Koineization in Medieval Spanish

WDE

G
Contributions to the Sociology of Language

88

Editor
Joshua A. Fishman

Mouton de Gruyter
Berlin · New York
Koineization
in Medieval Spanish

by
Donald N. Tuten

Mouton de Gruyter
Berlin · New York 2003
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter & Co. KG, Berlin

© Printed on acid-free paper which falls within the guidelines

of the ANSI to ensure permanence and durability.

ISBN 3-11-017744-7

Bibliographic information published by Die Deutsche Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche
Nationalbibliografie; detailed bibliographic data is available in the
Internet at <http://dnb.ddb.de>.

© Copyright 2003 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin
All rights reserved, including those of translation into foreign languages. No part of this
book may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopy, recording or any information storage and retrieval
system, without permission in writing from the publisher.
Printed in Germany.
Acknowledgements

During the several years that I have worked on this book, I have been for-
tunate to receive assistance from a variety of knowledgeable and talented
scholars. Of these, I must thank explicitly John Nitti, Tom Cravens, An-
drew Sihler, and Lesley Milroy, who gave particularly helpful advice at
different stages of the project. Jim Milroy and Jeff Siegel volunteered to
read and critique parts of the manuscript, and Ralph Penny and Ray Harris-
Northall generously agreed to review the entire manuscript. The final result
is much improved thanks to their detailed and thoughtful reactions. All
shortcomings, it need hardly be said, are entirely my own.
I must also thank others for their generosity: my colleagues at Emory
University for allowing me the time off to finish this project; Mikel Val-
ladares for his untiring help with bibliography; Emilia Navarro for proof-
reading the manuscript; and my friends and family, especially Pepe, for
their patience and support.
Contents

Acknowledgements ν

1. Introduction 1

2. Koines and koineization 9

2.1. Koine and koines 9
2.1.1. The Greek Koine 9
2.1.2. Modern use of the term koine 13
2.2. Models of koineization 22
2.2.1. Siegel's stage-based model 23
2.2.2. Trudgill's process-based model 28
2.2.2.1. Accommodation and salience 29
2.2.2.2. Interdialect 36
2.2.2.3. Focusing 39
2.2.2.4. Mixing, leveling, reduction 41
2.2.2.5. Reallocation 43
2.2.2.6. Simplification 45
2.3. Elaborating the process model: Conditions of speaker 47
activity
2.3.1. Norms, norm enforcement, and social networks 49
2.3.2. Language acquisition: Adults and children 54
2.4. Observing koineization: The Milton Keynes project 63
2.4.1. Interaction of acquisition and social networks 64
2.4.2. Focusing and the time scale of koineization 67
2.5. Koineization and other contact phenomena 73
2.5.1. Pidginization and creolization 74
2.5.2. Contact between stable dialects 79
2.5.3. Dialect leveling, homogenization, uniformity 80
2.5.4. Standardization and standards 84
2.5.5. Language shift and language death 86
2.5.6. Convergence 88
2.5.7. Borrowing 88
2.5.8. Interacting processes and the study of koineization 89
2.6. The model and its use 89
2.6.1. Definition of koineization 90
2.6.2. Methodological guidelines 91
viii Contents

3. The Burgos phase 94

3.1. Social history 94
3.2. Previous discussion of language change in early Castile 104
3.3. Linguistic changes 112
3.3.1. Leveling and simplification of articles and preposition 114
+ article contractions
3.3.2. Reorganization and simplification of the tonic vowel 119
system
3.3.3. Koineization and other changes of the Burgos phase 131
3.3.3.1. The aspiration of F- 132
3.3.3.2. The development of /tf/ 136
3.3.3.3. The development of CL-, PL-, FL- 138
3.4. Conclusion 143

4. The Toledo phase 145

4.1. Social history 145
4.2. Koineization and language spread/dialect leveling 153
4.3. Linguistic changes 159
4.3.1. Leveling and reallocation in the development of 160
apocope
4.3.2. Reanalysis and the rise of leísmo 173
4.3.3. Reorganization and simplification of the possessive 204
system
4.4. Conclusion 214

5. The Seville phase 215

5.1. Social history 215
5.2. Linguistic changes 222
5.2.1. Previous discussion of the origins of Andalusian 223
5.2.2. Documentary evidence 231
5.2.3. The decline of extreme apocope 233
5.2.4. Simplification of first-person singular possessives 238
5.2.5. Leveling and the rejection of leísmo 242
5.2.6. Seseo-. A 13th-century merger? 245
5.3. Conclusion 256

6. Conclusions 257

Maps 269
 Contents ix

Notes 274
References 302
Index 332
Chapter 1
Introduction

Approaches to the study of language and language change have long been
characterized as following in one of two tracks. The focus might be the
external aspects of the status and use of language, or, in line with what was
considered more properly linguistic, the internal structural features and the
changes they undergo. Little relationship was seen to exist between the
external (social and cultural) and the internal (structural), and an exclusive
focus on the internal features of language came to characterize linguistic
research, whether synchronic or diachronic. Although some scholars ques-
tioned the value of this approach, they lacked the theoretical base to
counter the views of others who believed and believe still that the sole
object of linguistic study should be the internal structure of the language.
Belief in the autonomy of language finds its origins in certain meta-
phors that have long governed scholars' views. In the nineteenth century,
the dominant view held that language was a biological organism, which
was born, grew, decayed, and died. This belief in the independent life of
language led to its study apart from the context of its use. In the twentieth
century, this metaphor, though not at all dead, has blended with and been
superseded by others, particularly that of language as machine.1 This view,
evident in Saussure's definition of a language as a system où tout se tient,
has led to ever more precise representations of language as efficient sys-
tem. While certainly enhancing descriptions of language structure, such a
position has left little room for language change, and has led, paradoxi-
cally, to the view that language systems must alternate between perfect and
flawed states.
James Milroy (1992: 23) has argued that such metaphors have actually
hindered research on language change. Continuing the machine metaphor,
he points out that internal combustion engines are also systems, but they do
not change themselves; they can only be changed from without. Still, the
understanding that speakers do not, in general, consciously effect changes
in the linguistic system has reinforced the idea that change can only be
explained system-internally. Therefore, even when attempts have been
made to include external factors in explanations of change, the view of
2 Introduction

language as an autonomous entity has tended to impede an accurate con-

ceptualization of how external factors might contribute to change.2
In recent decades, however, sociolinguists have strongly defended the
notion that it is not languages that change but rather speakers who change
language.3 To the non-specialist, such an observation may appear self-
evident, even trivial, but the biological and mechanistic metaphors so
dominated linguistic inquiry that the importance of speakers to language
change was for a long time largely ignored. Milroy (1992: 24) points out
that linguists who have worked within a wholly system-internal approach
have made notable headway in defining the linguistic constraints on
change, but that they have been unsuccessful in dealing with the Actuation
(and Transmission) Problem: "Why do changes in a structural feature take
place in a particular language at a given time, but not in other languages
with the same feature, or in the same language at other times?" (Weinreich,
Labov, and Herzog 1968: 102). In order to answer this question, perhaps
the fundamental problem in the study of language change, the researcher
must focus on what speakers in society do with their language. To do so is
not to reject the idea of language as system, but rather to supplement this
idea with a conception of language as a tool or pool of resources that is
used by speakers for specific purposes (Milroy refers to this as "speakers'
teleology"). From this perspective, linguistic change is seen as the product
of the interplay of speakers and systems in specific and changing social
contexts.
Keller (1994) follows a logic like that of Milroy in arguing for the use
of invisible-hand explanations in historical linguistics. Keller points out
that languages are social institutions, or phenomena of the third kind, and
that they cannot be explained in the same ways that natural phenomena or
human artifacts are explained.4 Rather, like other social institutions, lan-
guages arise and change as the unintended results of collective and inten-
tional human actions (as do traffic jams, or footpaths across lawns). An
invisible-hand explanation of linguistic change will therefore be composed
of two levels: a micro-level and a macro-level. The micro-level examines
intentional speaker activity and the constraints upon that activity. The con-
straints, or ecological conditions, include intralinguistic factors such as the
linguistic varieties and the specific features associated with each, as well as
extralinguistic factors, such as sociocultural, psychosocial, cogni-
tive/learning, and biological conditions that enable and restrict speaker
behavior. Speaker (interactions, summed and viewed collectively, then
 Introduction 3

lead to consequences on the macro-level, which is constituted by the lin-

guistic structures, processes, or outcomes that need to be explained.
Of the extralinguistic factors affecting speaker activity, only those relat-
ing to social conditions are normally open to change. For this reason, Mil-
roy emphasizes the link between social change and linguistic change, and
argues that the primary task of historical sociolinguistics is the establish-
ment of explicit causal links between these two domains (Milroy 1992:
222). To the extent that patterns of social change and linguistic change co-
occur in different situations, models and theories of change can be ab-
stracted from real cases of change. In this study, I critique, elaborate and
apply a sociolinguistic model of change, koineization, which provides a
means of linking certain kinds of structural changes to a specific type of
social change. Koineization is generally considered to consist of processes
of mixing, leveling, (limited) reduction or simplification, which occur in
social situations of rapid and intense demographic and dialect mixing. The
model has been developed primarily from studies within the variationist
paradigm of new towns (e.g., Omdal 1977; Trudgill 1986; Bortoni 1991;
Kerswill 1996; Kerswill and Williams 2000) and colonial and post-colonial
language varieties, often as an extension of research on other contact varie-
ties such as pidgins and creóles. These include discussions on the origins
of different varieties of overseas Hindi-Bhojpuri (e.g., Moag 1979; Gamb-
hir 1981; Siegel 1987, 1993; Barz and Siegel 1988; Mesthrie 1993), and, to
a lesser extent, colonial English (e.g., Trudgill 1986; Trudgill, Gordon, and
Lewis 1999). Unsurprisingly, the model has been applied to language
change in other colonial contexts, including Latin of the Roman Empire
(Wright 1996) and American Spanish (e.g., Fontanella de Weinberg 1992;
Granda 1994; del Valle 1998; Hidalgo 2001).
The model of koineization represents a significant theoretical advance
for our understanding of language change as influenced by dialect contact
and mixing. Traditional historical linguists, heavily influenced by
Neogrammarian, structuralist, or generativist/formalist principles, were
loath to admit explanations based on dialect contact or mixing in any but
anomalous cases of change; the ideal explanation was either internal to the
system (and thus little more than a description) or internal to the speech
community (if this aspect was considered). Indeed, as we will see in Chap-
ter 2, it is no accident that much early work on koines was carried out by
scholars already interested in contact varieties such as pidgins and creóles,
or variationist sociolinguists, who were drawn to the study of dialect con-
tact through their study of language/dialect variation and its relationship to
4 Introduction

linguistic change. Still, even in sociolinguistics, study of dialect mixing has

only now begun to enter the mainstream. 5
Although this model has most often been associated with colonial varie-
ties, it is my contention it should be useful for linking the unique social
consequences of the medieval reconquest and repopulation of the area we
now know as Spain (phenomena with no obvious analogues in medieval
France, Italy, and England) 6 with the particular formative changes of me-
dieval Spanish (or, more precisely, Castilian).7 In fact, as I discovered as I
began to research the topic, Ralph Penny had already proposed this idea in
his brief study Patterns of linguistic-change in Spain (Penny 1987). Penny
suggested that the model of koineization (as defined by Trudgill 1986)
could and should be applied to the history of Spanish, not only colonial
varieties, as in the work of Fontanella (1992), but also peninsular Spanish
itself. Penny pointed out that the medieval expansion of Castile and Castil-
ian had occurred in a series of geochronological stages of population
movement and dialect mixing (Penny 1987: 4—7):

1. Burgos and the early county of Castile (late 9th and 1 Oth centuries)
2. Toledo and surrounding regions (from 1085 and into the 12th century)
3. Seville and the valley of the Guadalquivir (mid- and late 13th century)
4. Granada and surrounding regions (from 1492)
5. The conquest and colonization of America, and the sudden conversion of
Seville into the sole port of entry/exit for the colonies (from 1492 and into
the 16th century)
6. Judeo-Spanish, in exiled Sephardic communities of the Mediterranean (fol-
lowing the expulsion of 1492)
7. Madrid, following its designation as capital of Spain (from 1561)

According to Penny (1987), there are numerous features of Peninsular

Spanish that might be ascribed to koineization, such as the absence of pho-
nemic splits and the small inventory of phonemes relative to other Ro-
mance languages. Other such features include:

— reduction from four to three degrees of aperture for tonic vowels

— reduction of final vowels from four to three
— merger of Ibi and /β/
— loss of voicing in sibilants
— leveling from four to three verb classes
— almost complete leveling of second and third verb classes
— extreme leveling of irregular verb forms
 Introduction 5

— leveling of perfect auxiliary verb to haber

— interdialectal solution to third-person object pronouns
— merger of /j/ and I7J in Andalusia (yeísmo)
— merger of the antecedents of Isl and /Θ/ in Andalusia (seseo/ceceo)
— replacement of vosotros with ustedes in Western Andalusia
— aspiration and loss of l-sl in Andalusia
(Penny 1987: 8-17)

However, Penny (1987) kept his comments speculative; he did not attempt
to tie most of these changes to any particular stage, provide evidence for
such linking, or enter into a detailed explanation of how particular features
originated, were selected or spread. Subsequently, Penny has returned oc-
casionally to the topic (e.g., 1992, 1995, 2002) and has published a recent
volume, Variation and Change in Spanish (Penny 2000), which includes
significant discussion of the effects of dialect mixing on the history of
Spanish. In his more recent publications, Penny has begun a more in-depth
analysis of some of the problems identified by him, and has begun to an-
swer the question he himself poses in his (1987) essay: "is it possible to
observe a correlated series of linguistic levelings and simplifications, in the
way the theory predicts?" (Penny 1987: 8). To this end, he has analyzed the
origins of Judeo-Spanish (1992) and the rapid propagation of the aspiration
of etymological Iii after the naming of Madrid as capital in 1561 (2002). In
Penny (2000), he discusses still more features of Spanish that may have
arisen as a result of koineization. Still, this volume was designed as a broad
overview and introduction to variation and change in the history of Span-
ish, and Penny therefore did not aim to link specific changes to specific
periods (though he does consider some changes that might be associated
with the Burgos phase), nor to engage in detailed reconstructions of par-
ticular changes. At this point, then, the questions that Penny posed in 1987
have yet to be answered, particularly for the early medieval periods of
koineization. Indeed, while the importance of the Burgos phase (or período
de orígenes as it is known to most historians of Spanish) has long been
recognized, the sociolinguistic significance of the Toledo and Seville
phases remains undemonstrated. Moreover, Penny (2000: 5), following
Wright (1999), expresses reservations about periodization of the history of
any language, including Spanish, and so avoids reliance on the historical
schema of geochronological stages of koineization he had proposed earlier
(though he continues to suggest them as possibilities).
6 Introduction

Penny's original suggestions and unanswered questions form the start-

ing point and organizational basis for much of this study. I have chosen to
study only a limited number of features, and, in the interest of exploring its
usefulness, I have chosen to maintain the geochronological framework
outlined in Penny (1987). The number of features has been limited to only
a few for each stage (not all of which were mentioned by Penny), because
my primary intent has been to show, in a thorough and detailed manner,
how certain changes can best be explained in terms of koineization, and in
so doing to offer evidence for the validity of the model as elaborated and
defined in Chapter 2. I have limited discussion to the levels of phonology
and morphology (or morphosyntax in the case of leísmo), in part because
these are the linguistic levels that have received most attention in Hispanic
historical linguistics, but also because these have been the components of
language least obviously related to social and cultural change. I have also
limited my discussion to only the first three stages proposed by Penny -
those which began and ended in the medieval period. This is so for two
main reasons. First, the year 1500, a traditional break-off point between
medieval and modern Spanish, marks a convenient break-off point for this
study, which would otherwise require entry into the even more complex
issues surrounding the origins of American Spanish (though the last chap-
ter includes discussion which will be of interest to those studying this
topic). Second, the late 15th century marks the advent of printing and with
it the beginning of widespread effects of standardization (Harris-Northall
1996b), a process which may erase or blur the effects of koineization (see
Chapter 2). Indeed, even though application of the model represents a clear
example of what Labov (1975) has called "using the present to explain the
past", medieval Spanish also represents an ideal context and variety on
which to test and explore the explanatory power and limits of the model, in
that the competing effects of standardization - so difficult to escape in the
modern world - were largely absent. And, even though we must still con-
tend with that perpetual dilemma of historical linguistics - relative paucity
of data - medieval Spanish is a variety that has been thoroughly studied,
and for which we possess fairly plentiful documentary evidence, some-
times (particularly in the Seville phase) much more than that available for
some of the colonial/post-colonial language varieties that have been stud-
ied as koines up to now.
The study is divided into four core chapters (Chapter 2-5). In the first
of these, I begin by reviewing the history and use of the term koine, from
which the term and concept of koineization is derived. Given the recent
 Introduction 1

development of this model, it is only to be expected that consensus about

all its defining features has yet to be reached. Even so, a thorough under-
standing of the model is necessary if it is to be used to reconstruct events at
great time-depths (1100 to 700 years ago in this study). Inevitably, my own
research has led to re-evaluation of earlier proposals on koineization. I
therefore examine key studies that discuss or apply a model of koineization
(particularly Siegel 1985; Trudgill 1986; and Kerswill and Williams 2000),
synthesizing this earlier work, relating it to more general theories of lan-
guage change and pointing to possible problems, limits, and refinements. I
also propose certain methodological procedures or guidelines that should
be adhered to when using the model to analyze and explain past changes.
From there, I move to investigate, in three further chapters, the first
three periods of possible koineization and rekoineization suggested by
Penny for Castilian: the Burgos phase (Chapter 3), the Toledo phase
(Chapter 4), and the Seville phase (Chapter 5).8 In each of these chapters, I
review the social and demographic history of the period and region, and
develop sociolinguistic reconstructions of certain changes that can and
should be linked to koineization as it is defined in Chapter 2. But I have
also varied the organization of each chapter as needed. For example, the
Burgos phase has received a great deal of attention from scholars, so Chap-
ter 3 includes a review and critique of previous discussions of the linguistic
significance of this phase, as well as the medieval period in general. In this
chapter I argue that there are at least two groups of changes for which
koineization was a primary cause, but I also discuss three other changes
less clearly related to koineization, and, in the interest of exploring the
explanatory limits of the model, consider to what extent koineization may
or may not contribute to our understanding of their development.
Unlike the Burgos phase, the Toledo phase has not generally been rec-
ognized as significant to the development of medieval Castilian, so my
primary aim in Chapter 4 is to show that there are several groups of
changes that can be attributed to koineization in this period. I also consider
the relationship between koineization and the spread of Castilian features
in neighboring regions, a phenomenon which appears to have accelerated
at this time. Until very recently, the Seville phase had been largely ignored
by scholars, but during the past decade the Spanish scholar Frago Gracia
has made strident claims that many of the features that today characterize
the Andalusian dialect of southern Spain had their origin in the dialectal
and demographic mixing of the 13th century. Chapter 5 is therefore, in
large measure, a response to the work of Frago Gracia (e.g., 1993), whose
8 Introduction

research and views I consider in some detail. In fact, I challenge his argu-
ments that a key modern feature of Andalusian, seseo, arose in the 13th
century. On the other hand, I do find convincing textual evidence of other
changes not considered by Frago which support his more general claim that
the Seville phase was an important period of (rapid) change. This particular
case will illustrate the importance of adhering to the methodological guide-
lines outlined in Chapter 2 when employing the model at a great time-
depth.
Chapter 2
Koines and koineization

The model of koineization, of fairly recent development, is based on earlier

metaphorical use of the term koine. We therefore begin with an overview
of the origins and modern use of the term in the linguistic literature, and of
the confusion that its varied meanings have sometimes provoked. The pri-
mary aim will then be to define, as thoroughly as possible, what koineiza-
tion is, and what it is not. Several scholars have sought to answer these
questions, though their responses do not agree in all respects, so I have
organized the bulk of this chapter as a critical review of previous discus-
sion of koineization, with the goal of synthesizing this earlier work and my
own views. Throughout, the various facets of koineization are put in rela-
tion to other theories of language use and change, but a special section
focuses on the differences between koineization and other processes with
which it may interact in real cases of change, or be confused in scholars'
discussion of change. The chapter concludes with the definition of a proto-
typical model of koineization, and the proposal of methodological guide-
lines for application of the model.

1. Koine and koines

Koine is a term with a long history and a wide variety of interpretations. It

has its origin in the name of a variety of ancient Greek that became the
common language of the eastern Mediterranean. Subsequent metaphorical
or technical use of the term has referred to a broad range of language varie-
ties that share some or all of the characteristics of the original Greek
Koine.

1.1. The Greek Koine

The κοινή (from koiné dialektos or koinè glòssa 'common tongue') was a
mixed dialect based largely on the prestigious Attic dialect of Athens.
From the middle of the fifth century B.C., when Pericles converted the
10 Koines and koineization

Confederacy of Delos into an Athenian empire, the influence of Attic

spread rapidly throughout the Aegean. Most of the other city-states in this
empire spoke Ionic dialects (to which the comparatively archaic Attic was
closely related)9 and resented the control of Athens, but the emerging
Koine, usually referred to as Great Attic at this early stage (Bubenik 1993:
12; Horrocks 1997: 29), was useful for commerce and general intercourse
and was also employed as the (written) language of administration (Hor-
rocks 1997: 33). It has been suggested (Thomson 1960: 34; Hock 1986:
486) that a likely birthplace for the Koine was the Peiraieus, or port of
Athens, where Attic speakers and Ionic speakers from other parts of the
empire interacted, along with Doric speakers from the neighboring Pelo-
ponnesus. However, its use as written "standard" and spoken vernacular
was never restricted to the Peiraieus, since contact between Attic and Ionic
speakers occurred in a variety of contexts. The city-states in the Attic
League had to provide soldiers for the Athenian armies, as well as deal
with Athenian officials in their territories and Athenian administrative
documents composed in official Attic (Horrocks 1997: 31).10 Athens also
sent out numerous Attic-speaking colonists to the colonial territories,
where they interacted with Ionic speakers. Many speakers of Ionic also
took up residence in Athens, and through their interaction with Athenians
may have contributed to changes in the speech of "middle-class" residents
of the city. Great Attic thus developed in part as a second dialect of Ionic
speakers, but it became the native dialect for following generations in some
of the Ionic cities. Eventually, Philip of Macedón adopted Great Attic as
his language of administration and it later spread throughout the eastern
Mediterranean as a result of the conquests of his son, Alexander the Great.
The early Koine may have benefited from its ambiguous relationship to
traditional Attic; its difference from Attic may have made it more accept-
able to the dominated Ionic speakers of the empire (Hock 1986: 486),
while its similarity perhaps lent it prestige and made it acceptable to Philip
of Macedón. It has been characterized as a "de-Atticized Attic" (Hock
1986: 486) and as a "de-Atticized Ionicized Attic" (Bubenik 1993: 13). It is
interesting to note that this mixed and simplified form of Attic was decried
from the beginning as being impure and corrupt (Palmer 1980: 175), and
centuries later, under the Romans, a campaign of "Atticization" was
launched to improve it (Buck 1933: 22). The following features have often
been identified as typical of the mixed and simplified nature of the original
Koine:
 Koine and koines 11

1. Highly distinctive Attic -it- was largely replaced by the more widespread
(Ionic) equivalent -ss-, thus:
Attic Koine
glötta glossa 'tongue'
phulattö phulassô 'guard, watch'
tettares tessares 'four' (Hock 1986:486)
2. Distinctive Attic -rr- was replaced by more widespread (Ionic) -rs-:
Attic Koine
arrën arsën 'male' (Hock 1986: 486)
3. Attic -ä- (<*-ayw) was replaced by more widespread -ai-:
Attic Koine
eläa elaia 'olive' (Hock 1986: 4)
4. Dual number, a feature of Attic, was abandoned in the Koine, as in most
other Greek dialects (Hock 1986: 486).
5. Attic -eös and Ionic -êos were replaced by Doric -äos in läos 'people' and
nâos 'temple', leading to a more regular declension for these nouns (Hock
1986: 487).11
6. Pitch accent was lost, replaced by a stress accent (Thomson 1960: 35).
7. Phonemic vowel quantity was abandoned (Thomson 1960: 35) and distinc-
tive consonant length was lost (Horrocks 1997: 113);
8. The number of vowels was reduced; diphthongs became monophthongs
(Palmer 1980: 176-177).
9. Final -n was regularized in the accusative (Thomson 1960: 35).
10. The optative disappeared (merged with the subjunctive); the infinitive be-
came common in use with prepositions; the imperfect and aorist were reor-
ganized on a new uniform basis; numerous irregular verb forms were regu-
larized (Thomson 1960: 35).
11. The particle äv was replaced by a more transparent periphrasis (Thomson
1960: 36).
12. In some cases new words replaced both Ionic and Attic equivalents:
Attic Ionic Koine
naûs nëûs ploîon 'ship'
(Bubenik 1993: 15)

This list of characteristics is attractively simple and clear - deceptively so

- but not all who have used the concept and the term koine have agreed on
the features that characterized the original Koine. This has led to varying
and problematic interpretations of the term's meaning. Indeed, the great
distance between the present and the period in which these social and lin-
guistic changes occurred has made it difficult to define the features of the
Koine, much less a clear notion of how the Koine was produced. One prob-
lem has been that this temporal distance (and lists like the one above) has
12 Koines and koineization

tended to give a falsely static impression of the Koine. Many scholars ap-
pear to have conceived of it as a finite state, but in reality the Koine was
constantly developing. Palmer (1980: 177) points out that precise dating of
attestations of these changes shows that they did not all occur concurrently,
but rather appeared and spread at different times over the course of centu-
ries, along with the social and geographical spread of the Koine. For ex-
ample, Horrocks (1997: 35, 27) discusses the replacement of -tt- by -ss-
and the loss of dual number as a feature of early Great Attic (presumably
lost even earlier in a prehistoric Ionic phase of dialect mixing), but believes
the loss of the pitch accent (and with it the resultant loss of distinctive
vowel and consonant quantity) to have begun in classical times and only to
have reached completion in the (Egyptian) Koine by 150 B.C. (Horrocks
1997: 109). Indeed, many of these phenomena were attested in one or sev-
eral contributing dialects prior to the formation of the Koine itself.
Another assumption, not unrelated to the view of the Koine as a static
entity, has been that the Koine was uniform across the Hellenistic world.
However, this seems to have been true primarily of a conservative and
standardized Koine which was employed in official documents. Horrocks
(1997: 61) observes that the "very high grammatical and orthographic stan-
dards of even very ordinary 'official' papyrus documents from Egypt"
suggests that even low-ranking officials must have received rigorous train-
ing in this formal variety. On the other hand, more private documents re-
veal significant regional diversity, and there exist features of Egyptian
Koine which distinguish it from the Koine of Asia Minor, or that of Pales-
tine and Syria (Bubenik 1989: 175-252; Horrocks 1997: 60-64).
With regard to the causes of these changes, Thomson (1960: 35)
seemed to assume that the extension of Greek to non-native speakers
played a role, but he offered no further details. Others have seen the
changes that resulted in the Koine as examples of "normal" development.
Indeed, Buck dismissed out of hand the possibility that the changes in the
Koine were in any way unique:
But mixture in vocabulary is common to most of the present European lan-
guages. There were also changes in pronunciation, in syntax, and in the
meaning of words, similar to the changes that have taken place in the other
European languages. (Buck 1933: 22)

Buck was partially correct in making these assertions, but, as will be dis-
cussed below, there is reason to believe that there are distinct though gen-
eralizable processes which led to the formation not only of the original
Koine but also of many other language varieties that share similar histories
 Koine and koines 13

of dialect mixing and demographic movement. More recently, scholars

such as Bubenik (1993) and Horrocks (1997: 41) have come to view the
changes which characterize the Hellenistic Koine, especially in such new
urban centers as Alexandria, Antioch, and Pergamum, as arising from the
combined effects of top-down imposition of the Koine by the ruling dynas-
ties (favoring overall uniformity), dialect leveling resulting from the mix-
ing of the dialectally heterogeneous immigrant masses from old Greece,
and imperfect acquisition by indigenous populations of the language un-
dergoing koineization (favoring interregional diversity).12 Horrocks makes
the following useful observation on the issue of uniformity and variation in
the Koine:
It is essential, then, to see the Koine not only as the standard written and
spoken language of the upper classes (periodically subject to influences
from belletristic classical Attic), but also more abstractly as a superordinate
variety standing at the pinnacle of a pyramid comprising an array of lower-
register varieties, spoken and occasionally written, which, in rather different
ways in the old and the new Greek worlds, evolved under its influence and
thereafter derived their identity through their subordinate relationship to it.
(Horrocks 1997: 37)

1.2. Modern use of the term koine

Modern metaphorical or technical use of the term koine has grown as

scholars have attempted to identify commonalities between (the develop-
ment of) the original Koine and other language varieties. In fact, this has
been only too easy to do, as different scholars have identified different
features of the original koine as being key to its nature. The meanings as-
signed to metaphorical uses of the term koine became increasingly diverse
as use of the term grew during the 20th century. According to Cardona
(1990: 26), modern use dates from Meillet's (1913) discussion and analysis
of the original Koine. Meillet reported three meanings for the term: for
Hellenistic Greeks, the language of everyday use; for Hellenistic gram-
marians such as Apollonius Dyscolus, the language of reference for use in
grammars, and possibly the base from which new dialects arose; for mod-
ern Hellenists, the base for modern Greek (Meillet [1913] 1975: 253-275).
Meillet suggested that it was easier to define the structure of a koine by
what is was not (the dialectal features it lacked) than by what it was,
thereby establishing the problematic notion that koines are merely the
14 Koines and koineization

"least common denominator" of contributing varieties (see below). He also

emphasized the long and apparently punctuated development of the Koine:
la κοινή n'est pas une langue fixée, ce n'est pas non plus une langue qui
évolue en obéissant régulièrement à certaines tendences; c'est une langue où
il y a une sorte d'équilibre, constamment variable, entre fixation et évolu-
tion. (Meillet 1975: 256)

Most importantly, however, Meillet argued that the features of the Koine
were not unique to it, and suggested that Vulgar Latin, among other lan-
guages, showed a similar history of social expansion and structural reduc-
tion (Meillet 1975: 257).13 Meillet's discussion thus identified the useful-
ness of "ce terme commode et nécessaire", as he calls it, and thereby
initiated its more general use as a means of categorizing language varieties.
Jakobson ([1929] 1962: 82) was another early user of the term, and ob-
served that dialects which serve as vehicles of communication in large
areas and gravitate towards the role of koine (by which he seemed to mean
lingua franca; see below) tend to develop simpler systems than dialects
which are restricted to local use (these ideas were further explored in An-
dersen 1988). Despite such early use, the term apparently remained highly
specialized and rarely used until the second half of the 20th century
(Cardona 1990: 27). Cardona offers as another early example the following
passage from Tagliavini's Origini delle lingue neolatine·.
Probabilmente il francone, parlato alle corti dei re merovingi e carolingi, era
una lingua mista, una specie di koiné formato da elementi franchi salì e
franchi ripuarì, nonché da elementi romanzi e germanici assai vari.
(Tagliavini 1949: 206)

Talgiavini uses the term to refer to a variety that results from the mixing of
not only related but also unrelated languages, thus employing it in a way
that seems justified only in the broadest sense (i.e., if the feature of mixing
is the only one picked out by the metaphor; but see below for discussion of
the potential impact of non-native speakers).
Although not all scholars would use the term with such liberty, it has
nevertheless received a tremendous variety of interpretations in the linguis-
tic literature. Siegel (1985) argues that this is so because the original Koine
had six different features which scholars could highlight (or ignore) in
making comparisons. According to Siegel, the Koine:

— was based primarily on one dialect

— had features of several dialects
 Koine and koines 15

— was reduced and simplified14

— was used as a regional lingua franca
— was a standard
— was nativized to some extent (Siegel 1985: 358-9, 362)

In order to determine the dominant interpretations of the term, Siegel ana-

lyzed references to some 36 language varieties as koines (Siegel 1985:
359):

1. Literary Italian (Pei 1966: 139)

2. Church Kikongo [Congo] (Nida and Fehderau 1970: 152)
3. Standard Yoruba (Bamgbose 1966: 2)
4. Bahasa Indonesian (Pei 1966: 139)
5. High German (Germanic Review 1 (4): 297 [ 1926])
6. Bühnenaussprache [Stage German] (Dillard 1972: 302)
7. Hindi (Hartmann and Stork 1973: 123)
8. Latin in the Roman Empire (Hill 1958: 444)
9. Belgrade-based Serbo-Croatian (Bidwell 1964: 532)
10. Mid-Atlantic koine [England] (Times Literary Supplement, 22 April 1965)
11. Network Standard English [U.S.A.] (Dillard 1972: 302)
12. Melanesian Pidgin (Ervin-Tripp 1968: 197)
13. Fourteenth-Century Italian of Naples (Samarin 1971: 134)
14. Town Bemba (Samarin 1971: 135)
15. Fogny [Senegal] (Manessy 1977: 130)
16. Kasa [Senegal] (Manessy 1977: 130)
17. Congo Swahili (Nida and Fehderau 1970: 152)
18. Lingala [Congo] (Nida and Fehderau 1970: 153)
19. 'Interdialects' of Macedonian (Lunt 1959: 23)
20. Koineized colloquial Arabic (Samarin 1971: 134)
21. Ancestor of modern Arabic dialects (Ferguson 1959a: 616)
22. Vernacular of north China, seventh to tenth centuries (Karlgren 1949: 45)
23. Calcutta Bazaar Hindustani (Gambhir 1983)
24. Israeli Hebrew (Blanc 1968: 237-51 )
25. Eighteenth-century American English (Traugott 1977: 89)
26. Fiji Hindustani (Siegel 1975: 136; Moag 1979: 116)
27. Trinidad Bhojpuri (Mohan 1978)
28. Guyanese Bhojpuri (Gambhir 1981)
29. Surinam Bhojpuri (Gambhir 1981: 184)
30. Mauritian Bhojpuri (Gambhir 1981: 184)
31. Slavish [U.S.A.] (Bailey 1980: 156)
32. Italian-American (Haller 1981: 184)
33. Slave languages [Caribbean] (Dillard 1964: 38)
16 Koines and koineization

34. English-based nautical jargon (Hancock 1971: 290n)

35. Black Vernacular English (Mühlhäusler 1985: 8)
36. Canadian French (Gambhir 1981)

Siegel reports that very few of these language varieties could be said to
have all the properties of the original Koine, and he found wide variation in
the meanings assigned to the term itself. Studies 1-22 used the term to
refer to a lingua franca (any variety used for intergroup communication);
studies 1-11 used it to refer to regional standards. A majority of the studies
indicated that several dialects must contribute to the formation of a koine.
Only a few studies included reference to a base dialect, reduction and sim-
plification, or to nativization (Siegel 1985: 362).
Though Siegel restricted himself to studies published in English, his
general conclusions appear valid for studies published in other languages
as well. Still, further variation in meaning does crop up. For example, Ro-
mance philologists have long used the term koine to describe certain me-
dieval literary varieties, such as the Provençal of the Troubadours and the
"Sicilian" dialect of the court of Frederick Π, praised by Dante in De Vul-
gari Eloquentia (Elcock 1960: 399, 459). These varieties certainly show
mixing and the elimination of dialect features, but they appear to have been
the result of conscious selection and limited to use in writing by a tiny
elite. They have also been labeled, perhaps more appropriately, literary
standards (Elcock 1960: 455).15 In Italian linguistics, the term has also been
used to describe certain (probably spoken) regional varieties that arose
from the Middle Ages around principal urban centers (e.g., Venice, Turin,
Milan, Genoa, Naples, Palermo). This use follows those that emphasize
dialect mixing, use as lingua franca and/or regional standard. More re-
cently, koine has also come to be used as a sociolinguistic label for a cer-
tain level in the dialect continua that characterize most regions of Italy
(Berruto 1989: 13). Pellegrini ([I960] 1975: 37) divided these continua
into four levels: dialect, regional koine, regional Italian, Italian standard.
The regional koines are thus seen as distinct from the regional standards,
but their lingua franca function remains significant, as does, at least for
some authors (e.g., Cardona 1990), the mixing, reduction, and simplifica-
tion of dialect features.
Given such wide variation in actual usage, it is unsurprising that explicit
definitions of the term have also varied widely. The following give some
idea of this variation (some of these are quoted in Siegel 1985):
 Koine and koines 17

— "a form of language resulting from a compromise between various dialects

and used as a common means of communication over an area covering all
the contributing dialects." (Graff 1932: xxxvii)
— "a compromise among several dialects" used "by a unified group in a self-
contained area within a larger linguistic area". Pei also considers a koine to
be a planned language: "a deliberately sought sublimation of the constituent
dialects rather than an unconscious and accidental merger". (Pei 1966: 139)
— "Koine is the term for a 'common' dialect which lacks prominent features
of the more conventional dialects of a language. It is the end result of dia-
lect levelling." A koine is often considered "good" speech in the language
and is most often a standard dialect. (Dillard 1972: 302)
— "KOINES. A standard normally has its origin in the dialect of some particu-
lar territory, which comes to enjoy superiority over those of neighboring
regions, for non-linguistic reasons (usually political, less often economic or
social, never purely literary). Such a favored dialect comes to be the com-
mon language or KOINÉ . . . used throughout its region, where it is usually
comprehensible to most of the speakers of the neighbouring dialects. In the
course of its spread, the koiné retains its basic relationship to the dialect on
which it is based, but takes in features from related dialects, as in the in-
stance of Span, /xuérga/ juerga 'spree' from Andalusian . . . or French fa-
bliau 'animal-fable' from Picard (Φ ONFr. fablel 'little fable')."
(Hall 1974: 104)16
— "The spoken language of a locality which has become a standard language
or lingua franca." (Crystal 1992)

Though a more precise definition of the term has been developing since the
publication of Ferguson (1959a), widely varying interpretations still
abound, even in more recent studies such as those in Sanga (1990) and
Knecht and Marzys (1993), where, for example, the terms koine and stan-
dard are frequently conflated.
The different interpretations given to the term have produced a situation
in which its use often produces more confusion than clarity. Siegel (1985:
363) sets out to resolve this problem by specifying a technical meaning for
the term. He claims that the concept of dialect mixing is fundamental, and
specifies that the contributing varieties must be language varieties that are
either a) mutually intelligible or b) share the same genetically-related su-
perposed language (1985: 375-376). 17 These may include regional dialects,
sociolects, and "literary dialects". For the last category, Siegel based his
claim on the development of Israeli Hebrew, which Blanc describes as a
18 Koines and koineization

result of the mixing of "a variety of literary dialects, several substrata, and
several traditional pronunciations" (Blanc 1968: 238-239). But this defini-
tion raises the problem of "non-native" speakers in the demographic mix:
should learner interlanguages be included among the contributing varieties
of a koine? The impact of non-native speakers has also been identified as
important to the development of the Hellenistic Koine (e.g., Horrocks
[1997] reports Coptic substrate features in Egyptian koine texts) and the
early Arabic koine (Ferguson 1959a). Mesthrie (1994: 1865) defends their
potential importance in the development of any koine, since the variants of
native speakers of unrelated languages are less likely to be perceived as
"foreign" in the mixed linguistic pool of the prekoine (cf. LePage 1992).
However, certain constraints need to be placed on this broad view of con-
tributing varieties, at least for prototypical cases. First, adult interlanguage
features may form part of the pool, but these speaker-learners must have
easy access to input and interaction with native speakers. This in turn im-
plies that such "foreign" speakers do not form a majority in the commu-
nity, since their dominance would reduce the likelihood of their obtaining
sufficient access to the language (varied though it may be). Thus, the range
of contributing varieties or subsystems must be expanded to include inter-
language varieties of second language learners.18
Siegel also warns that many of the definitions given to the term koine
are either too broad or too narrow. Thus, using koine as a synonym of lin-
gua franca or common language robs it of usefulness, as does restricting
koine to the meaning of "planned, standard, regional, secondary" variety or
one based primarily one dialect. Perhaps more controversially, Siegel's
explanation could be read as favoring a close identification between koines
and standards:
unplanned, nativized, or transported languages may be koines if they exhibit
the mixing of any linguistic subsystems such as regional dialects, literary
dialects, and sociolects. However, although a koine may or may not be a
formal standard, it is implicit in all definitions that a koine has stabilized
enough to be considered at least informally standardized. (Siegel 1985: 363)

In reality, Siegel meant socially-based language norms rather than the codi-
fied language norms that characterize standard languages, and Siegel
(1987: 201) clarifies this issue by abandoning use of the term "informal
standardization". The definition might also be improved by emphasizing
that prototypical koines not only may be but necessarily are unplanned,
nativized, and transported varieties (see below).
 Koine and koines 19

According to Siegel, most koines are characterized to some extent by

reduction and simplification, though he comments:
requiring a koine by definition to exhibit these features would be too restric-
tive, as the amount of reduction or simplification may differ between koines
according to both the conditions under which they developed and their cur-
rent developmental stage. (Siegel 1985: 363)

Recent research (e.g., Kerswill and Williams 2000) shows that there are
cases of koineization without obvious examples of simplification; this is
due to the pre-existing similarity between the contributing varieties, in
which most variation is allophonic. Mohan (unpublished paper; reported in
Siegel 1985: 361-2) points out that koines are of two types: those based on
dialects with great structural similarity (such as that studied by Kerswill
and Williams), and those based on more highly differentiated dialects.
While I think these "types" have to be viewed as extremes on a scale,
greater difference between the contributing dialects can be expected to lead
to greater perceived simplification in the resultant koine. On the other
hand, Siegel's reference to the "current developmental stage" is problem-
atic, since it implies that a koine, once formed, continues to be in some
way identifiable as a koine; as will be emphasized below, koines are only
identifiable in a historical sense.
Siegel concluded his discussion of koines with the following definition:
a koine is the stabilized result of mixing of linguistic subsystems such as re-
gional or literary dialects. It usually serves as a lingua franca among speak-
ers of the different contributing varieties and is characterized by a mixture
of features of these varieties and most often by reduction and simplification
in comparison. (Siegel 1985: 363)

The claim that a koine normally serves as a lingua franca requires some
qualification. A koine would only serve as a lingua franca for non-native
speakers, since for native speakers it would serve as a primary (perhaps
even sole) means of communication. The function of lingua franca may be
important in the development of regional koines. Siegel explains that:
a regional koine usually results from the contact between regional dialects of
what is considered to be a single language. This type of koine remains in the
region where the contributing dialects are spoken. (Siegel 1985: 363)

Petrini (1988: 34, 42) points out that regional koines with no native speak-
ers can be extremely unstable, varying from speaker to speaker and from
situation to situation, and may be no more than an abstract perception of
20 Koines and koineization

the linguist who observes the frequently similar results of multiple

accommodations by speakers. Petrini claims too that a regional koine as a
clearly distinct variety is likely to arise only as it gains native speakers,
most often in urban centers, who serve to stabilize the norms of the koine.
This regional koine is then used as a lingua franca by speakers of rural
dialects, but such use is secondary to its use by native speakers.19
The notion of koine as lingua franca is more problematic in the case of
immigrant or colonial koines. According to Siegel, an immigrant koine:
may also result from contact between regional dialects; however, the contact
takes place not in the region where the dialects originate, but in another lo-
cation where large numbers of speakers of different regional dialects have
migrated. Furthermore, it often becomes the primary language of the immi-
grant community and eventually supersedes the contributing dialects.
(Siegel 1985: 364)

In this case, it seems that the lingua franca function would only exist for a
short time, until the speakers of the contributing dialects die off. After that,
all or most speakers of the koine are native speakers. However, there is a
larger issue here: emphasis on the use of koines as lingua franca may re-
veal an assumption that koines develop primarily in order to facilitate clear
communication. This is a partly valid assumption in the case of language
subsystems that are sufficiently different to impede mutual comprehensibil-
ity (as seems to have been the case in many socially subordinate koines,
such as those used by workers in the Bhojpuri-Hindi diaspora), but most
dialects are in fact mutually comprehensible (or become so quickly with
interaction), so effective communication cannot be identified as the only or
even the most important factor in koine formation. This issue is discussed
in greater detail below.
Another valuable effort to define koine as a technical term is that of
Mesthrie (1994). Mesthrie, like Siegel, analyzes modern use of the term in
relation to the original Koine, for which he identifies four key features:

— (a) its development as a new, common variety based on existing dialects of

the language (common is taken in the sense of "shared");
— (b) its use as a common (or "vulgar") medium of communication between
speakers with different first languages or speakers from different dialect ar-
eas;
— (c) its use as the standard/official language of a politically unified region;
 Koine and koines 21

— (d) changes in its structure on account of its wide use as both first and sec-
ond language (involving a synthesis of these at some stage).
(Mesthrie 1994: 1864)

For modern uses, Mesthrie explains that in one stream of thinking, the for-
mal criteria of (a) and (d) are considered primary, and in another, the
functional properties of (b) and (c) are considered primary. Mesthrie re-
jects (b), (c) and (d) as criteria for definition of koine:
The major objection to (b), (c), or (d) alone as a defining criterion is that on
its own each defines a language variety or linguistic process that has a well-
established label: (b) is synonymous with lingua franca (and the process of
language spread); (c) is better described as 'standardization'; and (d) de-
scribes the phenomenon of substrate influence in second language acquisi-
tion or in language shift. (Mesthrie 1994: 1864-1865)
Mesthrie identifies (a), or the incorporation of features from several (re-
gional) varieties of a single language, as the only necessary feature of a
koine (however, see below for consideration of the impact of language
acquisition). In effect, Mesthrie rejects the synchronic functions - lingua
franca or standard - as defining features of a koine, and accepts only those
aspects that are essentially diachronic in nature, resulting from the process
of dialect mixing:
While the processes involved in koineization are of considerable interest to
the linguist, once a koine has formed there may be nothing to distinguish it
from older dialects of the language. (However, subordinate immigrant
koines do often show a significant reduction in inflections.) Generally, the
designation koine might be appropriate at a particular stage in the history of
the language, but loses significance once the variety becomes established as
the first language of a new generation. Like any other natural language a
koine may in time develop new regional subdialects, as shown by the history
of Greek. (Mesthrie 1994: 1865).
Hence, koine has become, in its technical sense, merely a convenient label
for those language varieties and states that result from the social and lin-
guistic processes of koineization.20
22 Koines and koineization

2. Models of koineization

Most recent discussions of koines have shifted from a focus on the resul-
tant state to a focus on the processes of koine formation. Though Samarin
(1971) was the first to use the term koineization, others before him had
already begun to shift focus to the diachronic study of koine formation.
Ferguson's (1959a) study of the Arabic koine, which he claimed was the
common base for modern spoken dialects of Arabic, was essentially an
exercise in reconstruction of a stage of the language. He attributes the for-
mation of this variety (perceived as uniform) to "a complex process of
mutual borrowing and leveling among various dialects", while most of the
14 features he discusses show some sort of loss, reduction or simplifica-
tion. Given the time depth of this study and the lack of documentary evi-
dence, no further study of processes was possible. Blanc (1968) argued that
modern Israeli Hebrew was "gradually given a definite shape by a slow
'koineizing' process drawing on several pre-existing sources . . . Usage had
to be established by a gradual and complex process of selection and ac-
commodation which is, in part, still going on, but which now has reached
some degree of stabilization" (Blanc 1968: 238-239). Samarin (1971) was
only indirectly concerned with koineization, but he suggested use of the
term as a means of differentiating a unique process, distinct from dialect
leveling or borrowing, that leads to the formation of a new dialect. Samarin
(and Dillard 1972: 300) also emphasized that koineization involves the
suppression of localisms or prominent stereotypable features as speakers of
different dialects mix together in new social contexts, particularly in cases
of migration.
None of these studies engaged in detailed discussion of the process or
model of koineization. However, the growth of studies of pidginization and
creolization also drew scholars' attention to other types of colonial/post-
colonial languages, among them the numerous varieties that arose as a
consequence of the Bhojpuri-Hindi diaspora. In the aftermath of the aboli-
tion of slavery, European colonial powers shipped hundreds of thousands
of Indian peasants on indentured contracts to other colonies. The immi-
grants spoke primarily genetically-related Indie languages from the north,
but in some cases there were also speakers of Dravidian languages from the
south. The Indie varieties included dialects of Bhojpuri, Avadhi, other
eastern and western varieties of Hindi, Bengali, Rajasthani, Panjabi and
Calcutta Bazaar Hindustani, with widely varying degrees of mutual com-
prehensibility between the different varieties. In each colony, a compro-
 Models of koineization 23

mise variety arose that was used as an in-group language among the Indian
laborers. The areas where these new dialects have been identified and stud-
ied include Fiji, Surinam, Natal (South Africa), Trinidad, Mauritius and
Guyana (Siegel 1988a; Mesthrie 1993: 26-29).

2.1. Siegel ' s stage-based model

As these different varieties received more scholarly attention, efforts to

define common principles of koineization began to appear. An early effort
is that of Gambhir (1981), but the most well-known and influential in this
tradition is that of Siegel (1985), who was investigating the development of
Fiji Hindi (or Hindustani; e.g., 1975, 1987, 1988b). In his (1985) study, he
synthesizes notions of koines and koineization from other studies in order
to arrive at a technical definition of koine (reported above) and a more
precisely defined model of koineization, based on his own findings and
that of others. His model is based on a sequence of four possible stages of
koineization:

1. Prekoine. "This is the unstabilized stage at the beginning of koineization.

A continuum exists in which various forms of the varieties in contact are
used concurrently and inconsistently. Levelling and some mixing has be-
gun to occur, and there may be various degrees of reduction, but few forms
have emerged as the accepted compromise."
2. Stabilized Koine. "Lexical, phonological, and morphological norms have
been distilled from the various subsystems in contact, and a new compro-
mise subsystem has emerged. The result, however, is often reduced in mor-
phological complexity compared to the contributing subsystems."
3. Expanded Koine. A stabilized koine "may become a literary language or
the standard language of a country. This extension of use is often accom-
panied by linguistic expansion, for example, in greater morphological
complexity and stylistic options."
4. Nativized Koine. "A koine may become the first language for a group of
speakers . . . This stage may also be characterized by further linguistic ex-
pansion (or elaboration), but here some of it may be the result of innova-
tions which cannot be traced back to the original koineized varieties."
(Siegel 1985: 373-374)

Siegel emphasizes that not all these stages need necessarily occur in any
particular case of koineization, and provides examples of such variable
development (see Table 1). Siegel consciously modeled this presentation
24 Koines and koineization

on then-current approaches to the study of pidginization and creolization,

and borrowed his paradigm (Table 1) from the developmental continuum
for pidgins and creóles (Table 2) constructed by Mühlhäusler (1980: 32).
At first glance, the proposed relationship seems eminently reasonable,
since pidgins, creóles, and koines all result from language contact and
demographic mixing, and they are often found in colonial or post-colonial
regions. However, as Siegel himself has argued in later work (see below),
there are significant differences between pidginization/creolization and
koineization, and these differences underlie some problematic implications
of the (1985) stage-based model.

Table 1. Developmental contunua of koines. Source: Siegel (1985: 375).

prekoine prekoine prekoine

4 i ι
i stabilized koine stabilized koine
4 4 I
J, J. expanded koine
4 I 4
nativized koine nativized koine nativized koine
(Fiji Hindustani) (Guyanese Bhojpuri) (Greek Koine)

Table 2. Developmental contunua of creóles. Source: Siegel (1985: 375), based on

Muhlhausier (1980: 32).21

jargon jargon jargon

4 4 4
4 stabilized pidgin stabilized pidgin
4 4 4
4 4 expanded pidgin
4 4 4
creole creole creole
(West Indian English (Torres Strait Creole) (TokPisin)
Creole)

Pidgins are generally understood to result from contact between typologi-

cally distant varieties, while prototypical koines (such as Spanish) result
from contact between linguistic subsystems that show high degrees of mu-
tual intelligibility. Since speakers in a koineizing context can usually un-
 Models of koineization 25

derstand each other, the need to communicate information - which plays

the key role in pidginization - cannot be a generalizable primary motive for
alterations in speaker production (though it may play a greater role when
contributing varieties show greater structural differences, as in the case of
Fiji Hindi). More importantly, prototypical pidgins and creóles arise in
very specific social circumstances in which speaker-learners are separated
from models - native speakers of the target variety - at the same time that
they must communicate with the socially-distant model speakers or, more
importantly, other speakers with whom a common language is not shared.
The social context of koineization could not be more different, for the
speaker-learners, be they native speakers of a related dialect, second lan-
guage learners, or children, must be assumed to have easy access to abun-
dant, if highly variable, input. Indeed, within the Thomason and Kaufman
(1988) model, koineization is properly categorized as change with lan-
guage maintenance (but see below).
As a result, it becomes difficult to accept an unintended implication of
the stage-based model: that both pidginization and koineization are charac-
terized not only by mixing (which remains undefined but which we may
assume means the appearance of features from several source dialects in a
resultant koine) and simplification, but also by reduction/impoverishment.
Siegel borrows Mühlhäusler's (1980: 21) definitions of simplification ("an
increase in regularity or a decrease in markedness") and reduction ("a de-
crease in the referential or non-referential potential of the language"). But
while simplification is indeed a linguistic process of koineization, reduc-
tion of this sort cannot be, for reduction as defined here includes the ex-
treme structural/lexical reduction of pidgins, which makes full comprehen-
sion difficult or impossible outside of contexts of direct oral
communication, where gestures, intonation, and the possibility of clarifica-
tion substitute for structural complexity. Since learners in a koineizing
situation are not deprived of input, there is no reason for extreme reduction
to occur.22
If extreme reduction does not occur, then there is no need for a stage of
structural expansion, which in pidgins is associated with the expansion of
functions and/or creolization of an existing pidgin; this occurs as the pidgin
is extended to use in new communicative contexts, and it therefore requires
new vocabulary and more systematic marking of grammatical relations to
make it less context dependent and more fully functional as a primary
means of communication. Since no radical reduction such as that affecting
pidgins is present in koineization, expansion must be reinterpreted in the
26 Koines and koineization

context of koineization. It is certainly true that any expansion of contexts

of use is likely to require an expansion of the lexicon, but this is true of any
language that acquires new functions. In fact, Siegel exemplifies this third
stage with the use of the koine as a literary or standard language. The in-
clusion of standardization is not entirely unwarranted, for it reflects a fre-
quent reality: koines tend to be selected as standards, since standards also
require the minimal variation in form that characterizes koines. However,
while standardization does include a process of elaboration of the lexicon
and syntax, particularly of written language (Haugen 1966: 933; Lodge
1993: 26), this is not quite the same as expansion in pidginiza-
tion/creolization, which includes especially an increase in morphological
complexity. Moreover, standardization may enter into competition with
koineization. For instance, Fontanella (1992: 42-54) argues that in the
history of American varieties of Spanish, standardization has sometimes
(partially) impeded koineization, as in the interior of Mexico, and some-
times reversed its effects, as in Buenos Aires (see below). Rather than in-
cluding processes such as lexical expansion or standardization within a
model of koineization, it is probably best to see them as interacting with
koineization.
Another problem with the parallel stages of pidginization and koineiza-
tion is the timing and significance of nativization. In pidginization, relative
stabilization of grammar and lexicon may occur before nativization (as in
the well-known case of New Guinea Tok Pisin), but pidgins are not native
languages, and structurally they are very simple and therefore easily
learned by adults; nevertheless, they are relatively unstable with regard to
phonology, since each speaker's version will be affected by his/her native
language phonology. In large measure, it is the nativization stage of creoli-
zation that leads to full stabilization of a pidgin (although the question of
how many and what types of features need to be stabilized in order to con-
sider the variety stabilized will depend largely on the perspective taken).
Here then, there is a problem with Mühlhäusler's original proposal. In fact,
while adults do play important roles in the selection of features, nativiza-
tion by children is probably key to full stabilization or focusing of a koine
(Mesthrie 1994: 1866; Kerswill 1996). Petrini (1988: 42), as mentioned
above, argues that the developing koine or prekoine of the Italian region of
Ticino has so far failed to stabilize because there are no native speakers of
this variety, and Kerswill (1996) argues that in a koineizing community the
first signs of the new koine will become evident among the older members
of the first generation of children (see below).
 Models of koineization 27

In later publications, Siegel has discussed the similarities and differ-

ences between koineization and creolization that I highlight here (e.g.,
Siegel 2001), and he has accepted many aspects of Trudgill's work into his
own approach (Siegel 1993b), but such reformulations have not led to a re-
evaluation of the original stage-based model, which is still frequently cited
in the literature (e.g., Kerswill and Williams [2000] situate their own dis-
cussion within this framework). The adherence to the pidginization para-
digm which underpins the four-stage model may reflect the particular his-
tory of Fiji Hindi and other post-colonial language varieties, since some of
these varieties show the interaction of koineization with other processes.
For example, Siegel himself (1987: 196) points out that one of the contrib-
uting varieties of Fiji Hindi was itself a pidgin, so we should not be sur-
prised to find some effects of pidginization in the resultant koine (also
suggested by Trudgill 1986: 106). Of course, the challenge to defining
koineization is the necessity of distilling a prototypical model from com-
plex and varied cases of real change.
Siegel's early model has helped to draw the attention of scholars to the
study of koineization and from it we can retain useful insights, including
the concept of prekoine (the highly variable, diffuse, even chaotic, initial
stage of demographic and dialect mixing); the importance of stabilization
or focusing of new norms; the definition of contributing varieties as mutu-
ally intelligible linguistic subsystems (which allows inclusion of child and
adult learner language); and the recognition of the interaction of koineiza-
tion with other processes. The definition of koineization in terms of these
four stages does not appear, however, to be the best means of furthering
our understanding of the process. In fact, the generalizable stages of
koineization can probably be limited to just two: the prekoine, character-
ized by the co-existence of numerous varieties and variants, and the stabi-
lized (or focused) koine, which has become the native language variety of
at least some speakers; these two stages can of course be separated by a
period of variable length of norm selection and enforcement (e.g., Trudgill
[1998] and Kerswill [2002] suggest three stages based on progressive fo-
cusing over three generations; see below). Nevertheless, the restriction to
two basic stages (or three generational stages) does not preclude the de-
scription of stages of development in a particular language variety, in
which case one could refer to the interaction of koineization with processes
such as pidginization or standardization. Siegel makes an observation that
has important implications for application of the model to actual language
histories, and in particular to that of Spanish:
28 Koines and koineization

It should be stressed that the developmental continuum of a koine is not

necessarily linear. At any stage, for example, 'rekoineization' can take place
if there is continued contact with the original closely related varieties, or ad-
ditional contact with different ones. (Siegel 1985: 375)
For those varieties that have seen repeated phases of koineization, the attri-
bution of particular changes to particular phases will become an added
objective of historical sociolinguistic research (see Chapters 3, 4, and 5).

2.2. Trudgill's process-based model

The single most influential discussion of koineization is that of the socio-

linguist Peter Trudgill, who was drawn to the topic by his broader interest
in contact between dialects and how this affects language variation and
change. His primary discussion of koineization appears in his (1986) book
Dialects in Contact, which includes analysis of a range of issues associated
with contact between individual speakers of dialects, contact between sta-
ble dialects, and dialect mixing in koineization. In the final chapters of this
book, Trudgill defines a model of koineization and tests it on the history of
certain varieties of colonial English, particularly Australian English.
Trudgill's model is based on the isolation of the processes that occur dur-
ing koineization rather than stages of development. The following is his
summary of what happens during koineization:
In a dialect mixture situation, large numbers of variants will abound, and,
through the process of accommodation in face-to-face interaction, interdia-
lect phenomena will begin to occur. As time passes and focusing begins to
take place, particularly as the new town, colony, or whatever begins to ac-
quire an independent identity, the variants present in the mixture begin to be
subject to reduction. Again this presumably occurs via accommodation, es-
pecially of salient forms. This does not take place in a haphazard manner,
however. In determining who accommodates to whom, and which forms are
therefore lost, demographic factors involving proportions of different dialect
speakers present will clearly be vital. More importantly, though, more
purely linguistic forces are also at work. The reduction of variants that ac-
companies focusing, in the course of new-dialect formation,23 takes place
via the process of koinéization. This comprises the process of levelling,
which involves the loss of marked and/or minority variants; and the process
of simplification, by means of which even minority forms may be the ones to
survive if they are linguistically simpler, in the technical sense, and through
which even forms and distinctions present in all the contributory dialects
 Models of koineization 29

may be lost. Even after koinéization, however, some variants left over from
the original mixture may survive. Where this occurs, reallocation may oc-
cur, such that variants originally from different regional dialects may in the
new dialect become social-class dialect variants, stylistic variants, areal
variants, or, in the case of phonology, allophonic variants.
(Trudgill 1986: 126; italics in original)
An important first observation must be made here: Trudgill explicitly
equates koineization only with the "more purely linguistic forces" of level-
ing, simplification, and reallocation. Though he recognizes these as signifi-
cantly related to speaker activity, we will see that his early view of koinei-
zation is both enhanced and limited by this focus on linguistic outcomes.
What follows is a discussion of each of the particular features of koineiza-
tion highlighted in the above passage.

2.2.1. Accommodation and salience

Trudgill borrows the concept of accommodation from the work of the so-
cial psychologist Howard Giles, who developed what is known as Speech
(or Communication) Accommodation Theory. According to Giles (1973:
90), "if a sender in a dyadic situation wishes to gain the receiver's ap-
proval, then he may adapt his accent patterns towards that of this person,
i.e., reduce pronunciation dissimilarities".24 This process is known as ac-
cent convergence. Its opposite is likely to occur when a speaker wants to
dissociate or signal disapproval.25 Accommodation may affect any linguis-
tic level (e.g., lexicon, syntax, morphology, phonology, as well as speaking
rate and style), and it is hypothesized to be a universal tendency of human
behavior (Trudgill 1986: 2).26
Trudgill's emphasis on accommodation reveals rather novel assump-
tions about why dialect contact leads to change. Given that most contribut-
ing varieties in a prekoine linguistic pool are mutually intelligible at least
to some degree, many of the alterations in speech that take place are not
strictly speaking necessary to fulfill communicative needs, although some
comprehension difficulties may occur (Trudgill 1986: 1). Rather, speakers
accommodate to the speech of their interlocutors in order to promote a
sense of common identity. This focus on the identity-marking function of
language (cf. Milroy 1992, 1993) is critical to an understanding of how and
why koineization occurs; it will be further discussed below.
30 Koines and koineization

Though accommodation is generally studied as a phenomenon affecting

the immediate performance of individuals (or short-term accommodation),
Trudgill identifies and analyzes what he terms long-term accommodation,
which results in a relatively permanent effect on the speech patterns of an
individual. Trudgill sees this as the basic mechanism underlying linguistic
change in two-dialect contact situations and in koineization resulting from
multidialectal contact. Making long-term accommodation the driving force
for linguistic change in situations of dialect contact has further implica-
tions. First, there are important differences between short-term and long-
term accommodation. Most social psychologists up to the time of
Trudgill's writing had focused on speech accommodation as stylistic varia-
tion, where the speakers presumably have pre-existing knowledge of the
variants they use. For example, Trudgill himself adduced data from his
Norwich studies which showed that his informants accommodated with
markers (variables subject both to social class and stylistic variation) but
did not do so with indicators (variables subject only to social class varia-
tion) which were not characteristic of their own speech; presumably they
lacked the knowledge to produce these forms. However, long-term accom-
modation places a greater burden on the speaker:

accommodation beyond the speech community will often be a rather differ-

ent process from accommodation within it. Accommodation within the
speech community, as in my Norwich interviews, involves altering the fre-
quency of the usage of particular variants of variables over which the
speaker already has control. Accommodation beyond the speech community,
on the other hand, may well involve the adoption of totally new features of
pronunciation. (Trudgill 1986: 12)
At this point speakers appear to enter into a process of language or dialect
acquisition, the significance of which is explained by Beebe and Giles:
All speakers have, to some extent, a limited linguistic repertoire, both in
their native languages and in any second languages they might speak . . .
Limitations in repertoire are central to research within SLA [second lan-
guage acquisition], whereas they are peripheral in studies of native speakers
. . . It is important, when extending social pyschological theories to SLA
data, that limitations in repertoire be considered. For it is the tension be-
tween limitations in ability to converge toward a native-speaking interlocu-
tor and motivation to converge that makes second-language data unique.
(Beebe and Giles 1984: 22-23)

Nevertheless, Trudgill hesitates to associate long-term accommodation

with second language acquisition as such. In part, this may be because of a
 Models of koineization 31

possible tendency on the part of speakers to approach dialect and language

learning differently (particularly in cases of contact between standardized
languages). If the dialect to be learned is viewed by speakers as simply a
different variety of the same language, then they may take a more piece-
meal approach to adopting or rejecting features of the target variety. If, on
the other hand, they believe the target to be a separate system, they are
likely to adopt strategies more typical of second language acquisition, and
attempt to learn the new variety as a whole. In Nordenstam's (1979: 24)
study of Swedish women in Norway, it was found that some of the women
perceived Swedish and Norwegian as separate languages and tried to keep
them apart, while others adopted features on a one-by-one basis, believing
Swedish and Norwegian to be varieties of one language. Still, no matter
what strategy a learner of a similar variety adopts (piecemeal or whole),
errors due to constraints imposed by the acquisition process are likely to
crop up, so a useful distinction is unlikely to stand on this criterion.27
Chambers (1992: 675-676) also prefers to maintain a distinction be-
tween long-term accommodation and second dialect acquisition. Chambers
bases his distinction on the behavior of the informants in his study of sec-
ond dialect acquisition. These were Canadian children/adolescents (aged
9-17) who had moved to England and were adopting features of Southern
English English. Though Chambers is himself Canadian, the children did
not accommodate toward him by using fewer British variants and more
Canadian ones. For Chambers then, the innovations in the children's
speech represented "irrepressible acquisitions rather than ephemeral ac-
commodations". This may be true, but the important point is that the chil-
dren had extended the newly adopted English features from their most
careful style to even their most colloquial vernacular style (Ellis 1985: 95).
Chambers seems to want to restrict the use of long-term accommodation to
those cases in which the speaker continues to shift between different ac-
cents in different situations and styles. This may be a useful distinction, but
it is a distinction of degree rather than kind, and Ellis makes the point that
acquisition of a feature in a learner's careful style is probably the first step
toward acquisition in all styles. Moreover, the examples of long-term ac-
commodation given in Kerswill (2002: 682-685) would seem to confirm
this analysis, since they show that the speech of adults engaged in long-
term accommodation is characterized by simplifications and great intra-
speaker variability, as is predicted in Ellis' model. I return to the issue of
acquisition below, but from here on I assume that long-term accommoda-
tion forms part of a process of dialect acquisition (although, in the koineiz-
32 Koines and koineization

ing context, what is being acquired is generally a set of the most fre-
quently-occurring features, rather than a pre-existing variety; see below).
If second dialect learners (those engaged in long-term accommodation)
do not learn all features, which features do they modify and/or learn? First,
it has been shown that (adult) speakers in more stable dialect contact situa-
tions tend to accommodate (to the extent that they can or want to) to the
most salient features. Following Nordenstam (1979), who was studying
contact between speakers of two different varieties, Trudgill (1986: 11-27)
argued that the most salient features are those which represent differences
in the lexicon and morphology (no mention is made of syntax; see below).
But, aside from the lexicon, it is in the phonology that dialect differences
are most consistently found, and this is where Trudgill focuses his atten-
tion. After analyzing evidence of accommodation from a variety of studies,
he suggests that the following factors contribute to salience (these are
summarized and listed as follows in Kerswill 1994: 154):

1. Phonological contrast. For example, English English speakers who relocate

to the US often adopt the contrasting phoneme /ae/ for their native /a:/ in
words of the dance class.
2. Great phonetic difference. English English speakers in the US tend to
adopt the American [A] for their native [D] in words of the hot class, and
Americans in England do just the reverse.
3. Naturalness. English English speakers in the US begin to flap intervocalic
Ν more quickly than Americans in England acquire the intervocalic [t] for
their flap, since flapping is a kind of "natural" weakening.

However, Trudgill also indicates a number of factors which appear to im-

pede accommodation and either reduce or override salience:

4. Extra-strong stereotyping. For example, accommodating Northern English

English speakers tend not to adopt the Southern English English /a:/ for
their native /ae/ in dance-class words, despite the phonological difference
(cf. Number 1 above).28
5. Phonotactic constraints. For example, English English speakers in the US
tend not to acquire American non-prevocalic /-r/ as in cart, bar (even
though the phonological difference is evident) since in most Southern Eng-
lish English dialects syllables of the shape VrC or Vr# are not allowed.
6. Homonymie clash. For example, English English speakers may avoid the
American pronunciation [hat] since it sounds too much like their own pro-
nunciation of heart.
 Models of koineization 33

These few examples make it clear that long-term accommodation cannot be

seen as an entirely predictable process, since accommodation sometimes
occurs or fails to occur for no obvious reason, and is also influenced by
cultural factors, such as in cases of stereotyping. Kerswill (1994b: 154-
155) is quick to point out the circularity of the notion of salience, for it is
defined precisely in terms of those features which are or are not adopted.29
Still, Kerswill too is unwilling to abandon salience, since it seems that
speaker-learners do find some features more "striking" than others. In a
recent and very detailed discussion of this very issue, Kerswill and Wil-
liams (2002) consider the value of numerous factors that have been pro-
posed as contributing to salience, and test several of those proposed by
Trudgill (1986) for their value in explaining the spread of features in dia-
lect leveling in southeast England. In the end, they conclude that language-
internal factors are an essential pre-condition for salience; these include
phonological contrast, great phonetic difference, internally-defined natu-
ralness (which favors weakening or loss of distinctions), semantic trans-
parency, and particular syntactic and prosodie environments (2002: 105;
see below). All these factors are likely to draw speakers' attention to par-
ticular features. However, no single linguistic factor, or set of factors, is
capable of guaranteeing salience. They suggest therefore that it is extra-
linguistic conditioning which is crucial in determining salience: cognitive,
pragmatic, interactional, social psychological, and sociodemographic fac-
tors. As such, salience cannot be predicted, and particular reasons for its
existence must be sought in each case, with special attention to relations
between speakers, their attitudes, and their sociocultural contexts of inter-
action.
In the discussions of Trudgill (1986) and Kerswill and Williams (2002),
salience is seen as the primary factor determining which features are ac-
commodated to and, therefore, which features become more frequent. This
of course is absolutely necessary in non-koineizing situations (how else
can one explain changes in frequency of particular variants?). However, in
nearly all discussions of koineization, it is argued that majority forms win
out in the final dialect mix (e.g., Trudgill 1986; Kerswill and Williams
2000). This argument rests on the assumption that frequency and consis-
tency of use of particular forms and form/function combinations play the
fundamental role in determining what forms are acquired, and this fre-
quency is determined primarily by the original demographic mix (Siegel
1997: 139). Indeed, the importance of frequency as a factor distinct from
salience is shown by Trudgill, Gordon, Lewis, and Maclagan (2000), who
34 Koines and koineization

provide clear evidence that most features of New Zealand English selected
during koineization are features which represented majority use among the
sum total of original settlers (a claim which is further supported by the
selection of similar majority forms in other Southern-Hemisphere Eng-
lishes which saw similar population mixes).30 When frequency and consis-
tency are seen as primary, salience is invoked as an explanation for those
features which cannot be explained as the result of greater frequency alone
(in some cases salience can plausibly be viewed as contributing to a fea-
ture's perceived frequency). It would seem therefore that older children
and adolescents - those responsible for the formation of a new koine (see
below) - do accommodate to each other and learn from each other, but that
over time it is the most frequent forms that are consistently favored in this
process (rather than any particular target variety). In a koineizing context,
salience as discussed by Kerswill and Williams (2002) is probably most
important for exceptional adaptations in adult, adolescent, or child speech,
which can then alter the frequency of certain variants and thereby affect the
learning of children and adolescents.
Nevertheless, the potential cumulative effects of perceptual and cogni-
tive salience should not be discounted. For example, Siegel (1997: 139)
emphasizes that stressed words are both phonetically and cognitively sali-
ent. Of the perpectives on perceptual salience reviewed by Kerswill and
Williams (most post-date Trudgill's early work), the most important is that
of Yaeger-Dror (1993: 203-206), who, following a comprehensive review
of studies of cognitive and phonetic factors which contribute to salience,
comes to the following conclusions about universally salient positions:
the beginning of a syllable, word, or sentence is most salient. A vowel nu-
cleus or intervocalic position is also salient (because more acoustically
prominent), as is the semantic nucleus (the focus) of the sentence. The coda
(of a syllable, of a word, or of a sentence) is the most redundant and least sa-
lient. (Yaeger-Dror 1993: 206)

Although, as Kerswill and Williams (2002) indicate, it is impossible to

make predictions about outcomes in particular cases, it may be that percep-
tual salience shows its influence in long-term trends. If speaker-learners are
most likely to learn perceptually salient items (other things being equal)
and to follow "natural" (ease-of-articulation) tendencies, then we would
expect a language variety which has undergone (repeated) koineization to
show a preference for CV syllable structure. In fact, this is exactly what we
find in the history of Spanish. Similarly, Ohala, in a review of acoustic
phonetic studies, finds that "place cues are generally less salient in VC
 Models of koineization 35

rather than in CV transitions and are even less salient in VN transitions"

(Ohala 1993: 159). If the effects of such "CV dominance" are combined
with articulatory weakening in syllable-final position, then we might also
expect a heavily koineized variety to show weakening of syllable-final
nasal consonants, with assimilation to the place of articulation of a follow-
ing consonant. In fact, Spanish (or most varieties of Spanish) is literally a
textbook-case of the regular assimilation of nasals to following consonants.
These same factors may have influenced the loss of non-prevocalic /-r/
in New Zealand (as in car, cart). Trudgill, Gordon, Lewis and Maclagan
(2002: 120-124) claim that a majority of immigrants to New Zealand were
rhotic (though many were probably variably so), while those of the second
generation were also rhotic (30% completely so, while most others were
variably so). However, New Zealand is today almost completely non-
rhotic. This case can be partly explained by a certain frequency of loss of/-
r/ in the prekoine linguistic pool, but the complete loss of non-prevocalic /-
r/ must certainly have been favored by its perceptually non-salient position.
Moreover, as these authors point out, the same phenomenon occurred and
was carried to completion at roughly the same time in Australia, South
Africa, and London itself, areas which were all characterized by both dia-
lect mixing, variable use of this feature, and this structural/cognitive factor.
As these authors argue, it is the repeated combination of similar factors and
structures which underlies the common "drift" of these dialects.
Finally, it is important to note that Kerswill and Williams (2002: 103)
extend the concept of salience (and the potential impact of perceptual sali-
ence) to discussion of syntax and discourse. Specifically, they analyze rec-
ognition of the (often commented upon) like, used as a focus marker and
quotative before key segments of utterances (e.g., Like Wow!), and com-
pare it to the clause-final discourse marker like (e.g., I did that when I got
home, like), which shares some of the functions of the more notorious like.
While participants in their study showed a uniformly high recognition rate
of the first, they showed a corresponding low recognition rate for the sec-
ond. While the first is spreading around the world, the second is showing a
decline in use. Following Yaegor-Dror, Kerswill and Williams suggest that
one reason for the low recognition rates and decline of clause-final like is
its lack of prosodie prominence.31
Finally, it is important to discuss briefly one last aspect of Trudgill's
(1986) presentation of accommodation: its limitation to adults. In later
work, Trudgill highlights the importance of adult speakers in situations of
dialect contact and mixing (particularly in regard to simplification; see
36 Koines and koineization

below). While certainly not ignoring the importance of child language ac-
quisition and use, to which he attributes focusing, he does not pay much
attention to child language when discussing the sources of innovations that
become established as changes in a koine. While it seems unquestionable
that accommodation by adults is important in the process of koineization,
in purely logical terms it need not be so: adults could refuse to alter their
speech patterns in a situation where speakers of different dialects were
mixed together, but the process of koineization (with all its features) would
occur unhindered so long as children learned from their parents and older
siblings, and then accommodated/learned among each other, as they gener-
ally do (there is evidence of this in the Milton Keynes study; see below).
Indeed, Mesthrie has argued that the primary importance of accommoda-
tion by adults during koineization is that it leads to a neutralization of the
social meaning attached to the linguistic variants affected, since variation
ceases to correlate clearly with non-linguistic factors such as region, social
status, or style (Mesthrie 1994: 1866). The question of the relative impact
of adults and children will be returned to below.

2.2.2. Interdialect

Features may appear in new dialects arising from dialect contact that are
not present in any of the contributing dialects; Trudgill (1986: 62) identi-
fies these as interdialect features. His thorough discussion of interdialect
focuses on the result of contact between two stable dialects, and he pro-
vides numerous examples of new phonetic and lexical interdialectal forms
as well as novel grammatical form-function reanalyses (presumably new
functions for established forms would also be interdialectal, but these can
also be analyzed in the overlapping category of reallocation). With regard
to the first type, Trudgill claims that there are two principal types of inter-
dialect phonetic features: fudged and mixed. In a fudged dialect, speakers
exposed to two different pronunciations opt for a phonetically intermediate
sound; in a mixed dialect, speakers substitute one sound for another, gen-
erally through the process of lexical diffusion, in which words are trans-
ferred one-by-one from one lexical set (with a characteristic pronunciation)
to another. It should be pointed out that fudging and mixing as contrasting
concepts are uncontroversial in the phonetic component of grammar: the
blending or analog nature of the phonetic system allows fudging, but the
 Models of koineization 37

phonology, morphology, and syntax are generally considered discrete com-

binatorial systems where only mixing is an option.32
Other types of mixed interdialect features result from "incomplete ac-
commodation". For example, Trudgill identifies the hybrid lexical form
/juba/ 'to work', used by residents of Oslo who immigrated from Sunndal,
Norway, as a cross between the older Oslo Norwegian form /jaba/ and the
Sunndal form /jub/ (Trudgill 1986: 63). Form-function reanalyses (cf. Croft
2000) lead to interdialect forms that may not be intermediate in any simple
or straightforward way. A prime example of this is the development of
forms (and corresponding functions) of the verb to do in the contemporary
English of younger residents of Reading, England (Cheshire 1982, reported
in Trudgill 1986: 65). Here, interaction has occurred between the tradi-
tional Reading dialect, in which different do forms were used to distinguish
between main and auxiliary do, and Standard English, where do forms are
used to distinguish person and number (see Table 3).

Table 3. Form-function reanalysis of Reading do. Source: Trudgill (1986: 65).

Standard English Original Reading Younger Reading
I do it, do I? I dos/does it, do I? I does/do it, do I?
He does it, does he? He dos/does it, do he?, He do/dos/does it,
do/does he?

In the process of shifting towards Standard English usage, some younger

speakers have created the novel construction he do it. In the traditional
Reading dialect, do is restricted to use as an auxiliary form, but as speakers
accommodate towards the standard, they begin to extend its use to mark
the 1st and 2nd persons singular and plural, as well as 3rd person plural. In
a typical case of overgeneralization, they extend it to the third person sin-
gular, thereby using it in a way that it is never used in either the standard or
in the original Reading dialect (however, in this case it remains a minority
usage). It is clear from this example that interdialectisms are often simply
hypercorrections. Trudgill gives a straightforward example of Northern
British English hypercorrection toward the perceived (largely southern)
standard; for the word but, northern speakers "correctly" shift their /but/ to
southern-standard /bAt/, but they hypercorrect when they shift butcher
from northern and southern /but/a/ to /bAt/a/.
Interdialect can also include statistical Labov-hypercorrection, in which
speakers use a variant more frequently than they normally would (Trudgill
38 Koines and koineization

1986: 123). Such frequency-enhancing changes in production are probably

the most significant interdialect influence on outcomes in koineization.
Among such changes are hyperdialectalisms, yet another category within
Trudgill's taxonomy of interdialect. These are forms (or form-function
reanalyses) that come to be seen as markers of community identity and
whose use is therefore altered or exaggerated (which in turn may further
contribute to the identity-marking function). As an ideal example of hyper-
dialectalism, Trudgill refers to Labov's (1963) well-known study of Mar-
tha's Vineyard. Labov showed that the islanders who identified strongly
with the local community and wished to remain there had more centralized
realizations of the diphthongs /ai/ and /au/, already centralized in the local
dialect, than those who did not so identify. These islanders took a pre-
existing feature of the local dialect and exaggerated the difference between
their articulations and the uncentralized ones of the mainland as they re-
acted to and resisted the penetration of mainland norms and identity.
Trudgill's identification of interdialectal features in the prekoine lin-
guistic pool is an important addition to the model. Still, his discussion of
interdialect is limited almost entirely to the effects of contact between sta-
ble dialects. Though on some occasions he claims to see the development
of interdialect features as important to multidialectal contact situations
(and gives one example from Fiji Hindi in his discussion of koineization,
as well as a few potential cases of phonetic fudging and mixing), he also
states that "genuinely intermediate forms are . . . less likely" in "dialect
mixtures where more than two contact varieties are involved" (Trudgill
1986: 65). This may be partly true, but it has the unfortunate effect of rein-
forcing the traditional belief that koineization is nothing more than a kind
of reduction to a least common denominator (in Chapters 3 and 4 I explore
the origins of several features of Castilian which show that the introduction
and propagation of interdialectalisms may on occasion lead to the survival
of apparently marked forms in a koine).
Finally, we should note that the concept of interdialect is adapted from
Selinker's (1972) concept of interlanguage. This is a fundamental construct
for understanding of second language acquisition, and its integration into
Trudgill's model also calls into question any meaningful distinction be-
tween long-term accommodation and dialect acquisition. In addition, I
would argue that there is reason to maintain the original term and concept
of interlanguage in discussion of koineization. When Trudgill uses the term
interdialect to refer to novel forms and uses of speakers, these speakers are
necessarily assumed to be native speakers of one of the dialects in contact.
 Models of koineization 39

This is very likely a reasonable supposition for contact between two stable
dialects spoken by communities with close-knit social networks. However,
if we accept the definition of contributing varieties given above, it will be
necessary to include the features of interlanguages of "non-native" learners
of the "language" (i.e., the collection of dominant forms and features in
contributing varieties), since, in the mixed linguistic pool and social situa-
tion of the prekoine, the features of their speech may be less identifiable as
"foreign" and thus find easier acceptance in the developing linguistic
norms of the new community (cf. LePage 1992; see below for further dis-
cussion of interlanguage). Interlanguage can thus be used as a cover term
to refer to both interdialect and interlanguage per se.

2.2.3. Focusing

Trudgill discusses Siegel's concept of stabilization within the framework

of focusing, a concept closely associated with the research of LePage and
Tabouret-Keller (e.g., 1985):
Le Page and Tabouret-Keller have pointed out. .. that speech communities,
and therefore language varieties, vary from the relatively focused to the rela-
tively diffuse. The better known European languages tend to be of the fo-
cused type: the language is felt to be clearly distinct from other languages;
its 'boundaries' are clearly delineated; and members of the speech commu-
nity show a high level of agreement as to what does and does not constitute
'the language'. In other parts of the world, however, this may not be so at
all, and we may have instead a relatively diffuse situation: speakers may
have no very clear idea about what language they are speaking; and what
does and does not constitute the language will be perceived as an issue of no
great importance. (Trudgill 1986: 85-86)
Many factors influence the degree to which a language variety is consid-
ered focused or diffuse. Primary among them is a common cultural identity
shared by the members of a speech community. Numerous factors can con-
tribute to the development of a common identity and thus to focusing. In
modern nation-states, standardization (with nationalism) is clearly an
important cause of focusing, as the highly focused European languages
attest. Isolation from outside influences and competing national or ethnic
identities will also aid in the definition of a common identity and
corresponding linguistic norms. On the other hand, the effects of
continuing contact without the creation of a new clearly-defined common
identity are found in the classic diffuse language situation of Belize (cf.
40 Koines and koineization

classic diffuse language situation of Belize (cf. LePage and Tabouret-

Keller 1985), long a colony with little national heritage, and where an un-
standardized Creole competes with highly focused English and Spanish.
In cases of dialect mixing (at least), Trudgill also sees isolation as an
important factor, along with the effects of child language acquisition. He
suggests that focusing of koineized varieties will be more likely and there-
fore easy to study in cases of language transplantation to new towns, re-
gions or colonies which are distant from the source communities. One of
the clearest and best known examples is offered by the new town of
Hayanger (Omdal 1977; Kerswill [2002] describes two similar cases stud-
ied by Sandve 1976). Hoyanger is a Norwegian town that in 1916 had a
population of only 120 inhabitants, but through industrialization grew to
have 950 in 1920 (but only 3000 by 1986). In 1920, 28% of the population
came from the immediate vicinity of H0yanger, 32% from other parts of
the surrounding county (split into two dialect zones), and the other 40%
came from other parts of Norway. Over the course of three generations
(first = adult immigrants), the Norwegian spoken in the town moved from
non-systematic variation that reflected the origins of the original immi-
grants, to more limited variation among the second generation, to a new
relatively unified and distinctive set of norms that characterize the speech
of the grandchildren (see below for further discussion). Similarly, Bortoni
(1991) includes a detailed description of the speech of young people of the
new Brazilian capital of Brasilia. Brasilia was founded in 1960 (in its sec-
ond generation at the time of Bortoni's study), and the speech of young
people there shows indications of incipient focusing and stabilization of
new norms (see below for more recent discussion of focusing).
It would thus appear that child language acquisition must play a signifi-
cant if not critical role in focusing or stabilization, but Trudgill (1986) does
not direct further attention to this issue.33 In fact, his only mention of the
social causes of this focusing is the following:
The focusing process may have been aided by the fact that, while workers
and managers originally lived in different parts of the town, this is today no
longer the case. As a consequence, there is little social dialect differentiation
amongst the youngest generation. (Trudgill 1986: 96).

Of course, as Trudgill himself pointed out, he was primarily interested in

the linguistic aspects of focusing, rather than the particular social factors
that lead to it. Nevertheless, it has become abundantly clear in subsequent
years that focusing and rate of focusing can only be understood if attention
is paid to the importance of child learning in focusing/stabilization as well
 Models of koineization 41

as the development and solidification of peer groups and social networks

(as also suggested in Trudgill's later publications; see below).

