHONY R. WARNER: English auxiliaries: structure and history
67 P.H. MA1THEWS: Grammatical theory in 01e United Stales from Bloomfield
to Chomsky
68 UIUANA PROGOVAC: Negative and positive polarity: a binding approach
.,.69 R. M. W. DIXON: Ergativity
YAN HUANG: The syntax and pragmatics of anaphora
KNUD LAMBRECHT: Infonnation structure and sentence form: topic. focus.
71
and the mental ~rcscntations of discourse referents

Supplnnmtary volumes
RUDOLF P. BOTHA: Form and meaning in word formation: ll. study of
Afrikaans reduplication
AYHAN AKSU·KO<;: The acquisition of aspect and modality: the case of past
rcf'mnce itt Turki1h
MICHAEL 0 SlADHAIL: Modi=m Irish: grammatical structure and dialectal
variation
ANNICK DE HOUWER: The acquisition of two languages from birth: a case
study
LILIANE HAEGEMAN: Theory and description in generative syntax: a case
study in WC!t Flemish

DCAMBRIDGE
V UNIVERSITY PRESS
Published by the Press Syndicate of !he University of Cambridge
The Pin Building, Trumpington Street, Cambridge CB:Z IRP
40 Wes1 20th Street, New York, NY IOOll-4211 USA This book is dedicated to the memory or my parents:
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
Annie and Hans Lambrecht
©Cambridge University Press 1994

First published 1994

Reprinted 1995

Printed in Great Britain by Athen~um Press Ltd, Gateshead, Tyne & Wear

A <:afa/ogue record for this book is a"ai/abie from /hf;' Bri11sh Library

Lihrary a/ Congress cataloguing irt pub/icauorr data

Lambrecht, Knud.
Infonnalion structure and sentence form: lopic, focus. and ihe
menlal representations of discourse referents I Knud Lambrecht.
p. cm.
Includes bibliographical references and 1nde:<..
ISBN 0 521 38056 I
J. Grammar, Comparative and genera! - Senlences. 2. Grammar,
Comparative and general - Syntax l DiS\;ourse analysis.
4. Pragmatics. 1. Title.
P29S.L36 1994
415-dc20 93-30380 CIP

ISBN 0 521 38056 l hardback

. . . . . . . . . . . .lmllmllmllml. . . . . . . . . . . . .

Contents

Preface xiii

Introduction
I.IWhat is infonnation structure?
1.2 The place of information structure in grammar 6
1.3 Information structure and sentence form: a sample
analysis 13
1.3.1 Three examples 13
1.3.2 A note on markedness in information structure 15
1.3.J Analysis 19
1.3.4 Summary 24
1.4 Information structure and syntax 25
1.4. I Autonomy versus motivation in grammar 26
1.4.2 The functional underspecification of syntactic
structures 29
1.4.3 Sentence types and the notion of grammatical
construction 32

2 Infonnation 36
2.1 The universe of discourse 36
2.2 Information 43
2.3 Presupposition and assertion 51
2.4 The pragmatic accommodation of presuppositional
structure 65

3 The mental representations of discourse referents 74

3.1 Discourse referents 74
3.2 Identifiability 77
J.2.J Identifiability and presupposition 77
3.2.2 Iden ti Ji a h1 Ii t y and definiteness 79
 Preface

This book proposes a theory of the relationship between the structure of

sentences and the linguistic and extra-linguistic contexts in which
sentences are used as units of propositional information. It is concerned
with the system of options which grammars offer speakers for expressing
given propositional contents in different grammatical forms under
varying discourse circumstances. The research presented here is based
on the observation that the structure of a sentence reflects in systematic
and theoretically interesting \vays a speaker's assumptions about the
hearer's state of knowledge and consciousness at the time of an utterance.
This relationship between speaker assumptions and the formal structure
of the sentence is taken to be governed by rules and conventions of
sentence grammar, in a grammatical component which I call INFORMA-
TION STRUCTURE, using a term introduced by Halliday (1967). In the
information-structure component of language, propositions as concep-
tual representations of states of affairs undergo pragmatic structuring
according to the utterance contexts in which these states of affairs are to
be communicated. Such PRAGMATICALLY STRUCTURED PROPOSITIONS are
then expressed as fonnal objects \vith morphosyntactic and prosodic
structure.
My account of the information-structure component involves an
analysis of four independent but interrelated sets of categories. The first
is that of PROPOSITIONAL INFORMA Tl ON with its l\\'O components PRAGMATIC
PRESUPPOSITION and PRAGMATIC ASSERTIO~. These have to do with the
speaker"s assumptions about the hearer's state of knowledge and
a\vareness at the time of an utterance (Chapter 2). The second set of
categories is that of IDE'NTIFIABILITY and ACTIVATION. \Vhich have to do
\Vi th the speaker"s assumptions about the nature of the representations of

the referents of linguistic expressions in the hearer's mind at the time of

an utterance and with the constant changes v.'hich these representations
undergo in the course 0f a conversation (Chapter 3). The third category is
 Preface xv
xiv PrejO.ce
concepts. My definitions of the notions of "topic" and "focus" serve· a
that of TOPIC, which has lo do with the pragmatic relation of aboutness
between discourse referents and propositions in given discourse contex:ts similar goal.
(Chapter 4). The fourth category is that of Focus, which is_ that ~lement in The book contains relatively few analyses of linguistic da~ and most
a pragmatically structured proposition whereby the assertion differs from of these data are from English. However, the principles discussed have
the presupposition and \\'hich makes the utterance of a senten.ce wide crosslinguistic applicability. The volume also contains little by way
informative (Chapter 5). Each of these categories or sets of categones of formalism. I believe, however, that the presented theory is explicit
is shown to correlate directly with structural properties of the sentence. enough to be amenable to formalization. I hope my attempt -will
The theoretical orientation of this study is generative, if"generative" is encourage other researchers to pursue the task and to correct whatever is
understood as referring to linguistic analyses which do not merely wrong or misguided in my approach.
describe observed structures but \\'hich also attempt to explain why The present volume was originally planned as an introduction to a
certain structures do not occur in a grammar. Hov.:ever, in analyzing the book dealing with the relationship between syntax and discourse in
facts of information structure 1 \\'as often led to an alternative, non- spoken French. It was to provide the theoretical foundation for the
generative, approach to grammatical analysis, in which the function of_ a analysis of a number of pragmatically motivated French constructioii
given lexicogrammatical structure is not interpreted con1pos1tionally, ln types. As my work on that introduction progressed it became clear that
terms of the meanings of its parts, but globally. in terms of the formal the theory was too complex to be dealt with in the same volume as its
contrast between the entire structure and semantically equivalent application to a particular language. At the same time, in working on the
alternative structures provided by the grammar. In terms of Saussure's analysis of spoken French data, I realized that more space was needed to
fundamental dichotomy, the study of information structure requires an present a coherent picture of the manifestation of information structure
analysis not only of the SYNTAGMATJC relations bet\veen the elements of a in that language. The analysis of spoken French will therefore appear as a
sentence but also, and importantly, of the ASSOCIATIVE relations bet\veen separate volume (Lambrecht, in preparation).
different sentence structures as they are stored in the memory of speakers The present book is a continuation of research presented in my Ph.D.
and bearers. Methodologically, this study is an attempt to combine dissertation, which was completed in 1986 in the Department .. of
insights from formal and from functional approaches to grammatical Linguistics of the University of California at Berkeley, under the title
. . "Topic, focus, and the grammar of spoken French." I would like to
analysis.
My ambition in the present \Vork was not to define the 1nformat1on- thank again the members of my Ph.D. committee, Charles Fillmore,
structure component in such a \\'ay that it v.'ould fit one or another of the Suzanne Fleischman, Paul Kay, Johanna Nichols, and Karl Zimmer for
established generative or functionalist frameworks but rather to lay the their help and encouragement. Among them, I would like to single out
theoretical groundwork \\'hich v.'ill nlake such integration possible and my thesis director Charles Fillmore, v.•ho more than anyone has shaped
meaningful. (Such integration will no doubt be easier \Vith fra_meworks my ideas about language and linguistics. His influence is manifested in
which do not postulate a hierarchical ordering and strict separation of the many aspects of the present book. I am also grateful to Wallace Chafe,
different components of gram1nar.) Throughout the book, the emphas_is \vho allov.·ed nle to pursue my research v.·hile I was his research assistant
is on the notional foundations of tht! theory of information structure in at the Institute of Cognitive Studies at Berkeley. Much of the material in
natural language. Special importance has therefore been attributed to Chapter 3 of the present volume \.'.'as conceived in relation and reaction to
elaborating the basic concepts and term~ needed to describe and define his ideas. And I would like to thank my former fellow graduate students
this underexplored part of gra1n1nar. In particuL1r, I have trieJ to provide Farrell Ackerman, Claudia Brugmann, Giulia Centineo, Amy Dahl·
definitions of, and alternative l.ibel~ rL)f, the concepts of "new slro1n, Pamela Downing, !\1ark Gawron, Tom Larsen, Yoshiko
information" and "old information,, .,, htch \'.ii! help prevent some of f\.1atsumoto, Sh.igeko Okamoto, David Solnit, and especially Cathy
the confusion that tends to creep intL• analy :-.es .,, h1cb make use of these O'Connor for many stimulating discussions concerning my work and
theirs. Thanks are due also lo Ruth Berman, Martin Harris, Peter Pause,
....
r.
Ff
l'Rll'<*o l. Annin Sdnregler. and Sandy Thompson for their WililDCUts
on "ftrious parts of my dissertation, and to my friend Lenny Moss for
fMking me aware of connection s between my work and research in other
8cicntific disciplines.
The present book has greatly benefitted from discussions with Lee
l Introduction
:.Baker. Charles Fillmore, Suzanne Fleishmann , Danielle Forget, Mirjam
Fried. Paul Kay, Manfred Krifka, Jean~Pierre Koenig, Dale Koike,
Fra~is Lagarde, Ellen Prince, Carlota Smith, and Tony Woodbury. I
am particularly grateful to Robert Van Valin for his interest in my work
and his encouragement in difficult times, and to Laura Michaelis for her 1.1 What is infonnation structure?
Joving help and her faith in the value of this enterprise. Special thanks are
due ,aJso to Matthew Dryer for his penetrating criticism of certain There has been and still is disagreeme nt and confusion in linguistic theory
.sections of the book, and to Randy LaPolla, who provided much help about the nature of the component of language referred to in this book as
INFORMATIO N STRUCTURE and about the status of this componen t in
the
with a careful reading of the manuscript. I am also grateful to a no-
fonger.ano nymous reviewer for Cambridge University Press (Nigel overall system of grammar. The difficulties encountere d in the study of
Vmccnt) for his comments on parts of the manuscript . But above all I informatio n structure are in part due to the fact that grammatic al
want to thank Sue Schmerling for her detailed comments on various analysis at this level is concerned with the relationship bet\veen linguistic
versions of the manuscript , for her gentle yet inexorable criticism, and for fo~ and the mental states of speakers and hearers and that the linguist
many hours of stimulating discussions in the cafCs of Austin. And last, dealing with informatio n structure must deal simultaneo usly with formal
but certainly not least, I would like to thank my friend Robby Aronowitz, and communica tive aspects of language. Infonnatio n-structure research
neither offers the comfort \vhich many syntacticia ns find in the idea of
linguist turned MD, for his love and support over the years. Finally, I
thank the University of Tel(as at Austin for a Research Grant which st~dying an autonomou s formal object nor provides the possibility
enjoyed by sociolingui sts of putting aside issues of formal structure for
faalitated work on this book during one air-conditi oned summer.
the sake of capturing the function of language in social interaction .
Negative or defeatist views of informatio n-structure research are
therefore not uncommon . even among linguists who emphasize the
importance of the study of linguistic pragmatics . The following quote
concerning the role of topic and focus in linguistic theory illustrates such
views: "Terminol ogical profusion and confusion. and underlying
conceptual vagueness, plague the relevant literature to a point where
little may be salvageable .. (Levinson 1983:x). In his own book on
pragmatics , Levinson explicitly excludes the analysis of the relationshi p
between pragmatics and sentence form, in particular the analysis of
topic-comm ent structure. Yet interestingly, he contradicts his own
negative appraisal later on in his book \Vith this comment:

Perhaps the most interesting [kinds of interaction between conversa~

uonal s1ruc1ure and s;..ntax] lie in the area subsumed by the (rather
unclear) notion of topic. for many of the syntactic processes called
mf'H'emcnr rules seem to have the function of1nd1cating how information
 Whal is informalion structure? .3
l '11r(}(lUt'l1011

the tailoring of an utterance by a sender to meet the particular assumed

in the clause relales to what has been talked about before ... Perhaps the
needs of the intended receiver. That is, information packaging in natural
great bulk of the derivational machinery in the syntax. of natural
languages can be functionally explained by reference to the specialized l~nguage refl~ts the sender's hypotheses about the receiver's asswnp-
uons and beliefs and strategies. (Prince 198la:224)
conversational jobs that many sentence structures seem to be designed
to perform. (1983:373)
As the word "tailoring" suggests, informatio n structure is concerned with
It seems to me that any theoretical research that has even the slightest the FORM of utterances in relation to assumed mental states of speakers
chance of eventually explaining "the bulk of the syntactic processes called and hearers. An important part of the "hypothese s about the rece.iverrs
'movement rules'" is worth pursuing, however discouragin g the present assumptio ns" are hypothese s about the statuses of the mental
state of our knowledge may be. I hope this book will reduce some of the representat ions of the referents of linguistic expression s in the mind-of
"confusion and vagueness" which "plague the relevant literature" and the receiver at the moment of utterance. About these statuses Chafe
(1976) writes: '
thereby help reduce the gap between "formal" and "functiona l" , ·'·'
1
approaches to the study of language. !he statu_ses to be di~ussed here have more to do with how the cOn~pt
The difficulties encountere d in the analysis of the informatio n- is tra_nsrwtted than with the content itself. Specifically, they all haV~. .to
structure componen t of grammar are reflected in certain problems of do with t_he s~aker's ~ssessment of how the addressee is able to proCCss
terminolog y. In the nineteenth century, some of the issues described here, what he IS saying against the background of a particular context. Not
in particular the issue of word order and intonation, were discussed in the oo.ly do people's minds contain a large store of knowledge, they.arc also
at any one moment in certain temporary states with relation io that
context of the relationshi p between grammar and PSYCHOLOGY, as knowledge ... Language functions effectively only if the speaker takes
manifested in the difference between "psycholog ical" and "gramma- account of such states in the mind of the person he is tal.king"tO ' (Chafe
.
tical" subjects and predicates (see e.g. Paul 1909, especially Chapters 6 1976:27)
and 16). Among the labels which have been used by twentieth-c entury
linguists are FUNCTIONAL SENTENCE PERSPECTIVE, used by scholars of the Crucial here is the ob.servation that the study of informatio n structure is
Prague School of linguistics, INFORMATION STRUCTURE or THEME (Halliday n~t concerned with lexical and proposition al content in the abstract but
with the v.'ay such content is transmitted .
1967), INFOR~1ATION PACKAGING (Chafe 1976), DISCOURSE PRAG~tATlCS, and
and most recently INFORMATICS (Vallduvi 1990b). What unites linguistic An imp_ortant caveat is in order here. Even though informatio n
research done under one or another of these headings is the idea that structure JS concerned with such psychologi cal phenomen a as the
speaker's hypotheses about the hearer's mental states, such phenomena
certain formal properties of sentences cannot be fully understood without
looking at the linguistic and extralingui stic contexts in \Vhich the are relevant to the linguist only inasmuch as they are reflected in
sentences having these properties are embedded. Since discourse involves G~MMATICAL STRUCTURE (morphosy ntax, prosody). The importance of
this caveat cannot be overempha sized. l take informatio n structure to be
the USE of sentences in communica tive settings, such research is clearly
a co~ponent of GRAM.MAR, more specifically of SENTENCE GRAMMAR, i.e. I
associated with the general area of PRAGMATICS. The general domain of
take It to_ be a deteffi1ln1ng factor in the formal structuring of sentences.
inquiry into the relationshi p between grammar and discourse is therefore
lnf?nnat1o n structure is NOT concerned with psychologi cal phenomena
often referred to as "discourse pragmatics ." The reason I have adopted
which do ~ot have correlates in grammatic al form. This important
Halliday's tenn "informati on structure" is because in the present book · r
limitation 1mpose d on 1n1ormat1o ·
n structure research is stressed by
special emphasis is placed on the STRUCTURAL implication s of discourse- Prince:
pragmatic analysis. Occasional ly I \\"ill alsl• use Chafe's more vivid
"infonnati on packaging, " whose part10Jl1) n<Jn-latin.:tte character makes \Ve may now word the basic problem as follows. From I.he point of view
it less appropriat e for internation al use." of_ the speaker/writer, what kinds of assumptions about the hearer/
What then, is informatio n structure 1..H 111rormalion pa...-:kaging? rl."ader have a bearing on the fonn of the text being produced ... ? From
According to Prince ( 198la), informJ.t!l•n p,;1.l·kaging has to do v.·ith the point of view of the hearer/reader, what inferences wills/he draw on
 What is information structure? 5
the basis of the partictllar form chosen? We are, therefore, NOT given conversational contexts, but rather with the discourse circum-
concerned with what one individual may know or hypothesize about
stances under which given pieces of propositional information are
another individual's belief-state EXCEPT insofar as that knowledge and
thote hypotheses affCfZt the fonn~ and understanding of t1Nou1snc expressed via one rather than another possible morphosyntactic or
productions. (Prince 1981.,233) prosodic fonn. Oversimplifying a little, one could describe the difference
between conversational pragmatics and discourse pragmatics as follows:
Thia limitation makes it necessary to draw a theoretical distinction while conversational pragmatics is concerned with the question of why
_between the domain of infonnation structure (or discourse pragmatics) as one and the same sentence form may express two or more meanings,
understood in this book and the general domain of PRAGMATICS, which is discourse pragmatics is concerned with the question of why one and the
·Often understood to be a subdomain ofsemantics. 3 Indeed since the early same meaning may be expressed by two or more sentence forms. In the
tieventies, when the work of such language philosophers as Austin and former there is no necessary relationship between the particular context-
Grice became integrated into mainstream American linguistics, the tenn dependent interpretation of a proposition and the morphosyntactic or
""pragmatics" has been intimately associated with the study of h1EANING. prosodic structure of the sentence expressing it; in the latter the
More particularly, "pragmatics" has been used to refer to the study of relationship between a given sentence form and the function of the
those aspects of the meaning of sentences which cannot be captured with sentence in discourse is directly determined by GRAMMATICAL CONVEN-
the tools of truth-conditional semantics. 4 Pragmatics in this sense, or TJON. 5
..conversational pragmatics" (as one might call it in contradistinction to There is thus an importanl though by no means always clear-<:ut
..discourse pragmatics"), is not so much concerned with grammatical di!Terence between the two areas of pragmatics, che "conversational" and
structure as with the INTERPRETATION of sentences in relation to the "discourse"' area. In the forn1er a~ Grice has emphasized, the
1

·conversation~! settings. It was in order to account for this relation inferences which a hearer dra\l,'S on the basis of the relationship between
between interpretation and setting that Grice (1975) developed the the form of a sentence and the particular conversational context in which
concept of CONVERSATIONAL IMPLJCATURE . the sentence is uttered are determined by general principles of goal-
. The concern with meaning in conversational pragmatics is predomi- oriented behavior, \Vhich are applicable to language as well as to other
nant not only in the study of conversational implicatures but also in the domains of mental activity. In the latter. the pragmatic interpretation
-study of certain aspects of language use which have more traditionally triggered via a particular association between a sentence form and a
: been referred to as .. pragmatic" and which are t1early refletted in discourse context is determined by rules or principles of grammar, both
-linguistic form. What I have in mind is the study of the pragmatic language-specific and universal. If in this hook references to conversa-
:strut.ture of individual lexical items, which for the purpose of the present tional pragmatics are relatively scarce, it is not because I underestimate
1di!cussion we may refer to as "lexical pragmatics." A good example is the the importance of Gricean pnnc1ples of interpretation or the explanatory
.shl:dy of deixis. the domain par excellence in which language structure po\ver of speech-act theory but because I think that information structure
land language use are inseparably intertwined. The study of the inherent relates only indirectly to such principles.
pragmatic properties of deictic expressions is essentially the study of the I propose, then, the fol]o\l,·1ng definition of "information structure" as
contributions which these expressions make to the MEANING and understood in this book:
'INTERPRETATION rather than to the STRUCTURE of the sentences containing
ll'lfOR~1A"TION STRl1cnrRf: That component of sentence grammar in which
.~em. Nevertheless lexical pragmatics differs from <:onversalional propositions as conceptual representations of states of affairs are pa.ired
pragmatics-and in this respect is related to information structure-in \\1th lexicogrammatica! structures 1n accordance .,,,,ith the mental states
that the interpretation of sentences containing such expressions is not of 1nlerlocutors who u~e and interpret these siructures as units of
""cfcfenninCd by conversational inferences but by lexical form 1nfonnation in given d1;;c0urse contexts.
The student of information structure, on the other hand, is not The information structure of a sentence is the formal expression of the
primarily concerned with the interpretation of \\'Ords or sentences in pragma11c structuring of a rropos1t1on 1n a discourse. A proposition
 The place of information structure in grammar 1
6 Introduction
[The level of utterance] "makes it possible to understand bow the
which has undergone pragmatic structuring will be called a PRAGMATI·
semantic and the grammatical structures function in the very act of
CALLY STRUCTURED PROPOSITION. The most important categories of
communication, i.e. at the moment they are called upon to convey some
information structure are: (i) PRESUPPOSITION and ASSERTION, which have extra·linguistic reality reflected by thought and are to appear in an
to do with the structuring of propositions into portions which a speaker adequate kind of perspective" (Firbas). Further all extra-grammatical
assumes an addressee already knows or does not yet know; (ii) means of organizing utterance as the minimal communicative unit are
IDENTIFIABILITY and ACTIVATION, which have to do with a speaker's contained at this level as well. Such means are: rhythm, intonation ... ,
the order of words and of clauses, some lexical devices, etc. (Dand
assumptions about the statuses of the mental representations of discourse ·.~
1966:227)
referents in the addressee's mind at the time of an utterance; and (iii)
TOPIC and FOCUS, v•hich have to do with a speaker's assessment of the In a similar vein, and acknowledging the influence of the Prague Schoo~
relative predictability vs. unpredictability of the relations between Halliday (1967) defines what he calls THEME as the third of three areas in
propositions and their elements in given discourse situations. the domain of the English clause, the two other areas being TRANSITIVrIY
Information structure is formally manifesled in aspects of prosody, in (roughly the study of syntax and semantics) and MOOD (roughly the study
special grammatical markers, in the form of synlactic (in particular of illocutionary force):
nominal) constituents, in the position and ordering of such constituents
Theme is concerned with the information structure of the clause; with
in the sentence, in the form of complex grammatical constructions, and in the status of the elements not as participants in extralinguistic processes
certain choices between related lexical items. Information structure thus but as components of a message; with the relation of what is being said
- intervenes at all meaning-bearing levels of the grammatical system. to what has gone before in the disrourse, and its internal organization
Information-structure analysis is centered on the comparison of into an act of communication ... Given the clause as domain~
transitivity is the grammar of experience, mood is the grammar of
semantically equivalent but formally and pragmatically divergent
speech function, and theme is the grammar of discourse. (Halliday
sentence pairs, such as active vs. passive, canonical vs. topicalized, 1967:199)
canonical vs. clefted or dislocated, subject-accented vs. predicate-
accented sentences, etc. Using a term introduced by DaneS ( 1966), I A threefold division of grammar is also postulated by Dik (1978, 1980),
will refer to such sentence pairs as pairs of ALLOSENTENCES. Differences in who in his model of "Functional Grammar" distinguishes the three levels
the information structure of sentences are ahvays understood in terms of of "semantic functions," "syntactic functions," and "pragmatic
contrasts between allosentences, i.e. against the background of available functions" (1980:3). It should be noted that for Danes, Halliday, and
but unused grammatical alternatives for expressing a given proposition. Dik, the fonnal domain of information structure (functional sentence
perspective, theme, pragmatic function) is the SENTENCE or the CLAUSE.
Thus for these linguists, as for the author of the present study,
information structure belongs to SENTENCE GRAMMAR. It is not concerned
1.2 The place of information structure in grammar with the organization of discourse, but with the organization of the
Linguists who have concerned themselves \vi th information structure and sentence within a discourse.
its status within the overall system of grammar have often described it as A somev.·hat different threefold division of grammar is found in
one of THREE COMPONENTS (or LEVELS) of gramn1'1r. For example, ID a Fillmore 1976. Although Fillmore's notion of pragmatics is much
paper summarizing the approach to gran1n1ar t'1k..:n by linguists of the broader than my notion of information structure, his definition for
Prague School (Mathesius, Firbas, Benes. Vachek, DaneS, and others), linguistics of the notions SYNTAX, SE/'.fANTICS, and PRAGMATICS is.
Frantisek DaneS (1966) distinguishes the following three levels: (i) the nevertheless relevant:
level of the GRAMMATICAL structure of sentt:nL-C$. (it) Lhe level of the Syntax, in short, charactenzes lhe g.rammatical forms that occur in a
SEMANTIC structure of sentences, and (in) the le\'el of the organization of language, whereas semantics pairs these forms with their potential
uTIERANCE. 6 Concerning the third level, DancS \\rites (quoting Firbas): communicative functions. Pragmatics is concerned with the three-

'j
.................
I Introduction
--------------~~~~
The place of information structure in grammar 9

termed relation that unites (i) linguistic form and (ii) the communicative One remarkable early exception to the lack of concern among
functiom that these forms are capable of serving, with {iii) the contexts generative linguists for the function of language in discourse is research
or9Cttings in which those linguistic forms can have those communicative
on the focus-presupposition distinction within the framework of the so-
functions. Diagrammatically,
called "Extended Standard Theory'" (Chomsky 1970, Jackendoff 1972,
Syntax [fonn] Akmajian 1973). Characteristically, most of this research concerns
Semantics [form, function}
pragmatic distinctions which are marked phonologically only, i.e. which
Pragmatics [form, function, setting] (Fillmore 1976:83)
do not involve alternative syntactic structores. The possibility of syntactic
Fillmore's diagram provides a further explanation for the earlier structures having specific communicative functions is acknowledged in
mentioned difficulties encountered in the study of information Chomsky 1975 (p. 58). In more recent work, Chomsky (1980:59!1)
structure. Indeed the diagram suggests that pragmatics, since it suggests that matters of stress and presupposition may fall within
presupposes the other two levels, is the most complex of the three "grammatical competence" rather than "pragmatic competence," both
levels of grammar, hence the most difficult to be clear about. The types of competence being part of "the mental state of knowing a
diagram also suggests an explanation for why syntax, being in a sense the language." The notion of pragmatic competence is left rather vague by
simplest level, has been given such preference in modem as well as Chomsky, but it seems that it is closer to what I have called
traditional linguistics. Although I do not think that in order to engage conversational pragmatics than to discourse pragmatics, leaving open
successfully in discourse-pragmatic research one must first have a the possibility that information structure is indeed part of grammar.
complete account of the levels of syntax and semantics, I do believe In the present study, the question of the function of allosentences, i.e.
that such research requires awareness of the intricate relationships of multiple structures expressing the same proposition, is given primary
betWeen the three levels and of the various ways in which they interact. theoretical importance. The functional linguist's concern with the
An illustration of the complex ways in which syntax, semantics, and diversity of competing grammatical structures is comparable, mutatis
information structure interact with each other in different languages will mutandis, to the ecologist's concern \\'ith the diversity of organisms. To
be presented in Section 1.3. quote the biologist Stephen Jay Gould:
··If-we accept a model of grammar containing a subdivision into
different domains along the lines indicated in the above quotes, we may In its more restricted and technical sense, ecology is the study of organic
diversity It focuses on the 1nteractJon of organisms and their
ask· ourselves whether these domains are autonomous subsystems or
environments in order to address what may be the most fundamental
whether they are interdependent. It is well known that in the Chomskyan question in evolutionary biology "\\lhy are there so many kinds of
view the level of syntax is an autonomous level of linguistic structure hving things'.'" (Gould 1977:119)
wliile semantics is a component which 'interprets' syntactic structure. In
generative grammar. the theoretical problem posed by the existence of If ecology focuses on the interaction of organisms and their environ-
different 0 cognitively synonymous" (Chomsky 1965) formal expressions ments, the study of information structure focuses on the interaction of
of a given proposition has mostly been addressed in terms of the question sentences and their contexts. It addresses the fundamental question of
ofhoW'sUch different structures are to be generated. Since the business of why there are so many kinds of sentence structures.
gtnerative syntax is seen as that of specifying which structures arc Reacting to the vie\v of syntax as an autonomous structural component
permitted by a grammar, the fact that such semantically related of grammar. contemporary linguists from various schools have proposed
structures have different communicative functions has received little models of language in 1,1,,hich the level of syntax is not the most basic level
attention. In particular, one theoretical question is not asked: \VHY should and in which syntax ts not. or not to the same extent, considered
grammars provide the means of generating so many di!Tercnt syntactic aut~1nomous. The n1ost radical departure from the belief in the autonomy
and prosodic structures for expressing one and the same propositional of syntax is found in the \"an0u" ··runct1ona!ly·· 0riented approaches to
content? grammar \\'h1ch have heen de1,·ct(1ped in Europe and the United States
 The place of information structure in grammar 11
Ullt UUUt. I IUll
'"
over the past twenty years or so, either in direcl reaction to explanation for some grammatical phenomenon does not in principle
transformatio nal-generativ e grammar or as continuation s of well- obviate a formal account of it, and that a formal explanation does not
established older linguistic trends. Analyses of the relationship between make a functional account superfluous or irrelevant.
syntax and discourse are often said to be "functional" rather than I see my own research as located somewhere in between the ..form.al"
"formal" insofar as they are primarily concerned with explaining the and the "functional" approaches to syntax. I do not believe that linguistic
communicative function of morphosyntactic or intonational structure in form can be exhaustively accounted for in terms of its communicati ve
discourse rather than with developing formal models of the structure of function in discourse. Nor do I believe that syntax is autonomous in the
sentences. A clear statement concerning the importance attributed to sense that it does not directly reflect communicati ve needs. As I said
functional considerations is the following: before, this book is based on the assumption that there are aspects of
grammatical fonn Y.1hich require pragmatic explanations. But it is also
In terms of the well-known distinction between syntax, semantics, and based on the understandin g that there are many formal phenomena for
pragmatics, the functional approach to language regards pragmatics as
which such explanations are not readily available. As I see it, the
the all-encompassing framework within which semantics and syntax
must be studied. It regards semantics as subservient to pragmatics, and interesting theoretical question is not whether or to what extent syntactic
syntax as subservient to semantics. Syntax is there in order to allow for form can be studied in isolation from communicati ve function, but to
the construction of formal structures by means of which complex what extent a research agenda based on the idea of the autonomy of
meanings can be expressed; and complex meanings are there for people syntax can further our understandin g of the workings of human
to be able to communicate with each other in subtle and differentiated language. The issue which ultimately divides the "formal" and the
ways. (Dik 1980:2)
"functional" approach is not so much disagreement about facts but the
The common characteristic of these often heterogeneou s approaches is question of what constitutes explanation in linguistics.
that the syntactic component and the information- structure component A view of the relationship between the different levels of grammar
of grammar are seen as connected to each other rather than as which I find appealing is expressed in the following passage from
independent subsystems. Sometimes certain syntactic phenomena whose Fillmore's already mentioned paper "Pragmatics and the description of
discourse'':
discourse function cannot be clearly established synchronically are
explained in diachronic terms as grammaticiz ations of erstwhile
I assume three ways of looking at linguistic facts, the three viewable as
functional distinctions. In some cases, the difference between forn1 and independent from each other or not, depending on whether we arc
function has been minimized to an extreme degree, so that the two levels thinking of classes of facts or explanations. In the broadest sense
I believe that syntactic, semantic, and pragmatic FACTS can ~
7
have been interpreted as ultimately identical.
It is in my opinion an unfortunate outcome of certain tendencies in distinguished from each other, but I also believe that some syntaclic
structuralist and post-structur alist linguistics that the so-called formal facts require semantic and pragmatic explanations and that some
semantic facts require pragmatic explanations. Put dilTerently, inter-
and functional approaches to grammatical structure are seen as being
preters sometimes use semantic and pragmatic information in making
diametrically opposed rather than as complementi ng each other. The judgm~nts about the syntactic structure of a sentence, and they
antagonism of form and function 1n linguistics 1s not one of necessity but sometimes use pragmatic facts in making semantic judgments.
rather of methodologic al and often ideological preference. If there exists (Fillmore 1976:81)
some level of autonomous structure :.il \Vhicb any appeal to such non-
structural notions as "communica tive functl!ln .. 1s e;..cluded, this does not Syntax may be autonomous in its O\\'n domain, but by its nature it must
entail the non-existence of another le\'el Jl \vhich autono1nous structure is provide the resources for expressing the communicati ve needs of
indeed connected with communic..1tn·e functi~1n. nor Joes it entail that all speakers. Therefore its nature cannot be fully underslood unless we
of grammatical structure must be e4uall:- Jut..:1n<:•n1uu;. Jr I use the term explain the principles \Vhich detem1ine its function in discourse. [n my
"functional" in this book, 1t is with th.: L1nJ...:r'>t.111d1ng that a functional view, the n1ost promising but perhaps also the most difficult approach to
 12 11ttrodr«tion A sample analysis 13

lflDUl18tical analysis is one in which the different components of of such an attempt I would like to mention Comrie's shon typological
grammar are seen not as hierarchically organized independent sub- comparison of Russian and English clause structure as two grammatical
systems but as interdependent forces competing with each other for the domains in which semantic and pragmatic roles, grammatical relations,
timited coding possibilities offered by the structure of the sentence. 8 I and morphological cases interact with each other in different language-
take a linguistic theory of high explanatory value to be one in which these specific ways (Comrie 1981: Section 3.5). GivOn's Syntax (J984) also
forces are not only analyzed in isolation but also in their multiple deserves to be mentioned here. Integrated descriptions have been
dependence relations to each other. In such a theory the grammatical attempted within the frameworks of "Functional Granunar" (Dik
structures found in particular languages would then be seen as language- 1978, 1980) and "Role and Reference Grammar" (Foley & Van Valin
specific manifestations of the interplay between the different components 1984 and especially Van Valin 1993; the latter has integrated parts of the
of grammar. theory presented in this book), Attempts at dealing with notions of
If we conceive of the structure of the clause as a domain in which the information structure have been made also within the framework of
different components of grammar- syntax, morphology, prosody, 'Lexical Functional Grammar' (see e.g. the djscussion of "topic" in
semantics, information structure-compete and interact with each other, Bresnan & Mchombo 1987). Most recently, "Construction Grammar"
regulated by universal principles and language-specific constraints, we has developed a descriptive framework. in which morphosyntax.
can understand why, for example, the notion of ''subject" has given rise to semantics, and pragmatics are lreated as integrated aspects of
so much discussion and controversy in the recent and less recent history of grammatical constructions (see Fillmore 1991 and Fillmore, Kay, and
linguistics. The label "subject" has been applied to phenomena from any O'Connor 1988, as well as the "constructionist" analyses in Lambrecht
of the four components of syntax, morphology, semantics, and 1984b, I986a, 1988b, 1990, and 1992). A brief infonnal characterization
information structure. In "subject-prominent" languages (Li & Thomp~ of Construction Grammar wlll be presented in Section 1.4 below.
son 1976) like modem English or French, "subject" has been defined as a
prominent grammatical relation which is crucially involved in certain
syntactic phenomena such as verb agreement, passive, "raising" 1.3 Infonnation structure and sentence fonn: a sample analysis
constructions, etc. But it has also been identified via the semantic notjon
My own objective in this book 1s not to offer a comprehensive description
of agent, the pragmatic notion of topic, and, in so-called "case
of the grammatical system. Rather I \viii concentrate my attention on the
languages," via a morphological case (the "nominative"). All of these
relatively unexplored level of information structure, relating it whenever
definitions have their independent justification, but also their problems. It
possible to the other levels, and offering here and there samples of fonna!
is easy, indeed, to provide examples of sentences in which one or the
analyses. However, I \vould like to illustrate here, with examples from
other of the various suggested criteria for subjecthood does not apply. 9
three languages, some of the in1plications or a multi-level approach to
, An important insight to be gained from the difficulties encountered in
grammatical analysis of the kind mentioned in the preceding section. This
def"ming the notion of subject is that certain grammatical phenomena
sample analysis \viH also aflow me to demonstrate the possibilities and the
cannot be fully captured unless elements from different levels of grammar
limits of information-structure research. The analysis involves the use of
are seen as mutually determining each other. This insight has been
a fev.• information-structure concepts v-•hich I will brieny explain as I go
commonplace in many good traditional grammars and has been kept
along but \~·hich \Viii be fully explained only in later chapters.
alive in work by such linguists as Paul (1909), Mathesius (1928) and orher
Prague School scholar.;, Bally (1932), and more recently Halliday (1967)
and Dixon (1972). In modern generative or typological studies it is only
1.3. 1 Three examples
relatively recently that attempts have been made at providing integrated
descriptions in which the levels of synta.,, morphology, semantics, and Consider the fo!Jo,\"ing real-life ~ituation At a bus stop. the departure of
information structure are dealt \Vi th on a par. As an outstanding example a crammed bus is delayed hecau<;e a woman loaded down '~·ith shopping
--- •• JlllrUtlUl'l IUll

bags is boarding very slowly. Turning to the impatient passengers in the

A sample analysis

argument designating the non-functioning entity, i.e. the speaker's car.

IS

(Strictly speaking, the French predicate est en panne does not express an
bus, the woman utters the following sentence with an apologetic smile:
event but a state; the event interpretation is pragmatically inferred.) Let
(1.1) My CAR broke down. us call "theme" the semantic role of the noun argument car (macchina,
In tenns of the taxonomy of focus structures developed in Chapter 5, the voilure) in the given proposition. 11 The sentences, at least in English and
example in (l.1) is of the "sentence-focus" type, i.e. of the type in which French, also contain a secondary proposition, within the syntactic
the domain of the "new information" extends over the entire proposition, domain of the noun phrase, expressing the relationship of possession
including the subject. The communicative function of such sentence-focus between the speaker and the car. Let us call the speaker semantically the
structures is either to introduce a nev,i discourse referent or (as in this "possessor," expressed in the possessive determiners my and ma, and the
case) to announce an event involving a new discourse referent. Sentences car the "possessed."
with the latter kind of communicative function will be called "event- In addition to their propositional meaning, which is derivable from
reporting sentences" (see Sections 4.1.l and 4.2.2). In (1.1) the new their lexicogrammatical structure, the sentences also have an utterance--
discourse referent is the woman's car. The constituent in small capitals is specific conversational meaning. We can understand the com.m.uniccltive
the locus of the main sentence accent, which is characterized by pitch function of the woman's utterance only if we understand its relevance in
prominence. The sentence accent serves as the FOCUS MARKER, i.e. as the the situati?n i_n the bus. The point of the woman's remark is to explain
formal indicator of the focus structure of the sentence. Notice that her behavior 1n the bus, not to tell her audience about the mechanical
although in (I.I) the focus accent falls on the subject only, the focus state of her car. The state of the car is relevant only inasmuch as it
domain as the domain of "new information" extends over the entire explains the woman's present situation. Even though the car is the subject
proposition (see Sections 5.1.2 and 5.6.2). 10 argument at the conceptual level of the proposition, and even though the
The utterance in (I. I), an example of actually observed speech, could noun phrase expressing this argument is the grammatical subject of the
have the follo\.\ ing natural equivalents in Italian and in French, given the
1 sentence (at least in English and Italian), the expression my car does not
same real life situation: correspond to a TOPJC at the level of the pragmatically structured
proposition. Rather the topic is the speaker: the woman, not the car, is
(1.2) Mi si i! rotla la MACCHINA.
"what the utterance is about." This pragmatic fact will turn out to be a
to-me itself is broken the car
crucial factor in the grammatical structuring of the sentence in all three
(1.3) J'ai ma VOITURE qui est en PANNE. languages.
I have my car that is in breakdown

The communicative intention and situation being identical in the 1.3.2 A note on n1arkedness in information structure
three examples, we may say that the three sentences have the same
MEANING, both semantically and pragmatically. Semantically (truth- Before I proceed to the analysis of our three examples, I would like to
conditionally) they are synonymous in that they express the same state make explicit two assumptions v.•hich I will be making concerning the role
of affairs in a given world. They presuppose, via the use of the definite of ~fAR~EDNES~ in the expression of information structure. (These
possessive noun phrase, that the speaker has a car - a pragmatic assumptions \V11l be justified and further developed in Chapters 4 and
presupposition made possible by the cultured situatil1n in \~·hich it is 5.) The first assumption is that in our three languages the PRAGMATICALLY
UNM.ARKED CONSTITUENT ORDER for sentences V.'ith full Jex.ical arguments is
expected that people have cars - and they a'.>-..ert th..il this car is presently
not in working condition. (The U)e l1r the tern1~ ··presuppose" and Subject-Verb-Object . (Pronominal arguments obey rather different
"assert" will be explained in Chapter 2.J A.I! three sentences express a syntacuc and prosodic constraints.) The second assumption is that in
simple proposilion composed uf l\v<1 elenh:nts· :i ((>ne-placeJ predicate these languages the PRAG MA Tl CALL y UNMARKED SEt-.'TENCE.ACCENT POSffiON
denoting an event involving '.>C1n11.: n1ech~1n1,_,i\ n1alfunction and one is clause-final (or near-final, 1f the clause contains "deaccented" post-
 A sample analysis 17

focal material; see Section 5.3.3). Assuming the existence of a relationship need be even screws. The difference is of course not that hammers have
between sentence accent and focus, these two assumptions, taken no special function or are functionally neutral but simply that their
10gether. entail that in the unmarked case a clause-initial subject will potential domain of application is larger, hence that they tend to be used
have a topic relation and a clause-final object a focus relation to the more often.
proposition (the terms "topic relation" and "focus relation" will be Concerning the pragmatic markedness status of grammatical struc-
_explicated in Chapters 4 and 5). The UNMARKED INFORMATION-STRUCI1JRE tures, we can state the following general rule: given a pair of
SEQUENCE for lexical arguments is thus topic-focus. (I am ignoring here allosentences, one member is pragmatically unmarked if it serves two
the pragmatic status of non-argument constituents, in particular of the discourse functions while 1he other member serves only one of them.
verb; see Section 5.4.2 for justification of this procedure.) Given these \Vhile the marked member is positively specified for some pragmatic
assumptions, the constituent order in the Italian sentence (1.2) and the feature, the unmarked member is neutral with respect to this feature. For
position of the focus accent in the English sentence (I. I) must be example, the canonical SVO sentence She likes GER,\f.~Ns is unmarked for
characterized as marked. the feature "argument focus" while its clefted counterpart It is GERMANS
These assumptions concerning the markedness status of the syntactic that she likes is marked for this feature (see Section 5.6 for details). The
and prosodic structure of our sentences are not uncontroversial and call canonical version may be construed bolh \vith a broad (or "normal") and
for some justification. In assuming that languages have a pragmatically \Vith a narrow (or "contrastive") focus reading, 1.e. the sentence may be
unmarked (or canonical) constituent order and an unmarked focus- used to ans\ver either the question "What kind of person is she?" or a
accent position, I am by no means suggesting that sentences having these question such as "Does she like Americans or Germans?" The clefted
formal properties are "pragmatically neutral." The widespread idea of allosentence, on the other hand, only permits the narrow-focus reading.
the existence of pragmatically neutral syntax or prosody is misleading In other \.\'Ords, while the fonner can be used in the reading of the latter,
-.because it rests on the unwarranted assumption that grammatical form the latter cannot be used in one of the readings of the forrner. 12 This
• normally" has no pragmatic correlates. (A terminologically more
0
approach to pragmatic markedness entails that the marked member of a
elaborate version of this idea is that unmarked word order or accent given pair of allosentences may in turn be the unmarked member of
,position is used in discourse situations which lack "particular pre- another pair. For example, the Italian inversion construction in (1.2),
suppositions"; such statements remain vacuous as Jong as they are whose syntax is marked in comparison to its canonical counterpart (see
not accompanied by a definition of "normal presuppositions.") The (1.2') below), is unmarked \.\'ith respect to the feature "argument focus."
assumption that certain sentence forms are pragmatically neutral Herein it contrasts \Vlth 1he clefted allosentence Ela rnia .MACCHINA quc si
naturally leads to the view, which I take to be misguided, that the task e rotra "It is my CAR that broke dov.'n": the VS sentence has both a
of linguists interested in information structure is at best that of figuring broad- and a narro\.1,1 -focus reading Clike its subject-accented English
out which "special constructions" are in need of a pragmatic inter- counterpart in ( 1.1 ll, but the cleft sentence can only be construed as
pretation. Just as there are no sentences without morphosyntactic and having narrow focus .
.phonological structure, there are no sentences \Vithout infonnation In calling S\'(0) constituent order and c!ause·final focus-accent
structure. Saying that some syntactic or prosodic structures "have a position "pragmatically unmarked" in our three languages I am
special pragmatic function" while others do not is somewhat like saying refernng to the fact that this pattern has greater DISTRIBUTIONAL FREE-
that some mechanical tools have a special function \vhile others are DOM than alternative patterns and, as a corollary, that it has greater
functionally neutral. According to this logic, a scrC\\•dn\'er for example overall frequency of occurrence. I am NOT implying that alternative, i.e
would be said to have a "special functinn'" because the objects marked, patterns are somehO\\' "stylistically remarkable" or "abnormal."
manipulated with it (i.e. scre..,,.·s) must have a special shape, \vhile a For example \V1th a certain class of intransitive predicates (the so-called
hammer would be said to be functiona!ly neutral hecau~e it may be used "unaccusatt\cs" as \\'ell as impersonal s1-predicate~l \'S order in Italian 1s
to drive in various kinds of objects including n::11I~. fence roles, and if in fact ciften rercc1\'eJ b~ ncll!\ e '>pe,11-.er.., t<J he n1dre nJ.tural than S\1
18 Introduction A san1p/e analysis 19

order, \\'hen no context is provided. This native intuition is comparable to / .3.3 Analysis
that of many speakers of English who in the absence of contextual clues Let us return to our three examples. How is the semantic and pragmatic
find focus-initial prosody in such sentences as ~fy CAR broke down or Her
structure of these sentences related to the level of morphosyntax and
F.~THER died more natural than the focus-final prosody of Afy car broke
13 prosody and, more specifically, what is the role of information stnlcture
DOlf'N or Her father DIED. Such intuitions result from the fact that in the shaping of these utterances? ·
certain propositional contents are most frequently expressed under
In the English sentence My CAR broke dou•n, the semantic role of THEME
certain pragmatic circumstances. hence tend to be associated in the minds
is associated with the syntactic relation of SUBJECT in the subject phi~
of speakers with those grammatical structures which are appropriate for
my car. Within this phrase, the determiner my plays the semantic role· ~f
those circumstances. A structure like Her fall1er died is perhaps more
the ~ossess~r~ ~nd the ~oun car that of the possessed entity. The subj~i
often used to announce the death of a previously unmentioned individual
NP is the 1nit1al constuuent in an intransitive sentence, resulting in a
(subject accentuation) than as a comment in a conversation in which the
seq~ence of the form NP-V. As for the information structure of(l.l), wC
individual is already the topic under discussion (predicate accentuation).
notice that the linguistic expression designating the topic of the utterance
They have no bearing on the status of SV(O) constituent order or clause-
(the speaker) is the initial pronominal element my. The pragmatic relatio~
final focus accentuation as unmarked.
It is a distributional fact that in Italian all predicates permit SV{O) of TOPI.c is th~s ~apped with the non-phrasal syntactic category of
det~r:n1n~r, \l.'h1ch 1s not an argument of the main predicate, and wh.ose
order while only a restricted set of predicates permit the alternative VS
pos1t1on 1s fixed within its phrasal domain. The sentence accent falls on
order in such sentence-focus structures as (l.2). Similarly, focus-final
the subj~ct nou~ car, marking the designatum of this noun as having the
prosody in English is permitted with all predicates, while focus-initial
prosody in sentence-focus structures such as (I.\) is permitted only with a P_ragmat1c rel~uon of Focus rather than topic to the proposition and-
relatively small number of. mostly intransitive, predicates (see Lambrecht g1ven the particular. focus structure of this sentence-indire ctly marking
1987a and forthcoming). In other words, in both languages there are all subse.quent constituents as part of the focus domain (see Section 5.6.2). ·l
many predicates which require the subject to be a preverbal topic and the Thus in (I. I) both the semantic role of theme and the pragmatic role of
object a postverbal focus constituent, but there are no predicates which focus are associated \.11ith the grammatical role of subject in a constit~eni
require the reverse situation. It is in this distributional sense that I call of ~~pe NP, and this subject NP occupies its unmarked preverbal
position. Moreover this NP is also the only nominal constituent in the
focus-final prosody and SV(O) order unmarked in the three languages.
sentence. Hov.·ever the position of the focus accent on the noun car is
Cognitively, the marked pattern receives its value not from some inherent
marked. Instead of being coded syntactically, the information structu.TC
feature specification but from the fact that it is perceived as a deviation
of the utterance is coded prosodically. It follows that the SYNTACTIC
from the unmarked pattern. This. I believe, is the reason why traditional
pattern in (I.I) is 1'.'0T directly motivated by the pragmatics of the
grammar has characterized VS structures in languages in which the
utleran~e. Rather the sequence NP-V is an independently motivated
subject normally precedes the predicate as "inversions," i.e. as deviations
syntac_tic stru_cture in _the language. Indeed, the same syntactic sequence,
from what is perceived to be the norm. By analogy. in languages with
but \v1th a different intonation contour, could be used under different
focus-final prosody one might call focus-initial sentences like (I.I)
pragmatic conditions, as V.'hen I ask "What happened to your CAR?" and
"prosodic inversions" (see Section 5.6.2). This approach to markedness
you ansv.·er, \Vith perhaps some\vhat unnatural explicitness:
in word order and prosody is based on the cla$sical notion of markedness
elaborated by Trubetzkoy and Jakob-:,on (sec e.g. Waugh 1982). It is (] ]') tvfy car broke DOWN

different from the \videspread u~agc 1n \\h1ch "unn1Jrked" designates any

pattern which is perceived to be 1nore nc1tural lhJ.n some alternative (I. I') conveys a pragmatically different piece of infonnation, in which the
pattern in a given discourse context referent of the noun phrase niy car is already established as a topic under
d1scuss1on. Sentences such as (I J'), in v.·hich the domain of the .. new
20 br1roduc1r'on A sample analysis 21

information" ex.tends over the predicate to the exclusion of the subject (1.2') Lamia macchina si C ROTTA.
wilt be referred to as "predicate-focus structures" (Chapter 5) and the
pragmatic articulation of the proposition will be called the "topic- Sentence (l.2') has the the canonical (unmarked) form, with the topical
comment articulation" {Chapter 4). The syntactic structure of ( 1.1) is thus subjecl NP in initial position and the focus accent on the predicate. 15
neutral with respect to the expression of information structure. What There is another interesting difference between the Italian and the
distinguishes (1.1) from (1.1') is not its syntax but its PR.osoo1c structure, English example with respect to the formal manifestation of the
and this prosodic structure is marked. Symbolizing the accented information structure of the proposition, having to do with the syntactic
constituent with the letter Z and the non-accented part of the sentence status of the pronoun mi. Like the English possessive my, mi has the
with the Jetter A (a simple representation introduced by Bally, l 932:53fD, pragmatic role of topic. But in Italian this topic is a personal pronoun
we can represent the prosodic sequencing in the English sentence (I. I) as bound to the verb rather than a determiner bound to a noun. By
Z-A (ignoring the role of the detenniner) and that in (1.1 ') as A-Z. replacing the ordinary NP-internal possessive relation (as in la mia
Let us now consider the Italian sentence Mi si e rotra la MACCHtNA in macchina .. my car" in (1.2')) by a relation between a personal pronoun
(1.2). 14 Concerning the mapping relation between SYNTAX and SEMANTICS, and a non-possessive NP, Italian is able to maintain the topic constituent
we notice that it differs from that in (I.I) in one interesting respect: the in its unmarked initial position rather than have it follow the verb. Notice
posses.sive relation between the car and its owner is left une:'ipressed that this expression of the "topic-first principle" occurs again at the
within the subject NP. Instead, this relation is indirectly conveyed via the expense of the unmarked. canonical syntax
rCtation between the clause-initial dative pronoun mi and the lexical NP To summarize, even though the Italian example (l.2) resembles the
/a, m~cchina. The semantic role of the pronoun mi is perhaps best English ( 1.1) in that the theme, the subject, and the focus are all
described as that of an "experiencer" since the event is described as combined in the same NP constituent, the manifestation of the
happening to the speaker. In spite of the presence of the dative pronoun information structure of the proposition in the form of the sentence is
mi, the sentence is intransitive in that it contains neither a direct nor an radically different in the two languages. In Italian it is not the unmarked
.indirect object (the reflexive si is not an object argument but a "middle syntactic SV sequence but the unmarked prosodic sequence that is
vOicen marker). In Italian, as in English, the semantic role of theme is maintained, with a topic constituent as the initial and the constituent
expressed as the subject NP of an intransitive predicate. carrying the focus accent as the final element in the clause. The Italian
· More interesting within the present argument is the radical difference sentence contains nvo argument constituents, v"hile its English counter-
bCiween English and Italian with respect to the way in which the part contains only one. 16 Using BaJ/y·"s schema£ic representation, \Ve may
INFORMATION STRUCTURE of the proposition is reflected in the SYNTAX of symbolize the prosodic sequence in the Italian sentence as A-Z.
the sentence which expresses it. In Italian the canonical SV(O) constituent In the discussion of the Italian sentence I have assumed, with
sequence in which the subject NP is a topic and the object part of the traditional grammar, that the postverbal constituent la macchina is the
focus is changed to fit the pragmatic requirements of the utterance. by subject of the sentence. albeit an "inverted" one. This has become a
inverting the order of the subject \Vi th respect to the verb. By placing the controversial assumption 1n generative syntax. One can argue that fa
subject after the verb, Italian respects the unmarked prosodic sequence in macchina is in fact not a full-ncdged sub1ect because It shares certain
which the constituent carrying the main sentence accent occupies linal formal properties v.·ith direct objects (in particular its position), an idea
position. We can see that a syntactic adjustment has taken place by \vhich has been much discussed in recent years in connection with the so-
comparing sentence (1.2) with the corresponding sentence in ( 1.2'), in called "unaccusative hypothesis., (Perlmutter 1978, Burzio 1981. etc).
which (as in the English sentence (I. I') above) the car is the lo pie of a According to this hypothesis. the post verbal subject in ( 1.2), as in other
statement intended to increase the addressee's i:n('•\~'ledge about the car, VS constructions containing certain intransitive prcdicZ1tes, \vould in fact
not a previously unmentioned entity depicted a:; participating in a not be a subject hut an oh1ect at a deeper level of analysis. I consider it a
reported event: major ad\'antage of the approach to grammatical analysis advocated in
 ,.
A san1ple analysis 23
the present study that the postulau on of an abstract level sentence such as Est en panne nia 1·01TURE, which would have the focus
of in
represen tation at \\'hich the postvcrb al subject is an object is made the right place, would be syntactic ally ill-forme d). 19
superflu ous. By adopting a fran1c\vork in \Vhtch the categorie s
of In the bi-clausa l structure thus created, the function of the first clause
informa tion structure are recogniz ed as gran1matical categories on a par J'ai ma 1•oiture, which appears to express a semantic ally indepen dent
\.\'ith the categorie s of syntax ;_ind semantic s. we are in a position proposit ion, is in fact not to make the (tautolog ical) assertion that
to the
preserve the tradition al insight that the NP in ( 1.2) ts indeed the subject
of speaker "has her car." Rather the sole function of the avoir-ciause is
the sentence , albeit not a canonica l one In such a framewo rk, we tO
can pragmat ically POSE the referent of the NP in the discours e in such a way
account for the differenc e bet\vecn the canonica l and the inverted
that its lexical manifest ation does not coincide with the gramma tical role
structure in terms of different niapping s of prag1natic and syntactic
of subject. The subject position of this clause is occupied by the first
relations . An inverted subject can then s1n1p!y be defined as one with
the person subject pronoun je which, like the Italian dative mi, has
marked informat ion struc1ure status of a focus consutue nt. 17 the
pragmat ic role of topic. The semantic role of this topic argumen t may
Let us OO\.\' look at the French exan1ple 111 (I 3). Here the situation 1s be
describe d as locative (see Lambrec ht 1988b). The semantic relation of
rather complex . In the sentence J"oi 111u i Olll'Rf 9ui est en P.4SSE, both the
the referent of the lexical NP ma l'Oiture to the predicat e est en panne
semantic structure and the syntactic structure are adjusted 111 order is
to expresse d in the qui-claus e, whose pronomi nal subject qui is anaphor ic
accomm odate the informat ion structure of the proposit ion. To apprecia to
te the object NP in the precedin g clause. This qui-clause, even though it
the extent of the adjustn1 ent it is useful to con1part! (1.3) "'ith has
the the internal structure of a relative clause, differs in crucial ways both
canonica l SV structure in (l.3'): 1 ~
from the restrictiv e and the appositiv e relative clause type. Not only
could the antecede nt NP in the a~·oir-clause be a proper name, thereby
(1.3') Ma voiture est en PANNF.
excludin g the modifyin g function associate d \'>1 ith the restrictiv e relative,
but the informat ion expresse d in the qui-claus e in (1.3) is neither
Due to a powerfu l gramn1a tical constrai nt against the co-mapp ing of
the (pragma tically) presuppo sed, as in the restrictiv e relative, nor parenthe
pragmat ic relation focus and the gramma tical relation subject -
(see tical, as in the appositiv e relative. In fact it is the predicat e of the qui-
Lambrec ht ! 984a and l 986b: Ch. 6), spoken French rnakes abundan t
use clause, not of the i:n·oir-clause, which expresse s the main assertion
of such CLEFTED construc tions as in ( 1.3) to avoid focus-ini tial
SV expn;sse d by the sentence (see Lambrec ht l 98Sa for further discussio
structure s. Constru ctions such as ( 1.3) n1ay be called clefts because n)_
the Thus in the French sentence both gramma tical relations and syntactic
proposit ional meaning expresse d by the l\vo-clau se sequence is identical
constitu ent structur e are accon1m odated to fit an indepen dently
to that expresse d in the canonica l ( l .3'), \VLlh no differenc e in truth
motivate d informa tion structure , at the price of complex formal
conditio ns. In spoken French, a canonica l sentence such as /t.fa IDtTt'RE
a<ljustn1e111s. \Vhile in English the proposit ion is expresse d with
est enpanne , with the accented NP in prevcrha l subje•.:t position , \\'ould be one
predicat or and one argun1en t, Lind in Italian with one predicat or and two
unaccep table because prosodic atly i!l-fL)rn1cU In 1 l 3). \\'h1ch is used
argun1en ts, in French it is expresse d \.\'ith two predicat ors and three
instead of this ill-forme d sequence . the 1..·(\n-.,t1tucnt carrying the focu:';
argun1cn ts (l\\O of which obligatn rily designat e the same entity)_ Since
accent, VOJTURE, docs not appear a~ the suhJcCt NP uf an 1ntr..1nsit the French sentence is synonym ous v.:ith the nionucla usal English and
ive
clause, as in English and Italian. hut ..1::- th1..' :.ynl<10.:t1c 1)1HH-r 1.if the
\·erb Italian sentence s, and since the French c<inonical monocla usal version
avoir, in a clause of its 0\\'11. -rhe prat:1n.111L· tu11L·t1un L1 f thi.:: clefting in
t 1.J'J is syntactic al!y anU semantic ally \.\'Cli-fom1ed, this prolifera
structure is to create an addition al p<1:-!\'L'rh.d ,1r1_..'lln1rnt p\i-.,111nn in v.·h11.::h tion of
argun1cn ts can only be expL.uneU by the requ1re1nents of informa tion
the focus NP may appear. prt::\ent111t'. 1t tr .. in ,•Le,LHr1ng 111 ::-<:nlcnl'(:-1n111al structure . \Ve can :.yn1bolile the sequence in the French sentence as A-Z-
position . The structure 111 ( l ..;l thu~ 1n,d.v, ll[1 1-.. 1lh l.ir the un<1cl..'ep A-Z, 1.e. as a gra1n111atil'~d cornpron usc betv. een the English sequence Z-
t-
abiiity of subject-a ccented SV ~lrtJL.lLHc' "I tf1,.- !·11;;l1~h l}'PL' an.J fL1r 1\ :..inU the ltJ!1an sc4ur.>11l'e A-Z. 1·1ie accented NP rna i·oirure is final
the in its
ungramm aticality of accent-fi n,d \ S ,111klli:L °' ··I 1)11..· lt-d1,in t~pc (a
VS
 24 /11/roductlon Information structure and syntax 25

own clause. as in Italian, but it precedes the main predicate .. break constituent order in this language, we may say that the formal structure
down:· as it does in English. of (1.2) is MOTIVATED by the pragmatic function of the utterance.
As for the French structure in (1.3), syntax and information structure
both win and lose in the competition. The constituent order in the French
1.1.4 Summary sentence being strongly grammatically controlled, the language does not
The purpose of the preceding analysis was to deFend the view that the freely permit subject-verb inversion or other types of word·order
grammatical patterns illustrated in our three examples can be understood variation found in languages with pragmatically controlled word order.
in all their complexity only by interpreting them as the result of multiple Nevertheless the global structure of sentence {l.3) DIRELnY reflects its
language-specific dependencies between the various components of pragmatic function. Even more so than in Italian, the syntactic structure
grammar. semantics, infonnation structure, morphosyntax, and pro- of the French sentence may be said to be pragmatically MOTIVATED, since
sody. As the examples show, the interaction between the components this cleft construction involving the verb avoir has as its unique function
may lead to quite different formal results, even in languages as closely to express a single proposition in bi-clausal form under the specific
related as English, French and Italian. In the case of the English example pragmatic circumstances discussed above. As shown in Lambrecht 1986b
(1.1). information structure "loses out" on the syntactic level. However (Section 7.2.2), certain formal and semantic properties of the construc-
this loss is compensated for by the fact that in English the sentence accent tion (such as the use of the verb a\'oir in cooccurrence with a possessive
can in principle "move" from right to left, allowing for prosodic focus object NP) can be made sense of only if reference is made to its pragmatic
marking in any position in the sentence. Because of the importance of function. By using grammatical constructions of the clefting type, spoken
sentence accentuation in English, syntactic expression of information French achieves several things at once. It substitutes structures of a
structure is often unnecessary in this language, or, phrased differently, certain pragmatically preferred type for the pragmatically unacceptable
sentence accentuation makes up for the rigid word order constraints of SV(O) sequence; it preserves its syntactically controlled basic word order
English grammar. Sentence prosody is thus pragmatically highly without violating the information-structure constraint which maps topic
motivated in English. Typologically, English presents itself as an with subject and focus with object; and it avoids violation of its strict
example of extreme ..subject prominence" (Li & Thompson 1976), i.e. oxytonic accent pattern. The "mixed strategy" of cleft formation allows
as a language in which a great variety of semantic and pragmatic the language to have its cake and eat it too. It represents one of the
functions may be associated with the invariant syntactic function of specific solutions in French to the competition between syntax and
subject and in which word order is to a large extent grammatically and pragmatics.
not pragmatically controlled. 20
The competition of grammatical factors has difTerent consequences in
the Italian example (1.2). Here, it is syntax that "yields" in the
competition between formal structure and information structure: the 1.4 lnfonnation structure and syntax
canonical constituent order is altered to accommodate the requirements
of discourse. Just as English is reluctant to tolerate a violation of its In the previous section I illustrated some of the implications of a view of
canonical SV order, Italian is reluctant to tolerate a violation of the natural language in \\'hich information structure is seen as a component
information·structure constraint that places a focus argument in post- of sentence grammar on a par \\'ith morphosyntax, semantics, and
verbal position. 21 In Italian, word order is thus to a greater extent prosody and in \\'hich these components are seen as interacting with each
controlled by information structure than in English, even though the other in various language-specific v,iays. Let me no\v address again the
syntax of Italian is far more rigid in this respect than the syntax of so- theoretical issue raised earlier concerning the place of information
callcd free word order languages, like Russian or Latin. Given our structure 1n grammar and in particular of the relationship between
assumption that the SV(O) pattern In (J _2') illustrates the unmarked information structure and syntax
 Information slructure and syritax 21

1.4.1 Au1onon1y W?rsus niotivatio" in gra111111ar clearly are determined by the structural properties of the individual
languages, which in turn follow from, or are at least ~nsistent with,
To avoid a certain misunderstanding which radical .. functionalist" views
universal typological principles and perhaps universal constraints on
of syntax have allowed to arise, I want to emphasize from the outset that
possible syntactic structures. It is in this sense that the syntactic structure
I would find it misleading to say that the con1municative requirements of
discourse directly determine not only the content but also the form of
utterances and that information structure therefore can in some sense
of sentences may be said to be "autonomous."
Having said this, I wish to emphasize that I find it equally misleading j
I
to say that information structure plays no role in the formal organization
EXPLAIN the structure of sentences. Taken to its extreme, such a view
of sentences. In this view, which has informed various versions of
would imply that sentences expressing similar propositional contents in
generative grammar, information structure- to the extent that it is
similar communicative situations must have similar forms across
acknowledged as a component of grammar-plays the role of an
languages, an idea which is blatantly contradicted by our three
interpretive mechanism which checks fully formed syntactic structures
examples. It is true that our three sentences have an important feature
for their appropriateness in given utterance contexts. Such an interpretive
in common, i.e. they depart in certain systematic ways from their
view of inrormation structure seems difficult to reconcile with some of the
canonical counterparts, but the form which this departure takes does not
follow in any predictable way from the function they serve. Moreover, facts discussed in the preceding section, for example the facts of focus~
since the number of states of affairs which one can talk about is infinite accent assignment. It has often been observed, both by generative and by
but since the number of different structures which the grammar of a functional linguists, that focus prosody may not only have an influence
language makes available to speakers is severely limited, there can be no on the pragmatic interpretation of sentences (as in the difference between
one-to-one relationship between communicative requirements or inten- example (I. I) and (I. I') above), but also on certain phenomena
tions and the grammatical forms of sentences. There can only be a traditionally called "semantic," such as the interpretation of anaphoric
mapping from types of situations to preestablished formal types. relations between nouns and pronouns (see e.g. Kuno 1972, Akmajian
Speakers do not create new structures to express new meanings. They 1973, Bolinger 1979, and the useful swnrnary in Van Valin 1990b). 23 The
make creative use of existing structures in accordance with their facts of focus-accent assignment therefore strongly suggest interaction if
not between syntax and phonology, at least between the "interpretive"
communicative intentions.
components of phonology and semantics.
Within the "competing-motivations" view illustrated with our model
sentences it would therefore be inconsistent to claim ·that information To accommodate the facts of focus prosody within a modular
structure by itself determines the differences in formal structure between approach to grammar, the claim has been made (see Culicover &
sentences. To go back to our examples, information structure cannot
Rochemont 1983:123fT, Horvath 1986:94ff) that focus accentuation is in :I
fact not a prosodic feature but an abstract SYNTACTIC feature which is
explain, for example, why in expressing the propositional content of (I. I)
assigned at the level of S-structure and which has no syntactic realization 'I
~
only Italian may resort to the inversion construction in order to mark the
whatsoever. This S-structure feature is then translated into stress
non-topical status of the subject, while the analogous English and French
assignment in "Phonetic Form" and semantically interpreted in
VS sentences •Broke don·n tny CAR and *Est en panne n1a f·o1ruRE are
ungrammatical, even tbough clause-final position is the unmarked focus- "L?gical Form" and "Discourse Grammar." As I understand it, the
main purpose of such interpretations of focus prosody is to reconcile the
argument position in all three languages. Nor can information structure
facts of focus with the notion of modularity and of the autonomy of
explain the internal syntax of the relative clause in the French example or
syntax. They seem to do little to elucidate the nature of the relationship
the grammatical mechanism which al!O\\'~ this clause to enter into a cleft
betv.•een prosody and syntax. They constitute examples of the recourse to
construction with the matrix. clause. If it could, it \vould be dillicult to
"placeholders" in generative theories as a way of preserving the internal
understand why in Italian and English such a cleft construction \vould be
inappropriate if not ungrammatical~.:- Such d1ffer.::nccs in grammaticality consi.stency of_a model in the face of strong empirical evidence against
certain predictions made by it. The use of such placeholders is cogently
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .___.._............~,...,,,,.,,...,,,,.,=====
ln.formalion structure and syntax 29
~'. 28 Introduction

criticized by Woodbury (1987:688fl), who observes that abstract grammatical structures arise diachronically under pressure from
placeholder morphemes were invoked at various points in the history of infonnation-structure constraints, which I take to be universal. It is the
generative grammar. both in phonology and in syntax, whenever some pressure to distinguish two types of pragmatically structured propositions
gramlilatical phenomenon threatened to invalidate given views of the that has led to manifestation of the fonnal contrasts discussed in the
organization of grammar. According to Woodbury, "a placeholder can previous section. What would be needed to account for the mapping of
be seen as a symptom that the formal organization of grammar has been infonnation structure and sentence form is something which to my
<listortedh (ibid.). knowledge does not exist, and according to some cannot exist: a theory of
Interpretive views of information structure seem even more difficult to GRAMMATICALIZATJON with predictive power. Such a theory could show us
defend in the face of the Italian and especially the French facts of focus how universal discourse requirements get expressed in grammatical form
marling. As we saw before, the sentence-focus structure which in English in accordance with the structural and typological properties of individual
is expressed prosodically, via changes in accent pJacement, is expressed in grammars. 24 But even though we cannot rely on such a theory, we can
French syntactically, via the combination of two canonical clauses in nevertheless rely on the notion that the complex relationship between
each of which the focus accent has its unmarked clause-final position. In form and function is not arbitrary but motivated within the grammatical
the French example, the focus reading is clearly a property of the system of individual languages. While the form of constructions Jike the
complex grammatical construction itself, not an interpretation imposed French avoir-cleft cannot be PREDICTED on the basis of communicative
on an independently motivated syntactic configuration. While the lexical needs, this form can be shown to be MOTIVATED within the grammatical
and phrasal elements which make up the construction are familiar from system of the language. Given the formal constraints of French grammar,
other parts of the grammar, and while these elements are for the most and given the need to express a universal pragmatic category, it "makes
part assembled according to general combinatorial principles, the global sense" that the avoir-construction looks the way it does in spoken
tonstruction resulting from the combination of these elements is unique French. The notion of the PRAGMATIC MOTIVATION of grammatical form is
and serves a unique function in discourse. In this sense, syntactic form one of the major theoretical concepts underlying the present study. 25
may be said to correlate directly with discourse function, hence cannot be
fully understood without reference to it. To use an analogy, claiming
total independence of grammatical form from discourse function is like
1 .4.2 The functional undcrspecifica1io11 of syntac1ic structures
claiming independence of the fonn of the automobile from its locomotive
purpose on the grounds that the form is determined by the laws of Having made explicit the notion that the grammatical form of sentences
mechanics only and not by the desire to get from one place to another. is motivated by the requirements of information structure, I must
While the reasoning is sound, it obscures the crucial fact that there would introduce a caveat concerning the extent to which SYNTACTIC STRUCTURE
be no automobile, and hence no form, if people didn't have the need to per se may be said to be motivated. Let us first remember the fact,
travel. Since the automobile owes its existence to this need, and since illustrated for example in the two sentences in (I. I) and (l. l '), that
existence is a logicaJ prerequisite lo form, the existence of a logical link syntactic pallerns, such as the NP-VP pa!lern of English, may be
between form and function is undeniable. underdetermined \vith respect to their discourse function and that
Even though information-structure analysis allo\vs us to recognize the functional differences may be expressed by non-syntactic means. That
pragmatic motivation of grammatical form, it must be ackno,vledged syntactic structures may serve more than one function is well known. As
that it docs not account for the process \Vhereby the constraints of a useful example of the functional diversity of unmarked syntactic
information structure are translated into, or mapped onto, grammatical patterns one might mention the \Vell·known fact that the canonical SV
structure, resulting in the CREATION of such constructions as the French sequence of English (and other languages) may not only be used for
avoir~left. While it is true, as I said earlier, that speakers do not create declarative hut also for interroga1ive sentences, by providing the sentence
new structures to express new meanings, it is nevertheless the case that with the appropriate non.declarative intonation contour (compare He is
....................... .......
'1uruu1u·11011
--------~~~-~-
Information structure and syntax 31

exclamations (Boy, is he hungry!). Here again, intonation crucially

hungry \\.'ith He i:S hungry?). Herein the SV pattern contrasts with the
interacts with syntax to produce the narrow fit between communicative
auxiliaJ)·-inversion pattern (Is he hungrv?), which cannot be used in
declarative contexts, no matter how th~ intonation is modified, and function and grammatical form.
which must therefore be characterized as marked for the feature "non- Another case in point is the so-called topicalization construction
declarative." known from English and other languages, whose pragmatically m8.riced
An example in French of a highly general syntactic pattern which is character has often been noticed. The term "topicalization" is commonly
used in the formation of various functionally divergent sentence used with reference to syntactic constructions in which an object noun
constructions is the [pro· V (XP)] pattern which is discussed in phrase whose canonical position is after the verb appears in clauso-U:ii~
Lambrecht 1986b (Chapter 6) under the label "preferred-clause position before the subject (or 4.irectly before the verb in languages witJI,
construction." Although this syntactic pattern typically serves to code V-2 order, in which case the subject appears in the position of the object).
propositions Y•ith a topic-comment relation between the subject and the As the name suggests, the discourse function of the "topicalized ..
predicate, it may also be used for the expression of propositions involving sentence is assumed to be different from that of its canonical counterpart,
no such relation. Compare the [pro-V] structure // pieure ··He is crying," the object noun phrase now being a topic (rather than being part of the
in which the predicate ex.presses a property of the subject and \Vhich can focus domain). What is often not recognized is that this syntactic type
therefore function as a topic-con1ment sequence, with the structurally serves in fact two very different discourse functions. As Stempel (1981)
identical JI pleut "It is raining," where no topic-comment relation is and Prince (1981b) have demonstrated for French and English respec-
possible because the pronoun has no referring function, the sentence tively, the "topicalized" phrase may stand either in a topic relation or in a
expressing instead a kind of existential statement: "there is rain falling." focus relation to the proposition expressed by the sentence. (In terms of
Or compare the two (pro· V NP] structures Ii bait une goutte "He is the framework developed in Chapter 5 below, the first has "predicate-
drinking a drop" and Ii tombe une goutte "A drop is falling": the first focus" and the second "argument-focus" structure.) And this clear
expresses a proposition with topic-comment articulation, in which the difference in pragmatic function correlates with an equally clear pro·
predicate is construed as a comment about the referent of the subject; the sodic difference. At the level of syntax, however, the difference.is nOt
second has an "eventive" articulation, in which the pronoun has no marked. 26
referent and in which the domain of the focus extends over the entire This lack of overt syntactic differentiation of the topic-focus contrast
proposition (see the discussion of "thetic" propositions in Section 4.2.2). in sentences with topicalized NPs does not entail that no form-function
The example of this preferred-clause construction shows that such correlation can be established for such sentences. It simply confirms the
fundamental semantic-pragmatic distinctions as the one between subject- observation that syntax is not the only formal level at which information
predicate statements and ex.istential statements (or, in the terms of the structure is coded. What syntax does not code, prosody does, and what is
present study, between predicate·focus and sentence-focus structures) not coded by prosody may be expressed by morphology or the lexicon.
may go unexpressed at the level of morphosyntax. \Vi thin the interactive view of grammar suggested here, such facts do not
The possibility of multiple form-function correspondences is not come as a surprise. lo fact it would be surprising if the opposite were true,
restricted to highly general, semantically or pragmatically unmarked i.e. if all syntactic patterns were uniquely paired with specific discourse
syntactic patterns. Even with marked patterns there 1s oflen no one-to- functions. Since the morphosyntactic resources of a language are limited,
one relationship between a specific s' NTACTtC forn1 and a specific and since the number of communicative distinctions is potentially
communicative function. A useful eAan1ple, discussed Dy Akmajian infinite, economy of form is a logical necessity in the expression .of
(1984), is the already-mentioned auxiliary-1nvers1on pattern in English. functional differences in natural language. In the case of the OSV (OVS)
While this pattern is functionally rt:stricted in th.:1.t it cannot be used to pattern, the reason for its dual function seems fairly clear: since sentence·
express simple statements, it may nevertheless be used in t\vO clearly initial position is a cognitively highly prominent position, it is ideally
different functions, i.e. to e.\press LJucstions (/s he lno1gry?) and suited to express the contrast between unmarked and marked structures.
 32 l1ttroduction Information structure and syntax 33

Marted topics and marked foci nacurally compete for this cognitively Type Schema." Jn the case of English, ir is the presence and the position
-privileaed position.27 of the auxiliary which determine a significant set of formal sentence
types, in conjunction with a set of intonation features which directly
interact with the syntax. Although Akmajian acknowledges the
J.4J Sentence types and the notion of gran1rnatical construction
theoretical possibility of a one-to-one form-function fit (for example in
We must draw the conclusion that there is often no one-to-one such constructions as Down with X! or Off \Vith X's Yl, which he calls
•correspondence between syntactic form and discourse function, even in "highly marked"), his main claim is "that something along the lines of
t !tfiC case of non--canonical sentence patterns. 2 ~ The general tendency the Formal Sentence-Type Schema, based on a small and restricted set of
f' '8.tross languages seems to be that the fit between form and discourse formal parameters, provides the input from formal grammar to the
·llint:tion involves multiple correspondences between the various pragmatics" and that across languages "the task will be to specify a set of
'6omponents of grammar. In what sense, then, may specific syntactic correspondence principles that relate certain ronnal sentence-types and
·Cori.figurations, such as the topicalization construction, the "dative" certain pragmatic functions" (p. 21).
'tiiiistruction, the passive construction, as well as other, less commonly Akmajian's theoretical stance may be characterized as follows. Given
-analyzed, patterns be considered to be pragmatically motivated if no the fact that there are clear cases of one-to-many form-function
unique discourse function may be assigned to them? What exactly is the correspondences, i.e. given the fact that in many cases a single syntactic
nature of the relationship between pragmatic function and syntactic structure serves more than one pragmatic function, let us assume a
form? syntactic component which is as simple and general as possible and let
One clearly articulated though tentative answer to this question is this componenl generate a small set of highly general sentence types. Let
-provided by Akmajian in his earlier-mentioned paper "Sentence types us furthermore allow this component to interact \Vith certain aspects of
·aild the form-function fit" (1984). Discussing the syntactic and phonology, and let a sophisticated pragmatic component, in the form ofa
pragmatic structure of such sentences as What, me worry? or Him 1-vear universal theory of speech acts, provide principles of pragmatic
a iuxedo?!, which he ca1ls .. Mad Magazine sentences," Akmajian notices interpretation which \viii rule out undesirable surface configurations.
a number of fonnal similarities between such sentences and the class or Any formal phenomena which are not accounted for in this way will have
imperative sentences. On the basis of these similarities, he argues that to be specified as a set (small, \Ve hope) of exceptions to the general
imperatives and Mad Magazine sentences may in fact be generated by the system, e.g. in the form of special syntactic rules.
same, highly general, phrase structure rule, "with the proviso that Although there is an undeniable theoretical appeal in this idea of a
pragmatic principles for the use of imperatives will in fact limit mapping function bet\veen highly general syntactic types and equally
imperatives to a subset of the structures in question" (p. 14). Akmajian general pragmatic principles, I believe that this approach does not
tOncludes that neither the Mad Magazine sentence type nor in fact provide a realistic picture of the relationship bet\veen form and function
iniperative sentences have a special status in a syntactic theory. Rather in natural language.~ 9 Even though it is true that a great many syntactic
such notions as ..imperative," "interrogative," "assertive," and so on are patterns cannot be uniquely paired \\'ith specific u~es. I believe that the
to be determined in a theory of speech acts, i.e. in Che pragmatic number of "highly marked" and idiosyncratic fonn-meaning-use
component of language. correspondences in natural languages is much greater than assumed
Ak:majian then raises the question whether "particular clusterings of in most current approaches. With Fillmore and other proponents of
:rormal properties should be singled out as constituting significant Construction Grammar. I take 1t to be impossible to draw a dividing
sentence-types" (p. 18). He suggests that the ans\ver to this question is line on principled grounds bet\\·een idiosyncratic (or "idiomatic") anJ
'*yes" and that the notion "sentence-type" indeed has theoretical status in general or ('"regular") types of construction~ One of the most important
formal grammar. However, according to Akma;ian such sentence-types tenets of Construction Grammar is the belief that the distinction
belong to a highly general, and perhaps unive-rsal, "Formal Sentence- bet\veen "idiomattcity" and "regularity·· fsyntact1c generativity, semantic
 1111IVUU1.-llVll !nfor1na1ion structure and .syntax 35
compositionality) has been overemphasized in generative grammar and typically to the syntactic category "sentence." However. information·
that an adequate linguistic theory must be able to account equally well structure contrasts may in principle be expressed within any syntactic
for idiomatic as for regular aspects of a grammar (see Lambrecht 1984b domain which expresses a predicate-argument relation, for example
and in particular Fillmore, Kay, & O'Connor 1988). within the noun phrase (see the information-structure contrast betW~n
According to Construction Gramn1ar, linguistic theory can bridge the 1ny CAR and MY car or French ma vo1TURE and ma voilure ti MOI).
gap between idiomaticity and regularity by recognizing as the Among sentence-level constructions it is further necessary to
fundamental unit of grammar the GRAMMATICAL CONSTRUCTION. JO A distinguish three major types. The first is represented by constructions
grammatical construction is defined as "any syntactic patlern which is whose purpose is to express varieties of SPEAKERS' ATTITUDE.s (such as the
assigned one or more conventional functions in a language, together with let-alone construction analyzed in Fillmore, Kay, & O'Connor 1988 or
whatever is linguistically conventionalized about its contribution to the Akmajian's aforementioned "Mad Magazine" type). These constructions
meaning or the use of structures containing it" (Fillmore 1988:36). In are often categorized as "idiomatic" and do not necessarily have analogs
Construction Grammar. complex grammatical constructions arc not across languages. The second type is made up of ronstructions expressing
viewed as being derived fron1 more general or simpler structures via SPEECH-ACT DIFFERENCES (such as interrogative vs. imperative vs.
generative rules of the type familiar from phrase structure grammars, declarative sentences). Unlike the first type, this type is entirely
even though in some cases the principles for the combination of smaller productive and can be easily identified across languages. 32 The third
constructions into more complex ones may be fairly general. Rather they type, which is the one I am concerned with in this book, comprises
are seen as ready-made templates used as such by the speakers of a constructions whose function is to express differences in INFORMATION
language. STRUCTURE proper, i.e. which, for a given proposition and a given speech-
In this book, I will assume the existence, and theoretical imporlance, of act type, express differences in the respective scope of the presupposition
a large number of more or Jess specific form-meaning-use correspond- and the assertion, differences in topic-focus structure, or differences in
ences in the grammars of individual languages, expressed in the form of a the cognitive status of the referents of argument expressions. Llke the
variety of more or less complex grammatical constructions. Grammatical second type, this type is entirely productive and identifiable across
constructions may appear at different levels, as lexical, phrasal, clausal, languages. As I 1nentioned earlier (Section I. I), these constructions come
or sentential structures. At the end of this book, I will argue lhat in pairs of ALLOSENTENCES, i.e. semantically equivalent but formally and
grammatical constructions can also be defined al the level of prosody. pragmatically divergent surface manifestations of given propositions. The
They may be highly productive, in the ~ense that their structural pragmatic contrasts in question are always Interpreted against the
descriptions may provide a relatively large number of positions which background of available, but unused, grammatical alternatives.
may be freely filled with smaller constructions or large classes of lexical
items. Or they may be more constrained, 1n the sense that the number of
open phrasal or lexical positions v.;hich they provide and of the
expressions capable of filling these positions 1s relatively small. As a
general rule, the fewer substitutions a con~truction permits \Vithin the
31
structural positions it provides, the more it i" pt:rce1vcd as idiomatic.
For the purposes of Lhe stud;.' nf 1nfurinc1tion structure, it 1s useful to
distinguish grammatical constructtcin.., ,11 \he phras::il level from
ronstructions whose syntactic J(•rna111 t\ the clau'.c or sentence. Since
information structure. hc1-; \(1 du 11 llh the pr,1grn;it1c structuring of
propositions in d1scour\c, I .in1 111.11111\· LllllL·crncJ 1n thVi book \vllh
constructions capable ut c.\prc..,..,111~ pr"r'1"1l1t111'i. and these belong
 The universe of discourse 31
(b) the TEXT-INTERNAL WORLD, which comprises LINGUISTIC EXPRESSIONS
(words, phrases, sentences) and their MEANINGS.