2.2.4. Mixing, leveling, reduction

Trudgill uses these terms to refer in different ways to the effects of what is
essentially the primary linguistic process that accompanies focusing of the
koine. Mixing emphasizes the selection and incorporation of linguistic
features from different dialects into the resultant koine. Leveling, on the
other hand, is used to focus on the exclusion of certain features, and
Trudgill defines it as the "reduction or attrition of marked variants" present
in the initial dialect mixture (where marked is used to refer to forms that
are in the minority in the set of contributing varieties). It is reduction not
with reference to any one contributing variety, but rather to the sum total of
all variants in all contributing varieties, including not only the pre-existing
features of established dialects but also the novel features of interdia-
lects/interlanguages (Trudgill 1986: 97).34 He identifies the primary source
of mixing/leveling/reduction as the accommodation by adults who suppress
marked features of their speech. Clear examples of this can be found in Fiji
Hindi (Table 4), where the surviving forms are those which were shared by
a majority of contributing dialects (or, rather, used by a majority of speak-
ers).

Table 4. Lexical mixing in Fiji Hindi. Source: Trudgill (1986: 101).

Standard Bhojpuri Avadhi Fiji Hindi
Hindi
'what' kja: ka: ka: ka:
'who' kaun kaun~ke kaun~ko kaun
'one's own' apna: a:pan a:pan apan

Trudgill supposes that "commonness" is the primary reason for the survival
of any linguistic feature. In fact, the organization of the above chart reflects
a perhaps too simple metric:
As far as levelling is concerned, we can note that. . . forms that occur in a
majority of the contributing dialects win out and survive in the emerging fo-
cused dialect. (Trudgill 1986: 143)
42 Koines and koineization

Siegel (1993b: 116) agrees with this general idea but criticizes Trudgill ' s
exact conceptualization, and argues that it is the overall frequency of oc-
currence that most matters, with this being initially dependent on the total
number of speakers who use any particular form, be they from the same or
different dialects. In more recent work (Trudgill, Gordon, Lewis, and
Maclagan 2000), Trudgill and his colleagues follow Siegel's suggestion,
and even find that the large size and cohesiveness of any particular group
(its possible use as a target) is unimportant in determining the final results
of koineization: those features that are used by the maximum number of
speakers are those that are generally selected.
Trudgill's original emphasis on number of dialects, as opposed to
speakers, may have reflected a earlier preference for emphasizing systems
over speakers (but note that it sometimes represents a necessity in histori-
cal work at great time depths, where more specific information may not be
available). Trudgill (1986) does recognize, however, that speakers must
have the opportunity to use variants they know. He points out, for example,
that socially marked variants, even very common ones, are likely to be
eliminated along with the contexts of use for which they were reserved. A
clear example of this phenomenon is found in Guyanese Bhojpuri (Gamb-
hir 1981: 255). Whereas Indian Bhojpuri is characterized by a complex
pronoun and suffixation system that is used to indicate respect for the inter-
locutor, in the new social environment of Guyana, the immigrants, largely
of the same caste (and in a situation requiring mutual support and solidar-
ity), no longer had need of the inter-caste cultural conventions of the caste
system or the linguistic forms that were associated with that system. As a
result, forms marking the respect feature were lost in modern Guyanese
Bhojpuri.
When minority forms survive, Trudgill suggests that this is because
other factors interfere, but these exceptions are often difficult to explain.
Sometimes interdialect forms are the ones chosen, particularly when there
is no clear majority form (Trudgill 1986: 101). In other cases, ad hoc ex-
planations become the only recourse; Trudgill provides an example where
the shortest of several competing forms was chosen; in another case, forms
with oral vowels were selected over forms with phonologically-marked
nasal vowels (Trudgill 1986: 102). In many cases, however, there is no
clear explanation of why a particular form was chosen, though perhaps
further analysis would reveal such factors in these cases. Siegel (1993b)
also considers Trudgill's inability to deal with such exceptional cases:
 Models of koineization 43

it seems that there would have been greater chance for success if he had
based his generalizations not primarily on linguistic characteristics (compar-
ing forms from different dialects) but also more on the demographic charac-
teristics he mentions, including information about the social context.
(Siegel 1993b: 116)
Attention to social context has in fact been the component of Trudgill's
original model most in need of further development, although Trudgill
does not completely ignore it (see below for further discussion).

2.2.5. Reallocation

Sometimes not all the variants (presumably only pre-existing variants in

this case) in the prekoine linguistic pool are leveled or reduced to just one.
Trudgill (1986) makes the important observation that variants that had the
same function/meaning in different established contributing varieties may
survive, but the function/meaning of each variant is altered, or reallocated,
so that they no longer compete directly. Thus, in socio-stylistic reallocation
(Britain and Trudgill 1999), originally regional, social or stylistic variants
may acquire new functions as social, stylistic or areal variants. Trudgill
also suggests that different phonetic variants in the prekoine linguistic pool
may be retained as allophonic variants in a koine, in a process known as
phonological reallocation (Britain 1997a; Britain and Trudgill 1999).
Particularly frequent is the reallocation of regional variants to a stylistic
function; the examples in Table 5 are taken from Mauritian Bhojpuri.

Table 5. Stylistic reallocation in Mauritian Bhojpuri. Source: Domingue (1981),

reported in Trudgill (1986: 65).

eastern central western high Mau- low Mauri-

India India India ritius tius
'big' /bara:/ /ba^a:/ /ba^a:/ /bara:/
'temple' /mandir/ /mandil/ /mandil/ /mandir/
'road' /ra:hta/ /ra:sta/ /ra:sta/ /ra:hta/

Siegel (1997: 127) provides evidence of a similar occurrence in Fiji Hindi,

in which the different variants appear to mark social class. In this case,
there exist two forms of the third person possessive, /okar/ from the Bho-
jpuri dialects, and /uske/ from Hindi varieties. The Hindi form has become
44 Koines and koineization

the norm in Fiji Hindi, while the Bhojpuri form is considered rustic and
employed in drama and comedic routines to characterize rustic characters.
Socio-stylistic variation can affect phonological variants as well. For ex-
ample, Britain and Trudgill (1999) report that in Australia the original Brit-
ish pronunciations of the vowel in the lexical set dance, sample, plant have
been reallocated with new stylistic functions. Original northern British /ae/
has become the lower-status norm in Australian English, while original
southern English /a:/ is retained as a high-status variant in formal registers.
Reallocation can also lead to the creation of novel areal variants in a koine.
Siegel (1997: 127) points to a case in which a dialect difference in India
becomes a marker of island residence in Fiji Hindi. Speakers who live on
the northern islands of Vanualevu and Taveuni insert a back glide before
the perfective suffix -ä while speakers from the main island of Vitilevu
insert a front glide, so that 'sang' is gäwä on the northern islands and gäyä
on the main island.
Cases of phonological reallocation are less easy to identify, but Trudgill
(1986), Britain (1997a), and Britain and Trudgill (1999: 251-254) have
argued that Canadian Raising (and similar phenomena found in the English
Fens and other English dialects that have arisen through dialect mixing) is
the result of phonological reallocation during koineization. In Canadian
Raising, the /ai/ diphthong is subject to allophonic variation. It is pro-
nounced with an open onset before voiced consonants and word-finally
(e.g., time, tie), but with a centralized onset (e.g., [ai]) before voiceless
consonants (e.g., night). The wide diphthongs are and were typical of
southern English dialects, while the narrow diphthongs are and were typi-
cal of Scottish dialects (among others). According to Trudgill and Britain,
as speakers from these areas mixed in Canada, their allophones were re-
tained, but according to regular phonetic principles of English: the shorter
narrow diphthongs were confined to the pre-voiceless consonant environ-
ment where all English vowels have shorter allophones; the broad diph-
thongs were retained elsewhere.
Trudgill (1986) also adduces as evidence of phonological reallocation
the wide range of phonetic variants that characterize the articulation of /a/
for many working-class speakers in Belfast, many of whom have rural ori-
gins and moved to the city to find work. He argues:
the very wide range of allophones of /a/ in working-class Belfast English,
found also in the case of some other vowels, is typical not so much of ver-
nacular varieties [as James Milroy argues] as of mixed varieties. My sugges-
tion in fact is that different phonological variants present in the dialect mix-
 Models of koineization 45

ture situation, and not leveled out during focusing, may be retained as allo-
phonic variants. (Trudgill 1986: 125)
Milroy (1982, 1992) holds very different views on the matter, and suggests
that it is the close-knit social networks of the inner-city neighborhoods that
allow them to retain and add variants in their speech, which serve as mark-
ers of identity against other groups and against middle class speakers, who
have much more loose-knit social networks and who generally show far
more restricted allophonic variation for /a/. In fact, Trudgill (1986) treats
the case of Belfast and its developing city-wide vernacular(s) as if it were
the same type of process as that which occurred in Heyanger (or Milton
Keynes), but, as the Milroys' research shows, this is not quite the case. For
koineization to occur, there must be a massive influx of immigrants at
roughly the same time, with a simultaneous breakdown in social networks.
This is not what occurred in the history of Belfast, where the in-migration
was more gradual. Moreover, in mixed, koineized dialects, we would ex-
pect the general tendencies toward reduction and simplification to operate
against the survival of high numbers of allophones.35 Thus, although
Trudgill (1986) views reallocation as a process distinct from leveling, it
may be better to understand it as a sub-process of mixing or leveling (or of
acquisition; see below), since in general, as the above examples show,
there is still reduction in the number of regional and social variants, but
more than one survives.

2.2.6. Simplification

Trudgill considers simplification to be an alternative process of variant

reduction (as against leveling), and it is frequently cited as a primary result
of koineization.36 Simplification, however, is a term that has been used by
many scholars with a variety of meanings. Trudgill provides the following
brief review and definitions:
Simplification is . . . a difficult and perhaps dangerous notion. Mühlhäusler
has argued (197[4]) that simplification can be taken to refer to 'an increase
in regularity', and that it is a term that should be used relatively, with refer-
ence to some earlier stage of the variety or the varieties in question. There
are, according to Mühlhäusler, two main types of simplification. The first
type involves an increase in morphophonemic regularity, and would include
the loss of inflections and an increase in invariable word forms. Mühlhäusler
also points to Ferguson's (1959[b]) discussion, where the following are
46 Koines and koineization

listed as indications of simplification: symmetrical paradigms; fewer obliga-

tory categories marked by morphemes of concord, and simpler morphopho-
nemics. The second aspect of simplification involves an increase in the
'regular correspondence between content and expression', which is intended
to refer to an increase in morphological and lexical transparency.
(Trudgill 1986: 103)
Though we will see cases in which it is necessary to distinguish between
regularity and transparency (at least when different levels are involved; see
below), they most often work in tandem, and the range of surface effects
listed above can generally be seen as part of a more general phenomenon:
the reduction in inventories of units and rules within the most systematic
components of the language (phonology, morphology, and syntax). The
elimination of irregular forms and exceptional rules can easily be seen to
increase both regularity and transparency.37 For example, the loss during
the Middle Ages of Spanish strong preterite forms and their replacement by
predictable forms results in greater regularity and transparency: e.g., es-
crise Ί wrote' has become regular escribí, cf. infinitive escribir, mise Ί
put (in)' has become regular metí, cf. infinitive meter (Penny 2000: 52
gives numerous other examples). Nevertheless, there also exist cases in
which regularity in one component, such as phonology, can lead to greater
complexity (opacity) in another, such as the morphology (this is a well-
known Neogrammarian principle), and this seems to be what Mühlhäusler
and Trudgill are referring to when they specify the two types of simplifica-
tion. For example, Kerswill and Williams (2000) show that the Ro-
land/rolling distinction, a case of relatively complex allophonic variation
(not found in all contributing dialects) has been maintained during koinei-
zation in Milton Keynes (the vowel of the open syllable of each word is
pronounced with a different allophone), in part because its maintenance
allows for a more transparent relationship to be maintained between the
morpheme roll and its dimorphemic derivative rolling (pronounced with
the same vowel; Britain [1999] presents a similar kind of argument; see
below for detailed discussion).
The reduction in inventories of units and rules is clearly the result of
imperfect language learning (see below). It is favored by discrepancies
between contributing dialects, which lower frequency and consistency of
input, but the simplified forms themselves, generated through overexten-
sion of dominant patterns, must arise in learner language. If this is so, sim-
plification should be understood as part of the leveling process, but with
 Conditions ofspeaker activity 47

the inclusion of novel (learner-generated) simplified variants in the linguis-

tic pool.
Trudgill (1996: 10-11) also extends simplification to include the proc-
ess of lenition, which he finds is frequent in looseknit societies affected by
dialect contact and mixing. He bases this claim on Harris' (1990) formal
definition of lenition, made within the framework of Government Phonol-
ogy. According to Harris, all lenition phenomena can be characterized as
"segmental decomposition", the subtraction of one or more elements from
the internal structure of a segment. This approach to lenition would support
Fontanella's (1992) inclusion of weakening of syllable-final /-s/ as a typi-
cal result of koineization in American Spanish (but note too that final /-s/
lacks perceptual salience; nevertheless, the variable aspiration and loss of
this consonant could be seen as not well justified, since such weakening
can compromise transparency and regularity). Thus, [s] can be character-
ized as composed of a coronal articulation and a narrowed articulatory
stricture; the move from [s] to [h] is a loss of coronal articulation, and loss
of narrowed stricture leads to loss of the segment. Nevertheless, lenition
itself is not always so easily characterized. A more generalizable analysis
is possible: when frequently-produced "natural" phonetic variants are
added to the prekoine linguistic pool, they may survive the process of lev-
eling as additive features of the resultant koine.

3. Elaborating the process model: Conditions of speaker activity

Trudgill's (1986) presentation is most noteworthy for its detailed specifica-

tion of the macro-level linguistic processes or outcomes that characterize
koineization. For the micro-level, he focused his discussion almost exclu-
sively on the process of accommodation by adults, but largely ignored de-
tails of the social context. As an explanatory model of language change, it
requires further elaboration. Siegel (1993b) discusses this problem in his
review of Trudgill (1986). He points out, for instance, that various dialects
can be in contact for extended periods without any mixing or leveling tak-
ing place:
the contact status quo may end . . . with certain political, social, economic,
or demographic changes . . . This may bring about either increased interac-
tion among speakers of various dialects or decreased inclination to maintain
linguistic distinctions. (Siegel 1993b: 117)38
48 Koines and koineization

This is, in fact, an allusion to the Actuation problem (see Chapter 1). Ac-
cording to Siegel, most treatments of the issue have been cursory.
Domingue (1981: 150), for example, vaguely ascribes koineization to "the
need for unification among speakers of different dialects in a new envi-
ronment". Others simply assume, without further comment, that it occurs
most frequently in cases of migration and colonization. About this Siegel
comments:
But social and demographic changes, as in migration, do not necessarily
bring about koineization. For the processes to begin people have to give up
(consciously or unconsciously) their own speech distinctions and conform to
the speech of others. Part of a theory of dialect contact would need to be
able to predict when this would occur . . . Trudgill uses the accommodation
of social psychology to show that people may modify their speech in face-
to-face dialect contact situations. But . . . [h]e focuses on the linguistic
changes themselves rather than on the aspects of the theory that attempt to
explain why sometimes these changes occur and why sometimes they do not.
(Siegel 1993b: 117)

The gaps outlined by Siegel were noticed by Trudgill himself, and in a

series of later articles (e.g., Trudgill 1992, 1994, 1996, 2002) he shifts his
attention to the influence of social structure, particularly in terms of social
networks, on speaker activity (and on language structure), and the con-
straints imposed on this activity by the language acquisition process. His
repeated claim in these publications is that smaller societies characterized
by closeknit social networks, overall stability, and isolation, are more ca-
pable of promoting the learning of a language variety with minimal change,
and that over time speakers of such varieties are more likely to retain and
add complex features. Conversely, (usually larger) societies characterized
by more looseknit social networks and high levels of contact are more
likely to tend toward simplification, hybridization, and higher rates of
change. These factors, and how they affect koineization, will be discussed
below. First, it is necessary to take into account the importance of norms,
norm enforcement and social network structure - or lack thereof - at the
micro-level of koineization. Even more importantly, one must understand
how the specific learning context of koineization affects language/dialect
acquisition by children and adults.
 Conditions ofspeaker activity 49

3.1. Norms, norm enforcement, and social networks

An understanding of norms is essential to the development of an integrated

model of koineization, since they define the social conditions that limit
speaker actions as well as the end-products of collective speaker activity on
the macro-level. Norms (or conventions as they are sometimes known, cf.
Croft 2000) are socially-determined limits on behavior; in the case of lan-
guage, they define correct linguistic products for a particular community.
Norms exist and are enforced because they are socially useful. Bartsch
(1987: 126) outlines two primary functions of norms. First, they ensure
communication by restricting the choice of actions, making them predict-
able, and thereby supporting the information-bearing function of language.
Second, they serve to indicate group identity:
From the point of view of group identity, adherence to the norms is watched
closely, much more than would be necessary for the mere functioning of a
language as a means of communication. This is especially true of acts of
correction that apply to those linguistic forms that symbolize the internal so-
cial order of a group or society: special morphological and syntactic forms,
as well as vocabulary and styles that are connected to roles, position and so-
cial status. (Bartsch 1987: 87)

At the level of language change, these two functions of norms enter into
conflict. The communicative function is conservative and promotes stabil-
ity and uniformity. The identity function, on the other hand, often leads to
the development of conflicting norms and overall fragmentation (Milroy
1992: 38), at least in more stable societies.
Norms, which are natural to all human societies, need to be distin-
guished from prescriptions, with which they are often confused. A pre-
scription may arise when a norm is codified and then extended beyond the
community where the norm originated, or when a norm changes but an
earlier codification of the norm does not. Norms also need to be distin-
guished from rules. Norms describe the linguistic system of a community,
the langue of Saussure or the Ε-language of Chomsky, and though they are
clearly abstractions based on dominant patterns of behavior, speakers (and
linguists) are aware of them and influenced by them. Linguistic rules de-
scribe the internalized systematizations (or competence) that speaker-
learners construct based on the norm-governed output of other speakers (=
learner input). Speaker-learners are exposed to norm-governed behavior
and abstract the best they can from this behavior, but their rule systems
often fail to match the norms exactly. This can be seen in the fact that these
50 Koines and koineization

abstract rule systems tend to overgenerate, producing patterns (or innova-

tions) that are unacceptable (see discussion on acquisition below). Con-
straints on pattern generation - or learning - are thus determined by norms.
Speakers constantly deviate from norms (for varied reasons), so norms
must be enforced, but how is this done? Bartsch sees this as an active proc-
ess and suggests that norms are enforced through correction, criticism,
ridicule, stigmatization or even exclusion. Such emphasis on overt correc-
tion may be valid in the case of salient and socially-stereotyped features,
but indirect correction through constant and consistent presentation of
"correct" models is certainly fundamental for less salient features. The
issue of norm enforcement has been a principal topic in the research of
Lesley and James Milroy, who, in numerous sociolinguistic studies of the
speech of inner-city Belfast, focused on the issue of language maintenance
in non-standard working-class speech communities (many of the results of
which are summarized in Milroy 1987). The Milroys have made certain
fundamental and useful recommendations that allow for clearer demarca-
tion between the micro-level conditions of speaker activity and macro-level
linguistic processes or outcomes. Namely, speakers introduce innovations
(through fluctuation in pronunciation, accommodation and mixing of
codes, incorrect learning, processing errors, etc.), but changes are seen to
occur only at the level of the linguistic system, understood as the product
of speaker activity in social contexts. They too view the linguistic system
as synonymous with the consensus norms of the community, so linguistic
change is change in agreement on the norms of usage. Like Bartsch, the
Milroys emphasize the functionality of norms, but the norms of dialects
depend far less on communicative needs:
There would be little point in having these different norms (which are arbi-
trary in linguistic terms) if they did not carry social meaning, distinguishing
between one community and another and carrying a sense of community
identity for speakers. (Milroy 1992: 83)

The conservation of different norms in low-status urban dialects in the face

of pressure from the standard and other varieties is inexplicable unless one
takes identity into account; the mechanism which allows such conservation
of minority identity and norms is the close-knit social network (Milroy
1987).
For the individual, a social network is the sum of the relationships
which he or she has contracted with others. At the societal level, social
network amounts to a boundless web of ties that permeate an entire society,
and generally extend beyond it as well. Any particular social network will
 Conditions of speaker activity 51

have particular structures and properties. In sociolinguistic research, Mil-

roy (1987) has defined social networks in quantifiable terms of density and
multiplexity. A social network is maximally dense if everyone knows eve-
ryone else, and multiplex if each person knows each other in a variety of
capacities (e.g., as family member, friend, neighbor, colleague, schoolmate,
fellow parishioner). Multiplexity is a useful measure of the strength of any
particular tie because it can be measured, but it is only an indirect indica-
tion, and is most useful because it reflects other potentially more signifi-
cant factors:
The strength of a tie is a (probably linear) combination of the amount of
time, the emotional intensity, the intimacy (mutual confiding) and the recip-
rocal services which characterise a tie. (Granovetter 1973: 1361)
It is clear, too, that strong ties must be limited in number. In contrast, weak
ties are characterized by their uniplexity and high number. All social net-
works are built on a combination of strong and weak ties, but close-knit
social networks characterized by a high number of strong ties serve as
norm enforcement mechanisms that tend to impede change of all types,
including linguistic change. The obvious corollary to this is that loose-knit
social networks with numerous weak ties permit change (Milroy and Mil-
roy 1985: 370-375; Milroy 1993: 227; Trudgill 1992). Indeed, the Milroys,
following Granovetter (1982), assert that weak ties serve as the very chan-
nels of transmission for all types of innovations.
According to the Milroys, the weak-tie model is sufficiently abstract to
be applied to other situations besides the close-knit urban communities
investigated in Belfast. In fact, the greatest advantage of the social network
model is its universality:
The fundamental [advantage] in an investigation of this kind is that we do
not need to accept any prior assumption about how society at large is organ-
ized or structured . . . it is a universal that all individuals in all societies have
contact with other individuals. (Milroy 1992: 85)
Contact situations are one area where an understanding of network rela-
tions is crucial. Milroy suggests that where speakers of different varieties
come into contact but also maintain strong social networks, stable bilin-
gualism may result (1992: 200; cf. Siegel's comments quoted above). How-
ever, mixing of speakers of different varieties with the breakdown of social
networks, as we would expect in koineization, will strongly favor linguistic
change, since the number, frequency and relative proportion of weak ties
increases dramatically.
52 Koines and koineization

The model of weak ties provides an important support to the model of

koineization. When there is great social or geographic mobility, weak ties
will predominate. Whereas strong ties promote local cohesion and overall
fragmentation, weak ties promote greater overall uniformity and, in certain
contexts, simplification (Milroy and Milroy 1985; Milroy 1992: 178). In
the inner city, the Milroys found that complex allomorphy functioned to
mark identity in close-knit social networks. However, middle-class speak-
ers in outlying suburbs of Belfast generally had low network indices (i.e.,
loose-knit social networks), and it was found that the speech of many of
these informants was characterized by a radical reduction in phonetic in-
ventory (Milroy 1992: 98). For example, the six inner-city allophones of
the phoneme /a/ (mentioned above) were often reduced to just one or two
in the middle-class and more socially-mobile outer city (Milroy 1992: 102).
Initially, this reduction occurs as a response to the change in speakers'
social context and the concomitant loss of functionality of certain variants.
This is demonstrated in another example given by Milroy:
one man of 27 commented that during his years away from Northern Ireland,
he had stopped using what I have called the in-group alternants of the (pulí)
set. The most obvious explanation for this is that during his years away from
home, he would develop a large number of relatively weak ties: in such cir-
cumstances, the in-group alternant would cease to have any function for him
and so could be abandoned. This functional explanation seems to be more
satisfactory than, and logically prior to, one based on prestige: as it happens,
this individual's activities away from home had not been upwardly mobile.
(Milroy 1992: 99-100)

The breakdown of consensus (found in outlying middle-class neighbor-

hoods) may lead to patterns of usage that are less predictable by linguistic
rule or by sociolinguistic pattern, with some apparently random variation,
or some lack of agreement on norms of usage (Milroy 1992: 104-105).
This reduction of variants and apparent randomness resembles in some
respects that of the sociolinguistically chaotic prekoine, and this resem-
blance is not casual:
resistance to merger, lexical transfer and restructuring is promoted by the
existence of strong network ties, and . . . types of change that result in sim-
plification are encouraged by weakening of ties. Complex patterns, which
are functional, are maintained by the norm-enforcement function of dense
and multiplex networks. For those whose ties are uniplex and (relatively)
open-ended, these patterns are no longer functional and it is for this reason,
 Conditions of speaker activity 5 3

and not primarily because of speakers' desire for 'prestige', that they disap-
pear. (Milroy 1992: 107-108)
This loss of social functionality of linguistic features is especially impor-
tant in language varieties used by larger and more mixed communities:
As Jakobson perceived long ago, those varieties that have supra-local func-
tions and that tend to develop in the direction of koines display simpler pho-
nemic systems than varieties that have purely local functions . . . It may well
be that, in varieties that have supra-local functions, a high degree of com-
plexity (at any level) is indeed dysfunctional. (Milroy 1992: 108)
In koineization, speakers of different dialectal origin come together, leav-
ing behind their communities and established social networks. In this con-
text, many features of their speech lose their functionality, and, of course,
may become counter-functional in the new context to the degree that they
mark a speaker as different (through accommodation they may be sup-
pressed). Norms not shared by most speakers cannot be consistently en-
forced, and new norms can only be created as new social networks crystal-
lize. This is, however, a multigenerational process. When immigrants to a
new mixed community first come together, their social relations will be
based primarily on weak ties. After one or two generations, however, new
more close-knit social networks will often arise (particularly in more tradi-
tional societies, where such networks were probably more necessary for
survival) with the development of new social and linguistic norms. Why is
this the case? According to Mitchell (1986: 74), individuals create personal
communities, or social networks, which provide them with a meaningful
framework for solving the problems of their day-to-day existence. Indi-
viduals in a new context are likely to work to set up new social networks in
order to meet their problem-solving needs. These new networks will of
necessity begin as loose-knit networks based on weak ties (and may remain
as such, as in the case of many middle-class persons in mobile post-
industrial societies), but with time - measured in generations - may de-
velop into close-knit social networks. The development of new social net-
works and the focusing of new norms has been of primary interest in the
research of Kerswill and Williams, and it will be discussed below.
Another important aspect of the Milroys' research has been their at-
tempt to determine which persons act as innovators in close-knit social
networks, and they suggest that this is likely to be a person who is only
related to the group through weak links. Significantly, the innovator cannot
be someone who is thoroughly integrated into a close-knit social network,
54 Koines and koineization

although an early adopter must be (Milroy and Milroy 1985: 367-370;

Milroy 1993: 225-226). In a koineizing social situation, however, all the
speakers become innovators vis-à-vis speakers of other dialects. As a re-
sult, the number of relative innovations peaks at the same time that the
strength of norm enforcement mechanisms (social networks) declines to a
minimum. It is clear that one way in which the variation is reduced is
through accommodation in the strict sense: the simple avoidance by speak-
ers of features that mark them as different. However, as we have seen, ac-
commodation, by both adults and children, can also imply the learning of
new features, and this process can itself lead to the introduction of new or
additive features. In fact, language acquisition also plays a significant role
in the selection of innovations and the stabilization of new norms.

3.2. Language acquisition: Adults and children

Speakers in contact with speakers of different dialects do not continue to

speak the same way; at times they have to learn new features that are alien
to their normal ways of speaking. Trudgill refers to this as long-term ac-
commodation when it is successful, and as interdialect when the learning is
incomplete. This incomplete learning may lead to the creation of new
forms that are not present in any of the contributing dialects. This, in fact,
is true of all types of adult second language learning, and it is probably not
necessary to distinguish between second dialect learning and second lan-
guage learning, except as a difference of degree: the number of features to
be learned in the acquisition of a second dialect (and thus of potential in-
novations) will necessarily be smaller than in the case of a typologically
distinct language, and the dialect learner may not have to employ strategies
such as paraphrase with the same frequency as a prototypical foreign lan-
guage learner. Otherwise the similarity between second dialect acquisition
and second language acquisition is neatly captured in Selinker's concept of
interlanguage (1972, 1992).39
Interlanguage describes the range of intermediate systems that charac-
terize the second language speech of individual adults. These systems,
which in the early stages of acquisition are progressively restructured, of-
ten stabilize permanently or fossilize before the learner system reaches full
identity with the target language. Thus, interlanguage systems are unique
systems which are different from both the native language and the target
language. There are five psycholinguistic processes that shape interlan-
 Conditions ofspeaker activity 5 5

guage. The first is native language transfer. Selinker, following Weinreich

(1968: 7), suggests that this happens as speakers make "interlingual identi-
fications" between the native language, their interlanguage, and the target
language. In other words, units are not only significant within any one sys-
tem, but also across systems. For example, a speaker may perceive a taxo-
nomic phoneme as being the same in all three systems, even though it pat-
terns differently or has different phonetic realizations in each system. The
second process is overgeneralization. In this case, the learner masters a
general rule, but fails to learn exceptions to the rule. For example, the fre-
quent production among learners of English of *goed, *drinked, and
*hitted for went, drank and hit shows overgeneralization of the rule which
specifies suffixation of -ed (in any of various phonetic realizations) when
the meaning 'past' is intended. A third process is transfer of training,
which occurs when the second language learner applies rules learned from
instructors or textbooks. This is significant only in cases where schooling
is available to a sizable proportion of the population, and in the discussion
of language change is particularly relevant to the process of standardization
and the impact of literacy. However, transfer of training essentially means
that the normal inferencing process is undercut and the learner is provided
with rules (whether correct or not) that can be deductively applied. This
may occur in non-classroom contexts when speakers engage in metalin-
guistic commentary on what is acceptable language practice in their com-
munity.
Communication strategies are used by the learner when the interlan-
guage system seems unequal to the task; thus, unlike transfer and overgen-
eralization, such strategies are less likely to affect the interlanguage of
second dialect learners, where mutual comprehensibility is high. Con-
versely, they will be especially frequent in the speech of learners of typo-
logically distant languages. Communication strategies include: avoidance,
paraphrase (approximation, word coinage, circumlocution), conscious
transfer, appeals for assistance, and mime (Tarone 1977). Certain of these
strategies overlap with those of accommodation. Thus, avoidance and
paraphrase remove tokens of an item or feature from the linguistic pool
(and sometimes replace them with new alternatives, though the reason for
doing so may be different in the koineizing context). Conscious transfer, as
in literal translations or code-switching, might also lead to greater overall
influence of certain speakers' native language variety in the prekoine mix.
Finally, learning strategies are, by and large, of little interest in discus-
sions of language change, with one exception: the cognitive strategy of
56 Koines and koineization

analysis. In order to produce forms and structures, the learner must cor-
rectly analyze them in the input he or she receives. If this input is incor-
rectly analyzed, the speaker may produce "incorrect" forms. These nor-
mally will be eliminated in the face of contrasting evidence, but when this
is absent or not perceived, or the speaker-learner has no further need to
continue altering output, then the errors may become fossilized in the
learner's speech (see below).
Second language learning can be quite difficult and almost never "per-
fect" for adults, even when they have plentiful input and interaction with
native speakers of a stable target variety (i.e., there is a difference between
input and intake). However, in a koineizing community with great variation
(i.e., without a clear target) and low norm enforcement, "errors" in learn-
ers' speech will be even harder to correct. Their incorrect production will
add further variation to the linguistic pool, and can also become input for
others. Of course, extreme departures from the shared basic system of all
speakers will probably not survive, but where the transfers, misanalyses,
and, especially, overgeneralizations are matched by many speakers (or
exist in one of the established contributing varieties), then the interlan-
guage forms/features stand a chance of survival in the koine.
The interlanguage model allows us to predict the introduction by adults
of innovative mixed forms, novel form-function analyses, as well as the
overgeneralized forms and uses that, when adopted by other speakers, lead
to simplification of the linguistic system. It also underlies an important and
controversial claim made by Trudgill. In his later publications (1989, 1992,
1994, 1996, 2002), he argues repeatedly that simplification (in all types of
language contact) is exclusively the result of adult second language (or
dialect) acquisition. For example:
simplification is due to the imperfect learning...by non-native adults and
post-adolescents. Everything we know about young children indicates that in
general they are such good language learners that they normally learn per-
fectly any language variety that they have sufficient exposure to. Imperfect
learning, and thus simplification, does not result from non-native language
learning as such but from adult non-native language learning.
(Trudgill 1992: 197)

Interestingly, he makes this claim in the midst of a discussion on the rela-

tion between the structure of social networks and the structural develop-
ment of dialects and languages. However, if we consider more closely the
tendencies of first language acquisition and how these are influenced by
 Conditions of speaker activity 57

the koineizing social context, we will see that Trudgill's conclusion must
be rejected.
To begin, adults are not the only ones who must learn in a koineizing
environment: children play a doubly significant role, in that they can pro-
duce innovations and it is child learning in conjunction with their estab-
lishment of new social networks that leads to eventual focusing (see be-
low). Indeed, as was argued earlier, were it not for the potential impact of
transfer, it would not be logically necessary to include adult accommoda-
tion and learning in the model of koineization: most of the same effects
would occur even if children were the only ones learning and introducing
innovations. Innovations in child language are often similar to those of
adult interlanguages. This is especially true of overgeneralizations, which
are likely to survive in child speech if children encounter inconsistencies in
adult usage and that of their peers.
The similarities between child and adult learning-based innovations
have been the object of both experimental and historical study. Bybee and
Slobin (1982) studied documented historical changes in the past forms of
English verbs and compared these to child and adult innovations in the
formation of past tenses of irregular verbs. For both adults and children
they found that the verbs most frequently regularized were those with the
lowest frequency of occurrence. Differences between child and adult over-
generalizations were minor. For instance, they found that about half of the
adult innovations were in the choice of the vowel in strong past tenses of
the ring, sing, swim class (e.g., shrunk for normal shrank), while nearly all
the child innovations for strong pasts in this class were of this kind, which
indicates that children are regularizing the change (or turning similar, fre-
quent adult innovations into changes).40 In a situation of weak norm en-
forcement, the combined tendencies of both adults and children to over-
generalize can lead to significant linguistic changes.
The fundamental similarity between child and adult acquisition is also
discussed by Ravid (1995), though she emphasizes that it is children and
non-literate adults - "naive speakers" - who are most prone to overgeneral-
ize or regularize (Ravid 1995: 91).41 More generally, however, child speak-
ers seek regularity and transparency in language far more consistently than
do adults, and this tendency causes them to produce "errors" that enhance
such regularity and transparency (Ravid 1995: 25). Children and non-
literate adults are seen to react against structural opacity in the sense of
Kiparsky (1982) and Lightfoot (1979). In so doing they rely on Operating
Principles of acquisition and processing (Slobin 1985):
58 Koines and koineization

— formal simplicity (leading to preservation of base forms when modified by

affixes, and back formation)
— formal consistency (Kiparsky's allomorphy-reducing trend or trend to
grammatical regularity)
— semantic transparency (seeking one-to-one correspondence between mean-
ing and form)
— saliency (perceptual distinctiveness of a feature makes it less subject to the
other principles, since salient structures are more easily learned and re-
tained; also marking information through the use of analytic structures, or
separate words which are by nature salient).

It is implicit in Ravid's discussion that the innovative results of such

strategies - "errors" - are likely to be eliminated only in the face of fre-
quent and consistent norm-governed input with which the child can com-
pare his or her output. When such frequent and consistent input is lacking,
then children will continue to produce "errors" that enhance regularity and
transparency. Of course, such behavior will either produce innovations and
lead to change, or, more likely, complete changes that have been initiated
in the adult community (which in the case of Israeli Hebrew is directly
linked to its status as a "resurrected" language whose first speakers were
second language speaker-learners; these speakers reduced the phonemic
inventory and in so doing made input for the complex morphology more
opaque for learners).
Simplification then is the result of both child and adult overgeneraliza-
tions which go uncorrected. This is also true of the mixed forms and form-
function reanalyses that Trudgill refers to as interdialectalisms. Like adult
second language learners, children must analyze the linguistic uses they
hear around them as they attempt to construct a grammar that corresponds
to community norms; how they go about this is explored in Andersen's
(1973) discussion of abduction and deduction (the flaws in this model have
recently been pointed, but it is worth considering Andersen's proposals
before discussing the problems). In the process of grammar construction, a
learner must construct a grammar based on the surface output of other
speakers (other speakers' output is by definition the input for learning).
According to Andersen, this is done through applications of the three
Peircean modes of logical inference: abduction, induction and deduction.
Deduction and induction are better known, but abduction is hypothesized
to be most important for understanding the relation between learning and
language change. The general relation can be shown in the following illus-
 Conditions of speaker activity 59

tration (from McMahon 1994: 94). Given a particular result (e.g., Socrates
is dead), and a law (e.g., All men are mortal), one abduces that something
may have been the case (e.g., Socrates may have been a man). Abduction is
notoriously unreliable (consider the consequences of invoking the law 'all
dogs are mortal'), but Andersen considers it immensely important because
it is the only one of the three modes of logical inference that introduces
innovations. In language learning, abduction operates on the output of
other speakers, and the other two modes are used as testing devices (pre-
sumably these abductions are made in accordance with something like
Slobin's Operating Principles). Induction occurs as the initial abduction is
tested against further examples in the output of other speakers. If this out-
put is perceived as inconsistent with the initial abduction, then it must be
revised with further abductions. Deduction occurs when the learner at-
tempts to produce utterances based on the abduction, testing her own out-
put (based on the abduction) on other speakers; if listeners misunderstand,
correct or reject the speaker-learner, then the initial abduction may (or may
not) be revised. This can be illustrated with a hypothetical case based on
Trudgill's example of the Northern English English speaker who attempts
to accommodate to Southern English English usage. Knowing that northern
/but/ is equivalent to southern the speaker-learner abduces "incor-
P O A X I ,

rectly" (through an overgeneralization of a pre-existing pattern) that north-

ern /but/a/ should be substituted with a previously non-existent /bAt/a/. By
hearing others use this word with the correct pronunciation, the learner
may be able to correct herself, but this will depend on both the frequency
and salience with which she hears the word. In addition, she may begin to
use the pronunciation /bAtJa/, and she may be overtly corrected or simply
receive less than positive feedback and therefore begin to determine that
there is some problem with her production.
However, Deutscher (2002) has recently pointed to certain flaws in An-
dersen's conceptualization of learner analysis. First, he points out that An-
dersen conflated earlier and later proposals in Peirce's discussion of the
modes of inference, and therefore the meanings of the term abduction shift
in this discussion. This, in itself, might not be a problem, but Deutscher
also shows that abduction as discussed by Andersen only explains how
learners match surface forms/features to a "wrong" pre-existing rule; it
does not recognize that learners invent new rules and attach them to exist-
ing forms (which they often do). He provides the following example:
The surprising fact is observed: an instrument is called 'fork'. But if it were
true that there was a rule: 'An N-pronged instrument is called N-k', the form
60 Koines and koineization

'fork' would be a matter of course. Conclusion: Hence, I have reason to sus-

pect that there is such a rule (so a three-pronged instrument is a 'threek').
(Deutscher 2002: 483).
Crucially, the production of the form threek is dependent on the invention
of a new rule (based in turn on pre-existing general knowledge, general
cognitive tendencies, surface and schematic analogies, and specific con-
texts). Deutscher goes on to point out that there is no useful distinction
between abduction and induction (both are induction). Finally, he suggests
that linguists have better terms to describe the inferencing processes of
induction and deduction: reanalysis (for induction, or association of an old
or new rule/explanation with a surface form) and extension or overgener-
alization (for deduction, which through analogy extends an existing rule -
whether old or new - to the production of new surface forms).
Nevertheless, one aspect of Andersen's model which remains useful is
its emphasis on the testing process that allows confirmation or correction
of initial analyses (abductions). This occurs through what Andersen re-
ferred to as induction and deduction (of course, Andersen's abductions are
simply incorrect inductions, or inductions based only on partial evidence or
input), in which the speaker-learner continues to receive input and try out
innovative forms/features when interacting with other members of the
speech community. This testing process, which allows restructuring of
learners' grammars and replication of grammars of previous generations,
cannot be successfully completed without the stable norms that are en-
forced through close-knit social networks (or the institutions of standardi-
zation). When these norms are lacking, as in the prekoine linguistic pool,
learners will learn enough to match majority (or salient) forms most of the
time, but such learning may well be "imperfect" and leave room for multi-
ple reanalyses and overgeneralizations (which, if produced by enough
speakers, can lead to change). Indeed, this was observed by Henry, who
studied how children handle syntactic input from parents and the surround-
ing community: "What is clear from this [situation] is that language learn-
ing does not involve selecting a grammar which fits all the data. Rather, it
must involve selecting . . . the grammar which best fits the majority of the
data" (Henry 1995: 79). Significantly, Henry's observation is supported by
more general research in cognitive and functional linguistics, which has
begun to emphasize the key role that frequency of use has in the develop-
ment - or emergence - of the grammars of all speaker-learners. Such stud-
ies have shown the impact of frequency effects on phenomena such as
phonetic reduction, functional reanalysis and grammaticalization, hyper-
 Conditions of speaker activity 61

corrections, and preservation of conservative properties (Bybee and Hop-

per 2001). More fundamentally, such studies show that frequency is key to
the construction and design (or emergence) of all speaker's grammars.
Learning, then, is the source of innovations such as form-function re-
analyses and overgeneralizations, and when these are sufficiently frequent,
and are propagated through the community, they become systemic-level
simplifications and "interdialect" features. They are especially important in
koineization, for the high degree of variation and low level of norm en-
forcement impedes the learning of features that are inconsistently produced
and/or perceived as infrequent.
The preceding models of learning are particularly useful because they
highlight similarities between child and adult language acquisition (though
young children generally are not conscious of the process of grammar con-
struction), as well as the importance of the social context to successful
learning. However, we must also ask whether there are any differences
between child and adult learning that could impact on language change
during koineization. It is widely recognized that children are ultimately far
more successful than adults in language learning, and it is often assumed
that only children can learn a language "perfectly" (cf. Lenneberg's Criti-
cal Period Hypothesis). It is now frequently claimed that children lose the
ability to learn "perfectly" from between the ages of 8-14 (Chambers
1992; Kerswill 1996).42 Kerswill (1996) includes a review of the literature
on this topic, and concludes that adults and children can learn some fea-
tures with equal ease (e.g., new vocabulary). However, other features can
be acquired only by children. The features vary according to age of ac-
quirer, the linguistic level concerned, and relative complexity. Kerswill
summarizes his findings in a hierarchy of learning difficulty (Table 6).
The ranking shows that both adults and children can learn certain kinds
of features equally well, assuming they have access to sufficient input and
interaction. However, even with such input and interaction, only children
can acquire the complex features at the top of the hierarchy. What is the
significance of this ranking to koineization? When simplification occurs in
koineization, we can be sure that it is partly attributable to adults, and
when complex features or patterns survive, we can be sure it is because
children have received sufficient input. However, this does not mean that
children can learn "perfectly" in the koineizing context. Children also sim-
plify, but they do so not because of any lack of inherent abilities, but rather
because the social context of learning often does not provide the sort of
consistent input that they need in order to construct (and restructure) inter-
62 Koines and koineization

Table 6. Hierarchy of learning difficulty for children and adults. Source: Kerswill
(1996:200). Rank of 1 represents most difficult features to learn.
Rank Feature type Age acquired
1 i lexically unpredictable phonological rules, by3(?)
which may reflect lexical diffusion nearing
completion and which are not socio-
linguistically salient
ii new phonological oppositions by 3-13
iii grammatical oppositions by 8 (?)
2 iv prosodie systems by 12-15
3 V grammatical change: new morphological peaks in adoles-
classes (in creóles, may be tied to lexical cent years
acquisition)
4 vi morphologically conditioned changes not before 4-7,
then lifespan
5 vii reassignment of words or lexical sets to other lifespan
morphological classes
6 viii mergers lifespan
7 ix Neogrammarian changes (exceptionless lifespan
shifts, easier if they are connected speech
processes)
8 χ lexical diffusion of phonological changes, lifespan
especially those which involve an existing
opposition and are salient
xi borrowing: new lexical forms of old words; lifespan
new phonetic forms of existing morphologi-
cal categories
9 xii borrowing: vocabulary lifespan

nal grammars that (eventually) allow the reproduction of complex patterns

or features (of course, older children and adolescents may simplify because
of a combination of decreasing ability/motivation and lack of appropriate
input).
Child language acquisition thus plays a unique role in the propagation
of certain complex features. But child learning and use of language must
also play a primordial role in the establishment of all new linguistic norms.
This is because it is only the features that are consistently learned by suc-
ceeding generations of children that will survive to characterize a new
koine. Adults, through their accommodation and acquisition, merely alter
 Observing koineizatìon in Milton Keynes 63

the output that serves as the starting point for child learning. And it must
be assumed that accommodation takes place between children, too, who
leam from each other as much as from adults. This process, which occurs
over the course of one or two generations of children, is what allows the
eventual focusing of new norms, though it is inextricably linked to the
solidification of social networks. In fact, these questions are at the core of
the Milton Keynes research project organized by Paul Kerswill and Ann
Williams, to which we now turn.

4. Observing koineization: The Milton Keynes project

Most studies of koineization (including this one) are post hoc studies of
change in previously stabilized dialects. As a result, it has been difficult to
do more than theorize about how adults and children behave and learn in a
koineizing environment, and how they fashion new norms. In order to un-
derstand better the micro-level processes and conditions of koineization,
Kerswill and Williams initiated a quantitative sociolinguistic research pro-
ject to study the developing dialect of the British New Town of Milton
Keynes, located in southeastern England (just to the northwest of London
in Buckinghamshire). They have reported on this research in numerous
publications throughout the 1990s (e.g., Kerswill and Williams 1992, 1999,
2000; Kerswill 1994a, 1996), one of the most recent of which (2000) in-
cludes their own overview of koineization, expressed in terms of eight
"Principles of Koineization", and exemplified with the results of their re-
search in Milton Keynes.
To discuss Kerswill and Williams' Principles and the examples support-
ing them, it is necessary to know something of Milton Keynes and the
sample used in the project. Milton Keynes was officially designated a Brit-
ish New Town in 1967, and its population grew rapidly, reaching 145,000
by 1990, when the project was initiated (Kerswill and Williams 2000: 78).
Migration accounts for 80% of the city's growth, while a relatively high
birth rate accounts for the other 20%. The proportion of residents over 60
is very low in comparison with other English towns, and the proportion of
children relatively high (as might be expected in most immigrant communi-
ties). The incoming population included persons from all parts of the UK,
but 76.2% were from the southeast of England, with some 35.2% from
London alone. The great majority of inhabitants work in manual occupa-
tions, as do all the families in the sample used.
64 Koines and koineization

Kerswill and Williams chose to investigate the development of nine

phonetic variables among Milton Keynes children (some of which are dis-
cussed below). To do so, they recorded (first in 1991, then again in late
1992) a socially homogeneous group of 48 children (eight boys and eight
girls from each of three age groups - 4, 8, and 12), who had been born in
Milton Keynes or who had moved there within the first two years of life.
They also recorded the principal caregiver of each child (in most cases the
mother). An analysis of the migration patterns of the sample families re-
vealed that they were representative of the Milton Keynes population as a
whole.
From their research in Milton Keynes and the work of other scholars,
Kerswill and Williams derived their eight Principles of Koineization.
These are grouped according to the aspect of koineization they describe.
The first group has to do with the outcomes in post-contact varieties:

1. Majority forms found in the mix, rather than minority forms, win out.
2. Marked regional forms are disfavored.
3. Phonologically and lexically simple features are more often adopted than
complex ones.

This group of principles recapitulates (some of) the conclusions of Trudgill

about macro-level linguistic processes and outcomes, so I will not discuss
them further here, though it is important to note that in Milton Keynes
there are no obvious cases of simplification, so the data are most useful for
studying leveling. The other two groups of principles, however, refer to
domains in which the Milton Keynes data have proved useful for elaborat-
ing our understanding of koineization: how acquisition and social networks
interact, and, more importantly, focusing and the time scale of koineiza-
tion. These are discussed below in separate sections.

4.1. Interaction of acquisition and social networks

The second group of principles is intended to describe micro-level proc-

esses and conditions, and more specifically the factors that affect the lin-
guistic behavior of the migrants and the first generation of native-born
children:

4. Adults, adolescents, and children influence the outcome of dialect contact

differently.
 Observing koineization in Milton Keynes 65

5. The adoption of features by a speaker depends on his or her network char-

acteristics.

Principle 4 is important for two reasons. First, children and adolescents

play central roles in defining new norms; this aspect relates to the third
group of principles, so it is discussed below. Second, children must play a
role in the learning of complex features which differs from that of adults.
Above I argued that both children and adults favor and are responsible for
simplifications during koineization, but Kerswill and Williams also draw
attention to the fact that (minority) simplified features in the prekoine lin-
guistic pool are sometimes rejected in favor of (majority) complex features.
Adults are far less capable of acquiring certain more complex features (cf.
Kerswill 1996: 200), so we must assume that even in the koineizing envi-
ronment some children can and do learn features that their parents cannot
or do not use. The example that Kerswill and Williams provide from their
data is the acquisition (and propagation) of the morphologically-
conditioned Roland/rolling vowel split, a pattern found throughout south-
eastern England and therefore dominant among the adult immigrants to
Milton Keynes. In these persons' speech, monomorphemic words such as
Roland, pola, cola, sola are syllabified with the syllable division between
the first vowel /au/ and the following lateral (e.g., Ro-land). These are
realized with a front allophone [aeu]. On the other hand, dimorphemic
words such as roll+ing are articulated with a back allophone [OU], even
though they share the same syllable division as Roland (i.e., ro-lling). Cor-
responding monomorphemic and monosyllabic forms, such as roll, are
realized with the back allophone [DU]. It was found that 35 of 41 children
recorded had acquired the distinction, even though many of their parents
did not use it. The exceptions tended to be the very youngest children who
remained oriented towards their parents and had not yet acquired the dis-
tinction from their peers. Does this mean that a complex majority variant
will be acquired and become a norm before a simple minority variant? Not
necessarily, for Kerswill and Williams also point out that this alternation is
favored by the principle of morpheme invariance, itself a relation of trans-
parency (and salience) that highlights the semantic relation between related
items such as roll/rolling. The model of koineization may allow no sure
predictions about outcomes when a simple minority variant and a complex
majority variant enter into competition; secondary factors relating to sali-
ence will need to be adduced to explain the preference for one or the other.
66 Koines and koineization

In this case, contact with surrounding dialects, in which the split is fre-
quent, seems to have been the deciding factor.43
Principle 5 relates to findings about the impact of individuals' social
network characteristics on the adoption or acquisition of features. Kerswill
and Williams use quantitative findings on the articulation of (ou) variable
(phoneme = /auf) to discuss this principle (see Table 7).

Table 7. Distribution of variants of (ou) across sample (%). Data for children
obtained in elicitation tasks, data for adults in interviews. Source:
Kerswill and Williams (2000: 93).
[ei], [aei] [BY], [asY] [au], [ay], [eu],
l'evi» Γο:1, [oui
4-year-olds 13.5 30.2 55.7
8-year-olds 12.9 53.6 33.3
12-year-olds 3.0 68.6 28.2
Caregivers 3.5 37.3 60.0

It can be seen that the younger children have both the widest range of vari-
ants and that the 4-year-olds are the most likely to use one of the variants
(in the righthand column) that is dominant among the adult caregivers.
However, the 8- and 12-year-olds are already focusing on the fronted vari-
ants in the middle column, and the 12-year-olds have abandoned both the
extreme fronted and unrounded variants of the 4-year-olds and, more
slowly, the conservative variants preferred by the adults. An essential point
to be gleaned from these data is that it is impossible to lump children into
one group. Children of different ages have different social orientations as
they grow older, and these differing social orientations have a significant
effect on their use and acquisition of language. Young children, repre-
sented by the 4-year-olds, tend to be oriented towards their parents, par-
ticularly the primary caregiver (Kerswill 1996), and they tend more than
others to reproduce the features of their parents' speech (though their
speech may also be affected by older siblings and other children). Older
children, represented by the 8- and 12-year olds, are already well integrated
into mainly school-centered peer groups, and as they grow older they tend
to abandon features of their parents' speech and adopt those of their peers,
to the degree that they can.44 This is especially true of the most sociable
and peer-oriented children, who have the most social contacts.
 Observing koineization in Milton Keynes 67

4.2. Focusing and the time scale of koineization

The last group of principles is not unrelated to the second, but all concern
the time scale of koineization and the process of focusing. Trudgill (1986)
ascribes the rise of a new system of norms to a process of linguistic focus-
ing, but does not address the social roots of this phenomenon. It is here,
however, where the Milton Keynes Project has proved most informative.
The principles are:

6. There is no normal historical continuity with the locality, either socially or

linguistically. Most first and second generation speakers are oriented to-
ward language varieties that originate elsewhere.
7. From initial diffusion, focusing takes place over one or two generations.
8. Because of sociolinguistic maturation, the structure of the new speech
community is first discernible in the speech of native-bom adolescents, not
young children. (Kerswill and Williams 2000: 84-85)

Principle 6 deals with the question of normal vs. abnormal transmission,

and is discussed below with pidginization/creolization and dialect leveling.
Principle 7 confirms what has been supposed about focusing, and Principle
8 elaborates on how this process occurs. The central claim is that focusing
of new norms can occur within one generation of settlement, even more
rapidly than the H0yanger data suggested. As the preceding table for the
(ou) variable shows, the oldest children (pre-adolescents) are already fo-
cusing on a new norm that differs substantially from that of their parent's
generation. In this case the new norm is not the majority articulation
among the adult migrants, but rather a fronted articulation that is a feature
of regional dialect leveling. This tendency is probably communicated
through weak ties between adolescents in southeast England, and Kerswill
and Williams suggest that it may also be favored as a sign of adolescent
autonomy. In this scenario, the first generation of children negotiates the
new norm among themselves, since they generally have no obvious target,
and their norms become most clearly defined when they reach adolescence
and their orientation toward their peer group is strongest. Their speech then
becomes a target for younger members in the speech community, such as
the 8- and 12- year-olds in the Milton Keynes study, who further focus the
developing norms. Young children are primarily oriented toward their par-
ents, but they may also be influenced by older children (as well as show
developmental features), and their speech tends to show the lowest degree
of focusing.
68 Koines and koineization

It is clear then that older children and adolescents play key roles in de-
termining new norms in the koineizing community. It is assumed in
Kerswill and Williams' approach that even though both children and adults
simplify, only children will be able to accommodate sufficiently well and
in sufficiently large numbers to focus the most difficult items. Still, as we
have seen, complex features are likely to survive only if they are majority
features, and even then only when other factors intervene to favor their
selection and production by speakers. However, Kerswill and Williams
leave open the possibility that adults might be responsible for focusing of
some types of features. These are likely to be those that Kerswill (1996:
200) identified as the easiest features for all speakers to learn (and the most
salient), such as the introduction of new lexical items and new forms for
existing morphological categories. Unfortunately, these are precisely the
items that have suffered the greatest degree of dialect leveling in the Mil-
ton Keynes contributing varieties, so there is no evidence of simplification
in Milton Keynes, at least in comparison to southeastern varieties (in Chap-
ter 5, I provide textual evidence of a possible adult-sponsored norm: the
rejection of leísmo [a morphosyntactic feature] in early Andalusia, which
appears to have happened very quickly).
The few (salient) norms negotiated by adults will be the first to appear
in a new community, but the set of norms negotiated by the first generation
of older children and adolescents may be the first to define a new relatively
stabilized variety. Kerswill and Williams claim that focusing is nearly
complete within this first generation of children in Milton Keynes, and
they originally claimed that this might be typical of koines (but Kerswill
[2002] adopts Trudgill's model). However, Trudgill (1998) and Trudgill,
Gordon, Lewis, and Maclagan (2000) have argued (based partly on their
study of New Zealand English) that koineization typically occurs over the
course of (at least) three generations in which: 1) the first generation of
adult migrants shows rudimentary leveling (loss of some salient minority
features); 2) the second generation (first of children) continues to show
extreme variability but also further leveling (loss of still more minority
variants); and 3) the third and subsequent generations realize focusing,
leveling, and reallocation.45 Still, this leaves two problems: 1) why do some
features resist focusing even when most other features in the variety have
focused (resistance to focusing of isolated features), and 2) why does fo-
cusing of all features occur more slowly in some cases of koineization,
such as in H0yanger, than in others, such as in Milton Keynes (general
resistance to focusing)? I discuss each of these problems below.
 Observing koineization in Milton Keynes 69

Kerswill and Williams identify a few isolated phonetic features that

have resisted focusing in Milton Keynes. For instance, the vowel of the
thought class still shows considerable variation between a more common
monophthong [o:] and a diphthong [ou] (Kerswill and Williams 2000:
100). The diphthongal articulation is a well-established London variant,
and over one third of adults in Milton Keynes are from London. Children
with London parents tend to show higher percentages of this articulation:
52.4% for children with both parents from London, 29.1% for those with
one parent from London and one from southeast England, 20% for those
with one parent from London and one from outside southeast England, and
only 10% for those with parents who are not from London or southeast
England. The fact that a very large minority of adult speakers use this vari-
ant has certainly favored its continued use among the children, but it is also
likely that weak-tie contacts with London and surrounding regions have
contributed to the survival of each variant. In cases such as this there is no
way to predict how the competition will be settled, though Kerswill and
Williams argue that the third generation will probably decide the matter. In
that case, it may be that one variant will win out, but it is also possible that
the variants will both survive but suffer socio-stylistic or phonological
reallocation.46 It has also been suggested that a feature may resist focusing
over a long period; this possibility is discussed below.
There are several reported cases of general resistance to focusing
(though this depends in part on what the normal time scale is considered to
be). Kerswill and Williams (2000) point to the well-known example of
Heyanger, where it was not until the third generation (the grandchildren of
the first migrants, or second generation of children) that focusing could be
discerned (Omdal 1977: 7, reported in Kerswill and Williams 2000: 73).
Kerswill and Williams argue that two factors were largely responsible for
the slower focusing in Heyanger. First, they suggest that the relatively
great differences between contributing varieties may have slowed accom-
modation, presumably because a) such differences might themselves help
reinforce differences in identity between children, and b) the greater differ-
ence between varieties may have encouraged children to favor a communi-
cation strategy of neutrality such as code-switching.47 Much more impor-
tant, however, was the early social separation of the town's population.
Some neighborhoods and schools were settled predominantly by the east-
ern-dialect-speaking managerial class, while others were settled by a west-
ern-dialect-speaking working class.48 This social separation did not break
70 Koines and koineization

down until the third generation, when children from both communities
began to form new mixed peer groups and with them a new koine.
Kerswill and Williams also discuss the case of the Israeli Hebrew koine
(or quasi-koine). According to both Blanc (1968) and Ravid (1995), Israeli
Hebrew has been slow to focus, even though the majority of Hebrew chil-
dren now have native-speaking parents (Yaegor-Dror [1993], however,
believes the variety is focused enough to be clearly defined for most speak-
ers). Perhaps the major factor slowing focusing in this case is the competi-
tion that exists between the spoken Hebrew koine and the officially-
sanctioned standard language, which retains a relatively complex phonol-
ogy and morphology (Ravid 1995; Kerswill and Williams 2000: 71).
Kerswill and Williams also suggest that this slow focusing is partly due to
the nature of the first generation adult input, which was in every case an
adult interlanguage. Supposedly such interlanguage input would have im-
peded the children's learning, but this seems unlikely since children must
always contend with great variation during koineization, and the differ-
ences between these varieties tended not to impede communication. On the
other hand, the tendency among Mid-Eastern Jews to maintain a distinct
ethnic identity certainly retarded their accommodation toward the Euro-
pean-dominated koine. The maintenance of their varieties of Hebrew, heav-
ily influenced by an Arabic substrate, has helped maintain a high degree of
variation. The constant arrival of new immigrants for many years may also
have slowed focusing.
Kerswill and Williams identify other factors that may retard or affect
the direction of focusing. The first relates to the relative proportions of
adults and children in the original mix:
Where there is a high proportion of adults, simplification and reduction will
occur more readily than otherwise, and focusing will not take place until the
third generation. Where there is an unusually high proportion of children (as
in most cases of migration), there MAY be a lack of simplification, as well as
the presence of focusing in the second generation.
(Kerswill and Williams 2000: 75)

I have already argued above that children and adults are probably both
responsible for simplification and that the key factors in determining
whether or not simplification occurs are the nature of the differences be-
tween contributing varieties and the degree to which learners are denied
frequent and consistent input for any particular feature. Without access to
such input, accurate learning of a feature is impossible. It does seem likely
that a higher proportion of children will favor faster focusing, since it in-
 Observing koineization in Milton Keynes 71

creases the proportion of the population actually engaged in negotiation of

new norms, and it increases the possibility of forming new social networks
among children and young people. If population density is low, or univer-
sal schooling is absent, then focusing may be retarded. But the presence of
more children will facilitate contact between them and increase overall the
number of speakers with greatest peer-group orientation. Schooling can
certainly facilitate the formation of peer groups among children and with
them the development of new norms (cf. Amery 1993; Britain 1997b), but
we need to keep in mind that in other societies different traditions may
exist that allow regular contact among young people. If the tendency for
older children and adolescents to be oriented more towards peers than par-
ents is indeed universal, then we would expect them to take advantage of
any possibilities for contact, particularly in immigrant societies where ex-
tended families (which are themselves social networks) are less likely to be
a source of support/restraint, at least initially. Of course, the exact ages at
which such peer groups are allowed to form (if at all) will certainly be
affected by particular cultural factors.
Finally, we should consider the evidence of the development of New
Zealand English, which has heavily influenced the three-generation model
proposed by Trudgill (1998) and his colleagues (e.g., Trudgill, Gordon,
Lewis, and Maclagan 2000). These researchers have managed to assemble
an unusual corpus of recordings of elderly people made in 1946-1948 by
the National Broadcasting Corporation of New Zealand. The persons inter-
viewed were residents of small towns and rural areas, and they were inter-
viewed as part of an oral history of pioneer reminiscences. Significantly,
all were children of adult immigrants to New Zealand (= members of the
second generation). Trudgill and his colleagues have closely analyzed
these recordings as part of their Origins of New Zealand English project
(ONZE). They have found that this group of speakers was characterized by
tremendous inter-speaker and intra-speaker variability (although, in gen-
eral, the most frequent features manifested in their collective speech were
later selected as norms of New Zealand English). From this they conclude
that focusing had not occurred in the second generation of New Zealand-
ers. This is clearly the case among the persons recorded, but it is worth
remembering that the recordings were of people living in small towns and
rural areas (some of them quite isolated), and do not directly reflect devel-
opments in urban areas. Presumably, focusing would have occurred first in
those urban areas, with rural areas following urban trends only later.
72 Koines and koineization

The above discussion on focusing forces one last question: is there an

upper limit on the time scale of koineization? Most scholars seem to con-
sider one or two child generations as the bottom and usual limit for this
time scale, but an upper limit on focusing has not been set. If no limit is
set, then one could conceivably relate changes that are occurring now with
cases of population movement that occurred centuries before. But there is a
methodological danger here, for the model could then be used to "explain"
just about any change in a variety that has at some time undergone dialect
and demographic mixing. Unfortunately, a model which explains every-
thing will in fact explain nothing. Given the cases of general resistance to
focusing we have seen above, there may be no simple answer to this ques-
tion, but it may be worth exploring one particular case of proposed centu-
ries-long focusing in order to better appreciate the methodological implica-
tions of such a strategy. I refer specifically to the case of koineization in
Fenland English studied by Britain (1997a, 1997b). The Fens, located in
eastern England, were once a wet, marshy area that was largely unpopu-
lated, but in the 17th century the area was drained and settlers moved in
from both east and west. The Fens happen to lie between two major dialect
zones (to the east and west), so settlement of the region led to the mixing
of speakers of different dialects. Britain analyzed the development of two
features: (ai) and /Λ/. The first feature is essentially the phenomenon of
Canadian Raising (see above), and Britain argues that the different variants
were brought by speakers of the neighboring dialects and then suffered
phonological reallocation as focusing occurred. There is evidence that "Ca-
nadian Raising" of this sort has existed for at least 200 years in the Fens,
and Britain argues convincingly that it can be attributed to 17th-century
koineization (though, as I will argue below, such claims are more
convincing if more than one such koineized feature is identified).
The other variable is more problematic. To the (north)west of the Fens,
all words of the butcher, cushion, luck, up class are (and were at the time
of Fenland koineization) realized with [u] (=/u/). To the east, this class is
and was split into two different sets, with the set butcher, cushion articu-
lated with [u] (=/u/), and the set luck, up articulated with [Λ] (=/Λ/). How-
ever, it appears that koineization did not lead to a clearly focused new
norm in the Fens. Britain shows that at least since the 19th century no sta-
ble realization of /Λ/ has existed in the central Fens. Rather, ItJ is articu-
lated with a range of phonetically intermediate articulations between [Λ]
and [u] (at least), including intermediate or fudged [γ]. Only recently have
younger speakers in the town of Wisbech begun to focus a norm by regu-
 Koineization and other contact phenomena 13

larly using the interdialectal [r]. Britain (1997b: 160-165) identifies a

constellation of factors that may have colluded to impede focusing of this
feature: lack of salience for local speakers, an originally sparse population,
lack of schooling and thus contact during initial koineization, the complex-
ity of the /U~A/split, the continued contact with the neighboring dialect
regions, ongoing change in neighboring southeastern regions (with increase
in number of allophones). This approach to the data seems quite reason-
able, though we cannot be certain that an earlier norm did not exist, for this
variation may have arisen subsequent to koineization through contact be-
tween the neighboring dialects. However, there is a problem in the larger
interpretation of this phenomenon. After showing that in Wisbech younger
speakers are focusing a new norm, Britain concludes that we are observing
"koineisation-in-progress in the Fens, despite the fact that the original con-
tact began over 300 years ago" and that this demonstrates a differential rate
of focusing for different features (Britain 1997b: 165). There does seem to
be evidence for a differential rate of focusing for individual features, but it
is not prudent to assume that the focusing that is now occurring in Wisbech
is in any way explicable as a result of koineization. This would simply
overextend the explanatory usefulness of the model, since these speakers
are in no way directly affected by the original social circumstances of some
300 years ago. We should rather view this as a case in which complete
focusing was prevented (or complex allophony was maintained) by a com-
bination of factors, not least of which was the continued contact with
neighboring varieties (as discussed in Britain 2002). In fact, it is normal for
varieties that have undergone or are undergoing koineization to be affected
by other processes of language change and contact.

5. Koineization and other contact phenomena

The preceding sections have attempted to define what koineization is and

how it works. However, given the uncertain history of meanings of the
terms koine and koineization, it is also necessary to consider what koinei-
zation is not, for it has been confused or conflated with numerous other
phenomena, particularly other sorts of language and dialect contact. In the
following sections, I distinguish koineization from other kinds of language
contact and change, but also explore some of the ways in which these dif-
ferent processes might interact with it.
74 Koines and koineization

5.1. Pidginization and creolization

Koines, pidgins, and creóles have not always been clearly distinguished, in
part because each process was not adequately defined, but also because
similarities do exist between them with a certain degree of overlap (Siegel
1985, 1997, 2001). Koines, pidgins, and creóles are new varieties of lan-
guage used by new speech communities, and they result from the mixing of
speakers of different pre-existing language varieties, and the need of such
speakers to negotiate and acquire new structures and lexicon. The results in
each case also show the results of mixing and leveling, but there are impor-
tant differences.
All observers (e.g., Siegel 1988a; Thomason and Kaufman 1988; Muf-
wene 1997) agree that pidgins arise rapidly as "emergency" varieties when
adult speakers of typologically-distant languages need a simple but conven-
tionalized means of communication. In prototypical cases (Thomason
1997), communication is limited to a few contexts and functions (e.g., bar-
gaining), and can be supplemented with non-verbal communication (e.g.,
gesturing), but is frequent enough to require a conventionalized language
variety. The language of the dominant group, or a simplified version of it
known as foreigner talk, serves as the target for other groups. Contact be-
tween groups, often more than two, is limited, and contact may not even be
with the native speakers of the target variety. As a result, the input, interac-
tion, and motivation needed to learn the full target variety, or even a sim-
plified version, is lacking. The resulting pidgin is therefore characterized
by a radically reduced lexicon (that needed for its few functions) and a
"no-frills grammatical system" lacking in inflectional morphology and
complex syntactic structures. The radical structural and lexical reduction
makes it possible for adults to learn, stabilize, and even expand significant
components of the pidgin (e.g., New Guinea Tok Pisin before it gained
native speakers). The lack of native speakers implies, however, that the
phonology in particular is likely to remain less stable across the commu-
nity, reflecting rather the native language background of each individual.
Prototypical creóles, like pidgins, develop when "no group has the need,
the desire, and/or the opportunity to learn the other groups' languages"
(Thomason 1997: 78), but a conventionalized means of communication is
still needed for both inter- and intra-group communication. Creoles draw
the bulk of their lexicon from a dominant (though inaccessible) target or
lexifier language, which is usually typologically distant from other varie-
ties in the mix. However, unlike pidgins, creóles are learned as native lan-
 Koineization and other contact phenomena 75

guages and come to be used as the main language of a speech community.

The greater number of contexts and functions for which creóles are used
leads to the creation of an extensive lexicon and richer structural resources
than those of pidgins. As such, they cannot on structural grounds be distin-
guished from other natural languages (though they do generally lack com-
plex morphology).
Koines differ from both pidgins and creóles primarily in terms of the
social situation in which they develop. Contact (or interaction) among
speakers is frequent and unfettered and input is easily accessible. More-
over, that input, though highly variable, is generally intelligible, since the
contributing varieties share much linguistic structure and lexicon. Speak-
ers, adults, adolescents, and (older) children, usually alter their speech not
in response to an urgent need to communicate information, but rather to
accommodate and project altered identities in the new community. Most
importantly, the resultant koine shows no radical structural discontinuities
with source varieties, and is mutually intelligible with them (though this
may be asymmetrical, with the koine being more immediately and com-
pletely intelligible to speakers of conservative varieties than the reverse).
This last characteristic implies that koineization must be considered as
normal transmission within the Thomason and Kaufman (1988) frame-
work. In this view, normal transmission, the object of much traditional
historical linguistic research, contrasts radically with change through con-
tact, best exemplified in the processes of pidginization and koineization. It
is characterized by "complete and successful transmissions, by native
speakers [either parents or peer groups] to child or adult learners, of an
entire language, i.e., a complex interlocking set of phonological, morpho-
logical, syntactic, semantic, and lexical systems" (Thomason 1997: 74).
Croft sees the ideal case of such normal transmission as dependent on a
very specific social structure, in which:
There are no significant communicative interactions with individuals outside
the society, and the only significant changes in the membership of the popu-
lation are through biological reproduction and death, not through individuals
from other societies entering the population. If these social conditions hold,
then the lingueme lineages [forms and features] will be traced back in time
to that society alone, and back through that society's unique parent society,
and so on. Any new linguemes will arise only through altered replication of
the existing linguemes, and hence will belong to lineages confined to that
society of speakers. (Croft 2000: 200)
76 Koines and koineization

This description matches exactly the isolated societies with close-knit so-
cial networks that Trudgill (e.g., 1996, 2002) sees as most capable of suc-
cessfully transmitting a variety with a minimum of change (or of success-
fully completing a change once implemented). According to Croft, such
communities are assumed in the traditional family tree model of genetic
relationships (even though no pure examples exist), in which parent lan-
guages are followed in time by diverging daughter languages, and each
daughter language has one and only one parent. As argued by Penny (2000)
and Croft (2000), the tree model does not allow for the convergence and
hybridization between dialects that characterizes koineization. But since
koineization most certainly leads to the successful transmission of "a com-
plex interlocking set of phonological, morphological, syntactic, semantic,
and lexical systems", it must be included within the Thomason and Kauf-
man category of normal transmission. In fact, koineization is distinguished
from pidginization and creolization by allowing normal transmission con-
currently with a degree of hybridization, and either can be emphasized. For
example, it is the rapid (but limited) hybridization of koineization which is
emphasized in Kerswill and Williams' (2000) Principle 6, which denies
continuity between the koine and contributing varieties to the prekoine
linguistic pool (cf. also Kerswill 2002: 695-698). Thomason herself (1997:
85) sees koineization as an example of a borderline case between normal
transmission and contact, but, as she also argues, categories and models
such as those discussed here are abstract and to some degree arbitrary. I
will discuss this issue in relation to Mufwene's (1997) critique of the value
of koine as a technical term.
Mufwene has argued quite forcefully that there is no useful distinction
to be made between the terms koine and pidgin/creole. 49 This is based in
part on his questionable redefinition of creole as any variety which arises
through intimate contact (1997: 56; an issue which I will not address di-
rectly here). If we maintain the more regular criterion that pidgins and cre-
óles arise in social situations characterized by relatively inaccessible input
and interaction with native speakers, then koines can be safely distin-
guished from pidgins and creóles on this basis. But Mufwene also claims
that the results of contact in creolization and koineization are in no way
distinct:
In much of the literature, koineization is . . . talked about as involving "lev-
eling" of dialectal features. As is suggested in Siegel's work, those who de-
velop varieties called "koine" did not do this by simply dropping from their
varieties structural features that distinguish them from other varieties.
 Koineization and other contact phenomena 77

Rather, they select from a pool of competing features those that are less, or
the least, marked relative to the contact setting. They restructure one target
variety, as in the prototypical case of Attic Greek, subject to the influence of
the varieties in contact. (Mufwene 1997: 48)
According to Mufwene, the exact same thing can be said of creolization.
There are indeed similarities between these phenomena, but Mufwene's
claim is not completely true. As we have seen above, koineization leads to
the passing on of "a complex set of interlocking systems" (dependent, ad-
mittedly, on pre-existing correspondences between the contributing varie-
ties), but this is never true of pidginization and creolization, in which the
lexicon is derived largely from the dominant target language (the lexifier)
and the grammar is often not clearly relatable to any one language in the
input. Moreover, as Siegel (2001: 182) points out, there may be no target
variety in koineization (though there may be a dominant or majority vari-
ety, and in certain social situations salient features of particular varieties
may become a target), since the changes in speech are not generally made
in order to enhance communication, but rather to project and define new
identities.
Mufwene's third argument, however, is potentially more convincing. He
claims that the distinction between language and dialect, on which the
model rests, cannot be maintained systematically enough for it to justify
unique categories such as koine and koineization (Mufwene 1997: 44-48).
The problems with making the distinction between dialect and language are
well known, and Mufwene reviews them all: structural differences do not
suffice for making the distinction; mutual intelligibility is often difficult to
determine; political and cultural factors always play a role in determining
what is same and different. He also points to many cases of language con-
tact in Africa that are not easily classifiable as either koineization or creo-
lization on this basis, since the contributing varieties are genetically related
but are a) not mutually comprehensible for most or all speakers, and/or b)
not considered the same language by the speakers.
Siegel (2001: 180-184, following Thomason [1997]; see below), ac-
knowledges that the boundaries between dialect and language are "fuzzy",
but he does not accept that the categories are therefore meaningless or use-
less. Siegel argues that mutual intelligibility is still a useful criterion for
distinguishing dialects (or related subsystems, as he prefers), though in
some cases it will be difficult to apply the criterion. Significantly, he ac-
knowledges that mutual intelligibility also depends on factors such as atti-
tudes, beliefs and goodwill. One must ask what speakers, both adults and
78 Koines and koineization

children, believe about the varieties in contact and about their own chang-
ing identity, for this may determine whether they maintain their own varie-
ties, attempt to shift to another established variety, or (inadvertently) move
towards creating a new hybrid variety (through code-mixing and accom-
modation). It will also be affected by the particular situation in which
speakers find themselves, and in the koineizing environment mutual intel-
ligibility will be favored with accommodation and continuing interaction.
Nevertheless, Siegel emphasizes that there are structural constraints on
mutual intelligibility: only varieties with enough similarity have the poten-
tial to be mutually intelligible. Such constraints are lacking in pidginization
and creolization. Finally, it is worth pointing out that in many cases there is
no doubt about the status of the contributing varieties, as in Milton Keynes,
or, as I will argue, most of the contributing dialects to the medieval phases
of koineization in Spain.
It is clear, however, that in many cases we will have to struggle with the
"fuzzy" boundaries between dialects and languages, koines and creóles (or
other contact varieties). Mufwene (1997) questions the existence of such
boundaries and emphasizes the commonalities between these contact phe-
nomena. He is not wrong to do so; Milroy (e.g., 1997) too has argued co-
gently that all language change is a contact phenomenon, since a change
(in his framework) only occurs when an innovation of one person has been
adopted by at least one other, and this propagation is dependent on so-
cially-mediated contact. But this does not mean that all contact situations
are exactly alike, or that no recurrent patterns can be found that will allow
the definition (and better understanding) of different types of contact.
Thomason (1997) responds to this question by emphasizing that catego-
ries and models such as pidgin/pidginization, creole/creolization (and pre-
sumably koine/koineization) are by definition abstract and arbitrary, as are
all the other classifications that linguists make, but that they are based on
observable regularities and patterns of co-occurrence. Models of change
are derived from study of particular cases of language contact and change,
but this is not to say that they are easily applicable to all cases. This is not
a problem with the classificatory criteria per se, but results from the "fluid
nature of language history, and indeed of human language behavior as a
whole" (Thomason 1997: 71). Such categories and models of linguistic
contact are best conceived of as prototypes, and the boundaries between
them will be fuzzy, but they are useful because it makes it easier to talk
about observed phenomena and to compare them. Real cases of change can
be measured against the models. Some will match the categories and mod-
 Koineization and other contact phenomena 79

els more closely and thus be considered prototypical; others will be less
easy to classify, but the different models will still serve to help pinpoint
their unique characteristics. A final question, of course, is this: how many
models of language contact do we need? Thomason, like Mufwene, argues
for parsimony. I would agree, but, for cultural and political reasons, mixing
between speakers of mutually intelligible dialects has been a frequent phe-
nomenon in world history. I would thus argue that a model of koineization
is a particularly useful prototype.

5.2. Contact between stable dialects

The results of dialect mixing as in koineization and those of contact be-

tween stable dialect communities have not always been clearly distin-
guished in earlier work. For example, Thomason and Kaufman (1988) cite
and criticize Mühlhäusler's assertion that "mixing of linguistic subsystems
tends to lead to leveling or a kind of common-core grammar" (Mühlhäusler
1980: 28). They comment:
Now, dialect leveling is one of the oldest traditional notions in historical lin-
guistics. But in its traditional sense 'leveling' merely refers to change to-
ward greater similarity of dialects and not, as Mühlhäusler's use of the term
suggests, to change toward a less marked overall system. All of the evidence
we have indicates that the traditional notion is correct and Mühlhäusler's
more constrained prediction is not. Dialect interference, like cross-language
interference, very often leads to replacement of a structure in one dialect by
a partially or entirely corresponding structure from another. This replace-
ment may of course constitute an overall change [more marked] > [less
marked], but it is just as likely to result in a change [less marked] > [more
marked], or one that is equivalent in markedness. Most changes that are not
simple replacements are partial reinterpretations, with the same variation in
their effects on the markedness of the system. Many of the clearest (recent,
well documented) examples involve pressure from the standard on the non-
standard dialect. (Thomason and Kaufinan 1988: 30)50

The confusion here rests on the use of the term "dialect leveling". Thoma-
son and Kaufman seem to have in mind contact between stable dialects (or
dialect leveling as discussed in the next section), while Mühlhäusler ap-
pears to refer to mixing of dialects as in koineization. The results of con-
tact between established dialects with stable and close-knit social networks
are likely to differ significantly from those of koineization, since such
80 Koines and koineization

communities are more likely to promote the successful learning of rela-

tively complex features. Though they may resist the introduction of innova-
tions, when they do accept them, such communities will be far more suc-
cessful in propagating them (Trudgill 1992, 1996; Andersen 1988). On the
other hand, it would not be correct to claim that koines can have no marked
or additive features. As I will argue in Chapters 3 and 4, these are usually
due to incorrect inductions (reanalyses) based on conflicting input in the
linguistic pool (i.e., they represent attempts of speaker-learners to make
sense of very complex input, and are in effect simplifications from the
speaker-learners' perspective).

5.3. Dialect leveling, homogenization, uniformity

Dialect leveling in its broadest sense occurs when two dialects become
more alike by sharing features. However, it is now frequently used to refer
to a more specific process which occurs when communication networks
over entire regions become dominated by weak ties. Each looseknit com-
munity of speakers is therefore incapable of resisting the combined impact
of numerous weak ties with other communities, which are themselves char-
acterized by weak ties. Such leveling is frequent in modern urban and sub-
urban middle-class society, and is due to both social mobility and cultural
changes which underscore the importance of personal autonomy. Kerswill
and Williams (2000) show that dialect leveling has a pervasive influence
on both new koineizing and old towns in southeast England. It is difficult
to distinguish dialect leveling from koineization because they share fea-
tures and because koineization over large areas is generally accompanied
by dialect leveling. Dialect leveling can easily be considered a weak form
of koineization, but it does show some distinctions. First, it does not lead to
the formation of a clearly-distinguished new dialect, but rather seems to
favor a general spread of features and the loss of peripheral dialects. Sec-
ond, it does not necessarily favor simplification. Third, dialect leveling
leads to gradual rather than abrupt changes of use of features between gen-
erations (Kerswill and Williams 2000).51
When koineization occurs simultaneously (or nearly so) in many places
over a wide geographical region, it must be supposed that dialect leveling
between each koineizing center (towns and cities, with surrounding areas)
also occurs. As evidence, Trudgill (1986: 145) pointed to the case of Aus-
tralian English, which shows clear social and stylistic variation, but which
 Koineization and other contact phenomena 81

is also noted for its general lack of geographical diversity. The differences
that do exist are limited to certain lexical items and fine-grained phonetic
differences. Trudgill attributed such geographical leveling to similar com-
binations and proportions of speakers of different varieties in the input of
different regions (with a predominance of speakers from southeast Eng-
land), which tended to lead to similar results. He highlights the fact that a
small number of seaports received the bulk of immigration (which would
become urban epicenters of leveling), and that these ports stayed in close
contact with each other and the hinterland. Trudgill (1986) also considers
the roughly similar case of Canadian English, and concludes that "uniform-
ity appears to be quite typical of the initial stages of mixed, colonial varie-
ties... with degree of uniformity being in inverse proportion to historical
depth" (Trudgill 1986: 145). Still, the evidence from the Origins of New
Zealand English project would seem to indicate that in the earliest stages a
good deal of interspeaker and intraspeaker variation is maintained in rural
areas, but that the similar combinations of settlers and contact between
communities then lead to similar results (Trudgill, Gordon, Lewis, and
Maclagan 2000).
Dillard (1985: 51-72) attempts to argue exactly along these lines in his
discussion of the origins of American (U.S.) English. He claims that, from
the founding of the first colonies to around the middle of the 18th century,
American English developed as a koine which was remarkably uniform
from New England to Georgia; it is only in the 19th century that signs of
diversity appear. Unfortunately, he bases this argument not on textual evi-
dence but on comments about colonial American English made by contem-
porary British travelers. The comments are quite positive and praise the
uniformity and "purity" of American English up to the end of the 18th cen-
tury, but turn quite negative afterwards. As Montgomery (1996) argues,
this probably tells us more about the culture and politics of the observers
than it does about the language observed, though he acknowledges that
differences in the colonies probably struck British observers as minimal,
given the high degree of dialect variation in the British Isles. In fact, most
scholars seem to agree that regional dialects formed fairly quickly, and that
this differentiation resulted from the differing origins of the immigrants,
their immigration patterns, and patterns of contact with Britain and be-
tween different colonies. Wolfram and Schilling-Estes (1998: 92-113),
who summarize earlier research, explain that the heavy immigration of the
Scotch-Irish to the Midland regions and Appalachia (fanning out from
Philadelphia) contributed greatly to the survival of non-prevocalic /-r/
82 Koines and koineization

(pronunciation of /-r/ in car, cart) in inland varieties of American English.

Other (older) areas such as eastern New England and Tidewater Virginia
were populated primarily by persons from southeastern England (with
some admixture of persons from southwestern England), and the dialects of
these areas became characterized by the typically southeast-of-England
loss of non-prevocalic /-r/, which was probably present in the original im-
migrant mix and reinforced by continued contact with London (as well as
factors discussed above for other varieties). Other dialect regions were also
clearly established during the colonial period, and it is supposed that initial
lack of contact contributed to the distinction between the dialects of east-
ern and western New England. Hence, it seems we should speak not of
koineization in American English but rather of multiple regional koineiza-
tions, some of which happened concurrently (e.g., Virginia and New Eng-
land), and others which occurred later (e.g., the Midwest, followed by the
American West, with a progressive decline in dialect variation from east to
west).52
There are then factors which favor koineization with regional dialect
leveling, sometimes over great expanses, as in Australia and the American
West, and factors which favor the creation of local differentiated koines.
Widespread homogeneity will be favored by:

1. A similar mix of immigrants in different locales: the same simple and ma-
jority solutions will tend to win out in each case.
2. Concurrent population of the zones in question: the contributing varieties
will also change over time, so a later mix with similar proportions of im-
migrants may lead to different results.
3. Extensive contact throughout the region and looseknit social networks in
all areas: numerous weak ties between speakers occupying urban centers
and between speakers occupying other urban and rural areas will help
spread changes, and the changes will not be easily resisted even if they
come from outside the community, since norm enforcement mechanisms
will initially be weak everywhere. In Milton Keynes, for example, children
are not developing a variety based simply on the majority forms among
their parents, but are rather developing a variety which is following the
dominant tendencies of widespread dialect leveling in southeast England
(favored not by koineization per se but by social mobility and autonomy).
In cases of regional koineization, we must assume that speakers of each
urban center affect speakers of other urban centers through weak ties, with
the largest centers of population being the most influential (cf. Callary
1975; Wolfram and Schilling-Estes 2003: 725). Especially in pre-modem
 Koineization and other contact phenomena 83

societies, we must also assume that contact and weak ties will be facilitated
by smaller size of the region undergoing koineization.
4. Development of transregional identity: to the extent that new settlers in dif-
ferent cities, towns and rural areas believe that they share a common iden-
tity, they will be more likely to resist innovations from outside the immedi-
ate community (cf. LePage and Tabouret-Keller 1985). Common culture
and political unity will favor the development of a common identity, as
will the perception of opposing identities outside the group.