2 Information The text-internal world is the abstract world of linguistic REPRESEN-

TATIONS created in the minds of the interlocutors in the process of
communication. It is the manipulation of such representations that
allows for the conveying of INFORMATION. In accordance with the
definition provided in Section 1.1. the INFORMATION-STRUCTURE compo-
In this and the next three chapters I will analyze the concepts which I nent of language necessarily involves both text worlds, since it matches
consider fundamental to the study of information structure. These form-meaning pairs with mental states of interlocutors.
concepts arc: ·(i) PROPOSmONAL INFORMATION and its two components I will not attempt to define the notion "meanings of linguistic
PRE!UPPOSmON and ASSERTION (Chapter 2); (ii) the IDENTIFIABILITY and expressions" in the above characterization, as this is not of primary
. ACTIVATION states of the representations of discourse referents in the concern in the study of infonnation structure. However I would like to
·minds of the speech participants (Chapter 3}; (iii) the pragmatic relations make three distinctions having to do with meaning which arc important
roPic (Chapter 4) and Focus (Chapter 5). Many of the observations in for the following discussion. The first is a distinction between two KINDS
these chapters have been made by other Linguists before me, and I will of meaning: (i) LEXICAL, which is the meaning inherently expressed in
acknowledge my predecessors whenever possible. Other portions, I lexical items (words and word-like expressions), and (ii) RELATIONAL,
believe. con~in new insights, such as the analysis of the pragmatic which is the meaning that arises by establishing relations between words.
relations .. topic" and "focus" and of the relationship between the two. In For the purposes of this study, the most important kind of relational
particular. what I believe is new in my treatment, and what prompts me mearting is that between ARGUMENTS and PREDICATES, expressed in
to call.it loosely a .. theory," is the idea that an account of information PROPOSITIONS. The distinction between lexical and relational meaning
structure must include alt three of the sets of concepts listed above and will be of importance in the discussion of information in Section 2.2,
must explain how they relate to each other. 1 where I will argue that information necessarily involves the meaning
expressed by propositions. An analogous distinction will be drawn in
Chapters 3, 4, and 5 between the pragmatic slates of discourse referents
2.1 The universe or discourse and the pragmatic relations between discourse referents and propositions.
The second distinction relating to meaning is that between the meaning
I will begin by sketching a simple model of the UNIVERSE OF DISCOURSE. In
of linguistic expressions and the things DESJGl'<ATED or DENOTED by these
this model, I presuppose the primacy of spoken language over other
expressions in particular utterances. The entities and states of affairs
forms of linguistic communication (see Lambrecht 1986b: Ch. I). I will
designated by linguistic expressions in particular utterances are their
therefore always refer to "speakers" and "hearers" (or "addressees") not
REFERENTS. In the case of expressions which do not designate entities or
to "writers" and "readers." The model makes no claim to originality but
states of affairs but attributes or relations. such as small. in, 1-1-·enr home,
simply serves to establish certain background assumptions for the
etc., I will use the term DESIGNATUJl.t or DENOTA_Tl.IM instead of "referent."
discussions to follow.
The third distinction I \vould like to make is that between the referents of
The universe of discourse is divided into t\vO parts: 2
linguistic expressions and the abstract REPRESENTATIONS of these referents
(a) the TEXT·EXTl'.RNAL WORLD. which comrnses Ii) SPfE("H PARTICIPANTS, i.e. a in the minds of the speech participants. Information-structure analysis is
speaker and one or several addressees, and (ii) a SPEECH SETTING, i.e. the primarily concerned \Vith such mental representations. This distinction
place, time and circumstances in which a speech event takes place; between referents and their mental representations, though theoretically

36
 The wiiverse of discourse 39
11!f<JTt11atlOll

imporlant, is tenninologically cumbersome and I will often neglect it pronoun is used, the question of the stress it will receive, where it will
when it is not required for the clarity of the argument. appear in the sentence, or (depending on the language) which pronoun
It is an inherent property of our bipartite model of the universe of type it will belong to is also determined by text-internal criteria. Most of
discourse that elements of the text-external world do not have to be the information-structure concepts used in this study, such as '"topic."
established by speakers via discourse representations but may be taken "focus," "aboutness," "information," etc., are categories of the text·
for granted by virtue of their being present in, or recoverable from, the internal world. They have to do with the discourse representations of
speech setting. Linguistic expressions designating such text-external entities and states of affairs in the minds of the speech participants,. not
elements are referred to as DEICTIC expressions. Deictic expressions with the properties of entities in the real world.
allow a speaker to directly designate elements of the text-external world Particularly revealing from the point of view of the formal
by ··pointing" to them (Greek deiknynii "l show, designate"). Among the manifestation of the categories under analysis are linguistic situations
deictic expressions of a language are those which denote (i) the speaker where the two discourse worlds come together or overlap. This happens
and the addressee(/, you. etc.), (ii) the time of the speech event and points for example when an element in the text-external world (e.g. the speaker
in time measured with reference to it (e.g. now, yesterday, romorro1\', etc.), and/or the addressee) is at the same time a topic in the ongoing
(iii) the place of the speech event and places situated In relation to it (here, conversation. In such cases, one and the same entity can be expressed in
there, etc.), and in general all expressions whose meaning can only be different grammatical forms depending on whether it is referred to by
understood with reference to some aspect of the text-external world (see virtue of its presence in the speech setting or by virtue of its role as a topic
in particular Fillmore !971a and 1976). The text-external world is also in the text-internal world. Let us consider one example of this kind of
coded in certain elements of form usually not labeled deictic, e.g. in the correspondence between the two discourse worlds and oftbe grammatical
feminine adjective ending in the French sentence Je suis co111e11te "lam changes which the transfer from one to the other can bring about.
happy," which owes its occurrence to the fact that the speaker is a English, like other languages, has a special "presentational" construc-
woman. 3 Moreover certain aspects of the social interaction between tion, involving a small number of intransitive verbs like be and come, the
speech participants may be expressed \.Vith linguistic categories relating to subjects of these verbs, and the deictic adverbs here or there. The point of
the text-external world, such as the grammaticalized expression of using this construction is to call the attention of an addressee to the
politeness via deictic categories.
4 hitherto unnoticed presence of some person or thing in the speech setting.
Elements of the discourse which pertain to the TEXT-INTERNAL v.·orld This construction is called "presentational" because its communicative
cannot be taken for granted in the same way by a speaker. In the text- function is not to predicate a property of a given entity but to int,roduce a
intemal world referents are not designated deictically by "pointing" but new entity into a discourse. (The notion "presentational const~ction" is
indirectly, via REPRESENTATIONS which the speaker must set up for the not restricted to deixis, as in the case I have in mind; such constru~tions
addressee (Section 3.2). The form of expressions denoting entities in the may also function to introduce a new entity into the text-internal world,
text-internal world often depends on whether, and bo\v recently, mental
representations of these entities have been established in a discourse.
. '
Section 4.4.4.1. ) Let us assume a speaker wants to draw her addressee's
.
in Y.'hich case they are usually- and misleadingly - called "existential"- see

Entities for which a representation has been recently evoked or ACTIVATED attention to the fact that a hitherto absent entity, say someone's cat, is
(Section 3.3) in the text-internal world are often denoted or "related to" now arriving at the speech setting. She can do this by uttering the
in more abstract fonn via a special class of ANAPHORJ( expressions (Greek sentence
anaphero "to bring back, relate") like she. ir. so, rhere, rhen, etc. For (2.1) Here comes the CAT
example the decision to designate a partH.:ul~ir discourse referent \Vith the
indefinite noun phrase a wo111a11, the definite fll)un phrase the wonian, the In this sentence, the subject noun cat is placed after Lhe verb and its
proper name }.fary, or the pronoun she i:'.> Jetennined by the status of the prosodic prominence characterizes it as having a FOCUS RELATION to the
discourse representation of this ref.:renl in the text-internal \Vorld. If a pro.position. Now 1f at the time of the utterance the entity newly
- 40 h!fomration
- :=-"__!...._ =-----

introduced into the text-external world happens to be already represented

The universe of discourse

who was sitting in the house of a cat owner and who was hoping the
41

6
as a topic in the text-internal world, the speaker can express this fact animal wouldn't show up:
grammAtically by coding the introduced referent in such a way that its

.
(2.2') And here the cat COMES!
topic status in the dis.course prior to its appearance in the text-external
; world is apparent: In (2.2'). the expression designating the animal appears in the same
~ position and with the same lack of prominence as the pronoun he in (2.2),
t" ~) • Here be COMES. but this time it is an NP with lexical content. Its external syntax and its
prosody are the same as that of the anaphoric pronoun, but its internal
1be passage of the referent from the text-internal to the text-external syntax (or its morphology) is different. As in the previous cases, the
world is linguistically expressed in (2.2) by combining a deictic device (the difference is pragmatically motivated. Even though in both cases the
presentational here-construction) with an anaphoric topic-marking referent is already topical in the text-internal world prior to its physical
device (the unaccented pronoun he), resulting in a construction which emergence at the speech setting, there is a difference in pragmatic
is both presentational and predicating. Using a theoretical concept from salience: in (2.2") the speaker does not consider the referent salient
Construction Grammar, we can say that the presentational construction enough in the interlocutors'" consciousness to warrant the use of a
in (2.1) "inherits" the pragmatic features of the pronominal expression pronominal variable. To use a concept to be introduced in Section 3.3,
/h!';in particular the feature "established topic" (Section 4.4.3). the mental representation of the referent has a lower degree of
Notice that the different status of the animal as previously established "activeness"' in (2.2') than in (2.2), requiring lexical rather than
in the universe of discourse or not is expressed not only morphologically, pronominal coding.
by the choice of lexical vs. pronominal coding, and prosodically, by the The use of the deictic here-construction is not restricted to third person
placement of pitch prominence, but also syntactically, by the position of referents. It is possible for a speaker to announce his or her presence to a
the word in the sentence. While the subject expression designating the hearer \.Vith this construction, thereby creating a change in the text-
newly introduced referent (the noun phrase the cat) appears after the external world. The speaker can do this e.g. by uttering the following
verb, the subject expression designating the referent which was sentence:
represented in the text-internal world prior to its arrival at the speech
(2.3) Here I AM.
setting (the pronoun he) appears before the verb. Since this difference in
word order bas no semantic import, we may conclude that it correlates It is also possible for a speaker to ackno\l;ledge the arrival at the speech
directly with the ·difference in the discourse status of the referent in the setting of a previously absent interlocutor by uttering either one of the
tWo utterances. Like the choice of pronominal vs. nominal coding and of sentences
pfetence vs. absence of prosodic prominence, the syntactic choice is
(2.4) a Here you ARF
determined at the level of information structure. The contrast bet\veen
b HERE you are.
(2.1) and (2.2) may be seen as another manifestation of the principle
enunciated in Section 1.3 according to which the topical vs. focal status of (The choice of (a) or (b) depends on "''hether the arrival of the
an expression may be reflected in the syntax of the sentence by preverbal interlocutor \Vas expected or not, a distinction to v.·hich I will return later
YI. postverbal position of the constituent. on.) Notice that even though in these examples the speaker or the hearer
-It is important to acknowledge that the difference in NP position in the is introduced into the text-external "'·arid via the presentational
two examples is not an automatic grammatical consequence of the construction. the status of I and you 1s fundamentally different from
morphological difference between a noun (cal) and a pronoun (he). The that of the car in (2.1) since the speaker and the hearer are necessary
difference is the result of a meaningful choice. To see this, consider the participants in the text-external \~·orld. This fact is again reflected
following attested utterance. made by someone \vith an al!ergy to cats grammatically The expressions referring to the speech participants are
 Information 43
42 lnfo1mation
external .and the text-internal world may be formally reflected in the
pronouns. they are unaccented. and they appear BEFORE the verb. They
grammatical structure of the sentence. This Is especially clear in the
are thus formally similar to the anaphoric he in example (2.2). The
contrast between Here comes 1he CAT and Here the cat COMES, or between
pronouns are not given prosodic prominence and cannot appear in the
Here you ARE and Here's rou. The difference between the members of
postverbal focus position which the NP the cat occupied in (2.1 ). The
these pairs of aUosentences cannot be captured with a ruJe which merely
sequences Here's ME or Here's rou would be unacceptable under the
relates syntactic form and semantic interpretation since the members of,
circwnstances (but see below), and the sequences Here an1 1 or Here are
each pair are syntactically well-formed and semantically equivalent. Any
YOU are ungrammatical.
impression of ill-formedness disappears once the right context is
ro spite of the intonational and positional similarity between the deictic
established. The formal difference can only be explained with reference.
I/you in (2.3), (2.4) and the anaphoric he in (2.2) there is an important
to the component of information structure.
difference between the two kinds of pronouns. With J and you the
contrast between the two discourse worlds is 10 some extent neutralized, In more general terms, the examples discussed in this section
because oftbe dual status of speakers and addressees as interlocutors and demonstrate the existence of a series of systematic correspondences
as possible topics of discussion. Speakers can talk about themselves as between grammatical and pragmatic factors. While preverbal position
well as about other referents, and an addressee can simultaneously be and lack of pitch prominence correlate with topic status and previous
talked TO and talked ABOUT by a speaker. Nevertheless, the contrast in presence of a referent (or its mental representation) in the universe of
question is sometimes gran1matically expressed with first and second discourse, postverbal position and pitch prominence correlate with focus
person pronouns. This happens v"henever the presence of the speaker or status and previous absence of a referent from the universe of discourse.
the hearer, in spite of their role as interlocutors in the text-external world As with our English, Italian, and French model sentences in Chapter I,
is unexpected in a world which is not that of the ongoing discourse. Fo; the grammatical contrasts illustrated in (2.1) through (2.6) are
example, when a speaker discovers herself or the addressee in a group illustrations of the pragmatic MOTIVATION of grammatical form,
pborograpb she might utter sentence (2.5) or (2.6):
2.2 Information
(2.5) Look, here's ME!
In my brief sketch of the universe of discourse I characterized the tex.t·
{2.6} Look, here's you!
internal world as the abstract world of linguistic representations in which ,
INFORMATION is created in the minds of the interlocutors. As a fust step
In these sentences the pronoun now carries the focus accent and is placed
in postverbal position, as in the case of the car in example (2. l ). The towards understanding the notion of information in natural language let
similarity is due to the fact that in both cases a referent is newly us carefuJJy distinguish the information conveyed by the utterance of a
established in a discourse world, a situation which entails prosodic sentence from the MEANING expressed by the sentence. While the meaning
of the sentence is a function of the linguistic expressions which it contains
prominence of the NP (see Section 5.7). Notice that, unlike the topic
pronouns in (2.2) through (2.4), the focus pronoun rne has oblique case- and thus remains constant, the information value of an utterance of the
sentence depends on the mental states of the interlocutors. Whether a
mark.ing and the verb does not agree with it. ( rLJC in (2.6) is also oblique,
given piece of propositional meaning constitutes information or not
but formally ambiguous between nominative and oblique case.) This
depends entirely on the communicative situation in which it is uttered.
illustrates the often noticed (Chough by no n1eans absolute) correlation
One useful \Vay of characterizing information is to say that by
between subject, topic, and agreement on the •)n':' hanJ, and non-subject,
informing the hearer of some situation or state of affairs, the speaker
focus, and lack of agreement on the (>ther (~ee GivL)n 1976, Lambrecht
7 influences the hearer's. MENTAL REPRESENTATION of the world. This
1984a, Bresnan & Mchombo 1987J.
represenlalion is formed by the sum of "propo:>itions" which the hearer
The various examples involving the de1ct1.._· ht'ft·-construction show that
knows or believes or considers uncontroversial at the time of speech. (My
pragmatic differences having to do with thl· contrast bel\veen the text-
 44 biformatio• Information 45

me of the tenn "'proposition" will be justified in the next section.) We in the sense that their application to states of affairs in given worlds may
may refer to this sum of propositions loosely-and perhaps somewhat be correct or incorrect, the mental representations of events, situations,
misleadingly-as the hearer's KNOWLEDGE. "To have knowledge of a or states which we think of in tenns of propositions and which are
proposition" is understood here in the sense of "to have a mental picture communicated in sentences can hardly be characterized as having truth
of its denotatum," not in the sense of "to know its truth" (see below for
,.•· further discussion). To infonn a person of something is then to induce a
values. Such representations simply exist, or do not exist, in the heads of
speakers and hearers. One can know, or be ignorant of, a certain event
change in that person's knowledge state by adding one or more denoted by a proposition, i.e. one may, or may not, have some mental
propositions. The notion of information which I am concerned with "picture" of the event; and one may be thinking of the event, or be
here is well described in the following quote from Dahl (1976): oblivious of it, at a certain time, i.e. one may, or may not, have that
;.
picture at the forefront of one's consciousness. But to characterize the
... ,. Let us consider one important use of declarative sentences, namely as
means to influence the addressee's picture of the world. In such cases,
event, or the picture of it, as true or false seems incongruous .
the speaker assumes that the addressee has a certain picture- or model- If someone informs me that "The cat in the hat is back," my
of tbe worJd and he wants to change this model in some way. We might representation of the world is increased by one proposition, indepen-
;. then identify THE OLD or THEGIVEN with the model that is taken as a point dently of whether what I'm being told is true. If later I find out that the
of departure for the speech act and THE NEW with the change or addition proposition "The cat in the hat is back" was not true in the situation in
that is made in this model. OLD will here be equivalent to PRESUPPOSED in
\vhich it was used, the representation of the cat being back may
one sense of the term. We can say that the addressee receives "new
information" in the sense that he comes to know or believe more about nevertheless linger in my mind. And this representation does not become
the world than he did before. What he believes may be true or false- the false just because it does not correspond to the world as it is. It just
information he gets about the world may be correct or incorrect. If we becomes outdated. To take another example, if someone says to me "I
accept that last statement, it follows that the object of his belief or the just found out that Sue is married," and I happen to know that in fact she
new information must be something which is capable of being true or is not married, it is certainly possible to say that the speaker has a false
false-that is what is usually called a proposition. Let us therefore call
belief about Sue's marital state, hence that the proposition "Sue is
this kind of information PROPOSmONAL INFORMATION. (Dahl 1976:38)
married" is false under the circumstances. But this \Yay of phrasing things
It should be noted that when a speaker influences the hearer's "picture" does not seem to contribute much to our understanding of the utterance
or the world by adding to it, only a small portion of that picture is as a piece of information. If I correct the speaker by saying "But it's not
normally affected, namely the portion which is "under discussion" and true that she is married," I am still evoking the same mental
with respect to which the piece of information conveyed is meant to be representation "Sue is married." and I assume my addressee still has
.RELEVANT. Pictures, like human knowledge, are structured. For example, this representation in his mind. even though the proposition is not true.
when someone talking to me about her brother says that he is studying From the point of vie\\' of the infom1ation structure of the sentence, it is
linguistics, it is my knowledge of the speaker's family that is affected and the existence and cognitive state of this representation in the mind of the
not my knowledge of, say, the economic situation of the United States. I interlocutors that counts, not the question of the truth of the proposition
will come back to this point in the discussion of presupposition in the in terms of v.'hich it is conceptualized. \Vhat v>'e are concerned with is the
next section. fit between states of minds and sentence structures. not bet\J\.'een states of
While I fully agree with the importance attributed in the above quote affairs and propositions.
to the propositional nature of "ne\V information," I depart from Dahl's Let us 1um lo the notions "o!d" and "ne""' 1n Dahl's quote. It is a
characterization in one important point, at least as far as terminology is fundamental property of infonnation in natural language that whatever
concemed. 8 I believe that in the linguistic analysis of information, hence is assumed by a speaker to he NE\\' to a hearer 1s infonnation which is
in the grammatical domain of information structure. the logical concept ADDED lo an already e.x1st1ng stock of knov.:ledge in the hearer's mind.
OfTRum has no place. While propositions may be said to he true or false, The hearer's mind is not a blank sheet of paper on which new
 JnJorma11on 47
•l,J-····-··~··

propositions are inscribed. Conveying information therefore requires Thus it is necessary to distinguish the PROPOSITIONAL INFORMATION
constantly changing hypotheses on the part of the speaker about the state conveyed by a sentence, as characterized in the quote from Dahl above.
of knowledge of the hearer as speech progresses. Information can from the ELEMENTS OF INFORMATION in a sentence, i.e. the contributi~~
normally be conveyed only if the hypotheses made by the speaker made by individual words or phrases to the propositional information.
concerning the hearer's state of knowledge are correct, Le. if the One might want to call such contributions ..lexical information.. or
information the speaker is trying to convey is not already stored in the "referential infonnarion." However, to avoid confusjon I will use the
hearer's mind. This condition on the successful transmission of term .. information" only for the creation of knowledge via propositions.
information has been called the "Principle of the Presumption of By insisting on the distinction between information as expressed via
Ignorance" by Strawson (1964). Now, since the stale of ignorance of a propositions and the elements or building blocks with which propositions
hearer is never total- new information always being added to already are formed, we may avoid one type of confusion which often ·arises ip.
e:\isting knowledge-this principle must be complemented by another discussions of new vs. old information. It is often said that certain
principle, which Strawson calls the "Principle of the Presumption of constituents of a sentence, in particular the subject, ''convey- old
Knowledge." This second principle is based on the idea that information," meaning that they are known to the addressee or have
been mentioned in previous discourse or are inferable from Previouslf
statements, in respect of their inforn1ativeness, are not generally self· mentioned elements, whereas other constituents, in particular the
sufficient units, free of any reliance upon what the audience is assumed predicate, "convey new information", meaning that they are not known
to know or to assume already, but commonly depend for their effect
or inferable in that way. Implicit in such statements is the notion that the
upon knowledge assumed to be already in the audience's possession.
(Strawson 1964:97) information expressed by a sentence is segmentable, i.e. that it can be
divided up among the various sentence constituents, each carrying' a
To use two popular terms. there is normally no "new information" subportion-either old or new- of this total information. It is Casy to see
without already existing "old information." why this idea has some intuitive appeal. Consider the following question-
The notions of NE\\' INFORMATION and OLD INFORMATION have given rise answer pair:
to great confusion in the literature and it is my main concern in the
present chapter to clarify and differen1iate them. As a first step towards (2. 7) Q: Where did you go las! night?
clarification, lel us recall the difference bet\veen INFORMATION and
A: I went to the movies.
MEANJNG. While meaning is expressed either in individual words or via
relations established between words, information can strictly speaking
only be conveyed reiationally, via propositions. Informing a hearer of ll is tempting lo say that in the answer to the question in (2.7) the
something means informing "him or her of some state of affairs, i.e. of constituent the tnovies or perhaps to the n1oi•ies expresses "the new
something which necessarily involves not only participants but also information" because the remaining portion of the sentence, made uP of
something to participate in. One can inform someone e.g. of the price of a the words I and went, was already contained in the question (and in the ,,1
book but not of a book or of ten dollars. The expression the price of a case of l taken for granted from the text-external world). Therefore, the
book codes the proposition "The book has a price," i.e. it codes a relation reasoning goes, the referents or designata of these words may be assumed
between a predicate and an argument, but the expressions u book or ren to be present in the speaker/hearer's mind, therefore they cannot count as
dollars code only quantities of entilies. 11 1s /rue that a proposition can new, and therefore these words "convey old infonnation."
state the mere existence of an entity, but such a staternent still involves a In spite of its intuitive appeal, I consider this account of the
predicate and an argument. If I say '·t-.,1oney 1'" upon seeing a dL1llar bill in information structure of (2.7) incorrect, or at least misleading. If "new
the street, I am infonning my addres~ee of .:i state of affairs, i.e. thal there information" were equated with "new constituent," i.e. a constituent
is a dollar bill in the street or that I h:.ivc DLHttcJ this bill whose referenC or designarum is "new" to the hearer in a particular
 48 llf/ormation Information 49

disccn11Set it would be difficult to account for the information structure of used deictically). Nevertheless, in an appropriate ullerance context this
the answer to the question in (2.8): sentence clearly may convey new information in the sense that it may
change the addressee's representation of the world. The conveying of
(is) Q: When did you movt:: to Switzerland? information is in principle independent of the previous mention or non-
A: When I was seventeen. mention of the designata of the different constituents in a sentence. As in
the previous examples, the conveying of information comes about .h~re
What constitutes the information conveyed by this answer is of course via the establishment of relations between the elements of the propos1tton
not the fact that at some point in his fife the speaker was seventeen (or, as in one interpretation of (2.9), via a change in the polarity of the
(expressed by I was se\•enteen). let alone some abstract time indication proposition). The sentence in (2.9) may convey some piece of new
(e:xprCssed in when and H-'as), but the RELATION established between an act information as readily as !he follov.:ing (pragmatically somewhat
Of moving to Switzerland, the person involved in that act, and the time at peculiar) sentence cited by Allerton {1978):
which the moving occurred. It is the role of the time expression as an
aiiument 'cor ..adjunct") in an open proposition that is unknown to the (2.10) A clergyman·s opened a betting shop on an airliner.
addressee; hence it is the indication of this role that makes the answer
In this sentence the three constituents a c/erg}'man, a betting shop, and an
informative. The fact that in (2.8) the mere use of the temporal clause is airliner have referents \vhich v.rere presumably not mentioned in the
sufficient to express the requested information does not entail that it discourse preceding the utterance. Moreover the predicate has opened
expresses by itself "the new information." The conveyed information is may also be "ne\v .. to the discourse_ (The question to what extent
not ..when I was seventeen" but (clumsily expressed) "The time when I predicates may be said to be "new" or "old" \vill be discussed in Sections
moved to Switzerland is the time when I was seventeen." The information 3.4 and 5.4.2.) The relevant difference between 12.9) and (2.10) is not a
is the establishment of a relation between terms in a proposition. difference in the "ne\\·nesS' or '"oldness" of the information but rather a
In a similar vein, the information conveyed by lhe answer in (2. 7) is not difference in the assumed states of the representations of the referents or
"'to the movies" but something like "The place I went to last night was designata of the various sentence constituents in the addressee's mind at
the movies." That to 1he movies in (2. 7) and u•hen I u•as sevenleen in (2.8)
the time 0f utterance. 9
cannot constt"lute the information in themselves is clear from the fact that To sum up, the informa(ion conveyed by a proposition cannot be
they could not function as interpretable answers without the associated factored out and matched with individual sentence constituents. In
full propositions. This is not to say that there is no grammatically particular. the difference betv. een "old tnformatton" and "new informa·
1

relevant difference between the answers v.•hen I •~·as seventeen or to the tion" e-annol be equated '.-vith the difTerence bet\veen "old" and "new"
movies and the elements of the propositions which were already contained referents (see Section 5.4.1 for further discussion)- I will therefore reject
in the questions. This difference will be described in the next section in the segmentation viev.-· of information and replace it with an account of
ierms of the notions "presupposition" and "assertion" and later on the information structure of sentences in \Vhich a distinction is made
(Cba'pter 5) in terms of the notion of "focus." bet1,veen (i) the PRAGMATIC STATES of the denolala of individual sentence
The dilemma for the segmentation vie\v of information is particularly constituents in the minds of the speech participants and (ii) the
striking in the case of a simple sentence like (2.9): PRAGMATIC RELATIONS established betv.-·een these denotata and the

(2.9) She om it. propositions in \vhich they play the role of predicate$ or arguments. It
is the establishment of such pragmatic relations that makes information
In this sentence all constituents must be equally "old" because otherwise possible
they could not aU appear in anaphoric pronominal (and "pro-verbal") The need (O drav.-• a (heorefn:a! disC1nct1on bct\veen the pragmatic status
form: to be able to interpret these cons!ituenls \ve must know from of individual items in a proposition and the infonnation conveyed by the
previous discourse who or what they refer to (none of these expressir1 ns is proposition as a \\'hole is n11\ a new 1dea. It is stated. for example, hy
............................... ----------~~~~
Presupposition and assertion Sl

Jespersen in bis Philosophy oj· Gran1111ar ( 1924). In his discussion of the 2.3 Presupposition and assertion
concepts "subject" and "predicate," Jespersen writes:
In my discussion of the dual nature of information as expressed in
Strawson's two principles of the "Presumption of Ignorance.. and the
The subject is sometimes said to be the relatively familiar element, to "Presumption of Knowledge," I mentioned that the information
which the predicate is added as something new ... This may be true of
conveyed by a proposition is itself normalJy a COMBINATION of old and
most sentences, but not of all, for if in answer to the question "Who said
that?" we say "Peter said it," Peler is \he new element, and yet it is new elements, insofar as what is new is normally new with respect to
undoubtedly the subject. The "new information" is not always something which is already given. This property of information is
contained in the predicate, but it is always inherent in the connection reflected linguistically in the fact that sentences typically contain som~
of the two elements, - in the fact that these two elements are put lexical or grammatical manifestation of the information assumed to be
together. (1924:145) already given in the hearer's mind, as a verbal point of departure or. basis
for the new information to be added. ·
Jespersen's statement contains both a clear distinction of the two The point that new information is made up of a combination of old
categories "new referent" and "new information" and the germ of their and new elements is by no means trivial. If our goal as speakers is to
confusion. For even though he emphasizes the propositional nature of increase the knowledge of an addressee, why should we ever have to say
information by insisting that information arises through the CONNECTION things that \Ve assume the addressee knows already? In a naive model of
established between the elements of a proposition and not through the information as a way of helping an addressee acquire knowledge, there
elements themselves, he blurs this distinction by saying that "the 'new would be no reason for such redundancy. However if it is understood that
information' is NOT ALWAYS contained in the predicate" (emphasis mine), information arises by RELATING something n..::w to something that can
thus leaving the door open to the interpretation that sometimes, or in already be taken for granted this apparent redundancy becomes a
most cases, it is indeed contained in the predicate. The distancing effect necessity.
Jespersen creates by putting the phrase "new information" in scare Let us look at an example. It is often said that the propos1t1on
quotes seems to indicate that he was aware of the possible confusion but expressed by a restrictive relative clause is "presupposed" (in one sense of
trusted the reader to make the necessary adjustments. I will return to the this word), meaning that it is assumed to be already known (or believed
particular issue of the information status of focal arguments like "Peter" or otherwise taken for granted) by the addressee. Thus when I say
in Jespersen's example in Section 5.2.3.
(2.ll) I finally met the woman who moved in downs.lairs
In my own terminological practice I will restrict the use of the terms
"old information" and "new information" to aspects of information \11hat I want to communicate to my addressee is that I met my new
associated with propositions. "Old information," then, is the sum of neighbor (v,rhose existence and sex I assume my addressee is aware oO,
"knowledge" (in the above-stated sense) evoked in a sentence which a not that someone moved in do,,..·nstairs. By using 1he restrictive relative
speaker assumes to be already available in the hearer's mind at the time clause who n1oi,ed in downstairs I express the fact that I take for granted
of utterance- "the old," ''the given,'' or ''the presupposed'' in the quote that my addressee already knows that someone moved in downstairs. If I
from Dahl-while "new information" is the information added to that "''anted to \NFOR1'-1 my addressee of the proposition expressed in the
knowledge by the utterance itself - "the ne\v" in Dahl's tenns_ Both old relative clause, I would have to say something like Someone moved in
and new information correspond to propositions and cannot b..:: equated downstairs. It's a won1an or perhaps This woman moved in downstairs. 10 So
with the lexical or phrasal elements out l)f \vhich propositions ar..:: formed. why do I bother to utter the relative clause, if my addressee already
Because of the confusion attached to the t\vO tem1s, I \vi!l r..::place them in knows the proposition expressed by it? The answer is, of course, that the
most contexts with the more spec1fic1ll;. llngu1st1c ter1ns "presupposi- relative clause helps the hearer determine the referent of the phrase the
tion" and "assertion," to \vhich I \Vtll Lurn no\\' 11-01nan, by RELATING this referent to some already given piece of
- 52
-----------------~-~~------
lrifomration

knowledge. which I assume the hearer happens not to be thinking of at

Presupposition and assertion

the hearer at the time of utterance, by indicating what is assumed to be

53

the time l utter the sentence. already given and lvhat is assumed to be new.
To confirm the claim that in using the restrictive relative clause in A comment first about my use of the tenn PROPOSITION in (2.12), whose
(2.11) the speaker indeed takes for granted that the addressee knows the Iogico·semantic connotations may seem inappropriate in the present
proposition expressed in it we can apply to this sentence what Erteshik- context. What a speaker assumes a hearer knows or takes for granted are
Shir & Lappin (1979, 1983) call the "lie-test'. Let us assume the addressee strictly speaking not propositions but states of affairs, situations, events
were to challenge the statement in (2.1 I) with the reply That's not true. etc., i.e. the kinds of things which may be DENOTED by propositions. Since
This reply would be understood as challenging only the proposition that I I know of no generally accepted simple tenn for the denotatum of a
nict my new neighbor, not that someone moved in downstairs from me. tr proposition, I will simply use the term "proposition" ambiguously,
he were to make his challenge more explicit, he could say That's nor true, making the distinction explicit in cases such as this where I fear confusion
you didn't, but hardly That's not true, she didn 'r. In saying That's not true may result. I am not suggesting that the knowledge shared between a
the addressee would be understood as challenging only that portion of speaker and a hearer has the status of a set of propositions or "latent
the utterance which is presented as new, not the portion which is sentences" in the interlocutors' minds. In particular, I am not suggesting
grammatically marked as to be taken for granted. If he wanted to indicate that the pragmatic presuppositions evoked in an utterance must be
that the proposition which the speaker is treating as known can in fact linguistically represented by some verbal or other predicating expression
not be assumed to be known, he would have to modify the in the sentence, allhough they may of course be_
prcsuppositional situation explicitly, by saying e.g. I didn't know that To take one example, the use of the definite article in the noun phrase
you had a new neighbor or What are you talking about. you /h•e in a one- the u•oman 11'ho mo1'ed in downstairs in example (2.11) evokes the
story building! 11 presupposition that the addressee can identify the individual designated
Let us refer to the "old information" contained in, or evoked by, a by that noun phrase (see Section 3.2). The definite article is a
sentence as the PRAGf.tATIC PRESUPPOSITION (or simply the PRESUPPOSITION, grammatical symbol for an assumption on the speaker's part, and this
see the comments below), and let us refer to the "new information" assumption can be represented in the fonn of a proposition. i.e. the
expressed or conveyed by the sentence as the PRAGMATIC ASSERTION (or proposition "the addressee is able to identify the individual in question."
simply the ASSERTTON). 12 ''Presupposition" and "assertion" are defined in (This is whal philosophers call the "existential presupposition" uf the
(2.12) (the definition of "presupposition" will be extended below and definite description expressed by the noun phrase the woman 1vho mo1'ed
slightly modified later on): in doirnstairs.) This does not entail that the definite article expresses a
proposition or should he viev..-ed as a kind of sentence. Notice also that
(2.12) PRAGMATIC PRESUPPOSITION: The set of propositions lexicogrammatically
evoked in a sentence which the speaker assumes the hearer already while the definite article may be said to symbolize an assumption, it is
knows or is ready to take for granted at the time !he sentence is uttered. misleading to say, as is often done in the literature on discourse
pragmatics, that the referent of a definite noun phrase, let alone the noun
PRAGMATIC ASSERTION: The proposition expressed by a sentence which
phrase itself, "is presupposed." Just as I cannot "infonn you of a
the hearer is expected to know or take for granted as a result or hearing
woman" (see Section 2.2 above), I cannot "presuppose this woman" (see
the sentence uttered.
the discussion of the relationship bel\veen presupposition and topic in
Recall that "to know a proposition,. is understood here in the sense of Section 4.3).
"to have a mental representation of its denotatum."' The expression is To the presuppositions evoked by a sentence \vhich concern the
neutral with respect to the question of \Vhether the proposition is true or assumed KNO\VLEDGE state of the addressee ...,..e must add those ,.,1hich
false. In making an assertion. a speaker expresses a PRAGMATICALLY have to do \Vith the speaker's assumptions about the state of
STR.ucrtJkED PROPOSITION, i.e. a proposition \Vhich renects not only a state coNscrousNE<;<; 11r A\V·\RENrss of the addressee at the time of utterance.
of affairs but also the speaker's assumptions about the state of mind of I \Viii refer to such presupposti0ns as CONSCl01.!SNESS PRESUPPOSITIONS.
 /11}Qr'111111011
Presupposition and assertion SS

Consciousness presuppositions are evoked in particular by the usage in which "assertion" is synonymous with ··statement," i.e. in which
differences between lexical vs. pronominal (or phonologically null) the term refers to a kind of speech act, expressed in DECURATIYE as
codings of denotata or by dilTerences in pitch prominence. These kinds of opposed to INTERROGATIVE, IMPERATIVE, or EXClAMATJVE sentences. From
presuppositions will be dealt \vith in Section 3.3 under the heading of the point of view of information structure, questions as well as orders and
"referent activation." For example the use of the personal pronoun she in requests convey infonnation, even though they are not statements. For
the sentence She is my FRIEND evokes the speaker's assumption that the example by asking a question, a speaker may inform his addressee of his
addressee is in a certain state of awareness with respect to the individual desire to know something; by giving an order he may infonn his
in question, i.e. that some mental representation of that individual is at addressee of the obligation to do something, etc. Within the present
the forefront of the addressee's consciousness at the time of utterance. 13 framework, non-declarative sentences, like their declarative counterparts,
Last, but not least. I \viii count among the presuppositions evoked by a are viewed as having pragmatic presuppositions and as being used to
sentence the assumptions a speaker has concerning the contextual make assertions. This extension is necessary because many of the
RELEVANCE or TOPICAL1n· of a referent in the discourse, i.e. the degree to grammatical phenomena analyzed in this book are found in questions
which a referent can be taken to be a center of current interest with and negated sentences as well as in statements (see example (5.13) and
respect to which a proposition is interpreted as constituting relevant discussion).
information (see Section 4.3). I will refer to such presuppositions as Of special importance in the definition of the pragmatic presupposition
RELEVANCE PRESUPPOSITIONS. For example, the above-quoted sentence She in (2.12) is the phrase LEXJCOGRAM~tATICALLY EVOKED. Unlike the more
is my FRIEND evokes not only the presupposition that the hearer is general cognitive notions "representation of the world" and "knowl-
presently aware of the particular female individual denoted by the edge," which I discussed in the section on information, "presupposition"
pronoun she (a consciousness presupposition) but also that this is understood here as a specifically LINGUISTIC concept. To count as a
individual is ropicaJ in the discourse. i.e. !hat !he proposi1ion expressed pragmatic presupposition in the sense of (2.12), an assumption made by
by that sentence can be contextually construed as constituting relevant the speaker concerning the hearer's state of mind must have some actual
information with respect to this individual. This topicality assumption manifestation in the grammatical or lexical structure of the sentence, i.e.
would NOT be evoked by the allosentence SHE is my friend, even though the presupposed proposition must be in one way or another FORMAUY
this sentence would still evoke the given consciousness presupposition EVOKED by the speaker in the sentence. Any assumption on the part of the
(see Section 5.2.3). One may object to applying the term "presupposi- speaker which has no formal manifestation in a sentence is irrelevant for
tion" to matters of consciousness and relevance since these do not seem the study of information structure.
to be appropriately described in terms of a hearer's knowledge or beliefs. Let us apply the definitions in (2.12) to example (2.11) I }molly met the
However, since the phenomena in question clearly have to do \\'ith a 11•01nan who rnoi•ed in downstairs. The pragmatic presuppositions
speaker's assumptions about the slate of mind of the hearer I v.'ill lexicogrammatically evoked with the utterance of this sentence may be
subsume them under the general heading of "pragmatic presupposition." loosely stated as the following set of propositions:
A comment is necessary also about my u::.e of the term ASSERTION. I use
this term as nearly synonymous \vith \\'hat I called "new information" in (i) tho add"'"' can identify the female individual designated by the
definite noun phrase;
the previous section. Nevertheless the t\vO notions are distinct. While
"information" has to do with the communicative act \\'hereby a speaker (ii) someone moved in downstairs from lhe speaker;
increases a hearer's kno\v]edge- or enricht.:s her representation of the (iii) one would have expected the speaker to have met that individuaJ at
world-by adding a new proposition to 1L "assertion" is the added some earlier point in time.
proposition itself. I should emphasize th:it n1y u~e L)f··assertion" does not
coincide with the common usage in \\'hich ··asserting" a proposition The first presupposition is evoked by a grammatical morpheme, the
contrasts with denying or questioning it. Nl1r does it coincide \vith the definite article the; the second is evoked by a grammatical construction,
c, 56 lnfomwtion Presupposirio11 and a.ssertion 57

the relative clause K'ho moved in do·wnstairs; and the third is evoked by a Often the presuppositions evoked in an utterance are fully or partially
lexical item, the adverb finally. 14 To these three presuppositions ex.pressed in the preceding linguistic context, either in already
COJJCem.ing the knowledge state of the addressee we must add the presupposed or in asserted ronn. For example, in the question-answer
consciousness presuppositions evoked by the personal pronoun I and the pair in (2.?) the presupposition evoked by the answer, i.e. that the speaker
relative pronoun Yt'ho: went somewhere, \Vas already evoked in the question ~Vliere did you go
fast night? (see the discussion of the presuppositional structure of WH-
(iv) the addressee is aware of the referents of the pronouns I and who at the questions in Section 5.4.4). 16 In (2.8) one or the presuppositions required
time these pronouns are uttered. by the answer iVhen I u•as sel'enteen is the proposition .. the speaker
moved to Switzerland," which was explicitly stated in the question and
Finally the sentence evokes the following relevance presuppositions via which may be lert grammatically unexpressed in the answer because it is
the two unaccented pronouns:
assumed to be still ··active" in the hearer's consciousness. Even though
this presupposed proposition is not over!!}' expressed in the answer. it is
(v) the proposition expressed by the sentence is construable as rele\'ant
information about the referent of /; the proposition expressed by the nevertheless grammatically evoked, as a phonologically null string.
relative clause is construable as relevant information about the referent Notice, incidentally. that the proposition .. I was seventeen" itself
of who. represents a piece of knov,rledge already shared by the interlocutors.
The assertion thus consists here in establishing a time relation between
(The status of the relevance presupposition evoked by the relative two pragmatically presupposed propositions (see example (2.13) below
pronoun is somewhat special; see the remarks about relative clause and discussion.)
presuppositions in Section 4.1). The assertion expressed by (2.1 I) may Similarly, in Jespersen's above-quoted passage, the presupposition
then be infonnally stated as rollows: "Taking for granted the pro- evoked by the answer PETER said ir is the proposition .. Someone said it"
positions in (i) through (v) above, the speaker has now met the individual which was already evoked in the question lt'ho said that? (see Section
in question.' 5.4.3 on presupposed .. open propositions"). The presupposition of the
Now the speaker who utters (2.J J) surely assumes that he and the ans\ver is again evoked by non-lexical means, in this case prosodica.!ly.
addressee share knowledge in addition to the above set or pragmatic Another presupposition evoked in this ans\ver is the assumption that the
presuppositions, e.g. the knowledge that moving is a hassle, or that two referent of the pronoun it is presently at the forefront of the addressee's
and two makes four. However, such shared knowledge is not evoked in consciousness and, a fortiori, that the addressee can identify this referent.
the utterance, i.e. has no lexicogrammatical manirestation in the sentence, The speaker also presupposes that the addressee can identiry the
hence is irrelevant to the analysis of the inrormation structure or this individual referred to as "'Peter." <The presupposition attached to
sentence. There is thus a difference between the gramn1atically relevant definite descriptions, \Vhich has to do \Vith the mental representations or
notion of pragmatic presupposition needed in inrormation-structure discourse entities, will be discussed in Section 3.2 under the heading of
analysis and the notion or presupposition round in many discussions on "identifiability.") \Vith her answer, the speaker then makes the AS.<;ERTION
pragmatics, such as, ror example, Kempson's notion or the "Pragmatic that the particular individual \vho said the thing referred to as "that" is
Universe of Discourse" (1975:166ff), \vhich she defines as the entire the individual "Peter." A more explicit characterization of the
"body of facts which both speaker and hearer believe they agree on·· in a presupposition-assertion relation in sentences such as this will be
conversation (see also the rererences on p. 345, note 12). Only the fonner presented in Section 5.2.3.
has a direct bearing on lhe formal structure of a sentence. The distinction The presupposition and the assertion are thus propositions \.Vhich
I am emphasizing here echoes the distinction bet\veen "infonnation coexist in the same sentence_ To make an assertion is to establish a
structure" and "(conversational) pragmatics" \vhich I emphasized at the RELATION between a presupposed set of prnpn<;ition'i (\vhich, as we shall

beginning of Chapter l. 15 see, may he empty) and a non-presupposed proposition. the latter being
 Presupposirion and assertion 59
58 Jnjor111a1ion

consists in the establishment of a relation of causality between two

in some sense added to, or superimposed on, the former. The assertion is
previously unrelated presupposed propositions. We can again apply
therefore not to be seen as the utterance .. minus the presupposition" but
Erteshik-Shir & Lappin's lie-test to make this clear. If speaker A were to
rather as a combination of two sets of propositions. In view of the claims
to be made later about the grammatical signaling of the presupposition- challenge speaker B's explanation by saying "That's not true," the
assertion relation it is important to understand that the superimposition challenge would normally be understood as involving the causal relation
of the asserted proposition on the set of presupposed propositions often between the two presuppositions ("That's not true, you didn't do it
CMXUrs in such a way that the two cannot be lexically factored out and because of that") not the presuppositions themselves. The reply "That's
identified with specific sentence constituents (see also my remarks to this not true, I'm not your friend" would of course not be impossible but it
effect in the section on information above). For example in (2.11) the would constitute an explicit modification of the presuppositional
presupposition "someone moved in downstairs" does not exactly situation. The observation that the combination of known propositions
coincide with the meaning of the relative clause who moved in downstairs can result in an assertion is related to the earlier observation (see example
since the relative pronoun H'ho and the indefinite son1eone have different (2.9) and discussion) that a piece of new infonnation may result from the
referential properties, nor does it coincide with the meaning of the combination of expressions whose referents are entirely given by the
complex noun phrase the woman who mo1·ed in downstairs since that noun preceding context. Failure to recognize this fact has often led to
phrase evokes several different presuppositions. Rather the grammatical confusion in analyses of "new" and "old" information.
domain for both presupposition and assertion is the sentence or clause as Consider now the following utterances. Sentence (2.14a) is to be
a whole. This fact will be of special importance in the discussion of focus imagined in a situation \vhere the speaker has just noticed that the
in Chapter 5, where "focus" will be defined as that portion of an addressee recently had his hair cut; (2.14b) was said to me by my three
utterance whereby the presupposition and the assertion differ from each year old son (who knew that I knew what was in my kitchen drawers);
other. Since that portion can often not be identified with a particular and (2.14c) is self-explanatory:
sentence constituent, the relationship between focus MEANING and focus (2.14) a You got a haircut!
MARKING will be shown to be rather indirect.
From the characterization of "assertion" as the proposition which the b. There's some candy in the kitchen dra'A-·er.
hearer is expected to know as a result of hearing a sentence, it follows (as c. You lied to me!
a truism) that the asserted proposition must differ from the set of
propositions which are presupposed. One cannot INFORM an addressee of It is clear that the propositions expressed in these sentences are
something she already knows (although one can obviously TELL an pragmatically entirely presupposed, in the sense that the addressee
addressee something she knows already). However, while an assertion obviously was assumed to know them before hearing the utterances. But
cannot COINCIDE with a presupposition, it may consist in RELATING two or it is equally clear that these utterances are assertions, in the sense that
more presuppositions to each other. This possibility v.'as hinted at in the after hearing them the addressee knov.·s more than he did before. The
discussion of the presuppositional structure of (2.8). As another example apparent contradiction is easily resolved. What is communicated with
consider the following conversational exchange: these sentences is not their propositional content but the fact that the
speaker knows a proposition which he assumes the addressee did not
(2.13) A: Why did you do that? think he knew. The communicative point of the utterances in (2.14) is to
make explicit that the speaker and the addressee now have the knowledge
B: I did it because you're my fricnJ
of those propositions in common. To use Stalnaker's expression, the
Even though both the proposition ··1 did 1t'" and the proposition utterances have created "common ground" between the interlocutors.
"you're my friend" may be considered pragmatically presupposed, This last observation concerning the difference between he.arer-
speaker B's answer clearly is informative The a)sertion it expresses presupposition and speaker-hearer-presupposition makes it necessary to
 60 lnfomwtion Presupposition and asse.rtion 61

slightly revise the notion of pragmatic presupposition in (2. l2). What The last aspect of the definition in (2.12) in need of explanation is the
counts for this notion arc not only the speaker's assumptions about the modifying adjective PRAGMATIC in the term "pragmatic presupposition."
hearer's state of mind but also the speaker's assumptions about the This adjective is meant to differentiate the phenomenon under analysis
hearer's assumptions about the speaker's state of mind. Instead of here from a different kind of presupposition, which has often been
reformulating the definition in (2.12), I will simply append to it the one referred to in formal semantics as SEMANTIC or LOGICAL presupposition,
proposed by Stalnaker: and which has to do with the effects of certain lexical items on the truth
conditions of the sentences containing them. 18 Even though the difference
A proposition P is a pragmatic presupposition or a speaker in a given between "pragmatic presupposition" and "semantic presupposition" is
context just in case the speaker assumes or believes that P assumes or by no means clear-cut (the terminological, if not the conceptual,
believes that bis addressee assumes or believes that P,and assumes or
distinction has in fact been al! but abandoned in the literature), it is
believes that his addressee recognizes that he is making these
assumptions, or has these beliefs. (1974:200) necessary to emphasize one point of divergence, which has to do with the
above-mentioned difference between INFORMATION and MEANING. While
Stalnaker's definition, like that of other philosophers and linguists, is pragmatic presupposition in the sense of information-structure analysis
phrased exclusively in terms of propositions, not in terms of the has to do with the assumptions of speakers concerning the information
Jexicogrammatical manifestation of propositions in sentences. As I status of propositions in utterance contexts, i.e. with communication,
mentioned before, such manifestation is crucial for the purposes of semantic presupposition, at least in one common use of this term, has to
information-structure analysis. do with the semantic relations between sentences or propositions, i.e.
While all utterances must express pragmatic assertions in order to be with logical meaning and truth conditions.
informative, it is less clear whether all assertions require presuppositions, According to one widespread notion of semantic presupposition, "one
i.e. whether the set of presupposed propositions in a sentence may be sentence presupposes another just in case the latter must be true in order
empty. Perhaps the best candidates for assertions without presupposi- that the former have a truth value at all" (Stalnaker 1973:447). To cite a
tions are ..thetic" propositions (see Section 4.2.2) like !r's raining or There common example, it has often been observed, since Kiparsky & Kiparsky's
is going to be afighl. Another candidate for a presuppositionless assertion seminal article "Fact" (1970). that sentences containing certain "factive"
is a discourse-initial utterance like our model sentence (I. I) My CAR broke verbs presuppose the truth of the complements of these verbs, i.e. that the
down, made under the described circumstances. As I observed in the truth value of these complements is not affected by matrix-clause
discussion of that sentence in Chapter 1, the interpretability of this differences in polarity or modality. For example, both the sentence John
utterance depends heavily on the situational context (the situation in the regrers that he lied ro A1ary and its negative counterpart John doesn't regret
bti.s)~ which determines in particular the RELEVANCE of the utterance. that he lied to Mary are said to presuppose the truth of the proposition
However, nothing in the lexicogrammatical structure of this sentence "John lied to Mary." lfth1s last proposition is not true, i.e. if John did in
evokes knowledge shared by the speaker and her audience, except for the fact not lie to Mary. then both the positive and the negative version of
·~accommodated" presupposition that the speaker has a car (see Section the sentence are said to lack a truth value because neither the claim that
2.4 be)ow). The sentence may therefore count as (quasi-)presupposition- they are true nor the claim that they are false seems to make much sense.
less for the purposes of information structure. Finally one might consider Let us look at some linguistic phenomena 'vith respect to y,:hich the
such examples as the earlier-discussed "hot ne,vs" sentence (2.10) A "pragmatic" and the "semantic" approaches to presupposition differ.
clergyman's opened a betting shop on an airliner. \Vhat makes such Consider the follo\\•ing variants of example (2 l4c)·
sentences pragmatically so exceptional is precisely that they lack any (2.15) a I didn't reali7e that you 1.1ro to me
overt presuppositional reference point and therefore violate Stra,vson's h I didn't reil!JZt' th:il ;n1· lied to n1e
"Principle of the Presumption of Knowledge'' (see 2.2). 17
c I didn't IU-\Ll7F rh,11 ~ou lied to me
................................._..... ----~~~~
Presupposilion and asserlion 63

S\ntt \he verb reali:e is factive, an account of {2. l 5) in terms of the within the pragmatic framework adopted here. From the defmition of the
semantic notion of presupposition v.•ill have lo stale that if the three pragmatic presupposition of a sentence as a (lexicogrammatically evoked)
sentences are to have truth values al all, the complement clauses must be set of propositions which the speaker and the hearer are assumed to have:
true in all three cases. What is semantically presupposed is the in common at the time of utterance it follows naturally that the truth of
"factuaJity" of the proposition expressed in the that-clauses. This any pragmatically presupposed proposition is simply taken for granted
presupposition is entirely determined by a lexical feature of the by the inLerlocutors and therefore cannot be affected by an assertion
sentence, i.e. the presence of the verb realize, and does not change with (unless the point of the assertion is to make the addressee aware that
the conversational circumslances under \Vhich the sentence is uttered. The some presupposition was faulty). As we saw with the application of the
meaningful distinctions expressed by the different accent placements in lie-test (example (2.1 l) and discussion), any aspect of a sentence which
the sentences in (2.15) (a), (b), and (c), remairi unaccounted for. affects the truth value of the proposition expressed by it must be an
A pragmatic account of the presuppositional structure of these element of the assertion, not of the presupposition. For example, let us
•
sentences is rather differenl. First, we notice that in the (a) sentence the assume a state of affairs in which Jespersen's sentence PETER said ii would
proposition expressed in the complement clause need not be pragmati- be false as a reply to the question JVho said that? The falsity of this reply
cally presupposed at all, since, as in the case or the original utterance would NOT affect the pragmatic presupposition required by the false
(2.14c), the kno\vledge of that proposition may not yet be part of the answer, namely that a particular person said a particular thing. What
common ground between the interlocutors (i.e. the meaning of the \Vould be affected is rather the assertion that the person who said it is the
sentence can be similar to "I've just found out that you lied to me"). individual named Peter. As a result, the proposition as a whole would
Second, assuming a discourse situation in which the fact that the speaker cease to be true and, if believed by the addressee, would constitute false
was lied to is indeed shared kno\vledge, the presuppositional status of the information.
complement clause is nonetheless different in each of the three sentences. \Vhat is interesting from the point of view of information structure-
In (a) it is presupposed that the addressee lied to the speaker (and and what further distinguishes a pragmatic from a semantic analysis of
asserted that the speaker didn't realize that fact at some earlier point in the presuppositional structure of this sentence-is the pragmatic status of
time); in (b) it is presupposed that someone lied to the speaker (and the NEGATION of the answer, i.e. PETER didn't say it. from the point of view
asserted that the speaker didn't realize that that person is the addressee); of two-valued logic, if the proposition expressed by the affirmative version
in (c) it is not only presupposed that the addressee lied to the speaker, i.e. of the sentence is false, its negation must be true, and that is all there.is to
that both speaker and hearer kno\.1.r this fact, but the sentence also evokes say. However it is obvious, from a communicative point of view, that this
the assumption that this presupposed proposition was recently touched negative sentence, though true, would normally be inappropriate as an
upon or "activated" in the conversation. This is shown by the fact that answer to the question "\Vho said that?" By its prosodic structure:(in
one can imagine a situation where the con1plernent clause in (2. ISc) could particular the lack of pitch prominence on some element in the verb
be replaced by an unaccented anaphoric pronoun (/didn't REALtZE that), phrase, see Section 5.6) the sentence PETER didn't say ii evokes the
while such a substitution \vould be impossible in ta) or (b). This last pragmatic presupposition underlying another question, i.e. the question
distinction, which I mentioned earlier, bet\veen propositions touched "Who didn't say that?", i.e. it pragmatically presupposes that one or
upon or not touched upon in preceding discourse is of n1ajor importance several individuals did NOT say a particular thing (and it asserts that Peter
from the point of view of the 1nfom1at1l)D structure of sentences. The is among these individuals). This, however, is not the presupposition
relationship between presupposition ;ind act1v.1tion \~·ill be analyzed in evoked in the original question "\Vho said that?'', hence the striking
Section 5.4.3. inappropriateness of the ans,,..,·er. This i111portant fact of communication is
The semantic observation concerning tht lfLllh-cond1tional stability of unaccounted for in the logico-semantic view of presupposition.
presupposed propositions unJer J1fferen..:e~ 1n pulcinty or inudallly in the It has been observed that in natural language negative sentences are
non-presupposed portion of a sentence c,1n he easily accommodated ordinarily uttered only if the speaker assumes that the addressee believes,
-
et hrfonnation
- -- -~ - -:: -

or at ltast entertains the possibility, that the corresponding affinnative

~- - - --

The pragmatic accon1modation of presuppositional structure

objection against the then prevalent view of presuppositions as purely

65

...,len<:e is true (sec Givon 1975b, Gazdar 1979:67, Horn 1989: Ch. 3). semantic phenomena, it conflicts, at least tenninologically, with the
For example, when l answer the question Hott• H'as your afternoon? with definition of pragmatic presupposition in (2.12) above. lnasmucb as
the statement /took a nap. my ans\ver leaves open whether I assume that presuppositions are evoked via lexicogrammatical structure it is justified
the bearer believes that I normally nap in the afternoon. But if I answer I to say that they are indeed properties of linguistic expressions, including
didn't take a nap, my statement normally does evoke that assumption. sentences. However, instead of saying that linguistic expressions (whether
While the positive answer is unmarked "'ith respecl to this presupposi- words or constructions) "have certain presuppositions," 1 will say that
tional feature, the negative answer is marked. It is interesting to notice they have PRESUPPOSlTIONAL STRUCTURES. These presuppositional struc-
'that the above observation does not apply to such "narrow-focus" tures, which are used to EVOKE speaker presuppositions, must be matched
sentences as our example P£TER didn't say it. This sentence does not with PRESUPPOSITIONAL SITUA"tlONS, i.e. the actual presuppositions of
conjure up the presupposition that Peter did make the remark in question interlocutors in discourse situations. Presuppositional structures are then
but rather that someone did not make this remark. grammatical conditions on the appropriate use of ,.,.ords and construc-
To summarize the foregoing observations about the differences tions in given discourse situations.
between pragmatic and semantic presupposition, not only is it the case
that a proposition which in the logico-semantic view counts as
presupposed may count as asserted in the pragmatic vie\v, but one and
2.4 The pragmatic accommodation or presuppositional structure
the same proposition expressed by one and the same complement clause
may or may not be pragmatically presupposed, depending on the context The notion of presuppositional structure as the lexicogrammatical coding
ofutterance. 19 In most cases, differences in pragmatic presupposition will of pragmatic presuppositions entails that presuppositions are not merely
correspond to differences in GRAM~fATICAL FORM, whether prosodic, as in a matter of the assumptions of speakers and hearers in a discourse; they
example (2.15), or morphosyntactic, as shown in later chapters. When are also a matter of GRAMMAR and of the LEXICON. And as stated at the
presuppositional differences are not grammatically expressed but merely beginning of this book, it is only to the extent that the mental states of
compatible with the form of the sentence, the sentence or clause evoking speakers and hearers are renected in linguistic form that they are relevant
the pragmatic presupposition will be said to be UNMARKED for the given to the study of infonnation structure. The theory of infonnation
presuppositional feature (as e.g. the that-clause complement in (2.1 Sa) structure presented in this book involves the assumption that a great
atiove; see also examples (2.16) through (2.18) below and discussion). number of simple or complex expressions in natural languages have
While I do not claim to have done justice to a semantically oriented presuppositional structures in the sense just discussed. i.e. that the
view of presupposition, I do hope to have shov·.'n that it is not the relationship behveen these expressions and the presuppositional situa-
semantic but the pragmatic notion which is relevant for information· tions in which they can be used appropriately is regulated by grammatical
structure analysis. With Stalnaker (!973, 1974, 1978). l believe that •'the convention. This assumption is empirica!ly justified in Lambrecht (in
basic presupposition relation is not bet,veen propositions or sentences, preparation), "·here I analyze the presuppositiona\ structures of a number
but between a person and a proposition" (1973:447), or, perhaps more of compk·x grammatical constructions in French
appropriately. between two persons and a proposi1i0n Stalnaker Jn this section, I \\'OU!d like lo discuss a number of phenomena t.1.'hich
emphasizes that speakef'j, not sentences, have presuppositions and he seem to contradict the notion of presuppositional structure as a direct
suggests that, instead of saying that a sentence "has" <:l presupposition, and conventional association bet,veen a grammatical fonn and a
linguists ought to say that it REQUIRES a presupposition, \\•1thout 1Nhich it presuppositional situation. These examples are all instances of the
cannot be used appropriately ( 1973:45 I J conscious or unconscious E'\PLnlTo\TION of presuppos1t1ons for special
Stalnaker's statement that presuppositions belong to speakers rather communicative purpo<;es. I ,~·11! c0nclude that presupposi1ional structures
than sentences requires a proviso. While this ~!.:Hemen\ cxrresses a valid are indeed inherent pr0perties of linguistic expressions and that such
 The pragniatic accon1modatiot1 of presupp ositiona / structur e 67
ln}Or1na1io11
l
using an expression which requires this situation, then presupp ositiona
apparen t counterexan1ples can be accounted for in terms of a general IONS,
struc,tures must indeed be INHERENT PROPERTIES OF LINGUISTIC EXPRESS
cognitive principle called PRAG~tATIC ACCOMMODATION. whether words or constructions.
In his discussion of pragmatic presupposition, Stalnaker (1973)
makes
in
t Stalnaker's views of presupposition, in particu lar the idea expressed
the following import ant observa tion concern ing certain apparen develop ed by Lewis (1979). Lewis observe s
the above quote, are further
violations of pragmatic rules or principles: ce which would be unaccep table
that it is difficult to think of an utteran
If, in a normal context, a speaker uses a sentence which requires
a only because it lacked a required presupposition. If the presupp osition
l
presupposition then by that very act, he does make the required evoked by some expression does not corresp ond to the presupp ositiona
presupposition. Whateve r his actual beliefs and assumptions, he
does tically supplie d by the
situatio n in the discourse, it is normally automa
the proposit ion for granted, and as if he osi·
speech participants: "Say something that requires a missing presupp
ACT AS IF he takes the truth of
assumes that his audience recognizes that he is doing so. (1973:451) existenc e, making
tion, and straightway that presupposition springs into
or of what you said acceptable after all" (Lewis 1979:172). The newly created
This possibility that speakers have of "makin g a presupposition," tic
n, by using a sentenc e that REQUIRE S it presupp osition will then become part of the set of pragma
creating a presuppositional situatio a·
o,vn type of convers ational presuppositions in the universe of discourse of that particu lar convers
is illustrated by Stalnak er with a well-kn
tion. Le\vis then formulates the following rule, which he _calls the RULE
OF
'she
exchange: "Someo ne asks of my daughter, 'how old Is he?' I answer,
an example seems to ACCOMM ODATION FOR PRESUPPOSITION;
is ten months old'" (1973:449). At first glance such
e. Indeed if the
contrad ict the notion of presuppositional structur If al lime t something is said that requires presupp osition p to be
be an accept.able, and if P is not presupposed just before t thcn-ce
teris
presuppositional structure of a linguistic expression is taken to existence
pragma tic certain limits- presupp osilion P come; into
inheren t propert y of that expression and if we define paribus
at t.
and within
und
presupp osition s as gramm atical evocati ons of shared backgro
assumptions of interlocutors, then how can the answer she is ten months In what f_ollows I v.•ill present a few examples illustrating this rule,
using
old be correctly underst ood, given that the presupposition required
by the e is relative ly well under·
constructions whose presuppositional structur
for
pronou n she (i.e. that the referent is female) is NOT the one taken stood. 21
of
granted in the question? Let us first consider the case of the presupp osiliona l structur e
a
The answer is that by using the pronou n she the speaker CREATES adverbial clauses involving such adverbi al conjunc tions as when, after,
which differs from the one
presuppositional situatio n in the convers ation before, because, since, airhough, etc. Here is a simple example:
newly
the addressee took for granted at the tin1e of her question. This
n can then be used as the required (2.16) A: \Vhat did you do before you sat dov.·n to eat?
created presuppositional situatio
actual
backgro und for the assertion about the age of the child. The B: (Before I sat down to eat) I washed my hands.
pronou n remain s unchan ged
presup positio nal STRUCT URE of the
point could be made, mutatis In speaker B's rt:ply, the proposition that B sat down to eat which
through out the conversation. The same osed
e appears in_ the fo~ of a dependent clause, is pragmatically pres~pp
mutand is, about the use of the definite noun phrase 111y car in exampl be omitted altogeth er without influenc ing the
\vhich no one in the (to the point that It could
(l.l) (My CAR broke do11·n), i.e. in a context 1n e). The presupp ositiona l status of the
fact interpretation of the sentenc
audience could be expectt:d to kno\\' that the speaker had a car. The left
the use of she in proposition expressed in the main clause, on the other hand, is
that the latter example seems less striking 1n L(IOlC,\t than have known as a fact that B washed his
that some unspecified. Speaker A may well
Stalnak er's exampl e may be explain ed by :.1ssun1ing such knowle dge is irreleva nt in this
hands at some point in time, but
presuppositional structures are ea~ier to e:-;.pll1tt than others, for reasons at the
nt is c_ontext. Speaker A clearly did not know that B washed bis bands
which I cannot go into hert:. \\'h•l.l L•)unts fpr the present argume
20
testing B's sincerit y or
a nL'\\. presurr• )<;ition al )ituat1o n n1erely by time before he sat down to eat, unless he was
this: if a speaker can CREATE
............................................. ._------------~~~

QI flr/omtation The pragmatic accommodation of presuppositional structure 69

memory. 22 The assertion in B's reply consists in establishing a relation not quite true, the reader would understand that it is the proposition
between a presupposed proposition and a proposition whose presupposi- expressed in the main clause, not in the be/are-clause, whose truth is
tional sta!Us is left open. being challenged. The explanation for the appropriateness of {2.18) is
While in (2.16) the presuppositional status of the main-clause provided by Lewis" rule of accommodation for presupposition. By the
proposition "B washed his hands" was unspecified, it is clearly act of using the clause which required the presupposition, the writer
specified. as presupposed, in the followi~g syntactically different CREATED the presupposition in the reader's mind and made it available as
cumple: a background for the assertion in the following main clause.
The phenomenon illustrated in (2.18) is not restricted to literary
_(2..17) A: When did you wash your hands?
discourse. Consider Che two English adverbial conjunctions because and
B: (I washed my hands) before I sat down to eat. since. Both indicate a causal relation between two propositions, but they
differ from each other in their presuppositional structure. As a rough
.ln (2.17) the main-clause proposition that B washed his hands is shared characcerization of this difference let us say that the presuppositional
knowledge at the time the question is asked and a fortiori at the time B
structure of since is such that the proposition expressed in the clause
answers it (again to the point that it could be omitted). The fact I am
which it introduces can be taken for granted in the reasoning process that
interested in here is that the proposition expressed in the before-clause in
links this proposition to the proposition expressed in the main clause. 24
B's reply has not itself ceased to be pragmatically presupposed in (2.17),
Because, on the other hand, does not require such a presupposition.
even though it appears in the same position and with a similar prosodic
While since is marked for the presuppositional feature in question,
structure as the main clause in B's reply in (2.16), whose presuppositional
because is unmarked in this respect. The basic difference is clearly
status was unspecified. As in (2.16), the knowledge that B sat down is
illustrated in question-answer pairs such as the following (the# symbol
assumed to De shared by the interlocutors. What has changed is the TOPIC
indicates unac.ceptability on the discourse level):
and FOCUS distribution in the sentence as a whole, i.e. the relation between
the presupposed proposition and the rest of the proposition, an issue to (2.19) A: Why did you hit him?
which I will turn in later chapters. B: Because he insulted me I #Since he insulted me.
The two examples above show that the propositional content of before-
clauses is regularly interpreted as being pragmatically presupposed, It is clear from the word why in the question that speaker A does not
independently of the discourse context. Such adverbial clauses may know (or purports not to kno\\•) the reason for speaker B's action. The
tb.erefore be said to be grarnrnaticaJJy 1.fARKED with respect to their difference in acceptability be!lveen the tn'o ans\1;ers sho"'·s that the
presuppositional structures, while main clauses tend to be UN,,.fARKED proposition "because P" can be used to make an assertion while the
with respect to the presupposition-assertion contrast. 23 But consider now proposition "since P" cannot. Now consider the following dialogue:
the following imaginary beginning of a short story:
(2.20) A: Where are you going on vacation this summer?
(2.18) Before I moved to Switzerland l had never setn a Rolls Royce. B· \Veil since my wife can't take more than tv.·o weeks off, we're not
going to Europe this time
Since (2.18) is assumed to be the first sentence of the story, the reader
cannot be expected to know that the protagonist moved to S\'1ltzerland at The use of the conjunction since \vould nonnally signal that speaker B
one point in his life. Nevertheless, the use of the ht:fore-clause is assumes that speaker A already knows that B's \vife has only two weeks'
appropriate and causes no difficulty of interpretation (at least not \vi thin vacation. However, in spite of this presuppositional requirement, the
the. given literary genre). The important fact here is that this does NOT answer in (2.20) is felicitous even if B assumes that A does in fact not
invalidate my claim concerning the presuppositional structure of be/are- have that knon'ledge. (In face it \vould be felicitous even if A did not
clauses. If the short story were to continue \vlth the sentence In.fact, that's know that B is married, in '.Vhich case the "existential" presupposition
 -· -- - -=---',-.::;;:;_

.. ,,, ............ .. The pragmatic accomniodarion of presuppositiona/ slructure 71

expressed in the possessive noun phrase 1ny wife also has to be (2.21) It was George Orwell who said that the best books are those which tell
accommodated). By the very use of the linguistic expression requiring you what you already know. -l
these pragmatic presuppositions, the speaker CREATES them and can use
them as a background for his staten1ent. In the discourse situation in which this sentence was uttered, it could'.
A somewhat special kind of exploitation of presuppositional structure, not be assumed as a fact known to the audience that some person ~d
which I take to be different from the cases of accommodation discussed made the statement expressed in the who-clause. The lecturCr May well'
above, is the one illustrated in the hackneyed ex.ample Ha~·e you stopped have thought that no one in the audience was familiar with this statement
bearing your wife? This kind of presupposition exploitation is discussed (nor with the author, for that matter). Strictly, then, it made no sense to:
by Clark and Haviland (1977) under the name of"bridging." The authors assert that the person who had made that statement was George Orwell.
observe that if I ask you the question '"Do you admit to \.,.ritlng this Nevertheless the utterance did not become unacceptable for lack of the
letter?" you are in trouble whether your reply is "Yes" or "No." In either required presupposition because the speaker could count on the
case you do not escape the presupposition that you did something bad. 25 audience's willingness to accommodate it.
What distinguishes this kind of devious exploitation of presuppositional There is a qualitative difference between (2.21) and the cases of
structure from the phenomenon of pragmatic accommodation is that it pragmatic accommodation discussed so far. As Prince (1978) and Bork.in
does not serve to indirectly convey information but merely to create a (1984) have shown, instances of it-clefts where the proposition expressed
fictitious presuppositional situation. by the relative clause is actually not assumed to be pragmatically
The pragmatic accommodation of certain presuppositional structures presupposed occur so regularly that it seems psychologically unmotivated
may to a greater or lesser extent becon1e CONVENTIONALIZED and to assume that in such cases the relative-clause proposition has to be
eventually GRA.~iMATICALIZED, a fact which makes the phenomenon more accommodated via Lev.:is' rule. Example (2.21) is not necessarily
complex than the preceding discussion suggests, but \Vhich also increases interpreted by the audience as an invitation to act as if the proposition
the range of facts it may be called upon to explain. It can happen that the expressed in the thar-clause were pragmalically presupposed in the strict
presuppositional structure of a frequently used construction_ is exploited I sense. Ralhe_r il _can be seen as a conventionally established indirect way
so regularly that it loses some of its force, sometimes resulting in a ne\v rof commun1cat1ng the content of that proposition. Because of the
meaning for the construction. regularity \11ith v.;hicb this convention is used, Prince and Borkin
As a case in point, consider the use of the ir-cleft construction in postulate t'.'.'O types of ir-cleft constructions, v.:ith different presupposi-
English. It is generally assumed that in order for this construction to be tional structures. Ho..,,•ever, given their fonnal similarity, it is important
used appropriately, the proposition expressed in the relative clause must to emphasize the relatedness of the t'.'.'O types. Indeed, in au instances of
be pragmatically presupposed, i.e. assumed by the speak_er to be known it-clefts, the proposition in the rhar-clause is GRAMMATICALLY MARKED as
to the addressee. This is what Borkin (l984:Append1x B) calls the factual and non-asserted. In what I take to be the original case, the
"grammatical meaning" of the it-cleft construction. (Typically, this proposition of the thar-clause is assumed to belong to the common
proposition is not only assumed to be kno\vn, but also to have been ground bel\\•een the interlocutors; in the second case it belongs to the
activated in the addressee's consciousne$S at the time of the utterance; see common ground bet..,,·een the speaker and some third party, and the
Section 5.4.3.) Thus, if I utter the sentence It's 111_1· keys that I losr. I addressee just happens not yet to be included in this party. The common
normally presuppose in my addressee the knO\\'lcdge that I lost syntax and the overlap in presuppositional structure between the two
something and I assert that the thing whic·h I lost is my keys. Hut as in types n1ake it possible, I believe, to interpret the second type as an
the previous cases, the presuppositional structure o..,if the .construcllon may extension of the first via conventionalized pragmatic accommodation.
be exploited for special con1n1un1c<J.t1,·e purr,1~e~. C1.in)1Jer the. fo..,)!low1ng Another, perhaps less controversial, case of the conventionalizalion
sentence uttered by a lecturer \() h1\ ~1ud1e11u;;> .Jt the beginning of the and gran1maticalization of pragmatic acco111modation is the so-called
lecture: emphatic do-construction in English. In ..,,,hat I take to be its original use,
 72 lnfomranon The pragmatic accommodation of presupposilional structure 13

llJis construction requires the presupposition that the truth of the time to counter in a principled way certain arguments raised against
proposition expressed by the sentence containing do was questioned in presuppositional analyses. The analysis of the presuppositional structure
the.immediately preceding discourse context. In this case, the verb phrase of a given expression or construction cannot be falsified simply by
following do is entirely unaccented. For example the statement pointing to examples in which an actual presuppositional situation does
not correspond to the presuppositional structure postulated by the
(2.22) I om pay you back
analysis. For e.i:;ample by pointing to examples like (2.18) or (2.20) one
is nonnally appropriate only if it was suggested in the immediately cannot purport to have demonstrated that the pragmatic analysis I
preceding discoune that the speaker did not pay the hearer back. or at cursorily suggested for before-clauses or for the difference between the
least if the possibility was raised that no payment was made. (In this use, conjunctions since and because is wrong. Moreover, by allowing for the
the do-construction has a function somewhat similar to that of the possibility of conventionalization and grammaticalization of pragmatic
particles doch and si in German and French.) Now, as with the accommodations we can account for systematic extensions and changes
constructions mentioned before, the presuppositional structure of the in presuppositional structures (see e.g. the analysis of the French ya-cleft
emphatic do-construction can be exploited via the principle of construction in Lambrecht 1986b, Chapter 7). In the case of the above-
accommodation, as e.g. in (2.23): mentioned it-cleft construction, this means that we do not have to
postulate two different.constructions whose presuppositional structures
(2.iJ) I was afraid to hit him; I oro INSULT him, though.
are the opposite of each other, but one basic construction and one or
In order to make the second part of (2.23) appropriate it is not necessary several cognitively motivated extensions of it.
for someone to have explicitly claimed earlier in the discourse that the
speaker did not insult the person in question. Rather, by saying I did
insult him the speaker merely suggests that someone could have made that
claim or might be tempted to make it. That the original pragmatic
presupposition is missing in the context of (2.23) is phonologically
expressed in the fact that the portion of the sentence following did is not
unac:c:cntcd, as it was in (2.22). Now consider the sentence in (2.24):
(2.24) I do hope that doggie's for sale.

In the context of the popular song in which this sentence occurs there is
no previous suggestion that the speaker did NOT have that hope. What
diJtinguishcs this last example from the previous ones is that there seems
to.·be no presupposition left at all. The use of the emphatic do-
constntction has become a conventionalized grammatical \vay of
expressing emphasis. The word do in (2.24) acts as a mere intensifier.
equiYalent to an adverb like really. so that I d(I hope is equivalent to I
mU!y hope. 26
, Leaving aside the issue of conventionalization, I \vould like to
emphasize the importance of the phenomenon of pragmatic accommoda-
tion for the theory of information structure. By recognizing the
~cal status of this principle of interpretation, \Ve are in a position
to-simplify the description ofpresuppositional stn1ctures and at the same
 Discourse referents 75

iJ is the proposition (or state of affairs) expressed in the antecedent clause

3 The mental representations of it does not appear full wizen opened. (The mood operator if does not enter
into the antecedent-anaphor relation.) While the representation of the:
discourse referents entity exists in the mind of the addressee prior to its linguistic expression
on the box, the representation of the propositional referent is created via;
the clausal antecedent itself (unless the state of affairs described is alreadYJ
known to the reader, in which case it is being reactivated). By the tim,e.
they are anaphorically referred to with definite pronouns, both constitute
discourse referents, which may serve as arguments in a predica~
3.1 Discourse referents argument structure. ~ -

Discourse referents are syntactically expressed in ARGUMENr (including

In the present chapter I will be concerned \Vith lhe nature of the
adjunct) categories, such as noun phrases, pronouns, various kinds of
representations of the referents of linguistic expressions in the minds of
tensed or non-tensed subordinate clauses, and certain adverbial phrases
interlocutors. Jn particular, I \\'ill be concerned with the changes \Vhich
(those that can be said to refer to the circumstances of a predication).2
these mental representations may undergo in the course of a conversation
They cannot normally be expressed in phrases which serve as PREDICATES.
and with the linguistic fonns which code these changes. The set of
Predicates by definition do not denote discourse referents but attributes
representations which a speaker and a hearer may be assumed to share in
of, or relations between, arguments. For example a finite verb phrase
a given discourse will be called the DJSCOURSE REGISTER. As indicated in
cannot play an argument role in a sentence, unless it is made into a
the remark in Section 2.1. I V.'ill tend to neglect the terminological (but
referential expression by being "nominalized" (in the sense of traditional
not the conceptual) distinction between REFERENTS and the ~tENTAL
grammar), i,e. by being stripped of its tense and person markings. This is
REPRESENTATIONS OF REFERENTS in a discourse. IL is primarily the latter
shown in the follov•ing contrasts:
that I will be concerned with in the follo\ving discussion.
Discourse referents may be either entities or propositions. 1 A (3.2) a. \Ve went to the movies yesterday.
proposition may acquire the status of a discourse referent once it is b. ft was a mistake.
assumed by a speaker to be known to the addressee, i.e. once it has been c. Our going to the tnories yesterday v.·as a n1istake.
added to the set of pragmatic presuppositions in the discourse register. d Going ro the 1no11ies yesterday v.·as a mistake.
The mental representation of such a propositional referent may then be e. •ive111 10 the moi·ies yesterday was a mistake.
stored in the register together with the representations of entities. Like
In (3.2b), the subject it refers to lhe propositional content of the entire
expressions denoting entities, those denoting presupposed propositions
preceding sentence (3.2a). The function of this pronoun is similar to that
may serve as arguments of a predicate. Propositional referents may be
of the second it in (3.1) above. Sentence (a), or the proposition denoted
expressed via various kinds of subordinate clauses (including non-finite
by it, appears in PRONOl'.t!NAL form in (b) and the proposition referred to
verb phrases) or they may be expressed by pronouns, as in the follo\ving
by the pronoun no\v has the status of a discourse referent. Sentence (a)
short text (from a cereal box):
c~n also be NOl'>HNA.LlZED, as in the subject expression our going to the
(l. l) This package is sold by wi.:1gh1. 11,11 by \1~lume .. II" 11 does not appear mo1'ies in (c). It is also possible to nominalize the verb phrase alone, as in
full when opened, it 1s bec<.luse l'onk·nt~ ha\t settled dunng sh1pp1ng and (d) (going co 1he rno1•ies). This non1inalization involves an understood
handling. subject, hence counts semantically as a proposition, hence may serve as
the argument of another predicate. Hoviever, as (e) shows, the finite verb
In (J.J), the referent of the firs1 1t 1~ the cntlly Jesign,:lled by the phrase u·ent 10 the mo1·ies cannol itself funccion as an argument. The
antecedent NP rhis package, i r.: the ..:r.:real b.:i:-.. the referent of the second morphosyntacuc difference betv.1een non-finite and finite (or tensed)

74
 ~ trvntal repnsentations of discourse referents Identifiability 77

dames is another grammatical correlate of the fundamental commu· various categories mentioned above involve independent sets of concepts.
nicative distinction between pragmatic PRESUPPOSmON and ASSERTION. they will be shown to be related to each other as well as to the previously
Nominalizing a proposition is one way of marking it as non-asserted. discussed categories of presupposition and assertion.
A potential problem for my account of the difference in discourse·
referential status between arguments and predicates is the occurrence of
3.2 Identifiability
expressions with predicate morphology in topic position and function, as
e.g. in the Getman sentence Arzt ist er nicht "He's not a doctor" {lit. "A When a speaker wishes to make an assertion involving some entity which
doctor he is not"), used to answer the question /st er Arzt? "ls he a she assumes is not yet represented in the addressee's mind and which
doctor?" In the reply, the bare noun Arzt functions as an anaphoric topic cannot be referred to deictically, it is necessary for her to create a
expression, hence its denotatum must have the status of a discourse representation of that entity via a linguistic description, which can then
referent (see Section 4.3), even though it has the grammatical appearance be anaphorically referred to in subsequent discourse. The creation of such
of a predicate nominal (it lacks a determiner). That Arzt must indeed be a a new discourse representation for the addressee can be compared to the
referential argument expression is demonstrated by the fact that it could establishment of a new referential "file" in the discourse register, to
be replaced by the definite anaphoric pronoun das "that," as in Das ist er which further elements of information may be added in the course of the
3
nicht '"That he isn't." Nevertheless, both Arzt and das function as the conversation and which can be reopened in future discourses.
non~subject complements of the copula ist, i.e. they correspond to the To account for the difference between entities for which the speaker
traditionaJ definition of predicates, except for their position. Similarly assumes a file has already been opened in the discourse register and those
problematic for my account is the occurrence of anaphoric pronouns for which such a file does not yet exist, I will postulate the cognitive
referring to predicate adjectives, as e.g. in the French sentence Ca ii /'est category of IDENTIFIABILITY, using a term once suggested by Chafe ( 1976).
-ibat' he is" (lit. "That he is it") used as a reply to Est-ce qu'il est Chafe observes that to designate referents for which a representation
intelligent? "'ls he intelligent?", where both the free topic pronoun pa and exists in the addressee's mind the term "identifiable" is preferable to the
the bound direct object pronoun l(e) seem to refer to the predicate sometimes suggested terms "kno\vn" or '"familiar." As we shall see, what
inteOigent. I must leave this issue unresolved here. counts for the linguistic expression of the cognitive distinction in question
Two infonnation-structure categories will be discussed in this chapter. is not that the addressee knO\\' or be familiar \\'ith the referent in question
The first is IDENTIFIABILITY, which has to do with a speaker's assessment (a newly opened file may contain no more than a name) but that he be
of whether a discourse representation of a particular referent is already able to pick it out from among all those \\'hich can be designated with a
stored in the hearer's mind or not (Section 3.2). The second is ACTIVATION, particular linguistic expression and identify it as the one which the
which has to do with the speaker's assessment of the status of the speaker has in mind.
representation of an identifiable referent as already "activated", as
merely "accessible," or as "inactive" in the mind of the hearer at the time
3.2.1 lder11~(iahiliry and presuppnsir1011
of the speech act (Section 3.3). At the end of the chapter, I will introduce
the categories TOPIC and FOCUS, which have to do \Vith the pragmatic roles The distinction bet\veen identifiable and non-identifiable referents is
which rd'erents with given identifiability and activation properties can conceptually related to the distinction between pragmatically PRE.SUP·
play within propositions. While identifiability and activation are POSED and ASSERTED propositions. A presupposed proposition is one of

categories of MEMORY and CONSCIOUSNESS, having lo do \Vi th the assumed \Nhich the speaker and the hearer are assumed to have some shared
STATES of the mental representations of discourse referents in the minds of kno\v!edge or representation at the time of utterance. An asserted
the speech participants at different points in a discourse, topic and focus proposition is one of \vhich only the speaker has a representation at the
are REIATIONAL categories, having to do with the pragmatic relations time of utterance. Similarly, an identifiahle referent is one for which a
between denotata and propositions in given conte'(tS. Although the shared representation alread~' exists in the speaker's and the hearer's
 Identifiability 79
l 1/t: '"t''UUI tt'presl!"lallons oj t11scvurse rejerent.s
children. (Compare also the remarks on the function of the deflllite article
mind at the time of utterance, while an unidentifiable referent is one for
in the discussion of pragmatic presupposition in Section 2.2.) Moreover1
which a representation exists only in the speaker's mind. Moreover, as we
it seems reasonable to assume that the representations of given entities in
saw earlier (examples (3.1) and (3.2)), when a presupposed proposition
people's minds are associated with sets of "propositions" corresponding
becomes a discourse referent and serves as an argument in another J
to various attributes of these entities. Nevertheless the referent of that '
proposition, it may be linguistically designaled with the same expression
noun phrase is mentally represented as an entity, not as a set of
type as an entity (i.e . .,..·ith a "persona!" or demonstrative pronoun). It is
propositions. For the purposes of the present study, I will therefore not
also \\'ell kno\\'n that in many languages the morpheme used as an
count existential presuppositions as pragmatic presuppositions in the
identifiability marker, e.g. the definite article or a demonstrative
sense of (2.12) and I will treat the notion of identifiability as a category in
determiner, is the same as, or is at least historically related to, the
its own right. The concepts of identifiability and of existential
subordinating morpheme used to introduce a nominalized sentence
presupposition do not necessarily exclude each other; they merely
(compare e.g. the Gem1an neuter definite article and demonstrative
represent difTerent theoretical perspectives on the same or a similar
pronoun das "the, that" with the complementizer dass "that").
phenomenon. I will return to this issue in the discussion of the
The relationship between presupposition and what I call identifiability
relationship between topic and presupposition in Section 4.3.
has long been recognized by philosophers, \Vho speak of the "existential
presupposition" expressed in. or required by, "definite descriptions."
However, there is an important difference in perspective between the 3.2.2 Identifiability and definiteness
logical (or semantic) viev.' of existential presupposition and the
An frnportant grammatical correlate of the cognitive distinction between
information-structure (or pragmatic) view of identifiability. This
identifiable and unidentifiable referents is the formal distinction made in
difference is analogous to the previously discussed difference between
many languages between DEFINITE and INDEFINITE noun phrases. The
meaning and information or between the truth of propositions and the
grammatical category of definiteness is a formal feature associated with
mental representations of states of affairs (see Chapter 2). In the domain
~ominal ·ex_pressions which signals whether or not the referent of a phrase
of information structure, the relevant property of an identifiable referent
is ~~~urned ~Y the speaker to be identifiable to the addressee. In man:i
is not that it is presupposed to exist, but that the speaker assumes that it
languages this category is regularly expressed via the ·contrast between a
has a certain representation in the mind of the addressee which can be
definite. and an indefinile ARTICLE or olher determiners (typi~}r
evoked in a given discourse. For example if I use the expression the King
~ossess1ve or demonstrative). In other languages, definiteness, or rather
of France in an utterance, I signal to my addressee, via the form of the
its cogn~tive correlate identifiability, may be marked by other
expression, that I assume that she has some mental representation of the
grammalical means, such as word order, the presence or absence of'a
individual designated by that expression \Vhich allo\vs her to identify it as
nwneral, a case-marking particle, etc (see below). Certain languages
the one I have in mind. What \Ve have in com1non is not a presupposition
arguably have no grammatical category for the expression of identifia·
of existence but the mental representation of an entity. The question of
bility, for example Russian (Johanna Nichols, p.c.). This does of course
whether the individual exists or not is irrelevant v.'ithin the conversational
~at ~ea~ . that speakers of Russian have no CONCEPT of pragmatic
exchange (see Section 4.3).
1dentifiab1hty and could not signal it in some indirect way.
From the point of vie\v of natural language use. it seems counter-
It must ~e e~pha_s~zed that the correlation between the cognitive
intuitive to assume that the ment ..tl representations of the referents of
category of 1denttfiab1hty and the grammatical category of definiteness is
noun phrases like John or n1y children \)f rht· K111g of Franct' have the
at be_st an. in1perfect one. There is no one-lo-one correlation between
status of propositions in the minds ofspedkLrs and addressees. This is not
1dent~fiab11ity or _non-identifiability of a referent and grammatical
to deny, of course, that element~ ()r noun phrases n1ay evoke
defin1teneness or indefiniteness of the noun phrase designating that
presuppositions. For ex.an1ple. the pu:.'>t'\'_,J\e detcrrnint:r 111.r in the
referent. Obvious evidence for this lack of corr·espondence is found in the
phrase my children evokes th~ pre~urrr.:.->111on that the speaker has
 Tlw m~tal representations of discourse referents Identifiability 81
80

fact that the use of the definite and the indefinite article varies widely "specific indefinite NP .. is one whose rererent is identifiable to the
from language to language, in idiosyncratic and sometimes quite subtle speaker but not to the addressee. while a .. non·specific indefinite NP'" is
ways, while the mental ability to identify referents is presumably the same one whose rererent neither the speaker nor the addressee can identify at
for speakers of all languages. Moreover, languages which have the time or utterance. This is tantamount to saying that a non-specific
definiteness markers often differ with respect to the grammatical option indefinite NP is one which may have no rererent at all.
of not using them. Certain languages offer a three-way distinction The specific/non·specific construal or an indefinite noun phrase may be
between a definite, an indefinite, and a zero article. Such a three-way influenced in subtle ways by the modality or the predicate or which the
contrast is found e.g. in English and in German, but not in French, where noun phrase is an argument. For example if, in order to put an end to a
referential common nouns must ordinarily be accompanied by a drav.·n-out phone conversation, I want to tell my interlocutor that a
determiner. Moreover the types of nouns with which the three options certain obligation is waiting ror me, only the first or the two versions in
can be used are not the same across languages. For example, while (3.4) is rully appropriate:
English allows for a three-way distinction between the man, a man, and
man (as in Man is a dangerous anima(), German has only der Mensch and (3.4) a. I have to go to a meeting nov.·. It starts in five minutes.
b ? I'd better go to a meeting now. It starts in five minutes.
ein Mensch, prohibiting *Afensch. But German does permit such three-
way contrasts as die Grammatik, eine Grammatik, and Grammatik (as e.g.
Even though (3.4b) is intended to be more polite. it is a strange thing to
in Grammatik.is1 nichl seine Sltirke, "Grammar isn't his rorte"). French,
say. The non-assertive form I'd better strongly suggests that there is in
however, permits only /'homme and un homme, la grammaire and une
ract no specific meeting which I have to attend.
grammaire, normally prohibiting both *homme and •grammoire, except
In some languages, the semantic distinction betv.·een specific and non-
when the noun is used predicatively, i.e. non-referentially. 4
specific rererents of indefinite noun phrases has grammatical correlates.
An important semantic distinction having to do with identifiability
For example in French the l\~·o readings or sentence (3.3) above are
which has no direct correlate in the grammatical definite/indefinite
fonnally distinguished ir a restrictive relative clause is added modirying
contrast is that between SPECIFIC and NON-SPECIFIC rererents or indefinite
the noun !il're ' book," as in (3.5):
0

noun phrases. It is often said that in an English sentence like

(3.3) I am looking for a book. (3 5) a. Je cherche un livre qui csr rouge. "'I'm looking for a book that's red ...
b. Je cherche un livre qui soir rouge "J"m !ooktng for a book that's red."
the indefinite NP a hook can be either specific or non-specific (i.e. can
have either a specific or a non-specific referent), depending on whether In (3.5a) the indicative n1ood of the verb of the relative clause indicates
the.speaker is looking for a particular book or whether "any old book" that the rererenl or the NP is specific, i.e. that the speaker is looking ror a
will do. The semantic difference between specific and non-specific particular red book v>hich exists but \\•hich she assumes the addressee
indefinite NPs can be made explicit in anaphoric contexts. Ir the cannot yet identiry. \vhile the subjunctive mood in (3.5b) indicates that
referent is specific, the anaphor must be a definite pronoun or noun the rererent is non-specific, i.e. that the speaker \\'Ould like to find a book
phrase. For example the speaker in (3.3) could go on to say I found it or I \vhose color is red, but or \vh1ch there n1ay not exist an instance in the
found the book l was looking for (but not I found a book or J.fnund one). If given universe or discourse. The correlation bet\\'een indicative mood and
the referent is non-specific, the anaphor must be an indefinite expression. specificity on the one hand and subjunctive mood and non-specificity on
In this case, the speaker might go on to say I.found one or I jOund a book the other is a result or the diO-erent semantic functions of the two moods.
(but not /found it or I found the book). Notice that in the anS\\ er I found a
1 \Vhile the indicative treats the relative clause pr0pc1sition as a matter or
book the referent has necessarily become specific and must be ract, the subjunctive marks 1t ;1s being '>UhJeCt to 1ncert1tude or doubt. The
subsequently referred to as ir or the book. One \vay of describing the use of the suhjunctive in the relative clause 1s rnotivated by the racl that it
specific/non-specific distinction in pragmatic tenns is to say that a is not possible to a'>-;1gn \v1th certitude a property (e.g. redness) to
 4 ............... tLy•L.><.l••U•lV/j,) VJ Ul.)LVUl.)t' rt:;t:rt:tll!i Jdentifwbility 83

something which may not exist, hence the necessarily non-specific To complicate matters further, the DEFINITE article, which is normally,
interpretation of the indefinite noun phrase. 5 used to designate specific identifiable individuals out of a particular~
What is expressed via mood variation in French may be expressed via can sometimes also be used with noun phrases that refer generically to the
word order variation in colloquial Gennan: whole class. For example, if, talking about a certain zoology student, I
say
(3.6) a. lch suche ein Buch, das rot isl. ··rm looking for a book that's red."
b. Ich suche ein Buch, das ist rot. "'I'm looking for a book that's red." (3.8) She is now studying the whale.
I can either mean that she studies the species "whale.'' or I can mean that
The standard version in (3.6a), which has the verb in final (subordinate
she studies a particular whale, for example one which got stranded on a
clause) position in lhe relative clause, has both the specific and the non-
specific reading. The colloquial version in (b), however, which has the beach, and which has become identifiable because of its prominence in
verb in second (main clause) position, has only the specific reading. The the real world situation. 7
fact that main clause word order only yields the specific reading is a Yet another cognitive distinction which cross-cuts the defmite/ .1
consequence of the correlation between main-clause status and ASSERTION indefinite contrast and which also varies from language to language
(cf. Section 2.4). In (3.6b) it is asserted that the object designated by the involves the coding difference between specific unidentifiable referents
pronoun das is red. For such an assertion to make sense, the existence of which are meant to become topics in a discourse and those which play
the object must be taken for granted. Hence the necessarily specific only an ancillary narrative role. The difference, in colloquial English,
reading of the noun phrase. 6 between the phrases this guy and a guy in the sentence
A grammatically indefinite noun phrase may have yet another (3.9) I met {this/a guy} from Heidelberg on the train.
semantic value, as in the sentence
is an instance of this discourse-pragmatic distinction (cf. e.g. Prince 198lc
(3.7) A book is a useful thing to have in a doctor's waiting room. and Wald 1983). By using the phrase this guy, the speaker signals her
intention to add further information about the person in question, while in
where the indefinite noun phrases a book and a doctor( 's) are said to be the version containing a guy such an intention is not expressed. The
GENERIC, meaning that their referents are either the classes of all books or morphologically definite noun phrase this guy is thus in fact "seman-
doctors or perhaps some representative set of members of these classes, tically indefinite" in the sense that it designates a not-yet-identifiable
but not specific or non-specific individuals. Since such noun phrases discourse referent, which in other languages (e.g. German and French)
merely require that the addressee be able to identify the semantic class could only be expressed in the form of an INDEFINITE noun phrase.
designated by the lexical head, generic indefinite NPs may be said to have Following Prince and other authors, I will refer to the determiner in
identifiable referents, further weakening the correlation between the question as "indefinite this" and l will categorize noun phrases
formal category of definiteness and the information-structure category of containing it as indefinite.
identifiability. That the referents of generic indefinite noun phrases must In certain languages the presence vs. absence of a NUMERAL expression
be considered identifiable is confirmed by the fact that they may be in association with the noun can have a function analogous to the
anaphorically referred to either \Vith anoth!!r indefinite NP or \Vith a function served by the contrast between a and this in English. In
definite pronoun, without a clear difference in interpretation. For languages with numeral classifiers it is often the case that the noun phrase
example after uttering (3.7) I can go on to say A book is also so1nething \Vhich is preceded by a classifier is marked as topical for subsequent
easy 10 carry o.round as \\'ell as It is u/s(I 1onk'thi11g easy to carry around. discourse (cf. e.g. Downing 1984:Ch. 7 for Japanese, Hopper 1986 for
This possibility distinguishes genenc 1nJefinites from indefinites with Malay, and Chaofen 1988 for Chinese). In other languages, the same
specific or non-specific referents \\'hich, J5 we S.ll\'. permit only one or the distinction is expressed via the contrast between the presence or absence
other kind of anaphoric expression_ but not both. of the numeral one, as in the case of Latin unus (\Vehr 1984:39fI), Turkish
- 84 7Tw f!Mfttal upresentotions of discourse referents

6IP (combined with the accusative case suffix on the noun, see Comrie
Identifiability

nevertheless identifiable TO A DEGREE by being a member of an identifiabJe

85

1981:128 and example (3.11) below), and Hebrew exad (Giv6n 1983:26). set (the three windows of the room). This may explain the coocurrence of
In ract. this function of the numeral one is attested in English, as e.g. in I the definite and the indefinite article. 8 A related case is that of the
.Jaw this one woman or I was introduced to one John Smith. German correlative expression der eine-der andere "one of them-the
The theoretical distinction between grammatical definiteness and ' other one," in which a definite article (der etc.) precedes an indefinite
cognitive identifiability has the advantage of enabling us to distinguish "pronoun" (einer, eine, etc.) to indicate unidentifiable members of
between a discrete (grammatical) and a non~discrete (cognitive) category. identifiable sets. The same correlative expression exists in French (l'un-
While the definite/indefinite contrast is in principle a matter of yes or no, l'autre), where it can be used either as in German or, with different
identifiability is in principle a matter of degree. Referents can be assumed syntax, to express the reciprocal meaning of "one another."
to be more or less identifiable, depending on a multitude of psychological A particularly rich system of grammatical contrasts coding degrees of
factors, but articles cannot be more or less definite (but see below). The (un)identifiabllity is found in Turkish. 9 Consider the data in (3.J J):
differences in the grammatical marking of definiteness among those
(3.11) a. Ahmet Okllz - U aldi
la.aguages whose grammar codes this category should perhaps be seen as Ahmet ox . ACC boughl "Ahmet bought the ox ..
reflections of different language-specific cut-off points on the continuum b. Ahmet bir Okfu. - 0 aldi
of identifiability. The fact that grammatical definiteness is a (relatively) Ahmel one ox· ACC bought "Ahmet bought this (one) ox"
'"arbitrary" category with (relatively) unpredictable cut-off points across c. Ahmet bir OkGz a\di
languages may account for certain difficulties in second language Ahmet one ox bought "Ahmet bought an ox"
d. Ahmet Okiiz aldi
acquisition. The difficulties which Russian speakers encounter in the Ahmet ox bought "Ahmet bought an ox"
acquisition of definiteness in e.g. English or German are notorious, but
serious difficulties also often arise for speakers who pass from one to In (3. l la) the ox is presented as identifiable, via th.e accusative case
another of two relatively similar systems of definiteness marking, such as marker (here -il), which could also be called a definiteness marker. In (c)
those in German and English. it is presented as unidentifiable, via the numeral bir "one" before the bare
Even though grammatical markers of definiteness are normally noun t/kiiz. Jn (d) it is also marked as unidentifiable, but via the bare
indicators of two-way distinctions and cannot mark degrees, there are noun alone. The contrast bet,veen (c) and (d) 1s reminiscent of the
interesting formal hedges between definiteness and indefiniteness specific/non-specific contrast. 1n (dJ the noun 1s unmarked for number,
marking, which seem to stem from a psychological need for the hence entirely non-specific: the sentence would be appropriate in a
grammatical expression of intermediate degrees of identifiablity. As one situation where Ahmet has bought one or more oxen, somewhat like in
instance of such hedging we can mention the above-mentioned three-way the English sentence "Ahmet did some ox-buying," ·.vhere the referent
distinction between a definite, an indefinite, and a zero article in English has lost its individuality via "incorporation" of the noun into the verb.
and German. Another instance is the French expression l'un( e) "one (of Particularly interesting for the present argument is example (b), in which
them)" (NOT "the one"), in which the definite article le, la serves as a the NP is hoth marked as ··definite, .. via the case marker, and as
determiner in a noun phrase whose head is the indefinite article (and "indefinite," via the numeraL The force of (b) 1s similar to that of (3.9)
numeral) un(e), as illustrated in (J.10): above, i.e. the di!Terence bet,veen (b) and (c) is that in (b) the referent of
the noun phrase is pragmatically salient in the context of utterance and is
(3,10) La cha.mbre avait trois fenetres; l'une d'elles kta1t ouverte. likely to be talked about in subsequent discourse, while in (c) the referent
~e room had three windO\\'S: one of them (lit "the one of them'') was
has no such salient role.
open.'
A common \Vay in which the relativity of cognitive iclentifiahility is
Although the window described here as being open is unidentifiable in the structurally reflected 1n grammar is the phenomenon discussed by Prince
sense that we are not told \\'hich of the three \vindo,vs is the open one, it is (\98\b) under the name of ··anchoring." Discussing the pragmatic
 • ,,~ ,,.~,.,.,, •<.Jll<..l<./ljUl/Ull.l v_/ tl/.'JtOllf.'Jt' ft'jt'rt'll/~·
Jden1!fwbi/i1y 87

differences between the indefinite noun phrases in the two sentences I got In (3.13). the name Pavel, which by virtue of being a proper noun phrase
a bus yesterday and the dri~·er n·as drunk and A guy I work n·ith says he
011 is necessarily "definite" (i.e. whose referent is necessarily assumed to be
knou·s your sister, Prince writes: 10 identifiable), appears both postverbally and preverbally. It follows that
the different English glosses of kniha in ex.ample (3. 12) do not constitute
Brand-new entities themselves seem to be of two types: ANCHORED and
sufficient evidence for recognition of a category of definiteness in Czech~
UNANCHORED. A discourse entity is Anchored if the NP representing it is
LINKED, by means of another NP, or "Anchor," properly contained in it, The difference between preverbal and postverbaJ position in Czech must
to some other discourse entity. Thus a bus is Unanchored, or simply correspond to some other grammatical distinction. which, in. thC
Brand-new, whereas a guy I 11·ork with, containing the NP/, is Brand- particular case of (3.12), happens to coincide with a difference in
new Anchored, as the discourse entity the hearer creates for this definiteness in English. I believe that the relevant contrast in (3.12) aDd
particular guy will be immediately linked to his/her discourse entity for
the speaker. (Prince 198ia:236)
(3.13) is the contrast between topic and non-topic, which, as we will see in
Chapter 4, correlates with, but cannot be equated with, the definite/
I will return to the notion of anchoring in Section 4.4.2, \vhere I will show indefinite contrast. 11 1
· .·'
that the various types of anchoring, \\•hich reflect degrees of identifiability Even though I will continue to use the familiar term "definiteness'~
of a referent in a discourse, may have effects on the acceptability of when referring to the language-specific expression of identifiability
sentence topics. known under this label, I prefer not to think of definiteness as a universal
Finally, the conceptual and terminological distinction betv;een a linguistic category. What is presumably universal is the COGNITIVE
grammatical category of definiteness and a cognitive category of category of IDENTIFIABILITI', which is imperfectly and non-universally
identifiability enables us to avoid a certain confusion which sometimes matched by the grammatical category of definiteness.
arises in discussions of the manifestations of definiteness across
languages. lt is not uncommon to find the tenn "definite NP" applied
3.2.3 The estahlishnient of identifiability in discourse
to some noun phrase in a language other than English only because the
English gloss of the sentence in which this noun phrase occurs contains a What are the pragmatic criteria according to which a speaker can assume
definite NP. It has been suggested, for example, that the distinction that a particular referent is identifiable by an addressee? In the clearest
between definite and indefinite NPs can be expressed in Czech (which is case, the referent of a noun phrase may be considered identifiable because
similar to Russian in this respect) via the difference between preverbal in the universe of discourse of the interlocutors or of the speech
and postverbal position of the NP (Kr<imsk.Y 1968). Consider the community as a whole there exists only ONE referent which can be
following examples: approp;iately designated with that noun phrase (see Chafe 1976:39). Such
(l.12) a. Kniha je na stole "The book is on the table" noun phrases with uniquely salient referents are expressions like mom,
b. Na stole je kniha "On the table (there) is a book" John, the President of the United Suites, the sun, etc. Each of these foul
expressions has specific referential properties which distinguish it from all
In (3.l2a), the preverbal NP k111ha "book" is appropriately glossed as the others, but they all have in common the fact that the individual
"the book,'' while in (3.12b) the postverbal k11ih111s appropriately glossed designated by the expression may be assumed to be uniquely identifiable.
as "a book." Ho\'lever. this difference 1n thi;: English glosses should not be The referential properties of such expressions often entail certain
taken as evidence for the existence of a gran1n1atical definiteness/ constraints on their grammatical coding, either by restricting the use of
indefiniteness contrast in Czech, i.:xpn:s:.cd ,·ia preverbal vs. postverbal the indefinite article or by excluding any kind of definiteness marking. as
position of the NP. To see this. consider thc-se ...1ddition;iJ Czech exan1ples in the case of unmodified proper names in English and many other
cited by Benes (1968) (the focu~ tn<.1rl\1ng 1n the English glosses is mine): languages. (The fact that some languages, e.g. Greek, do use the definite
(3_13) a Venku je Pa\'el ··ouh1Jc: 1;. I'\'- I I 1·4.1'LL'" <.iuto11.l<'., article with proper names is more evidence for the non-universal,
b. Pavel ie venku "P;nd 1~ n1 Ts1r,r'· language-specific character of grammatical definiteness.) Since referents
 Identifiability 89
88 nr~ mmtal representations of discourse referents
In the case of anaphoric reference, the status of some referent as
which are uniquely designated by some NP in the universe of discourse
idertGiiible~n.-be taken for granted because the referent was nieritioned
are particularly easy to identify for a hearer, they can also be assumed to
in previous -discourse~ Notice that once a previously unidentifiable
be, pragmatically more easily ACCESSIBLE (cf. Section 3.3) than other
rderent-~Ii introduced into the discourse register in the form of an
rcrerents. As a result, such NPs often exhibit exceptional behavior with
indefinite NP, it must from this point on be referred ..to_ with a..definite
~t to certain rules ~oveming the marking of topic expressions (see
noun phrase or a pronoUil. (As mentioned earlier, this requirement d?es
Givon 1983, and the discussion of French subject NPs in Lambrecht
not hold-for·non~specific and generic indefinite NPs.) For example, 1f I
1987a). Among the expressions with uniquely identifiable referents we
may also count GENERIC noun phrases, whether definite or indefinite. say to someone
Identifying the class of all entities which can be designated with an (3.14) I'm going to a meeting tonight.
expression is identifying a unique referent.
_. In the case of noun phrases which denote classes of entities rather than my interlocutor and l must later in the conversation refer to this
individuals, a particular referent may be assumed to be identifiable particular meeting with a definite description, i.e. among the following
because it has a salient status in the pragmatic universe of the speaker utterances only (a) or (b), but not (c) can be used:
and the hearer. I have in mind phrases like the kids, the cleaning lady, the (3.14') a. How long is {the/your} meeting supposed to last~
car. etc, as used e.g. by members of a family. The intended referents of b. How long is it supposed to last?
such phrases can be easily picked out of the respective classes because in c. #How Jong is a meeting supposed to last?
the universe of the interlocutors there is one salient referent or set of
(The symbol #in (c) indicates unacceptability under specific interpreta-
referents which is normally designated by such a definite noun phrase.
tion of the NP referent.) Note that this constraint holds even if the only
lit both case_s__1!1~!1-~i.~n_ed so_ far, i.e. in the case of NPs with unique
feature identifying the meeting in question for my addressee is the fact
referents and ofNPs whose refer~~t-~.ar~_un~quely idenilfi<lble because 0 -r that I am going to attend it. The fact that identifiability can be created
aomn;trarnd-knOWTOOge-betWeen the speake;· ~~d -th~~ ~-d-dres-see~ the
identifiability-of1he refetenr (and the use of the definite articlifiri English) through mere mention of a referent in the discourse, without any further
semantic specification, confirms our observation that identifiability of a
i&-d~..!!!_~_!_~cf t~a:t:-~rteferent is more or less permanently stored in the
meniory of the speaker/hearer and c.an be retrieved without difficulty at referent (and corresponding definite coding in English) does not
any-particutat time-;· given the appropriate discourse conteXt. -A rath~~ necessarily entail familiarity \Vith, or knowledge about. the referent.
dtfl'ettnt reason ·rot- assuming that a heaier. can identify a particular The identifiability status of a referent is normally preserved throughout
a discourse, and from one discourse to another. unless the speaker
referent obtains when a referent is saliently present in the external or the
assumes that the addressee has forgotten the existence of the referenl.
_internal discourse world, i.e. in cases of DE1cnc or ANAPHORIC reference.
Chafe ( 1976:40) mentions the example of the indefinite noun phrase a
In the case of deictic reference, a referent may be assumed to be
fetter used on page 13 of a novel and v.:hose referent is not mentioned
identifiable because it is visible or otherwise salient in the speech setting. I
again until page 118. \\'here it appears in the form of the definite noun
can deictically identify an entity by saying those ugly picrures or the
phrase the note. Once the referent \Vas introduced into the discourse
woman in the green hat oi-·er there, using the demonstratives those and
register, its identifiability status 'vas preserved over 105 pages. As Chafe
thue. whose interpretation is determined by the text-external setting in
remarks, "it '''ould appear that context or scene is all-important, and that
'W'hich the NPs are uttered. A referent can also be dcictically identifiable
definiteness can be preserved indefinitely if the eventual context in which
because it is ..inalienably possessed" or other.vise anchored in the
the referent is reintroduced is narrow enough to make the referent
individuality of one of the interlocutors, as in your le_(r leg or m)' sister's
~Cond ex-husband. In such cases, the referent is detennined by the identifiahle"
In the various examples of identifiability discussed so far, a referent is
semantic "frame" evoked by the possessive expression designaun.g the
caused 10 hecome identifiahle for 'vhat seem to be quite heterogeneous
interlocutor. 12
 Jdentl}zabilily ~I
lhe mental representauons o_I d1scuurse re_lerents
expression the doctor in this sentence refers to an unspecified individual
reasons. From a psychological point of view, the status of the
permanently established referent designated by the noun nrom seems out of a specific subgroup. This subgroup is not coextensive with the
rather different from that of the referent designated by the meeting whole category (thus precluding generic interpretation). nor is .. the
tonigh1, for which a file is opened only for the purpose of a particular individual designated by the NP entirely non-specific (thus precluding use
discourse and which may get permanently erased from the hearer's of the indefinite article). In cases like (3. J5}, semantic categories such as
memory at the end of the conversation. One may therefore wonder what specific, non-specific, or generic are not of much help. What explains the.
motivates the use of a single gramma!ical category (definiteness} for such occurrence of the definite noun phrase in (3.15) is the fact that the
apparently divergent instances. I believe that the common cognitive individual in question is identifiable as an element in a semantic frame..in
property which unites all instances of identifiability and therefore justifies this case the world of the clinic. 11, _, l · ·
expression by a single grammatical category is the existence of a cognitive A striking instance of frame-determined identifiability is discussed .by
SCHEMA or FRAME within which a referent can be identified. The concept Hawkins (1978:Ch. 3). Hawkins observes that under particular circum-
of "frame" is defined as follows by Fillmore ( 1982: l 11 ): stances the same referent, with apparently the same pragmatic status. can
be coded as a definite or as an indefinite NP, depending on whether it is
By the term "frame" I have in mind any system of concepts related in viewed as part of a cognitive schema or not. For instance (modifying
such a way that to understand any of them you have to understand the Hawkins" example slightly) in explaining the workings of a car to. an
whole structure in which it fits; when one of the things in such a
structure is introduced into a text, or into a conversation, all of the
entirely ignorant addressee, I can point to different parts under the hood
others are automatically made available. and say
(3.16) This is the air filter, this is the fan belt, this is the carburetor.
(See also FiJlmore 1976 and l 985a.) The frame within which a referent
becomes identifiable can be so broad as to coincide with the speaker/ i.e., I can use definite noun phrases for my explanation even though the
hearer's natural or social universe, accounting for the identifiability of the designated objects were not previously identifiable by the hearer. This is
referents of NPs like the sun or the President of the United States. It can possible because the various car parts are all indirectly identifiable as
be narrower, as the personal frame within which the referent of the elements of an already identified frame or schema, which is the car itself_
cleaning lady or the car becomes identifiable. Or it can be the physical However, if the same ignorant addressee wonders about some
environment in which a speech act takes place, making it possible to unidentified object on a shelf in my garage, I cannot say to him This is
jdent.ify the referents of such noun phrases as the u•o1nan over there or 1he carburetor but only This is a carburetor, because the object is not
those ugly pictures. Finally, the text-internal discourse \Vorld itself can be interpreted as an element of an already established frame.
such a cognitive frame. For example the referent of the NP the meeting An intriguing case of definiteness motivated by relations between
tonight in (3.19) is identifiable to the hearer by virtue of the frame of elements of a semantic frame is the phenomenon whereby the possessee in
reference established by the ongoing discourse alone, independently of a possessive noun phrase gets marked as definite even if both the
whether such a meeting actualJy exists or ,~·ill exist in the real world.JJ possessor and the possessee are unidentifiable in the discourse. For
The concept of frame-linked referent identification enables us to example I may say
account in a straightforward fashion for cerlain occurrences of definjte
(3.17) I met the daughter of a king
noun phrases which otherwise might see1n mysterious. One instance is the
use of definite noun phrases in contexts like the follo\ving: {in~tead of a daughter of a king) even if I assume that my addressee can

14 identify neither the king's daughter nor the king himself. This suggests
(3.15) Every time I go to the chn1.: the dc,ctur 1s son1eone dillerent.
that for the purposes of grammar an entity may be categorized as
Unlike the expressions the cle1.1ni11g luJy ()f the cur, \vhich are assumed to identifiable merely by virtue of being perceived as st.anding in a frame
designate just one specific individu..i.l for the speaker and the hearer, the relation to some other entity, whether this other entity is itself identifiable
 Acti~·arion 93
92 T1te mental representations of discourse referents
3.3 Activation
or not. Notice furthermore that (3.17) would be appropriate even if it
turned out that the unidentified king in question has more than one In my discussion of the knowledge or set of representa~ions ~hie? a
daughter. (On the other hand, if the king were already identifiable, the speaker judges to have in common with an addressee at a given po1nt 10 a
phrase the daughter of the king would not be appropriate if the individual discourse, J was assuming as a convenient simplification t~at ~hatever an
-bad more than one daughter.) The peculiar pragmatic structure of the NP addressee knows is something she can be assumed to be th1nk1ng of a~ the
in (3.17) is shown also in the synonymous alternative version I met a time of an utterance. Given the vastness of the knowledge stored I~ a
king's daughter, in which the phrase a king's acts as a possessive-hence
definite-determiner, even though it consists of an indefinite (genitive)
NP. The phenomenon in (3.17) is related to, but interestingly different
person's mind it is clear that this assumpti~n is not warranted. Knowing
something and thinking of something are different mental states. In order
for an addressee to be able to process the presuppositions evoked by an
t
I

from, the cases of pragmatic anchoring discussed by Prince. While in utterance it is not only necessary that she be aware of the relevant set of I
Prince's examples an unidentifiable referent is anchored in an already presupposed propositions but that she have easy access to these '
identifiable one, in (3.17) an unidentifiable referent is anchored in propositions and to the elements of which they are composed. In other
another unidentifiable one. One might call (3.17) an instance of words as Chafe has repeatedly emphasized (1974, 1976, 1987), the
"'ancborless anchoring" or "pragmatic boot-strapping. " 15 conve~ing of information in natural language not only involves
17
Interestingly, a grammatically definite complex noun phrase such as KNOWLEDGE but also CONSCIOUSNESS. The difference between these two
the one in (3.17)-whose definiteness is motivated by pragmatic boot- mental states has important grammatical consequences. A large part of a
strapping-may behave like an indefinite NP with respect to certain speaker's assumptions concerning the representations of referents. in the
grammatical requirements which many linguists take to be of a SYNTACTIC mind of an addressee at the time of an utterance has to do with the
order. Consider the so-called "impersonal" ii-construction in French. limitations imposed on the short-term memory of speakers and hearers. ;:

''
Like the "existential" there-construction in English, this French My description of these psychological phenomena is base? on Chafe's
construction is traditionally said to welcome only indefinite noun (1976, 1987) account of the interaction bet\.,.een consc1ousnes~ a~d
phrases, at least in the standard language (for exceptions, see Lambrecht verbalization. Certain differences bel\~·een my account and Chafes will
1986b, Section 7.5). Thus while the (a) example in (3.18) is grammatical, be discussed from case to case. Generally speaking, while Chafe
the (b) and (c) examples are not: emphasizes the importance of the cognitive states of concepts i~ ~he
hearer's consciousness at the time of an utterance, 1 will be emphas1z1ng
(3.18) a.nest entrC un roi. "There entered a king." the importance of the hearer"s \'{illingness and ability to model her state
b. •nest entrC le roi. "There entered the king." of consciousness according to the requirements expressed by the
c. •nest entre la fille du roi. "There entered the daughter of the king." presuppositional structures chosen by the speaker. A partial revision
d. II est entrC la fille d'un roi. "There entered the daughter of a king." concerning the concept of ··activation" discussed belo\.,. \viii be presented
at the end of the book (Section 5.7)
Example (d). however, is grammatical even though it contains a definite
noun phrase. Syntactically, the definite possessive NP in (d) patterns with
the simple indefinite NP in (a). Lafille d'un roi is thus both definite (from 3.3_1 The acti1'arion states t~( re_(erenrs
the point of view of morphology) and indefinite (from the point of view
Taking as his point of departure the idea ··1hat our minds contain very
or syntax). 16 While such facts pose a problem for an analysis of large amounts of knov.'\edge or information. and that only a very small
definiteness in purely morphosyntactic terms, they are easily e:i:plained if
amount of this infonnation can be focused on. or be 'active' at any one
definiteness is understood as the imperfect grammatical retlection of the
time." Chafe ( ! 987:22ff) argues that a particular ··concept'· may be in any
non-discrete pragmatic category of identifiability. one of three ACTIVATION STATES. \Vhich he calls ACTIVE. SEMl-ACTrYE (or
 Ac1;vation 95
•1 ••J•••••••""''J '"-' ...... , ...... ~ ''-J'-''-'"J

concepts are spoken with an attenuated pronunciation. The attenuation

ACCESSIBLE) and INACTIVE respectively. An ACTIVE concept is one "that is involves, at the very least, weak stress. Typically, though not always, .it
currently lit up, a concept in a person's focus of consciousness at a involves either pronominalization or omission from verbalization
particular moment." An ACCESSIBLE/SEMI-ACTIVE concept is one "that is in altogether. (1987:26)
a person's peripheral consciousness, a concept of which a person has a
The cognitive category "activeness" thus has grammatical correlates in
background awareness, but one that is not being directly focused on." An
PROSODY (phonological attenuation) and MORPHOLOGY (ptonon:µnal,
INACTIVE concept is one "that is currently in a person's long-term
inflectional, or zero coding). It also has correlates in syntax, bu~ ~~
memory, neither focally nor peripherally active." In the following
discussion I will refer to what Chafe calls here "concepts" as "(mental are not as easy to demonstrate (cf. Section 4.4.4 and in particular th.e
analysis of unaccented French pronouns in Lambrecht 1986b Section
representations oO referents," for reasons to be explained later on. 18
The psychological factors detennining the activation states of discourse 6.1). An unaccented referential expression is necessarily ass~ed.to
referents are thus CONSCIOUSNESS and the difference between SHORT-TERM have an active referent (barring certain cases of pragmatic accommoda·
MEMORY and LONG·TER~I ME~tORY. An item is active if it is "currently lit
tion to be discussed below), i.e. the prosodic evidence of "attenuated
up" in our consciousness, to use Chafe's expression, and activation pronunciation" is a sufficient condition for assumed activeness. of a
normally ceases as soon as some other item is lit up instead. It is possible, referent. (We will see, however, that attenuated pronunciation is .not a
for example, to use the unaccented pronoun she to refer to a particular
female referent only as long as that referent is the current center of
attention of the speech participants (or, as I will say later on, as long as it
necessary condition for activeness.) But the clearest evidence for ass~ed
activeness is no doubt the morphological evidence of PRONOMINAL CODING ..
with the possible exception of generic pronouns like English you, they oi
.
is one of the TOPICS under discussion). Once the attention of the speech German man and of certain deictic uses of pronouns. In the former case,
participants has shifted to another item, it is no longer felicitous to use the referents are so general ("people in general") that they may always_~
that pronoun to refer to that person. An inactive item, on the other hand, taken for granted and need not be activated. The special status of such
keeps its cognitive status for much longer periods of tin1e, independently generic pronouns is reflected in the fact that they can generally not be
of what is presently at the forefront of the interlocutors" attention. accented. In the latter case, activation may take place with the very
Depending on the nature of the referent, inactive status-and the utterance of the pronoun, often accompanied by a gesture. I can
possibility to use a definite noun phrase in languages like English-may felicitously say I want THAT without assuming that my addressee was
last for the period of a particular discourse, it may last for a stretch of previously aware of the object designated by the demonstrative pronoun.
time beyond a particular discourse, or it may last indefinitely (see the As hinted by Chafe at the end of the passage quoted above,.referents
remarks to that effect in Section 3.2.3). which have been activated and which in principle satisfy the condition for
What makes the different activation states of referents relevant for the unaccented pronominal or inflectional (or zero) coding are sometimes
NOT coded as pronouns but as unaccented lexical noun phrases. This
study of information structure is that they have formal correlates in the
structure of sentences. A short text example illustrating the various happens for example y,•hen more than one referent is activated at the
activation states and their formal manifestations v.'111 be discussed in same time and pronominal coding would lead to ambiguity. For example
in the sequence
Section 3.4 (example (3.27)). Further examples, illustrating in particular
the function of differences in prosodic prominence, V.'iil be discussed in (3.19) I saw John and Bill this morning. He's sick.
chapter 5. Concerning the \Vays ACTIVE referents -:an be expressed in a
sentence, Chafe writes: it is not clear which of the two activated referents, John or Bill, the
pronoun he is intended to designate. Barring pragmatic disambiguation
Those concepts Y•h1ch are already .;ct1>e for the sp~aker, and which the (e.g. John might be known as a sickly person so that he would most likely
speaker judges to be acl1\'e for the hi:an:r as well. are verbalized in a
be interpreted as referring to him), it would therefore be preferable, in the
special way, having rroperties which have often been discussed in terms
of"old" or "given" inform;.it1<:>n Tht· g..:n<:ral thing to say is that given given context, to refer to the one who is sick with a lexical noun (John or
-
t. 1' 71w- ffttlltal "pre$e'ttations of discourse referents

lllll).. There arc various other semantic and stylistic reasons for not
Activation

the speaker and the hearer than keeping an already established referent in
97

coding alrtady activated referents as pronouns, which I cannot discuss a state of activeness. As a result, it involves higher acoustic intensity and
here. Languages also seem to differ 'videly with respect to the possibility typically more phonological material. The iconic nature of intonation has
or. or tolerance for, non-pronominal coding of active referents. 19 been emphasized in much work by Bolinger, in particular in his 1985
Nevertheless~ in spite of various kinds of exceptions, the overall essay on "The inherent iconism of intonation," where we find the
correlation between assumed activeness and pronominal coding is following statement: "Suppose we take the obvious emotive correlation
extremely strong on the discourse level and has important consequences as basic: high pitch symptomizes a condition of high tension in the
for the structure of sentences. It can be sho,vn to play a major role in the organization, low pitch the opposite" ( 1985:99).
structure of the clause in spoken French (see Lambrecht 1986b). The Important though this iconic principle may be, there are severe
lbnnal category ..pronoun" is no doubt the best evidence for the limitations on its applicability. While it is true that the referent of a
grammatical reality of the information-structure category of "activeness." pronominal expression or of a nominal expression spoken with
-To summarize, assumed active status of a referent is formally expressed attenuated pronunciation is always taken to be active (again, barring
via lack of pitch prominence and typically (but not necessarily) via certain cases of pragmatic accommodation), it is NOT the case that an
pronoiil.inal coding of the corresponding linguistic expression. In the expression coding a referent which is assumed to be active is necessarily
terminology established at the end of Section 2.3, Jack of prominence and also spoken with attenuated pronunciation. In other words, weak
pronominal coding are to be seen as features of the PRESUPPOSITIONAL prosodic manifestation is only a sufficient, not a necessary condition
Sn.ucruRE of a given expression. The term "pronominal coding" will be for assumed activeness of a discourse referent. Under certain circum-
tibderstood in a very general sense, applying to free and bound pronouns, stances, constituents with clearly active referents, including anaphoric
inflectional affixes, and null instantiations of arguments. Any non-lexical pronouns, may receive prominence. Compare the tv.•o examples in (3.20):
eXpression of a referent counts as pronominal.
(3.20) a. I saw MARY yesterday. She says HELLO.
As for the formaJ expression of the INACTIVE status of a reft"rent, it is
b I Sa\\' MARY and JOHN yesterday. SHE says HELLO, but HE'S still
the opposite of that of active referents. Inactive marking entails ANGRY at you
ACCENluATION of the referential expression and FUlL LEXICAL coding.
An inactive referent cannot be expressed pronominally (again, with the The referent of the pronoun she is equally active in both sentences, but
possible exception of deictic pronouns). The grammatical correlate of the pronoun is prosodically more prominent in the second example. The
inactiveness is thus the coding of a referent in the form of an ACCENTED difference between accented and unaccented pronouns has often been
LEXICAL ·PHRASE. Even though the relation between accented and accounted for in terms of the notion of '"contrastiveness," which I will
unaccented ·and between pronominal and lexical coding is one of simple discuss in detail in Section 5.5. Suffice it to say here that in (3.20b)
opJjosition, we will see below that this relation is functionally prosodic prominence has a distinguishing, or disambiguating, function
«sYmmetr:ic, One of the members being marked and the other one which is different from the simple marking of an activation state.
triunarkCd With respect to the category of activation. Active referents may also be coded as lexical noun phrases with pitch
' -The relationship postulated by Chafe between activeness of a referent prominence, as shov.•n in (3.21), an example originally discussed by Kuno
and attenuated pronunciation" of the expression designating it, on the
0
(1972),
one hand, and between inactiveness of a referent and "strong
(3.21) (= Kuno's (1-5\)
pronunciation," on the other, may be called ICONIC, in the sense that
Q: Among John. ~1ary. and Tom. who 1~ the oldest?
there exists a direct correlation between different mental states and A· 10~1 is the oldest
differences in phonetic intensity or \VOrd length (pronouns tend to be
shorter than lexical NPs). Creating and interpreting a ne\v discourse The referent of the focus noun Tom in the ansv.·er is clearlv active, but
this noun is ~ilso clearly the prnsod1c peak of1ts sentence. According to
211
representation ofa referent requires a greater mental effort on the part of
 ._ •-••• ··~··-•'"'"••V••,.> v_1 .. •..>•V•••.J• '•)•'•"'-'
Activation 99

Kuno, the accent on Tom in the answer is due to the fact that the referent corresponding unambiguous marking for the status INACTNE, at least not

f of this noun is ··unpredictable" as an argument in the proposition, an

explanation which I will adopt and develop in a later chapter (Section
via prosody or morphology (certain important SYNTACTIC correlates of
inactiveness will be discussed in the section on referent promotion in
5.7). Let me say, for the time being, that an active referent is coded with Chapter 4).
prosodic prominence for reasons having to do with the marking of a A remark is in order about the status of ACCESSIBLE or SEMI-ACTIVE
RELATION between it and the proposition in \.vhich it occurs, rather than as discourse referents, According to Chafe, semi·active (accessible) referents
a result of its activation state in the disourse. I will return to the can be of two kinds. A referent ("concept") may become semi-active
distinct.ion between actjvation marking and focus marking at the end of ejtber "through deactivation from an earlier state, typically by having
this chapter and, in more detail, in Chapter 5. been active at an earlier point in the discourse," or it can become semi-
'' The preceding observations allow us to draw an important conclusion active because it "belongs to the set of expectations associated with a
concerning the interpretation of prosodic prominence in general. While schema." A schema is defined by Chafe as follows:

I the ABSENCE of prosodic prominence on a constituent necessarily indicates

active status of the coded referent or denotatum, the PRESENCE of
prominence has no analogous distinguishing function. The function of
the one is not simply the opposite of the function of the other. There is a
fundamental functional asymmetry between accented and unaccented
constituents. This is the asymmetry of MARKEDNESS. An unaccented
A schema is usefully regarded as a cluster of interrelated expectatiOD.S.
When a schema has been evoked in a narrative, some if not all of the
expectations of which it is constituted presumably enter the semi·'active
state. From that point on, they are more readily available to recall th3.n
they would have been as inactive concepts. (1987:29)

constituent is MARKED for the feature "discourse-active denotatum," As an example of schema-related accessibility, Chafe mentions the
while an accented constituent is UNMARKED with respect to this feature. expectations associated with the typical schema of an undergraduate

i The characterization of unaccented constituents as marked may seem

counterintuitive, since it is normally the presence rather than the absence
class, which includes the concepts "student,'' "instructor," .. teaching
assistant," "classroom," etc., all of which become accessible and can be
I of a feature that designates one member of a pair as the marked one. coded accordingly once the general classroom schema is lit up via
Moreover, if Bolinger's claim concerning the emotive correlation between mention of one of its components. Chafe's notion of a schema and its
high pitch and high tension is correct, interpreting high pitch as associated expectations, which he takes from cognitive psychology; is
• unmarked would entail that high rather than low tension is considered closely related to the Fillmorean notion of a SEMANTIC FRAME (see Section
the unmarked state of affairs in speech, a conclusion one may be 3.2.3 above). Various types of inferences that can cause a referent to be
reluctant to accept. However, there is nothing unnatural about such a accessible are also discussed by Clark (1977) and Prince (198la). Bes.ides
conclusion. The negative feature "absence of sound" can be phrased frame relationships, Prince mentions culture-based inferences involving
positively as "presence of silence." And in the use of language, silence is stereotypic assumptions and logical set relations (set to subset, set to
indeed the marked state of affairs. Thus the generalization concerning lhe number, number to set).
unmarked status of prosodically prominent constituents is to be I believe that it is necessary to add a third kind of accessible referent to
maintained. It will be shown to be of great importance in the the two mentioned by Chafe. These are referents whose accessible status
21
interpretation of focus prosody in Chapter 5. is due to their presence in the text-external world. For example, sitting in
An analogous generalization n1ay be made in the case of the an office room with a friend I might say Those pictures sure are ugly with
morphological contrast between PRONOMINAL and LEXICAL coding of a reference to some photographs on the wall which I assume my addressee
referent. A pronoun is marked as having an active referent; a lexical noun is not presently aware of but which I take to be easily accessible to him.
phrase is unmarked for the activation state of its referent. To sum up, Notice that to utter such a sentence it is not necessary to assume that the
while ACTIVE referents can be unarnbigu1..1usly 1narked as such, via absence accessible referent was in the addressee's "peripheral consciousness"
of prosodic prominence, or pronon1inal coding. or both, there is no (Chafe) at the time of the utterance. Such externally accessible referents
 - :n._..,,._ -oT.lisaw>< -.fn'O<U Activation 101

,._ • - Jliulllaa for a dtfiuilioo of ..,..,,Dility solely in b:rms of majority of subjects in spoken English have active or accessible b.ut not
• - badground • ............., (sec below). inactive referents. 22 Since subjects are normally the leftmost constituents
Aoce:5Sl.bility (semi-activeness) of a referent can thus be ascribed to in an English sentence, such facts point to the existence of a co~lat~on
three factors: deactivation from an earlier state, inference from a between left to right order in the English sentence and the act1vat1on
cognitive schema or frame, or presence in the text-external world. In states of the referents of syntactic constituents. I will return to the issue of
Jlte case of deactivation from some earlier active state in the discourse, [ the cognitive category "accessible" at the end of the next section.
W'iII call the accessible referent TEXTIJALLY ACCESS1BLE; jn the case of
access:tDility via inference from some other active or accessible element in
3.3.2 Principles of pragmatic consrrual
the universe of discourse I will call it INFERE1'ITIALL y ACCESSIBLE; and in the
case of accessibility due to salient presence in the text-external world I While the category of activation accounts for the relationship between
will call the referent SmJATIONALLY ACCESSIBLE. The two categories the assumed cognitive states of discourse referents and types of
·~tua11y accessible.. and "situationally accessible" correspond to the grammatical forms, it does not account for the principles of interpreta·
text·intetnal and the text-external world respectively, while the category tion whereby particular syntactic constituents are construed as designat·
~"'inferentially accessible" is neutral with respe<::t to this distinction: a ing particular referents. In addition to the Chafean account of the
n:ferent can be inferred from an element in the linguistic as "'ell as in the relationship between consciousness and verbalization, independent
extra·linguistic oontext. pragmatic principles of CONSTRUAL are required in order to explain how
Chafe's distinction between active, accessible, and inactive referents is linguistic expressions which code discourse referents with certain
:based on the idea that there are different types of mental effort or "cost" activation properties are INTERPRETED in particular discourse contexts.
~involved in ~e processing of mental representations. We should keep in In what follows, I will brieny discuss some of these principles of construal
mind. however, that from a strictly grammatical (phonological andjor on the basis of a number of examples which pose apparent problems for
morphosyntactic) point of view, only a binary distinction is justified, the activation analysis. These examples should not be read as evidence
namely the distinction between referents which are MARKED AS BEING against the activation approach itself but only as evidence for tbe
ACTTVE (attenuated pronunciation and/or pronominal coding), and those existence of additional principles of interpretation which must be taken
which are NOT so MARKED- As I pointed out earlier, accented non. into account within the larger framework of information structure.
pronominal constituents may have referents in any activation state, i.e. Example (3.19) showed that it is sometimes impossible to code an
-they arc unmarked for the feature "active referent." From the active referent pronominally because of the referential ambiguity
psychological point of view, there is no theoretical upper limit to the resulting from the presence of t\vO or more competing active referents
~ber and kinds of cognitive states which mental representalions may in the context. But in certain cases, the ambiguity may be resolved
·have in the course of a conversation. But from the point of view of through conte;i;tual semantic clues. In the following example, quoted by
grammar, only two such states are recognized and given the status of Dahl (1976), who attributes it to Lashley ( 1951 ), correct construal of an
. formal categories. This is not to say that the postulation of an unaccented pronoun, in spite of the presence oft\vo competing referents,
intermediary category "accessible" has no grammatical reality. For is made possible on the basis of the semantic content of a proposition
example. we will see in the discussion of lopic in Chapter 4 that the FOLLO\VING the utterance of the pronoun:
difference between accessible and inactive referents can have syntactic
(3_12) a. ( = Dahl's 8) Peter went to see Bill. but he was not at home.
consequences; in particular it can inOuence the position of a constituent
b. ( = Dahl's 9) Peter went to see B1lL but he had to return.
in the sentence or the choice of one rather than another grammatical
comtruction. Different syntactic constraints on the coding of inactive and It is clear that in lhe (a) sen\c:nce \he pronoun he will nom1ally be
of ac:ccssible referents have been observed by Prince ( 1981 a) and Chafe construed as referring to "Bill ... \Vhereas in (bl it \\'ill be construed as
(1987), who both conclude on the basis of text counts that the vasr. referring tc1 "Peter.·· This i:. o.;n 1ecause the hearer is able to keep the
 AcJivalion 103
lVl J he n1enra1 represe"tallons o_I discourse re}ererlls

interpretation of the pronoun on hold until a referent can be assigned to contract" between the interlocutors was not yet negotiated with respect
it on the basis of the entire sentence in which it occurs. As Dahl observes, to this referent. By violating this contract, John impaired the normal
it is often the case that we cannot interpret a constituent until we have communication process. In Section 5. 7 1 will propose a revised account of
heard part of what follo\vs upon it. This phenomenon of delayed the appropriateness conditions on the use of unaccented referential
construal is shown to be of great importance in the analysis of the French constituents, based on the added notion of "expected topic," which will
"antitopic" construction in Lambrecht 1986b (Chapter 8). Example predict the oddness of (3.23) (see example (5.80) and discussion).
(3.22) demonstrates that the use of a pronoun cannot ahvays be Notice that (3.23), though strange, is not uninterpretable. Mary could
understood as evidence rhat the speaker assumes that a representation have correctly interpreted her husband's utterance by means of pragmatic
24
ofa specific entity is already ''lit up" in the hearer's mind at the time the accommodation. This indicates that it is not only, and perhaps not even
pronoun is uttered. Rather, the use of the pronoun indicates the speaker's mainly, the use of the unaccented pronoun that causes the utterance to be
assumption that the hearer is able to INFER the referent from contextual anomalous. If John had used a proper name instead, i.e. if be had said
clues. Where is Bill? or even Where is a1u?, his utterance would still have been
The next example demonstrates rhat even in a situation \vhere there is inappropriate under the circumstances. Rather than constituting evidence
only one active referent in the speaker's and the hearer's minds, the use of against the activation approach to the use of pronouns, this example only
an unaccented pronoun to designate this referent may still be illustrates the intricacies of pragmatic presupposition and the importance
infelicitous. 23 Imagine the following scenario. John is returning home of shared knowledge in the processing of information in discourse.
from a trip during which he had no contact with his wife Mary or with The last example which [ would like to discuss concerns an apparent
anyone in his home town. During his absence, Bill, an old friend who bad exception to the above-postulated necessary correlation between use of
left the country years ago, arrived unexpectedly and is now waiting with an unaccented pronoun and assumed activeness of the pronominal
Mary for John to come home. It so happens that al the airport a fourth referenl. Allerton (1978) cites the case of a man who sees another man in
person bas told John of Bill's arrival. When John arrives at the house, a tennis outfit coming back from a tennis court and who says to that
both Mary and he know that Bill is in the house and both are thinking of man:
him. In spite of th.is common state of awareness concerning the referent, (3.24) Did you BEAT him?
it would nevertheless be inappropriate (though not impossible) for John
to say, upon entering his house: Even though the referent coded by him has the typical grammatical
characteristics of an active referent (lack of accent and pronominal
(3.23) Where is he?
coding), it is unlikely that this referent is in fact "currently lit up" in the
If felicitous use of an unaccented pronoun depended exclusively on the hearer's (the tennis player's) consciousness at the moment of utterance.
state of activation of a referent in the mind of the speaker and the hearer, Rather the referent is either totally inactive or it is in a state of inferential
the inappropriateness of (3.23) would be difficult to explain. The accessibility. (The decision whether to call the referent inactive or
explanation for this inappropriateness is round in our definition of accessible is difficult in this case, a problem to which I will return.} What
pragmatic presupposition in terms of mutually shared assumptions (cf. intuitively seems to justify the coding of the referent in this form is the
Section 2.2), or in what Clark and Haviland (1977) have called the fact that it is EASY TO {RE)ACTIVATE. As I see it, the interpretation of this
"given-new contract" bet\veen the speaker and the addressee. \Vhat is utterance involves two cognitive steps. The first is the pragmatic
wrong with (3.23) is the absence or an agreement bel\~'een the speech accommodation to the presuppositional structure of the pronoun him..
participants concerning the state ..._,f the mental representation of the Eveo though the speaker does not assume that the addressee is thin.king
referent "Bill" in the discourse. \Vhile John knev.' that ~'lary \Vas thinking of his tennis partner, he ACTS AS IF he were making that assumption,
of Bill at the time of his utterance, J\L:ir;, \\JS not aware that John was forc.ing his addressee lo go along with Lhat fiction in a cooperative
thinking of him. The utterance is inappropriate because the "given-new manner (or else to reject the dialogue). The second step is the reaction of
 Summary and illustration 105
104 11w mental representations of discourse referents
world-if she is invited to do so ON THE BASIS OF THE PRESUPPOSmO NAL
the addressee to the implicature created by the accommodat ed
STRUCTURE OF THE SENTENCE. She will NOT look for a referent if such an
presuppositio n. Having agreed to accommodat e the referent, i.e. acting
invitation is not grammatically expressed. I believe that the main criterion
as if he was indeed thinking of a particular male person, the addressee
in manipulating the pragmatic states of referents in a discourse is not
must now assign to the pronominal variable an actual referent. He is able
whether some referent is "objectively" active or inactive in a hearer's
to do this by inferring the intended referent from the semantic frame
mind but whether a speaker assumes that a hearer is willing and able, on
evoked by the word beat, from elements in the text-external world (for
the basis of grammatical forms with particular presuppositi onal
example the tennis outfit which his addressee must have noticed), and
25 structures, to draw certain inferences which are necessary to arrive at
from bis memory of the tennis match.
Far from constituting evidence against the activation analysis, example the correct interpretation of a referent.
(3.24) indirectly supports it. If the activeness assumption were not an
inherent feature of the presuppositio nal structure of the unaccented 3.4 Summary and illustration
pronoun, i.e. if the use of the pronoun him did not by itself signal that the
Even though identifiability and activation are independent cognitive
referent of the pronoun was assumed to be active in the mind of the
categories, one having to do \Vith knov"Jedge, the other with conscious-
addressee, it would be impossible for the addressee to take the utterance
ness, the two correlate with each other in certain predictable ways. It is
as an invitation to draw the required inferences in order to arrive at the
clear that a referent v.·hich is assumed to be unidentifiable to an addressee
correct interpretation . Here again, recognition of the psychological
is necessarily outside the domain of the activation parameter, since an
mechanism of pragmatic accommodat ion allows us to preserve the simple
activation state requires the existence of a mental representatio n in the
8Daly!is of the presuppositio nal structure of an expression type (in this
addressee's mind. To characterize such a referent, "unidentifiab le" would
case, pronouns) and to account for apparent counterexamples in a
therefore be a sufTiciently e;'(plicit label. Nevertheless, for the sake of
principled way.
parallelism with other terms to be introduced below, I will sometimes
iJlowever, I do think that this and similar examples to be discussed later
refer to an unidentifiable referent as BRAND-:-<EW, following Prince's
on suggest a necessary modification of the concept of ACCESSIBILITY as
(1981a) terminology. Still follov.•ing Prince, I will distinguish within this
viewed by Chafe. In (3.24) it seems futile to detennine whether the
category between UNANCHORED and ANCHORED brand-new items.
referent of him was inactive or accessible at the time the sentence was
The status of a referent as UNJDENTJFIAB LE correlates with certain
tittered. There arc few reliablC criteria a speaker can use to evaluate the
fonnal properties of the expression coding it. Prosodically, a noun phrase
states of referents in the mind of an addressee. I would like to argue that
with an unidentifiable referent is necessarily prominent, since lack of
acCesstDility (semi-activeness) of a referent, in particular accessibility of
prominence is reserved for constituents with active referents. In languages
the--mf'erential.. or '"situational" type, does not have to entail that the
which possess a grammatical category of definiteness. an unanchored
aa:essible referent is somehow present, indirectly or peripherally, in the
brand-new item typically appears in the form of an indefinite noun
bearer~s consciousnes s, as Chafe seems to assume. Rather what seems to
phrase (a guy. a bus). v.•hile an anchored brand-new item is a syntactic
mate·a referent accessible is the fact that, due to the existence of certain
combination of an indefinite and a definite phrase (a R'IH' I n·ork with, a
Jemantic relations within an invoked schema, due to presence in the
friend o._f mine).
situational context, or due to other contextual factors. the referent is
The correlation bet\veen unidentifiability and formal INDEFINITENESS,
easier to conjure up in the addressee's mind than a referent \vhich is
though strong, is not absolute Certain unidentifiable referents are coded
entirely inactive.
\vith definite NPs, and certain 1dentifiah!e ones may be expressed \.Vith
I·suggest, then, that we think of cognitive accessibility a~ a roTENT!AL
indefinite NPs_ 1 n1ent10ned the case of "pragmatic boot-strappin g:·
FOllAcnvATIO N rather than as the STATE OF A REFERENT in a pers0n's mind
v.·here an unidentifiahle referent 1s ended as <i cnmp!ex definite NP (the
Given accessibility of a referent, a hearer \vi\1 e-.:ploit th1s potential--by
dau.ehtcr (If u k1n.e or iJ k1n~·s daughter. example (3 17) and discussion).
drawing inferences or by searching the text-external or text-internal
iVV ~ 111: nlt'IJIU/ rt'µfl;;'.}l;;'l//UllUll:l Uj UO'COllf.}t' rejt'rt'nls
Summary and illustration 107

Furthermore, English noun phrases wilh lhe determiner this, which are The apparent exception may be explained as follows. In uttering a,
fonnaUy definite, may have unidentifiable referents (example (3.9) and lexical noun phrase, whether definite or indefinite, a speaker necessarily
discussion). On lhe other hand, generic referenls, which are always activates the CATEGORY denoted by the lexical head in addition to.
identifiable, may be expressed via indefinite noun phrases (see (3. 7) and activating an individual in that category. The active status of this-
discussion). The lack of a one-lo-one correspondence between category may then be reflected in an anaphoric expression, independently
(un)identifiability and (in)definiteness entails that there can also be no of whether the addressee can identify the particular referent the speaker
absolute correlation between indefiniteness and the presence of PROSODIC has in mind or not. Expressed in different terms, while the lYPE has
PROMINENCE. For example, generic indefinite NPs may be unaccented become active, the TOKEN may not be. In the competition for formal
when used anaphorically. marking, the type wins out over the token. Notice that this process of
Once a referent is assumed to be IDENTIFIABLE, it is necessarily in one of category activation applies whether we go from one individual to another
the three activation states "active," "inactive," "accessible." These (as in the above-mentioned I'm looking for a BOOK. - I FOUND one), from
activation states have a variety of formal correlates. An ACTIVE referent is an individual to a category (I'm looking for a BOOK - I LOVE books). or
typically, but not necessarily, coded with an unaccented expression. All from a category to an individual (I love BOOKS; in fact I'm reading one
unaccented referential expressions have active referents, but not all active right NOU').

referents appear as unaccented expressions. Unaccented expressions are An identifiable referent which is INACTIVE is necessarily relatively
marked for the feature "active referent" but accented expressions are prominent prosodically (/saw your Bil.OTHER yesterday) and it is typically.
unmarked for this feature. Similarly, all pronominal expressions (free or coded in English as a definite lexical noun phrase, except in the case of·
bound pronouns, inflectional markers, null elements) have active generic indefinite NPs and in certain cases of deixis, where an inactive,
referents, but not all active referents are expressed pronominally: they referent can appear as an accented pronoun (e.g. I want THAT). (In the
may appear as lexical noun phrases, and these lexical phrases may be latter case the referent might also be called accessible.) Even though the
definite or indefinite. Pronouns are marked as having active referents, label "inactive" is sufficient to designate such a referent, I will sometimes
while lexical phrases are unmarked for the active/inactive distinction. To use Prince's more vivid term UNUSED. This term has the advantage of
designate an active referent, the label "active" is sufficient. An often- distinguishing the category "inactive" more clearly from the category
encountered alternative label for "active" is "given," a term \11hich I will "unidentifiable" (calling an item "unused" implies more strongly that it is
generally avoid because of its ambiguity. already stored in the addressee's mind than calling it "inactive").
There is an apparent exception to the one-to-one relationship between As for the cognitive category ACCESSIBLE, it has no direct phonological
Jack of accentuation and/or pronominal coding on the one hand, and or morphological correlates, though it may have indirect correlates in
activeness of the coded referent on the other. In the discussion of the syntax. Accessible referents may be coded either like inactive or like
specific/non-specific distinction in Section 3.2.2 I mentioned that the active ones, depending on various factors to be discussed below.
anaphor to an indefinite noun phrase with a non-specific referent must be Accessible referents are subdivided into "textually accessible," "situa~
an indefinite pronoun or lexical NP. For exan1ple \\.'e saw that sentence tionally accessible," and "inferentially accessible.'
(3.3) ['m looking for a BOOK COUid be roJIO\ved by [FOUND one or [FOUND a Let me summarize the correlations between the cognitive states of the
book. The referents of the anaphor~ one and tJ huok in these sentences are mental representations of discourse referents and the formal properties of
unidentifiable, in the sense that the addre~see 1$ not assumed to be able to referential expressions. The most important formal contrasts are (i)
identify the particular book the spe;_iker h,t\ in n1ind. l'et the referent is presence vs. absence of an accent; (it) pronominal vs. lexical coding; and
treated as active- hence necess:inly ;_is idenllfiable - .JS indicated by the (iii) in some languages definite vs_ indefinite marking. The relationship
lack of accentuation on the anaplH1r dnd. 1n thc- c'1~e <1f one, by the use of between cognitive states and formal types can be looked at in two ways,
a pronominal form. depending on \\'hether \Ve are describing the former in terms of the latter,
or the latter 1n terms of the fom1er. Going first from the cognitive state to
 Summary and illustration 109
108 77w ntmtal npresmtations of discourse referents
:r. b ·r
. of J'dentljla d activation
Table 1. The grammatical expression i ity an
its formal exp~on, we notice that an active referent may be coded as
an unacx:ented or accented, pronominal or lexical, definite or indefinite Pragmatic category
Formal category
exp1t&ion, while a non-active (identifiable or unidentifiable) referent Identifiable referent Active referent
" rily appears as an accented, lexical noun phrase, which may be +
definite or indefinite. Thus while all formal types are compatible with the Pronoun I+ J
cognitive state "active," only a subset of formal types is compatible with Lexical XP
the other cognitive states. The selection of one or another formal type for +
Unaccented constituent [+J
expressing an active referent depends on various discourse factors too
complex to summarize here. Accented constituent
Going now from formal type to cognitive state, we notice the following
Definite NP ( +)
conclations: (i) pronominal coding and absence of pitch prominence are
sufficien'9 but not necessary, conditions for activeness of a referent; (ii) Indefinite NP
presence of an accent and lexical coding are necessary, but not sufficient,
conditions for inactiveness of a referent; and (iii) definite vs. indefinite
coding is neither a necessary nor a sufficient condition for either
The various terms in the systems of identifiability and activation are
identifiability or activation state, at least in English, even though the
tendency is strong for unidentifiable referents to be coded as indefinite summarized in the diagram in (3.25):
noun phrases. The lack of any necessary correlation between grammatical ( 3 .2S) unanchored (I)
definiteness and a cognitive state is consistent with the fact that

<
unidentifiable< d (')
anchore ~
definiteness, unlike prosody and the pronoun/noun contrast, is not a
IDEtrrlflA.BlLITY inactiv_e (3)~texrually (4)
universal grammatical category (see Section 3.2.2). In sum, the only one- ACTIVATION/ aro:ssible situationally (5
to-one correlation between a formal category and a cognitive state is the identifiable___. ~ . inferentially (6.
active (7)
one between lack of prosodic prominence and/or pronominal coding and
activeness. In other words, activeness is the only cognitive state which can
Using the numbers after each terminal la~el in the diagram, . the
be unambiguously expressed by grammatical means in English. The
terminological conventions, including alternal1ve terms, are summanzed
eXtent to which this latter statement applies to other languages is subject
to empirical verification. in (3.26):
-The various correlations between cognitive states and formal categories (J.26) (I) unidentifiab\efbrand-new
are summarized in Table I. In' Table I, a "+" symbol indicates that the (2) unidentifiable anchored/brand-new anchored
category on the left is formally marked for the pragmatic feature above it. (3) inactive/unused
(4) textually accessible
The"[+)" symbol in the identifiability column indicates that the given
(5) situationally accessible
feature is not directly coded in the formal category but is merely an (6) inferentially accessible
entailment from the .. +" feature in the activation column. The "( + )" (7) active/given
symbol indicates that there exist significant exceptions to the correlation · represente d ·tn t h e d"agram are exemplified in the
betw&n identifiability and grammatical definiteness, at least in English The categories t ·. .
follo,vine short (attested) text."" The relevant rererent1al express1~ns are
(see the discussion of "indefinite this," example (3.9), and of "pragmatic
under!in~d. Small capitals 1ndica!e main points of_ pttch pr~m1nence.
bootstrapping;~ example (3.17)) but that this correlation is nevertheless
Finer pitch accent variations are ignored I am also 1_gnonng differences
significant enough to deserve mention in the table.
in INTONAT!O~ co....;TOl'R, such as thut bet\veen ,HAR/\ (ns1ng, interrogative)
--
 Summa ry and illustration 111
110 The mental representations of discourse referents
to the
while about the person who has died of AIDS, might shift back
not
and ..f.IDS (falling, declarative). Such intonati onal distinctions are person called Mark with the sentenc e
sense of this book
indicato rs of infonna tion-str ucture categories in the
Mark is terribly UPSET.
but of speech-act distinctions or of differences in the speaker's attitude (3.27')

towards a proposi tion (see Section 5.3. l ): In this modified context , the referent of the noun Mark, which is by
now
the form of a pronou n
textually accessible, is unlikely to appear in
(3.27) ! heard somethi ng TERRIBLE last night. _(Ql rememb er the guy \ve™, because the intervening discourse has deactiv ated this referent , withou t.
t
went HIKING with{£} , who's GAY? His LOVER just died of AIDS. however, erasing it from the current discourse register, i.e. wiiliou
Mark may or may not be
returnin g it to the status of unused. The noun
ry
those accented, depending on the mental effort the speaker assume s is necessa
Among the referents of the underlined expressions in (3.27), to reactivate the referent as well as on other factors to be discusse d in
stands for you in the
designated by the deictic pronou ns /, o (\vhich Chapte r 5.
salient presenc e
omitted sequence do .vou), and 1ve are active, given their The reader will have noticed that in the charact erizatio n of
the
pronon1 inal
in the text-external world. The referents of the anapho ric discourse statuses exemplified in (3.27) and (3.27') I have said nothing
1~·ho,
expressions 0 (the relativized argun1ent in the first relative clause), under-
ernal about possible differences among the constitu ents that are not
and his are also active, due to their anapho ric status in the text-int lined. One might wonder in particu lar about the status of a verb· like
Mark.
world, having been "lit up" with the previous mention of the noun heard or hiking, an adjective like gay, or a preposi tion like with.
The
of these referent s is unambi guously express ed via
The active status reason for this omissio n has to do with the informa tion-str ucture
and absence of prosodi c promine nce. Among the
pronom inal coding distinct ion I drew in Section 3. I between DISCOURSE-REFERENTlAL and
NON-
" is
referents of the underlined lexical noun phrases, ··somet hing terrible DlSCOURSE-REFERENTIAL categori es. While it seems relative ly straight for-
tifiable ) and unanch ored, while .. Mark" and "AIDS "
brand-n ew (uniden interloc utor may have the referent of a noun
ent, as ward to assume that an
are unused (inactive). All three NPs are prosodically promin phrase like Mark or AIDS present in his consciousness or that
he can
exposit ory reasons , I am ignorin g the status of the comple x
required. For rnentall~ acc.ess such a referent , and while I think it makes sense to say
which,
apposit ional noun phrase the guy we 11"e111 hiking wilh, who's gay, that an interloc utor can identify the referent of such a phrase once it has
in apposit ion to the noun flr/ark, presum ably has the in
by virtue of being been introdu ced into the discour se, it is not clear what gets "activa ted"
referent of that noun and GAY are similar in
same status as the (!I/KING
a verb, an adjectiv e, a preposi tion, or
referent of the time the h.earer's mind when he hears
pitch accent and intonat ion contour to .lf.~RA")- The in_ his
d certain adverbs and what it is that can be assume d to be
expression last night is situatio nally accessible, being deictically anchore y makes no sense to apply
to its deictic status, lt may go consciousness after he has heard it. It certainl
with reference to the time of utteranc e. Due There are no definite or
noun the category of identifiability to such words.
unaccen ted (see Section 5.6.1 lP· 303)). As for the referent of the indefin ite verbs or adjectiv es, etc. (unless , of course, they are
t that it is inferent ially
phrase his lo1•er, let us say for the mo1nen nominalized, i.e. made into discourse-referential expressions).
evoked by the
accessible, both via its associa tion \\'ith the frame ~cc_ording to Chafe (1976, 1987), it is the .. concep t" of
the verb,
previously uttered \.vord gay and \ ia the anapho ric link to the nO\V
the hearer
27 adjective. adverb, or preposi tion that Is "lit up" in the mind of
active referent "Mark " 1nstant1att?d in the possessive detem1iner
his. activate d i~
I \Vill just as it is the "conce pt" of "Mark" or .. AIDS" that is being
However, it is not active, hence its ro.:l.1t1\"C prns(•dic pron1inence. the nl_ind of the speech particip ants during the speech act. This
way of
return to the status of the referent 1_1r /11s lurer shortly speaking is useful in the case of certain syntactic processes in
which
The text in (3.27) contain s n1_1 ex:unrk 1_1f ,1 textually c1c1.X'.Ssiblc referent lexi_cal items which are clearly underst ood from the context may be
ing
(i.e. of a previously active referent thctt i:il·(J.111e di.:a\_"(J\'ateJ by interven omitted from verbalization. One such case is the .. gapping " phenom
enon
discourse). An exampl e of such a refcren t ..:·<.1 n e:1.~ily be added to (3.27) by
to the restaura nt,
in a sentence like John went to the tnoi·ies and J.,,/ary
Sj'.'l.'ctker . b;iv1ng talked for a
extendi ng the text. For ex<.11nple the art('T
 /dent~{iahifity. activation. and the topic-focus parameter 113
112 T1te mental representations of discourse referents
ACTIVATION and IDENTIFJABLITI) and those indicating RELATIONS between
where the verb form went is omitted under identity with an immediately
referents and propositions (the categories of TOPIC and Focus). These two
preceding occurrence of the same form; another is the elliptical omission
types of information-structur e categories often coincide, but the
of larger sentence portions in answers to questions, as when I answer the
distinction is nevertheless crucial. For example, although in (3.27) the
question Where did she go? with the prepositional phrase To the mo\•ies.
accents fall in most cases on underlined items, i.e. discourse-referential
In such cases, it seems reasonable to assume that it is the identity of two
expressions associated with various activation states, they sometimes also
dcnotata which accounts for the possibility of omitting the second
occurrence, 28 fall on non-underlined items (HIKING. GAY), i.e. on non-discourse-
I see an important difference, however, between the infonnation- referential expressions. In the next section, I \Viii discuss further
structure status of predicators like heard or gay or with on the one hand, examples which shov.· that a distinction must be drawn between the
_and that of referential argument categories like Mark or AIDS or his or grammatical manifestation of activation and identifiability, on the one
JOU on the other. Only the latter designate entities-whether real,
hand, and that of topic and focus. on the other. This distinction is of
possible, or imaginary-and involve the creation and interpretation of fundamental importance for a proper understanding of the role of
,discourse representations via lex.ical phrases or pronouns. This difference prosody and morphosyntax for the informational structuring of
jn status between discourse-referential and non-discourse-referential propositions in discourse. The difference between activation accent and
categories has important implications for the interpretation of the role focus accent \vill be discussed in detail in Section 5.4. A unified functional
of prosody in information structure. While in the case of referential account of sentence accentuation \o,·ill be presented in Section 5.7.
categories prosodic differences may clearly indicate differences in
activation states, prosodic differences involving non-referential cat- 3.5 Identifiability, activation, and the topic-focus parameter
egories cannot, or cannot as clearly, be attributed to differences in
"concept activation." For example it would make little sense to argue Let us take another look at the status of the noun phrase his lo1·er in
_that the prosodic difference between the (relatively) unaccented verb (3.27), v.•hich I provisionally characterized as "inferentially accessible."
,forms heard and went and the (relatively) accented form Hlf{JNG in (3.27) is Consider the fol\O\Ving (constructed) variant of that text·
due .to a difference in the activation state of the designated concepts since (3.28) I just heard something TERRIBLE Remember MA.RK, 1he guy v.·e went
D;One of tbesC verbs were mentioned previously in the discourse. This is HIKING with, who's GAY'l I ran into his LO\'ER yesterday, and he told me
not to say that the cognitive states of the designata of predicators never he had AIDS
matter for their grammatical coding in the sentence (as seen in the cases
of gapping and ellipsis above). but only that they do not play the same The activation status of the referent of hH /ol'cr is necessarily the same in
prominent role as referential expressions in the coding system in \'.'hich (3.27) as in (3.28): in both examples the referent \vas not previously
meaningful contrasts arise with variations of word order, prosody, lexical mentioned but is to be 1nferred from another referent in the context. But
vs. pronominal coding, etc. I will return to the issue of the relationship there is a subtle difference in the pragmatic construal of his lover in the
between prosodic prominence and the pragmatic statuses of predicators two versions. It seems as if in (_t~kl the referent is given greater pragmatic
in Section 5.4.2. salience than in (3.27). And this greater salience correlates v.·ith the fact
It is necessary, then, to postulate l\vo functionally (though not that in (3.28) the NP is likely to be perceived as being prosodically more
necessarily phonetically) distinct types of prosodic contrast, one prominent. 29 The inferentially access1ble status of the referent is
expressing differences in the activation states of referents, the other grammatically EXPLOITED in L~-~7), \vhere the NP 1s a subject, but not
differences of another kind. I 'Nill argue that the existence of these hvo in (3.28), \~·here it occur~ in a p1)">!\erhal prepositional phrase. Through
functionally different types of accent correlates with the existence of t\vo its ~)NTACTlt organ1zat1on. (3.::'.7) suggests that the referent is already
different types of information-structure c:itegories: thos~ indicating acces"ih!e in the d1~cc•ur">e But the <,) :'-iT11.CT1c organization of (3.28)
suggest~ that the referent 1~ hetn& evnkcd 1n the hearer's consciousness as
temporary cognitive STATES of discourse referents (the categories of
 114 The )t1e111al represe111ations of discourse ref"erents JdenlljJ.abili1y, ac1iva1ion, and the topic-focux paronu!ler 11.5

a previously inactive, unused, discourse referent. In fact the sentence I ra11 the following question-answer pairs involving personal pronouns (from
into his LOVER yesterday (just like the preceding sentence Remember Gundel 1980):
MARX.?) may be seen as a type of "presentational" construction, whose
(3.29) A: Has Pat been called yet? _j
purpose it is co introduce che referent of the NP into the discourse rather
B: a. Pat said they called her TWICE.
than predicating something about the subject r30 b. Pat said she was called TWICE.
The apparent contradiction concerning the activation status of the
"
referent of the phrase hi.s lol'er in the two contexts is explained if we In the two answers to A's question in (3.29). the pronouns her and she are
accept my earlier suggestion to think of cognitive accessibility not as the anaphorically linked to the antecedent Pat. Their referent is clearly ·acti.Ve
state of a referent in a person's mind but as a POTENTIAL FOR EASY and they show the expected lack of accent. These pronouns stand in a
ACTIVATION, which may be exploited on the basis of clues from the TOPIC relation to their propositions. The constituent receiving the main.

presuppositional structure of the sentence in which the referent is focal accent in these answers is the adverb tn•ice. But consider now
expressed. The difference between (3.27) and (3.28) is that (3.27), by its example (3-30).
structure, conveys a request to the hearer to act as if the referent of the (3.30) A: Who did they call?
NP were already pragmatically available, whereas (3.28) does not convey 8: a. Pal said SHE was called.
such a request b. Pat said they called HER.
Facts such as these show clearly that the parameters of identifiability
The pronouns SHE: and HER in the answers in (3.30) are as anaphoric and
and activation do no! ex.hauslively determine 1he infonnation struclure of
their referents are as active as in (3.29), but this time they are accented,
sentences. Indeed if activation has to do with the stales of the mental
contrasting with the low tone on the other constituents in the sentence.
representations of discourse refen::nts, hu'v can tht: posilion of a noun
They do not stand in a topic but in a Focus relation to their propositions.
phrase in a sentence have an influence on our perception of the activation
The distinction drawn here on the basis of the prosody of English
state of its referent? We must conclude that the syntactic structure of
pronouns is particularly clear in languages in which the difference
i
I
sentences and the assumed discourse representations of referents
between pronouns with topic function and pronouns with focus function
correlate with each other and that this correlation is determined by an
has nol only intonalional but also morphosyntactic correlates. Thjs is tbC
independent factor. I will argue that this independent factor is the TOPIC
• and FOCUS structure of the proposition in which the referent is an
case e.g. in Italian and French. Rather dramatic syntactic differences
belween a discourse-active referenl with topic function and one with
argument. The reason why his lover in (3.27) is perceived as accessible and
focus function are illustrated in these Italian and French sentences:
why his lover in (3.28) is not so perceived has to do with the fact that in
the first case the referent plays the pragma1jc role of TOPIC, whjJe in the (3.31) <l. lOPAGO. -MOJJePAYE, "J'JJ PAY,.

second case it is part of Lhe Focus of the utterance. This points to the b. Pago 10. -C'esl MO! qui paye. "1'11 pay"
existence of a three-termed relation bet\veen accessibility. subject, and In example {3.3Ja), !he preverbal pronouns io and nioi are topic
topic on the one hand, and inactiveness. object. and focus on the other. expressions. Their intonation contour is nsing, indicating that an
In languages like English, 1n which the pragmatic categories topic and assertion is follov,1ing. Example {3.31a) is appropriate in a context in
focus are only weakly coded at the level of n1orphosyntax, the need for which the proposition "I pay" is construed as conveying information
the topic-focus distinction is perhaps n1ost convincingly demonstrated in about the speaker, in particular '~·hen contrasted with another
situations where a given referent which is already marked for its proposition expressing inforn1ation about someone else (as e.g. in 1'1/
activation state may in addition receive d 4uite different marking for its PAY, the OTHERS n1ay do as they PLEASE). In example (3.3lb), on the other
role as a topic or focus. \Ve S:J\~ e<.1rll~r lhJt the cle;in~st marking or the hand, the pronouns are focus expressions, and appear eilher in sentence-
status "active" is unaccented pron(l1n1n;.il <.:<.'ding ~·r a referent. Consider final pos1t1on (Italian io) or in clause-final .. clefted" position (French
n101i. Their intonation contour is falling, indicating the end of the
 f 16 T1tt- ~nlal "presen1a1iorrs of dlscourse refererrrs

assertion. Example (3.3lb) is appropriate in a situation in which it is

pragmatically presupposed that someone has to pay and in which it is
asserted that the person who is to do the paying is the speaker. (This
situation could present itself e.g. at the end of a dinner party where one
member of the party ostentatiously takes out his wallet and asks for the
4 Pragmatic relations: topic
check, whereupon another member protestingly utters (3.31b).) The
activation state of the referent of io and moi is obviously the same in both
cases, since the referent is one of the speech participants. What changes is
the information FUNCTION of the referent as a topic or as a focus of the
proposition expressed in the sentence. I will return to this issue in Section
S.7, where I will argue that the formal contrast between ac.cented and 4.1 Definition of topic
non-accented constituents can in fact be entirely explained in terms of the
4.1.l Topic and aboutness
grammatical expression of such information functions and that the
activation states of discourse referents are only PRECONDITIONS for the Let me begin with a few remarks about what I will NOT take "topic" to be
expression of these functions. in this chapter. First, in keeping with the decision to restrict my research
Notice that while in the English glosses in (3.31) the pronoun I is to pragmatic phenomena with grammatical, in particular syntactic,
obligatory, though subject to prosodic variation, the Italian pronoun io correlates in sentence structure (cf. Section l.1 ), I will restrict my
in (a) and (b) and the French pronoun moi in (a) are syntactically attention to SENTENCE TOPICS or CLAUSE TOPICS. I will have little to say
speaking optional. Unlike English •Pay, Italian Pago and French Je paye about the notion of 01scouRSE TOPtC, which has more ro do with discourse
are complete sentences, the subject being expressed in the inflectional understanding and text cohesion than with the grammatical form of
sufftx ·O and the bound pronoun je respectively. These unaccentable, sentences (cf. Halliday & Hasan 1976, Ochs Keenan & Schieffelin 1976b,
morphologically fixed morphemes are not capable of marking the same van Dijk 1977, Reinhart 1982, Barnes 1985, Van Oosten 1985), although
pragmatic contrasts as the free pronouns io and moi (and as the English I will sometimes informally use that term to designate a topic expression
pronoun /).which can function either as topic or as focus expressions. In whose referent is pragmatically salient beyond the limit of a single
Italian and French, as in many languages, the linguistic expression of a
referent can be split up, so to speak, into a pragmatically motivated part
.. sentence .
Second, I would like to emphasize from the outset that the concept of
(a pronoun of the "strong" or "independent" series io, tu. Jui ... or moi, • topic developed here does not coincide with that of topic (or "theme") as
toi, lui... ) and a semantically motivated part (a subject morpheme bound the "element \\'hich comes first in a sentence." In the framework adopted
to the verb). This dual coding of referents by means of functionally and here. sentence-initial elements may either be topics or foci, hence cannot
formally distinct expression types will be discussed in more detail in the be identified \vith either of these categories. The notion of topic/theme as
next chapter. the first element in the sentence is extensively discussed in Prague School
research (cf. e.g. the summary in Firbas 1966a) and has been adopted e.g.
by Halliday ( 1967) and Fries ( l 983). The relationship between sentence-
initial position and topic: function 'Nill he discussed in Section 4.7. Finally,
my notion of topic differs from that of GivOn and other linguists (cf. e.g.
GivOn 1983), who often use the term "topic" to refer to any "participant"
in a discourse and \\'ho do not dra'""' a principled distinction between
topical and non-topical participant<;, a distinction v.•hich is essential in my
O\VTI approach

117
 Defmition of topic 119

I
118 Prag1'tatic relations: rapt<

The definition of topic which I will adopl is related to the definition of What does it mean for a proposition to be ABOUT a topic? A
"subject" in traditional grammar (v.:hich goes back to Aristotle). The characterization of the notion of "aboutness" which I find helpful is
topic of a sentence is the thing \.vhich the proposition expressed by the given by Strawson (1964), who writes:
' sentence is ABOUT. The definition of topic 1n terms of the relation of
"aboutness" between an entity and a proposition has been adopted in Statements, or the pieces of discourse to which they belong, have
subjects, nol onJy in lhe relatively precise sense of logic and grammar,
one fonn or another by various contemporary linguists, including Kuno
but in a vaguer sense v•ith which I shall associate the words ••topic" and
(1972), Gundel ( 1976), Chomsky ( 1977), Dik ( 1978), Reinhart ( 1982). "about" ... Stating is not a gratuitous and random human activity. We
Even though this topic definition is derived from the traditional do not, ex.cept in social desperation, direct isolated and unconnected
definition of "subject." the two notions "topic" and ··subject" cannot pieces of information at each other, but on the contrary i.nteod in
be connated. Topics are noc necessarily gramrnatical subjects, and genera] to glve or add information about what is a mallet of standing
grammatical subjects are not necessarily topics. at least in languages like current interest or concern. There is great variety of possible types of
answer to the question what the topic of a statement is, what a statement
English. For example non-subjects may act as topics in !opicalization i:s ''about" ... and not every su~h answer exdudes every other ina given
constructions. and subjects may act as non-topics in such accent-initial case. (1964:97)
sentences as (I. J) (frfy c.~R brvke doH·n). The issue of non-topical subjects
will be taken up in Section 4.2.2. The principle expressed here, that statements normally are statements
Topic has sometimes been defined as a "scene-setting" expression, or about "what is a matter of standing current interest or concern," is called
as an element which sets •·a spatial, temporal or individual framework the "Principle of Relevance" by Strawson. If a topic is seen as a matter of
within which the main predication holds" (Chafe 1976). This definition of standing interest or concern, a statement about a topic can count as
topic applies mainly to what Chafe calls "Chinese style topics" (following infonnati\le only if it conveys information which is RELEVANT with respect
the description in Li & Thompson 1976) and to certain adverbial phrases to this topic. The Principle of Relevance may be added to Strawson's
often found in sentence-initial position across languages. Chafe previously discussed principles of the Presumption of Knowledge and the
distinguishes between topic in this sense and "subject," which he Presumption of Ignorance (cf. Section 2.2). Taken together, the three:
characterizes metaphorically as the "hitching post for the new knowl- principles are essential components of a theory of linguistic information.
edge."1 Chafe's distinction between "topic" and (topical) "subject" is Strawson's remark that "there is great variety of possible types of
motivated, I believe. by a desire to distinguish between topics which are answer to the question what the topic of a stalement is" stresses the
arguments, i.e. which are syntactically and semantically integrated into inherently vague character of the notions of aboutness and relevance.
the predicate-argument structure of a clause, and topics which are only This inherent vagueness has consequences for the grammatical coding of
loosely associated with a proposition and whose relation to the topics in sentences. If the topic is seen as the matter of current interest
proposition is a marter of pragmatic construal (see examples (4.50) ff \Vhich a statement is about and with respect to which a proposition is to
and discussion). The difference bet\veen these t'>.'O types of topic is drawn be interpreted as relevant, it is clear that one cannot always point to a
explicitly by Dik, who distinguishes beJ\veen "topic" and "theme," only particuJar element in a proposition, Jet alone to a particular constituent
the former or which is "a constituent or the predication proper" or a seutence, and determine that this element and nothing else is the
(1980:15). In the present study. I \Viil use ··1op1c" as a cover term for all topic of the sentence. As there are degrees of relevance, I.here are degrees
types of topic expression and I \\·tl! n1J!...e ;idditional distinctions in to which elements of propositions qualify as topics. It is this fact. I
morphosyntactic rather than pragrn01t1c lt'nns (cf. t-specially Section 4.4). believe, \Vhich accounts for the absence of unambiguous formal marking
Dik also distinguishes bet\veen '"then1t:"' (a n(1n-Jrgu1nent topic lo the left of the toplc relation in many languages. And, as a corollary, it accounts
of the clause) and "tail" (J nl1n-argunie11t \c>rtL· tl1 the nght of the clause). for the fact that in those languages which do have formal topic marking
These two notions closely 1natch 1ny 111-'L11i11~ 1u1'll r-.r c1nd ANT1Tor1c NP this tnarking 1eOects only imperfectly the relative degrees of topicality of
(see Sections 4.4.4 and 4.7J given referents.
 =
Definition of topic 121
120 Pragmatic relations: topic
vague does not alter the fact that they are crucial in understanding a great
1be definition of topic in terms of the pragmatic concepts of aboutness
number of formal grammatical phenomena. .
and relevance explains that it is sometimes not possible to determine the To see that even in a sentence like (4.1) the topic-comment structure ts
topic of a sentence on the basis of the syntactic structure of that sentence
to some extent formally (though not syntactically) e~pressed let us
~one, at least in languages like English, in which neither grammatical
contrast it with some possible a\losentences by embedding the statement
~lation nor linear constituent order are reliable topic indicators. In order 2
expressed in it in different discourse contexts:
to·determine whether an entity is a topic in a sentence or not it is orten
~cccssary to take into account the discourse context in which the sentence a. (What did the children do next?) The children went to SCHOOL
(4.2)
~ em~dcd. Consider the canonical subject-predicate structure in (4.1 ): b. (Who went to school?) The CHILDREN went to school.
c. (What happened?) The CHILDREN vtent to SCHOOL!
~4.1) The children \\'ent to school. d. (John was very busy that morning.) After the children v.•ent to
SCHOOL, he had to clean the house and go shopping for the party.
ID._ ?rder to determine whether (and to what degree) the subject noun
pfuase the children is the topic of, or is a topic in, this sentence, we must
Only in the reply in (4.2a) can we say that the referent of the subject NP
know whether the proposition expressed in this sentence is to be
the children is properly ··what the sentence is about." hence that this NP
pragmatically construed as being about the children, i.e. whether the
represents the topic of the sentence. In this context, the statement
children designated by the noun phrase are "a matter of standing current
expressed in (4. l) is intended to increase the addressee's knowledge about
interest or concern" (Strawson) and whether the proposition expressed in
the children as a previously established set or entities. The statement
(4.1) can be construed as relevant information about this matter. And to
pragmatically presupposes that the children in question are a "matter of
fmd·· out whether this is the case, we must know the context, the
standing current interest and concern·· (Strawson) and asserts about
communicative intentions the speaker had in making the statement, and
these children that they went to school. \Vith traditional logic, we might
the state of mind of the addressee with respect to the referent in question.
say that the predicate "\vent to school" expresses a property attributed to
·•1 The fact that in cases like this the topic relation is not unambiguously
the subject "the children." Ho\vever, the information-structure analysis
expressed at the syntactic level of the sentence is perhaps the main reason
differs from the traditional one in one important respect: both the
for the objections often raised against recognizing topic as a category of
"subject" relation and the ··predicate" relation are seen not as logical
grammar (and sometimes against recognizing the study of information
properties of the proposition expressed in the sentence but as pragmatic
structure as a valid theoretical enterprise; see Section 1.1 ). I believe such
properties of the sentence used in discourse. This distinction is crucial
objections are due to a failure to recognize the proper domain of
since, as \Ve will see. the same syntactic structure, expressing the same
information-structure analysis. As I mentioned in Chapter I, discourse-
logical proposition, can have a different information structure in \vhich
pragmatic categories are often most clearly manifested in pairs of
the "subject-predicate" distribution is not the same as in (4.2a). Hence
alloscntences, i.e. in the FOR~1AL CONTRASTS between alternative sentence
the need for the terms ••topic" and "comment." To characterize a
structures expressing the same proposition. Contrasts of this kind \vere
sentence such as (4.2a) in information-structure terms. the label ··topic-
illustrated in the English, Italian, and French allosentence pairs discussed
comment sentence" is therefore preferahle to the term "suhject-predicate
in Sections 1.3 and 3.5. While the morphosyntactic and prosodic
structure of individual sentences can be analyzed \vithout recourse to sentence." 3
Fonna!ly, topic--comment sentences are minimally characterized by the
the categories of information structure, only infonnation structure can presence of a focus accent on some element of the verb phrase, at least in
explain the DTFFERENCES BCIWEEN ALLOSENTENCES. The reluctance to
languages like English (see Section 5.2.21. In (4.2a). the topic--comment
recognize pragmatic relations as categories of gramm;:ir has a parallel
structure is expressed by the accent on the noun schnnl and. in addition,
in the reluctance (and often refusal) in early generative grammar to
by lov.' pitch prominence on the topic noun childrr11. (LO\~' pitch
recognize the relevance of semantic roles (Fillmore's "case roles") ror
prominence is. h(1\~·ever, not a neces'.'-ary condition fnr topic marking: see
linguistic theory. The fact that categories like ",:igent" or ··patient" are
 Definition of topic 123
122 Pragmatic relations: topic

Sections 4.4.4.2 and 5.4.2.) As I will show later on, topic-<.:omment The characterization of sentences such as (4.2b) as "identificational"
sentences like (4.2a) are syntactically and prosodically UNMARKED with does not embody the c;laim that the identified referent must be unique, Le.
respect to their infonnation structure, i.e. their formal structure is that it must be the only one to fit the open argument position (see the
compatible with other pragmatic construals, in which the subject is not a discussion of the so-called "exhaustiveness condition" in Hom 1981).
topic. Sentence (4.2b) is compatible both with a situation in which the children
Lel us no'v look al (4.2b). Here, the statement in the ansv.•er is NOT to exhaust the number of individuals that went to school and with a
be construed as a statement about the children. Rather its communicative situation in which the children went to school among other individuals.
function is to provide the referent solicited by the word who in the The former reading can be paraphrased as The ones who wen/ lo school
preceding question. Jn context (b), the reply pragmatically presupposes are 1he children, the latter as Among those »-'ho went ro school are.· J,he
the proposition that "someone went to school" and it asserts that this children (this latter reading is sometimes referred to as the "listing"
"someone" is "the children." I will call sentences such as (4.2b) reading; see Rando & Napoli 1978). The semantic distinction between the
IDENTIFICATIONAL SENTENCES, since they serve to identify a referent as ''exhaustive'' and the "listing'' interpretation of identificational sentences
the missing argument in an open proposltion. 4 The subject NP the is syntactically expressed via two different kinds of cleft constructions in
children is not a topic but a particular type of FOCUS expression (to be spoken French. The exhaustive reading is expressed via the ubiquitous
called "argument focus" in Chapter 5), i.e. its referent is not in the c 'est-cleft construction (C'est les ENFANTS qui sonr al/is ci l'ico/e), while the
domain of the presupposition. The non-topic status of this subject is listing reading is expressed via a type of a~·oir-cleft construction in which
formally marked by prosodic prominence on the subject noun, .,.,•hile the the relative Clause is Unaccented (Ya Jes ENFANTS qui SOnl a/Jes Q J'eco/e lit.
inclusion of the rest of the proposition in the pragmatic presupposition is "There are the children that went to school"; see Lambrecht 1986b,
marked via absence of prominence on the verb phrase. (This last Section 7.2.1 ).
statement will be modified somewhat later on; see Section 5.4.3 on the It is important to distinguish identificational sentences such as (4.2b)
relationship between activation and presupposition). from certain superficially similar topic-<.:omment sentences with copular
If we were to look for a topic in (4.2b), the best candidate would be the predicates. The semantic difference between the two types is made
presupposed open proposition ··x went to school," concerning which the apparent in the following set of examples:
asserted proposition can be said to add a relevant new piece of
(4.3) a. The ones who did that are my FRIENDS.
infonnation. It is the pragmatic status of this open proposition that
b. My FRIENDS did that.
bas prompted linguists in the "'language-psychology" tradition to call the c. It's my f"RIE.NDS that did chat.
verb phrase in such sentences the "psychological subject" and the subject d. They're my FRIENDS, the ones who did that.
the "psychological predicate." 5 I see, however, two reasons for NOT
calling the open propositions of identificational sentences "'topics" (or Sentence (4.3a) is an1biguous between an identificationaJ and a topjc-
"psychological subjects"), one semantic, one syntactic. First, since the comment reading. In the identificational reading, the noun phrase my
open proposition "X went to school" is se1nantically incomplete it cannot friends, which identifies the n1issing argument in a presupposed open
be said to have a referent, therefore the asserted proposition cannot be proposi!ion, is a referring expression. Under this reading, (a} is an
construed as being ABOUT its referent {see Sections 4.3 and 4.4 below). instance of a WH-cleft (or "pseudoc!eft") construction. As such, it is
Second, since the presupposition cannot be identified \\•ith a syntactic synonymous with the subject-accented (canonical) and the jt-cleft
constituent-the finite verb phrase 11·c11t to school does not express a allosentences in (4.3) (b) and (c). Jn the topic-comment reading, (a) is a
complete proposition-there ts Ill) structural element \vhich can be simple copular sentence in which the subject the ones who did that refers to
identified as a topic expression. "'Presuppos1t1on" and "topic" are an identifiable set of individuals and in which n1y friends is a non-referring
related, but not synonyrnous. I will return to this i~sue in Section 4.3. predicate nominal. Under this reading, ta) is not synonymous with (b)
and (c) but w·ith the anlitopic (right-detachment) construction in (d).·
 Definition of topic 125
124 hagmatic ~lations: topic

Let us now tum to example (4.2c). In this sentence, the subject NP is The nature of such sentences will be discussed in some detail in Section
also non..topical, as in (4.2b). But in contrast to the latter example, the 4.2.2 and again in Section 5.6.2.
proposition tbs< someone went to school is not pragmatically Context (4.2d) also exemplifies a situation in which the subject the
presupposed here. The answer in (c) is not, or at least not primarily, children is not clearly the topic of the sentence. But the reason here is not
construed as conveying infonnation about the children. Rather its that the NP is part of the focus of the assertion, as in (b) and (c), but that
function is to infonn the addressee of an EVENT involving the children as it is an argument in a proposition which itself is pragmatically
participants. The pragmatic function of sentences Jike (4.2c) wjJI be called presupposed (or to be accommodated as such). As we saw in Section
6 2.4, most adverbial clauses are marked for expressing pragmatically
EVENT·REPORTING. In (c), the pragmatic presupposition required by the
reply is merely that something happened. Since the focus of the assertion presupposed propositions. Now since in (d) the entire proposition "the
covers the entire proposition "The children went to school," the sentence children went to school" is assumed to be already known to the
is conte:\tually relatively independent and could be felicitously uttered addressee, it obviously does not constitute relevant new INFORMATION
..out of the blue." This is not to say that the answer in (c) requires no about the referent of its subject. Rather, this proposition provides the
shared knowledge between the interlocutors. For example, the speaker temporal background for the proposition expressed in the main clause. In
must minimally assume that the referent of the definite noun phrase the this sense, the subject tire children is not a topic. Nevertheless it is a topic
cfrildren is identifiable to the addressee. But this assumption has no direct in the sense that the proposition in the adverbial clause is indeed ABOUT
bearing on the question of the topicality of the subject. From the point of the children. The difference between (4.2) (d) and (a) is that in (d) this
view of grammar, what counts for the infonnation structure of the aboutness relation is not new to the addressee; it is not asserted, but itself
sentence as a whole is that the reply in (4.2c) is not, or not primarily, to be presupposed. The presuppositional structure of the adverbial clause in
construed as a statement about the referent of the subject noun phrase. (4.2d) is similar to that of the restrictive relative clause in example (2.11) I
, The non-topical status of subject NPs in event-reporting statements finally met the woman who n1oved in downstairs, whose proposition is
such as (4.2c) is not unambiguously marked in English. In the conte'lts understood to be about the referent of the pronoun who without however
chosen for (4.2), the subject noun in (c) is prosodically more prominent being ASSERTED about this referent. Prosodically, (4.2) (d) is similar to (a),
than in the topic-comment sentence in (a), but the presence of this accent the accent being assigned by default to the same constituent as in the
docs not in itself constitute grammatical evidence that the NP is not a corresponding topic---{'.omment structure (see Sections 5.1.2 and 5.3.3).
topic. Pronounced: with two prosodic peaks, as indicated in (4.2c), the The pragmatic status of the subject referent in (4 2d) can go formalJy
sentence could still have a topic-comment reading, for example if used in unmarked because it is entailed by the presuppositional structure of the
reply to the question "What did the children and the parents do?" The entire clause, which is marked by its external syntax, i.e. by the fact that it
reply. to the latter question could involve two subjects as "contrastive appears in topic position before the main clause.
topics" (see Section 5.4.2), as in (4.2c"): Notice that in (4.2d) the entire adverbial clause 1.1_/fpr the children ~i·ent
to school functions as a ··scene-setting'" topic for the matrix clause, whose
(4.2c') The onLDllEN went to SCHOOL, and the PARENTS v.·ent to BED. topic is he (i.e. ··John"). Thus in (4.2d) the noun phrase 1he children
is a non-topical or "semi-topical" expression \vhich appears within a
English sentences like (4.2c') can therefore not be said to belong to a sentential scene-setting topic expression. \Vhich itself is embedded wilhin
formal category "event-reporting sentence." Ho'A•ever. in other languages a matrix clause whose suhject is the primary topic or the sentence. This is
the non-topical status of the subjects of event-reporting sentences is a good example of the complexities of the concept of aboutness hinted at
consistently marked, as e.g. in Japanese or spoken French. In English, in the quote from Strav.·son above.
event-reporting sentences constitute a formal category only in the case of Because of the hybrid character of the pragmatic relation between the
certain intransitive sentences containing predicates such as die (Her subject referent .Jnd !he proposition in conte.'<t (d). this !ype of
HUSaAND died'), break down (M_v CAR broke do\\'n). call (JOHN called), etc. presuppos1tiona! structure is not as consistently marked across
 l.26 Pragn1atic relations: topic Definition of topic 121

languages as the other types, at least as rar as the grammatical marking or Chapter 5, the types in (a), (b), and (c), which are analyzed here in terms
the subject NP in the adverbial clause is concerned. 7 Nevertheless, some of their TOPIC STRUCTURE (or lack thereof), will receive a complementary
languages do mark this difference in topicality by morphosyntacic means. analysis in terms of their FOCUS STRUCTURE, and will be labeled ''predicate-
For example in Japanese the topic marker 11·a is normally not suffixed to focus", "argument-focus", and "sentence-focus" structures respectively.
noun phrases in embedded clauses (see e.g. Kuna l 972). In spoken In Lambrecht (in preparation) I will show that in spoken French these
French, topic marking via ''left-detachment" of the topic NP is information-structure categories are systematically distinguished via
incompatible with gram1natical subordination, except in cases where different syntactic structures. :i
the proposirional content of the subordinate clause is asserted rather than The characterization of "topic" adopted here may be summarized'as
presupposed (cf. Lambrecht 198!:58ff and J987a). Background-provid- follows. A referent is interpreted as the topic of a proposition if INAOIVBN
ing subordinate clauses as in (4.2d) are among the rare syntactic 01scoURSE the proposition is construed as being ABOUT this referent, i.e. as
environments in \Vhich canonical SV(O) syntax is used in spoken French. expressing information which is RELEVANT TO and which increases the
It is perhaps the kind of background-providing clause in (4.2d) to addressee's KNOWLEDGE OF this referent. Following Reinhart (1982), we
which one coufd most reasonably apply the label "pragmatically may say that the relation "topic-of" expresses the pragmatic relation of
unmarked" (see Section l.3.2), since in such sentences the topic-focus aboutness which holds between a referent and a proposition with respect
articulation is neutralized or maxin1ally reduced. But if we \vere to apply to a particular discourse. The term "pragmatic relation" should be
the term "pragmatically unmarked" to such sentences, this term would understood as meaning "relation construed within particular discourse
NOT be synonymous with "most normal" or "most frequently used." And contexts." Topic is a PRAGMATICALLY CONSTRUED SENTENCE RElATION. In
most importantly, it \.vould apply to sentences which show a maximum of what follows, I will try to make this somewhat vague characterization of
pragmatic presupposition and contextual dependence, rather than a "topic" more precise.
minimum, as usually assumed in discussions of pragmatically unmarked
or "contextually neutral'' structures (see e.g. Lyons 1977: 503 ). In my O\Vfi
4.1.2 Topic referents and topic expressions
usage, I wiJl not apply the term "unmarked" to such sentences but rather
to the topic-<:omment type illustrated in (4.2a), ror reasons which I v.•ill It is necessary to point out a certain ambiguity in the term .. topic" as I
make explicit in Section 4.2. I and in Chapter 5. have used it so far. Since the topic relation is a relation between a
Let me summarize the discourse functions of the four information- REFERENT and a PROPOSITION, it is natural that the term "topic" should be
structure categories illustrated in (4.2). In the topic-comment type in (a), understood as designating the ENTIIT which the proposition is about. i.e.
the purpose of the assertion is to pragmatically predicate some property the discourse referent itself about which information is being conveyed in
of an already established discourse referent. In the identificational type in a proposition. This is also one of the everyday uses of the term. To take
(b), the assertion has the purpose or establishing a relation betv.•een an an example, it would be natural to say that the topic of the propositions
argument and a previously evoked open proposition. In the event· we saw expressed in the sentences in (3.29) and (3.30), repeated below for
reporting type in (c), the purpose of the assertion is to express a convenience, is in all cases "Pat," i.e. the person designated with the noun
proposition which is linked neither to an already established topic nor to Pat and anaphorically referred to with the pronouns she and her:
a presupposed open proposition \this characterization \.••i!l be slightly
(3.29) a. Pat said they called her TWICE.
modified later on). Finally, tn the background-establishing type in (d). a b, Pat said she \'.'as called TVr'ICE.
pragmatically presupposed propo~ition scr\'eS as a scene-setting topic for
(3.30) a. Pat saidSHE was called.
another proposition, y,:hich may itself he of .iny of the other three types.
b Pat said they called HER.
With the possible exception of Lhe 1.:ist type, the exarnples in (4.2)
illustrate major 1nformation·strui:turc c .\ 1CG(1R1Es. \vhich tend to be To designate a topic in this sense I will use the term TOPIC or sometimes
morphosyntactically or prO':>Od1cal!y J1~l111gu1shcd across languages. In
8
TOPIC REflRENT, depending on whether I am more concerned with the
- 128 ~tic ~lotions: topic
- -:;:__ - -----=-"""

pragmatic relation between the entity and the proposition or with the
-~ __,,:'_- -· =-- - -- ~ ::..._ - -- ·-

Definition of topic

(b) it is a focus expression, i.e. its referent has a focus relation to the
129

enriry itself. Thus in the above-mentioned sentences, the individual proposition.