Resistance to regional dialect leveling also occurs. This has been reported
for some areas of Canada:
Inland rural dialects often differ sharply from standard speech, especially in
the numerous rural communities in which the founders were Irish and Scots,
and inland working-class dialects differ not only from standard speech but
also from one another, with the ethnic origins of the founders cutting across
social class. (Chambers and Hardwick 1985, cited in Trudgill 1986: 146)
The factors favoring local koineization with resistance to regional dialect
leveling tend to be corollaries of those listed above:

1. A different mix of immigrants: this will tend to lead to different results.

2. Homogeneity: a relative lack of mixing will favor the successful transmis-
sion of a particular variety, and will reinforce ethnic identity.
3. Strong ethnic identity: this will favor resistance to accepting a broader re-
gional identity and speech.
4. Maintenance of closeknit social networks: if a whole community moves at
once, then its members may be able to maintain stronger ties and resist al-
tering identity and speech.
5. Political divisions within the region: this will tend to create distinct identi-
ties within the region
6. Different date of arrival: if a particular locale is settled either a generation
or more before or after other areas in the region, then the settlers in that lo-
cale may resist adopting changes current in other areas.
7. Isolation: if the population of the community is relatively isolated from
other koineizing communities, then a unique local koine is most likely to
arise (this underlies Trudgill's identification of new towns, frontiers, and
colonies as prototypical sites of koineization).
8. Continued contact with source dialects: if some locales in the koineizing
region remain more oriented towards the source varieties, then their con-
tacts will differ and they will develop differently.
84 Koines and koineization

These lists highlight the primary linguistic, social, and cultural factors that
contribute to resistance to koineization with dialect leveling, but it is im-
possible to list all the factors, and the researcher will need to consider
purely local factors as well.
While it is useful to have a conceptual distinction between dialect level-
ing and koineization, and to see that they interact, there remain cases where
no clear distinction may be possible. I refer to so-called "regional koines".
Are these the result of dialect leveling or prototypical koineization? No
general answer to this question is possible, since "regional koine" has re-
ceived no technical definition, and regional koines, whatever they are,
probably result from the interaction of both processes. The problem is
made more difficult by the fact that immigration to regional urban centers
is often (but not necessarily) slow, and in-migrants tend to come from sur-
rounding territories and thus speak similar dialects. Again, one possibility
for making a distinction is to use Kerswill and Williams' Principle 6,
which suggests that in koineization changes are rapid and abrupt, with
speakers of the oldest and youngest generation showing little overlap (for
changing features), while dialect leveling leads to more gradual change
across generations.

5.4. Standardization and standards

There exists long-standing confusion between the terms koine and standard
and a tendency to conflate or integrate the processes of koineization and
standardization. This is not surprising, as there are, indeed, similarities
between the two processes. Haugen (1966), in his classic essay, explains
that standardization occurs for functional and ideological reasons. Stan-
dards arise as supra-local language varieties in nation-states in order to
assure internal cohesion and external distinction. Socially, the process of
standardization begins with the selection of a dialect to serve as the base of
the standard. Following this, measures are taken to ensure this dialect's
acceptance by the population. Linguistically, standardization is character-
ized by codification, leading to minimal variation in form, and elaboration,
which is tied to maximal variation in function. Moreover, codification gen-
erally begins and is most effective in written versions of the language, so
literacy is a necessary concomitant of standardization. Unsurprisingly,
phonetic variation tends to be most resistant to standardization, even when
phonographic writing systems are employed.
 Koineization and other contact phenomena 85

An analysis of each of these aspects reveals a major source of confusion

between koineization and standardization. First, the process of selection is
a socially and politically thorny issue in any community. The group whose
variety is selected is automatically favored. If a widespread, stable koine
has already come into existence (and it is used by the most powerful
group), and no previous standard exists (or is rejected because of its struc-
tural distance and difficulty, as in some cases of diglossia), the choice of
the koine for the standard is a fairly easy one to make. Moreover, since the
bulk of the population will already use the koine, or will find learning it
relatively less difficult, choice of a koine will aid acceptance by the popu-
lation. In addition, a koine will already include a mixture of features from
other dialects and the elimination of much variation, which will aid the
task of codification. Finally, the koine is a structurally complete language,
so it will already be used for a wide range of functions and can be easily
extended through lexical and syntactic expansion to meet new ones. Obvi-
ously, if a koine has arisen spontaneously before standardization has be-
gun, the process of standardization (whether evolutionary or planned) will
be greatly facilitated. Nevertheless, the processes are conceptually distinct.
The distinction between standardization and koineization is highlighted
by Fontanella de Weinberg (1992: 42-54), who analyzes cases in which a
pre-existing standard or written prestige norm is present in the mix. Fon-
tanella argues that standardization can impede or erase the results of
koineization, and bases this claim on the unique developments in three
varieties of American Spanish: Mexican Spanish of the interior (relatively
conservative, standard), Paraguayan Spanish (highly non-standard), and
Argentine Spanish (an intermediate variety where later standardization
erased some of the effects of an earlier koineization). In the case of Mex-
ico, the pre-existent prestige norm of central northern Spain was a domi-
nant component in the prekoine linguistic pool of the colonial capital,
home to large numbers of prestige-oriented representatives of the vice-
regal government, church, and university. As a result, many (though not
all) features of other American koines (such as weakening and loss of syl-
lable-final /-s/, and confusion of liquids) were rejected in Mexico. On the
other hand, the early Buenos Aires koine was characterized by a strong
tendency to neutralize syllable-final /-r/ and /-l/ (found also in Andalusian
and Caribbean Spanish), but increasing literacy and influence of the stan-
dard (whose spelling was based on the pronunciation of northern central
Spain) during the 19th century led to the suppression of this simplified
feature in Argentine Spanish.53
86 Koines and koineization

Fontanella's claims about the interaction of koineization and standardi-

zation in Mexico find some support in Scholtmeijer's research on the de-
velopment of Dutch in the new polders of the reclaimed Zuider Zee,
founded between 1930 and the 1950s (Scholtmeijer 1992, 1997, reported in
Kerswill and Williams 2000: 74). In these areas, children speak very dif-
ferently from their immigrant parents and grandparents, but no new variety
or koine has developed in these areas (though Kerswill and Williams note
that Scholtmeijer may be overstating the case). Rather, children speak a
highly standardized form of Dutch, and Scholtmeijer ascribes this to the
combined influence of schooling in the standard and a need for a language
variety for external communication (1992: 145). On the other hand,
Kerswill and Williams note that the inhabitants of the polders have re-
mained in continued contact with surrounding regions, so dialect leveling
has probably played a role in the development of speech in the polders.
Significantly, move toward the standard is part of the more general process
of dialect leveling in the Netherlands.54

5.5. Language shift and language death

Language shift occurs when an entire speech community acquires a new

language variety and abandons its former community language. Substrate
theory has focused on the effect of shift on the target language into which
features from the source are accepted. According to Thomason and Kauf-
man (1988: 119-119), substratum interference tends to lead to alteration of
phonological and abstract syntactic patterns, but not to the borrowing of
morpholexical items. Presumably, dialect shift is not essentially different
from language shift, and koineization can certainly be seen as a particular
instance of complex language shift. However, koineization involves the
mixing of vocabulary from different sources, which is not predicted, and,
unlike prototypical cases of shift, there may be no one target. As we have
seen, this tends to favor processes of leveling and simplification, and ulti-
mately the creation of a new variety. On the other hand, since majority
features tend to be selected from the linguistic pool, it is essential to con-
sider the effects of shift (acquisition) by large groups of speakers in the
demographic mix, particularly when this involves language shift in the
strict sense, i.e., the learning of the "language" by speakers of unrelated
languages.
 Koineization and other contact phenomena 87

When the speakers of an unrelated language in the mix form a signifi-

cant but minority presence (e.g., Basque learners of Romance in medieval
Castile), it is appropriate to include their production in the prekoine pool,
for they (or, rather, their children) will have adequate access to input and
features of their speech will not necessarily be more marked than those of
speakers of established dialects. Nevertheless, clear influence of a minority
of "non-native" speakers will tend to be washed out. The unique social
context of koineization might seem to provide ideal circumstances for sub-
strate influence by weakening potential social resistance to substrate-based
innovations, but it also weakens the possibility of selection of substrate
features unless a) they are also found in the speech of others, and b) the
substrate features are themselves simplifications that are favored by other
learners in the community, particularly (older) children. I explore this issue
with reference to Basque influence on the development of Castilian in
Chapters 3 and 4.
Language death studies have looked at shift from the opposite perspec-
tive: the effects of shift on the minority language during its abandonment
by a speech community (Dorian 1981). In the best known version of this
process, the language suffers a progressive reduction in its functional do-
main as parents cease to use certain forms and children do not learn them;
extensive simplification and even impoverishment occurs. Silva-Corvalán
(1994) argues that contact between English and Spanish in Los Angeles
(where the language is maintained at the community level thanks to new
immigration, but dies within families) has led to an exaggeration of certain
"internal" leveling and simplifying tendencies of Spanish, rather than
wholesale transfer of English forms and structures to Spanish. The effects
of language death can therefore be difficult to distinguish from those of
koineization. Moreover, In Los Angeles, immigrants have come from many
parts of Mexico and other Latin American countries, and it is probable that
koineization has occurred at the same time as language death. Gambhir
(1981) indicates that Guyanese Bhojpuri, as it is gradually replaced by
Guyanese Creole, is also undergoing language death, making it impossible
to clearly distinguish (in the contemporary language) the original effects of
koineization from the subsequent effects of language death.
88 Koines and koineization

5.6. Convergence

Koineization is often considered to be an example of linguistic conver-

gence, but only in the broad sense of the term. Convergence as a technical
term is now often used to refer to the effects of stable bilingualism or mul-
tilingualism on the structures of clearly distinct languages. The classic
example is the case of Kupwar, India (Gumperz and Wilson 1971). In this
multilingual community, the languages of Marathi, Urdu and Kannada
(unrelated to the first two) each retain a distinct lexicon and morphological
markers, but their use by multilingual speakers over generations has led to
complete parallelism in the abstract syntactic structure, so that in general
sentences can be translated word for word. The motivation for this conver-
gence is strikingly different from that of koineization, since accommoda-
tion appears to have played little part. Rather, a conscious desire to retain
separate identities and languages was reinforced by close-knit social net-
works, while patterns of intergroup relations forced multilingualism on the
members of the community, and in their interlanguages they (probably
unconsciously) transferred patterns across the different languages despite
their efforts to keep them apart.

5.7. Borrowing

Thomason and Kaufman (1988: 74-75) define borrowing on a five-level

scale from 1, "casual contact" with only lexical borrowing, to 5, "very
strong cultural pressure" characterized by heavy structural borrowing. The
case of Kupwar, for example, would be placed on this scale at 4, "strong
cultural pressure and moderate structural borrowing" (due to what is seen
as the relatively low typological distance between the languages involved).
Normally, however, the term borrowing has a more limited meaning close
to "1" on the Thomason and Kaufman scale. In a language that has under-
gone koineization, the results of mixing could easily be confused with ex-
tensive borrowing, although both will be present. Any study of koineiza-
tion will need to distinguish, to the degree possible, between the results of
mixing and the results of borrowing. For many items or features this may
be impossible, but for others knowledge of the time of their appearance and
the contemporary social circumstances may allow a determination of their
origin as borrowings or as items that survived the process of mixing in
koineization.
 The model and its use 89

5.8. Interacting processes and the study of koineization

From the above discussion we can conclude that koineization must be un-
derstood as a prototype of change, and that we should not expect to find
"pure" examples of koineization in real cases of linguistic change. We will
find koineization being affected by a range of varying linguistic, social,
and cultural factors (aside from more stable cognitive factors), and it will
interact with other processes of contact-induced change. A central chal-
lenge will therefore be to distinguish effects of koineization from those of
other processes, or, at least, to show in a given case why this cannot be
done (i.e., how two or more processes might all favor a similar outcome).
This is a problem in any investigation of causes, but, as we have seen, there
are places and situations in which the effects of koineization are likely to
be particularly clear-cut. Still, the researcher is only too likely to find in-
teraction with other factors. Siegel highlights the influence of a pidginized
variety of Hindi on the development of Fiji Hindi, Blanc the effects of
multiple language shifts on Israeli Hebrew, Fontanella the impact of stan-
dardization on the development of American varieties of Spanish, Gambhir
the effects of language death on Guyanese Bhojpuri, and Kerswill and
Williams the influence of regional dialect leveling on Milton Keynes
speech. The frequency with which we find such combinations of influence
means that they probably represent the messy norm, but the model is de-
signed not to exactly simulate reality, but to help us tease out patterned
threads of cause and effect that might otherwise remain hidden.

6. The model and its use

In the preceding I have sought to review previous conceptualizations of

koineization, and to elaborate those views with clarifications and sugges-
tions of my own. The intention, of course, has been to arrive at a thorough
definition of koineization and an understanding of how koineization works.
This is particularly important for the purposes of this study, for my goal is
to apply the model at relatively great time-depths of between 1100 and 700
years, to reconstruct the sociolinguistic contexts of these periods (to the
degree necessary and possible), and to reinterpret previous scholarly dis-
cussion of the changes and periods under study here. Still, it seems that
there may exist potential abuses or hazards in applying the model to the
historical record, so I have also sought to define basic methodological
90 Koines and koineization

guidelines for its use. Below I present the definition and a discussion of
methodological guidelines.

6.1. Definition of koineization

Koineization is best conceived as a two-tiered model or process. On the

micro-level speakers of different linguistic subsystems interact in a new
community. Such mixing is especially likely during large population
movements to new towns, frontiers, and colonies, and is accompanied by a
sudden breakdown in social networks and, consequently, norm enforce-
ment. The language varieties are mutually intelligible (or quickly become
so through exposure), and include regional varieties, sociolects, adult inter-
languages, and child language. In this mixed social environment, variation
in the linguistic pool peaks at the very time that social networks are weak-
ened and norm enforcement declines. As adult speakers attempt to interact
in the new community and establish social ties, they accommodate to the
speech of their interlocutors, often eliminating non-functional or minority
variants from their speech; they may also seek to reproduce new features
that are perceived as frequent and/or salient in the prekoine, thereby neu-
tralizing the original social value of many variants, and possibly introduc-
ing interlanguage (interdialect) variants (developing frequency of variants
can be favored by high salience, both perceptual and, especially, sociolin-
guistic). In this contact- and input-rich environment, children generally
learn what they perceive to be the most frequent and consistently produced
features, including those found in their own learner language (such as sim-
plified, transparent, regular forms and/or those based on incorrect induc-
tions or analyses). As older children shift their social-psychological orien-
tation from parents to peers, they accommodate their speech to that of their
peers to the degree that they can, introducing more interdialectal forms as
they learn and generally favoring majority (and/or salient) features, what-
ever their origin may be. It is among the older children and adolescents of
the first generation that regularities of usage begin to appear, but it may
take another generation of children for all norms to focus.
Over time, therefore, language acquisition, accommodation, and re-
solidification of social networks (beginning with adolescent peer groups)
lead to the establishment of new linguistic norms, or systemic changes at
the macro- or linguistic level. These include mixing/leveling and simplifi-
cation. Mixing refers to the survival in the resultant koine of variants from
 The model and its use 91

different contributing varieties. Leveling refers to the survival of one vari-

ant from a pool of competing variants. Reallocation occurs when more than
one competing variant survives but each with a different function (social,
stylistic, geographical, phonological) or meaning. Simplification is under-
stood as limited reduction in inventories of units and rules, generally seen
as greater regularity or transparency in the grammar. These macro-level
phenomena are explained as the unintended results of micro-level speaker
activity, such as accommodation, constrained by the process of acquisition
and the particular social context. It is the unique social context that allows
for the formation, in only one or two generations, of a relatively mixed,
simplified (though socio-stylistically variable), and focused koine.

6.2. Methodological guidelines

A well-defined model is of great use for reconstruction of the social condi-

tions of past changes. However, there remain potential problems. First, any
and all of the particular types of changes that characterize koineization can
occur in language varieties which clearly have not been affected by it. For
example, the effects of mixing and borrowing could easily be confused,
and examples of simplifying changes can certainly be found even in con-
servative varieties used by isolated closeknit communities. The only way to
be (relatively more) certain that a particular change is causally related to
koineization is by situating it within a larger analysis of both macro-level
linguistic phenomena and micro-level social conditions and constraints on
speaker activity.
The researcher must therefore begin with an accurate description of the
linguistic changes themselves, specifying (minimally) that:

1. the changes reflect mixing/leveling and simplification;

2. several (preferably numerous) such changes co-occur (and changes which
add opacity and complexity are few, and explicable within the constraints
of theframeworkdescribed above);
3. the features are selected and focus rapidly, normally over the course of just
2 or 3 generations (=1 or 2 generations of children).

The primary prediction of the model is that all changes will tend toward
mixing and simplification. Therefore it is necessary to show that numerous
changes correspond with this prediction. Analysis of an isolated case of
change in terms of koineization is not sufficient to show that the change in
92 Koines and koineization

question is causally related to koineization. Of course, such focused analy-

sis may be necessary on a provisional basis (it is impossible to do every-
thing at once), but the goal should be to identify broad patterns of change.
Only this will lend validity to the argument that any particular change is
the outcome of koineization. Moreover, if numerous counterexamples to
predicted patterns are found, then it will be difficult to analyze even sim-
plifying changes in terms of koineization (though relative chronologies and
interaction with other processes of change may be considered in such
cases). Finally, if a change is not complete within the prototypical time
scale of koineization, then it will be far more difficult to establish a causal
relation between social changes and linguistic changes. If a change does
not occur (or a feature focus) until hundreds of years after the original
demographic mixing, it will be difficult if not impossible to establish a
convincing causal link between the events. If such extensions in the time
scale are allowed, then any recent changes in any modern variety of
American Spanish and English might be linked to koineization, but this, of
course, is of no real explanatory value (but see Chapter 6 for discussion of
chain reactions or drift).
Clearly, however, the core of an explanation based on koineization is
the specification of micro-level conditions of speaker activity and the link-
ing of (collective) speaker activity to the linguistic results. This must in-
clude an understanding of the constraints imposed on speakers by the ac-
quisition process (which remains constant), but the explanation of
linguistic change will depend on how speaker-learners are affected by the
changing social context. Thus, the variable aspects of the micro-level that
must be demonstrated include the following:

1. an increase in variation due to demographic and dialect mixing;

2. a breakdown in social networks and consequently in norm enforcement;
3. specification of the contributing varieties and the relative proportions of
speakers of each;
4. a specification of the lexicon and structures of the contributing varieties,
including possible learner varieties.

But as the above review has shown, numerous other factors may have to be
taken into account, such as the isolation of the koineizing community from
the source communities, contacts with new communities, the potential
(cultural) impact of a "foreign" threat that might contribute to community
solidarity, and, in general, any factor which affects the development of
community identity. Kerswill and Williams also assume that an exact un-
 The model and its use 93

derstanding of the social and linguistic history of the koineizing commu-

nity is necessary; for them, an ideal social history will include the follow-
ing information:

1. The original population of the area: its size, its social characteristics, and
its speech forms.
2. The size of the incoming population in relation to the original population.
3. The abruptness of the settlement: Was it sudden and fmite, or did it con-
tinue over a long period? Did it continue after koineization had taken
place?
4. The proportion of children to adults among the incomers and the original
population, and the rate at which children were bom to the incomers after
migration.
5. The continued contacts of the incomers with their place of origin: Did they
break off relations with their original home completely, or did they main-
tain links with it to the exclusion of new, local contacts?
6. The social characteristics and ethnicity of the incomers: Did they come to
take up specific jobs, e.g., in a new industry? Were they socially mixed?
Were they an ethnically distinct group?
7. The speech of the incomers: Was it diverse or homogeneous? Was it simi-
lar to that of the native population? Were some social dialects better repre-
sented than others?
(Kerswill and Williams 2000: 70)

Kerswill and Williams acknowledge that finding precise answers to all

these questions is probably not possible, even in contemporary cases of
koineization. However, even when they cannot be answered, it is important
to keep such questions in mind when exploring the impact of the social
context on language use and change during koineization. Indeed, the same
can be said for several of the methodological guidelines listed above. For
example, in following chapters I analyze three periods of koineization that
affected medieval Castilian (or varieties of Castilian). For each case or
stage, I analyze and reconstruct several groups of changes that (I believe)
were linked to koineization, as well as some which were not (or only indi-
rectly). Such groups of changes are investigated in accordance with the
methodological restriction that particular koineizing changes must occur as
part of a broad pattern of koineizing change.
Chapter 3
The Burgos phase

It has long been recognized that many linguistic features which came to
characterize Castilian first arose between the 9th and 11th centuries in
central northern Spain, in the region surrounding the city of Burgos. In this
chapter, I argue that the model of koineization contributes significantly to
our understanding of several of the key changes that took place in that time
and place. In order to do so, the corresponding social and demographic
conditions must be shown to apply; the chapter therefore begins with a
social history of the Burgos phase. Since this period has long been an ob-
ject of historical linguistic study, it is also necessary to review and critique
previous scholarly discussions of issues related to dialect mixing or koinei-
zation in this period, and, more generally, early medieval Spanish. In line
with the methodological guidelines outlined in the preceding chapter, I
analyze and explain as koineizing developments two groups of linguistic
changes of Burgos-phase Castilian: loss of preposition + article contrac-
tions, and simplification of the vowel system. Finally, in order to explore
the explanatory limits of the model, I discuss three other changes often
associated with this period and consider to what degree the model may - or
may not - lead to a better understanding of them.

1. Social history

The Moorish invasion of Visigothic Hispania in 711 initiated a series of

events that were to have tremendous linguistic effects on the Romance
spoken there.55 The Moslem Arabs and their Berber allies overran the en-
tire Peninsula in only a few years, and the great bulk of the (Christian)
Romance-speaking population chose to remain under their relatively toler-
ant control. However, the Moors failed to subdue the less attractive moun-
tainous regions in the North. This was particularly true of the areas which
were (or had been) inhabited by the Asturs, Cantabrians, and Basques. It is
often assumed that great numbers of Visigoths fled to these northern re-
gions; however, their numbers were probably not at first large, since north-
ern groups had long resisted Visigothic dominion. On the other hand, there
 Social history 95

is some evidence that a pact of mutual cooperation and resistance was

reached between the Asturians led by Pelayo and the local Visigothic com-
mander, Duke Pedro of Cantabria. His son, wedded to the daughter of
Pelayo, would become Alfonso I, king of the mountain kingdom of As-
turias.
The Cantabrian Shelf had never been thoroughly integrated into His-
pano-Roman culture, and remained thoroughly rural even after the Moorish
invasion: only Oviedo and Lugo in Galicia could have been considered
administrative centers, and even these were founded or repopulated only in
the mid-eighth century. It is assumed today that Latin/Romance had been
established in some form in the areas of Galicia, Asturias, and Cantabria
before the Moorish invasion, but increased contact with refugees from the
South may have contributed to even greater use of Latin/Romance and the
final elimination of any surviving pre-Roman languages, save Basque
(Ridruejo 1995: 241-2). 56 The northeastern Mediterranean coastal area of
Catalonia had long been heavily romanized and Romance-speaking, and
remained so throughout the brief Moorish occupation of the region.
Later in the 8th century, the Asturian kingdom grew strong enough to
organize raids into the Duero valley to the south of the mountains, which
after mid-century had no strong Moorish presence. According to medieval
chronicles, Alfonso I cleared the Duero valley and brought its remaining
inhabitants back to the kingdom of Oviedo. Still, nearly a hundred more
years would pass before significant population movement towards the
south took place, and the Duero Valley was left as a defensive "almost-no-
man's-land". During this time, the Asturians were able to extend their con-
trol to the adjoining mountainous regions of Galicia to the west and Can-
tabria to the east, even though these regions were not well connected by
land. Other northern Christian kingdoms and counties also appeared. The
late eighth century saw the revolt of the population of Navarre, its rise as a
separate kingdom (aided in part by the Franks), and the integration of the
eastern Catalan counties into the Carolingian Hispanic March.
The isolation imposed by life in mountain valleys - found across the
whole of Christian territory, from the Pyrenees in the east to the Cantabrian
Mountains and northern Galicia in the west - probably laid the base for the
subsequent political split into different kingdoms and counties, as well as
contributing to the gradual development of divisions between the linguistic
varieties of the different communities (Ridruejo 1995: 247). Wright
(1994d;1994f) claims that at this time, and even for centuries after, the
linguistic differences between Romance-speaking regions consisted pri-
96 The Burgos phase

marily of statistical variation in the frequency of occurrence of certain

variables: differences of degree rather than type. Nevertheless, social isola-
tion and closeknit social networks in the northern mountain regions proba-
bly contributed over time to ever-growing divisions among the different
varieties, differences that would need to be overcome when the inhabitants
of the mountain kingdoms began to move south and mix together.
Population movement to the south was initiated by inhabitants of the
regions of Galicia and Asturias. It began for a variety of reasons. The size
of the northern populations began to rise (especially in Galicia) and out-
strip the ability of the surrounding landscape to support it. Moreover, the
Duero valley which lay to the south of Asturias and Cantabria was largely
uninhabited, and contained vast areas of land that could be used for farm-
ing or grazing of livestock. Still, the region was open to attack from the
Moors, so the first attempts at repopulation occurred in the west, farther
from the Moorish threat, and Galicians began to move south into the Miño
valley even in the early ninth century. At roughly the same time, the As-
turians initiated repopulation in the eastern end of the kingdom, in the
small river valleys of eastern Cantabria. However, the more northern of
these valleys were open to attack from northern Basques (not Christianized
and controlled until about 1000), and the valleys towards to south (along
with the Basque area of Álava to the east) were open to attack from the
Moorish-controlled Ebro valley. Fortifications were therefore built to de-
fend these valleys, and settlement continued southward to the Ebro. The
region acquired the name of Castilla 'land of castles', and this name was
eventually extended further southward with reconquest and repopulation
(in what follows it is referred to as primitive Castile).
In the second half of the 9th century the Asturian kings began to spon-
sor repopulation of the southern regions, and they ordered the
(re)establishment of such cities as Astorga (in 853), Oporto (in 869), and
León (in 866, after a failed attempt in 856). The majority of the population
of each of these frontier regions tended to be drawn from the areas just to
the north. In the west, settlers from northern Galicia constituted the most
significant portion of the new population near the Miño, though Asturians
also arrived. These immigrants may have found some of the "reconquered"
land occupied by communities surviving from before the Moorish invasion,
and it is known that Mozarabs escaping the progressively less tolerant
Moorish south also relocated to Galicia. In the repopulation of Leon, the
settlers from Asturias were socially and numerically dominant, but they
were joined by numerous westerners from Galicia. A few settlers from
 Social history 97

Navarre (and/or Basques) also moved into this zone, but the Mozarab im-
migration (particularly from Toledo) was especially intense in Leon, where
political and ecclesiastical institutions where centered.57 It is generally
believed that Mozarab monks were responsible for creating the idea of the
Christian reconquista and "salvation of Spain", as well as the belief that
the Leonese kings were the direct inheritors of the Visigothic tradition.
Reconquest and repopulation in Leon and Galicia were heavily pro-
moted by official institutions: the king, the nobility, and the church. As in
all areas of repopulation, there were also many independent presuras by
hombres libres (see discussion below), but the official presence was espe-
cially intense in this zone:
los últimos reyes astures dedicaron una gran parte de su actividad desbor-
dante a la reorganización de la tierra leonesa, efectuando personalmente la
repoblación de viejas ciudades abandonadas e impulsando eficazmente la
del medio rural a través del estímulo o la confirmación de las presuras reali-
zadas. (Moxó 1979: 56)

This official presence was reinforced when the Asturian court moved south
to the city of León in 914. The activity of the kings was reinforced by that
of the nobles in both Leon and Galicia, though particularly in Galicia no-
bles entered into conflict with the king. The establishment of large num-
bers of monasteries also reinforced the important institutional presence in
Leon and Galicia.
To the immediate east of Leon lay the region of Castile (to which we
will soon turn), and beyond it, the Christian states or regions of Navarre,
Aragon and Catalonia, which saw little demographic movement in the 9th
century. The Catalan counties, or Spanish March, had close ties to the
Carolingians and their Occitan neighbors, and they sought to maintain
peaceful relations with their powerful Moorish neighbors. In the central
Pyrenees, the county of Aragon, dependent on Navarre, appeared in the
early ninth century. A thinly populated region throughout its history, its
expansion took place extremely slowly, with significant advances only
being made in the eleventh century, during the second or Toledo phase (as
I term it here) of the reconquest.
Closer to the center, Navarre was prevented from making any signifi-
cant expansion to the south by the powerful Moorish presence, particularly
in the kingdom of Zaragoza, with whose rulers it maintained a truce
throughout the 9th century: "el fenómeno de repoblación es más débil y
atiende más a la restauración de los sectores devastados por los ejércitos
musulmanes en sus campañas que a una actividad repobladora meditada y
98 The Burgos phase

sostenida" (Μοχό 1979: 80). Moreover, the idea of reconquest only began
to penetrate Navarre in the 10th century, and the slow pace of the recon-
quest kept the kingdom of Navarre limited to its original confines until its
conquest of La Rioja in 923 (until then Castile would remain a destination
for Navarrese seeking land). Navarre was split between a Romance-
speaking southeast near or in the Ebro valley (an areas which grew with the
reconquest of La Rioja) and a Basque-speaking Northwest. The Basque
presence was extremely important in early Navarre:
Observando . . . el área geográfica del que fue reino de Pamplona, vemos
fácilmente cómo dentro de su territorio habitaban gentes de estirpe vascona,
que ofrecían un distinto grado en su proceso de romanización, por cuanto
éste se había desarrollado por la zona llana hasta Pamplona, sin alcanzar to-
dos los medios rurales y muy especialmente los de la Navarra alta o húmeda,
donde perduraba su arcaica organización tradicional, la lengua vasca, sus
estructuras socio-económicas e incluso su paganismo. (Moxó 1979: 80)
When La Rioja (a region bordering Castile) was taken, the Basques moved
in significant numbers, and there is evidence that some isolated communi-
ties continued to use Basque there until the 13th century. Nevertheless, La
Rioja was dominated at all times by Romance speech, probably brought by
southern Navarrese and used by Mozarab communities that had remained
in the region under Moorish rule.
Between Navarre (including the present-day Basque Country) and As-
turias proper lay the region of Cantabria. Cantabria, though a territory of
Asturias-Leon, was on the periphery of the kingdom and not closely con-
trolled by the institutions of the Leonese monarchy. Repopulation focused
first on eastern Cantabria (different parts of which became known as As-
turias de Santillana, Asturias de Transmiera, and Asturias de Laredo, al-
though Castilians would use the name La Montaña to refer to most of the
area now known as Cantabria). Settlers moved in from Asturias and west-
ern Cantabria (probably joining some residual inhabitants) to take advan-
tage of the land that became available, sometimes as part of efforts organ-
ized by small monasteries located in more secure areas (Mestre Campi and
Sabaté 1998:15). But the south-facing valleys of this region (in primitive
Castile) also attracted settlers and their movements are attested to in docu-
mentary evidence; for example, in 814 a group of colonists arrived from
the valleys of the Picos de Europa, a region located to the west at the bor-
der of modern Asturias and Cantabria (Moxó 1979: 64-65). Settlers such
as these were joined by Basques (parts of the southern Basque region of
Alava were also under the nominal control of Oviedo) and romance-
 Social history 99

speaking Mozarabs from farther down the Ebro valley. The area of primi-
tive Castile was repopulated by the beginning of the ninth century, and the
open region further south then began to draw immigrants on a larger scale.
This area between the Ebro and Duero rivers had also been known to the
Romans as Bardulia, but new settlers extended to it the name of Castilla
(Castile). Though perhaps well populated in Roman times, archeological
evidence indicates that in the Visigothic era it was largely abandoned, ex-
cept for a few small population centers such as Clunia and Osma, near the
Duero (Pastor 1996: 40-41). After the Moorish invasion, it was left yet
more decimated, since Arab armies moving to attack Asturias-Leon pre-
ferred to enter through this area, either from the east along the Ebro valley
or from the southeast, through the mountain passes of the Central System
and across the Duero.
Repopulation had been successful in the mountain valleys of primitive
Castile, but the relative isolation of the more southerly regions and their
exposure to Moorish attacks were significant detractors to repopulation.58
In an effort to protect an important route and encourage repopulation, Or-
doño Π ordered the population of the new fortress-city of Amaya (founded
in 860), which lay south of the mountains. Further south, Alfonso ΙΠ or-
dered the founding in 884 of Burgos, destined to become the new county's
main city. Reconquest and repopulation continued, and the inhabitants of
Castile secured control of the north bank of the Duero River by 912 (forty
years after the Galicians reached the Duero). Repopulation around the cen-
tral area of Burgos appears to have happened extremely rapidly, and was
especially intense during the first years of the 10th century. Most inhabi-
tants moved to small villages, which typically had some 25-35 property
holders and their families (though size differed greatly). Such inhabitants
were not isolated from other communities. Demographic studies of Castile
in the year 1000 indicate that on average there was one village every 5 or 6
square kilometers. Some areas were much more densely populated; for
example, in the Burgos area the distance between neighboring villages was
on average about two-thirds of a kilometer (Pastor 1996: 64).
The first and most numerous settlers of Castile are presumed to have
come from Cantabria and primitive Castile (at each stage of repopulation,
the largest number of colonists in each area tended to come from areas
directly to the north). Asturians and Leonese also settled in the new county,
especially in the western portions, and, at least for later phases (when at-
tempts began to repopulate regions south of the Duero), there is evidence
that Galicians participated in the settlement of Castile (Barrios 1985). As
100 The Burgos phase

in Galicia, Leon and Navarre, Mozarabs from the south and, more proba-
bly, from the east, moved to the new county of Castile (a few Mozarabic
names such as Abolmóndar are frequent in the documentation), but their
institutional monastic importance is less evident here. From the east there
also arrived Romance- and Basque-speakers from Pamplona/Navarre, as
well as Basque-speakers from nearby region of Álava; their presence in the
demographic mix of Castile has often been emphasized by historians of
Spanish, most notably Menéndez Pidal.59 Finally, we should note that it is
sometimes argued that the region of Castile retained some inhabited com-
munities from before the Moorish conquest, but, if so, their numbers must
have been quite low.
The rapid arrival of immigrants of such regional and dialectal diversity
(particularly near Burgos and areas south and west) increased the variation
in social and linguistic norms (i.e., the number of competing variants in the
linguistic pool) at the same time that it weakened social networks: the very
conditions that promote koineization. The weakening of norm enforcement
was probably exacerbated by the relative isolation of Castile from Leon.
The monarchy, though exercising influence, was unable to maintain the
control that it held in Leon proper. It also had difficulties maintaining con-
trol in Galicia, but for very different reasons. In Galicia, the relative stabil-
ity of the society had allowed for the maintenance and reinforcement of a
dominant nobility and this group resisted royal authority, but did not break
away from the Leonese crown. Though Castile benefited from royal initia-
tives, these tended to be controlled by local counts (who fought among
themselves for power), and even the monasteries (e.g., the famed San
Pedro de Cardeña founded near Burgos at the beginning of the 10th cen-
tury) tended to support the local counts more than the Leonese kings.
Castile is said to have expanded much more spontaneously than did
Galicia, Leon or Navarre, in that it tended to attract settlers who were will-
ing to take greater risks, and to repel those who were not:
El carácter fronterizo de Castilla no anima, al menos hasta época tardía, a
instalarse en ella ni a la vieja nobleza visigoda ni a los clérigos mozárabes
huidos de Córdoba, y en Castilla no existirán grandes linajes ni proliferarán,
como en León, los monasterios y las grandes sedes episcopales que son los
dueños de la tierra, de la riqueza, y poseen la fuerza necesaria para someter
a los campesinos libres. (Tusell, Martínez Shaw, and Martín, 1998: 102)
This unique development (limited almost surely to only the initial stages)
was reflected in its social structure. The weak presence of the monarchy
and the lack of an established nobility in Castile led Sánchez Albornoz
 Social history 101

(1966) to characterize early Castilian society as socially "horizontal" -

with relatively little vertical or social differentiation and with most repopu-
lation being carried out by hombres libres under the system of presuras, or
small private settlements of land. The predominance of the hombres libres
('free men', normally small landowners) was accompanied by a corre-
sponding decline in serfdom (or slavery) in Castile.60
Though Sánchez Albornoz emphasized social leveling, it is perhaps
more appropriate to speak of social mobility, since early Castilian society
was clearly hierarchized. It must be remembered that Castile, more than
Galicia and Leon, was dominated by constant warfare from its beginning,
for it was directly in the path of the armies sent by the Moors each summer
(and later it entered into conflict with other Christian kingdoms). This con-
stant warfare required that the local counts be given a great deal of author-
ity, and it made indispensable the presence of "free men" who could and
would fight. All free men were required to contribute service towards the
defense of the county, including preparing for and fighting in battle, serv-
ing as lookouts, and helping with building and repair of defensive struc-
tures. However, military service required a horse and arms, and only the
better off could afford these. The most successful soldiers tended to be
perceived as the most important and useful members of society, and they
acquired more booty from raids in Moorish territory. These conditions led
to the appearance of a new influential class of low-level aristocracy, the
infanzones, a social rank that would spread from Castile to other areas
(Μοχό 1979: 148).61 Just below them in this functional social hierarchy
there appeared a class of peasant-soldiers known as caballeros villanos,
who owned horse and arms but also worked the land. The least influential
free men were assigned the less prestigious duties of infantry service,
building, and maintenance. With time, these social distinctions became
hereditary, and large estates became more common (though still frag-
mented), as small property owners were forced to sell their land to monas-
teries or large landowners. Still, in the early stages Castile was a socially
mobile society.
The social structure of early Castile was reflected in and reinforced by
changing family structure. In the northern mountain regions, as in most
stable rural societies, the extended family constituted a closeknit social
network that protected and restrained the individual. However, the move-
ment southward had a disruptive effect on such extended families in all the
Christian regions:
102 The Burgos phase

Parece evidente, asimismo, en la esfera general de la sociedad campesina el

deslizamiento - en la propia alta Edad Media - desde la familia indivisa ha-
cia la familia conyugal, cada vez más típica del círculo social no aris-
tocrático, limitada a los cónyuges y a aquellos hijos que vivían en la casa pa-
terna, en cuya transformación influyó la dinámica de la repoblación al
disgregar territorialmente a antiguos miembros de la comunidad familiar,
que se instalaban en lugares dispersos, debilitando la lejanía el vigor de los
originarios lazos de sangre. (Moxó 1979: 129)

Extended families were more strongly maintained, even in the face of mi-
gration and over distances, by the aristocratic families of Leon and Galicia,
and by peasant families where there were collective presuras or aprisiones,
as in the Pyrenees. However, given the other features of Castilian society,
the increase in the number of nuclear families must have contributed even
more in this region to the disruption of traditional norms.
The greater heterogeneity in the origins of early Castilian colonists, and
the dominance of looseknit social networks in a mobile society are features
that, by their nature, link early Castilian society (and language, of course)
to those of other regions, particularly Leon, and at the same time, by their
degree, set Castile apart. For Castile, in effect, became a prototypical fron-
tier society.62 Its location on the periphery of the kingdom of Leon, as well
as that of Pamplona-Navarre, and its openness to attack by the Moors along
both the Ebro and Duero, fed a sense of in-group solidarity and separate
identity which eventually manifested itself in opposition to outside norms.
This can be seen in the 10th-century rise of count Fernán González, who
managed to gain control of the entire county of Castile and openly opposed
the kings of both Leon and Navarre (even as he arranged marriages be-
tween the royal houses of each and his own family).
This oppositional identity is illustrated in the ancient legends of rivalry
between Castilians and Leonese. These are found in many sources, from
chronicles to sung ballads. One such ballad establishes clearly the conflict
between the two groups:

Castellanos y leoneses tienen grandes divisiones,

el conde Fernán González y el buen rey don Sancho Ordóñez
sobre el partir de las tierras, ahí pasan malas razones:
llámanse de hideputas, hijos de padres traidores.
 Social history 103

In the lines where Fernán González addresses the king, the poem sets up a
series of contrasts between the urbane or sophisticated Leonese and the
intentionally down-to-earth Castilian, ready for battle:

vos venís en gruesa mula, yo en ligero caballo

vos traéis sayo de seda, yo traigo un arnés trenzado
vos traéis alfanje de oro, yo traigo lanza en mi mano
vos traéis cetro de rey, yo un venablo acerado;
vos con guantes olorosos, yo con los de acero claro;
vos con la gorra de fiesta, yo con un casco afinado;
vos traéis ciento de mula, yo trescientos de caballo.
(Díaz Roig 1987: 123-124)

Though this ballad was certainly composed much later than the period we
are now studying, it probably does reflect quite accurately the kinds of
myths that developed in the period and which contributed to the formation
of a Castilian identity.63
There are other legendary traditions which reflect a breakdown in norm
enforcement and rise of a regional identity. Menéndez Pidal is well known
for having emphasized the unique nature of Castilian society (but see be-
low); one of the clearest examples he gave was the supposed rejection by
Castilians of the traditional legal code of Visigothic Spain, the Fuero
Juzgo, which was used in all the other Christian regions (Menéndez Pidal
1964: 474-475). Though in fact this rejection may never have occurred in
the dramatic way that the myths tell us, it is true that Castile tended to rely
much more on customary law than other regions, and in this sense showed
once again a greater willingness to depart from traditional norms.
From the above, we can conclude that the unique social environment of
Castile was conducive to koineization with rapid focusing. First and fore-
most, there was mixing of Romance speakers from different regions across
the north (especially Cantabria, but also Asturias, Navarre, Leon, and, later
and to a lesser extent, Galicia), as well as speakers of Basque from the
regions to the northeast. Though in-migration began in the north in 9th
century (and perhaps even before), its intensity peaked later (in the year
900 and the years following) and further south (near and beyond Burgos).
Settlers tended to locate in villages which were near other villages, in a
dense network of small populated centers. These conditions led to the ini-
tial dominance of looseknit social networks in a larger society which facili-
tated contact between speakers of all ages, particularly within and between
104 The Burgos phase

villages, but also throughout the county (older boys and men from across
the region certainly interacted frequently as they fulfilled their military
duties). On the other hand, there were pressures favoring the rapid focusing
of new regional norms. Castile was located in a peripheral zone not easily
controlled from the prestige centers (and less influenced by prestige
norms),64 but its inhabitants were engaged in constant warfare, first with
the Moors, and soon with neighboring Christian regions. Moreover, local
leaders, notably count Fernán González, managed to convert the region
into a distinct political entity during the mid 10th century (an event which
was followed by the establishment of a separate kingdom in the 11th cen-
tury). These factors would have favored the development of a certain re-
gional identity, which in turn would have contributed to rapid focusing of
regional norms.65

2. Previous discussion of language change in early Castile

While the preceding social history is conceived in terms of koineization,

awareness of the unique social situation of early Castile is not new. In fact,
Castilian exceptionalism has been a favorite theme of traditional Spanish
historians such as Sánchez Albornoz and Menéndez Pidal. Menéndez Pidal
also sought to relate the linguistic development of the Castilian dialect to
his views on the place of Castile in Spanish history. Influenced by the
Generation of 1898, Menéndez Pidal came to identify the "spirit" of Cas-
tile as the principal unifying and stabilizing force of Spain (cf. Mitre 1999:
94). Similarly, he came to see the Castilian dialect as an exceptional, unify-
ing, and stabilizing force in the history of Spanish. He first presented his
views on these links between socioculturel history and linguistic history in
his famous Orígenes del español, first published in 1926 (and subsequently
revised and reprinted several times). Throughout this work, he reiterates
his views on the unique nature of Castilian. The following passages are
illustrative:
en [Castilla] la evolución está más adelantada. Así, por ejemplo, la fijación
de los diptongos ué y ié es ya completa en el siglo X en Castilla, mientras
León, la Rioja, y Aragón vacilan mucho todavía, usando ua, uo, ia . . . Cas-
tilla muestra un gusto acústico más certero, escogiendo desde muy tempra-
no, y con más decidida iniciativa, las formas más eufónicas de estos sonidos
vocálicos . . . (Menéndez Pidal 1964: 486)
 Language change in early Castile 105

El dialecto castellano . . . representa una nota diferencial frente a los demás

dialectos de España, como una fuerza rebelde y discordante que surge en la
Cantabria y regiones circunvecinas... (Menéndez Pidal 1964: 487)

Cuando la lengua es adulta, como tiene clara conciencia de su personal ca-

rácter, repugna esa indecisión, halla pronto unidad de criterio y tras una lu-
cha breve viene la estabilidad. (Menéndez Pidal 1964: 528)

Ciertos países muestran una orientación espontánea hacia la estabilización

más decididamente que otros. Castilla se adelanta a todos los dialectos her-
manos. (Menéndez Pidal 1964: 529)
Menéndez Pidal's favorite historical themes of exceptionality, unity, and
stability are here applied to the early Castilian dialect. It is also important
to note Pidal ' s assumption (in the first quote) that the more "advanced"
evolution is the one showing the least variation. In fact, this passage and
others like it reflect, in an extreme form, the standard ideology which col-
ors so much discussion of historical linguistic change. Standards require
minimal variation in form, and Milroy (1992: 124) argues that the standard
ideology leads to a tendency to project modern standard (i.e., less variable)
norms backwards onto earlier, more variable states of language, and to
ignore or discount much variation when it is encountered (particularly
when that variation does not lead directly to the modern standard). For
example, the relatively greater degree of variation that Menéndez Pidal
identifies with neighboring Leon ("dentro de la región central del reino, la
propiamente leonesa . . . hay poca unidad"; 1964: 448) does not mean the
Leonese dialect was backward or even unstable. After the initial stages of
mixing, new norms were established, but Leon retained greater systematic
variability for many features than did Castile (e.g., the articulation of the
spontaneous diphthongs; see below) and greater geographic (and social?)
variability, showing more significant variation from east to west and north
to south than did Castile within its borders (though even Menéndez Pidal is
forced to distinguish Cantabria and primitive Castile from the more south-
erly Burgos region). The maintenance of a greater degree of complex
vowel variation can be attributed to a lesser degree of heterogeneity among
the new colonists (in comparison with Castile): Asturians and (usually
Toledan) Mozarabs clearly dominated in the mix, and, as Menéndez Pidal
himself argued, these varieties showed similar features (which, in the case
of variable diphthongs, younger Galicians must have acquired from their
majority peers in the center of Leon). On the other hand, the survival of the
etymological pronoun system in western and central Leon seem to be at-
106 The Burgos phase

tributable to the combined impact of Galicians (including western Asturi-

ans) and Mozarabs (see discussion of leísmo in Chapter 4). A more impor-
tant factor is perhaps the growing influence of Castile on Leon, and the
progressive takeover by Castile of more and more territories of the Duero
valley between Burgos and Leon. Ridruejo (1995) points to the fact that
there were no sharp dialect boundaries in the Duero valley, and the iso-
glosses that developed in the earliest centuries tended to move westward as
Castilian dialect features were adopted with ever-growing frequency within
Leon proper (see Chapter 4 for discussion of dialect spread).
The nationalistic and standardizing bent in Menéndez Pidal's work
might lead us to question all his claims. Were Castile and its dialect really
exceptional? Certainly one could argue for "unique" and "exceptional"
features for any region and dialect (depending on the criteria used), and
one can argue that the early sociolinguistic history of Castile is very like
that of neighboring regions. Numerous scholars (e.g., Wright 1994f;
Ridruejo 1995; Penny 1987, 2000; and even Menéndez Pidal 1964: 444)
recognize that dialect mixing occurred in all the more southerly regions
that underwent repopulation:
in each case the new inhabitants of the expanding city came mainly from
several valleys to the North, and a sociolinguistically simpler interdialect
formed, usually only perpetuating in the prestigious capital city the highest
common factor of such regional diversity as already existed in the northern-
most mountains (which is why the speech of the mountain valleys is even
now linguistically more complex than their related dialects to the south).
The continuing repopulation movements were later to reinforce mutual intel-
ligibility several times through the Middle Ages. (Wright 1994f: 159)
This is essentially true, but if we place the dialects on a continuum from
the most complex and systematically variable to the least, then Castilian
certainly seems to appear at the latter end of the continuum (with the nota-
ble exception of leísmo), and given its history we can assert that mixing
was more intense and that focusing on distinctive norms more favored in
early Castile than in some other regions. We must ask too whether the
"unique" linguistic features of 10th-century Castilian really existed, for it
could be that Menéndez Pidal saw or invented what he needed to, but the
consensus among experts seems to be that Menéndez Pidal was largely
correct in his description of specific features of early Castilian and his
claim that at least some of these changes serve to distinguish Castilian
from neighboring dialects. I return to this issue under the discussion of
evidence for changes.
 Language change in early Castile 107

How did Menéndez Pidal deal with the question of causation? What
was the connection between the social history and linguistic history of
Castile? Here Menéndez Pidal had access to no theories that would allow
him to establish explicit links between the two. In fact, he was left with no
option other than to ascribe these changes, in a rather circular fashion, to
the inherently exceptional nature of the Castillans (or, in this case, Can-
tabrians, whom he sometimes identified with Castilians), who were in gen-
eral "more revolutionary, more inventive, more original" (1964: 487), and
at the same time, more given to "stabilization" (1964: 529). In his discus-
sion of the development of customary law in Castile, Menéndez Pidal
(1964: 475) noted the fact that customary law also dominated in the Ile-de-
France, as a result of Germanic influence. He saw this not just as a coinci-
dence, but as evidence of the "fermento germánico" (mixing between Ro-
man and German peoples) which he believed was the source of the distinc-
tive features of Castile and Castilian. Such arguments are clearly
unacceptable for early Castile. Still, while we must be wary of Menéndez
Pidal's unsuccessful efforts to explain the social and linguistic develop-
ments of Castile, based on his nationalistic and standardizing sentiments,
his efforts to describe them retain great value.
Menéndez Pidal is not the only scholar to have considered the potential
importance of dialect leveling, mixing, or koineization in the history of
Castilian and Spanish. Others too have dealt with the topic, though with
widely differing approaches, differing time frames, and varying degrees of
success. For example, in 1950 the Spanish dialectologist Garcia de Diego
published a well-known article called "El castellano como complejo dialec-
tal y sus dialectos internos". In it he made special claims for Castilian vis-
à-vis the principal Peninsular dialects and Romance languages:
El castellano sólo tiene conciencia defensiva frente a los grandes dialectos
conservados, como el gallego o el catalán. Sobre los dialectos inconsistentes
barridos por él, y, en parte, solapadamente subsistentes, el castellano obra
sin cautela, aceptando lo que encuentra. Esta es la paradoja de que el
español es dialectalmente pobre frente al francés y el italiano y es más rico
en dialectalismos. Mientras elfrancésy el italiano en su denso ambiente dia-
lectal reparten bien las dos hablas, el francés para el uso oficial y el patois
para el familiar, el castellano ha mezclado, ya sin separación posible, sus
elementos y los regionales, dejando de considerarlos extraños . . . el caste-
llano, al instalarse en zonas nuevas, ha perdido la conciencia de la distin-
ción, dando el mismo valor de oficialidad a los elementos provinciales.
(García de Diego 1950: 107)
108 The Burgos phase

This is as strong a statement as exists on the mixed nature of Castilian, but

Garcia de Diego's explanation for mixing relies on the suspect notion of
borrowing from some twenty-odd "hidden" geographical dialects, which
supposedly were overlaid and absorbed by Castilian as it invaded new re-
gions. Here we encounter a view of dialect contact as contact between
competing yet stable dialects (rather than speakers), and as part and parcel
of the political expansion of Castile, which is equated with the linguistic
expansion of Castilian. Unsurprisingly, this view has been criticized for
exaggerating the importance of dialect borrowing:
Garcia de Diego, en su trabajo sobre 'el castellano como complejo dialectal'
. . . utiliza el criterio del dialectalismo para explicar algunos términos tradi-
cionalmente considerados como cultismos, puesto que su hipótesis central
consiste en afirmar que el castellano está formado por un 'complejo dialec-
tal' . . . y 'acepta todo lo que encuentra'. Así,flamaes una forma vulgar de
la Rioja y del Sur, clavija es también una forma riojana, frente al soriano y
burgalés [*]llavija-, glera es un aragonesismo. Esta solución puede ser con-
veniente para algún caso, pero resulta obvio que no se puede tomar como
panacea para todos los términos problemáticos. (Clavería Nadal 1991: 35)
The particular examples of borrowing given by García de Diego (and
Clavería Nadal) actually reveal a more complex form of mixing. For exam-
ple, clavija (< CLAVICULA) is a form that is difficult to force into the mold
of direct borrowing. This item contains one of the distinguishing features
of Old Castilian, / j / < C'L, as well as the conservative retention of CL-,
generally associated with Latinisms or the normal development in eastern
Iberian regions. Given the mixed or intermediate nature of the form, Garcia
de Diego is forced to classify it as Riojan, which is geographically inter-
mediate between Castile and eastern regions which maintained CL-. It is in
order to account for each unique combination that Garcia de Diego must
postulate so many hidden dialects.
Other scholars have claimed, more directly, that Old Spanish was in fact
a koine, but problems with the meaning of the term have clouded discus-
sion. In 1985, López García published El rumor de los desarraigados, the
central thesis of which is that early Castilian, which he labels a "koine",
arose as a mixed language. The mixing was based on Basque-Romance
bilingualism, which in turn arose from the need of native Basque speakers
to communicate with Romance-speaking neighbors. López García assumes
that Basque was used over a wide territory, that the mixed language or
koine developed and spread to many communities beyond Castile (includ-
ing those of Mozarabs and Jews throughout Spain), and that it was only
 Language change in early Castile 109

nativized in Castile, and thus came to be called Castilian. In order to sup-

port this argument, López García identifies 14 features of the koine (re-
named vascorrománico), that can be traced to Basque. In so doing he sur-
passes even the strongest supporters of substrate theories in advocating
simple external causes for the development of Castilian.66 There are nu-
merous flaws in this thesis. First, López García's main purpose in El rumor
de los desarraigados is political: by claiming that Castilian was once a
koine and therefore nobody's native language, he hopes to make it more
palatable as a second or official language in Spain's modern autonomías.
Since Rumor is fundamentally a political treatise, it re-molds the (linguis-
tic) past in support of a political stance (in a way that Menéndez Pidal did
not):
El español nació como la lengua de los otros, la lengua de los desheredados
que no conocían otra nación que la que ellos mismos y su trabajo pudiesen
edificar sin restricciones de raza, sexo, clase social o lugar de nacimiento:
henos aquí ante el mensaje que quisiera transmitir con estas líneas.
(López García 1985a: 54)

Second, López García does not clearly or consistently define the term koi-
ne. Though in general he appears to consider it as combining the function
of a lingua franca with changes in form, particularly simplification (1985a:
62), his discussion is full of internal contradictions:
Resumamos: el español no fue un simple dialecto de transición - algo impo-
sible entre el euskera y el latín, como es obvio, y de ahí que no se produjese
nada similar en Francia, por ejemplo - , sino que nació con voluntad de
constituirse en lengua mixta que podrían adoptar también quienes no habla-
ban ni vasco, ni la variedad romance del Alto Ebro; propiamente surge co-
mo un créole, como una lengua simplificada que retiene la forma interior
gramatical de un idioma y el apresto léxico y morfológico de otro.
(López García 1985a: 54)

In this passage López García equates koine with creole, thereby gutting the
term koine of any useful independent meaning and confusing the theoreti-
cal basis of his argument. Finally, the features that López García claims
mark Basque influence cannot all be described as simplifying (e.g., the
split of the primary copula into ser and estar),67 and, more damning still,
Trask and Wright (1988), have argued convincingly that, of the fourteen
examples of "concordancia gramatical entre el castellano y el euskera"
outlined in López García (1985b),68 not one can plausibly be identified as a
source of changes in Castilian. Besides these problems with factual details,
110 The Burgos phase

Trask and Wright point out that López García's arguments suffer from the
same problems as all substrate theories, in that they attempt to attribute a
given phenomenon to a single cause. On the other hand, they allow that
Basque influence may still be seen as a contributing factor for certain
changes, such as aspiration of /f-/ (Trask and Wright [1988: 371]; but see
below for my discussion of later arguments presented by Trask 1997).
A more recent work which constitutes a potential contradiction to the
hypotheses defended in this chapter is Gimeno Menéndez' Sociolingüística
histórica: siglos X-XII (1995). This monograph contains no reference to
Penny's (1987) lecture, nor to the work of Trudgill (1986), a surprising
lack for a text which contains an otherwise exhaustive bibliography. No
discussion of koines or koineization is made in the entire work, and there is
only a footnote devoted to López García's arguments (Gimeno Menéndez
1995: 123). Other details of this study do address issues pertinent to
koineization, however. Gimeno concludes with a discussion of the devel-
opment of the personal a, or marker of personal accusative, in early Ibero-
Romance. After reviewing the literature on the topic, he presents the re-
sults of a quantitative study of texts from the 10th to the 12th centuries
from the regions of Leon, Castile, Navarre, and Aragon, in which he seeks
to identify the origins of the increased use of this particle. He finds a
higher absolute number of occurrences of use of the particle in documents
of Leonese origin. However, the potential significance of this finding is
effectively nullified by the fact (reported by Gimeno himself) that there is
an "ausencia de covariación significativa entre la variable dependiente
[appearance of ad/a] y las independientes" (Gimeno Menéndez 1995: 160).
Although Gimeno argues at great length that quantitative textual research
is necessary for an accurate description of medieval language, he chooses
to ignore the results of his own statistical analysis and defend the signifi-
cance of this finding. This, in turn, leads him to challenge Menéndez Pi-
dal's characterization of Leonese and Castilian history:
Ya hemos aludido a que, según R. Menéndez Pidal . . . Castilla se carac-
teriza por su derecho consuetudinario local y se opone al derecho escrito
dominante en el resto de España, al mismo tiempo que representa el fer-
mento germánico frente a León, región arcaizante y conservadora de la en-
vejecida cultura latina. No obstante, este reino también presenta influencias
exteriores árabes y caracteres innovadores en el latín notarial de los escribas
mozárabes (en su mayoría) de los siglos X y XI. A este respecto, en nuestro
análisis de la variable sintáctica ad en el registro de los fueros municipales
de derechos consuetudinarios hemos observado que el mayor impacto del
 Language change in early Castile 111

vernáculo en los textos escritos latinos se da en León y no en Castilla, aun-

que no sea estadísticamente significativo. (Gimeno Menéndez 1995: 164)

This line of reasoning leads Gimeno back to an interpretation of Leonese

and Castilian history presented earlier in his study (Gimeno Menéndez
1995: 99), in which he argues that it was the maintenance/development of
strong social ties in early Castile that allowed for the shift to customary
law and the rejection of the Fuero Juzgo (Liber Ludiciorum) :
Desde este punto de vista, la intensidad relativa del vínculo de retícula en su
extenso territorio [León] presentaría retículas sociales difusas (abiertas a la
influencia externa) que facilitarían la innovación lingüística, mientras que la
comunidad cohesionada del reino de Castilla, afirmada su supremacía, ofre-
cería retículas densas, que impedirían un mayor desarrollo del latín forai.
(Gimeno Menéndez 1995: 164)
Gimeno thus argues for a complete rewriting of Leonese and Castilian
history, with a kind of role reversal, in order to justify the greater (though
statistically insignificant) frequency of appearance of one syntactic feature.
Still, while the Castilian speech community certainly began as more per-
meable to innovations, resolidification of social networks, along with the
creation of an oppositional regional identity, may indeed have made the
Castilian koine resistant to rapid change after koineization had taken place.
Leonese varieties, for which there is plenty of evidence, remained more
conservative than Castilian (and more varied across physical and probably
social space), but also participated in and were influenced by changes
which dominated in and spread from Castile.
In recent years, the notion of Castilian (and Spanish) as a koine has be-
gun to gain currency in the work of some historical linguists. Penny (e.g.,
1987, 2000) leads the way here (see Introduction), and Ridruejo (1995)
follows suit, but not all are in agreement. In her (1996) introduction to the
study of the Romance languages, Posner makes frequent use of terms such
as koine, koinai, koineized, koineizing and koineization to analyze and de-
scribe numerous other Romance varieties, yet claims, surprisingly, that:
the Castilian dialect is deviant phonologically and innovating morphologi-
cally, and far from a koineized conpromise between Iberian varieties.
(Posner 1996: 208)

This very direct claim relies on the erroneous assumption (which has its
modern source in Meillet's pronouncements) that koineization does not
lead to innovation, but merely to the elimination of variation (or a very
simple form of innovation). As was made clear in Chapter 2, the contribuí-
112 The Burgos phase

ing varieties are not limited to pre-existing dialects, but also include the
interlanguages/interdialects of individuals and groups that are attempting to
accommodate to and/or acquire a new variety or varieties. In a context of
fluid norms, the innovations of such speakers, if consistent and frequent
enough, stand a chance of being learned by the entire speech community
and thereby becoming fixed in the resultant koine. Moreover, some of the
most "deviant" and innovating features of Castilian can be analyzed as the
results of mixing, leveling, and simplification. We turn now to analysis of
some of those features.

3. Linguistic changes

In order to show that koineization has occurred, structural changes in the

linguistic system must be pinpointed and linked to the social changes and
other factors that enable and constrain speaker activity. This section fo-
cuses on certain changes in Castilian (and, to some degree, in other dia-
lects) which appear to result from the koineizing social conditions de-
scribed above. The changes discussed are those for which there appears to
be reasonably good evidence from the período de orígenes. This issue
forces some comment on the sources and quality of the evidence.
Labov is well known for having characterized historical linguistics as
the art of making the best of bad data (Labov 1972: 100). In the case of
early Ibero-Romance, the general paucity of phonographic vernacular texts
before around 1200 underscores, in some respects, the validity of Labov's
observation. On the other hand, Menéndez Pidal's Orígenes del español
(1964) brought attention to the value of early Latinate texts, particularly
legal documents, that survive from the period as rich potential sources of
evidence of changes in spoken Romance. In these texts the scribes often
attempted to represent features (particularly phonological features) of their
spoken language within the much more conventionalized and increasingly
logographic written language. Menéndez Pidal's analysis of these texts,
reported in Orígenes, remains a primary source of information on the lan-
guage of this period, and I rely on it heavily in the following discussion.
Given his desire to show Castilian as exceptional, we might question Pi-
dal's analyses, but even his harshest critics have ended up largely endors-
ing the value and overall accuracy of his philological work. For example,
Torreblanca (1986b, 1988, 1989) analyzed the documents used by Mené-
ndez Pidal and his assistants, and was able to show that in some cases Pidal
 Linguistic changes 113

relied on later copies from the 11th to 13th centuries rather than originals
from the 10th century (which are relatively scarce). Even so, for most
changes, Torreblanca supported Menéndez Pidal's conclusions (even as he
criticized him for not always indicating if he was using copies or original
documents). For example, Torreblanca (1986b: 233-234) analyzed the
early spelling variation between ie and i, and ue and u (e.g., tyerra vs.
tyrra), and concluded that the appearance of / and u was merely the result
of scribal omissions and carelessness (Menéndez Pidal had claimed they
were early and imperfect attempts to represent the novel diphthongs).69
However, Torreblanca also found that the diphthongs are represented in
Castilian texts with great consistency as ue and ie, exactly as Pidal
claimed. We should also note that manuscripts from the Burgos region,
dated to the 11th century and even early 12th century, can still be used to
see the results of early koineization in Castilian, since once the initial stage
of repopulation and mixing had drawn to a close, the society and its dialect
remained relatively stable (though variable; but see Chapter 4 for discus-
sion of changes which progressed after initial koineization). Most scholars
today accept the bulk of Menéndez Pidal's description of developments in
early Castilian. Nevertheless, textual evidence has to be interpreted and
criteria are needed for making interpretations (Wright 1982, 1994a; Alva-
rez Maurin 1994). In the following sections, we will use the model of
koineization as a tool for interpreting the evidence offered up by the docu-
mentation.
The following sections include analysis of two groups of changes that
occurred during the Burgos phase in Castile: the simplification (reduction
in variants) of article forms and preposition + article contractions, and the
simplification and stabilization of the tonic vowel system. In each case I
review the evidence for each change as well as previous attempts at expla-
nation, and I argue that koineization played the key role in the actuation
and transmission of each of these changes (though additional causal factors
identified previously are often incorporated into an explanation based on
koineization).701 also show that the features which survived can be found
in at least one of the contributing language varieties, including learner lan-
guage, and that selection and stabilization (focusing) of the new fea-
ture/norm occurred relatively rapidly.
114 The Burgos phase

3.1. Leveling and simplification of articles and preposition + article

contractions

Loss of variation in forms of articles and in the contractions or fusions of

prepositions and articles into single lexical units are features of early Cas-
tillan. Menéndez Pidal (1964) discussed them long ago, but since then they
have received little more than brief mention in the linguistic literature.
Their early disappearance has received still less attention and few attempts
at explanation, yet this is one of the most dramatic results of koineization
in the Burgos phase. One can only conclude that this is so because these
reductions are seen as the "normal" development from the perspective of
the modern standard (see below for further evidence of this view).
Most northern Romance dialects show a variety of written forms for the
articles themselves. Evidence for the early, frequent and regular use of
contracted forms is abundant for most dialects of the northern Iberian dia-
lect continuum. Tables 8-13 show the frequently-attested forms of articles
and preposition + article contractions for each of the main dialect areas of
the Middle Ages.71 Faced with this great variety of forms, and the general
simplicity of the Castilian results, Menéndez Pidal concluded: "El artículo
en Castilla se muestra más adelantado en su evolución" (1964: 337).
Moreover, the lack of variation in Castilian is evident in the earliest texts:
Desde la fecha más antigua de nuestros documentos aparece el artículo
completamente formado en su estado moderno . . . Las grandes vacilaciones
de forma que el articulo tiene en León . . . y en Aragón . . . son en Castilla
poquísimo usadas o totalmente desconocidas. Castilla, como en otros fenó-
menos que ya hemos notado, es la región en que la evolución se ha realizado
más rápida y decididamente, logrando pronto una mayor fijeza lingüística
que las otras regiones. (Menéndez Pidal 1964: 337-339)

In Castilian we observe a selection of transparent forms with a greater re-

duction in inventory (in comparison with neighboring Leonese, Riojan and
even older, more northern dialects of Castilian itself). The other dialects
were not unstable, however; the variation in written forms probably re-
flected a more complex and variable system in which the phonetic context
of the forms determined their actual articulation.
Not unrelated to this stabilization and simplification of the forms of ar-
ticles is the loss of preposition + article contractions. The geography and
chronology of these losses can be exemplified with Pidal's discussion of
the form enno/a/os/as:
 Linguistic changes 115

Table 8. Forms of articles and preposition + article contractions in Galicia and

Portugal. Sources: Huber (1986); Ferreiro (1996).
Basic por + de + a+ con + en +
articles article article article article article
Masc. Sing. o, lo72 polo, do, du ao, au, ό co73 no
pelo
Fem. Sing. a, la pola. da aa, á coa na
pela
Masc. Plur. os, los polos, dos, dus aos, ous, cos nos
pelos ós
Fem. Plur. as, las polas, das aas, ás coas ñas
pelas

Table 9. Forms of articles and preposition + article contractions in Leon. Sources:

Menéndez Pidal (1964: 330-332); Menéndez Pidal (1962a: 87, 97);
Staaff (1907: 253-258).
Basic por/per + de + a+ con + en +
articles article74 article article article article
Masc. Sing. ele, elo, pollo,76 del al conno, enno,
el, le, l polo, cono, eno,
lo, lio75 pello, collo enne,
pelo ene, no
Fem. Sing. eia, lia polla, de la a la conna, enna,
pola, cona, ena, na
pella, colla
pela
Masc. Plur. elos, líos pollos, de los a los connos, ennos,
polos, conos, enos, nos
pellos, eolios
pelos
Fem. Plur. elas, lias pollas, de las a las connas, ennas,
polas, conas, enas, nas
pellas, collas
pelas
116 The Burgos phase

Table 10. Forms of articles and preposition + article contractions in Aragon. Sour-
ces: Menéndez Pidal (1964: 333-336); Alvar (1953: 215-218).77
Basic por + de + a+ con + en +
articles article article article article article
Mase. Sing. lo, o, elo, por lo, del a lo con lo enno, no
el, ero, polo

Fem. Sing. la, a, eia, por la, de la a la con la enna, na

era, ra pola
Mase. Plur. los, os, por los, de los a los con los ennos
elos, polos
eros, ros
Fem. Plur. las, as, por las, de las a las con las ennas
elas, polas
eras, ras

Table 11. Forms of articles and preposition + article contractions in La Rioja. Sour-
ce: Glosas Emilianenses and Glosas Silenses, Menéndez Pidal (1964:
330-340); Alvar (1976: 60).
Basic por + de + a+ con + en +
articles article article article article article
Mase. Sing. elo porlo del al cono eno
Fem Sing. eia porla de la a la cona ena
Mase. Plur. elos por los de los a los conos enos
Fem. Plur. elas por las de las a las conas enas

Table 12. Forms of articles and preposition + article contractions in northern Cas-
tile. Source: Menéndez Pidal (1964: 337-339).
Basic por + de + a+ con + en +
articles article article article article article
Masc. Sing. el, elo por el del al con el en el
Fem. Sing. la, eia, I por la de la a la con la ena
Masc. Plur. los por los de los a los con los enos
Fem. Plur. las por las de las a las con las enas
 Linguistic changes 117

Table 13. Forms of articles and preposition + article contractions in Castile (Burgos
and areas to the south). Source: Menéndez Pidal (1964: 337-339).

Basic por + de + a+ con + en +

articles article article article article article
Masc. Sing. el por el deln al con el en el
Fem. Sing. la (el)19 porla de la a la con la en la
Masc. Plur. los por los de los a los con los en los
Fem. Plur. las por las de las a las con las en las

El mayor uso de enna, ennos es en la Montaña (DL, 1191, 1202, 1210,

1223), en Campó (DL, 1204, 1219, 1219, 1220, enno, neutro, 24°26, 1259,
1284) y en Castilla del Norte (DL, 1102, 1144, fines del siglo XII40°, 1202,
1212). Después, más al sur del reino, esta asimilación decrece en uso, y sólo
se hallan muy escasos ejemplos, que desaparecen ya en los primeros años
del siglo XIII: Burgos 1197, 1209 (DL, 153013 y 160°I2), Toledo 1206 y
1206 (DL, 265°n y 266°27). Esta asimilación enna, ennos era, pues, un ar-
caísmo, que en el siglo XIII no tenía uso frecuente sino en la Montaña y en
Campó. (Menéndez Pidal 1964: 338)

Significantly, Pidal includes this change in what he considers the broader

analytical tendencies of the language:
La asimilación enna, corna que hoy parece peculiar del leonés, se usó tam-
bién antiguamente en Castilla y Aragón, pero en Castilla se desechó pronto
como arcaísmo, obedeciendo a la tendencia analítica del idioma, que pro-
cura mantener la l- inicial intacta. (Menéndez Pidal 1964: 339)

However, Pidal makes no attempt at an explanation for this phenomenon,

other than including it among the analytical changes which tended to char-
acterize Castilian. Of course, "tendency" (like drift) is a descriptive, not an
explanatory, notion, and it is completely circular when used for explana-
tion.
The contracted forms in question, the result of elisions and assimila-
tions, came into existence before the periods for which we have written
documentation. They are the result of slow phonetic changes - the accre-
tion of phonetic rules - that led first to increasingly complex phonetic re-
alizations of underlying forms and eventually to the creation of new forms,
as the original surface phonetic realizations crystallized and themselves
became forms that had to be learned. Changes by slow accretion are typical
of stable communities with close-knit social networks (e.g., Milroy 1992:
99-100; Trudgill 1992, 1996, 2002), and they permit a more economical
118 The Burgos phase

encoding of information (Liidtke 1980). Trudgill (1996) points out that

such opaque forms benefit the speaker by allowing the efficient packaging
of information, but disadvantage the listener, since the listener must know
and do more to understand the message. While communicatively efficient,
such forms are more difficult to learn. However, the northern Christian
regions of the eighth and ninth centuries (before the reconquest began in
earnest), would have been ideal incubators of such changes. The ability to
use contracted forms requires that the speaker learn either a) the exact eli-
sion and assimilation (morphophonological) rules required to produce such
forms as well as the exact base forms and combinations on which they
operate or, in later stages, b) a learned list of such contracted forms and
knowledge of the base forms to which they correspond.80 For example, a
speaker of Riojan (itself already simplified) would have had to learn sev-
eral allomorphs of the morpheme elo, including elo and combining forms
lo, el, I, o. Similarly, the possible allomorphs of other forms would have
been: eia, la, a; elos, los, os; elas, las, as. Such complex knowledge is
most effectively passed on by stable communities (or through literacy,
which was not widespread at the time), which are able to enforce norms
and thus limit the output of speakers (which, in turn, constrains child and
adult language learning).
Explanations for the early and rapid loss of such contracted forms in
Castilian are wanting. However, Lleal associates the loss of contracted
forms with the rise of literacy in Romance:
En el plano morfofonológico, se dan frecuentes contracciones del artículo
con la preposición contigua: in illu > [eno], cum illu > [kono]. Estos fenó-
menos son un fiel reflejo de las contracciones que se daban (y se dan) en la
lengua hablada y que, al parecer, tenían un ámbito de aplicación muy ex-
tenso por toda la Península. Posteriormente, la aparición de unas normas de
escritura que dieron prioridad a la palabra como unidad gráfica eliminaron
esas formas de algunos romances hispanos. (Lleal 1990: 163-164)

However, there are two problems with this view. First, the changes in ques-
tion clearly predate the establishment of a written Castilian standard, which
does not begin to occur until the 13th century. Second, it ignores the fact
that contracted forms have been maintained in modern standards such as
Portuguese and French. The Portuguese say and write no, pelo, not *en o
or *por o\ the French say and write du, au, not *de le or *à /e.81 Literacy
and standardization cannot be seen as the cause of loss of such contractions
in Castilian.
 Linguistic changes 119

Speaker activity in koineization offers a better explanation. The rapid

expansion of Castile and the massive demographic mixing that followed
led to the removal of speakers from established social networks and their
insertion in new, mixed ones. As social and linguistic norms were weak-
ened, speaker-learners were exposed to a wide variety of possible forms of
articles and contractions of prepositions and articles. For example, what
are now en el and por el would have been part of a pool of competing vari-
ants: no, enno, eno, enne, ene, en el and polo, pelo, pollo, pello, por el. In
order to accommodate to their new and dialectally mixed communities,
some speakers may have begun to eliminate those forms which were unfa-
miliar to many other speakers; no and polo could have been replaced with
the easily generated en el and por el. More likely still, both adult and child
speaker-learners would have tended to search for and reproduce forms
which were consistently and frequently produced, such as the easily ana-
lyzed and generated en el or por el, whose component parts also appeared
regularly in other contexts. In fact, research on contemporary varieties with
such contracted forms has shown that they are difficult for children to
learn. For example, French-speaking children as old as 5 still do not consis-
tently contract de le to du, de les to des, à le to au, or à les to aux (Clark
1985: 727). All these factors would have led to the production of still more
regularized forms (such as en el and por el). With such massive variation
and weak norm enforcement, there would have been little need or possibil-
ity - given the competing norms - of "correcting" such output, which
would have then served as input for other learners.82

3.2. Reorganization and simplification of the tonic vowel system

Castilian is well known for its classic five-vowel system. However, this
exists in marked contrast to that of both western Ibero-Romance (Galician-
Portuguese) and eastern varieties (Catalan), where the seven vowel pho-
nemes of the typical Vulgar Latin/Early Romance system are retained, as is
the case in most Romance varieties. This reduction in phonological inven-
tory can be dated from the earliest days of Castilian and is, in fact, the indi-
rect consequence of another of the most characteristic developments of
early Castilian: the stabilization of the diphthongal realizations of the Vul-
gar Latin/Early Romance low mid vowels h i < Ö and /ε/ < Ë (Menéndez
Pidal 1964: 486).
120 The Burgos phase

The spontaneous diphthongs that developed from Latin tonic Ö and Ë

have long been the source of controversy in Romance linguistics. Past re-
search has focused on the dating and linguistic reasons for their initial ap-
pearance, as well as how they should best be analyzed within the phono-
logical system of each language variety affected. The widespread
occurrence of such diphthongization throughout the Romance-speaking
area indicates that it must have begun in Late Latin/Early Romance, cer-
tainly being established in the Iberian Peninsula before the Moorish Inva-
sion. The varieties to the extreme West, Galician and Portuguese, were
largely unaffected (though traces of incipient diphthongization can be
found in some dialects), as was Catalan in the East. However, the Hispano-
Romance dialects of the center of the Peninsula were all (eventually) af-
fected by spontaneous diphthongization of Ö and Ë. Moreover, both histori-
cal textual and modern dialectal evidence points to tremendous variation in
the realization of these diphthongs (and related monophthongs) in the ma-
jority of central Iberian dialects, namely, Astur-Leonese, Navarro-
Aragonese and at least some of the principal dialects of Mozarabic.
In Leon, early Latinate texts (written in what Menéndez Pidal called
latin vulgar leonés) show a mixture of orthographic forms: o, uo, ua, ue <
Ö; e, ie, ia < Ë (Menéndez Pidal 1964: 113-115, 144-146). Later 13th-
century texts written in Leonese continue to show this same variation
(Staaff 1907: 191-206), though ua declines significantly, leaving uo, ue
and o as the dominant forms; ia was less frequent, found principally in
documents from western Leon (Menéndez Pidal 1964: 146). With regard to
the phonetic value of these orthographic representations, Wright (1994c)
has emphasized that the novel combinations of letters are probably more
significant than those that correspond to the written (Latin) orthographic
norm. This is so because the scribe must depart from an orthographic norm
to produce innovations which reflect phonetic realizations. This could
mean that there remained only diphthongal realizations in speech, but in
Leon and Aragon representations with o or e continued many centuries, so
the survival of the monophthongs as variants is plausible. Aragonese texts
from the 11th century and later show vacillation similar to that of Leonese
texts, between o, uo, ua, ue<0 and e, ie, ia < Ë, though in these texts uo is
relatively infrequent compared to ua, ue, o, and ia was frequent, though not
as common as ie (Menéndez Pidal 1964: 118-121). For Mozarabic, Ro-
mance forms transliterated into Arabic script indicate widespread vacilla-
tion between a monophthong and diphthong for both Ö and Ë. Galmés de
Fuentes (1983: 70, 130, 188) indicates vacillation between monophthongs
 Linguistic changes 121

and diphthongs in Toledan Mozarabic and a preference for diphthongal

articulations in Seville (while Peñarroja Torrejón [1990: 177-182] shows
that monophthongal articulations were the norm in Valencian Mozarabic).
Menéndez Pidal (1964: 132-139) offers evidence for diphthongal realiza-
tions of Ö in the Mozarabic of Zaragoza, Toledo and Andalusia, but claims
that monophthongs appeared to predominate over diphthongs:
El botánico sevillano de hacia 1100 nos da, mejor que ningún otro, idea del
estado en que se hallaba la diptongación. La mayoría de los casos por él ci-
tados carecen de diptongo . .. total, 18 nombres, repetidos 38 veces, frente a
13 nombres con diptongo, repetidos 19 veces.
(Menéndez Pidal 1964: 137)

He also claims such widespread variation was typical of reflexes of Ë. For

the variation between diphthongs, various botanical texts give transcrip-
tions which indicate forms such as yärba and yérba < HERBA, yádra and
yédra < HEDERA (Menéndez Pidal 1964: 148-149).
Modern-day rural dialects of Leonese and Aragonese have preserved -
or at least preserved into the twentieth century - many of these vacillating
diphthongal articulations. Relic forms such as ya < ET, yara < ERAT still
occur in western Astur-Leonese, though generalized use of [ja] has disap-
peared (Zamora Vicente 1967: 96). However, vacillating diphthongal out-
comes of Ö could still be found in these same zones during the 20th cen-
tury:
todavía hoy perdura algún resto viviente de esas condiciones primitivas de
convivencia de formas españolas, y lo encontramos en una zona occidental
del leonés, siempre tan arcaizante y conservador: en Cangas de Tineo (Astu-
rias) el diptongo castellano ué se sustituye muchas veces en Villaoril por un
diptongo que suena casi ua, con tendencia a uo, alternando arbitrariamente
esta forma dialectal con la forma ue, aún en una misma palabra pronunciada
por un mismo individuo; cuando empecé a observar el lenguaje de otros
pueblos de esta región, yo anotaba unas veces uo y otras ue\ después en va-
rios de estos casos dudosos oí wç, con una 9 que tiende mas a o en proximi-
dad de consonantes labiales . . . y se hace e en otras circunstancias.
(Menéndez Pidal 1964: 122)

Zamora Vicente adds: "Todos los investigadores coinciden en señalar una

gran riqueza de timbre al segundo elemento" (1967: 91). Krüger (1923)
described in San Ciprián de Sanabria a diphthongal realization whose se-
cond element was no longer an "o clara, sino una vocal bastante imprecisa,
que hace el efecto de una e cuyo timbre es difícil de analizar y que se pare-
ce con frecuencia a una vocal media entre o y e". This could indicate [woe]
122 The Burgos phase

or [wa]. Catalán (1956-1957) provides evidence from several isolated As-

tur-Leonese dialects of even greater variation, with articulations ranging in
a continuum from [wo] to [wa], [woe], [we], and (apparently) [wa] for the
back vowel, and from [je] to [ja], [joe], and [ja] for the front vowel
(Catalán also points out that the [a] is "palatal", and thus probably slightly
raised and fronted to [as]). Moreover, in these same dialects, the stress does
not always fall on the second element; rather, it can be spread over the
entire diphthong or fall on the first element, giving such alternative articu-
lations as [uo], [úo], [uoe], [úoe], [ua], [úa], [ue], [úe], etc., and [ie], [ie],
[ia], [ia], etc.83
Aragonese has preserved evidence of the early variation in placenames
such as Laguarre, Lascuarre and Banaguás and present-day (though mori-
bund) articulations such as buanas tardes and puarta in the Pyrenees
(Menéndez Pidal 1964: 120, Zamora Vivente 1967: 216). Far more fre-
quent here than in Leonese, the diphthong [ja] survives (as a lexicalized
form) in some areas in words such as piadra, tiada (= Castilian tea), bian,
cadiara < CATHEDRA (from Sercué in Sobrarbe, Menéndez Pidal 1964:
217), tiampo, hiarba, diande (< DENTE) (Zamora Vicente 1967: 217).
One may safely conclude that medieval central Hispano-Romance dia-
lects were characterized by the appearance of variable diphthongal reflexes
of Latin Ö ([a, wo, woe, we, wa, wae, wa]) and Ë ([ε, je, ja, ja]). But what
was their phonological status? At least in their initial stages, these articula-
tions represented merely a phonetic lengthening and subsequent diph-
thongization of the mid-open phonemes of Late Latin/Early Romance, hi
and /ε/ (Lloyd 1987: 184).84 This early phonetic variation, however, did not
initially alter the underlying 7-vowel system of Vulgar Latin. Indeed,
Menéndez Pidal pointed out that speakers he interviewed in conservative
Galician dialects (with incipient diphthongization) were rarely aware of
purely phonetic diphthongs:
En el Miño portugués, donde hay una extensa diptongación, oí también
vacilación entre resposta, respwosta y respwesta. La forma dominante en
Baixo-Minho es [wo] y [wç] . . . En Áncora uno que pronunciaba sinhuora
(o), aguora, tuoda, pwortu, pwocu, no tenia conciencia del fenómeno de
diptongación y me decía que pronunciaban solamente 'o muito fechado'.
(Menéndez Pidal 1964: 122)

Doubtless this was the case in early Hispano-Romance (as it is in many

languages), and both phonetic diphthongal and monophthongal articula-
tions of hi and lei could have been safely recorded in the orthography as o
and e. In the case of diphthongs, the high first element (vowel or glide) was
 Linguistic changes 123

articulated with greater tension, and this was the element that remained
stable. The second element was lax, and could therefore vary freely
(Menéndez Pidal 1964: 125). As long as this second element fluctuated in
different contexts and with different speakers, learners could continue to
perceive (and learn) these diphthongs as allophones of a single underlying
phone (i.e., [wo], [wa], [we], [woe], [wa] = hl < Ö, [ja], [je], [ja] = Id < Ë).
However, as the diphthongal realizations grew dominant, gradually dis-
placing monophthongal realizations, speakers - and scribes struggling with
the orthographic system - would have become increasingly aware of the
double nature of the diphthong (the endpoints of movement) as the phono-
logical space of different vowel phonemes was crossed in the articulation
(whether [wo], [wa], [we], [woe], [wa] or variants with stress spread or on
the first element; and whether [ja], [je], or [ja] or variants with stress
spread or on first element). The very attempt to transcribe such diphthongs
indicates an incipient change in the analysis that speaker-learners were
making of them.85 This gradual reanalysis of the phonological system
would have come about slowly in areas that showed greater social stability
(e.g., Asturias) and/or homogeneity for this feature (e.g., Leon), since these
factors would have tended to enforce traditional linguistic norms with a
wide range of phonetic variants (e.g., [D], [wo], [wa], [we], [woe], [wa]
reflecting Romance hi) which learners needed to infer and thereby pre-
serve the increasingly opaque 7-vowel phoneme system of early Ibero-
Romance.
In early Castile, however, this rich and complex variation was elimi-
nated quickly. Menéndez Pidal appeared to see these changes, which he
had so carefully researched and described, as the result of the influence of
some literary or prestige model: "Es decir, que donde la diptongación no ha
logrado fijarse por obra de un decidido cultivo literario de la lengua, se
halla en estado vacilante y en gran indeterminación" (Menéndez Pidal
1964: 122). This is a typical example of the tendency to see any reduction
in variation as a result of increased literacy and standardization. While it is
true that literacy and standardization lead to the fixing of (prescriptive)
norms, such standardization for Romance is anachronistic for the period
under study.
Alarcos Llorach also sought to explain the stabilization and phonologi-
zation of the diphthongs. He suggested (1958, 1996) that these phenomena
had their origins in the Basque substrate:
el vascuence posee también un sistema de cinco vocales . . . La diptongación
de las abiertas sugiere que esos indígenas procuraron distinguirlas de las
124 The Burgos phase

cerradas, porque, aun siendo incapaces de reproducirlas exactamente, nota-

ban que no sonaban igual. (Alarcos Llorach 1996: 16)
Alarcos seems to want to ascribe the entire process of phonetic diphthongi-
zation to Basque substrate influence; given the widespread occurrence of
the phenomenon this is unlikely. On the other hand, it seems plausible that
the Basques, with their five-phoneme vowel system, would have heard the
diphthongs not as allophones of single phonemes, but rather as realizations
of two sequential phonemes.86 In turn, this analysis would have determined
their own output, which thereby would have entered the pool of competing
linguistic variants and served as input for other speaker-learners. However,
this hypothesis cannot explain why the minority Basques should have had
such a significant influence on the entire Romance speech community. Of
course, one could argue that Galicians in Castile would have reacted to the
diphthongs in a way similar to Basques, thereby increasing the frequency
of such variants in the linguistic pool. Galicians maintained the original 7-
vowel system but (generally) without diphthongization; they too would
have perceived the diphthongal articulations as sequences of two pho-
nemes. Based on their previous knowledge, these western speaker-learners
would have induced a different underlying vowel system from that of
speakers of central Hispano-Romance dialects. Having arrived at this
transparent analysis of the diphthongs, they would then have produced
articulations that more closely and consistently reflected the underlying
phonemic sequences of /uo/ and /ie/ (or /ue/ and /ie/), and further contrib-
uted through their own output to the predominance of [je] and [we] in the
linguistic pool.
Nevertheless, Galicians were clearly a tiny minority in the early social
mix of Castile, so the combined impact of Basques and Galicians does not
suffice to explain this change. In all likelihood, native speakers of central
Hispano-Romance dialects contributed to the predominance of certain
variants. As Romance speakers accommodated to one another they may
have found their original in-group variants to be non-functional in the new
environment (in the identity-marking sense; cf. Milroy 1992). For example,
easterners favoring the articulation [wa] and westerners favoring [wo] may
have found it convenient, in conversations with persons not sharing their
home dialect, to abandon marked usages of their home communities. In this
way, Romance-speaking adults, Basque adult learners of Romance, and
Galicians (if present) would have all contributed to a decline in the number
of variants and/or a favoring of some variants over others.
 Linguistic changes 125

This reduction in the number of phonetic variants employed would have

had an important impact on child language acquisition and older children's
negotiation of new norms. Children must have played a key role in the
reanalysis of the diphthongs. Lacking consistent exposure to the numerous
phonetic variants of the more conservative dialects, they would have
lacked the necessary input to construct the old 7-vowel phoneme system.
Child learners would have had to depend on the most frequent and consis-
tent surface output of other speakers. Since they would have encountered
(or perceived; see below) more consistent articulations of [je] and [we],
their first abduction (based on the five monophthongal vowel phonemes
that they were also learning) would have been that these diphthongs repre-
sented articulations of sequences of two vowel phonemes. As with adult
learners, this very analysis would have constrained the output of children,
and probably led them to exaggerate the diphthongal quality of the articula-
tion as it was phonologized. In the koineizing environment, children would
not have been obliged by stable norms to modify either their original infer-
ence or their subsequent production. And as succeeding generations repro-
duced these same articulations, the 5-vowel phoneme system could rise as
the new norm.
Before doing this, however, young speaker-learners would have had to
settle on an ideal realization of each diphthong. We would expect the
choice to fall on the most frequently employed variant, which appears in all
dialects to have been [je] for the front vowel. However, the diphthongal
realization of Romance h i was far more variable, and the selection of a
particular variant may have been less obvious, though most observers
would agree that [wo] was the most likely candidate in terms of absolute
frequency (and in relation to the results in other varieties of Romance). In
contrast, [we] is among the most extreme of all the variants. The diphthong
shows movement from high to mid, from back to front, and from rounded
to unrounded. Badia Margarit (1962) suggested that it was perhaps the
perceptual salience of [e] that led to the selection of [we], since [e] is the
vowel (in a 5-vowel system) that combines the greatest degree of percepti-
bility (sonority) with the greatest degree of sharpness.87 Unfortunately,
there is not much other evidence to favor this argument. Corbett (1970-
1971: 289) suggested that surface analogy with the already dominant [je]
diphthong led to the selection of [we] over other variants, especially since
they appeared frequently in words felt to be related, such as bien < BENE
'well' and bueno < BONU 'good'. Interestingly, both [wa] and [ja] show
higher frequency in Aragonese, a fact which lends some support to Cor-
126 The Burgos phase

bett's analogical argument. It is true that the koineizing context would have
favored such analogies, but other analogies were possible (as the Ara-
gonese results indicate) so this could not be the deciding factor. The selec-
tion of [we] may have been partly favored as speakers reanalyzed the input
at the same time as they were learning the simplified vowel system. In that
case, several of the diphthongal vowel articulations ([was], [woe], [we],
[wa]) could have been analyzed as realizations of /ue/, since the phonetic
variants were articulated in or near the phonological space of /e/. This
process of reanalysis may have favored a certain increase in frequency of
[we], but it still seems unlikely to have made [we] the dominant articula-
tion.
The key factor in the selection probably relates to the tendency of
speakers to exaggerate articulations in order to mark a distinct social
and/or generational identity (hyperdialectism). Early residents of Castile,
especially younger members of the community, may have exaggerated use
of this articulation as a way of emphasizing their identity as frontier people
and signaling their difference from residents of other regions. Such an ex-
aggeration of the diphthong would provide a close parallel to Labov's
(1963) description of the identity-related reactions of the inhabitants of
Martha's Vineyard, who also exaggerated their diphthongal articulation in
an effort to mark a local and oppositional identity (conversely, as John
Ohala has argued, the exaggeration of the diphthong would favor its pho-
nologization; see Janda [2003] for discussion of these different factors).
In the end, these factors combined with the koineizing process to pro-
duce the rapid selection and stabilization of the diphthongs in Castile.
While Castilians apparently used ue almost exclusively in the 10th century,
ue did not predominate in (north) eastern Leon until the 12th and 13th cen-
turies (Zamora Vicente 1967: 89), which coincided with the Toledo phase
of repopulation, widespread koineization and dialect leveling, and increas-
ing influence of Castilian. In the same period, Aragonese continued to al-
ternate ue with wa.88
Though it has been suggested here that the reanalysis of the diphthongs
by itself results in the reduction of the vowel system to the five vowels of
Castilian, this is only partly true. In fact, the reduction in the vowel inven-
tory can only be understood if one also considers another of the hallmark
developments of Castilian: lack of diphthongization before yod (or front
glide [j]). This feature, in fact, represents an inconsistency (or mixed re-
sult) in the development of the low mid phonemes h i and /ε/, since these
show diphthongization in Castilian everywhere except before yod. This
 Linguistic changes 127

effect can be observed in the following examples, where a diphthong might

otherwise have been expected (Lloyd 1987: 194-196):

— /ε/ > /e/: lecho, DERECTU/DIRECTU > derecho, INTEGRU > entero,
LECTU >
MATERIA > madera, SUPERBIA > soberbia, CERESIA > ceresa, NERVIU >
nervio, PRAEMIU > premio, GREGE > grey, SEDEAT > sea
— h! > loi·. OCTO > ocho, NOCTE > noche, OSTREA > ostria, NOViu > novio,
HODIE > hoy, MODIU > moyo, PODIU > poyo, FOLIA > hoja, CORDOLIU >
cordojo, OCULU > ojo

Menéndez Pidal emphasized that this development set Castilian apart (as a
feature of the Castilian cuña or wedge) from other surrounding Romance
languages and Iberian dialects, where diphthongization before yod did
occur (Menéndez Pidal 1964: 495). It is generally believed that the follow-
ing glide had an assimilatory effect on a preceding stressed low mid vowel
(Lloyd 1987: 185,194), raising it to a high mid vowel (and into the space of
another phoneme): [ε] < Ë, but [e] < Ë + yod; [d] < Ö, but [o] < Ö + yod.
Supposedly, as the tongue was raised in anticipation of the following glide,
the vowel too was raised, thus bleeding the environment of the diphthongi-
zation rule. However, this analysis leaves unanswered the perplexing ques-
tion of why the yod would have a metaphonic effect in Castilian, but not in
Leonese and Aragonese.89
In order to explain the Castilian developments, it is necessary to under-
stand the nature of diphthongization before yod in Leonese and Aragonese,
which is in fact a rather complex phenomenon. The most convincing evi-
dence of Leonese and Aragonese diphthongization before yod comes from
the modem dialects. Zamora Vicente comments on diphthongization before
yod in Aragonese:
Las vocales breves (Ö, Ë) diptongan ante yod comportándose así en estrecho
paralelismo con el leonés y frente al castellano: FOLIA > fuella; OCULU >
giiello; ROTULU > ruello·, cueto < COCTU; HODIE > güe (güey en lo antiguo);
PODIU > pueyo-, HORDEU > giierdio·, FOVEA > fueva, despuellas, etc. Casos
de Ë: TENEO > tiengo; viello, retienga, devienga, etc.
(Zamora Vicente 1967: 218)

This gives an impression of regularity that may indeed be true of the mod-
ern isolated and conservative dialects. Such regularity is also present in the
results of Leonese diphthongization of hi before yod:
128 The Burgos phase

Hoy, esta diptongación es uno de los rasgos más característicos del habla
viva. Es general en Asturias central y occidental: nueche o nuechi; ueyu,
giieyu; uoy, mueya, muecha; fueya, fuöya; cuechu; adueitar, etc. No se dan
casos en Santander y, hacia oriente, vacila ya en Lianes, donde se oye en
unos casos y se ignora en otros; en los Picos de Europa, el límite parece
estar en Cabrales, donde hay vacilación. En León, casos de este diptongo
(cada vez más en retirada ante el empuje de la lengua oficial) se oyen en Sa-
jambre (mueyo, jueya; Güeyo, en topónimos), los Argüellos (ueyo), etc.
(Zamora Vicente 1967: 93)
Significantly, Zamora views diphthongization as normal and the lack of
diphthongization as a result of increasing influence of the Castilian stan-
dard. However, modern Leonese diphthongization of /ε/ is less frequent
than that of loi, and its lesser frequency does not appear attributable to
Castilian influence:
La diptongación de Ë + yod no se ha producido con tanto rigor y abundancia
como en el caso de Ö. La Ë no se ha diptongado {pecho, entero) ni en los
textos antiguos ni en el habla moderna, excepto en el verbo: tiengo, viengo;
yes < EXIT, en el Alexandre. Hoy, las formas verbales, con o sin diptongo, se
reparten confusamente en el dialecto. (Zamora Vicente 1967: 98)
What we observe is a great deal of variability in the appearance of these
diphthongs (as opposed to monophthongs).
Nevertheless, lack of diphthongized forms could be the result of modern
Castilianization. What was the situation in Old Aragonese and Old
Leonese? Zamora Vicente is quite certain about the ancient regularity of
diphthongization before yod. The following discussion of Leonese diph-
thongization before yod is indicative:
La diptongación de ö ante yod es normal en los antiguos textos: nueche 'no-
che'; mueyo 'mojo'; cueya < COLLIGAT; ue, vué < HODIE; vuedia < HODIE
DIEM; ueyo < OCULU. Todavía Lucas Fernández usa duecho < DOCTU. Otros
ejemplos son: Aradue, moderno Araduey (Sahagún, 1096); Pedro Abrueyo
< APERIOCULU (Sahagún, 1171); Pedro Redrueyo < RETROCULU (Sahagún,
1253); uecho 'ocho' (León, 1260); etc. La toponimia confirma la vitalidad y
extensión antiguas del cambio. Arguelles < ARBOLEIS, 1064, en Oviedo, y
Argüellos en La Vedila; Los Fueyos < FOVEU, en Asturias, León y Sanabria;
Sigüeya en La Cabrera Baja (frente al castellano Segovia); Cirigiiello, En-
truello, en Asturias. (Zamora Vicente 1967: 93)
It should be noted, however, that much of this evidence is actually quite
late, post-dating the Burgos phase of mixing in Castile by several centuries.
For the period of Orígenes, Menéndez Pidal is able to provide only a few
 Linguistic changes 129

examples of Leonese, Aragonese and Mozarabic diphthongization before

yod (as compared to far more abundant examples in other positions):
El caso especial de Ö + yod ofrece pocos ejemplos.
(Menéndez Pidal 1964: 139)

La Ë diptongada ante yod ofrece casos más oscuros y raros que la Ö.

(Menéndez Pidal 1964: 158)
Moreover, Menéndez Pidal himself was much more circumspect in his
claims. In fact, he claimed that diphthongization before yod was simply far
less frequent in Castile than in neighboring (or, in my view, contributing)
dialects. The evidence of diphthongization before yod appears to grow in
Leonese and Aragonese texts only over time, but lags behind that of diph-
thongization in other positions. It is only for later stages that Pidal is able
to offer significant evidence of Leonese and Aragonese diphthongization
before yod (Menéndez Pidal 1964: 139). Even at later stages, it is not clear
that diphthongization before yod became regular; Staaff (1907: 207) found
only four orthographic examples of diphthongization of loi before yod in
62 13th-century documents from eastern Leon and only 6 examples in 30
documents from Western Leon (described as "tres veces mas abundantes").
Of course, this irregularity in textual evidence could also be due to a lack
of phonologization of the diphthongs (and thus lesser inclination to repre-
sent them in the orthography), to growing Castilian influence or to Latiniz-
ing tendencies, but if diphthongization before yod had been absolutely
regular, it seems that Staaff would have found more examples.
A plausible explanation for the above situation would be that spontane-
ous diphthongization spread through the process of lexical diffusion (e.g.,
Wang 1969), in which speakers adopt features on a word-by-word basis.
This adoption is affected by other factors: verb morphology and phonetic
environments can favor or constrain the adoption of new sounds, as can
word frequency. In early Hispano-Romance, it would appear that words
with low mid vowels followed by yod were less likely to be articulated
with a diphthong (tautosyllabic nasals also seemed to impede diphthongi-
zation).90 In this sense, they were at the "end of the queue" for the change
(with a few high frequency exceptions, such as viejo). In areas where mix-
ing was less and social networks more stable (such as Asturias, Leon, and
mountain regions of Aragon), speakers retained a large number of variants
in the articulation of loi and Id. For example, hi, whatever its position,
was probably articulated with a number of these variants: [D, WO, we, wa,
woe, wa]. As a result, speakers, over many generations, could continue to
130 The Burgos phase

induce the same phoneme for frequent [a] of oyo (< OCULU) and the fre-
quent [wo, we, wa, woe, wa] of bueno (< BONU), because either word could
be articulated with the less frequent variant (i.e., a diphthong for oyo and
monophthong for bueno). It is the maintenance of this rich but systematic
variation that allows speaker-learners to construct and maintain an abstract
phonemic connection between the surface variants, and to continue the
analogical process of lexical diffusion (in the Leonese case, this is exem-
plified by the extension of diphthongization to words with underlying hi
and /ε/ and following yod, such as in Leonese ueyo < OCULU).
These connections through variation were broken in Castile, and led to
an abrupt interruption of the process of lexical diffusion. At the time of the
Burgos phase repopulation, phonetic diphthongization was clearly well
under way, but it had not yet spread (consistently, at least) to words with
low mid vowels followed by yod.91 In the koineizing environment of Cas-
tile, however, adult speakers were reducing the non-functional variation in
their speech as they accommodated. The most frequent and consistent ar-
ticulation of words with low mid vowels before yod would have been with
a monophthong (e.g., ojo < OCULU), and as adults accommodated out in-
frequent variants, the monophthong would have become even more preva-
lent for such words.
However, as we have seen, diphthongal articulations were at the same
time being reanalyzed as sequences of two phonemes, with a reduction in
the number of diphthongal variants. With the loss of a wide range of vari-
ants, children could not have perceived any connection between the [D] of
ojo (or its equivalent) and the [we] of bueno (or its equivalent). Given such
input (the output of other speakers), a reasonable induction for child (and
adult) learners to make would be to relate the monophthong with the sur-
viving monophthongs closest to it in phonetic terms: the (formerly high)
mid vowels loi (in the case of [o]), and Id (in the case of [ε]). This merger
between hi and loi, and between /ε/ and /e/, produced by enough speakers,
would not have encountered resistance in the environment of fluctuating
norms and weak norm enforcement. It is the combined effects of this
merger with the phonological reanalysis of the diphthongs in the koineizing
context that leads to the reduction of the 7-vowel Romance system to the
simplified 5-vowel system of Castilian. But these phenomena also result in
a mixed pattern that is clearly marked in comparison to the surrounding
dialects - one of those patterns that Posner (1996) might label "deviant".
This may be so, but the changes in question are best understood as the ef-
fects of koineizing reanalysis and simplification.
 Linguistic changes 131

3.3. Koineization and other changes of the Burgos phase

The two groups of changes analyzed above show how the model of koinei-
zation can contribute significantly to our understanding of how and why
certain changes arose when and where they did. Though we cannot claim
to have fully accounted for the actuation and transmission of these
changes, the model does allow for a far broader range of factors to be in-
cluded in our necessarily partial reconstructions. Indeed, the preposition +
article contractions and the particular changes in the tonic vowel system
are phenomena that remain quite mysterious if not analyzed as koineizing
changes. Still, having established that koineization certainly occurred dur-
ing the período de orígenes, one must also ask: what was the impact of
koineization on other changes that are frequently associated with this pe-
riod? Menéndez Pidal (1964) is well known for having defined a group of
ten changes that characterized the early Castilian of Burgos, and then were
carried southward in a progressively widening band, creating what is
known as the Castilian cuña or wedge that supposedly broke the previous
dialectal unity of the center of Visigothic Iberia. These characteristic fea-
tures include:

1. Development of F- > [h-]

2. Development of Vulgar Latin [-lj-] > [3]
3. Development of Latin -CT- > [tj]
4. Development of Latin hi > [we]
5. Lack of diphthongization before yod
6. Lack of palatalization of IVI (> [λ-])
7. Coincidence in a number of features between east and west that are not
present in Castilian, e.g., phonetic variation in the realization of the spon-
taneous diphthongs was frequent to both the east and west, but not in
Castilian; many forms of the verb ser are diphthongized in Navarro-
Aragonese and Leonese {tú yes, él ye), but not in Castilian (tú eres, él es).
8. The assimilation of Latin /mb/ > /m/.
9. The monophthongization of Vulgar Latin [aj] > [ej] > [e]
10. The development of Latin consonant clusters PL-, CL-, FL- > [λ].
(Menéndez Pidal 1964: 490-502)

Scholars do not now understand all these changes in the same way that
Menéndez Pidal did, but his list of features, restricted though it may be, is
widely accepted as representing key changes of early Castilian. Some of
them are certainly much easier to understand when analyzed in terms of
132 The Burgos phase

koineization. This is the case of items 4, 5, and 7, which have already been
discussed above (Menéndez Pidal did not include the development of the
preposition + article contractions in his list). In theory, however, all the
changes should be analyzable (and must be analyzed) in terms of koineiza-
tion, even if koineization merely preserves features found in one or several
established contributing dialects. Such changes may be seen as only tan-
gentially related to the process of koineization, but as consistent with it.
Indeed, one would not expect to find changes that increase complexity and
opacity, and such a finding would weaken the plausibility of other explana-
tions based on koineization. Below, I discuss briefly three of the changes
mentioned by Menéndez Pidal in order to highlight three different issues
relating to application of the model. First, although the model cannot al-
ways be used to construct explanations such as those found above, it may
still be useful in judging the relative value of other explanations. Second,
some apparent increases in complexity (e.g., the addition of a new pho-
neme) may in fact result from a decrease in variation (and reanalysis).
Third, even changes which seem (at first glance) to fit the model exactly
may in fact be better explained as the result of other processes.

3.3.1. The aspiration of F-

One of the best known and most studied features of early Castilian is the
change in pronunciation of (normally word-initial) Latin F-. It is generally
accepted that by the 10th century the word-initial /f-/ of Latin/Romance
developed into [h-] in Castilian. This [h-] then spread south along with
reconquest, and in later centuries was itself lost in many varieties of Castil-
ian, including the standard.92 Scholars have tended to take one of two
stands regarding the origin of this feature. Menéndez Pidal (1964) and his
followers have argued strongly that the change must be understood primar-
ily as a result of a Basque (and Cantabrian?) substrate on early Castilian.
Others, including Penny (1972, 1990), believe the change was primarily an
internal development of the Romance of Cantabria and/or Castile. Trask
(1997) strongly favors an internal explanation, and vehemently rejects the
Basque substrate thesis. Lloyd (1987) positions himself between these two
extremes. As we will see, an analysis based on the model of koineization
tends to support Lloyd's position. Let us consider first the Basque substrate
hypothesis and the arguments leveled against it. Since Castilian was clearly
in some kind of contact with Basque, many assume that this seemingly
 Linguistic changes 133

unusual change (at least within Ibero-Romance) is due to Basque influence,

particularly since the same change in found in Gascony, which borders the
modern Basque Country to the North, and where a form of Basque was
spoken among the Aquitanians when the Romans arrived (Trask 1997: 36).
It has long been known that Pre-Basque (the precursor to modern Basque
spoken in Roman times) had no Iii, and it is argued that when Basque
speakers adopted Romance speech in (or near) Castile, they could not pro-
nounce the [f] and replaced it with [h], the closest phone available in their
own speech. Though the existence of [h] has been questioned, Trask (1997:
127, 157) indicates that it is certain that Pre-Basque included a frequent
and prominent phonetic aspiration which accompanied stressed vowels
(this was later phonologized in many varieties of Basque).
Nevertheless, many authors have rejected this hypothesis. Most re-
cently, Trask (1997: 424-429), following Izzo (1977), adamantly rejects
the possibility of Basque influence on the change, but his rejection is
grounded on a very conservative notion of substrate theory. For instance,
he assumes that Basque must once have been spoken in Castile as the pri-
mary language in order to speak of direct influence, and he stresses that
there is no evidence that Basque was ever a stable language variety of Cas-
tile. This is true enough, but it is not necessary that this be so if one con-
siders potential Basque influence during dialect mixing as opposed to lan-
guage shift by a stable population. Basque settlers most certainly moved
into early Castile in significant numbers and would have had to learn Ro-
mance. Their linguistic production would then have formed part of the
prekoine linguistic pool. We might therefore accept that Basque speakers
had some impact on the change F- > [h-].
However, Trask includes other arguments (also made by Izzo and
scholars before him) that might call into question even a less rigid view of
Basque influence on this change:

1. Romance /f-/ became [h-] in most cases, but not before the novel diph-
thong /ue/ (realized as [we]) or before liquids (cf. Castilian fuente 'foun-
tain', frío 'cold'). Why would the Basques pronounce some instances of
/f-/ well and others incorrectly?
2. Latin and Romance words with /f-/ were borrowed into Basque, but in
such cases the /{-/ was regularly reproduced with Basque [b-] or [p-].
Thus, Latin/Romance festa became Basque besta or pesta. Why would
Basque speakers reproduce the words one way within their own commu-
nities and another when outside those communities?
134 The Burgos phase

3. As Menéndez Pidal pointed out, the change /f-/ > [h-] is shared by Castil-
lan with Gascon. On the other hand, the Romance of Navarre, where
Basques dominated in demographic terms, retained the /f-/. If Basques
were responsible for the change, why did it not occur in Navarre?
4. The change /f-/ > [h-] is found in other varieties of Romance (e.g., in
southern Italy) far from the Basque Country. Lloyd (1987) also points out
that the change is frequent cross-linguistically, so there seems to be a de-
gree of naturalness to the change. If this is so, why do we need to resort
to explanations based on Basque influence?

Such arguments cannot be dismissed, and they oblige us to consider two

related factors: the linguistic models presented to learners and the social
context of learning. Item 2 should be understood as referring to cases of
borrowing, in which speakers often have only limited access to models and
the items they learn are for use within their own speech community. They
are therefore free to adopt foreign words such as festa and adapt them to
their own phonology; this is a normal phenomenon in borrowing. On the
other hand, item 3 requires an understanding that Basques learning Ro-
mance in Castile or southern Navarre would have needed to learn a variety
of Romance for communication with native speakers of that variety (or
those varieties). When Basque (and other) learners were presented with
clear and consistent models, used by a majority of speakers in the target
community, it is probable that they (or their children) acquired the target
forms correctly. We know that the rate of expansion was much slower in
Navarre than in Castile, and that the degree of demographic and dialect
mixing was far less in there than in Castile. Basque learners of Romance
were therefore presented with more stable and homogeneous input than
they might have received in Castile, where the rapid mixing of speakers
from several regions to the north, west, and east, would have made the
input more unstable and heterogeneous. Moreover, the relative demo-
graphic stability of Navarre would have contributed to much stricter norm
enforcement, which would have served to suppress Basque interlanguage
features not shared by native Romance speakers.93
If Basque learners of Romance had heard Romance speakers in Castile
consistently articulating [f], then they (and, more importantly, their chil-
dren) probably would have learned this articulation. But the change to [h]
did occur, but not in all possible contexts (as indicated in Item 1), so it is
likely that at least some native speakers within the early Castilian speech
community showed some variation in their speech. As indicated above in
Item 4, other regions of the Romance world far distant from the Basque
 Linguistic changes 135

country also show the change ItΊ > [h], so it is quite plausible that native
Romance speakers introduced the variant [h] into their speech.
What sort of variation might this have been? A first consideration is that
Latin/Romance /f-/ was probably articulated as both a labiodental [f] and
as a bilabial [φ]. The labiodental was favored in those varieties of Ro-
mance that also developed a labiodental voiced frivative [v], whereas in
central Iberia the bilabial would have been favored through (unconscious)
analogy with the bilabial voiced fricative [β] (Lloyd 1987: 213). According
to Penny (1972, 1990, 1991a), the articulation of /f-/ (or /φ/) was probably
once limited to [φ] in all contexts. Later the appearance of the spontaneous
diphthong with [w] would have led to the addition of a rounded variant [M]
before the glide [w]. At a very early stage, possibly in Cantabria or primi-
tive Castile, before expansion to Burgos and beyond (but this is impossible
to say, for Cantabrian usage might have followed Castilian in this regard),
this was complicated by the development of another allophonic variant in
complementary distribution with the earlier variants:

— [h] before syllabic loi, Ivrf: FURNU > [horno] horno 'oven'
— [M] before [w]: [Mwérte] fuerte 'strong'
— [φ] elsewhere (before other vowels, the glide [j], /r/, IV)·. FARINA > [eparina]
harina 'flour' (Penny 1991a: 80)

Penny suggests that the use of [h-] before back vowels arose from an ar-
ticulatory conflict between the slit fricative [φ] and the following rounded
back vowels. Given that [φ] is lax and inherently unstable (Penny 1972:
466-467), an easy and probably automatic solution in fast speech would
have been to simply relax the articulation of the lips even further, which
would lead to the articulation (and/or perception) of a silent vowel similar
to a glottal aspirate. Once [h] was introduced as an allophone (probably
through listeners' reanalysis of the weakened bilabial), it could be general-
ized to other contexts, leading to the following distribution that was proba-
bly characteristic of Old Spanish (and is retained in many rural dialects
today):

— [M] before [w]: [Mwérte]/weríe 'strong'

— [φ] before hi, IV\ frente, frío
— [h] elsewhere; before all syllabic vowels and glide [j]: FERRU > [hjéro]
(adapted from Penny 1991a: 80)
136 The Burgos phase

How might koineization have affected this development? If some Romance

speakers (perhaps of Cantabrian origen) began to pronounce [h] in at least
one phonetic context, then the door was open to the extension of this phe-
nomenon. If the group of innovators formed a sizeable bloc of the early
Castilian population, then it is easy to see how the first stage described by
Penny could have been favored in the koineizing context. It is possible that
overgeneralization of the [h] allophone occurred during Burgos-phase
koineization among native Romance speakers, particularly since it is a
more natural change. Nevertheless, the maintenance of /f/ was typical of a
number of contributing dialects, and this fact might have worked against
extension of the phenomenon. Basque learners of Romance may have
played a role in extending the phenomenon by increasing the relative fre-
quency of the [h] allophone in the prekoine linguistic pool. Presented with
varying realizations of the phoneme /f/, Basque speakers may have tended
to reproduce more easily and more frequently the one (perceived) allo-
phone which existed in their native language: [h]. Basque learners of Ro-
mance may thus have contributed to "tipping the scales" in favor of the [h]
allophone.
This reconstruction is just one of several that are possible (see Lloyd
1987: 220-223 for a slightly different version). The extreme lack of evi-
dence means that no definitive explanation of the change /f/ > [h] will ever
be possible, and applying the model of koineization gets us no closer to a
definitive explanation. However, if we view this change as part of the
process of Burgos-phase koineization, then we must reject extreme argu-
ments such as that of Trask, in which even the mere possibility of Basque
influence is rejected out of hand.

3.3.2. The development of /t//

The development of -CT- to [tj] (= orthographic ch) is a particularly salient

feature of early Castilian. In most Ibero-Romance dialects, the syllable-
final [k] in Latin words such as FACTU, NOCTE, LACTE, OCTO, developed
first to a fricative [x] or [c], and then voiced in assimilation to the preced-
ing vowel (which itself was raised if a mid-vowel), producing diphthongal
forms such as feito, noite, leite, oito, in Galician-Portuguese, western
Leonese, and Aragonese (Lloyd 1987: 252-253; Zamora Vicente 1967:
150, 242). Latin words with the sequence -ULT- showed a similar develop-
ment; for example, MULTU > muito, in Portuguese, western Leonese, and
 Linguistic changes 137

Aragonese. However, in Castilian and neighboring eastern Leonese, the

Romance group [-jt-] regularly produced [tf] as the realization of the new
single phoneme /t//. This is surely one of those Castilian changes that is
considered "phonologically deviant" by Posner (1996: 208) and therefore
not to be associated with the dialect leveling of koineization.
It might also be seen as introducing a new phoneme which existed in
none of the contributing varieties and thereby leading to an unexpected
increase in the inventory of phonemes. However, I view this change as
neither deviant nor contrary to the process of koineization. We have little
or no evidence of the articulations of the sequence [-jt-] in most of the con-
tributing dialects in early Castile. Still, assimilation of the following con-
sonant to a neighboring palatal glide is likely to occur at least sporadically
in all varieties showing this sequence.94 If, as is often assumed, the dialect
of Cantabria had previously developed an allophonic variant [-jt/-], then
this would have entered the Castilian prekoine linguistic pool. Still, this
variant would have entered into competition with the conservative and
common articulation [jt]. In the koineizing environment, however, it may
have been favored by a natural articulatory tendency toward (and/or per-
ception of) palatalization of the consonant in this context (especially in fast
speech), and norm enforcement mechanisms would not have limited its
extension. As such, koineization might actually have favored the extension
of this variant and the eventual loss of the original sequence [-jt-] (includ-
ing loss of the glide). The phonologization of the palatal variant of the
original Ν is then no more than the end result of the decrease in allophonic
variation that we expect to find in koineization, although in this case we
have the addition of a new phoneme. We must therefore conclude that it is
possible to occasionally find the development of new phonemes in koinei-
zation, but only when this is the result of the rapid elimination of numerous
variants.
Nevertheless, this is not the perfect case to introduce this argument, for
the phoneme /tj/ was almost surely introduced into all Romance varieties
of the northwest and northcenter of Iberia (before Burgos-phase koineiza-
tion in Castile) through the common development of post-nasal consonant
clusters (e.g., AMPLU > ancho·, see below). As such, many speakers in early
Castile may have perceived (and reanalyzed) a slightly palatalized variant
of the consonant in [-jt-] as an exponent of the pre-existing phoneme /t//,
and then begun to reproduce it as [tj]. Over a generation or two, this could
easily have led to the regular development of [-jt-] to [tj], and then as /tj/.
Of course, this is just one potential reconstruction of events. As in the case
138 The Burgos phase

of the change /f-/ > [h-], the lack of contemporary evidence does not allow
us to prove this or any version of events. However, what I hope to show
here is not that the change in question is necessarily the result of koineiza-
tion (though it may be); rather the important conclusion is that this change
(and others like it, such as the development of medieval Castillan /g/) are
perfectly consistent with the process and expected outcomes of koineiza-
tion, and therefore do not constitute counter-evidence to my more specific
claim that koineization played a key role in the development of the preposi-
tion + article contractions and the system of tonic vowel phonemes. The
general implication for the model of koineization is that, during the level-
ing out of phonetic variants, very frequent allophones could be selected by
speakers and phonologized. This in turn would actually allow an increase
in phonemic inventory, but at the expense (simplification) of phonetic
variation.

3.3.3. The development of CL-, PL-, FL-

The Latin consonant clusters /kl-/, /pi-/, /fl-/, were all affected by a com-
mon change in Late Latin/Early Romance: palatalization of the lateral,
which produced the clusters [ρλ-] [kl-] [ίλ-]. These feature-heavy and un-
stable clusters were nearly everywhere simplified in some way.95 Accord-
ing to Menéndez Pidal, the regular Castilian development of word-initial
PL-, CL-, FL- was to a bare palatal lateral /λ/. As we have seen, he identified
this as one of the defining features of early Castilian (Menéndez Pidal
1964: 502 and 1962b: 126). There are numerous examples, including the
following (from Penny 1991a: 63):

— CLAMARE > llamar ' t o call', CLAUSA > llosa ' e n c l o s e d field', CLAVE > lla-
ve ' k e y '
— PLAGA > llaga ' w o u n d ' , PLANU > llano ' f l a t ' , PLICARE > llegar ' t o a r r i v e '
— FLAMMA > llama ' f l a m e ' , FLACCIDU > llacio (later lacio) ' l a n k ' , FLAMMU-
LA > Lambra 'personal name'

On the other hand, it has been observed that other words which must be-
long to the ancient wordstock of Castilian do not show the change (Wright
1980; Anderson 1992; Corominas and Pascual 1981):
 Linguistic changes 139

— PLAGIA > playa 'beach', PLATEA > plaça > plaza 'plaza', PLICARE > plegar
'to fold', PLANGERE >plañir 'to wail', PLANTA >planta
— FLACCU > flaco 'skinny', FLORE > flor 'flower', FLUXU > floxo > flojo
'loose', FLOCCU > flueco > fleco 'fringe'
— CLAVICULA > clavija 'peg', CLAVU > clavo 'nail'

Menéndez Pidal explicitly rejects these as popular Castilian forms: "en

época posterior o por influencia culta, se conservó el grupo" (1962b: 126).
However, these words cannot be shown to be simple borrowings (whether
dialectalisms or Latinisms), since they often combine "non-Castilian" fea-
tures (conservation of the cluster) with archetypal popular Castilian fea-
tures, such as the palatals /·$/ and /ji/ in clavija and plañir. More damaging
still for Menéndez Pidal's argument is Badia Margarit's (1972: 148) obser-
vation that the empirical base needed to extend the palatalization rule to
FL- is lacking; there are only six generic Latin etyma beginning with FL-, of
which the majority conserve the cluster (the exceptions are llama <
FLAMMA, lacio < FLACCIDU). To further complicate matters, Malkiel
(1963-1964: 146-155) and Penny (1991a: 63) argue that the following
forms should be considered regular popular developments in Castilian:

— *PLATTU > chato 'snub-nosed'

— *PLOPPU (< POPULU) > chopo 'black poplar'
— PLUTEU > chozo > choza 'hut'
— CONCLAVARI > conchabar 'to join together'
— MACULA > *mancla > mancha 'stain, spot'
— AMPLU > ancho 'wide'
— IMPLERE > fenchir > henchir 'to fill u p '
— INFLARE > flnchar > hinchar 'to inflate, swell'

When one takes into account the fact that Cantabrian and central and east-
ern Asturian varieties tend to agree with Castilian in the development of
CL-, PL-, FL- > /λ/, that eastern varieties (Catalan, Riojan and Navarro-
Aragonese) preserve the initial clusters, and that western varieties
(Galician-Portuguese and western Astur-Leonese) show the development
CL-, PL-, FL- > /ι//, it is tempting to analyze these changes as the results of
mixing during koineization. However, a closer examination reveals that
koineization played no more than a secondary role (in preserving these
developments), and that the basic patterns arose before the period of
Leonese and Castilian koineization.
140 The Burgos phase

The inventory of words affected by palatalization is surprisingly similar

across the northwest and north-center of the Iberian Peninsula; words in
Castilian that show palatalization to /λ/ tend to correspond to words in
Galician-Portuguese that show palatalization to /tj/ (as in Group I of Table
14), while words preserving (more or less) the original cluster tend to cor-
respond as well (as seen in Group ΙΠ). As the examples in Group Π make
clear, however, western varieties appear to palatalize more items in the
lexical inventory than does Castilian. Varieties of Western Astur-Leonese
tend to follow the Galician-Portuguese distribution of forms and specific
outcomes, while central and eastern Astur-Leonese and Cantabrian varie-
ties generally agree with the Castilian patterns. There is some evidence that
even central and eastern varieties of Astur-Leonese followed the western
patterns more than Cantabria and Castile. For example, Asturian retains the
form llantón (cf. Castilian planton), which corresponds to Galician (and
Old Portuguese) chantao. There is also contemporary dialect evidence
which seems to show that a form *llor was once in use in Asturias and
Leon (Corominas and Pascual 1981: 917).
Nevertheless, the fundamental coincidence in these inventories reveals
that the changes in Castilian and other northwestern and north-central va-
rieties must surely be directly related to those in Galician-Portuguese. In
fact, what we observe here are almost surely the outcomes of several proc-
esses of lexical and social diffusion that occurred throughout northwestern
Iberia, but whose epicenter was Galicia itself. The first change certainly
occurred before the Burgos phase of koineization, and probably prior to the
Moorish invasion. This first regular change to diffuse through the lexicon
and the region was the elimination of the consonant preceding the palatal
[λ] in many words containing the original Latin clusters. However, some
words were left unaffected by this change, and actually preserved or re-
stored the original clusters. This process of lexical diffusion was affected
by various factors: CL- and PL- were more often affected than FL- (Malkiel
1963-1964 suggests that this was because the fricative was weakly inte-
grated into the group affected); word-internal post-consonantal position
favored the change most strongly (as in Group V), word-initial position
less so.96 The process of lexical diffusion was also one of social diffusion
across geographical space, and the existing evidence indicates that more
forms were affected in the extreme northwest (probably the geographical
center of the change) than in areas to the east (Asturias and Cantabria, and
areas south that would later be depopulated and repopulated).
 Linguistic changes 141

Table 14. Comparative development of words showing CL-, PL-, FL-, in Old
Galician-Portuguese and Castilian.

Category Galician-Portuguese Castilian

Group I chave llave 'key'
chamar llamar 'to call'
chäo/chan/chao llano 'flat'
chaga llaga 'wound'
chegar llegar 'to arrive' (note
doublet: plegar)
cheio lleno 'full'
chanto/pranto llanto 'weeping*
chuva/chuvia/choiva lluvia 'rain'
chama llama 'flame'
chantagem, chan- llantén 'plantain'
taxe/chanchagem
Group II changer/chanzer plañid1 'to wail'
chantar/prantar plantad* 'to plant'
chevella/c(a)ravilla clavija 'peg'
chumbo plomo 'lead'
chôr/frol/flor flor 'flower'
Group III cravo clavo 'nail'
prazer plazer 'to please'
praça plaça 'plaza'
prazo plaz(d)o 'term, period'
praia playa 'beach'
pregar plegar·" 'to nail'
fraco flaco 'flabby, weak'
froco fl(u)eco 'lock of hair'
Group IV chato chato 'snub-nosed'
chopo chopo 'black poplar'
choza choza 'hut'
Group V ancho ancho 'wide'
inchar hinchar 'to inflate'
encher henchir 'to fill up'
mancha mancha 'stain'

Subsequently, the palatal lateral was devoiced and affricated in some posi-
tions, producing [tf]. This occurred with absolute regularity in word-
internal post-consonantal position throughout the region (Group V). In
142 The Burgos phase

Galicia, this voiceless affricate was then generalized to initial position in

any words that had already been shifted to initial [λ-] (Lloyd 1987: 226).100
However, in regions to the east, the older palatal lateral was retained in
initial position in those words that had been affected by the earlier change.
During the Burgos phase, the settlers from Asturias and Cantabria carried
[λ] southward, and it survived as the dominant form in Castile and central
and eastern Leon.
Still, the results in Leon and Castile were not exactly alike, for
Galician-Portuguese and Leonese were heavily affected by yet another
change: rhotacism of the lateral in remaining conservative clusters. This
process led to forms such as cravo and praça (Group ΙΠ). Rhotacism was
very regular in Galician-Portuguese (i.e., phonologized), but was restricted
(at most) to mere phonetic alternation between [r] and [1] in Asturias and
Cantabria.101 Leon, heavily influenced by the norms of its many settlers of
Galician origin, followed the western tendency of converting remaining
/kl-/, /pi-/, /fl-/, to /kr-/, /pr-/, /fr-/. Castile, on the other hand, shows differ-
ent results: the remaining conservative clusters in words such as clavo and
plaça remained unchanged. Such results are due in large part to lesser in-
fluence of westerners among its settlers (particularly in the earlier stages).
We may thus conclude that, during the Burgos phase of koineization in
Castile, the pre-existing distribution of forms found in Cantabrian, As-
turian, and Leonese varieties was used by the majority of settlers and there-
fore accepted into the koine. On the other hand, the strong western ten-
dency towards rhotacism in conservative clusters was rejected, due to its
relative weakness in Asturias, Cantabria, and the east.
Finally, we must consider the word-forms chopo, choza, and chato. If
we accept that these words do form part of the ancient wordstock of Castil-
lan (as argued by Malkiel 1963-1964 and others), then these forms must be
analyzed as the only (surviving) forms affected by the change [λ] > [tj] in
the north-center; in other words, they were at the "beginning of the queue"
for this change in north-central regions, but no other words were affected
by the change. On the other hand, the earliest attestations of these words
are from the 13th century or later (Corominas and Pascual 1981), so it is
possible that they were introduced by Galician and (western) Leonese set-
tlers to Castile before intentionally phonographic Romance writing became
common in the 13th century. This may have happened at any time, even
during the Burgos phase (Barrios [1985] shows that Galicians did partici-
pate in early attempts at repopulation south of the Duero), but speakers
from these regions were probably much more influential during the later
 Conclusion 143

Toledo phase of koineization, and even more so in Andalusia (a similar

interpretation is suggested by Lloyd 1987: 225).

4. Conclusion

Though Menéndez Pidal described well both the social and linguistic
changes of early Castile, he lacked a means of showing a causal relation
between them. Koineization allows us to do just that. During the Burgos
phase of Castillan repopulation (9th to 11th centuries), immigrants from
Cantabria, Asturias, the Basque Country, Navarre and La Rioja, Leon (and
Galicia), as well as Mozarabs from the south and east, settled in the border
county of Castile. The bulk of these immigrants spoke different dialects,
generally mutually intelligible and considered to be the same language,
though the Basques, who spoke an unrelated language, also moved into
Castile in significant numbers and there had to learn the Romance speech
of the majority. The rapid influx of settlers led to a breakdown in social
networks, as seen in the rise of importance of the small landowner, the
nuclear family, and customary law. At the same time, the conflicting
speech habits of the settlers led to a rapid increase in the number of vari-
ants in the linguistic pool, including all the variants of the contributing
dialects, as well as innovations in learner language. Accommodation be-
tween speakers frequently led to the suppression of variants in favor of
some common alternative, but also motivated learning of new features.
Cultural and political factors also appear to have favored rapid focusing of
this dialect.
These conditions had significant linguistic effects on the Romance spo-
ken in early Castile. The most obvious changes include the reduction of the
many difficult-to-leam preposition and article contractions, as well as vari-
ants of the articles themselves, to invariant and transparent sequences of
prepositions and articles. Burgos-phase koineization led to the reanalysis of
the originally phonetic spontaneous diphthongs as sequences of two pho-
nemes and the merger of the vowel of undiphthongized forms (i.e., words
containing low mid vowels followed by a front glide) with the originally
high mid monophthongal vowels, with the larger result that the vowel pho-
nemes were reduced from the seven of Romance to the five of Castilian.
Surface analogies, phonological reanalysis, and identity factors led to the
choice of the characteristic /ue/ diphthong.
144 The Burgos phase

Given the impact that koineization had on early Castilian, other changes
must also be analyzed in relation to koineization (since it is only too easy
to attribute all changes to koineization). Discussion of the development of
/f/ > [h] allowed us to conclude that Basque influence cannot be ruled out,
as Trask (among others) has attempted to do, but the lack of data makes
firm conclusions impossible. The development of /tj/ shows that new allo-
phonic variants can be added, and, once allophonic variation decreases, the
remaining variants can, in principle, be phonologized. Nevertheless, the
fact that this phone or phoneme was probably already used in other con-
texts and in some contributing varieties means this is probably not a case of
introduction of a new phoneme (though it shows how that might happen
during koineization). Finally, the varied results of Latin CL-, PL-, FL- might
seem to be prototypical effects of mixing, but in fact the broad pattern is
better explained as a result of earlier lexical diffusion and areal spread
across the northwest, and koineization had only a secondary effect on out-
comes in Castile.
Chapter 4
The Toledo phase

The slow advances of the Burgos phase largely ceased at the end of the
10th century with the defeats inflicted by the Moorish general Almanzor,
but were resumed and greatly intensified in the Toledo phase, beginning in
the late 11th century. Unlike the Burgos phase, the Toledo phase has
hardly been considered of any significance to the development of Castilian.
This chapter, therefore, focuses on demonstrating that there are indeed
changes in medieval Castilian that can best be explained as linguistic ef-
fects of the demographic and dialect mixing that occurred in this period.
Moreover, the rapid simultaneous advance of Leon and Castile (under one
king at the time) across the entire central and western Moorish-Christian
frontier favored not only koineization, but the spread of features of the
already mixed and reduced Castilian koine.

1. Social history

Following the initial occupation of the northern half of the Duero Valley,
reconquest and repopulation occurred only slowly and on a relatively small
scale, being limited in large part to the attempts to repopulate the southern
bank of the Duero. However, several factors would combine to propel the
reconquest into a new phase of rapid expansion along the whole Chris-
tian/Moorish frontier. Bishko (1975: 356-360) specifies three main factors
that contributed to this intensification. First, the Christian kingdoms had
attained a level of development that could support such an effort. Signifi-
cantly, the steady rise in the population allowed the Iberian kingdoms to
produce more commodities, to field more men for battle, and to hold re-
conquered territories through repopulation. Second, the Moors had been
weakened by the breakup of their territory into competing taifas, with con-
stant internecine strife. This allowed Alfonso VI - king of both Leon and
Castile - to take the central and key kingdom of Toledo in 1085, the fall of
which held huge symbolic import for the Christians of all kingdoms since it
had been the Visigothic capital and the seat of the Primate of Spain.
146 The Toledo phase

However, this victory caused the remaining Hispanic taifas to invite Af-
rican forces to aid them against the newly successful Alfonso VI. The Afri-
cans, first the Almoravids (from 1086) and later the Almohads (from
1146), came to fight a jihad, took control of the Moorish territory and by
uniting it provided the fiercest threat the Christians had seen since the days
of Almanzor, a hundred years earlier. This in turn coincided with and pro-
voked the development of a similar attitude on the part of the Christians,
who began to view their struggles more clearly as crusades. Indeed, the
third factor Bishko cites is the development of the concept of crusade or
holy war against the so-called "infidels" not only in Palestine - where the
First Crusade took Jerusalem in 1099 - but in Iberia as well.
The taking of Toledo, the African threat, and the invention of the cru-
sades turned the Iberian Peninsula into the second front in the war on the
infidel. It attracted Europeans from beyond the Pyrenees, particularly Occi-
tans and northern French, zealously dedicated to the ideal of holy war and
the extermination of the enemy (unlike many Iberian Christians, who often
took a much more secular view of war with the Moors). This coincided
with and contributed to a general rise in French (i.e., from the area now
associated with the state of France) influence in the Peninsula, with many
persons of Gallic origin coming to hold important positions in the govern-
ment and the church, and large communities of merchants of the same ori-
gin being founded in the Spanish cities.
Another important effect of this expansion of the reconquest was the
constant militarization it imposed. Towns and cities (see below) were re-
quired to send militias and cavalry for large armies that had to be fielded
again and again, and to a degree that few other European areas would
know. These armies promoted the development of numerous weak social
ties among those involved. In addition, the constant need for military activ-
ity led to the establishment in the mid-12th century of the military orders,
seen nowhere else in Europe save the Balkans. These included the Tem-
plars and the Hospitalers, but also native orders based on these, such as
Calatrava, Santiago, Alcántara, and San Juan. These orders played a key
role in defense and repopulation of the new frontier, and are discussed
below.
The most important effect of the rapid expansion of the reconquest was
the necessary repopulation effort that followed in its wake, with concomi-
tant migration and demographic mixing. The taking of Toledo turned it
suddenly into the center of (a temporarily) united Castile and Leon, and
even of the entire Peninsula, and it attracted new inhabitants from all over
 Social history 147

the Peninsula and beyond. The sudden shift of the frontier from the south-
ern rim of the Duero valley to that of the Tagus left a wide, largely empty
gulf between Castile proper and the kingdom of Toledo. The area -primi-
tive Extremadura - extending from the south bank of the Duero to the
mountain ranges of the Central System, also needed to be occupied and
repopulated. During the course of the 12th century, the frontier moved
further south and the newly reconquered regions to the east (La Alcarria),
the southeast (La Mancha), and the southwest (the Montes de Toledo and
Leonese Extremadura), had to be repopulated. Finally, the interior regions
of the north, which suffered a drain on their own populations, constituted
another important zone of repopulation. In each area, there was great
demographic movement and mixing.
Toledo itself was one of the principal Moorish cities, with a large popu-
lation probably numbering in the several tens of thousands, and well-
populated surrounding areas, including the towns of Guadalajara, Madrid
and Talavera. When Alfonso VI conquered the city, many of its residents
as well as those of the surrounding region chose to remain. These included
the members of the sizable Mozarab community, who were spread
throughout the kingdom and who had maintained use of their own dialect
of Romance, but who had also become proficient in Arabic. Some Mos-
lems, having received guarantees of protection, also chose to remain (at
least for a time). These Mudejars were primarily people of humble position
who continued to work the fields or who worked in the building trades
(Μοχό 1979: 219). There was also a large Jewish community, and both the
Mozarab and Jewish communities saw their numbers increase as Christians
and Jews fled persecution under the Almoravids and Almohads in Al-
Andalus. There is evidence that (parts of) the Mozarab community retained
a separate identity up to about 1150, from which time it declined until be-
ing completely absorbed by the general community, probably by 1300
(Μοχό 1979: 220).
Significant numbers of northern Christians were also needed to defend
the new region and replace the many Moors who had left. The new arrivals
to Toledo and its surrounding area - who came throughout the 12th century
- included a core group of Castilians, but also many Leonese, and signifi-
cant numbers of Asturians and Galicians: These were also joined by
Navarrese, Aragonese, and Catalans (Μοχό 1979: 221). Yet another group
must also be counted among the new settlers:
Un nuevo y peculiar contingente de pobladores cristianos en Toledo va a es-
tar representado por los francos que hasta allí llegan para establecerse tras
148 The Toledo phase

su conquista por Alfonso VI, pues, aunque no se perciben sus huellas en el

ejército cristiano del monarca, gozaban ya de influencia en su corte con el
apoyo de la reina Constanza de Borgoña. Los francos establecidos en
Toledo debieron representar un número apreciable, para gozar de fuero y
barrio propios . . . Dentro de la población franca toledana resalta singular-
mente por su influjo el elemento clerical, en torno al arzobispo don Ber-
nardo, que llevó a cabo la reorganización de la archidiócesis de Toledo tras
su conquista por castellanos, pudiéndose considerar, a su vez, como de
origen marsellés la primera generación monástica toledana establecida en
San Servando . . . Los francos de Toledo, aunque recibieron su fuero propio,
disolvieron pronto su personalidad ante la mayoría castellano-mozárabe.
(Moxó 1979: 221)102
In sum, reconquered Toledo (particularly the city itself), though dominated
by Castilians, represented a population mix that was far more heterogene-
ous than that of primitive Extremadura and other regions.
Nevertheless, the heavy mixing that occurred in Toledo was echoed,
though on a reduced scale, in the cities that were founded to consolidate
the new "behind-the-lines" territory of primitive Extremadura. Towns just
to the south of the Duero, such as Olmedo, Medina, Coca, Iscar, Cuéllar
and Sepúlveda, were apparently repopulated "spontaneously" in the 11th
century, as settlers moved in and established themselves under the old pre-
sura system (though fueros were also granted). However, the urgent need
to secure this region led to an organized effort to found several new cities
(or re-establish old ones) as foci for repopulation. These included Sala-
manca, Ávila, Segovia, and Soria, each of which was founded as an admin-
istrative center with control - and linguistic influence - over the villages
and towns in the surrounding alfoz. The task of organizing these communi-
ties was undertaken by Alfonso's son-in-law, Raymond of Burgundy (a
franco), who initiated repopulation of Segovia and Ávila in 1088, just three
years after the fall of Toledo. Documentary evidence reveals that the set-
tlers of Segovia included not only Castilians (who represented a clear ma-
jority) but also settlers from Galicia, Asturias, Cantabria (montañeses),
Leon, La Rioja, Aragon and Gascony (Moxó 1979: 210). Similarly, the
Crónica de la población de Ávila reports the arrival of a majority of Castil-
ians, but also of many serranos (from the eastern areas of Castile and La
Rioja), Mozarabs, Aragonese, Jews, Mudejars, and Burgundians (followers
of Raymond). The repopulation of Salamanca, within the kingdom of Leon,
apparently began somewhat later than that of Segovia and Ávila, but it also
received a mixed group of settlers, though in this case westerners (Leonese,
Galicians, Portuguese) represented the majority; they were joined by many
 Social history 149

Castillans, but also Mozarab and Jewish refugees from Al-Andalus, fran-
cos, as well as some Mudejars. The definitive repopulation of Soria, to the
southeast of Burgos (and directly south of La Rioja), was completed by
Alfonso el Batallador of Aragon in order to support his encroachment upon
the Moorish kingdoms to its east. Some spontaneous repopulation had al-
ready been initiated by Riojans and Navarrese from the immediate north,
and to these were added the new settlers attracted by Alfonso, which in-
cluded a mix of Castilians and easterners.
In spite of the evident social mixing that took place in all these cities,
Μοχό accords preeminence to the Castilians:
Tales gentes procedieron de diversas regiones hispánicas, preponderante-
mente -claro es- del interior del reino de Castilla y León -como burgaleses,
riojanos, 'serranos', vascos, leoneses, palentinos, portugueses, asturianos y
gallegos-, pero también procedentes de otros reinos peninsulares, como
navarros y aragoneses, así como mozárabes, al igual que los que había lle-
vado Sisnando Davídiz a repoblar Coimbra con Femando I, e incluso - a lo
largo del siglo XII- de fuera de la Península, como los 'francos' de que se
tiene noticia a través de la documentación y que formaron en conjunto un
ingrediente importante de la población peninsular durante la plena Edad
Media. Pero ante todo, la nueva población de la Extremadura nació por en-
jambramiento del castellano situado al norte del Duero. (Moxó 1979: 205)

Still, in general this phase of repopulation shares a key feature with the
Burgos phase: in any area of repopulation, the majority of settlers appear to
proceed from regions directly north, with the next largest number of set-
tlers coming from neighboring regions. In all cases mixing and contact
occurred, and the expansion of Castilian into these areas cannot be seen as
the simple extension of the speech of the county of Burgos to regions
south. Neither can one argue for the simple political imposition of Castilian
speech, since, the expansion of Castilian features does not always coincide
with the borders of Castile (see discussion of language spread below).
Although the Castilians had pushed the frontier far south in 1085, the
counterattack initiated by the Almoravids in 1086 led to a retreat, including
the evacuation of existing Mozarab communities. Initial repopulation was
consequently limited to areas extending along and behind the Tajo River,
from Talavera to Guadalajara, running through Toledo and Madrid. Later
this line was gradually extended south and eastward, particularly in the
middle part of the century, with the taking of the fortresses of Calatrava
(1147) and Uclés (1157), although Calatrava was lost again from 1195 to
1212. The instability and centrality of·this area was to slow repopulation
150 The Toledo phase

and lead to a greater demographic mixing between Castilians, Mozarabs,

and easterners (in some areas). Indeed, the eastern frontier of Castile came
into direct contact with Aragon in the early 12th century. Zaragoza fell to
the Aragonese in 1118, and the Aragonese advance (after merger with
Catalonia) reached its southern limit with the founding of Teruel in 1171,
but was cut off after the conquest of Cuenca by Castile in 1177. The Fuero
de Guadalajara (1133) calls specifically for "pobladores . . . de Castilla . .
. de León . . . de Galicia . . . y mozárabes" (Μοχό 1979: 235), but it is
known that repopulation of (Castilian) areas east of Guadalajara was or-
ganized by the Aragonese king Alfonso el Batallador from 1125, and the
Fuero de Guadalajara seems to represent a response to the heavy eastern
influence in the region. Nevertheless, Aragonese and Mozarabs from Ca-
latayud were invited to settle Zorita in 1156. Mozarabs were also well rep-
resented in the later-resettled regions near Guadalajara and south of the
Tajo, with many settlements dominated by them (e.g., the town of Orgaz,
just to the southeast of Toledo; González 1975-1976: 119-218). On the
southwestern frontier, Plasencia was taken and settled by Castile (separate
from Leon from 1157-1230) in 1186. Further to the west, the frontier en-
tered Leonese Extremadura, which remained lightly populated and plagued
by warfare, although Leon continued to advance slowly. Coria, for exam-
ple, was taken in 1143, and, for a limited time, Cáceres fell under Leonese
control. At the western extreme, Portugal was recognized as an independ-
ent kingdom from the 1140s, and though periodically joined by Crusaders
from northern Europe, accomplished much of its reconquest and repopula-
tion on its own, particularly the notable conquest of Lisbon in 1147.
Further east and south of Toledo, La Alcarria and La Mancha were for a
century the scene of constant warfare, especially in the valley of the
Guadiana. Warfare, a lack of natural barriers, and an arid landscape made
repopulation of these zones difficult, and repopulation was never as suc-
cessful here as in Old Castile and interior areas of the kingdom of Toledo.
Indeed, repopulation south of the Guadiana was only secured after the bat-
tle of Las Navas de Tolosa in 1212 (Μοχό 1979: 230). Elsewhere, detailed
lawcodes, fueros extensos, were granted to make settlement here more
attractive, and the military-religious orders were given direct administra-
tive control in most frontier zones to facilitate the process. These orders
tended to operate autonomously, and their outposts were spread along the
frontier from Portugal to Aragon, where they were responsible for defense
and repopulation. The Order of Calatrava, for example, was founded in
1158 by the Cistercians of Fitero in Navarre (but with the support of the
 Social history 151

Castilian king Sancho ΠΙ), to replace the Templars at Calatrava la Vieja,

located on the southern frontier of Castile. It then extended its holdings to
Aragon, Leon and Portugal. Similarly, the Order of Santiago developed
from the earlier fratres de Càceres (in Leonese Extremadura), sponsored
by the Leonese king Fernando Π and the Archbishop of Santiago de Com-
postela. After the loss of Cáceres (retaken by the Almohads in 1174), the
Order was invited to transfer its main frontier operations to the fortress of
Uclés in Castile, to the east of Toledo. From then on, the Order was doubly
administered from Uclés in Castile and San Marcos in the city of León.
Significantly, these orders were composed of members who owed primary
allegiance not to any one kingdom or region, but rather to the order, which
existed across borders, and they actively recruited settlers from across ter-
ritories in areas controlled by them, including Cuenca (Bishko 1975: 418;
Fernández Ordóñez 2001: 454).
Much of the coastal north, including the coast of Cantabria) was not
well populated during the Burgos phase, and continuing migration to the
south necessitated efforts to restore both rural and urban populations in
inland regions. The repopulation of the interior is largely associated, as in
other regions, with the founding and development of cities. In particular,
cities were founded or repopulated along the northern coast, such as La
Corufia in Galicia, Avilés in Asturias, and Santander, Castro Urdíales,
Laredo, Santillana del Mar, and San Vicente de la Barquera in Cantabria
alone (Lapesa 1951: 206). Raymond of Burgundy also encouraged the
growth of Zamora and Valladolid at the time he was founding Ávila and
Segovia to the south. As elsewhere, this repopulation was aided by a rise in
the birthrate and the granting of fueros extensos, but other important fac-
tors were the increased security provided by the then distant frontier, the
decline of Norman raids on the northern coasts, and the growing impor-
tance of the Camino de Santiago (or camino francés).
Early in the 11th century, Sancho el Mayor had changed the Camino de
Santiago from its original coastal route to a route skirting the southern edge
of the Cantabrian Mountains. This change increased traffic, improved
commerce in the cities it traversed (such as Pamplona, Burgos and León),
and led to the founding of new cities along the way, such as Estella in Na-
varre. The importance of the route may have contributed to east-west
movement and contact in the region. Toponymie evidence indicates that
Catalans and Galicians established villages in Leon and Castile in this pe-
riod (Μοχό 1979: 266). More importantly, significant numbers of francos,
largely merchants, moved into these cities. The francos often established
152 The Toledo phase

separate quarters and received special fueros. Indeed, they appear to have
formed a significant portion or even majority of the populations of Pam-
plona and Estella. Gallic influence in the region was also enhanced as the
Cluniac movement took over many monasteries and foreign clerics came to
Spain.
The large influx of francos helped increase the relatively low urban
population of the mountain regions of Navarre. The same appears to have
occurred in sparsely populated Aragon (Lacarra 1972: 164), where notable
growth only began after 1063, when Jaca was selected as its primary city.
As the Aragonese expanded from Jaca to Huesca to Zaragoza, their lower
numbers forced them to seek settlers wherever they could, and the francos,
many of whom had participated in the taking of Zaragoza, came to make up
a significant portion of the new population (Bishko 1975: 404).103 The
Mudejars were also encouraged to remain, and Alfonso el Batallador even
prohibited them from emigrating. When Alfonso took the Jiloca and Jalón
river basins (close to the border with Castile), with the cities of Calatayud
and Daroca, he repopulated them (and neighboring areas of Castile) with
Aragonese, Navarrese, Catalans and francos. In 1126, Alfonso returned
from an expedition to Andalusia with many thousands of Mozarabs. Note-
worthy, too, is the presence of Castilians in the south of Aragon. Μοχό
(1979: 313) reports that Castilian knights were awarded villas in this area
and that the Orders of Calatrava and Santiago were also present in the
southern region. Μοχό concludes:
la extensión del Derecho de la Extremadura castellana que se efectúa en esta
época por la zona del bajo Aragón -partiendo del núcleo propulsor del
Fuero de Sepúlveda- encuentra fácil explicación en la presencia de elemen-
tos castellanos que le favorecieron. (Moxó 1979: 313).
It is clear that the Toledo phase of repopulation led to a similar kind of
demographic movement and mixing as found in the earlier Burgos phase.
However, in some areas it involved an even greater variety of groups - and
thus language varieties - than those found in the Burgos phase. In particu-
lar, the presence offrancos must be stressed, for as a group they possessed
tremendous prestige, and could therefore influence prestige norms.
Nevertheless, important differences in social organization between the
early and Toledo phases have been pointed out by historians. In the Toledo
phase, clearly marked hereditary class distinctions appear in Castile, with
an aristocracy divided between great nobility (ricos hombres) and lesser
nobility {hidalgos). Though such class distinctions often indicate a more
rigid society, their appearance during the Toledo phase can be attributed to
 Koineization and language spread 153

the opportunity to acquire new wealth as the kingdom expanded, and the
penetration of feudal ideology accompanying the arrival of the francos
(Μοχό 1979: 403). Along with feudal hierarchization, Castilian repopula-
tion became much more official in nature, with the old system of presuras
disappearing, and new settlements organized by the king, nobles, bishops,
monasteries, and military orders. Of course, such social class distinctions
would have favored the development of social and stylistic linguistic varia-
tion. Even with these changes, however, there is evidence that contrary
trends towards social leveling and mobility continued. For instance, in
order to attract new settlers, fueros promoted equality for non-nobles
within new communities; lots apportioned to settlers often were required to
be of equal size and in some fueros infanzones/hidalgos were required to
give up their special status before being allowed to settle in a new town
(Μοχό 1979: 229, 412). The fueros, of course, were associated with cities,
and a hallmark of this period is the increased level of urbanization, in
which cities were developed as magnets of colonization. But even as the
cities served as stable administrative centers, they also served as demog-
raphically heterogeneous bases for the development of a widespread net-
work of weak social ties, along which changes initiated in one area could
spread to others.

2. Koineization and language spread/dialect leveling

Migration and mixing during the Toledo phase was far more rapid and
widespread than in the earlier Burgos phase. These demographic changes
contributed not only to koineization, but also to the rapid and early spread
of multiple features which we now identify as Castilian. The spread of
Castilian features can be understood as part of a process of language spread
or dialect leveling. Significantly, these ways of viewing the spread of fea-
tures conflict with traditional views.
Menéndez Pidal described the expansion of Castilian as assuming the
form of a wedge (cuña), which fanned out towards the south from the early
county of Castile. However, it is clear from his writings that he did not
view this wedge as developing immediately with the expansion southward
of the Toledo phase. Rather, he appears to have seen it as developing its
full form only over time, largely as a result of the political domination of
Castile:
154 The Toledo phase

como en estos siglos XII y XIII la reconquista se activa semejantemente en

los otros reinos, no sólo Castilla, sino León, Portugal y Aragón propagan en-
tonces hacia el Sur sus dialectos respectivos mucho más activamente que en
las épocas anteriores y los propagan en sus formas meridionales ya algo
mezcladas por la reconquista anterior y por las repoblaciones consiguientes.
Estos dialectos de invasión rápida, propia de los siglos XII y XIII, comien-
zan en León al sur del Duero. (Menéndez Pidal 1964: 514)
From these comments it is clear that Menéndez Pidal was aware that the
differences between some dialects diminished with the initial move south-
ward. However, he clearly defends the idea that the wedge would only
appear later as dialect speakers in contiguous territories shifted to categori-
cal use of dominant Castilian (see also Chapter 6). Underlying this inter-
pretation is the image of the column, in which each of the main language
varieties spread south with political conquest, assuming the form of paral-
lel columns (as the shape of each of the political territories could be de-
scribed). Still, if these columns were only vaguely defined for Menéndez
Pidal (a position which merits some support), for others they have become
clearly defined entities. This conception is clearly represented on linguistic
maps in some manuals (e.g., Diez, Morales, and Sabin 1980: 169) and indi-
rectly in others (e.g., Lapesa 1981: 192). Zamora Vicente's comments
clearly support the existence of sharply-defined dialect borders that only
disappear with the later adoption of the official language of Castile:
La primera frontera entre el castellano y sus laterales aragonés y leonés, ha
sido o muy alejada o difuminada por la acción secular de la lengua oficial.
(Zamora Vicente 1967: 12)
However, the massive mixing that took place after the taking of Toledo
(and Zaragoza in the east) makes it difficult to argue that sharp divisions
for numerous features - to the extent that they ever existed - would have
held during the Toledo phase. Indeed, the social environment would not
have been able to support them, with norm enforcement mechanisms
weakened and exposure to variation increased. In addition, metalinguistic
conceptions of dialectal distinction were probably weak at the beginning of
the Toledo phase (Wright 1994d: 41). As discussion in the previous chap-
ter reveals, Castillans were probably among the first to begin to define
regionally-based norms and a regional conception of their speech variety,
but there is significant evidence that this did not exist for northern varieties
of Leonese, where locally-based norms predominated. Speakers certainly
would have been aware of stylistic and geographic differences, as in all
times and places, but a lack of easily categorized difference may have
 Koineization and language spread 155

made individual speakers even more permeable to innovation and encour-

aged them to engage in variant-shifting. This would have made the mainte-
nance of clearly-defined dialect divisions impossible. Indeed, it is at this
point in the process of reconquest and repopulation that it becomes clearly
obligatory to accept the widespread homogenization that Ridruejo (1995)
claims for the early stage.
Though this spread can be analyzed as dialect leveling, it is also useful
to analyze it from the perspective of the model of language spread, which
has received attention from sociologists of language. Cooper (1982: 21-
22), for example, identified certain factors that generally contribute to the
process of language spread. These include low cost, high anticipated prof-
itability, easy trialability, low complexity, high pervasiveness, and positive
attitude. During the Toledo phase of repopulation, speakers in the repopu-
lated zones faced minimal cost if they adopted Castilian-like features, since
in many cases they already had these features in their own repertoire (ex-
clusively or in variation with others) and did not need to learn any new
features, but rather eliminate some of the marked features of their native
dialects. The Burgos-phase Castilian koine would have presented less
complexity (with one notable exception; see below) than other competing
varieties, with, for example, the reduced and stabilized vowel system or the
transparent preposition + article combinations. In the new mixed popula-
tions, the consistency and frequency with which Castilian speakers pro-
duced these features, which were also variably present in the speech of
other dialect speakers, would have contributed to their pervasiveness, as
speaker-learners were more consistently exposed to these features. Adopt-
ing the speech patterns that were most frequently used in the developing
speech community would clearly have been socially profitable, and speak-
ers would have been free to use these features without great fear of censure
(itrialability), since norm enforcement mechanisms were weakened. In ad-
dition, speakers' attitudes may have ranged from positive (prestige of
speech associated with victorious Castile and the city of Toledo) to, more
likely, at least non-negative, since weak norm enforcement mechanisms
would have relaxed any pressure to maintain features from other, possibly
more complex, dialects.
A particular example of the effects of this process may be found in the
12th- and 13th-century language of Salamanca, a major city repopulated
with Ávila and Segovia, but farther to the west within the kingdom of
Leon. The early language of Salamanca is often considered to have been a
simple southward extension of Leonese, and it is common opinion that the
156 The Toledo phase

speech of the city was not Castilianized until the 14th century. However,
several factors might make one suspect this simple characterization. First,
at the time of reconquest, Leon and Castile were under one crown, which
may have weakened - if only temporarily - perception of collective differ-
ences between Leonese and Castilian. Toledo itself had fallen under Castil-
lan hegemony, so, to the extent that prestige was a factor, it would have
favored the spread of Castilian features. We also know that the city of
Salamanca was settled by people of greatly varied background (see above),
so it cannot be argued that colonization by a homogeneous group led to a
simple transplantation of Leonese. Finally, Leonese, even in northern
Leon, seems to have always shown a great degree of geographic variation,
and was apparently less focused as a regional variety than Castilian (i.e.,
norms were often locally restricted). It is impossible that political factors
alone could have supported the simple southward extension of these varie-
ties (to the extent that such variation could be transferred to a new zone).
The only extant collection of early Salamanca documents is that tran-
scribed by Martín Martín, Villar García, Marcos Rodríguez, and Sánchez
Rodríguez (1977). This collection contains 113 documents from the 12th
century and 363 from the 13th, the majority of which were prepared in
Salamanca. The editors point out that most of the texts are in Latin or are
very Latinate, but that in texts prepared by local notaries and scribes, the
Romance (or more phonographic) element appears in ever growing propor-
tions (starting with words and phrases) until the 13th century, when whole
sections of some texts are clearly in revised (Romance) spelling (1977: 42).
The first entirely phonographic text is from 1240 (Doc. 199). Since the
editors' primary intent is to make texts available for historians (1977: 78),
the transcriptions are not paléographie. For example, filled-out abbrevia-
tions are not indicated in the transcriptions, and the collection is thus not as
useful for linguistic study as might otherwise have been the case. Neverthe-
less, the texts, such as they are, do reveal significant mixing of Leonese
and Castilian elements. One finds, for example, either the clearly Latinate
spelling pectet or the Castilian (and eastern Leonese) spelling peche (Docs.
144-1222, 201-1242, 211-1245, 222-1247, etc.) with no examples of the
Leonese forms with -it-, the same is true of factu/fecho. Little effort is
made in the texts to represent diphthongs until the 13th century, but when
they do appear, they are the entirely stable and Castilian ie and we: castielo
(Doc. 174-1225), Azogue Viello (Doc. 193-1236), sueldos (Doc. 195-
1237). From early on the preposition and article combinations show fluc-
tuation between the transparent combinations typical of Castilian and the
 Koineization and language spread 157

more conservative contractions: en la zoc velo, en la aldea (80-1181), en

la Zoc Veio (93-1187), en lo vostro poder, in lo mes (177-1227), in las dos
partes (192-1235), en la Azogue Viello (193-1236), con los parientes
(199-1240), en el vostro (222-1247), con los (199-1240), con la (204-
1242), but enno castro (171-1225), enna carera (172-1225), ena collation
(179-1228, 180-1228), enna se (181-1229), enno poderío (183-1229),
ena, conna (199-1240). Normal apocope is regular in these texts, though
there is also a preference for the forms heredade and edade, with final -e
maintained as in western varieties. In these texts up to the mid-13th cen-
tury, I have found only one example of that most characteristic feature of
Leonese: syllable-final consonant weakening to a lateral: duldanza
(246:1253, where the lateral substitutes b).
Still, there are two possible exceptions to these patterns. The Castilian
/·$/ ( < CL, LI) was not adopted, at least in the orthography: zoc velo (80-
1181), Azoe Veio (89-1185, where -i- probably represents [j] since we also
find Clemens lalfaiat in the same document, which presumably represents
the Arabism alfayate 'tailor'), Carvalosa, Carvaiosa for Carbajosa (79-
1181 and 68b-1177), and numerous later examples of filio, fiyo, muler,
conceyo, conzello. In addition, count/mass pronoun distinctions and leísmo,
which became so typical of Castilian were not established in Salamanca.
The rejection of Castilian / j / may be only apparent (i.e., a failure to use a
distinct grapheme), but it may be that it arose only later in Castile, and/or
became a salient marker of developing Castilian identity (particularly in
the later 12th century when Leonese and Castillans were in frequent con-
flict). The lack of leísmo and count/mass distinctions is first and foremost
attributable to lack of a Castilian majority in the original population of
Salamanca, but its developing salience as a feature of Castilian identity,
may have made it socially difficult for speakers to adopt this feature. In
either case, leísmo and count/mass distinctions are among the few features
of Castilian which are more complex and marked than equivalent structures
in Leonese, and they therefore would have been difficult to learn even if
speakers had wanted to learn them. It is interesting to note that today, it is
precisely Salamanca that shows the fewest Leonese features, even in rural
dialects, while zones to the north and south show stronger retention of such
features. Northwestern Extremadura, of course, was reconquered and re-
populated later in the 12th and 13th centuries when Leon and Castile were
ruled by different monarchs and in frequent conflict. Southern pockets in
this zone may represent transplants of Leonese by relocation of more ho-
mogeneous groups than those which settled Salamanca.
158 The Toledo phase

Similar arguments can be made for Aragonese south of the Pyrenees. As

is the case with Leonese, there are few texts from outside the northernmost
areas - Pyrenean areas of Aragon in this case - that are considered as rep-
resentative of the so-called "pure" dialect. Indeed, Frago Gracia (1991a:
120) has argued that the Aragonese of the Ebro Valley was never simply
transplanted to the Ebro valley, where a simplified and mixed language
variety - closer in appearance to Castilian - developed, primarily as a re-
sult of dialect mixture, but also as a result of simultaneous dialect contact
and spread. Significantly, the Toledo phase marks a period of simultaneous
demographic mixing for Castile, Leon, and Aragon. In the context of this
discussion, it is easy to envision the increase in weak social ties that oc-
curred and the concomitant increase in permeability to innovations that had
already been consolidated as changes in the center of the Peninsula.
Language spread or dialect leveling of this sort was certainly facilitated
by koineization at the same time that the spread to new speakers increased
the possibility of koineization. Still, the spread of features that we now
associate with the developing Castilian koine was not limited to the south-
ern or lateral areas that were suddenly opened to repopulation. There are
two reasons for this. First, as was discussed above, there began a massive
program of internal repopulation in the northern areas, focused on though
not limited to urban centers. In these areas, traditional norms would have
been weakened as social networks were weakened, though quite clearly not
to the same degree as in the south, since to this day these northern regions
show the greatest conservatism. Still, these areas would have become more
permeable to changes that were first consolidated in the newly repopulated
zone through the process of spatial diffusion of features.
Spatial diffusion has generally been understood as occurring as a kind
of wave that spreads geographically. This may be a valid model for the
largely rural and agricultural societies of the very early Middle Ages (even
early Castile), in which most inhabitants lived in small villages separated
from other villages by short distances. In such societies, innovations
probably do spread from village to village. However, modern sociolinguis-
tic research has made clear that once larger urban centers appear, particular
features do not spread in this way (though historically they may still appear
to do so). Rather, features seem to jump from urban center to urban center,
and thence to progressively smaller urban centers before slowly spreading
into intervening rural areas (Chambers and Trudgill 1980: 189). These
authors cite as a particularly stunning example the spread of uvular [R]
from France to urban centers in Germany and Scandinavia, across national
 Linguistic changes 159

and linguistic borders (Chambers and Trudgill 1980: 189-191). This pat-
tern can also be seen in the spread of front-vowel raising (e.g., bag /baeg/ >
[beg]) from Chicago into outlying areas, where some urban centers were
participating in the change, but intermediate rural areas remained unaf-
fected (Callary 1975). The Milroys (1985; Milroy 1992: 197) have ex-
plained that the impact of geography or urbanity is indirect, and that the
real factor underlying this parachute effect is the conglomeration of weak
social ties between urban centers that allow the spread of innovations.
In 12th- and 13th-century Iberia, the same kind of mechanisms must
have been present. Koineization was certainly taking place in the center-
south around Toledo, in a variety of northern areas (though to a lesser de-
gree), and in reconquered Zaragoza, but, unlike cases in distant colonies,
contact was maintained with the contributing communities and their lin-
guistic varieties. In particular, we might expect changes to spread back
from southern regions (or cities) into northern areas (or cities), even if
these areas saw a lower degree of social mixing. The Camino de Santiago
must also have promoted the development of weak social ties and thus
served as a conduit for innovations across northern Castile and neighboring
areas. These considerations must be kept in mind when interpreting the
documentary evidence.

3. Linguistic changes

We now examine the evidence for three Toledo-phase changes in Castilian

and, to some degree, other dialects: apocope (reduction of variants in nor-
mal apocope, mixing with stylistic functional reallocation in extreme
apocope), early leísmo (mixing showing the survival of interlanguage or
interdialect features with distinct solutions for different areas, according to
different population mixes), and reorganization of the possessive system,
through social and functional reallocation and simplification. Though some
of these changes are sporadically attested in Burgos-phase Castilian, and
early leísmo certainly originated in the Burgos phase, I argue that each of
them was consolidated or significantly altered as an effect of Toledo-phase
koineization.
With regard to evidence, this period is in some ways easier and in some
ways more difficult to investigate than the Burgos phase. Available docu-
mentation for this period is more abundant, but no comprehensive study
such as Orígenes has been prepared for it, in large part because it has not
160 The Toledo phase

generally been identified as linguistically significant. Moreover, in the 12th

century most documentation tends to be highly Latinate, and even less
useful as evidence following the reforms of the Council of Burgos in 1086
(Menéndez Pidal 1964: 460; Wright 1982), which helped to reinforce pre-
scriptive norms of writing. Romance only began to be written phono-
graphically with some regularity in the 13th century, so there is little direct
evidence available for the Romance of the 12th century. Of course, it is the
12th century that is of greatest interest in this chapter, since it is during this
period that the linguistic changes resulting from the Toledo-phase repopu-
lation were being completed. Nevertheless, if we assume that the early 13th
century documents from the regions around Toledo preserve the results of
the koineizing process that must have occurred during the 12th century, we
can, in most cases, deduce from this later evidence changes that must have
occurred during the preceding century. Unfortunately, even today relatively
few texts can be accurately located both chronologically and geographi-
cally. As a result, the richest source of evidence for this period continues to
be the texts assembled by Menéndez Pidal in his Documentos lingüísticos
de España. In addition, researchers have closely analyzed many of the first
texts written in Castilian (often literary or historical works), and their find-
ings will be referred to where appropriate.