named Pat" is the topic which the different propositions are about. The
0
A similar point is made by Reinhart ( 1982) in an argument against a
referent of the expression Pat is a topic referent. definition of topic in terms of "old infonnation." Reinhart discusses the
This common use of the term topic to denote a referent ''"ith a t following simple question-answer pairs (focus marking added):
particular relation to a proposition should be sharply distinguished from
(4.4) a. A: Who did Felix praise?
the use of the term to refer to a L1Nou1ST1CEXPRESSION designating a topic
B: FeliJt praised MAX.
referent in a sentence. To refer to such an expression, I will use tenns like b. A: Who did Felix praise?
TOPIC EXPRESSION, TOPIC PHRASE, or TOPIC CONSTITUENT, instead of the B: He praised HIMSELF.
simple ..topic." When referring to particular syntactic categories with
topic function I will use terms like TOPIC NP or TOPJC PRONOUN. In Section Since B's answer in (b) can be construed as conveying infonnation about
4.S. an additional distinction will be drawn between LEXICAL TOPIC the previously mentioned referent "Felix," the pronoun he in this answer,
l!XPRBSSIONS and PRONOMINAL TOPIC EXPRESSIONS. To designate topic NPs which refers to Felix, is a topic expression. However, the reflexive
which are grammatically marked as such by their position or their form pronoun himself is NOT a topic expression, even though it is anaphorically
and which cannot be identified with the grammatical relations subject or linked to the topical antecedent he. The reflexive pronoun is a focus
object. I will use the category labels TOP (for left-detached topic constituent, whose referent just happens to coincide with the referent of
constituents) and A-TOP ("Antitopic," for right-detached topic the topic expression. 9
constituents). The ambiguity of the term "topic" is reminiscent of the It often happens that a referent which is topical on the discourse level is
well-known ambiguity of the tenn "subject," which tradltionalJy denotes coded as a focus expression upon its first appearance ln a sentence (for
both a grammatical or logical relation between an argument and a reasons having to do with its identifiability or activation status; see
proposition and the syntactic constituent in which this relation is Section 4.4.4) to become a topic expression only in a subsequent clause.
instantiated in a given sentence. This situation arises regularly in certain bi-clausal topic-introducing (or
If the distinction between "topic referent" and "topic expression" were presentational) constructions, such as the following:
not drawn, it would be impossible, for example, ro account for the (4.5) Once upon a time there was an old king who lived in a beautiful castle.
different information-structure statuses of the constituents Par, her, and
she on the one band and SHE and HER on the other in (3.29) and (3.30). The phrase an old king 1n the first clause of this sentence designates an
lbe two sets of expressions refer to the same entity, namely the person individual which has topic status in the discourse (the fairy tale is likely to
called ..Pat," who is the topic of both propositions. But while the be at least in part about this king). HO\\'ever, at the point in the discourse
expressions in the first set are topic expressions, those in the second set where this referent is first mentioned in the form of a lexical noun phrase,
are not. Similarly, in example (3.3 I) this noun phrase is not a topic expression. because the clause in which it
occurs cannot be said lo be ABOUT !he referent of this phrase; rather the
(3.31) a. fOPAGO. -MO! jc PAYE. clause INTRODUCES this referent in order to make it available as a topic for
b. Pago 10.-C'est MO! qui paye. subsequent predication. It is only with the relative pronoun who in the
relative clause that the refert"nt enters an aboutness relation with a
it would be impossible to account for the ditTerent infonnation-structure proposition, making who a topic expression in that clause. (Notice that
statuses of the preverbal constituents ro and {f(/f in (a) and of the this relative clause differs from the one discussed in (2.1 l) I finally met the
postverbal constituents 10 and .i.101 in (b). All of these rronouns designate ll'On1an who mo1'ed in downstairs in that its proposition is not presupposed
the same individual, i.e. the speaker, but in (3) the pronoun 1s a topic but asserted.) Even though it n1ay be natural and convenient to say that
expression, i.e. its referent has a topic relation t0 the proposition, \Yhile in the old king is "the topic of the sentence." insofar as that sentence is part
 Topic and subject 131
130 Pragmatic relations. topit·
I propose the following definitions of the pragmatic category "topic"
of a discourse about this individual, it must be emphasized that the NP an
and the grammatical category "topic expression": , o,;
old king is not a topic expression in that sentence.
A related issue arises with restrictive relative clause constructions. It (4.6) TOPIC: A referent is interpreted as the topic of a propositioo"if ui ~.gfv'en
bas been claimed or implied by researchers interested in the discourse- situation the proposition is construed as being abolit ibis refCMii,'T.e~
pragmatic functions of relative clauses (Schachter 1973, Kuna 1976) that expressing information which is relevant to and which ·incrcaseS-.,.-thc
addressee's knowledge of this referent. '.~ .3
the antecedents of relative clauses are topics or themes. For example,
TOPIC EXPRESSION: A constituent is a topic expression if. the p~opp_si_ti,gn
Kuno postulates the following '"Thematic Constraint on Relative
expressed by t~e clause with which it is associa~ is pragma~)'
Clauses": "A relative clause must be a statement about its head noun" construed as being about the referent of this constituent.
(1976:420). 10 However-assuming for the purpose of the argument that
the topic relation is extended to non-asserted propositions- a necessary As we will see later on, for a proposition to be construable as being about
topic relation obtains only bet\veen the proposition expressed by the the referent of a topic expression this referent must be pragmatically,
relative clause and the REFERENT of the head noun. Only the relativized accessible in the discourse. The somewhat vague formulation of the topic
element (which can be null) in the relative clause is necessarily a topic constituent being "associated" with a clause is necessary in order-:tq
expression. Both the head noun and the complex noun phrase account for topic constituents which bear no grammatical relation ·.t.o.a ..
CONTAINING the relative clause may V.'eli be focus expressions, as e.g. in predicate and whose semantic relation to the proposition is deterzcin~
the above-mentioned 1 finally mer the won1a11 who mo~·ed in downstairs. by pri~ciples of pragmatic construal only. Finally, by using-··tqe
This shows that topic constituents may be embedded not only within expre~s1on "being about the referent" rather than ..expressing. ·an
other topic constituents but also \\•ithin focus constituents (see the asse~t.Lon about the referent" I am allowing in principle for· the topic
analysis of the presuppositional structure of NPs like 1ny CAR, example relation to hold not only \\'ithin asserted but also within presupposed
propositions.
(I.I), or himsELF, p. 350, note 9; see also Sections 5.1.2 and 5.4.3). The
topic relation, holding within propositions, may in principle be expressed
in any syntactic domain capable of expressing a proposition. 4.2 Topic and subject
Let me summarize the remarks made in this section. A conceptual and
terminological distinction must be drawn bet\veen "topic" as the entity or The examples in (4.2) show that, in English at least, it is not possible to
referent which stands in a topic relation \vith a proposition and "topic" cqua.te topic with a ~ingle grammacical cacegory like subject. If this were
as the linguistic expression which designates such an entity or referent in possible, no separate category ··topic" would be needed. These examples
a sentence. While a topic expression always necessarily designates a topic also confirm something we had noticed earlier (Section 1.3), namely that
referent, a referent which is topical in a discourse is not necessarily coded ~he _syntax of the English SV(O) construction is unmarked with respect to
as a topic expression in a given sentence or clause. This asymmetry is due its i_nformation structure (the "O" in "SV(O)" stands here for any non~
to the simple fact that a referent is an entity \\'hich ex.is ts independently of ~UbJect argument). Nevertheless the correlation between topic and subject
its linguistic manifestation. The di~linction bet\veen topic referents and IS extren:iely strong on the level of discourse and has important

topic expressions is particularly important in the analysis of clause-level gra.mmat1cal co~sequences, in English as \\.'ell as in other languages.
or sentence-level constructions in \vhich one and the saine referent This JS "'hat I will show in the next section.
appears both as a focus expression and as a topic expression. The
structural domain in which the topic status of an expression is determined 4.:l.J Subjects as u11111arked topics
is the minimal syntactic domain coding the proposition of \\·hich the
referent of the expression 1s the topic. 1.c tht' clau~t> or phra-;e. not the Evidence for the strong correlation beh~·een subject and topic in English
sentence or the discourse. and across languages can be found in the way in which sentences whose
syntactic structure is unn1arked with respect to Information structure are
 132 Pragmatic rtlatioru: topic Topic and subject 133

interpreted in the absence of context. Since sentences are primarily used below). Other reasons invoked in the literature to explain the subject-
as units of information in coherent discourse, and since information relies topic correlation- such as the triple correspondence between subject,
"on presuppositions (see Chapter 2), language users have an unconscious topic, and the semantic role of agent (cf. Kirsner 1973, 1976, Hawkinson
·inclination to impose presuppositional structure on isolated sentences in & Hyman 1975, Comrie 1981, Lambrecht forthcoming, etc.)-are in my
order_ to be able to conceive of them as pieces of information. Now if opinion secondary, though by no means unimportant.
English speakers interpret canonical SVO sentences such as (4.1) in ._. There are certain \\'ell-known exceptions to the unconscious tendency of
isolation_ without contextual or prosodic clues, they are more than likely language users to construe the subjects of isolated sentences as topics.
to construe them as topic-comment sentences, i.e. they will unconsciously ;I'hese exceptions have to do with the lexical nature of certain predicates,
conjure up contexts of the kind given in (4.2a). Empirical evidence the propositional content expressed by the sentence, and the semantic role
;supporting this observation will be provided later on. . of the subject argument. In the discussion of markedness in Section 1.3, I
This psychological fact suggests that in English, as in other languages, ·observed that sentences like Her father died or My car broke down tend to
.subjects are UNMARK.EDTOPICS and that the topic-comment articulation is · be interpreted as event-reporting sentences, whose subjects will naturally
'the UNMARXEDPRAOMATIC SENTENCE ARTICULATION. This is easily explained be construed as being in focus. This is due to the fact that in the minds of
if we, assume that topic-comment structures, i.e. structures which are speakers and hearers certain propositional contents are strongly associated
used to convey information about some topic under discussion, represent with certain types of discourse contexts. The properties of such event-
.communicatively speaking the most common type. With Strawson and reporting sentences v.'iil be discussed in the next section. It is also well-
two thousand years of largely unchallenged grammatical tradition, I take known that certain experiential predicates v.'ith strongly non-agentive
this assumption to be a reasonable one. It is more common for speakers subjects, as \\•ell as certain types of passive constructions, favor non-topic
to convey ~ormation about given discourse entities than to identify status of the subject. For example a sentence like A strange thought just
arguments in open propositions, to introduce new entities into the occurred to me is likely to be read \\'ith the focus accent on the subject noun
discourse, or to report events out of the blue. Strong empirical evidence in rather than on the verb. In many languages, the non-topical status of the
favor of this assumption can be found in the fact that in coherent subject NPs of such sentences is marked syntactically (for example via
discourse the overwhelming majority of subjects are unaccented subject-verb inversion). Typicdl examples are Spanish ,\fe gusra NP "l
pronouns, i.e. expressions which indicate topic continuity across like NP" and Se 1·ende /1/p "NP is sold .. or German A-fir fdllr NP ein "'NP
5entenCCS (see Prince 198la, Chafe 1987, and the text counts in Chapter occurs to me" and Hier 1rird ,\'P 1·erka1{/i '"NP is sold here," etc.
6 of Lambrecht 1986b). The topic-comment articulation is then The unconscious tendency of language users to equate gran1matical
communicatively speaking the most USEFUL pragmatic articulation. It is subject-predicate structure \\·llh pragmatic top1c----(:omment structure in
therefore the one to which speakers will most naturally resort for the the absence of grammatical clues to the contrary is documented in the
pragmatic construal of isolated sentences. following real-life example_ Sentence (4.7) \\'as \vntten v.·ith a felt pen
If we accept this assumption, the fact that it is the SUBJECT and not across a poster protesting the \\·ar in Central America_ The poster had
Some other argument that tends to be interpreted as the topic hardly been partly ripped do,vn from the ,,·all it had been glued onto and (4.7)
requires an explanation. Since the subject is the most common argument was \Vritten in reaction to this presumably politically inspired attempt to
in the sentence-most predicators have at least a subject but not remove the poster. The sentence 'vritten by hand across the remainder of
necessa.rily an object complement-it is necessarily also the argument the poster reads·
which will be most readily identified "'ith the pragmatic role of topic.
(4.7) Nans tear dcn\'n antiwar roster-;
This, I believe, is the primary reason for the often-postulated universal
correlation between subject and topic. It is no douht this reason \-vhich Because of the tendency to interpret tht:> .;uhject of a sentence as its topic.
has led traditional and modern gramman;:in~ to C(•nsider the subject- the first interpretation of (4 7) \\'h1ch comes to the mind of a not overly
predicate (or NP-VP) sentence type the mo~! Fl\51(. one (see Section 4.5.2 attentive reader is the generic 1n1erpret.111,1n, \\'herehy the referent of the
 Topic and subject 135
134 Pragn1atic rela1io11s: topic

subject noun phrase Nazis functions as a topic about which the subject in (4.8); but (b) shows that no difficulty of construal arises if the
proposition expresses some generally accepted truth ("Nazis are people subject-predicate relation is reversed, i.e. if the postverbal NP in (4.8) is
who tear down antiwar posters"). However, given the situation in which construed as the topic. The pragmatic articulation of the sentence in (4.8)
the sentence was read, this natural first interpretation must be rejected, as presents itself as a reversal of the one most commonly associated with the
it violates the most elementary requirement of relevance. There is no clue given syntactic structure, leading to a certain increase in processiDg
whatsoever in the linguistic and extralinguistic context which would make difficulty. But unlike in the case of the highly anomalous (4.7), in. whiCh
the referent "Nazis" in any way pragmatically accessible as a topic and topic-focus indeterminacy led to severely diminished interpretability,_tl:Jp·
which would allow the reader to interpret the proposition expressed in topic-focus reversal in (4.8) may be seen as a (more or less conveuti9nal)
(4.7) as relevant information about this referent. Moreover the violation of an information-structure principle for rhetorical purpo~.
interpretation of the sentence with Nazis as a (generic) topic would Example (4.8) is acceptable within the conventions of a written ge~ fu
require that the predicate phrase tear down antiwar posters be read as the which added processing difficulty is made up for by the stylistic value of
COMMENT about this topic. But this reading conflicts with the fact that this unexpectedness. 12
predicate in fact evokes the presuppositional situation in the external I would like to emphasize that the kind of topic-focus indeterminacy
world of the discourse (i.e. the fact that the poster was partly torn off). manifested in (4.7) and (4.8) is NOT a common feature of spontaneous
The puzzled reader must therefore reinterpret the sentence with the spoken discourse. As far as written discourse is concerned, such
subject Nazis as a Focus rather than a TOPIC expression. This is indeterminacy seems to be tolerable to different degrees in different
presumably the interpretation intended by the person who wrote the languages. The fact that examples such as (4.7) and (4.8) do occur in
sentence. The intended meaning of(4.7) would then be equivalent to that English no doubt has to do with the previously mentioned fact that word
of the alternative versions People who tear down antiwar posters a;e N.~z1s order in modern English is to a high degree grammatically (and not
or perhaps Only NAZIS tear down antiwar posters. pragmatically) controlled and that in principle both the topic and the
Another attested example of an unusual match between pragmatic focus relation can be associated with the grammatical role of subject. As
topic-comment structure and synlaclic conslituent structure is shov.1n in the discussion in Section 1.3 showed, there is a striking difference between
the following sentence at the beginning of a course description for a modern English and modern French in this respect. In French the literal
linguistics class: translation of (4.7) in (4.7')

(4.8) The heterogeneity of linguistic com1nunities is the lopic of this course. (4.7') # Les Naz.is arracbent les affiches anti·guerre.
Even though the notion of the "heterogeneity of linguistic communities"
is pragmatically accessible in the context of a linguistics course would not only be puzzling, as it is in English, but downright
announcement (depending perhaps on the university), the referent of uninterpretable in the given context. The main reason for this is that in
the subject noun phrase is clearly less topic-\vorthy than the predicate NP French the pragmatic articulation of a proposition cannot be modified
the 1opic of this course. This becomes clear if we apply the question- simply by changing the prosodic structure of the sentence. To express the
semantic content of (4.7) in pragmatically acceptable form in French, a
answer test: 11
writer would most likely resort to a cleft structure such as C'esl /es Na2is
(4.8') a. Q: What is the heterogeneity of !ingu1st1c con1n1unities~ qui arrachent !es a/fiches anti-guerre ("It is Nazis who tear down antiwar
A: # The topic of this cour;.e posters") or perhaps II n'y a que /es Nazis qui arrachenr /es affiches anli-
b. Q: What is the topic uf llus cour~e·'
guerre ("Only Nazis tear down antiv.·ar posters").
A: The heterogeneity of linguistic C<.•rnrnun1t1..:s
Evidence for the strong tendency of English speakers to identify subject
Example (a) shows that it is difficult tu cc>n~1rue thi: pred1c3te "be the with topic may also be seen in certain facts of ANAPHORA and ElllPSlS.
topic of this course" as relevant inf0rm.it1un abuut the referent of the Consider the l\vo examples in (4.9):
 Topic and subject 137
136 hfwrnatic rdations: topic
(43) L John married ROSA, but be didn't really LOVE her. presuppositional structure, in which the subject is part of the focus of the
b. John married ROSA but didn't really LOVE her. sentence. 15
Ifwc conceive of these sentences as answers to the question "What ever
16
happened to JohnT' the subject constituents John and he in (a) are both 4.2.2 Non-topical subjects and the thetic-categorical distinction
topicexp~ons. As sentence (b) shows, it is possible to omit the subject
In the discussion of (4.2c), I characterized the statement ..The children
pronoun Ir~ in the second clause. However, if the subject NP in the first
went to school" when used to answer the question .. What happened?" as
d_ause were a FOCUS expression, omission of the topic pronoun he would
an instance of "event-reporting", and I argued that in such contexts the
iiSult 1i:i:i strongly diminished acceptability of the sentence. This is
focus covers the entire proposition, hence that the subject is not a topic. I
demonstrated in the variant of (4.9) in (4.9"):
also observed that in English the difference between event-reporting and
(4:9) . Q: Who married Rosa? topic-comment sentences is not unambiguously marked in sentences such
A:. a. JOHN married her, but he didn't really LOVE her. as (4.2c). However, there exists a class of sentences, in English and across
b. ..,JOHN married her but didn't really LOVE her.
languages, in which the contrast between the two pragmatic sentence
The status of John as a focus constituent in the answer makes omission of articulations is made formally explicit.
the anaphoric pronoun he in the second clause impossible in (4.9'). The The fonnal distinction which I have in mind is the one illustrated in the
difference between (4.9b) and (4.9'b) shows that our initial judgment pairs of allosentences discussed in Section 1.3 (examples (1.1) through
concerning the omissibility of the pronoun he in (4.9) was made under the (1.3) and (I. I') through (1.3')) Here is another set of examples, this time
unconscious assumption that the subject John is the topic of that including Japanese·
sentence. 13 B. Hov.··s your neck'!
(4. 10) A. \\'hat's the maHer"
The contrast between (4.9) and (4.9') is explained if we make the a I\.·1y NECK hurts a I\.1y neck HURTS.
functionally reasonable assumption that for an argument to appear in b. Mi fa male il COLLO b 11 col!o mi fa MALE.
phonologically null form in English the referent of the argument must c. J'ai mon cou qui me fait MAL c Mon cou ii me fait MAL.
d. KUBl ga !TAI. d Kub1 v.•a ITAi
have been established as a topic in previous discourse. The validity of this
assumption is demonstrated for Japanese by Kuno ( 1972), who observes It goes without saying that under the minimal context provided here
that when deleted arguments in Japanese texts are made overt via lexical the sentences on the right-hand side \Vou!d normally not contain full
material the overt NPs must be marked with the topic particle wa and can lexical subjects. Ansv.·ers such as those in (4.11), which contain
never be marked with ga. An analogous observation is made by Ochs pronominal or null subjects. v.·ould no doubt be more appropriate in
(1979) concerning argument deletion in spontaneous English discourse.
most situations:
In this study, I will take the preponderance of the topic-comment
sentence type and the strong correlation between subject and topic to be (4. \ \) a. It HURTS

universal features of natural language. 14 Across languages, the subject of b. 1\.11 fa MALE.
c. II me fail MAL.
a sentence will be interpreted as its topic and the predicate as a comment
d. 1T Al
about this topic unless the sentence contains morphosyntactic, prosodic,
.; Nevertheless unaccented suhject NPs are pragmatically POSSIBLE in such
or -semantic clues to the contrary. The subject can therefore be
characterized as the UNMARKED TOPIC EXPRESSION and the topic-comment sentences. They may therefore be used to emphasize the formal contrast
strocture as the UNMARKED PRESUPPOSITIONAL STRUCT1.!RE of a sentence. het\veen the t\vo types
The characterization of the subject as the unmarked topic :::dlo\VS for the The relevant grammatical contrasts het\~·een the event-reporting
possibility that in some sentences the subject is NOT a topic_ In the next sentences on the left-hand side in (4_ JO) and the topic-comment
section, I will discuss one important class of sentences \vith marked sentences on the right all have to do \vith the formal marking of the
 Topic and subject 139
138 Pragniatic relalions: topic

NP argument designating the body part. They may be characlerized as inspired by the philosophical distinction between THETIC and C.AT~
follows: (i) accented vs. non-accented subject NP in English; (ii) EGORICAL statements. According to this approach, the contrast
postverbal vs. preverbal subject NP in Italian; (iii) clefted vs. detached expressed in our examples is seen as the manifestation of two differeQ_t,
NP in French; and (iv) ga-marked vs. wa-marked NP in Japanese. The LOGICAL REPRESENTATIONS of the same propositional content. Topio-

four grammatical structures illustrated in the left-hand examples comment sentences are logically complex, while event·reporting
represent four major structural types which are attested in many sentences are logically simple. This approach is represented in work
languages (see Sasse 1987). Subject accentuation without concomitant by Kuroda (1972, 1984, 1985), it is hinted at in Dahl 1976 and Vattuonc
syntactic change is found e.g. in German (which also has type (ii), see 1975, and it has more recently been taken up in work by the Ge~.
Section 5.3.3); postverbal subject position is found in Romance, Slavic, scholars Ulrich (1985) and Sasse (1984, 1987), who have added a
and Chinese, etc.; cleft structures are found e.g. in \Velsh and Arabic; and pragmatic dimension to this logical approach. The proponents of.the
special morphological marking exists e.g. in Bantu (cf. Giv6n l975a). thetic-categorical approach claim (or imply) that thetic sentences
Sasse (1984, 1987) also mentions a fifth type: subject incorporation, represent a category of their own which cannot be captured with
which is attested e.g. in the Cushitic language Boni. principles of information structure.
There is no established terminology concerning the sentence type The distinction between thetic and categorical sentences was first
represented by the left-hand examples in (4.10). Terms which have been proposed by the nineteenth-century philosopher Brentano and further
used in the literature include "presentational sentences" (Bolinger and developed by Brentano's student Marty as a cognitive distinction
others), "neutral descriptions" (Kunc 1972), "news sentences" (Schmer- between two types of human JUDGMENT. Reacting against the generally
ling 1976), "event-reporting sentences" (Lambrecht i 987b etc), and accepted Aristotelian view that all judgment is categorical in nature_. i.e.
17
"thetic sentences" (cf. below). These terms all correspond to notional consists in predicating (or denying) some property of some ~~tity,
definitions of the phenomenon under analysis. In Chapter 5, I \Vil! Brentano and r-.1arty claimed that sentences can express two distinct types
introduce the terms SENTENCE-Focus STRUCTURE (for event-reporting of judgment. 18 The CATEGORICAL judgment, which is expressed in the
sentences) and PREDICATE-FOCUS STRUCTURE (for topic--<:omment sen- traditional SUBJECT-PREDICATE sentence type, involves both the act of
tences), two terms which emphasize the structural implications of the recognition of a subject and the act of affirming or denying what is
categories involved. expressed by the predicate about the subject. Since it involves these two
One of the first linguists to recognize the theoretical importance of the lndependenl cognitive acts, it is called a "double judgment" (Doppe.
grammatical contrast under discussion was Mathesius (1929). In post- lurleif) by Marty ( 1918, passim). The logical structure of the categorical
World-War-11 linguistics, the discussion has been pursued by numerous judgment can be represented as "'A is B" or '"A is not B." As illustrations
European and American scholars, all of \\'horn \vere more or less directly of sentences expressing categorical judgments Marty cites such examples
influenced by the Prague school: Bolinger (1954 etc.) on English and as the follo\ving:
Spanish; Hatcher (1956) and Contreras (1976) on Spanish; Firbas
(4.12) a. Diese Blume isl blau. "This Oower is blue...
(1966b), Halliday (1967), Chafe (1974), Schmerlrng (1976), Faber b !ch bin wohl. "l am (feeling) well."
(1987), etc. on English; Fuchs li980) on English and German; Kuno c Mein Bruder isl abgerei~t. "~1y brother left on a trip."
(1972) on English and Japanese; \Vandruszka {1982) on Italian; and Wehr
(1984) on Romance. In spite of many individual differences. the analyses In contrast, the THETJC judgment involves only the recognition or
presented by these scholars share the baste pren1ise that the contrast rejection of some judgment material, without predicating this judgment
between the two types has to do \\'ith 1nfonn.JL1on structure, in particular of son1e independently recognized subject Its basic logical structure is "A
with the activation and identifiability stJ.tc of the subject referent. is" or '·A is not." It is therefore also called a "simple judgment"
Another approach to th~ gran1mat1cal cl1ntr.i.st in (4.10). v.·hich has (einfaches {lr1eif). r-.tarty cites the German and Latin sentences in (4.13)
been less prominent in Am.:rican hngu1~t1L'S, is the seinantic approach belov.' a.s typical examples of propositions expressing thetic judgments. I
 Topic and subjecl 141
140 Pragmatic nlation.s: topic
Kuroda explicitly rejects the explanation of wa as a topic ~arker. I will
have grouped them into two sets, according to semantic and fonnal take Kuroda's notion of a "subjectless" sentence to be equivalent to that
of a "topicless" sentence, or, more accurately, to that of a ~ntence i_n
criteria which I will make explicit later on:
(4.13) a. Es rcgnet. I Pluit. "It is raining." which the subject is not a topic. It is equally clear that Kuroda s scmant~c
b .. Gott ist. "God exists." analysis of the thetic-categorical contrast is related to my pragmatic
Es gibt gelbc Blumen. "There arc yellow flowers."
analysis of the contrast between topic-comment utterances_ and eve.nt-
Es fmdet ein Markt statt. "'A market is being held."
reporting utterances. In what follows, I will argue that an 1nformat1on
In the thetic type in (4.13a). often exemplified with weather verbs it structure approach to the thetic-categorical contrast, which is based on
s;ems rel~tivcly uncontroversial to assume that the propositi~ns pragmatic, not logical, categories, is better suited to capture the nature of
~~~ 10. such sentences are logically simple in Marty's sense. They . . h 19
thetic sentences than a logico·semant1c approac .
~o not predicate a property of some entity but they simply assert or As I mentioned before, the German and Latin sentences Es regnet and
!>°sc" (hence "thetic") a fact or state of affairs. The type in (4.13b) P/uit in (4.13a) are cited by Marty as typical examples ofthetic sentences,
corresponds to the so-called "existential" sentence type, but as we shall expressing logically simple judgments. Notice, however, that these
sec it is not restricted to existentials. sentences are FORMALLY indistinguishable from subject-predicate
To my knowledge, the first systematic attempt to apply Brentano's and (topic-comment) sentences with pronominal or inflectional subjects.
MartYs logical dichotomy to linguistic theory was made by Kuroda For example in (4.15) there is no morphosyntactic or prosodic difference
(1972). According to Kuroda, the logical distinction between thetic and between the thetic sentence in (a) and the categorical sentence in (b):
categorical judgments is empirically confirmed in Japanese grammar in
~c formal distinction between the particles "'a and ga. For example, the (4.15) a. It 1s raining.
difference between the two sentences in (4.14) b. It is leaking

(4._14) a. Inu ga hasitte iru. "The dog is running." In (b), the prvperty of "leaking" is predicated of the argument expressed
b. ]nu wa hasitte iru. "The dog is running." in the subject pronoun it (for example a pot or a water pipe). Thus in (a),
a subject-predicate STRUCTURE is used to express a proposition which
is analyzed by Kuroda as follows. Example (4.14a), which contains a ga-
logically speaking has no subject (hence no predicate). This possibility of
marked NP. is a thetic sentence. It represents "the fact that an event of
expressing a thetic judgment \vi th a structure v.1 hich is normally reserved
running ... is taking place, involving necessarily one ... participant in the
for categorical judgments is possible because of the earlier-mentioned fact
event." The speaker's intention is directed in (a) toward the entity
participating in the event, i.e. the dog, "just insofar as it is a constituent that the topic--<'.omment structure is UNMARKED.
Marty's logical distinction betv.:een thetic and categorical judgments is
of an event." In the categorical sentence in (b) however, which contains
thus not necessarily translated into corresponding grammatical distinc-
wa, .. the speaker's interest is primarily directed towards the entity ... and
tions, a fact which Marty himself pointed out. Notice, however, that the
the reason why he wants to give an expression to the fact that he
use of subject-predicate (or categorical) structures for non-predicating
recognizes the happening of the event ... is precisely that he wants to
(thetic) propositions is possible only under one condition: the subject
relate the occurrence of the event to this entity" ( 1972: 16211). The entity
cannot be a FULL LEXICAL NOUN PHRASE. For example in those languages
to which an event is related by the speaker in this way is referred to by
(like Japanese or Russian) in which the meaning of (4.1 Sa) is expressed by
Kuroda as the "subject," which is grammatically manifested in Japanese
saying something like "Rain falls" or "Goes rain,"' the lexical noun
asa wa-marked NP. Thetic sentences such as (4.14a), on the other hand
meaning "rain'" must be grammatically marked as a NON-TOPIC, via
in which the entity is only a necessary particip::int in 30 event. are called
accentuation. post verbal position, ga-n1arking, etc. In other \NOrds, among
..subjectless" by Kuroda.
the sentences expressing thetic Judgments only those v.'hich contain full
It is clear that Kuroda's notion of "suhject" 1~ clo~elv rel3ted to the
\cxic:1l subject NP~ constitute a distinct formal category. Only they can,
information-structure notion of "topic"' or "theme.:, even though
.................... ..
142 Pragmatic rt'/arions. toph·
_. _. ________ ~~~

Topic and subject 143

.•1
are inherently PRESENTATIONAL, i.e. that they serve to introduce not-yet
at least in principle, be paired with categorical allosentences, allowing for activated referents into a discourse (see Section 4.4.4. l below). This is
grammatical contrasts such as those illustrated in (4.10). I
The fact that only thetic sentences with full lexical subject NPs form a
confirmed by the fact that in many languages certain constructions .I
expressing thetic propositions are restricted to, or at least strongly
grammatical category is clearly demonstrated in the following example:
preferred for, "indefinite" NPs, i.e. NPs with unidentifiable referents
l4.16) a. Her f"1'THER died. (English "existential" there.sentences, Chinese inverted word order. etc.).
b. He DIED. In contrast, topic-comment sentences have a strong tendency to tolerate
C. #HE died. only "definite" NPs (cf. Section 4.4 below). For example, Kuroda
observes that a ga-marked NP in Japanese can translate either as an
In (4.16), both (a) and (b) can be interpreted as reports of some striking
event and can be used as ans\vers to the question "What happened?" But
indefinite or as a definite NP in English, while a wa-marked NP can only
correspond to a definite (or a generic indefinite) NP in English.
"
the sentence cannot be GRAMMATICALLY marked as expressing a thetic
There is thus a direct link between theticity of a proposition and
proposition if the subject is a referential pronoun. The sentence in (4.16c),
in which a pronominal subject carries prosodic prominence, can only be presentational discourse function. Kuno (1912), in his functionally
oriented analysis of the wa/ga-contrast, observes that ga-sentences of .. ,
construed as an identificational or "argument~focus" sentence, in which
the kind discussed here, which he calls "neutral description" sentences,
the proposition that someone died is pragn1atically presupposed.
tend to be intransitive, containing predicates indicating the existence or
Another (attested) example, illustrating the same contrast, is (4.17).
coming into existence of some referent, or the appearance of a referent in
The sentences in (4.17) were uttered independently of one another and at
the external or internal world of the discourse. 21 Jt is well known that
different times by two waitresses in a restaurant when each passed near
these are among the verbs found in presentational clauses across
the door to the kitchen:
languages. Examples of presentational sentences in English, lta14m,
(4.17) Waitress A: Something's BURNING! French, and Japanese are listed in (4.18):
Waitress B: The TOAST'S burning!
{4.18) a. JOHN arrived.
The situation of utterance is virtually identical but the prosodic form of b. E arnvato GIOVANNI.
c. Y'a JEAN quj est arrive.
the two sentences is radically different. In the first sentence, in which the
d. JOHN ga kita.
subject is an indefinite pronoun, the main accent is on the predicaie; in the
second sentence, in which the subject is a lexical noun phrase, the accent is The utterances in (4.18) could be used by a speaker to introduce the
on the subjecl. If the first sentence \\'ere pronounced with an accent on the referent "John" into the discourse, from which point on it could be
subject (i.e. soJtETHING's burning). it could again only be construed as an anaphorically referred to in (unaccented) pronominal or null form,
identificational sentence. as in (4.16c). It appears, then, that there exists a depending on the language (see Sec Lion 4.4.4.1 ). The presentational
direct correlation between the grammatical marking of theticity and the interpretation is particularly obvious in the bi-clausal French construc-
presence of an overt accentt:<l lexical NP in the sentence.~ 0 tion, which consists of a presentational ("existential") clause follov.·ed by
What is the reason for this correlation? The ans\ver must have to do a relative clause in which the referent is coded as active via the
with the functional difference bet\veen leA1cal J.nd pronominal coding of unaccented anaphoric pronoun qui.
referents. As we sa\v in Chapter 3, full accenteJ NP coding is a necessary Let me illustrate the formal similarity between event-reporting
(though not a sufficient) condition for the eApre:-:.sion L1f a referent which sentences and presentational sentences with another example, illustrat-
is NEW in a discourse, i.e. \~'hich 1s either un1dcnt1fiable or inactive. It ing a con1mon utlerance type:
follows that the grammatical n1ark1ng .:,f thet1ci1y 1<. rc~tricted to
(4.19) a. The PHONE'S ringing! (subJec! accentuation}
discourse contexts in v.'hich the referent .:•f the NP has not yet been
b. Squilla ii TELEf-ONol (subjecl-verb inversion)
pragmatically activated This 1s tantam!.>unt tn ~..i.ying that such sentences
 Topic and subject 145
144 Pragmatic relations: topic

c. Y'a le n.LEPHONE qui SONNE! (y'a..clerting) constituents to have topic status in thetic sentences, in particular
d. DENWA ga NATTE iru yo! (go-marking) unaccented pronominal arguments and constituents below the phrasal
level. For example, we saw that the possessive determiner .my_ i~ (4.10~)
My NECK hurts is topical, even though the subject N~ of':h1ch 1t IS part IS
The sentences in (4.19) are not strictly presentational, in the sense that
a focus constituent. The same is true of the determiner 1n example (I. I)
they do not serve to introduce the telephone as a referent into the
Mr CA/I. broke down, a sentence which we can now categorize as a~ even.t-
discoune. Rather they serve to announce an event of ringing, in which
ce~tral thetic sentence. Examples of pronominal topic expressions tn
the telephone is merely a necessary participant. Nevertheless, in each
thetic sentences are the Italian pronoun nti in (4.JOb) or (1.2) and the
sentence the non-topical status of the NP is indicated with the same
French pronoun je in the bi-clausal construction ill_ustrated in (4.l~) or
prosodic and/or morphosyntactic features as the "presented" NP in
(1.3). (Recall that this bi-clausal French construction ~xpresses a single
(4.18).
proposition, in which the subject is NOT a topic: see Secllo~ 1._3 abo~e and
I suggest the following explanation for the fact that the same
further discussion in Lamhrecht 19R6b. Section 7.2.2.) S1m1larly, in the
grammatical category is used to express both the presentational and
the event-reporting function. What both functions have in common is second clause of the reply in example (4.20)
that the sentence expressing the thetic proposition introduces a new
(4 20) Q \\!hat harpened to J\.1:iry'
element into the discourse without linking this element either to an A She loq her j(lh. Jnd then her H\ ''>BA.ND left her
already established topic or to some presupposed proposition. The thetic
sentence thus has an "all-new" character which distinguishes it both from both the possessive determiner and the homorhonous object pronoun her
the categorical (or topic-comment) and from the identificational sentence are topic expressions Thetic ~entenccs may alsn contain locative topics,
type. The difference between the presentational and the event-reporting such as the prepositional phrase in nty soup in the sentence There we~e
type is that in presentational sentences proper the newly introduced 0

thrct' FLIES in 111 1• J(!llf, \\'hose topic slat us is e:-;pressed via lack of prosodic
element is an ENTITY (a discourse referent) while in event-reporting prominence (s~e Section 5.3.3) \Vhat counts for the definiti.on of the
sentences it is an EVENT, which necessarily involves an entity. (Even the formal category ••thetic sentence·· is that the constituent which would
events ofrai~ing or snowing involve entities, i.e. rain or snow.) I will use appear as the subject (or dist1ngu1-;hed argu1nent) NP in a corresponding
the term "thetic sentence" (i.e. sentence expressing a thetic proposition) categorical allosentencc gets fonnally marked as .1 t-;nN-.TOP!C", resu~t1ng in
to designate a superordinate information-structure category which a departure fron1 the unmarked prag1na11c articulation 1n which the
includes the categories "event~reporting sentence" and "presentational subject i'> the topic and thc prcd1ca1e the c0mment As v.·1th other
sentence," the latter including a deictic and an existential subtype. information-structure categories, the formal 1dent1ficallon of this
Following Sasse (1987), I will sometimes also refer to event-reporting category is made on the has1s ()f the ..::nntrast \~'ith a p~ssible
sentences as "event-central" and to presentational sentences as "entity- allosentcnce. I \vdl return \•) the ft•rn)~d d1~t1n..::t1nn het\\'cen thct1c and
central" thetic sentences. categorical sentences in the d1su~s1nn of fncus structure in Chapter 5
1would like to emphasize that the format contrast bet\l;een the marked (especially Secllon" 5.2.4 and 5 6.2).
category of thetic sentences and the unn1arkcd category of topic- An apparent prohlem for the definition nf the11c sentences in terms of
comment (or categorical) sentences crucially involves the grammatical the grammatical man1festat1nn of the ~uhJect NP arc thet.1c -ca_tcgoncal
relation SUBJECT (or "distinguished argument"; sec r. 150, note 14). his pairs \vh1ch in\"oh·e nn gran1n1at1cal suh.iect -~t, all. a~ in th1~ Czech
nbt the absence of any topic relation that charactcri1es thetic sentences example. \~'hich pJ.ra\leb the e\amrk~ in !-1 Jni:·-
but the absence Of a topic relation bC\V.'CCn the rrnrn~itJClO Jnd that
argument which functions as the topic in the c;:itcgc1riccil countcrp::irt. A.s (4 !fl") a V 7:i\Jed1 m0 h,)11 ·r-,1\' h.n::h 111 PT"'<'
we saw in Section 4.2.1, 1n the un1narkcd c;1o;c thio; categorical topic 1n h:ick-1.0C rne-1\( ( huri-.''¥
h Bnli m0 \ 1,i·dech '"'-l\ 1•1r ~- hurtc.
argument is the SUbJCCt. It io; in rrincirle rn~~1hlc for nnn·~llhJeC!
146 Prog,nalic relations: ropic Topic and subject 147

The information-structure contrast bet\\.•een (4.10') (a) and (b) is exactly but in the corresponding Spanish verb gustar it is an object (see (4.2la)).
the same as that between the allosentence pairs in (4.10), but there is one The same situation holds for the English predicate 10 be missing s1h. and
structural difference: what is a subject in English appears as a locative the French verb manquer (see (4.21b)):
prepositional phrase in Czech. (The locative case marking is semantically (4.21) a. Q: What kinds of th.ings do you like?
motivated by the fact that the body part argument is the "locus" of the A: I like WINE. - Me gusta el VINO.
pain.) Even though this prepositional phrase does not have the typical b. Q: What's the matter?
coding properties (case marking, agreement) we expect of a subject, it A: I'm missing a PAGE.-11 me manque une PAGE.
nevertheless has the required relational properties: like the subject in the Since topic is presumably a universal pragmatic category, it would be
corresponding English sentence, it has the semantic role of the locus of
absurd to claim that in the answers in (4.21) only the English subject
the pain and it bears the logical subject relation to the predicate boli
pronouns, but not the Spanish and French object pronouns, are topic
"hurts." Moreover, it appears in the positions \Ve expect, given the Italian
expressions, given that the relevant sentences may be used under the~sa.me
model (preverbal in the categorical version, postverbal in the thetic
discourse circumstances. Finally, we can mention the case of the
version). The prepositional phrase has the distinguished argument
topicalization construction, in which a non-subject constituent is
properties of a subject. (4.IO"b) can therefore be subsumed under the
"topicalized," i.e. marked as a topic expression by being placed in the
category "thetic sentence." sentence-initial position normally occupied by the topical subject.
The topicalization construction allows us to settle another issue, i.e. the
4.2.3 Topical non-subjects and n1ultiple-1opic sentences question whether a sentence can contain more than one topic. The fact
that in topicalization a non-subject beco1nes a topic does not entail that
The analysis of thetic sentences in the preceding section has confirmed
the subject must lose its topic status in the process. Therefore such.;,i
that subjects are not necessarily topics. I \viii now show that topics are
sentence may have two topic expressions. To see this, let us look at the
not necessarily subjects. An example of a topic expression \vhich is not a following atlested example:
subject was found in the thetic example (4.20) above. Another example
was shown in the sentence pair in (3.29), \Vhich I repeat here for (4.22) Why am I in an up n1ood? Mostly it's a sense of relief of having finished
a first draft of my thesis and feeling OK at least about the time I spent
convenience: writing this. The product I feel less good about.
(3.29) a. Pat said she was called TVr'ICE.
The last sentence in this text can be said to have t\\.'O topics and two
b. Pat said they called her TVr'!CE.
corresponding topic expressions: the topicalized object NP the product
Since these two sentences are (approximately) synonymous-they are and the subject J. Both are formally marked as such (non-canonical
both about "Pat" and they may be used under (approximately) the same position of the lexical NP, lack of accent on the pronoun). The subject /is
discourse circumstances- the object pronoun her in the second example topical because the whole passage in (4.22), including the last sentence, is
must be as much a topic expression as the subject pronoun she in the first. about the letter-writer and his feelings. We may call it the PRIMARY TOPIC.
A third example illustrating a non-subject topic expression is Jespersen's But the last sentence, in addition to conveying information about the
sentence PETER said ii (anS\Vering the question H'ho suid 1hi:u:1). in \vhich \Vriter, is also intended to convey information about the "product" (i.e.
the subj~ct is the focus and the objo:cl Lht: topic, Cl1rrespond1ng to the the thesis) in relation to the writer. The reader learns as a fact ABOUT the
pronoun that of the question product that the \\Titer is not happy \Vith it. We may call the thesis a
That non-subjects can be topics i::. ahc• t\·idc:nt from the fact that the SECONDARY TOPIC.

subject arguments of certain verbs in one l.1ni;uJge son1ellrnes appear as No\v since both the wnter and the product are presupposed to be
object arguments in another langu.igc To LbL an Jlready nh!ntioned topics under discussion at the time the sentence is uttered, the two
example, in the English verb like the e\pt:ri..:ncer arguino:nt is a subject, referents can be expected to stand in a certain relation to each other in the
 148 Pragmatic relations: topic Topic and subject 149

sentence. We can therefore say that the point of the utterance is to infonn noun phrase in canonical object position. The fonnal difference between
the addressee of the nature of the RELATION between the referents as the focal and the topical object argument is particularly clear in a
arguments in the proposition. The situation in (4.22) can be loosely language like French. In French the object argument appears after the
paraphrased as follows: given the writer and the thesis as topics under verb when it bears a focus relation to the proposition (fl a epouse Rosa)
discussion, the reader is infonned that the relation bet\veen the two is that but before the verb, together with the subject, when it has a topic relation
24
between the subject (or experiencer) and the object (or theme) of the (mais ii ne l'aimait pas i·raiment).
predicate/eel less good about. Thus a sentence containing two (or more) That it is possible and natural to pragmatically construe the clause he
topjcs, in addition to conveying infonnation about the topic referents, didn't really love her as conveying information about the referent of her as
conveys infonnation ABOUT THE RELATION that holds behveen them as well as that of he is shown by the fact that it may appear in the contexts in
arguments in the proposition. The reason the proposition can be said to (4.24):
be ABOUT this relation is because the existence of some relation between
(4.24) a. As for Rosa, John didn't really love her.
two (or more) topics is already established before the sentence is uttered. b. John said about Rosa that he didn't really love her.
The assertion in such a sentence is then the statement of the nature of the
relation. 23 In these versions, the status of her as a topic expression is made explicit
One might object to the view of there being more than one topic per via the first part of the t\vO sentences, whose grammatical purpose is to
clause or sentence by saying that it makes the concept of topic vacuous or mark the following propositions as being about "Rosa." The use of the
near-vacuous. Do we want to say, for example, that in the ans\ver as-for construction and the about construction to test the topic status of
sentence in (4.23) (a variant of (4.9)) an expression will be further discussed in the next section.
The claim that a sentence can have more than one topic is explicitly
(4.23) Q: What ever became of John?
rejected by Reinhart ( 1982). Reinhart argues that even though, when
A: He married Rosa, but he didn't really love her.
talking about a given topic, it is obviously possible to mention individuals
both he and her are topic expressions? No doubt the answer in (4.23) is who were mentioned in previous discourse or \.vho are otherwise
intended primarily as information about John, therefore the two pragmatically available in the context, only one expression can be the
oCcurrences of the pronoun he must be topic expressions. But this does topic of a given sentence. To demonstrate this. Reinhart discusses a short
not entail that the unaccented pronoun her is not a topic expression as text example from a recorded conversation (Shimanoff, transcription
well. Although the sentence primarily adds to our knowledge of John quoted by Ochs 1979:63 ):
(John being the primary topic), it also has the effect of increasing our
(4.25) A Je\\ish Grandfather (G! has !:ieen talking about the fact that his
knowledge of Rosa, by informing us that she \Vas not loved by her grandson is difficult to please. He gi\'es one example-oatmeal
husband. Both John and Rosa are under discussion at the time the clause
Ire didn't really love her is uttered. The communicative point of uttering G: And n's uh got ta good taste, !ls good And the cereal- grandma e
this clause is to inform the addressee of the nature of the relation bet,veen don't like cereal but she finished to the last (dish! and I enjoy- I hke it
100_ It's tasty' And I uh ( L2l He didn't want the cereaL doesn't eat. I
the two topic referents.
said. ·'Todd. it wouldn't kill ya, taste it' ..
Notice that while the clause he didn 'r really lo1'e her may be said to be
ABOUT the relation between the hvo arguments, the same ts not true of the Reinhart argues that in each of the beginning sentences, up to the
clause He married Rosa, in V.'hich Rosa is mentioned ror the tirst time. In parenthesis indicating the pause length, the topic is the cereal and that
the latter clause, Rosa does not bear a topic but a rocus relation to the after the pause the topic nf all sentences ts the grandson. In the second
proposition (see Section 5.1). The pragmatic difference bel\veen the t\vO part, the cereal, even though it i<; ~till vividly present in the interlocutors"'
clauses is morphosyntactically marked. the L\vo topic expressions are a\vareness and even th11ugh it is mentioned several times, is no longer the
unaccented pronominals, while the focus e-.:pres~ion is an accented lexical topic "\\1hi!e hefore 1t was the pr0pert1e~ of the cereal that the speaker
 150 Prag,,101ic rt'larions: topic Topic, presupposition, and semanlic inlerprelation ISi

was concerned with (e.g. hO\\' everybody likes it) here he is concerned referent of this NP. Since the referent was touched upon in the questio~
with the properties of the grandson (his rejection of the outstanding it was possible to construe it as an element of the pragmatic
cereal)" (Reinhart 1982). presupposition required by the answer. The proposition "The children
As far as I can see, there is no grammatical evidence supporting are under discussion" or "The children are to be predicated something
Reinhart's claim. It seems to me that the difference Reinhart is trying to of" is evoked by the presuppositional structure of the answer sentence in
capture on the basis of this text is a difference in the pragmatic SALIENCE (4.2a). Notice that this is not equivalent to saying that the referent of the
of the various topic referents at given points in the discourse, not the subject NP is ACTIVE in the discourse. The topichood of the referent is the
difference between topics and non-topics. It seems clear that the cereal is pragmatic relation it bears to the asserted proposition; the activeness of
more salient in the first part of the text and that the grandson is more its referent is a feature of the communicative setting. That the pragmatic
salient in the second part. And this difference in salience is reflected in the relation is not identical to the pragmatic property follows from the fact
fact that the more salient topic tends to be coded more often (though not that an active referent may also enter into a FOCUS relation with a
exclusively) as a subject. But it is not clear on what grounds topic status proposition (cf. Section 3.5). The relationship between topic and
of e.g. the pronoun it in it wouldn't kill ya can be ex.eluded. By the same activation will be further discussed in Section 4.4.
token, there does not seem to be any principled reason besides pragmatic It is no doubt the inherent relationship between topic and pragmatjc
salience to exclude the NPs grand1na and I from topic status in the presupposition that has led to the widespread terminological habit of
sentences grand111a e don't like cereal but she finished and I like it too. All calling the topic of a sentence "presupposed." This habit is as misleading
sentences containing two topic expressions have in common that the as that of calling a definite noun phrase, or even its referent,
referents of both expressions can be considered to be "under discussion" "presupposed" (cf. Section 2.3, example (2.12) and discussion). Any~
at the time of utterance and that some relationship is known to exist thing presupposed is propositional in nature (such as some shared belief
between them. The sentences then convey infonnation about the nature or knowledge), but topic referents are for the most part not propositions
of this relationship. but entities. Moreover, even propositional topics are not predicates but
arguments of, or adjuncts to, predicates. The fact that topic and
4.3 Topic, presupposition, and semantic interpretation presupposition cannot be identified with each other was mentioned
earlier, in the discussion of example (4.2b). What is presupposed in a
The definition of topic in terms of aboutness and contextual relevance topic-comment relation is not the topic itself, nor its referent, but the fact
entails that there is an inherent relationship between topic and pragmatic that the topic referent can be expected to play a role in a given
PRESUPPOSITION. Since the topic is the already established "matter of proposition, due to its status as a center of interest or matter of concern
current concern" about which new infonnation is ADDED in an utterance, in the conversation. It is this property that most clearly distinguishes
for a proposition to be construable as being about a topic referent this topic arguments from focus arguments, whose role in the proposition is
referent must evidently be part of the pragmatic presupposition, i.e. it always unpredictable at the time of utterance (see Section 5.1.1). One
must already be "under discussion" or otherv.'ise available from the therefore ought not to say that a topic referent .. is presupposed" but that,
context. We can say that the proposition ··xis under discussion" or "Xis given its discourse status, it is presupposed to play a role in a given
to be predicated something or·· is a relevance presupposition of a proposition. To indicate the fact that an item is in the domain of the
sentence containing x as a topic (see Section 2.3).: 5 presupposition, or belongs to the presupposition, I will say that it is IN
le is this relationship belv.'een topic anJ presupposition that motivates THE PRESUPPOSITION. The expression "in the presupposition" is the analog
the use of the question-ansv:er test as a \\ ay or d<:term1ning the topic of a of the expression "in focus" which I v.•ill introduce in Section 5.1.1.
sentence. For example in (-1.2:.i) \\C \\ere ahk t0 determine that the subject The correlation between topic and presupposition is what has
NP of the sentence The children 11·c111 ro \( 1!011/ was a topic expression by motivated the use of the above~mentioned as-for construction (cf. Kuno
construing this sentence a'> an anS\\·er L~1 ...1 4ucsl!on inqu1nng about the 1972, Gundel 1976) and about construction (Reinhart 1982) as tests for
 Topic. presupposition. and semantic interpretation 153
152 Pragmatic r~lations: topic
. h t ·c is an element of the
t d" by the interlocutors. Since t e opi . .
determining the topic status of an expression. In the as-for test, the gran e .. k d b th sentence. there is a sense in
referent of the putative topic expression is first coded in pre-sentential pragmatic presupposition evo e Y e must be outside
which the topic itself must be taken for grant~d, hence
position as the complement of the expression as for and then repeated in · odality in an assertion.
pronomina1 form in the sentence, typically-but not necessarily, as shown the scope of negation or m · p ne (1985·199fl)
a description of different types of negation, . ay .
in example (4.24)- as its subject. The as-for construction is thus a subtype ob~~rves that the way we understand sentential negation depends on the
of the detachment or dislocation construction (see Section 4.4.4.2 below).
"contextual articulation" or the sentence:
Applied to (4.1), application of the as-for test results in the structure in
. f h tence varies so does the
(4.1 '): As the contextual artt_culat1on -o the sen that what i~ negated is lhe
:~~;::;~a~~;p;r~~ ni~gf~l::·t~on: ~~u~n~e:~i:I ~eg~!~~,d~~; ~:~~~;~
1
(4.l') As for the children. !hey went to school.
element stands semantically therefore at t e
In the about test, the sentence containing the putative topic expression is contextually bound and contextually free elements
embedded under a matrix containing the preposition about, whose
. l · .. · losely related to \vhat is
complement is the putative topic NP, as in (4.1"): p ne's concept of "contextual artlcu auon is c d"
ay 5 d h. t rm "contextually boun
called "focus structure" in Chapter an is e .. "
{4".i") He said about the children that they went to school. may be considered the equivalent of my "in the presupp.os1uon. hAs a
ests a performallve parap rase
Both in the as-for test and the about test, the topic referent is formally diagnostic for negation scope, P ayne sugg . th
, t r "where x contains e
marked as being in the presupposition by being coded in a portion of the 0
f the type I sar of .¥ that it is not true I 1a • .
· · d the negative relates the
sentence which PRECEDES the clause expressing the proposition about it. b nd elements y contains the free e1ements, an
Notice that the phrase as for NP (as well as similar phrases in other ou .. In the cas'e \vhere the \Vho\e sentence is contextually free, X may be
two.
languages) can be appropriately used only if the NP referent is already a
potential topic in the discourse at the time the phrase is used, i.e. if the ze~ith Payne's diagnostic in mind. let. us consider the fo\lo:i~g negative
referent is contextually accessible (cf. Ochs Keenan & SchiefTelin l 976b). counterparts of the ans\ver sentences in (4_2) (a), (b), and ( ).
Not only would it be impossible to use the as-for construction for a (4.26 ) a. The children didn't go to SCHOOL
brand-new referent (•As for a strange guy, I satt• him last night) but it b. The CHILDREN didn ·1 go to school
would be highly inappropriate to use it with an inactive referent as well. c. The CHILDREN didn't go to SCH00L 1
An utterance like As for your brother, I satt' him last night is appropriate
For the topic-comment sentence in (4.26a). the perforrnative para~~~~:
only if the brother belongs to the set of referents under discussion. It is
ls "l say of the children that it is not true that they w~nt to ~choo\.
worth observing, in this context, that the as~for phrase can ONLY be used
b' ct NP as the contextual\v bound topic element, is outside the scope
in this topic-establishing function, as witnessed by the unacceptability of
~~ Jtehe ne~al!on. For the ident1ficational sentence in (4.26b), the
sentences like •As for V.'HOM did they go to school? or •They H'ent to school
araphrase is "I say of x going to school that it is not true that _the
as for the CHILDREN. As far as I know, as/or is the only phrasal constituent
phildren are x" (the unnaturalness of this paraphrase i.s due to the. earher-
in English which may not function as a focus expression. c ·t-on 1s not a topic about
mentioned fact that a presupposed open propos1 I . - h
By acknowledging the inherent relationship (but not identity) bet\veen
\vhich something can be properly asserted). Here _the subject NP is int. e
topic and pragmatic presupposition \Ve are in a position to understand scope of the n;gation. As for the _event-reporting (thct1c.) sentence in
certain correlations between the topic status of a sentence constituent and . h . "l say that it is not true that the children went to
the semantic interpretation of the sentence containing this constituent. In (4_26c).1ts parap rase is · · ontains
h I " The bound element X '~ zero here since this sentence c
Section 2.3, I mentioned that the truth of a pragmatically presupposed sc oo - -· (4 26c) the subject
neither a topic nor a presuppo~ed open proposition. 1n - -.
proposition cannot be affected by an element of negation or modality, is in th; scope of the negation togciher v.·1th the rest of the propos1t1on.
because the content of presupposed propo51tions is neces~arily "taken for
 154 Pragn1a1ic relatio11s: topic Topic, presupposition, and se1na111ic inlerpreta/jon 155

Notice, incidentally, that this semantic analysis is consistent with Marty's presupposition is attached to noun phrases in predicate, i.e. unmarke4
definition of thetic statements as expressing "logicaJJy simple" judg· focus, position. This has interesting consequences for the semantic
ments. 26 interpretation of negated topic-comment sentences. Consider. the! two
The fact that topics are outside the scope of negation is syntactically examples in (4.28):
reflected e.g. in the behavior of topicalized noun phrases in German.
(4.28) a. John isn't my FRIEND.
Consider the examples in (4.27); b. My friend isn't JOHN.

(4.27) a. Er ist keinARZT. "He isn't a doctor/He LS no doctor." Sentence (a) presupposes semantically that a certain individual named
b. Ein Arzt isl er NICHT. "A doctor he's not."
"John" exists (and it presupposes pragmatically that this individual is
c. *Keio Arzt ist er. "No doctor he is."
d. Kein Arzt kann dir HELFEN. "No doctor can help you."
identifiable to the addressee and that it is a topic under discussion): But·it
e. Ein Arzt kann dir nicht HELFEN. "A doctor can't help you." does not presuppose that the speaker has a friend. This sentence could be
truthfully uttered by someone who has no friends. Sentence (b), on the
When the noun Arzt is in predicate (or comment) position, as in (4.27a), other hand, presupposes that the speaker indeed has a friend and could
the negative morpheme must bi: fused with the indefinite article eiri into not be utlered felicitously if this presupposition didn't hold. It is certainly·
the negative detenniner kein. This sentence can be paraphrased as "I say possible to CANCEL the presupposition evoked in (b} via a metalinguistic
of him that it is not true that he is a doctor." If it is topicalized, as in (b), speech act, by saying fdr example, with special intonation, My friend isnil
resulting in postverbal position of the topical subject er, the negative John; I don't ha~'e any friends. But this simply means that the speakec feels
morpheme remains in predicate position, in the form nicht, and the the need to correct a wrong assumption on the part of the addressee. It·
topicalized constituent appears as a regular (generic) indefinite noun does not invalidate my claim that the topic~omment structure of the
phrase. Sentence (4.27b), which has two topics, could be paraphrased as sentence requires that presupposition. 21
··1 say of the relationship between him and being a doctor that it does not I believe that the requirement of existential presupposition (see Section
exist." As example (c) shows, it is impossible to topicalize the negative 3.2.1) for topic expressions is best explained pragmatically, in teems of
NP kein Arzt in (a), i.e. to use it in a structure in which its topic status is the discourse function of topics. It is obvious that for a proposition to be
marked. Example (d) shows that it is not the position of the NP v.'hich about some topic, and for this topic to be a matter of concern in the
makes (c) ungrammatical but indeed the fact !hat negation scope and discourse, there must EXlST an entily or set of entities which can be
topic status are incompatible. Sentence (d) is grammatical because in designated by the topic expression. However, what counts from the point
German (as in English) subjects are unmarked for the topic-focus of view of information structure is not that the entity simply_ eJlists but
distinction. The meaning of (d) is similar to that of the bi.clausal that it is PART Of THE UNIVERSE OF DISCOURSE OF THE INTERLOCUTORS. A topic
sequence There is no doctor who can help you, in "'hich doctor is a focus expression must not only be referential; it must designate a DISCOVRSB
constituent. Finally, sentence (e) illustrates the unmarked topic-comment REFERENT. The difference between expressions which designate discourse
structure in which the (generic) topical subject precedes the negative referents and expressions which don't is well demonstrated in this
morpheme nichr. This sentence is paraphrasablc as "I say of the minimal pair cited by Karttunen ( 1969)·
relationship betv.'een a doctor and you chat it LS not true that one can
t4.29) a. Bill has a car. II is black
help you." b. Bill doesn't have a car. •rt is black.
From the fact that topic~ n1ust he 111 the rresupposit1on, 1.e. taken for
granted as elements of the pr;ign1atic presupposition evoked by a Unlike in (a), in sentence tb) the NP a cur is not a referential expression in
sentence, it follov.·s that the referents ur tcipic expressions are necessarily the given discourse. As Karttunen observes, an indefinite NP establishes
presupposed to exist. Therefore, \(l th<-· .::-..tent that subjects are topics, the a di~course referent only if the NP justifies an anaphoric pronoun. The
referents of subject NPs n1ust be pr<-'supro~i:J to exist. l1ov.-·ever, no such pronoun 1t in t4.29b) is inappropriate because no entity has been
- IS6 Pragmatic relations: topic

previously established in the discourse which can be anaphorically

j

Topic, presupposition. and semantic interpretation

about a topic if one doesn't know what the topic is. Sentences whose
157

referred to with this pronoun. topic referents have an insufficient degree of pragmatic reality for the
The requirement that topic expressions designate discourse reFerents interlocutor are therefore difficult or impossible to interpret.
entails that only REFERRING EXPRESSIONS can be topics. For example such A famous example of such a sentence is The present King of France is
unaccentable expressions as the it in It is raining or the "existential" there bald which was first discussed by Russell ( 1905} in connection with the
in There's nobody in the room cannot be topic expressions. Application of pro~lem it poses to logicians. Since I am not concerned with the meaning
one of the above-mentioned topic tests will easily prove this point. (ln and truth conditions of sentences but with the information value of
contras~ the accentable deictic adverb there in There is John! is a topic utterances, I will not enter the debate over the so-called "truth-gap"' issue
expression.) The restriction against non-referring expressions applies also arising with this and similar sentences. Instead I will look at the sentence
to so-called "indefinite pronouns" and other QUANTIFIED expressions, like from the point of view of infonnation structure. In one of his
nobody, everybody, many people, etc. This explains why in example (4.27c) contributions to the debate. Strawson ( 1964} makes the following
the NP kein Arzt "no doctor, not a doctor" could not be topicalized, i.e. observation:
~old not occur in a position in which its topic status is marked, while it
Assessments of statements as true or untrue are commonly. though not
was acceptable in the unmarked subject position (cf. (4.27d)). However only, topic-centred in the same .,..,ay as the statements assessed; and
unive~ly quantified noun phrases can sometimes be topics, provide~ when, as commonly, this is so, we may say that the statement is assessed
that their i:ferents are coextensive with the entire class designated by the as putative infoi-mation ;\BOUT ITS TOPIC. ( 1964:97)
NP (cf. Remhan, 1982). Thus one can imagine a sentence beginning with
As for all his fr;ends, they ... but hardly •As for some people, they ... This Strawson's observation is related to the observation I made earlier about
also explains why the generic indefinite ein Arzt in (4.27e) is a possible the way language users unconsciously provide isolated sentences with
topic expression. 28 discourse contexts which a\IO\\' them to interpret subjects as topics.
Thus the referents of topic expressions must be discourse referents, i.e. Following Strawson's analysis (\vith minor terminological adjustments), I
they must have a certain pragmatic reality for the interlocutors. This I interpret the difficulty \.ve experience in assessing the truth of Russell's
believe, is the point of Keenan's (1974) analysis of what he calls the sentence as a consequence of the fact that the sentence, whose
"Functional Principle." Keenan is not concerned with topics and presuppositional structure is syntactically unmarked, insinuates as a
pragmatic relations but with subjects and the semantics of sentences topic a referent \\'hich, in the absence of context, does not have a
but his observations can easily be recast in infonnation-structure tenns'. sufficient degree of pragmatic salience to be considered a possible subject
Taking the subject (phrase) of a sentence to be an argument expression of discussion. In this sense, Russell's sentence 1s comparable to the
and the predicate (phrase) to denote a function, Keenan states as part of example of the tom-do\vn anti,var poster in (4.7).
his Functional Principle that the reference of an argument expression Nevertheless, it is not impossible to imagine a discourse situation in
must be determinable INDEPENDENTLY of the meaning or reference of a \\'hich the present King of France is indeed a topic. It is sufficient to
function symbol. He then concludes: imagine that the discourse participants believe. or act as if they believe,
that the individual designated by this phrase actually exists. In such a
(The function) associates with the referents of the subject a sentence- situation, \ve \vould have no difficulty interpreting the statement "The
meaning (say a truth value in a state of aflairs). So to evaluate the truth present King of France is bald" as a statement about this individual. As a
of a simple sentence we must mentally identify the referent of the subject
consequence, the sentence \Vould cease to be semantically and
and then determine whether the predicate holds of it or not. ( 1974:299) . 'O
pragmatically <inorn<ilous.-
One cannot assess the truth value of a proposition if one cannot identify 1 would like to mention. 1n thi~ connection. th<it 1n a language like
the entity of which the predicate is said to hold fn pragmatic terms: one (spoken) French. in \\'hich difference~ 1n information structure tend to be
cannot assess the infonnation value and the rele\'ance of a statement marked syntact1c<1lly. the log.1l-al ru11le :issoc1ated \V1th Russell's sentence
 158 Pragn1a1ic rela1io11s: topic Topic, presupposition, and semantic interpretarion 159
(4.30) A: Is Alice there?
·-
could hardly arise. If the phrase /'acruel Roi de France "the present King
B: a. There is no Alice here.
of France" were lo be construed as a topic expression, it would likely b. #Alice isn't here.
appear in right-detached or left-detached position, i.e. the sentence would c. #She isn't here.
be Le Roi de France ii est chaiH·e or //est chauve, le Roi de France. As d. #No.
shown in Lambrecht 1981, both of these topic-n1arking constructions can
be used felicitously only in discourse situations in which the referent of
the NP is pragmatically accessible. In such a situation, the truth value of What is remarkable in this exchange is that even though the individual
the sentence could be normally assessed. If. on the other hand, the named Alice was mentioned in speaker A's utterance, i.e. even tho.ugh tile
sentence were uttered "out of the blue," without the King of France referent would normally count as having been pragmatically established
being an accessible topic referent, this lack of accessibility would also after its first mention (and hence as being discourse-active for the purpose
have to be expressed syntactically and the proposition would be likely to of pronominal anaphora), speaker B can refer to this individual neither
appear in the forn1 of the tht!tic sentence II y a le Roi de France qui est with an anaphoric topic pronoun nor with a lexical topic NP. Rather, for
chau\•e (literally "There is the King of France who is bald"), in which the the sentence to be understood in the intended meaning, the NP Alice must
referent is formally n1arked as a non·toplc. In this case, it again seems appear in postverbal focus position, as the subject of a thetic sentence, as
possible to assign a truth value to the sentence, by determining whether in reply (a). This is so because the required pragmatic presupposition
the ··event" announced by that sentence is taking place or not. which would allow the individual "Alice" to serve as a topic is lacking in
It is worth pointing out that the bi-clausal, clefted, structure of this the conversational exchange. Even though the noun Alice is a ·referring
spoken French sentence, \Vith its initial existential (or presentational) expression, it cannot serve as a topic because it does not designate a
clause 11 _vale Roi de France, is remarkably similar to the logical structure discourse referent in the speaker's universe of discourse.
originally proposed by Russell to account for the semantic problem posed Particularly important for the present discussion is the fact that
by the English sentence. Expressed in plain English (with one minor answers (b) through (d) in (4.30) are inappropriate and misleading even
simplification concerning the uniqueness of the referent) Russell's logical though they express TRUE STATEMENTS (speaker B can truthfully say· that
structure is: ··There is a King or France, and he is bald." The important the individual in question is not in b..is house even without knowing the
difference between Russell's structure and the spoken French sentence is individual). The inappropriateness of these replies is reminiscent of that
that in French the noun phrase le Roi de France is DEFINITE, indicating of example (3.23) Where 1s he? in the discourse situation described thCre.
that its referent is treated as an identifiable entity in the discourse. This As I have emphasized in Chapter 2, what counts for the information
entails that the existence of the referent is pragmatically taken for structure of a sentence is not the truth value of the proposition ex.pressed
granted. What counts for the proper use of this thetic sentence by it but its lNFORMATION VALUE in a particular discourse. This
construction is the activation state of the referent in the discourse (see information value depends not only on the meaning of the sentence but
Lambrecht 1988a). also on the presuppositional situation in which the sentence is uttered.
That an expression must not only have a referent in order to serve as a One might object to this analysis by saying that the kind of
topic but that this referent n1USL be prag1natically established in the inappropriateness illustrated in (4.30) is not a matter of information
universe of discourse of the interlcicutL•rs 1~ de111on~trated in the following structure, i.e. of grammar, but of conversational implicature: answering
real-life example of a telephone exch..1ngc Speaker A has dialed a v..•rong A's question with "No" would be saying too little in the given situation.
number and is asking to speak tL1 :.i pcr"un unkno\\'ll to speaker B who i.e. it would be a violation of the maxim of quantity. I believe such an
receives the phone call. interpretation is misguided, because it puts the cart before the horse.
Indeed the reason the answer constitutes a violation of a conversational
maxim is precisely BECAUSE of the information structure of the
understood sentence Alice is not here in which Alice is a topic
 160 Pragmatic relations: topic Topic and tire mental representations of referents 161

expression. If the syntax of the sentence did not insinuate topic status of (1983:6). Similar activation states can be attributed to the other
the NP, no violation \VOuld be perceived. topicalized expressions in (4.31) and (4.32). . .
Even though I think that this characterization of the act1vatton
properties of topicalized noun phrases in English is corr~t and
4.4 Topic and the mental representations of rt>ferents
illuminating, I believe that the general definition of the d1scoui:se
4.4.1 Topic relation and aclil'ation state function of topicalization proposed by Prince is flawed in one crucial
respect. It seems misleading to characte~ze as one of the functions.. of
The definition of a topic as a referent which stands in a certain RELATION
topicalization that of "marking" an entity represented by an ~ as
to a proposition makes the topic concept intrinsically different from the
being" in a particular activation state. If one of the functions .of
concepts of identifiability and activation, which have to do with the
topicalization were to mark a referent as inferable (or otherw1se
PROPERTIES of (the representations of) discourse referents in the
accessible), we would not be able to explain the status of the non·
interJocutors" minds at given points in a conversation. The distinction
topicalized NPs sciences in (4.31) and the spoon and tire fork in (4.32?. The
between the mental representations of referents and the pragmatic
referents of these NPs have exactly the same activation properties as
relations which these referents enter into as elements of propositions is
those of the topicalized constituents, but instead of being fronted, they
related to the distinction between "old/new referents" and "old/new
Occupy canonical object position in their sentences. All of the relevant
information" discussed in Section 2.2. And like that distinction, it has
NPs in these texts, whether topicalized or not, have referents which the
often been neglected in the discourse-pragmatic literature.
speaker can assume to be in one \vay or another accessible in the hearer's
As a case in point, let us look at Prince's (1983) analysis of the English
mind. The cognitive state of these referents is a (temporary) PROPERIT
topicalization construction. According to Prince, one of the hvo
which they have in the particular discourse context, independently of the
discourse functions of topicalization is that it "marks the entity
RELATIONS they enter into as elements of a proposition.
represented by the NP as being either ALREADY EVOKED in the discourse
As I see it, the relevant function of topicalization is not to mark an
or else in a salient SET RELATION to something already evoked or inferable
activation state of a referent but to mark the referent of an NP as a
from the discourse" (1983:4). Among several other attested examples of
(particular kind o0 TOPIC in the proposition in which it is an argument
topicalization on which she bases this characterization, Prince cites the
and, as a corollary, to mark the proposition as being about the referent of
following (the topicalized NPs are in italics):
this topic. Such syntactic marking is necessary because in sentences with
unmarked presuppositional structure accented object NPs are not topics
(4.31) (=Prince's 22a) fl graduated from high school as] an average sudent.
My initiative didn't carry me any further than average. History I but focus constituents. (In English, topical object NPs may also appear as
found to be dry. J.fath courses I was never good at. I enjoyed unaccented NPs in canonical postverhal position, but being unaccented
sciences ... Football was my bag. (Terkel 1974:590) they Jack the "contrastive" or "referent-establishing" value of fronted
(4.32) (=Prince's 22b) Sunday I y,'as taking paper and pasting it together and constituents; see the discussions in Sections 5.5 and 5.7.J
finding a method of how to drop spoons, a fork, a napkin, and a straw Here is another attested example of multiple topicalizations which
into one package. The napkin _feeder I got. The straw feeder we made shO\VS that the discourse function of this construction cannot be
already. That leaves us the spoon and the fork. (Terkel 1974:516) described in terms of activation states alone. The example in (4.33) was
uttered by my six-year-old daughter \\"ho \vas sho\\'1ng me a number of
The referents of the topicalized NPs in these examples exhibit what I have recent additions to her sticker album
termed the activation states of textual and inferential accessibility
·(Section 3.3.I). According to Prince. the referent c1f the NP history "is
(4 33) Thi~ one we traded. this one we traded. this one she let me have, this one
inferable, via a set-to·element inference, from a set that is not mentioned she let me have, this one we traded. she let me have this one. this one we
but that is itself saliently infera hie fron1 the high sch not 'frame'., trade-d
 . . ·"'" -~· --
162 Pragrnaric relarions: topic Topic and the mental representations of referents · 16l

All occurrences of the phrase this one in this text have referents which are information, i.e. as to the entity that she wishes to convey informatio~
highly accessible in the speech situation (the pictures were displayed on about. But before making this communicative decision, the speaker must
the table in front of us and pointed to during the conversation) and all make certain hypotheses concerning the status of the referent of the topiC:
but one are topicalized. It is clear that the difference between the in the mind of the addressee at the time of the utterance. On the..basis~or
topicalized phrases and the canonical occurrence cannot be explained these hypotheses, the speaker then decides upon the form of.the-senlC2.nce
here in terms of the cognitive states of the respective referents. Rather it in which the topic is to be coded. However, the fact that a particu.1¥
has to do with the nature of the relation between the referent and the referent has the activation properties required for topic· function.:inra
proposition. By leaving the object NP this one in canonical position in the sentence does not entail that it must be coded as a .topic. Co~
second-to-last clause the speaker marks the referent as having not a topic accessibility is only a necessary, not a sufficient condition for the use ofa
but a focus relation to the proposition, i.e. as being an unpredictable construction such as topicalization. This important point was emp~
element within the proposition, thereby drawing special attention to it. in the preceding chapter in the discussion of the two syntactic codings of
(In this particular case, the switch from the topic to the focus relation the NP his lover in examples (3.27) and (3.28) and it is clearly illustrate<l
may have been motivated by the desire to exploit the unexpectedness in the cognitive similarity of the topicalized and the non-topicalized NPs
inherent in the focus relation for rhetorical purposes.) in (4.31) through (4.33) above.
In insisting on the distinction betv.·een pragmatic relations and The indirect but necessary relationship between topic function on the
pragmatic properties I am NOT denying the existence of a correlation one hand and the temporary cognitive states of referents on the other has
between the topic function of a referent and its cognitive activation state. a revealing parallel in the area of semantics. In semantic analysis it is
Indeed in order to make a referent INTERPRETABLE as the topic of a necessary to distinguish the semantic case roles associated with. the
proposition and in order to make the proposition INTERPRETABLE as arguments of a predicate, such as "agent," "experiencer," .. patient," etc.,

[,
presenting relevant information about this topic, the topic referent must from the inherent semantic properties of the noun phrases used for these
have certain activation properties, which, in the case of the English roles, such as the property of ANIMACY. However, there is a necessary
topicalization construction, are precisely the properties pointed out by relationship behveen semantic role and semantic property. The nature of
Prince. Prince claims- correctly I believe- that the sentences in (4.31) and this relationship is weil explained by Comrie (1981) in the following
(4.32) could not be processed effectively if the hearer were incapabie of passage. Interpreting the case roles "agent," "force," ..instrument/'
I
making the necessary inferences concerning the status of the topicalized "experiencer," and "patient" not as a set of discrete semantic roles but as
NP referents in the discourse. But the reason these sentences could not be various points on what he calls a "continuum of control," Comrie writes
r processed effectively is that a topic relation between a referent and a (emphasis added):
proposition can be effectively construed only if the topic referent has a
certain degree of pragmatic accessibility. By its presuppositional
structure, the topicalization construction acts as an invitation to the It might seem Lhat the conlinuum of control and the distinction of
e.\.periencer from patient are concerned with animacy, but in fact it is
hearer to EXPLOIT the cognitive accessibility of a particular noun phrase in
crucial to keep these two parameters apart. Nol.ions like control and
a particular syntactic configuration. {Recall that pragmatic accessibility is experiencer refer lo a RELATION BETWEEN TiiE PREDICATE AND ONE Of' ITS
seen here not as the cognitive ~tale of a referent in a person's mind but as ARGUMEi-.1s. The :><:ale of animacy, however ... is concerned with an
a potential for activation; cf. Sections 3.3 and 3.4.J INHERENT PROPERTY OF THE NOUN PHRASE, irr~spective of their role within a

I believe that it is the condition 1.>f interpretJbility that provides the best parlicu!ar construction. Thus the noun phrase the man is always high in
animacy, although it may vary in degree of control, having high control
explanation for the relationship bel\\Cen tc•pic function on the one hand
in the n1an deliberately hit n!e, minimal control in I hiI 1he man, and
and the activation and ident1li;,i.bil1t! pr(•p<.'rtics of topic referents on the either high or low control in 1he man rolled down the hill, depending on
other. In selecting a tL•pic fc•r ~1 ~cntence, a speaker makes a the particular interpretation assigned ... More generally: a high degree
communicative decision as t•J the "r1.•1nt l)f departure" for the new of an1macy is nece.ssary for a noun phrase lo be interpreted as having a
I 165
AM "1aptotic nlations: topic Topic and the mental representations of referents

high degree of control or as an experienccr, but is not a sufficient topic, that the topic of a sentence is "the old information" is, to say th.e
condition. (Comrie 1981:55() least, misleading. However, it is equally misleading to assert th~t ~here IS
no necessary relationship at all between the two parameters. llus JS what
Comrie's distinction of the semantic relations between predicates and I will demonstr~te in the next section.
arguments on the one hand and the inherent semantic properties of noun
phrases on the other clearly parallels my distinction between pragmatic
4.4.2 The Topic Acceptability Scale
relations and pragmatic properties of referents. Just as a degree of
animacy is a necessary condition for a high degree of control, a degree of from the requirement that topic referents have a degree of pragmatic
activeness or at least accessibility is a necessary condition for a referent to accessibility it follows that sentences with insufficiently accessible topic
,be interpreted as having a high degree of topicality. But just as animacy is referents must pose certain difficulties of interpretation, hence will tend
1
DOt SUFFICIENT to guarantee a high degree of control, activeness or to be perceived as ill-formed. Such difficulties of interpretation can be
iaccessibility arc not sufficient conditions for topic function of a referent accounted for by postulating a general correlation between the activation
in a proposition. An active or accessible referent may appear either as a and identifiability states of topic referents and the pragmatic acceptability
topic or as a focus expression, depending on the pragmatic role it is of sentences. This correlation can be expressed in the form of a SCALE OF
·meant to play in the proposition. ACCEPTABILITY. Allowing for a certain amount of cross-language
~ The fact that it is necessary for a referent to have a degree of variation, we can measure the degree of pragmatic well-formedness of
accessibility in order to be interpretable as a topic fol\o\vs from the very a sentence containing a topic expression by the position of the topic
defmition of topic in terms of pragmatic aboutness and relevance. For a referent on the follov,ring scale:
statement to count as information about some topic, the speaker must
assume that the hearer finds this statement relevant with respect to this (4.34) THE TOPIC ACCEPTABlUn· SCALF

topic in the context of the speech situation. But for a statement to be active most acceptable

I
relevant with respect to a topic, this topic itself must be of CURRENT accessible
unused
INl'EREST. Now for some topic to be of current interest, it must obviously
brand-new anchored
be assumed to be "current," i.e. it must either be already established in brand-new unanchored least acceptable
the discourse or it must be easily relatable to one that is already
established. Of course, the extent to which a topic may be considered The most easily processed and therefore cognitively speaking most
current ultimately depends on the speech participants, and under certain acceptable sentences are those v.'hose topics are highest on the scale, i.e.
conditions topics may be interpretable as current even though they whose topic referents are ACTIVE in the discourse. Such referents are the
haven't been brought up in the current discourse. Thus the requirement PREFERRED topics because the mental effort necessary to process sentences
of cognitive accessibility follows from the very definition of topic in terms containing them is not increased hy the additional ta'>k of assessing the
of the relation of aboutness. One cannot "add" information about a topic REFERENT, by retneving it from long-term memory or by drawing
·referent unless th.is referent is in some important sense already available inferences leading to its assessment Chafe ( i 9R7J ca!Js the cognitive etTort
in the discourse as a starting point. necessary to interpret a discourse-active referent a .. lov.' cost" effort
, To conclude this section on the relationship bet,veen topic and the Since active referents are normally unaccented and pronominal {cf.
cognitive states of referents, I would like to emphasize again the Section 3.3), the preferred topic expre<;.:;i0n !San U~ACCENTED PRONOMINAL
fOllowing points. Given the fact that linguistic expressions \vith "old" (or inflectional or zero) morpheme. It is this cognitive preference for
,rCfercnts .can be either topics or foci in a sentence, there CJ.n be no nne-to- active topic referents that ;:iccount<; e.g. for the preferred clause type in
._one correspondence between pragmatic relations and pragmatic proper- spoken French. \vhich has a bound pn1n(1un rather th;in a lexical NP in
ties of referents. Therefore to as5erL as is often done in discussions of initial subject po.:;i1ion (see Lamhrccht !986h. c-hapter 61
 166 Pragn1atic relations: topic Topic and the menial representalions of referenu 16l•

Less easily interpretable but still acceptable and indeed frequently formedness on the sentence level, even in a language like English. It:~
occurring topic expressions are those \vith ACCESSIBLE referents. In the been observed (Perlmutter 1970, Kuno 1972) that in English a subject'Nl"
case of accessible topic referents. the mental effort necessary to interpret with an indefinite article cannot occur with certain stative predicates~.
Perlmutter cites this example: : -~
the proposition which expresses the new information about the topic
• .' J !°I'
must be performed simultaneously with another processing task, the task
(4.35) ( = Perlmutter's (25)) *A boy is tall
of remembering, inferring, or otherwise determining the referent of the
topic expression. I will argue later on that these two cognitive tasks, that
I believe the reason for the unacceptability of this sentence lies in. the 'fi~
of interpreting the information conveyed by a proposition and that of
that it is difficult to imagine a context in wbfch it would be inform~-~~?!
determining the referent about which the infonnation is conveyed, are
predicate tallness of an unidentifi_e~ subject referent. Such ~.n~~~
best carried out separately, i.e. not within the same clausal processing
violate the most elementary cond1tlon of relevance. The unmterprsi...
unit (see Section 4.5.1 )_ In spoken (and to a certain extent in written) ability of (4.35) is aggravated by the fact that the indefinite noun phl;asC·(i
language, the separation of these two tasks is often reflected in the syntax ·· ··LI
boy cannot easily be construed as having a generic (and therefo~,
of the sentence,
identifiable) referent, which could then be interpreted. as the topic of a
A borderline case of pragmatic acceptability arises when new
gnomic statement. If the referent were interpretable as generic;. the
information is expressed about an UNUSED (i.e. identifiable yet inactive)
sentence would become acceptable, even with a stative predi~tp.
topic referent. The acceptability of sentences containing topic expressions Sentences like A boy is a boy or even A boy wants to be ta.II do riOi',
with unused referents varies widely with the language, the type of pose the same difficulties of interpretation as (4.35). · · .....
discourse, and the speech situation. The cognitive effort required in this
Thal the unacceptability of (4.35) is indeed due to the pragmatic
' case is of relative "high cost" because, in addition to processing
indeterminacy of the topic referent, i.e. to a difficulty of interpretation~
propositional information about some topic, the interpreter must
and not, as originally claimed by Perlmutter and others, to some'
I determine the referent of the topic itself, which was not previously
semantic, let alone syntactic, incompatibility between indefmite noun'

I:.
I
made available in the discourse. Of course, some unused referents may be
easier to access for an interlocutor than others, and the acceptability of
the sentence will vary accordingly.
phrases and certain types of stative predicates becomes clear if we'
con1pare (4.35) with the following modified version:

Clearly unacceptable as topics are BRAND-NEW referents, i.e. referents (4.36) A boy in my class is real tall.
' which are unidentifiable for the hearer at the time the new information is
conveyed. about them. This type of unacceptability is easily accounted for Example (4.36), with its anchored brand-new referent (see Section 3.2.1)1 ·
in terms of Keenan's Functional Principle. lf a hearer cannot mentally is clearly more acceptable in isolation than (4.35), even though its subject
identify the referent of the topic, she cannot determine whether the noun phrase (a boy ir1 my class) is still formally indefinite, and even
I predicate is true of this referent or not. This is but a fancy way of saying
that the hearer cannot make sense of the piece of propositional
though its predicate is still stative. The difTerence in acceptability between
(4.35) and (4.36) is predicted by the Topic Acceptability Scale. By adding
information she is presented 'sith. Senti;;nces containing such topics are the phrase in my class to the indefinite noun phrase a boy, the unspecified
in a sense incomplete pieces of 1ni"om1ation. A sentence with an set of all boys in the universe of discourse of the speech participants is
unidentifiable topic referent forci.:\ :.i heJ.r<.:r to "put the predication on reduced to the much sn1aller set of all boys in the speaker's class, i.e. a set
hold," so to speak, until she find:- out wh:.it )he is receiYing inforn1ation which is referentially linked (or anchored) to the identity of the speaker
about. This explains V.'hY 1nany languages have grammatical constraints herself As a member of this set, the referent becomes more identifiable,.
against indefinite NPs 1n initial \Ul_1J<-'l'l (1.t' unn1Jrk.ed topic) position. hence more easily interpretable as a topic. The pragmatic-semantic
In some cases. the pragn1at1c--."n1.1nl 1c un:u.·-ceptability resulting from difference between the two indefinite noun phrases in (4.35) and in (4.36).
insufficient recoverubility uf J t..:ip1c rr:ft:r<.'nt can lead to perceived ill- and the diO'crence in acceptability it entails, is another piece of evidence
 Topic and the mental representations of referents 169
168 Pragmatic relations: topic

for the need to distinguish the cognitive category of identifiability from Instances of acceptable sentences with brand-new subject referents may
the more language-specific category of definiteness (cf. Section 3.2). also be found in thetic sentences of the EVENT·REPORTING type. While the
Notice that the Topic Acceptability Scale in (4.34) is meant to account scale in (4.34) predicts the low acceptability of sentences such as (4.35)
only for differences in sentence acceptability which are due to differences (•A boy is raff), it does not not make the same prediction for utterances
in the mental representations of topic referents in a discourse. It does not such as the following:
account for certain other factors which have been shown to influence the
(4.38) A eov was run over by a CAR!
choice of topics in discourse, such as the animacy hierachy discussed in
Comrie (1981). the natural topic hierarchy discussed by Hawkinson and The greater acceptability of this sentence compared to (4.3_5), at ~east
Hyman (1975), the case hierarchy discussed by Giv6n (1976), etc. These' when considered in isolation, is due to the fact that a dynamic predicate
hierarchies, which have to do with the inherent semantic properties -'Q'f such as be run over is more readily interpreted as expressing an event than
~· ~ccount primarily for the likelihood of a referent's becoming a topic a stative predicate like be tall. The subject NP in (4.38) is not construed as
m a, discourse. They do not account for the different degrees of sentence a topic but as a participant in an event. Since no aboutness relation is
acceptability captured in (4.34). · intended in this sentence. the interpreter of the sentence does not feel the
It is important to keep in mind that the constraints expressed in the need to mentally identify the referent of the subject NP in order to assess
Topic Acceptability Scale are only meant to account for those sentences the relevance of the information expressed in the predicate. Hence the
which contain TOPIC EXPRESSIONS. They do NOT hold for NPs whosd greater naturalness of the sentence.
referents are not topics. With respect to non-topical NPs, (4.34) makes a· The acceptability of sentences \Vith initial indefinite subject NPs, such
different prediction, which is a corollary of the prediction on topic. as the English examples in (4.37) or (4.38). varies from language to
acceptability: if a constituent has a referent which is clearly NOT accessible language. The more a language associates topic function v.'ith subject role
in the context, in particular one that is unidentifiable, and if the sentence' and initial position, the less acceptable such sentences will be. For
is nevertheless of normal acceptability, there is a good chance that the example in those Romance languages v.'hich permit subject-verb
constituent is not a topic expression in the sentence. This prediction is of inversion, the non-topical subject NPs of thetic sentences must appear
particular interest in the case of SUBJECT NPs, because of their unmarke·d' in post-verbal position (cf_ e.g. Hatcher 1956 and Contreras 1976 for
topic status. Spanish, \Vandruszka 1981 for Ital!an. and \Vehr 1984 for Romance in
Instances of acceptable sentences whose subjects have unidentifiable or general). In French, \vhere subject-verb inversion is syntactically
otherwise highly inaccessible referents are commonly found in THETJC constrained, the bi-clausal avo1r-construction 1s often used instead (see
sentences, in particular those of the presentational type. Given the Section 1.3). in which the non.topic NP appears p0st·verbally in the first
discourse function of presentational sentences, the occurrence of clause. Notice that in all these lang.uage'i the position after the verb is the
unidentifiable subject referents needs no explanation. An example of an position normally reserved for ORJECTS, \vh1ch are the unmarked focus
acceptable sentence with an initial subject NP \vhose referent is brand- constituents. Marking a subject NP syntactically as non-topical is thus
new is the fragment in (4.37): tantamount to stripping it 0f its most important unmarked-topic feature.
which is preverbal position. by providing it v,·ith morphosyntactic and
(4.37) ... and then a BOY came in . prosodic features normally found on objects (see Lambrecht 1987c).
Even though (4.37) and (4.38) illustrate perfectly acceptable English
In (4.37), the NP a boy is not a topic because the communicative purpose utterances, there is a tendency even in English to mark the non-topical
of the sentence is not to convey information about some bov. but to status of brand-nev.· subject referents by syntactic as v,•ell as prosodic
introduce an individual into the text-internal v.•orld.Jo As -in other means. The grammatical conqructions perhaps mo'it frequently used in
examples we have analyzed, the non-topical status of the subject is English for introducing hr::1nd-nc\~- referents are the Jc1ct1c and existential
expressed prosodically (see Section 5.6.2). rherr-construct 1on and the deictic hcrc-ct1n~truction. in \Vhich the subject
 170 Pragn1a1ic re/a1io11s: topic Topic and the mental representations of referents 171

NP appears after, rather tban before, the verb, the preverbal position ban"ane "There ate a boy a banana" is ungrammatical, the structurally·
being filled by a locative element. I will return to sucb constructions in identical // esl arrivi un garpon ce malin "There arrived a boy this.
the discussion of presentational sentences in Section 4.4.4.l. For unused morning" is unobjectionable. . .
referents, English may also resort to locative inversion constructions In those English constructions in which the non-topical stat~ of thf;:
involving verbs of motion, such as Here co1nes rhe sun or In hopped rhe subject NP is expressed via prosody alone, the constraint ag&n.st.
rabbit (see e.g. Bolinger 1977, Van Oosten 1978, Green 1980), although simultaneous introduction of two or more NPs with iUactivc; · referCnts
such sentences tend to be stylistically marked. is also to some extent grammaticalized. As we noticed -~~ef (SectiOµ,
There is a natural restriction on the number of unidentifiable or 4.2.2), formally unambiguous thetic marking is possible onIY ,wi~ ce~
inaccessible non-topical referents which can be introduced within one IITTRANSITIVE predicates (see also Section 5.6.2 below)." ~ple· (4.3?)1

sentence or clause. In Section 2.2, I quoted the sentence A clergyman's lists a number of attested English utterances containing intransitive
opened a belling shop on an airliner (example (2.10)), in which all three event-reporting sentences with brand-new or unused .(i~~cf:ive · oC
arguments are indefinite NPs with brand-new referents. Such sentences accessible) subject referents:
are pragmatically so anomalous that they can be used only in special
contexts (see p. 345, note 17). In spontaneous discourse, sentences with (4.39) a. E\ery time I went over to his house a major CATASTROPHE happened.
non-topical subjects strongly tend to be IITTRANSITIVE, as in (4.37) and in b. I had a problem with my car. The BATTERY went dead.
the examples oftbetic sentences discussed in Section 4.2.2. c. (SJCting ar a computer terminal:) Oh shir! The
SCREEN'S &oing dead.
d. If you was back East and you saw a sky like that you'd know SNOW
In many languages this cognitive constraint on the number of
was coming.
inaccessible referent~ per clause is grammaticalized. For example spoken
French has a constraint on event-reporting sentences such that one NP
argument (the subject of the corresponding canonical sentence) must If the sentence has a direct object (other than an unaccented pronoun).
appear in a clause of its own. Any additional NP must be an argument of and if the subject is a definite NP, the sentence becomes am~iguous
a subsequent clause. In Italian and Spanish inversion sentences of the bet\veen the event-reporting and the unmarked topic-<:omment reading
event-reporting type no lexical direct or indirect object NP may cooccur (The BOY is thasing the CAT again). lo those cases where more than one
with the postverbal subject (see Wandruszka 1981, Lambrecht l987c). 31 lexical NP carrying a focus accent occurs in an event-reporting clause, the
The same restriction holds for the thetic constructions involving second NP is typically in a preposiliooal phrase (as in example (4.38)), in
"impersonal" es and ii in German and French and for the English particular one with a locative case role. Some attested examples are
inversion constructions mentioned above. For VS constructions, the quoted in (4.40):
constraint against cooccurring NPs with inactive referents can be
explained structurally, as a result of the fact that the language in (4.40J a. ~fommy, mommy~ Diego's SHOE /ace is stuck in his BIKE!
question does not tolerale the sequence V-NP-NP. There is no b. You know Tim redid lb.e storefront? A CAR went through the WINDOW.
c BACK L!GHT's out, on RIGHT HAN[) St[)EI
contradiction between such a structural account and the pragmatic
account presented here. Notice hov•ever that a pragmatic explanation is
needed to account for the fact that PRONOMINAL NPs, i.e. NPs \Vith active (Sentence (4.40c) was shouted by a bus driver to the driver of another bus
referents, often cooccur more freely \\'Ith non-topical post verbal subjects he had been rollO\~'ing.) Notice that in these examples the locative focus
than lexical NPs (see example (4.20) and discussion, and Lambrecht expression could be on1itled and the result would still be a possible thetic
1987c). A purely syntactic account of the unJcceptab1Lity of V-NP-NP sentence. The natural occurrence of such t\\'O-accent sentences is a feature
sequences i.s also unable to account for Lh"°" fact that such sequences are distinguishing English. whose constituent organization is to a high degree
acceptable if the second NP is an ad June\ rather than a complement of the grammatically controlh:d, from a language like (spoken) French, in which
verb. For example, while the French :.enlencc ;,//a niangd un garron une sentences \Vlth pr~vcrbal focal subjects do not naturally occur.
 174 Pragmatie relations: topic Topic and rhe mental representations of referents 175

information about the individuals who did the calling but only about the are topic expressions; in marked presuppositional structures: they appear
person who received the call. The interpretation of they as non-topical is . by default. The dual function of subject morphe~es is no doubt
co,_nfirmed in the approximately synonymous passive version in (3.29b) explainable in tenns of very general principle~ of coding econ~my (or
Pat said she was called twice, in which the agent doing the calling is left "grammatical ecology," as it is now sometimes called). It IS more
unexpressed because it is unknown or unimportant. Sentences (3.29a) and economical for the grammatical system to use the same morpheme for
(3.29b) are semantically equivalent in a way in which the members of the two functions than to have a different morpheme for each function. This
structurally similar pair John called Afary and J..fary li'as called are not. does of course not entail that there cannot be languages in which the
· To fmd examples of unaccented pronominals which could be topics by 32
distinction is fonnally marked.
the semantics of the predicate but which cannot be topical because of the The problem of the functional ambiguity of unaccented ~rono~~als
presupposi~onal structure of the clause in which they occur we may again also arises with OBJECT pronouns. Consider the two superficially s1m1lar
tum to ~hetic sentences. Such sentences may contain unaccented subject ·French sentences in (4.42):
pronom.inals even though their lexical subject NPs are not topics. For
example in the Spanish VS sentence Lleg-6 JUAN "JUAN arrived," the NP (4.42) a. Je t'a1 vu TOI "I saw YOU."
b. Je fai vu, toi. "I SAW you."
Juan is not a topic constituent; therefore the third person suffix -6, which
counts as an unaccented pronominal in my analysis, has no topic referent Example (4.42a) can be thought of as ansv.·ering the question "Who did
to refer to, hence itself cannot be a topic expression. The same you see?", \\'hile (4.42b) might ro1lo\v the command "Stop hiding:" In
observation applies to the inflectional suffix -a in the Italian event- sentence (a), the unaccented object pronoun t" (re) cooccurs with a
repOrting sentence Squill-a ii TELEFONO in (4.19b). strong pronominal focus NP \\'hich occupies the position of_direct object
Such apparent counterexamples to my claim that unaccented of the verb voir. Given the function of the sentence, and given that the
pronominals are the preferred topic expressions can be accounted for English gloss of (a) contains no overt or covert unaccented pronominal in
by interpreting the subject pronominals in such sentences as DEFAULT the second person, it would seem misleading to interpret the pronoun 1 ·
morphemes which are required by the grammatical system of the as a topic expression. Instead, r' functions here as an agreement marker
.language. Most linguists would agree that the subject pronouns of of sorts, the object argument being toi. In sentence (b) on the other hand,
weather verbs in English, French, or German are required not for the same unaccented pronoun, occupying the same position. is itself the
semantic or pragmatic but for structural reasons. The reason for the direct object argument, v.•hile the post-focal strong pronoun toi is a
j>resence of the subject pronoun in lt 's cold is not the same when this syntactically optional antitopic NP. In (b), t' is thus a topic pronoun of
sentence is used to describe a meteorological condition as \\'hen it is used the preferred type.
to describe the temperature of someone's hand, even though the position The status of unaccented pronominals. in particular inflectional
and the morphology of the pronoun are the same in the two situations. A morphemes and bound pronoun<>. as either non-argumental agreement
similar situation obtains with sentence-initial there in English, \vhich may markers or fully argumental pronominal morphemes, or as both, has
function either as a subject place-holder (There arri1·ed three soldiers) or been the topic of some debate in generatt\'e syntax (see in particular the
as·a·Jocative argument expression (There is your brother). analyses in \'an Valin 1985, Jelinek 1984. and Bresnan & :-...1chombo 1987,
Whatever explanation is used to account for the dual function of such \\'hich try to overcome certain shortcomings 1n earlier formal analyses in
morphemes will in my opinion also account for the dual function of the \Vhich unaccented pronominals \Vere not considered argument expres-
_person-number morphemes in in!lectional Janguage5. The default status sions). It seen1s to me thdt a satisfactory explanation for the inherent
Of'the unacceilted pronominals in the above-quoted thetic sentences is functional ambiguity of such morphen1es is possible only within a
Consistent with my analysis of subjects as unmarked topics (Section theoretical frame\\'Ork \Vhich doc\ not force the linguist to decide on
4.2.1). In sentences with unmarked presuprositional structure, 1.e. in FORMAL grounds alone \vhethcr an un.iccented rronom1nal is an
topic-comment sentences, these unaccented suhject-marking morphemes agreement-marker (ir J. rront1m1na! argun1ent hut \\·h1ch takes into
- 176
g- -z
Pragn1atic relations: topic
·7· -z E

account semantic and information·structure factors. The markedness

Topic and ihe menial represen1a1ions of referen.ls

Two such topic~promoting constructio ns are the previously discussed

J 77

approach suggested above makes an explanation of the status of such PRESENTATI ONAL constructio n and the constructio n often called (left"'",a.J1,(i
pronomina ls possible to the extent that it can account for the inherently right-) DETACHMENT or DISLOCATION. Both constructio ns are illustrated in
vague or "ambifunct ional" nature of unaccented subject pronomina ls by the following text example, which was originally used by Giv6n (l976)..iil
treating them as unmarked topic expressions. Without pursuing th.is his discussion of the diachronic rise of grammatic al agreement ..The
complex issue any further here, I will assume that unaccented italicized expressions indicate the referent whose activation state is being
pronomina ls are NORMALl Y topic expressions and that exceptions to promoted from non-active to active state: n.
this general tendency can be accounted for in a principled way.
• •• J
The above observation s regarding the argument or non·argurn ent (4.43) Once there was a wizard. He v.:as very wise, rich, and was ma.med to a
·"'If'-~
beautiful witch. They had two sons. The first was tall and brooding, he

.
status of "agreemen t" pronouns leave open the question of whether
spent his days in the forest hunting snails, and his mother was afraid of
bound pronouns which cooccur and "corefer" with an argument NP in a
him. The second was short and vivacious, a bit crazy but always game.
single clause are REFERF.N'TIAL expressions or not. This question is relevant
with regard to the theoretical problem of whether a single semantic
Now the wi;;;w-d, he lived in Afck:a. . r> . ,
"
argument can be instantiated twice in a single clause by referential
constituent s. Without being able to go into the necessary detail here, I
The first sentence in this text, Once there was a wizard. is ·''a
PRl:SENTATl ONAL sentence. The last sentence, A'ow the »'izard, ~ live'd i_Jf
(
.
would like to suggest that the answer to his question may be "yes." The Africa, is an example of LEFT-DETACH/i.tENT. I will start with the discussion
cooccurren ce of two coreferenti al pronomina l expressions in examples like of the presentatio nal constructio n type.
(4.42a) could be explained in tenns of the model of two discourse worlds
sketched in Section 2. l. For example, the bound pronoun t' in (4.42a) 4.4.4. l Presentatio nal constructio ns
could be seen as representing the referent in one discourse world while the As I obmved earlier (Section 4.4.2), the proposiuons expressed in
free pronoun toi would represent it in another. Both expressions would presenlatio nal sentence's are THETIC. The basic communica tive function Or
refer to the same individual, but by virtue of its role in two different such sentences is not to predicate a property of an argument but to
discourse worlds. Hence the need for two referential expressions in the introduce a referent into a discourse, often (but not always) with the
same clause. purpose of making it available for predication in subsequent discourse,
The basic discourse function of presentatio nal sentences is defined by
Hetzron (1975) as that of
4.4.4 Topic promotion
calling special attention to one element of the sentence for recall in the
The view of unaccented pronomina ls as the cognitively preferred topic subsequent discourse or sJtuation. This recall may be needed because the
expressions makes it possible to interpret a number of crosslinguistically elen1ent is going to be used, dHectly or indirectly, in lhe ensuing
widely attested grammatic al constructio n types as pragmatica lly discourse. because what is going lo be said later has some connectiOo
motivated structural devices "'hose basic function is to PROMOTE with the element in ques!ion. -or because thal eJeruen! is relevant to
referents on the Topic Acceptabil ity Scale from non-active (i.e. brand- what is going lo happen or be done in the reality. (Hetzroo 1975:374)
new, unused, or accessible) to active stale in the d1s.::ourse and hence from Hetzron does not n1ake use of the concepts of actJvation and referent
lexical to unaccented prononuna l coding in the sentence. By promoting promotion, but his notion of "recall in the subsequent discourse" is
the state ofa referent in this \\'ay. the'.>e constructio ns 1nake it possible for clearly related to these concepts. The reason \vhy lhe referent of the NP a
speakers to adhere to 1he preferred unaccenied pronomina l topic type. tt.rizard in (4.43) can be expressed at the beginning of tbe second sentence
Their grammatic al function is lo n1atcf1 the n;quiren1ents of syntactic in the preferred topic form he is that this referent was lexically expressed.
structure and informatio n :-Lru..:turc 1n <:ascs \~·here the two do not ·and thereby prag1natically activated, 1n the immediatel y preceding
naturally coincide. sentence. The purpose of the first sentence in the text is thus to
 178 Pragmatic relations: topic Topic and the mental representations of referents 179

introduce, or "present," the previously inactive, brand-new referent "a · · 1n· sect1ons
discussions of existential presupposition · 3.2· I and 4 ·3) · Under
wizard" in the text-internal world and thereby to make it discourse-active this account, the communicative function of the sentence One~ there was
and ready for recall in subsequent sentences. a wizard \vould be to assert that a particular wizard once existed. Such
It is an empirical fact of natural language use that sequences consisting sentences are therefore often referred to as "existential" sentenc_es. F~o~
of a presentational clause followed by another clause expressing the point of view of information-structure analysis, the label "ex1sten_t1al.
infonnation about the newly introduced referent are strongly preferred is somewhat misleading. Mere assertion of the existe_nce of some entity ts
over such syntactically \Veil-formed sentences as the following, in which a rather special kind of speech act \Vhich is of li~1ted use in .everyday
an inactive topic referent appears directly as the subject NP of the communication. It is difficult (though not impossible) to conjure up a
sentence: situation in i,vhich a statement like "There are cockroaches" would be
made with the unique purpose of stating the existence of such creatures.
(4.44) A wizard once v.·as very wise, rich, and married to a beautiful witch.
Such a statement would be most naturally used in situations \vhere the
The strangeness of(4.44) follo\vs from the fact that topic expressions with existence of cockroaches may already be taken for granted and \vhere the
unidentifiable referents are least acceptable on the Topic Acceptability purpose of the speech act is to introduce .the. NP referent_ into. the
Scale in (4.34). If the sentence is not perceived as ill-formed, especially if discourse world of the interlocutors by asserting tts PRESENCE 1n a given
the word once is removed, this is due to the fact that the scale in question location ("Don't go into the kitchen. There are cockroaches''). From the
measures pragmatic, not necessarily syntactic acceptability. Moreover, as discourse-pragmatic point 0f \·ie\\", it 1s therefore p_referable to 1nte_rpret
we will see later on, it is possible to pragmatically accommodate the function of such sentences as that of presenting or tntroduc1ng a
unidentifiable referents to some extent. 33 referent into the ·'place" or ··scene·· of the discourse and thereby of
Because of the discourse function of presentational clauses, \Vhich is to raising it into the addressee's c0nsc1ousnes<>, rJther tban of asserting its
promote brand-new or unused referents to active status, the expressions mere existence '~
used to code the "presented" referents are indefinite or definite ACCENTED This interpretation ha~ the addition:..1\ advantage nf explaining the
LEXICAL NOUN PHRASES. Presentational NPs may not normally be formal simi\anty bet\veen the exi'>tcntial c0nstruct1on :..1nd the DEICTIC
pronouns, since the referents of pronouns are already active. 34 In some presentati 0 nal construction discussed in Section: l 81)th constructions.
~anguages, presentational clauses are used exclusively or \vith strong the deicttc and the existential. are presentational 1n that both serve to
preference for the introduction of brand-new (i.e. unidentifiable) introduce previously unidentifiable nr inactive referents into a discou_rse.
referents. For example in the so-called "inverted word order" construc- The main pragn1at 1c difference het\\een the \\\'O 1s that de1ctK· 1herc points
?on in Chinese, in the English existential there-construction, and in those to a referent in the te.xt-E>.:TFRN \l \\'orld. ,,·hereJs existential there
Gennan and -French presentational constructions which involve the introduces a referent into the l"'ll·R'-'AL \\'orld (lf the text. This
dliMn:iy subject markers es/ii, only or mainly NPs with brand-new interpretation of both construcuons a~ presentJ.t10nal 1s consistent \\'ith
f'cfeients may occur. (For the French ii-construction see Section l 2 the alreadv mentioned fact that in snme languages !e.g spoken French)
C'~mple (3.18) and Lambrecht 1986b, Section 7.4.4.) In the case-·of the "pres~nted NP" of J.n existential (l1nstruct1on can he a definit_e
EOglish, German, and French, this entails that mainly indefinite NPs are description and even a proper name. i.e. an e;..press1on \\'hose referent ts
tolerated in these constructions. This kind of quasi-grammatical not only presupposed to e-.;ist \:iut al\() hi he kno\\'n h' 1he addre~see._ ln
constraint is directly explainable in terms of the Topic Acceptability such cases, mere assertion of the existence 0f the referent \\r1uld be a kind
Sea.IC. Given that brand-new topic referents are lo\vest on the scale, the of tautol0gy. It 1s also \\'ell-kno\\·n that existential clauses often begin
need to avoid sentences having such topics is greatest Therefore \\'ilh a place adverbial, such as English there. Germ3n da. French y, _etc.
grammaticalization is most likely to arise in thn~e case<:. making the claim e\'cn more con1pell1n~ that the presentational. locat1on-
·. One common account of the meaning or rre~entJli(1n~d clause~ is that oriented function of the CL'n5truc11,1n 1s 1n fact the fundamental
they assert the EXrSTENCE of the referent (lr the p<1'd\crh-il !'.'P (cf the communicative function ,1f e.x1stent1al ~cntcnres ·"'
 J80 Pragn1a1ic re/oJions: topic Tapic and the mental representations ofreferenls l~l

Often the grammatical relationship between the presentational clause or state, is merely made available for predication in subsequent clauses by
and the subsequent clause in which the referent just introduced appears being raised into the addressee's consciousness.
as an unaccented pronominal topic expression is one of syntactic Presentational sentences sometimes contain intransitive predicates (or
dependency, the second clause being grammatically subordinated to the transitive predicates with unexpressed object arguments) whose subject
first. One common construction type illustrating this phenomenon is arguments can be said to be agentive to a certain degree. In such cases.
shown in this variant of our fairy~tale beginning: the agentivity of the predicate is subordinated to the presentational
function of the proposition and the predicate is in fact pragmatically
(4.43") Once there WilS a wizard who was very wise and nch. construed as non-agentive. As an example of such a "pseudo--agentive ..
presentational sentence consider the Italian inversion sentence (4.45):
(see also example (4.5) above). Unlike the beginning of (4.43), where the
(4.4S) Ha telcfonato GIOVANNI. "GIOVANNI called."
t\.,.'o clauses are juxtaposed, in (4.43'') the second clause is grammatically
subordinated to the first and appears in the form of a relative clause This sentence could be uttered for example to infonn an addressee that in
whose antecedent is the presented NP. The preferred topic expression is her absence the person named Giovanni tried to reach her. The utterance
now the relative pronoun who in the second clause. I will refer to the is a way of introducing "Giovanni" into the universe of discourse by way
complex construction illustrated in (4.43") as the Bl·CLAliSAl. PRESEf'ITA· of mentioning the fact that he called. Example (4.45) does NOT have the
TIONAL CONSTRUCTION. The relative clause in this construction, even purpose of conveying information about the caller as an agent involved in
though it is grammatically marked as dependent, exhibits a number of some action. If such information were intended, the utterance would have
syntactic, semantic, and pragmatic fea1ures '"'<'hich characterize it as a to be of the topic-comment type and would likely be of the form
special type which could perhaps be characterized as "dependent main
(4.46) Giovanni ha TELEFONATO.
clause" (see the formal analyses in Lambrecht I988b for English and in
Section 7.3 of Lambrecht 1986b for French). These features characterize or perhaps
the two-clause structure in (4.43") as a grammatical construction which is
c: uniquely dedicated to avoiding violations of the Topic Acceptability
(4.47) Ha TElEf'ONATO, Giovanni.

Scale. In English. bi-clausal presentational constructions may also (with the lexical NP in antitopic position). Notice that there is a limit to
involve non-finite {participial) clauses, as in Titer<: 1vas a dog rur1ning the degree of agentivity a predicate can have to be exploitable as
down the s1ree1 or There was a man arrested by the police. In Chinese, a presentational and thus to be able to appear with presentational syntax.
serial verb construction is used whose first verb is the existential predicate or prosody. This upper limit is hard lo define, but it clearly exists. For
you (see Li & Thompson 198 L61 I ff, LaPolla 1990: l 1511). example while the gloss .JOHN called in (4.45) may be understood as
The most common and grammatically most clearly marked presenta- presentational in the sense described, a transitive sentence with subject
tional clause type is characterized across languages by the presence of a focus such as JOHN called his wife cannot be so construed but can only be
limited set of predicates whose argurnents have a highly non-agentive and understood as an identificational sentence, with 1011N as an "argument
often locative case-role, such as "sE", "BEAT," "LI\'£," "ARRJVE," "HAVE:," focus" (Section 5.2.3) and the rest of the proposition pragmaticall;i
17
"sE:E," etc. (see Section 4.:2..2, example (4.18) and discussion). The presupposed. 3a
crosslinguistic predominance of such predicates is a natural consequence
of the basic discourse function \~·hich al! presentariona! senleaces, 4.4.4.2 Detachment constructions
whether deictic or existential, have in comn1on: they do not predicate In order to promote the representation of a referent from non-active to
some property of the NP referenl bu\ they assert the presence of the active slate in the addressee's mind and thus to allow a speaker to code
referent in the (extcrna! or internodJ !e.\t v.-or!d. The nen·Jy introduced the referent as a preferred topic expression, tt is not always necessary to
referent, rather than being depicted a~ participating in some action, event introduce it in a presentational clause of its own. From a certain degree of
 183
Topic and the n1en1al represe111a1ions o.( referenls
182 Pragmatic relations: topic
. h t 0 pie 40 But while in the
constructions serve to esta bl is a new
resentational sequence the referent of the NP is b~and·new or at least
pragmatic accessibility on, it is possible in many languages to code a not-:; ·

yet-active topic referent in the form of a lexical noun phrase which jg

placed in a syntactically autonomous or "detached'' position to the left
~nused in the detachment case it is usually cogn1uve\y ACCESS~BLE. (ln
· . f h d t ched NP referent 1s due to
or. less commonly. to the right of the clause which contains the (4 .4 3) the textually accessible status o t e_ ea .
. . "[Chafe 1987] from its previous active state, caused by the
" d eact1vat1on · · · h
activation of other referents in the context, in our case_ the beaut1f~I \\'1tc
propositional information about the topic referent. The semantic role of
the referent of such a lexical noun phrase as an argument in the
and the two sons.) The detachment or marked to_p1c construction can
proposition is usually indicated via an intra-clausal "resumptive'~
then be defined pragmatically as a grammatical dev1c~ used to promote a
pronoun or other unaccented pronominal which is construed as
referent on the Topic Acceptability Scale from access1bl~ to active_ stat~s,
coreferential with the detached lexical constituenl. This intra-clausal
from \vhich point on it can be coded as a preferred topic expression. 1.e.
pronominal morpheme is of the preferred unaccented pronominal topiC
1
type (see Section 4.4.3), while the extra-clausal lexical NP is a MARKED as an unaccented pronominal. . .
It must be emphasized that this pragmatic char_actenzat1on accounts
type of topic expression. Traditionally, this syntactic construction has
39 only for \vhat I take to be the basic discourse function of the detachment
been referred to as LEFT and RIGHT DETACHMENT (or DISLOCATION). :I
construction. As we sa\\' in Chapter 3. it 1s not ahvays p_oss1ble_ to explain
Detachment constructions are common in the v.•orld's languages, even
variations in the morphosyntactic coding or referents d1rect_ly in terms or
in languages whose typological properties would seem to make NP,
corresponding variations in their activation or 1dent1fiab1\ity states. It
detachment difficult. For example Japanese and Turkish have well~
often happens that already actt\'e referents are coded 1n the form of
developed right-detachment constructions, even though they are strict
lexical noun phrases rather than unaccented pronom1nals. for example
SOV languages, whose verb strongly marks the right clause boundary
when it is necessary 10 avoid the ambiguity created by the presence of t\.vo
(see Kuna 1978 for Japanese. and Erguvanli 1984 and Zimmer 1986 for
or more competing active referents (cf. Section 3.3). Thus _the left-
Turkish). Detachment constructions are often considered substandard oi
detachment construction is often used to mark a shift .in at~ent1on from
at least inappropriate in formal registers. This is no doubt a consequence
one another or t\\'O or more alre<idy actl\'e tl1p1c re!erents. This
explains the frequent occurrence_ of _rRo~o..,nN,\L NP_s tn detached
of the fact that the canonical sentence type. in '""hich the subject is a full 10
lexical NP in argument position, has traditionally served as the basiC
positions (Ji,fe. /'ni hungry . .\fn 1 ,·ai fa11nl. Such detaLhed lexical or
pronominal NPs often have a "contra<.t1ve·· funct10n. 1n \\'h1c~1 case they
model of sentence structure (see below). In order to accommodale the
functional need for NP detachment, written languages often resort to
may be referred 10 as coNTR;\511\'E ror1c NP-; (see exan1ple (3 ..:.0) and the
constructions of the as-for type (see Section 4.3 above), which are . · · Sect,·on ').;;-. he\O\\') A.s f0r the right-detachment
detachment constructions in disguise. Interestingly, the markers introdu- d 1scuss1on 10 - .. .... · ·
construction {He {il'ed in Africa. the 111:ard). 1t is also often used_ for
cing such standard constructions are often both syntactically irregular
alreadv acti\·e or qua..,i-actl\'C referent'>. but 1t cdn nc\'er be used in a
contr;stive function The difference bet\\'eCn .!ef\-de1~1chcd_ ; 1nd ng.ht-
and semantically opaque (cf. English as for NP. French quant d NP or
pour ce qui est de NP, or German 1\·as NP anbe1riff1, all of \vhich are non·
detached constituents \viii he further d1~cu<;..,ed in Sccth'n 4
1

compositional). As is often the case in normative grammar, syntactic s metimes ::ictl\'C referents .1re coded a<; detached \e:-.ical n,)un phrases
0
even \Vhen no ambiguity anse.;; and \\'hen unaccented pn1n~Jm1nal c~d1ng
irregularity and semantic opacity are more readily tolerated than
transparent violations of canonic.al structures \vou\d tie sufficient te> identify the referent l)ne such ca..;e ol ~p
detachment lffvo\\·ing :1 n alre~1dy act1\'e referent I\ d1<>cu;;,.;;ed hy En\:
An example of a detachment constructinn not 1n\'0[\·1ng the 11s-.fOr
marker was presented at the end of the sh~1r1 te:-.t 1n (4 4Jl: 1'.'o\\' the (\986). llc,ing. the same text I quoted 1n 14.4:1l. En(,": 1....,h..;er\'e\ that th1<; text
wizard. he Jived in Africa. The text in (4.431 is ;:i gch1d 1llustra11nn nf \vhat I can he changed in <.uch a \\',1y \h<it ~1 \cft-dctach111..-:nt construction
becOnlL:S apprl1pri:-ite c\'t\1 w1th1-..ut ,1n~ 1nlCf\'Cn1n~ .te\\ _that \\OUld
take to be the basic pragmatic difference bet\>.Cen the detachment
construction and the presentational cnnstrucL1nn In be>th c(instruct1(1ns,
deaCl!\'ate the tL1p1C refcr<:nl 1\CfC I~ fn~·\ \,1fLl111 \ll (Jl\\l\1 \text
a referent is promoted from non-.:1ctiYc h' ;h:L1\'t' st;:itu~. and both
 184 Pragniaric relations: topic Implications for syntactic theory 185

(4.48) COl'llTEXT:Once there was a wizard. He \11as very wise. rich, and \l.'as of the relational network of the clause, appears either in a special;
married 10 a beautiful witch. He lived in a magnificent mansion by the intransitive clause of its own {as in the presentational construction) or is
lak.e. had for1y-nine servan1s, and owned an impressive collection of rare placed in a non·relational position altogether (as in the case of NP
books.
detachment»
TOPIC SHIFT: Now the wizard, he was \'ery ambitious. He had been
planning for years to conquer the world and finally he was ready. I will caU the grammatical principle whereby the lexical representation
of a topic referent takes place separately from the designation of the
According lo En~, the derachment construction is appropriate in {4.48) referent's role as an argument in a proposition the PRINCIPLE OF lHB
because it signals a SHIFT in what she calls the "topic of discourse," here a SEPARATION OF REFERENCE A.ND ROLE (PSRR) for topic expressions. The
change from the general description of the \Vizard 10 his plans to conquer communicative motivation of this principle can be captured in the form
the world. En~ here follows the approach of other linguists who define of a simple pragmatic maxim: "Do not introduce a referent and talk
.. discourse topic" in propositional terms. 42 about it in the same clause." There are two processing reasons for
Such subtle variations in the cognitive state o( the referents of detached adhering to this maxim, one speaker.oriented, one bearer-oriented. From
NPs do not affect the basic pragmatic distinction between the the speaker's point of view, it is easier to construct a complex sentence if
presentational and the detachment construction. Despite some possible the lexical introduction of a non·active topic referent is done
overlap, especially in the shady area of accessibility, the two construc- independently of the syntactic expression of the proposition about the
tions are in complementary distribution as far as referents at the extreme referent. (This is particularly evident in example (4.49) below.) From· the
ends of the Topic Acceptability Scale are concerned: active referents may hearer's point of view, it is easier to decode a message about a topic if the
not occur in presentational clauses, and brand-new referents may not task of assessing the topic referent can be performed independently of the
occur in detachment constructions. This distributional difference is task of interpreting the proposition in which the topic is an argument.
formally reflected in the fact that presentational NPs may not normally These processing reasons may ex.plain why detachment constructions are
be pronouns and that detached NPs may not normally be indefinite. so often restricted to the domain or spoken language. Indeed, the planned
character or written discourse makes such processing requirements less
stringent.
4.5 Implications for syntactic theory
• An example of spontaneous manifestation of the Principle of the
The above analysis of the relationship between topic and the mental Separation of Reference and Role is sho\.\.'Il in the following fragment
representations of discourse referents has certain implications for the from an election campaign speech by Ronald Reagan (San Francisco
study of syntax. In the present section I would like to discuss, in rather Chronicle, August 25, 1984). The constituents expressing the relevant
general terms, some of these implications. referents are italicized and indexed with subscripts:

(4.49) H'e 1 are lhe party of the new ideas. li'e 1 are the party of the ruturc. We 1
4.5.J The PrinL·iple oj. lhe Sepi.lri.llion r~l Rejl·rence 1111il Role are the party whose philosophy is vigorous and dynaatic. The old
stereotype of the kind of pudgy, s1ohd. 11ega1n·e Repubilcan 2-thcre may
As I observed earlier, the bi-clausal pn:'.'>entational construction and the be a fell' cartuon1SIS 3 around who 3 s.till want to portray us 1 as thal 2, but
detachment construction have 1n C<Jrnn11.1n thJ.1 they c.:iuse a referential rhey 3're lying through rhcir 3 teeth if rhe.v 3 do
noun phrase to appear else\vherC" than in the p1.•'.'>it1on assigned to it by the
canon1cal sentence 1nodel, in\\ hich c1l! .1rgun1cnls of a predicate appear as Of special interest here is a comment by the author of the newspaper
grammatical arguments at Lhc lc\'i:\ ._,r cL.1.use structure. These non- article concerning the portion of the text starting with the words The old
canonical configuration:::. thus ,1!lnw sp1.',1ker\ to separate tht: REFERRtNG stereotype . , v.•hich is the portion I am concerned with: "Somewhat
function of noun phrJ.ses fr•)m 1h..:: i.:.u -. 111 >:-..".L rc•!c their Jcnotata play as snappishly, the President departed from his prepared speech to add ... "
arguments in a prop1.1s1t1~1n. The k'.\IL"il <...'1.1n'.>t11ucnt. in::.tead of being part This com1nent indicates that this porlion of the text is an example of
 187
/n1p/ications for syntactic theory
186 Pragmatic relations: topic
. al hrase, whose distribution and co~stituent
(relatively) spontaneous speech, compared to the planned rhetorical referent, by means of a lex1c p S h lexical topic expresslons are
. h t of a noun phrase. uc .
clichCs at the beginning of the quote. structure is t a . Their semantic role in a propos1-
In (4.49) the first topic referent (and the discourse topic) is "the . REFE.RENCE-ORJENTED expression types. . f or their position in the
· b\e from their arm .
Republican Party." This referent is coded in the fonn of the preferred tion is often not recogn1za . ession designates the topic
d case the topic expr .
topic expression M'e throughout the first three sentences. The second topic sentence. In the secon · . ,. pronominal expression.
. Of DE1CT!CALL'' \ 1a a
referent, introduced in the detached NP the old stereotype of the kind of referent A.NAPHOR!CALL 1 . ORIENTED expressions. They
· · pressions are ROLE· ..
pudgy, stolid, negati-..·e Republican is an inferentially accessible referent, Such pronom1na 1 topic ex . ferent and the proposition,
t" 1 \"nks betv.ieen the topic re
whose inferable status is due to the relationship of polar opposition serve as gramma ica I . r h 'ferent as an argument. i.e. as a
between it and the preceding concepts "ne\v ideas," "future," "vigorous," by indicating the semantic role o t e re ssed bv the proposition.
. h · event or state expre - .
and ..dynamic." It is because of this inferable relationship that this topic participant 10 t e ac1ion. ' . · s are often not noun
. h minal topic expression
can be expressed in the form of a detached NP constituent. The third Syntactically, sue prono h d to another constituent of the
topical referent in this text, first introduced in the NP a few cartoonists, is phrases but bound morphemes attac e
entirely new to the discourse. Accordingly, it is first expressed as the focus clause. . ind !et us consider again the
NP of a presentational there-construction, after which it is coded with the With the above distinction in m . . (4 4~) In this
. our model discourse · - ·
unaccented pronominals who, they, and their. Notice that presentational detachment construclion in . d I, names the referent about
h d NP rlie i1·1~ar mere'
structure and topic-comment structure are combined here in a single discourse. t h e d e t ac e - . · e ne\\' information. The
construction, in which the presentational !here-construction expresses a v.•hich the fo\lo~ing proposition ~~:v:~~e;oh:nd represents the referent
comment about the referent of the left-detached topic NP (see the pronominal topic expression. he on . ·1 I . the Reagan text in (4.49)
· a propos1t1on. Simi ar y, in .
remarks in Section 5.2.5). as an argumen t in
· h 0 /d / COl~C0 11€~1
(
1
,.,,J ,,,~~ pud&r. ~.
stolid. 11ega111·c
I interpret the use of the two pragmatically motivated grammatical the lexical topic '. e s er · · 1--, ·ng it therebv making it
the topic referent l•) nami , .
constructions illustrated in this short text as a manifestation of the Republican esta bl is. h es. . Th rile plaved by this
d · subsequent clauses. e l ·
Principle of the Separation of Reference and Role: those topic referents in available for pre 1cat1on. ln . d. -ated via the rronorninal
(4.49) which are not yet active, and which therefore cannot be directly referent as an argume_nt in a propos1tton i:. in ll· I I
h -\a use 1r/io por1rar us a~ I id ·
coded. as preferred topic expressions, appear as lexical NPs OUTSIDE the topic expression t Iiat in 1 e o,; . •_ . e"><>ion tvpes is reflected in a
clauses which express the propositional information about them. In the The differe.nce bet\\·een the \\\·O topiL ex-pr . d b:h ··or. To mention
r d ff n grammatical form an c a\ l
presentational construction, the not-yet-active referent is introduced as number o ' . erenc~ t ted in the fact \hat a \ex1cal top1..: can l)c..:ur at a
the postverbal focus NP or a separate clause (there may be a few but two here, it is re ec . the ro i)sition about its
cartoonists around); in the detachment construction, the referent (the old certain distance from the c\:1usc expre.,h.s1\ng·1.n u:ac:C~ted pronominal
le (4 49) above), \\' 1 e '
stereotype of 1he kind of pudgy. stolid, negatil'e Republican) appears in a referent (see examp · . f the -\·:1.use acrQss languages,
. . ·1y a con<.tituent o
symactically autonomous, non-argument position to the left of the topic is necessan l d L' ·
to he morphosvnt;-1ctic~1lly integrate_d into
clause. In my analysis of spoken French (Lambrecht in preparation) I \vill unaccented pronom1na s ten S .ton 4.7 belov.·). It is also
. t or a <.entence (see ec \
Show that the Principle of the Separation of Reference and Role, along the predicate por ion · _ d to be noorrho\ogt-
. h 11 known fact that pronL)un-.. 1en
~th the formal distinction bet\\'een topic and focus, is a major factor reflecte d in t e \\'e - h c·:1.\ NP-.. often gL) \\-ithout
k d - <>S !':1.nguage<> \V i 1e 1ex.1 , -
determining the shape of the French sentence. cally case-_mar ·e acr0 .. er, \~961·96). This difference in mL)rphosyntact1c
The formulation of the Principle of the Separ3llt)n of Reference and case m.ark1ng. (cf. Greenh g .-. , Snee the ni:.un function of a
·t~nt \\1th n1\ ,tna 1\'> 1" · 1
Role makes it possible, and indeed nece~sary, to dra\\" a theoretical behavior is LL)n'> 1". c _ . ,_: h th., 1 'Pll' referent in thc J1<.course by
~pt" \P e,L\l' 1i"
distinction between two grammatically quite different strategies \\'hereby j
lt'tac. h e d \ex1"1l
· L, ' , .
c \
·f pr n)niin,i\ \L)pi..: e"-rn::s-.,1,)n
I, 't' 1he niain lunlttc'n (' ' 1 ' 1 l
a referent in the universe of discourse may bl" ..:oded as a l\)pi..: expression naming it, anL -..1nl . . - . r ,un1ent in ,1 rr•.lr•1,\11nn. the
is to indicate the rp\;: nl the l•'PlL ,1-. .111 ,1 !..!
in a sentence. In the first case, the l\)pic e.\pr.:~s1l•n "·\\H·~ the topic
 Implications for syntaclic 1heory 189
188 Pragn1atic relations: topic

positional and case-marking differences can be seen as natural 4.5.2 The PSRR and the canonical sentence model
consequences of the above-mentioned principle. There is an old and respectable grammatico-philosophical tradition
The above-made distinction between two kinds of topic expressions according to which certain types of sentences are better suited as models
requires a terminological proviso. Since a detached lexical topic of grammatical and logical analysis than others. The preferred sentence
constituent does not occupy an argument position in a clause, it is model in this tradition is one in which all arguments of the verb arc full
strictly speaking nol with the lexical topic NP but with the anaphoric referential lexical noun phrases. The tradition based on this descriptive
pronominal topic expression 1hat the pragmatic aboutness relation model goes back to Greek, Latin, and Medieval grammatical theory.
between the referent and the proposition is expressed (cf. the from Plato on, the preferred sentence type of grammatical analysis has
terminological observations in Section 4. l.2 above). It is therefore been the type Socrates currit "Socrates is running." This type was called
slightly inconsistent to call such a detached lexical constituent a "topic by Latin grammarians the "oratio perfecta," the sentence which expresses
NP." Rather it is a "topic-announcing" NP. However, I will continue to a "complete thought. " 43 The reason why the oratio perfecta came to be
follow the terminological convention (to whose establishment I have the preferred sentence model has no doubt to do with the fact that the
contributed), referring to detached constituents as topic NPs, i.e. NPs study of grammar was essentially the study of the laws of though~ rather
which appear in special syntactic positions, labeled TOP (left-detached than of the formal properties of sentences. It is clear that the mearung and
NPs) and A-TOP (right-detached NPs) in the present framework. the truth conditions of a sentence like Socrates currit are easier to state,
There is a striking analogy bet\>,•een the pragmatic principle postulated \.vit.hout context, than the meaning and truth conditions of a sentence like
here and the logical claim underlying the distinction between categorical Currit "he/she/it is running," \.vhich has no lexical subject. In the
and thetic sentences postulated by Brentano and Marty (see Section "subjectless" version, the act of running is predicated of an entity which
4.2.2). Recall that according to Marty the categorical judgment of the is not named v.•ithin the sentence. We therefore cannot tell, from the
subject~predicate sentence involves both the act of recognition of a
sentence alone, whether the judgment expressed by it is true or false.
subject and the act of affirming or denying what is expressed by the Lexically explicit sentences of the type mentioned above have typically
predicate about the subject. Since it involves these two independent been used also in modern linguistic and philosophical argumentations.
judgments or cognitive acts, Marty calls it a "double judgment." The Archetypal examples are Russell's The present King of France is bald
thetic judgment, on the other hand, involves only the recognition or (1905), which I discussed in Section 4.3, Chomsky's The man hit the ha!I
rejection of some judgment material, without predicating this judgment (1957), or Sapir's The farmer kills the duckling ( 1921 :Ch. V). S~p~
of some independently recognized subject. We can easily translate explicitly called bis model "a typical English sentence." It is c~a~actenstJ.c
Marty's logical argument into the present cognitive-pragmatic argument of the preoccupation v.·ith form in tv.:entieth-century lingu1sllcs that a
by saying that the "act of recognition of a subject" (Marty) corresponds great linguist should choose as a grammatical model a sentence whic~
to the act of establishing a topic referent in the discourse and that this act blatantly violates our intuitions concerning ordinary language use. It is
and the "act of affirming or denying what is e:'l..pressed by the predicate difficult (though of course not impossible) to imagine a situation in which
about the subject" (Marty) are t\vo cognitive tasks v.'hich for processing this sentence could be uttered felicitously. It is no doubt for this reason
reasons are best carried out separately, i.e. not v.·ithin the sa1ne n1inimal that linguists tend to misquote Sapir's example as The farmer killed the
clausal unit. Hence the emergence of presi.::ntational and detachment · duckling, v."ith the verb in the past tense, in an unconsci~s attempt to
constructions in natural languages. The preferred topic expressions, · bring grammar and the real v.•or\d a little closer together. . .
which indicate the semantic role c•f th.: \L1p1c ;.irg.u1nent in the proposition, I will refer to such sentences, in v.·hich all argument pos1llons of the
are unaccented pronominal~" 1.e e.\pre~s1ons for \\ h1ch the "act of,. verb, in particular the subject position, are filled v.'ilh lexical NPs, and in
recognition" of the subject i::. reJuceJ l\1 ,1 n11nin1un1 which these NPs appear in their unmarked position. as CANONICAL
SENTENCES. In languages like French or English they may also be
 190 Pragmatic relations: topic /n1plicationsfor syntactic theory 191

referred to as SV(O) sentences, if "S" and .... the rule what the information-structure-based approach takes as an
understood as FULL LEXICAL N 0 are-as usual-
tenn ..canonical sentence" is u~u~ ~HRA~Es (see Greenberg 1963). The exception.
I do not regard it as an embarrassment for my analysis that the
fromthewayinwhichlhav beee _ere.in a way which differs slightly
e n using 1t until n · II . Principle of the Separation of Reference and Role conflicts with a well-
I .3.2, where "canonical" was a quasi-synonym
. ow, espec1a yin Section
of" k d established descriptive convention as well as a fundamental theoretical
uses, the adjective "canonical" e k h . unmar ·e ." In both assumption of many syntactic framcv.rorks. As I observed, the canonical
.. . h , vo es t e notion of a no (G k
stra1g t ltne, ruler rule law") B 1 h"I . nn ree kan~n sentence model v.•as established according to logical rather than syntactic
b ' ' · uw1e1noneu th ·
y the grammatical system itself in th h s_e . e norm is provided criteria, therefore a departure from this model does not entail a departure
i.e. decided upon by the gramma '. f e ot er use It is metagrammaticaf,
. nan or the purpose 5 f . from the scientific methods of structural analysis but only from certain
Iog1cal ·
analysis. The amb· . . . .
iguity ts rem1n1scent f h
o grammat1cal and
pretheoretical assumptions. Moreover. there is statistical evidence that
ambiguity of the notion .. ru 1e o f grammar " o t e well-known. languages which require a subject NP are a small minority among the
From the point of view of the Princi le ~ft . languages of the world. For example, Gilligan (1987) demonstrates that
and Role and of the dichotorn of .p h_e Separation of Reference
"Pro-Drop languages" represent the vast majority of the world's
preceding section, the occur:enc:o~~c expre~s1on types postulated in tb.C
languages. Gilligan sho\vs that of a sample of one hundred widely
subject NPs constitutes canonical sentences with topicil
an anomaly Indeed h I . . diverse languages only seven do not allo\v "null subjects" in finite clauses.
expressions combine the t\\'O . f. sue ex1cal subject That the occurrence of lexical subject NPs is indeed a relative anomaly
semantic unctions the ~ ·
.an d the role-oriented function-which the . . - re erence-oriented across languages (compared to the occurrence of "detached" NPs) is
Reference and Role strives t Pnnc1ple of the Separation of
o separate They are in confirmed by the fact that there exist languages in \\'hich lexical NPs
expressions. Now it is precisely such . . a sense aggregate never function as subject arguments {Van Valin 1985. Jelinek 1984,
as basic sentence compo t . agg.regate expressions that are taken
nen s 1n syntactic theo · b ·i · Mithun 1986). On the other hand. languages \Vhich prohibit NP
sentence model Since in su h th . nes u1 t on the canonical detachment do not seem to exist. Languages \vhich lack the category
. · c eones the structural s b. ·· .
occupied by a full lexical NP d . . . u 1ect pos1hon 1s "subject NP" (but not necessarily the category "subject") are called
. , an since this NP 1s t k b
component of the sentence Ian a en to e a necessary "pronominal argument" languages by Jelinek. Such languages are of
presence of a b. , guages whose syntax does not require the
su 1ect NP tend to be interpret d I course excellent evidence in support of the validity of the Principle of the
surface structure LACKS an .m e as anguages whose Separation of Reference and Role. i\·1oreover. it has been repeatedly
. 1 por1ant e1ement.
To cite one example which has fi d . observed-and it is strongly confirmed for spoken French in Lambrecht
discussions, the postul~tion of the s~g:~1e%~~m1n~ntly ~recent syntactic 1986b- that even in "non-Pro-Drop" languages canonical sentences with
Government-and-Binding theo f ro- rop arameter" in the lexical topic NPs are rare. hence "statistically anomalous." 1n
conse uence f h . ry o syntax can be seen as a direct
spontaneous discourse. Fina!ly, as v.:e sa\\' earlier. the use of lexical
som: ~evel of~a~a~y:~ss.u~:nt~~~~::~i~;e;~:~~:h:s~t~~:ena :uh~~~ctd~p a: NPs is often motivated not by lack of activene~~ 0f the NP referent but by
require a subject NP, are called Pr -D ' no other factors such as disambiguation among tv.:o or n1orc active referents.
"'drop" an ele tf . . _o rop lan_guages because they may
. ~en rom a pos1t1on in \Vhich it is assumed to be nonnall Many naturally occurring lexical argument !()pie-. have pragmatically a
present. W1th1n the present approach wh· h . b d - y "quasi-pronominal'· character. The existence of ~uch topic NPs therefore
that grammati I . , - ic is ase 0n the assumptton
does not invalidate the general dichotomy of topic express1l1n types \~ hich
1
the postulatioca stfructupre and infonnatton structure are interdependent
disadva .n °
. a. ro-Drop paramete r represents a theoretical' I am postulating
ntage since it ignores the issue ofv.·hy th~ el , . .
not pr · h · '= emt:nt in ques!lon 1s
esen1 in t e first place (see examrlc ( 1 1 J) ·i d d. .
Secti 5 5 2 be! . - · "n 1scuss1on and
on . . ow). The postulation or the Prl'i-Dror raramet~·r takes as
 192 Prag111atic relations. topic Jmplica1ions for syntaclic theory 193

4.5.3 The syntactic status of detached constituenrs element, whose relationship with the clause is not the grammatical
relation of subject or object but the pragmatic relation of aboutness and
It has often been argued by linguists interested in the relationship relevance (see Gundel 1976, Dik 1978). More specific arguments in favor
between language change and language typology (see e.g. Wartburg 1943,
of the position that topic NPs do not occupy argument positions are
Hyman 1975, GivOn 1976, Harris 1976) that the frequent use of NP
presented in Lambrecht J986b for spoken French. 46
detachment constructions in a language is a sign of the rise of a new verb-
The first argument involves a construction which I call the UNLINKED
agreement paradigm in which a resumptive pronoun is reinterpreted as a
TOPIC CONSTRUCTION. In the languages with which I am familiar. this
grammatical agreement marker and in which the detached NP is
construction occurs frequently in spontaneous spoken language but is not
becoming an intra-clausal subjec! NP. While the diachronic reanalysis
considered acceptable in writing, It involves detached lexical noun
of detached NPs as subjects seen1s to be a solidly attested historical fact,
phrases which have no anaphoric link with a pronominal topic expression
it is important to acknov..•ledge that there is no necessary link between the
inside the clause. Here are a few examples from English conversations
regular use of NP detachment and the rise of a new verb-agreement
(for convenience, I have separated the unlinked topic NP wjth a comma
paradigm, even though language change may lead to increased frequency
from the clause with which it is associated; this comma does not
of use. There are many languages, such as German or Turkish, which
necessarily indicate a pause):
have fully functional systems of subject-verb agreement but which
nevertheless make frequenr use of NP detachmenl. In such languages (4.50) (Six year old girJ, explaining why the African elephant has bjgger ears
there is no evidence that the detached topic NP is being reinterpreted as a than the Asian elephanl)
subject NP. Indeed, reinterpretation of Turkish right-detached NPs as The African elephant, it's so hot there, so he can fan himself.
subjects would be a rather surprising typological event, given the strong (4.51) (From an article in the San Francisco Chronicle about a wealthy town in
SOV character of this language. Dade County, Florida, becoming a "fort against crime")
I would like to emphasize, against what seems to be a widespread ''What we are trying to do here is keep this community what it is, a
beautiful, safe place to live," said police chief Dick de Stefani. "Dade
assumption, that the grammatical topic-marking construction referred to
County, you just can't believe the rise in crime."41 Ii'
as NP "detacbmen!'' or "disloca1jon" is no! some kind of structural
'i;
(4.52) (From a TV interview about the availability of child care)

..
anomaly which tends to develop under the pressure of historical change
'" That isn't the typical family anymore. The typical family today, the
and which grammars strive to eliminate by absorbing it into the canonical
husband and lhe wife both work .
sentence model in which all semantic arguments of a predicate appear as
syntactic arguments in a clause. 45 Within the present framework, in (4.53) (Talking about how to grow flowers)
Tulips, you have to plant new bulbs every year?
which the clause is analyzed as a processsing unit for spontaneous speech,
the opposite view is closer to the truth: it is the reinterpretation of (4.54) (Lecturer in an introductory linguistics course)
detached NPs as "regular" subjec!s that constitu!es !he anomaly. Its Other languages, you don't just have straight tones Ii.le Lhat.
generalization across languages v..'ould contradict the functional motiva-
In all these examples, the referent of the detached NP in the relevant
tion for the detachment construction, v;hich is precisely to keep lexical
sentences is to be interpreted as a topic since one or several of the
topic constituents OUTSIDE the clauses in which their referents play the
following proposilions can be c-onsfrued as conveying information about
semantic and syntactic role of arguments.
it. Moreover, in all cases the referent of the topic NP has the required
I would like to present here somi.: observations, both cross-linguistic property of pragmatic accessibility. But, with the exception of the second
and language·specific, v..'hich confinn Lhe vie\~' that, in some languages at
clause in (4.50), the topic NP is not anaphorically linked to an argument,
least, the detached topic NP -.:<.1rir1ot bed 1.\)rl.'>ti1ucnt- 1rhether argument whether overt or null, in any of the propositions about the topic. It
or adjunct- of the c!ausl'. \\·Jth \\'h1ch it is pragmatically associated. follows that the topic phrase cannot be an argument in the clause with
Rather it must be analyzed a-. a ":>ynl.lct1cal\y autonomous, extra-clausal which it is associated. No\~' since the unlinked topic NP appears in the
 194 Pragmatic relations: topic Topic and pragmatic accommodation 195

same position as the linked one, it follows that the latter does not have to . between the topicalized object NP den Apfel in (b) or den in (d) and (e) on
be an argument NP within the clause. the one hand, and the detached topic NP den Apfel in.(e) on the other.
The second piece of evidence for the extra-clausal, non-argum~-~ Both the topicalization and the detachment construction mark an NP
status of detached lexical topics comes from the syntax of German. It-is grammatically as a topic, but only the detached topic falls clearly under
well known that in German the finite verb of a main clause is always the the Principle of the Separation of Reference and Role. I am not aware of
second constituent. Therefore any phrase which precedes the verb musl . any language-independent functional explanation which woul~ account
be a sing1e syntactic constituent. If a constituent other than the subject for the difference between the two construction types and which would
appears in initial position, the subject must follow the verb. Consider now allow us to assign to each of them its own invariable information-
the following e:icamples: ·,; structure properties. The difference seems to be at least in part one in the
cognitive accessibility state of the NP referent. Topicalization generally
(4.55) a. Hans isst den Apfel. "Hans eats the apple" (SVO) "'!
b. Den Apfel isst Hans. (OVS) (Ii
seems to require a higher degree of accessibility than left detachment, but
c. •Den Apfel Hans isst. (OSV) much empirical research is necessary before any substantive claims can be
d. Den isst Hans. "Hans eats it" (OVS) ·! made to this effect.
48

e. Den Apfel den isst Hans. "The apple Hans eats it" (TOYS) .c Whatever the exact differences between the various lexical topic-coding
f. •Den Apfel isst Hans den. (TVSO) strategies may turn out to be, such differences \Viii not affect our ge~eral
g. Jetzt isst den Hans. "Now Hans eats it'' (AdvVOS) !
h. • Jetzt den isst Hans. empirical observation: languages with an apparently well-established
(AdvOVS)
SVO or other canonical constituent pattern have a strong tendency to
As the comparison of (4.55) (a) or (b) with (c) shows, the verb must violate this pattern under specific pragmatic conditions by placing lexical
occupy second position in its clause in order for the sentence to ~ topic NPs, especially potential subjects, outside the clause. This tendency
grammatical. However, in (e) two constituents appear before the verb is not due to historical pressure but to a fundamental functional need. In
and nevertheless that sentence is grammatical. It follows that only the certain languages, including spoken French. this tendency is so strong
second constituent (i.e. the topicalized object pronoun den) can be a that the canonical pattern is hardly ever used in spontaneous speech.
constituent of the clause, as it is in (d). The contrast between (g) and (h) Such languages, in which the basic or canonical pattern has restricted
shows that the detached topi~ NP does not have the status of an adjunct distribution in language use, are of great theoretical interest for word-
to the predicate like the adverb jetzt "now', which does occupy intra- order typologies and for syntactic theory in general. They pose in
clausal position. The detached constituent den Apfel is therefore an extra- particularly acute terms the general problem of the relationship betv.;een
clausal lexical topic NP. abstract models of grammar and actual sentence structures.
The reader may have noticed that in example (4.55e) both the lexical
NP and the topicalized pronoun have accusative case marking. This case-
4.6 Topic and pragmatic accommodation
marking is optional on the lexical NP but obligatory on the pronoun.
This phenomenon of dual case marking is reminiscent of the dual case- The point of the preceding sections has been to demonstrate the existence
marking pattern described by Jelinek 1984 for \Varlpiri, except that in of a svstematic, though flexible. correlation bct\veen the topic status of an
Warlpiri the lexical NP has ergative-absolutive marking \vhile the intra- expre.ssion and the presumed cognitive state of the topic referent in th_e
clausal "agreement" marking is nominative-accusative. \Vhatever hearer's mind at the time of an utterance. I have tned to account for this
explanation is given for the accusative marking on the NP in (e), it correlation v.'ith the Topic Acceptahihty Scale in (4.34) and I have
does not affect the syntactic argument that the lexical topic constituent observed that a nun1ber of syntactic constructions found across
cannot be a constituent of the same clause as the pronoun. languages have the function of pron1ot1ng referents on this scale,
The German examples just examined raise a problem \\'htch I cannot allo\ving speakers to preserve as much as possible the cognitively
discuss here in any depth_ This is the problem of the pragmatic di!Terence preferred topic type. \vhich 1s C\pressed as an unaccented pronominal.
 196 Pragma1ic relations: topic Topic and pragmalic accommodation 191

The claim that a degree of pragmatic accessibility is a necessary to the indefinite subject phrases .students in lhe science club al Mark Twain
condition for topic function has been rejected by some linguists. An Junior High School of Coney Island and a child of my acquaintance
explicit rejection of this claim is found in Reinhart (1982). 49 Although respectively. This kind of cataphoric reference to pragmatically non-
Reinhart concedes that topics strongly tend to represent "old informa· accessible items is a rhetorical convention, which is based on the rule of
tion," she argues that this tendency has nothing to do with the nature of accommodation for pragmatic presuppositions. Because of the nature of
topic. To substantiate her claim, she quotes examples from English texts the reading situation, readers can more readily accommodate as active
in which topical subject noun phrases have referents which are not (or accessible) certain referents which are in fact new in the discourse
accessible in any clear sense (they occur at the beginning of newspaper context.
articles). Here are two of Reinharfs examples (the relevant topic This common phenomenon whereby a writer introduces a referent via
expressions are italicized): a linguistic expression or grammatical construction which nonnally
requires the presupposition that the referent is already introduced is
(4.56) ( = Reinharrs 2la) Because they wanted to know more about the
discussed by Clark & Haviland (1977) under the name of "addition." For
ocean's current, students in the science cfub al Mark Twain Jwiior High
School af Coney ls/and gave ten bottles with return address cards inside example if a reader finds the sentence The old wonwi died at the
to crewmen of one of New York's sludge barges. (The New York Times) beginning of a story, she knows, consciously or unconsciously, that she is
dealing with an intentional violation of a principle of information
(4.57) ( = Reinbart's 2lb) When she was five years old, a child of my
acquaintance announced a theory that she was inhabited by rabbits. structure. Such a violation is acceptable because the author of the story
(The New York Times) can expect the reader to act cooperatively AS IF the referent of the NP the
old woman were already present in the reader's awareness by constructing
The exceptions mentioned by Reinhart happen to be restricted to certain an antecedent for the NP which then can be "added" to the text world of
types of written discourse and therefore do not directly affect my the story. In the case of (4.56) and (4.57), the cooperative effort necessary
argument, as they do not pertain to the domain of sponlaneous spoken on the part of the reader to interpret these sentences is accomplished all
language which is considered the basic language use in this study. the more easily since the referents of the subject noun phrases are
However, by observing that an exception is genre·specific we have not pragmatically ANCHORED (Section 3.3) in the modifying prepositional
... , explained the exception, nor have we explained why some (written) phrases in the science club at ... and of my acquaintance. Notice that the
genres are more tolerant of exceptions than spoken language. To the acceptability of the two sentences would be severely diminished if these
extent that the Topic Acceptability Scale is an expression of a gener~l prepositional phrases were missing (cf. examples (4.35), (4.36) and
cognitive constraint on information processing in natural language, any discussion). 1 therefore do not think that examples such as these can be
violation must in principle be accountable for on general, genre- invoked as arguments against the postulated inherent connection between
independent, cognitive grounds. topic function and cognitive accessibility of the topic referent. Rather the
1 beJjeve that vJolatjons of lhe Topic Acceptability Scale such as those conventionalized character of these exceptions indirectly confums this
in (4.56) and (4.57) can be explained as genre-specific instances of the connection. Such exceptions are interpretable precisely against the
principle of the PRAGMATIC ACCOMMODATION of presuppositional structure background of the presuppositional structure conventionally associated
(see Section 2.4). Notice that Reinhart's examples of topic noun phrases with topic-comment sentences. 50
with brand.new referents contain unaccented anaphoric pronouns that It is interesting to observe that the pragmatic accommodation of an
occur BEFORE the topic NPs \vhich they refer to. Without these pronouns, unidentifiable referent is easier when the referent is grammatically
the sentences could in fact be con~trued as event-reporting sentences, marked as a topic than v.•hen it is marked as a focus constituent. While
comparable to the journalistic sentence in (2. I0\ (A clergyman's opened a the above-quoted story opening The old wo1nan DIED, which has topic--
betting shop on an airliner), in \~h1ch cJ.:::.e the subject NPs \vould not be comment structure. is rather conventional, the corresponding opening
topics. The pronouns they 10 (4.5t1) :ind she 10 (4.57) refer cataphoricaUy involving the thetic sentence The old ~'OM.~N died would be rather
 198 Pragmatic relations: topic Topic and H'Ord order 199

. "tion evoked by the topic-comment

unusual. At first glance, this difference seems surprising. Given that topic to the clerk; and (1v) the presup:~s1 . that this letter is a topic under
referents must be accessible while focus referents have no such structure of the sentence as a w o e, i.e. ence
requirement (sec Section 5.4.1), we might expect the second version to . . ·n
d 1scuss1on 1
the text By its presuppositional structure, the sent
· t to the reader to
be more natural. That it is not is due to the nature of subjects as containing this complex subject phrasefconv~dyscaomremq:~ication as if the
unmarked topics (Section 4.2.1). Since a subject is expected to be a topic, . 1 c the purposes o rap1 '
cooperal!ve y act'. or . d d etrievable from the discourse. The
and since topics must be pragmatically acxessible, the invitation for
accommodation is more strongly expressed-hence more readily followed referent i~ ~u:~~~;ri;a~~; :o:~at:d shortcut for some more exp~cit
by the reader-in the topic-comment version than in the thetic version. ::~:~:~~ s~:h as 1 sent a letter of protest 10 the Sete cci;e Se~v'.cet~~t~:,::;~
This fact also shows, incidentally, that the often-made claim that 1 , an '~ .
I handed my registration card to the post ~/flee ~er ·n the
presentational sentences are typically found at discourse beginnings is d l that A certain processing difficulty ans1ng I
unwarranted. state exact Y · indication that in this case the
interpretation of the sentence mt~ bpea;tnof the reader to understand this
Herc is another striking example of a violation of the topic accessibility cooperative effort necessary on e · th vlolation of
constraint which I think is best explained as such a rhetorically motivated · crht to be Be that as it may. e
sentence is greater t h an 11 oui. · . · he revlously
deviation from normal usage. Consider the following passage from a the topic accessibility constraint in this sentenc~. as ~n t th:r than as
newspaper article on the problem of US army registration (the relevant discussed examples, can be seen as a confi~at1on o , ra
topic expression is again italicized): evidence against, the existence of this constraint.

(4.58) Of equal importance is the fact that objecting registrants can say they
oppose the policy of registration and will not cooperate with a draft. The 4 .7 Topic and l\'ord order
lei/er of protest I sent 10 the Selectfre Service at 1he time I handed my . .
r~i.stratlon card to the post office clerk stated exactly that. (The Daily The theoretical d1stlnct1on dra,vn .in Sect1on 4 . 5.2 het\\'een two . categones
d II ,
Californian, September 1982) . . one role-oriented and one reference·onente , a ov.s
of toplC expressions, f topic
us to c an Y I ·r a much debated issue: that of the ros1TtON o
d h h s a
The prime candidate for topic function in the second sentence is the . . the sentence lt has often been cla1me t at t ere I
expressions in · · d C)' for topic
sentence-final pronoun that, whose referent is the propositional content . . . at least a strong crosshngu1st1c ten en ,
un1versa 1 pnncip 1e, or . . Section 4.1.1 ).
of the last two clauses in the first sentence, a referent which is discourse- . l be the first constituents in a sentence (see
expressions o h . t" of the
active at the time the pronoun occurs. Nevertheless that is not a topic The issue is as old as the traditional debate over t_ e post ion I
expression because of its association with the "rhematizing" adverb subject since subjects and topics \Vere equated in trad1t1onal gram:~rth;
exactly and the ensuing (implicit) focal accent. Without this adverb, the croes b~ck at least as fa. r as the e1ghteenth-.century debate over R~ a l's
sentence could have been formulated as That was stated in the lener ... or o"best" \Vord order is 10 universa 1 gramm<1r. For example · :. in 1varo d
I stated that in the letter ... The intended topic in (4.58) must therefore be • 1· . d l 1· 1ngue frani;a1se (quote in
1784 "Discours sur l universa tte e .:i • , .. , d
the italicized complex subject noun phrase. This phrase is stylistically . VO order of French p; taken to be !ht: tdt:a 1 v.or
Grev1sse 1959) the S . f . ·al thought. In nineteenth·
peculiar in that it requires the pragmatic accommodation of at least four 1
order because it IS a direct expresSJl)n o odgi·~ c0nnection \\'tth the issue
presuppositions: (i) the presupposition evoked by the definite article, i.e. . · t" the debate \~'as conunue l
century I1ng.u1s ics, . _ b ~-i-· (as opposed to the
.that the reader knows of and can identify a certain letter \vritten by the f the "psychologica 1 su J1:L
of the p_os1tion o '. ~ 1 1 ·the l\\'Cntieth century, the
author of the article (the letter \Vas not mentioned earlier in the article); grammatical subject) in the. sentefnce. n h·1s been emphasized in
(ii) the presupposition evoked by the proposition expressed in the f t l p )S1t1on or topics '.
importance o int ia . l h p . School and bv linguists influenced
restrictive relative clause starting with I senr, 1.e. that the \\'nter protested partic.ular hy scholars oft e raguer,· b ~ 196',a) .Indeed. evidence for
h mn1arv in 1·ir a~
the draft by sending a certain letter to the Selective Service at a certain bv .this school (c f · t e su
l•
· _ll d ··free word order"
time; (iii) the presupposition that the v.-riter handed his registration card s~ch a tendency is p.:irticularl~' abundant in so-c.l e
 200 Pragrnaric relations: topic Topic and word order 201

languages, like Russian or Czech, in which any constituent can be placed (Mithun 1987) that in certain languages the very notion of "basic word
in sentence-initial position, without causing the resulting sentence order" is not applicable. According to Mithun, in such languages no
structure to be marked in the way a sentence with a topicalized object known pragmatic principle governs the choice of the various alternative
is marked in English. For such languages, convincing arguments have word order possibilities offered by the syntax. If this claim is correct, the
been made that sentence-initial position of a constituent normally marks topic-first principle does not apply to such languages.
this constituent as the topic of that sentence The occurrence of focus-initial sentences such as the one mentioned
The putative universality of the "topic-first principle" has been above leads us back to the discussion concerning the relationship between
questioned by various scholars, and for various reasons. One reason is syntactic structure and information structure (see Section 1.4.2). Given
the existence of VOS or VSO languages, i.e. languages in which it is the that sentence-initial position is cognitively speaking an eminently salient
verb that occupies the first position in what seems to be the unmarked or position, it would be a priori surprising if the prominence associated with
basic sentence type. A theory according to which sentence-initial position this position could only be exploited for a single function, such as the
is a natural, cognitively based, requiren1ent for topic NPs would not be marking of the topic relation. As I observed earlier, in English, German,
able to account for the fact that such languages normally require a and French, and no doubt in many other languages, it is possible to use
naturally non-topical constituent, i.e. the verb, to appear in initial the construction traditionally referred to as "topicalization" both for
position. This problem would persist even if it could be shown that all "topicalizing" and for "focalizing" the fronted non-subject NP, the
verb-initial languages have a topicalization rule which allows noun difference being marked only via accent placement. It has also been
phrases to occur before the verb. Indeed application of such a rule would observed by Prague School scholars that even in Slavic languages non-
presumably result in a marked construction; however, the topic-first thematic constituents may occur sentence-initially for reasons having to
principle is meant to account for the most general case of topic do with "emphasis" (see e.g. Firbas 1966a).
placement. Without going into much detail here, I would like to point out that
A second argument against the universality of the topic-first principle some of the apparent differences among languages with respect to the
can be made on the basis of languages like English or German, in which adherence to, or disregard for, the topic.first principle disappear if we
focus constituents can freely occur as sentence-initial subjects and in make the suggested categorial distinction between lexical and pronominal
which topical non-subject constituents may appear in canonical argument topic expressions. From my characterization of the preferred topic
"' position after the verb, i.e. without any concomitant SYNTACTIC expression as an unaccented pronominal argument, v.·hose function is to
" markedness of the construction, the information structure of such express the grammatical and sen1antic role played by a pragmatically
52 ALREADY ESTABLISHED topic referent in a clause it follo\vs that the position
sentences being only marked prosodically (see Section 5.3.3 below).
Recall the various examples of thetic sentences discussed in Section 4.2.2. of such a pronominal expression is functionally speaking IRRELEVANT.
Or consider a sentence such as Jespersen's PETER said iI (Section 2.2), in Once a topic referent is pragn1atically established, i.e. once the function
which the subject is a focus constituent and in which the clause-final of the topic expression is no longer to ANNOUJ\'CE the topic referent but to
pronoun it is a topic expression. Notice that topic constituents in mark its role as an argument in a proposition, there is no longer any
postverbal argument position do not have to be pronouns, as in functional reason for the topic to appear at the beginning of the sentence.
Jespersen's example. It is easy to substitute a lexical phrase for the For the preferred-topic expression it is funcllonally speaking more ,,I
pronoun, as in PETER r11adc that re111ark Such unaccented lexical topic important to be in close association \vith the predicate than to appear in :1
constituents have an important prl1pcr1y in Cl1mmon \\·ith pronouns: their sentence-initial posit11Jn, since 1t is the predicate that governs the semantic
and synL..1ctic relations in the clduse. L!naccented pronominal topics
[:
referents must be acti\'e or qu:.is1-Jcli\'c 1n the d1~course. In some
languages, unaccented topic NPs 111 ubJCCL J:K1s1t1l1n are avoided, as in therefore tend to oc...'ur in or nedr the position 1n v.·hich the verb itself
spoken French, v.'here such NP':> typ1c.dl~ appcdr in right-detached occurs, i.e. tvv.arJs the beginning of the sentence 1n verb-initial or verb-
position (see Section 5.3.3 belo11·1 F1n,dl). the clai1n has been 111ade second languages and llllvarJs the end 111 \'erb·finJl J..inguages. 53 When
 Topic and word order 203
202 Pragmatic relations: topic
r 1984) "extraposition" (Jespersen I 933/1964:
unaccented pronouns develop diachronically into bound pronominal and
inflectional morphemes, they tend to be affixed to the verb or an auxiliary
sc~~a)rs;· ~e:.!~~u;i~~O::ation.': The detached constituent itself has ?een
1 . , an .. f d NP" "afterthought NP," "post-predicate
of the verb, rather than to some other constituent of the sentence. As the
referr~d to .~~.ta~~-(;i~s~ 980, ~allduvi I 990b), and "'antitopic" (Chafe
semantic center of the clause which serves as a point of reference for
argument constituents, the tensed predicate is fixed in its position. As a
result. the position of unaccented pronominals will also tend to be FIXED
~~~~tt~«~';he last
Curiou~ly,
term, "an:~~~~~~tit~:t ~::::c:~::;e~,;~: :;,e~:::~~
thts i~pohrtant s.i The antitopic construction is illustrated in
within the sentence, contrasting with the relative positional freedom of generative syntactic t eory ·
phrasal constituents such as NPs, PPs, and adverbial phrases. Now, as I this English example:
emphasized in Chapter I, a fundamental requirement for any pragmatic (4 -5 9 ) He is a nice GL"Y, your brother
word order analysis must be the potential for CONTRASTS between
alternative ordering possibilities, i.e. the possible occurrence of In this sentence, the intra-clausal unaccented pronominal t~pi~ expres~
alloscntences. Since the position of unaccented pronominals in the . h recedes the lexical topic expression your bro~h~r, which is place
sentence is essentially fixed, this fundamental pragmatic criterion is not ~~opnos~-~ocal position. Notice that this antilopic NP ts itselfl~N~cct~::E:y·
What makes the antitop1c . construe t"10 n pragmatical\v
_ _ - pecu 1ar . 1s. .
satisfied with such expressions. Therefore no functional claims about
sentence-initial position should be made in the case of the preferred topic the time the referent of the anti topic expression is mentioned in .'ts I1e;1ca 1
expressions. it has already been referred to in una_ccented pronom1na arm
'.o~, clause \vhich expresses the proposition about the referen_t. .
ins;~~s \~portant ant1top1~
The situation is quite different with ACCENTED TOPIC EXPRESSIONS,
whether lexical or pronominal. Only with these expressions can - and to realize that the detached constituent in the
. n AFTERTHOL'GHT in the proper sense o
should-the case for initial topic position be made. Since they have the construction does NOT express al d s~ Right detachment is a fully
primary function of announcing a new topic or of marking a shirt from th. \Vord as has often been c a1me
one topic to apother, it is cognitively speaking important for such topic co,~ventio~alized grammatical construction which permits speak~~d:~
expressions to occur AT THE BEGINNING OF, or preferably BEFORE, the dh to the Principle of the Separation of Reference and Role , ....I
t· 'es ~ speaker \\'ho uses an antitopic
sentence which expresses the information about their referents. It is
a ere . -
·ric discourse c1rcums anL
spec1 I . .
·
. II. fullv ··a\\·are" that the n1ere mention o t e
f h ....
difficult to imagine an e!Tective topic-coding strategy whereby the construction is norma ) in ~he clau'e \\ insufficient for the hearer to
pragmatic establishment of a topic referent would always take place unmarked topic pronoun _- _ s ., under normal
simultaneously with, or subsequent to, the conveying of information d d \\'hO or \\·hat the prc>pt1,1uon !' "1bout . llll<:,
un erstan. . . un·iccented pronl1minal is
d 1 "lflS the tl''-C n 1 an ·
ABOUT this referent. Such a strategy would run counter to the Principle of cooperat1\·e co~ -i P .. ~ l"- acti\·e. the antttl... plL ...:onstruction, once
the Separation of Reference and Role. I conclude that the topic-first approrn3te on!;. tf ito;, reft.:r~~ta~ ·1n in1rlic1t re4ueq frpm the speaker to
principle can be maintained as a universal ordering tendency, as long as it conventionalized, can he ust.: _- • _ .. I !LI'" until the
h rop0s11ion·tl 1ntonnat1on nn 10
is.only applied to accented lexical and pronominal topic expressions with the hea.rer. to rut t e P - · ' ·tru'ture of the ant1top1c
a topic-announcing function. d The prcsupp(°l\ll1Pna 1 ' l
antitopic is uttere _ . 1 th·it the !ll't-vct-;icll\C lnpic referent 1~
It is necessary to mention here one apparent exception to this revised construction 1nvo]\'CS a ~ign.i ' _ ·Th. request fl1r temporary
and severely restricted version of the topic-first principle. This exception, b d ·:1t the end of the sentent..e i:
going to e name , . f -our ·e eaqest to con1ply \\'ith if the
for which there is ample cross-linguistic evidence, is the car!icr-n1ent10ned "h Id. ·· of the proros1t1on 1s l1 t.. ~
o 1ng . . ·actn·c or at 1e.:1~t .. h.1g hi\ ··:1ccessib!e. This exp 1a1ns
referent. is alrea d Y .q". as\.-
r the referent I~ ;i~ g"ner
RIOHT-DETACHMENT or AITTITOPIC construction. in \Vhich a le'\tcal topic NP . II Cl1nd1t1on for
hI ... '
is positioned AT THE END of the clau~e containing the inforn1at1on about \vhy hig.h acce5s1 i it;. l1 - , truction across languages (see
if the 1nt1tl1r1c c1 1 ns ~
the topic referent. This construction has been var1ou~lr referred to in the appropnate use t , . • R the ~11uation in spoken French.) 1n
literature as "epexegesis" (a 1ern1 from classical grammar). "inverted Lambrecht. J986h .. Ch.1ptcr , , on I • It 1nJependent-pronom1nall
word order" (a translation of the tenn dcl'nk climlc u~ed hy Turkish contrast \Vtt h Icrt- J e tJL-hment , the ex1~a or
 204 Pragn1a1ic: relations: topic Topic and word order 205