3.1. Leveling and reallocation in the development of apocope

Apocope in medieval Spanish has been most thoroughly analyzed by

Lapesa (1951, 1975, 1982). He divides cases of apocope into two catego-
ries: normal and extreme. Normal apocope refers to the permanent loss of
word-final /-e/ after the phonemes represented with the letters r, s, I, n, z, d
(Lapesa 1975: 13). Lapesa also uses the term to refer to the occasional loss
of final /-o/ (or /-u/?) in enclitic or proclitic position, if it left r, s, I, n, z, d
in final position (e.g., Martino Gonçalvez > Martín Gonçalvez, solo > sol),
whether or not they are maintained in this form to the present day. Extreme
apocope refers to the (what historically seems) "temporary" phenomenon
of loss of/-e/ after other single consonants (e.g., nuef [< nueve], quiçab,
dix, noch, axac) and consonant+dental consonant clusters (e.g., adelant,
part, fizist, ond, estonz). It also refers to loss of final /-o/ and even some-
times /-a/ in forms used as clitics after the same consonants and clusters
that define extreme apocope of /-e/ (e.g., todo > tot, Lobo > Lop, Diago >
Diag, Diac, Fernando > Fernand, Fernant, como > com). Lapesa consid-
 Linguistic changes 161

ered extreme apocope to be a phenomenon that occurred intensely from the

end of the 11th century to the end of the 13 th, and, in some areas, up to the
mid-14th century. Both normal and extreme apocope appear to have been
affected by Toledo-phase rekoineization, with this period marking the con-
solidation of normal apocope (or, better, the rapid reduction of variation
through the stabilization of phonological forms) and the functional reallo-
cation of extreme apocope.
Normal apocope - as the name might suggest - has received far less at-
tention than extreme apocope. It is true that the loss of final /-e/ following
many of the apical consonants became a widespread and regular phenome-
non in western Romance (though in Galician-Portuguese it failed to occur
after /d/). Menéndez Pidal (1964: 186-187) reports it was probably an an-
cient phenomenon, and provides the example of a 6th-century coin bearing
the inscription LEONES MONETA. Nevertheless, he points out that examples
of normal apocope are infrequent even in 10th-century texts, where the few
examples he finds occur after n, r, s, z. In the Glosas emilianenses and
Glosas silenses the final vowel was regularly conserved: muliere, stier-
core, promissione, meretrize, uece, uoluntate, but clear examples of
apocope of /-e/ are found in the future form following s in alongarsan (<
alongar se han), and following liquids in tal, quai, leuator (Pidal claims
that the last of these was influenced by the nominative form being glossed;
the others are words whose high frequency would have favored their reduc-
tion). Pidal concludes from this that "en la segunda mitad del siglo X la
apòcope tenía muy escasa cabida en la lengua común" (1964: 186). If one
accepts the later dating now assigned to the Glosas by most scholars (e.g.,
Wright 1994e: 209; Díaz y Díaz 1978; Bezler 1991), the implications of
this evidence would seem to grow in significance, for they would indicate a
very rapid change in use of final /-e/. It seems more likely, however, that
early full forms hide alternative manifestations that occurred in speech,
where these forms regularly suffered phonetic reduction (in a kind of
phrasal syncope), but without the phonological reanalysis that underlies
adaptations in phonographic spellings. Pidal reports that the process of
normal apocope grows stronger in written texts during the later 11th cen-
tury, when the fluctuating loss or retention of final -e after apicals occurs
alongside increasingly frequent manifestations of extreme apocope and
syncope. Such forms indicate that phonetic reduction of /-e/ had become so
frequent that it was beginning to lead to phonological reanalysis. Such
reanalysis would be consolidated during the Toledo phase (although, as we
162 The Toledo phase

will see in the discussion of leísmo, at least one very high frequency form
suffered full reanalysis during the Burgos phase).
Lapesa links the reasons for the development of normal apocope to
those he gives for extreme apocope (see below):
En el tránsito del latín al romance español del siglo X, la simplificación, pa-
latalización o asibilación de dobles consonantes y grupos consonánticos la-
tinos había reducido mucho el número de fonemas capaces de cerrar sílaba
interior. Así habían desaparecido las consonantes implosivas en SEPTEM >
sete o siete-, FACTU > faito o feito; LAKSARE > laisare o lesare; PUGNARE >
puñare; CUPPA > copa-, SAGITTA > saeta; BUCCA > boca. Los únicos fone-
mas consonánticos que podían tener función implosiva eran r, l, s y m/n.
Precisamente fue tras estas consonantes donde se inició la apócope de la e
latina; recuérdense los ejemplos ya citados qual, tal, segar, aliatoti, o el as-
toricés de una escritura de 1030. (Lapesa 1951: 192-193)
At the phrase level the originally phonetic phenomena of syncope and
apocope would have had similar effects. In addition, speakers were under
less paradigmatic pressure to maintain final /-e/, since, unlike final /-o/ (or
/-a/, to the extent that it was affected), it was not a gender-marking mor-
pheme.
Such arguments can explain the origin of the variation between full and
apocopated forms in many Romance varieties, but do not explain the chro-
nology of the development in any particular variety. Normal apocope - in
its different manifestations - may have developed and been consolidated in
different areas at different times and at different rates. When did this hap-
pen in Castilian? Lapesa (1951: 204-206) seems to view the mostly regular
use of final -e after r, s, I, n, z, d as continuing in Castilian only up to the
12th century, after which its appearance must be ascribed to particular
causes. He states:
En Castilla, aparte de las rimas en la poesía épica, era muy rara ya [12th
century]. No faltan casos sueltos, ora debidos a procedencia dialectal de los
notarios, ora a un apego individual por gustos viejos. (Lapesa 1951: 205)
Lapesa mentions specific examples of retention of final -e in the Fuero de
Valfermoso de las Monjas (1189) and Fuero de Madrid (1202), where they
appear mixed with examples of apocope: andadore, menestrare, piede,
Madride, but segur, sayal, fiador, Maydrit (1951: 205). Nevertheless, these
are both documents that intentionally Latinize many parts of the text, and
these particular forms may simply show that Latinizing influence. Lapesa
also gives examples of retention of final -e in the Fuero de Avilés (particu-
 Linguistic changes 163

larly after d, as in potestade, lide), but ascribes them to conservative As-

turian influence (today the conservation of /-e/ after /-d/ occurs only in
western Asturias and Galicia). Likewise, the Mozarabs of Toledo adopted
use of final -e - with "coherente unanimidad"- during the 12th century,
perhaps in an overt conservative reaction against the dominant trends,
which may have served to mark a separate identity within the broader
koineizing community (Lapesa 1951: 192).
Perhaps the most significant evidence of the 12th century shift in status
of normal apocope is the disappearance from the texts of a certain kind of
hypercorrection:
Ultracorrecciones como matode por matod 'mató', kede por ked ( < QUID),
bédene por veden (evident), stane < stant, abundan hasta principios del si-
glo XII. (Lapesa 1951: 188)104
The disappearance of such hypercorrections would appear to indicate the
general loss of alternating full and apocopated forms associated with a
fundamentally phonetic process of vowel deletion. The end of confusion on
the part of scribes indicates the end of the variation produced by the pho-
netic process of normal apocope, and it reveals speakers' definitive phono-
logical reanalysis of these forms as lacking underlying /-e/.
Álvarez Rodriguez (1996), in a study of normal apocope, confirms that
it began early after the single consonant phonemes /r, z, 1, n, d7, but argues
that it was only extended later to /-d/.105 He adduces as evidence a docu-
ment from Zamora (dated 1050) that contains within 15 lines numerous
tokens of normal apocope (e.g., puçal, aliatoti, pergaminar, sal, segar,
caral, segur), but none after -d. This is, of course, a Leonese document, but
Lapesa agrees that the same was true of Burgos-phase Castilian (1951:
193): "En la época primitiva no se registran ejemplos de e apocopada tras
d; y tras ζ, sólo en patronímicos, donde casi siempre es dudoso que la e
fuese etimológica." (in other words, -e in patronyms was a result of hyper-
correction). Thus, in the 11th century d maintains some resistance to
apocope, but the situation changes abruptly in the 12th:
es razonable pensar que la apocope de la /e/ después de /r, z, 1, n, d, dz/ a-
rraigó profundamente en la mayor parte de la población durante los siglos
XII y XIII. (Alvarez Rodríguez 1996: 40)

The rapid consolidation of normal apocope following all voiced apicals

was apparently favored by the process of Toledo-phase koineization. In the
late 11th century the phenomenon had become more frequent throughout
northern Iberia (and, according to Lapesa [1951: 190], was probably not
164 The Toledo phase

absent from Mozarabic speech, despite the later reaction against it).106 The
mixing of different dialect speakers in Toledo-phase repopulation certainly
would have favored their accommodating out of the less frequently occur-
ring forms (in this case, those that maintained the final vowel). Older child
and adolescent speaker-learners could then reanalyze the more frequently-
occurring apocopated forms as what they seemed to be: transparent realiza-
tions of phonological forms lacking final /-e/. This is a process that proba-
bly would have occurred slowly with or without koineization, but in the
koineizing environment, it occurred much more rapidly and was thus in-
corporated into the new speech norms of the Toledo phase, particularly as
they spread north along the Toledo-Burgos axis, but also west into Leon.107
In effect, koineization was not critical to the origin of the process itself, but
it very likely did contribute to its sudden completion in the early 12th cen-
tury.
The development of extreme apocope is more difficult to understand,
though koineization was critical to its constitution as a Toledo-phase norm.
In some respects, such as the chronology, its development parallels that of
normal apocope. Menéndez Pidal (1964: 190-191) gives a few examples of
loss or hypercorrect use of -o, -a, and -e in proclitic forms from the 10th
century: don Paulo apostolo vs. dueno Cristo ( < duerno or dueño, with
the nasal possibly articulated as a conservative geminate or as a palatal;
examples from the Glosas Emilianenses)·, Seemen Didaci (< Ximeno), Fre-
denande Albarez (< Ferdinando), Bal carcere (examples from eastern
Leon); Lope Garsea (< Lop < Lobo < LUPU), Fortuni Sancio (< Fortunio),
ual de ripa Hibre (< valle) (examples from Castile). These kinds of exam-
ples - showing loss or confusion - become more frequent from the second
half of the 11th century, though alternation with full forms continues: Nun
Sanciz, Fernán Gonsalbez, DuenfaJ Elo, font de caballos (from a docu-
ment of 1009), allend presa / allende parte del aqua (1084), quern quadra
(1091), (examples from Leon); Fortun Sanz, Lop Garcez, duanne Ezo (<
DOMINU), al mont Sancì Mikael (1097), Ajerb (1118, from Aragon), Nun
Albariz, Fortun Sangiz, ual de uascones, la sierra adelant (1057), duos
uasos de argent (1082), sachestme (1107) (examples from Castile). For
word-final -o, Menéndez Pidal (1964: 173) suggested that the apocopated
forms spread analogically from the frequent proclitic use to stressed forms,
as in de comité don Lop et sua uxore (1107). Apocope of enclitic pronouns
also appears more frequently in 11th century texts: lexol, levos, quem
quadra, quet dare (Lapesa 1951: 194). It is worth remembering, however,
the apparent increases in frequency of apocope between the 10th and 11th
 Linguistic changes 165

centuries may simply be due to the greater number of texts retained from
the 11th century.
The causes of extreme apocope have long been the topic of debate.
Lapesa (1951: 193) argued that there was a direct causal relationship be-
tween syncope and apocope. Continuing the arguments he applied to the
case of normal apocope, in which he suggested that normal apocope could
occur because the consonants that it occurred after were already acceptable
in syllable-final position, he suggests that the results of syncope opened the
door to apocope by regularizing the appearance of many new consonants
and consonant clusters in syllable-final position:
La caída de las vocales intertónicas convirtió en implosivas muchas conso-
nantes que antes precedían a la vocal interior elidida. Los primeros ejemplos
se dan ya en el siglo X (limde, año 934; Fretnando, 940; semdero, 964;
Mamlas, 972; trepde, 984); pero la síncopa no logra afirmarse hasta fines
del siglo χι. Entonces quedaron como nuevos fonemas aptos para función
implosiva la ζ (plazdo, aztor); la d (judgar, cadnadó)·, la ρ (trepde, riepto);
la b (cobdo, cibdad)·, la m {semdero, comde); la t (setmana); y hasta grupos
poco estables, como los de comptar, Antfossus, *antnado, *vendgar,
sangne. (Lapesa 1951: 193)

There are problems with this argument, however. Harris-Northall (1990)

argues cogently that syncope was a process that began in Latin in a limited
number of phonetic contexts (e.g., between /s/ and /t/, as in POSITU >
POSTU) and which gradually affected other phonetic contexts (and vowels)
over the course of centuries. While it is likely that syncope was still operat-
ing in some of these contexts at the time that apocope began, it is impossi-
ble to argue that it led to a sudden change in the phonotactics of the lan-
guage. This is evident in Lapesa's own examples. The starred forms are not
attested and cannot serve as evidence of new phonotactic norms. These
items are only attested in forms - like their modern reflexes alnado, vengar
- which show adaptation of the novel consonant clusters produced by syn-
cope to the existing phonotactic constraints of the language. The other
cases were highly unstable - probably representing sporadic speaker inno-
vations - and they were replaced by phonotactically more acceptable alter-
natives: contar for comptar, sangre for sangne, candado for cadnado,
sendero for semdero. While it is true that some of these forms lasted for
centuries (e.g., cobdo, cibdad), these may have been the last affected by
syncope and they too were reformed over time (Harris-Northall 1990a:
153). As Harris-Northall (1990a: 152) argues, syncope did challenge the
166 The Toledo phase

phonotactic constraints of the language, but most of its products were

modified in the end.
On the other hand, syncope and apocope cannot be seen as unrelated:
they are both processes that represent a normal effect of fast or relaxed
speech. In the case of syncope, the output of speakers (or its effects, accu-
mulated over centuries) was reanalyzed by learners without the original
underlying vowel, but also with consonant sequences conforming to the
phonotactic constraints - or norms - of the language. They could do this
because the same speakers who produced the syncopated forms also articu-
lated a much larger number of forms that continued to reflect the phonotac-
tic constraints of the language. As a result, the phonotactic norms of the
community were not greatly altered by syncope (except perhaps for the
voiced apicals affected by normal apocope), but the results of syncope
were greatly altered by the phonotactics. Like syncope (and normal
apocope), extreme apocope in Castilian began as a phonetic process. In-
deed, this is argued for by Lapesa:
La fonética sintáctica contribuiría a este desbordamiento de la apocope,
omitiendo como intertónicas las vocales finales de palabras agrupadas en
torno de un solo acento o constitutivas de una sola unidad significante. La
proclisis favoreció, sin duda, la pérdida de la e en el topònimo «Font de Ca-
ballos» (1009, León, Arch. Episc., 84), y en el «allend presa» que otro di-
ploma leonés de 1084 emplea en alternancia con «allende parte». Los pro-
nombres enclíticos apoyados en palabras con terminación vocálica
empezaron a perder su e (feto/, 1055; leuós, 1100). (Lapesa 1951: 193-194)

But in Castilian, what distinguishes extreme apocope from syncope is the

degree to which extreme apocope was able to alter the phonotactic norms
of the community - at least for a time - in a way that syncope did not. So
syncope cannot be seen as a cause of extreme apocope, but it is a similar
kind of process that disturbed the phonotactic norms of the community by
permitting phonetic innovations in the output of many speakers. Only to
the extent that it weakened norms (with respect to syllable-final conso-
nants) can it be seen as a cause of apocope.108 However, as Harris-
Northall's arguments make clear, it is by no means certain in what phonetic
contexts syncope was still operating in the 11th century, and to show a
direct causal link one would have to show that all the consonants affected
by extreme apocope were also affected by syncope and preserved in final
position. In at least one case this was clearly not so: extreme apocope is
frequent after the cluster -nt, but the same cluster is exceedingly rare word-
internally (see quotation from Lapesa below).
 Linguistic changes 167

It seems likely that at the earliest stage (corresponding to what I have

called the Burgos phase, though examples of early extreme apocope were
not geographically limited to the immediate region of Burgos), phonetic
factors were essential in allowing and determining the appearance of
apocopated forms. However, in many studies phonetic factors have been
seen not only as contributing to the origins of extreme apocope, but also of
continuing to govern its use in the later periods of the 12th and 13th centu-
ries. In this koineizing environment, we might have expected either a) full
maintenance of regular phonetic constraints, b) the selection and stabiliza-
tion of apocopated forms, as in normal apocope, or c) the selection and
stabilization of full forms of words that were prone to extreme apocope,
(particularly given the existing phonotactic constraints that remained so
strong in cases of syncope, and the lower frequency of extreme apocope).
What we observe in the documentation of the 12th and 13th centuries -
a period of some 200 years - is the very frequent appearance of forms
showing extreme apocope as well as those that do not, with some variation
in the relative frequency of apocopated forms from text to text. However,
this phenomenon has escaped convincing description and explanation.
Most scholars have seen the continued use of apocope as governed princi-
pally by phonetic factors. Catalán argued, along the lines of Lapesa before
him, that the development of extreme apocope - from the end of the 11th
century to the 13th (Catalán 1971: 81) - was simply the extension and
regularization of the closed syllable structure that he saw as dominating at
the time, concluding:
No veo razón alguna para seguir considerando a la 'apócope extrema' como
extraña al genio del idioma. De otra parte, la geografía del fenómeno fa-
vorece el carácter autóctono que creemos que hay que conceder a la
apócope, pues su intensidad disminuye gradualmente según pasamos del
catalán al aragonés, del aragonés al castellano, del castellano al leonés y del
leonés al gallego-portugués. (Catalán 1971: 79)

While both his assertions - that extreme apocope was "native" to Castilian
and that there is a gradual decrease in intensity from east to west - are cer-
tainly true, Catalán oversimplifies in two ways. First, like Lapesa he as-
sumes that syncope and apocope are one and the same. Second, he incor-
rectly equates the early phonetic conditions that allowed the rise of
apocope with the conditions that determined its later use in the 12th and
13th centuries. Surprisingly perhaps (given the above discussion), Lapesa
(1975: 17) responded to Catalán by pointing out that there were important
168 The Toledo phase

mismatches {"desajustes") between word-internal syllable-final consonants

and word-final consonants left by extreme apocope:
No todos los fonemas y grupos registrados como finales de palabra a causa
de la apócope extrema se encuentran como finales de sílaba interior ni en
conglomerados fonéticos-sintácticos de palabras. Así ocurre con la Iti... la
/g/ o /k/. . . / n s / . . . M . . . En otros casos la correspondencia es excepcional
o escasa: la /-f/ . . . y la / § / . . . en los pocos frecuentes frexno, trexnar. La
cronología de los fonemas y grupos implosivos es distinta según estén en in-
terior o en final de palabra. El grupo /-nt/ + consonante (antparar,
*antnadó) se resolvió muy pronto (amparar, atinado, alnado) en posición
interior, mientras que en final de palabra perduró largamente, con ejemplos
hasta muy avanzado el siglo xiv. El paso de *astmar (<AESTIMARE) a asmar
y de *mastcar (< MASTICARE) a mascar fue tan rápido que no dejó -que se-
pamos- testimonio escrito de su /-st-/ implosivo; en cambio est, huest, to-
mest, dixist aparecen profusamente hasta 1280 y con menos frecuencia más
tarde. En sentido inverso: riepto, reptar continuaban usándose en el siglo
XV, pero princep y Lop habían sucumbido mucho antes ante principe y Lope.
La Ibi y NI implosivas de recabdar, cobdo, dubda, cibdad tenían mucho
vigor en tiempo de Juan II y de los Reyes Católicos. (Lapesa 1975: 17-18)

Lapesa, as we will see, began to look elsewhere for the factors that gov-
erned the appearance of extreme apocope.
Nevertheless, other attempts at analysis of later extreme apocope have
also assumed that it is fundamentally a phonetically-constrained phenome-
non. Allen (1976) defined phonological rules for the appearance of ex-
treme apocope in 12th- and 13th-century texts. Extending Lapesa's ideas
on the importance of stress and syncope (in the origin of apocope), Allen
attempted to define the effects of phrasal stress on the appearance of
apocope in noun phrases, verb phrases and prepositional phrases; so, in
certain sandhi environments: 1) an unstressed vowel falls when it is word-
final in a word which is proclitic to the principal word in the phrase, or to
that word and one or more proclitic words; 2) an unstressed vowel falls
when it is word-final in the principal word or in a word enclitic to it, and is
phrase-final before pause (Allen 1976: 17). Though these rules are valid
for many cases of apocope, they are not consistently so, and Allen's own
conclusions show a certain frustration with the results of his analysis:
Needless to say, apocope does not occur at every occurrence of these envi-
ronments; it is a sporadic phenomenon which sometimes occurs where we
expect it to, and sometimes fails to occur where we expect it. It does occur
with great frequency in these two environments, however, and I have not
 Linguistic changes 169

found it in other environments . . . It is my contention that, when apocope is

narrowly defined, as here, it will be seen that stress plays a crucial role, pos-
sibly the only role, in this phenomenon. (Allen 1976: 17)

But, how can this be so? Obviously, Allen must be referring to the purely
linguistic constraints on apocope (which might coincide with the phonetic
contexts of origin, as described by Lapesa). But his own comments make
clear that this is not an entirely phonologically-determined phenomenon. If
it were, the rules would work in all instances (or, at least, in the great ma-
jority of instances).109
The inability of phonological rules to fully describe or predict the oc-
currence of extreme apocope (of the 12th and 13th centuries) has been
shown by Harris-Northall (1991) in a statistical analysis of the variants of
(a)delant(e) in eight large texts of the Alfonsine corpus.110 His initial hy-
pothesis (somewhat different from Allen's) was that, at this late stage in
the use of apocope, variation in usage would be determined by a phonetic
factor: the presence or absence of a following vowel.111 In this view,
apocopated forms should appear before vowels, full forms before conso-
nants and pauses. Indeed, this is exactly what happens in the case of mucho
and muy/muyt/much, where the apocopated form appears almost exclu-
sively in pre-vocalic contexts (Harris-Northall 1993: 189). However, the
results of the study do not confirm this hypothesis for -e. The following
commentary on the results obtained from the first and fourth parts of the
General Estoria (labeled as GEI and GE4) is illustrative:
In GEI, the tokens of (a)delant followed by a vowel or & total 125, those of
(a)delante in the same context, only 51; before a consonant (a)delant ap-
pears 173 times, adelante 114. There is, therefore, a clear preference for the
apocopated form, even when a consonant immediately follows. In GE4, the
preference for the apocopated form persists even more clearly: delant ap-
pears 25 times before a vowel or &, delante only once; delant 26 times be-
fore a consonant, delante only nine. Adelante is found only three times in
the whole text, while there are over 200 tokens of adelant ~ adeland. Gen-
erally speaking,... it would be true to say . . . that there is a certain prefer-
ence for the apocopated form, restricted on occasions by the presence of a
following consonant. (Harris-Northall 1991: 35)112

Clearly then, phonetic context was not determining the appearance of these
forms (though, as Harris-Northall suggests, it seems to have had some con-
straining effect). Some other factor must have been at work. Of course, one
could view such variation as free variation, but it has become increasingly
clear that true free variation does not generally exist: some factor - Unguis-
170 The Toledo phase

tic or social - is generally implicated in the appearance of nearly any lin-

guistic variant. The relative stability of extreme apocope as a variable phe-
nomenon during two centuries indicates that social factors must have been
involved. This idea is partly suggested by Harris-Northall (1991: 37, 1992:
189-90), who suggests that the scribes, following a "sociolinguistic bias",
perhaps exaggerated their use of apocope in written texts (but that in
speech their use of extreme apocope may have been more constrained by
the pre-vocalic context).113
Dissatisfaction with purely internal explanations also led Lapesa to look
for an external cause for what he perceived as the increase in extreme
apocope in the 12th century: "la intensificación del fenómeno en el español
arcaico ha de atribuirse principalmente a un nuevo factor: la influencia
francesa y provenzal" (Lapesa 1951: 195). Lapesa points to multiple mani-
festations of the sudden rise in French influence in the late 11th century:
the re-routing of the Camino de Santiago through Burgos and other north-
ern cities, the establishment of colonies of francos in these cities, and the
influx of Gallic prelates, nobles and merchants, particularly during and
following the reign of Alfonso VI (1065-1109). This, then, is for Lapesa
the primary cause of the rise of extreme apocope:
En los setenta años que transcurrieron desde que Alfonso VI comenzó a re-
inar hasta que apareció el Cantar de Mío Cid hacia 1140, el influjo de la
inmigración franca había actuado sobre los finales de palabra españoles mu-
cho más poderosamente que los cuatrocientos años de contacto con los ára-
bes, y más también que las incesantes oleadas de mozárabes venidos al norte
durante los siglos IX al XI. (Lapesa 1951: 197)

To explain why Castillans would be open to such imitation or borrowing,

Lapesa dedicates some four pages to discussing demographic movements
of the Toledo phase and their disruptive effects on society - much in the
manner that I employ here. Indeed, his awareness of the particular perme-
ability of the center to innovations leads him to focus on an issue that
might otherwise not have received attention:
Habría sido de esperar que las regiones incorporadas a Castilla desde fines
del siglo XI, pobladas por gentes de tan diversa procedencia, fuesen las que
hubieran mostrado mayor predilección por la apócope y que las zonas del
Norte, con población hispana arraigada de antiguo y ligada a viejas tradi-
ciones locales, se hubiesen mantenido más refactarias a la innovación. Pero
no fue así. (Lapesa 1951: 204)
 Linguistic changes 171

Lapesa goes on to point out that francos colonized cities and regions all
along the north coast and the Camino de Santiago, which would explain a
strong French influence on extreme apocope in the north. As far as explain-
ing French influence on the origin of extreme apocope as a prestige norm
of Castilian speech, Lapesa's arguments can hardly be improved upon,
although it is worth pointing out that there was significant population
movement in the north as well during the 12th century, and this would have
contributed to a general weakening of resistance to the novelty that non-
phonetically-determined apocope represented. In addition, such weakening
of social networks would have made these northern communities more
permeable to new changes arising in the south. If the population had re-
mained stable, the constant presence offrancos might simply have had no
effect at all.
French (and probably eastern Iberian) influence was undoubtedly im-
portant to turning extreme apocope into something different from a primar-
ily phonetic phenomenon. However, it is clear from Lapesa's arguments
that he views extreme apocope as a kind of direct borrowing on the part of
Castilian speakers. For him, extreme apocope was a direct and continuing
reflection of Gallic influence; speakers were always conscious of imitating
French models when using apocopated forms, and they only ceased using
them when Gallic influence waned. Lapesa stressed again and again (1951,
1975, 1982) the importance of this direct French influence on Castilian
speakers. In the late 11th and early 12th centuries, there can be little doubt
about the importance of this influence. However, it is quite another thing to
view the apocope of the 13th century as being the result of direct imitation
of a French norm (in opposition, apparently, to a "native", non-apocopating
norm). The major problem with this notion is the lack of chronological
correspondence between the decline of French social influence and the
decline of apocope. Surprisingly, Lapesa himself argues that French influ-
ence began to wane gradually even in the 12th century:
la corte de Alfonso VIII parece haber sido menos afrancesada que la de su
abuelo el Emperador: hay muchos menos nombres extranjeros entre los dig-
nitarios que firman los diplomas reales . . . La influencia política ul-
trapirenaica empezaba a declinar. Por excesiva, había provocado una reac-
ción. (Lapesa 1951: 210)

Lapesa also provides clear evidence of the declining number of French

prelates and nobles in the Peninsula, as well as the assimilation of the chil-
dren of earlier immigrants. He places the end of French prestige in Castile
in the year 1212, when the French armies literally couldn't "take the heat"
172 The Toledo phase

and abandoned the Hispanic armies just before the great battle of Las Na-
vas de Tolosa. What then are we to make of Lapesa's claim that apocope
reaches its zenith between 1200 and 1230, or, more damaging still, Harris-
Northall's (1991: 35) convincing evidence that preference for apocopated
forms actually survived (at least in Alfonsine texts) up to the very end of
the 13th century? Following Lapesa's own logic, the continued use of ex-
treme apocope in the 13th century makes no sense.114 How would most
speakers (and writers) even know that they were imitating the French when
they must have had progressively less contact with them, particularly in
regions distant from the Camino de Santiago and the Pyrenees?
A more reasonable analysis is to view extreme apocope, at least in its
later manifestations, not as a foreign intrusion but as a norm of Castilian.
As we have seen, extreme apocope appears to have challenged the phono-
tactic norms of Castilian - permitting frequent use of final /-nt/ for exam-
ple - but it never extended to the full array of phonetic contexts affected by
apocope in Gallo-Romance or even eastern Ibero-Romance. I would sug-
gest that this native Castilian norm of the Toledo phase arose as a result of
koineization. More specifically, it represents a clear case of functional
reallocation of a variant (see Chapter 2; Trudgill 1986: 125). Extreme
apocope began in central Hispano-Romance as a purely phonetic phe-
nomenon; in eastern Ibero-Romance and Gallo-Romance, it was an even
more regular feature of speech. From these various sources, extreme
apocope entered the linguistic pool of the Toledo phase prekoine. Its fre-
quency of use probably rose thanks to the outsiders present in the mix, and
its consistent use was probably associated with francos, who at the time
were numerous and often in positions of power in a society that showed
increasing consciousness of class/caste distinctions. It this context, extreme
apocope could have become marked as a prestige variant as native speakers
of central Hispano-Romance used it in situations where they wanted to
accommodate to the outsiders or take on for themselves some of the for-
eigners' perceived prestige. Of course, since apocopated variants already
existed within Castilian, accommodation in some cases meant increasing
the frequency of use of an option already available to Castilian speakers.
This increase in use in formal contexts contrasted increasingly with the
decline and near elimination of extreme apocope in informal speech. The
decline in informal speech can be attributed to the phonotactic challenge
represented by extreme apocope and its lesser frequency and consistency in
comparison with normal apocope. As a result, full forms were selected and
stabilized in informal speech. Over time, use of apocopated variants ceased
 Linguistic changes 173

to be associated with the francos, but the prestige-marking function sur-

vived in the resultant koine as the contrast with informal speech became
more marked. And of course, the stabilization of these new norms was
itself dependent on the stabilization of new social networks. In this way,
the mixing of speakers and variants in the prekoine led to the assignment of
a new stylistic (and possibly social) function to extreme apocope.
Lapesa himself was well aware of this stylistic function of extreme
apocope, as his comments on its particular association with writing reveal:
La apócope extrema nació en el habla del Centro peninsular, se incrementó
en el habla y luego decayó en el habla hasta extinguirse. Ahora bien, creo
necesario admitir que en la lengua escrita fue más arrebatado su crecimiento
y más rápido su declive. (Lapesa 1975: 22)
As stylistic variants associated with formal/literate language, apocopated
forms would naturally have tended to dominate in written texts, which in
all times and places tend to reflect more formal/literate norms. On the other
hand, the appearance of full forms shows the influence of orate or collo-
quial language (but, in some cases may also reflect Latinizing tendencies in
written language).
As Lapesa (1982) argues, there clearly was a "contienda de normas"
[clash of norms], but not between a French and a Spanish norm. Rather,
what we observe is a case of mixing with reallocation, in which two com-
peting variants survive in the resultant koine, but with new social and sty-
listic functions. Extreme apocope may be seen as the reallocation of a regu-
lar French-Occitan-Catalan-Aragonese variant, and northern Ibero-
Romance phonetic variants (probably associated with fast speech), to a
stylistic function marking formal or literate registers of speech (and proba-
bly upper social class as well). But the articulation of full forms also sur-
vived in the role of an unmarked, colloquial variant. Nevertheless, neither
one nor the other was the "true" Castilian norm: both were for a time,
though with different functions.

3.2. Reanalysis and the rise of leísmo

Leísmo, along with laísmo and loísmo, with which it is traditionally

grouped, is a phenomenon that has received the attention of many re-
searchers, both in regard to its synchronic diatopic and diastratic distribu-
tion and functioning, as well as its diachronic development. Leísmo repre-
sents a departure from the so-called etymological system of oblique
174 The Toledo phase

pronouns that developed from Latin. In the etymological system (still

dominant in most dialects of Spanish, and promoted as a prestige norm by
the Royal Spanish Academy), the Latin accusatives and datives developed
into the clitic forms presented in Table 15.

Table 15. Object pronouns: The etymological system. Source: Alvar and Pottier
(1983: 127-128).
Singular Plural
Masculine Feminine Neuter Masculine Feminine
Accusative lo < ILLUM la < ILLA lo < ILLUD los < ILLOS las < ILLAS
Dative le < ILLI le le les < ILLIS les

Examples of this usage include:

— lo/la vi en la fiesta '(I) saw him/her/it at the party'

— los/las vi en la fiesta '(I) saw them (m. or f.) at the party'
— le dieron un premio '(they) gave an award to him/her'
— les dieron regalos '(they) gave them (m. or f.) gifts'
— no me lo dijeron '(they) didn't tell me about it (= what happened)'

Leísmo has been understood (somewhat incorrectly) to alter this system in

any of several ways, which traditionally have been analyzed as follows. In
its strictest (and what is often believed to be its original) sense, leísmo
represents a novel usage in which the formerly gender-neutral dative pro-
noun le comes to be used to refer to masculine persons in the accusative,
thus encroaching on the etymological domain of lo. This use is known as
personal leísmo, and is still accepted as a prestige norm by the Royal Span-
ish Academy and by most Spaniards.115 However, there also exist animal
leísmo (a perceived extension of reference to masculine animals) and ob-
ject leísmo (a perceived extension of reference to any masculine object or
referent). One can also find, though by all accounts rare, neuter leísmo (in
which le rather than lo is used to refer to antecedents such as phrases or
whole concepts not reducible to a single term) and, in some areas, feminine
leísmo (in which le or les is used for a feminine direct object). Leísmo is
often accompanied by laísmo, a phenomenon in which singular la and plu-
ral las substitute le(s) for feminine dative reference. Loísmo has tradition-
ally been considered the least frequent of the three phenomena; it usually
describes the use of plural los (and less frequently singular lo) for dative
 Linguistic changes 175

masculine or neuter referents (Llórente Maldonado 1980: 23; Fernández

Ordóñez 1994,1999).
Leísmo can be found in many parts of the Spanish-speaking world. In
the Americas it has sometimes been seen as occurring with certain verbs,
but Fernández Ordóñez (1999) has pointed out that most of these cases of
putative American or Andalusian leísmo are only apparently so. Leísmo is
also associated with American varieties that have resulted from contact
between Spanish and one of the main indigenous languages, such as
Quechua in the Andean highlands and Guaraní in Paraguay. In such varie-
ties le(s) is often used for dative and accusative, masculine and feminine,
singular and plural reference. A similar kind of radical leísmo, also result-
ing from contact between typologically distinct languages, can be found
within modern Spain itself, in the Basque country (see below). However,
the kind of leísmo (and related phenomena) that is of interest here, primar-
ily associated with central-northern peninsular Spanish, has always pat-
terned very differently from the leísmo of contact varieties.
Llórente Maldonado (1980: 23-26) provided the first tentative sketch of
the modern geographical distribution of peninsular leísmo and related phe-
nomena, but recent research has led to both a revised understanding of the
interrelationships between these phenomena as well as a much clearer un-
derstanding of where different usages are employed. Indeed, the above
terms and categories do not suffice to describe the systematic nature of
many of the uses in question. For example, Klein-Andreu, in a long series
of articles (e.g., 1981, 1992, 1996, 2000), has argued that in the center-
north of the Peninsula (focused on Valladolid) there exists a referential
system of oblique pronoun use which differs greatly from the etymological
system. According to Klein-Andreu, the referential system has lost distinc-
tions between dative and accusative but exaggerates distinctions based on
gender, and also shows distinctions between count and mass (or discon-
tinuous/continuous) noun referents. Klein-Andreu's pioneering analysis
has been elaborated upon in the impressive research of Fernández-
Ordóñez. Her work includes a thorough review of previous discussion of
the question (1993), a detailed investigation of dialectal variation in the use
of pronouns within Spain (1994), a complete description of use and devel-
opments in etymological, contact, and referential pronoun systems, in
Spain and in the Americas (1999), and a recent essay on the origins of
leísmo and the referential systems (2001).
It is impossible to understand present-day leísmo within Spain without
reviewing the findings of Fernández-Ordóñez (1994). Aware that none of
176 The Toledo phase

the existing descriptions of referential systems really captured the variety

of use within them, she carried out field research among lower-class rural
speakers (less affected by standard and prestige uses) throughout the center
and north of Spain, with the goal of better defining the varied pronoun
systems and determining their exact geographical extension. Her work has
allowed her to map the geographical distribution of several different sub-
types of referential systems, as well as other pronoun systems (Fernández-
Ordófiez 1994: 125; Map 5 is based on information in Fernández-Ordóñez
1994 and 2001). Given the tremendous variation in actual usage, and the
competing influences on speakers, she has had to abstract these systems
from dominant usage (i.e., none of them can be said to be categorical for
entire communities).
Although Fernández Ordóñez (1994) begins her discussion with the
core referential systems (systems A, B, C), for historical reasons, it makes
sense to begin this discussion with the pronoun systems F, G, and H of
Asturias and Cantabria. Systems F and G (Tables 16 and 17) represent
particularly ancient and complex systems.

Table 16. Object pronouns: System F. Source: Fernández Ordóñez (1994: 101).

Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative lu la lo los les/las
Dative i i i yos yos

Table 17. Object pronouns: System G. Source: Fernández Ordóñez (1994: 103).

Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative lu la lo los las
Dative li~le li~le li~le lis~les lis~les

Systems F and G are characterized by the regular preservation of

count/mass distinctions, as well as distinctions based on case, gender (in
accusatives), and number. Though different from the so-called etymologi-
cal system, it is important to note that they remain etymological systems
nonetheless, and in this sense are quite conservative. Significantly,
 Linguistic changes 177

count/mass distinctions such as these seem to have survived only in As-

turias and Cantabria (except for some areas of Italy), while all other Ro-
mance varieties (except in the Basque Country) developed pronoun sys-
tems that correspond to the etymological system.
Within these systems, it is important to note that count/mass distinc-
tions (only in the singular, of course) take precedence over gender distinc-
tions. Thus, lo is used for both masculine and feminine mass reference
(including neuter reference), while the forms lu and la contrast only for
masculine and feminine count nouns, respectively. Otherwise, these sys-
tems are similar to the etymological system (the remaining differences are
not significant to the following discussion). System F is found in central
and eastern Asturias, while system G survives in eastern Asturias and the
upper river valleys of western Cantabria. The following sentences are ex-
amples of the count/mass distinctions which characterize these systems:

— a Bernardu llevàronlu preso 'Bernardo, (they) took him prisoner'

— el paquete olvidélu 'the package, (I) forgot it'
— a María tú no la conoces 'Maria, you don't know her'
— la piedra no la llevanta 'the stone, (s/he) doesn't/can't lift it'
— el café vendíalo en granu 'coffee, (s/he) sold it in grain [unground]'
— la herba primero hay que segalo 'the grass, first one has to cut it'
— aquí en pueblu, la leche no lo vende naide 'here in the village, milk (f.),
nobody sells it' (from Fernández Ordóñez 1994: 102)

As is clear from the examples, marking of distinctions in these systems is

dependent on the maintenance of at least four dintinct vowels in final posi-
tion, as opposed to the normal three of most varieties of Spanish.

Table 18. Object pronouns: System H. Source: Fernández Ordóñez (1994: 104).

Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative le la lo los las
Dative le le le les les

System H, the last of the Astur-Cantabrian systems, is found in coastal

Cantabria, and in central and eastern inland Cantabria, extending into bor-
der zones of the province of Burgos north of the Ebro.116 In comparison
178 The Toledo phase

with systems F and G, its unique feature is extension of le for masculine

count reference. While system H thus blurs one form/function contrast, it
generally maintains the distinctions of systems F and G, but with only three
final vowels, as in most Spanish dialects (even though Cantabrian dialects
maintain distinctions between final /-u/ and /-of). System H is spreading at
the expense of system G in western Cantabria, as speakers replace lu,
stereotyped as rustic, with le. While this system represents an innovation
with respect to systems F and G, it seems to preserve the forms of the pre-
cursor to the core referential systems to be described below.
Before moving to discussion of those referential systems, it is necessary
to describe the last of the pronoun systems to be found in the extreme north
of the Peninsula: system I (Table 19), found in the Basque Country and
northern Navarre.

Table 19. Object pronouns: System I. Source: Fernández Ordóñez (2001: 406).

Animates Inanimates (singular and plural)

Masculine Feminine Neuter
Accusative le(s) 0~lo(s) 0~la(s) 0~lo
Dative le(s) le(s) le(s) le

Like the etymological system and the Astur-Cantabrian systems, it is

probably among the oldest pronoun systems. It is clearly the result of
Basque-Romance bilingualism and is unique within the Peninsula. As de-
scribed by Fernández Ordóñez and others, the Basque system very much
depends on the degree of bilingualism of the individual. In initial stages of
acquisition, the speech of Basque learners of Romance/Castilian is charac-
terized by the complete suppression of accusative pronouns and exclusive
appearance and use of le(s) for dative reference. Apparently this occurs as
a result of interference from Basque (which lacks exact parallel forms) on
the Spanish interlanguage of these speaker-learners. When Basque learners
begin to include reference to accusatives, these speakers first extend the
use of le(s) to both feminines and masculines, failing to mark gender dis-
tinctions or employ lo(s) and la(s). Moreover, this extension is limited to
personal referents, following directly from the tendency of the dative to be
used primarily for personal reference. This leads to uses in which le(s) is
regularly used to refer to masculine and feminine personal antecedents,
both singular and plural. The use in the feminine is particularly surprising
to speakers of other varieties of Castilian:
 Linguistic changes 179

— para que a la hija no le violen 'so that (the daughter) they don't rape her'
— ¿la hija le va a dejar a la madre? 'the daugher ¿is she going to leave her
(her mother)?'
— pocas [chicas] les veo yo [que coman cocido] 'few [girls] do I see (them)
[who eat stew]' (Fernández Ordonez 1994: 111)

Given its clear origins in Basque/Romance language contact and language

learning, this system must certainly have existed in the medieval period. In
this regard, it is significant that this basic system has survived today even
in regions of the Basque country that have been monolingual for some
time.
To the south of these systems lie the core referential systems. The first
of these, system A (Table 20), is found in the modern provinces of Valla-
dolid, Palencia, and the west of Burgos (Fernández Ordóñez 1994: 84). In
traditional terms, system A is characterized, like system H, by personal
leísmo and, less consistently, object leísmo. Unlike system H, it shows
singular and plural laísmo, (inconsistent) singular loísmo, and plural
leísmo. Again, what the traditional terms do not recognize is the distinction
made between count and mass reference. As in Astur-Cantabrian systems,
mass reference overrides gender reference (which is itself exaggerated with
the weakening of case distinctions), and lo is used to refer to both mascu-
line and feminine mass referents. However, case distinctions are almost
completely lost, though a vestigial distinction between mass accusative lo
and mass dative le seems to survive as a last remnant of case distinctions.117

Table 20. Object pronouns: System A. Source: Fernández Ordóñez (1994: 84).
Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative le la lo les las
Dative le la lo~le les las

System Β (Table 21) is found in most of the southerly areas where the ref-
erential system occurs: the province of Ávila (including bordering regions
of Salamanca), the east of Cáceres, the west and north of Toledo, and (ap-
parently) much of the province of Madrid. This system is very like that of
system A, except that it shows plural loísmo, rather than plural leísmo.
180 The Toledo phase

Table 21. Object pronouns: System B. Source: Fernández Ordóñez (1994: 87).
Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative le la lo los las
Dative le la le~lo los las

Finally, Fernández Ordóñez identifies a system C (Table 22), which is

found in the southern extension of the modern region of Cantabria (Cam-
póo), in the north and east of the province of Burgos (excluding areas north
of the Ebro), and much of the province of Segovia. It differs from systems
A and Β in that masculine plurals also seem to preserve some distinction
between dative and accusative, with les dominating in the dative and les
and los competing about equally in the accusative, except when the refer-
ence is to masculine persons, when preference is given to les. Fernández
Ordóñez (1994: 114) attributes this to the fact that dative le and les are
normally used with personal reference, and where speakers have a choice
between two forms, this factor seems to have some influence.

Table 22. Object pronouns: System C. Source: Fernández Ordóñez (1994: 90).

Singular Plural
Count Mass Masculine Feminine
Masculine Feminine
Accusative le la lo los~les las
Dative le la le~lo les~los las

Systems A, B, and C constitute the core referential systems. They all show
count/mass distinctions, leísmo for masculine count nouns, use of lo for
masculine and feminine mass reference, singular and plural laísmo. They
differ only to the degree that case distinctions are lost in plural masculine
reference, and in what option is chosen when it is lost {los vs. les). System
C appears to represent an older state, previously shared by all core referen-
tial systems. Los for plural masculines has spread in the south, and les for
plural masculines has spread in much of the north, apparently spreading
from Palencia and/or Valladolid, an important city in the region since the
14th century.
Outside these core referential systems, there exist other systems that
Fernández Ordóñez identifies as "transitional". The first of these is system
 Linguistic changes 181

D (Table 23), found in the east of the province of Leon, just to the west of
the areas where system A dominates. This system is fundamentally similar
to neighboring system A, except that it shows a) loss of marked lo for
feminine mass reference, and b) some uses which seem to reveal either the
survival or influence of another system. For example, lo(s) occurs in up to
40% of masculine count accusative contexts, and le occurs in about 15-
20% of singular and plural feminine datives. As in systems C and T, plural
(but not singular) les occurs with slightly greater frequency for animate
reference (75%) than for inanimate reference (60%). The history of devel-
opment of this system is not entirely clear from these patterns. It could
easily result from contact with the etymological system (which dominates
in the modern provincial capital), but it could also reveal an earlier system
(like G, but with merger of lu and lo in lo) that has only partially accepted
the innovations of neighboring system A.

Table 23. Object pronouns: System D. Source: Fernández Ordóñez (1994: 95).

Singular Plural
Count Mass Masc. Fem.
Mase. Fem. Masc. Fem.
Accusative le- lo la lo la les~los las
Dative le la lo~le la~le les las

Another transitional system is found to the northeast of the core area, in

the small part of the province of Burgos found to the north of the Ebro (the
area I have identified as primitive Castile). It is not used in the most north-
western portion of this area (where system H dominates), but it is found in
border areas of the Basque provinces of Alava and Vizcaya, to the east and
northeast. Fernández Ordóñez (1994) did not include this system within her
original classification scheme, but in Fernández Ordóñez (2001) she refers
to it regularly as the northern transitional system, so here it is labeled sys-
tem Τ (Table 24). Since Fernández Ordóñez considers this system of great
historical importance, it is worth describing in some detail.
System Τ is characterized by the regular maintenance of case distinc-
tions: le(s) is always used for dative reference (i.e., laísmo and loísmo do
not appear), but le dominates for masculine count reference in the accusa-
tive. In this respect, system Τ is very similar to system H, and clearly more
conservative than system C, found just to the south of the Ebro. System Τ
shows the fairly frequent phenomenon of loss of marked lo for feminine
182 The Toledo phase

Table 24. Object pronouns: System T. Source: Fernández Ordóñez (1994: 114).
Singular Plural
Count Mass Masc. Fem.
Mase. Fem. Masc. Fem.
Accusative le (< lo) la lo la (< lo) los (< les) las
Dative le le le le les les

mass reference, with la now dominating lo (this is probably a recent phe-

nomenon and not of great significance to this discussion). Still there are
two features that seem to set system Τ apart from most others. First, it
shows an elevated (though not dominant) use of les for masculine personal
reference in the plural (but not the singular). Though system Τ shares this
feature with System C south of the Ebro, the more distant system D, and
the neighboring system I of the Basque Country, it is the only non-Basque
leista system that shows any plural leísmo without accompanying laísmo.
As I will argue below, there is some evidence that plural leísmo predates
laísmo, even though laísmo is likely to attain categorical status more
quickly once it appears. If this is so, then plural leísmo can be understood
as a normal analogical development, as it is for other referential systems.
Another distinctive feature, more difficult to explain, is the high frequency
of lo for masculine accusative count nouns, which approaches 40%. As in
system D, it is not clear whether this use of lo is innovative or conserva-
tive. Since this small area is bounded by system Η to the northwest, system
C to the southwest, system I to the northeast, and the etymological system
of La Rioja to the southeast, it is exposed to more conflicting norms than
almost any other pronoun system. Use of lo for count le could be an ana-
logical extension back to the singular (based on competing use of les and
los), but it could also represent the partial survival of a conservative system
(see below).
Finally, the most southerly of the transitional systems identified by
Fernández Ordóñez (1994) is system E (Table 25). System E occurs in a
long strip from the western areas of the province of Guadalajara, through
the eastern tip of the province of Madrid (apparently), down to the east of
Toledo, including some towns to the southwest of Toledo and even to the
north of Toledo (in areas where it competes with system B). The most dis-
tinctive feature of system E is maintenance of leísmo but with the replace-
ment of the count/mass distinction by an animate/inanimate distinction for
masculine singular referents. Although this system is used by rural infor-
mants over a wide area, it corresponds closely with the prestige norm of
 Linguistic changes 183

personal leísmo. Still, Fernández Ordóñez emphasizes that this system var-
ies according to its proximity to other pronoun systems. Those towns
nearer to etymological zones show consistent use of system E (personal
leísmo), while those which are nearer to referential zones (system B) show
increasing use of le for count reference and use of lo for mass reference
(tending toward a system like system H). Fernández Ordóñez (1994: 99)
argues that system E represents a reanalysis of an early referential system.

Table 25. Object pronouns: System E. Source: Fernández Ordóñez (1994: 97).
Singular Plural
Masculine Feminine Masculine Feminine
Animate Inanimate
Accusative le lo la los las
Dative le le le les les

The complex nature and geographical restriction of these phenomena

might lead one to conclude that they arose fairly late in the history of Cas-
tillan, at least post-dating the expansion into Andalusia in the mid-13th
century, yet textual evidence appears to indicate the contrary. Before dis-
cussing this evidence, it is important to consider some potential problems
in interpreting the data on leísmo and related phenomena in medieval texts,
an issue discussed with impressive clarity by Fernández Ordóñez (2001:
389-398). These include the following:

1. As is the case today, it seems likely that leísmo, laísmo and loísmo were
salient features, readily noticed and consciously altered by both speakers
and scribes. As a result, written texts often reveal the influence of compet-
ing norms of usage. Preference for a certain set of norms was probably in-
fluenced by text genre and by the learned norms of Latin. On the other
hand, scribes often worked in areas distant from their homeland (e.g.,
royal scribes) and would have been influenced by dominant usage in their
new communities. In addition, scribes working in a scriptorium or chan-
cery were certainly exposed to and influenced by the native or preferred
usage of other scribes. Still, as we will see, some texts do reveal clear pri-
mary influences, apparently based on norms of spoken language.
2. Many medieval texts (and modem editions of them) are based on later
copies (and copies of copies). For example, only some of the Alfonsine
texts (those emanating from the Royal Scriptorium under Alfonso X) are
184 The Toledo phase

preserved in 13th-century manuscripts. Clearly, these are of greater value

in dating the development of leísmo and related phenomema.
3. Literary or historical texts with a known author are usually preserved in
contemporary or later copies made by scribes (or, in later periods, by
typesetters and proofreaders). The original (or earlier copies) may or may
not have influenced the scribe's use. For example, as Fernández Ordóñez
argues, most of the works by the 13th-century Riojan poet Berceo are pre-
served in later copies of varied origin and date. Some of these texts show
some incidence of leísmo, others none. How can one know which or any
reveal the "true" usage of Berceo? Still, all observers agree that these texts
show a very low incidence of leísmo, rising no higher than 10% (Lapesa
1968: 527; Echenique Elizondo 1981: 134; Fernández Ordóñez 2001:
392-394), and they thus correspond surprisingly closely to the etymologi-
cal usage of La Rioja.
4. Scribes often worked in teams on long texts, with different scribes tran-
scribing different sections. Moreover, given the use of dictation in making
copies, one must also consider the influence of the scribe who read the
text aloud to the transcriber. This method of text preparation clearly fa-
vored at least some mixing of norms in every text, though sections actually
written by one scribe are likely to reveal the primary influence of one sys-
tem, as is seen in some Alfonsine texts such as the Primera Crónica Gen-
eral (cf. Echenique Elizondo 1979, 1980).

These are problems one expects when studying the development of any
sociolinguistic marker, but there are also problems particular to the study
of leísmo and related phenomena that complicate the task further:

1. Apocope of -e was normal following laterals, and we can be certain that le

was very often apocopated in early texts. However, apocope of -o also oc-
curred in high-frequency forms (e.g., mucho ~ muy/muyt/much), so it is
possible that some (or even many) cases of V represent lo. Lapesa (1968:
525), for example, recognized this last possibility, while Sanchis Calvo
(1992) has suggested that in texts with high rates of leísmo in unapoco-
pated forms only le was apocopated, but in others both le and lo were
apocopated. Fernández Ordóñez (2001: 412) assumes that le was the only
pronoun apocopated with any frequency in leista texts. In specific cases
consistent interpretation of /' as le could lead to different analyses (e.g.,
leísmo might be interpreted as based on the animate/inanimate distinction
rather than a count/mass distinction), but this is not always so. Though this
 Linguistic changes 185

issue is important when trying to determine usage in different places, it

does not prevent the determination of the broad patterns of variation re-
flected in 13th-century texts.
2. It is clear that some verbs in medieval texts vacillate in taking the dative or
accusative. Much of this vacillation can be attributed to the centuries-long
and Romance-wide tendency towards transitivization of verbs which in
Latin took the dative (Fernández-Ordóñez 1999: 1339). Indeed, Lapesa
(1968: 528-531) argued that most of the cases of leísmo in Berceo are
only apparently so, since they occur with verbs which in Latin took the da-
tive but in Castilian have shifted to the accusative. Even today, etymologi-
cal varieties in the north of Spain tend to maintain the dative with many
verbs which in (parts of) Andalusia and America have shifted to the accu-
sative (e.g., ayudarle vs. ayudarlo·, Fernández Ordóñez 1999: 1330-
1335). Such changes can make conservative usage look like leísmo.
3. There are some constructions in etymological varieties which alter or ap-
pear to alter the use of dative or accusative pronouns. These too can be
misinterpreted as cases of leísmo. Cuervo (1895: 236) pointed this out
long ago (though as a potential cause of leísmo), and Fernández Ordóñez
(1999: 1323-1330) has provided a clear overview of these structures."8

As the above discussion shows, any analysis of medieval texts must be

careful not to attribute too great a significance to any one example (or
small set of examples). Nevertheless, broad patterns of usage have been
identified in the research done so far that allows us to sketch an outline of
development of leísmo, laísmo, and loísmo.
Lapesa (1968) showed that leísmo was firmly established by the 13th
century. Using traditional classifications, he identified a predominance of
personal leísmo, accompanied by less frequent object leísmo. From this he
concluded that personal leísmo was the oldest type (Lapesa 1968: 542-
543). Echenique Elizondo (1981: 132) reported further clear and numerous
examples of leísmo (analyzed only as "masculine singular", without analy-
sis of person or object reference) in the majority of 13th-century Castilian
texts analyzed by her. More recently, Sanchis Calvo analyzed the use of
leísmo in the very early Fazienda de Ultramar, a long text preserved in a
manuscript from the first third of the 13th century (probably 1220), which
also shows heavy use of personal leísmo along with many cases of object
leísmo. Basing her analysis on unambiguous (i.e., unapocopated) forms of
lo and le, she found in a particular subsection of the text that personal
leísmo occurred at a rate of 67.4%, while object leísmo occurred at a rate
186 The Toledo phase

of 38.4% (Sanchis Calvo 1991: 266). Fernández-Ordóñez (1994: 100,

2001: 410) has argued that patterns of very high frequencies of personal
leísmo and lower frequencies of object leísmo are exactly what users of
referential systems are likely to produce, since human referents will nearly
always be considered count (discontinuous) and only objects will fluctuate
between count and mass categorization. Hence, this text does not reveal a
gradual extension of leísmo from person to object, but rather the presence
of a referential system in the early 13th century (albeit in an early stage,
rather like that which is preserved today in Cantabrian system H).
In an analysis of different parts of the Primera Crónica General (or
Estoria de España), Echenique Elizondo (1979, 1980: 206) also finds that
the fragments written by different scribes show different levels of use of
personal leísmo and object leísmo. For personal accusative reference, the
first fragment shows 159 cases of apocope, 2 cases of lo, and 108 cases of
le\ moreover there are 19 cases of object leísmo. Fernández Ordóñez
(2001: 395) interprets these figures as evidence of a user primarily influ-
enced by an early referential system, like system H. On the other hand, the
second fragment (chapters 117-385) shows 67 cases of apocope, 393 cases
of lo, and 38 cases of le (with only one case of possible object leísmo), and
this section seems to reveal a user much more influenced by the etymologi-
cal system.
Analysis of other texts shows that in the 13th century three basic sys-
tems were in use in Castilian: the etymological system, an early referential
system like system H (though in many Alfonsine texts often accompanied
by some suppression of the marked lo for feminine mass reference), and
another system based on animate/inanimate distinctions, as in personal
leísmo or system E. Fernández Ordóñez (2001: 414, following Matute
Martínez 1999) notes that there exist three dated manuscripts from the
13th-century Alfonsine scriptorium that reveal clear correspondences with
each of the three basic systems. Unsurprisingly, the Libro de las cruzes
(1259), long recognized for its Aragonese features, shows consistent use of
the etymological system. On the other hand, the fourth part of the General
Estoria (1280) uses le for reference to masculine count nouns and lo for
mass reference. Finally, the Libro conplido en los judizios de las estrellas
(1254) uses le to refer to animates and reserves lo for inanimates, as in
personal leísmo or system E.
Other texts seem to confirm this patterning. Etymological texts are
found in all periods. For example, the Fuero de Madrid (1202) reveals no
case of leísmo, and the Historia troyana very few (Echenique Elizondo
 Linguistic changes 187

1981: 132). Fernández Ordóñez (2001: 408^410) also considers the data on
pronoun use reported by Lapesa (1968) and López Bobo (1990) for the
three copies of the Libro de Buen Amor, which date from the late 14th and
early 15th centuries. All these texts are influenced by the etymological
system, but in competition with either system E or system H. The Sala-
manca manuscript (characterized by many Leonese features) reveals use of
personal leísmo at 30%, while the Gayoso manuscript (considered the most
Castilian version) shows personal leísmo at 60%. Neither of these texts
shows significant object leísmo, and thus reveal scribes heavily influenced
by system E (as do the 15th-century texts of Alfonso Martinez de Toledo,
the Archpriest of Talavera; Fernández Ordóñez 2001: 411). On the other
hand, the Toledo manuscript of the Libro de Buen Amor (also characterized
by Leonese features) shows personal leísmo at 55% and object leísmo at
33%, proportions which reveal the influence of an early system H.
From the above it is reasonable to conclude that two kinds of leísmo had
been firmly established as Castilian norms by the 13th century: one based
on the count/mass distinction, and another, less frequent, based on the ani-
mate/inanimate distinction, with both these systems in competition with the
etymological system. Clearly, leísmo was the first of the variant phenom-
ena to become established, while plural leísmo, laísmo, and loísmo devel-
oped later. In general, Lapesa (1968: 544) found few cases of plural leísmo
in early texts. For the Poema de Mío Cid (an early text preserved in a 14th-
century manuscript), he reported 26 cases of le versus 34 of lo for accusa-
tive personal reference, but only 4 cases of les versus 120 of los. Interest-
ingly, Lapesa suggested that plural leísmo developed as a more limited
regional use, just as Fernández Ordóñez has confirmed. Echenique
Elizondo (1981: 132, 136), in a study of many early medieval texts, re-
ported that 49% of all masculine direct objects appearing with full forms
show le, while only some 8% show les. Such low numbers might simply be
the result of problems of interpretation, but Echenique Elizondo did find
one text with high levels of plural leísmo: the Alfonsine Libro conplido de
los judizios de las estrellas, which showed plural leísmo at a rate of nearly
38%. Interestingly, Echenique Elizondo (1981: 147) found no case of plu-
ral object leísmo, in this or other texts, a fact which would seem to confirm
a claim of Fernández Ordóñez: that personal reference seems to favor use
of plural leísmo where a choice is possible. Still, it is also worth noting that
this combination of singular personal leísmo with plural personal leísmo
does not match any system existing today, and may therefore represent the
attempts of the scribe to accommodate to perceived dominant usage (see
188 The Toledo phase

below). Overall, however, the low percentages of plural leísmo in 13th-

century texts indicate that it was just beginning to gain currency.
Convincing and numerous examples of laísmo do not appear in 13th-
century texts (Echenique Elizondo 1981: 148). The only sure examples
date from the 14th century, and significant evidence is only found from the
15th century on. Lapesa (1968: 544), for example, found significant laísmo
in the 1499 Burgos edition of La Celestina (which also showed high per-
centages of personal and object leísmo, as one would expect in system C).
However, laísmo, singular and plural, grew ever more intense and reached
its literary height in the 17th century, when it was preferred by authors
such as Quevedo and Calderón. Loísmo, while apparently an ancient phe-
nomenon, is relatively infrequent in written texts of all periods (Lapesa
1968: 546). Fernández Ordóñez (1994: 89) adds, however, that plural
loísmo, found today in Ávila's system Β, was a regular characteristic of the
(rather colloquial) 16th-century writings of Saint Teresa of Ávila.
Although much work remains to be done on the details of the develop-
ment of these phenomena, the above contemporary and historical data
leave little doubt about the early appearance of leísmo. However, it is im-
portant to inquire into its geographical distribution.119 Fernández Ordóñez
(2001: 408-415), though aware of the difficulties, attempts to do this by
matching the percentages of occurrence of leísmo in the literary texts with
the percentages of occurrence found by her in contemporary systems.
Given all the potential influences on scribes, exact correlations are
unlikely. Still, the evidence seems to indicate that the limits of leísmo, as
indicated in Map 5, have remained extremely stable through the centuries.
Leísmo and related phenomena do not seem to have ever become estab-
lished in Andalusia and southern regions of Spain (see Chapter 5). Ara-
gonese texts are regularly etymological, and the texts of Berceo, such as
they are, also seem to confirm an age-old preference for etymological
forms in La Rioja (except its western edge, which in the early medieval
period belonged to Castile). Leonese texts seem to show a mix, but this is
unsurprising given the (present-day) limits of leísmo. The two Leonese
manuscripts of the Libro de Buen Amor reveal this variation: the Sala-
manca manuscript shows basic use of the etymological system, with efforts
to accommodate to system E (or, as may be the case, system H), while the
Toledo manuscript shows use of a count/mass distinction. Today, leista
count/mass distinctions can be found in eastern Leon and in eastern Sala-
manca, so we should not be surprised by such usage. All in all, there is no
 Linguistic changes 189

evidence to indicate that the medieval limits of (non-prestige) leísmo were

significantly different from what they are today.
We are left then with this fundamental question: how and why did
leísmo develop, in two separate systems, almost exactly within the confines
of the kingdom of Castile following the conquest of Toledo in 1085? Many
scholars have attempted to deal with this issue (see Fernández Ordófiez
1993 and 2001 for a thorough review), although usually with an emphasis
on internal or structural factors and only rarely (and recently) with consid-
eration for social or external factors. Cuervo (1895) was among the first to
study leísmo and to offer explanations for its origin and development. Most
scholars who have subsequently dealt with the causes of leísmo have, in
broad terms, adopted or adapted one of the arguments put forward by
Cuervo. In the most influential of these, Cuervo argued that leísmo arose as
the result of a tendency to confuse the case assignments of verbs in certain
kinds of constructions. Later, Lapesa adopted and extended this approach.
According to Lapesa (1968), the origin of leísmo can be found in the ex-
tension of the dative from groups of verbs which always or sometimes took
the dative in Latin to groups of verbs which usually took the accusative.
The shift occurs because the verbs affected belong to the semantic field of
human relationships, and their transfer helps to consolidate a "personal
sphere of reference" in Spanish (also manifested in the use of personal a, a
morpheme which marks personal direct objects which appear in full form).
Favoring this interpretation was Lapesa's belief that leísmo began with
personal reference and only later extended to object reference. Lapesa's
argument was not unrelated to that proposed by Fernández Ramírez (1951),
who suggested that the near equivalence of dative and personal reference
(on average, 90% of datives refer to persons) favored extension of dative
forms to accusative personal reference. Of course, this left unexplained the
restriction to masculine reference. Lapesa (1968: 541) therefore high-

Table 26. Apparent correlations between different pronouns of Spanish.

Singular Plural
Masc. Fem. Neuter Masc. Fem.
Subject el(e) ella ello ellos ellas
Demonstrative este esta esto estos estas
ese esa eso esos esas
aquel(e) aquella aquello aquellos aquellas
Leista accusative le la lo los las
190 The Toledo phase

lighted the effects of analogy with subject (él 'he', ella 'she', ello 'it', etc.)
and especially demonstrative pronouns (este 'this', ese 'that', aquel(e) 'that
over there', etc.), which show similar forms and regular distinctions be-
tween masculine and neuter (see Table 26).
Still, most scholars have seen the primary motivation in the supposed
extension of dative functions to the accusative. Fernández Ordóñez (1993,
2001: 419-421) has pointed out that the vogue for such explanations con-
tinues up to the present. For example, Garcia (1975, 1994) explained that
in etymological varieties the dative is used for more active referents and
the accusative for less active. In attempting to explain leísmo, she sug-
gested (in an argument that has not received much support; cf. Monge
1983) that the traditional machismo of Hispanic society led to an extension
of the more active dative to masculine accusative reference. Subsequently,
Flores Cervantes (1997), in a refinement of the approach first suggested by
Garcia, has argued that leísmo was favored and developed in contexts in
which the prototypical features of the dative (individuation, animacy) were
emphasized (e.g., when the object was animate or the subject lacked agen-
tivity, when a stative as opposed to dynamic verb governed the object, and
when pragmatic needs called for personification of or the showing of re-
spect to the referent). While Fernández Ordóñez recognizes the potential
impact of the first two of these contexts in etymological and transitional
systems, her own analysis of usage in the referential systems does not jus-
tify this interpretation, since many of the effects of transitivity are erased
rather than enhanced (Fernández Ordóñez 1999: 1366-1380; see below).
As we have seen, there is some evidence that the use of the dative for
personal reference has a small influence on variation today (and the same
may be true of other factors already mentioned). However, it is unlikely
that extension of dative function was a primary factor in the origin of
leísmo. First, as Fernandez Ordóñez (2001: 421—429) argues, the supposi-
tion that the functions of the dative were extended to the accusative goes
against the tendencies of most varieties of Spanish in Spain and the Ameri-
cas, as well as all other varieties of Romance, in which nothing like leísmo
exists (it also contradicts a typological universal that favors the consolida-
tion and extension of accusative case at the expense of the dative). Second,
such arguments do not explain the original extension of le only for mascu-
line singular count reference. In fact, they rest on the idea that case distinc-
tions were lost early, but in fact case distinctions were clearly marked in
feminines and plurals in the 13th century, and continue to be marked,
though only vestigially, in some referential varieties today. Lapesa's at-
 Linguistic changes 191

tempt to use surface analogy with subject and demonstrative pronouns is

interesting, but the contrast between masculine and neuter in these forms
fails to match the contrasts between count/mass (or animate/inanimate)
reference that are part and parcel of leísmo (García González 1978). Fi-
nally, and most importantly, arguments based on extension of dative func-
tions (and others like it, such as analogy with other pronouns, and vacilla-
tion in case assignment of certain verbs and constructions) try to use
features that are (supposedly) common to most varieties of the language to
explain changes that are limited to one or just a few varieties. While these
factors may have played a secondary role in the change, they are incapable
of dealing with the actuation of the change.
Recently, however, Fernández Ordóñez (2001) has attempted to do just
that. She has argued that internal factors and external factors must be con-
sidered in tandem in order to arrive at an explanation for the origin of both
types of leísmo. Fernández Ordóñez assumes - and I agree - that leísmo
originated in Old Castile with count reference, and that leísmo used for
animate reference arose only later, during the Toledo phase (see below). In
her view, the first reanalysis of le occurred as a result of language contact
in the area that I have labeled primitive Castile, where system Τ now ex-
ists. There, Cantabrian speakers using system G entered into contact with
Basques using system I. As a result of this contact, the Cantabrian speakers
borrowed the extensive use of le that they encountered in the Basques'
system I (i.e., use of le(s) for masculine and feminine, singular and plural,
dative and accusative, animate referents). This borrowing was favored by
an internal factor: the ambiguity of sentence structures employed by
Basque speaker-learners of Romance. This can be seen in the answer to the
following question/answer pair:

— Q: ¿Devolviste el libro a Juan? 'Did you return the book to Juan?'

— A: Ya le devolví. Ί already returned [it] to him' or Ί already returned it'

The structure of the answer would be understood differently by the Basque

speaker producing it and the Cantabrian Romance speaker trying to parse
it. For the Basque, the le would refer to Juan, and accusative reference to
the book would be understood as unexpressed, in accordance with system
I. On the other hand, for the Cantabrian Romance speaker, the le would be
understood as referring to the direct object, since direct objects with spe-
cific reference cannot be suppressed in other varieties of Romance, includ-
ing system G. It was the surface ambiguity in sentences of this type which
192 The Toledo phase

favored the reanalysis of le within the original Cantabrian system. More-

over, this reanalysis occurred in stages. According to Fernández Ordóñez,
the first stage is preserved in system T, which she claims still shows fluc-
tuation between animate/inanimate reference (as in Basque Romance sys-
tem I) and count/mass reference (as in Astur-Cantabrian systems). Only
later would adoptions of le in other areas lead to its exclusive association
with count nouns, as in system H.
Fernández Ordóñez (2001: 434) recognizes that this argument does not
explain the restriction of the reanalysis to masculine singular count refer-
ence: why didn't Cantabrians simply adopt Basque usage for singular, plu-
ral, masculine, and feminine reference? In response, she argues that bor-
rowing of le for anything more than masculine singular reference would
have endangered gender and count/mass distinctions, distinctions which
were also marked in other aspects of the grammar (e.g., certain pronouns
and adjectives). She also suggests that the limited extension of le was re-
lated to the desire to maintain clear formal differences between le for count
reference and lo for mass reference.
These arguments certainly represent the most serious attempt to date to
deal with the actuation of leísmo. However, there are problems with both
the external and internal factors which are claimed to underlie the reanaly-
sis of le. Let us begin with the last arguments presented. Fernández
Ordóñez' reasons for the limiting of le to masculine singular reference are
problematic. In fact, they are classic functional (and teleological) argu-
ments (like avoidance of homonymie clash) that have been thoroughly
criticized by Lass (e.g., 1997: 355-357). Why would either the system or
speakers care about what distinctions were lost or maintained? As Lass
argues, language histories are full of useful distinctions that have been lost,
and less useful ones which have been maintained. In this case, such at-
tempts at explanation fail to resolve the crucial question of why le was
reanalyzed with exclusively masculine accusative singular count reference.
The idea that leísmo arose in areas of contact (and, I note, repopulation)
in eastern Cantabria and primitive Castile is, I think, plausible (although
not necessarily for the reasons given; see below). Still, the suggestion that
system Τ is the oldest of the leista systems is problematic. Supposedly this
is so because it preserves both count/mass distinctions (from the Can-
tabrian systems) and animate/inanimate distinctions (from Basque Ro-
mance), as well as singular and plural leísmo without laísmo, a seemingly
odd occurrence that might be explained as due to early Basque influence.
However, Fernández Ordóñez' (1994: 114) original and thorough descrip-
 Linguistic changes 193

tion of this variety specifies only that (variable) plural leísmo is more fre-
quent with personal referents. This, of course, is a phenomenon that occurs
in other varieties where a choice is possible, and it is also supported by
historical data. With regard to the present state of system T, one can just as
easily argue that it was a conservative system that has subsequently been
affected by contact with the other pronoun systems near it (see below).
Fernández Ordófiez' argument that ambiguous syntactic structures were
misinterpreted is promising, but it depends crucially on the assumption that
speakers of Cantabrian Romance would have wanted to adopt features of
Basque Romance into their own speech. In and of itself, this is not an un-
reasonable argument, but, as we saw in the previous chapter, there is very
little evidence (and little likelihood in a koineizing context) of Basque or
Basque Romance having such a primary and direct effect on the speech of
Cantabrians. Indeed, this argument suffers from the same weaknesses as all
arguments which claim contact with Basque as the simple cause of changes
in Castilian (or Cantabrian): why would a majority of native Romance
speakers copy a highly marked feature produced by a minority of Basque
Romance speakers?; why do we not find similar sorts of influence in other
Romance-speaking areas which border the Basque Country (e.g., southern
Navarre and La Rioja)? Nevertheless, I would not want to argue that
Basque usage had no effect at all. To the extent that it existed, it was
probably only secondary, in that radical leísmo would have introduced
further variation into the prekoine linguistic pool and thereby weakened
traditional norms.
In what follows, I propose an alternative explanation for the origins of
early leísmo (as in system H) and then move on to consider changes of the
Toledo phase and the origin of system E, as well as the chronology and
motivations for other changes which were set in motion by the rise of
leísmo. To begin, following Fernández Ordófiez' emphasis on external
factors, it is reasonable to suppose that koineization (rather than strict lan-
guage contact) played the key role in the origin of leísmo, for it arose and
remained in the central Castilian areas where demographic and dialect
mixing were most intense. Moreover, it seems likely (though not absolutely
necessary) that leísmo first arose during the Burgos phase and was then
carried south during the Toledo phase. Of course, Burgos-phase koineiza-
tion began in primitive Castile north of the Ebro, where three main groups
came together: Asturians and Cantabrians (including perhaps some existing
population in the more mountainous zones), Basques, and speakers of Ro-
mance from the Ebro valley (Μοχό 1979: 65). Though dialect contact and
194 The Toledo phase

mixing began in this region, it continued and intensified to the south

around Burgos. The koineizing context(s) would have been ideal for ex-
tending an innovation like leísmo, but only if sufficient numbers of speak-
ers (not just a minority) were likely to produce and/or accept the same in-
novation. For this to happen, the dominant dialect systems (like modern
system G) must have had or developed some structural ambiguity that al-
lowed and even encouraged numerous speakers to make the same form-
function reanalysis of le. Below, I argue that the structural ambiguity arose
as a result of early apocope of object pronouns.
Cuervo (1895: 235) was among the first to mention apocope as a possi-
ble source of confusion, but seeing that neuter lo was rarely if ever apoco-
pated, he suggested instead that le for lo arose as an effect of analogy with
the first and second person clitics me and te, which are used for both dative
and accusative reference. However, years later, Lapesa would argue
(against Cuervo and Menéndez Pidal) that apocope of lo was perfectly
plausible in medieval Castilian:
Opiniones muy autorizadas consideran que -1' es siempre la variante
apocopada de le. Sin embargo, creo que no hay razón para descartar que en
función de acusativo proceda de lo: la -o final, convertida en intertónica por
fonética sintáctica, desaparece en los nombres propios ante patronímico o
apellido . . . tot siempre . . . como > cum . . . bueno > buen . .. Igual pudo
ocurrir, hubo de ocurrir, a la -o del lo intertónico. No es obstáculo el que la
apócope se produzca en el lo masculino y no en lo neutro, pues bien conoci-
dos son otros fenómenos fonéticos ligados a ciertas categorías gramaticales.
En la forma apocopada -/' confluyeron, pues el l(o) del acusativo masculino
y el 1(e) de dativos. (Lapesa 1968: 525)

Lapesa admitted that apocope of lo might have led to the generalization of

le, and he recognized that the alternation between full and apocopated
forms of the pronouns m(e), t(e), s(e) might have influenced that develop-
ment. Nevertheless, he showed little enthusiasm for this argument and in-
stead defended his broader proposal relating leísmo to the impact of the
personal semantic sphere (Lapesa 1968: 525). Others, such as Echenique
Elizondo (1981) and Sanchis Calvo (1992) have also assumed that there
might be some causal relationship between apocope and leísmo. For exam-
ple, Echenique (1981: 136) pointed out that high percentages of apocope
and leísmo tended to co-occur in 13th-century texts, and suggested that this
co-occurrence might be a reflection or relic of an earlier state of the lan-
guage during which leísmo had first appeared. Of course, the corollary is
that texts with significant use of lo showed little apocope. Fernández
 Linguistic changes 195

Ordóñez (1993: 69) has interpreted this as meaning that apocope of lo

could not have played a role in the development of apocope, and does not
mention leísmo as a possible factor in any of her later publications. How-
ever, if apocope allowed the rise of leísmo prior to the 13th-century, then
the 13th-century texts should show exactly the patterns they do (i.e., co-
occurrence of apocope of le and leísmo), since by that time the change was
established. Texts showing other patterns tend to come from the west or
Andalusia (with low levels of apocope and leísmo) or from the east (with
high levels of apocope, though not of lo, and little leísmo; see below).
In fact, as Fernández Ordóñez' own studies reveal, the rise of leísmo
probably first occurred in Old Castile during the Burgos phase, only to be
carried south (and altered) in the Toledo phase, and eliminated along with
apocope in the Seville phase (see Chapter 5). In what follows we will re-
construct the social and linguistic contexts that led to the rise of leísmo. It
is assumed that the dominant pronoun system represented in the early Bur-
gos-phase mix was Cantabrian system G (now found only in mountainous
areas of western Cantabria and eastern Asturias). Therefore, the extension
of le to accusative masculine count reference must be seen as replacing lu,
not lo (as Fernández Ordóñez has also argued). Previous discussions of the
possible effect of apocope have always centered on -o and -e, the only
forms found in later texts. However, it can be plausibly argued that an early
apocope of lu and le/li in primitive Castile was in fact the key phenomenon
in the development of early leísmo. Such apocope would have resulted in
surface output in which both lu and lo were realized as /', but in which
accusative forms la, lo, los, las remained distinct. As early Castillans
moved south into the Duero Valley, they would have carried this system
with them, and there it would have been especially susceptible to reanaly-
sis. Several factors support this argument:

1. Apocope of vowels after a lateral was a feature of normal apocope. Indeed,

it seems that post-lateral position was among the most preferred of all con-
texts for apocope. Even in the Glosas Emilianenses, where most words ap-
pear with a final vowel, qual and tal appear without [-e]. Penny (1969: 69)
finds relic forms in the Valle de Pas which show apocope of [-u] after liq-
uids: barril, badil (as in Castilian), pabil, incrédul, alar. Mozarab and
other Ibero-Romance dialects were also known for the apocope of [-u/o] in
the suffixes -iel, -el, -uel, -ol (e.g., Buñuel in Navarre, Cardiel in Aragon,
Xaramiel in Burgos, now Jaramillo\ Menéndez Pidal 1964: 180-185).
2. The Tobler-Mussafia effect also favored apocope of clitic pronouns, since
in this period they did not appear in absolute initial position and were al-
The Toledo phase

ways atonic and enclitic (e.g., vilu Ί saw him', or no lu vi Ί didn't see
him', but not lo/le vi, as in modern Spanish). Such positioning would have
favored increased levels of apocope or phrasal syncope, since the clitics
generally appeared between other stressed elements. This phenomenon was
frequent with preposed names and titles: Nun Albariz, duen Tello, don
Fernando (Menéndez Pidal 1964: 191).
The lesser degree of perceptibility of high vowels makes them more sus-
ceptible to elision (apocope or syncope). If [-e] was apocopated early, then
[-u] certainly must have been as well. Indeed, for the related phenomenon
of Romance syncope, Harris-Northall (1990b) found that u suffered syn-
cope with the greatest frequency, followed by i, e, o, and a, in descending
order of frequency. Interestingly, Montgomery (1975), who analyzed four
large 13th-century texts, found that the most frequently apocopated items
were conjugated or declined forms which had ended in Latin /-i/ (i.e., first
person of the strong preterite, second person of the preterite, and, signifi-
cantly, le). He suggested that the high frequency of apocope of these items
was the result (and a relic) of their having been the first forms to suffer
regular apocope.
The frequency with which function words such as lu and le/li occurred
would itself have favored their early inclusion in the process of apocope.
Bybee and Hopper (2001: 10) report that recent research has documented a
tendency in which words of higher frequency tend to undergo a sound
change at a faster rate than words of lower frequency. This tendency has
been most clearly identified in cases of phonetic reduction, including such
phenomena as English reduction to schwa, schwa deletion, t/d deletion,
and deletion of [5] in Spanish. Apocope of lu and le would seem to fit this
pattern closely.
Astur-Cantabrians moving into primitive Castile (north of the Ebro)
brought with them their system G, but they also entered into increased con-
tact with easterners and eastern norms, including more intense apocope of
final vowels. Indeed, given the long-standing association of the Upper
Ebro with eastern regions, such tendencies probably survived and devel-
oped among remaining inhabitants in eastern areas of Cantabria and south-
ern Alava. We know that distinctions between -u and -o survived in this
region up to and beyond the 13 th century, and it is probable that system G
dominated there as a popular norm. Menéndez Pidal (1964: 172) provides
numerous examples, such as un pedaçu 'a piece', so nietu don Johan ven-
diolu a don Diago 'his grandson sir Juan sold it to sir Diego', tienelu
'(s/he) has it'. But Menéndez Pidal also provides examples of apocope
from this region of -e and -u. His examples of possible apocope of -u in-
clude mulo rodan (1145), escriuan de illa regina (1111) 'scribe of the
queen', as well as the frequent names Lop and Martin. The Documentos
 Linguistic changes 197

lingüísticos are poor sources of data on the use of object pronouns, but the
following late example does appear in Doc. 70 (dated 1310,fromFrías, on
the Ebro near or in modern system T), in which the juxtaposition of the
accusative forms lu and / would seem to indicate apocope of lu: " . . . & ssi
aluor sse ssecare, quelli ayamos a meyas. Y este ssolar uos arrendamos po-
ra en todos vuestros dias con todas estas posturas, & que noi podades ven-
der njn empeñar njn mal meter..."

Thus, it is likely that lu and le/li suffered frequent apocope in primitive

Castile, and that frequent apocope of these forms survived as repopulation
continued in Burgos and regions farther south and west. Still, as Menéndez
Pidal (1964: 172) showed, distinctions between final -u and final -o were
lost in the more southerly zones. The frequent apocope of the masculine
accusative singular count pronoun and the dative singular would have ren-
dered the distinctions between their forms increasingly opaque, and new-
comers to Burgos (especially from farther west) and children would not
have known how to interpret these apocopated forms. Indeed, Menéndez
Pidal (1964: 173) reported especially numerous cases of -e for -o in docu-
ments form Sahagún, a monastery located in eastern Leon near the modern
border between systems A and D. Examples include Moratelle for Moratiel
(971), Lope (< lobu/o < LUPU), Fortuni (975), Ual Fartelle for Ual Far-
telo, Fredenande Albarez (969). While Menéndez Pidal interpreted this -e
as a relaxed vowel, it seems more likely that it represents a hypercorrec-
tion, as is indicated by the survival of the name Lope and the more general
tendency to add final -e when adapting consonant-final Arabisms, Mo-
zarabisms, and Gallicisms to the phonotactics of Castilian (Lapesa 1951).
In the case of /', whatever its source, Burgos-phase hypercorrections
such as those reported by Menéndez Pidal would have marked the begin-
ning of its systemic reanalysis as le. Such hypercorrections would have
been especially favored in and around Burgos, since the koineizing social
conditions would have allowed the spread of the innovations. Speaker-
learners (the majority of whom used systems like G) must have encoun-
tered frequent use of the apocopated form /', and they would have had to
interpret this in order to construct grammars and reproduce others' output.
We can be sure that they excluded several interpretations based on the
input they received. The restriction of apocope to the forms le and lu would
have prevented any initial confusion between /' and the other
forms/functions of la, las, lo, los, and les. These distinctive forms provided
speakers with clear and consistent input that served to maintain case, gen-
der, and count/mass distinctions. It was only with the forms of dative le
198 The Toledo phase

and count lu/lo that input was obscured, due to the effects of apocope (e.g.,
noi vi Ί didn't see him'). Of course, full forms were articulated in some
phonetic (and probably social) contexts (e.g., vimoslu or vimoslo 'we saw
him/it'), but other factors were at work that served to obscure the value of
this input: the variable use of the dative or accusative with particular verbs
and constructions; the generally high level of variation in the community;
and the fact that even in the best circumstances clitic pronouns are ac-
quired late, and remain unstable in the speech of older children (Clark
1985: 714).
In making their reanalyses, speaker-learners were certainly influenced
by the general tendency to add -e to words which were perceived to have
suffered apocope. Of course, they were also influenced by paradigmatic
pressures from the other pronouns which were frequently apocopated but
which ended in -e. These included the other oblique pronouns (me, te, se),
and the personal and demonstrative pronouns/adjectives (el/e, este, esse,
aquel/le), in which final -e was associated with masculine singular refer-
ence. Beyond these structural considerations, the impact of Basque bilin-
guals on the linguistic pool should not be disregarded. While children of
native speakers would have moved towards more frequent overgeneraliza-
tion of le for lu/lo, Basque bilinguals would have moved in the opposite
direction, towards limiting their overgeneralization of le to only those ref-
erents. Like all others, these learners would have received clear input on
use of la, las, lo, los, and les, but they would not have received clear input
on underlying forms of /'. As a result, Basque bilinguals (or their children)
probably increased the frequency with which le for masculine accusative
count reference was heard (in contexts where apocope was disfavored),
and their output may have even been critical to the success of the change.
As frequency of le for lu/lo increased, and communities stabilized, early
leísmo became established as a norm of Burgos-phase Castilian, outside of
more conservative areas in northern Castile. However, once established (or
during its establishment), it almost certainly became a salient marker of a
(Duero Valley) Castilian identity, and could then spread north into regions
that had not originally used it (a process still happening today in western
Cantabria). Still, its spread into mountainous areas of Cantabria probably
occurred from the north, after the founding and population of the coastal
cities during the Toledo phase. The area of primitive Castile north of the
Ebro (where system Τ is now found) appears to have remained as a conser-
vative holdout, and only began to accept the innovation later (indeed, it
does not seem to have accepted it completely even today, but social con-
 Linguistic changes 199

servatism and contact with varied systems may be responsible for this).
System D in eastern Leon, probably once more like system G, gradually
adopted leista usages as Castilian influence waxed (but has probably re-
mained under the influence of neighboring etymological systems).
The above reconstruction is coherent and consistent with the existing
data on leísmo, and has the advantage of explaining clearly why the exten-
sion of le was limited to masculine singular count reference. Nevetheless,
an alternative hypothesis remains possible. As seems to be assumed in
earlier discussions, it could be that apocope began with final /-e/ and then
extended to final /-o/. If this happened, it must be assumed that system G
had suffered merger of lu and lo (in Burgos and the Duero Valley), but that
distinctions between mass lo and feminine la made it possible for speakers
to recreate a distinction between masculine count and mass noun reference.
This, in turn, would have been dependent on the limitation of apocope of lo
to its use for masculine count reference, leaving mass lo unaffected. There
are cases of such morphological restrictions of apocope (of -e) in Castilian
(e.g., the lack of apocope in present subjunctives of -ar verbs) and in Ara-
gonese, which was generally more strongly affected by apocope of -e and -
o (cf. Catalán 1971: 79). In the Liber Regum (or Cronicón Villarense), an
Aragonese text preserved in an early 13th-century manuscript, we find,
unsurprisingly, no examples of leísmo. Dative le is regularly apocopated;
however, there is no apocope of the masculine singular accusative pronoun
lo (Cooper 1960: 56, 87). This appears to have been a regular tendency in
Old Aragonese, since Tilander (1956: 31) reports a similar phenomenon in
the Vidal Mayor (an Aragonese text from the 1250s). Nevertheless, both
these restrictions occur with all functions of a particular form, and do not
require speakers to distinguish between different functions of the same
surface form, as would have been the case with apocope of count lo. If the
reanalysis of lo occurred in this way, then it probably happened much later,
with the new round of dialect mixing of the Toledo phase. Further research
will have to judge the relative value of the two hypotheses, but the assump-
tion of early apocope of lu and le/li shows greater explanatory potential.
If system H was indeed established during the Burgos phase, then it
would have been carried south during the Toledo phase. Despite popula-
tion mixing, Castillans (and those using systems like G) represented the
majority in most regions conquered by Alfonso VI, and in most of those
regions system H became the norm. Fernández Ordóñez argues along simi-
lar lines, and points out that the limits of modern leísmo (and referential
systems) coincide almost exactly with the limits of Castile during the
200 The Toledo phase

Toledo phase. Unsurprisingly, Ávila and Segovia in primitive Extremadura

adopted system H (but not Soria, which was repopulated by Riojans and
Aragonese), as did most parts of the kingdom of Toledo retained after the
Almoravid counter-attacks and settled during the first 30 years or so. These
areas include most areas along and behind the Tajo River, from Talavera,
over to Toledo and Madrid (but not Guadalajara). The only modern refer-
ential areas that extend beyond the Castilian border of Alfonso VI are
found in eastern Salamanca (near Avila) and northeastern Cáceres and
parts of western Toledo (to the south of Ávila). Fernández Ordóñez (2001:
452-453, citing Barrios 1985), points out that Alba de Tormes, a Leonese
city southeast of Salamanca, was repopulated by a majority of Castillans
(an occurrence that was favored by the union of the crowns of Leon and
Castile under Alfonso VI), as was Béjar, whose settlement was overseen
from Ávila. Plasencia, further south in northeastern Cáceres, was recon-
quered only later in 1186, at a time when Leon and Castile were under
separate crowns, and its repopulation was directed by Castile. Unsurpris-
ingly, towns and areas under the control of Plasencia or which came under
its control (e.g., Trujillo) later adopted leista usage.
Of much more interest, however, is the establishment of system E,
which appears to date from the Toledo phase as well. It is found today in a
band along the entire southeastern fringe of the leista zone. This includes
the area of Guadalajara (one of the key Toledan cities reconquered under
Alfonso VI) as well as areas to the east of the city of Toledo. These areas
appear to have been characterized by greater influxes of non-Castilians
and/or dependency on the cities of Guadalajara or Toledo. Guadalajara was
repopulated with great difficulty; its fuero specifically mentions the need
to seek settlers from Castile, Leon and Galicia, and it also retained old
Mozarab communities and attracted new cohorts of Mozarabs fleeing the
south (Μοχό 1979: 232). But this eastern region certainly attracted settlers
from the east, for in 1085 and for years afterwards Navarre and Aragon
were blocked in their advance southwards. La Rioja was under Castilian
control at the time of the reconquest of Guadalajara (1076-1109; again in
1135-1162 and permanently from 1176). More significantly, repopulation
of areas between Guadalajara and Aragon was organized by the Aragonese
king Alfonso el Batallador from 1125 (a fact which explains the final
elimination of leísmo and the dominance of the etymological system just to
the east of Guadalajara, in the area known as La Alcarria).
This high degree of mixing in areas immediately adjacent to areas
where system H was established, seems to have favored the rise of system
 Linguistic changes 201

E. Speaker-learners who lacked an understanding of the count/mass dis-

tinctions would have found it difficult to understand and reproduce Old
Castilian pronoun usage (as have centuries of grammarians), even as social
circumstances urged them to accommodate to that norm. In interpreting
this apparently bizarre usage, non-Castilian speaker-learners applied a
more widely-shared criterion. Perceiving that le was always used for refer-
ence to masculine persons and only inconsistently for objects, speaker-
learners reinterpreted the perceived overuse of le (or / ") as referring to
masculine persons (or animates) only (and simultaneously rejected the odd
feminine mass lo). In this, they may have also been influenced by the
dominant use of dative le for personal reference (suggested by Fernández
Ramírez 1951: 198). Once again their production would not have been
inhibited in the koineizing environment, and the innovation, repeated by
numerous individuals, could become a change.
A scenario like this also seems to have happened near Toledo, espe-
cially in areas just south and east of the Tajo. This region was more charac-
terized by mixing between Castillans, who seem to have dominated overall,
and Mozarabs, who seem to have dominated in some communities (e.g.,
Orgaz), but it is also important to note that the towns showing system E
seem to have been founded by and were dependent upon the city of Toledo.
Many of the Castilians and Mozarabs who populated these zones came
from the city of Toledo itself (González 1975-1976: 209, 216). Today, the
towns near Toledo showing system E are found mostly to the south and
east (e.g., Consuegra, Lillo), but Fernández Ordóñez (1994: 97) also identi-
fies towns to the north of Toledo which show this system (Recas and
Camarenilla), even though they are surrounded by other towns showing
system B. Significantly, however, nearly all the towns showing system E
(including Recas to the north of Toledo) were founded by the city of
Toledo during the mid-12th century, after the first wave of repopulation.
The fact that they were all founded later and were clearly under the control
of Toledo - and thus socially oriented toward the city - may indicate that
what is preserved in these towns is an original pronoun system of the city
of Toledo itself. Indeed, no other city in Castile saw the degree of demo-
graphic mixing that characterized Toledo, and it is therefore the most likely
place for another reinterpretation of system H (resulting in system E or
personal leísmo) to have taken place (although in Toledo it would have
continued to compete with system H). This was a period of unprecedented
prestige and power for the Castilians, and non-Castilian newcomers and
existing residents would have sought to accommodate to the perceived
202 The Toledo phase

usage of the Castilians, producing another system E as in Guadalajara (it is

interesting to note too that the 15th-century author Alfonso Martinez de
Toledo made regular and consistent use of personal leísmo; Fernández
Ordóñez 2001: 411). This reconstruction is strongly favored by the fact
that system E - personal leísmo - has appeared independently in other ar-
eas as well. Fernández Ordóñez (1994: 118; 2001: 408,410) points out that
a few towns in La Rioja (near the Camino de Santiago) as well as border
areas of Soria and Salamanca show some personal leísmo as a popular
norm, and this same system seems to have been reinvented (or chosen?) by
modern grammarians who have been unable to identify the factors govern-
ing dominant pronoun use in Castile (Klein-Andreu 1992).
A final question on the Toledo phase is why leísmo ceased to prosper
beyond these zones. I address this question with regard to Andalusia in
Chapter 5, but it now seems clear that the limits of leísmo were being es-
tablished even before then. Fernández Ordóñez argues plausibly that the
end of leísmo in New Castile has to do with the change in organization of
reconquest and repopulation. After 1157, the military-religious orders, such
as Calatrava and Santiago, took over defense and repopulation of the fron-
tier areas. Indeed, Calatrava took its name from the castle of Calatrava just
south of the alfoz of Toledo. Since the military orders operated across
kingdoms, they sought settlers wherever they could find them, producing
even more heterogeneous communities that were not politically dependent
upon any city nor oriented towards urban social norms (as had been the
case earlier in the Toledo phase). The intense demographic and dialect
mixing reached a peak when the Order of Santiago, having lost Cáceres to
the Almohads, transferred its administrative center from Leon to the Castil-
lan castle of Uclés in 1174. Today, the former frontier areas repopulated by
the military orders show use of the etymological system (e.g., La Mancha
and Cuenca).
As effects of koineization, the development of the two types of leísmo
(systems H and E) represent especially interesting cases of interdialect
mixing. In the first {le for count reference), koineization actually favors the
introduction and spread of a marked feature not present in any of the exist-
ing contributing varieties. In the second, koineization favors a reanalysis
based on perceived majority usage. It is important to note, however, that
both reanalyses represented simplifications from the perspective of
speaker-learners, who simply sought ways to accommodate to and repro-
duce perceived majority norms. Once again we find ourselves before re-
sults that might be seen as "deviant" when compared to other varieties of
 Linguistic changes 203

Romance, and yet which are explained as a result of koineization. More-

over, the development of system-Ε leísmo during the Toledo phase seems
to show how different population mixes, differing organization of settle-
ment, and (small) differences in the exact dates of settlement can contrib-
ute to new dialectal variation within the broadly koineizing context.
Finally, the rise of leísmo shows how one surface reanalysis favored by
the koineizing context opened the door to an entire series of subsequent
changes that, over time, result in the modern and "advanced" referential
systems A and B. These changes, as Cuervo suggested long ago, are based
on a series of surface analogies (affecting system H but not system E)
which can be ordered chronologically:

1. Leísmo with count/mass distinction

2. Initiation of plural leísmo
3. Singular laísmo, plural laísmo
4. Regularization of plural leísmo or plural loísmo
5. Singular loísmo for mass reference, neuter loísmo

Though this chronology must be understood as a simplification (it also

differs slightly from that presented in Fernández Ordóñez 2001: 445), it is
consistent with historical and dialectal evidence. Historically, plural leísmo
clearly precedes laísmo as a variable phenomenon, and it is also found
today without laísmo in system T. However, it does not regularize quickly.
As Fernández Ordóñez (1994: 92) indicates, les is favored by its internal
consistency: le is used to refer to count nouns, and plural nouns are by
definition count. On the other hand, los is favored by surface anologies
with plural pronouns ellos, estos, esos, aquellos and the article los. Laísmo
follows plural leísmo. Once le was established for dative and masculine
accusative count reference, it served as a model for learners who began to
use la for feminine accusative count and dative reference. This was quickly
extended to the plural, and became categorical in most referential varieties,
since it was reinforced by all other paradigms (personal ella/s, demonstra-
tive esta/s, esa/s, aquella/s, articles la/s). Today we find variable plural
leísmo with regular singular and plural laísmo in system C. The next step,
regularization of plural leísmo or plural loísmo must have been decided by
other factors. Consistent plural leísmo appears to have spread from Valla-
dolid and Palencia in the north, while consistent plural loísmo developed in
Avila and most referential areas of New Castile. The final stage of singular
loísmo is "complete" only among some users of systems A and B.
204 The Toledo phase

3.3. Reorganization and simplification of the possessive system

During the 12th, 13th and 14th centuries, the possessive forms of early
Castilian suffered significant reorganization (Méndez García de Paredes
1988; Lyons 1993). The changes have generally been understood to have
occurred slowly and gradually, but I argue that they represent cases of
rapid reallocation and simplification due to koineization. Nevertheless, the
changes did not occur at the same rate in all styles, nor can the overall re-
organization of the possessive system be limited to the Toledo phase, since
some elements were not fully reorganized until the Seville phase. Realloca-
tion of forms and simplification of 2nd and 3rd person forms were first
stabilized in popular Toledan speech during the Toledo phase, and final
simplification of the 1st person forms occurred during the Seville phase
(see Chapter 5).
To understand these changes, it is necessary to begin with the basic sys-
tem of possessives used in the Burgos-phase koine. Unlike eastern Iberian
and Gallic varieties, other varieties of Ibero-Romance inherited a fairly
simple and regular system of possessives from Latin. These forms and uses
can be deduced from Menéndez Pidal's analysis of the Latinate documents
of the period. For the early part of the Burgos phase, the most basic system
of forms that can be reconstructed appears in Table 27 under the column
labeled "Basic forms".