topic expresssion in right-detached position cannot indicate a new topic The earlier-mentioned processing implications of the antitopic
or a topic shift. Given that antitopic constituents are always unaccented, construction have consequences for its syntactic structure. In many
any contrasting function is excluded. Since my revised topic-first languages, the antitopic constituent must immediately follow the clause
principle is meant to account precisely for new topics or topic shifts, which contains the coreferentiaJ pronominal topic expression, i.e. it is
the antitopic construction does not contradict my general claim about the syntactically marked as belonging to the clause containing the anaphoric
position of topic NPs. pronoun. 57 Herein, the antitopic construction differs markedly from the
The crosslinguistic tendency for right-detachment constructions to be left-detachment construction, in which the lexical topic constituent can
used in discourse contexts in which the topic referent is already highly appear at any distance from the associated proposition and in which the
salient, and for left-detachment constructions to be reserved for topic- anaphoric pronoun can appear in a clause of indefinite depth of
announcing or topic-shifting contexts, has grammatical reflexes in embedding. In French, the tight connection of the antitopic constituent
German. German has two sets of personal pronouns in the third with the proposition is also reflected in the fact that the antitopic NP
person, the set er, sie, es, sie and the set der, die, das. die, both meaning must agree in case with the pronominal topic, while this sort of agreement
"he, she, it, they." Both sets may be accented, although accentuation is is not required with sentence-initial topics (see Lambrecht 1981). A
more common with the der-series. As a general rule, pronouns of the er- similar situation obtains in German.
series are used when a referent is active AND already topical, while those The processing constraints associated with the antitopic construction
of the der-series are used when a referent is active bur not yet an may also explain why subject-final languages, ie. languages in which the
established topic. Therefore in most anaphoric contexts, only the er-series unmarked topic occurs at the end of the clause, are so rarely attested
can be used, as shown in example (4.60): among the languages of the world (see Greenberg 1963, Universal# 1).
An explanation along these lines is suggested by Keenan (1978:303fl), on
(4.60) Wenn er/der isst, macht er/•der so komische Gerausche. "When he eats the basis of his analysis of Malagasy (a VOS language). Even though I
he makes kind of funny noises."
find Keenan's cognitive explanation for the scarcity of subject-final
languages highJy suggestjve, J would hesitate to draw general conclusjons
....
:s:·

.•..
Assuming coreferenliaJiry with the subject of the subordinate clause, use
of the pronoun der in the main clause is prohibited. (The sentence is
grammatical if the second der indicates a topic shift.) Now consider the
of this sort. A problem with many proposed explanations involving
universal word-order typology is that they do not take into account
following examples involving left and right detachment: infonnation-structure distjnctions of the sort discussed in this study
""
(pronominal vs. lexical coding, topic vs. focus, etc.). Until such
(4.61) a. Die MU.llers, die wohnen irn dritlen Stock. "The MUiier's (they)
fundamental distinctions are carefully drawn and until it is firmly
live on the third floor."
estabJished whal the final subject constituenl in VOS languages actually
b. •?Die Miillers, sie wohnen im dritten Stock.
codes in discourse, any speculation concerning the cognitive status of the
(4.62) a. Die wohnen im dritten Stock, die Miillers subject in subject-final languages seems premature. 58
b. Sie wohnen im dritten Stock, die f'o..hiller~.

As (4.61) shows, left-detachment requires an anaphoric topic pronoun of

the der-series, i.e. a pronoun marking the referent as active but not yet
established as a topic. 56 No such constraint, ho\~'el/er, is imposed on the
right-detachment construction, as sho,~·n in {4.62J. The Gem1an data thus
confirm the existence or a gener.il inforn1ation-structure principle
according to which right detachment re4uires greater pragmatic salience
of the topic referent than left detachment, the latter being used mainly for
referents whose topic status 15 nol yet established in the discourse context.
 Definition of focus 201

h the focus ofa sentence,. or, more

. en
Within the framewor k developed ere,
sition expressed by a sentence in a gtv
precisely, the focus of the propo l t f ·informati on whereby the
t 1·s een as the e emen f o each other The focus is
utterance contex •d the s . ·
.. assertion DIFFER ram
5 Prag1natic relations: focus presuppos ition an ..
that portion of a propos1t1on w '
h- ch cannot be taken for granted at the
r pragmatic ally NON-RECOV-
. f h It is the UNPREDICT ABLE 0
speec . . an utterance. The focus is what makes an utterance
ttme o element
ERABLE 10

into an assertion. · k f cus and

This notion of focus is implicit in mu.ch prev_1ous_wor ~n o, rly
related phen'ome na. For example it is imphe~. tn Bolinger~ ea

5.1 Definition of focus definition of what he calls the "infonnat ion point of a sentence.
the
k lhe . "point" of that
We can say that the prosodic stress ... mar. s
Focus, presupposition, and assertion
~~ 1 ~~~~:;:~;· told.
1
sentence, \Vhhere there is dth~g;::~~s~i~~~~~e~~r~~ f~;
1
1
5.1.1
which the earer wou
In Oiapter 4, I used the term "focus" as a convenient shorthand to refer (1954c\52)
to the status of certain sentence constituen ts ·which systemati cally differed
d · H II da ,· 5 (1967) definition of
from topic expressions in their pragmatic function and in their formal More explicitly the notion is expresse in a ! )
expression . It would therefore seem natural to define focus as the focus:
"complem ent of topic." The complem entarity of the two notions is on focus is one kind of emphasis, that whereby the speaker
suggested e.g. by the alternativ e concept pair theme/rhe me, \Vhose ~~~~:~t a part (which inay be the \\hole) of a n~e5sag_e _b:ock ~~n~~.~
members are often seen as complementing each other. Using Chafe's which he.wishes to be interpreted a~ inforrnativ e. ~ h.~I ~1oi:apl ;:\ iousl\
. ot in the sense th.it it cc1nnot 3\1: en .
character ization of the (topical) subject as the "hitching post for the ne\v
knowledge" (cf. Section 4.1. l), we might then say that the focus of a ~~~:~~~~~\;hough it s often
1 the case thd~ ha~e~:~.:r:e~~ b~~; :~:
it
-
se that the speaker presents it <Is not eing
sentence is the "new knowledge hitched to the topic post" i.e. the ne\v sen . a·iscours e The focus of the n1essage. it is sugge~ted. ., .
is that
ll, (and
preceding ne\\. l~X!U3
informati on conveyed about a topic. . d ti.
)
th• . . pe.iker ,1.; he1ng , 41'"' "ll4r
) '
which ts represente 1:
~ 1
Within the present frame\vor k, there are at least two reasons for not situat10nal\y) non-den\·a hk inf,..,rn1.1t1» n tH:.i!hd.1'.I I ,

adopting such a definition . First, if we assume- as I do - that focus has to as the element of informati on in a sentence \\'herehy
Th e concep t of focus- · · ,, ·h ther l" a].;l1
do with the conveying of new informati on, and that all sentences convey .
t shared kllO\\'\edgc differ trom c.ic o . -
new informati on (Section 2.3), all sentences must have a focus. Ho\vever, shared and not-ye - ' h J -k , -r
u ed v ac ·enuo 1 1 ' - · '
I"?' 1
Jackendo ff \\·hose
·
h
not all sentences have a topic'(see Sections 4.2.2 and 4.4.4.1). Therefore closely relat.ed lot e one s il·d , . Chom.;k\' ( !970). defines the
1 . th·1t I"
analysis builds on those of Ha i ay anL
focus cannot simply be defined as the complem ent of topic. Second, in the . . . f lCOcc"" "IS "the 1nrnrrn;1t1nn 111 the scnlCTICe "
'
d the
"presuppo s1t1on o a sen '·
present framewor k the terms "new knowledg e" or ''ne\\' informati on" are , s eaker to be shared hy him and the hearer an
loose equivalen ts for the term "pragmat ic assertion. " \Vhich I defined in assumed b) the - P " . rorma 110 n •n th•'._ '"n\cnce that 1s assumed
"th· 1n h
''-
h , .. (197" "lOJ. For
of a sentence as c
Chapter 2 as a propositio n that is SUPER!MPO SEO ON and that INCLUDES the "focus
k t to he shared b\' h1n1 Jnd t e ean:r - ~-
· . of ihe presuppos 11on in a
pragmatic presuppo sition (see Section 2.3). The focus of a sentence. by the spea er no
Jackendof f. the focus is thus the l 0\1PI E-~1fr--. T
1

however, is generally seen as an element of informati on \vhich is ,\DDED

TO, rather than superimp osed on, the pragmatic presuppos ition. Ju-;t
as a sentence. ,. d . in m·in\'- re..;pcct" similar to
f "fncus I \\'l 11 a Pp 1 1"
_ But \\·hik c·1i, 1 n 1 sk~. Jackendof f. and
•
topic is included in the presuppos ition \\'ilhout bL'.1n~ identical to ii (see The concept 0 .-1 ,,.
"f
d by• mv- prcueL1:'>S< 11
~
is part of an J'~Crl1••n \Vith0ut c0incidin g h
t at U<iC • nh 111 ..;,n-called ocu<; -
Sections 4.1.I and 4.3), a focus ~1pp\1ed l11cu" n11t1••n m.11
other~ ha\'I? 1h1"
with it.

206
 208 Pragn1atic relations: focus Definition of focus 209

presupposition" sentences, i.e. to sentences in which the focus while a non-focused constituent is understood to be "old information."
corresponds to a variable in a presupposed open proposition, I will While I agree that correlations must be established between phrasal
generalize it to all types of presuppositional structure, building on the accents and the statuses of the denotata of such phrases in the minds of
theoretical concepts developed in the preceding chapters, especially the interlocutors, the simple equation between accent and new
the concepts of activation and topic. I will also show that "the pre- information is inadequate, especially if the concept of "new informa-
supposition" in the Chomsky-JackendofT tradition is in fact only one tion" is used as an unanalyzed primitive. 1
particular subtype of pragmatic presupposition and that the accent rules The approach to focus based on the segmentation view of information
proposed by these authors are insufficient to account for the focus-- conflicts with the analysis which I presented in Chapter 2. As I argued
presupposition relation in general (Section 5.4.3). there, information is not conveyed by lexical items or individual sentence
Finally, I will emphasize that certain prosodic phenomena which have constituents but only by establishing RELATIONS between denotata and
been subsumed under the general rubric "focus" are in fact not related to propositions. Consequently, if focus has to do with the conveying of
focus in the sense defined here but to the marking of different ACTIVATION information, the function of grammatical focus marking must be to
STATES of discourse referents, which in turn serve to indicate certain TOPIC express such relations rather than to attribute the property "new" to the
OJSCONTINUITIES in the discourse. What has been categorized and analyzed denotata of individual sentence constituents. Here, as in many other
as phonological focus marking in the generative literature is seen here cases, it is crucial to distinguish between PRAGMATIC RELATIONS and
only as one instance of a more general phenomenon: the marking of a PRAGMATIC PROPERTIES. Focus, like topic, is a relational pragmatic
pragmatically construed semantic domain by means of prosodic category. Grammatical devices which serve the purpose of indicating
prominence. At the end of this chapter (Section 5.7), I will argue that pragmatic properties of the denotata of individual sentence constituents
the overriding purpose of sentence accentuation is not to mark foci but to (such as the property of being "ne\v" in the discourse) are by definition
mark the establishment of RELATIONS between various kinds of denotata devices for marking differences in the IDENTIFIABILITY and ACTTVATION
and the propositions to which they belong. The interpretation of a given states of discourse referents, not differences in the relations between
activated denotatum as focal or topical is typically determined by denotata and propositions. The specific issue of the relationship between
morphosyntactic rather than phonological criteria. To the extent that the focus and the activation states of referents will be discussed in Section
present chapter is concerned with accent placement it will be useful for 5.4.
the reader to keep in mind that accent placement and focus marking are Let us look again at example (2.7), which I repeat here for convenience
not to be equated. as (5.1):
Many of the approaches to focus in the literature are characterized-
conceptually or at least terminologically - by what I referred to earlier as (5.IJ Q: Where dtd you go last nighfl
A: I went to the 1.1ov1Es.
the "segmentation" view of information (see Section 2.2). Recall that in
this view the information conveyed by a sentence is thought to be In some intuitive sense, \\'e are no doubt justified in saying, with Bolinger,
segment.able into "old" and "nev/' portions, \vhich can be directly that the \\'Ord n101'ies, or perhaps the phrase the mo1·ies, in the answer
identified with syntactic constituents, in such a way that some indicates the point "where there is the greatest concentration of
constituents express "o!d," and others "ne\v," infonnation. The function information," or, .,..,.·1th Halliday, that this \vord is the element "whereby
of focus prosody is then taken to be that of n1arking sentence constituents the speaker marks out [theJ part .. of /the] message block which he
as "new." For example, for JackcnJoO' ( 1972:240) the focus of a sentence \\'ishcs to be interpreted as informative." Nevertheless, it would be
is "the semantic nlatenal J)'iociJLeJ \>ilth surface structure nodes inaccurate to sav that this \\'ord, or this phrase, "1s the focus," if focus is
dominated by F," \\here F \(Jnd".> rL)f "nLin,:.hareJ 1nfonnation." For identified 'With ~e\\' inforn1ation. The expression f rhe) mo~·ies in (5. l) can
Selkirk (1984:206f) ··a focu~cJ ...·un\lltucn! [i.e. a .:on:::.tituent to \Vhicb a have information value only as an element of the proposition expressed
pitch accent 1s assign..::J] c ..1nLr1hutc~ "nc\\ 1nf0rn1ati0n" to the discourse, by the entire sentence. \Vhat is "ne\\'" is not the constituent, nor its
 210 Pragmatic relations: focus Definition of focus 211

designatum, but its role as the second argument of the predicate "go-to" . '
new 1n1orma ,.ion " let us consider anot h er, more natural . answer to the
in the pragmatically presupposed open proposition "speaker went to x." question in (5.2a), i.e. (5. 2'):
Equally inaccurate would be the claim that the new information is
(5.2') No, BILL. .
expressed in the prepositional phrase to the movies, since the directional
meaning of the preposition to is recoverable From the word where in the .
· formation . this answer • as in the answer inh (5.2b}.d
conveyed 1n
Th
question. e new in . . What makes uttering t e wor
1. h oun or constituent 8 ' 11· . .
s clearly not t e n h h h arer establishes a relat1onsh1p
Bill informative is the f~ct t ath t e b.e ct argument in the understood
The information conveyed by the answer in (5.1) is neither "movies,"
nor "the movies," nor "to the movies" but the abstract proposition "The between the individual B1~l and t e ;~ JC re technically, between the
place I went to last night was the movies." It is only as the predicate of .. " meone wntes poetry or, mo .
propos1t1on so .1. al function "x wntes
this abstract proposition that the expression the rno~·ies-or rather its h Bill and the propos1 ton
referent of t e noun h . ble x In the present case, where
denotatum-may be said to be the focus in (5.1). Thus when we say that .. h "B'll"
1 replaces t e vana ·
the phrase the movies is the focus of the answer in (5.1) what we mean is poetry, w ere . d main of the focus coincide, my
"" d" word and the semantic o
that the denotatum of this phrase stands in a pragmatically construed the 1ocuse . lo ical quibble. Ho\vever, as we
criticism may be no more tha~ a te~nn~~tv.·;en a focus accent and the
shall see later on, the relattonsl~tp . ften extremely indirect. In such
relation to the proposition such that its addition makes the utterance of
the sentence a piece of new information. This pragmatic relation between · d · which it symbo 1zes 1s o d
semantic omatn .. 'th ··ne\v information .. stan s
a dcnotatum and a proposition will be called Focus RELATION. In the reply t' of "focused elen1ent wt
cases, the equa ton d. f the mechanism offocus marking.
in (5.l) it is the establishment of such a focus relation between the in the way of a proper understan ing o "unpredictable" and "non-
Earlier I used the somewhath vagu: te:~ relation bet,veen the focus
dcnotatum the movies and the rest of the proposition that creates the new
state of information in the addressee's mind. The function of Focus
recoverable" to characte.nze t e pr ~ ue these terms seem to capture
MARKING is then not to mark a constituent as new but to signal a focus
element and the propos1tton. Though h·g th t~rrn "nev. .. Consider the
relation between an element of a proposition and the proposition as a the nature of the focus relation better t an e
whole. In those cases where no such relation exists, i.e. where the focus variant of (5.1) in (5. \ '):
clement coincides with the entire proposition, the function of focus the niovie..; nr l<) thr rest<iuranl '
(S. I') Q: Where d,d you go l as t n 11;>ht • tL)
marking is to indicate the ABSENCE of a focus-presupposition contrast (see
A: \Ve went to the RL~TAl'RA..,'T
Section 5. 7.2).
The intuitive appeal and terminological convenience of the notions " h denotatum of the NP 1hc resttJurant can safel)
In the ans\ver in (5.1 ) t e . h· - >f utterance since it \vas
"old information" and "new information" are such that these terms are d . ourse active at t e time <-
be assumed to be tsc - - .. ld .. t .. ne\v ··
often misleadingly used even in carefully worked-out analyses. Consider . . d.13 t 1. preceding questl\)fl. It is n , n0 .
mentioned tn the 1mme e) · h .. t d ,· "c<i"r to be efficiently
the question-answer pair used by Jackendoff (1972:229) to illustrate the . . d to e activa e 0 ~' '- ·
in the sense that it doe<:; not nee. .- •r Ne\'crtheless this
concepts of focus and presupposition: ent 10 the propo~1 1IL 1
processed as an argum . h . lion hence the constituent
denotatum has a f0cus relation to t e proro~1 b . - t ·d \\'hat makes
(S.2) a. Is it JOHN who writes poetry? (Jackendoffs (6.1 )) - n hence ii must e aci..en e
coding it is. a focus expres:.10 . h d cogn1t1\'e state of the
b. No, it is who writes poetry. (JackendoIT's (6.2)) f cal is not t e assu1ne - r
BILL
this const1tuen
.
t
f-
°• fr•ntinthea fC5SCCS
dd •·. mind nor the nature o
·
representation O its re e i.: d h, 'dicatc (the scmanllc role of
According to Jackendoff, in the question in {5.2al "the rresupposition is · I - bet\veen 11 an t i.: pre
the semantic re at1on h th, tis the nuH·1t•_1 0 r the rt'.stu11ran1..
n the replv \\' e er I . h
that someone writes poetry. John is the focus." Jn the ans\ver in (5.2b), the focus argumen t 1 _ - • h the fu.:t that 1t ts t ts
.. the presupposition is also that someone \\Tites poetry, and Bill is the d h dire'ti 0 nal) hut rat er '
can be exrecte to e i.. .·b·I , that is chosen to supply. th.e
focus, the new information being conveyed" ( 1972:230). To understand d t nother pt1S\l e dne, ·
denotatum. an no a - _ 1111 ··,peaker went tP x · It ts 10
why it is misleading to call the focu~ constituent Bill in this ans\ver "the .
missing argumcn t 1'n the open rrt•r··~ 1 t1
 212 Pragmaric re/atio'1s: focus Def111i1ion of focus 213
this relational sense that the focus element can be said to be what is novel is the fact that a particular denotatum is chosen as the agent
"unpredictable" or "non-recoverable" at the time of utterance. argument. This is what I referred to above as the FOCUS REU..TION. This
Notice, incidentally, that the accent on the noun restaurant in (5.1 ") is focus relation is perhaps better characterized as pragmatic than as
not "contrastive" in any clear sense. This observation will be of semantic.
importance in Section 5,5, where I wjlJ argue that the notion of The second proviso concerns the fact thar what Akmajian calls the
contrastiveness cannot be used to account for the occurrence of accented "focus component of the semantic reading" is closely related to what I
constituents whose denotata are discourse-active at the time or utterance have called the "(pragmatic) assertion" in Chapter 2. Akmajian thus
(such as the RESTAURANT in (5.J')). conflates here two notions which I think ought to be distinguished:
The fact that it is not the accented constituent itself that conveys the '.'pragmatic assertion" and "focus." The expression "[x = Mitchell]"
new information but the establishment of a relation between the indicates a relation between an element which is, and an element which is
denotatum of this constituent and an abstract proposition is clearly not, part of the presupposition, given that the x stands for "the (set of)
stated by Akmajian (1973) in his discussion of the role of focus in the individual(s) that urged Nixon to appoint Carswell," which is not a focus
interpretation of anaphoric expressions. Akmajian illustrates the denotatum. To capture the fact that the focus is the COMPLEMENT of the
relationship between focus and presupposition with the following presupposition it is necessary to separate the two terms of the relation.
example: · I propose then the following modified analysis of Ak.majian's example:
(5.3) MJTCHE.LL urged Nixon to appoint Carswell. the expression "[x urged Nixon to appoint Carswell]" in (5.3") is the
PRAGMATIC PRESUProsmON (the "old information"); the expression "[x =
Akmajian analyzes the proposition expressed by this sentence as the Mitchell]" is the ASSERTION (the "new information"); and the right-hand
complex semantic structure in (5.3'): side of the equation which constitutes the assertion is the Focus. In this
(5.3') "[ x urged Nixon to appoint Carswell J. { x = Mitchell]" analysis, the focus is indeed a term, but a term in a pragmatic relation.
The word "focus" is then to be underslood as shorthand for "focus of the
He then ...,.rites: · assertion" or "focus of the new information."

,., . Note lhat ... the focus component of the semantic reading is given as a
semantic relation, not a single tcnn .. The focus constituent of a
sentence represents novel information not because the constituent is
My definition of "focus" is given in (5.4). The terms "(pragmatic)
presupposition" and "(pragmatic) assertion" in (5.4) are as defined in
example (2.12) of Chapter 2:
,..
'
Ir necessarily novel, but rather because the semantic relation whlch the
'! constituent enters into is novel with respect to a given universe of (5.4) Focus: The semantic component of a pragmatically structured
discourse ... The entire expression ""[x = Mitchell]," then, is taken to be proposition whereby the assertion differs from the presupposition.
what the speaker assumes to represent novel infonnation to the hearer.
(1973'218) Example (5.4) implies that if a sentence evokes no presupposition, focus
I! . and assertion coincide. This situation obtains often (but not always) in
lj Akmajian here defines focus clearly as a semantic relation rather than as thetic sentences, such as It's RAINING.
a property of a constituent. 2
!l It is very important to understand that focus is defined in (5.4) as a
My concep1 of focus is s1milar to Akmajian 's, ~vi th t~vo 1enninological semaotico-pragmatic, nor a formal, category. It is defined a! the semantic
and conceptual provisos_ The first proviso concerns the expression .level of the (pragmatically structured) proposition, not at the gramma-
"entering a semantic relation." Since in Akmajian·s example we already .:tical level of the (syntactically structured) sentence. The pragmatic
know from the presupposition component of the semantic reading that Category "focus" must be sharply distinguished from its grammatical
the missing argument will have: the semantic role of agent in tho ··realization in the sentence, i.e. the syntactic domain in which it is
proposition it is strictly speaking not the SEMA!'.c-r1c RELATION entered into -, expressed and the prosodic means whereby this syntactic domain is
by the focus denotatum that is novel in the universe of discourse. Rather. marked, i.e. the means of sentence accentuation. The distinction between
 Definition of focus 215
214 Pragmatic relations: focus
. d Mitchell.3 It follows from my definition
focus and sentence accent is particularly important since-as mentioned the noun phrases the mov1es_an t be constituents whose denotata are
earlier-sentence accentuation is not a focus-marking device per se but a "f .. th t focus domains mus
of ocus a . . h added to presuppositions. As we
general device for the marking of semantic portions within pragmatically capable of producing assert1o~s w en ES or ARGUMENTS {including
structured propositions, whether focal or not. The focus construal of a d t are either PREDICAT .
shall see, such enota a Th. enta"tls that focus domains
J te PROPOSITIONS. IS
proposition is determined by a number of grammatical factors, only one adjuncts), or else comp e d. ,. e phrases noun phrases,
of which is prosodic. ·es (verb or a Jec iv •
must be PHRASAL categon . d sentences). Focus domains
A semantic element which is part of the focus component of a prepositional phrases, adverbial ~h~ases,ban se i"nformation structure is
ries Thts 1s so ecau
pragmatically structured proposition will be said to be IN Focus or cannot be LEXICAL ca t ego · . . with the relations
FOCAL, independently of whether the constituent coding it carries an
d "th words and their meanings, nor
not concerne WI d h f phrases or sentences, but
· r words an t ose o
accent or not. For example, if we were to use sentence (5.1) We wenr ro b~tween the mea~1ngs o f the relations between entities and. states
the MOY/ES as a reply to the question "What did you do last night?" the with the pragmatic construal ~ . E (ties and states of affairs are
designata expressed by the (relatively} unaccented constituents went and of affairs in given disco.urse s1tuat1ons. ~ I . I . I ·terns
1 t genes not 1n ex1ca 1 ·
10 wouJd be in focus together with that of the accented constituent syntactically expressed tn phi casWa h~~ ea predicate phrase can express the
movies. The expression "in focus" (or "focal") is the converse of the L t \ook at an examp e. 1 e . ).
e us .. a predicator by itself cannot. Consider ( 5.5 .
expression "in the presupposition" which I introduced in Section 4.3. focus of a propos1t1on, .
(Recall that I reserve the adjective "presupposed" for propositional And then. when we"d finished talking about pigs. we started talking TO
(5.5)
designata.) A denotatum which is not in focus is necessarily in the the pigs
presupposition. For example, in sentence (5.1) the topical subject . · ted and we could say,
pronoun we is in the presupposition. Jn the main clause. the preposition to alone is accen ome sense the
h t this preposition constitutes 1n s
The word or minimal constituent carrying the focus accent will be with Bolinger, t a . · . _ the remaining elements we,
referred to as the ACCENTED CONSTITUENT, e.g. Mitchell in (5.3) or mon·es "information p_oint'" of the ct\~~:e~e::~tic relation to each other. are
in (5.J). (For reasons of typographic clarity, I capitalize words, or talk, an~c~:;y~:Cs;v::a:~:~ ~severtheless the predicator ta cannot_ by ttsellf
sometimes morphemes within words, rather than the syllables, or pragrna . f that clause. Since its denotatum is pure y
segments of syllables, which carry the pitch accent.) I will avoid the be the focus constttuent o 1 t of infonnation iv hose addition to
tenn "focused constituent" which is sometimes found in the literature, relational, it cannot supply an.e emen . Th ·t ·ng ··,\'e talked x the
because accented constituents are not necessarily focus expressions and a Presupposition \Vould result in a_n assertion. e., n
. n Thus \~'htle the question-answer
pigs" is not a viable presuppost t 10 . .
because it tends to blur the distinction bet\veen a pragmatic category
. (5- .6)-though stilted-make 5en~e
. 1n
pairs
(focus) and a prosodic category (pitch prominence). It should be obvious
from what I have said so far that accented constituents are not necessarily (5 6) \Vhat was the relation betwe<:n you and the pig-; ~ A talking relation
coextensive with focal elements. While in our paradigm example (5.3)
~- \Vhat did you do to the p1g~·1 Talk to them
focus. focal element, and accented constituent all converge on the same the exchanges in ( 5.6") do not·
word (the proper noun Mirchelf). such convergence is by no means the a •\Vhat was the talking relatllHl between you and the pig~~ - A to·
norm_ One of the main issues to be dealt \vith in this chapter is the (5.6')
relation
£(>mplex problem of the relationship bet\veen focus domains and h •\Vhat did you talk the rigs~ To
prosodic prominence. We can refer to this problem as the problem of . d h VP talk to the,,1 are capable of
FOCUS PROJECTION, using a term introduced by Hl'lhle ( 1982). In (5.6), the NP u talking rellat1011 ant :h: presupposit1ons_created by the
The syntactic domain in a sentence 1vhich expresses the focus . ngful comp ement~ o . (5 6')
supplying meani , the bare preposition 10 in ·.
. h · ess1ons 1nVl1 1'ing
component of the pragmatically structured proposition '"ill be called questions. but l e expr · - '\ S _, the prepositional phrase to the pigs
the FOCUS DOMAIN. For example in (5. I) :ind ! 5 3) the focus domains are are not. The focus domain in ( _ . l i
 216 Prag111atic relations. focus Definition of focus 217

(or the verb phrase started talking to the pigs, see Section 5.6.l), not the It is important to state from the outset that focus domains may contain
preposition to. constituents denoting pragmatically presupposed propositions. We saw
The above remarks lead us to an important conclusion. Since, on the that the proposition which is evoked by the modification construction 1he
one hand, the referent designated by the complement of the preposition GR££N one in (5. 7a), i.e. "the shirt is green," is pragmatically presupposed.
(rite pigs) is not only discourse-active but also topical in the sentence- the Nevertheless, the NP which codes this proposition is in focus. Similarly.
sentence is about the relation betwt:en the speakers and the pigs-but in the following variant of example (5.3), in which the proper noun
since, on the other hand, this noun phrase is nevertheless part of the focus Mitchell is replaced by an indefinite noun phrase:
domain, it follows that focus domains must be aHowed to cont.ain aon-
(5.3") One of his close COLU.BORATORS urged Nixon to appoint Carswell.
focal elements. (As we shall see, the reverse is not true, i.e. focus elements
may not be part of topical domains.) I will return to this important the referent of the subject one of his close collaborators is in focus, but the
conclusion in Section 5.3.3 proposition evoked within the focus NP, i.e. that Nixon has close
Here is another example, involving a focus domain containing a collaborators, is pragmatically presupposed. If someone were to
modifier: challenge the utterance in (5.3") by saying "That's not true," this
challenge would involve the identity of the agent who did the urging, not
(5.7) a. Which shirt did you buy? - I bought the GREEN one. the notion that Nixon has close collaborators. Notice also that the
The GREEN one.
possessive dctenniner his in (5.3"), referring to the jndividual "Nixon," is
•GREEN
a topic expression, whose pragmatic status is unambiguously expressed
b. What color is your shirt? - GREEN. by the fact that it is an unaccented referential pronominal (see Section
4.4.3). The s.tatus of this deti:nniner is con1parable to that of the NP the
The question in (5.7a) may be answered either with a full sentence or witb pigs in (5.5) and the pronoun one in (5.7a), all of whom arc topical
a full noun phrase, but not with the adjectival modifier alone, even expressions within focus domains.
though the constituents which distinguish the second from the third Focus domains may not only contain constituents coding pragmati-
version, i.e. rhe and one, are fully predictable elements in the answer (the cally presupposed propositions but they may be coextensive with such
definite article is !he symbol of the identifiability of the referent, the constituents. A clear example is the adverbial clause When I was se~enteen
unaccented pronominal one is a topic expression with an active referent, in example (2.8), uttered in reply to the question When did you move to
see Section 4.4.3). Thus the focus domain of the answer must be the NP Switzerland? Both the time relation expressed by when and the
Ihe green one (or the VP boughr the gree11 one), not the adjective green. proposition "I was seventeen" are "ofd" in the discourse. As I
Indeed what the addressee is being infonned of is not the color of a shirt emphasized before, what creates the assertion is not the focus denotatum
but the identity of a purchased item. The string "I bought the x shirt" is 1 by itself but the establishment of a RELATION between the denotatum and
not a viable presupposition_ This is so because a modification relation the proposition. In the reply in (2.8), the assertion is created by
within a noun phrase is necessarily pragmatically presupposed, as an establishing a tin1e relation between two knov.'n situations. I will return to
appJicacion ofrhe lie-rest (Seclion ~.JJ 1-1.·iJJ reveal. Therefore aa adjectival the issue of presuppositiOll.'i \Vithin focu:i don1.:iins iu Section 5.4.J.
modifier alone cannot constitute a focus domain. Notice, however, thai Let me summarize the analysis presented in this section. The focus of a
the ill-formedness of the third :ins\vl.'.r in (a) is not due to the Fact that an proposition is that element of a pragmatically structured proposition
adjective cannot serve as a fucu>. don1ain. As (5.7b) sho'.vs, the adjective which makes the utterance of the sentence expressing the proposition into
phrase green does constllute .i well-fanned focus domain if its a piece of information. It is the balance remaining when one subtracts the
designatum is construed as the predicate of an asserted proposition. In presupposed component from a given assertion. When a sentence evokes
this sentence the addressee 1:> inJccd being infom1ed of the color of a no presupposition, focus and assertion coincide. Like the topic, the focus
shirt. The string "The shirt 1s x·..:(•kir"· fL1m1s a viable presupposition ..-, is an element which stands in a pragmatically construed relation to a
 Definition of focus 219
218 Pragmatic relations: focus
. . f this example in Section 4.1.1, the
proposition. But while the pragmatic relation between a topic and a As I observed in _the discussion oh \'" xpressed in the subordinate
·· "th children went to sc oo e .
proposition is assumed to be predictable or recoverable, the relation propos1tton e . f ( 2d) but informational 1Y
be . serted 10 the contex 1 o 4·
between the focus element and the proposition is assumed to be clause is not 1ng as d . d event which is assumed to be
. k f anted The es1gna 1e • .
unpredictable or non-recoverable for the addressee at the time of the speaking ta en or gr · temporal reference point for
utterance-. The focus relation relates the pragmatically non-recoverable to known to the addressee, mere\~ serves as a ·1·ons As we saw in that
d . th main clause propos1 i .
the recoverable component of a proposition and thereby creates a new the events expresse in e . . f the subordinate clause
. h positional status o
state of information in the mind of the addressee. Focus MARKING is then discussion, t e pre~up. h . status of its subject NP the
the formal mechanism for signalling a focus relation between a proposition has imphcattons for t e topb1c s a topic since the
this NP mav e seen a ·
pragmaticaIJy construed denotatum and a proposition. The focus of a children. In . one sense, f h- NP· in another sense, it is not a
proposition may be marked prosodically, morphologically, syntactically, proposition is about the referent o t e_ . , t ew to the addressee, i.e.
. . th comment about the topic ts no n
or via a combination of prosodic and morphosyntactic means. Following topic, since e . . \vied e of the topic referent.
the multiple-level approach to grammatical analysis sketched in Section does not increase the hea~er s knob gd bout the pragmatic status of
b hon can e ma e a
1.3, I wiJI emphasize in my discussion the relationship between prosodic An an.alogous o serva . , In one sense, this predicate is focal,
and morphosyntactic focus-marking devices, as well as the relationship the predicate went to school 1n (4.~d). b t the subi·ect referent· in another
·1 resses a comment a au '
between focus marking and other aspects of information structure, in inasmuc h as 1 exp -. . f , _ d) the comment does not
particular activation, identifiability, and topic. . · t f 0 cal since 1n the context o 4 2 I
sense .1t is no '
· The accent on sc 1ivo 1 can therefore not proper Y
constitute an assertion. h . oo;;odicallv similar to the focus
II d focus accent even thoug it is pr - . . I
be ca e a · ( 4 _,a) (The children 11 ·ent /o SCHOOL,
accent in the i_n~epe_ndent se~te;:ther~ the function of the accent is to
5.J.2 Focus and sentence accents \vhose proposition is asserte . . osition and to announce
reactivate the referent of the ~resupposed pro'.\ e proposition. As \Ve
It will be useful to elaborate briefly on an issue which I hinted at before · 1 c for the main-i..: aus
and which might appear to be a problem for the approach to focus and its role as a scene-setting o~1 I f t a)· be stored in long-tenn
. 1 positiona re eren s m
focus marking adopted here. Since the definition in (5.4) makes crucial saw in Section 3 ·. · pro . . h , _- 1 he activated in order
memory like entities and, ilke the latter. t e_;-_ ~~; Jccent in question is
reference to the contrast between presupposition and assertion. the ments in nev• a~-;ertion~
occurrence of focus accents (or other focus-marking devices) is to be use d as argu ·cENT I \vill return to the
er\)' called an .\CTIVAT!O~ "'
necessarily restricted to sentences whose propositions are ASSERTED. therefore more prop f · . _ ts and activation accents in
fT , ce betv:een ocu'.> aLcen
However there are numerous contexts- such as those mentioned at the issue oft h e d 1 ercn . - ·r;ed account of sentence
. ·h I \vdl suggest a uni •
end of the preceding section-in which the content of a proposition is Section 5. 7 , \.\ ere - ~ . en as t\\"O expressions of a
. . hich the t\vO tvpeo;; .lre se
presupposed, i.e. assumed to be kno\vn to the addressee, but in which the accentuation in ""'. · -h· r ><;tJhllshing a relation, either
single underlying discourse function t a 1 n c:_ -
constituent expressing this proposition nevertheless carries an accent. As - .• , de-.1gnatum and a prnr~1~1tion
we saw, one situation in \.vhich this happens is \vhen the constituent topical or foLal. hel\\~en a -. . · ·umo;;tances even the
h 1 under the nght discourse ciro... -
evoking the presupposition functions as a focal argument in an assertion. No.lice . t a 1 11 s ·ntence in (4.-'b ) ( Tl u, ( Hfl.DRF'- \\·ent 10 schoo 1l
0
But a constituent expressing a presupposed proposition may carry an ident1ficat1ona. a t: . • ,. _dverhl'il clause, as in After
·al 1 e non-loi.:a 1. a •
h
could occur 't'.'tt in a topK · · ih, HO!. .~f This sentence
accent even if it is NOT in focus. This situation arises often \vhen a t school he h,1d 10 c1e.in i
presupposed proposition serves as a TorrcAL ,1rgument (or adjunct) in an the CHtL{lRES went a · . ·h ·h the knO\~·n event of the
, in a contc:-;t 1n \\ IL
asserted proposition. One such contexr \\'.1S il!u~tratcd in (4.~d) \.vould be appropriate e.g . , ·t ·d with ;.t knov.·n event
. ~ fir schPol I\ L·L1ntra~ c
children s departure ' - F' . ll , the ·idverhial clause could
. else's departure. -1n.1 ;.. ·
(4.2} d. (John '-'"'as very busy that morning.) After the children wen! to involving someone · · h h- "pprnpriate in a discourse
_ ted T 1 ~ \\'OU 11 L l
also be entire I y unao...cen
SCHOOL, he had to clean !he H<i1 <;r 3nd go shorr1ng f<:•r the P-'.RTY
1
 220 Pragn1atic relations: }Ocus Focus structure and focus marking 221

situation in which the referent of the presupposed proposition would be Carswell was already known) and thereby invites his addressee to
both topical and discourse-active at the time of utterance. For example in accommodate this different presuppositional situation in the universe of
a situation where someone asked us about John's activities after his discourse. 4
children left for school we could answer He had to clean the HOUSE after
they went to school, where the postposed adverbial clause is entirely
unaccented. S.2 FocllS structure and focus marking
Here is another exan1ple illustrating the issue at hand. Consider again
5.2.1 Types of focus structure
sentence (5.3) with its NP focus domain. In a discourse context in which
the proposition expressed in (5.3) would constitute already-shared The approach to the grammar of focus adopted here is based on the idea
knowledge bety.•een the interlocutors, the accent might still fall on the that the focus articulations of sentences can be divided into in a number
subject NP Alitchell. Consider the possible replies to (5.3) uttered by of distinct TYPES which correspond to different kinds of pragmatically
speaker Bin (5.8): structured propositions. These focus types are used in different
.communicative situations and are manifested across languages in
(5.8) MITCHELL urged Nixon to appoint
Carswell. ( = (5.3)) distinct formal categories. Since in English the focus articulation of a
B: a. Well if MITCHELL did it, 1hen what's the problem? proposition is often expressed by prosody alone, and since accent
b. I wonder why MITCHELL did it. placement can be seen as a linear continuum from left to right, the
c_ Surpnse, surprise. So MITCHELL was the one. existence of such distinct focus types has gone widely unnoticed in the
By placing the accent on the same constituent as in speaker A's utterance literature on English. My approach to prosodic focus marking differs
i
I
(i.e. on the subject /t.fitchelf), speaker B uses the focus structure of A's then from most current approaches in that various focus-accent positions
' assertion as the point of departure for the assertion expressed in his reply, ..are not interpreted as signaling different points on a continuum leading
'

~I'
as a kind of "second-instance" focus articulation. Version (a) of speaker from the narrowest to the broadest kind of focus. Rather they are seen as
B's reply for example could be loosely paraphrased as follows: "Given the prosodic correlates for a small number of discrete information-structure
CATEGORIES, each of which expresses a different type of focus meaning.
previously known fact that some individual urged Nixon to appoin't
Carswell, and given the now known fact that the individual in question iS Now since there are more possible focus-accent positions than focus
~·. categories, it follo\';S that different accent positions can be manifestations
~ \ Mitchell, I ask what the problem is." The presuppositions evoked in such
of the same focus category.
' '
'
sentences are "layered," so to speak.
Such cases of layered presuppositions are nol limited to contexts in , One advantage of my approach is that it offers a way out of the
which a speaker literally reproduces a previously occurring focus "segmentation" problem I mentioned in Section 5.1.1 by identifying
prosody. B's responses in (5.8) would be possible also if A's utterance focus domains with major syntactic and semantic categories. Another
had been as in (5.9): . advantage is that it makes it possible to capture semantic correspon·
dences bet\veen formally divergent but functionally identical sentences
(5.9) A: ~1itchel! urged Nixon to appoint cARSV.ELL. across or \l.'lthin languages, in particular between sentences with prosodic
B: I wonder why MITCHELL did it and non-prosodic focus marking. The prosodic expression of focus
The pragmatic articulalion 10 speJker A's utterance to (5.9) is that of a ~ categories will be discussed Lo detail in Section 5.6. In the present section,
topic-comment sentence, in \\·h1ch the fact th:it Nixon was urged to my goal is mainly to establish the existence of such categories by
appoint Carswell is not pre:.upposed but asserted_ By choosing a different comparing a number of focus-marking mechanisms in different
focus articulation of the pr..:)pl1s1t1c1n 111 his reply, speciker B acts as if the languages.
presuppositional situation at the t1n1c of his utteran..::e \Vere in fact that 1 The sets of examples
V.'hich I \l.'ill discuss- many of which are familiar
evoked in (5.3) (in \vh1ch the f:i..::t th~it son1cone urged Nixon to appoint from previous chapters- are parallel to the examples which I used in
 Focus structure and focus marking 223
222 Pragmatic rela1ions: focus

. exampIes 1n
the English . (4 ·2) certain differences in .presuppositio nal
Chapter 4 to illustrate the category of topic (examples (4.2) (a) through
(c)). The three pragmatic categories established with those earlier structure were not overtly marked, the example sets dtsc~s~d here ~re
examples. i.e. the "topic-comm ent," the "identificatio nal," and the sets of allosentences proper, i.e. they express differences in info";atto~
'"event-repor ting" (or "presentation al") categories, will be reformulated structure via differences in prosody or morphosynta~. The ~1:"1ad
here in terms of their rocus STRUCTIJRE. By the "focus structure" of a identity of some of the examples in (5.11) and (5.12) will b.e exp atne)
sentence I mean the conventional association of a focus meaning with a later on as a case of functionally motivated homophony (Section ~.6.2.3 .
sentence form. The unmarked subject-pred icate (topic-comm ent) As usual, the questions preceding · t he examp Jes suggest
. discourse

sentence type in (4.2a), in which the predicate is the focus and in which situations in which the structures could be used appropnately :
the subject (plus any other topical elements) is in the presuppositio n, will (5.10) PREDICATE-FOCUS STRUCTURE
be said to have PREDICATE-FOCUS STRUCTlJRE; the identification al type
What happened to your car?
illustrated in (4.2b). in which the focus identifies the missing argument in
a. My car/lt broke DOWN.
a presupposed open proposition, will be said to have ARGUMENT.FOCUS b. (Lamia macchina) si ROTTA.e
STRUCTURE; and the event-reporti ng or presentationa l sentence type, in c. (Ma voiture) el!e est en PANNE
which the focus extends over both the subject and the predicate (minus d. (Kuruma wa) KosHoo-shi-ta
any topical non-subject elements), will be said to have SENTENCE-Focus
(5.1 J) ARGUMENT·FOCUS STRL'CTLIRE
STRUCTIJRE. S

The terms "predicate focus," "argument focus," and "sentence focus" I heard your motorcycle broke down'!
a My CAR broke down. . ,
play on the notorious ambiguity of the terms "predicate," "argument,"
b·. Si e rotla !a mia MACCHINA./E la mia MACCHINA che SJ e rotla
and "sentence" as referring both to semantic and to syntactic categories. c. C'est ma vorruRE qui est en panne
They evoke both differences in the respective syntactic focus domains, d. KURUMA ga koshoo-shi-ta
such as VP, NP, PP, S, and differences in the focus portions of the
(5. J 2) SENTENCE·fOCUS STRUCTURE
pragmatical1y structured proposition, i.e. predicate, argument, and
sentence. {The English word "sentence" is nowadays used almost What happened'!

exclusively to designate a Formal category, but its origin in Latin a. My ci1.R broke down
b. Mi si f rotta (ROTT A) ]a MACCHINA.
sententia "judgment" is still transparent in some of its uses.) Notice that
c. J'ai ma VOlTURE qui est en PANNE
the terms "subject," "predicate," and "sentence" themselves will be used d. KURUMA ga KOSHOO-shi-ta
in the traditional way, i.e. to designate semantic rather than pragmatic
categories (see the terminologica l remarks in Section 5.2.3 below). In Concerning the predicate-foc us examples in (5. \0), it is clear that in t_he
. . context prov1.d ed here the sentences
- \vould
_ be. most natural with
combining the semantico-sy ntactic terms "predicate," "argument," and minimal
. I or nu II subJ·ects_ The versions \Vlth. lexical ht_opic NPsJd are
be
"sentence" with the pragmatic term "focus," my intention is to capture pronom1na h
included here as possihle grammatical alternatives. \V Jc \~·ou
the correlation between certain fonnal and semantic categories and is pragmatically
required in discourse contexts 1n v.·h-ch 1 the. topic referent
_ f
certain types of communicati ve functions, such as the function of
less accessible. The t\VO Italian sen_te_nces in (5.1 lb) illustrate two ~7ad1
commenting on a given topic of conversation (predicate focus), of specified contex1 (the c te
e no!
d ccur in the m1n1mally
identifying a referent (argument focus), or of reporting an event or types whic h cou I o k I
. ·s perceived as some\vhal stilted by some spea ·ers). am
presenting a new discourse referent (sentence focus). There is thus a version i . d-.;;course contexts (sec
. . that these
c Ia1m1ng · \\\'Cl tvpes
- are equn a 1en! in a 11 L .
correlation between type of focus structure and type of comn1unicative
situation. that the opt111nality of the lexical
belov.·).
It is \~'orth noticing. in this conte:d.
The three focus-structu re categories are illu.;;tnited in the three sets of
NPs in (S. \0) is evidence for t I1eir topic::i.L 1 e non-focal. status
A
English, Italian, (spoken) French, and Japanese examples belo\v \Vhile in
 .!L4 I'ragniat1c rt'lat1v11s: jocus Focus structure and focus marking 225
constituent in focus can by definition not be omitted v"ithout depriving for example the marking of focus-structure distinctions within the
the utterance of some or all of its information value. One might argue morphology of the verb, as in various African languages {see e.g. GivOn
that the difference between a sentence with a lexical topic expression and 1975a, Sasse 1987, Watters 1979). Nor does the above account exhaust
a sentence \Vith a pronominal or phonologically null topic expression the formal possibilities offered within the four languages mentioned. For
should also be seen as a difference in focus structure, since in many example, under certain conditions both English and French permit the
discourse contexts topic NPs such as those in (5.10) may not be omitted placement of a focal object constituent in presubject ("topicalized")
without "loss of information." Ho\vever, what is lost in such cases is not position, resulting in the argument-focus construction called "focus
propositional information, in the sense in which this term is understood movement" in Prince (1981 b) (e.g. J.!ACADAJ.ftA NUTS they're called; see also
in this book, but rather the addressee's ability to activate the referent of Stempel 1981). Like French, though less freely, English may also use cleft
the topic with respect to which the information is to be construed as constructions; and French, like Italian, may use (a type of) subject-verb
relevant. Such differences in the pragmatic accessibility of the referents of inversion for a restricted set of thetic sentences (see Lambrecht 1986b,
topic expressions have no direct bearing on the focus category of the Section 7 .5). Determining the appropriateness conditions for the use of
sentence. alternative focus-marking devices for the same general focus category is a
Concerning the argument-focus example in (5.11) I should emphasize complex matter, which is beyond the scope of this chapter.
that the choice of a grammatical subject as the focus argument is an It is necessary to mention that focus-structure homophony may occur
artefact of the exposi1ion. The term "argument-focus structure" applies not only in prosodic systems, as in the English examples in (5.11) and
in principle to any sentence in \Vhich the focus is an argument rather than (5.12), but also in morphosyntactic systems. For example, in Italian and
a predicate or an entire proposition. Nevertheless, as we will see towards other Romance languages subject-verb inversion may be used not only
the end of this chapter, there is an intrinsic relation between the fonnal for thetic sentences, as in (5.12), but also for certain argument-focus
marking of argument focus and the role of the argument as a subject. It sentences (see the alternative Italian sentence in (5.1 lb) or the Spanish
i should be noted also that the word "argument" in "argumenl focus" is example in (5.51) below). The existence of such cases of homophony is
I used here as a cover term for any non-predicating expression in a consistent with the obser.•ations n1ade in Section 1.4.2 concerning the

,, .
:1 proposition, i.e. it includes terms expressing place, time, and manner. It is
neutral with respect to the issue of the valence of predicates
·· multiple discourse functions of syntactic structures.
The only focus-marking device v.·hich all of our examples have in

.. \.
~·
("subcategorization") and the argument-adjunct distinction.
The focus-marking mechanisms illustrated in these sets of examples
are in short the following: (i) exclusively prosodic (English); (ii) prosodic
common is PROSODIC PROMINENCE of a given syllable in the sentence. It is
also the only device which occurs by itself, without being complemented
t>y another coding system (as e.g. in English). It v.·ould seem therefore
and morphological (use of wa vs. ga with lhe subject/topic noun in that the role of prosody in focus marking is in some sense functionally
Japanese); (iii) prosodic and syntactic (\vord-order variation in Italian); more important than morphosyntactic marking. This is no doubt a
(iv) constructional (French and Italian)_ 6 I \vilt make no attempt to consequence of the iconic relationship bet\1,-·een pitch prominence and the
characterize different kinds of pitch accent or olher phonetic variations degree of communicative importance assigned to the focal portion of a
across languages or v.•lthin one language, as l do not consider these proposition (see 5.2.2). However, in insisting on the fundamental
relevant to the issue of focus marking (~ec 5.3.1 below). In general, I \viii importance of prosody in the marking of focus structure I do not wisb
also ignore secondary accents \Vh1ch n1ay accompany the main focus to claim that the relationship between focus and prosody is universal
accent, except when such accents SL'f\'C to 111ark particular information- Nor am I in a position to establish any systematic retatiollShip between
structure contrasts. language type and the use of one rather than another type of focus
The focus-marking strategics lllu-.tr..11t·J tn ('.\_IOI throut'h (5.12) do not marking. I believe that differences in the rhythmic structure of languages
exhaust the gran1ma1ical pos~1hil1L1e .. ftlunJ ~1cros5 languages. A n1ore account at least in p::irt for the use of particular focus-marking systems.
complete typology of focu~-n1arJ...in!_! rnc'-·h.1n1sn1s \V~iuld have to mention For. example 11 seems likely that in French the prevalent use of cleft
 226 Pragmatic relations: focus Focus structure and focus marking 227

constructions for the marking of focus differences is~! Jeast in part due to . marked for its topic function Oeft-detachment, wa-
the fact that this language has both a relatively riii"d constituent order morp~osyn~~:~c:~!y osition). In English, it is marked only prosodicall~,
marking, pr p . h t unlike the presence of prosodic
and a relatively rigid rhythmic structure (see my remarks to this effect in via the absence of an accent. Notice t la ,k f prominence on the subject
prominence on the predi~t~ phrase, a~ t: focus The topical subject of
Section 1.3.4). However, extensive typological research is necessary to
substantiate such ideas. In the remainder of this chapter, I will be NP is not a necessary cond1t1on for predtca . l (4 2c') and
concerned mainly with the nature of prosodic marking. a predicate-focus sentence may be accented (see examp e .
7
Section 5.5.2 below). · nt those
. . ( .. ) have only taken into accou
5.2.2 Predicate-focus structure Jn the representation in 5· 10 1 h' h necessary
aspects of the information structure of th.e sentences w J~h;r:ther focus
to distinguish the predicate-focus type .1n (~f l~e ~~or:rmation structure
In the PREDICATE-Focus STRUCTURE in (5. I 0), the relevant presupposition
evoked in the reply is that the speaker's car is pragmatically available as a
topic for discussion, i.e. that the proposition can be construed as a categories. A more complete .rep~~s~~;a~oonuld have to mention other
comment about this topic (see Section 4.3); the assertion is the of the different fseeantuter::e~s ':ell. .for example in the English sente~ce
Presuppositional ) h f 11 wing
establishment of an aboutness relation between the topic referent and . b' NP (Af, car broke DOWN t e o o
the event denoted by the predicate; and the focus is the predicate "broke involving a lexical su 1ect ; . ddition to that mentioned in
Presuppositional features are evoke tn a . . th t the
down." The sentence accent marks the predicate denotatum as the focus . NP evokes the presuppos1t1on a
(5.10'): (i) the possessive my car , evokes the
of the proposition. The information structure of (5.10) is schematically 11 th definiteness of the NP m; car
(")
represented in (5.10'): speaker has a car; e NP · ·d tifiable· (iii) the
resu osition that the referent of the is I en ,
(S.10') Sentence: p pp . I , evokes the presupposition that a second
My car broke DOWN. unaccented pronom1na m1 . d the
Presupposition: "speaker's car is a topic for comment x" referent i.e. the speaker, is. topical; (iv) both the determiner my ~n h
Assertion: "x == broke down" , k the presupposition that the r.eferents o t ese
NP my car evo e · d f th
Focus: .. broke down" expressions (i e the speaker and her car) are active in the min . o ~
Focus domain: VP addressee: the. first by virtue of being a pronoun, the second by virtue o
There is no contradiction between the representation of the (relevance) being unaccented.
presupposition as an open proposition- hence as lacking a truth value- What is not to be represented as a presupposition evoked by the
and the nature of the presupposed proposition as being pragmatically . (5 10"') is the connotation t h at some th in g happened -to the .
sentence 10 · _ _ - d b the question 1s
taken for granted since it is not the truth of the proposition that is at issue k , car This connotation \\·h1ch is create y ,
here but its assumed availability in the mind of the addressee at the time spea er s . rt of the discours~ context. not a grammatical prop~r.ty of
merely a prope .y _ ·ruation not to a presuppos1t1onal
of speech (see Section 2.2). the sentence. It ts due to a discourse si . . -. " .
Formally, the predicate-focus structure of (5.10) is minimally
. '
structure. Recall that the definition o
r " pra.grnat1c presuppos1t1on
- ll evoked"
tn

Chapter 2 requires that 3 . propos~~:~e~~;·~ef'\:~~;::::~ ~n!cture.

1
expressed in all four languages by prosodic prominence on an element
of the predicate, marking the predicate phrase as the focus domain. The to qualify as a presupposition in d. the present s1udy. the various
presence of such prominence is a necessary condition for predicate- Jn the theoretical frame\vork assume tn - (5.11) and (5.12) to be
focus construal of a proposition. (HO\\'ever, it is not a sufficient · ('\ 10) as v.'el\ as those tn -
sentence structures in - · - . CONSTRUCTIONS in the sense of
condition; see Section 5.6.1.) The focus structure of the sentence is also seen as GRA..MMAT!( ·\L •
discussed.bel ov.·-are - . l) · e as grammatical conlig-
expressed by the fact that the subject is marked as a topic, hence as being G (see Section 1 4 ·- , 1· ·
Construct.ion .rammar _ ' d osodic features concur to
rnorpho~vntact1c an pr
excluded from the focus domain. This topic appears either in the urations 1n \\' h ic h -. . nd"nces_ The information
f f 0 ·ti in use correspo ...
preferred form, i.e. as an unaccented pronominal (cf_ Section 4.4.3), or as produce particular orm_·_u L~·l - ~d n the sentences in (5.10), i.e. the
a lexical NP, or as both. In French. Japanese. and Italian, the lexical NP is structure of the propos1t1on t:xpresst: I
 228 Pragmatic relations: focus Focus structure and focus marking 229
pragmatic construal of the predicate as a comment about the topic- I should emphasize that the representation in (5.11 ')leaves unexpressed
subject, is seen as an intrinsic property of the grammatical construction. one important feature of the information structure of the represented
It is a fact of grammar, not merely a case of compatibility between a sentence: the designatum of the open proposition "X broke down" is not
pragmatically neutral grammatical structure and a given discourse only presupposed to be known to the addressee but it is also marked, via
function. The fact that the English construction in (5.IOa) may occur lack of pitch prominence, as being presently under discussion, i.e. as
also in discourse contexts in which no topic-comment or predicate-focus ACTIVE in the discourse. I will return to the important issue of the
meaning is called for (as e.g. in the case of non-referential subjects) is relationship between presupposition and activation in Section S.4.3,
explained by the fact that the predicate-focus structure is the UNMARKED where I will argue that the construal of the knowledge presupposition in
pragmatic articulation. The marked pragmatic articulations discussed the English version in (5.11) is in fact a consequence of the activation
below are distributionally more restricted. state of the open proposition.
From the point of view of the actual communicative situation in which
(5.11) js uttered, the schematic representation in (5.l l ') is considerably
5.l.3 Argumenr-focus s1ructure
underspecified. In the given situation, the common ground between the
The analysis of the AR.GUt>.tENT-FOCUS STRUCTURE.$ in (5.11) is similar to speaker and the addressee is not simply that something belonging to the
that of the previously discussed sentence (5.3) (,v1TCHELL urged Nixon to speaker broke down but that the addressee thinks that this something is
appoinl Carswell). The relevant knowledge presupposition evoked in the the speaker's motorcycle. The assertion made by uttering this sentence is
reply in (5.11) is that something belonging to the speaker broke down; the therefore not merely the identification of X with the speaker's car but
assertion is that th.is thing is the speaker's car; and the focus is "car." The also the correction of a mistaken belief on the part of the addressee.
accent marks the argument "car" as the focus of the proposition. However this aspect of the presuppositional situation is not formally
Schematically: evoked in the structure of the sentence: the utterance in (5.11) is
compatible also with a situation in which the addressee did not bold such
( 5.11 ') Sentence: My C.-!R broke down. a mistaken belief For example, (5.11) could serve as a reply to the
Presupposition: "speaker's x broke down"
alternative question "What broke dov.·n?" Therefore the representation
Assertion: "x = car"
Focus: "car" in (5.11 ') does not have to take such facts into account.
Focus domain: NP This is not to imply that the choice of a different context question could
not have an impact on the information structure of the reply. For example
As in the predicate-focus representation in (5.10'), the representation in the just-mentioned alternative context question "What broke down?"
(5. 11 ') ignores the various presuppositions evoked by the definite NP my could have as an answer the simple noun phrase My c.~R. an answer which
car, as these are irrelevant for the distlnction between focus types. The would be inappropriate in the context provided in (5. 11). In this case, the
focus domain in (5.11') ls given as NP rather than N in accordance with difference in appropriateness between tv.'o replies with respect to a given
our definition of focus domains as phrasal categories (Section 5. 1.1 ). The question is reO&ted in a difference in form (one is a full sentence, the
focal argument, and hence the constituent expressing it, is not "car" but other a sentence fragment). This difl"erence in form would entail a
"my car." But as in previously discussed cases, this focus constituent difference in presuppositional structure. The sentence fragment My CAR
8
contains a non-focal expression, nan1ely the possessive determiner 1ny, would therefore require a different information-structure representation.
which, by virtue of being an unaccented referential pronominal, is Grammatically, the focus structure of the sentences in (5.11) is
necessarily a topic expresS1l1n (see the d1scuss1011 of example t5.3") expressed in rather heterogeneous v.·ays. In English, we find focus
above). The topical status of the 1efcrenl of this determiner (i.e. the accentuation of the subject noun and lack of accentuation of all other
speaker) is expressed in (5 I l'J in the'. Lh'l th:c11 !he speaker .sippears in the consli!uents. In Japanese, we find focus accentuation plusga-marking on
presupposition line. a:. an en\1ty under di~cusston the subject noun, and lack of ac.centuation of the rest of the sentence (but
 230 Pragmatic relarions: focus Focus structure and focw marking 231

see note 6. p. 354). In French, and in one of the two Italian versions a dilemma that Jespersen had in mind in the passage from his Philos~hy of
clen construction is used, i.e. the semantic content of the proposition' is Grammar quoted at the end of Section 2.2. Recall that accord1n~. to
syntactically represented by a sequence of two clauses. (The cleft Jespersen (and other linguists of his time, seep. ~55,. note 9) the defirut~on
construction is obligatory in French but not in Italian, which also has of the predicate of a sentence as "the element which 1s added as something
the-more natural-option of using a subject-verb inversion construc- new to the subject" does not apply to a sentence like Peter said. that when
ti~n.) Notice tha: the first of the two clauses in the cleft construction (E /a used in reply to the question "Who said that?", a sentence whi~h we ~n
m1a ~ACCHINA, C est ma vorroRE) has the syntactic and prosodic form of a now categorize as having argument-focus structure .. (f~e _tenruno~~~cal
predicate-fo~us construction, while the second (relative) clause is entirely dilemma is Jess acute with the formulation "the subject 1s 1n focus since
unaccented, 1.e. has no focus at all. In other words, the focus articulation the term "subject," unlike "predicate," is commonly used to refer both to
of the pragmatically structured proposition, in which the focus a grammatical relation and to a syntactic sentence constituent.) .
corresponds to an argument in semantic structure, is grammatically The terminological dilemma mentioned here is is an express10~ ~f a
expressed by means of a sequence of two clauses neither of which is deeper, conceptual, problem which deserves to be e.lucidated, as 1t is a
formally marked as having argument-focus structure. The focus meaning potential source of misunderstandings and confusion. In the present
of these two-clause sequences is thus non-<:ompositional, in the sense that analysis, the term "argument-focus structure" applies to a sentence
it is not the computable sum of the meaning of its parts. Rather it is a construction in which a designatum v.•hich functions as an argument on
property of the complex grammatical construction as a whole. While this the semantic level of the proposition serves as the focus portion on the
construction is clearly motivated pragmatically, neither its form nor its level of information structure. Jn (5.11), the semantic theme argument
interpretation are predictable on the basis of general syntactic and required by the predicate break down, 'vhich syntactically appears as the
semantic properties of the grammar. subject NP my car, is the focus element of the pragmatically structured
The various focus-marking devices found in the constructions in (5.11) proposition "the thing that broke down is the spea~er's car." (~n
have o~e formal feature in common: in all four languages, the NP impoverished version of this proposition is represented_1n the Asse.rt1on
expressing the focus denotatum is the only accented constituent in the line in (5.l l') "X =car.") No\v notice that in this pragmatically
sentence. Prosodically, i.e. in terms of accent placement, the argument- structured proposition the focus is in fact construed as a PREDl.CATE,
focus structure can therefore be characterized as the REVERSAL of the namely the predicate "(is) the speaker's car.·· This means that in an
predicate-focus structure, in which only the predicate constituent argument-focus sentence like (5.11) the designatum of a subject NP (here
necessarily carries an accent. This prosodic difference between the mv car) is construed simultaneously as an argument on the level of
topic-comment type and the identificational type directly reflects the se.mantics and as a predicate on the level of information structure. .
diffe~nce in communicative function. In (5.10) it is pragmatically To capture tenninologically the conceptual distinctions I am drawing
predicated of the speaker's car that it broke down; in (5.11) it is here it may be helpful to use alternative labels for "f~cus"' and
pragmatically predicated of a broken-down thing that it is the speaker's "presupposition" which prevent the 1dentificat1on of predicate "".1th
car. (lb.is formulation is some,vhat misleading and 'viii be revised below; focus and of subject ,vith presupposition v•hile at the same time captunng
see also the remarks in Section 4.1.1 about the non-topic status of the perceived parallel. For lack (lf a better alternative, I _suggest the
presupposed open propositions.) In the former sentence, the semantic- expressions PRAGMATIC PREDICATE and PRAGMATIC SUBJECT. \\'hJch contrast
syntactic subject is in the presupposition and the semantic-syntactic v.:ith SEMANTIC PREDICATE and SE\1ANTIC SUBJECT. Thus ID (5.11) V.'e can say
predicate is in focus; in the latter sentence, the semantic-syntactic subject that the svntactic predicate phrase (or verb phrase) hroke_down codes both
is in focus and the semantic-syntactic "predicate" is in the presupposition the sema~tic predicate ·'broke d0v.'n" and the pragmatic subject "the x
The scare quotes around the word "predicate" in the preceding sentence that broke dov.'n," while the syntactic subject phrase my c~r co~~~ both
are the expression of a terminological and definitional dilemma: the semantic subject "speaker·~ car"' and the pragmatic predic~te ~~s) the j
"predicate" and "presupposition" seem to exclude each other. It is this spea k er •s car .·· N"t,·ce
" that the tern1<> .. pragmatic predicate and
,/
 232 Pragmalic rela1ions: focus Focus structure and focus marking 233

"pragmatic subject" are not synonymous with "comment" and "topic," but is directly reflected in the structwe of sentences. It is a matter of
which apply only to propositions with a particular type of pragmatic grammar, not merely of conversational inference. This is particularly clear
articulation. 9 in the case of the "clefted" construction in French (and in one of the
The non-identity between semantic predicate and pragmatic predicate Italian versions). In the cleft sentence, the pragmatic predicate (i.e. "my
in argument-focus sentences is captured in the information-structure car") appears as a syntactic predicate phrase, i.e. the right-band
representation in (5.11 ').In the Assertion line "X = speaker's car," which complement of the copula, while the semantic predicate (i.e. ubroke
symbolizes the pragmatically structured proposition .. The x that broke down") is syntactically expressed as a relative clause, i.e. by means of a
down is the speaker's car," the relation between the pragmatic subject and clause construction which is typically (though not necessarily) reserved for
the pragmatic predicate is NOT a topic-comment relation but an equational the coding of pragmatically presupposed propositions. The non-identity
relation. It would be incorrect to say that the pragmatic predicate "is the of semantic and pragmatic predicate is syntactically marked also in the
speaker's car" is a comment about (or "is true of") the pragmatic subject Italian VS version Si e rorta la mia MACCHINA, in which the pragmatic
"the x that broke down." Since this subject is semantically incomplete it predicate (the denotatum of la macchina) appears in the clause-final
cannot designate an identifiable discourse referent, hence it cannot serve position normally reserved for semantic predicates. Cleft constructions
as a topic. The relation between the pragmatic subject and the pragmatic and subject-verb inversions can be seen as grammatical strategies for
predicate is not one of predication but of IDENTIFICATION, as indicated by overcoming disparities between semantic structure and information
the "equals" sign (see the remarks in Section 4. I. I concerning the structure.
identificational sentence type illustrated in example (4.2b)).
The terminological and conceptual problem at issue here has its roots
5.2.4 Sentence-focus structure
in Greco-Roman grammatical tradition, from where it was imported into
generative grammar, in spite of Chomsky's attempt at defining Let us go on now to the SENTENCE-FOCUS STRUCTURES in (5.12). In these
grammatical functions in structural terms (1965:71ff) and in spite of his structures, no pragmatic presupposition is formally evoked, except for
later substitution of the term "verb phrase" to the term "predicate some of the non-distinctive presuppositional features also found in (5.10)

.. phrase." In the classical tradition the "predicate" portion of a

proposition is "what is said about the subject" (cf. Gr. karegOrema and
Lat. praedicatum ''what is revealed or expressed'), not ''what is true of the
or (5.1 l) (e.g. the proposition that the speaker has a car, that the speaker
is a topic, etc). One might suggest that (5.12) presupposes the proposition
"something happened." However, such a presupposition is merely
> subject," as in post-Fregean logic. The classical terminology was no
doubt dictated by unconscious pragmatic considerations. Since in most
situationally implied, not lex.icogrammatically evoked in the sentence.
What is formally evoked 10 (5.12) is an ABSENCE of the relevant
sentences subjects are topics (see Section 4.2.1), the subject-predicate presuppositions obtaining in (5.10) and (5.11): neither is the subject a
relation was naturally equated with the topic-comment relation for thC topic, nor is an open proposition '"X broke down" pragmatically
purposes of grammatical (hence logical) analysis. Unfortunately, this presupposed (the latter feature is formally neutralized in English and in
equation is misleading in the case of sentences with argument-focus the non-clefted version in Italian). Since the assertion extends over the
structure, as pointed out by modern grammarians such as Jespersen,1 entire proposition, assertion and focus coincide in these structures. It is
Paul, and many others. this lack of a presupposition that gives rise to the "eventivc"
To sum up, in sentences such as (5.1 la) the potential for confusion, and interpretation of the proposition. Schematically:
the need for terminological distinction. lies in {he non-isomorphic
(5.12') Sentence A1}' CAR broke down
mapping relation betv.•een syntactic and semantic categories on the one Presupposition
hand (semantic predicate "" verb phrase) and syntactic and information'"' Assertion ··speaker's car broke down"
structure categories on the other (pragn1at1c predicate: subject NP). The Focus "speaker's car broke down"
existence of this non-isomorphic n::latitin is not an artifact of my analysis Focus domain s
 Focus structure and focus marking 235
234 Pragmatic relations: focus
. t (or predicate-focus) structure,
As in (S.10') and (5.11'), the representation in (5.12') does not mention
the presuppositions evoked by the definite NP my car as they have no ~~;',~:~,~::~~ ~;~;;~:~ ::~~:,~;,:~~ed b\ th~;~::~:;,r i.'~;~~~
bearing on the determination of the focus domain, which in (5.12) is the focus structure. This observation comp emen s . f th
4.2.2 concerning the role of the subject for the formal expression o e
entire sentence. Notice that in (5.12'), unlike (5.1 O') and (5.11 '), the
assertion is not expressed as a relation, i.e. assertion and focus coincide, contrast between thetic and categorical sentences. . f f ns the
For the purposes of the grammatical coding of d1scours_e un~ l~t than
reflecting the non-binary pragmatic structure of the thetic proposition
(sec Section 4.2.2). marking of the subject as a non-topic appears to be n:io~e impo ad b an
the marking of the predicate as non-presupposed. T~s i.s s~g~e~te y
Grammatically, the focus structure of the sentences in (5.12) is
. f the formal similarities and d1ss1m1lant1es among
expressed in rather heterogeneous ways in the four languages. 10 In overall companson o · in
the redicate-focus, argument-focus, and sentence-focus versions
(5 10~ through (5.12). While alt argument-focus an~ sentence-focus
English, the sentence is prosodically (and syntactically) identical to the
argwnent-focus sentence in (5.10). The accent placement on the subject is
st~ctures differ clearly from the corresponding predicate-focus s;ruc-
an instance of what I referred to in Section 1.3.2 as "prosodic inversion"
tures in a given language, the sentence-focus and the argument- o~us
(see Section 5.6.2.3 below for further discussion). It would not be possible
structures in each language do not necessarily differ from each othe;, I.~.
to accent any other constituent without altering the focus category. In
th may be homophonous (as in English) or near-homophonous as in
Jae:nese and in one of the Italian versions). In other words, f?cus~
Italian, the focus structure is expressed via a combination of subject

st:Octure homophony seems to be tolerable only betw~n non-predicate

accentuation and subject inversion. Unlike the English sentence, the
Italian sentence may have a secondary accent on the predicate (the
focus structures_ The fundamental category boundary LS th~t between ~he
participle rotta) without affecting the focus category. This is so because in
resence vs. absence of a topic-comment articulation. This observation
Italian prosody is not the primary device for marking the focus structure; P - of the topic-comn1 ent structure as the
an accent on the verb is therefore not necessarily interpreted as a signal of corroborates my ana 1ysis
unmarked pragmatic articulation and of the other focus type~ as
predicate focus. French uses the avoir-cleft construction, in which the
marked departures from this basic type (see Section 4._2.1 ) ..As I ave
underlying simple proposition is expressed via a combination of two
repeatedly emphasized, in order to fully capture the re!auonsh1p bet\l.·een
clauses (see. Section 1.3), as in the case of the argument-focus
focus marking and focus meaning. it is necessary to u.nderstand the status
construction. In both of these clauses the accent falls on the predicate
of different sentential structures as ALLOSENTl::l"CES, i.e. as elements in a
phrase, i.e. in purely formal terms both clauses can be said to have
system of formal and pragmatic oppositions
predicate focus structure. Moreover, both the personal pronoun je and
the relative pronoun qui are topical subjects in their clauses and the
proposition in the relative clause is asserted rather than presupposed. The 5.2.5 Summarr
focus structure of the French sentence is expressed by the grammatical ns <1f tht:: thrt::e focus-5\rul·ture cacegones are
construction as a whole, in the same way as that of the cleft constructions The pragmatic articu Ia t io . .. .. " red1cate"
. d 1·n Table i_ Notice that the \~·ords argument and p
in (5.llc). Finally, in Japanese the structure of the sentence is similar to summanze ' ~ - d th semantic
in the top line refer to the semantic a.rgument an e
that in (5.1 Id) in that the subject is accented and marked \Vi th ga
d' cti\'e\v
However, unlike (5.l ld), the predicate constituent koshoo is also pre icate respe - - T· bl , renects the above-mentioned fact
prosodically prominent. 11
Th feature d1stnbut1on 111 a e -
e
th gument-· focus t\·pe. is. the rever''ll "'
of the predicate-.focus type
Despite the rather heterogeneous coding strategies employed in the t h at ear · 1 re!1eels. t h·i.: Th_·"n-binarv. sen1 anttc structure
tvpc
For the sentence· f ocu - '
1
four languages, the sentence-focus con..;tructions in (5.12) have one
5
• ~ 4 _2 2). The sentence-
\vhich charactcnzes .thettc propos1t1~1ns (st:e Seo._:tion . -,. lation of the
crucial formal property in common: the marking of the subject as a NON· h, the 1,1 pic -('Ommen! ar 1cu
TOPIC, whether by prosodic, morrihological. or syntactic means_ Since focus structure ex hi hi ts ne1 t t:hr· f) , us~pre"-u pposi t1 on art1cu lation of the
predicate-focus structure nPr t ~ \ L · -
topichood of the subject (or highest-ranking argument) is the defining
236 Pragmatic relations: focus Focus structure and focus marking 237

Table 2. Pragmaric arlicula1ion of 1he 1hree focus-slruclure categories comment articulation (predicate focus) with the thetic articulation
(sentence focus) or the identificational articulation (argument focus).
Argument in focus Predicate in focus
An example of a configuration in which two focus-structure types are
Predicate focus + combined is the structure illustrated in The kitchen, rou have to clean,
Argument focus +
which I will discuss in detail later on (examples (5.54) and (5.54')). In this
Sentence focus + +
configuration an argument-focus structure in which the focus ..you"
supplies the missing argument in the presupposed open proposition "x
has to clean the kitchen" serves at the same time as a comment for the
argument-focus structure. Since both in the predicate-focus and the topic "the kitchen," resulting in a sentence which combines predicate--
sentence-focus categories reference to the grammatical category "subject" focus and argument-focus elements. Another example of combined focus
(or highest-ranking argument) is criteria!, the header "Argument in types is the one discussed in Chapter 2, examples (2.1) Here comes the CAT
focus" in Table 2 necessarily refers to the subject argument in these two and (2.2") Here the cat COMES, where we saw a sentence-focus-marking
articulations. As I mentioned earlier. this is not true in the case of the device (the deictic here-construction) combined with a predicate-focus--
argument-focus category, since in principle any argument element of a marking device (the prosody and position of a subject NP). Predicate-
proposition can serve as the focus in a focus-presupposition relation. focus structure and sentence-focus structure are combined also in the
Nevertheless, as we will see later on, there is a formal correlation between Reagan-quote in example (4.49). An interesting Chinese example of the
the prosodic marking of an argument focus and subject status of the combination of sentence focus and predicate focus is discussed by
argument constituent (see Section 5.6. l). LaPolla (forthcoming). LaPolla observes that in a sentence like Ta si le
It is worth mentioning that the representation in Table 2 does not fuqin "His FATHER died," the sequence si le fuqin (die ASP father) is an
involve the feature combination "minus-minus," i.e. no allowance is event-central thetic sentence which serves as a comment for the initial
made for structures coding only pragmatically presupposed propositions. topic NP ta "he." In future work (Lambrecht in preparation) I intend to
Indeed for a structure to qualify as an independent sentence it must account for such focus-structure combinations in terms of the concept of
express an assertion, i.e. the proposition expressed by It must contain a INHERITANCE as used in Construction Grammar and other monostratal
focus. There are no independent sentences expressing only pragmatically syntactic theories. In such an account, various information-structure
presupposed propositions. (This is true even for sentences like I love you, features are analyzed as being passed on from one grammatical
which may have been said hundreds of times to the same addressee.) construction to another. 13
Sentences expressing pragmatically presupposed propositions typicaUy There is one important aspect of the grammar of focus which I have
appear as subordinate clauses, such as the unaccented relative clauses in not mentioned so far and about v.·hich I will unfortunately have very tittle
the French and Italian cleft constructions in (5.11). 12 ·to say in this book. This is the question of the amount and type of
The three major focus-structure types discussed In the preceding propositional information \.vhich can be indicated by a single focus accent
sections no doubt do not exhaust all possibilities. A likely candidate for a~ in a single sentence or clause. One example will suffice to illustrate the
fourth type is the "counterassertive" or '"counterpresuppositional" type issue I have in mind. Imaglne having a stimulating but very brief
proposed by Dik et al. ( 1980). v.·hich involves the polarity of a.· conversation with someone you didn't knov.• before and who you are not
proposition rather than some sen1antic domain \Vithin it (see example sure to see again. Imagine further that you V.'ant to express to that person
(5.25) below), It is also po~sible to con1b1ne different focus types v.'ithin a· your hope for some future meeting, v.·hich will be Jess brief. It is no doubt
single sentence construction, e.g by CL1mbin1ng prosodic and morpho~. possible in English to e:i.press the desired propositional content by means
syntactic devices. For example under certain conditions a propositioii of a single sentence, such as I hope we will meet again for more than five
with sentence-focus 0r argument·fucu~ articulation may serve as a minutes. Hov:ever, this sentence docs not express the pragmatically
comment for a given topic, re;.ult1ng in a combinalion of the topic-! structured proposition you have in mind. The syntax of this sentence only
 Prosodic accents: iconicity, rule, default 239
238 Pragmatic relations: focus
mark contrasts in the INFORMATION STRUCTURE of sentences, in accordance
seems compatible with pragm a f tc readings
situation. . which do not fit the given with the goal of this study as laid out in Chapter l. I am NOT concerned
with the use of intonation for other kinds of semantic or pragmatic
. In one reading of this sentence the fact tha . purposes, such as the marking of SPEECH-ACT DISTINCTIONS (e.g. question
will meet again is pragmaticall resu t you and .your interlocutor
your hope that the next meeti:gpwill ~p~sed and what is asserted is only vs. imperative intonation, see e.g. Culicover & Rochemont 1983:125) or
case the temporal adjunct phrase ·11 e o_nger than the first one; in this of SPEAKERS"" ATIITUDE.S toward propositions (see Section 1.4.3). One
will meet again fior more th w1 receive the focus accent (/hope we example should suffice to justify the distinction between speech-act
an five '-flNUTES) In h .
sentence what is asserted is th t h . anot er reading of the marking and focus marking:
in the undesirable presupposi~o~ot~atop; for a future me~ting, resulting
longer than five minutes· in th. hy ur present meeting was in fact (5.13) a. Your car broke DOWN.
we will meet AGAIN fior, more ISlhcase t e a~ent \viii fall on again (I hope
b. Your CAR broke down.
an five m1nutes) In ·rh c. Did your car break DOWN?
utterance conveys the · e1 er case, your d. Did your CAR break down?
misleading readings by ::::ti:e~;::·r~et us then. try to avoid the two
hope Hie will me I fi g evant portions of the sentence: I The accentual difference between the predicate-focus structure in (5. l 3a)
e AGAIN or more tha fl
utterance is less misleading than the rev~ ive Af!N~TES. Although this
and the sentence-focus structure in (5.13b) is preserved in the
odd. We have the desire to f p ious ones, it seems nevertheless interrogative versions in (c) and (d), even though the latter differ
pause a ter the first ace t . sharply in their intonation contour from their declarative counterparts.
prepositional phrase which follows is not en , suggesting that the
Sentences (a) and (c), on the one hand, and (b) and (d), on the other, have
belongs to an incomplete separate clause part _of the same clause but
The intended piece of · " . expressing a separate assertion. the same focus prosody, but they differ in intonation .
1n1onnat1on would be 1 Similar remarks can be made about the use of intonation for the coding
concisely) expressed in a bi_,..! I . more c early (though less
....... ausa sequence hke J ho .11 of speakers" attitudes to,vards propositions, which, following Ladd, I
and I hope it'll befior h ft 14 pe ive w1 meet AGAIN
more t an 1ve MINUTES will call the EXPRESSIVE use. As Ladd (1978:213) observes, "focus is
asserted information compatible wi~~e a_re constraints on the amount of
Facts such as these suggest that th . . signaled solely by the LOCATION of lhe accent; various intonational
leave this interesting topic for future given claus_al structures. I must characteristics such as greater volume and \videoed pitch range can also
is related to the remarks at th ~ese;rch. The issue brought up here be used to signal \vhat might be called 'emphasis." " A distinction along
number of unidentifiable refere~tse~ha~ c::tio~ 4.4.2 con~erning the
similar lines is made by Selkirk (1984:1980. \Vho suggests that we
partition the domain of "intonational meaning" into two components:
clause. It is also related to the p . . I be introduced in a single
and Role (see Section 4 5 I) S~nc1p e_ of the Separation of Reference the "expressiveness component'" and the ··focus-structure component"
. · · · timulating suggesf 0 . (see also \Voodbury 1987). To designate \vhat Selkirk calls the "focus-
existence of constraints on the . i ns concerning the

::~s~ ~/ single clause are ro::;~~tG~:6 ~ ~~~:.ti~~a;~a1t9 ;;,n a:~

0

9
structure component" I prefer the term "information-structure compo-
nent," for reasons which \\'ill become app;:irent in Secl!on 5.4, where I will
argue that not all of the accents covered by Selkirk's term are focus
accents.
Functionally distinct though they are. the various components of
S.3 Prosodic accents: iconicity, rule, default intonational meaning- the information-structure component. the speech-
act component, and the expressiveness comp•Jnent - have nevertheless an
5.3.J Accent, intonation, stress
important grammatical property in con1mon: all may be expressed by
"th to ~start my d.ISCUSS!On
I would like
accents · of the nature and f unc ,.ion o f sentence formal means other than prosody. depending on the language. For
reader t:t la ew general obs~rvati0ns. First. I \\·ould like to remind the example yes-no questions may lie m<lrked by question morphemes or
am concerned \Vtth prosndv· nnlv· inasmu"h
· '" 3"~ ,·, serves to
240 Pragmatic relations: )Ocus Prosodic accents: iconicity, rule, default 241

"'ord-order variation instead of, or in addition to, "question intonation"; are adYANlages and DJSad~·antages, involving a "contrastive" accent shift
a speaker's attitude may be marked intonationally as well as e.g. by on the second noun, with the Spanish Hay venTAjas y desvenTAjas, where,
various types of particles, such as the evidential particles (" AbtOnung~ in spite of identical discourse circumstances, no accent shift occurs.
partikeln") of modern German. This suggests that the different Another example: while English permits prosodic focus contrasts in such
components of intonational meaning are perhaps more closeJy related derivalional pairs as ls she ChiNESE? No, she's JAPAnese, no such contrast
than I am assuming in the present study. In a more exhaustive treatment; would be possible in German, even though the derivational structure of
the study of sentence accentuation and that of sentence intonation might the German adjectives is similar: next to the regular chiNesisch, jaPAnisch
be profitably combined. 15 there is no possible •curnesisch or • JApanisch. This difference between
Next, r should make explicit that my analysis of sentence accentuation English and German may be a result of the fact that Gennan has greater
takes for granled a theoretical distinction between meanJngful prosodic word-order flexibility than English, hence can express certain pragmatic
contrasts of the kind mentioned above and the non-meaningful contrasts syntactically rather than phonologically. 17
assignment of STRESS to individual lexical items via language-specific
phonological rules. Unlike the assignment of prosodic accents, the Iconic motivation versus grammatical rule
assignment of word stress is not meaningful. Therefore ungrammaticality
rather than meaningful pragmatic contrasts tend to arise when the stress Given the view of the role of information structure in grammar espoused
on a word is changed. To take a simple example, compare the difference in this book, sentence accentuation cannot be accounted for with the
in stress between the English word existence and its German equivalent same phonological rules as those which assign word stress. Nor can it be
ExistENZ. Changes in stress assignment in these words result in exhaustively accounted for with any rule which is uniquely formulated in
ungramn1aticality rather than in new meaning: English •exi.rtENCE and terms of constituent structure, i.e. without recourse to the communicative
German •£x1ste11z are phonologically ill-formed. Since sentence accents intentions of speakers.in given discourse situations.
are the formal expression of pragmatic relations bet\veen propositions The notion that sentence prosody is determined by communicative
and their elements, they necessarily mark relationships between (two or intention& rather than formal rules bas been repeatedly and forcefully
more) meaningful elements. And such elements typically appear as . expressed by Bolinger, as e.g. in the following quote from an early essay
different "'ords or groups of H ords.
1 (the tenn "prosodic stress" 1n the quote is equivalent to the term
Nevertheless, since words may contain more than one meaningful "prosodic accent'' in Bolinger's later \\'Ork):
segment (morpheme), focus-related contrasts are sometimes expressed, Prosodic stress does not HAVE to fall as I de.scnbed 1t. The heart of the
also WITHIN words, in particular in compounds and derivational .: matter is this very freedom 10 fall now here. now there, with the
formations. One example is the contrast in Thul 's not an adrANrage; speaker's attHude determining where ir v.·iJJ fall. A mechanical rule
that's a Dtsad~·anrage; another is the met.'.llinguistic distinction in I didn'ti demands that we predict directly where 11 will fa!L A functional rule
16 predicts indirectly: Jt will fall here. or there. !F the meaning is such-and-
say 11Ffinnation but cosjirmalio11 discussed in Chomsky 1970. In the
such; insfei1d of autoa1atism, we ba\·e a meaning. (1954:153)
unmarked case, the prosodic accent \Vi\l coincide \Vith the word stress as
assigned by the phonological stress rules. In the marked case, the accent view that focus prosody cannot be cx.clus1vely accounted for with
will override the \vord stress, <JC !eJ.st in languages Vr'ith "free accent semantically or pragina(ica!ly '"blind" phonological rules is now shared
position" like English. In a general language typology, the category ofa by most scholars, in spite of in1portant individual differences. 18 A famous
"free accent position langu3gc:'" n11ght be profitably contrasted with the example making this point forcefully 1s the (thctic-categorical) pair
established category of a ·'free \\'llrd order language," since accent JOHNSON died vs. Tru1nan DIED discussed by Schmerhng (1976:4lfl), in
position and conslituenL pos1L1t)n b..1\·.; comparable pragmatic fllnctions. '\Vhich l\\'O synta-::ui..:.ally identical surface strings express two different
The poss1bi11ty llf expr~-5~ing 11it~_1nnauon-s1ruc1ure contrasts \vithin focus meanings (i.e. predicate focus vs. sentence focus) via radicaJJy
words varies from \anguJg..: 11-, LinguJg.:: Comp:.ire the English pair There different prosodic structures.
 Prosodic accents: iconicity. rule. default 243
242 Pragmatic relations: focus
prosody must therefore be to show how prosodic prominen~ as an ico~c
The assi~cnt of different accentual patterns to a given syntactic information signal is converted into informational meaning by being
structure~1.e. the occurrence of prosodic allosentences-can no more be mapped onto grammatical structure, which is an essentially non.-iconic
captured in purely phonological terms than the use of different syntactic system for the expression of meaning. (This conversion procedure is what
structu~ to express a given propositional content-Le. the occurrence df H6hle 1982 refers to as "focus projection.") The iconicity of sentence
s~tact~c all.osente~ces- can be captured in purely syntactic terms (see the prosody is comparable, mutatis mutandis, to the iconicity of onomato-
discussion in Sec.t10~ 1..4). The interpretation of sentence prosody in poetic expressions. We know that the words for animal cries often
terms of communicative intentions is based on the notion of a correlation resemble, or are perceived to resemble, the sounds made by animals. But
between prosodic ~rominence and the relative communicative impor· to get from a rooster's crow to English cockadoodledoo or to German
tance of the pros~i~lly highlighted element, the prosodic peak pointing kikeriki we must first walk the constrained path through the grammars of
to the. commun1cattvely most important element in the utterance. these languages. Sentence accentuation may be iconic in its foundations,
Prosodic marking is thus in an important sense ICONIC, since it involves but it is filtered through the machinery of grammar.
a mo~ or Jess direct, ~ather than purely symbolic, relationship between The analysis of prosodic structure must therefore concern itself. among
meamng and grammatical form (see the remarks on activation accents in other things, with the formal principles \vhich determine where a
Section 3.3.1). meaningful accent \\'ill fall WITHIN A Gl\'EN STUCTURAL DOt.{AIN expressing
Ha~ng said this, I hasten to add that I do not consider it possible to a given denotatum within a proposition_ To account for the relationship
exp.lain sentence prosody without recourse to formal grammatical rules. betv.•een prosodic prominence and prosodically defined meaning it is
It IS clear that the relationship bet\veen prosodic prominence and necessary to assume the existence of a formal level that MEDIATES bet\veen
communicative importance can be at best PARTIALL y ICONIC. A sentence
the t\vO (see Selkirk 1984:197f0.
accent necessarily falls on a single word (or rather a single syllable within The need for such an intermediate level can be demonstrated by
a word), while the semantic domain marked by the accent typically comparing accent placement in languages \vith different word order
ext~d~ over a sequence of words, not all of which are accented (cf. the constraints. Consider these t\vo simple predicate-focus sentences in
prehm1nary remarks to this effect in Section 5.1.1). Thetic sentences like
Schmerling's above-mentioned JOHNSON died, are an excellent ca~e in
English and French:

point. In such.sentences, the focus extends over the entire proposition, yet (5.14) a. She doesn·1 have a panicubr\;. 1nteres11ng JOR
b E\\e n'a pas un n1et1er pJrt1cu\ierement tNTFRFSS-'1:-<T
only the ~ubJect of the sentence is accented. The prosody or thetic
sentences IS thus a problem both for a "purely pragmatic" and for a The t\NO sentences in (5 14) have the same meaning: and can be used in the
..purely formal" view of sentence accentuation. same discourse context to convey the same piece of infonnat1on. In both
''
~le accent assignment is motivated by pragmatic principles, it is not languages, the accent \vh1ch defines the focus Jorn.tin falls \~·1th1n the
free ID the sense of an iconic isomorphism bet\veen the communicative object noun phrase. ,~,hich 1s the last phrase in tht' sentence. and \V1th1n :i
I,

"
importance of a denotatum and the placement of the accent on a this phrase, It falls on the last word. But while 1n fngli<.h this !ast word is 1'
particular word. Just as the syntactic structure or, say, a cleh construction the head of this phrase, 1n French it is the adjective modifying the head I
cannot. be fully accounted for by explaining its discourse properties, the
prosoch~ structure of a sentence is not fully accounted for by explaining its
This difference is clearly not the result of a difference in commun1cat1ve
intentions. It is not the case that in Eng.li5h the noun iol> is the point of the i
appropn3teness in certain communicative situations. Bolinger's dictum infonnation \Vhi\e in French more importance is attributed to the
that the fOCUS accent has the "freedom to ral\ nO\V here, nO\V there, \Vith modifier intCrcssanl. If ,~·e \Vere to put the accent on 1nterc.Hing in English
the speaker's attitude determining \vhere It \viii fall .. ought not to be taken the result \VOuld he a different focus reading. (in French. the l\VO readings
too literally, either for English or for other languages. are compatible \\·ith the <.amc pr1)~odic structure.) \Vhat remains constant
There can be no one-to-one relationship bet\veen sentence accents and in the {\'.'O language~ 1s not the ~1~~fh.'.ta!1(1fl of the accent \nth a narrO\\'
communicative intentions. One of the tasks in the description of 5entence
 244 Pragn1a1ic rl:'/ations: .focus Prosodic accents: iconicity, rule, default 245

semantic denotatum but its final position \vithin the focus domain (here verb happens to be the last accentable element in the domain within
the verb phrase). ·Which the accent is assigned. 19
Further examples sho.,ving the need for grammarical rules of sentence~ Or consider exampJe (5.16). Imagi.ne a bricklayer on a ladder calling
accent placement are easy to adduce. Consider the English question in for a new brick for his wall:
(5.15a) and its spoken French equivalent in (5.15b);
a. every lime he NEEDS one.
b. cbaque fois qu'il en a besoin d'uNE.
(5.15) a. Who's THAT?
b. C'est QUI ~a? {lil. "it is who that?")
ne may argue that the accent on the predicate need in English is at least
The two sentences have the same meaning and can be used under the· ' directly motivated iconically: the pronoun one is unaccented because
e referent "brick" is active in the context; however no such argument
same discourse circumstances (the speaker may e.g. ask this questioq·
while pointing to the individual designated with the demonstrative be made for French. The presence of the accent on une runs counter
pronoun that). Nevertheless the sentence accent falls on a different word o any narrow iconic account. (Notice that the French sentence has no
in the two languages. If sentence prosody were entirely determined by "contrastive" connotation whatsoever: the point of (5.16b) is not to ask
iconic considerations - the prosodic point of prominence coinciding wit~ for ONE rather than two or more bricks.) It is true that the pronominal en
the pragmatic information peak- \~'e \.vould expect the same v.'ord to be· 'of it, of them" sho\vs the lack of prominence expected from topical
prominent in English and in French. In fact, given the presuppositioaa'i aphoric expressions; but the presence of the accent on une is not
similarly motivated.
structure of WH-queslions, in which the non· WH portion of the sentenc;
is normally pragmatically presupposed, we would expect the accent to rail · The limitations of a narrow iconic view of accent placement can be
. .
on the question \vord, as 1t does in French. However, here as before, th~
'I
emonstrated also within the prosodic system of a single language. A
accents are assigned on structural grounds, i.e. they fall on the las · triking example, which I have referred to before, is the prosodic
accentable constituent of the sentence (the French pronoun ;:.a is an xpression of the thetic-categoncal contrast in English, as illustrated e.g.
antitopic-i.e. post-clausal-constituent and as such not capable of ~ the earlier-mentioned pair JOHNSON died vs. Truman DIED, or in the
~ntrast between (5.IOa) lily car broke DOU'N and (5.J2a) }..fy CAR broke
carrying a sentence accent, see Lambrecht 1981). I \Viii return to th~
specific issue of the accentuation of \VH-questions in Section 5.4.4. 1
down. This contrast is clearly not amenable to an iconic expianation (an
To explain the differenci:: in accent placemi::nl in the two sentences in bservation which should not obscure the fact tha1 ii also resists any
(5.15) one might want to invoke some language-specific se1nantic. iplanation in purely syntactic terms). To take another example, consider
motivation. For example, since both who and rhat are (in a rathe( e French s~ntences in (5.17), illustrating three syntactic allosentences
vague sense) "new" to the discourse, one might argue that each language or the question "Where are you going?" Examples (a) and (b) represent
simply picks one of lhe two nev.' eJemi!nts as the bearer of the accent: ·poken French, \vhile (c) represents the standard \"'ritten form (notice
.
Besides the fact that such an explanation \~·ould introduce un..velcom
' at (a) is r-;or an echo question like the corresponding English sentence
randomness into the notion of iconic n1oti\'ation, it could not account f~{ ou're goins; WHERE.'):
the existence of the alternative French version of (5. l 5b) in (5. l 5b'); .. a. Tu vas ou? '"Where are you GOir->c..'1"
b. OU tu \'AS? '"\\'here are you GUl!'.G?""
(5.1 Sb') Qui c'EsT <;:a 1 c. OU vas-Tu1 "Where are you GOING?"' "
'

In (5.15b') the accent falls ne1thl'.r .._)n the dem.._)nstrative nor on While there n1ay be a subtle pragmatic difference betv.·een (a) and (b), and
interrogative pronoun. but on the \erh. which cJ.n h<.1rJly be argued to be hile there clearly 1s a register difference betv.'een (a)/(b) on the one hand
the ''nt:\•/' ele1nent or cnn1n1un1<.:,11!\c h1ghltght of the proposition. As in and (c) on the other, these differences are not differences in focus
the previous cases. the Jccent r~db where lt does in (5.15b'l because the structure. Nor arc they differences in en1phasis in any clear sense of this
 Prosodic accents: iconicity. rule, default 247
246 Pragmatic relations: focus
Finally, Ladd (1978:85), \Vho, like Schmerling 1976 and Selkirk 19~4.
te~ .. It ~ay be tempting to attribute pragmatic meaning to the accent rejects the very notion of ··regular stress rules," proposes the following
pos1t.J.on in (a), since the word oU represents in some sense the focus of the version which he calls the "Revised Focus Rule': "Accent ~oes on t~e
question_ (the fact that the addressee is going somewhere being most a~entable syllable of the focus constituent." Like Hall_1day's tome
pragmatically presupposed; see Section 5.4.4 below). But if the placement rule, Ladd's version of the phrasal accent ru\e-whtch I take to
designatum of oU is the pragmatic focus, Y•hy are the verb i•as and the hold, mu ta tis mutandis, for English, French, and many other languag~s ~
bound pronoun tu accented in (b) and (c)? The answer is of course that captures an important generalization, provided,. of course, that it ts
give~ the focus-structure type of these questions, French gramma; complemented with a principled account (not provided by Ladd) of wha.~
reqwres the accent to fall on a specific PLACE in the SENTENCE, i.e. at the constitutes a "focus constituent" and what the "most accentable syl.la~le
end, rather than on a specific WORD expressing a certain DENOTATUM. The within such a constituent is. One of the goals of the present analysis ts to
purely grammatical character of this requirement is all the more striking provide such an account. The reader should keep in mind. that in the
since in (c) the sentence accent falls on a bound morpheme (a so-called present framework any accent placement rule is seen as applying not only
"clitic" pronoun), which grammatical tradition has correctly identified as to focus domains but to pragmatically construed semantic domains in
being a tonic or unstressable. 20 general (see the preliminary remarks to that effe_ct in Se~tion 5. l.2).
The minimal conclusion to draw from such crosslinguistic and It is a generally acknowledged-though not uniformly interpreted-fact
language-internal observations is that while accent placement is that what Ladd calls the ··most accentable syllable of the focus
~ragmatically motivated, the prosodic expression of pragmatic meaning
constituent" strongly tends to be located at. or towards, the end of
ts nevertheless mediated by rules of grammar. 21 The assignment of the that constituent, at \east in languages like English and French. ~s a
sentence accent to a certain position within a phrase cannot be said to be general rule, we may say that a sentence accent. serves t_o mar_k the ng~t
uniquely determined by semantic or pragmatic principles. At the very boundary of a pragmatically construed semantic domain. This seman_tic
least, we must allow for certain PHRASAL ACCENT RULES which, given a domain may extend lefnvard to,.vards the beginning of_the sentence, 1.e.
semantic domain to be signaled, assign the accent to a certain position its major portion may PRECEDE the accented word It 1s a fundamental
within the phrasal constituent expressing this domain. principle of information structure, i.e. of grammar, that_ a sentence accent
An early proposal for such a phrasal accent rule is made by Halliday marks the END of a semantic domain, \\'hose BEG1Nl"'ING is marked by non·
(1967). After giving his definition of focus (quoted in 5.1.1 above), prosodic means. in particular by phrase structure. I 'viii call th~s
Halliday states the Following general rule: "The tonic falls ... on the last fundamental principle the GENERAL PHRASAL "'ilCENT PRINCIPLE. This
accented syllable of the item under focus" (1967:206). Halliday's rule
principle can be stated as fol\o\\·s:
captures an important generalization and is in my opinion essentially
correct, though it needs to be further specified in order to account for Gfr-IER.Al PHRASAL A(CENl PR!NCIPl.F A phra.;a\ accent marks the right
important exceptions. A modified version of Halliday's rule is proposed (5 18) boundarv of a svntact1c domJ1n expre~s1ng a pragma11call) construed
by Jackendoff, who restates the notion ••item under focus" In terms of portion of a proposition
syntactic phrase structure: "If a phrase P is chosen as the focus of a
sentence S, the highest stress in S \Viii fall on the syllable of P that is Notice that Ladd's simple term ··focus constituent" has been replaced in
assigned highest stress by the regular stress rules" (1972:232). As pointed (5.18) by the cumbersome "syntactic domain e.xpressing a pragmatically
out by Schmerling, Ladd, Selkirk, and others, Jackendo!Ts rule, \vhich is construed portion of a proposition.·· This 1s necessary because t_he
based on the notion that "regular stress rules" assign accents domain in question may be either a focus don1a1n c•r a topic domain.
automatically within syntactic constituents. 1s unable to account for Notice also that the "syntactic d1.1ma1n" men11oncd 1n (5 18) 1s not
sentences in which accents do not occupy the rredicted position \vithin necessarilv coterminous \vith "syntactic constilucni.·· :\s \\'e have seen.
focus phrases (see Section 5.3.3 bek.1\v) and as I -\vill she>\\' in mc•re detail in the next sccl1L1n. the activated
 248 Pragn1atic relations: ]Ocus Prosodic accents.· iconiciry. rule. default 249

designarum marked by a phrasal accent is not ahvays coextensive with As Ladd observes, "the accent in {19) is in no sense 'contrastive,' as it is
that of the syntactic constituent carrying the accent. often said to be: the meaning of B's reply is not the explicit contrast 'John
It is worth pointing out that the grammatical mechanism described in doesn't read books, be writes (reviews, collects, bums, etc.) them.'
the General Phrasal Accent Principle is not "natural," in the sense that it Rather, the point of the accentual pattern is that books is deaccented; the
does not follow from general extra-linguistic n1echanisms of perception focus is broad, but the accent falls on read by default" (p. 81). The accent
or interpretation. It is in fact the opposite of the natural mechanism on read is then neither iconic in the narrow sense, since it docs not
whereby a noise signals the beginning of an event, as e.g. the mechanism highlight a specific denotatum, nor is it directly defined on syntactic
whereby the sound of a starting pjstoJ marks the beginnjng of a dash, Jn· _structure, since the focus domain cannot be identified with a syntactic
the sports event, it is the beginning of the dash that is signaled. constituent. Instead, it is determined by the pragmatic status of the
acoustically, while its end. i.e. the crossing of the finish line, is measured denotatum of some other element in the sentence.
by non-acoustic (mechanical or electronic) means. fn sentence prosody,' The "deaccenting" phenomenon in (5.19) finds a simple and
the acoustic signal marks the end of the signaled domain, allowing for, straightforward explanation in terms of the concepts of activation.
elements following the accent to be interpreted as outside that domain.. presupposition, and topic, as defined in this book. 22 In the reply in (5.19),
The General Phrasal Accent Principle is a principle of grammar and must 1 the (generic) referent of the NP books was activated via mention of the
be stated as such. specific book "Slaughterhouse-Five" in the question (see the discussion
in Section 3.4 concerninE automatic coactivation of types, given
activation of tokens). It may therefore be coded as an unaccented noun
'• phrase, and its referent may bear the pragmatic relation of topic to the
5.3.3 Default accenruation proposition (the sentence is about the relationship between "John" and
In the preceding section I argued that prosodic accent "books"). The focus structure of the reply sentence in (5.19) can then be
determined by a combination of two factors: pragmatic analyzed as follows. Since the sentence is a topic-comment sentence, with
grammatical rule. It is now necessary to comment on a third factor the subject as the primary topic, the focus is necessarily expressed in the
influencing the location of a focus accent, \Vhich I have repeatedly hintedr predicate. The focus domain is therefore the verb phrase, and this verb
at in the preceding discussion. This factor is ~vhac Ladd (J978:81fl) has phrase must carry an accent. \Vithin the verb phrase, however, the object
called accent placement by oEFAUL T. Default accentuation is the prosodic- NP is marked, via lack of pitch accent, as a topic expression with a
phenomenon \vhereby an accent is assigned to a constituent neither for· ' discourse-active referent, hence as not being in focus. Now since the
pragmatic reasons (i.e. because the denotatum of the constituent is to oe sequence doesn't read in (5.19) is not a single constituent, and since the
highlighted). nor for structural reasons (i.e. because the constituent verb phrase nevertheless expresses the focus, we must identify as the focus
occupies the unmarked accent position), but because accentuation of an./. domain the verb phrase while exempting one of its constituents from
other constituent \vould result in a different - unintended -pragmatic' focus status. The representa1ion of B's reply in (5.19) is then as in (5.19'):
construal of the proposition I
(5.19") 10•[ John] n>< [doesn't READ 1(,,[ books l l
The phenomenon 1s n1c1"t e:.h\ lo obscr\'e in sen!ences in ,,,.hich a·
constituent v.·hicll coulJ reccl''c tl~e c.1ccent is '\k:accented," as Ladd cali'S! A similar analysis can be given to exan1ple (5.16) above. In the clause
it, for prag1natic (or '>01net1n1e~ phPnok1g1cal) rea~ons, resulting in a shifi. every 1i1ne he _.,.EEDS one, the post-focal object one, \vh.ich stands for "a
of the accent to a preccJ1n1? 11r r,.. \Ju,~·1ng constituent or syllable. One of brick," is a topic expression wilb an active referent of the same sort as the
Ladd's examples 11lu~1rc1t1n.i; JeL1ul1 ,1cccnlua11un ts the fol\o\ving: NP books in (5.19). The behavior of these unaccented topic constituents is
analogous to that of the possessive determiners his and my and the
(5.19) A H;i~ Juhn rc..icl ~L,u!o'hkrh•·U'·~·-F1,c'' ( ~ L;idd's (19)) indefinite pronoun one in (5.3"), (5.11'), and (5.5) respectively, which
B No. John d"csn·l vL·-;, h'"'k' have the function of topic expressions within lhe focaJ NPs one of his
 , default 251
Prosodic accents: iconicity, rule
Pragmatic relations: focus
clus ion is
the two dom ains . Thi s con
cl~ COLLJ.BOR.ATORS, ml' CAR
and th 4.2.2, map pin g rela tion betw een here. It is
e green one (see also Section roa ch to focus dev elop ed
example (4.20) and disc~ssio~). unp rob lem atic with in the app pos itio nal
the non -seg men tabi lity of pro
a.!: tedE ngli sh, . una~nted focus d
topic con stitu ents con sist ent with the view of mm atic al
may also PRE CED E the tion s 2.2 and 5.1. The gra
--· • con stitu ent in a . info rma tion exp ress ed in Sec betw een
oma1n, as in (5.20), which I bor row this asy mm etri cal rela tion ship
aga in from Lad d (1978:84): mechanisn1 whi ch allows for s diff eren ce
ent stru ctur e is the mar ked nes
c rete TOMORROW. ( focus stru ctur e and con stitu (see Sec tion
(5.20) I'm leaving f,or Lad d's (31)) ente d refe rent ial con stitu ents
betw een acce nted and non -acc
=
ked for the
stitu ents are positively mar
te, whose referent . 3.3.1). Since non -acc ente d con
!n (5.20) the nou n phr ase Cre mus t be discourse-active veness com bin ed with lack or
pro min enc e
JD the utte ranc e con
text . k .
• is mar ed v1a d eaccentuat1on as a topic
. feature "ac tive ," and since acti with in focus
. al stat us of such con stitu ents
con st1t uen t, hen ce as bein g exc luded fro
the focus enta ils topic stat us. the non -foc
I m the focus. As a result, dom ains is una mbi guo usly mar
ked .
accent must fall on some other e eme nt of the sen ten BY d e1au r erve the
ce. lt it falls ion is that it al\o\vs us to pres
f. d . .
on th e fimal con stitu ent of the _ocus oma1n, i.e. of the verb phr ase 2J The One adv anta ge of this con clus (see Sec tion
Lad d's Revised Foc us Rul e
focus stru ctur e of (5 ·20) 1s .
as 1n (5.20'):
· gen eral izat ions exp ress ed in acc ord ing to
eral Phr asal Acc ent Principle,
5.3.2) and implicit in the Gen on the LAST
(5.2 0") TDP[ I] me[ 'm leaving
for Top[ Crete J TOM ORR OW] . of cases the focus acc ent falls
which in the grea t maj orit y r, hO\\'ever. that the
- ACC ENT ABLE con stitu ent
of the focus dom ain. It is clea
~t so hap pen s that the acc ente d final con sti w1ng the topi c Crete rath er than
tuen t follo erst ood here in a prag mat ic
IS the deic tic adverb1·a1
tom o
rro1v an exp ress1on · term "ac cen tabl e" mus t be und s dom ains
s' - type which nor mal ly doe s topi cal elem ents \Vithin focu
not attr act a focu s acce nt ( see ect1on 5 6 I below) s·1nce suc h adv erbs gram mat ical sense. Una cce nted re prev ents
· · · sense that thei r lexical natu
arc usu ally una cce nted , th e acc ent on tomor · (5 are ··un acc enta b\e" not in the with in the
ro111 in .20" ) tend s to be Rat her they are una cce ntab le
. I
inte rpre ted as "contrast1"ve·.. n view of the
. I · them from receiving an accent. the com mu-
ana ys1s to be pres ente d in 1n \vhich they occ ur, given
Sec tion 5.6.1, it is imp orta nt
t lize
in (5.19) and part icul ar disc our se con text s be acc ente d
o rea that m (5.2 0). as ake r. The se elem ents can not
(5.16), the shif ted acc ent d oes NOT necessa ·1 · 1 nica tive inte ntio ns of the spe arti cula tion
n Y sign a the con tras tive result in an unin tend ed focus
t.
arg ume nt focu s of an iden tific a 1ona1 sen tenc e (as tn · I' bec aus e acc enti ng them \vould stit uen t" in
m not leaving for the con cep t of ··ac cen tabl e con
Crete TOD AY I'm going th ere TOMORROW) The sent e Id of the sent enc e. I \Viii retu rn to in Sec tion
' nee wou be equ ally bet\veen focus and pred icat es
app rop riat e in a con text like "W h t d . the ana lysi s of the rela tion ship acc oun t or
to Cre te; I'm \\'here I \v1ll pres ent a re\'ised
tha r. o you MEA N I nev er go . 5.4.2 and aga in in Sec tion 5.7.
goi ng ther e TOM ORR OW •"wh e re e 1act that the Speake · occ ur wit hou t
· r is goin g to Cre te a refe rent ial con stitu ent may
IS not prag mat ical ly pres
upp ose d b t the con diti ons und er whi ch
a~serted. I take the pro sod ic stat us
of suc h topi cal con stitu ents WI'th~In !OCUS dom ain t b e a stro ng an accent.
res510ns may occ ur \v1thin fC>c
us dom ains
fi eca te o "to - " s o The exte nt to \Vhich topi c exp
con irm .ati ono fthe rea lity ofth pie as a form al cate gor y in n_ F0r exa mpl e. \\'e sa\\' in the analysis of
the gra mm ar of Eng lish .
g ry is sub ject to typo logi cal variatio ,ess or-t opic which in
in Cha pter 1 that a pos
9) and (52 0 sen tenc es (I. I) thro ugh ( 1.3) phr ase
Exa mpl es suc h as (5.16a), ( (5.1 ' · ) con firm the obs erva tion ' ve dete rmi ner v.·-ithin a foca l nou n
first mad e in Sec tion 5 · 1· 1 exa mpl es (5 5) (I 6) d . Eng lish is cod ed as a pos5essi an and Fre nch as a
con stit uen t . h : , ~ - an disc ussi on). -that (the NP rny C4R in (I.I )) may
be cod ed in ltab
s
top ical non -sub ject s Wlt acti ve refe rent s may occ ur \Vtthin ressic•n (Ita lian rni. Fre nch Je). tv1y ana lysi
r..._ 1
1\J\;< l_J verb phr ase
s Thi s fact 11 ows us to dra\v an im rt sen tenc e-in itial arg ume nt exp pter s 6 and 8) reve als as a
· a
b po ant con clus ion of spo ken Fre nch (sec Lam bre
cht \9~6h. ('ha
con cern ing the rela tion sh· lang uag e that it systemallca!ly
avoids
syn tact ic stru ctur e, a concl~
ionet:~~c~ ir:s od1 c _ focu <; mar kin g and
general syn tact ic feat ure of this
ents app ear
15 ..:'0.). 111 \\'l11ch h1p1c con stitu
sible tom t ~1~e hint ed at repe ated ly in stru ctur es such as (5.19"') nr inal topi c
the prec edin g sect ions : it is pos stru ctur e and phr ase ~roken 1-'re nch, non -pro nom
stru ctur e only und er the
d. . a c ocu5 wit hin focu s dom ains In in righ t-de tach ed
s dom ains he ~ Jre reguL1rly pL.iced
allo wed to con tain non · f oca coin 11t1on th:} syn tact ic focu constituent<; \vith active referent
e ement-; Th ere c3n ~ue no one -to- one
 252 Pragn1atic relations: .focus Prosodic accents: iconicity, rule, <kfault 253