Table 27. Possessive pronoun/adjective forms of early Old Spanish. Sources:

Menéndez Pidal (1964: 345-346); Lyons (1993: 216-217).

Person Gender Basic Other Latin

forms forms etyma
1st singular Masc. mió(s) MEU 'my/mine'
Fem. mia(s) mie(s), mi MEA
2nd singular Masc. to(s) túo(s) TUU 'your/yours'
Fern. túa(s) túe(s), tu TUA
1st plural Masc. nuestro(s) NOSTRUM 'our/ours'
Fern. nuestra(s) NOSTRA
2nd plural Masc. vuestro(s) VOSTRU 'your/yours'
Fem. vuestra(s) VOSTRA
3rd singular Masc. so(s) súo(s) suu 'his/his,
and her/hers,
plural Fern. súa(s) súe(s), su SUA their/theirs'
 Linguistic changes 205

Unlike later varieties of Castilian (or even contemporary medieval varieties

of French; Lyons 1993), the basic forms were regularly used as preposed
possessives, with or without a preceding article (e.g., (el) so padre, (los) sos
padres, (la) sua madre, (las) suas ermanas), and almost certainly in all
other functions: as postposed adjectives (e.g., la madre tua, las ermanas
suas, los amigos míos), as pronouns (lo so, los sos, las tuas, las mías) and
as predicates with ser 'to be' (es suo, son tuas).
The historical development of these forms is not difficult to reconstruct.
Nuestro/a/s and vuestro/a/s developed (more or less) regularly from Latin,
and their form has not changed much since the earliest days of the lan-
guage. This is very likely due to the greater phonological substance of
these forms, which are disyllabic and contain spontaneous diphthongs and
consonant clusters. These factors certainly favored consistently tonic ar-
ticulation of nuestro and vuestro, and, by ensuring their perceptual salience
in speech, favored their retention. The other forms suffered more varied
fates. Menéndez Pidal (1962b: 257-258) claims that the feminine forms
mía, túa and súa developed as a result of phonetic dissimilation between
the two original Latin vowels, mió developed as would be expected (see
Chapter 4), and to and so also developed regularly as contractions of the
early Romance results (TÜUM > *[tóo] > to, SLJUM > *[sóo] > so). More
importantly, during the early period of Castilian (exactly when is not
known, but it may have occurred either with Burgos phase koineization or,
more likely, afterwards), there arose innovative masculine forms túo(s) and
súo(s), based on analogy with the feminine etymological forms túa(s) and
súa(s). Other forms which probably developed after the Burgos phase
koine had first stabilized were mie(s), túe(s), súe(s), which, according to
Pidal, suffered weakening and raising of the final vowel, probably when
used as preposed possessives (e.g., súa mádre > súe mádre).120 The forms
míe, tue, and súe then became candidates for the fast-speech phonetic
apocope which developed over time in early Castilian, leading to the for-
mation of mi, tu, su (Menéndez Pidal does not report any examples of the
plurals of these forms). The sum of these changes would have created an
incipient tendency to associate short forms such as mi, to/tu, so/su with use
as preposed possessives, and a concomitant tendency for longer forms to
predominate in the other functions of the possessive. Below we will see
how this played out in the Toledo phase.
The shortening of the feminine forms also removed information that
helped mark gender distinctions in the affected forms. However, in spite of
these changes, gender distinctions were regularly maintained in the Burgos
206 The Toledo phase

phase (Menéndez Pidal 1964: 345-6). Pidal does acknowledge an incipi-

ent, though limited, tendency for so to appear before both feminine and
masculine nouns. There are a few examples of this in the 11th-century
Glosas Silenses: so cosa, so membra. On the other hand, this was held in
abeyance by the continued frequent use of the full-form preposed posses-
sive adjectives, both analogical (e.g., túo, súó) and etymological (tua, stia),
as in: suo filgo, sua/sue parte, suos filios, suas bestituras.m Broader pat-
terns of gender distinction in the language, as well as the effect of Latinate
norms of writing, may also have contributed to maintaining the different
forms. Still, the most significant factor was probably the regular placement
of stress on the possessives, whatever position or function they fulfilled.
Today, stressed articulation of the preposed possessive remains character-
istic of the northern regions of Old Castile, Leon, Asturias and Santander
(Navarro Tomás 1957: 190). The tendency to stress even preposed posses-
sives would hinder the reduction of forms by favoring the continued fre-
quent use of salient variants.
How, then, did the incipient tendency to use to and so for both mascu-
line and feminine arise? No fully satisfying explanation is available, but
Méndez García de Paredes (1988: 535) suggests that in the early stages it
merely represents vacillation in the timbre of an atonic vowel [o~u], a
common enough phenomenon of the medieval period and later. This analy-
sis requires atonic articulation of the preposed possessives, which may well
have occurred variably in some contexts; lack of stress would help to ex-
plain both the tendency to shorten the feminine forms as well as the weak-
ening of the vowel distinctions between to/tu and so/su. Since the distinc-
tion is represented in the orthography, it might seem obvious that there was
a phonemic distinction; however, if these forms lost their tonic articulation,
even if only variably, then they may well have developed articulations that
included anything between [u] and [o]; nevertheless, Latin orthography
forces the selection of a specific vowel grapheme.122
On the other hand, speakers - at least some of the time - may have been
fully aware of the differences in timbre, but they may have come to see súo
and so, túo and to as equivalent forms. If so, they may have also become
accustomed to associating phrasally-syncopated su(o) (as in su(o) ermano)
with the form so, thus contributing to a certain neutralization of the gender
associations. Nevertheless, the documentation examined by Menéndez
Pidal clearly reveals that the dominant tendency was to maintain the dis-
tinction between forms during the Burgos phase, and this tendency was
maintained in the areas around Burgos until the 13th and 14th centuries
 Linguistic changes 207

(see below). Given the great regularity of gender distinctions in the mor-
phology of Romance, it cannot be surprising that the stabilized post-
koineization speech community of Old Castile initially resisted the simpli-
fication of su/so and tu/to.
The system was to undergo significant change, however. The sequence
of changes and its chronology have been outlined by Méndez García de
Paredes (1988), though she is unable to explain the actuation of the
changes, and I believe her chronology needs to be modified to take account
of geographical and social variation. Here I argue that Toledo phase
koineization led to the leveling and functional reallocation of the many
variant forms found in Burgos Castilian by the late 11th century. Funda-
mentally, monosyllabic (unstressed) forms are allocated to the function of
preposed possessive, and longer disyllabic (stressed) forms are allocated to
the other functions of postposed adjective, pronoun, and predicate. More-
over, the distinctive forms so/su and to/tu are regularized and established
as a norm of formal writing (and speech?) in Toledo, while so/su and to/tu
are simplified to gender-neutral su and tu in more popular varieties of
Toledo speech (mi/mio resisted simplification until the Seville phase; see
Chapter 5).
The evidence for these claims is not easy to find, for even early written
documents tended to favor the most conservative norms, and most of those
that have been analyzed by scholars such as Méndez have been literary
texts, which provide little sure information about regional variation. I have
relied principally on two sources. The best source is the early phono-
graphic Romance documents from Menéndez Pidal's Documentos
lingüísticos de España (DLE), of which I have analyzed texts from both
Toledo and Burgos which date from the late 12th century to the year 1250.
The other primary source are texts produced by the Royal Scriptorium of
Alfonso X between 1250 and 1280, which have been transcribed pale-
ographically and transferred to digital format (Kasten, Nitti, and Jonxis-
Henkemans 1997). Though all these texts post-date the actual period of
Toledo phase koineization, I assume that, in general, they conserve norms
established during the early 12th century. Still, as we shall see, the great
majority of texts also reveal the effects of contact between more formal
and informal registers, and between more conservative northern (Burgos)
dialects and more innovative southern (Toledo) dialects.
Evidence of the reallocation and simplification of forms/functions is not
to be had in every text. Even among the Toledo documents of the DLE, it
seems that some documents were prepared either a) in an intentionally
208 The Toledo phase

more conservative style, and/or b) by scribes from more conservative dia-

lect regions. In fact, with respect to possessive use, all texts can be put on a
scale from most conservative to most innovative. For example, DLE docu-
ment 274, prepared in Toledo and dated 1215, can be classified as ultra-
conservative. This text contains the following forms:

— Preposed Masculine: so coment, so poder, sos omnes

— Preposed Feminine: suas aguas, la sua meetad, la sua meetad, suas dere-
churas
— Other uses: son suas, son suas, lo so

This text uses the traditional article + possessive sequence, as in la sua

meetad, it maintains so as the masculine (stressed) pronoun, and it consis-
tently distinguishes masculine from feminine forms, even maintaining an-
cient and disyllabic sua as a preposed form. This use of sua(s) probably
reflects conscious effort to use the most conservative forms available.
More normal conservative usage (at least for Toledo, where sua, sue, suo,
suffer almost complete leveling) tends simply to distinguish masculine so
from feminine su, but to also make regular use of the analogical plurals of
these forms: sos and sus. An example of the Toledan conservative norm is
Doc. 266, dated 1206. The text is from Toledo, but stands out as an excep-
tion among the DLE Toledan documents (but it is similar to conservative
Alfonsine texts; see below). It contains the following forms:

— Masculine preposed: so coment, so coment, so coment, so coment, so

coment, so coment, sos nomres
— Feminine preposed: sus heredades, su generación, su voluntad, sus cartas
— Other uses: lo so

Typical features include the regular distinction of so/su, regular use of

analogical sus, and the use of lo so rather than lo suyo (even so, such use of
lo so tends to be very conservative; see below).
However, other Toledan texts reveal consistent use of an innovative
system that shows the reallocation of monosyllabic/unstressed forms to use
as preposed possessives, and disyllabic/stressed forms for all other func-
tions. For example, document 274, a long text from Toledo dated 1221,
includes the following relevant forms:
 Linguistic changes 209

— 1st person masculine preposed: mio auolo, mio anniuersario, mio fecho,
mios herederos
— 1st person feminine preposed: mi ermana, mi fin
— 1st person other uses: lo mio
— 3rd person masculine preposed: sus aueres, sus sucesores, sus sucesores,
sus sucesores, sus dias
— 3rd person feminine preposed: sus derechuras, sus derechuras, su casa, su
uida, su uida
— 3rd person other uses: de suyo

As is normal at this stage, mio/mi are maintained in contrastive use, but

other pronouns are used as they are today. Su(s) is used indiscriminately
with both masculine and feminine nouns (and, presumably, so is tu(s),
based on the evident analogy). The monosyllabic forms (even mio was
monosyllabic, at least originally) are restricted to use as preposed posses-
sives, and the disyllabic form suyo is assigned to other functions. This text
exemplifies the Toledan popular/innovative norm.
Among the 23 DLE documents for Toledo up to the year 1250 (Docs.
259-1146 to 281-1242), a few, such as 266 and 274, regularly maintain
gender distinctions, but there is significant evidence of reallocation and
reduction of forms in all other texts. First, monosyllabic forms dominate as
preposed possessives. In a few very early documents, one finds simplica-
tion with leveling to so(s) (in apparent parallel to Leonese and Aragonese;
see below). Doc. 262 from the year 1194/1198 includes numerous exam-
ples of exclusive use of so: so termino, so mugger, so uita, toda so uita,
todas sos derechuras, so passamiento. Document 268 from 1210 provides
still further evidence of leveling to so(s): so coment, so filia, so ration, so
padre, so uita, sos pertenenzas, sos rationes, toda so erentia, so ermana, so
mugger, so capelano. A few other documents mix forms, such as Doc. 267
from the year 1207: sos nomnes, sos morauedis, sua uita, sos derechuras,
suas pertenencias, sua heredat, so passamiento, sua uita, so seso. Given
that plenty of other documents show leveling to su(s), such heavy use of
so(s) almost certainly represents a hypercorrection or conservative reaction
to the prevailing popular norm. Since so was probably not used in collo-
quial registers of speech, but was recognized as a conservative form, Tole-
dan scribes preparing formal texts may have used it in an attempt to give
their texts a more ancient appearance and thereby lend the text greater au-
thority. This explanation seems to be confirmed by the coupling of consis-
tent use of so(s) and sua(s) in Doc. 267.
210 The Toledo phase

Leveling to so(s) did not predominate in either the spoken or written

language of Toledo or, for that matter, Burgos, but it was in Toledo that
su(s) first developed as the common form for both masculine and feminine
nouns. The Romance DLE texts from Toledo (1210 to 1250) show over-
generalization of su(s) in nearly every text where possessive forms appear
(except those mentioned above). The following are typical examples: sus
dias (269-1210), sus aueres, sus successores, sus dias (274-1221), todos
sus exidos (276-1228), sus dias (277-1235), su poder, sus molinos, sus
prados, sus alcaydes, su cuerpo (278-1236), sus hermanos, su hereda-
miento, su padre, sus ermanos, su fijo (281-1242). In fact, most of these
documents show exclusive use of su, unlike the Burgos documents for the
same period, which show very strong retention of conservative norms (and
only 3 examples of su + masculine). In all of 47 documents for the same
period from Burgos (Docs. 147-1100 to 194-1245, double the number
from Toledo), gender distinctions are regularly maintained between so/suo
and sua/sue/su, with only scattered examples of confusion of so(s) and
su(s). No Burgos document shows consistent simplification; the few exam-
ples are limited to: sos uinzedades (155-1200), su molinero (161-1209), su
otorgamiento (164-1214), sos casas (178-1227), sos pertenencias (179—
1227, 188-1235), sus términos (192-1241), so mugier (194, 1245).123 In
contrast, the Toledan documents reveal only one example of ultra-
conservative sua after 1215, and there are no examples at all of the inter-
mediate sue. For other functions of the possessive (pronouns, postposed
adjectives, object of preposition, or predicates) both Toledan and Burgos
texts (11th century to 1250) appear to reveal a preference for the longer
forms, but there are few examples. Aside from the ultraconservative Doc.
273, and the conservative Doc. 266, there is no example of so in construc-
tions like lo so. The other examples favor su(y)o: lo suyo (178-1227), lo
suyo (180-1228), lo suo (192-1241), de suyo (274-1221), suya (279-1239).
Clear evidence of the development of a Toledan popular or innovative
norm, and of a Toledan conservative norm (which is nonetheless less con-
servative than that of Burgos) can also be found in the texts of the Al-
fonsine corpus. Although these texts are slightly later than those of the
DLE (those I analyze are all dated between 1250 and 1280) and cannot be
associated with particular places, they are generally very long texts that
contain numerous tokens of possessives, and therefore allow for confirma-
tion of the patterns found in the DLE texts. In fact, the Alfonsine texts con-
tain dramatic evidence that very different geographical and/or stylistic
norms affected the scribes. These texts too can be placed on a scale from
 Linguistic changes 211

most conservative to most popular/innovative in terms of use of posses-

sives.
To do so, I have analyzed the form counts included in the alpha-indices
of each text included in Kasten, Nitti, and Jonxis-Henkemans (1997).
These counts are only rough measures, and certain forms are often used for
non-possessive meanings (e.g., so < SUB 'under', so < SUM, Ί am', tu < TU
'you', tos < TUSSIS 'cough'), but comparison with other forms usually al-
lows confirmation of general tendencies. The counts for each of certain 3rd
person and 2nd person forms are listed in Table 28.

Table 28. Frequency of occurrence of possessive forms in Alfonsine texts. 124

so SOS su sus to tos tu tus lo lo

so suyo
GE4 2454 1565 2276 1294 2 0 19 4 2 7
MOA 261 89 189 130 0 2 0 0 0 0
EE1 605 277 1979 910 28 10 180 14 2 11
GEI 225 77 4072 2534 0 1 1266 194 2 21
LEY 149 7 1076 638 2 0 19 4 3 26
CRZ 10 8 340 260 0 1 115 6 0 0
LAP 28 0 1033 377 0 2 0 0 0 1

First, it is important to note that conservative and innovative texts all re-
spect the allocation of short forms to use as preposed possessives and long
forms for other uses, exemplified here by the neuter pronoun lo suyo. In
addition, the numerous variant forms that long survived in the north are
almost completely absent. For all the texts in Kasten, Nitti, and Jonxis-
Henkemans (1997), the counts for s- forms are: sua 2, sue 8, suo 1 (on the
other hand, all texts, even innovative texts, maintain the mio/mi distinction;
for this reason, these forms are not included in the table).125 So for all
styles of Toledan speech, selection, stabilization, and reallocation of short
forms as preposed possessives is nearly complete, and long forms are allo-
cated to all other functions, with leveling out of many Burgos variants.
Nevertheless, there is also clear evidence of a distinction between a
conservative norm and a more innovative popular norm. The conservative
norm maintains gender-marking distinctions between so/su and to/tu, while
the innovative popular norm levels these distinctions to su and tu. The most
conservative texts are GE4 and MOA, the most innovative are GEI (where
a majority of the so forms are not possessives), LEY, CRZ, and LAP. EE1
212 The Toledo phase

represents a sort of mixed style, where significant effort is made to respect

the conservative norm, but popular usage predominates (but recall that
different sections were clearly prepared by different scribes; a more de-
tailed analysis is needed). It is important to note that the relative dating of
these texts is insignificant: the most innovative texts are among the oldest;
we can conclude that scribes of the innovative texts were responding pri-
marily to their own dialectal norms, though source texts and competing
norms certainly exercise a secondary influence. It is also important to rec-
ognize that even in a very conservative text such as GE4, usage of so and
su is not absolutely correct. For example, in GE4, usage correctly follows
the conservative norms during the first 9 folios (e.g., so nombre & su uidä),
but after that sporadic cases of confusion crop up: este su fijo, con su sen-
nor, su padre, su regno, so bondad, so poridad. These "errors" indicate the
constant tendency of the popular norm to affect the scribes' use of formal
language. In EE1, this tendency is even more evident, and confusion is
rampant, though the clear preference is for su. For example, there are 24
tokens of so padre, 59 of su padre, 1 of so madre, and 30 of su madre. In
the most innovative texts, so and sos are almost completely absent, and,
with the exception of the mio/mi distinction, we find a system very like that
used today.
How did this system develop? As changes occurred slowly in and
around Burgos in the late 10th and 11th centuries, analogical masculine
forms tuo, suo appeared, and short forms of singular feminine possessives
began to appear (mi, tu, su). These forms were the result of final-vowel
weakening and phonetic apocope, which was favored by novel but variable
atonic articulations of preposed possessives. During Toledo phase koinei-
zation, speaker-learners encountered significant use of short monosyllabic
forms as preposed possessives (mi, monosyllabic mio, tu, to, su, so) and
longer forms in other functions/contexts, which were normally stressed (lo
súo, es túa/e, la casa mía). Speaker-learners overgeneralized the use of the
forms mi, tu, su to produce plural mis, tus, sus, adding still more short
forms for preposed possessives. In addition, the use of preposed posses-
sives without articles grew in frequency (la mie madre > mi madré). Lyons
(1993) suggests that numerous atonic preposed possessive adjectives were
reanalyzed by speakers as determiners/specifiers, on the model of the
atonic articles. Such reanalysis favored even more consistent use of short
atonic possessives when preposed to nouns, which carried the primary
stress accent of the resulting noun phrase.126 At the same time, the disyl-
labicity of the stressed forms tuo and suo was reinforced with the insertion
 Linguistic changes 213

of an intervening glide: tuyo, suyo and related forms (the earliest attesta-
tions of suyo are from the 12th century). As has often been suggested, this
was favored by analogy with the form cuyo < CUIUS 'whose', which paired
pragmatically with possessives (e.g., question: ¿cúyo esΡ; answer: es suyo).
Such analogical patterning is strongly favored in koineizing contexts. At
the same time, the many variant forms such as tue, tua, sue, sua, were
largely rejected, and so, to were generally dispreferred for most functions
of the possessives.
So and to were left in only one context/function: atonic and preposed,
where they were supposed to contrast with su and tu. However, the increas-
ingly unstressed articulation of these forms tended to weaken the phonemic
opposition. The tendency for unstressed vowels to raise and form diph-
thongs with following vowels obscured input for learners. In many pre-
vocalic contexts, so was articulated as [sw], as in so amigo, and this [sw]
could be (mis)interpreted by adult and child learners as a realization of an
underlying form /su/ (rather than /so/, the correct option in earlier stages).
Moreover, the form so, which contains an apparent masculine morpheme -
o, might have been perceived as more marked than su, which had no overt
gender marking. For generic use, the less marked form may have been pre-
ferred. Finally, and most importantly, morphological analogy with the
stressed forms suyo, tuyo, certainly influenced speaker-learners, who
would have sought transparent and regular relations between the different
possessive forms and functions. Modeling the short atonic forms on the
first and stressed syllables of the tonic suyo and tuyo highlighted those
relations. As speaker-learners began to base their own production on incor-
rect abductions about the functions of su and tu, they increased the fre-
quency and consistency of occurrence of su for all speakers attempting to
learn the community norms (some of these causal factors are also discussed
in Méndez García de Paredes 1988: 536; Penny 1991a: 127).127
Thus, it is likely that koineization played a key role in the definition of
the Spanish possessive system and the timing of its development. In this
reorganization we see the interaction of leveling, functional (grammatical)
reallocation, and simplification. The development of the possessives during
the Toledo phase also reveals clear stylistic reallocation of variants, along
the lines of that of extreme apocope.
214 The Toledo phase

4. Conclusion

Though the expansion of Castile into Toledo is not normally considered

to have had an impact on the history of Castilian, there is significant evi-
dence for making such a claim. The late 11th and 12th centuries show a
renewed and enlarged process of demographic and dialect mixture within
Toledo (or New Castile) in particular and the Peninsula in general. In New
Castile, this mixing was greater than in the Burgos phase, with the novel
elements of francos, Mozarabs, Mudejars, and colonists from a greater
variety of regions (though Castilians dominated in most central regions).
As in the Burgos phase, there occurred a breakdown of social networks and
norm enforcement among newcomers, at the same time that linguistic
variation peaked. Three significant linguistic effects have been identified:
the establishment of extreme apocope as a native Castilian norm through
reallocation as a stylistic variant, apparently in most parts of Castile; the
reorganization of the possessive system, a process that was first completed
in the region around Toledo and only later spread north to Burgos and Old
Castile; and the extension of Burgos-phase leísmo to much of New Castile,
but also its reanalysis and simplification in some areas of New Castile. The
discussion of leísmo also involved exploration of its origins in the Burgos
phase; this change shows that an apparently marked feature in the resultant
koine can arise from speaker-learners' attempts to make sense of variable
and opaque input.
A significant feature of Toledo-phase rekoineization is its occurrence in
many cities at once, and its contribution to the spread of Castilian features
(those of the Burgos phase) to lateral varieties and the spread of new fea-
tures from the south (Toledo and New Castile) to the north (Burgos and
Old Castile). This spread, in turn, probably served to accentuate the effects
of koineization. Nevertheless, the period is also marked by the develop-
ment of internal dialect divisions, partly between North and South, since
the North adopted some changes more slowly (as indicated by early 13th-
century differences in use of the new possessive system), but also new
dialect divisions within New Castile due to different demographic mixes,
timing, and organization of settlement (seen in the differences between
pronouns systems H and E, and between these and the etymological sys-
tem). Finally, it should be noted that the resultant koine of the Toledo
phase is marked not only be geographic variation, but also by significant
social/stylistic variation, a fact which may be due to the greater social
class/caste distinctions which characterized this era and area.
Chapter 5
The Seville phase

The reconquest and repopulation of Betic Andalusia in the 13th century

represents the last great phase of population movement in Castile begun
and completed during the medieval period. As in preceding chapters, I first
address the social and demographic history of this period in order to iden-
tify the sources of variation in the new koineizing environment. Unlike the
earlier Toledo phase, language historians have commented on the possible
linguistic consequences of the original repopulation of Betic Andalusia,
though most of this attention has been recent and no consensus has yet
been reached. Some see this phase as representing the mere transplantation
of Castilian to the south while others view it as marking the very origin of
Andalusian as a distinct dialect. I therefore examine this recent debate and
how the model of koineization as here defined and applied may impact
upon it. From there, I move to offer evidence for three linguistic changes
that can reasonably be attributed to 13th-century koineization in Andalusia.
Finally, I discuss seseo, a typical feature of modern Andalusian which has
been attributed to 13th-century dialect mixing, show that this is unlikely,
and consider the value of this negative evidence in helping understand the
limits and usefulness of the model of koineization.

1. Social history

By the late 12th century the Toledo phase of reconquest and repopulation
had slowed. Though military activity continued all along the frontier, the
Guadiana valley had clearly become a stable dividing line between Chris-
tian north and Moorish south. In fact, the Moors made important incursions
into Christian territory later in this century, including, for example, their
victory at the battle of Alarcos in 1195. Still, this was their last great vic-
tory, and the later defeat by the Christians at Las Navas de Tolosa in 1212
marked a decisive reversal in their fortune. As the Almohads fell into civil
war, the Christians reinitiated their advance. This third phase of the recon-
quest, in which the Christians eventually conquered the whole of the valley
216 The Seville phase

of the Guadalquivir - Betic Andalusia - along with Murcia in the east, took
place from the 1220s up to 1264.128
Fernando ΙΠ of Castile initiated this phase of the reconquest as he
pushed south of the Sierra Morena into the Moorish kingdom of Jaén, tak-
ing and repopulating several strongpoints along the upper Guadalquivir
valley: Andújar in 1224, Martos in 1225, and Baeza in 1226. In 1230, Fer-
nando ΙΠ inherited the crown of Leon, which allowed him to combine the
might of both kingdoms. Before this, however, Alfonso IX of Leon had
already taken Valencia de Alcántara (1221), Cáceres (1227), as well as
Mérida and Badajoz just before his death in 1230. Fernando added to these
Extremaduran advances with the victories of Trujillo (1233) and Medellin
(1234). In Jaén, the Archbishop of Toledo took Cazorla in 1231, while the
King took the city and surrounding lands of Ubeda in 1233. However,
these successes paled in comparison to the taking in 1236 of Córdoba, for-
mer capital of the Omayyad Caliphate. From Córdoba, Christian forces
advanced both east and west. In 1243, Fernando sent his his son and heir,
Alfonso, to negotiate the capitulation of Murcia, where the local king de-
clared himself a vassal of Fernando and allowed the placement of Castilian
garrisons in the main fortresses of the kingdom. In 1246 the city of Jaén
itself, one of the best fortified of the Andalusian cities, was added to the
list of Christian victories and left Fernando in control of the entire upper
part of the Guadalquivir valley, which could then provide a secure base for
the siege of Seville, the economic and cultural capital of the region. When
its Moorish inhabitants finally surrendered in 1248, other cities belonging
to the kingdom of Seville - equally exposed to attack along the plain of the
river valley - ceased to resist, and the kingdom of Niebla (in modern
Huelva) became a vassal state of the Castilian king.
For a short time the reconquest slowed, as the victors went about con-
solidating and repopulating their newly gained territory. However, in 1262
Alfonso X took direct control of the kingdom of Niebla and in the follow-
ing year initiated repopulation of Cádiz, closing off access to the sea from
Jerez and other Mudejar cities just to the north. These actions contributed
to the Mudejar Revolt of 1264, in which the remaining Moors in Betic
Andalusia and Murcia rose up against their Christian overlords. Alfonso X
quelled this uprising, then expelled the Moors from most of Andalusia and
initiated repopulation of the newly vacated zones. The limits of the terri-
tory conquered up to this point would remain roughly the same until the
15th century, though the Christian-Moorish frontier shifted slightly as the
 Social history 217

Christians took and repopulated Tarifa (1292), Gibraltar (1309), Algeciras

(1344), and Antequera (1410).
The conquest of these new territories in so short a time necessitated yet
another massive repopulation effort. This was focused, as in the Toledo
phase, on large urban centers. Indeed, the newly conquered and vacated
cities were so large that they were difficult to repopulate. The concejos of
the main cities (Jaén, Córdoba, Seville) administered a surrounding alfoz,
in which they controlled the founding and repopulation of surrounding
villages. In this sense and others, Andalusian repopulation simply contin-
ued the traditions developed in the Toledo phase, but the size and speed of
the needed repopulation required the development of new procedures:
Resultaba claro que ya en el siglo XIII no cabía aplicar las viejas formas de
la presura castellano-leonesa o de la aprisio catalana, realizadas en buena
parte de manera espontánea e informal y cuando lo era dirigida, se efectuaba
ordinariamente con volumen de población reducido y con objetivos muy lo-
cales y predominantemente dentro de un ámbito rural y en tierra habitual-
mente abandonada, donde se efectuaban asentamientos de índole en buena
parte familiar. Tampoco las primitivas cartas-pueblas aparecían ya suficien-
tes por sí solas para las nuevas necesidades, y ni aun siquiera los propios
fueros municipales, puesto que las primeras resultaban fórmulas lentas para
la atracción de pobladores y con objetivos modestos, en tanto que los se-
gundos se hallaban encaminados fundamentalmente a regir la organización y
la vida ciudadanas una vez establecida la población cristiana que acudiera
para establecerse en las tierras meridionales. (Moxó 1979: 396)

In order to facilitate the new round of repopulation, the Castilian/Leonese

monarchy organized repartimientos, in which all properties abandoned by
the Moors were inventoried and reassigned to new settlers. This was coor-
dinated through commissions named by the king (known as the junta de
partidores in Seville). This close control of the distribution of property
heightened the official nature of the repopulation by allowing the monar-
chy to determine who would receive rewards of land. Fernando ΠΙ ordered
his first repartimiento in 1224, when he took Andújar. Afterwards, nearly
all the larger towns, such as Baeza, Jaén, Córdoba, Seville, Jerez and
Cádiz, received fuero and repartimiento (Moxó 1979: 353). This system
was so successful that it was also used in Murcia and Lorca, as well as
outside of Castile in newly reconquered Valencia (1238).
One might assume that the new system of repartimientos prevented so-
cial mobility in Andalusia (and thus would have retarded or limited the
effects of koineization). In Seville, for example, the upper nobility or ricos
218 The Seville phase

omnes (including members of the royal family), royal officials, and various
branches of the Church, received some of the largest and best properties,
generally in the form of donadíos (González 1951: 257; González Jiménez
and González Gómez 1980: xxx).129 Many of these lands were those most
in need of repopulation, and they were given out as donadíos in the hope
that their recipients would oversee their repopulation. However, many ri-
cos omnes never actually resided in Andalusia, and some eventually sold
their properties to the new local nobility (González 1951: 323), known as
caballeros de linaje (in Seville). These were nobles of originally lesser
rank who received larger properties in the repartimiento, but who were
obliged to provide military service and to reside in the city. The families of
these caballeros de linaje were destined to dominate the social hierarchy in
Andalusia, and, indeed, the powerful Guzmán family of Seville had its
roots in this group.
The military concerns of this frontier society were especially significant
in determining social status, which was often assigned not on the basis of
previous or existing status but rather on ability to contribute military ser-
vice. A very clear example can be found in the caballeros villanos (or ciu-
dadanos), who, though non-noble, formed the most privileged group
among the majority of the population, and whose privileges were directly
linked to their military capacities. The importance of this and other newly-
created groups is patent in the Repartimiento de Jerez (a frontier city con-
stantly exposed to attack), in which the most important social distinctions
are cast in socio-military terms. The relative importance of these newly-
privileged groups can be seen in Table 29.

Table 29. Social Structure of the Population of Jerez. Source: González Jiménez
and González Gómez (1980: liv).

Social Category Number % of total

Caballeros del feudo (linaje) 42 2.3
Caballeros ciudadanos 212 11.6
Other military groups: 109 5.85
Peones 1467 80.25

In effect, this widespread and rapid (re)assignment of social status to sig-

nificant numbers of settlers represents a tremendous fluidity of social class
at the beginning of repopulation (though paradoxically it would lead to a
rigid stratification). Given its occurrence at this initial phase, it would have
 Social history 219

contributed to the spread of koineizing linguistic changes throughout the

social hierarchy in spite of subsequent stratification.
Few of the libros de repartimiento have survived, but those that do,
such as those of Seville and Jerez de la Frontera, are rich sources of infor-
mation on the varied geographical origins of the first settlers, most proba-
bly the primary source of variation. Julio González' (1951) study of the
Repartimiento de Sevilla, considered one of the best studies on the Andalu-
sian repopulation, includes a detailed description of the origins of the set-
tlers. As one might expect, the early Christian inhabitants of Seville were
of quite varied origin. Still, the exact nature of the mixing was different
from that of the earlier Toledo phase. The Moorish inhabitants of the city
were exiled, being provided passage to Africa or allowed to move south to
Jerez or east to Granada (González 1951: 308). Elsewhere in Andalusia,
areas that were forced to capitulate through military actions were also emp-
tied of their Moorish inhabitants. Only in those areas which were acquired
through pacts - such as La Campiña of Córdoba, or regions south and west
of Seville - were the Moors allowed, at first, to remain (González Jiménez
1982: 131). However, after the Mudejar Revolt of 1264, nearly all the
Mudejars abandoned Andalusia for Granada or North Africa (only in
Murcia did a sizable Mudejar population remain). Still, a few Moors -
probably those that had served the king in some way and were therefore
referred to as moros del rey in the repartimientos - were allowed by the
king to remain in the new Christian communities of Andalusia (González
Jiménez and González Gómez 1980: lxxxi). In addition, a small morería
appeared in Seville, probably due to the extended contact and trade with
neighboring Granada (González 1951: 310).
Another factor which differentiated Andalusian repopulation was the
lack of surviving Mozarab and Jewish communities in Almohad Seville.
Gonzalez (1951: 303-305) finds no textual evidence whatsoever to indicate
the survival of a Mozarab community in Seville at the time of its defeat.
This is not entirely surprising, since the Almohads had decreed their expul-
sion under pain of death. The Jews had also been expelled from the city by
this time (1951: 311), though a new judería appeared after the reconquest.
According to González, the absence of Mozarabs and Jews was also char-
acteristic of other Andalusian regions reconquered in the 13th century
(1951: 308). Once again, however, Murcia was different, in that it retained
ancient juderías that remained isolated from the rest of society until the
time of the expulsion in 1492 (Μοχό 1979: 373).
220 The Seville phase

Even without these groups, however, the size of Seville and the forced
exit of its entire Moorish population must have made it the single largest
repopulation project that the Christians had ever undertaken. As was the
case of Toledo, much of the new population was made up of settlers from
nearly all of the northern regions. González (1951) was able to determine
the primary areas of origin of the settlers in Seville by analyzing indica-
tions of place of origin accompanying the surnames of beneficiaries listed
in the repartimiento. He found that Leonese Extremadura and Valencia
were weakly represented, since both had recently been reconquered and
were themselves in need of new inhabitants. The core group of settlers
were from the region centered on the cities of Burgos, Palencia and Valla-
dolid, in Old Castile. Leonese and Galicians were also prevalent, though
González warns: "Hay que tener en cuenta que en el repartimiento a veces
se llaman gallegos a los del reino de León" (1951: 321). Navarrese - lack-
ing new territories in which to expand - also came in significant numbers.
Many more Catalans than Aragonese (limited to a small group of nobles)
appear to have participated in the repopulation (1951: 317). There were
significant numbers of settlers from New Castile (Toledo and the more
recently repopulated Cuenca, Huete and Alcaraz), as well as previously
reconquered areas of Andalusia itself. Some areas of the city were charac-
terized by a heavy predominance of colonists from particular areas, as was
the case of the barrio de la mar, where many natives of Santander (Can-
tabria) settled (Μοχό 1979: 361). The francos were also represented in the
new mix, though far less significant in terms of absolute numbers and rela-
tive influence. On the other hand, the repopulation of Seville also included
at least one group that had been absent from Toledo: Italians, especially
Genoese, Lombards and Pisans, who came to set up commercial ventures
(González 1951: 312).
The mixing found in Seville also seems to have characterized smaller
communities, though in each community the proportions from different
regions were different. Jerez, for example, underwent repopulation when
the Mudejars were forced to abandon the city after the uprising of 1264.
Alfonso then ordered the city's repopulation and the preparation of a Libro
de repartimiento. In their study of this text, González Jiménez and Gon-
zález Gómez (1980) prepared a quantified analysis of the origins of its
earliest inhabitants, basing their counts on surnames and toponyms used as
surnames (see Table 30).
 Social history 221

Table 30. Origins of the Settlers of Jerez de la Frontera. Source: González Jiménez
and González Gómez (1980: xlvii).
Region Number of settlers % of total
Old Castile 307 30.27
Leon 155 15.28
New Castile 137 13.57
Galicia 82 8.08
Extremadura 61 6.01
Catalonia 44 4.33
Jaén 42 4.14
Navarre 34 3.35
Portugal 30 2.95
Aragon 27 2.66
Seville 25 2.46
Basque Country 22 2.16
Córdoba 20 1.97
Asturias 13 1.28
France 7 0.69
Valencia 4 0.39
Italy 3 0.29
England 1 0.09

As in the case of Seville, the Castillans form a sizable block, but they rep-
resent less than half the total population. There are some differences in the
proportions for Seville and Jerez; in Jerez, New Castile and Extremadura
appear more heavily represented, while the non-Iberian presence is practi-
cally nil, as we might expect in this much smaller border town. A few
Moorish families were included in the repartimiento, though they were
apparently new arrivals under royal protection. Jews of northern origin also
established a judería in the town.
Mixing of this sort must have occurred throughout Andalusia, though in
each locale there may have been particularly large or small contingents of
certain groups. This appears to have been the case of Cádiz:
Parece que el mayor número de los repobladores gaditanos procedió de las
villas cántabras -Laredo, Santander, San Vicente de la Barquera y Castro
Urdíales- y Vizcaya, pero aunque prevaleciera este contingente norteño -al
que hay que agregar un cierto número de gallegos y astures- encontramos
también en buena medida castellanos e incluso algunas gentes de la recién
repoblada Andalucía. No faltaron tampoco individuos de fuera de la Corona
de Castilla, especialmente francos, catalanes y genoveses.
(Moxó 1979: 364)
222 The Seville phase

There were also small towns that in their beginnings were settled by fairly
homogeneous groups. For example, Alfonso X gave land grants to 100
Catalan ballesteros in Camas (Seville) and 150 omes de Catalunna in Co-
ria del Río (González Jiménez 1988: 48). As in early stages of the recon-
quest, placenames reveal a certain regional dominance in some newly
founded towns. Alvar (1979: 1870), for example, reports 5 towns in Huelva
which include de León in their names, and a Zamoranos in Córdoba.
Even though the main effort of reconquest and repopulation ceased fol-
lowing the 1260s, Andalusia was not left fully populated. Enormous areas
along the frontier and even in the interior remained empty or underpopu-
lated. Indeed, many of the earliest settlers abandoned their properties dur-
ing the second half of the 13th century and returned to northern homes. As
a result, efforts at repopulation did not cease entirely, continuing weakly
into the 14th century. These efforts were divided between the internal rural
zones, where the great nobles and the church especially sought more peas-
ants to farm the land, and the frontier, where the military orders and the
Crown sought more residents to defend the borders. However, most of
those who participated in this repopulation appear to have been peones
already settled in Andalusia, particularly Seville. While this later phase of
repopulation did not really increase the overall population of the region, it
did maintain a certain degree of demographic movement within Andalusia
(González Jiménez 1975: 41-47) and perhaps extend the influence of the
speech then developing in key urban centers, most notably Seville, and
perhaps Córdoba as well.

2. Linguistic changes

Given the kind of demographic and dialectal mixing that clearly occurred
in 13th-century Andalusia, one must assume that koineization did in fact
occur. The results of this koineization may logically have been of two
kinds: changes which were restricted to the repopulated region and which
became characteristic features of the newly-formed dialect, and changes
which also affected the dialects of Castilian to the north through the proc-
ess of language spread and dialect leveling. Numerous scholars have men-
tioned or discussed the first (though frequently only with a tangential inter-
est, since only recently has medieval Andalusian come to be considered an
object itself worthy of study), while the second of these possibilities has
rarely been considered.130
 Linguistic changes 223

2.1. Previous discussion of the origins of Andalusian

The first serious discussion of the dating of the typical changes of Andalu-
sian arose as a result of the famous polemic on the origin of American
Spanish between Max Wagner (1920), on the one hand, and Amado
Alonso (e.g., 1953) and Henríquez Ureña (e.g., 1932), on the other. Wag-
ner held that the base of American Spanish was the Andalusian dialect;
Alonso and Henríquez Ureña defended the separate development of
American Spanish, and Alonso in particular tended to date the characteris-
tic changes of Andalusian to periods following the founding of the Ameri-
can colonies. But Alonso also proposed that it was a process of linguistic
leveling (nivelación) that led to the unique development of both Andalu-
sian and the American Spanish dialects. For him, leveling seems to have
implied mixing and homogeneization - concepts that are certainly funda-
mental to koineization - but did not necessarily include a notion of simpli-
fication. Interestingly, he indicates that such leveling had also occurred in
Betic Andalusian:
En realidad, los andaluces anticiparon tres siglos la misma operación lin-
güística niveladora que estamos explicando en los americanos: después de
reconquistada Andalucía se pobló con castellanos, leoneses, asturianos y
gallegos. El castellano dio el tono (o fue la base, según la tradicional ima-
gen), por análogas razones que en América. (Alonso 1953: 54-55)

This, of course, is consonant with the belief in the separate but similar
"polygenesis" of American and Andalusian, though the early dating im-
plied for changes in Andalusian is surprising within the general context of
his work. In another discussion on what he considered the early merger of
/z/ and /d z / and the supposed later extension of this merger to /s/ and /ts/,131
he refers to leveling once again, this time even briefly referring to proc-
esses like leveling and simplification:
Y esto ocurrió con mucha más profundidad y extensión en aquellas regiones
donde el castellano había sido transplantado: en el siglo XIII, a la Andalucía
occidental y Córdoba; en los siglos XIV y XV, al resto de Andalucía y a
Canarias . . . a América y Filipinas. En estas regiones el castellano no sólo
fue transplantado con la reconquista o la conquista, sino que los repo-
bladores o colonizadores procedían de regiones dialectalmente heterogéneas
y tuvieron que llegar a una nueva unidad idiomàtica por nivelación de las
diferencias, con renuncias y aceptaciones, si bien siempre a base del 'castel-
lano' oficial al que todos tendían. La formación de conglomerados humanos,
de componentes dialectalmente heterogéneos, tiene entre otros el efecto de
224 The Seville phase

aflojar o de agravar la flojedad de ciertos resortes lingüísticos hasta enton-

ces funcionantes, una especie de renuncia económica a ciertos 'primores'
del sistema: en las nuevas condiciones sociales se ven como de lujo y pre-
scindibles signos, funciones, valores y oposiciones que en la homogénea
área originaria se tiende a guardar como naturales. En este sentido Andalu-
cía tiene historia ejemplar, porque no sólo fue repoblada en los siglos XIII
al XV con gentes de distintas regiones norteñas, sino que en los siglos XV y
XVI otra vez se vio excepcionalmente agitada y demográficamente revuelta,
primero por los ejércitos españoles que acudieron a la conquista de los rei-
nos de Málaga, Almería y Granada; después con la formación de los ejérci-
tos del Gran Capitán, y sobre todo con la extraordinaria afluencia de espa-
ñoles de todas partes y de movimiento de andaluces cuando el descubri-
miento, las guerras de conquista y las mareas de la colonización de las
Indias hicieron de Sevilla la ciudad de más atracción y de más brillo de toda
España y el corazón de aquel movimiento de poblaciones.
(Alonso 1969: 108-109)

Though issue may be taken with Alonso's view of continuous repopulation

from the 13th to the 15th centuries, this passage does make the very impor-
tant point that Andalusia was affected by large-scale demographic mixing
not only in the 13th century but also in the late 15th and 16th centuries.
Alonso, however, could not afford to continue the parallel between the
origins of Andalusian and American Spanish, since one of his most pre-
cious theories was the separate and unique development of American Span-
ish. If Andalusian had completed many of the mergers that its shares with
American Spanish before the founding of the American colonies, Alonso's
anti-Andalusian conclusions would have proved unsustainable. He was
thus forced to backtrack with regard to the supposed effects of 13th-
century repopulation:

Los dialectalismos son mucho más difíciles de rastrear porque en el siglo

XVI los dialectos andaluces no habían desarrollado aún, o habían desarrol-
lado muy poco, los rasgos peculiares que hoy los separan del castellano . . .
[in footnote] En fonética sólo el seseo de la -z y de la -z-, no aún el de la c,
hasta entrado el siglo XVI; luego el trueque anárquico c-s, que arrastra en
seguida a la z, y que produce el ceceo en Andalucía la Baja al cabo de larga
gestación. (Alonso 1953: 57-58)

Though this statement would appear to allow some limited effects of 13th-
century mixing (which remain unspecified), we know that Alonso associ-
ates even these few phonetic changes with the 15th century at the earliest
(Granda 1994: 93).
 Linguistic changes 225

Alonso, as he himself made clear, did not have access to medieval An-
dalusian manuscripts or to paléographie transcriptions of such manuscripts
(Alonso 1969: 79). This had serious implications for his theories, since
these were based almost exclusively on evidence obtained from the less
precise transcriptions done by and for historians. As a result, numerous
linguists have questioned Alonso's stance on quite a range of issues, not
least among them the dating and sequencing of the changes that character-
ize Andalusian. Even in the 1950s, Catalán (1956) and Lapesa ([1957]
1985) began to question the late dating assigned by Alonso to the changes
of Andalusian, pushing the earliest dates for such phenomena as seseo (or,
in the terms of Catalán, çeçeo-zezeo) as early as the 15th century. Mené-
ndez Pidal (1962c) also sought evidence for typical changes (such as seseo)
in the 15th century, and argued for the existence of an estado latente in
which changes of this sort were incubated for centuries before fully devel-
oping in the 16th century.
In recent years, however, a polarization has developed in the dating as-
signed to many Andalusian phenomena,132 pushing the assigned dates to
much earlier periods or to much later periods than those previously as-
signed. Perhaps no one has assigned later dates to the general rise of Anda-
lusian features than the Granadan scholar José Mondéjar, who states:
Puesto que la mayoría de los rasgos señalados desde el XIV hasta el XX han
sido comunes al castellano (lengua medieval) y al español (lengua clásica y
actual) de Andalucía y del resto de España, y puesto que los más caracterís-
ticos y diferenciadores son los aparecidos entre los siglos XVIII y XIX, creo
que sólo es histórica y estructuralmente posible hablar de andaluz o de
hablas andaluzas a partir del siglo XVIII. (Mondéjar 1991c: 231)

In line with this late dating, Mondéjar clearly views the 13th-century re-
population as constituting no more than a simple extension of northern
Castilian:
Todos los rasgos que presentan los documentos del castellano medieval, en
la Andalucía del siglo XIII, son los mismos que se encuentran en otras re-
giones del dominio lingüístico peninsular. (Mondéjar 1991c: 227)

Although it seems likely that most of the features of the early Andalusian
koine were found (at least as incipient phenomena) in other contributing
varieties, it is not true that no changes occurred in Andalusia, or that no
differences developed between the new Andalusian koine and northern
(varieties of) Castilian (see below). Indeed, Mondéjar assumes that north-
ern Castilian was a homogeneous and uniform variety, capable of simply
226 The Seville phase

being transplanted to the south. Moreover, he offers no evidence for his

claim.
As we will see in the discussion of seseo, Mondéjar is not totally con-
sistent in his own claims, but the forcefulness of his assertions have led to
the articulation of an antithetical view, according to which the very origin
of Andalusian is to be found in its 13th-century repopulation. The leading
advocate of this theory is Juan Antonio Frago Gracia, who, from the 1980s
and into the 1990s devoted the bulk of his efforts to the study of the history
of Andalusian (and American) Spanish, particularly as revealed through
philological study of surviving medieval manuscripts. This effort led to a
long series of articles (e.g., 1983, 1984, 1989, 1991b), culminating in his
synthesis and magnum opus, Historia de las hablas andaluzas (Frago Gra-
cia 1993). He explained the origin of his thinking on this matter in the fol-
lowing passage:
Pues bien, son ya diez años bien cumplidos los que llevo a vueltas con el
acervo documental de Andalucía, y desde los primeros contactos con estos
corpus escritos y con los indicios lingüísticos que de ellos emanaban se fue
alumbrando en mí la sospecha de que los distintos modos de hablar andalu-
ces tenían una antigüedad mucho mayor de la que solía suponérseles, sospe-
cha que, conforme se hacía más acendrado e intenso mi trato con los viejos
textos del mediodía español, se ha ido convirtiendo en afirmación cada vez
más rotunda, lo cual no es sino el resultado de reposadas reflexiones dentro
de una concepción historicista, cuyo horizonte he procurado ensanchar y
profundizar lo más que he podido. (Frago Gracia 1993: 7)
Frago Gracia's insistence on the early origin of a broad spectrum of fea-
tures associated with Andalusian goes far beyond the claims of any other
scholar who has written on the topic (and he has continued to maintain
them; cf. Frago Gracia 1999: 65). His fundamental thesis in this regard is
more clearly defined in the following passage:
Indudablemente, es el habla de la Andalucía primeramente reconquistada -
vista en su conjunto y con sus variedades internas, y no por referencia a la
de Sevilla en particular o en exclusiva- el fundamento del dialecto que en el
cuatrocientos se extendería por el reino de Granada en competencia con
otras modalidades castellanas. Por consiguiente, a este inicial arranque dia-
lectal es al que obligadamente se debe acudir en los comienzos de la indaga-
ción científica, a fin de determinar en la medida de lo posible el surgimiento
de los distintos rasgos del peculiarismo andaluz, así como el proceso de afi-
anzamiento social. Parece lógico pensar, pues, que si el andalucismo ter-
mina triunfando con la totalidad de sus modismos en buena parte del territo-
 Linguistic changes 227

rio nazarí tomado en el siglo XV, y con varios de ellos en la completa exten-
sión de ese dominio oriental -algo en cierto modo similar ocurriría en la
siguiente centuria en la colonización de las Indias - , es porque tal diferen-
ciación dialectal se había producido antes en la vieja Andalucía cristiana.
(Frago Gracia 1993: 68)

While issue must be taken with some of the assumptions revealed here (see
below and Chapter 6),133 Frago's emphasis on the history of the medieval
language in Andalusia is laudable, for until recently this was an almost
untouched domain. Lapesa (1981), for example, does not even mention the
language of Andalusia in the 13th century, focusing exclusively on the
language of the Alfonsine corpus. More importantly, Frago's fundamental
hypothesis is thoroughly plausible within the model of koineization.
Though he does not mention koineization as such, he attributes the early
development of Andalusian features in the valley of the Guadalquivir to
external causes: the process of demographic mixing and concomitant dia-
lect mixing and leveling that he assumes must have occurred. Also, like
Amado Alonso before him, he indicates the importance of at least two
great periods of movement:
Si en el Nuevo Mundo americano el español moderno experimentó una ni-
velación sobre la base lingüística de variantes peninsulares entrecruzadas,
las que en sus hablares llevaban emigrados salidos de todos los rincones de
España, este fenómeno, que inevitablemente acompaña a una lengua de co-
lonización, se había producido antes en Andalucía, y repetidamente, ade-
más, en dos sincronías perfectamente marcadas por los momentos claves de
la expansión castellana en el sur de la Península, a saber, el siglo XIII,
cuando el primer impulso guerrero arrebata a los musulmanes el señorío de
la Andalucía bética y el XV, sobre todo en su último tramo . . . Una y otra
situación traería consigo la mezcla de gentes venidas de muy diversas par-
tes, con modos de hablar en mayor o menor medida, según los casos, diver-
gentes, y en ocasiones con lenguas distintas, lo cual traería consigo una serie
de entrecruzamientos culturales y de compromisos lingüísticos.
(Frago Gracia 1993: 55)

An even-handed consideration of the two periods of demographic move-

ment in Andalusia is essential (as a minimum) to an understanding of lin-
guistic change that occurred there. Despite paying lip-service to this impor-
tant axiom, Frago Gracia again and again returns to the 13th-century
repopulation as the source of nearly all the characteristic features of Anda-
lusian:
228 The Seville phase

Puede decirse sin temor a exagerar que el andaluz ha sido el más acendrado
crisol de las distintas variedades que el castellano medieval tuvo . . . Efecti-
vamente, a los territorios andaluces reconquistados en el siglo XIII afluyen
castellanohablantes de todas las procedencias, pero también asturiano-
leoneses y gallegos, así como, aunque en menor medida, vascongados y
navarros, aragoneses y catalanes, junto al aporte demográfico de portugue-
ses, franceses o genoveses, entre otras partidas de extranjeros. Y todos estos
elementos no castellanos, peninsulares y ultrapirenaicos, de aquella naciente
sociedad meridional en su seno hubieron de castellanizarse y de alguna
manera tendrían que influir en la germinación dialectal muy pronto
planteada en la novísima Castilla. (Frago Gracia 1993: 54)134

As we will see below, this tendency to place priority on the 13th century
comes to amount to an article of faith in the work of Frago, and it some-
times leads him to twist his interpretation of the data to make it fit this
claim.
Nevertheless, it must be emphasized that Frago's fundamental thesis -
that the most significant features of Andalusian can be traced to 13th-
century demographic/dialect mixing in Betic Andalusia- remains funda-
mentally sound as a starting hypothesis, albeit an extremely strong one.
One of the major problems with Frago's development of this thesis is his
refusal to modify it in the face of the evidence, or lack thereof. There is no
a priori reason to assume that this hypothesis is true, particularly since he
himself reminds us that Andalusia saw two great periods of demographic
movement: the 13th and the 15th/16th centuries. Other problems stem from
the assumptions that he makes. First, he assumes that the changes of Anda-
lusian (primarily phonetic in his studies) occurred throughout Betic Anda-
lusia: "el andalucismo fonético se fraguó al mismo tiempo en su completa
extensión geográfica" (Frago Gracia 1993: 228). This claim apparently
arises in response to the traditional insistence on the primordial importance
of Seville in the development of Andalusian, an importance which is
clearly greater during the 16th century, when Seville was the very center of
the great population movements. In reality, it is possible that particular
areas - even particular towns or villages- saw unique developments in the
early period that failed to spread until later (or which disappeared under
the effects of dialect leveling). Such may have been the case with villages
founded predominantly by Catalans or Leonese, for example, but this is a
far cry from the more extreme claim made by Frago. More importantly,
however, such an assertion ignores the fact that Jaén, also repopulated in
the same period, is characterized to this day by features more typical of the
 Linguistic changes 229

center-north of Spain (e.g., contrast of /s/ and /Θ/, loss of /h-/). I will return
to this issue in the discussion of seseo.
Another problem may be found in Frago's conceptualization of the
process of language change in the context of dialect mixing. To begin with,
Frago never enters into detailed discussion of the effects of dialect and
demographic mixing, restricting himself to vague references to the terms
mixing and leveling. He never mentions the term koineization and cites
none of the relevant bibliography, and his conceptualization differs from
koineization in at least one very significant and damaging way: his assump-
tions about the rate of language change. His view is made patent in a pas-
sage in which he criticizes the belief of some earlier scholars in the relative
suddenness of the "great phonological readjustment" of the 16th century:
En realidad, a veces parece como si con demasiada facilidad se olvidaran
los grandes principios que todos confesamos aceptar, entre los cuales está la
creencia en la enorme lentitud con que se verifica la difusión geográfica y
social del cambio lingüístico, particularmente cuando se trata de innova-
ciones de tanta complejidad como las que han cuajado en la fonética anda-
luza. ¿Cabe tal diversificación lingüística en los estrechos márgenes que a-
penas dos centurias marcan? (Frago Gracia 1993: 6)135

This statement assumes that there is always an extremely slow, perhaps

even constant rate of linguistic change (one is reminded of Swadesh's glot-
tochronology), but why should we believe any such thing? Indeed, the
model of koineization predicts a temporarily faster rate of change for a
community suffering simultaneous demographic/dialect mixing and weak-
ening of norm enforcement (as compared to a stable, homogeneous com-
munity). If koineization can in fact occur in the space of two or certainly
three generations (Trudgill 1986: 95; Granda 1994: 100; Trudgill 1998;
Kerswill and Williams 2000), two centuries becomes a very long period of
time.
The conceptualization of change as necessarily slow has significant ef-
fects on the interpretation of documentary evidence. In what way? First, as
is revealed in the quote above, Frago repeatedly emphasizes the need to
return to medieval documents in search of evidence of linguistic change,
even lambasting other scholars for not doing so (e.g., Frago Gracia 1993:
37). There is no question but that Frago is right to make this claim. In this
regard, his practice is consistent with his belief, and in his studies of the
language history of Andalusia he has examined untold number of manu-
scripts. In fact, so new is the close philological study of medieval Andalu-
sian that no collection of paleographically-transcribed 13th-century Anda-
230 The Seville phase

lusian documents exists, making Frago Gracia's contributions in this re-

gard irreplaceable (even some ten years after the publication of his book;
see discussion below). However, by prioritizing the value of textual evi-
dence, Frago obliges himself to provide such evidence for his primary the-
sis: that the (nearly entire) set of features that characterize Andalusian had
their origin in 13th-century Andalusia. As stated above, this is a reasonable
starting hypothesis, and Frago is right to seek evidence in its support in the
documentary record. However, the doctrine of the slowness of change
means that even meager evidence can be interpreted as significant, since
one might not expect changes stemming from the 13th century to appear in
the documentation until much later. I provide specific examples of the
problems occasioned by this belief in the discussion of seseo below.
Scholars currently working on these issues must situate themselves
somewhere on the continuum between Mondéjar on the one hand and
Frago Gracia on the other. Among the most outspoken of these is Manuel
Ariza, who continues to see the 16th century, the period traditionally as-
signed to the reajuste fonológico, as key to the formation of Andalusian:
Parece hecho indiscutido que la consolidación de los fenómenos lingüísticos
que conforman básicamente lo que se suele llamar andaluz se debe en gran
medida al hecho histórico del descubrimiento de América y, por ende, a la
importancia de todo tipo -cultural, política, económica- que adquirió
Sevilla. (Ariza 1992: 15)
This, of course, stands in direct opposition to Frago, who sees the 13th-
century repopulation and dialect mixing as the most significant. Ariza has
argued against many of Frago's views. We will see many of his arguments
below in the discussion of seseo, but there are two general criticisms he
makes of Frago that must be addressed here. First, he objects to Frago's
"colonizing" theory because Frago sees such phenomena as sibilant devoic-
ing, seseo, and final sibilant aspiration as being originally northern phe-
nomena that must have been carried south by 13th-century reconquista-
dores and then spread quickly in Andalusia (Ariza 1996: 44).136 Ariza
believes that Galician and Leonese minorities (seen by Frago as the source
of many changes, including early seseo) were too small to affect the Castil-
lan majority, who would have provided the prestige norm. While it is true
that the minority status of the Galicians and Leonese would have disfa-
vored the selection of features of their speech (even simplified features), it
is incorrect to assume that Castilians constituted a majority (the available
evidence would indicate the contrary), or that their speech served as a pres-
tige norm or target for other speaker-learners. As indicated in Chapter 2,
 Linguistic changes 231

young learners do not seem to aim at established "targets" or pre-existing

varieties in the linguistic pool (cf. Siegel 2001: 182), and absolute fre-
quency has been shown to be a much stronger determiner of outcomes (cf.
Trudgill, Gordon, Lewis, and Maclagan 2000). Moreover, neither Frago
nor Ariza recognize that learner language can also be a source of innova-
tions in the prekoine linguistic pool.
Ironically, Ariza also rejects Frago's arguments because he believes that
Frago views change as occurring too rapidly, asking, with regard to 13th-
century seseo/ceceo:
Si Sevilla se reconquista en 1248 y la zona de Huelva en 1262, ¿es posible
que en el período de menos de un siglo se haya conformado uno de los ras-
gos más peculiares del andaluz? ¿Es que, nada más reconquistarse estas
zonas, los repobladores castellano-leoneses decidieron hablar de forma dis-
tinta a como lo hacían en sus lugares de origen? (Ariza 1994: 229)

While adult speakers may or may not change their speech habits in a
koineizing environment (although the answer seems to "yes" with respect
to certain features of Andalusian), the effects of learning in this environ-
ment of great linguistic variation can easily lead to quite significant
changes in the space of 100 years (or much less). Although Ariza's com-
ment points to some interesting inconsistencies in Frago's arguments,
Ariza's a priori rejection of Frago's hypothesis must itself be rejected. The
hypothesis - qua hypothesis - is sound. On the other hand, whether or not
particular changes can be attributed to 13th-century koineization is another
matter, and such issues can only be decided by returning to the documenta-
tion, as Frago and Ariza both argue, and interpreting it appropriately.

2.2. Documentary evidence

Returning to the documentation may seem a simple matter, but in fact,

there are no collections of completely paléographie transcriptions of early
medieval Andalusian texts. Those collections that do exist have been tran-
scribed primarily for use by historians, and thus attention to many details
has often been ignored. The difficulties that the language historian encoun-
ters can be illustrated by considering one well-known and well-received
collection: the Diplomatario andaluz de Alfonso X (González Jiménez
1991). Several factors combine to eliminate this collection as a useful
source of data on medieval Andalusian linguistic changes. First, no attempt
was made in the transcription to indicate occurrences of the notoriously
232 The Seville phase

ambiguous sigma(s), which have been resolved as s or z, according to the

etymology. As a result, the collection is of little value for study of the de-
velopment of seseo. In addition, the editors filled in word-internal abbre-
viations (without indications) and often modernized the spelling of proper
names (González Jiménez 1991: ix). Nevertheless, the collection could be
used to trace the Andalusian development of apocope, leísmo, possessives,
or intervocalic distinctions between -ss- and -s- (expressly preserved for
their possible linguistic interest, González Jiménez 1991: ix), were it not
for another problem. The Diplomatario is restricted to documents that were
prepared by the Royal Chancellory (González Jiménez 1991: vi), excluding
private, municipal, and ecclesiastical documents, as well documents of
noble households and the military orders. The scribes of royal documents
were associated directly to the royal household through the Notaría mayor
del rey or the notaría de cámara, or less directly through one of the tradi-
tional notarías mayores of the kingdoms of Leon and Castile, or through
the newly created Notaría Mayor de Andalucía. These scribes often fol-
lowed the king, and, though all the documents in this collection pertain to
Andalusia, many were prepared as the king traveled and resided outside of
Andalusia. In fact, if one eliminates all documents prepared outside of
Andalusia and those listed without text, only 277 of the total 539 remain
potentially useful. Of these, the potentially most useful are the 25 docu-
ments ascribed to the Notaría Mayor de Andalucía (Sanz Fuentes 1991:
cc), but only 13 of the 25 were actually prepared in Andalusia. Clearly,
these scribes did not reside regularly in Andalusia, but rather held an ad-
ministrative responsibility over the preparation of Andalusian legal docu-
mentation. They certainly would have been less likely to conform to the
newly developing norms of Andalusia, being more directly influenced by
supra-regional or more traditional sociolinguistic and scribal norms.
On the other hand, there are document collections which do satisfy, al-
beit partially, the needs of the language historian. This is the case of the
document collection prepared by Ostos and Pardo (1989). As with the Dip-
lomatario, the texts were not prepared with linguistic ends in mind; the
editors resolve sigmas and fill out word-internal abbreviations in the
documents in order to facilitate their reading. On the other hand, it has
seemed reasonable to use them as sources of data in this study for two rea-
sons. First, the transcriptions are sufficiently precise to provide reliable
evidence on certain phenomena, including three which are discussed below
(extreme apocope, possessives, and object pronouns). More importantly,
however, these documents are restricted to texts prepared by scribes who
 Linguistic changes 233

lived in 13th-century Seville, who worked for the municipal concejo, who
prepared (and of course dated) the documents in Seville, and who consis-
tently identified themselves in the texts as escríbanos de Sevilla. As a re-
sult, it is much safer to assume that these scribes were directly affected by
the developing norms of 13th-century Seville and that the texts prepared by
them therefore reflect those norms.
Below, I discuss and provide textual evidence for three developments
that can reasonably be attributed to 13th-century koineization in Andalusia:
the end of extreme apocope, the simplification of the medieval first person
singular possessives to mi(s), and the restitution of the etymological object
pronoun system. The first two are changes that arose first in Andalusia, but
later developed in the north as well, probably under the intense influence
of weak social ties with the south, though the structure of the language
itself already favored the change to a certain extent. The third is a feature
of Andalusian that has remained restricted to Andalusian and other south-
erly dialects (and the lateral dialects of Aragon and Leon). Finally, in re-
sponse to Frago and in order to demonstrate the need for constraints on
claims of change caused by dialect mixing, I analyze his arguments and
evidence for the 13th-century origin of Andalusian seseo.

2.3. The decline of extreme apocope

As explained in the previous chapter, it appears that extreme apocope be-

came established in Toledo-phase koineization through reallocation of a
phonetic and regionally restricted variant to a stylistic function in Castilian.
This is evident in the fact that, long after French influence declined, ex-
treme apocope continued to appear in Castilian texts. Lapesa marks the
zenith of this phenomenon - as represented in texts - as the first third of
the 13th century (Lapesa 1982: 173), but we know that it continued quite
strongly until the end of the 13th century. One is thus led to ask why ex-
treme apocope began to decline and full forms began to appear more regu-
larly. Lapesa points to what he considers a primary reason:
desde 1230 aproximadamente su predominio se atenúa un poco, coincidien-
do con el creciente afán de la iglesia castellano-leonesa por independizarse
de Cluny y el Císter, y con la mayor acomodación de los oriundos ultrapire-
naicos a la sociedad española. (Lapesa 1982: 173)

However, as was made clear in the preceding chapter, the decline in French
influence began long before the decline of extreme apocope, and it is cer-
234 The Seville phase

tainly safe to claim that extreme apocope was decoupled from French in-
fluence by 1200. A more immediate cause, according to Lapesa, can be
found in the influence exercised upon the language by King Alfonso X
(Lapesa 1951: 219-220). Lapesa pointed out that the prologues to many of
the Alfonsine texts were supposedly written by the king himself, and these
prologues are notably lacking in extreme apocope, while the following
texts generally show heavy apocope. Eventually, the king - reacting nega-
tively to French political moves after 1276 - even decided to impose his
non-apocopating preference as a court norm. However, even if Alfonso did
effect some editorial prescription against use of apocopated forms, this
cannot be viewed as a cause of the change. Rather, it must have been a
mere reflection of a change already in progress in a wider community.
What then might have provoked the decline of apocope? Certainly its
limited status as a formal and therefore marked linguistic variant could
have been a factor. But this fact alone cannot explain why extreme apocope
declined at a particular time. I suggest that the decline is rooted in the
process of koineization that was initiated in Andalusia from the very be-
ginning of reconquest in 1224. In this military environment, the combina-
tion of popular Castilian norms with the generally conservative western
Galician-Portuguese and Leonese speakers and scribes (Azevedo Maia
1986: 524; Staaff 1907: 213), would have combined to drastically lower
the frequency with which extreme apocope was employed, and therefore
the chances of its being learned by succeeding generations.137
In fact, Lapesa himself made reference to the possible influence of
demographic mixing in Andalusia:
La reconquista de Andalucía da lugar a los últimos grandes desplazamientos
de población; la mezcla de gentes no se resuelve allí en dominio de la
apócope, que nunca llega a igualar el número de formas plenas en las es-
crituras andaluzas. (Lapesa 1951: 214)

Still, Lapesa mentions this fact simply as a passing comment, and sees in it
little causal significance. Other scholars have also made passing reference
to this phenomenon. Frago, much more recently, includes only this com-
ment in a footnote: "No cabe duda, pues, de que en el castellano de la épo-
ca alfonsi implantado en Andalucía la apocope apenas tuvo repercusión"
(Frago Gracia 1993: 86). Beyond these vague remarks, is there hard evi-
dence that full forms first began to be consistently restored in Andalusia?
Lapesa attempted to investigate the regional distribution of apocope in its
final stages by using the texts in the Documentos lingüísticos de España of
Menéndez Pidal. From those notarial documents not granted by the Royal
 Linguistic changes 235

Chancellory during the reign of Alfonso X (1252-1284), he was able to

obtain the following data:
El computo de ejemplos recogidos en documentos no precedentes de la can-
cillería regia arroja los siguientes resultados: Campó, 4 apócopes extremas
contra 2 casos de vocal conservada; Castilla del Norte, 39 apócopes y 17 ex-
tranjerismos sin vocal contra 14 ejemplos de vocal conservada o añadida;
Álava, 5 apócopes extremas sin contradicción; Rioja Alta, 15 apócopes con-
tra 29 casos de conservación o vocal subsidiaria; Rioja Baja, 7 contra 14;
Burgos, 2 contra 7; Ávila, 2 contra 8; Toledo, 19 contra 30; Andalucía, 25
contra 38; Murcia, 19 (incluyendo muchos nombres catalanes o catalanis-
mos como agost) contra 25. (Lapesa 1951: 216)

These data, if superficially interpreted, appear to indicate that during the

reign of Alfonso, the phenomenon of apocope was both at its weakest and
its strongest in the extreme north of Castile: weakest in Burgos and La
Rioja, strongest in Castilla del Norte and Álava. However, it must be re-
membered that Lapesa's only data on Burgos are reduced to just four short
texts that Menéndez Pidal happened to include in the DLE. Indeed, if we
look generally at the north (e.g., La Rioja, Castilla del Norte, Álava and
Campó), it seems that apocope was relatively strongly retained there, main-
taining rates of occurrence above 50% even in local notarial documents
during the reign of Alfonso.
The data on Andalusia, based on only eleven texts, indicate that
apocope was relatively weaker in this region than in the extreme north,
though it appears as roughly similar, in proportional terms, to that of
Toledo (39.6% in Toledo vs. 36% in Andalusia). However, this too may
simply be an effect of the relatively small number of documents studied by
Lapesa. In my own investigation of the 13th-century notarial documents
transcribed by Ostos and Pardo (1989), the results appear quite different.
Following Harris-Northall (1991: 31), I analyzed the forms of the
preposition delant(e) and the adverb adelant(e). These forms were chosen
because they are native Castilian forms frequently subject to apocope in
texts of Old and New Castile, and they therefore represent a fairly reliable
gauge of the general intensity of apocope. Since, as we saw before, Harris-
Northall (1991: 35) reports that the phonetic context was only a minor
factor in determining the appearance of apocopated forms in Alfonsine
texts, I have counted the number of occurrences of apocopated and full
forms in all contexts. The Ostos and Pardo collection contains a total of
134 documents prepared by scribes of Seville between the years 1253 and
1300. The great majority of these documents contain at least one occur-
236 The Seville phase

rence of (a)delant(e) (the results appear in Table 31). Indeed, the very last
occurrence of the apocopated form is in the year 1276 (Document 65).138
These data contrast markedly with the very strong preference for apoco-
pated forms in the Alfonsine corpus. For example, in the General Estoria,
the percentage of apocopated forms in the first part (GEI) reaches 64.4%, a
tendency which grows stronger in the later fourth part (GE4), with apoco-
pated forms rising to 95% (Harris-Northall 1991: 35).
Table 31. Occurrences of (a)delant(e) in 13th-century texts from Seville.
Form Number of occurrences % of total
(a)delant 8 6.4
(a)delante 117 93.6%

Still, the Andalusian figures given by Lapesa do not clearly indicate an

overwhelming majority of apocopated forms compared to contemporary
notarial documents in Toledo. In part, this might represent the lingering
influence of northern norms, since the first Andalusian scribes of the 13th
century were all settlers who had moved from areas in the north. More
important was the probably spread back north of the consistent use of full
forms (or more accurately, the effect that contact with the novel and con-
sistent Andalusian usage had on northern sociolinguistic norms). It must be
remembered that massive demographic movements began with the ex-
tended military mobilization of the second quarter of the 13th century
(1224—1248). In addition, repopulation of the province of Jaén dates to
1224 and that of Córdoba to 1236. From this time forward, the effects of
these movements must have been felt in the north as well, and, as revealed
in the roughly similar figures given by Lapesa for Toledo and Andalusia,
were probably having an impact on Toledo, the primary northern urban
center with which the communities of the newly reconquered Andalusian
cities were in contact. From Toledo, the rejection of extreme apocope
could have then jumped north to Burgos and other cities, leaving apocope
only in isolated areas. Lapesa may have inadvertently identified the pattern
of spread northward from Andalusia in the following passage:

Tras el reinado de Alfonso X se precipita la decadencia de la apocope ex-

trema, que a fines del XIII y principios del XIV sólo se mantiene con alguna
intensidad en La Montaña, Álava, la Rioja Baja y Murcia, en las dos últimas
regiones es indudable el influjo aragonés y catalán. Castilla del Norte pasa
al dominio de las formas plenas. En Burgos, Osma, Valladolid y Segovia,
así como en la parte castellana de la Extremadura actual y en Andalucía, los
ejemplos son excepcionales o no se dan ya. (Lapesa 1951: 220)
 Linguistic changes 237

The spread of the complete preference for full forms in the latter part of the
13th century was facilitated as the massive numbers of soldiers needed for
the extended reconquest of Betic Andalusia returned to their homes in the
north. This would have occurred after each compaign, but we also know
that many of the earliest settlers sold the properties they had obtained in
the repartimientos after a few years and returned to extended families and
easier lives in the north. In so doing, they would have acted as language
missionaries, spreading the new features that were being negotiated as
norms in the demographically-mixed and socially-leveled south.139
King Alfonso X himself may have been one of these language mission-
aries. As indicated above, the royal prologues to many of the texts in the
Alfonsine corpus are characterized by an almost total lack of apocope
(Lapesa 1982: 180). Lapesa, of course, believed that the King's influence
was decisive in bringing about the decline of apocope. Whether or not Al-
fonso actively discouraged the use of apocope, it is certain that he did not
invent a new "norm" as a reaction to French political and military manoeu-
vres, as Lapesa suggested. More likely is that Alfonso - if he actually did
have any influence on certain segments of these texts - was simply reflect-
ing the rejection of apocope that gained currency first in Andalusia. It
should be remembered that Alfonso, though raised in the outskirts of Bur-
gos, spent much of his youth in the Andalusian military environment,
which he apparently entered at the age of 10. Alfonso's padrino, in charge
of training the prince in military matters, had been named the commander
of the strongholds of Martos and Andújar from 1225, and in 1231 he took
the young Alfonso with him on an attack into Cordobán territory. Alfonso
later participated in the battle for Córdoba, led a delegation and troops to
take control of Murcia (1243), and participated in the siege of Jaén. Troops
entered Seville on his birthday and he was crowned in the same city in
1252, where he would also eventually die (González Jiménez 1993: 15-
22). Moreover, Alfonso was made king of Castile and Leon, in a city popu-
lated by many non-Castilians (at least half the population), and it may have
been socially (and politically?) effective to reduce use of such a marked
and localized Castilian feature when interacting with subjects from other
areas. Under such circumstances, Alfonso and those around him may have
learned to prefer the less socially-marked forms of language - more or less
as Lapesa claimed (1982: 189) - that were more frequently employed in
early Andalusia.
238 The Seville phase

2.4. Simplification of first-person singular possessives

Andalusian rekoineization also seems to have been the catalyst for the final
step in the reorganization of the possessive system: the simplification of
the mìo/mi contrast (e.g., mio padre, mi madre).140 Of course, this change
did not remain limited to Andalusia, but it only spread in the north after its
establishment in Andalusia. As we have seen, the Toledo phase of koinei-
zation led to the selection and reallocation of long and short possessive
forms in all Toledan styles, and simplification of the 2nd- and 3rd-person
preposed possessives to tu(s) and su(s) in the Toledan popular norm. Tole-
dan formal varieties maintained gender distinctions (but not with complete
success), while conservative northern regions, which maintained more
close-knit social networks, took longer to accept all the changes. Oddly,
however, the first person possessives showed very strong resistance to
leveling and simplification. Méndez, like other scholars, claims that the
first-person possessives are regularly maintained even into the 14th cen-
tury:
En el masculino, hasta la primera mitad del siglo XIII, sólo existia la forma
mio con una alternancia acentual mio ~ mió. A partir de entonces, la forma
femenina mi comienza a desplazarse hacia el masculino; estas confusiones
son algo más tardías que las ocurridas en segunda y tercera personas: en la
época de Alfonso el Sabio son raras y don Juan Manuel sigue prefiriendo la
forma mio aunque son numerosas las confusions.
(Méndez García de Paredes 1988: 539)

Nevertheless, although the mio/mi contrast did survive until the 14th cen-
tury, this was not so for all regions.
In an analysis of the collection of documents from 13th-century Seville
prepared by Ostos and Pardo (1989), I find occurrences of mi(s) + mascu-
line noun from the very beginning, with extremely rapid growth in the fre-
quency of usage (mi(s) is used exclusively before feminine nouns in these
documents, with no examples of mio(s) + feminine noun or of the older
forms mia(s), mie(s)). Table 32 shows the developing preference for the
use of mi(s) over mio(s) with masculine nouns in percentages and absolute
counts (in parentheses), broken down by decades.
Many tokens of mio, however, occur in the final formulae of these
documents, especially in the sequence mio signo (e.g., Et yo Roy Perez,
escriuano publico de Seuilla, la fìz escriuir e so testigo e fìz y mio signo,
from Doc. 83-1285). In fact, by the end of the century use of mio seems to
survive only in formulaic passages, and removal of these examples makes
 Linguistic changes 239

much more dramatic the sudden reduction to exclusive use of mi(s), as the
statistics in Table 33 reveal.