(antitopic) position, rather than directly under Sor VP. For example the the following two propositions had been activated in the utterance
spoken French versions of (5.19) and (5.20) might be as in (5.21a) and context and could therefore be taken as presupposed: "Bill Clinton
(5.22a). rather than as in the corresponding (b) sentences whose basic smoked pot twenty·five years ago" and "Bill Clinton did not inhale the
structure is similar to English in the relevant respects: smoke." (Recall that for a proposition to be taken as presupposed it does
not have to be considered true by the interlocutors; see Section 2.3.)
\5.21) a. Jean [ 1l en lit PAS] de livres. ··J_ doesn't REAO books." Given this conversational background, the second sentence in (5.23) can
Jean he of-them read not of books
b. 'l Jean [ ii lit PAS de livres ].
be said to evoke the presupposition "Jerry Brown breathed (pot smoke)
in x way." The assertion expressed by this sentence is then "x = out," and
t5.2:!) a. ( J'y \-'ais DEMAIN] en Crete. ··r1n going to Crete TOMORROW." the focus is "out." Since the bound morpheme ex- expresses the focus, it
I there go tomorrow to Crete
is the only constituent which can receive the focus accent. The mechanism
b. [ Je vais en Crete OEMAJN ].
whereby the focus accent is assigned in (5.23) is essentially the same as
In these examples, the topic constituents corresponding that in (5.19) and (5.20), even though the pragmatic status of the
complements books and Crete appear as antitopic ti.e. non-argument) unaccented element is slightly different. In both situations, it Ui the
constituents after the clause, thereby allowing the focus accents to fall in presence of a non-focal element within the focus domain that causes the
25
normal clause-final position. The information-structure function of the accent to occupy non-final position in this domain. What the denotata of
detachment construction in French is then not only to satisfy the the words books, Creze, and ·hale in our sentences have in common is that
Principle of the Separation of Reference and Role (see Section 4.5.l) but they are both discourse·active and in the presupposition.
also to allow for a closer match between focus structure and phras~ Notice, again, that the occurrence of the focus accent before the
structure, by eliminating topical NPs from the predicate or comment unaccented morpheme -hale is not a sign of "contrastiveness" but a
domain. The difference revealed in the contrast between (5.19) and (5.21)' simple consequence of English morphosyntax. To see this, it is sufficient
or (5.20) and (5.22) is the manifestation of an important typologic~f to replace the Latinate exhale with its quasi-synonymous Germanic
parameter, which I believe has received insufficient attention in studies of equivalent breathe out. In sentence (5.23')
language typology.
It is necessary to mention that not all unaccented constituents within (5.23'J He forgot to breathe OUT
focus domains are topic expressions. The placement of a focus accent
before the end of a focus domain may be triggered also by presupposed which is semantically close to (5.23), and which would be more or less
elements which do not function as topics. Let me mention here one· appropriate (though clearly less "''illy) in the given discourse context, no
particularly clear example, involving a left,~·ard accent "shift" \Vithin a accent "shift" occurs and no connotation of conlrastiveness arises. The
semantically complex "''ord (_see Section 5.3. I). In accounting for the presence of the focus accent on the prefix ex- in (5.23) is a case of default
accent pattern in the second sentence in {5.~J) (quoted on p. 356, note 16) accentuation, not of "contrastive accent shift." I will return to the issue
of contrastiveness in more explicit terms in Section 5.5.
(5.23) Jerry Brown also ~mok<:J pLl! twenty-five years ago. But he forgot tp
Exhale A revealing fact about the nature of default accentuation is that a
default accent may fall on certain grammatical morphemes whose
it would be counterintuitive to analyLe the postai::centual bound\ meaning (or lack thereof) does not seem !o be compatible with focus
morpheme -hale as a topic e:>.pres:.1..:,n since this nlorpheme does not status. Consider the pair of negative sentences ia (5.24) (a) and (b):
designate a discourse rel"erenL (for e:>.J.n1plc, it ...:ould not be represented
by an anaphonc pronDun l Thi:-. rnLlrphc1nc c<.1n h<.1rdly be said to be what {5,24) a I u1on't
the proposition 15 :1h11ut T11 undC"r\t;111cl !ht' rr(l!>Odtc pattern in (5.23)' b I did NOf
recall that 1n addition lo the pr• •r<l~1t1•Jn e.\pressed in the first sentence c. • 1 DID not
 Prosodic accents: iconicity, rule. defaull 255
254 Pragmatic relations: focus
focus accent assignment cannot be a matter of constituent structure, at
~!s ::n~::;:is~:~~l:~cexj~: the differenc~ between the accented and least not in all sentences. This is NOT to say that the accent on to is
and (b) iconically ,- e as) d"" of the negative morpheme in (5.24) (a) assigned randomly. It is motivated !NDIRECTLY, in the sense that any other
· · · a t11erence of em h · B
earlier, emphasis is not a matte f . p as.is. ut as I observed position would entail a different focus interpretation of the sentence.
expressiveness By th . ~ r o information structure but of What allows the accent on the \vord to to act as the locus of the focus
they both . e1r ocus structure, the two sentences are identical· marker is neither its denotatum nor its syntactic status in the sentence but
action. Thea~::~!;i:gm~~~ail~y that t_he speaker did not do a certai~ the fact that the accent pattern of (5.25) contrasts implicitly with possible
enclitic and the non-!nclitic fo~2~ is not the .difference between the alternative sentence patterns for the same proposition, each of which has
prosodic status of the verb . . f ~he negative morpheme but the its own focus construal (e.g. the pattern in There is nothing to EAT).
difference betwee th did. No 1con1c explanation is available for the The phenomenon of default accentuation allows me to emphasize one
in the two versi n e accented and the non-accented occurrence of did very important aspect of the grammar of focus which in my opinion has
::sence. of the ac~ent on in (a) is not
motivated by theo~;~ir~: did not been sufficiently emphasized in the generative literature. We have

~:.~e;:~~gr::~~: ~~~~~\~he :~c I~:tc :~':.::~7~~r~:~:n~:d~'1:.:::h:~

1
seen that the interpretation of the focus structure of sentences containing
a default accent is determined neither by format rules alone, i.e. by rules
of (5 24c) N t" h go, as demonstrated by the ill-formededness defined on some structural property of the sentence, nor by pragmatic
· o ice t at example (5 24) d'~
of default accentuation in that the.I k If ers r~m the previous examples
· f
principles alone, i.e. by weighing the relative communicative importance
ra . ac o prominence on n 'r is not due to of the various designata in a sentence. The cognitive mechanism for the
io t:;:~ ~::~~: (su~·h· as acti:eness and topic status of a referent) but interpretation of a default accent is of an entirely diITerent sort. Rather
A-default enc tt1c negative morpheme is inherently unaccentable than being interpreted within a given structure, the value of such an
. d accent can even fall on certain function words to h. h .
accent is understood against the background of alternative but unused
In ependent lexical meaning can be assi d . w ic no
dialogue in (5.25) gne · For example in the short grammatical structures expressing the same proposition. ln other words,
the interpretation of the accentual pattern is determined by the existence
(5.25) A.: Let's ~o to t~e KITCHEN and get something to EAT of one or more PROSODIC ALLOSENTENCES generated by the grammar of the
B. Theres nothing TO eat. · language (see Sections l.1 and l.4.3)_ Given this theoretical outlook,
Ladd's term "deaccentuation," \\'h1ch evokes the idea of a process
the focus
word to. 26accent i th~rep 1y. falls on the semantically empty function
In (5.;5) whereby one accentual pattern is changed into another one. is somewhat
unaccented. The funct' e e;~h1re sentence ex~ept the preposition to is misleading (though eminently convenient). Accent assignment by default
" ion o e accent on to is to mark the .. is not a process whereby a focus accent 1s assigned to a less rather than
There's. nothing to eat (in the kitchen)" as a correction
. of thee
propos1t1on

::~a~ ~~:k:r:i~i~?snu~~!:1ae~~.islts~r:e~~t~gmtaorkeatthein ~i~~:::.~

more normal position in the sentence As various researchers
thed (Schmerling, Selkirk, and others. including Ladd ) have pointed out,
new t · \\'OT to as the the very notion of "nonnal stress"" (or "nonnal accent") is misguided.
diffi es ~rf most rmportant element of the sentence. (This case is thus The assignment of a default accent to a constituent is the necessary result
talke;;:n ro rom ~at of .the. accented preposition to in (5.5) H'e started of independent gramm::itical fachirs. in particular (but not only) the
said t~ be,',~~ p1gs, w_hich in some pretheoretical sense could indeed be presence of a non-focal d1scourse-acti\'e designatum in phrase-final
e most important semantic element" in the sen ten ·e)
The prosody of (5.25) is clearly not amenable to any iconic ac~o~nt
of position
::e:t p~~cement. E~ample (5.25) is of special theor~tical interest as it
o s wtt ~eat clanty that the interpretation of an accented constituent
As evidence in support of the notion that focus meaning is often
construed against the background ~if a ~et of al!osentences I would like to
present some Gennan fact~. \\ h1ch to my knov.']edge have not been
as representing
sentence i · the . "new information'" _ or the "n'\\'
t: e Iement .. in a noticed t>efore In Genn;in. the accented ~uhJCCt !\"P of event-reporting
s mis 1ead1ng. At the ~ame time it sho\vs \vith equal clarity that
 Focus and the mental representations of referents 251
256 Pragn1a1ic rcfations. focus
the focus accent to appear in postverbal position, resulting in
sentences can either appear in clause-initial position, as in English, or,
preservation of the unmarked information-structure sequence topio--
under certain syntactic conditions, it can be inverted, as in Italian (see
focus. A syntactic reversal, or "re-inversion," of this inverted sequence
Sections 1.3 and 4.2.2). The t\vo options are illustrated in (5.26). Sentence
would be pragmatically unmotivated unless it also involved a change in
{a) parallels the English SV construction in (4.!0a) an<l sentence (b) the
focus con£lrual. To quote Anttila's famous dictum: "The mind shuns
Italian OVS construction in (4. I Oh)
purposeless variety."
{5.::!6) a. !V1ein HALS tut weh '"My THROAT hurts" These German facts show clearly that, in some cases at least, the focus
my throat does pain structure or a sentence can be correctly interpreted only within a system
b. t>.flr tut der HALS weh. "!Vty THROAT hurts"'
of contrasting allosentences. The notion of a prosodic allosentence will
me-OAT does the throal pain
prove particularly important in the analysis of the prosodic marking of
Both (5.26a) and {5.26b) have possible thetic interpretations, the word- the contrast between predicate-focus and sentence-focus sentences (see
order difference being motivated essentially by syntactic factors. \Vhat Section 5.6.2). The kind of interpretive mechanism postulated here
allo\vs the accented subject NP to appear postverbaUy in (b) is that the requires a "structuralist" (rather than "generativist") approach, i.e. an
possessor expression. \vhich \\'as "advanced" from the status of a approach in which the interpretation of a given structure is viewed as
determiner to that of a dative argument, no\v occupies the first position in being determined within a system of formal oppositions rather than by a
27
the clause; see example (4.49) and discussion. set of. rules. Su~h an approach to focus interpretation is comparable,
Now notice the follo\\'ing fact. If the positions of the preverbal topic ~utahs mutand1s, to Reinhart's (l983:Ch. 7) approach to the interpreta·
pronoun and the accented postverbal NP in (5.26b) are reversed, without tton of bound vs. free anaphoric expressions in English. Reinhart argues
however making the structure identical to (5.26a), i.e. if (5.26b) is that the phenomenon referred to as "disjoint reference" in the generative
changed into (5.26b'), which is also a \vell-formed grammatical structure: literature on anaphora cannot be accounted for on the basis or the
(5.26b') Der HALS tut mir weh. "My THROAT hurts" structural properties of a given sentence alone but on the basis of
pragmatic inferences determined by the unconscious awareness of
the resulting sentence has only an ARGUMENT-Focus interpretation, i.e. it possible alternative structures. 28 My approach differs from Reinhart's.
has only the identificationaL not the event-reporting (sentence-focus)~ ho\\' ever, in that I do not think that a speaker's choice among possible
reading. Even though (5.26b") is prosodically and lexically similar to allosentences, as well as a hearer's interpretation of this choice, is
(5.26a), the focus interpretation of the two sentences is different. , determined by pragmatic inferencing. Rather I take this choice to be
Pragmatically, (5.26b") contrasts both with (5.26a) and with (5.26b). directly determined by the speaker's and the hearer's knowledge of
It is clear that this difference in interpretation cannot be explained on · GRAMMAR. In the case of the phenomena under discussion, it is their
the basis of the syntactic, prosodic, or lexical properties of (5.26) alone. knowledge of the rules and conventions of information structure. No
No formal element in this sentence is per se incompatible \\'ith thetic recourse to conversational implicature is nect:ssary.
construal of the proposition For example the syntactically analogous :
sentence in (5.27)
Die POLJZEI ist h1nttr n11r htr ··The PUllt ~ .He after n1c" 5.4 Focus and the mental representations of referents
(S.27)
the police is aft.:r me-01\ T PA RT 5.4.1 Focus relation and aclirallon stale
does have a sentence-focus reading. ·rht.! re<.1son the thcuc interpretation is
In the preceding sections I described what I take to be the shortcomings
absent in l5.26b'") is bec<.1u;e 1t i;. rr.:etnpted by the existence of the two
of a narro\v pragn1atic vie\v of sentence accentuation, according to
alternative thetic struo.:ture~ 1n 15 21~1 (a) and (b). The point of using the
\~·hi~h prosodic prominence 1s a direct signal of ··new information," and
possessor-advancement ct>n~truct1l•n 1n (5.26b) is to al!O\\' the topical'
I argued that accent position is detennined by at least three factors:
element rnir 10 app~·.i.r 111 1n1t1,_1I r•);.ition. thereby making it possible for
 258 Pragmatic relations: focus Focus and the mental representations of referents 259

iconicity, grammatical rule, and default. In the present section I would As I noted in the original discussion of this example, it would make little
like to discuss one particular view of the pragmatic function of sentence sense to explain the prosodic difference between the unaccented forms
accentu.ation, whi~h i~ explicitly or implicitly shared by many linguists. heard and went, on the one hand, and the accented forms HI/ONG and GAY,
According to this Vle\v, there is a direct correlation between the on the other, in terms of"concept" activation, since the denotata of these
prosodic marking of "new information" and the assumed "newness" of expressions are all equally "new" in the discourse context. As I will show
the REFERENT or DENOTATUM of the accented constituent in the in the next section, these expressions do not have the necessary referential
addressee's mind. In this view, which has perhaps been most properties for their activation states to be a determining factor in the
consistently expressed in work by Wallace Chafe (e.g. 1974, 1976, prosodic structuring of the sentence.
1987), sentence prosody can be fully or partially explained in terms of Example (3.27) also illustrates the second fact (though it does so
the function of referent activation. I will call this the FOCUS-NEWNESS somewhat weakly), i.e. that an accent may fall on a constituent which is
~RR.ELATION VIEW. This view can be understood in two ways, which I not in focus. In one possible construal of (3.27), the accented subject NP
will _call the "stron~" and the "weak" version. According to the strong his lover in the last sentence is a topic expression, hence is excluded from
vcmon, any constituent that carries an accent necessarily expresses a the focus domain. (Topic construal of the subject NP is somewhat
referent or denotatum which is new to the discourse (i.e. to the unlikely in the given context, but il is nevertheless compatible with the
addressee's mind or consciousness). According to the weak version, any prosodic structure of the sentence.) In this case, the accent on lover is not
refere_nt or denotatum which is new to a discourse requires prosodic a focus accent but an activation accent. Its presence signals the
prominence of the corresponding constituent. assumption that the referent of the subject NP, though topical in the
In order to evaluate the two versions of the focus-newness correlation discourse, was not yet established as the topic of the sentence at the time
view it will be useful to recall the theoretical distinction which I drew at of utterance. The focus accent in the sentence is the one on AIDS.
the end of Chapter 3, and repeatedly thereafter, between two kinds of That the accent on the subject noun and that on the object of the
information~structure categories: those indicating the cognitive STATES of preposition in (3.27) are functionally not equivalent can be demonstrated
the mental representations of discourse referents (activation and by modifying the prosodic pattern of the sentence. If we take away the
identifiability) and those indicating pragmatically construed RELATIONS accent on AIDS, leaving only the accent on LOVER, the focus structure
between propositions and their elements (topic and focus). The two changes from predicate focus to argument focus, i.e. the sentence
categories correlate with each other, but they cannot be equated. (The necessarily evokes the presupposition that someone died of AIDS. But if
fact that they correlate with each other may explain why the difference we take a\\'ay the accent on lover, changing the sentence to His loi·er jusr
between them has often been neglected.) In the analysis of prosody, the died of AtDS, the focus structure remains the same, i.e. the sentence still
need for a distinction between the marking of cognitive states and that of has predicate focus. The only change is that in the assumed activation
pragmatic relations follows from two observations. The first is that status of the topic referent. (The sentence . .,.,·ith l\vo prosodic peaks, may
constituents whose designata have similar activation states often have in principle also be used in an event-reporting context, but this is beside
different accentual properties; the second is that constituents with the the point.) In a predicate-focus structure. an accent on the pred1cate is
same accentual properties may have diITerent pragmatic relations to the criteria!, but an accent on the sub1ect 1s not (see Section 5.6.1)
proposition, i.e. they may be either topical or focal. More clearly perhaps than in (3.27) the difTerence hetv.'een focus accent
That the presence or absence of an accent is often not due to a and activation accent appears in our paradigm example 13.31), whose
difference in the activation state of a designatum \vas first sho\vn in focus articulation is coded not only rrosodicall~, but morphosyntacti-
example (3.27), which I repeat here for convenience· cally:

(3.27) I heard something TERRIEILE last night Rememt.er MARI-., the guy we (3.3!) a IOP.'\t;o MOI Je P.\Yl

went HlKrNG with, who's G.O..Y') His LOVER juq died of A!DS. b Pago 10 ('"es( ~f<ll qui raye
260 Pragmatic relations: focus Focus and 1he menial represenlalions of referents 261
In the (a) examples, the pronouns 10 and MOJ are syntactically marked as (himself), or to its deictic status in the text-external world (you). 30 In (c),
topic expressions. The accents on the pronouns cannot therefore be focus the referent of the NP his brother may in principle be in any activation
accents. This is confirmed by the fact that the pronouns could be state, since accented definite noun phrases are unmarked for the active--
deaccented without a concomitant change in rocus structure (lo PAGO - inactive distinction (see Section 3.3.1). As for (e). the activation criterion
Moi je PAYE). In ract, these pronouns are semantically and syntactically simply does not apply, since no activation state can be assigned to the
speaking omissible altogether, making focus status or their referents denotatum of noone.
logically impossible (see Section 5.2.1, example (5.10) and discussion). In The referents designated by the direct object constituents in (S.28) thus
(3.31b), on the other hand, the pronouns are syntactically marked as range from active to brand-new in the discourse. Yet they are all
being in rocus, by appearing in postverbal position. The accents on the accented. The only pragmatic property which all examples have in
pronouns therefore are focus accents. These pronouns could not be common is that the designatwn of the direct object stands in a FOCUS
deaccented (let alone omitted) without causing the sentence to be RELATION to the proposition. This focus relation is exactly the same in (b),
prosodically ill-formed. Since every sentence must have a rocus to be where the referent is entirely active, as in (d), where it is entirely new, or
informative, every sentence must have at least one accent, and this accent in (e), where the constituent has no referent at all. Moreover it is
is necessarily a rocus accent. tThe last statement does not necessarily expressed by the same formal means, i.e. prosodic prominence on the last
apply to sentences with non-prosodic focus marking; see the discussion of constituent in the sentence. Since all accented constituents in (5.28) are
WH-questions below.) 29 focus expressions, and since only some of them have inactive referents, we
Let us now return to the focus-newness correlation view. It is clear that may conclude that, in these examples at least, focus prosody overrides
any claim concerning a correlation between focus and the cognitive state activation prosody.
"inactive" can be made only for focus constituents to which the The communicative purpose of the various replies in (5.28) is to
activation parameter can be applied, i.e. to referential constituents in· identify for the addressee the referent of a missing argument in a
the sense of Section 3.1. It is easy to demonstrate that even for such pragmatically presupposed proposition (hence my characterization of
constituents the strong version of the focus-newness correlation view sentences with argument focus as "ldentificational," see Sections 4.1.1
cannot be upheld. Referential constituents may carry a focus accent even and 5.2.3). And the function of the focus accent is to mark a particular
if their referents can in no way be said to be "new." Consider the constituent as the one designating that referent. Whether the referent of
sentences in (5.28) (a variant of (4.3)): the missing argument is "new" or "old" in the discourse is irrelevant
from the point of view of the prosodic form of the sentence. Not
(5.28) Q: Who did Felix praise?
knowing which referent fits a given argument position in an open
A: a. He praised HJMSELr
b. He praised You.
proposition is an entirely difTerent mental state from not being familiar
c He praised his IJROTHER with the referent or fro1n not thinking of it. In asking the question in
d. He praised a woman you don't KNOW (5.28), the speaker may v.·eU be familiar with the referent, as in (c); she
e He praised NOONL. may even be thinking of it, as in (a). \Vhat counts is that she does not
know that this referent corresponds to the ntissing argument in the
The foci of the ans\vers in (5.28) are argument foci, 1.e. they provide the proposition. What gives a focus constituent its flavor as a "new"
referent inquired about with the \\·ord who in the preceding question. element is not the status of its deootatum in the discourse but its
(The answers V.'ouid be \\Cl!-formcd if the subject and the verb were relation to the asserted proposition at the time of utterance. Focus and
deleted.) Notice that only in 1d~ 1;. the referent of the focus phrase inactiveness are independent 1nformation-structure parameters and their
necessarily nev.· to the Jio.cl•ur~..:: 1t 1s n1;.irked as unidentifiable, hen~e grammallcal n1anifestations must be carefully distinguished.
necessarily inactive. \'td th...: 1n.Jer1n11e art1ck In (a) and (b) the referent is I conclude therefore that the strong version of the focus-newness
active, due either tc• ih an ..qih~•r1c ;.t.itus in the text-internal \vorld correlation view cannot be upheld. Let us nov.· turn to the weak version,
 Focus and the rnental representations of referents 263
262 Pragmatic relations: focus

according to which a referent which is new to a discourse necessarily inactiveness marking and focus marking must be functionally (but not
involves focus status of the corresponding constituent. I believe that this necessarily phonetically) distinguished I depart on the one hand from
view is essentially correct, although it may not contribute much to our analyses like Chafe's (1976, 1987), in which prosodic prominence is
understanding of the grammar of focus. To see that there is a necessary explained solely in terms of the marking of activation states and in which
correlation between the newness of a discourse referent and focus no use is made of the notions of focus and presupposition; and on the
prominence of the expression coding it we must go back to the Topic other hand from analyses like Jackendoffs (1972) or Selkirk's (1984), in
Acceptability Scale in example (4.34) which predicts that referents which which pitch accent is analyzed solely in terms of focus marking. Notice
are low on the scale, in particular brand·new referents, are cognitively that both the "activation approach" and the "focus approach" to pitch
speaking "poor" topics. Such referents will therefore normally be coded prominence are based on the idea that sentence accentuation is a means
in non-topical form, i.e. as focus constituents. From a certain degree of of expressing "new information." One might therefore be tempted to
inactiveness on, referents are more acceptably coded as focus expressions identify the two approaches with each other by redefining the notion of
than as topic expressions. It is this correlation between inactiveness (or "new information" in such a way that all sentence accents can be
unidentifiability) and focus status that was shown to motivate the use of subsumed under this notion. One might say, for example, that activating
topic-promoting constructions of the presentational type (cf. Section a referent in an addressee's mind is an act of information of essentially
4.4.4.l). In this sense, a certain relationship of complementarity holds the same kind as conveying new propositional knowledge to an addressee
between the activation states of topic and of focus referents. But this and that all sentence accents are therefore expressions of the same
32
complementarity is only partial: while a topic constituent must have a function: the marking of nev.· information.
referent, and while this referent must be identifiable and have a certain Besides being incompatible with the notion of information I argued for
degree of pragmatic salience in the discourse, a focus constituent is in in Chapter 2, such an attempt at collapsing the inactiveness·marking
principle free with respect to the question of identifiability and activation. function and the focus-marking function of sentence prosody \vould fail
The theoretical observation that pragmatically inaccessible discourse to account for the fundamentally different effects which the two kinds of I
,I
referents are most likely to be coded as focal constituents is strongly accent have on the pragmatic interpretation of a sentence (cf. the
conflTDled by statistical observations concerning the distribution of topic discussion of the different status of the pronouns io and n1oi in (3.31)
and focus constituents in texts. For example the text counts from spoken above). Moreover it would neglect the fundamental markedness
French presented in Lambrecht I 986b (Chapter 6) reveal that subjects asymmetry betv.•een accented and non-accented constituents which I
overwhelmingly tend to be pronouns while objects overwhelmingly first pointed out in Section 3.3 L1naccented referential constituents,
appear as lexical noun phrases. Given the necessary correlation between whether pronouns or lexical noun phrases, necessarily have active
pronominal coding and activeness on the one hand and bet\veen referents, i.e. they are formally marked for the pragmatic feature ··active
inactiveness and lexical coding on the other (Chapter 3), and given the referent." Accented constituents. on the other hand, are unmarked with
correlation between subject and topic on the one hand and object and respect to activation, i.e. they may ha\·e either active or inactive referents
focus on the other (Chapter 4), \Ve can conclude that there must be a It is therefore in pnnciple imposs1hle to tel! fron1 the accent on a
strong discourse tendency for referential focus constituents to have constituent alone whether its referent 1s ··nev.-·" or "old." But it is equally
"new" referents. 31 And this tendency may have important consequences impossible to determine on the has1s of an accent alone. i.e. without :\
for the syntactic structuring of sentences. Nevertheless. there is no
necessary correlation between focus and the activation states of referents.
considering other grammatical aspects of the sentence, \•>hether the
accented constituent indicates a focus relation or an inactive referent. I,
Neither the focus paran1eter alone n0r the act1vat1on parameter alone are
An important conclusion to dra\v from the discussion in this section is
sufficient to account for the role of prosody in information structure.
that a point of prosodic prominence is neither necessarily an indicator of
\Vhat the t\\'O parameters have in i.:01nmon is that they involve the
a focus relation nor necessarily an indicator of inactivcness of a referent.
marking ofpr:::igrnatically cnnstrued portions ofproposttions via prosodic
It may be one or the other, or b0th at the ~ame time. \Vith the claim that
264 Pragn1alic relations.- focus
Focus and the n1entai representations of referents 265

structure. I will return to this issue in Section 5.7, where I will suggest an The different behavior of nouns and verbs with respect to focus
interpretation of the two accent functions which is neutral with respect to prosody can be observed also in anaphoric contexts. Unlike nominaJ
the focus-actjvation distinction. arguments, predicates are not necessarily unaccented on second mention.
Consider the contrast between (5.30) and (5.31):
(5.30) a. He promised to go SHOPPING but he forgot to GO.
5.4.:! Predicates ~·ersus argurnents
b. #He promised lo go SHOPPING but he FORGOT to go.
So far, the discussion of the relationship betv.'een focus and actjvation
(5.31) a. He promised to buy FOOD but he forgot to GET the stuff.
states has been centered on the status of discourst: referents. In the
b. #He promised to buy FOOD but he forgot to get the sruFF.
present section, I \\'ill examine the assignment of focus accents to
expressions which are not referential in the sense of Chapter 3, in In (_5.30a) the second instance of the verb go receives the focus accent, due
particular to predicating expressions. to its final position in the focus VP forgot to go, even though it is
That the discourse status of predicators has a fundamentally different anaphorically related to the VP go shopping in the first part of the
effect on focus prosody than that of referential expressions (noun sentence. As (5.JOb) shows, it would be inappropriate to deaccent go and
phrases, pronouns, nominalized clauses, non-finite verb phrases, etc.) can to put the main accent on forgot, in spite of the fact that the latter verb
be seen in sentences in which both a verb and a noun are in focus. \Ve saw expresses the "ney.•est" denotatum in the proposition. In (5.3la), on the
a clear example of this difference in (3.27) (repeated at the beginning of other hand, the noun stuff, which is anaphorically linked to the
5.3.1), where \Ve noticed that the referential expressions something ~ntec~dent -~ood, is necessarily unaccented, even though it is not lexically
terrible, f.fark, and A.IDS \vere accented, while the predicates heard, ~ identical to Its antecedent. If accented, as in (5.31 b), its anaphoric status
remember. wenr, and died were not, even though the designata of all these Y.'Ould be canceled and sruffwould refer to something other than the food
expressions were equally "new" in the discourse. The different prosodic in question.
behavior of nouns and verbs is not restricted to English, as demonstrated The same situation obtains in the follov.·ing short dialogues:
by the parallel between the English and German sentences in (5.29):
(5.32) A: I know what JNSTl/\CT means.
B- Oh yeah? What does {instinct/it} ME.AN?
(5.29) Q:Why is ht: so upset? Oh yeah'1 #\\'hat does 11\'ST!NCT mean?
A: a. He bought a CAR from one of his NEIGHBORS and now it's not
RUNNING. (5.33) A:I know when: ANNA is
b. Er hat von einem seiner NACHBARN ein AUTO gekauft, und jetzt B: Oh ye.ah'l Where 1s (Anna/she}
LJ.uvr es nicht Oh yeah? #\Vhere is At·,-l"A?

Taking these sentences to have predicate.focus structure (they provide In (5.32), the anaphoric arguments 111sr111cr or ir in B's reply must be
information about the referent designated by the pronoun he in the unaccented upon second mention. The predicate rnean on the other hand
question), i.e. taking the focus domain in both clauses to be the verb must receive the accent upon second mention, even though its denotatum
phrase (exempting from it the topical expression his/seiner), it is clear that is as clearly a<.:ti\'ated as that of the argun1ent expression. Accenting the
the designata of the verbs and the nouns are equally ne\v to the discourse. argument insr111ct. as 1n !he second version of the reply, y.•ould have the
But while in the second clause the \'crb (runningiliiu_(I) carries the main effect of canceling the anaphoric link between the tY.'O occurrences of the
accent in the first clau;.c onh the n<..>uns are accented. The difference is argutnent. leading tn this case to unacceptability since there is no other
partic~larly striking in the c,~..,c p( Lhc Gcmian past participle gekauft. ~. referent in the universe of discourse that could be designated with the
This verb form is un..1c•:cntcd 1_'\C!l thc1ugh it occupies clause.final noun in~tinL·t (the n0un designates the cL.J.ss, not an individual). Example
position, \Vhich in Gern1,1n . ...i'i 1n l:ngllsh ond French. is the unmarked (5.33) is sun1lar, except that here the order of the predicate and the
~
position for the predtcal<.>lo..icus "
,1c~·cnt --
argun1ent is re\'crscd 1n B"s reply, ~hO\\·ing that it is not the position oftbe
 Focus and the rnentaf representations of referents 267
266 Pragmatic relations: focus
Schmerling, who makes this observation in the context of her criticism of
constituent in the sentence but its function in the predicate-argument Chomsky & Halie's formulation of the N~clear Stress Rule, does not
structure that detennines where the accent will fall. make use of the notions of focus and activ~uon. To account for sentences
Such examples show that from the point of view or the expression of such as TRUMAN DfED or JOHN SURl'H'ED, wh~ch ~re_ co~nterexamples ~.::;:
information-structure distinctions the pragmatic status of verbs is in Princi le II, she postulates a pragmattc d1sunct1on between . .
some sense less important than that of nouns. It is true, as Bolinger p ••and .. topic--comment cases" and she argues that her Pnnc1ple
(1972) has pointed out, that verbs sometimes require accentuation, sen t ences · whose
II does not hold for topic-<omment sentences, i.e. sentences
depending on their relative "semantic weight." Bolinger discusses subject NPs are topics. Schmerling's distinction bel:'~en news ~~ten~s
sentence pairs such as I have a POINT to make vs. I have a POINT to and topic--comment sentences is clearly related in s~1nt to the d1st1nctlon
EMPHASIZE. In this pair, the verb make does not require an accent in order
I drew in Section 4.2.2 between thetic and categoncal sentences. .
to be in focus but the verb emphasize does. If the latter verb were Schmerling's analysis is taken up by Selkirk. who_ makes the foll.~w1ng
unaccented, it would necessarily be interpreted as being in the . (the term "focus" in the quote is equivalent to my focus
o bserva t ion
presupposition. Whatever the reason for this difference is, it does not
accent" and "focused" to my "accented"):
affect the point at hand: while Jack of prosodic prominence on a
referential expression is necessarily an indication of activeness or at least One important observation to be made about the interpretation of focus
is that lack of focus is not unifonnly interpreted. A nonfocused ~ ~s
high accessibility of the referent, lack of prominence on a predicating necessarily interpreted as old i_nfo_rmation, but a nonfocused ver 1~
expression does not necessarily have a similar implication. Phrased not \Ve believe the generahzal!on to be that only t~e foc~s o
differently, while unaccented referential expressions are marked for the con~;ituents that are argumenrs 1s relevant. to the as peel of 1ntonat1onal
feature "active referent," unaccented predicating expressions are meaning where the discourse.relevant dist1nct1on between old and new
unmarked with respect to the activation states of their denotata. infonnation is represented (1984·213)
Now since ACCENTED expressions, whether referential or not, are always Selkirk does not define "argument" but she contrasts argu~_ents with
unmarked for the active-inactive contrast, we may conclude that the "modifiers, quantifiers. and others," \\·hich (somev.·hat s~rpns1ngly) she
activation parameter is irrelevant, or at least relatively unimportant, for calls "adjuncts." She does not discuss the status of locative or.temporal
the prosodic coding of verbs. A verb with a "new" denotatum must adjunct phrases containing le:-.ica\ NPs. \Vhose focus ~ropert1es a~e 1n
receive an accent only if it is located in a syntactic position which attracts · ·1 ar to th0se
many respects s1m1 . of ·1rrruments
• c and .. \\h1ch she
.
\.\Ou\d
an accent for independent reasons, for example because no argument presumablv subsume under the category "argun1ent. Selk_irk s general- I
expression is present or because an occurring argument cannot be tion acc.ording to \.\'hich only the bcha\'tor of arguments is relevant for
accented due to its role as a topic with an active referent. Thus prosodic iza
the interpretation of focu;; ~tructure 1~ called !
. b. h
er
th "Focus
e
:I
I
prominence on predicating expressions is generally assigned by default
rather than by iconic motivation. The accent on a predicator indicates a
Interpretation Principle ..
The differenl prosodic beha\'1r1r ,1f noun~ and vcrh~ or arguincnts and
I
I
broader focal designatum in a proposition, not merely the focal status of predicates noticed bv Schmcrltng. and Selkirk (,ind other~) finds a .~atural
the predicator itself. lanation \Vithi~ my theory of. inf1.1rmal10.n ~tructure. \\"tth its
exp
· ' ertv. The
The fact that nouns and verbs behave differently \vith repect to distinction bet\veen pragmatic relation and pr.1g:mat1c prop - .
sentence accentuation was noticed early on by Chafe (1974) and r-osodic difference bet\';een nouns ,1nd verh~ can b.e seen a~ a
Schmerling (1976). Schmerling tried to account for this fact \Vith the p . f th ·nherent difference in the \\·ay in v.·h1ch d1scourse-
consequencc o e 1 Wh'J• he
following statement (her "Principle Tl"): referential vs. non-referentJal e\press1ons are prl1ccssed. -t~ l
referent1.il exprcss1c1ns involves the task of _L:reatJng. ,:
The verb receives lower stress than the suh.icct and Lhe direct object, if processing of I
l rnodif\'ing mental represenlat1ons, the
there is one; in other words. predicates receive lc11ver <;!re<;s th3n their identifying. remernhcring. an l . - f
arguments, irrespective of their linear p••~111c>n in surface structure. C\prc~" 1 , 1 n~ requ1rc~ at most the task o
proces:-ing of predicating
(1976'82)
!P'
--------------......
268 Pragn1atic relations.- focus
------~~~----
Focus and the mental representations of referents

topic referent. It is the denotatum of a propositi4n, not only of·a

269

remembering that a particular predicate occurred in a previously uttered

sentence (as in the cases of gapping and ellipsis mentioned in Section 3.4). predicate, that got activated by my utterance, and it is this propositional
The former involve long-term memory, the latter short-term memory denotatum which can be referred to with an (accented or unaccented)
only. In the next section, \Ve will see another important manifestation of anaphoric expression.
the distinction being drav.'n here, in the different prosodic behavior of The observations made in this section concerning the different prosodic
constituents expressing complete vs. incomplete (or open) propositions. behavior of arguments and predicates are comparable to observations
Thus it is the processing difference between arguments and predicates made in a different theoretical context by Hankamer & Sag (1984). The
that accounts for the fact that only the former require prosodic distinction drawn by these authors between n1odel·interpretive anaphora
prominence \\'hen their denotata are inactive, while the activation state (or "deep" anaphora) and ellipsis (or "surface" anaphora) is related. I
of the latter has no similar effect on sentence prosody. The fact that believe, to the distinction I am drawing between the discourse
verbs, adjectives, or prepositions do not require the mental identification interpretation of referential vs. non-referential categories. Hankamer &
of referents ex.plains e.g. why these categories are not grammatically Sag do not draw this latter distinction, nor do they operate with the
marked for DEFINITENESS across languages. except when they are notions of activation and identifiability, but it is no doubt not a
nominalized, i.e. when they function as arguments. It is also consistent coincidence that the ellipsis phenomena they mention all involve verbs
\\'ith the non-existence of PRESENTATIONAL or DETACHMENT constructions ("VP ellipsis," "sluicing," "gapping, .. and "stripping"). Hankamer & Sag
involving predicators (except, again, in nominalizations). Since predica- observe that such ellipsis phenomena are understood "by reference to the
tors do not code discourse referents \vith semantic roles in propositions, re~resentation of propositional structure of RECENT discourse" (emphasis
the Principle of the Separation of Reference and Role (Section 4.5. l) does mine). The restriction of ellipsis phenomena to anaphoric relations
not apply to them. The level at which the distribution of points of betv.'een elements in adjacent sentences or clauses is consistent with my
prosodic prominence is determined is not the level of semantics, with its observation that predicates are not stored in the form of mental
contrast between predicates and arguments, nor the level of syntax, with representations \>rhich can be activated and maintained over stretches
its contrast between noun phrases and verb phrases. Rather it is the level of discourse of indefinite length.
of information structure, at v.'hich the mental representations of entities
and states of affairs are marked for their statuses in the minds of the
5.4.3 Focus relation, acti1·ario11, and presupposition
interlocutors and for the pragmatic relations they enter into within
pragmatically structured propositions. The distinction v.·hich I dre\v in Section 5.4. I betv.·een the formal marking
The preceding remarks are not meant to in1ply that the discourse of "focus" and that of the cognitive state "inactive" allows us to shed
interpretation of predicates involves no anaphoric processing and has no light on a number of prosodic phenomena v.'h1ch have posed problems in
bearing on the structuring of sentences in discourse. For example, if I say previous treatments (see e.g. Schmerling 1976:74fl) or which have been
J'rn going HIK/t>'G ton1orrow and after my return you ask n1e How ~·as your ignored in the literature. The phenomena I have in mind all involve the
HIKE? it is the fact that I used the verb hike in my first utterance that difference betv.'een the marking of activation states and tha! of pragmatic
allows you to refer to n1y activity \vi th the phrase .rour hike. ~·1y utterance presuppositions. The observations presented here further develop the
has activated so1nething in ) Dur n1ind \\'hich n1akes the anaphoric preliminary remarks n1ade 1n Section 5.1 ::.
relation bel\\'CCO hiking anJ J11k,· r,1~~1bk But nL)lice that the thing The definition of focus as that ponion of a proposition whereby the
referred to in your repl;. 1~ n•:il ~1rnrly the denL1t;_itu1n of the verb hike. assertion differs from the presupposition entails that focus and
Rather it is a pn~suppL10.,eJ pr•_•p<:•..,1ti,1n. 1.e ··)ou hiked," v.·hich involves a presupposition exclude each other. Since a focus accent necessarily falls
predicate and an Jrburncnt rh1" 1> \1 h;. unJer the appropriate v.·ithin a focus dom.iin. it v<ould seem ni:ltural to conclude that a
circumstances you ,,:pu)J .1h,·· h,11e 1n4u1n:d ab(1ut 1ny outing \\'ith the : consutuent expre~sing a pragn1.ttically presurposed proposition cannot
question flo11 ~1 1 ,- u' 111 1'.h1':l1 1hL· .1n.1ph,_,ric rr,1rh1un u designates a be prosoJ1..:<.1!ly prl1n1incnt. This conclusion is indeed implied in much
 Focus and the n1en1al representations of referents 271
270 Pragmatic relations: focus
carry an
That a constit uent expressing a presup posed propos ition may
work o~ foc_us a~d presupposition in which the distinction betwee
n focus
and &ctivall~n IS not made (e.g. Choms ky 1970, Jacken
doff 1972). accent was first mentio ned in the discussion of the differen
"seman tic" and "pragm atic" presup positio n in Section
ce
2.3,
between
where I l
Ho~r. this con~lusion is patently false. To claim that constituentsd when they l
observed that sentent ial comple ments may be accente d even
expressing pragma ucally presupposed propositions cannot
be accente
d to be are pragma tically presup posed. Let us take anothe r look
at example I
on the ground s that the conten t of the propositions is assume \
definite noun (2.15), which I repeat here in modified fonn as (5.34):
already known to the addressee is as false as to claim that
their referen ts are
..
phrase s cannot receive an accent on the ground s that ''
ized, not (5.34) a. I didn"t realize [that you LIED to me].
assu~ed to be al~dy identifiable. As I have repeate dly emphas
b. I didn't REALIZE [that you lied to me].
nt mental
knowin g s.omethtng and not thinkin g of someth ing are differe c. I didn't REALIZE [that].
categor ies. In this
stat~, wh1c~ are expressed in different gramm atical
ence on a clause or
secu?n• I will show that the absence of pitch promin Jn the discussion of (2.15), I observed that the propos itional
conten t of
portion of a clause ~erely has the effect of markin g a propos itional ically
semant
3.3, and the comple ment of a factive verb like realize, though always
denota tum as ACTIVE, in accord ance with the analysis in Section presup posed, is not necessarily presup posed pragma tically. Let us
as pragma tically presup posed is
that the constru al of this denota tum situatio n in which the
discuss ion will proceed remem ber that (5.34a) is compa tible both with a
~erely a conseq uence of its activat ion state. The
propos ition "You lied to me" is already part of the commo
n ground
tn n:o parts. The first part deals with the accentu
al markin g of n in which this
between the speake r and the hearer and with a situatio
propos itions,
constit uents e~prcssing comple te or saturat ed presup posed
commo n ground is only being created with the utteran ce. But the form of
the second with the markin g of const1't uents ·
expressing presup posed ns. The prosod ic structu re of the
which are incomp lete or open. the sentence is the same in both situatio
propos itions st between
sentence can therefore not be explained in tenns of the contra
presup posed and non-pr esuppo sed propos itions.
to me"
In the (b) and (c) examples of(5.J4 ), the propos ition "You lied
5.4.3. l Compl ete presup posed propos itions is prosod ically and-in (c)-
either entities clearly ts pragma tically presup posed, since 1t
In Section 3.1, I observed that discou rse referents may be morpho syntact ically marked as having an active referen t. Someth ing that ,,
propos itions may be in
or ~roposi~ion~. Like entities, the referen ts of \\'e kno\v or at least
presup posed propos i- is active in our minds is necessarily someth ing that
~anous act1vat1on sta~es. Consti tuents express ing in (b) and
ts as noun phrase s can identif y. (The differe nce bet,vee n lexical coding
tions are therefo re su.bJeCt to .the same prosod ic contras pronom inal coding in (c) is dcteml !ned by discour se factors \\'hich
and.pr onouns . As with n~m.J.nal_constituents, any accent they carry may
exampl es.) In (b)
of a focus relation cannot be explained on the basis ~if these constru cted I
be either a focus accent-1nd1cat1ng the establi shment and (c), the comple ment of re£J!i:c is a topic c0nst1tuent v.:ith a discour se.
betwee~. the. refe~ent. ~f the presup posed propos ition and the larger
propos ition in which 1t LS embed ded-or an activat ion accent
-indica ting active referent of the same kind a" the unaccented
hooks and Crete in e,xamples (5. l9l and (5.201. Thc~e sentenc
topic constit uents
es are about
II
ted from
~hat ~he referent of the presup posed propos ition is being promo
the relation ship bet\~·cen the speake r and the fact that someon e lied to
se. As \Ne sa\\.' in
1nac~ve (or accessible) to active status in the discour hec
coincid e, i.e. the
Section 5.4.1, the two functions of pitch promin ence may Thus in (5.34) the status of the compl ement propos
ition as
oft he referen t of
establi shmen t of a focus relation may involve activat ion pragma tically presupposed LS forn1ally n1arked only 10 those cases v.·here
ver, as I briefly
the focus constit uent, but they cannot be equate d. tvforeo the presup posed denota tum 1s also discour se-activ e. \Ve may therefo re
ition ma~
menti~ned in ~ection 5.1.2, a pragma tically presup posed propos conclude that it is the cogn1tiYe status of the presup posed propos ition as •'
accents \vithin
be activat ed "1n layers, " so to speak, gi,·ing nse to derault ACTIVE 1n the discour se that detern11nes
the position of the focus accent in
the. sentent ial constit uents coding them. anal0gou~ to the default accents '
these sentences. not the fact that the propos ition is presupposed. The ~I

which may occur within noun phrases. >

.1
~:
 272 Pragrnatic rt•lations_· }Ocus Focus and the 111erital representations of referents 213

relevant aspects of the information structure of (5.34) (b) and (c) are Finally, (5.34a) is compatible with a dis.course situation in which the
represented in (5.34b') and (5.34c'): referent of the complement is already discourse-active (hence necessarily

l (5.34b') I n:•r[ I J ><x-1 didn'I REALIZE 1u•[ lhat you hed to me]]]
identifiable) at the time of utterance. This would be the case e.g. if (a)
were uttered in reply to the (admittedly somewhat bizarre) question
"Which of the two did you not realize, that I lied to you or that I was
I"
(5.3-k') ("-"'[I] , 0 c[ didn't RE;\LIZE »or[ that}]
cheating on you?" In this third situation, the complement proposition is
(5.34) (b) and (c) are instances of the prosodic type discussed in Section presupposed, active, and in focus. For all three situations, the

l
"
5.3.3. in v.·hich a predicate-focus do1nain includes a topical object
constituent, causing the focus accent to fall by default on the last
"ace!:ntable" syllable preceding the topical object.
Let us now take a closer look at example (5.34a), in v.·hich the
propositional content of the complement clause may or may not
information structure of (5.34a) can be represented as in (5.34a'):
(5.34a') [TOI'[ I J •oc [didn't realize that TOI'[ you J LIED to TOP( me J)]

(In the third situation, the denotatum "didn't realize" would not be in
focus, but this fact has no consequences for the prosodic structure of the
represent mutually shared kno\vledge. Let us first take a situation in sentence; see Sections 5.4.2 above and 5.6.1 below).
which the kno\vledge or the proposition expressed in the complen1ent The difTerences betv.·een the three communicative situations in which
clause in (a) is not assu1ned to be shared bet\vcen the speaker and the (5.34a) can be used thus have no influence on the form of the sentence.
addressee, i.e. in v.•hich the sentence is uttered with the intent to make the What the three readings of (5.34a) have in common, and what determines
addressee aY.'are of the fact that the speaker kno\vs that the addressee lied the prosodic structure of the sentence. is not the presuppositional or
to her. Jn this situation, the utterance of (5.34a) establishes a new shared identifiability status of the complement clause proposition, nor the
discourse referent, i.e. the referent of the proposition "You lied to me," activation status of the propositional referent, but the fact that the
which will then be added to the discourse register. After being referent of the complement clause has a FOCUS RELATION to the main
established. this discourse referent necessarily has the status "identifh proposition in all three situations, 1.e. that the role of this referent as an
able." In being established, the referent is necessarily also being activated argument in the proposition is not predictable or recoverable at the time
" in the hearer's mind, hence the necessary presence or an accent in (5.34a).', of the utterance. The information structure, and the prosodic form, of
Notice that this activation accent is at the same time the focus accent for (5.34a') is identical, muta1is mutandis, to that of (5.35), where the
the entire sentence. It marks the higher VP didn"t reafi::e that you lied to'· complement of the matrix predicate is a noun phrase:
nie as the focus domain of a predicate-focus sentence. In contrast to (b).
(5.35) [ r<>r [ I J '"' (didn't realize the DANG~R J]
and (c), the complement of realize in (a) has a Focus RELATION to the
matrix proposition. Both in (5.34a') and in (5.35) the position of the ac.:ent is determined by
Next let us assume a situation in \vhich the content of the complement, the General Phrasal Accent Principle, i.e. the focus accent falls on the last
clause in (5.J4a) is already pragn1attcally presupposed, i.e. in \~·hich the accentab!e syllable of the focus don1ain, •~'hich is tbe predicate phrase. In
fact that the speaker v.'as lied t1.1 by the ;:iddressce is shared knov.•ledg~ (5.34a) the accented syllable is not fin.JI 1n the focus domain because the
betv.•een the two. In this situation the accent on lied ;:igain has the clause-final argun1ent expression n1e has an active referent with a topic
function of establishing a flicus relc1Liun bct\veen the complement clause relation to the proposition, hence it cannot recen·e an accent. The same
and the rest of the propu~iti\1n, n1.-1rl-;1ng lhe higher predicate as the focus situation \vou!d hold 10 (5.35) if v.·e added an acllve topic referent to the
domain of the sentence anJ the L"1)n1pkmi.:nt :J~ heing in focus. At the. focus NP, as e.g. in I di.Jn ·1 rt:uli:::e the DASGER for you.
same time, the accent pr(11T1l1te~ th<: ,i!rc.1d:. identifiable referent of thf: 1' The necessity to drav.' a distinction betv.·een the pre:.uppositional status

complement clause frl1n1 1n,11."U\ L !•-' Jcti\ <: ~t:ni.: in the n1ind of the of a clausal di:nota!un1 and the prag1nat1c re!J.tion bet1,1,-·een this
addressee. The con1ph::n1..:n1 Ll.1u-,L" l' then ,1n C\<1n1ple of a focus denotatu1n and the rest of the proposition is particularly obvious in the
constituent wHh :J rrJgn1:111l·.1ih 1·q·c.-.u1•p·1~i.:d prup\1~1t1l 1 nal dcnotatum,: case of grdntmatic.J.! constructions V.'hose presupposit1onaJ structure
 274 Pragmatic relations: focus focus and rhe n1enral represenrations o.f referents 275

marks the content of a clause as pragmatically presupposed (unlike in is pragmatically presupposed. What distinguishes them is that in (5.36')
factive complement clauses, whose pragmatic status is unmarked). One the focus domain contains a discourse·active non·focal denotatum.
such case is the restrictive relative clause construction, which I discussed expressed in the relative clause, causing the focus accent to fall on the
in Section 2.3. Consider again example (2.11), which I repeat here for head noun. The infonnation structure of (5.36') can be paraphrased as
convenience as (5.36), provided with phrase structure brackets and _follows: "Speaking of the couple who has moved in downstairs, I've only
infonnation·structure labels: met the woman."
While focus accent and activation accent coincide in (5.34a) and (5.36),
(5.36) Tot'[ I] FOC( finally met [the woman [~[who J moved in DOWNSTAIRS]]].
the two kinds of accent are distinct in (5.37):
The relative clause in (5.36) carries an accent, even though it does not (5.37) Oh my God! My new do\vnstairs NEIGHBOR is a PIANIST!
express an assertion, i.e. even though the fact that someone moved in
downstairs from the speaker constitutes already shared knowledge. This In (5.37), the subject NP my new downstairs neighbor is a topic expression;
accent is required because the referent of the entire complex noun phrase; the sentence is to be interpreted as conveying relevant news about the
which includes the modifying relative clause, has a focus relation to the referent of this NP. The topic constituent is accented because its referent
proposition. The larger sentence in (5.36) is a topic-comment sentence is not discourse-active. (If it were active. it might appear as the
with predicate·focus structure, therefore a focus accent is required unaccented pronoun size.) The accent on neighbor, which may be
somewhere in the higher verb phrase met rhe 1roman who moved in perceived as less prominent than that on pianist, 1s not a focus accent
downstairs. The General Phrasal Accent Principle assigns this accent to but an activation accent. The relevant aspects of the information
the final constituent of the focus domain, which is the final constituent of structure of (5.37) are represented in (5.37'1: f '
the direct object NP, which happens to contain a clause coding a (5.37') ro•{ !'.1y new downstJir5 :'~ICHBOR l ''" [\Sa Pl-'.~IST J
pragmatically presupposed proposition. The accent is assigned to the
same position to which it would be assigned if the sentence were I finally
As in the examples of activattl1n accents discussed 1n 5.4. I (examples It
(3.27) and (3.31)). the accent on the topic constituent in (5.37) could be
mer my nett• downstairs N£fGHBOR, in which the object NP contains no
omitted \vithout influencing the focus structure of the sentence. On the
relative clause.
Now let us assume a discourse situation in which it is known that a
other hand, the accent on the i nc1n-referent1.:il) predicate nominal a ,,
pianist is a rocus accent. indicating that the verb phrase 1s the focus '
man and a woman have moved in downstairs from the speaker and in
domain. If this accent \vere on11ued. leaving the accent llO nt'1ghhor as the
which this fact has been recently mentioned in the discourse. In such a ()f the "entence \~'()uld
situation, the speaker might say: sole point of prominence. the rucus structure
change from predicate focu" tc1 ,1rg:un1en1 fl)cus
(5.36') I've only met the WOMAN v.·ho moved in dov.•nstairs Another grammatical context \\'h1ch all<.n\·s for re\·e,ding contrasts
bet\vecn pre~upposed pr0pc1s1!11)n" wi!h act1\·C \ s 1nact1\'e referents ts the
In contrast to (5.36), the relative clause in (5.36') is unaccented. The
construction (or family ofconstruction"l referred to as .. cx!raposillon" in
prosodic difference between the relative clause in (5.36) and that in (5.36')
the generative literature. Jn this construction. a sentential subject 1s
results from the different pragmatic relation het\veen the denotatum who
"extraposed'" from prcverbal subJCCt pos1t1on dOd appended 1n1med1at.ely
moved in dott•nsrairs and the rest of the propositinn in the (\VO situations.
tO the right or the \'Cfb phrase. the pOsl\l<.)TI L)f !ht: prC\"CTb,l! subjeCt being
While in (5.36) the presupposed relative clause rror0sition is part of the
filled by the pronoun 11 If the referent of ~uch an e.xtraposed sentential "
focus denotatum. in (5.36') it is topical in the discourse. \Vhat the t\~'O
subject is as:o.un1ed t<.i be ac\J\T ir1 the di~course C(1ntext. the extraposed
sentences have in common is that the referent or the entire con1plex noun
clause v.·ill be unaccented. JS c.g in (5 ~SJ
phrase the it·oman who ninred ifl do11·11srairs has a focus relation to the ,,
proposition, hence that this noun phr:ise nlust receive an accent. They (5 3R) A I'm afraid the pre'-tdi:nt might he lv1ng
B· \Vhal dn you niean [(, <'ll•I•"' lh3! he 1~ h1ng '
also have in common that the proros1tion expressed in the relative clause
 --------

276 Pragn1aric relations: foc10 Focus and the tnenral representations of referents 211

In the context in which speaker Buses the extraposition construction, thei' a Focus REU.TION to the main proposition. Only if the referent is both
referent of the proposition "The president is lying" is discourse-active and1 · discourse-active AND non-focal in the discourse can the constituent
bears a topic relation to the proposition. (B's utterance is a shortcut, viaJ · expressing the proposition occur without an accent. (Recall that prosodic
pragmatic accommodation, for some more explicit sequence like "Whatj prominence, being unmarked, is compatible both with active and inactive
do you mean 'The president MIGHT be lying.' He is lying. It's oav1ous."), status of a referent; see Section 3.3.1.)
Now consider the follo\.\'ing (attested) example. The speaker is The prosodic status of a constituent coding a pragmatically
woman v•ho was used to carrying a bag on her shoulder but who had.:· presupposed proposition is then detennined by two factors only: the
given the bag to a repairman that morning: ·1 activation state of irs referent and the pragmatic relation which it bears to
the matrix proposition. The prosodic status does NOT depend on the
(5.39) It's so STRANGJ:: not carrying a PURSE around.
"knowledge" presupposition of the proposition itself. Prosody is not a
{The utterance \vas accompanied by a gesture of the shoulder, indicating distinguishing factor \vithin the discourse categories of presupposition
a funny sensation.) In the situation in \vhich (5.39) was uttered, the fa ·and identifiability. These categories are expressed by lexical or
that the speaker was not carrying a purse was pragmatically presuppos morphosyntactic means alone, especially by nominalization of the
Ho\vever, this fact had not yet been activated in the conversation, i.e. the verbal constituent expressing the presupposed proposition.
speaker "''as not assuming that her addressee \Vas presently thinking ofit:· !
The fact was kno\\'n, but not yet a topic under discussion. Therefore th . 5.4.3.2 Open presupposed propositions
nominalized VP constituent evoking the pragmatic pH:supposition (not The examples of presupposed propositions discussed in the previous
carrying a purse aroand) was given prosodic prominence. Again, it is not section involved subordinate clauses functioning as arguments of higher
the presuppositional status of the denotatum but its pragmatic relation to·· predicates (complement clauses) or as noun modifiers (relative clauses).
the rest of the proposition that determines the presence or absence of an Such propositions, \vhich are semantically COMPLETE or SATURATED, must
accent As for the location of the poinr of prominence \Vi thin the clause be distinguished from presupposed propositions which are semantically
expressing the presupposed proposition, it is detennined, as in previouS" INCOMPLETE or OPEN, like those expressed by the VPs in (5.3) (MITCHELL)
examples, by the General Phrasal Accent Principle. The prosodic~ urged Nixon to uppoint Cars~rell or (5.1 la) (Aly C.~R) brokedott•n. 35 In the
structure of this clause is the same as that of the corresponding asserted saturated type, the entire clausal denotatum is presupposed, hence the
34
predicate-focus sentence /'1n not carrying a Pl'RSE around. I' clause itself does not evoke a contrast betv•een a focal and a presupposed
To sum up, in order to understand the accentuation facts involvin portion. Instead, the presupposed proposition as a \\·hole functions as an
constituents \vhich express pragmatically presupposed propositions W filgument or modifier in an asserted proposition. If such a clause carries
must carefully distinguish among the different inforrnation-structu an accent it is because of the pragmatic relation it has to the matrix
categories discussed in this book. First ,...,e must distinguish the questio proposition, not because of a focus-presupposition contrast within its
of whether a proposition is pragmatically PRESUPPOSED or ASSERTED-'i.. OWn denotatum.
whether or not it represents an already t.:NO\VN or IDENTll-'IABLE discours It is the second type, the one involving incomplete or open
referent-from the question of whether the referent of a give· presupposed propositions, \l.'hich has re..:eived most allention in the
pragmatically presuprosed propL:isiliL1n is .\CTl\'E or 1r-.:ACT!VE (UNUSED generative !Jterature on focus and presupposition since Chomsky 1970.
at a particular tirne-1.e. \1,·hether cir nL:it it is <issun1ed by the speaker to bC What has r-.:or received sumcient attention is the question of the
at the forefront of the addn.:-,-;..:c\ L"'nsc1l1usncss (or at least cognitive! ACTIVATION STATL'S of such open propositions in the discourse. In

highly accessible) at the t1n1c .:•r u1tcr.1nce Only the lat1er category hasl contrast to saturated propositions, the discourse presuppositions evoked
potential eITect on the rn·,,,,,J1~· structure "'f the -.,entence Second, once w in open propositions are typically interpreted as Hl::CEr,rrLYACTlVATED in
have determined that .J t'l\·en pre~ur'r'''<.:d JentJL.llun1 is active or inacµv the discourse context (except in \VH-questions, 1,1,·hose presuppositional
in the discourse \\e !DU,[ cl,\..•.\ \i..:lh~·r th1~ r..:rerenl h.JS a TUPlC 1-ll:LAT!ON 0 strucrure is marked syntactically; see Section 5.4.4 below). The grammar
 Focus and rhe ntental representations of referents 279
278 Pragmatic relations: focus

of English does not provide for the unambiguous marking of is necessarily construed as recently activated (hence presupposed), in the
presupposed open propositions whose designata are INACTIVE in the predicate-focus sentence in (5.40'b) the denota.tum of ~he open
addressee's mind. Consider the contrasts in the following set of examples: proposition "he was urged by x," which is semantically equt~alent to
"x urged him," may or may not be active and presupposed. (5.40 .b) could
(5.40) a. urged Nixon to appoint Carswell. ( = (5.3))
MITCHELL be used to answer either "Who was he urged by?" or "Why dtd he do
b. urged NIXON to appoint CARSWELL
MITCHELL
that?" This shows that the interpretation of the predicate denotatum as
c. The one who urged NIXON to appoint CARSWELL was MITCHELL
being discourse-active is not determined by the prosodic status of the
d. It was MITCHELL who urged NIXON to appoint CARSWELL.
verbal constituent itself but by the presence of an accent ELSEWHERE in the
In the (a) sentence in (5.40), the open proposition "x urged Nixon to sentence. In (5.40'a) it is not the ABSENCE of prominence on the verb urged
appoint Carswell" is both pragmatically presupposed and active in the but the PRESENCE of a focus accent preceding the predicate phrase that
discourse. It would be odd to use this sentence in a situation in which determines the focus construal of its denotatum. Since predicate-focus
someone's having urged Nixon to appoint Carswell was not presently construal requires an accent on some portion of a predic~te phrase,
under discussion in one way or another. (Of course, it would not be absence of such prominence necessarily indicates lack of predicate focus
impossible to use it in such a situation; this would simply sho\v that the (this last point will be further developed in Section 5.6). .
speaker assumes that the addressee is able to accommodate the required \Vhile presupposed open propositions cannot b~ unamb1guous~y
activeness presupposition.) In contrast, it seems difficult to interpret the marked as inactive in English. it is nevertheless possible under certain
(b) sentence as evoking the same presupposed open proposition, but conditions to construe the denotata of accented finite verb phrases as
where the denotatum of this proposition would be in an INACTIVE state. In presupposed. Compare the (attested) French cleft sentence in (5.41) (a)
order to mark the predicate portion in (a) as both presupposed and less \Vith its German and English counterparts in (b) and (c). Example (5.4la) I
than fully active it is necessary to evoke the presupposition morpho- was uttered by a factory manager \\'ho employed workers 'vho \vere being
syntactically, in the form of a clause expressing a saturated proposition, retrained at the expense of the French government: the clause in
as e.g. in (c), where it appears as a nominalized topic argument to the left parentheses provides the necessary context·
of the matrix predicate, or in (d) where it appears as a relative clause to its
(5 411 a. (ils TRA\'AILLEl'.'l pour "()L'S) n1a1;; c"est le (Jfll'\ER"IE~H.NT qu1 PAYE
right (for a discussion of it-cleft sentences with accented relative clauses
the\' work for us hut l\ 10 the go\ernment which pays
see Prince 1978). b (di~ -'IRBEJTEN fur I.NS) ciher BF7.AHLE" tul dit REC"olERL~G
Notice that it would be inaccurate to say that in a sentence like (5.40a) they v.'ork for us hut pay-INF does the government
the discourse-active state of the predicate denotatum is marked via lack c (they WORK for l'Sl hut the (;()\'l:R'.'.tF!'>T P\Y'i
of prominence on the verb. To say this would contradict the claim I made
in Section 5.4.2, according to which prosodic prominence is not In the situation of utterance, the propos111on °'x pays for the v.'ork done
distinctive with predicates in the way it is with arguments. Only the by the \VOrkers .. is taken as pre'>upposed but not necessarily ac.tivale.d and
arguments contained WITHIN the predicate (i.e. Nixon and Cars1\'elf) are so the fact that xis the government is asserted. The pragmatic art1culat1on of
marked. To see that it is indeed not the prosodic status of the verb that the proposition is made fomia!ly explicit in French. The den_otatum of
accounts for its pragmatic construal in presupposed open propositions let the predicate NP le ,:rou1•ernen1cnl. v.'hich is the semantic subject of the
us look at the following pair· proposition. is unambiguou5ly marked as focal (\'1a syntax). and that of
the relative clause qui paye. \Vhich is the <;emant1c predicate of the
(5.40') a. (MITCHELL l urged him proposition. is unanlbiguously marked as presupposed (via syntax) and
b. He (was urged [ by MrT<HFLL l l as inactive (Yia prl)soJ 1c prominence). A. '>in1i!ar unambiguous marking
· ih~c Gemian
· 1·oun d in \·ersion 1n (bl \\'Jth 11s ••top1cal1zed (hence
In both sentences the verb is unaccented, but while in the argument-focus c ffec\ 1s J · .

presupposed) yet accented rrt'dicatc fic:ahlen ""to pay and its inverted
sentence in (5.40'a) the denotatum of the ()ren rroposit1on ··x urged him"
 280 Pragn1atic relation.I: )Ocus
Focus and Jhe mental represen talions of referent s 281
(hence focal) subject die Regieru ng "the governm ent." As for the (5.42) A: Where's my pencil?
English
sentenc e in (c), it is syntacti call)' and prosodi cally unmark ed B: a. JOHN'S go! it
for its
pragma tic articula tion. Since it can be used under the same discours b. JOHN took it.
e
circums tances as (a) and (b} the predicat e phrase pays can be interpre c. /JOHN stale it./JOHN STOLE it
ted d.• JOHN put it in his pocket/JOHN put it in his
as non-foc al, the accent indicati ng the not-yet- active status POCKET,
of its
denotat um. But this pragma tic constru al is n1erely compat ible
with the Among the replies in (5.42), the one contain ing the unaccen ted
given structur e; it is not marked by it. The same structur e can also noun
receive pocket in (d) is clearly the least accepta ble. This reply can only
(and normall y does receive) a predica te-focus interpre tation. occur in
the form of a predica te-accen ted sentence. The locative phrase
The above obsen•a tion concern ing the marking of the activati on in his
states" pocket provide s the referent inquired about with the questio n word
of the denotat a of presupp osed open proposi tions is consiste nt where,
with the' hence it must appear in the focus domain in Ihe reply. But
observa tions made in Section s 3.1 and 4.1 concern ing the how to
fact tha( account for the contras t between (a} and (b) on the one hand, and
proposi tional referent s are express ed in ARGUME NT categor (c) on
ies (see'. the other, none of which contain a lexical noun phrase? This
exampl es (3.1), (3.2), (4.2b) and discussion). Since open proposi contras t
tions of indicate s that the possibil ity of taking a predica te denotat um as
the type in (5_40a) are semanti cally incomp lete, their denotat a active is
do not not only determi ned by the presenc e or absence of a referent ial express
constitu te referents \vhich could be stored in the discours e register, ion
i.e. in' but also by the semanti cs of the predica tor.
the long-rer m men1ory of the speech particip ants. Only those activati
on The question ~Vhere 's n1y pencil? sets an expecta tion for a reply
changes which involve the 1nental represe ntations of discours e whose
referents' focus will be the indication of the place of the pencil and whose topic
are reflected 1n the formal contras ts discusse d in Chapte r 3, in particul will
ar· be the pencil. In languag es in which the relation ship between gramma
the morpho logical contras t hetween lexical and pronom inal coding. tical
Noa·' relation s and sentenc e position s is less fixed than in English a reply
argume nt categori es, such as the tensed VP in (5.40a), are subject to this
to question 'Nill tend to have the form of a topic-co mment sentenc e
differen t cognith 'e manipu lations, invofving short-te rm memory .in which
pro.. the pencil ls the initial topic NP and the locative expression the final
cessing, such as those describe d by Hankam er & Sag in their focus
abov~" expression. For exampl e the most natural German equival ent of
mention ed study. (5.42a)
would be Den hat H.~Ns, lit. "It has John," where the sentenc e-initial
What has been called the "presup position " in !he Chon1sky-.Jack topic
endoff' is an accusati ve object and the final focus NP a nomina tive subject.
traditio n is then only one kind of presupp osed proposi tion, i.e. In
an open English , an argume nt·focu s sentence: is used instead, in \vbich the
proposi tion \Vith a recently activate d denotat um. The \vord "presup location
po of the pencil is expressed by the initial subject John.J 6
sition" is used here in a rather special sense. Being 1ncomp lete,
sue The reply in (5.42b) is similar to that in (J) bur il adds some semanti
proposi tions by definitio n have no truth value, or- using the concept c
s of content to the predica te. Rather than simply stating the place
the present framew ork- they are not represe nted as discour se referent of the
s in pencil, (b) also indicate s how the pencil got from its former to its
the minds of the speech parucip ants. Since they have no indepen present
den location . This semanti c change from be to take does not require a
referent ial exis!ence outside lh>: sentences in which they occur they change
cannO in rocus structur e, presum ably because of the commo n.sense inferenc
be stored as iden!1tiable entitie~ in the discour se register. Their e
denotata: that when certain kinds of objects are not at their usuaJ place it is
can therefor e not be properly pre~upposed. i.e. they cannot be conside because
r they have been taken away. The designa tum "take" can therefor
part of the commo n ground be! wo:en the speaker and the address e be
ee. pragma !ica/ly accommodated as discourse-active, hence the verb took
The above an;.i\ysis of Lhe act1,·.:1.tion status of presupp osed can
ope remain unac-cented in the sentence. But conside r now the reply in
proposi tions. raises an 1n1port.1n1 yue~tll•iL \~·hich unfortu nately (c). Th.is
I c reply is similar to (b) in that. in addition to indicating the new location
only touch up<:in briell;. hL·r..: v. hL'll is .111 open proposi tion of
accessible the penci!, 1t also provide s son1e explana tion of \\·hy the pencil is not
enough in the d1scour-.e 1·l·r th..: (1)nst1tucnt c.\press ing at its
it to normal place. llo\\'e\·e r this time the indicall on of the transfer
uuaccen ted? c,1n"1der lhl· "11l•l'l J1dk1!,!U(:5 1n ().-l2l of the
obJect from its old io its nev,: locarjon entails a difference in
focus
 283
Focus and the n1ental representations of referents
282 Pragmatic relations: focus
. . . k' i e I take the proposit ion "Someo ne ate the
structure : unlike the denotatu m "take," the denotatu m "steal" cannot be indtv1dual ate thts coo ie. ·~I .(unless it is a rhetorical question suggesting
k d b the NP
cookie" to be uncontrovers1 .. .
taken for granted as active in the discourse. Thererore it is naturally N ) The first presuppo sition is evo e y
construe d as Focal. the answ.er oone . entence construc tion as a whole. The
construction, the second by t.het~en the ex ression or my desire ror my
While the semantic difference between take and steal is easy enough to
characterize, this difference does not in itself explain the difference in
focus structure . Is there some cut-off point beyond which the denotata or
i:
assertion expressed by (5.43h) is. d' 'd l . (Recall that in the present
addressee to tell me who t at lfl 1v1 ua .
framework assertions are not limited to declarative sent~nces; seed~:~:· s:
verbs, unless recently activated, must be coded as rocal instead of being t at my a .
2 3 ) In asking my question. I normally also assume
pragmatically taken for granted and coded without an accent? I have . . . h t he can. answer
.
my dquestion
k b the
.
unfortun ately no answer to this question. I can merely make a very knows the identity or the refe_rent. 1.e. t a s
However that assumpt ion ts not a presuppos1t1on evo e y ..
tentative suggestion as to the direction in which the answer may lie. The s but mere!y a re~1c1t~
grammatical structure of inrormat ion question
difference between "focus-a ttracting " and "non-foc us-attrac ting" pre-
conditio n on the use of quest1ons in general. One normally oes no as
dicates may have to do with the cognitive difference between so-called . get an answer. I am not
questions without assuming that one c~n J\I
"basic level" and "subord inate" categories discussed in much work by assumption. .
Co ncerned with the latter kind of
Eleanor Rosch (e.g. Rosch 1977, Mervis & Rosch 1981) and applied to which I h ave use d b e fore ' 1n asking the
To take another example,
linguistic analysis e.g. in work by George Lakoff (in particula r 1987). It
question in (5.15a)
seems possible that basic-level categories, being cognitively more easily
accessible, can be more easily taken for granted pragmatically than (5.lSa) Who's THAT~

subordin ate categories. The difference in prosodic behavior between take

I normally presuppose not only the presence of a particula r individu~l
and steal in (5.42) could then be explained by the fact that the former
would be a basic-level category while the latter would be non-basic. I
the universe of discourse <ror example in the tex_t-exte.mal world) w ~
i.n
l:
identifiable to my addres:::.ee but also- and in .this case som.ew a
must emphasize that this is no more than speculat ion at this point.
37
trivial\ -that this individual has a certain identity. The q~est.1o_n in
(5.JSa)y presuppo ses the open proposi" °.n ··That identifiable i~d'.,~'~~::
5.4.4 Focus and informat ion questions has x identity" and asserts the speakers desire to find out v. h
·de t·ty ,-c- That (5.15a) indeed evokes these presuppo sitions is shown by
In this section, I would like to discuss a particula r construc tion which I 1 n 1 ~. f \VH-que ·
_
stions in \\'hich the
-d
1
c b·rt, of the
ent111a 1t )
. . 'h . . ·er
have repeated ly alluded to before and in which the relationship between th e b 1zarrenes 5 o ·
o s a gu,1 oi
referent is not taken for granted, like U hn 's sun1cone , or fJ
formal structure and inrormat ion structure has a somewh at exceptional
'bl "II of
status. I have in mind the construc tion expressing informat ion questions, there.? _- th-,\ t e\·o\..:es the set ofposs1 e 11. ers
or WH-que stions. Since it .15 the \VH-express1on
_- _ in . th e nr"suppo
~ t:.
sed open propositidon, the
the empty argumen t position
As a general rule we can say that the use or an informat ion question is . (5 -.g) and (-'i · \Sa) v.·htch qualifies JS the focus tyamain e of
appropri ate only if the open proposit ion resulting from removal or the only const1tuen t in ..:.
.
. the question \\OT
ts . d 11·/10. WH-ques!1ons are thus a part1cu 1ar p
question expression (the \VH-expression in English) from lhe sentence is t-focus construct1on_ \Ve \\'OUld therefore expect the main
pragmat ically presuppo sed in the discourse. For ex:ampte, tr I ask the argumen f 11 the \VH-phrase. Hov.·ever, in our examples
sentence accent to a on . . - It ·s clear that
question in (5.43)
f JI th final constitue nts ~·ook1e and I /iar i

(5.43) Who ate the COOKrE? the accent a .s on ~e, -- I he a fnclls accen! hut only an activation
this sentence-final aCLt:nl 1.:anno . _ ,d , ferents of the NPs the
A h· (me the quesuon i:::. utter~ . t1l\.: re -
my question not only evokes the assumpt ion (conjured up by the
definiteness of the noun phrase) that my addressee can identify the
;~~:~;~ a~dtin t~at'. though 1denf1Gabk to t~e a~~:::e:~::,:~:i~.t ~:~eb~:
addre~sec s mind. nr. per ap
0

activated the
particula r cookie I have in mind but alsl• that she kno\vS that some
2ts4 J'ragn1a1ic rt?iations: }Ocus Focus and the n1en1ai representalions of referents 285

yet been established as objects of inquiry in the discourse. The NPs The prosodic structure of the sentences in (5.45) is identical to that of the
therefore require activation accents (or, as I will say in Section 5.7, argument-focus sentences discussed in the preceding section, in which
"topic-establishing" accents). If the referents were already-established, post-focal open propositions are marked as discourse-active and as a
active topics, the sentences would likely be of the form Who ATE them and~ result interpreted as pragmatically presupposed. That the open proposi-
Who ts Ihat, with them and that as unaccented topic expressions of the tion is discourse-active in examples like (5.45) is demonstrated by the fact
preferred type and default accents on the verbs. But in either situation, that, given the appropriate context, the WH-expression could appear all
the final NPs are TOPIC rather than FOCUS constituents, since they by itself, as in I'm going son1ewhere. - Where? or Someone ate the cookie8,
designate the referents about which information is requested. This - Who?39
analysis is corroborated by the French versions of the question in (5.l5b) In WH-questions like (5.43), (5.44), and (5.15a), the prosodic marking
and (5.l5b'), i.e. C'esl QVtfa? and Qui c'EsT fa?, which have the syntax' Of of activation thus takes precedence over the prosodic marking of focus.
topic-comment sentences, with the pronoun fa in A-TOP position. This fact constitules an exception to a general principle stated earlier in
Nevertheless, it would be misleading to say that the function of the this chapter, according to which a single accent in any sentence is
final aa.:ent in a WH-question is always to activate the referent oLa necessarily a focus accent. The ex.ception has. a natural functional
sentence-final (or clause-final) topic constituent. Indeed, in the French explanation. Since the presuppositional structure of WH-questions is
examples, which can be used under exactly the same discourse marked constructionally, i.e. by the form and position of the question
circumstances as their English counterparts, the pronouns referring to word, the accent does not need to mark the focus and can be used for its
the individual inquired about, i.e. the bound c' and the antitopic ;a, are other main function, the coding of inactiveness of a denotatum.
unaccented. What is being activated in these questions is not, or at leas For the sake of completeness, I should mention that there are
not primarily, the individual designated by the sentence-final pronou'n exceptions to what I characterized above as the normal presuppositional
but rather the entire presupposed proposition, in the case at hand the'. structure of \VH-questions. While a question like (5.43) normally
proposition "That person is someone" or "That person has a certain presupposes that someone ate a particular cookie, a question like (5.46)
identity." This accounts for the prosodic structure of a question like (5.46) Who wants a COOKJE'.'/WHO wants a coor::1e'.'
(5.44):
does not necessanly presuppose that someone in the audience wants a
(5.44) Where are you GOlNG'l cookie. The pragmatic diITerence bet\l.'een WH-questions which do and
those v.1 hich do not evoke a presupposed open proposition can be made
In (5.44), the accented constituent is clearly not a topic expression. explicit in French via the form of the answer. To answer a question like
Rather it is the last accentable constituent of the domain marked by tlie (5.43), a speaker V·•lll nonnally use a cleft construction, as in (5.43'):
accent, i.e. the synt<ictic domain evoking the presupposed proposition (5.43') Q: Qw (c'est qui) a mange le biscuit'l "Who ate the cookie?"
"You are going somewhere." It is the denotatun1 of this proposition tb'3.t A: C'est moi. "ME," "1 did."
is being activated in the discourse. .,.
Notice that in contexts in \~·hich the presupposed proposition of ari' On the other hand, the question in (5.46), in which the final NP 1s
information question has J.]ready been activated in preceding discoursC:~ indefinite, \.Vil! typically be ansv.:ered \•.:ith a simple NP, as in (5.47):
the accent \vill necessarily fJ.11 on the focus ;irgument, i.e. the Wl;l·.: (5.47) Q: Qui \'eut un biscun? '·\\'ho wants a cookie?"
expression, as in (5.45J A: f',1oi. "r-.1e," "I do."

In sentences like (5.46) and t5.47), the question is perhaps best analyzed
(5.45) a WHO ale the C<)Ok1~/'
b WHO'S lhJ.t"'
as a conventionalized shortcut for a more cumbersome sequence such as
C. \\·HERE :1rc: )'(•U ),'.L•Jng" "Does anyone \11ant a cookie, and if so, who?" I believe it is the
 Conrrastiveness 287
286 Pragmatic relations: focus
Sentences containing such accented anaphoric or deictic pronouns or
~b~ty of making such shortcuts that acounts also for the
1n Enghsh, of questions like ( 5 _48 ) occurrence, nouns have often been characterized as involving CONTRASTIVE accents
(see e.g. Halliday 1967:206, Chafe 1976, Schmerling I 976:Ch. 4). The
(5.48) Where's a piece of paper? notion "contrastive" is defined by Halliday as "contrary to some
predicted or stated alternative." Halliday calls contrastive foci involving
as uttered by someone who needs a . f
that there is one availabl . h piece o. pa~er but who is not sure accented pronouns "structurally new." Clear examples of such
questions stran e tn t e speech situation. What makes such contrastive foci are the postverbal pronouns in (3.31 ), which are used
ge (and for the present author unacceptable) is the clash by the speaker to contrast himself with the person who was trying to pay
.betw
d fieen. two mutually exclus'v ··
1 e presuppos1t1onal structures: that of the
tn e 1n1te noun phrase and that of the WH- . . . the bill in his stead.
th I t · d' question construct1on While Even though accented pronouns or accented nouns with active
d' e a ter in irectly ev?ke.s the existence of the referent in the uni;erse of referents are often contrastive in Halliday's sense, they do not have to
iscourse, the fonner 1nd1rectly questions it.
be. Consider the utterance in (5.49), for ..vhich the reader can easily
40
conjure up an appropriate context:
S.S Contrastiveness
(5.49) (Sherlock Holmes to the butler) The murderer is YOll.
::tio~ 5.4. I I argued against the idea of a necessary correlation While it is conceivable that in the detective's utterance the focal referent
een ocus and the "newness" of a referent in a discourse and I rou is interpreted as contrasting \l.'ith some other previously entertained
~otnclud~ that focus and activation must be seen as independent though alternative suspect of the committed murder ('"It wasn't the pet alligator
in eracllng, par~meters. One of the reasons for drawing the di~tinction after all; the murderer is YOt/'). the utterance would be equally
was that co~shtue?ts with "old" {active) referents often carry focus appropriate if the detective had no other suspect in mind. The latter,
~~ents. In this s~tJon, I _would like to take a closer look at the category non-contrastive, reading is perhaps even the likelier of the two. This is
ccented
1 · 1 ·constituent with an active referent ·.. On th eas1soth1s
b . f . suggested by the fact that in (5.49) the focus accent occupies the ordinary
ana ys1s ~II then suggest a revision of the notion "activation accent"
presented In Chapter 3. This in tum will allow me to present a unifi:~ final position in the focus domain.
The impression of contrastiveness in (5.49) may be largely due to the
account of_sentence accentuation, in which focus prosody and activation somewhat unusual syntactic and pragmatic configuration of this
prosody -WI~ be sho~n to be two different manifestations of a sin le sentence, anything unusual being potentially perceived as contrasting
commurucatJve function (Section 5. 7). g \vith a more usual alternative. Example (5_49) is unusual in the sense that
it involves contact between the t\vo discourse worlds (see Section 2.1).
5.5.l Contrastii•e foci The referent of the deictic pronoun you. \~·hich is taken for granted as an
element of the te:w;t-external world. plays at the same time the role of the
Amon~ the ~reviously discussed examples illustrating accented constitu- missing argument in a proposition v.'hich belongs to the text-internal
ents with active referents were (3.30) and (J.J l ): \vorld. i.e. the open proposition ""The murderer is X."" (Cf. the discussion
(3.30) a. Pat said SHE was called. of examples (2.5) and (2.6) in Section 2. I.) \Vhat counts for the present
b. Pat said they called HER discussion is that {5.49) need not be percei\'ed as contrastive. The
accentual pattern of the sentence can therefore not be explained in terms
(3.31) Pago 10. - C'est MOJ qui paye
of the pragmatic notion of contra.;;ti\'eness.
An example of a non-pronominal focus con~titucnt \Vith an active If the speaker in (5.491 had a specific alternat1\'e 1n n11nd. i.e. if he v.'ere
referent was found in Kuno"s sentence qu<..""ited in (3 2!\ explicitly contrasting the propos1t1on expres~ed by this sentence v.·ith
some a\ternati\'e pr0p\_1sit1on in \\'htch it \\'<:I\ as~erted that someone else
(3.21) Among John. f\.1ar;,r. and Tom. wh11 1~ the olde-;(' rr)~\ 1s the olde<;t
 - - . -·- -_·--:-;: . .

288 Prag,'1atic relations: focus

Contrastiveness 289
was the murderer, the sentence in (5.50) with
s<ruclure "'ould perhaps be more appropriate: focus domain containing a "contrastive" focus argument may even be the
entire sentence, as in the thetic structure (2.5):
(5.50) vov ar(' the murderer! (2.5) Look, here's ME.

In the same vein, the marked structure in (3.21) above (ro.~1 is the oldest)1 Halliday's notion of contrastiveness does obviously not apply to such
seems to convey n1on: ::.trongly the notion "contrary to some stated· ' sentences with broad foci since the focus domain covers here not only a
alternative" than the equally possible The oldest is TOA!, where the focus' referent (the "alternative candidate") but also a state of affairs. The verb
accent is in its unmarked final position and lhe rest of the propositioll phrase met rou and the sentence Here's /.IE do not express "predicted. or
may or may not be in focus. Hov.·ever, intuitions are not clear·cut with stated alternatives."
respect to such sentence pairs and it seems in1possible lo determine whic To aecount for those occurrences of accented pronouns or nouns with
structure is contrastive and which one is not. A similar situation obtaid; active referents which are not captured by Halliday's definition of
in the tv.ro versions in (3.30) above. Here 100, HallJday's characteriiatio4 contrast, Chafe ( J 976) develops a notion of contrastiveness that differs
of contrast does not necessarily apply. The assertion in these sentences from Halliday's (besides being more explicit) mostly in that it does not
does not imply the existence of some previously entertained candidate tO take the notion "contrary to some predicted or stated alternative" t be
•vhich the referent of the focus NP is the correct alternative. dfi" o a
'i e 1n1ng criterion. For Chafe, contrastiveness involves three factors: (i) a
That contrastiveness in Halliday's sense cannot be the only factoi background knowledge of some sort, e.g. the awareness shared by the
explaining the use of accented pronouns is shown also in the followinS.1. speaker and the hearer that someone did something (a pragmatically
Spanish example (from a conversation reported in Silva-Corvaliii pres~pposed open proposition, in the present framework); (ii) a ret of
!982: 107; prosodic. marking. added: -~

possible candiJates for the role played by the element which is bein
(5.51) Q: Quien hizo el queque, IU 0 !U mama? - A. Lo hize 1·0 contrasted; and {iii) the assertion of which of these candidates is th~
"Who baked the cake, you or your mother? - 1 did" cor.rect one. Chafe interprets contrastiveness as an exceptional feature
which cancels \Vhat he considers to be the norm.ally holding correlation
In (5.51) the assertion that the person who made the cake 1s the speaker is ·' between Che oe<::urrence of anaphoric or deictic pronouns, activeness of a
not a contradiction of some other previously stated or imagined. ref~ren_t, and low pitch (see Section 3.3). According to Chafe, discourse.
alternative. It is a neutral reply \\•hereby the speaker picks out one of_, actJve items can receive an accent only v.·hen contrastive in th.is revised
two candidates under consideration. In the inversion structure, with the sense.
focaJ subject YO fo1JoH•1ng !he verb and Che co pica/ object lo preceding it,i' . The pro_blem I see l\'ilh Chafe's definition is lhat the cognitive category
both the topic and the focus argument appear in their ll de~nes 1s ~ot reflected Jn a corresponding grammatical category. The
unmarked position:;. By its information structure, (5.51) crux 1s cond1t1on (1i), concerning the set of possible candidates for the
argument-focus sentence but a topic-comment sent1:nce in focus role. Chafe 1~·rites that "contrastive sentences typically appear
00
object is the topic and the subject tn focus ~ the surface to be indistinguishable from answers to so-caJ.led WH
For a sentence to be perceived as contrastive the proposition need nol, questions" {1976:36), the latter NOT being contrastive for him. He then
have argument-focus structure This 1~ shoY.'n JO exi.11nple (5.52): observes that his model of a contrastive sentence, RONAW made the
luunburgers, does in fact not need to be contrastive but can be used also
(5.52) My life was MEANl~<~Lf5S, un1il 1 n1e1 .., ,_,1
as an ansv.'er to the question IVho made the HA.t18URGERS?, in which no
limited ser of candidates is implied. Chafe's criterion for distinguishing
In the second clause of thi~ e.\Jn1pk. the focus Jornain is the VP n1c1 you,
the two cases is that in the contrastive reading "the speaker assumes that
rather than the NP you alone 1unle\~ 1t 1\ere kno\\'n from the context tba '
a li~ired number of candidates a.re available in the addressee's mind"
the speakt:r h.iJ meL vanuu~ re1•rk hefOJre n1i.:et1ng the nght one). The~
(p. 34), \\'h1le in the non-contrastive reading such an assumption is not
'1';
<,I
I
 Contrastiveness 291
290 Pragmatic relations: focus

present. However, sentences uttered with such an assumption on the unmarked focus·structure type, instances of accented pronouns, being
~·
speaker's part are formally indistinguishable from sentences in which this departures from the norm, are naturally interpreted as special
' assumption is not made. While in the Spanish example Lo hize YO and in communicative signals.
the Italian example Pago 10 the number of candidates is indeed limited in To conclude, contrastiveness, unlike focus, is not a category of
the speech situation (there are exactly two), the same statements would grammar but the result of the general cognitive processes referred to as
also be appropriate if no limited number of alternatives were suggested. "conversational implicatures." In what follows, the term "contrastive"
In fact, the sentence Las hize YO "I did," "I made them" (with a plural (in such expressions as "contrastive focus" or "contrastive topic") is to be
feminine object pronoun) could serve as an ans,ver to the above-quoted understood in this general, non-grammatical sense. My conclusion
question Who made the HAJ.fBURGERS? in its NON-CONTRASTIVE reading. concerning the nature of contrastiveness is related in spirit to a general
According to Chafe's definition, the sentence Lo (las) hize ro would then argument made by Horn (1981). Horn argues that the so-called
be either contrastive or non-contrastive, depending on the situation. But "exhaustiveness" condition on it-clefts, \\'hich has been claimed by
the difference is not grammatically marked: in either case the pronoun yo other linguists to be an entailment or a conventional implicature, is in
is accented. 41 fact a generalized conversational implicature which naturally arises with
To sum up, the presence of an accent on a constituent with an active all "focusing constructions" (read: "argument-focus constructions") in
referent cannot be explained in terms of the notion of contrastiveness the absence of a contextual trigger or block
(whether in Chafe's or in Halliday's sense) but only in terms of focus
structure. Given the problems involved in the definition of the notion
"contrastive," I prefer not to think of this notion as a category of 5.5.2 Contrasti1·e topics
grammar. Instead I suggest that the impression of contrastiveness which So far I have looked only at accented pronouns (and lexical constituents
we receive when we hear such sentences arises from particular inferences with discourse-active referents\ \vhich are 1:--i Focus. As we know,
which we draw on the basis of given conversational contexts. Herein I however, the referents of accented constituents may also be TOPICS. It is
follow Bolinger (1961), who views contrastiveness as a gradient notion. useful to distinguish bet\~·een CONTRASTIVE FOCI. such as those expressed in
Bolinger writes: the accented pronouns and nouns ment1oned in the previous section. and
what I referred to in Section 4.4.4.2 as CONTRASTIVE TOPICS The
In a broad sense, every semantic peak is conlraslive. Clearly in Let's distinction between the t\\'O types \\'as first hinted at with example
have a picnic, coming as a suggestion out of the blue, there is no specific
contrast with dinner party, but there is a contrast between picnicking (l 20b):
and anything else the group might do. As the alternatives are narrov;ed
do-wn, we get closer to what we think of as a contrastive accent. (3.20b) I saw Mary and John ye~terday sHE says HFl to, bu! 11E's still
(1961:87) ANGRY at you

This gradient approach to contrastiveness has the advantage of allo,ving In this sentence. the accented pronouns in the l\\'O clauses code t\\'O active
for clear and for Jess clear instances of contrastiveness, and it accounts topic referents '.vhich are contrasted \vith one another. The function of
for our intuition that the clearest instances are those in ,,·hich a focus such contrastive topics is entirely different from that of contrasllve foci,
designatum explicitly contradicts a stated or predicted alternative, i.e. even though some pretheoretical notion of contra.s11vene_ss may a_pply t~
both. Indeed the notion of topic 1s 1ncompat1hle \\'Ith the idea ot
those which Halliday had in mind '"'hen he fonnulated his definition. 42
correction or contradiction associated \Vith contrasllve foci. Contra-
The fact that accented pronouns are especially likely to be perceived as
dicting or correcting a staten1ent entails NEliATING 11 or some pare of it
contrastive finds a natural explanation \\'ith1n the present frame\vork.
HO\Ve~·er. as "'e 5 ;,,. in Sectilln 4.3. topics are nutside the scope of
Since pronouns are most often unaccented in discourse, due to the
prevalent role they play as the preferred topic expressions in the negation. 43
 291 Pragniatic relations: focus
Contrastiveness 293
The difference bet.,.,'een contrastive topics and contrastive foci is easy to M: Non, c'est MOI qui fais la cuisine, ro1 tu peux faire autrc O:IOSB.
discern in languages in \~'hich it is expressed not only prosodically but no it is 1 who do the cooking YOU· TOP you-SUB can do other thing
also morphosyntactically. Consider again our old standby (3.31):
In (5.53') the clause·initial accented constituents are contrastive· top"ic
(3.31) a. 10 PAGO.~ MOJ je PAYE expressions while the postverbal accented constituents (whCther Pio.
b. Pago ro. - C'est MOI qui paye.
nouns or nouns) represent contrastive foci.
In (3.31 a) the accented pronouns are contrastive topic expressions; the,y In Japanese the first part of Mary's reply in (5.53) could also take the
are placed in preverbal position and are necessarily followed by a secon4 form in (5.54):
accented constituent, which indicates the focus. In (3.31b) the accented
(5.54) M: le, oryori-wa watashi·ga shimasu kara.
pronouns are contrastive focus expressions; they are placed in postverbal no cooking-TOP 1-NOM do CONJ
position and the sentences contain no other prosodic peak. "No, the COOKING, 1'll do."
The difference between the two kinds of contrastive elements is
formally expressed also in Japanese, in the contrast between what KunQ In (5.54), the initial (contrastive) topic NP oryoriMwa is followed by the
(1972) calls "contrastive wa" and "exhaustive-Listing ga," the fonneJ (contraslive) focus NP watashi·ga, resulting in a structure in which the
indicating a contrastive topic, the latter a contrastive focus in our terms: topic-comment articulation (predicate focus) and the identificational
Consider the examples in (5.53): 44 articulation (argument focus) are combined in a single proposition (see
Section 5.2.5). While the English gloss of (5.54), with the NP 1he cooking
(5.53) Roommates Hanako and Mary discussing household chores: ~
in topicalized position, may sound somewhat unnatural, the following
H:Mary-san, anata-wa osoji shite kudasai, watashi-wa oryori shimasu ,.
Mary-VOC you-TOP cleaning do please I-TOP cooking do_,! spoken French version matches perfectly the Japanese utterance:
kata. (5.54') ~1: Non, laCUISINE, c'est MOI qui la fa.is.
CONJ
no the cooking·TOP it is I who do it
"Mary, ~"Oll do the CLEANING, 1'1J do the COOt\.ING." 11.
~Lie, watash1-ga oryori shimasu kara; anata-wa hoka-no koto shite The left-detached NP la cuisine expresses the lo pie and the clefted NP moi
no 1-NOr-.1 cooking do CONJ you-TOP other thing do the (argun1ent) focus, corresponding to the ga·marked NP in Japanese_
kudasai.
please Like Japanese, French distinguishes the two types of contrastive
"No, 1'1J do the cooking; YOU do something ELSE" expression morphosyntactically. French also permits the alternative
I·: version in (5.54"}:
Hanako's utterance is a sequence of l\vo topic--commenl clauses. Tbe ·
pronouns ana1a and watashi arc contrastive topics, n1arked v.:ith the topic:·: (5.54") M Non, c'0>1 MO! qui fais 1, crn1'c

particle wa. In Mary's reply, hov.·ever, the pronoun warashi, which is::·
Example (5.54"), like (5.53') and (5.54'), c0ntains a cleft construction,
marked with the "non1inat1ve" particle ga, c:-.presses a contra~tive focus,,
marking the clefted pron0un AIOI as a focus argument. But unlike the
the open proposition ··x \viii do !he cooking" being no.,.,, pragmatically~·
previous versions, the relative clause in (5.54") carries an accent. The
presupposed. The second clause in /\fary's utterance is again a topic...:._.,
information structure of ( 5.54") 1s rather similar to that of (5.54'}, except
comment sentence. in which the prunoun a11a1a plays the role of a
that in the lattl!r the ac1.'.ented lexical noun phrase appears in initial topic
contrastive topic .'\ very \1n11L1r nh~rrho'>yntact1c distinction is made in'
·position, \~·ith a prono1ninal anaphor in the relative clause, while in the
spoken French· . ~
former it appears only as an arguml!n! of the relative clause. The accent
' j 1
(5-53') H l<)I tu f..11s !cs NEnOYAGES, ' 00 cu1s1NE 10 the relative clause in (5.54") is an activation accent,
MOI Jt' f;i1~ IJ ,_1·1-.,p·,t ' indicating that th!! state of affairs expressed in the relative clause is not
~L.ir:- .,,,, ·l(ll' · ,,u.~:1_ lJ JL' the: Lk:inir1g~ 1-l()p I-SUB do the fully d1scourse-act1ve (the latter formulation will be slightly modifed in
Section 5. 7). 4 ~
 Contrastiveness 295
294 Pragmatic relations: focus
"unaccen table" in the discourse for pragmatic reasons (see Section 5.3.3).
A ~cre~t type of two-accent sentence is illustrated in the following
text. 10 which a contrastiv e interpreta tion arises INSIDE a nominalized Since the accent does not fall in the final position assigned by the General
to~icNP. The example is taken from an article in the Washington Post, in Phrasal Accent Principle, and since lack of prominen ce on an argument is
a marked prosodic feature (see Section 3.3. l), the accentual pattern
which the author talks about his experience as a student of French in an 46
English high school: within the subject constituen t is perceived as contrastiv e.
The distinction between contrastiv e topics and contrastiv e foci has
(S.SS) Our French teacher, a crusty character named Bertram Bradstock made often been neglected in discussions of contrastiv eness centered on
clear that SPEAKING French was an unnecessary LUXURY: foreigne;s were English. As a case in point we may mention Chafe's (1976:49) analysis
expected lo speak ENGLISH.
of the English topicaliza tion constructi on. Chafe argues that in a sentence
(In the original only the word speaking is highlighted.) In (5.55), the like (5.56)
com?leme nt clause speaking French »'as an unnecessary luxury has Chafe's
(5-56) The PLAY John saw YESTERDAY. ( = (13))
pred1cate--focus. articulatio n, the finite verb phrase expressing a comment
abou~ the topic "speaking French." (The entire compleme nt clause "the so-called topic is simply a focus of contrast that has for some reason
functions as a focal argument within the larger VP made clear that been placed in an unusual position at the beginning of the sentence. "
speaking French was an unnecessary luxury, wh.ich expresses a comment While the noun phrase the play may be intuitively felt to be contrastiv e, it
a~u~ the Fre~ch teacher; we can ignore that for the point at hand.) cannot be contrastiv e in the sense intended by Chafe. Among his three
~I~~ the subject constituen t speaking French, the denotatum "speak- definition al criteria- a backgrou nd knov,1\edge (a pragmatic ally presup-
mg ts naturally interprete d in (5.55) as contrastin g with another posed open propositio n), a set of possible candidate s, and the assertion of
d~notatum, i.e. "writing" or "reading. " However, as in the previously which candidate is the correct one- only the second applies. (Recall that
~1scussed cases of contrastiv eness, this interpreta tion is due to an this is precisely the criterion which makes his definition unoperati onal.)
inferen.ce from the context, perhaps aided by the reader's own experience In the topicaliza tion construct ion illustrated in (5.56) no backgrou nd
of foreign-la nguage learning; it is not directly determine d by the prosodic knowledge is taken for granted. i.e. the open propositio n minus the
structure of the utterance. In the constituen t speaking French the accent
topicalize d argument is not pragmatic ally presuppos ed. The focus
falls on the participle by defauh, due to the fact that the referent
domain is the predicate phrase minus the topicalized constituen t, the
"French" e~pressed in the object NP is an already activated topic.
latter being positional ly marked as being outside the focus. The
The . topic-foc us articulati on of the sentence in (5.55) is made
topicalize d NP can therefore not he said to provide a "correct
syntactica lly e;{plicit in the following (admitted ly clumsy) paraphras e:
candidate ," i.e. the missing argument in a presuppos ed open propositio n.
(5.55') (Our teacher made clear that) ro~ [ in studying FRENCH ] mr [SPEAKING it] Chafe's above-qu oted characteri zation (\vith the proviso concerning
l'OC( was an unnecessary LUXURY]. the second criterion) applies only to the syntactically similar but
prosodica lly different "focus-mo vement .. constructi on (Prince 198lb),
Example (5.55') bas two topic constituen ts, one scene-sett ing adverbial, illustrated in (5.57) (an attested utterance\ , in \vhich the fronted
one argument ; both contain the referent "French." In the first constituen t constituen t indicates an argument -focus domain
this C:ferent is being promoted from inactive (or accessible) to active
~tate; in the second constituen t the referent is already active, hence coded (5.57) Fl~! SIX HL'NDRED DOU ~R~ 1~e ra1Sed ye~terd:i~

lil pronomin al form. Notice that there is no necessary contrastiv

eness in
the paraphras e in (5.55'). The same is true of ( 5.55 ): the reason the accent In (5.57). the initial object NP is in fl1cus, 1n (5.56! 1t is a (contrastive)
falls on .speaking is not because the denotatum of the verb is to be topic. In neither case, h(nVe\-er. 1s the accented in1l!J[ NP necessarily
perceived to be cnntra:;;ti\'e. a~ Prince (I 9S l bl anJ \Va rd ( 1988) have
highlighte d (although in the context such highlighting is a desirable
consequen ce) but because the .:irgument constituen t follO\\'ing it is conv1nc1ngly dcrnonqra ted
 296 Prag111atic relations: ]Ocus Alarked and unmarked focus structure 291

5.6 !\'larked and unmarked rocus structure 5.6./ Predicate focus and argwnenI focus
In Section 5.2. I I introduced the notion of "focus category" and I argued· As we have seen earlier, it is often possible to determine the focus of a
that the facts of focus prosody are best understood by analysing acce~t' proposition by asking an information question whose WH-constituent
positions as correlates of a small number of such categories-predicat9" corresponds to the presumed focus constituent in the answer. This kind
focus, argument focus, and sentence focus-rather than as points on a of question-answer test is analogous to the test used for determining 'the
continuum from the narrowest to the broadest focus type. In the presen~­ topic of a sentence by asking a question in which the presumed topic
section, I would like to make a specific proposal as to the way in whic~ referent is a matter of inquiry (see Section 4.1.1). Comrie (1981:57), in his
these focus categories are prosodically manifested in English, an"1 discussion of the role of focus in language typology, uses the following
mutatis mutandis, in other languages with prosodic focus marking. I w: question-answer pairs to illustrate certain major differences in focus
show that most accent positions are compatible \vith two focus readin&s structure (his examples ( 13) through ( 16); accent markings added:
one "broad" and one "narrow," and I \vill argue that this compatibilityi
(5.58) a. Who saw Bill? - JOHN saw Bill/him.
best analyzed in terms of the concept of MARKEDNESS. The predicate-focuS
b. Who did Bill see? - BiU/he saw JOHN.
structure will be analyzed as the UNMARKED focus structure while the~ c. What did Bill do? - Bill/he went straight HOME.
argument-focus and the sentence-focus structures will be analyzed as. d What happened? - BILL v.rent straight HOME.
'-lARKED.
Since predicate-focus sentences are unmarked for . , According to Comrie, the foci (or, in our terminology, the focus domains)
articulation, they systematically have more than one interpretation. in the replies in (5.58) are John 1n (a) and (b), went straight home in (c),
Alternative readings for given predicate domains are an automatic and Bill went straight honze in (d). Example (5.58) illustrates the well-
consequence of this unmarked status. Such readings result froni known fact that sentence accents may mark semantic domains which are
implicatures drawn on the basis of conversational contexts, not from larger than lhat of the accented constituent: "''hile in (a) and (b) the
grammatical rules of focus construal. When alternative focus readings o intended focus domain is coextensive with the accented word, in (c) and
predicate-accented sentences are to be made formally explicit, prosodic (d) the accented \\'Ord represents only part of the focus domain, which,
focus marking bas to be supplemented with, or replaced by~ according to Comrie, is the verb phrase in (c) and the entire sentence in
morphosyntactic marking, by n1eans of word-order variation or special (d). (Comrie does not mention the issue of the accented subject in {d),
which I will discuss later on.)
grammatical constructions, such as various types of cleft constructions!
dative shift, focus fronting, etc. .f There is an important difference in focus interpretation between (5.58)
The status of the predicate as the unmarked focus domain correlates (a), on the one hand, and (b), (c), and (d), on the other. In (a), where the
with the status of the topic-comn1ent structure as the unmarked accent falls on the subject, there is only one interpretation, with "John"
pragmatic articulation (see Section 4.2.1 )_ Unless special circumstances as the argun1ent focus of an identificational sentence; the information
obtain, a VP-accented sentence \.Viii be interpreted as having topic~ structure of this :.enteni.:e is the same as that of (5.l I) (Aly CAR broke
comment structure. Jn order to preclude topic-comment interpretation of' do11'11) anal1zed 111 Section 5.2.3. In sentences (bJ, (c), and (d), on the other
a sentence, the predicate domain mu:.t be prosod1cally marked via·, hand, \\'here the accent falls on the (last syllable of the) verb phrase, there
absence of prominence_ This in turn entails in most cases presence of an._ is more than one possible interpretation depending on the context
accent on the subject. For sin1plic!!y";. 5akc, I ,~·ill deal in this section provided. For exun1plc, sentence (b), which is analyzed as an argument-
mostly with the prag111.Jt1c arlH:ulal1un t•f ..l.'>'>ertcJ proposition~ focus sentence by Comrie, \\'Ould be equally uppropriate as an answer to
the context question in (c), i.e. it can abo receive a predicate-focus
Ho\Ar'ever, the principles of dl.'l'l'O\ r!.:1ccn1ent di.:;,i..:rtbed here apply to,
interpretation. (The app-ircnt 1nco1npa1ibili!y btt\\·een the question in (c)
pragmatically construcd '>t:n1:11H1~ d•"111-i111:. 1n genera!, 1vhether asserteQ
and the reply 1n (bj 1s purely sen1antic, having to do \~·1th the agentive case
or presupposed
role asSOL'iateJ with the verb Ju but not see: it is not a matter of
 Jt..farked and unmarked focus structure 299
298 Pragmatic relations: focus

information structure.) Sentence (b) could also answer the question in The above-mentio ned focus ambiguity of (b), (c), and (d) in (5.58") is
(d); the question What happened does not necessarily require a sentence- represented by the repeated focus labels on the verb phrase constituents.
focus reply _but is in principle compatible with any focus type in the The focus can be either the predicate, or the argument within the
answer (e.g. it could also be answered with (c)). Furthermore the reply in predicate. (Later on I will argue that the embedded focus label is in fact
(c) does not need to have predicate focus but could have argument focus unnecessary.) Concerning the separate representatio ns in (d) and (d'), I
e.g. if ~sed to answer the question "Where did Bill go?'" Finally th~ am not claiming that the difference between the topic-comme nt reading
::;~wer in {d), ~ven. with a seconda'!' accent on the sub}ect, is compatible and the eventive reading is formally marked in sentence (5.58d). Rather,
the question 1n (c) as well, given an appropriate context. (Such a the same prosodic structure is compatible with two focus construals. In
context could be created e.g. by adding the follow-up sentence But his sentences in which both the subject NP and the VP are accented, the
SISTER stayed at the PARTY). Sentence (d) can receive either a sentence- accent on the subject may indicate either lhat the referent is in focus or
foc~s or~ predicate-focus interpretation ; the subject may be a contrastive that it is topical but inactive in the discourse (see Sections 4.4, 5.4, etc).
topic or tt may be in focus. Let us look at some additional data. An amusing example of focus
Th~s in (b), (c), and (d) of (5.58) the different focus construals ambiguity is provided by Jackendoff ( 1972:225) with the question-ans wer
mentioned by Comrie are not uniquely determined by the prosodic pairs in (5.59) (Jackendoffs (6.51) through (6.53)):
structu~e of the various sentences. Rather they are, in part at least, (5.59) a Was The Sound Pattern or
English reviewed by the New York TIMES?
determi°:ed by the expectations created with the context questions.47 This b. No. it was reviewed by the Reader·s DIGEST
observa~on leads us to an important generalization. Since (b), (c), and (d) c No. it was made into a MOVIE.
arc amb1~ous, .or vague, but {a) is not, we can tentatively conclude that
Both ans\vers are compatible with the question. but they rely on
sentences 10 which the predicate phrase is accented permit two or more
different pr<Igmatic presuppositio ns. In (b) the presuppositio n created by
focus. readings, while sentences in which the predicate is unaccented
the question is taken to be "The SPE was revie\l.·ed by x"; in (c) it is taken
penrut only one. (For the time being, I will ignore the issue of possible
to be "Something \vas done \vith the SPE .. or simply "The SPE is the
sentence-f~cus co~strual of subject-accen ted sentences.) The availability
topic for a comment." The reply in (5.59b) is playful only because of its
of alte~~tiv~ readings for VP-accented sentences is a consequence of the
semantic content, given what v.'e know about the nature of the book in
compe~1tion be~ween the inherent iconicity of prosodic prominence,
question and the nature of the Reader's Digcsl. The ansv.'er in (5.59c) is
according to which any accented constituent can be interpreted as focal,
playful both because of its se1nantic content and because of the fact that
and .the ~n~ral Phrasal. Accent Principle, which is non-iconic (or only
the pragmatic presuppositio n chosen for the ansv.·er clashes with the most
partiaUy 1con1c), and which allows the domain signaled by an accent to
likely presuppositio n of the question. i.c that the medium for
extend ~ver pre.ceding non-accented constituents. The fact that such
alternative readings are found with predicate domains rather than promulgation is publication in book f0rm
A playful exploitation of focus ambiguity is found also in the example
arguments is of course a consequence of the status of predicators as
in (5.60), a first-grader jokl' 10ld by n1y daughter. Nl1tice that A·s first
unmarked for the activation states of their denotata (Section 5.4.2).
utterance is a \VH-quest1on. hence the propo<>it1on expressed in the
Th.e focus structures of the different replies in (5.58) are represented in
sentence minus the \VH-clcrnent t'i tJ.ken to he ,;;hared kno\~·ledge at the
(5.58 ?· Item. (d) represents the topic-comme nt rc 3 ding and (d') the
time the question 1s asked. the accentuation f<>cls 1n this utterance are
evenbve reading of the subject-accen ted sentence:
thus a matter of reactivation 11r "'second-instance focus .. (see Sections
(5.58') a. !'OC( JOHN J saw,,,.( Bill/him] 5.12 and 544)
b. '"°" [ Billjhe ] ,oc ( saw "~ [ JOHN J
c. TOI'[ Bill/he] ,.,x[ went,·,·~ [ str::iight H0\1[ (5 fl)) :\ D:id. wh: dn bird~ f1: \<11 11('
d. TOf'( BILL J ,,.... [went"" f S(r;11gh1 HC1\\F l A I g1\"e ur
d'. .uc[ BILL went straip:ht HO~.ff) A H1xau~c 11·~ !<'''!Jr 1" \\.\ti..
 300 Pragmatic relations: focus Afarked and unmarked focus structure 301

The (mildly) funny effect of the answer to the question in (5.60) is due to reading, in which flying is taken to contrast with some other kind of
the fact that the information structure of the reply is not consistent wi_th, locomotion, as suggested in the joking answer in (5.60). In this read.in&.
that of the question. (The joke is thus built on uncooperative the sentence could be used e.g. to correct the previously expressed
conversational behavior.) By uttering her question, A activates in B's 1 mistaken belief that birds migrate on foot. In the second- .. broad"-
mind the referent of the presupposed proposition "Birds fly south." Since reading the sentence could be used e.g. to contradict someone's claim that
the predicate phrase is accented it allows for two readings: one "narrow,"~ birds have stopped migrating altogether, in which case the interpretation
in which the matter of inquiry is the direction of the flight-south. of the sentence is similar to Birds DO fly south.
contrasting e.g. with north- the other "broad," in \vhich the matter of,· The second reading is perhaps easier to grasp in a different discourse
inquiry is the behavior of birds in the fall- flying south contrasting e.g. setting. Imagine a traveler in a New York airport walking up to an airline
with staying home. In the narrO\\-' reading, the principle of accent. ticket counter and asking for a flight to Dallas. The airline employee
interpretation is iconic, the focus coinciding with the smallest accented. might give the answer in (5.62):
constituent. In the broad reading, the interpretation is based on the I'm sorry, Sir, we don't soulh.
(5.62) FLY
General Phrasal Accent Principle.
In both readings, the denotatum of the directional argument south i~ 1 The point of the utterance in (5.62) is not to contrast flying with some
being activated in the question together with that of the verb. The force other manner of transport ("We don't fly south, we only use buses for
of the joke is that the answer given by A requires a pragmatic situation i~. those destinations") but to convey general infonnation about the airline.
which the direction of the birds' migration has in fact already been It is a topic-comment sentence whose predicate-focus domain contains a
activated and in which the matter of inquiry is instead the manner of- topic element with an active referent. The two readings of (5.6 lb) are thus
locomotion. The required presupposition would be properly evoked by analogous to the two readings of (5.19) John doesn't RE.AD books or of ! 11

the alternative question in (5.60'): (5.55) SPEAKJNG French was an unnecessary LUXURY. In each case, an
accented verb is followed by a topic argument inside a predicate domain.
(5.60') Why do birds FLY south?
In the case of (5.6ia). the t\vo readings are easily explained as a result
Both in (5.60) and in (5.60') the question presupposes knowledge of the of the fundamentaUy different behavior of verbs and nouns with respect
entire proposition "Birds fly south." However in (5.60') it is not thi.S. to sentence accentuation, as described in Section 5.4.2. While unaccented
entire proposition that is being activated by the question but only part of. arguments n1ust be topics, the pragmatic status of unaccented predicators
it. The sentence requires a discourse situation in which the directional, is left open. Hence the two possible interpretations_ More remarkable is
argument has been activated prior to the time the question is uttered. :J the fact that (5.6lb) also has two interpretations, even though the
Let us take a closer look at the interpretation of the l\.\'O prosodic~· accented verb is the only focal element in the predicate domain. In fact,
patterns in the questions in (5.60) and (5.60'). To simplify matters, I will~ the sentence would have t,,.,•o readings even if it Y.'ere simply Birds FLY.
use the declarative counterparts of the C\vO sentences in (5.6l): It js important lo see !hat this focus ambiguity (or rather vagueness) of
predicate-accented sentences is not simply a result of the fact that the
(5.61) a. Birds fly SOUTH. denotatum of a verb phrase is semant1caUy "n1ore complex" than that of
b. Birds FL\' south
a subject, thereby leaving room for a greater range of focus
As stated above, sentence (5.6la) h;,i:. tWL' readings. It cou!d ans\ver thi interpretations. Similarly, the non-ambiguity of subject-accented
question "What do birds do·i·· ur ··\\'here Ji:• birds tly?" But what about sentences is not a simple consequence of the fact that being the leftmost
(5.6lb)? According to the genl.'ra!izal1•_>11 rncntioncJ abo\e, \\'e also ex.peel element in the sentence the focus domain cannot include anything
this sentence to have (\\'ll 1ntcrprl'lJ!11)!l~ ~incc the :u..:ccntcd verb phrase, preceding it. Rather the difference in focus construal is inherent in the
indicates an unmarked fL1CU'.> d1.1nLJ1n Thi~ c.\pccta11on is indeed borne l\vo focus-::.tructure types. This is confirmed by languages which perm.it
out. The first reading.\\ hich '-'1n1.:\ 1,_i rnnid 111u~1rcaLlily,1::. the "narrow" e
subject-verb inversion, like Italian. Senten..;es like (5.11) Si rolla la mia
 Afarked and unmarkedf ocw structure 303
302 Pragmatic relations: focus

"My .
broke down .. or H a mang1ato As in earlier-disc ussed cases (examples (5.19) or (5.62) and others), the
MACCHINA . CAR GIOVANNI "JOHN ate"
tendency to interpret the focus or such sentences as "narrow" or
have ~o P~lcate-fo cus readings, even though the NP is not the leftmost
to some linguists, "contrastiv e" is not due to a rule of focus interpretati on but to a
COD~tituent 10 the sentence (and even though, according
generalized conversatio nal implicature . Since the presence of the
the inverted ~ is part of the verb phrase). Thus the fact that Birds FLY
unaccented topical element within the predicate-fo cus domain causes
has two readings while BIRDS fly has only one (ignoring again a possible
thetic reading _of that sentence) is a natural consequen~e of the
inherent the actual focus designatum to be narrower than the syntactic structure
would allow it to be, and since contrastive ness implies relative
mark~ness difference between the subject domain and the predicate
dom8.ln. narrowness of a semantic domain, the contrastive interpretati on tends
to be the one that comes to mind first. In Ladd's example (10), as in
~e above-stated generalization concerning the inherent ambiguity of
(5.20), the tendency towards narrow-foc us construal is reinforced by the
pred1cate-~ocus sentences contradicts a widely held belief. Consider e.g.
fact that deictic adverbs like yesterday or tomorrow are most often
the _following statement made by Ladd (1978:75) in his discussion of the
unaccented and tend to have a topic relation to the proposition .
nobon of "normal stress.·· Referring to work by Halliday ( 1967)
Chomsky (1970), and Jackendoff (1972). Ladd writeso ' Sentences in which they are accented and focal are therefore perceived as
special. And anything special is potentially perceived as contrasting with
The most important point that emerges from these works is that while a norm. Such observation s do not alter the basic fact that predicate-
most of the possible accent placements in a sentence signal a narrow accented sentences have t\vo focus readings, one of \vhich is necessarily
focus, one leaves the f?cus bro~d o~ unspecified. While focus is hardly a
well~efined c~ncept, its effect 1n dialogue provides hard data for those "broad."
unsatJsfied by intuitive definitions. Halliday's examples will illustrate. In At the risk of overstating my case, I would like to argue that the
(8) John painted the SHED yesterday the focus can be the shed, or painted situation is in fact the opposite of the one claimed by Ladd. Instead of
the shed, or th~ whol.e sentence, etc. Thus (8) could be used to reply to a saying that "\\'bile most of the possible accent placements in a sentence
ra~ge of questions hke ivhat's new, ~Vhat did John do. H"hat did John signal a narro\\• focus, one leaves the focus broad or unspecified " \Ve can
paint yesterday, etc. By contrast, other possible accent placements
say, with greater justification , that \•,rhile most of the possible aC{:ent
narrow the focus. so that for example (9) JOHN painted the shed vesterdai•
could o~ly answer the question ivho painted the shed yesterday: and (lci) placements in a sentence leave the focus broad or unspecified , one signals
John painted the shed rESTERn.~ r could only answer H'hen did John ainr a narrow focus. This narro\v focus placement is the one on the subject.
the shed. p
An interesting question. \vhich I cannot pursue here in any detail. is
\Vhy adverbs like yesterday or 1on1orrow n1ay occur unaccented in "out-
While the interpretati on of Ladd's example (9), in which the accent falls
of-the-b\ue " utterances \\'hile other adverbial phrases require prior
on the subject, seems uncontrove rsial, his example (IO) does not confinn
activation of their denotata in the discour$e in lirder to occur \Vithout
the "most important point" mentioned at the beginning of the quote. It is
an accent. In a discussion of this is.;;ue. Halliday ( 1967:207ff) suggests that
true .that (10) can have the indicated function, \vith the accented
the difference bet\veen J(lhn sall' the Pt. { t _rl"stcrday (n11 prior activation of
cons~tuent yesterday as an argument-f ocus domain. But (10) also has a
the adverbial denotatum needed) and J11f1n sc111· the Pf.~)· t"n June (prior
predicate-f ocus ~~ding. For example (10) could be used in a reply to
activation required) can be explained by saying that the de1ct1c yes1erda1
someone compla1n1ng that John doesn't take good care of the shed in his
is equivalent to a deictic pronoun. However. deictic <>Latus alone does not
backyard: What do .vou mean. John's not doing anrrhing ahour the shed.
seem sufficient to explain these accentuatio n facts. For example. the
He just painted it YESTERDAY! (In this case. the verb rain red mav receive a
temporal post position 11g1> is deictic. but a sentence like John sa11· die PL 41"
secondary acc:ent, but this accent is non-distinc tive.) On thi.s reading,
1hrcc hours ago seems to require prior activation of the te1nporal referent
sentence (10) 1s parallel to example (5.201 !'111 going tn Cri:te TO.\tORRoii·.
I must leave this issue for future research
To take another example, upon seeing a ::i. strange-k>ciking pcrs 0 n in the
The abO\'e quote from Ladd cc>nta1n~ :inc•ther often-heard statement
street I can say I saw that gur YESTERD.~ r. \•.·ithnut nece5~arilv contrastinc \vhich needs clarific,1tion .t\cc0rd1ng !<) L.idd. 1f Hall1dJy·~ ~entence John
yesterday with today or some other dJ! . ~
.)l.H rrugfflcllll rl:'lU.llV/lj. jvc:u:s Marked and unn1arkedfocus slructure 305
pa111ted the SHED vesterday 1s used to ans\ver lhe question .. What's new?'• the case of non-referential or quantified subjects (see Section 4.3) or of
its focus is '"the \Vhole sentence." The same belief is expressed in Comrie;! empty subjects in event·central thetic sentences {see Section 4.2.2).
statement that sentence (5.58d) (Bill went straight ho111e) has senten Finally, as we saw with example (5.41). VP-accented sentences may have
focus when used as a reply to the question "What happened?" Howev~r, construals in which the entire VP denotatum is in the presupposition. In
as I observed earlier. context questions do not require specific foy all cases, however, sentences with such alternative focus readings have the
structures for their replies; they merely suggest preferred readings. If · grammatical fonn of topic-comment sentences.
subjects John and Biil in the t\VO sentences above are unaccented, the. An important claim embedded in the principle in (5.63) is that
necessarily function as topics, even if these sentences are uttered in repl alternative focus construals of VP-accented sentences are not determined
to the question "\Vhat happened?'' As a result, these sentences canno by alternative focus STRUCTURES, i.e. do not result from conventional
have sentence-focus structure, t.e. lheir subjects cannot be in focus .. T form-meaning pairings. Rather they are the natural consequence of the
take another exan1ple, I can use the sentence I lust n1y 1-f'ALLET either a~ unmarked nature of the predicate domain. Such readings do therefore
reply to '"Ho\v are you doing?" or to "What happened?" In bqt not have to be accounted for viilh RULES of infonnation structure. In the
situations, I am using a sentence with topic--comn1ent articulation a11d... case of argument-focus construal of a VP-accented sentence it would
predicate-focus structure The preferred "eventive" interpretation ' . therefore be misleading to say that the pragmatic articulation of a given
merely a function of the semantic content of the proposition; it is J:\O · sentence has "changed" from predicate focus to argument focus. The
determined by the prosodic structure of the sentence. For a sentence to accent on the argument constituent does not MARK the argument as the
qua~ify as having sentence focus its subject must be marked via prosodi focus. The given accent position, v.•hich is motivated for independent
prominence. But subject accentuation is not a sufficient condition fa reasons, is merely compatible with an alternative-iconic-reading which
sentence-focus construal, as \ve have repeatedly seen before. In English. is conversationally induced. Instead of speaking of "'focus ambiguity" of
only sentences v·:hich have both accented subjects and non·accente<)., predicate domains it .is therefore more accurate to speak of "focus
predicates can, under certain semantic conditions, be said to belong to. vagueness. " 48
formal category '"sentence-focus structure" (see Section 5.6.2 belov.•). To better understand the nature of tbe interpretive principle that
On the basis of the above observations, I would like to propose th allows for alternati\'e pragmatic construals of an unmarked prosodic
following general principle of interpretation for VP-accented senten~:· structure let us compare the facts of focus construal with the
interpretation of unmarked SYNTACTIC configurations (cf. the discussions
(5.63) THE PRINCIPLE Of PREDICAfE Flli..US lt..:TERPRf.fATlON Sentences whose vef in Sections l.3.2 and l.4.2). A simple English example of an unmarked
phrases carr,: an accent ha\'e predicate-focus structure. The predicate:.
focus struc~ure is the unmarked focus structure and allows f0( and a marked syntactic structure is the contrast belv.·een the canonical SV
alternative focus readings.. Such alternative readings are contextuall, pattern and the corresponding Juxiliary-inversion pattern with respect to
determined. 'l the mood feature "declarative." A canonical SV sentence like She i.J
beaut1j'ui n1ay be 'iaid lo be unn1arked v,-lth respect to this feature, since it
In VP-accented sentences, the semantic and Lhe pragn1atic can also be used in an interrogative or exclan1ative function, given
coincide (entirely or in partL \\'hi le 1n subject-accented sentences the tw appropriate changes in intonation (see the interrogative She is beautiful?
levels of interpretation dtverge (sec Section 5.2.J c1bove). The status of th and the exclamative She is beautijU!.'). The 1n\'erted sequence Is she
predicate as the unn1c1rkeJ fL)CUS gue~ hanJ in hJ.nd \~ith the status of th beaurijU!, however, is negatively marked for the feature in question since
subject as the unmarl-.ed tori~·. 1 e .JS .t cc1nst1tuent \\'h:ch is normall)\ it cannot be u~ed as a sin1ple declarative sentence. no matter how the
topical but \\·hich a!l1)w, fur n,_,n-ll•p1l' C\>nstruc1l of its denotatum. intonutton is 1nodified. (!n~tead of saying that the canonical pattern is
Among the poss1bk· ;,ilter11.111'.L' rL·.1d1n~ . . 1·,,r \'P-.Jc-:enll·d sentences are· unn1;,irked for the re.uure ··cteclarat1\·e." w·e could also say that the
not only the earlier-1ne11t1••ned ,1r~u111~'nt·i"11l·u, r..:..1J1ng~ l1ut also readings relevant feature is "non-declarative," and th..:1.t the inversion pattern is
in \vhich the preJio.;ul..: 1~ 111 (,_.,.u.., \•.1!11 .. ul th.: ~uhJt:l\ heing a topic, as in positi\·eJy mJ.rked for this feature. J
 306 Pragmatic relations: focus Marked and unntarked focus structure 307

Notice that in the example of syntactic markedness mentioned above it 5.6.2 Sentence focus
would seem counterintuitive to characterize the canonical SV pattern as
..ambiguous" between a declarative and a non-declarative reading. A In the precechng . section . I es t a bl"shed
1 the distinction between markedd
and unmarked focus struc ' t ure and 1 observed that predicate-accente
h ~
more insightful characterization would be to say that the SV structure is
semantically vague or UNDERSPECIFIED with respect to the given sentences as the unmarked type, systematically have more t an o?~ ocus
distinction. A speaker who uses a canonical SV structure in an interpret;tion. Since predicate accentuation is a necessary condition for
. k.10 g of any other focus structure
interrogative function does not exploit a separate form-meaning Predicate-focus structure, proso d1c mar Th. .
correspondence. She merely exploits a vagueness. The principle of entails absence of prominence . on th e predicate phrase. 1s. ts
interpretation at work here is similar to that used in the interpretation of tantamount to say1n . g that in non-predicate-focus sentences . · the subject
d.
such well-known LEXICAL pairs as dog and bitch or German Katze "cat" t b ted Now since the topic-comment art1culat1on (pre icate
mus e accen · · t. ( rgument
and Kater "tomcat." The first members of these pairs are unmarked with 1
1' ocus ) contras t s n ot only v.1ith the identificational articu· a ton a· I t"
respect to the sex of the animal while the second members are positively focus) but also \vith the event-reporting or presentatl~n~ 1 ar~1c~ a ton
specified in this respect. If I use the word dog to refer to a female animal I (sentence focus), sentences \vith unaccented predicates w1\l 1n_pr1nc1pl~ ~e
am not using the word in a different SENSE as when I use it to refer to a ambiguous betv.·een the last t\\'O readings in languages hke. E~g IS •
male dog. I am merely using a lexical structure which leaves the semantic although semantic and pragmatic factors usually preclude ambiguity.
distinction unspecified in a particular way. The principle of interpretation v»h1ch accounts for argument-focus
I suggest, then, that the principle of semantic interpretation which is at construal of a sentence \\•ith an unaccented verb phrase h~s been
work in such syntactic or lexical markedness oppositions is also at work . . d in
repeate di v d 1scusse . th.is book and . needs . · (see
. no further elaboration d
in the case of alternative focus construals of predicate-accented sentences. especially. the discussion of the relat1onsh1p bet\-veen pre~uppos1t1~n, an
Consider again sentence (5.61a) Birds fly SOUTH. Whether I utter this .
act1valion. . Sec t.10 0 ,- . 4 . 31 . \\'hat remains to be explained is v.h) .the
10
sentence with the intent to correct someone's mistaken belief about the same prosodic structure which expresses argument focus on the s~b1ect
direction of the flight of birds, or with the intent to inform someone of also expresses sentence focus, contra d ic t in.g the general rule according . to
their migratory behavior, the information STRUCTURE of the sentence is which a predicate phrase \Vith a focal designatum must be. prosod1cally
the same. My sentence is vague, not ambiguous. The difference in prominent. The follovoing sections are devoted to this last issue.
interpretation does therefore not have to be accounted for with a rule.
Thus in order to represent the two readings of (5.6 la) \\'e do not need the
two structures in (5.64): 5.6.2.1 The theoretical issue
In a remarkable early essay comparing the variable pu"ition of sentence
(5.64) a. '°" [ Birds J ny SOUTH J
F"OC [ accents 1n · E ng 1·is h \\·1·th th' vanJh!e rosit1on of phr..1"al. constituents in
b. lOI' [ Birds] ny For; f SOUTH l S pan1s . Bo 1·1nger (19<')
· h . '-+ ·,\)uqr·1te"
• the n1:lJOr ;icccnt pn"1!1l)ns 1n English
\\•ith the ~entences 1n (5.66)
The representation in (5.64b) is unnecessary, since the represented
reading is included in (a). Similarly, among the t'vo representations for (5.66) \Vhv didn"! she come tl) work today"' (Bl)linger l'l:'\41
the alternative focus readings of Ladd's example (9), given in (5.65): :i Her husband is <;tck
b Her hu~h,1nd made a scene
(S.65) a. TOO'[ John J FOC[ painted TD•[ !he shed] iL<;TERD'\l" ]. c Her hu~hand !<; \0 hbme
b. ror[John J painted To"f the 5hed Jr·.- f iESTFRD.\'l) d Her hu~h~ind fell ,)ff a laJdo:r
e Her hu.;h~1nd hrokt' h1~ neL-k
Her hu~hanJ haJ .in <.JC(llknt
only (a) is needed. The interpretation represented 1n (b) ""comes for free,"
g Her hu<;hand d1eJ
given the general role of predicates in sentence prosody (Section 5.4.2) 11 Ht•r hu~band 1~ re'>r<'n~1hlo:
.:SUIS Pragmatic rela1iv11s: jocus Marked and wimarkedfocus structure 309
1. Her husband is irresponsible.
(5.69) SENTENCE FOCUS:
j. Her husband is in jail. a. Her HUSBAND is sick.
b. Her died.
Bolinger notes that, given the minimal context created by the questiq~
HUSBAND

"'prosodic stress" in the answers will most likely fall on the subject no:_" With respect to the predicate-focus sentences in (5.67) it should be noted
husband in (a), (c), (g), and (h), but on the sentence-final Y.'ords (see, · that in the mimimal context provided both the predicates and the subjects
ladder, neck, accident, irresponsible and jail, respectively) in the ·5 are likely to be accented, given the fact that the subject referent is not
remaining answers. The tv.·o patterns are contrasted in (5.66') cln ·discourse-active (see Sections 5.4.1, 5.4.2, 5.6.1). Parallel to (5.67) (a), (b),
(5.66"): (c), etc. there are also the versions in (5.67'):
(5.66') a. Her HllSBAND is sick. (5.67') a. Her HUSBAND made a SCENE.
c. Her HUSB.S..ND is to blame. b. Her HUSBAND fell off a LADDER.
g, Her HUSBAND died. c. Her HUSBAND broke h.is NECK.
h. Her HUSBAND IS responsible etc.

(5.66") b. Her husband made a scE.NE I

Notice that the phonetic difference between the two-accent sentences in
d. Her husband fell off a L"'DDER.
e. Her husband broke his NECK
(5.67') and their single-accent counterparts in (5.67) is perceptually much
less clear-cut than that betv.·een the predicate-accented sentences in (5.67)
,,'!
'

f. Her husband had an ACCIDENT

i. Her husband is lRRfSPONSlBLJ:. and (5.67') on the one hand and the single-accent structures in (5.69) on
j. Her husband 1s in JAIL the other. This perceptual difference has to do with the different
functional impact of activation marking and focus marking. In predicate-
It is clear that the subject-accented sentences in (5.66') must 1 • accented sentences, the presence or absence of a second accent on the
subdivided into two sets: (c) and (h) have argument focus, and (a) an subject does not have the same category-defining effect as the presence or
(g) sentence focus. In (c) and (h) the denotata of the predicates ~­ absence of an accent on the predicate in subject-accented sentences. For
pragmatically recoverable since the notion that someone or something· example, changing (5.67a) into (5.67'a) does not entail a change of focus
responsible for (or to blame for) the \voman's absence is implied by th structure; it merely marks the subject referent as inactive in the discourse
question. We can say that the accented subjects identify the missin (making it possible to use the sentence in a relatively context-independent
element in the proposition "She is absent for x reason." In (a) and (g), o~_ way). Hoy.•ever, adding an accent to (5.69a), i.e. changing it into Her
the other hand, the predicates are in no \Vay predictable. The t~ HUSBAND is SICK, entails changing the focus category. The latter sentence is
sentences have an event-reporting function. The various replies in (5.6 construed as a topic-comment sentence v.•ith an inactive (or contrastive)
may then be divided into the follo\ving three sets: ' topic referent, not as a sentence-focus sentence v.·1th an inactive (or
contrastive) predicate.
(5.67) PREDICATE FOCUS
a. Her husband n1ade a Sl ~NE To take another exan1ple, while the contra~t betv.·een the two versions
b Her husband fell off a LAlJIJLk in (5.70a) is one of activation only, that bctv.·eeo the t'.'.'O versions in
c. Her husband hrokc his NLl f... (5.70b) is a contrast bel\veen two focus categories·
d Her husband hJJ an -\llll>l1"<!
e Her husband i~ IRkL\l',_1~·,11111 (5.70) a Johnson DJED./JOHNSON DIED

f. Her husbJnJ 1> 111 1-1c11 b. JOJ-H>SON died.,.'JOHNSON DIED

(5.68) ARGL'M~NT 1-•_H:l ~ While both versions 1n (a) have predicate focus, the versions in (b) do not
a Her Ht,~\'•l• 1, l<' hL.11nc: both have sentence focus. This is so because in predicate-focus sentences
b Her HI. ~~.-1cr.r• " r~·;r,-•1h101c
the category-defining feature is the accent on the predicate, leaving open
 Marked and unmarked focus structure 311
310 Pragmatic relations: foc11s

the possibility of an activation accent on the subject argument. In marked The second theoretical question, which is the one I am concerned with
~ntencc-focus sentences, on the other hand, the category-defining feature
in this section, is that of the RELATIONSHIP BETWEEN INTERPRETATION AND
IS both the accent on the subject AND the absence of an accent on the
FORM in sentence-focus sentences. It can be formulated as follows: given
predicate. The contrast in (5. 70) confirms the existence of a grammati- the existence of a universal semantic and pragmatic category of thetic
cally relevant category boundary between predicate focus and sentence (presentational and event-reporting) propositions, why do sentences
focus in English. expressing such propositions have the prosodic form they do (in
The contrast in (5. 70) also confirms the observation, \Vhich I discussed languages in which the category is expressed prosodically), i.e. how
at the end of Section 5.3.3, that prosodic focus construal is often does their prosodic form relate to their meaning? I will argue that the
determined by the contrast with potential prosodic allosentences. Since principle which governs the interpretation of prosodic sentence-focus
Her. HUSBAND is sick or JOHNSON died have possible event-reporting marking is of an entirely different sort as that which accounts for the
readings, the VP-accented allosentences Her HUSB.4ND is SICK or JOHNSON interpretation of the predicate-focus and the argument-focus structure.
DIED are necessarily interpreted as having a focus structure other than
To the three principles which I have shov•n lo motivate the location of a
sentence focus. Eventive construal of the two-accent sentences is sentence accent, i.e. iconicity, rule, and default (see Section 5.3), I will add
preempted by the existence of the eventive single-accent sentences. This a fourth one: the systemic contrasl between CANONICAL and INVERTED
explains the subtle difference in interpretation between Her Hvsa.4No is sequences.
SICK and Her HUSB.~ND had a HEART attack. The second sentence is easier to
construe as eventive since it does not have an eventive single-accent 5.6.2.2 Previous approaches
allosentence (Her HUSBAND had a heart attack can only have argument In his discussion of the examples in (5.66), Bolinger observes that in the
focus). sentences which I grouped under example (5.68) (Her HUSBAND is to
The contrasts in Bolinger's examples above raise t\vO distinct blarne, Her HUSB,~SD i~ responsible) lhe predicates ··are logically only
repetitions of the initial question ·why,'" i.e. he implicitly categorizes
theoretical questions. The first is a question of tNTERPRETATroN: \vhy do
the various propositional contents in (5.66) favor or force the focus these sentences as argument-focus structures Concerning the sentence-
readings in (5.67) through (5.69), and in particular why do the predicates focus structures in (5.69) Bo\Jnger then \\Tiles·
in (5.69) permit "eventive" construal of the propositions while those in (The predicates in (5 66) (al and {gl) do gi\·e some 1nfonnation. The
(5.67) and (5.68) do not? The ans\ver to this question involves a number information that they give. hl1v.-evcr. 1s of a hackneyed sort - sickness
of factors, which include the lexical nature of the predicators, the number and death are maJor causes of absenteeism, and one or the other could
almost be expected as an excuse for absence The real information !1es in
of arguments associated with these predicators, and the morphological.
the idenuty of the person who was sick or who died .. The remaining
semantic, and pragmatic constraints on these arguments (see Lambrecht predicates all relate to some occurrence which is out of the ordinary -
forthcoming). The issue is too complex to be dealt with satisfactorily here making a scene, fa!hng off a ladder. landing 1n j<lll (Bolinger 1954·152)
(see the preliminary remarks in the section on thetic propositions in
It ls unclear to me \vhy the death of one·s husband should count as an
Chapter 4). There is no doubt a relationship bet\veen the class of
ordinary. hackneyed, event compared to e.g. !us making a scene or falling
predicates pennitting eventive readings of subject-accented sentences and
off a ladder. It is also unclear \vhy the identification of the husband is to
the class of "unaccusative" predicates (Perlmutter 1978), but the tv.·o are
be considered the ··real information"' in (5 66) \al and (g) hut not in the
not coextensive. I should also point out that the class of "eventive
predicates" is much larger than has hecn as'>umcd by most linguists \\'ho other subject-accented sentences.
The na\V in Bolinger's explanat!llO lS that he interprets the prosodic
have dealt with the focus structure of such sentences (sl'e the summaries
peak of the sentences icon1cal!y as the ··inf0rmat1on p1>1nt:· as he calls it
below). In fact, it seems to be an oren cl::i~s FL)r illuminating discussions
(see the summary in Section 5.1 l ). tie interprets 111-ohand 1n ( 5.66) (a) and
of this issue the reader is referred Lo the analyse<. in Fuchs 1980 and Faber
(g) as more informati\'C than in (hl. (dl. (el. (f). and {J) only hecause this
1987.
 3 12 Prag1111111c rt•lariv11s ·_!Ocus Marked and unn1arkedfocus structure 313

noun has relative prosodic pron1inence in the former but not in the latter. preserve the basic insight of Halliday's approach while avoiding its flaws.
sentences. Since the sentence-focus and the narrow-focus patterns a~ The default accent principle allows the "most accentable syllable" of the
formally identical, he concludes that they must be interpreted in the same focus constituent to be located further towards the beginning of the
\vay. If they \Vere not, i.e. if two identical prosodic pallems could haVO constituent, if the syllable which would normally carry the accent is
t\\'O different meanings, the iconic principle of accent assignment woul "deaccented" for pragmatic reasons (see example (5.19) and discussion).
be jeopardized. It is important to be aware of this view because ·i However, Ladd's approach does not account for the accent pattern in
reappears, under one form or another, in many of the subsequeri sentence-focus sentences. Unlike the unaccented object nouns in topic-
analyses, including more formally oriented ones. comment sentences like John doesn't RE..-tD books, the unaccented
The next analysis \vhich deserves to be mentioned is the one bf predicates in sentence-focus sentences are not "deaccented." Their
Halliday (1967). Halliday does not discuss the sentence-focus pattern pef denotata are not discourse-active or otherwise recoverable from the
se, but it is clear that the rules he formulates cannot account for tbiJ discourse.
pattern. As indicated in the quote at the beginning of this chapte;,:"· The next analysis which I would Like to mention is the one by Culicover
Halliday defines the "information focus" as "that "'hereby the speakei! & Rochemont (1983). These authors follow the Chomsky-Halle tradition
marks out a part (\vhich may be the \\'hole) of a message block as tha .. in that they treat stress assignment as a purely formal matter. Focus stress
which he \vishes to be interpreted as informative." Halliday is careful t assignment takes place in the syntactic component of the grammar, by
observe that the focus n1ay sometin1es be "(the v.'hole) of a message, means of a silent morpheme or placeholder v.'hich gets phonologically
block," 1.e. he in principle allov.'s for sentence focus. Ho\vever, his accent realized at the level of surface structure (comparable to the silent
placement rule (quoted in Section 5.3.2), according to v.·hich "the toniC<i syntactic question morpheme of some early versions of transformational
falls ... on the last accented syllable of the item under focus," does not generative gran1mar). The authors claim that "the identification of the
allow for the sentence-focus pattern since it prevents the subject froni- constituent in focus cannot be stated in tenns of either the prosodic
being accented if the predicate is also under focus. · pattern, or the contextual beliefs that are implicated in the interpretation
The same problem mars Jackendoffs ( 1972) analysis, \Vhich i~ · of focus; and the assignment of stress cannot be a function of the
influenced by Halliday's (see Section 5.1.\). JackendofT atten1pts t~· contextual beliefs" (1983:123).
reconcile the Nuclear Stress Rule of Chomsky and Halle (1968) \vith the Culicovcr & Rochemont's argument-v.·hose main goal is to defend a
insight that accenl assignn1ent correlates directly v.·ith the mental states of hierarchical and "n1odular" approach to grammatical theory against the
speakers and hearers. His accent rule. which I quoted in Section 5.3.2, is evidence froin focus prosody adduced by other linguists-is too complex
formally more explicit than Halliday's in that the '"item under focus" of to be summarized here. For the purposes of the present discussion, it is
Halliday's definition 1s nov.' characterized in tenns of phrase structure: sufficient to quote the follo\\'ing observation v.·hich the authors make in a
Like Halliday, Jackendo!T does not deal \Vith sentence-focus structures footnote:
per se. It is clear. ho\~·e\·er, that his strcss-ass1gn1nent rult: fails to account
An unfortunate consequence of this chardctcnz.ation of generalized
for the sentence-focus r:.iuern, for the san1e reas,in .is HalliJav's toniJ pres.entational focus is that ll allows (a), b<::low, in contexts where I.he
placement rule fails to do so. The "regular ~lre~s rules" \Vhich Ja~kendotf sentence is usr.:d to inillate a d1~oun.e, but excludes cases like (b--<:) io
refers to \vould assign stn:ss to the final ~yll.1blc of the verb phrase, not t~ similar contexts
the subject noun Tills 1~ vi be..:,tthc these rules do not make th
(a) The COr-<STRL"CllON crew L~ DYNAMlllSG.
distinction betv.een \·erbo JnJ llPUll\. ,ir prcJ1cate;, and argun1ents, latC (b) A ~trange THOUGHT JUSI occurred to me.
postulated by Schn1erl1n!:' .111d s._,Jkirk (~ec: Se._tiun 5.4.:'J (c) A MAS appeared.
Ladd's ( 1978) ;,:1pprLJ.J,_'h l1' .J'-(•:111 pL1u:rncn1 .J\ o..1ids the pitfalls of th
nuclear-stres~-b.Jsed .1rrr•1,1,_·h h: r''''1ul-it111.l'. the 1heC>rct1cal construct o Jn (b ;.:;), the pred1ca1e c::innot be prcscnt...Lt1ona!l) fo.::u.<.ed, since it is
the "default ,Hxent"· (-,cc St..·1..'l1<'11 :'.\.ii Thi~ c11nqruct :..ililn~·s Ladd fr, not strc~~ed H~,wever, in a context where, e.g. (b) is being used to
 Marked and unmarked focus structure 315
314 Pragmatic relations: focus
riateness of the utterance in {5.71') cannot
initiate a discourse, the predicate of appearance meets our conditions for It seems clear that the approp II believed non-occurrence of a
interpretation as presentational focus. In our tenns, then, it should be be considered contingent upon the mutua y-
loud noise in the s~eech. situt:::ld like to mention here is the one .by
stressed ... Examples like [(b)] might lead us to the conclusion that true
verbs of appearance, like occur or appear need not be focused in order to
be introduced. In other words, in [(b)] used to initiate a discourse, it The next analysts which .. · Focus Rule" according
must be c-construable [=contcxt-conslruable] that something has just Selkirk ( l 984:Ch. 5) .. Selkirk po~~u~ates .:chBa:~;nt is assigned is a focus"
occurred to the Speaker, in order for the predicate not to be focused. to which "any constituent ~o ~.JC a p~ t is a focus" is equivalent to
Given our definition of "c-construable," this proposition must be (p. 206). (Selkirk's expression. a co~s~1tufe:us "')To account for the fact
inferrable from the mutual beliefs of Speaker and Hearer. Let us say
that the mutual beliefs or speaker and hearer include a set of principles
my "the denotatum of a c?nst;t~e;~:sd,~m~ins .of varying length, Selkirk
that a single accent can s1gna o .
of discourse, along the lines or Grice's conversational maxims (1975). "Phrasal Focus Rule" in example (5.72).
We might then assume that the c-construable proposition associated proposes the
with [(b)J in the context with which we are concerned falls under some
version of Grice's CoOperative Principle. (Culicover & Rochemonl, (5.72) Selkirk's PHRASAL FOCUS RULE { \9~4:207~: . .
19UI56) A constituent may be a focus if (1) or (11) (or both) is !rue.
(i) The constituent that is its HEA.o ~s a foe.us. , or the head
The above footnote says essentially the same thing as Bolinger in the (ii) A constituent contained \~'1th1n it that is an ARGLMENT
earlier-quoted passage, except that this time Gricean principles of is a focus.
interpretation are invoked. The lack of accent on the predicate in the
two sentences in question, V>'hich constitutes an exception to their rules, is I hich has a recursive effect, allows the domain
explained by Culicover & Rochemont by saying that the denotata of The Phrasal Focus Rue.'. w d" from smaller to larger constituents.
of a focus accent to sprea ' unt for the
these predicates are "context-construable," i.e. pragmatically recoverable .. ..) f (5 72) is motivated by the need to acco
from the discourse context. Cond1t1on (11 o . uments and predicates \vith respect to the
try bet\.veen arg · k ·
Besides the fact-established beyond doubt by Schmerling (1976), asymme - ce (see the quote from Selk1r in
interpretation of prosodic prom1nenl le the Phrasal focus Rule
Fuchs (1980), Faber (1987), and others-that the sentence-focus pattern . 542 b ) To take a s1mpe ex.amp ,
is by no means restricted to what Culicover & Rochemont (following Section . · a ove · . (5 7 1,) can have the t\vo focus
accounts for the fact that the sentence tn . -
GuCron 1978) call "natural verbs of appearance," I do not see how interpretations in (a) and (b)·
Grice's Cooperative Principle can be used to justify the claim that the
denotata of such verbs can be considered context-construable. The
(5.73) She watched "KOJAt<..
weakness of the approach to sentence-focus construal in tenns ofGricean a. She watched 1oc[ "KOJAf.. .. l
maxims is apparent also in Culicover & Rochemont's discussion of the b. She ,rx:( watched "KOJAt;." l
following example:
I (S 733 ) the object argument Ko;a .. k a 1one is ·nterpreted
i
as the focus byd
(S.71) My exploded. ( = Culicover & Rochemont's (64))
~ . of, the Basic focus Rule, according to_ which any acce~te
STEREO

The authors claim that (5.71) is appropriate only in a context in v.·hich the virtue . ,. resumablv also bv virtue of condition (1) in
constituent "1s a focus, (~nd. P d · f h N.P Koiak). In (5.73b), both
previous occurrence ofa loud noise is "mutually believed" by the speaker 5 72) . the noun Ko1ak ts the hea o t e .J . ..

and the addressee. This is clearly false. The falseness of this claim ( . , since d the redicate are in focus by virtue of cond1t1on (11) .of
appears, for example, in the fact that. in the appropriate universe of the argumentFan R pl The accented NP is able to include under its
the Phrasal ocus u e. ~<i
discourse, the corresponding NEGATED proposition can also have focus the predicate of V>'hich it is an argument. ct· . . the case
sentence-focus construal: Selkirk's Phrasal Focus Rue I ma kes the v.Tong pre 1ct1ons in
didn't cxrlode 1
of the sentence-focus pattern_ Con'>ider the structure in (5_74)
{S.71') Guess what 1 My STEREO
-------------------~------~~-- ---------

rri1gn1aric relations. focus Marked and wunarkedfocus slructure 317

(5.74)
s Assignment Rule" (SAAR) in (5.75), which "operates over FOCUS

~VP
NPFOC
DOMAINS." A focus domain is defined by Gussenhoven "as one or more
constituents whose [+focus) status can be signaled by a single accent..
(1983:391) 51 :
~I
DetNroc V (5. 75) Gussenhoven's "Sentence Accent Assignment Ru1e" (SAAR)

I
Her
I
HUSBAND
I
died
Domain assignment P(X)A - [P(X)A]
(a)
A(X)P - [A(X)P]
y - [YJ '·
(b) Accent assignment: [) -+ [•].In AP/PA, accent A.
Sel~rk's rule predicts that in the sentence Her HUSBAND died only th (A = argument, P = Predicate; X,Y = any of these;
subject NP may be a focus, i.e. it allows only for argument focus an bold race = [ + focU5]; [I ~ focus domain: • = sentence accent.)
excludes the sentence-focus interpretation. The focus of the subject
·Example (5.75) says, among other things, that in any focus domain
cannot spread to the verb since the accented subject is neither the head 0
involving a predicate and an argument the argument receives the accent.
the VP nor c~ntained within the VP as an argument of V. Noticing th
'This is meant to capture the difference between predicates and arguments
problem, Selkirk makes the following comment:
with respect to accentability. The X in parentheses in the domain·
It has ~en claimed that some sentences having a prosodic promine •'.
·assignment lines alloWs for topical arguments within accented· focus
only w1t~n the subject NP, ltke these [i.e. The SUN is shining, M domains of the kind discussed in Section 5.3.3. Gussenhoven illustrates
UMBRELL.~ s been found, Aly MOTHER's conzing]. are perfectly appropri the application of the SAAR with example (5.76):
when uttered out of the blue ... Does their appropriateness when ut ·
out of the blue require us to consider the VP to be focused here ~~ (5.76) Example: AP - [A• P] Our DOG's disappeared. (Gussenhoven's (30))
there~ore to al!ow for the_ possibility that a focused argumen; · ~Q
Given the argument our dog and the predicate has disappeared, and given
conta_1ned ...,.tthin the conslttuent with its head [sic] may focus t "
co~s.tnuent? We think not. It may simply be that it is possible in"
sentence-focus construal of the proposition, the two constituents form a
feh~1tous discourse to utter sentences out of the blue where onJy 1lJC single focus domain, in \vhich the accent falls on the argument.
subJecl NP, and not the entJre sentence, is being focused. (Selkir Gussenhoven 's analysis faces difficulties in the case of two-accent
1984:217) ·-
sentences of the type discussed earlier. Since both the argument and the
predicate may receive an accent, as in (5.58d) or (5.67') above, or as in the
Selkirk's ~omment embodies the same claim as the one made in the quo~ examples in (5.77) (from Faber 1987)
from Cu!Jcover & Rochemont above. By suggesting that it may be
"possible in a felicitous discourse to utter sentences out of the blue whe~ (5.77) a. TRESPASSERS WLJl be PROSECUTED
b. JOHN'S WORKING
only the subject NP, and not the entire sentence, 1s being focused" Selkirk
c. JOHN PROTE.STED. etc
claims in fact that a non-focused constituent can constitute "ne~·
information" and "old infonnat1on" at the same tin1e, invalidating h_~· such sentences will have to contain tv•o focus domains in order to receive
own focus definition. 50 1 t\VO accents. 52 While this is formally easy to accomplish-the first focus
The last analysis \\'hich I \\·ould like to mention here is the one b~~ domain contains only an argument (using option Yin part (a) of(5.75)).
Gussenhoven (1983). Gussenh1Jven char::ictcri?es focus .. as a binil' the second only a predicate- it begs the question of v.'hy the predicate
variable which obltgatonly n1ark<; .ill or part of a sentence ··as constitutes a separate focus domain in (5.77) but not in (5.76).
f +focus] ... [+focus] m;irk<; thL· ~PC'<Jf...er\ dL·c!arcd contribution · tri · Gussenhoven 's analysis, though formally correct, is notionally inad-
the conversation, \vh1!t: [-f,:icu~] con\l1tute5 hh cngn1tive starting point" equate in that it lea\'es unexplained how a single argument constituent
(pp. 380, 383). Gus~enho\·cn thcri f1)rn1u!cttcs the "Sentence Accelit" can constitute a focus don1ain. Even though Gussenhoven does not
 Marked and unmarked focus structure 319
318 Pragmatic relations: focus

appeal to the notion of "new informatio " h . As I observed in the discussion of these examples in Section 4.2.2, the
sentence accents his anal sis . n to c aractenze the function of common feature of the sentence-focus constructions in the four languages
those by Chafe •Jackend:fT ~s 1~-a~ed by the same notional problems as is that the NP is formally marked as a non-topic, i.e. as included in the
in terms of ne~ informal·, e Bir , an~ many others, who define focus
ion. Y treating the s b. · focus domain.
sentence as a separate focus domain t u .Ject _in a two-accent Concerning the morphosyntactic focus-marking devices in the Italian,
referent such as e g th b" he author imphes that a single
' ·· esu~ecttre · French, and Japanese sentence pairs in (5.78), few linguists, if any, would
constitute "a speaker's decl d . spassers in (5. 77), can by itself
argue nowadays that the contrasting structures in each pair are related to
is, to say the least co f . areA contnbution to the conversation ... This
' 0 using. referent can be said 1 b C each other by some transformationa l or other grammatical RULE
to "contribute to the conversation " onl . . . o ~a ocus, hence explaining their similarity in meaning. The members of these various
as a predicate, as explained in Section 5 ~ ~ tt IS _pragmatical~~ construed sentence pairs are independently generated structures. each with their
have two foci. I will return to this a· . ·.. A single p~opos1t1on cannot own semantic properties. The same is true of the information-stru ctural
(5.85") and d' . p int in more detail below (example
1scuss1on). Gussenhoven h0 ... properties of the various sentences. The different pragmatic interpreta-
distinction between .. ' w criticizes Schmerling's
. news-sentences" and "topic .. tions given to an SV and a VS sequence in Italian. a n•a-marked and a ga-
circular, in fact confirms im licitl t --<:omm~n~ cases as marked NP in Japanese, or a detachment construction and a cleft
state that at I d. p y he need for such a distinction To
others is tantoca pre icate nee~s. an accent in some sentences but n~t in
construction in French, cannot be captured \vith any kind of rule. Instead
structure typeamount to recogn1z1ng the existence of two separate focus- the interpretation of each member of a pair is based on the FORMAL
s. CONTRAST which distinguishes it from the other member. Each member is

wh~~:;n;.o: ;t appears tha~ none ?f the. analyses summarized above, interpreted in terms of some grammatical feature which is absent 53
other, one member being typically taken to be the unmarked one.
in the
a or pragmatic, or 1nvolv1ng a combination f b
approaches, can satisfactorily explain the p d' o oth The same reasoning concerning the pragmatic interpretation of the
sentence~focus h roso IC structure of the
pairs of morphosyntactic allosentences in (5.78) fb), (c). and (d) must be
assumption tht:i~h~ e weaknes_s of these various analyses is the implicit
grammatical mechanism whe b . applied to the pair of prosodic a!losentences in (a). There is no single rule
structure is paired with f . re Y a prosodic or principle \Vhich could account for the interpretation of the two focus-
. a ocus meaning must be the same for all
sentences, i.e. that all focus types can be accounted for wit.h . accent positions in English. The ""'ay we understand the difference
In the t · . a single rule
nex section, I will present an analysis in which th . beh\'een the t\vo members is by \vay of contrast. Now if we look at the
structure is treated as a e sentence-focus \Vay in \Vhich the contrast bctv.•een predicate-focus and sentence-focus is
. separate formal category or, as I would like to
call It. . a PROSODIC
. CONSTRUCTION · A "prosod·IC construction" . is the formally coded in English and Italian, \\'e notice an interesting fonnal
prosod IC equivalent
. of a "grammatical t .
cons ruction" in the sense of parallel. ln both languages, an element appears in a POSITION in which it is
C
onstrucl!on Grammar (see the definition in Section 1.4.3). not found in the other member of the pair. In Italian. this element is a
syntactic constituent; in English, it is a feature of prosody. Symbolizing
the subject 'vith the Jetter "'S" and the predicate \.Vith the letter "P," and
5.6.2.3 Prosodic inversion symbolizing prosodic prominence \.\'ith an acute accent, the system of
Let
~
. the various pairs of sentence-focus and pred· t -
us cons1•ct er _again contrasts in the \\.\'O languages can be sumn1arized as in (5.79):
, convenience as
1ocus sentences in (4 . 10) , wh·IC h I repeat here !or 1ca e
· (5.78):
Pred1c3\C Focu~ Sentence Focus
(5 79)
(5.78) A. What's the matter? B How's your ned.; doing? ltahan· SP PS
a. My NECK hurts. J. l\ly neck H\IRTS Engh sh sP SP
b. Mi fa male ii COLLO h Tl coll1• m1 f 3 M.i.L r
Both languages resort to the system of pennutat1on traditionally referred
c. J'ai le cou qui me f3i1 \tAl l' ~fnn C••U il me friil M.i.l
d. KIJBI ga ITAi d Kuhi wa IT·\I
to as "inver~ion," v.here the non-inverted sequence i~ interpreted as
 Marked and unmarked focus structure 321

unmarked. ln bo1h languages, the senlence-focus pattern is a FOR French, which has neither syntactic nor prosodic flexibility, it can be
REVERSAL of the unmarked predicate-focus pattern. The only difference.~ achieved by reorganizing the syntactic structure or the sentence in such a
that in Italian the syntactic sequence is inverted and the accent stays put,' way that both clause-final accent position and clause-initial subject
\\•hile in English the accentual se4uence is inverted and the syntax stay position are preserved (see Lambrecht 1988a). In all cases, the
put. l suggest therefore that v.·e refer to the formal patlern of the En · relationship between the structure of the sentence and its pragmatic
sentence-focus construction as PROSODIC tNVERSEON. Both types, 1 construal is not provided by rule but is given as a property of the
54
inversion, syntactic and prosodic, are formally highly distinctiv grammatical construction as a whole.
indicators of a contrast bel\\'een a marked and an unmarked memb · To sum up, in the English sentence-focus pattern the subject is
of a pair. . accented and the predicate unaccented not because the subject is
The above analysis of the interpretive n1echan1sm which determines. "newer," more "important," or more "focusworthy" than the pre-.
interpretation of pairs of expressions which are perceived as related~·· dicate, as Bolinger and other Linguists have it, but because accenting the
based on the intuition which traditional grammarians tried to capt ' predicate would necessarily result in predicate-Focus construal of the
with the notion of ''inversion," which was applied to various phenome propositio11. Since sentence focus is pragmatically defined by the absence
involving the reversal of sequences \vhich were perceived to be "normal' of a topic-comment relation between the subject and the verb, accenting
TIDs traditional view of inversion is fundamentally different fro the verb phrase in a sentence-focus structure would result in intolerable
(though not necessarily incompatible \Vith) the now common vix ambiguity. The fact that ambiguity rs tolerated between sentence-focus
associated with the so-called ··unaccusative hypothesis," according tJr and ARGUMEl'.fT-Focus structures can be explained as a result of two
which certain VS structures in Italian and other languages are in fact VO~ factors. The first is that these tv.·o articulations have one crucial
structures at a deeper level of analysis. \Vhile the traditional vie · pragmatic feature in common which sets them off against the
emphasizes the idea that an inversion construction marks a deviati~n predicate.focus articulatlon, i.e. the non-topical relation of the subject
from some structural norm, the unaccusative viev• emphasizes the id.~ to the proposition (see the feature representation in Section 5.2.5, Table
that an apparent structural deviation can 1n fact be reduced to.JV\\, 2). The second factor is the unmarked status of predicates with respect to
underlying formal regularity. Even if the latter vie\Y is correct, it cannq activation, allov..·ing unaccented verbs to be interpreted as either focal or
replace the traditional analysis since it does not capture the perceptual non·focal, depending on the construction {see Section 5.4.2).
nature and communicative function of inversion. The formal identity of sentence focus and argument focus in English
It should be noted that \Vhile the 1nterpretat1on of the accent positio does not entail that subject-accented sentences are v,o..uui:, in the way in
in the sentence-focus pattern is not predictable by rule, it is neverthel v.·hich we said that predicate-accented sentences are vague between
not ARBITRARY but fonnally highly MOTIVATED. The minimal distinctive: "narrow" and "broad" focus construaL This formal identity is better
feature of a predicate-focus structure in English [and many oth.' characterized as an instance of HOMOPHONY, where tv.·o distinct meanings
languages) is the presence of an accent on some ele1nent of the predica~. · are encoded in one form. Partial or total homophony of sentence-focus
phrase. In order to mark the AB~ENCE of predicate focus in a sentence it.is and narro\v-focus sentences is a common occurrence across languages. It
therefore necessary to alter tlu~ 1ninirr,..d d1s!inct1ve feature by removing can be observed for example in Japanese. v.·here the "uhject of both types
the accent from the predicate phrase Since the n11nimal constituents in a" is marked \Yith gll. Unlike English, ho\1..·ever, Japanese does permit
sentence-focus construct1l n .ir..: a \ub_1ec1 anJ a predicate, marking the
1 accentuation of the predicate In sentence.focus sentences, since in this
absence of predicJ.te focus fL'LJUJres the f'RE'iEt~~T l)r an accent on the·, language the forn1a! variable is morphological, not prosodic (a common
subjecL In a flexible-aci:ent l.1nl:'u.Jge 11\..,e Englt~h. the m::i.pping of subject·· example is (4.!9d) Dt:SH'.~ ga .JVA.TTE ir11 yo "The PHONE's ringing," where
position and sentence accent pc1'.>111l1n c,111 he :11:h1eved by changing the the predicate 11arte is accented). SimiLirty, in ltali::i.n VS structures the
position of the accerl! In a llc_\1hJ,··~\ nld\ !.Jngu..1ge like Italian it can b.e predicate n1::i.y be prosodically pron1inent in addition to the subject, given
achieved by changing the ru'.>1\11111 ,_.( 111<..: ~ub_1ec1 1\nJ in a language like that· in Italian the vanable is not prosodic but syntactic (compare e.g.
!.."'--;

A unified functional account of sentence accentuation 323

322 Pragmatic relations: focus
will propose a unified account or the function of sentence accents in
(S.78b) with the possible variant Mi fa MALE ii coiw). Since in the
~apanesc an~ the Italian construction absence of prominence on the verb general.
IS not run~ooally distinctive, fluctuation in the prosodic pattern does
not lead to 1ntolerable ambiguity, as it would in English.
!o. sum ~p, in the. analysis I propose here, the basic cognitive 5.7.J Activation prosod}' rel'isited
pnnctple w~ch determines the interpretati on of the prosodic contrast
To make such an account possible, it will be necessary to refonnulate
between subject-accented and predicate-accented sentences is the same
parts of the analysis presented in Chapter 3 concerning the relationship
as that whic~ ope~ates in the pragmatic construal of the pairs of
allo~ntences 1n Italian, French, and Japanese. The interpretation of the between prosody and the activation states of discourse referents. The
English sentence-focus pattern is determined neither by fonnal rules, i.e. need for such a revision was hinted at repeatedly in the course of this
chapter in the fonn of a recurring exception to the simple correlation
~ r:i!es defined o~ s~me structural property of the sentence, nor by
ICO~ctty, e.~. by we~gh1ng the relative communicative importance of the
between prosodic prominence and referent activation. This exception is
:a-nous d:Stgnata 1n a sentence. The cognitive mechanism for the the occurrence of sentence accents on constituents whose referents are
tnterpretab on of the sentence-focus structure is of an entirely different clearly discourse-active in the utterance context, in particular on
~ort. The meaning of the accent on the subject, rather than being pronouns (see Section 5.5). It is clear that the function of prosodic
interpreted solely on the basis of a given structure, is interpreted against prominence on a pronoun cannot be that of referent "activation " proper,
the backgroun d of an unused alternative structure in which the accent is if activation is understood as in Section 3.3. I as the cognitive process
whereby a previously inactive discourse referent is "lit up" in the
on the predicate. It is a function of the existence of one or more
consciousness of an addressee, as Chafe puts it. Since pronouns by their
P~osoorc.ALLOSE~NCES provided by the grammar of the language. This
kind of 1nterpret1ve mechanism requires a "structural ist" rather than nature are assumed to have active referents, independently of their
"ge?erativi st" approach, i.e. an approach in which the interpretati on of prosodic manifestation, the function of an accent on a pronoun cannot be
a given structure is viewed as being determined tvithin a system of to activate the referent in the hearer"s mind.
formal oppositions rather than by a set of rules. To understand the function of accents on constituents with already-
active referents, in particular pronouns, it will be useful to emphasize
again the theoretical distinction bet\\'een, on the one hand, the cognitive
properties of the mental representations of discourse referents in given
5.7 A unified functional account of sentence accentuation discourse situations and, on the other hand, the pragmatic relations
\vhich these mental representations enter into with given propositions (see
In this chapter, I have repeatedly emphasized t\\'O facts about the
in particular the discussion in Section 4_4.1 ). \Vhile in the case of accented
grammar of focus. The first is that the focus of a proposition may be
pronouns the cognitive STATE of the referents as discourse·active is
marked not only by prosodic but also by morphosyn tactic means. The
already established at the time they are verbalized, allowing them to be
second is tha.t n~t all sentence accents are focus accents; accents may also
coded pronominally, their ROLE as topics or foci in the proposition is
mark t~e a~ttvat1on .state of a referent. It is no.,.,, time to address a general
established only via the utterance itself The accent can therefore strictly
theore~cal issue which arises as a result of this functionally non·unitary
speaking not be a function of the assumed state of the referent in the
analysts of sentence accentuatio n. Since focus and activation are different
interlocuto r's mind. Rather it must be a function or the pragmatic ROLE of
lnformatio n:st:ncture categories, \vhy are they often expressed by the
the referent in the given proposition. The function of an accent on
same, or a similar, phonetic fonn? Phrased differently, given that the l\vo
constituents .,.,·ith active referents, .,.,,hether pronomina l or nominal, is
categories have the same, or a similar, phonetic manifestation, is this
then to establish the role of a given referent as a topic or a focus
phonetic similarity the expression of a common functional basis'> In the
argument in a pragmatically structured pr0positi0n
present section, I will argue that the ans\~·er to this question is "yes" and I
 324 Pragmatic relations: fOcus A unified functional account of sentence accentuation 325

Now since there is no reason to assume that an accent on a constituen of sentence accentuation the accent marks a NON-RECOVERABLEPMOMATIC
with an active referent has a different function from an accent on.a' RELATION between a denotatum and a proposition. By "non-recoverable
constituent with an inactive referent, we can extend the abov relation" I mean a relation which an addressee cannot be expected to take
explanation to ALL constituents carrying activation accents. This lea for granted, in one sense or another, at the time a proposition is
us first to a revised definition of the category of "activation." 'I;, verbalized. The general discourse function of all sentence accents can now
ACTIVATE a referent is then not simply to conjure up a representation ofi be characterized as in (5.81):
in the mind of the addressee but to ESTABLISH A RELATION between it and!'
proposition. The assumed mental state of a referent is only• (5.81) THE DISCOURSE FUNCTION OF SENTENCE ACCENTS: A sentence accent
PRECONDITION, not the REASON for accenting the constituent expressin indicates an inslruction from the speaker to the hearer to establish a
it. Referent activation is then not only a psychological but also a properl pragmatic relation between a denotatum and a proposition.
linguistic fact.
This revised account of the nature of activation accents makes According to (5.81), sentence accentuation is relational by definition.
necessary to reformulate the conditions under which a referen ~'• We can then distinguish two major kinds of accents: focus accents and
constituent may occur WITHOUT an accent. Since accenting a constitu~ topic accents. Both can be characterized as "activation accents," if
entails establishing a relation between its referent and a proposition, N activating a denotatum is understood as establishing a relation between a
accenting a given constituent entails considering this relation to denotatum and a proposition. In the case of a focus constituent, the
already established. Instead of saying, as Chafe does, that a constituent!' relation being established is that between the focus and the non-focal
unaccented if its referent is discourse~active, we must postulate the riv portion of the proposition (which may be null). The accent marks the
conditions in (5.80): designatum of the accented constituent as entering a focus relation with
the proposition, i.e. it marks it as the element whose presence makes a
I
(5.80) DISCOURSE CONDITION ON UNACCENTED CONSTITUENTS: A
proposition into an assertion (see Section 5.1.1). In the case of a topic I
constituent is unaccented if and only if the speaker assumes: (i) tha,~ . 1,
constituent, tbe relation being established is that between the topic
mental representation of the referent is active in the addressee's rnind (o
can be accommodated by the addressee as such); and (ii) that the referent and the proposition. The accent marks the referent of the
addressee expe.:ts this referent to be a topic in the proposition at'·t:hc accented constituent as entering a topic relation with the proposition, i.e.
time of utterance. .t: as being the element which the proposition will be about or which serves
to establish the temporal, spatial, or instrumental framework within
The formulation in (5.80) requires a modification of my which the proposition holds (see Section 4.1.1 ). In both cases, the
characterization of unaccented referential constituents as bein &entence accent symbolizes an instruction from the speaker to the hearer
"marked for the feature active" {see Sections 3.3. l and 3.4). Strict!~ to establish a relation betv.,een a denotatum and a proposition. Notice
speaking, such constituents are not marked for the feature "active" bU: that a topic accent necessarily cooccurs with a focus accent in a sentence
for the feature ESTABLISHED TOPIC. This analysis is consistent with while a focus accent does not require a cooccurring topic accent.
definition of unaccented pronon1inals as the ''preferred topic expr _ The establishment of a pragmatic relation between a denotatum and a
sions" (see Section 4.4.3). It lends further support to my claim proposition includes the marking of a SHIFT from one already-activated
"topic" is indeed a formal grammatical category in English. 55 designatum to another (such a shift may or may not be perceived as
Given the revised formulation of th~ condition for non-accentuation contrastive). In the case of topic constituents with active referents, in
referential constituents we can no\v charJcterize the occurrence1(b particular pronouns, prosodic prominence may be seen as an iconic signal
sentence accents in general in ti.:rms 1)f the notion of DEFAULT: uni of some sort of D!SCONTlNUITY in the expected anaphoric relations in the
(5.80) applies, a constituent \~·ill carry .in dccctH, \\'hcther it is in focus·, · sentence. The prosodic marking of a topic shift is related to the
in the presupposition. This chJracti.::nz;iuon implies that in all instan phenomenon of $1,l,'ITCH REFERENCE. 56
 326 Pragmatic relations: focus A unified functional account of sentenc:e accentuation 327

S.7.2 Topic accents and focus accents: some examples mind the bi-accentual pattern referred to by Ladd as "reciprocal."
Consider example (5.83) (Ladd's (17))o
The establishment of a common functional basis for topic accents and
fOcus accents allows us to better understand certain cases or infonnation- (5.83) A: Hey, come HERE.
structure underspecification in sentences in which the topic-focus B: No, YOU come HERE.
distinction is not made by morphosyntactic means. One such case of · accen t , ,· .e . .it. serves to
In B's reply the accent on you 1s · a topic . establish
.
underspecification, which I mentioned before, is found in sentences in the role of the referent as the topic of the pro~os1tton .. This accent is
which both the predicate and the subject receive an accent (see e.g. required because the referent of the topic expression you ts not .the s~~e
example {5.67") and discussion). As we saw, such sentences have in as that of the understood topic "you" in A"s utterance. Since it is
principle two readings, an eventive reading, in which the subject is part of
different, it cannot be considered the establishe~ topic._ The accent ~n
the focus, and a topic-comment reading, in which the subject is a topic
with a non-active referent. As we saw with example (5.41), such sentences here, o 0 the other hand, is a focus accent. It ts required ··
because its
"x come
absence would mark the designatum of the open propos1t1on . .
may in principle also have a third reading, in which the subject is focal here" as pragmatically presupposed. Since the_ referent of. t?e deictic
and the predicate presupposed but inactive; I will ignore this possibility
"shifter" here is different in the l\vo sentences, this ~resu.p~os1t1on wo~Jd
here. Let us look at another (attested) example:
not hold. As in previously mentioned cases, it is the 1mphc1t contrast with
(S.82) That was a STUDENT of mine. Her HUSBAND had a HEART attack. a possible a\losentence \vith a different information-structure that
detennines the presence of the accents tn (5.83). .
The purpose of the utterance in (5.82) was to explain why the speaker had A similar situation obtains in the hackneyed example in (5.84):
left a- discussion among colleagues in order to talk to the student in
(5.84) John hit BILL and then HE hit HIM
question. In the given situation, the proposition expressed in the second
sentence in (5.82) was not to be construed as conveying information In (5.8 4), the preferred interpretation of he as refe_rri~g to "Bl_ll" rather
about the subject referent since this referent \Vas not topical at the time of than "John" follo\vs directly from the topic-estabhsh1ng function of the
utterance. What was topical was the student, expressed in the possessive ince the accent on he marks this constituent as a not-yet-
accen t . S d-~
determiner her. But the same sentence, in a different utterance context, established topic, the topic referent is naturally interpreted as 1 erent
could be used to convey information about the husband (What abora her from that of the preceding clause (i.e. "John"). If he v.'ere not accent~d it
family? - Her HUSBAND recently had a HEART attack but her KIDS are doing inte..-nreted as referring to John, by \'lrtue of (5.80), since
wou Id be "t' . . d .
FINE). In this case, the first of the two accents would be a topic accent, "John" is the already-established topic for the propos1t1on an since
serving to establish the referent as the topic with respect to which the continuation of the same topic 1s the expected strategy. The accent on
proposition is to be interpreted as relevant information. What both him, on the other hand, is a focus accent of the same kind as that on here
readings have in common is that the accent on the subject indicates the in (5.83). lfthis accent v.•ere absent, i.e. if the sentence v.•ere ... and then.HE
establishment of a pragmatic relation between the referent of the subject hit him, an open proposition "x hit him" v.·ould be under_stood as being
constituent and the rest of the proposition. It is this relation that is pragmatically presupposed. Since such .a presuppos1t1on v.•ould be
expressed by the accent, not the topic-focus distinction. The latter is left inherited from the preceding clause, and since in that clause the person
underspecified by the structure of the sentence.~ 7 being hit is Bill, an unaccented J11 rn \\'ould naturally be interpreted as
The formulation of the appropriateness conditions for the use of referring to Bill. To preclude this unnatural 1nterpretat1on, the pronoun
unaccented referential expressions in example (5.80) and the functional must be accented. Notice that I am not c\a1rn1ng that the 1nterpretat1on of
distinction between topic accent and focus accent allcl\v us also to the l\VO accented pronouns is necessarily the one indicated above. tv1y
account in a straightforward fashion for a class 0f sentences \vhich have accentuation principles only account for the fact that such an
posed problems for previous analyses (see Ladd 1978:7Sff). I have in interpretation is the preferred one
328 Pragmatic relations: focus
A unified functional accoW1t of sentence accentuation 329

A related example is the follo\.ving, taken from an article in the :· candidate in the universe of discourse. The implicature in question is
J.t'ashington Post about traveling in France: quantity-based (an "R-based implicature," in Horn's 1984 terms) in that
the accented pronoun says more than is needed to designate the given
t5.85) The American travel writer Paul Theroux once defined an Englishman
referent. 58
as someone who apologizes if YOU tread on HIS foot. To extend the It is necessary to mention an alternative analysis of the two-accent
analogy, a Frenchman could be defined as someone who expects you to
apologize if HE treads on YOUR foot. sentences in (5.83) through (5.86), which-though appealing-is excluded
by my theory of focus, correctly, I believe. This analysis, which is
Unlike in (5.84), no pragma(ic inferencing is required in this example to proposed by Selkirk (1984:200ff) and by Gussenhoven (l983:380ff),
interpret the referents of the pronouns since the second and third persons among others, is based on the assumption that a single clause may
have different forms, but the general information structure of the express more than one focus, hence may have more than one focus
sentences is the same in the two examples. The accents on the various accent. \Ve may call this the "multiple-focus analysis." The occurrence of
pronouns in (5.85) are motivated by the fact that in each case the referent multiple foci in a single clause is not allowed for by my focus definition,
of the pronoun is not the one most naturally expected to fill the giveq. since a given assertion cannot have more than one focus and since a given
argwnent role in the propositon, given generally accepted rules of proposition cannot express more than one assertion. (However, nothing
politeness. In each case, the accent signals that the relation between the in the present framework prevents the occurrence of multiple~acceµt
referent and the proposition cannot be taken for granted from preceding sentences or clauses, since there is no one-to-one relationship between
discourse. ' "accent" and "focus.") A multiple-focus analysis of the last clause iD
The use of topic-shifting accents is not restricted to contexts in which-~ (5.85) could be represented as follo\vs:
given topic is different fron1 that of a preceding sentence or clause. The
discontinuity marked by the accent may be the unexpected selection of an (5.85') Sentence: H£ treads on }"OUR foot.

already established topic referent over some other, more expected, Presupposition: "x treads on y's foot"
referent which is available in the universe of discourse. Consider first the Assertion· "x = he; y = you"
exchange in (5.86): Focus: "'he; you"
Focus Don1a1n. NP; NP
(5.86) A: What is Mary's job going to BE?

B: She's going to do the cOOKJNG. One possible argument in favor of the analysis in (5.85') is the fact that
The lack of pitch prominence on the subject pronoun she in the reply the sentence in (5.85') could be conceived of as an answer to the multiple-
signals the speaker's assumption that the role of the referent as the topiC \VH question IVho treads on 1\'hose foot? Since a WH-word in a question
of the proposition is expected or taken for granted by the addressee. (The can be used to determine the (argument) focus portion of an answer
topic referent was mentioned in the preceding sentence.) Now considef (Section 5.4.4), one could argue that those denotata in (5.85) which
the alternative version in (5.86'): correspond to the l\VO WH-words in the question must be foci.
The main reason v.'hY the multiple-focus analysis is to be rejected is
(5.86') A: What is ~.fary's job going to Bf'.' that the Assertion line in (5.85') is ill-formed. A single proposition cannot
B: SHE'S going to do the (()Qt...lr"'G
express t\vo assertions, therefore it cannot have t\\'O foci (see lhe criticism
As in the previous examples, the pitch accent on the subject pronoun in of Gussenhoven·s sentence-focus analysis in Section S.6.2.2 above). My
the reply in (5.86') signals that the role of the ~uhje..:t referent as the topic rejection of the analysis in (5.85') is supported by t\vO grammatical
of the sentence is not yet e~LJ.bli).ho:d c1t the tln1c iJf utterance. However, observations The first is that a single sentence cannot be clefted twice.
since the referent is tn fact an Jlre:.iJy-e,.t:.iblished topic (it was mentioned For example, a structure like (5.87) is ill-fanned:
in the preceding sentence\. the Jcccn! i;1\c~ ri~e to Jn i1nplicature: the
(5 87) ~1t 1~ 1nt R foot th.it is HE that treads on.
referent "Mary" is betnt'. ~e1e~·1l·J u\·cr ~on1t' potential alternative 1t
 A unified functional account of sentence accentuation 331
330 Pragmatic relations: focus
the contrastiveness effect conveyed in (5.85) is a consequence of the
Ipart
believe that the ungram
explained · rtty of sentences like (5.87) is at least in
. mauca unusual state of affairs described, not of the information structure of the
pragmatically as a result 0 f th · ' sentence. The subject pronoun he is accented because it is not yet the
violation which ·t ' . e iniormation·structure
1 presents A single ·· established topic of the sentence. The NP your foot is accented because it
.. He treads on your foot .: propos1t1on, here the proposition
contain two pra f • ~nno:9 be clefted twice because it cannot is a referential expression located in the focus domain. Within this focus
tion which . gm!eda IC assertions. The second grammatical observa- NP the determiner your is accented by default since the referent "foot"
' ts re at to the first one . th t . satisfies the condition in (5.80), being a topic expression with an active,
cannot contain a cleft co st . 6-0, is a a mult1ple·WH question
n ruction This constrai t · h
naturally illustrated in French than i. E J' h . n is per aps more i.e. expected, referent.
freer use of clef . . n ng is , since French makes much The category "topic accent" allows us to better understand another
examples: t constructions in questions. Consider the following accentual pattern which has often been discussed in the literature on
focus in English. This is the pattern found in sentences in which the VP in
a predicate-focus sentence contains two prosodic peaks. Consider the
(S.88) a. ~·~t qui qui a mange le fromage? "Who ate the cheese?..
I~
it who who bas eaten the cheese . contrast in (5.89). which is discussed by Selkirk (1984:21 lfO:
b. Qui a mange quoi? "Who ate what?"
c. •C'est qui qui a mange quoi? "It is who who ate what?" (S.89) a She sent 3 BOOK to MARY. (Selkirk's (5.6))
b. She sent ajthe book to MARY. (Selkirk's (5.7))
rel ti shows
(5.88c) l that it . is impos s1·ble to use a c ,est-cleft construction if the
Selkirk argues- against \vhal she refers to as the ··NSR-focus analysis"
a ve c ause
perfectly natural contains
if th a b.question. wo rd . H
. owever the same sentence is represented by Chomsky, Jackendoff, and others-that the pragmatic
. e o Ject in the relative clause is a lexical NP as
~S.88a).sbhlows. Like (5.87), sentence (5.88c) is ungrammatical because it is
difference bet,veen the t\l.'O sentences in (5_89) requires an approach to
focus structure in \\'hich focus domains (such as the VP in this example)
hunposst e to construe
. . a single proposition
·· as expressing two assertions
are a\\o,ved to contain argument constituents \1,rhich may or may not
ence
questi as conta1mng . two foci. One of two or more WH ·WOrd s ID . a,
themselves be in focus. Her argument is in agreement with the analysis I
on n~anly cor:esponds to a topic, not a focus, in the answer
The ~mmatical facts in (5.87) and (5.88) confirm that multiple WH.
presented in the section on default accentuation above (see examples
(5.19), (5.20) and discussion). Nevertheless I belie,•e that Selkirk's
questions
sentences. cannot be used 1o argue } or the existence
. of multiple·focus• account of the contrast in (5_89) is f1a,Yed. Recall that for Selkirk,
\vhose frame'.vork includes neither the concept of topic nor that of
th Tuel analysis represe.nted in (5.85') must therefore be rejected. Instead
e re eva~t se~tence in (5.85) is to be interpreted as a predicate-focu~ activation, any constituent to "'hich a pitch accent is assigned is a focus
constituent by virtue of her "Basic Fcicus Rule" (see Section 5.6.2.2
~c;;re, 1n.which th.e subject is the topic and the predicate the comment
. . ~t infonnallon·structure representation of the sentence in (5 85).
above). a focus constituent being defined as one which "contributes 'ne\V
infonnation' to the discourse .. ( !984:206) Selkirk does not allow for
IS given m (5.85"): ·
accented constituents \vhich do not eApress foci
(5.85") Sentence: ..HE treads on rouR Joor To account for contra~ts such a~ that in (5.89J. Selkirk uses her
Presupposition: ..referents "be.. an d "foot ·· are topics for comment x" "Phrasal Focus Rule .. (see example (5.72) ahove 3nd discussion), \Nhich
Assertion: x = treads on your (foot)" al\O\VS for the en1bedding of focu-:. consti\t!ents \vllhin larger focus
Focus: "treads on your (fool)" constituents. In her analysis of (:\~lla\. hl1!h a hc)(1k and fl,fary are
Focus Domain: VP
necessaril) in fc>cus. the fLirmer being en1bedded in the larger focus
~ parentheses
The . . d" around the design 'a tun>. .. 1.,
Lo ,... in t h c a.ssert1L1n and the consutuent sen! a hook le' .\/ar1· It i~ true th~1t the accented direct object
~OCUS 1d1nes tn. 1cate that this denot·a t um is a tor1c.il constituent \\·ithin the NP in (5_:-;93) 1s ffiL)st n,1turally interrre1ed a5 being part of the focus
;:is~ oma1n, whose presence is required f1_1r the semantic and svntactic don1ain. The ~entence .. eeni-. :.in ,1pprL1priate replv L1nly to the question
. ormedness of the predicate i\s in prev11)usly di~cussed ex«1n1ples,

------~
 A unified functional account of sentence accentuation 333
_,_,.:. l'ragrnatic relations. jocus

"What did she do?" and not '"Who did she send a book to?" The latter dog. For the purpose of the principle in (5.80), the rel~tion of the r~ferent
question seems appropriate only in the context of(5.89b). In (5.89a), both to the proposition does therefore not count as established at the ume of
accented constituents are then in focus. Within the present framework utterance. In the reading under consideration, the two nouns dog and cat
these accents are predictable on the basis of the principle in (5.80), sin~ are not focus expressions but topic expressions. The focus accents in the
the referents of both constituents have a pragmatically non-recoverable sentences are the final ones, i.e. those on parents and outside.
relation to the proposition. In the analysis I am proposing, Selkirk's recursive focus-embedding
But the pragmatic construal of (5.89a) in which both accented NPs are rule quoted in (5.72) becomes unnecessary, even [or the alternative
in the focus domain is not the only possible construal of the sentence. The interpretation of (5.89a) in which both book and Mary are in focus. In the
reason this construal is strongly preferred is that the object NP is latter case, even though book is in focus, the focus DOMAIN, which is the
indefinite and therefore most likely to be interpreted as having a "new" predicate, ls marked by the final accent on Mary alone. The accent on
(i.e. unidentifiable) referent. Since unidentifiable referents cannot serve as book is required by virtue of (5.80), since this constituent does not express
topics, the accent on the NP cannot be a topic accent. But topic construal an established topic with an active referent. In the analysis of f~
of an accented object NP is not impossible. To see that the prosodic adopted here, a clause can have only one focus domain. It would be
structure of (5.89a) is in principle compatible \vith an interpretation in redundant to mark this domain more than once.
which the direct object is in the presupposition let us compare (5.89) with
the exchange in (5.90):

(5.90) Q: Whal are you going to do with the DOG and the CAT while you're
av.ray?
A: I'll leave the DOG with my PARENTS and the CAT can stay OUTSIDE.

In the reply in (5.90) both the dog and the cal are (contrastive) topic
expressions with active referents. The only difference is that the topic is a
direct object in one clause and a subject in the other. The first clause is
about the relation between the speaker and her dog; the second clause is
about the cat only. (In spontaneous speech, the first clause might be more
natural if the direct object were topicalized- The DOG /'II fea\!e with my
P.~RENTs-but this is not an absolute requirement in English.)
In the reply in (5.90), both topic expressions must be accented, even
though their referents \Vere mentioned in the preceding sentence: since
there are two referents competing for argun1ent status in the proposition
the relation which the one chosen enters into \Vi th the proposition is non-
recoverable at the time the sentence 1s uttered. It is true that by the time
the referent "the dog" is mentioned in the first clause, the referent "the
cat" has in a sense becon1e a predictahle topic for the second clause.
However, it cannot be cnnsidert:d .in e:.t,1bl1shed top11: fro1n the point of
view of information structure Thi~ i~ sll<..>wn by the folJo,ving
observation. If the NP the (i.ll 111 the scc\inJ cL.iuse \\·ere replaced by
the anaphoric pronoun it. tht'.-> pr"n•"Jun \'.ould naturJlly be interpreted as
referring to the already estJbl1shed tc1~)1l· ~>t the preceding clause, 1.e. the