Table 32. Frequency of use of mi(s)/mio(s) + masculine noun in documents from

Seville (1253-1300).
Period (total tokens) mi(s) mio(s)
1253-1260(36) 41.7% (15) 58.3% (21)
1261-1270 (34) 58.8% (20) 41.2% (14)
1271-1280(20) 70% (14) 30.0% (6)
1281-1290(51) 60.8% (31) 39.2% (20)
1291-1300(119) 62.2% (74) 37.8% (45)

Table 33. Frequency of use of non-formulaic mi(s)/mio(s) + masculine noun in

documents from Seville (1253-1300).
Period (total token) mi(s) mio(s)
1253-1260(36) 41.7% (15) 58.3% (21)
1261-1270(32) 62.5% (20) 37.5% (12)
1271-1280(16) 87.5% (14) 12.5% (2)
1281-1290(39) 79.5% (31) 20.5% (8)
1291-1300(83) 89.2% (74) 10.8% (9)

These data make clear the rapid dominance of mi(s) in the newly-
repopulated community of Seville (along with nearly categorical use of
su(s) for masculine and feminine nouns, even in formulaic phrases). The
early evidence of shift to mi(s) shows the initial importance of adult usage
and accommodation, but that overall progression underscores the impor-
tance of the first generation of children to the definitive selection and stab-
lization of the change. These data contrast sharply with roughly contempo-
rary documents from the north, even those from the early 1300s. For
example, though sporadic confusions do appear, there is generally regular
maintenance of gender distinctions in the DLE documents that show lst-
person possessives:

— Burgos (205-1315): mjos tíos, mjos tutores, mjos tutores, mjos cogedores,
mi carta, mjo vassallo, mjo mayordomo, mj carta, mjo seello
— Burgos (206-1367): mjo fijo (no other examples)
— Toledo (289-1302): mj don Johan, mjo vassallo, mjos ganados, mjo ssee-
llo
240 The Seville phase

— Toledo (290-1302): mj don Johan, mjo vassallo, mjs vacas, mj cabanna,

mi carta, mjo seello
— Toledo (291-1306): mj guarda, mj acomienda, mjo deffendjmiento, mis
cosas, mjo sseello

Evidence for Andalusian-style reduction can be found in the following

documents:

— Toledo (292-1324): mj muger, mj muger, mj poder, mj muger, mj marido,

mj signo
— Burgos (207-1414): mjplaser, mj otorgamiento, mjo signo

We can conclude that reduction of masculine mio(s) and feminine mi(s) to

gender-neutral mi(s) was completed as a general change first in Andalusia,
from where it then spread north.
Two questions remain to be answered: a) why did this pair resist reduc-
tion until the Andalusian phase of koineization?; and b) what factors con-
tributed to the final simplification? Méndez García de Paredes (1988: 533)
suggests that it was the larger number of competing variants that led to
quick reduction of 2nd and 3rd person adjectives, while the fairly regular
masculine mio(s) helped to preserve distinct input for learners. The very
high relative and absolute frequency of mio(s) probably did help to pre-
serve it during Toledo-phase koineization, but other factors played impor-
tant roles as well. The original monosyllabic articulation of mió made it
less susceptible to phrasal syncope or elision (a significant factor in the
earlier changes), which also served to protect the clarity of the input for
language learners. The reallocation of forms in the Toledo phase led to a
strong association of monosyllabic unstressed forms as preposed posses-
sives, and disyllabic stressed forms with other functions of the possessive:

— Preposed unstressed: mio(s), mi(s), tu(s), su(s)

— Stressed: mio(s), mia(s), tuyo/a(s), suyo/a(s)

The monosyllabic mio was partly consistent with this pattern, and therefore
was less likely to be eliminated.
Nevertheless, innovations did creep in. Although the masculine form
was originally pronounced as a monosyllable, mió(s), rhyme and assonance
in medieval poetry reveal that with time it came to be pronounced more
and more frequently like the disyllabic feminine mia(s), to give mio(s)
 Linguistic changes 241

(Menéndez Pidal 1962b: 256; Malkiel 1976: 472). This is unsurprising,

since even speaker-learners in the more stable social context of 13th cen-
tury Burgos and Toledo would have tended to seek and produce forms of
greater paradigmatic regularity and transparency. Speaker-learners cer-
tainly compared mio with mía, though by itself this analogy would not have
led to change of the articulation. Also influencing speaker-learners was the
larger possessive paradigm, in which all other stressed forms were disylla-
bic, and all other unstressed forms were monosyllabic. Mío may have ap-
peared first in stressed contexts (in unstressed contexts there was more
likelihood of a diphthongal articulation), particularly when paired with the
other possessives (e.g., lo tuyo y lo mío), since it was then that the evident
analogy was most likely to be made. By the beginning of the Andalusian
phase at least, mio(s) had come to be pronounced as disyllabic mio(s) much
of the time (Lloyd 1987: 279). Nevertheless, this change in articulation of
the mio/mio did not lead to loss of the mio/mi distinction (i.e., through real-
location of mio to stressed position only), since the stable social conditions
of post-koineization Toledo and Burgos then favored overall resistance to
innovations.141
However, during the Seville phase, the less constrained tendency of
speaker-learners to seek and produce greater regularity and transparency
led to the final simplification of the mio/mi distinction in preposed con-
texts, and to the reallocation of mio to stressed contexts only. Some of the
factors that affected the simplification of so/su and to/tu may have played a
part in this change. As with su(s) and tu(s), the form mi(s) was favored for
generic use because it was less-marked morphologically (Méndez García
de Paredes 1988: 536, 539). The definitive loss of stress on preposed
possessives probably favored some coalescence of the final -o of mio with
initial vowels in following words (e.g., mio honor > mi honor). Undoubt-
edly, however, the dominant factor affecting speaker-learners' analyses and
production was the overwhelming analogical pressures created by the sta-
ble Toledan popular norm (in even the earliest Andalusian texts su(s) is
used with stunning regularity for masculine and feminine). In Andalusia,
speakers were able to develop the innovative overgeneralization of mi(s) to
masculine use, and to reserve disyllabic mío for use in stressed contexts.
With the weakening of social networks and of norm enforcement, the
scales were tipped towards those factors favoring the reduction. Signifi-
cantly, at this stage the opacity of surface phonetic realizations played at
best a minor role. Rather, it was the establishment of clear morphological
patterns that most influenced these changes. Once the change was estab-
242 The Seville phase

lished in Andalusia, the multiple weak social ties functioning along the
Seville-Toledo-Burgos conduit facilitated its subsequent spread, and, along
with the existing structural influence, ensured that the change was regular-
ized.
Finally, it is worth noting that the structural reorganization of the pos-
sessive system is normally perceived as a unitary phenomenon. Lleal
(1990: 242), for instance, simply postulates that the gender distinctions for
all persons (save 1st and 2nd person plural) began to disappear in the mid-
13th century. However, from a sociolinguistic perspective, the changes of
the Seville phase are quite distinct from those of the Toledo phase. More-
over, in the constellation of phonetic/phonological, morphological, cogni-
tive and social factors contributing to the changes, phonetic/phonological
factors played a far more important role in the Toledo phase than in the
Seville phase.

2.5. Leveling and the rej ection of leísmo

Lack of leísmo has long been seen as an unusual feature of Andalusian

Spanish (as well as either inexplicable or in no need of an explanation).
Indeed, in Ariza's discussion of the origins of Andalusian, he lists it as one
of its most distinctive morphosyntactic features:
Dejando fuera las neutralizaciones morfológicas derivadas de la pérdida de
/s/ final, la principal diferencia del andaluz se produce en el sistema pro-
nominal. Efectivamente, en el andaluz no suele darse el leísmo, que tanta di-
fusión tiene por el resto de la Península, lo que viene a ser un mantenimiento
del sistema medieval castellano. (Ariza 1997: 64)
He also indicates that this is one of the most widespread features of Anda-
lusian. However, aside from the unlikely claim that Andalusian continues
the pronoun use of medieval Castilian (see Chapter 4), he makes no refer-
ence to either the causes or the chronology of its development, or lack
thereof. Elsewhere, scholars who have studied the medieval origins of
leísmo have paid little attention to its lack in Andalusian. This may be due
to the perceived normality of the lack of leísmo, since Andalusian pre-
serves the etymological system of object pronoun reference (see Chapter
4). At first glance, therefore, this appears to be a phenomenon in need of
no explanation, for it is exactly what is expected from the perspective of
the modern standard language. However, if one remembers that personal
leísmo and object leísmo were well established phenomena by the early
 Linguistic changes 243

13th century (see Chapter 4), then the lack of leísmo and the stable use of
the etymological pronoun system becomes a phenomenon in need of expla-
nation. Why did Castilian leistas not impose their usage as the norm in
Andalusia? Or, if they did, why has it disappeared?
It would appear that only two of the scholars who have studied leísmo
have also considered its absence from Andalusian. Fernández-Ordóñez
(1994: 123) first addressed this issue in her sociolinguistic study of leísmo
and the referential systems, in which she located the origins of leísmo in
the Basque substrate. In that scenario, the Basque influence was considered
strongest in the north, weaker in the center (Toledo), and absent in the
south, due to a tendency of syntactic features to resist borrowing. The
problems with this argument are evident, and more recently, Fernández
Ordóñez (2001: 462) has suggested that dialect mixing is likely to be the
cause, as I argue here.
Lapesa (1968) too addressed the issue of leísmo in Andalusia, though
only tangentially (he refers to Andalusian in his discussion of the possible
causal relation between apocope and leísmo).142 In his interpretation,
leísmo is first extended to Andalusia, but is later extinguished:
No es que falte algún ejemplo de le acusativo en la Andalucía medieval. Re-
gistro 'este pedazo de tierra calua, todo uos le uendemos . . . e nos vos so-
mos fiadores. . .de todos los omnes del mundo que uos le quieran demendar
. . . e os fe fagamos sano' 1263 Córdoba . . . 'el bien e las vebras que el om-
ne faz en este mundo le seguirán e le serán prueua . . . ante la faz del Sen-
nor' 1270, Jaén . . . Pero en 1303, Sevilla, encuentro 9 lo acusativo . . . sin
ningún le. (Lapesa 1968: 527)

Lapesa was hard pressed to provide evidence, and the examples le seguirán
e le serán prueua actually contradict his point, since seguir takes the dative
regularly in many dialects, as does ser in this kind of construction. Still,
Lapesa's basic argument is valid: the elimination of leísmo in Andalusian
dates from its very origins.
It is clear that what is lacking is data on the development of pronoun
use in early Andalusian. In order to address this problem, I have carried out
an analysis of the use of object le(s) and lo(s) in the Ostos and Pardo
(1989) collection of 13th-century documents from Seville. The results are
clear in their implications. In the 134 documents of this collection, there
appear a total of 403 tokens of unapocopated lo or los. These appear with
both neuter and masculine accusative reference (though few refer to mas-
culine persons):
244 The Seville phase

— como la carta ... lo dize e lo departe (11-1254)

— todo uos lo uendo (1-1253)
— dos maravedís alfonsis... que los paguedes cada anno (22-1263)
— el heredamiento .. . que me lo de (28-1264)
— aquel donadío ... que lo tengades & lo lauredes (69-1278).

On the other hand, there are only 26 cases of les, 10 cases of le, and 18
cases of apocopated / ' . I n the great bulk of instances, these forms are em-
ployed with clearly dative reference:

— una mezquita quel diemos (68-1277)

— que uos el cabildo quel fagades cada anno aniuersario (45-1266)
— assi les do poder (78-1284)
— la pregunta que les fiz (90-1289)
— etpreguntel si le plazia... dixo quel plazie (90-1289)

Nearly all the possible cases of leísmo involve verbs or constructions that
alternate in taking the dative or the accusative (all of which are discussed
in Lapesa 1968, Cuervo 1895, and Fernández Ordóñez 1999):

— Elvira . . . que ninguno de nuestros herederos noi puedan demandar (13-

1255)
— Belenguer... α pleyto quel quitemos desta fiadora sin danno (24-1263)
— que cunpliere mandado del arçobispo segunt el le enuiaua mandar (97-
1285)
— vn capellan ... et... vn sacristan quel ayude (127-1300)

Indeed, there is only one token of le(s) in all these documents that may be a
clear example of leísmo', los heredamientos . .. que les metades (84-1285).
However, even this example is questionable, since the exact reference may
not have been totally clear to the scribe: some 73 words appear between the
antecedent and the anaphoric pronoun. Moreover, a similar document from
the same year gives: los heredamientos . . . que los metades (87-1285).
This leaves the few examples from Córdoba as Lapesa's only clear evi-
dence of leísmo in 13th-century Andalusia. Clearly, however, that evidence
merely reflects the usage of a scribe who must have moved to Córdoba as
an adult and who maintained his own leísta usage for a time. In fact, the
tokens from the Cordobán document point to the contemporary vitality of
object leísmo not in Andalusia but rather in more northerly regions.
 Linguistic changes 245

With this evidence it becomes relatively easy to reconstruct the factors

which contributed to the establishment of the etymological pronoun system
in Andalusia. First, even though large numbers of the new settlers must
have been leistas, they do not appear to have represented a majority (set-
tlers from Old and New Castile seem to have represented not quite half the
total population, and not all New Castillans were leistas). In most koineiz-
ing situations, the results of dialect mixing tend to reflect simple majorities
(i.e., if there were more users of the etymological system in the mix, then
their usage would be selected, no matter how varied their origin), as op-
posed to the potential (social and cultural) impact of one particularly large
and influential group (cf. Trudgill, Gordon, Lewis, and Maclagan 2000).
Moreover, teista usage represented an added complexity to all those
speaker-learners who used the etymological system. Without an absolute
majority of leista speakers, it was even less likely to survive into the new
koine. Still, one might ask why an interdialectal solution like system E did
not appear in Seville, as had occurred in or near Toledo. Such an outcome
might even have been expected given the level of mixing and the frequent
presence of the Castilian court in the city. However, the social situation
had changed somewhat. Previously, Castilians had represented a clearly
defined and relatively homogeneous high-prestige group. However, by the
time of the conquest of Seville, the combined kingdom of Castile and Leon
had come to encompass large non-leista areas to the west (Galicia and
Leon), to the east (Cuenca, La Alcarria) and areas south and east of Toledo
controlled by the military orders (La Mancha). As such, the clearly-
desirable target that leista usage had represented in Toledo had lost much
of its social value. Indeed, the evidence from the Seville documents seems
to indicate that leista usage had become so marked that it was consciously
avoided even by those who had grown up using it.

2.6. Seseo\ A 13th-century merger?

As we saw above, Mondéjar (1991c: 227) has claimed quite insistently that
13th-century Andalusian is a unaltered transplant of the Castilian of
Toledo and Burgos (under the assumption that these were uniform). How-
ever, the data on the above phenomena provide fairly certain evidence to
the contrary. Moreover, the first two (the elimination of extreme apocope
and the reduction of lst-person singular possessives) eventually affected
dialects to the north, and the last (the elimination of leísmo) survived as
246 The Seville phase

one of the key characteristics of the Castilian of Andalusia (as well as

southern Spain).
On the other hand, we are left with the dilemma of deciding the dates
and causes of many other features, particularly those that now characterize
Andalusian varieties. There can be little doubt but that the environment of
social change and demographic/dialect mixing in 13th-century Andalusia
was propitious to linguistic change, and certainly other changes might rea-
sonably be attributed to the mixing that occurred during this period.143
However, we must consider the strong claim articulated by Frago Gracia:
was Andalusian indeed "born" - with the great bulk of its features - in the
13th century? Principal among the features that Frago believes can be
traced to the 13th century is seseo, perhaps the single most characteristic
feature of Andalusian (and American Spanish) vis-à-vis northern varieties
of Castilian (1993: 308).144 This is both a most daring and most reasonable
claim: daring, because it directly contradicts the much later dating for the
phenomenon given by many eminent scholars, and reasonable, because the
type of merger represented by seseo is exactly the kind of simplification or
reduction of phoneme inventory that one expects to find in a case of
koineization. The only reasonable response to this claim is to seek to
evaluate the textual evidence in its favor. As we have seen, however, this is
not an easy thing to do, particularly in the case of seseo, where there is
little option but to return to the manuscripts themselves. To his great credit,
Frago Gracia, in accordance with his own oft-repeated claim (e.g., Frago
Gracia 1993: 37), has done just these things. Therefore, his studies should
offer us the best available textual evidence for the early appearance of
typically Andalusian features in general, and seseo in particular. It is
largely this evidence that I examine here.
Frago Gracia addresses the issue of seseo primarily in his large (1993)
monograph and an earlier article (1989), in which he argues not only that
seseo must have already been well-developed in medieval Andalusia for it
to have been transplanted to America (an idea itself worth questioning; see
below and Chapter 6), but that its spread must be included among the ef-
fects of the 13th-century demographic movement and mixing within Anda-
lusia. In these two studies, he presents his own evidence as well as that of
other researchers. Of particular importance here is the reference he makes
to studies of Menéndez Pidal, Mondéjar, and Lapesa, all of whom have
also offered apparently valid textual evidence of seseo from medieval An-
dalusian texts. Paradoxically, Frago rejects the evidence offered by both
Pidal (1962c: 112-113) and Mondéjar (1991b, 1991d: 114-115), since
 Linguistic changes 247

much of this is based on forms appearing with the famously ambivalent

sigma(s) (a symbol or symbols often used to represent more than one sibi-
lant phoneme):
Resulta peligroso, sin embargo, fiar a la igualación de s y ζ en un solo
dibujo, e incluso al uso indiscriminado de la sigma con la zeta en un mismo
corpus, la atestiguación del fenòmeno seseo-ceceoso, cosa que Menéndez
Pidal hace con grafías andaluzas de finales del siglo XV como aseyte, asey-
tuno, disen, faser, fiso, hasen, honse, entre otras. Del mismo modo continúa
argumentando J. Mondéjar cuando propone como «los más antiguos testi-
monios del seseo en Andalucía» las formas -de 1302, dice- alguasil, arros,
cafis, nues, ves, veses, vesino, contrastadas con registros de cafizes y vezino,
con el apoyo de un quinse de 1423 que parece serle sumamente significa-
tivo. Curioso planteamiento metodológico: por la misma regla de tres dicho
seseo habría estado extendido a lo largo y a lo ancho de todo el dominio
castellano y hasta por otros ámbitos peninsulares, y, ni que decir tiene, en
fuentes andaluzas ejemplos como los precitados podrían amontonarse ad
nauseam. (Frago Gracia 1993: 332-333)

On the other hand, Frago apparently accepts much of the evidence from
15th-century Andalusia offered by Lapesa; though he cites only some of
this evidence directly (Frago Gracia 1993: 282), the generally positive tone
in which he refers to Lapesa's studies (e.g., Frago Gracia 1993: 332) leads
one to conclude that Frago assumes that the reader is already familiar with
the examples of 15th-century Andalusian seseo that were presented by
Lapesa ([1957] 1985: 252-3). These include:

— 1419 (Sanlúcar o Niebla, Huelva) : dief, diefmo

— 1460s (Cancionero de Baena)·. escaçeza for escaseza, çatan for Satan(as),
çedal for sedal, çenado for senado, azaz for assaz, bruçelas for Bruselas,
Amadiz for Amadis. (Frago Gracia 1993: 282, from Lapesa 1981: 283)
— 1475 (Pero Guillén de Segovia, Seville): çemençera, çenzilla, çenzillo
— 1487 (Pedro de Toledo, Seville): Roblez, inglez, fijoz, Andrez, Blaz, vis-
cayno, Sanches, Gomes, durasnos, Beatris
— 1492 (Seville): sirios for cirios, fiçieçe for fiziesse
— 1495 (Seville): çufriendo for sufriendo

Assuming, as Lapesa and Frago do, that the forms with s and ζ are cor-
rectly transcribed (and not resolutions of sigmas), they may represent sig-
nificant evidence of some kind of change. Lapesa viewed these merely as
evidence of early confusion resulting from the deaffrication of /t7 and /d2/,
248 The Seville phase

seseo as it existed in its supposed estado latente. His goal was merely to
provide roots for a phenomenon that can only be well-documented from the
mid-16th century, and which Lapesa clearly linked to the second great
phase of Andalusian population movement:
¿Qué había ocurrido en Andalucía para que el cambio [incubado segura-
mente durante siglos] lograse tan amplio desarrollo en el que va de 1470 a
1570? Hubo una doble sacudida en la vida andaluza: primero con los des-
plazamientos acarreados por la guerra y conquista de Granada: después con
el descubrimiento de América, la intensa emigración y, a la vez, el cre-
cimiento de Sevilla en pobladores y riqueza. (Lapesa 1985: 255)
On the other hand, Frago's arguments are much more extreme than
Lapesa's. Since it would clearly be difficult to use this evidence alone to
argue for the 13th-century origin of seseo, Frago provides further evidence
from older texts, often original manuscripts. Frago Gracia (1989) and
(1993) include what appears to be the sum total of textual evidence discov-
ered in his analysis of the medieval Andalusian corpus. However, Frago
presents this evidence in such a way that it is difficult to see the evidence
in its entirety and, thus, to evaluate it. This is so for several reasons. First,
he mixes evidence for seseo from different time periods, jumping, for ex-
ample, from the 14th century to the 17th without hesitation (Frago Gracia
1989: 282). Second, since he is convinced that devoicing of sibilants was
brought into Andalusia by new settlers in the 13th century and that the old
system of four sibilants (-ss- Iii, -s- Iii, ç Iti, ζ /dz/) began to collapse to
one (Isl) from that time, he mixes evidence for seseo with evidence for
devoicing. Since there is in fact substantial orthographic evidence for con-
fusion of -ss- with -s-, as well as confusion of ç with z, it becomes difficult
to accurately gauge the quantity of the evidence for seseo.H$ Third, he re-
ports this evidence in different places, without providing a clear summary.
Even in his (1993) monograph, some orthographic evidence of seseo is
provided in chapters devoted to other topics, and some is simply not in-
cluded, such as when Frago refers the reader to a previous article (i.e.,
Frago Gracia 1989) for further examples.
It therefore seems prudent to list all the tokens found by Frago of confu-
sion of originally apicoalveolar -ss- Isl and -s- Iii with originally dental
affricate ç Iti and ζ ldzl. Below I include the sum total of tokens from be-
fore the year 1500,146 presented in roughly chronological order, that I have
collected from Frago Gracia (1993) and (1989). Frago is quoted where this
is necessary for inclusion of pertinent details:
 Linguistic changes 249

1272 (Seville): 3 ençenz, 1 ençienz for censo, encenso (Frago Gracia 1993:
74)
1275 (Seville): 2 ençenz (Frago Gracia 1993: 74)
1293 (Seville): ensensarios for incensarios, susepçores for sucessores
(Frago Gracia 1993: 46). Frago points out that, elsewhere in the work of
this scribe, there is: "una constante atestiguación del trueque de c por ss en
el ant. suçessores: sosçepçores, soçepçores, sopçeçores, sopçepçores,
susçepçores, suçepçores" (Frago Gracia 1993: 226).
1338 (Repartimiento de Jerez): Almança for Almansa (Frago Gracia 1993:
228)
1347 (Huelva): fessieron for fezieron (Frago Gracia 1989: 282)
1352 (Santa Clara de Moguer, Seville): hesebçion for excepción (Frago
Gracia 1989: 282)
1353 (Santa Clara de Moguer, Seville): exssebçiones (Frago Gracia 1989:
282)
1366 (Seville): exepçion, esepçion (Frago Gracia 1993: 233)
1366 (Palma del Río, Córdoba): exepçion, efepçion (Frago Gracia 1993:
340)
1375 (Palma del Rio, Cordoba): conno/co for conozco (Frago Gracia 1993:
340)
1379 (Seville): beçola for besóla (Frago Gracia 1993: 230)
1369-1421 (Palma del Río, Cordoba): "En dos de estos textos es significa-
tiva la existencia del cultismo excepción con las grafías seseosas esepcion
(Doc. de 1369) y essebcion-esebcion (Doc. de 1374). Algunas décadas
después el análisis de dicha fiiente documental continuará arrojando los
mismos resultados, no siendo raros los escritos que por completo descono-
cen la ss, por ejemplo imo de 1403, que también incluye tres casos de esep-
cion, otros dos de 1412 y 1413, respectivamente, con sendas muestras de
exsebçion, y uno más de 1421 en el que se repite la forma excepçion."
(Frago Gracia 1993: 233)
1370-1400 (Alcalá de Guadaira, Seville): asas for assaz, pescueso for
pescuezo (Frago Gracia 1993: 278)
1402-1426 (Morón de la Frontera): "en el corpus ACM I (años 1402-
1426), aparecen los lapsus calami y variantes asas-aças, caliçes-caliaejo,
caliçes-caliaes, ençençario "incensario", ençienfo al lado de dos ençienço,
enpefca, e/quierdo, ofiçiales-ofiaiales, Velajques, vifcayno, Çahara-
Sahara ["Zahara"].' (Frago Gracia 1993: 343)
1407 (Seville): "En carta de donación dictada en Sevilla el año 1407 se po-
ne eglecia junto a indusimiento, y dos ejemplos de eglecia se encuentran en
250 The Seville phase

otro documento nobiliario dado en el mismo año y lugar." (Frago Gracia

1993: 342)
— 1411 (Carmona, Seville): resebryades for recibiríais (Frago Gracia 1993:
342)
— 1413, 1414, 1420 (Peñaflor, Seville): esebçion-esepçion (Frago Gracia
1993: 233)
— 1417 (Puerto de Santa María, Huelva): fesiese (Frago Gracia 1993: 234)
— 1419 (Homachuelos, Córdoba): esepçioti (Frago Gracia 1993: 233)
— 1425 (Santa Clara de Moguer, Seville): ençenço 'censo' (Frago Gracia
1989: 282)
— 1430 (Santa Clara de Moguer, Seville): exebçion, ecleçiasticas (Frago Gra-
cia 1989: 282)
— 1449 (Santa Clara de Moguer, Seville): ençienço (Frago Gracia 1989: 282)
— 1460 (Segovia, Andalusian copy): contumasia, yndusido, perteneser-
pertenesçen, 5 exepçion (Frago Gracia 1993: 342)
— 1467 (Huelva): Medina Çidonia for Medina Sidonia (Frago Gracia 1989:
282)
— 1470 (Huelva): Çiguença for Sigüenza (Frago Gracia 1989: 282)
— 1478 (Santa Clara de Moguer, Seville): juysio (Frago Gracia 1989: 282)
— 1480 (Seville): quatroçientos-quatrosientos, Vyçente-Visente, aluaseas for
albaceas, desendientes for descendientes, mayorasgo for mayoradgo, may-
orazgo (Frago Gracia 1993: 235)
— 1484 (Huelva): Medina Çidonia (Frago Gracia 1989: 282)
— 1484 (Seville): Aluaref for Alvarez (Frago Gracia 1993: 337)
— 1486 (Seville): syçion for cesión (Frago Gracia 1993: 235)
— 1489 (Seville): desernimiento for discernimiento, preçion for prisión
(Frago Gracia 1993: 235)
— 1489-1500 (Málaga): Ançurez for Ansurez, Lezmez for Lesmes, Yçrael for
Israel (Frago Gracia 1993: 350)
— 1490 (Córdoba): aniz for anis (Frago Gracia 1993: 342)
— 1492 (Granada): Yzmael for Ismael (Frago Gracia 1993: 234)
— 1494 (Pilas, Seville): "Haznalcaçar-Haanalcaçar-Hasnalcaçar, el ceceo
gráfico de Çeçilla "Secilia" -tal vez lexicalizado, en vista de su frecuencia
en muchos textos de la época-, y las formas Vifcaya y efsebçiones "excep-
ciones", siendo digno de notarse el hecho de que la formación escolar del
pendolista, sin duda apreciable, se manifieste en la corrección que introdu-
ce sobre un posision "posesión", en primera instancia con la segunda sibi-
lante escrita ç (posiçion) y luego retocada con ese sigma" (Frago Gracia
1993: 344).
 Linguistic changes 251

— 1495 (Seville): 5 sedula for cédula, asebtada for aceptada (Frago Gracia
1993: 235)
— 1495 (Francisco de Cisneros, Sevilla): sojusgar, Calis for Cáliz, topasios
for topaçios, Topasia (Frago Gracia 1993: 235)

Frago employs this evidence to argue that seseo was brought to Andalusia
in the 13th century, where it then spread rapidly. Of particular interest is
the following assertion:
Antiquísimos son los inicios de las transformaciones fonéticas conducentes
al seseo y al ceceo; las hemos visto nítidamente apuntadas en diplomas nota-
riales sevillanos del último tercio del siglo XIII, algunos de ellos redactados
por escribanos públicos de procedencia noroccidental, y no es descartable
que un posible primitivo seseo y ceceo gallego hubiera colaborado en el
desencadenamiento del cambio andaluz. Pero también . . . gentes de lengua
catalana, que ya, desde tiempo atrás practicaban el seseo . . . y con ellos . . .
vasconavarros . . . portugueses, probables apoyos asimismo de la tendencia
confimdidora. (Frago Gracia 1993: 362)
This sounds convincing, until one filters out the true examples of 13th-
century seseo: six cases of enc(i)enz for encens(o) and one of susepçores.
Given the massive numbers of documents that Frago Gracia claims to have
examined, are we to assume that a couple of examples of confusion show
the rapid spread of a merger? According to Frago, yes:
el que en un diploma hispalense de 1293 figure la grafía susepçores
'sucesores' es indicio irrefutable de la iniciación, insospechadamente tem-
prana, del proceso seseo-ceceoso. (Frago Gracia 1993: 34-35)
I will return to this particular example below. Here it must be emphasized
that in general Frago has little to offer for 13th-century seseo. Indeed, this
couple of examples of supposed 13th-century seseo stands out in opposi-
tion to the much better evidence he is able to provide for confusion be-
tween -ss- and -s-. Overall, the relatively low number of examples of me-
dieval orthographic representations of seseo that Frago (and others) are
able to provide is perhaps the most telling flaw in the argument (particu-
larly when compared to the greater evidence for devoicing).147 Frago him-
self is aware of the problem, and offers several explanations. For example,
he argues that it was easier for scribes to remember distinct spellings with
ç, ζ and -s-, even when the phonemes they represented had merged (Frago
Gracia 1993: 281). This, of course, leads one to ask why scribes suddenly
stopped remembering in the 16th century (when evidence of seseo becomes
252 The Seville phase

abundant). Perhaps in reply to this question, Frago offers another explana-

tion for lack of evidence of seseo in the medieval documentation:
Lo que sucede es que entre los últimos decenios del siglo XV y los primeros
del XVI se opera en la documentación andaluza una transición bastante
brusca en cuanto a la manifestación de las grafías incorrectas, que pasan en
poco tiempo de ser relativamente dispersas a abundar extraordinariamente,
en lo cual mucho tuvo que ver, quizá, el relajamiento de las normas ortográ-
ficas más introducidas en la scripta medieval, progresivo distanciamiento a
su vez favorecido por la misma evolución lingüística, que acaba siendo
asumida como una nota de afirmación regional. (Frago Gracia 1993: 355)

It may be that there was a relaxation of spelling norms at this time (just as
other social norms probably were relaxed during this period of movement),
but Frago cannot make clear why the norms seemed to prevent indications
of seseo while allowing significant examples of confusion of voiced and
voiceless sibilants to appear.148 More significantly, Frago is forced to argue
against himself in making the above affirmation:
Por otro lado, es muy arriesgado hablar de normas escriturarias de semejan-
te rigidez en relación a la ortografía antigua, pues el trato directo con los
textos manuscritos no da esa visión de los hechos: en no pocos casos de lo
que se trata es de tendencias, cumplidas con mayor o menor exactitud, pero
susceptibles de ser rotas incluso por los autores más cultos.
(Frago Gracia 1993: 155-156)

Clearly then, the lack of medieval orthographic evidence is a problem for

Frago, a problem which he is forced into the difficult position of having to
explain away.
Another problem with Frago's insistence on the early origin of Andalu-
sian seseo is that it forces him into another dilemma already commented
upon. Since he insists repeatedly that the characteristics of Andalusian,
including seseo, appeared simultaneously throughout Betic Andalusia in
the 13th century (i.e., they did not spread from Seville), he has to explain
why certain areas of Betic Andalusia, such as Jaén and the north of Huelva,
have never accepted seseo. His response is to suggest that this may be due
to the (small minority of) Galicians who participated in the reconquest of
Córdoba and Seville after 1230 (Frago Gracia 1993: 361), but who were
not well represented in eastern Jaén.149
Unfortunately, the problems do not end here. As Ariza has pointed out
in a series of articles (1992, 1994, 1996, 1997), Frago appears to misinter-
pret - or, at least, over-interpret - many of his data. One the greatest prob-
 Linguistic changes 253

lems is that evidence similar to that found by Frago for Andalusian crops
up in the north of Spain. Ariza cites evidence from the 14th-century Fuero
de Huete: fisieron, asension, consejo 'conçejo' (Ariza 1994: 230, from
Diaz Montesinos 1987). 150 Ariza (e.g., 1994: 228-229) also offers a few
other examples from northern regions, seeing in these very sporadic confu-
sions signs of deaffrication of the affricate sibilants, which became more
prone to confusion as they became distinguished by only one distinctive
feature.151 Indeed, the clear tendency of marginalized groups, such as Black
Africans and Gypsies (who arrived in Spain in the late 15th century), to
acquire Spanish with seseo (Ariza 1996: 62) would indicate a situation
already conducive to merger.
Other over-interpretations and misinterpretations on Frago's part can be
grouped into several categories:

1. Syllable- or word-final confusion. Ariza (1996: 53) suggests that sporadic

confusion of -s and -z in implosive position probably indicates not seseo
but rather deaffrication and devoicing (or, rather, the scribes' attempts to
represent a phone that was not clearly assignable to only one phoneme).
Harris-Northall (1992) has argued that final -z was used to represent a de-
voiced and deafiricated sibilant in 13th-century Alfonsine texts, so it
would come as no surprise to find examples of orthographic confusion in
this period. Indeed, these are the very cases that also show up in the non-
Andalusian Fuero de Huete. Lapesa, of course, offers us dies, diesmo, and
many of Frago's forms fall into this category: encenz, connosco, asas, en-
pesca, esquierdo, Velasques, viscayno, mayorasgo, Yçrael, Yzmael, Lez-
mez, Hasnalcaçar, sojusgar. For later periods, such as the end of the 15th
century, it does seem reasonable to hypothesize an initial neutralization of
-s and -z in implosive position.152 Lapesa offers some of the best: Roblez,
inglez, fijoz, Andrez, Blaz, viscayno, Sanches, Gomes, durasnos, Beatris.
2. Post-nasal confusions. Confusion in this position appears frequently in
Frago's examples: encenz, Almança,153 ençienço, Ançurez, conçejo-
consejo (and in many other examples after 1500). Terrado Pablo, commen-
ting on the forms seguransa and fiansa in a 1380 document from Barcelo-
na offers the following explanation: "Acerca del valor fricativo o africado
de esta s no vamos a emitir juicio alguno, y ello por una razón de fonética
general: la nasal η supone una oclusión en la cavidad oral, que seguida de
una fricación puede equivaler a (§). El uso de s en lugar de ç no creemos
que posea aquí valor probatorio en favor del ablandamiento de las sibilan-
tes africadas, pues el rasgo oclusivo puede estar implícito en la nasal que
precede a i . . . [footnote] Se pudo comprobar esto mediante sonogramas
realizados por el autor de este artículo en el laboratorio de fonética Pere
254 The Seville phase

Barnils. En una pronunciación enfática, una secuencia como 'con sus' pa-
recía poseer una realización (§): era patente la barra de explosión que pre-
cedía a la zona de turbulencia fricativa" (Terrado Pablo 1986: 177).
Given such phonetic similarity, we should not be surprised to find occa-
sional confusions in post-nasal position (cf. English speakers' confusion of
prince and prints, mince and mints).
3. Metathesis, assimilation, dissimilation. In discussing the examples çemen-
çera, çenzilla, çenzillo, Lapesa commented that they were "casos todos ex-
plicables por asimilación o disimilación" (Lapesa 1985: 252). Other exam-
ples which show the appearance of two sibilants can easily be explained in
the same manner, or by reference to the related phenomenon of metathesis.
Forms used by Frago that can be explained as resulting from these phe-
nomena include: susepçores, encenz (among Frago's most significant to-
kens), ensensario, asas, pescueso, ençençario, ençienço, fesiese,
Çiguença, Hasnalcaçar, ecleçiasticas, syçion, desernimiento, Ceçilla. In
addition, the numerous variants of the Latinism excepción are clearly af-
fected by a combination of these phenomena (consider the number of sibi-
lants in [e(k)stseptsjón]), as well the strong tendency to adjust this alien
surface form to the phonotactics of the language.154
4. Influence of etymological forms. Ariza (1996: 57) points out that the form
connosco actually corresponds to the etymological form of the word, and is
thus best not included as evidence oí seseo.
5. Misreadings. Correct interpretation of the data depends, first and foremost,
on correct transcription of original manuscripts. In the case of the sibilants,
this is especially difficult, since the sigma (or sigmas) was often used to
represent both ζ and s in even the most conservative dialects and, in the
case of devoicing, could represent all four sibilants. As we have seen
above, Frago chastises Mondéjar for using forms with sigmas as evidence
of early confusion. However, Frago himself falls to the same temptation.
To wit, he offers the following forms as evidence of seseo, even though in
each case he explains that they appear with sigma: caliaejo, caliaes, oficia-
les, Sahara (Frago points out that the initial sibilant could be a sigma),
Calia, topamos, Τορασΐα (Frago indicates that the last three are all tran-
scribed with sigma). Ariza has also returned to some of the same manu-
scripts used by Frago in order to check the transcriptions. He suggests that
Frago has made two important misreadings. First, Frago suggests that the
formposision (actually posiaion) 'posesión', was originally written with ç
(posiçion) and then corrected with the sigma. Ariza does not agree: "Es
verdad que hay una sigma, pero no es claro que debajo haya una ç" (Ariza
1996: 47). More importantly, Ariza also checked the form beçola from a
1379 document. Given that the use of the ç would provide incontrovertible
evidence of confusion, it is an especially clear example. Ariza includes a
 Linguistic changes 255

photocopy of the original in his paper, in which he reads a clear s (Ariza

1996: 54).
6. Substitutions. Alonso (1947) pointed to the frequency of substitutions of
one sibilant phoneme for another. Though there are claims in his article
that we must now question, his assertions about confusion of s and ç are
validated by the survival in northern peninsular dialects of lexicalized
forms showing such confusion (e.g., çurcir 'zurcir' for original surcir).
With regard to Frago's claims, the following observations are of particular
interest: "ençienço alternaba con encienso en la Edad Media . . . sencillo;
cenzillo en Mingo Revulgo . . . sedaço hasta el siglo XV, luego cedaço . ..
acechar alterna con assechar (*ASSECTARE) desde muy antiguo . . . reçuçi-
tar aparece algunas veces. . . macizo sobre MASSA . . . çoçobra (SUBSU-
PRARE) alternaba todavía con soçobra en el Cancionero de Castillo . . .
bisma-bizma, lesna-lezna, mezquino-mesquino, brosno-brozno, biscocho-
bizcocho, respe-rezpe, chosne-chozne, Velasco-Velázquez, cascorvo-
cazcorvo, mescolanza-mezcolanza." (Alonso 1947: 1-2). Pascual (1991)
has also pointed out that sufrir-çufrir is a form that shows a long history of
phonemic vacilation, with lexicalization of çufrir in some dialects. Other
forms reported by Frago can also be interpreted as lexicalized substitu-
tions, particularly such repeated examples as egleçia, seduta and Medina
Çidonia.
7. Errors. Ariza (1996: 47) indicates that some forms may simply be scribal
errors. He points to the case of conçejo for consejo, in which the scribe
had just written conçejo 40 times and then should have written consejo, but
perhaps wrote conçejo by dint of repetition.

After considering the alternative explanations, Frago's use of the limited

number of forms he can supply to support his thesis appears unjustified. Of
course, the evidence here reviewed may point to a very limited and gradual
development of neutralization of final sibilants (or perhaps even merger) in
the 15th century (as Lapesa suggests), but it clearly does not serve to sup-
port Frago's more extreme claims of 13th-century origin for this phenome-
non.155
Given the small number of forms offered as evidence, and the existence
of very plausible counter-explanations, Frago's hypothesis that Andalusian
seseo originated during the 13th-century repopulation of Andalusia must be
rejected. On the other hand, it may very well be related to a later stage of
demographic movement and koineization in the 16th century (see Chapter
6). Moreover, the rejection of this particular hypothesis does not imply the
rejection of other hypotheses linking particular features to 13th-century
koineization in Andalusia. The point is that each case must be investigated,
256 The Seville phase

and confirmed or denied on the basis of the historical record. The model of
koineization makes no predictions about such particular developments. To
the degree that it is predictive, it is only so in the broadest sense: given
certain sociohistorical conditions, including but not limited to demographic
mixing, dialect mixing and breakdown of close-knit social networks, one
can expect to find the co-occurrence of changes that show mixing and sim-
plification. In this sense, the model can be compared to a guide map, telling
the researcher where there is likelihood of discovering certain types of
change. Of course, future research may be able to identity more specific
types of change as characteristic of koineization. Beyond that, only histori-
cal research and analysis of the particular social and linguistic situation can
provide further illumination.

3. Conclusion

The Seville phase represents the last great phase of koineization begun and
completed in the medieval period. During this phase, there occurred, once
again, great demographic and dialect mixing in the valley of the Gua-
dalquivir. This mixing, amply attested in the repartimientos, presented an
ideal social context for the occurrence of koineization. Analysis of a col-
lection of 13th-century documents prepared by municipal scribes of Seville
reveals three changes that can plausibly be attributed to this rekoineization:
the elimination of extreme apocope (which probably had an impact on sub-
sequent northern elimination of the phenomenon, particularly through
weak ties between south and north); the elimination of the minority feature
of leísmo (with some evidence that adult leistas suppressed this marked
feature in their speech); and the swift reduction of the first person singular
possessives to the invariant form mi(s) (which also began to affect northern
usage). These changes are in some respects less "dramatic" than those of
earlier phases, but this is unsurprising given the extensive leveling or
spread of features that had already occurred in the north by the time of the
Seville phase. The evidence of such changes serves to rebut the claims of
some who see no significant effects in the 13th-century repopulation of
Betic Andalusia. On the other hand, analysis of the case of seseo reveals
that it cannot be plausibly linked to 13th-century koineization in Betic
Andalusia. This case in particular shows that efforts to link dialect mixing
and language change must be grounded on appropriate application of the
model and careful interpretation of the evidence.
Chapter 6
Conclusions

This study has had two fundamental objectives: first, to apply the model of
koineization to the study of early Castilian, in order to gain a clearer under-
standing of the nature and causation of characteristic changes of this vari-
ety (or group of varieties), and, second, to refine our understanding of the
model itself and to explore its potential uses in the explication of language
change, particularly in the reconstruction of changes which occurred in the
distant past. For this reason, I have focused on the earliest phases of devel-
opment of Castilian, beginning at a time-depth of over a thousand years.
To complete this task it was necessary to start with a well-defined and
coherent model. I therefore examined the literature on koineization, with
particular attention to the key studies of Siegel, Trudgill, and Kerswill and
Williams, and arrived at a definition of the model which is a critical syn-
thesis of existing research. It is important to note that this model is explicit
in distinguishing a micro-level of analysis, that of individual speaker-
learners in a particular type of social and linguistic context, from a macro-
level of analysis, which consists of a description of the predominant or
systemic linguistic results of collective speaker activity (i.e., the commu-
nity norms, or the abstraction known as a dialect or a language). The proto-
typical social context of koineization is characterized by relatively free and
unfettered interaction of speakers of mutually intelligible dialects (which
by definition share most lexical and structural features), with rapid increase
in variation accompanied by equally rapid decline in norm enforcement. As
adult and especially child/adolescent speaker-learners accommodate to one
another and re-establish social networks, they develop new linguistic
norms, which most often favor the features which have the highest fre-
quency in the prekoine linguistic pool. In addition, speaker-learners some-
times favor features because of their relative salience. Salience depends
necessarily on some structural factor (e.g., transparency or regularity) or
cognitive/perceptual prominence (e.g., stress or initial position in an utter-
ance, constituent, or word), but also depends, often crucially, on the attri-
bution of sociocultural significance to a particular variant (e.g., identity-
marking and stereotyping). Some items (e.g., morpholexical items) are
inherently more salient than others (e.g., some phonetic variants). In ac-
258 Conclusions

commodating to others, adult and child speaker-learners must analyze the

varied output of other speakers (converting learner input into intake) and
base their own production on these inductions or analyses. In most cases
they overgeneralize frequent, regular, or transparent variants and patterns
to the detriment of others. Opaque features in others' speech can lead
learners to misanalyses, leading in turn to innovations not present in any of
the established contributing varieties. When this is done by large numbers
of speakers (either through simultaneous innovations, or through spread
from a smaller group), then even such misanalyses can be expected to enter
the resultant koine. Koines can generally be characterized by simplification
and mixing/leveling, but we need not expect all features of a koine to be
directly inherited from existing contributing varieties, particularly those
which can be shown to have represented simplification from the speaker-
learners' perspective, or those which are exaggerated and/or selected in
order to mark new identities.
This model has allowed the identification of Burgos-phase demographic
and dialect mixing as a key causal factor for some, but not all, of the char-
acteristic changes of the período de orígenes, changes which Menéndez
Pidal described long ago, but for which he was unable to offer any but
circular explanations. One of these, the leveling out of the variant forms of
articles and of contractions of prepositions and articles, reveals clearly the
tendency of koines to avoid fusional opacity, which results from the weak-
ening of social constraints on speaker-learners' tendency to perceive and
reproduce the most frequent, transparent, and regular variants. Another
change, the decrease from the seven vowel phonemes of Romance to the
five of Castilian, reveals the effects of speaker-learner reanalyses of the
most frequent variants in the surface output of other speakers. In addition,
the reanalysis of the Romance low mid vowel phonemes as either two-
phoneme diphthongs or single-phoneme monophthongs reveals how
koineization can lead to abrupt interruption of lexical diffusion. When
speakers begin to abandon non-functional variants in the new social con-
text (or simply favor one to the exclusion of others), the rich variation re-
quired for the continuation of the analogical process of lexical diffusion is
lost, and even changes in progress shared by most contributing dialects are
likely to be frozen by the rapid selection and reanalysis of the most fre-
quent forms. This, of course, leads to results that seem irregular when
compared to (usually) more stable contributing dialects, where lexical dif-
fusion is more likely to progress. Another significant finding is that the
rather marked development of Romance h / into Castilian /ue/ (as [we]) is
 Conclusions 259

best understood as the combined effect of surface analogy, reanalysis of

frequent forms, and most important, its exaggeration and use as a marker of
generation and local/regional identity.
Examination of the Burgos phase also obliged discussion of other
changes which Menéndez Pidal identified as hallmark features of early
Castilian. In the case of /f-/ > [h-], koineization may have favored the over-
generalization of a pre-existing Cantabrian variant, but the lack of evidence
makes more definitive claims impossible. The development of Latin -CT- >
/tj/ (= orthographic ch) was analyzed as a possible introduction of a new
phone thanks to the overgeneralization of a natural assimilatory/perceptual
variant, and of a new phoneme thanks to the reanalysis of the new palatal
phone. However, close analysis of dialectal evidence reveals that this
phone was a pre-existing, possibly phonemic, variant in other contributing
dialects (though in different phonetic contexts). Nevertheless, in more con-
servative dialects these innovations were resisted, so koineization probably
did favor the natural phonetic variants of -CT- (or [jt]) by allowing them to
be freely produced and quickly stabilized by speakers, as well as associated
with pre-existing /tj/ (a similar argument can be made for the development
of -C'L- > ! γ). Finally, the development of the Latin obstruent + lateral
consonant clusters, PL-, CL-, FL-, shows mixed results in Castilian and
would seem to be a prototypical result of mixing. However, comparison
with contributing dialects, including Galician, reveals that the patterns of
development in Castilian follow closely those of western varieties of Ibero-
Romance, and that the changes probably occurred before the Moorish inva-
sion. The development of these clusters, so often cited as a unique feature
of Castilian, is not in fact so typical of Castilian as it is of the entire north-
west of the Peninsula. Koineization in Castile merely preserved the major-
ity features of the contributing dialects (though it is possible that the par-
ticular demographic and dialect mix favored preservation of the lateral in
those words where the original consonant clusters had not been altered).
While koineization may not have played the key role in most of these de-
velopments, it is important to note that these changes, sometimes perceived
as marked or even "deviant" (Posner 1996), are not in fact contradictory to
the tendencies of koineization, and therefore do not serve as counterevi-
dence to my arguments (as specified in the methodological guidelines out-
lined Chapter 2).
At some time or another, nearly all of the hallmark features of early
Castilian have been attributed to the impact of Basque learners of Ro-
mance. In Chapters 2 and 3, I discussed some of the proposals that have
260 Conclusions

been made, and examined them in light of the model of koineization. Over-
all, it seems clear that koineization would indeed allow Basque learners of
Romance to exercise some influence in the prekoine linguistic pool. Larry
Trask's complete rejection of Basque influence on Castilian must itself be
rejected. However, it is equally clear that marked features of the Basques'
interlanguages were unlikely to survive the selection process of koineiza-
tion, since minority features are generally leveled out. Only when features
produced by the Basques were also produced by native Romance speakers
were they likely to be selected in the resultant koine. In every potential
case of Basque influence examined in this study (e.g., vowel simplification
and stabilization of diphthongs; /f/ > [h-]; the rise oí leísmo), the influence
of Basques is not strictly speaking necessary to explain the changes in
question, but in most cases it is possible to see their linguistic output favor-
ing or disfavoring particular Romance features in the prekoine linguistic
pool. Therefore, each case of putative Basque influence must be reconsid-
ered in relation to koineization.156
A major objective of Chapter 4 was the establishment of the Toledo
phase as a significant period of koineization in the history of Castilian,
through analysis of three groups of changes. It was also a period of wide-
spread dialect leveling and language spread (but also dialectal differentia-
tion; see below). For example, typically Castilian features such as the /ue/
diphthong and transparent preposition + article contractions appear to have
became common in parts of Leon and Aragon in this period. This occurred
as these regions to the west and east of Old and New Castile underwent
nearly simultaneous koineization at the same time as they were in contact
with Castilian(s). Such outside influence certainly favored Castilian-style
variants in the new and less distinctive Leonese and Aragonese dialects.157
Though the specific changes of this period include such simple effects
as the stabilization of normal apocope, they also reveal other effects of
koineization which I did not find in the Burgos phase. For instance, the
development of the rich variety of early Castilian possessive forms shows
functional reallocation and simplification of (some) atonic forms, appar-
ently as a result of reanalysis of pre-posed possessives as determiners and
their simultaneous deaccentuation, in contrast to the continued salience of
stressed pronominal and post-posed adjectival possessive forms. This is an
interesting case because frequent but less salient forms suffered simplifica-
tion (loss of gender differentiation), while stressed forms were actually
built up and their regular differences reinforced. In addition, the realloca-
tion of possessives affected all styles in Toledo, but the simplification of
 Conclusions 261

preposed possessives developed only in more popular registers. Yet an-

other case of stylistic reallocation can be found in the development of ex-
treme apocope, in which a phonetic variant typical of fast (and probably
informal) speech in early Castilian, but which was also a regular feature of
the speech of eastern Iberians and francos, suffered reallocation to a formal
stylistic (and possibly social) function. This retention of extreme apocope
as a stylistic variant was influenced by the heavy presence of prestigious
francos in the early Toledo demographic mix.
One of the most interesting changes analyzed in this work is the rise of
leísmo, which began in the Burgos phase, but which spread (and was reana-
lyzed) during the Toledo phase. Its origins reveal the impact of apocope,
which obscured the input for speaker-learners, and allowed the reanaly-
sis/overgeneralization of dative le for use as a marker of singular masculine
accusative count reference (it must be recognized that the reanalysis repre-
sented, from the learners' perspective, a simplification of complex input).
The conversion of this innovation (probably shared by many speaker-
learners) into a change was favored by the koineizing context, in which
such innovations were not suppressed or corrected. Such marked use of le
later spread north into Cantabria (with internal repopulation) and was car-
ried south into Toledo. However, in areas of Toledo (including Guadala-
jara near Aragon, and Toledo itself, which saw very high levels of dialect
mixing), this original extension of le was itself reanalyzed. In these areas,
speakers constructed their grammars based on the perceived most frequent
and consistent use of le: for reference to masculine persons. This reanaly-
sis, along with loss of count/mass distinctions, also represented a simplifi-
cation, and in the koineizing context it too became established as a norm in
some areas.
Leísmo is clear evidence of my claim that koines should not be viewed
as mere reductions to a "least common denominator" of contributing dia-
lects. This idea, prevalent from the days of Meillet, but alive and well to-
day (e.g., Posner 1996; Mufwene 2001: 5), relies on a view of koineization
which does not take into account the role of learning and the importance of
induction/reanalysis. Leísmo (or better, leísmos) also appears to reveal the
development of internal dialectal variation during the Toledo phase of
koineization (though further research of the textual record needs to be done
to confirm some of the details). This is found in the different norm (per-
sonal leísmo) that developed in eastern (but not central or western) regions
repopulated during the Toledo phase. Though it has often been assumed
that koineization necessarily leads to widespread uniformity or homogene-
262 Conclusions

ity, as in Australia, this is not necessarily so for all features, nor for all
cases of koineization. In this case, the causes are fairly clear: the mix of
settlers was distinct, and most of the areas that developed personal leísmo
were settled not immediately but rather slightly later in the early to mid-
nth century. Indeed, even more mixed settlements settled in the late 12th
and 13th centuries abandoned leísmo and favored the etymological system.
Leísmo also presents an interesting example of the long-term effects
that koineization can have. Leísmo itself is clearly tied to Burgos- and
Toledo-phase koineization. However, in the following centuries, the intro-
duction of leísmo into the system presented succeeding generations of
speaker-learners in Old Castile and Toledo with a structural anomaly. Even
in the more stable social environment of those centuries, speaker-learners
appear to have innovated, and to have done so to such an extent that over
time new norms were established. For example, by analogy, speakers who
used le for both (masculine) dative and masculine accusative count refer-
ence, began to use les for plural reference. This was followed by use of
la(s) for both feminine accusative count and feminine dative reference. In
some varieties, such analogies eventually led to the regularization of les or
the extension of los for plural masculine datives and accusatives. Finally,
singular lo was extended to the dative, thus eliminating the last remnant of
case distinctions. We can see that the initial change provoked by koineiza-
tion set up a structural irregularity that speaker-learners sought to regular-
ize by altering other components of the pronoun system (in a kind of chain
reaction), until slowly arriving at the most advanced system we know to-
day, which lacks case reference. This example points to the potential im-
portance that "catastrophic" processes of change such as koineization can
have even after social networks resolidify. A parallel claim is to be found
in Dixon (1997) and in Labov's recent volume Principles of Linguistic
Change: Social Factors. Though Labov is primarily interested in "changes
which emerge from within the linguistic system" (Labov 2001: 20), or
changes in stable communities, he mentions several times that the catastro-
phic social mobility and dialect contact occasioned by World War Π (simi-
lar to - though different from - prototypical koineization) may have served
as a trigger for many of the sound changes under way in Philadelphia
(Labov 2001: 227, 318, 509).
Not unrelated to the preceding discussion is the fact that many changes
that are initiated in a koineizing region later spread to regions where con-
tributing dialects are spoken. In such cases, koineization is once again a
catalyst for changes that do not progress in other regions until after the
 Conclusions 263

original phase of koineization is well under way. Examples of this can be

seen in the spread of the reorganized possessive system from Toledo back
north to Burgos, as well as the spread north from Betic Andalusia of regu-
lar simplification of mio/mi and the decline of extreme apocope in all
styles. Several factors can be cited that favor such spread: the pre-existence
of the innovations in contributing varieties (originally resisted by stable
norms); the probability that the structure of the system already favors the
change; the effects of numerous weak ties between innovative and conser-
vative regions (including the impact of language missionaries). Still, these
are only partial explanations for the spread of innovative koineizing fea-
tures, since resistance is maintained to others, such as when leísmo was
preserved in the north even after its loss in La Mancha, and, more impor-
tantly, Andalusia. Questions of identity certainly come into play, and in the
case of leísmo, both its sudden disappearance in Andalusia and its surpris-
ing preservation in the North (with subsequent analogical changes in the
pronoun system) argue for its high sociocultural salience (stereotyping).
Changes that do not spread outside the original place(s) of koineization
become new regional dialectal features. Frago Gracia has claimed that a
whole range of hallmark features of Andalusian can be traced to 13th-
century dialect mixing in Andalusia. In my analysis of his data relating to
the sibilant merger known as seseo, I have been forced to conclude that the
evidence offered by Frago does not support the hypothesis. The analysis of
Frago's claims and data shows once again that the establishment of causa-
tion in cases of dialect mixing must be subjected to the methodological
constraints outlined in Chapter 2. On the other hand, I have found solid
evidence for the rapid loss of leísmo in Andalusia (as compared to Old
Castile and Toledo), so it is plausible to speak of the early formation of a
distinct dialect and some notion of regional identity in Andalusia (although
this same rejection was apparently occurring in other southern varieties as
well).
The findings on seseo have further implications. Given the analysis of
available evidence, it seems likely that seseo - as a regular feature of the
dialect - dates from the 16th century, and was probably initiated by the
massive demographic movements which followed the fall of Granada and
the discovery of America in 1492. If so, then the common claim that seseo
(along with some other Andalusian features?) was already present and
simply transported to America needs to be re-examined. It seems that at the
end of the 15th century the difference between the phonemes in question
had been reduced to a minimum (based only a small difference in place of
264 Conclusions

articulation), and seseo may have existed in some communities as an in-

cipient syllable-final neutralization. These features would have been car-
ried by colonists to all parts of the Americas. As such, it is possible that the
regularization of seseo as a feature of Andalusian and all American varie-
ties of Spanish occurred as part of a particularly far-flung process of
roughly simultaneous koineization(s). Indeed, a similar kind of analysis has
recently been proposed by Trudgill, Lewis, and Maclagan (2000), who
have argued that some features of southeast England and of Southern-
Hemisphere Englishes developed simultaneously as a result of "inheri-
tance" of common tendencies or "drift". In the case of seseo, the inherited
tendencies or structural features were the weakly-marked phonemic dis-
tinction and incipient neutralization. In the koineizing context(s), speaker-
learners everywhere could easily have generalized the merger and thereby
established it as a new norm.
The changes of the Seville phase are significant for other reasons as
well. For example, the better textual evidence of this period makes it pos-
sible to identify the effects of both adult and child accommoda-
tion/acquisition. Adults appear to have rejected use of stereotyped Castil-
lan leísmo from the very beginning of repopulation. They also dropped use
of extreme apocope fairly quickly (though the formal written texts proba-
bly maintain it longer than it was maintained in speech). On the other hand,
the data for the simplification of the mio/mi contrast show that it began
with adults, but was only carried to completion and stabilized by the first
child generation. Indeed, the changes for this period seem not only to con-
firm the rapid time scale of koineization suggested by Kerswill and Wil-
liams (2000), but suggest that adults can quickly negotiate new norms for
certain very salient features from the very beginning of the process. Of
course, by the time of the Seville phase, many of the contributing varieties
had already suffered significant leveling and simplification, so the dialects
in contact were probably not as distinct (or opaque to learners) as in earlier
phases.
It is also useful to consider the three stages of koineization in medieval
Castilian in relation to each other. One very significant difference is the
tendency in the Toledo phase to establish markedly different stylistic
norms. For example, both extreme apocope and maintenance of gender
distinctions in to/tu and so/su were features of the conservative Toledan
norm, while weakening or elimination of both marked the popular norm.
Examples of such marked stylistic distinctions are not present in my analy-
ses of the Burgos and Seville phases, though this may simply be an effect
 Conclusions 265

of the data and/or the features chosen for analysis. Still, the Toledan ten-
dency to maintain a conservative norm even in the face of the usual
koineizing tendencies may reflect the unique combination of factors found
in the repopulation of Toledo: the arrival of the court to the symbolically
important city; the great institutional presence, and with it, the arrival of
aristocrats, royal officials and servants, and ecclesiastical bureaucrats, in-
cluding many francos in high positions; as well as the maintenance of an
influential and tightknit Romance-speaking Mozarab community, which
may have shared and favored some of the more conservative features of
Castilian and other varieties.
The perspective across different phases of koineization also allows us to
note that, on the one hand, changing conditions can lead to very different
results (cf. the establishment vs. loss of leísmo and extreme apocope in the
Toledo and Seville phases), while, on the other, what might be considered
unitary processes from a structural perspective are in fact completed only
over the course of several periods of punctuated or catastrophic change.
From a sociolinguistic perspective, therefore, they are not unitary changes.
This is well exemplified by the reallocation and simplification of the pos-
sessive forms, which begins in the Toledo phase, but is not "completed"
until the Seville phase with the loss of the mio/mi contrast. Beyond this, it
is not unreasonable to claim that the repeated series of koineizations in
Castilian/Spanish (including those not studied here) are at the root of the
long-recognized "drift" of Spanish toward more analytical, transparent, and
simplified structures.158 Such tendencies have also been identified in other
"central" or urban varieties, and would seem to warrant further investiga-
tion of dialect mixing in the history of other language varieties, even those
less clearly affected by prototypical koineization (cf. Milroy 2002).159
Throughout this study I have assumed that the periods of koineization
originally suggested by Penny are in fact valid and useful periods for the
history of Spanish. Indeed, one of my principal aims has been to substanti-
ate Penny's broad claim that the repeated periods of demographic and dia-
lect mixing had a significant effect on the development of medieval Span-
ish. Still, there is presently general disagreement on the proper division of
the history of Spanish into periods (e.g., Eberenz 1991; Harris-Northall
1996b; Martinez Alcalde and Quilis Merin 1996), and even over whether
any periodization of the language is possible or desirable (Wright 1999;
Penny 2000: 5). Of course, as Penny and Wright argue, if one establishes
temporal divisions in the history of a language and assumes that these rep-
resent absolute boundaries between the language of one period and that of
266 Conclusions

another (rather like absolute divisions between dialects), then periodization

must be abandoned. Wright argues that the lack of such absolute differ-
ences means that any periodization based on internal features is nearly
meaningless, and he suggests that periodization of a language be based
exclusively on external, metalinguistic factors. He argues that there is
probably only one really important period division in the history of Castil-
lan: the 13th century, when the distinction between Latin and Romance
penetrated most Iberian societies, along with sharp conceptions of differ-
ence between Romance "languages" (as a result of a change from a more
logographic "Latin" script to a more phonographic one). He also shows
some willingness to accept Harris-Northall's (1996b) demonstration of the
standardizing effects of printing, from about 1500. Otherwise, nothing
counts.
I would argue, however, that to establish such an extreme dichotomy
between internal and external factors is to fall into, in part, the very "struc-
turalist fallacy" that he criticizes in his essay. Of course, Wright's rejection
of periodization has to be understood as a rejection of traditional and
largely arbitrary divisions such as "Old Spanish", "Golden Age Spanish",
"Modern Spanish", as well as Eberenz' slightly less arbitrary alternatives, a
"fase expansiva" from 1050-1250, a "fase antigua" from 1200-1450, and a
"castellano medio" from 1450-1650 (Eberenz 1991: 94, 105). Interestin-
gly, however, Eberenz asks if "la evolución de la lengua se produce a un
ritmo siempre igual o si, por el contrario, los cambios se acumulan en de-
terminadas épocas. Si la segunda hipótesis se revelara como exacta, se
resolvería el problema de la periodización" (Eberenz 1991: 93). I would
argue that the analysis presented in this study implies just this: periods of
rapid change punctuate periods of slow change in the history of Castil-
ian/Spanish. Koineization allows us to establish a periodization that links
both internal and external factors (cf. Eberenz 2000: 17), and which can
improve our understanding of the patterning and causation of changes in
ways that previously were not possible. Of course, this periodization is
rather different from earlier types of periodization, since it shuns absolute
divisions, and is sensitive not only to temporal but also geographical varia-
tion. As such, a periodization of Spanish based on koineization can help
avoid certain pitfalls of historical linguistic research, since researchers will
be required to specify not only when and where changes took place, but
also when and where they did not. The failure to comment on such varia-
tion in the past has been explained by Milroy as yet another effect of the
 Conclusions 267

standard ideology, or tendency to project the modern conception of uni-

form standard back onto earlier language states:
when dates are suggested for changes . . . they are usually given, without
comment, as the dates at which the changes took place in this standard vari-
ety or its unilinear precursor, and not the dates at which they might have
taken place in some other variety. Despite this, however, they are then usu-
ally cited as the dates at which the changes took place 'in English'.
(Milroy 1992: 125)

Although Milroy focuses on the study of English, the discussions of previ-

ous research contained here reveal that similar problems plague the study
of Spanish (cf. also Fernández Ordóñez 2001: 399).
Surprisingly, however, there remain scholars who continue to subscribe
to the belief that language change is never rapid or sudden, but rather is
necessarily gradual. We have seen evidence of this belief in the work of
Frago Gracia and even Ariza, but they are not alone; recently the creolist
Salikoko Mufwene (2001) has come out in defense of this position. Muf-
wene rejects completely the idea that the rate of language change can vary:
"Creoles did not develop more rapidly than other languages.. .All of them
developed gradually" (Mufwene 2001: 130). For Mufwene, all linguistic
change is necessarily gradual. This is a stunning conclusion, given that
throughout this work Mufwene argues that scholars must investigate "eth-
nographic ecologies" in order to explain language change, and many of his
ideas are similar to those I defend here (e.g., the tendency of linguistic
contact to favor cognitive and linguistic processes of simplification, trans-
parency, and regularity). Why then does he insist on the gradualness of
change? Mufwene rejects the dominant prototype model of creolization
(i.e., sudden change or language creation caused by children's learning of a
pre-existing adult pidgin or pre-pidgin); instead, he argues that creóles
developed slowly over many generations, gradually developing away from
their lexifiers as speaker-learners were socially distanced from native-
speaker models (he also completely rejects the importance of children to
the development of creóles).160 Although this may well have occurred in
some cases, Mufwene does not present this as another possible model of
creole formation; rather he presents his as the only model (and though he
argues for gradual distancing, he rejects the gradual rapprochement implied
by the concept of post-creole continuum). In addition, he argues throughout
the book that creolization, koineization and all contact-based change works
in fundamentally the same way (certainly true at deep levels of abstraction,
but I address the value of distinct prototypes of koineization and creoliza-
268 Conclusions

tion in Chapter 2). As a result, his belief in the gradualness of creole for-
mation forces him to view all language change as gradual.
Nevertheless, these claims strike me as untenable, for the research of
many other scholars (not least that of Kerswill and Williams) stands as
strong counter-evidence to Mufwene's view, as does the research presented
here (although the value of the evidence depends on how "gradual" is de-
fined; in language change, its meaning can only be relative). It is clear that
the rate of language change can vary along with change in the socio-
cultural context, and it is equally clear that both adults and children influ-
ence language change, though not necessarily in the same ways, and per-
haps in different ways in different social contexts. Still, in every case, chil-
dren's cognitive and psychosocial development play key roles in stabilizing
new norms and new linguistic varieties.
These questions lead back to the more fundamental goal of this study:
to respond to the Actuation problem, through the establishment of explicit
causal links between social conditions, speaker activity, and linguistic
changes. This is increasingly recognized as a fundamental goal of research
on language change. However, not all scholars are agreed on the value of
the approach adopted here. For example, Roger Lass agrees that "the most
important act of theoretical integration that could be performed in histori-
cal linguistics is somehow establishing . . . a clear and intelligible nexus
between short-term individual behavior and long-term linguistic evolution"
(Lass 1997: 336). Though this is exactly what Milroy (1992), Keller
(1994), Mufwene (2001), and others set out to do, Lass argues that the
"hermeneutic approach" of these scholars can only be effective for expla-
nation of non-structural lexical change. For structural changes, he suggests
that all attempts to establish such a nexus are doomed to failure. However,
it is my view that analyses of change which take into consideration struc-
tural, cognitive, psychosocial, and sociocultural factors, such as those in-
cluded in this study, show that Lass' pessimism is not completely war-
ranted, and that so-called hermeneutic approaches do indeed offer scholars
a promising means of developing cogent and plausible explanations of
linguistic change.
Maps
270 Maps
Maps 271

I
.s

aCO

•43

•υ
O
m
O
m

!
272 Maps

IS
o
oU
0
(O
1
1

Β[β

•ν

Λ
ο
CQ

Ι
Maps 273

H
1
3

§o

υ
41
S

O
δ
^
O
α
0

1Vi
o
&
&
oυ
00
10
1υ
I
υ

£
Notes

1. Substrate theory in historical linguistics was based on a geological metaphor.

Though the approach did acknowledge the importance of external factors
through its emphasis on language contact, such contact was viewed as contact
between systems, rather than between speakers.
2. Among historians of Spanish, Menéndez Pidal and Lapesa stand out as advo-
cates of the close relation between external and internal history. However,
their views ranged from reasonable claims of foreign influence in the case of
borrowings, to broad and sometimes questionable claims of substrate influ-
ence, to clearly Humboldtian ideas of language as embodying the spirit of the
nation. Lapesa, in particular, sometimes employs concepts that are similar to
those employed here (see Chapters 4 and 5), but my overall approach and con-
clusions differ markedly from his.
3. A particularly clear and well-known example of the impact of speakers on
language change was provided in Labov's (1963) study of Martha's Vineyard.
4. Recently, scholars such as Lass (1997), Croft (2000), and Mufwene (2001)
have developed models of language change based on biological evolution.
These rely on yet another biological metaphor, but in this case languages are
seen not as organisms but rather as species. This work has had a significant in-
fluence on recent discussion of language change. For example, references to
"competing variants" within a "linguistic pool" are influenced by this meta-
phor. The comparison of linguistic and biological evolution is clearly an intel-
lectually fruitful endeavor, and I refer to this work where appropriate (particu-
larly as discussed in Croft 2000). Nevertheless, I remain wary of the
evolutionary metaphor, since it is difficult to reconcile with factors such as
human learning, human intentions, and human culture.
5. For example, Labov' (2001) recent volume Principles of Linguistic Change:
Social Factors includes little discussion of dialect mixing, and focuses instead
on change within stable communities, though he does at times acknowledge
the potential impact of "catastrophic" social mobility and dialect contact (see
the Conclusion for further discussion). On the other hand, there is growing
emphasis on study of dialect contact/mixing and language change (e.g., the
special issue of the Journal of Sociolinguistics on dialect contact and mobility
edited by Lesley Milroy (2002), and Paul Kerswill's (2002) recent article on
"Koineization and Accommodation" in the Handbook of Language Variation
and Change).
6. Note, however, that Nevalainen (2000), Mufwene (2001), and Milroy (2002:
6) suggest that dialect mixing in London had a significant effect on the
development of English English. Lodge (1999) argues similarly for Parisian
French. Smith and Sneddon (2001), following in part Tuten (2000), argue that
 Notes 275

Smith and Sneddon (2001), following in part Tuten (2000), argue that dialect
mixing in medieval Paris led to the loss of some of the preposition + article
contractions of French (in medieval Spanish they were completely eliminated).
The model of koineization is applicable to these cases, but Castilian presents a
more nearly prototypical case of (repeated) koineization.
7. Spanish is used in the title of this book for two reasons. First, it is the name
most frequently used to refer to the modern standard language which devel-
oped from the medieval Hispano-Romance variety known as Castilian (castel-
lano). Second, in informal English usage, Castilian (spoken in central northern
Spain) is sometimes contrasted with Andalusian (spoken in southern Spain), so
use of Spanish avoids any possible confusion. Nevertheless, I use both labels
in this book, though I employ Castilian whenever it is necessary to distinguish
the variety under study from other varieties of Hispano-Romance, such as
Leonese and Aragonese.
8. In each case, I have used the name of the principal urban center of the re-
gion/period as a label, both for ease of reference and since these urban centers
probably played key roles as centers of mixing and of diffusion of change to
surrounding regions, and thus in determining the features that would make up
the regional varieties in each period. Note that I have given placenames in
their English form only if those forms are frequently used and likely to be rec-
ognized (Seville, Lisbon, Castile, Catalonia). Otherwise, I have used Spanish
forms of placenames (particularly for cities and rivers).
9. It now seems that the Ionic dialects, particularly Eastern Ionic, were them-
selves the results of dialect mixing and leveling, for the Ionic coastline had
earlier been colonized by speakers of varied origin (though early Attic speak-
ers probably dominated). Attic is thus often identified as a conservative or ar-
chaizing form of Ionic (Horrocks 1997: 7, 27).
10. Horrocks (1997: 29) suggests that the new written standard of Great Attic
strongly affected speech.
11. Horrocks (1997: 36) rejects the supposed Doric origins of this change (fre-
quently claimed by other scholars) since the early Koine (Great Attic) was
based almost exclusively on Attic and Ionic varieties. Instead, he argues that
the forms selected in the Koine were favored because they allowed speakers to
avoid the morphological irregularity which resulted when the Attic or Ionic
forms were inflected (the propagation of such regularizing innovations is fre-
quent in koinezation). But Horrocks also believes that speakers may have
opted for the minority Doric forms precisely because they allowed avoidance
of a "parochial morphological anomaly"; in other words, the avoidance of
marked dialectal forms through a "strategy of neutrality" (see below). Still,
Horrocks (1997: 19-20) grants that these forms may have been recognized
and accepted as a result of their use in the heavily Doric choral lyric of Athe-
nian drama. It probably is not necessary to choose any one of these factors,
276 Notes

since all would have contributed to increasing the frequency of the forms that
survived.
12. Some regional differences in the Greek homeland can also be attributed to
imposition of the Koine and to "substrate" effects of dialects such as Boeotian
or Doric (indeed, Bubenik [1989] and Horrocks [1997] use the term Koineiza-
tion to refer to the spread of the Koine into the regions where traditional dia-
lects were spoken). It is also worth noting that these same factors can be seen
as crucial to the development of Latin American varieties of Spanish (though
with different factors weighing more heavily in some regions than in others).
13. See Wright (1996) for a more recent discussion of dialect mixing and the
development of Vulgar Latin.
14. Siegel (1985) cites brief definitions of these terms proposed by Mühlhäusler
(1980: 21) for pidginization (see below).
15. The very existence of the medieval Provençal koine as a uniform linguistic
variety has been brought into question (Blanchet 1992, Zufferey 1987). Its
perceived uniformity is quite relative, since a multitude of dialectal variants
are attested in the medieval chansonniers. More importantly, the appearance of
uniformity, to the degree that it exists, probably reflects the fact that surviving
texts were copied largely between 1250-1350, towards the end or after the
great period of Troubadour production. Moreover, they were copied by scribes
in Northern Italy who would have known Occitan or Provençal only as a for-
eign language, particularly as it existed in the poetry. As a result, they may
have perceived it as more uniform than it actually was, and thus regularized in
their copies dialectal variants they perceived as incorrect.
16. The Andalusian form juerga is not useful to Hall's argument unless one as-
sumes that the development of Andalusian was clearly distinct from that of
Castilian. This view cannot now be accepted. Hall (1974) also describes nu-
merous peninsular dialects as koines, but it is clear that he uses the term only
to mean shared or common language of a region: "In the twelfth and thirteenth
centuries, the dialects of Aragon, Asturias, León, Castile, and Galicia were
about equal to each other in prestige and administrative and literary use . . .
Concomitantly with the Reconquista, the Castilian dialect became the standard
for the regions which came under Castilian rule, gradually overlaying the other
regional koines such as Asturian, Leonese, Aragonese, and the conservative
Mozarabic spoken in the central area" (Hall 1974: 121). Only some of these
varieties could be considered koines in the technical sense of the term devel-
oped here.
17. In light of Mufwene's (1997) criticisms of the second criterion, Siegel (2001)
restricts the definition of contributing varieties to those that are mutually intel-
ligible (a criterion which allows for the inclusion of second-language varie-
ties). Of course, speakers in a contact situation may at first believe that their
 Notes 277

varieties are mutually unintelligible, but later change their attitudes as they be-
come more familiar with different speech forms and their users.
18. The assumption that native speakers need to be present (and dominate) in the
mix excludes Israeli Hebrew from the category of "prototypical koine" (see
discussion of prototype models below). However, that is not to say that the
koineization is not useful in explaining its development, a fact which is recog-
nized in Blanc's (1968), Siegel's (1997) and Kerswill's (2002) discussion of it
as a koine. Still, Israeli Hebrew escapes (and will continue to escape) easy
classification, for its development (i.e., its resurrection as a native spoken lan-
guage based on use as a learned literary language) is surely unique in history.
Other cases in which non-native speakers dominate are probably best classi-
fied and studied primarily as cases of language contact (and only secondarily
as cases of dialect contact). Such varieties will, of course, show clear multiple
influences of the source language(s) on the target language. In the discussion
of Basque and Castilian in Chapters 3 and 4,1 show how contact between two
languages in a prototypical case of koineization differs from that in other
situations.
19. More recent work (e.g., Milroy 2002) tends to analyze the development of
regional varieties (or standards) as the result of dialect leveling (see below for
further discussion of the relation between koineization and dialect leveling).
20. Definition of a technical sense for koine (and koineization) implies that some
varieties often referred to as koines will no longer be classified as such in spe-
cialist discourse.
21. Thomason and Kaufman (1988: 170) also question Mühlhäusler's develop-
mental model of pidgins and creóles, since the expansion of a pidgin depends
on social factors that may or may not be present. In other words, there is no
necessary developmental model of pidgins into creóles.
22. In his explicit discussion Siegel himself indicates that reduction must be less
frequent in koines (1985: 370-371), and Siegel (1987: 187) does not consider
radical reduction as a feature of koineization. I discuss it here because the
(1985) the stage-based model seems to imply its importance.
23. Koineization is not the only type of new dialect formation, for new dialects
certainly arise through slow accretion of (opacity-producing) changes and the
development of new community identities in socially-isolated communities.
However, koineization is certainly the most rapid.
24. Speakers may assimilate to indexical features other than social or geographical
dialect, such as age, gender, and emotional state.
25. These two processes are not necessarily mutually exclusive, nor completely
conscious. Yaegor-Dror (1993) describes a situation in modem Israel in which
ethnic minority speakers of Hebrew mark their identity by diverging from
mainstream Israeli koine norms for /r/ in salient positions, but accommodating
unconsciously towards the Israeli koine for /r/ in non-salient positions.
278 Notes

26. Subsequent research confirms many early hypotheses about (short-term) ac-
commodation. For example, studies have demonstrated that indexical cues (in-
cluding regional or social accent) can influence listeners' perception of the
speaker's attractiveness and intelligence, their classification of the individual
into a particular group or class, and even the persuasive power of the speaker
(e.g., Oskenberg et al. 1986, DePaulo 1992, Gallois and Callan 1988, Gallois
et al. 1992). Interestingly, women have consistently been shown to accommo-
date more than men, and both men and women tend to accommodate more to
men than to women (e.g., Willemyns et al. 1997). This greater tendency to ac-
commodate fits well with the fact that women are more often leaders of lin-
guistic change (see Labov 2001). Nevertheless, Communication Accommoda-
tion Theory is now often subsumed within Bell's (1984) Audience Design
model. For many cases of face-to-face accommodation, accommodation theory
generally remains an adequate framework for explaining speaker actions and
motivations. However, Bell's emphasis on "initiative style shifts", in which
speakers change their language in order to alter the situation and/or their pres-
entation of themselves (accommodating to a perceived outside norm as op-
posed to accommodating to any particular interlocutor) is necessary for ana-
lyzing cases in which speakers attribute a social value to a variant and then
adopt/exaggerate its use (Trudgill's hyperdialectalism). It also shows that cul-
tural factors will affect short-term accommodation, thus making the results of
accommodation less predictable (see Kerswill [2002] for further discussion).
27. It is well known that the efforts of speakers to imitate and/or learn second
dialects are are plagued by errors of analysis and production. For example,
speakers of Southern American English often notice the mis-analysis that non-
Sourtherners make of the 2nd-person plural pronoun y 'all·, this can be inter-
preted by northerners as simply a southern variant of singular/plural you. Non-
Southerners attempting to accommodate to a Southerner will sometimes greet
a potential interlocutor with a perplexing "How are y'all?", when no one else
is in sight and the speaker has no previous knowledge of the interlocutor's
family and social relations (which might otherwise justify such usage).
28. This seems similar to the case described by Yaeger-Dror (1993). Avoiding the
stereotyped pronunciation would allow Northern English English speakers to
both accommodate and continue to mark their northern identity.
29. Trudgill's original list of examples of accommodation influenced by salience
does not constitute a theory of salience, and, though interesting, remains slip-
pery as first presented. Speakers can adopt or avoid a feature because of its sa-
lience. For example, speakers often do not adopt the contrast mentioned in
item 1. Items 2 and 5 seem to conflict, because if English English speakers do
not acquire post-vocalic /-r/, then there is homonymie clash and speakers pro-
nounce hot and heart the same way, since in item 2 we are told that speakers
adopt the American pronunciation of hot. Homonymie clash itself does not
 Notes 279

seem to be something that speakers worry about (with any consistency), since
any conversational breakdown can be repaired with a repair strategy.
30. Trudgill, Gordon, Lewis, and Maclagan (2000: 309) also reject Lass' (1990)
thesis of swamping, according to which South African English developed
along the lines of southeast-of-England English because colonial speakers
consistently selected forms from the English southeast (as a prestige target va-
riety) when faced with a number of options. They find, rather, that majority
features tended to win out, but that majority features shared by all contributing
dialects (in Southern-Hemisphere Englishes) tended to coincide with south-
east-of-England features.
31. Kerswill and Williams (2002: 103) also attribute the decline of clause-final
like to its lack of pragmatic or interactional salience, since it is customarily
used in propositions that are not foregrounded. This claim is founded on pro-
posals made by Cheshire (1996), who argues that certain variables are favored
in interactionally prominent or foregrounded constructions. For example, she
claims that interrogative and negative clauses are inherently interactive syntac-
tic environments where non-standard forms tend to occur with greater fre-
quency.
32. Wolfram and Schilling-Estes (2003: 722) argue, however, that a fudged lect is
something of an anomaly, since speakers in an area where an innovation is dif-
fusing generally show alternation (mixing) between two or more variants, and
they suggest that (some) apparent cases of fudged vowels simply show inter-
mediate stages in the movement of the vowel (at least in English).
33. Mesthrie (1994: 1866) also suggests a key role for children: "While accom-
modation is a characteristically adult process, selection of accommodated
forms, and stabilization, are more likely to be associated with child acquirers
of the koiné." However, Kerswill and Williams (2000) show that older chil-
dren accommodate to and learn from peers, and, for psychosocial reasons,
probably do so more frequently and consistently than do adults.
34. Trudgill also uses reduction as a cover term for the processes of leveling and
simplification.
35. Trudgill (1996: 12-13) also defends this analysis, but this is particularly sur-
prising since in this article he is at pains to demonstrate that small isolated
communities with closeknit social networks are likely to develop and retain
complex changes, while larger communities with frequent contact with other
communities and looseknit social networks are more likely to favor simplify-
ing changes (including reductions in inventories). Of course, the Belfast com-
munities described by the Milroys fit neither of these categories. They are
small communities with closeknit social networks, but they are in constant
contact with other varieties in the city. It is far more likely that the many dif-
ferent allophones of /a/ have been adopted from outside the community
280 Notes

through weak ties and that the closeknit social networks have allowed their re-
tention in a relatively complex system.
36. Siegel (1985) and Mesthrie (1994) have both questioned whether simplifica-
tion necessarily occurs in koineization. When contributing varieties show sig-
nificant differences in linguistic structure, simplification is sure to occur (at
least in comparison to some of the contributing varieties). When differences
between varieties are limited to allophonic variation (as Kerswill and Williams
found in Milton Keynes) then only leveling will be found.
37. Note that transparency is conceived of here as an effect of speaker activity and
learning, and as an influence on speaker-learners, but not as a systemic objec-
tive, as in the work of Lightfoot (e.g., 1979) and Kiparsky (1982).
38. In fact, koineization is dependent on the co-occurrence of "increased interac-
tion among speakers of various dialects" and "decreased inclination to main-
tain linguistic distinctions".
39. Chambers (1992) presents a very informative analysis of the acquisition of
English English by six Canadian children/adolescents who had recently moved
to England with their families. Chambers defines and provides evidence for
eight principles of dialect acquisition (which focus primarily on the learning of
phonetics/phonology and lexicon, precisely those components of the grammar
that tend to differ most between dialects). There are some differences between
SLA and SDA, but these are fairly easily explained by differing contextual
needs of speakers. For example, Principle 1 (Lexical replacements are ac-
quired faster than pronunciation and phonological variants) is exactly in line
with SLA, in which learners begin by learning vocabulary items, which are by
definition more salient than phonological items. Principle 2 (Lexical replace-
ments occur rapidly in the first stage of dialect acquisition and then slow
down) indicates a difference, but this is partly due to the fact that learners do
not have to learn more in order to communicate. Other aspects fit neatly in line
with the predictions of SLA theory: Principle 3 (Simple phonological rules
progress faster than complex ones) is expected if learners overgeneralize
common rules and delay the learning of exceptional rules. Principle 4 (Acqui-
sition of complex rules and new phonemes splits the population into early ac-
quirers and later acquirers) reflects the fact that children will normally learn
"perfectly" any variety to which they have sufficient exposure and interaction,
no matter how complex it may be. Principle 5 (In the earliest stages of acquisi-
tion, both categorical rules and variable rules of the new dialect result in vari-
ability in the acquirers) reflects the fact that all interlanguages are highly vari-
able, and that learners often switch between a "careful" style and a
"vernacular" style. Principle 6 (Phonological variants are actuated as pronun-
ciation variants) reflects the fact that adult and child learners begin by learning
isolated lexical items, and only when knowledge of a critical mass of instances
has been attained do they become rule-governed. Principle 7 (Eliminating old
 Notes 281

rules occurs more rapidly than acquiring new ones) is dependent on the struc-
tural similarities between dialects and is predicted by accommodation theory,
but it is also the most controversial of Chambers' principles. This is so be-
cause in order to explain his evidence, he is forced to argue that "absence" of a
historically novel merger (=acquisition of a lexically unpredictable phonologi-
cal split) is an example of elimination. This reasoning seems difficult to ac-
cept, and it directly contradicts the findings of Kerswill (1996), who argues
that new phonological oppositions and lexically unpredictable phonological
rules are among the most difficult features to acquire. Principle 8 (Ortho-
graphically distinct variants are acquired faster than orthographically obscure
ones) reflects the impact of salience (provided in this case by the writing sys-
tem) on learning, common to both SLA and SDA.
40. It should be noted, however, that adults may overgeneralize more than chil-
dren for some features. Thus, Bybee and Slobin found that adults tend to
"double-mark" past tense and produce forms such as hitted, which small chil-
dren tended to avoid, perceiving the final -t of hit as a past tense marker itself.
41. Ravid's study of psycholinguistic processes and their relation to language
change in modern Hebrew is especially pertinent to the model of koineization,
since many of her observations about change in Hebrew take for granted the
quasi-koineizing social situation of Israeli Hebrew (cf. Blanc 1968). In this
context, Ravid shows literacy (and/or standardization) is the primary conserva-
tive force that preserves irregularities.
42. It is not certain whether this limitation is due to cognitive or social factors, or
a combination of both. Payne's sociolinguistic study (1976, 1980) of children
in King of Prussia, Pennsylvania revealed that certain features of the Philadel-
phia dialect (fronting of /ow/ and /uwΓ) were only acquired regularly if chil-
dren moved to the area by age 8 (and the complex short a pattern was only ac-
quired if the children and both of their parents were from Philadelphia). This
is regularly cited as evidence for a cognitive or maturational constraint on the
age of language learning, but Labov (2001: 430) carried out a multiple regres-
sion re-analysis of Payne's data and found that the most significant independ-
ent variable was the degree of integration in the peer group.
43. Other evidence indicates that frequency of occurrence remains a strong predic-
tor. For example, Trudgill, Gordon, Lewis, and Maclagan (2000: 310) report
that the English phonemes /w/ and /MJ had merged in southeast England and
probably in Australia at the time of the main immigration to New Zealand in
1840-1880. Although many colonists came from southeast England and Aus-
tralia, they were far outnumbered by colonists from Scotland, Ireland, and
northern England. These speakers consistently preserved the distinction, and it
survived as a norm of the New Zealand koine. On the other hand, Britain
(1999) analyzes a minority complex feature propagated during dialect mixing
in New Zealand English. This is the split in pronunciation of (-own) words
282 Notes

into a monosyllabic groan class and a disyllabic grown class. Britain argues
that this change was favored because the novel disyllabic articulation in-
creased the degree of correspondence (transparency) between a grammatical
category (participle) and its expression.
44. The discussion of psycho-social stages of development is based on Giddens
(1989: 82-85).
45. As the discussion in Kerswill and Williams (2000) indicates, the term genera-
tion is best understood as a convenient shorthand for reference to the succes-
sion of overlapping cohorts of children and adolescents.
46. This sort of lingering competition between two strong variants may be exactly
what is required for reallocation to occur.
47. Mashlum (1992) studied the linguistically "chaotic" society of Spitsbergen,
Norway (where no permanent community exists), and found that the children
employed a variety of "strategies of neutrality" to negotiate the mixed and un-
stable sociolinguistic context. Such strategies include code-switching, mutual
accommodation and code-mixing, and use of the relatively neutral standard-
like East Norwegian accent.
48. There may have been incipient processes of koineization within each group,
which if maintained would have led to the appearance of two koines, one mid-
dle class and the other working class.
49. Nevertheless, Mufwene (2001) betrays a certain ambivalence: he criticizes the
distinction between koineization and creolization, but he also contrasts koines
and creóles throughout the work. See Chapter 6 for further discussion of these
issues.
50. This passage reflects Thomason and Kaufman's larger concern with the circu-
larity of appeals to markedness as a kind of linguistic universal. Here, I follow
Trudgill in assuming that markedness is dependent on the structure of the par-
ticular language varieties in question, and can be equated with "exceptional"
or "less frequent".
51. Kerswill and Williams' (2000) describe the different results for koineization in
a new town (Milton Keynes) and strict dialect leveling in a relatively stable
and established old town (Reading). The actual results seem to be the same in
the long run, but the changes are more gradual in the old town, and relatively
sudden in the new town, where the first generation of children has already de-
fined new norms and is not familiar with many of the traditional features of the
region. In the old town, the youngest generation retains some features typical
of the oldest generation.
52. Population movements toward the west tended to group settlers according to
their place of origin, so that the North (Great Lakes region) was populated
primarily by New Englanders, the Lower South by lowland Southerners, and
the Midlands by a great mix of people from all areas. This created three hori-
zontal dialect tiers stretching from the eastern dialect zones to about the Mis-
 Notes 283

sissippi River, where differences between them begin to disappear (Wolfram

and Schilling-Estes 1998: 103-108). The development of dialect areas in me-
dieval Spain looks very much like this, but shifted to a north-south axis.
53. See Granda (1994), Rivarola (1996), and the studies in Hidalgo (2001) for
other factors that may have played a role in koineization and regional differen-
tiation of American Spanish.
54. It seems that these new communities were founded with the very intention of
weakening resistance to the standard. Versteegh describes a case of settlement
of the reclaimed land of the Zuider Zee (in 1936); the Dutch government de-
cided to select new inhabitants from a variety of regions, religions, profes-
sions, and dialects: "It was thought that a representation of all Dutch regional
accents would gradually lead to the disappearance of the regionalisms and the
emergence of a new interregional standard, which would be better suited to
serve as the means of communication of these carefully selected inhabitants"
(Versteegh 1993: 65).
55. The social history given here represents a synthesis of earlier research. I have
consulted Sánchez Albornoz (1966), Pérez de Urbel (1945), Menéndez Pidal
(1964), Menéndez Pidal (1960); however, given the extreme positions adopted
by some of these, I have also relied heavily on Μοχό (1979), still recognized
as an excellent and balanced account of the social history of medieval Spain
(cf. Mitre 1999: 103), as well as more recent manuals and reviews: Riu Riu
(1989), Minguez (1989), Iradiel et al (1995), Torres Fontes et al. (1990),
Tusell et al. (1998), and García de Cortázar (1991). I have included some use-
ful references from Pastor (1996), but his central thesis - that Castile was re-
populated by natural internal growth and that immigration played only a minor
role - remains unconvincing in the face of significant evidence to the contrary.
Pastor's larger aim seems to be to reject the tradition of Castilian exceptional-
ism by emphasizing parallels between early Castilian history and that of other
parts of Europe. However, to do so he is forced to ignore the large demo-
graphic movements that are so characteristic of Iberian history.
56. I leave aside the questions of how and why use of Basque persisted. See Trask
(1997) for an excellent history of this language.
57. Menéndez Pidal (1960, 1964: 442-443) provides toponymie evidence of the
mixed origins of settlers, gleaned from contemporary documents of the period,
such as the name of the village of Toldanos, near the city of León (and others
like it), indicating the arrival of a group of Mozarabs from Toledo, or simi-
larly, placenames such as Gallegos, Galleguillos, Astureses, and even Castel-
lanos.
58. One might also ask what attracted settlers to this region rather than to the safer
lands to the west. Certainly proximity to home was key for settlers from the
east and north. However, Castile's proximity to the Moorish regions played a
role, since the Moors represented both a threat - for their military power - and
284 Notes

an attraction - for their wealth and sophistication. Soldiers also had the oppor-
tunity of gaining booty in raids on Moorish communities, and it is known that
at least one Castilian count dressed in Moorish style (Menéndez Pidal 1964:
472-482).
59. Menéndez Pidal lists names of towns, found in contemporary documents, that
attest to the Basque migration: Báscones, Basconcillos, Billabáscones, Bas-
cuña, etc. (1964: 473), as well as placenames of Basque origin, such as Zaldu-
endo. These tend to appear in the northeastern part of the county. Such
toponymie evidence is difficult to interpret. It may be that some such homoge-
neous communities resisted the dominant regional norms and developed their
own (e.g., maintenance of Basque, which we know occurred in the nearby Rio-
jan valley of Ojacastro). However, it appears that relatively few communities
were founded by homogeneous groups of immigrants (including non-
Basques): Pastor (1996: 76) reports that only 2% of village names in southern
Castile indicated a place or group of origin. Moreover, in the case of the
Basques, Romance names such as Villabâscones highlight their minority
linguistic status.
60. Sánchez Albornoz (1966) refers to serfs, but some recent scholars argue that
workers tied to the land (and without the rights of hombres libres) in this early
period were probably slaves in a system carried over from Roman and Visi-
gothic times, while serfdom only appears later, with the penetration of feudal
ideology (cf. Pastor 1996: 280).
61. The infanzones would rise further in social status in 11th-century Castile. For
example, the Cid was an infanzón, but in his youth he was a close advisor of
the king of Castile and, after his conquests, he married his daughters to roy-
alty.
62. The social effects of the frontier have often been overestimated (see the col-
lection in Hofstadter and Lipset 1968), but its effects are (in part) those which
accompany mixing and the breakdown of social networks. The frontier is
therefore a specific manifestation of more general kinds of social change,
which may interact with the cultural values of the immigrants (Lipset 1968:
12). Lee (1968: 69-70), for example, pointed out that frontier societies are
simply examples of the process of migration, which often leads to a decline in
the importance of hereditary elites and extended families. However, the isola-
tion produced by migration to a frontier is a feature that may further social
change and contribute to the formation of new identities. With regard to Cas-
tile, the linguistic features which came to be seen as prototypically "Castilian"
may actually have begun as variants which distinguished the frontier people of
the Duero Valley (including much of the flat, open zone between Burgos and
León) from other Christians who lived to the hilly west, north, and east.
63. Wright (1994b) has argued that some ballads may be much older than previ-
ously thought, and many perhaps survive from the early Middle Ages.
 Notes 285

64. The general lack of literacy also favored a weakening of influence of written
prestige norms. Even among the literate, the writing system of these centuries
had become highly logographic (see the comments in Wright 1994c) and was
thus less likely to exercise a conservative influence on pronunciation.
65. Wright asserts that it is anachronistic to visualize separate Ibero-Romance
dialects before the 13 th century, since speakers probably did not conceive of
each other as speaking different language varieties (e.g., Galician, Leonese,
Castilian, etc.) during this time (Wright 1994f: 160, 1994d). When I use such
terms I generally do so for the sake of convenience. Nevertheless, speakers
certainly would have been able to identify differences between their own
speech and those of other communities, whatever label they did or did not give
them. Moreover, the development of a regional (linguistic) identity may have
occurred at different time in different places. The oppositional identity of early
Castile along with its political separation would have easily favored an earlier
development of Castilian dialect identity (based on a Burgos-area norm) than
for other varieties such as Leonese (assuming Leonese as a regional variety
has ever been anything more than a creation of modern dialectologists).
66. Alarcos Llorach (1982: 18) had earlier made a more tentative claim that the
rural dialect of early medieval Cantabria was almost a creole or lingua franca
that was utilized by Basque-Romance bilinguals. This, along with López Gar-
cía's work, reflects a general trend in the 1970s of applying early pidginiza-
tion/creolization theory to the development of Romance (e.g., Whinnom 1980;
Schlieben-Lange 1977).
67. Surprisingly, López García tries to use the fact that ser and estar exist in all
Ibero-Romance varieties as proof of the broader existence and use of vascor-
románico beyond the bounds of Castile.
68. López Garcia (1985b) takes a decidedly more linguistic approach to discuss-
ing the linguistic features of Basque that the author claims appear in vascor-
románico.
69. Torreblanca's main point is that spellings with only i and u did not reflect
pronunciation as well as ie and ue, but his attribution of this to carelessness
and omissions implies some kind of written norm which at the time was lack-
ing for these innovations. I am therefore inclined to agree with Menéndez Pi-
dal's interpretation.
70. Penny (2000) identifies other features that can reasonably be seen as the re-
sults of koineization in early Castile. However, he tends not to enter into de-
tailed reconstruction of their development, as I do here.
71. Mozarabic is excluded. Since Mozarabic forms are generally limited to single-
word citations or brief lines of text, there is little evidence of how Romance
articles were treated. Moreover, the Arabic article al came to be used before
both masculine and feminine Romance nouns.
286 Notes

72. The forms without the lateral appear with much greater frequency. Also at-
tested in Old Galician-Portuguese texts are the forms elle, ele, el, apparently
derived from the nominative ILLE.
73. Also found are contractions with the indefinite article: cun, cunha, cuns, cun-
has and contractions with the preposition ca: co, cós, cá, cás.
74. Per was found only in western and central Leonese.
75. Normally, the forms ele, el (< ILLE) appear as nominatives; elo, lo as obliques.
This indicates an added degree of complexity that was lost from Castilian. The
great variety of forms reported for Leonese reflects (in part) the great variation
in local norms within Leon.
76. Though many of the 13th-century documents consulted by Staaff contain only
contracted forms, he points out that the combinations in question are not nu-
merous relative to text size. Moreover, while most long documents show a
mixture of contracted and full forms, some documents contain only iull forms.
One must note, however, that the documents consulted by Staaff were from the
13th century, a period in which Castilian influence was increasing. Indeed,
Staaff (1907: 256) points out that the overall variation in forms increases in
the east (near Castile) and decreases in the west (where contracted forms were
preferred).
77. The contracted forms of both Old and Modern Aragonese are often considered
to be restricted to the combination of en + o. However, in the mountain dia-
lects of Aragonese forms such as the following are found in speech: do 'del' ,
dos 'de los', to 'al', tos 'a los', da 'de la', das 'de las', no 'en el', ñas 'en las'.
78. The forms del and al are the only contracted forms to survive in Castilian.
Note, however, that they are found in a majority of the contributing dialects (a
rough measure of their frequency in the linguistic pool), are among the most
frequently-used preposition + article combinations in any variety (and thus
more likely to suffer reduction), and reflect the coalescence of vowels (com-
mon throughout the history of Castilian) rather than the assimilation of conso-
nants. It is true that the coalescence of a+e normally leads to loss of the first
vowel (at least today), but in the koineizing context speaker-learners probably
expected and produced a combination that preserved transparency. In contrast,
Galician phonetic reductions allowed the (near) erasure of transparency in
some combinations.
79. The article el was used before feminine words beginning with a vowel: el
estrella, el alma. This feature was shared with nearly all contributing dialects
and, unsurprisingly, it was maintained in Castilian. Throughout the centuries it
has suffered progressively greater morphosyntactic and lexical restrictions on
its occurrence. See Janda (2003: 407-408) for a clear and succinct explanation
of the reanalyses that characterize its development. A more conservative form,
ell, the precursor to el (both masculine and feminine), also survived in the
Burgos-phase koine. This form preserved a geminate or palatal articulation be-
 Notes 287

fore words beginning with a vowel. Like "feminine" el, it was found in all con-
tributing dialects, and its frequency made possible its survival. Nevertheless,
by the 13 th century, its use had become restricted to certain high-frequency
lexical items. See Pensado (1999) for a detailed description and explanation of
its development.
80. In forms of the type cono, eno, the article has suffered apheresis and assimila-
tion of the lateral to the preceding nasal of the preposition, resulting in the
production of an opaque form, in which the underlying representation must
differ from the surface realization, at least initially. Thus, cono, articulated
[kono], would have had an underlying representation of /kon elo/ or /kon lo/,
depending on whether the particular speaker had come to consider /elo/ or /lo/
the base form. Speakers of the time would have varied in their perceptions of
underlying structure; some may have learned /kono/ as a separate form that
corresponded to the lexical combination con + lo. Whether learned as forms
generated by rules or as suppletions, the contracted forms still differ from their
conservative or underlying representation. As a result, the analysis made by
learners of the surface output might vary.
81. Smith and Sneddon (2001) have also related the loss of (some) preposition +
article contractions in French to dialect mixing in Paris.
82. In his discussion of conservative dialect features of Cantabria, Penny (2000:
84-88) suggests (directly in some cases, indirectly in others) that many of
these were lost in early Castile. One of these is very similar to the change I
have discussed here. In conservative Cantabrian dialects the final /-r/ of infini-
tives is deleted when followed by a clitic pronoun, so medirla is realized as
[me5ila]. This feature is shared along the northern dialect continuum, from
Galicia, through Asturias and Santander, to La Rioja and Aragon (and was
more frequent in the past). According to Penny it was lost in Burgos, most cer-
tainly as part of the general tendency to reproduce and learn easily-analyzed
transparent units.
83. There is evidence of phonetic conditioning for these different articulations.
The disyllabic [ie], [ia] are found primarily in normally monosyllabic words
(e.g., diez > díaz), but disyllabic [úe], [úo] are found in a wide range of con-
texts. The different vowel qualities of the second element are also (partly)
conditioned by the phonetic context. For example, a following /-r/ favors [ia]
and [ua], following labials and velars favor [uè] and [ie], a front glide favors
[ie] and [ue]. Unsurprisingly, some varieties show a narrower range of varia-
tion, having lost variants with [a], for example, while others, such as those of
Sanabria (in western Zamora) show the full range. Catalán finds three sepa-
rated areas in Leon and Asturias maintain very similar patterns of complex
variation and concludes that these dialects probably preserve the ancient varia-
tion that medieval scribes struggled to represent.
288 Notes

84. Lloyd (1987: 184) also adduces the developments CORIU > cuero, QUO MODO
> cuerno as evidence of the fluctuating realizations of the diphthongs, where
the earlier diphthongs [oj] and [wo] respectively alternated with other diph-
thongal realizations and were later stabilized along with the spontaneous diph-
thongs.
85. Zamora Vicente (1967: 90) indicates that orthographic o and e continued to
dominate in early Leonese texts, for which he suggests the following causes:
1) the proximity and influence of culturally prestigious Galician; 2) influence
of Latin orthographic norms; 3) inability of scribes to represent the diphthongs
(which could, of course, be related to their uncertain phonological status).
These arguments prefigure those offered so insistently by Wright (e.g., 1994c),
who emphasizes that conservative spellings are often opaque to interpretation,
while novel spellings imply a conscious effort on the part of the scribe to adapt
the orthography to the developing sound system. Still, it is not clear that mo-
nophthongal articulations, at least in conservative dialects, need have disap-
peared entirely, given the kind of phonetic variation that has been found by
modern researchers in conservative dialects, and, more specifically, the evi-
dence of Menéndez Pidal cited above.
86. "Pre-Basque had just five vowels i e a o u. So far as we can judge, it probably
also had the modern set of diphthongs, ai ei oi ui au eu. These diphthongs
were not distinguished from sequences of the corresponding vowels, but the
point is that they counted as single syllables for such processes as aspiration
assignment, whereas the other vowel sequences counted as two syllables"
(Trask 1997: 149). Note that there are no falling diphthongs, as in Romance
and Castilian, so Basque speakers may have been especially likely to interpret
these as sequences of two phonemes and produce them as such in their inter-
languages.
87. The vowel [a] is considered the most perceptible while [i] is the sharpest.
Perceptibility is based on the position of the first formant and sharpness is
based on the position of the second formant.
88. What cannot be decided from the evidence is the exact order of these events:
was the articulation of the phonetic diphthong exaggerated before phonologi-
zation (perhaps in primitive Castile), or was the diphthong selected from the
pool of variants at the very time it was being reanalyzed by speaker-learners?
89. Perplexing too is Pidal's insistence that diphthongization before yod is the first
or most ancient form of diphthongization in Romance: "la diptongación ante
yod de e y o abiertas (pueyo . . . ) . . . es la diptongación más general a la Ro-
mania y sin duda de fecha más antigua que la diptongación de e y o en otras
circunstancias . Esa diptongación de y + palatal, muy arraigada en Galia . . .
se extendía por España ininterrumpidamente desde Cataluña hasta Asturias, a
través del mozárabe" (Menéndez Pidal 1964: 495). This view was subse-
quently adopted and defended by Schürr (e.g., 1970), who saw all Romance
 Notes 289

diphthongization as resulting from metaphony with subsequent generalization

to other phonetic contexts. This does appear to describe the development of
diphthongization in many varieties of Gallo-Romance and Italo-Romance, but
it does not fit the Hispano-Romance data (Alarcos Llorach 1996: 17).
90. It is well known that Castilian shows mixed results for diphthongization in this
context as well: cf. monte < MONTE and puente < PONTE. Old Spanish some-
times shows competition between such variants: cf. comde and cuemde (<
COMITE), with the first winning out in the modern language. Catalán (1956-
57) shows that this phonetic environment tends to impede diphthongization in
western Astur-Leonese as well, but it is worth noting that both monte and
conde were often used as preposed (atonic)forms, and in such contexts diph-
thongization would not have been favored.
91. The exceptional Castilian form viejo 'old' (< VETULU) indicates that diph-
thongization before yod had at least begun as a variable phenomenon in His-
pano-Romance at the time of Burgos-phase koineization.
92. Penny (2000: 45-46; 2002) analyzes this spread of the loss of/h/ and its ac-
ceptance as a prestige form as a result of dialect mixing during and following
the move of the Court to Madrid in 1561.
93. Lloyd (1987: 222) interprets these situations rather differently: "In Navarre,
where the Basques remained at home and did not mingle actively with Ro-
mance speakers to form a new linguistic community as in Castile, the Ro-
mance dialect was not impelled to adopt the innovation."
94. This tendency also favored the development of [-jt-] to [tf] in other regions.
Commenting on evidence of medieval Mozarabic forms, which most often
show forms with [-xt-] and [-jt-], Galmés de Fuentes (1996: 107) claims: "en
otras areas de los dialectos mozárabes, en donde no es pensable un influjo cas-
tellano, encontramos, a veces, la palatalización: felech, felecho, felecha, /ele-
ction < FILICTU, ejemplos del cordobés del siglo X Ibn Juljul, de Pedro de Al-
calá y del andaluz Ibn al-Baytar, que prueban, sin duda, la autoctonía del
inicio de la palatalización en los dialectos mozárabes". Of course, the devel-
opments in Castile and eastern Leon were much more regular.
95. The history of changes of CL-, PL-, FL- (and GL-, BL-) begins with a widespread
alteration in the articulation of the lateral. It is generally agreed that speakers
of Late Latin/Romance began to produce a palatalized articulation of the lat-
eral in some or all of the obstruent + lateral clusters (this articulation is pre-
served today in some Pyrenean dialects). One of the most widespread palatali-
zations was that of CL-, and it is often hypothesized that palatalization began
as a reciprocal assimilation of the velar stop and the dentoalveolar lateral in
CL- ([kl-] > [kft.-]), with subsequent partial or complete extension through
analogy to other clusters. In some dialects, such as Tuscan, all the obstruent +
lateral Latin clusters CL-, PL-, FL-, GL-, and BL- were affected similarly. In oth-
ers (e.g., Rumanian), only the clusters with velar consonants palatalized, or
290 Notes

only the clusters with voiceless/tense consonants (as seems to be the case in
Hispano-Romance). It is also generally assumed that nearly everywhere these
"feature-heavy" clusters were eventually "reduced" in some way (Repetti and
Tuttle 1987: 54). In Tuscan, for example, this was done through regular delat-
eralization of the second element (e.g., CLAVE > *[kXáve] > [kjáve]).
96. Coraminas and Pascual (1981: 724) also report Ilaner in some medieval texts.
This could be a Leonese form, however, given the western-style conservation
of -er.
97. The form llantar is preserved today in Asturian, but Coraminas and Pascual
(1981: 573) find few medieval attestations of the form. They also report that
the etymologically-related placenames Llantada and Llantadilla are found in
Cantabria and the province of Palencia (between old Leon and Castile). The
related noun forms pranta and planta are the normal outcomes in Portuguese
and Castilian, respectively. However, there are medieval attestations of llanta
meaning 'plant', though planta was normally used in this sense, while llanta
was used in Castile with the narrower meaning 'cabbage'. Doublets of this sort
are a common outcome of incomplete lexical diffusion.
98. Plegar and pregar meant 'to nail' in early medieval texts. These forms appear
to be doublets of llegar/chegar 'to arrive'. The conservative forms were once
explained away as upper-class or urban forms (e.g., Piel 1931), or as later bor-
rowings from Latin. Wright (1980) and Anderson (1992) have argued that
doublets of this sort are a normal result of the incomplete lexical diffusion
which seems to underlie the development of these clusters.
99. Malkiel (1963-1964) also suggests that particular results may have been influ-
enced by a host of minor factors, including (unfortunately) avoidance of
homonymie clash (clavo/cravo vs llave/chavé), and (more likely) dissimilation
between palatals in the same word (llave vs. clavija, llanto vs plañir).
100. Lloyd's interpretation (i.e., that the change progressed from CI > CX > λ > tj)
is the only one that is consistent with the linguistic outcomes presented here.
The theories put forward by Malkiel (1983) and Repetti and Tuttle (1987)
must therefore be rejected. Torreblanca (1990) argues similarly, but suggests
that the initial consonant in each cluster first aspirated and was then lost. This
seems unlikely in word-internal post-consonantal position.
101. Today Cantabria and eastern and western Asturias (but not central Asturias)
show phonetic alternation between [1] and [r] in /CI-/ and /Cr-/ clusters
(Zamora Vicente 1967: 138; Ñuño Álvarez 1996: 188).
102. Franco was a term used to designate anyone from outside the Peninsula, pri-
marily persons of Gallic origin, but also including people from the areas of
present-day England, Germany and Italy (Μοχό 1979: 263).
103. Alvar (1968) presents a detailed discussion of the linguistic effects of French
settlers in Aragon. See Frago Gracia (1991a) and Imhoff (2000a) for discus-
 Notes 291

sion of the effects of dialect mixing in Aragon. Imhoff (2000b) also compares
possible effects of dialect mixing in Aragon and Castile.
104. Excepting, of course, forms with paragogic -e used in verse.
105. This argument broaches the interesting side question of the merger of /d/ and
/δ/ (assuming these were indeed separate phonemes in Hispano-Romance).
Alvarez Rodríguez (1996: 33, 39) argues that normal apocope occurred only
in ungrouped apical consonants that were "originally voiced" and "not folly
occlusive". Thus, the loss of /-e/ following /-d/ (< τ) could not take place until
the consonant had spirantized. He hypothesizes that this is a 12th-century
change, since he finds no evidence for apocope following this consonant until
1146 (in a document from Toledo). Torreblanca (1986a: 15) also points out
that, up to the 12th century, Hispanoarabic transcribers of Spanish used the
letters dal and δαί to represent the sounds [d] and [δ], respectively. In the 12th
century, however, the transcribers begin to use these two symbols inter-
changeably. Torreblanca believes that this signals the loss of distinction be-
tween the two sounds. Certainly, such a merger would have been favored by
the massive movement and mixing of the 12th century, but there is no ortho-
graphic evidence available, unless one sees this as part of a chain, the second
lenition. In this case, the loss of intervocalic -d- (<D), as in suor < sudar, can
also be included as evidence, and such forms begin to appear in the 12th cen-
tury as well. Another problem is that when these forms suffered apocope, they
were often represented in the orthography by final -t (normally representing a
voiceless stop). Since other alternations seemed to show maintenance of the
manner of articulation (e.g., nieve > nief, Lobo > Lop), this might lead one to
conclude that the -d was still - at least variably - occlusive. Still, at the time
there was no means of indicating orthographically the devoiced equivalent of
/δ/.
106. Galmés de Fuentes argues against significant apocope in Mozarabic, since he
finds that scribes normally made an effort to represent Romance words with
the final vowel, at least in Toledo and Seville (1983: 71, 192). However,
Peñarroja Torrejón (1990: 241) argues that apocope was prevalent in Valen-
cian Mozarabic, which would support the pre-Invasion origin of apocope, at
least in the east.
107. Some areas were slower to accept the new norm. Staaff (1907: 210) points out
that in Leonese the tendency to maintain final -e after apical consonants was
retained into the 13th century (e.g., sacare, abere).
108. Lapesa (1951: 188-190) also points out that Arabic borrowings and Mozarab
names ending in final consonants could have contributed to these develop-
ments in the 10th and 11th centuries. However, he finds that most of the terms
were regularized to the dominant phonotactic structure through the addition of
final -e. Still, such borrowing continued unabated and may have also contrib-
uted to opening up the system to new word-final consonants and clusters.
292 Notes

109. There were clearly other constraints on appearance of apocopated forms. For
example, there was an almost total lack of apocope in the 1st and 3rd persons
singular of the present subjunctive of -ar verbs. Luquet (1992) offers a possi-
ble explanation for this restriction.
110. These texts date from the latter half of the 13th century, but they probably
represent a traditional usage within the area of Toledo, what Hernández (1991 :
330) refers to as the norma culta of the time - although this norma was clearly
variable.
111. It seems reasonable to conclude that such alternation would have represented
the penultimate stage to the loss of apocopated forms, for, as Harris-Northall
points out: "In such circumstances, it is questionable from a phonetic point of
view whether apocope had actually taken place; we might just as accurately
speak of coalescence of vowels, a common enough occurrence in the history
of Spanish, and, of course, in the modern spoken language, though not nor-
mally reflected in the standard orthography" (Harris-Northall 1991: 34). The
results of Harris-Northall's analysis discontinued this hypothesis.
112. Harris-Northall excluded apparently phrase-final tokens from his tabulations.
113. Moreno Bernal (1993), in his study of Escorial Bible 1-1-6, reports a signifi-
cant correspondence between phonetic context and apocope: "dans des condi-
tions favorables (devant une pause ou une voyelle), nous trouvons plus de
75% des cas d'apocope, tandis que, dans ces conditions moins favorables
(devant une consonne), nous relevons précisément le contraire: 75% des cas
de conservation de la voyelle finale" (1993: 195). This still leaves in need of
explanation some 25% of instances of apocope or lack thereof. Going beyond
Lapesa's observation that apocope was probably more frequent in writing,
Moreno claims that apocope was completely restricted to written language and
the variation between full and apocopated forms simply represents conscious
manipulation of the rhetorical device of variatio (1993: 201). While this last
may be true (and is conceivable within the approach I take), his rejection of
apocope as a phenomenon of the spoken language stands in clear contradiction
to his own assertions that phonetic context was an important factor in deter-
mining the appearance of apocopated forms. In addition, such a claim fails to
consider an important phonetic effect of apocope: the devoicing of final con-
sonants (e.g., nief for nieve).
114. Staaff (1907: 213) observed in 13th-century Leonese documents that apocope
occurred principally in eastern regions near Castile, but was little used in cen-
tral and western Leonese. Lapesa pointed out, however, that "en los comienzos
de la apocope extrema los ejemplos leoneses no son menos que los castellanos
. . . aunque luego escaseen" (Lapesa 1951: 186). It may be that a simple rela-
tion (French influence > extreme apocope) was valid for northern Leon, but
not for Castilian zones.
 Notes 293

115. Personal leísmo is the only one of those related to leísmo, laísmo or loísmo
that today receives the approval of the Royal Spanish Academy (Abad 1985:
14). The prestige associated with this acceptance by the Academy has appar-
ently contributed to its modern spread, at least in formal registers, to regions
which traditionally maintain the etymological pronoun system (Quilis et al.
1985: 37).
116. Interestingly, the narrow area of primitive Castile showing system H (includ-
ing the town of Espinosa de los Monteros) once formed an administrative unit
with the neighboring Cantabrian Valle de Pas, which also shows system H
(Penny 1969: 27).
117. System A is modified even further by some speakers (Fernández Ordóñez
1994: 86). This system, labeled system A', results from the loss of marked use
of lo for feminine mass reference (a tendency of variable strength in nearly all
the systems where it exists), and a concomitant loss of its use for masculine
mass reference. System A' thus loses completely both case and count/mass
distinctions, and the pronoun lo disappears except for neuter reference. This is
the most altered or advanced of any of the pronoun systems, but given that it
clearly arises as an adaptation of System A, it will not be further discussed
here.
118. Examples of variable constructions include those in which differing degrees of
agentivity are expressed by use of the accusative or dative: Aquellos amigos lo
interesaron en la política 'Those friends interested him in politics.' vs. A Juan
le interesa la política 'Politics interest Juan'. Certain constructions with infini-
tive clauses can also affect pronoun use, depending on whether or not the in-
finitive itself takes a direct object: Su madre no la dejó subir 'Her mother
didn't let go up' vs. Su madre no le dejó conducir el coche 'Her mother didn't
let her drive the car'. Some verbs can omit their direct object, though it is un-
derstood from the context: Los enseña '(s/he) teaches them (=the facts)' vs.
Les enseña '(s/he) teaches them (= the children) (the facts)'. In Northern Spain
the verb llamar takes the accusative for permanent names, but the dative for
nicknames and such: Cuando nació mi hija, la llamamos María 'When my
daughter was born, we called her Maria' vs. Aunque se llama María, todos le
llaman Irene 'Although her name is Maria, everybody calls her Irene'. Signifi-
cantly, Fernández Ordóñez (1999: 1335) indicates that use of the accusative
has been generalized to both contexts in Zamora, Salamanca and all of south-
ern Spain.
119. The best potential source of this information (at least for geographical distri-
bution in the late 12th and 13th centuries) is the Documentos lingüísticos of
Menéndez Pidal. However, as Echenique Elizondo (1981: 135) and Sanchis
Calvo (1991: 811) have pointed out, these texts offer relatively few examples
of /', lo or le, and are thus unusually poor sources of information on leísmo.
This probably stems from the need for clarity of reference in legal texts, which
294 Notes

leads to an extreme redundancy in which full names and nouns are repeated
and the pronouns le and lo avoided.
120. The exact source of this form is not known. Atonic articulations of the pro-
posed adjectival forms may have favored the weakening of the final vowel.
But raising of final -a is a well-known feature of central Asturian (la vaca >
les vaques), and it may have retained a place in the earliest Castilian koine
(giving mies, then mie as a backformation). On the other hand, the medieval
variation between -ia and -ie in the imperfect verb forms may reflect a broad
phonetic trend extended even to possessive forms, producing variants mía-míe,
which in turn may have affected the development of túa-túe, súa-súe.
121. The relative lack of early documentation of second person forms requires that
discussion of these forms be based on the evidence for the third person. Pre-
sumably the second person forms would have followed analogically the devel-
opment of the third person.
122. An interesting question, not yet investigated, is the origin of the
loss/weakening of phonological distinction between non-high vowels /e-i/ and
/o-u/. It may be that this was yet another result of Burgos phase koineization
but which Latin orthography, literacy/standardization, and the pattern of tonic
vowel distinctions have tended to erase in recent centuries. See Penny (2000:
133-134) for discussion.
123. For the documents in Orígenes Pidal commented: 'No he recogido ejemplos
de la preferencia por la forma femenina mi, su que prevaleció en castellano'
(1964: 346). The leveling to su(s) and tu(s) appears to be a Toledan innova-
tion.
124. The abbreviations are those established by the Hispanic Seminary of Medieval
Studies (manuscript production dates are included in parentheses) GE4 = Ge-
neral Estoria IV (1280); MOA = Moamyn - Libro de las animalias (1250?);
EE1 = Estoria de España I = Primera Crónica General (1270-1284); GEI =
General Estoria I (1272-1275); LEY = Libro de las leyes (12567-1265?);
CRZ = Libro de las cruzes (1259); LAP = Lapidario de Alfonso Χ (1250?-
1279?).
125. For example, LEY includes 5 tokens of mio + masculine noun, and 9 of tokens
oîmi(s) + feminine noun. There are no cases of confusion.
126. Scholars have recognized but do not agree on the causal relation between
shortening and deaccentuation. With specific reference to the 1st person,
Malkiel comments: "The shortening was, in turn, conducive to loss of stress in
tone-setting New Castile (hence, in standard Spanish, proclitic mi-padre), with
the further consequence that the definite article before the possessive adjective
became dispensable, except in archaic formulas." (1976: 473). Penny, on the
other hand, implies that the loss of tonicity was key: "When used adjectivally
before a noun, the possessives lose their tonicity, shed their final vowels, and
(in the case of to(s), so(s)) suffer raising of their newly atonic vowels' (1991a:
 Notes 295

127). These phenomena may have been simultaneous and mutually reinforcing
in the koineizing context. Penny (1969: 114) provides some evidence of the
impact that stress could have. In Cantabria (Valle de Pas), where stress on the
possessives was maintained, all the possessives were eventually leveled to
long forms for all functions: míyu, míya, míyu, túya, túyu, túyif, etc.
127. The value of the reconstruction and explanation for Castillan possessives can
be highlighted by briefly comparing them to developments in neighboring
Aragon and Leon. In Aragon, where massive demographic mixing also oc-
curred during the 12th century, there is documentation that reveals complete
leveling to so; the Liber Regum (Cronicón Villarense), the manuscript of
which dates from before 1211 (Cooper 1960: 7), contains only the form so(s),
used for both masculine and feminine reference, with 33 cases of so + femi-
nine (Cooper 1960: 88-89). Gallic and Catalan influence may have been the
deciding factor in Aragon. These immigrants, arriving in the 11th century,
would have used forms such as masculine son/ton, whose vowel corresponded
to that of the Hispanic form so, even as the feminine forms they employed -
favoring -a - tended to weaken the frequency and consistency with which the
equivalent Hispanic forms were used. For Leon, Staaff (1907: 273-274) re-
ports that 13th-century northeastern Leonese documents reveal limited evi-
dence of leveling: 3 occurrences of sos + feminine, but 4 cases of sues + femi-
nine and 5 cases of sus + feminine. The leveled forms to so(s), to(s) later
predominated in central Leonese and spread to central and eastern Asturian,
where they survive today (Menéndez Pidal 1962a: 96). In Asturian and north-
ern Leonese, the broader association of -u with masculine reference and -o for
neuter reference may have contributed to the selection of so and to as a less-
marked variant.
128. Basic events and dates referred to here can be found in the standard histories
on the topic. I have relied primarily on Μοχό (1979: 349-382), González
Jiménez (1982: 97-127), and Bishko (1975: 422-439).
129. Properties were generally distributed as donadíos or as heredamientos.
Donadíos were generally made by the king, primarily to noble lords or eccle-
siastical institutions as vassals, who then became responsible for overseeing
the repopulation of these areas. Heredamientos were generally smaller, went
to those lower on the social scale (<caballeros de linaje, caballeros ciudada-
nos, and peones), could only be sold after a stipulated period of several years,
and required military service and residence in the city (González 1951: 256).
130. However, Andalusian has often been associated with yeísmo, or merger of /λ/
and 1)1, which has spread in the north of Spain during the 20th century.
131. The symbols are used to represent the four non-palatal sibilants of medieval
Castilian: /á/ = voiceless apicoalveolar, Izl = voiced apicoalveolar, /ts/ = post-
dental voiceless affricate, /d7 = post-dental voiced affricate. In all dialects of
Castilian the last two deaffricated, possibly during the period under study.
296 Notes

This reduced the distinctions between the phonemes, and made their merger
more likely. In central and northern Spain (and in eastern Betic Andalusia), the
voiced and voiceless sibilant phonemes merged, leaving the precursors to to-
day's /s/ and /Θ/ (the articulations of which have subsequently been differenti-
ated by increasing the distance between their points of articulation). In Anda-
lusia and all American varieties, even this distinction between apicoalveolar
and dental sibilants was lost - in the process known as seseo - reducing the
sibilants to just one: /s/. Ceceo refers to this same merger, but with selection of
the post-dental (or interdental) allophone as the prototypical articulation. In
some discussions, çeçeo (merger of voiceless sibilants) is distinguished from
zezeo (merger of voiced sibilants). This distinction is based on the assumption
that devoicing occurred only later. The relative chronology of these develop-
ments is one of the principal issues discussed by historical linguists studying
the history of Andalusian.
132. The most frequently discussed features are phonetic or phonological, including
seseo (çeçeo-zezeo), aspiration and loss of /-s/, neutralization and loss of /-V
and /-r/, yeísmo or merger of /y/ and /λ/, as well as sibilant devoicing, mainte-
nance of aspiration of Latin F- and the use of [h] instead of northern [χ] for /x/
(as in the first sound in jamón).
133. Convincing as this argument first appears, it is riddled with problems. First,
the repopulation of Granada cannot be limited to the 15 th century; in fact, it
only began from the late 1480s, and there were significant repopulations after
the forced conversion (and partial exodus) of the Mudejars in 1501-1502 and
after the Morisco rebelión of 1568-1570, in which the Moriscos were exiled
to other areas and the whole of Granada repopulated with settlers from Betic
Andalusia and the rest of Spain, particularly the northwest (Salvador Salvador
1978: 99). Moreover, Frago ignores the likelihood that the population move-
ments associated with Granada and America may have helped provoke many
of the changes that he claims must have been transplanted wholesale. Would
not simultaneous and similar demographic and dialect mixing in Granada,
Seville and America have led to similar outcomes? This is precisely what
Trudgill, Gordon, Lewis, and Maclagan (2000) argue for Southern-
Hemisphere varieties of English and southeast-of-England (London) English.
134. Elsewhere, Frago argues against Alonso's assertion (based on 16th-century
observations by Arias Montano) that seseo and devoicing spread between
1546 and 1566 in Seville. This is certainly extreme, but Frago denies any sig-
nificance to 16th century social and demographic change: "en modo alguno es
posible admitir que en tan breve lapso pueda alterarse radicalmente la pronun-
ciación de una ciudad" (Frago Gracia 1993: 24). Still, this time lapse amounts
to (nearly?) one generation, so it is conceivable for the effects of koineization
to have first been noticed by older observers in this amount of time.
 Notes 297

135. Frago Gracia (1991b) includes a discussion of his views on language change,
in which he argues for the Pidalian roots of his thinking on this matter. It must
be pointed out, however, that Menéndez Pidal clearly saw change as occurring
at different rates (as his comments in Orígenes make clear). Other scholars,
including Mondéjar and Ariza sometimes echo Frago's view of the absolute
slowness of change.
136. Penny (1983) and (1991b) also suggests a northern origin for many Andalu-
sian features. He also suggests a western influence for the loss of leísmo in
Andalusian (1991b: 39-40), since the relative simplicity of the etymological
system would have favored its selection.
137. Staaff described the restitution of full forms that marked the end of extreme
apocope as the imposition of the Leonese tendency over that of Castilian. This
view has been scoffed at, but seen in the context of dialect mixing it acquires
new relevance.
138. Even if we restrict the analysis to documents from the reign of Alfonso (1252-
1284), the results are only slightly altered, with 8 occurrences of (a)delant
(11.8%) and 60 of (a)delante (88.2%).
139. Trudgill (1986: 56) explains Steinsholt's (1962) concept of language mission-
ary or spràkmisjoner as a person who has left his or her home community,
travels or lives elsewhere and thereby acquires new linguistic features, but
who eventually or periodically returns to the home community. As an insider
to the group, such a person is especially likely to spread innovations to speak-
ers in close-knit social networks who otherwise might reject the innovations.
140. Evidence for the appearance of feminine mi(s) and the disappearance of
mia(s)/mie(s) corresponds closely to that of the leveling of other forms to tu(s),
su(s). In DLE documents from Toledo, I find mi mugier (269-1210), mi fin
(274-1221), la mi heredat (276-1228), with the last examples of older forms
being mie muger and mias eguas (270-1212). In the more plentiful Burgos
documentation, I find a very early example, miuida (147-1100), and mifuessa
(154-1200), mi uida (157-1206), mi heredad (160-1209), with the final ex-
ample of the older forms being mie muger (181-1228). Oddly, even in Bur-
gos, the variation in use of first person feminine forms is eliminated faster than
in the third person, even though the opposition between mi/mio is maintained
longer.
141. There is evidence of early sporadic innovations in Toledo and Burgos. Mén-
dez García de Paredes identifies a handful of occurrences of generalization of
mio to feminines: "Durante los primeros años del siglo XIII escasean los ejem-
plos de posesivo masculino referido a sustantivo femenino -«. . . che fue de
mio padre de mio madre» (Burgos, 1206), «ego Michael Gonzaluez mio
mulier Maria Martin . . .» (Soria?, 1225)- similares a los encontrados de so +
femenino, y a que ahora sí parecen apuntar ima incipiente confusón de géneros
orientada hacia el femenino" (Méndez García de Paredes 1988: 539). In Tole-
298 Notes

dan texts from the first half of the 13 th century there appears only one exam-
ple of use of mi with a masculine noun: mi fijo (279-1239).
142. Lapesa (1968: 526) used the apparent lack of leísmo in Andalusia as evidence
for his argument that apocope and leísmo may not have been causally linked,
since for him apocope was fully extended to early Andalusian, while leísmo
failed to survive there. He assumed that if apocope had played a role in the
development of leísmo, it should also have led to leísmo in Andalusia.
143. For example, lexical mixing in Andalusian has been frequently commented
upon, with some awareness that some of this variation must have originated in
the 13th century: "Los efectos de la reconquista conjunta de Andalucía occi-
dental por Castilla y León se observan también en el vocabulario. Los prime-
ros textos andaluces registran voces de aire occidental, como prato, branco·,
hoy se usa en Córdoba prata. También es frecuente la d- protética, corriente
en numerosas formas leonesas: dir, dalguno, dambos, o algunas voces como
esmorecerse 'trasponerse de ira, desmayarse', usual en las comarcas del occi-
dente peninsular" (Zamora Vicente 1967: 327). Other studies have also
pointed to the strong western influence on the western Andalusian lexicon,
such as Alvar (1964) and Navarro Carrasco (1985). However, these are all
synchronic studies from which only assumptions can be made regarding the
historical situation. Even the study by López de Aberasturi (1992), which re-
veals the evident antiquity of many Leonesisms in Andalusian, cannot show
clearly that they date from the 13th century, since there have been later periods
of influx of westerners, particularly the 16th century. The two great population
movements that have taken place in Andalusia, and the effects of its subse-
quent administrative separation into a western half focused on Seville and an
eastern half focused on Granada (Ariza 1992: 18), have tended to blur any
clear patterns in the contemporary data. Lack of knowledge about the demo-
graphic history can cause serious problems; for example, Navarro Carrasco
(1985: 77) attributes use of the western form lamber in eastern areas of Anda-
lusia to Mozarab influence, even though few if any Mozarabs remained in
Andalusia or Granada at the time of their reconquests.
144. For example: "En la Andalucía primeramente reconquistada se formarían áreas
de seseo, áreas de ceceo y puntos de coexistencia de las modalidades fonéti-
cas" (Frago Gracia 1993: 366). I am primarily concerned with the phonologi-
cal merger that seseo represents and do not discuss the development of the
phonetic variants known as seseo and ceceo (in which one or the other of the
main allophones of the earlier phonemes was selected and generalized for the
new merged phoneme). Frago's discussion, however, is perplexing, since he
assumes that ç written in place of both s and ç/z indicates ceceo, and that s
written in place of s and ç/z indicates seseo. But, even assuming that seseo and
ceceo originated as clearly distinct articulations, can one assume that medieval
writers would necessarily have associated seseo with s and ceceo with c?
 Notes 299

145. Frago points out numerous early examples of confusion in the orthographic
representation of voiced and voiceless phonemes of each pair (Frago Gracia
1993: 213-300), not only in Andalusia, but elsewhere as well. Nevertheless,
he hints at a special role for Andalusia when he comments on the (apparent?)
spread of sibilant devoicing in the 13th century (Frago Gracia 1993: 283).
Still, Lloyd (1987: 329) offers numerous contemporary examples from Burgos
that are quite similar to those presented by Frago.
146. Given Frago's contention that the change is slow, it seems fair to allot a sub-
stantial time period. Extending the time frame of analysis to 1500 should pro-
vide a sufficiently long period for the evidence to accumulate (over 250
years). At that point, the second great phase of Andalusian population move-
ment was beginning and would be a more likely catalyst for change.
147. The 15th-century evidence offered by Lapesa ([1957] 1985) suffers from the
same problem. Particularly noteworthy are the examples from the Cancionero
de Baena; these amount to a tiny handful, yet the Baena includes some 28,000
lines of text. On the other hand, as Harris-Northall (1996a: 322-323) points
out, there is significant orthographic evidence of confusion of -ss- and -s-, and
I have found numerous and varied tokens of confusion of ç and ζ in the con-
cordance prepared by Morgan (1974). For example: 2 razion - 1 ragion, 4 ra-
ça - 1 raza, 2 satisfaçe - 4 satisfaze, 1 neçios - 1 nezios, 48 fazen - 2 façen, 5
coraçones -1 corazones, 6 viçios -1 vizios.
148. The same argument is implied in the following passage: "Ejemplos aislados de
seseo-ceceo . . . en un texto como el que hacia 1495 daba el Doctor Cisneros,
sin duda significan lingüísticamente tanto como un inventario considerable-
mente mayor de cacografías debidas a escribanos de menor cultura" (Frago
Gracia 1993: 31). However, as we have seen, Frago himself is unable to pro-
vide a convincing inventory of errors made by less educated scribes.
149. Ariza clearly views seseo as originating and spreading from Seville in the 16th
century: "Si echamos una ojeada al mapa . . . podemos comprobar que la con-
fusión tampoco depende de la época de la reconquista, pues abarca a zonas re-
pobladas en el XIII o en el XV . . . el seseo/ceceo irradia de un epicentro an-
daluz -seguramente Sevilla- y se extiende sin alcanzar las zonas más
extremas" (Ariza 1997: 62). Interestingly, Kauffeld (2002) has found far more
evidence of incipient confusion in texts from Córdoba (see below), but this
does not change the fact that 16th-century Seville was a more likely place for
such incipient phenomena to become regularized as seseo.
150. Curiously, Frago also cites this same evidence for widespread devoicing, but
ignores its implications for his arguments about seseo. The specific examples
also differ: doz 'dos', miez 'mies', trez 'tres'. Frago indicates that the first two
are corrected to -s. (Frago Gracia 1993: 260, from Díaz Montesinos 1987:
287-321).
300 Notes

151. In fact, one can see in these northern examples a case of limited neutralization
or near merger, perhaps prevented from developing by the greater social sta-
bility of the north.
152. There are several ways of interpreting forms showing confusion of -s and -z.
Ariza himself (1997: 62) points out that Lapesa's diesmo may be interpreted
as a sign of either seseo or confusion resulting from deaffrication. Given the
sporadic nature of early examples, Ariza prefers the second possibility, but
there are others: seseo or, more likely, implosive neutralization may have de-
veloped first in restricted areas or social groups (as perhaps is indicated in
Kauffeld 2002; see below).
153. Frago himself points this out as a doubtful case, since he assumes this place-
name refers to Almansa in Albacete. However, the same toponym appears as
Almanza in Leon.
154. Ariza (1996: 65) also points out that Menéndez Pidal (1954) found examples
similar to these in the Poema de Mio Cid: çeruiçio (verse 1535) and San
Çaluador (verse 2924).
155. Kauffeld (2002) recently completed a careful analysis of seseo in a collection
of paleographically-transcribed Andalusian legal documents dating from
1324-1500 (Kauffeld 1999). She includes a lucid discussion of the difficulties
involved in interpreting sigmas, and her analysis therefore excludes all cases
of sibilants represented with sigmas. In 31 of 89 texts, she finds a total of 67
clear examples of graphic confusion. The first of these are 3 tokens of rrason
(for razon) in a 1398 Seville document. Most of the examples are best ex-
plained as the general results of the phonetic proximity of the main allophones
of these phonemes (with growing neutralization in syllable-final position)
combined with the particular factors outlined in the preceding discussion. Ex-
amples include: exsepcion (1398, 1465, 1482)), dies (1409, 1454), fasemos,
jfaser, Fis, fase, fise, jfis (8 tokens from 1428 to 1500) - all of which are para-
digmatically linked to the frequent apocopated relic forms faz and fiz - quinse
(1437), jues (5 examples from 1454 to 1487), etc. Of particular interest is
Kauffeld's finding that most of the examples of confusion were found in
documents from Córdoba, rather than from Seville (although it must be recog-
nized that more of the documents were from Córdoba).
156. One of the most important of these is sibilant devoicing, which refers to the
merger of the medieval voiced and voiceless dental sibilant phonemes, as well
as the merger of the voiced and voiceless apicoalveolar sibilant phonemes. It
has been attributed to Basque influence by numerous scholars (e.g., Lantolf
1974; Lloyd 1987). This is not an unreasonable claim, since Basque has never
had voiced sibilants in its phoneme inventory. Still, merger of these phonemes
in Basque interlanguages would have represented a minority feature in the
prekoine linguistic pool (and minority mergers are not automatically favored
in koineization). However, it does not seem likely that devoicing spread dur-
 Notes 301

ing the Burgos phase (it certainly was not adopted as a norm in Toledo). One
possibility, of course, is that sibilant devoicing is not at all related to Basque
influence (Penny [1993] presents an possible internal factor), and that it arose
in other areas simultaneously (cf. Cano Aguilar 1998: 138). Another is that
Basques influenced sibilant devoicing not during koineization in Castile, but
rather during subsequent language shift to Castilian in their home communi-
ties. The question requires further investigation.
157. For Leonese and Aragonese, one could argue for independent development
based on learners' similar response to similar dialect mixes (what Trudgill,
Lewis, and Maclagan [2000] refer to as drift). This is particularly true for the
loss of preposition + article contractions and the phonologization of the diph-
thongs. On the other hand, the selection of the marked [we] as the preferred
realization of the diphthong would seem to favor direct contact with Castilian.
158. This is not to say that only koineization can favor such outcomes. Silva-
Corvalán (1994), for example, claims that contact between English and Span-
ish in Los Angeles has led to an exaggeration of "internal tendencies" of Span-
ish, usually towards greater analyticity - but she does not discuss the sources
of those tendencies.
159. This global perspective of the multiple (re)koineizations of Castilian (and
other varieties) of the medieval period is also relevant to observations made by
Cano Aguilar: " . . . desde muy pronto, el castellano fue más, bastante más, que
la inposición de un dialecto, el castellano, a costa de los demás. Naturalmen-
te, a lo largo del proceso hubo casos en que ciertas peculiaridades castellanas
relegaron a las vecinas . . . pero del mismo modo algunos procesos acabaron
triunfando en castellano porque también eran las formas propias de los otros
dialectos." (Cano Aguilar 1998: 139-140). These comments conclude a cri-
tique of the Pidalian concept of the cuña or wedge, in which Castilian norms
were seen as simply imposed on other dialects. As Cano notes, many Castilian
features (often simplifications, but also consistent selection of /ue/ and other
more marked features) did spread, but Castilian itself ended up accepting
many features of other dialects (e.g., rejection of extreme apocope and, in
southerly varieties, the etymological pronoun system). The concept of the
cuña depends (in its strict sense) on an essentialist understanding of the Castil-
ian dialect (reflected too in Lapesa's view of extreme apocope as an alien
norm). As we have seen, some features do appear to have stabilized and
spread from very early on, but it is also clear that the Castilian dialect(s) was
changing and variable from its beginnings, particularly as affected by the dif-
ferent stages of koineization.
160. Mufwene is right to reject the simple dichotomy "adults create pidgins, chil-
dren create creóles", but I think he overreacts when he denies children any
significant role in creole formation (or, for that matter, in the creation of any
non-pidgin language variety).
References

Abad, Francisco
1985 Leísmo y laísmo: explicaciones y datos históricos. In Philologica
Hispaniensia in honorem Manuel Alvar, Vol. 2, 11-20. Madrid:
Gredos.
Atareos Llorach, Emilio
1958 Quelques précisions sur la diphtongaison espagnole. In Omagiu
lui Iorgu Iordan, A. Rosetti (ed.), 1-4. Bucharest: Academiei
Republicii Populare Romîne.
1982 El español, lengua milenaria (y otros escritos castellanos). Va-
lladolid: Ambito.
1996 Reflexiones sobre el origen del sistema vocálico español. In Actas
del III Congreso Internacional de Historia de la Lengua Españo-
la, A. Alonso González, L. Castro Ramos, B. Gutiérrez Rodilla,
and J. A. Pascual Rodríguez (eds.), 15-20. Madrid: Asociación
de Historia de la Lengua Española, Arco/Libros, Fundación Du-
ques de Soria.
Allen, J. D. H.
1976 Apocope in Old Spanish. In Estudios ofrecidos a Emilio Alarcos
Llorach, Vol. 1, 15-30. Oviedo: Universidad de Oviedo.
Alonso, Amado
1947 Trueques de sibilantes en antiguo español. Nueva revista de filo-
logía hispánica 1: 1-12.
1953 La base lingüística del español americano. In Temas lingüísticos.
Temas hispanoamericanos, 7-72. Madrid: Gredos.
1969 De la pronunciación medieval a la moderna en español. Vol. 2.
Madrid: Gredos.
Alvar, Manuel
1953 El dialecto aragonés. Madrid: Gredos.
1964 Estructura del léxico andaluz. Boletín de filología de la Universi-
dad de Chile 16: 5-12.
1968 Historia y lingüística: 'colonización' franca en Aragón. In Fest-
schrift Walther von Wartburg, Kurt Baldinger (ed.), 129-150.
Tübingen: Niemeyer.
1976 El dialecto riojano. Madrid: Gredos.
1979 Hablas meridionales: el andaluz. In Gran enciclopedia de Anda-
lucía, José Maria Javierre (ed.), 1868-1880. Seville: Promocio-
nes Culturales Andaluzas.
 References 303

Alvar, Manuel, and Bernard Pottier

1983 Morfología histórica del español. Madrid: Gredos.
Álvarez Maurín, María del Pilar
1994 Diplomática asturleonesa: terminología toponímica. León: Uni-
versidad de León.
Álvarez Rodríguez, Adelino
1996 ¿Irregularidades en la apócope 'normal' de la /e/? Intento de
explicación. In Actas del III Congreso Internacional de Historia
de la Lengua Española, A. Alonso González, L. Castro Ramos,
B. Gutiérrez Rodilla, and J. A. Pascual Rodríguez (eds.), 33-41.
Madrid: Asociación de Historia de la Lengua Española, Ar-
co/Libros, Fundación Duques de Soria.
Amery, Rob
1993 An Australian koine: Dhuwaya, a variety of Yolgu Matha spoken
at Yirrkala in North East Arnhemland. International Journal of
the Sociology of Language 99: 45-64.
Andersen, Henning
1973 Abductive and deductive change. Language 49: 765-793.
1988 Center and periphery: Adoption, diffusion, and spread. In His-
torical Dialectology: Regional and Social, Jacek Fisiak (ed.), 39-
83. Berlin/New York: Mouton de Gruyter.
Anderson, James M.
1992 Doublets, cultismos, and their relation in Castilian Spanish. Orbis
35:166-170
Ariza, Manuel
1992 Lingüística e historia de Andalucía. In Actas del II Congreso
Internacional de Historia de la Lengua Española, M. Ariza, J.
M. Mendoza, R. Cano, and A. Narbona (eds.), 15-33. Madrid:
Pabellón de España.
1994 Sobre fonética histórica del español. Madrid: Arco/Libros.
1996 Reflexiones sobre la evolución del sistema consonàntico. In
Actas del III Congreso Internacional de Historia de la Lengua
Española, A. Alonso González, L. Castro Ramos, B. Gutiérrez
Rodilla, and J. A. Pascual Rodríguez (eds.), 43-79. Madrid: Aso-
ciación de Historia de la Lengua Española, Arco/Libros, Funda-
ción Duques de Soria.
1997 Historia lingüística del andaluz. Demófilo. Revista de cultura
tradicional de Andalucía 22: 59-68.
Azevedo Maia, Clarinda de
1986 Historia do galego-português. Coimbra: Instituto Nacional de
Investigaçâo Científica.
304 References

Badia Margarit, Antoni

1962 Nuevas precisiones sobre la diptongación española. Revue de
linguistique romane 26: 1-12.
1972 Por una revisión del concepto de 'cultismo' en fonética histórica.
In Studia hispánico in honorem R. Lapesa, Vol. 1, 137-152. Ma-
drid: Gredos, Cátedra, Seminario Menéndez Pidal.
Bailey, Charles-James N.
1980 Old and new views on language history and language relation-
ships. In Kommunikationstheoretische Grundlagen des
Sprachwandels, Helmut Lüdtke (ed.), 139-181. Berlin/New
York: Walter de Gruyter.
Bamgbose, Ayo
1966 A Grammar of Yoruba. Cambridge: Cambridge University Press.
Barrios, Ángel
1985 Repoblación de la zona meridional del Duero. Fases de ocupa-
ción, procedencias y distribución espacial de los grupos repobla-
dores. Studia Histórica 3 (2): 33-82.
Bartsch, Renate
1987 Norms of Language: Theoretical and Practical Aspects. London:
Longman.
Barz, Richard K., and Jeff Siegel (eds.)
1988 Language Transplanted: The Development of Overseas Hindi.
Wiesbaden: Otto Harrassowitz.
Beebe, Leslie M., and Howard Giles
1984 Speech-accommodation theories: A discussion in terms of sec-
ond-language acquisition. International Journal of the Sociology
of Language 46: 5-32.
Bell, Allen
1984 Language style as audience design. Language in Society 13: 145-
204.
Berruto, Gaetano
1989 Sociolinguistica dell'italiano contemporaneo. Rome: La Nuova
Italia Scientifica.
Bezler, Francis
1991 De la date des gloses de Silos. Revista de filología española 71:
347-354.
Bidwell, Charles E.
1964 The Serbo-Croation verb. Language 40: 532-556.
Bishko, Charles Julian
1975 The Spanish and Portuguese Reconquest, 1095-1492. In A His-
tory of the Crusades, Kenneth M. Seiten (ed.), Vol. 3, 396-456.
Madison: University of Wisconsin Press.
 References 305

Blanc, Haim
1968 The Israeli Koine as an emergent national standard. In Language
Problems in Developing Nations, Joshua Fishman, Charles Fer-
guson, and Jyotirindra das Gupta (eds.), 237-251. New York:
John Wiley and Sons.
Blanchet, Philippe
1992 Le mythe de la 'koinè' en ancien provençal. Lou Prouvençau a
¡'escolo 129 (1): 6-7.
Bortoni, Stella Maris
1991 Dialect contact in Brasilia. International Journal of the Sociology
of Language 89: 47-59.
Britain, David
1997a Dialect contact and phonological reallocation: 'Canadian Raising'
in the English Fens. Language in Society 26: 15-46.
1997b Dialect contact, focusing and phonological rule complexity: The
koineisation of Fenland English. University of Pennsylvania
Working Papers in Linguistics 4 (1): 141-169.
1999 Locating the baseline of linguistic innovations: Dialect contact,
the founder principle and the so-called (-own) split in New Zea-
land English. Cuadernos de filología inglesa 8: 177-192.
2002 Space and spatial diffusion. In The Handbook of Language
Variation and Change, J. K. Chambers, Peter Trudgill, and Nata-
lie Schilling-Estes (eds.), 603-637. Oxford: Blackwell.
Britain, David, and Peter Trudgill
1999 Migration, new-dialect formation and sociolinguistic reiunction-
alisation: Reallocation as an outcome of dialect contact. Transac-
tions of the Philological Society 97 (2): 245-256.
Bubenik, Vit
1989 Hellenistic and Roman Greece as a Sociolinguistic Area. Am-
sterdam/Philadelphia: John Benjamins.
1993 Dialect contact and koineization: The case of Hellenistic Greek.
International Journal of the Sociology of Language 99: 9-24.
Buck, Carl Darling
1933 Comparative Grammar of Greek and Latin. Chicago: University
of Chicago Press.
Bybee, Joan, and Paul Hopper (eds.)
2001 Frequency and the Emergence of Linguistic Structure. Amster-
dam/Philadelphia: John Benjamins.
Bybee, Joan, and Dan Slobin
1982 Rules and schémas in the development and use of the English past
tense. Language 58: 265-289.
306 References

Callary, Robert E.
1975 Phonological change and the development of an urban dialect in
Illinois. Language in Society 4: 155-170.
Cano Aguilar, Rafael
1998 Los orígenes del español: nuevos planteamientos. In Estudios de
lingüística y filología españolas. Homenaje a Germán Colón,
Irene Andrés-Suárez and Luis López Molina (eds.), 127-140.
Madrid: Gredos.
Cardona, Giorgio Raimondo
1990 II concetto de koinè in linguistica. In Koinè in Italia dalle origini
al cinquecento, Glauco Sanga (ed.), 24-34. Bergamo: Lubrina.
Catalan, Diego
1956 El çeçeo-zezeo al comenzar la expansión atlántica de Castilla.
Boletim de filologia 16: 306-334.
1956-57 El astuiano occidental: examen sincrónico y explicación diacròni-
ca de sus fronteras fonológicas. Romance Philology 10: 71-92.
1971 En torno a la estructura silábica del español de ayer y del español
de mañana. In Sprache und Geschichte. Festschrift fur Harri
Meier, Eugenio Coseriu and Wolf-Dieter Stempel (eds.), 77-110.
Munich: Wilhelm Fink.
Chambers, J. K.
1992 Dialect Acquisition. Language 68: 673-705.
Chambers, J. Κ. and Margaret Hardwick
1985 Dialect homogeneity and incipient variation. Sheffield Working
Papers in Language and Linguistics 2.
Chambers, J. K., and Peter Trudgill
1980 Dialectology. Cambridge: Cambridge University Press.
Cheshire, Jenny
1982 Variation in an English Dialect. Cambridge: Cambridge Univer-
sity Press.
1996 Syntactic variation and the concept of prominence. In Speech
Past and Present: Studies in English Dialectology in Memory of
Ossi Ihalainen, Juhani Klemola, Meija Kytö, Matti Rissanen
(eds.), 1-17. Frankfurt/New York: Peter Lang.
Clark, Eve
1985 The acquisition of Romance, with special reference to French. In
The Crosslinguistic Study of Language Acquisition, Volume 1:
The Data, Dan Slobin (ed.), 687-782. Hillsdale, NJ: Lawrence
Erlbaum.
Clavería Nadal, Gloria
1991 El latinismo en español. Barcelona: Universität Autònoma de
Barcelona.
 References 307

Cooper, Louis
1960 El Liber Regum: estudio lingüístico. Zaragoza: Instituto Fernando
el Católico.
Cooper, Robert L.
1982 A framework for the study of language spread. In Language
Spread: Studies in Diffusion and Social Change, Robert L. Coo-
per (ed.), 5-36. Bloomington: Indiana University Press.
Corbett, Noel
1970-71 Reconstructing the diachronic phonology of Romance. Romance
Philology 24: 273-290.
Coraminas, Joan, and José Antonio Pascual
1981 Diccionario crítico-etimológico castellano e hispánico. Madrid:
Gredos.
Croît, William
2000 Explaining Language Change: An Evolutionary Approach. Har-
low, UK/New York: Longman.
Crystal, David
1991 A Dictionary of Linguistics and Phonetics. 3rd ed. Oxford:
Blackwell.
1992 An Encyclopedic Dictionary of Language and Languages. Ox-
ford: Blackwell.
Cuervo, Rufino José
1895 Los casos enclíticos y proclíticos del pronombre de tercera per-
sona en castellano. Romania 24: 95-113, 219-263.
DePaulo, Bella M.
1992 Nonverbal behavior and self-presentation. Psychological Bulletin
111: 203-223.
Deutscher, Guy
2002 On the misuse of the notion of 'abduction' in linguistics. Journal
of Linguistics 38: 469-485.
Díaz Montesinos, Francisco
1987 Sobre las sibilantes en el Fuero de Huete. Revista de filología
española 67: 287-321.
Díaz Roig, Mercedes
1987 El romancero viejo. Mexico: REI.
Díaz y Díaz, Manuel C.
1978 Las primeras glosas hispánicas. Barcelona: Universität Autòno-
ma de Barcelona.
Diez, Miguel, Francisco Morales, and Ángel Sabin
1980 Las lenguas de España. 2nd ed. Madrid: Ministerio de Educa-
ción.
308 References

Dillard, J. L.
1964 The writings of Herskovits and the study of the language of the
Negro in the New World. Caribbean Studies 4: 35-41.
1972 Black English: Its History and Usage in the United States. New
York: Random House.
1985 Toward a Social History of American English. Berlin/New York:
Mouton de Gruyter.
Dixon, R. M. W.
1997 The Rise and Fall of Languages. Cambridge: Cambridge Univer-
sity Press.
Domingue, Nicole
1981 Internal change in a transplanted language. Studies in the Linguis-
tic Sciences 4: 151-159.
Dorian, Nancy
1981 Language Death: The Life Cycle of a Scottish Gaelic Dialect.
Philadelphia: University of Pennsylvania Press.
Eberenz, Rolf
1991 Castellano antiguo y español moderno: reflexiones sobre la pe-
riodización en la historia de la lengua. Revista de filología espa-
ñolan·. 79-107.
2000 El español en el otoño de la Edad Media. Madrid: Gredos.
Echenique Elizondo, María Teresa
1979 Apócope y leísmo en la Primera Crónica General: notas para una
cronología. Studi Ispanici 1-16.
1980 Sobre pronombres afijos en español antiguo. Neophilologische
Mitteilungen 81: 201-207.
1981 El sistema referencial en español antiguo. Revista de filología
española 61: 113-157.
Elcock, W. D.
1960 The Romance Languages. London: Faber and Faber.
Ellis, Rod
1985 Understanding Second Language Acquisition. Oxford: Oxford
University Press.
Ervin-Tripp, Susan M.
1968 An analysis of the interaction of language, topic, and listener. In
Readings in the Sociology of Language, Joshua A. Fishman (ed.),
192-211. The Hague: Mouton.
Ferguson, Charles A.
1959a The Arabic Koine. Language 35: 616-630.
1959b Diglossia. Word 15: 325-340.
 References 309

Fernández-Ordóñez, Inés
1993 Leísmo, laísmo y loísmo: estado de la cuestión. In Los pronom-
bres átonos, Olga Fernández Soriano (ed.), 63-96. Madrid: Tau-
rus.
1994 Isoglosas internas del castellano: el sistema referencial del pro-
nombre átono de tercera persona. Revista de filología española
74: 71-125.
1999 Leísmo, laísmo y loísmo. In Gramática descriptiva de la lengua
española, Ignacio Bosque and Violeta Demonte (eds.), 13 Π -
Ι 397. Madrid: Espasa-Calpe.
2001 Hacia una dialectología histórica: reflexiones sobre la historia del
leísmo, el laísmo y el loísmo. Boletín de la Real Academia Espa-
ñola 81: 389-464.
Fernández Ramírez, Salvador
1951 Gramática española. Madrid: Manuales de la Revista de Occi-
dente.
Ferreiro, Manuel
1996 Gramática histórica galega. Santiago de Compostela: Laiovento.
Flores Cervantes, Marcela
1997 Individuación de la entidad en los orígenes de leísmo, laísmo y
loísmo. In Cambios diacrónicos en el español, Concepción Com-
pany (ed.), 33-63. Mexico: Universidad Autónoma Nacional de
Mexico.
Fontanella de Weinberg, Maria Beatriz
1992 El español de América. Madrid: Mapire.
Frago Gracia, Juan Antonio
1983 Materiales para la aspiración de la /-s/ implosiva en las hablas
andaluzas. Lingüística española actual 5: 153-171.
1984 La fonética del español meridional y sus fuentes históricas. In
Miscel.lània Sanchis Guarner, Volume 2, 131-137. Valencia:
Universität de Valéncia.
1989 El seseo entre Andalucía y América. Revista de filología españo-
la 69: 277-310.
1991a Conflicto de normas lingüísticas en el proceso castellanizador de
Aragón. In / Curso de Geografia Lingüística de Aragón, José
María Enguita Utrilla (ed.), 105-126. Zaragoza: Institución Fer-
nando el Católico.
1991b Viejos y nuevos dialectos en la evolución lingüística: el caso del
andaluz. In Actes du XVIIIe Congrès International de Linguisti-
que et de Philologie Romanes, Dieter Kremer (ed.), 22-32. Tü-
bingen: Niemeyer.
1993 Historia de las hablas andaluzas. Madrid: Arco/Libros.
310 References

1999 Historia del español de América. Madrid: Gredos.

Gallois, Cynthia and Victor J. Callan
1988 Communication accommodation and the prototypical speaker:
Predicting evaluations of status and solidarity. Language and
Communication 8: 271-283.
Gallois, Cynthia, Victor J. Callan, and J. M. Palmer
1992 The influence of applicant communication style and interviewer
characteristics on hiring decisions. Journal of Applied Social
Psychology 22: 1041-1060.
Galmés de Fuentes, Alvaro
1962 Las sibilantes en la Romania. Madrid: Gredos.
1983 Dialectología mozárabe. Madrid: Gredos.
1996 Mozárabe. In Manual de dialectología hispánica: el español de
España, Manuel Alvar (ed.), 97-110. Barcelona: Ariel.
Gambhir, Surendra K.
1981 The East Indian speech community in Guyana: A sociolinguistic
study with special reference to koine formation. Ph. D. diss., Uni-
versity of Pennsylvania.
1983 Two koines compared: Guyanese Bhojpuri and Calcutta Bazaar
Hindustani. International Journal of Dravidian Linguistics 12:
471-480.
García, Erica
1975 The Role of Theory in Linguistic Analysis: The Spanish Pronoun
System. Amsterdam: North-Holland Linguistic Series.
1994 Extralinguistic conditioning of grammatical change. Linguistische
Berichte 153: 341-371.
García González, Francisco
1978 El leísmo en Santander. In Estudios ofrecidos a Emilio Alarcos
Llorach, Vol. 3, 87-101. Oviedo: Servicio de Publicaciones,
Universidad de Oviedo.
García de Cortázar, José Ángel
1991 La repoblación del Valle del Duero en el siglo IX: del yermo
estratégico a la organización social del espacio. In Actas del Co-
loquio de la V Asamblea General de la Sociedad Española de
Estudios Medievales, 15-39. Zaragoza: Departamento de Cultura
y Educación.
García de Diego, Vicente
1950 El castellano como complejo dialectal y sus dialectos internos.
Revista de filología española 34: 107-124.
Giddens, Anthony
1989 Sociology. Cambridge: Polity Press.
 References 311

Giles, Howard
1973 Accent mobility: A model and some data. Anthropological Lin-
guistics 15: 87-105.
Gimeno Menéndez, Francisco
1995 Sociolingiiistica histórica: Siglos X-XII. Madrid: Visor.
González, Julio
1951 El Repartimiento de Sevilla. Madrid: Consejo Superior de Inves-
tigaciones Científicas.
1975-1976 Repoblación de Castilla la Nueva. Madrid: Universidad Com-
plutense.
González Jiménez, Manuel
1975 La repoblación de la zona de Sevilla durante el siglo XIV. Sevi-
lle: Secretariado de Publicaciones, Universidad de Sevilla.
1982 Orígenes de la Andalucía cristiana. In Historia de Andalucía,
Vol. 2, 97-303. Madrid/Barcelona: CUPSA and Planeta.
1988 En torno a los orígenes de Andalucía. La repoblación del siglo
XIII. Seville: Servicio de Publicaciones, Universidad de Sevilla.
1993 Alfonso X: 1252-1284. Palencia: Diputación Provincial de Palen-
cia.
(ed.) 1991 Diplomatario andaluz de Alfonso X. Seville: El Monte.
González Jiménez, Manuel, and Antonio González Gómez
1980 El Libro de Repartimiento de Jerez de la Frontera: estudio y
edición. Cádiz: Instituto de Estudios Gaditanos.
Graff, Willem Laurens
1932 Language and Linguistics: An Introduction to Linguistics. New
York: Appleton.
Granda, Germán de
1994 El proceso de koineización en el período inicial de desarrollo del
español de América. In El español de América en el siglo XVI.
Actas del Simposio del Instituto Ibero-Americano de Berlín, 23 y
24 de abril de 1992, Jens Lüdtke (ed.), 87-108. Berlin: Vervuert
and Iberoamericana.
Granovetter, Mark
1973 The strength of weak ties. American Journal of Sociology 78:
1360-1380.
1982 The strength of weak ties: A network theory revisited. In Social
Structure and Network Analysis, Peter Marsden and Nan Lin
(eds.), 105-130. London: Sage.
Gumperz, John J., and Robert Wilson
1971 Convergence and creolization: A case from the Indo-
Aryan/Dravidian border in India. In Pidginization and Creoliza-
312 References

tion of Languages, Dell Hymes (ed.), 151-167. Cambridge:

Cambridge University Press.
Hall, Robert A.
1974 External History of the Romance Languages. New York: Ameri-
can Elsevier.
Haller, Hermann W.
1981 Between Standard Italian and creole: An interim report. Word 32:
181-192.
Hancock, Ian F.
1971 A provisional comparison of the English-derived Atlantic creóles.
In Pidginization and Creolization of Languages, Dell Hymes
(ed.), 287-291. Cambridge: Cambridge University Press.
Harris, John
1990 Segmental complexity and phonological government. Phonology
7: 255-300.
Harris-Northall, Ray
1990a The spread of sound change: Another look at syncope in Spanish.
Romance Philology 44: 137-162.
1990b Weakening Processes in the History of Spanish Consonants.
London: Routledge.
1991 Apocope in Alfonsine texts: A case study. In Linguistic Studies in
Medieval Spanish, Ray Harris-Northall and Thomas D. Cravens
(eds.), 29-38. Madison: Hispanic Seminary of Medieval Studies.
1992 Devoicing, deaffrication, and word-final -z in medieval Spanish.
Hispanic Linguistics 4 (2): 245-274.
1993 Algunos aspectos de la variación ortográfica en los textos alfon-
síes. In Actas del Primer Congreso Anglo-Hispano. Tomo I: Lin-
güística, Ralph Penny (ed.), 181-193. Madrid: Castalia.
1996a Linguistic aspects of the Cancionero de Baena. In Nunca fue pena
mayor: estudios de literatura española en homenaje a Brian Dut-
ton, Ana Menéndez Collera and Victoriano Roncero López
(eds.), 321-331. Cuenca: Universidad de Castilla-La Mancha.
1996b Printed books and linguistic standardization in Spain: The 1503
'Gran Conquista de Ultramar'. Romance Philology 50: 123-146.
Hartmann, R. R. K., and F. C. Stork
1973 Dictionary of Language and Linguistics. London: Applied Sci-
ence Publishers.
Haugen, Einar
1966 Dialect, language, nation. American Anthropologist 68: 922-935.
Henriquez Ureña, Pedro
1932 Sobre el problema del andalucismo dialectal de América. Buenos
Aires: Instituto de Filología de la Universidad de Buenos Aires.
 References 313

Henry, Alison
1995 Belfast English and Standard English. Oxford: Oxford University
Press.
Hernández, César
1991 Acercamiento al castellano del siglo XIII. In Scripta Philologica
in honorem Juan M. Lope Blanch a los 40 años de docencia en
la UNAMy a los 65 años de vida. II: Lingüística española e ibe-
roamericana, Elizabeth Luna Traill (ed.), 329-344. Mexico:
Universidad Nacional Autónoma de México.
Hidalgo, Margarita (ed.)
2001 Between Koineization and Standardization: New World Spanish
Revisited. International Journal of the Sociology of Language
149.
Hill, Trevor
1958 Institutional linguistics. Orbis 7 (2): 441-455.
Hock, Hans Heinrich
1986 Principles of Historical Linguistics. Berlin/New York: Mouton
de Gruyter.
Hofstadter, Richard, and Seymour Lipset (eds.)
1968 Turner and the Sociology of the Frontier. New York: Basic
Books.
Horrocks, Geoffrey
1997 Greek: A History of the Language and its Speakers. London and
New York: Longman.
Huber, Joseph
1986 Gramática do portugués antigo. Lisbon: Fundaçlo Calouste
Gulbenkian.
Imhoff, Brian
2000a Socio-historic network ties and medieval Navarro-Aragonese.
Neuphilologische Mitteilungen 101: 443-450.
2000b Dialect contact and historical sociolinguistic variation: The Old
Spanish "-ie imperfect". Hispanic Review 68: 381-396.
Iradiel, Paulino, Salustiano Moreta, and Esteban Sarasa
1995 Historia medieval de la España cristiana. Madrid: Cátedra.
Izzo, Herbert
1977 Pre-Latin languages and sound change in Romance: The case of
Old Spanish /h-/. In Studies in Romance Linguistics, M. P. Hagi-
wara (ed.), 227-253. Rowley, Massachusetts: Newbury House.
Jakobson, Roman
1962 Remarques sur l'évolution phonologique du russe comparée a
celle des autres langues slaves. In Selected Writings, I. Phono-
314 References

logical studies, 7-116. The Hague: Mouton. Original edition,

1929.
Janda, Richard D.
2003 "Phonologization" as the start of dephoneticization - or, on sound
change and its aftermath: of extension, generalization, lexicaliza-
tion, and morphologization. In The Handbook of Historical Lin-
guistics, Brian D. Joseph and Richard D. Janda (eds.), 401-422.
Oxford: Blackwell.
Karlgren, Bernhar
1949 The Chinese Language. New York: Ronald.
Kasten, Lloyd, John Nitti, and Wilhelmina Jonxis-Henkemans
1997 The Electronic Texts and Concordances of the Prose Works of
Alfonso X, El Sabio. Madison: Hispanic Seminary of Medieval
Studies.
Kauffeld, Cynthia J.
1999 Electronic Texts and Concordances of Andalusian Documents
(1324-1500). CD-ROM. Madison: Hispanic Seminary of Medie-
val Studies.
2002 Textual evidence of seseo in Andalusian texts. In Two Genera-
tions: A Tribute to Lloyd A. Kasten (1905-1999), Francisco
Gago-Jover (ed.), 157-168. New York: Hispanic Seminary of
Medieval Studies.
Keller, Rudi
1994 Language Change: The Invisible Hand in Language. London:
Routledge.
Kerswill, Paul
1994a Babel in Buckinghamshire? Pre-school children acquiring accent
features in the New Town of Milton Keynes. In Nonstandard Va-
rieties of Language: Papers from the Stockholm Symposium,
Gunnel Melchers and Nils-Lennart Johannessen (eds.), 64-84.
Stockholm: Almqvist and Wiksell.
1994b Dialects Converging: Rural Speech in Urban Norway. Oxford:
Oxford University Press.
1996 Children, adolescents, and language change. Language Variation
and Change 8: 177-202.
2002 Koineization and accommodation. In The Handbook of Language
Variation and Change, J. K. Chambers, Peter Trudgill, and Nata-
lie Schilling-Estes (eds.), 669-702. Oxford: Blackwell.
Kerswill, Paul and Ann Williams
1992 Some principles of dialect contact: Evidence from the New Town
of Milton Keynes. In Working Papers 1992, Irene Philippaki-
 References 315

Warburton and Richard Ingram (eds.), 68-90. Reading, UK: De-

partment of Linguistic Science, University of Reading.
1999 Mobility vs. social class in dialect leveling: Evidence from new
and old towns in England. Cuadernos de filología inglesa 8: 47-
57.
2000 Creating a New Town koine: Children and language change in
Milton Keynes. Language in Society 29: 65-115.
2002 'Salience' as a factor in language change: Evidence from dialect
leveling in urban England. In Language Change: The Interplay of
Internal, External and Extra-Linguistic Factors, M. C. Jones and
E. Esch (eds.), 81-110. Berlin/New York: Mouton de Gruyter.
Kiparsky, Paul
1982 Explanation in Phonology. Dordrecht: Foris.
Klein-Andreu, Flora
1981 Distintos sistemas de empleo 'le', 'la', Ίο'. Perspectiva sincróni-
ca, diacrònica y sociolingüística. Thesaurus 36: 284—304.
1992 Understanding standards. In Explanation in Historical Linguis-
tics, Garry Davis and Gregory K. Iverson (eds.), 169-178. Am-
sterdam/Philadelphia: John Benjamins.
1996 Anaphora, deixis, and evolution of Latin Ille. In Studies in
Anaphora, Barbara Fox (ed.), 305-331. Amsterdam/Philadelphia:
John Benjamins.
2000 Grammatical and lexical behavior in the development of the
Spanish third-person clitics. In Between Grammar and Lexicon,
Ellen Contini-Morava and Yishai Tobin (eds.), 159-183. Am-
sterdam/Philadelphia: John Benjamins.
Knecht, Pierre, and Zygmunt Marzys (eds.)
1993 Écriture, langues communes et normes: Formation spontanée de
koinès et standardisation dans la Galloromania et son voisinage.
Geneva: Droz.
Krüger, Fritz
1923 El dialecto de San Cipriân de Sanabria. Madrid: Sucesores de
Hernando.
Labov, William
1963 The social motivation of a sound change. Word 19: 273-409.
1972 Some principles of linguistic methodology. Language in Society
1:97-110.
1975 On the use of the present to explain the past. In Proceedings of
the Eleventh International Congress of Linguists, Aug. 28 — Sept.
2, 1972, Vol. 2, Luigi Heilmann (ed.), 825-851. Bologna: Il Mu-
lino.
316 References

1994 Principles of Linguistic Change: Internal Factors. Oxford:

Blackwell.
2001 Principles of Linguistic Change: Social Factors. Oxford: Black-
well.
Lacarra, José Maria
1972 Aragon en el pasado. 2nd ed. Madrid: Espasa-Calpe.
Lantolf, James
1974 Linguistic change as a socio-cultural phenomenon: A study of the
Old Spanish sibilant devoicing. Ph. D. diss., Pennsylvania State
University.
Lapesa, Rafael
1951 La apòcope de la vocal en castellano antiguo: intento de explica-
ción histórica. In Estudios dedicados a Menéndez Pidal, Vol. 2,
185-226. Madrid: Consejo Superior de Investigaciones Científi-
cas.
1968 Sobre los orígenes y evolución del leísmo, laísmo y loísmo. In
Festschrift Walter von Wartburg, Kurt Baldinger (ed.), 523-551.
Tübingen: Niemeyer.
1975 De nuevo sobre la apòcope vocálica en castellano medieval. Nue-
va revista de filología hispánica 24: 13-23.
1981 Historia de la lengua española. 9th ed. Madrid: Gredos.
1982 Contienda de normas lingüísticas en el castellano alfonsi. In Ac-
tas del coloquio hispano-alemán Ramón Menéndez Pidal, Wido
Hempel and Dietrich Briesemeister (eds.), 172-190. Tübingen:
Niemeyer.
1985 Sobre el ceceo y el seseo andaluces. In Estudios de historia lin-
güistica española, 249-266. Madrid: Paraninfo. Original edition,
1957.
Lass, Roger
1990 Where do extraterritorial Englishes come from? Dialect input and
recodification in transported Englishes. In Papers from the 5th
International Conference on English Historical Linguistics.
Cambridge, 6-9 April 1987, Sylvia Adamson, Viven Law, Nigel
Vincent, and Susan Wright (eds.), 245-280. Amster-
dam/Philadelphia: John Benjamins.
1997 Historical Linguistics and Language Change. Cambridge: Cam-
bridge University Press.
Lee, Everett S.
1968 The Turner thesis re-examined. In Turner and the Sociology of
the Frontier, Richard Hofstadter and Seymour Lipset (eds.), 65-
72. New York: Basic Books.
 References 317

LePage, Robert B.
1992 'You can never tell where a word comes from': Language contact
in a diffuse setting. In Language Contact: Theoretical and Em-
pirical Studies, Ernst Hàkon Jahr (ed.), 71-101. Berlin/New
York: Mouton de Gruyter.
LePage, Robert B., and Andrée Tabouret-Keller
1985 Acts of Identity. Cambridge: Cambridge University Press.
Lightfoot, David
1979 Principles of Diachronic Syntax. Cambridge: Cambridge Univer-
sity Press.
Lipset, Seymour
1968 The Turner thesis in comparative perspective: An introduction. In
Turner and the Sociology of the Frontier, Richard Hofstadter and
Seymour Lipset (eds.), 9-14. New York: Basic Books.
Lleal, Coloma
1990 La formación de las lenguas romances peninsulares. Barcelona:
Barcanova.
Llórente Maldonado, Antonio
1980 Consideraciones sobre el español actual. Anuario de letras 18: 5 -
61.
Lloyd, Paul M.
1987 From Latin to Spanish. Philadelphia: American Philosophical
Society.
Lodge, Anthony
1993 French: From Dialect to Standard. London: Routledge.
1999 Convergence and divergence in the development of the Paris
urban vernacular. Sociolinguistica 13: 51-68.
López Bobo, Maria Jesús
1990 Sobre el leísmo en el Libro de Buen Amor. Verba 17: 343-361.
López García, Ángel
1985a El rumor de los desarraigados: conflicto de lenguas en la Penín-
sula Ibérica. Barcelona: Anagrama.
1985b Algunas concordancias gramaticales entre el castellano y euskera.
In Philologica Hispaniensia in Honorem Manuel Alvar, Vol. 2,
391-405. Madrid: Gredos.
López de Aberasturi Arregui, Ignacio
1992 Leonesismos léxicos de carácter migratorio en Andalucía. In
Actas del II Congreso Internacional de Historia de la Lengua
Española, Vol. 2, M. Ariza, J. M. Mendoza, R. Cano, and A.
Narbona (eds.), 179-186. Madrid: Pabellón de España.
318 References

Lüdtke, Helmut
1980 Sprachwandel als universales Phänomen. In Kommunikations-
theoretische Grundlagen des Sprachwandels, Helmut Lüdtke
(ed.), 1-19. Berlin/New York: Walter de Gruyter.
Lust, Horace G.
1959 The creation of Standard Macedonian: Some facts and attitudes.
Anthropological Linguistics 1: 19-26.
Luquet, Gilles
1992 De l'apocope verbale en espagnol ancien (formes subjonctives).
Cahiers de linguistique hispanique médiévale 17: 33—46.
Lyons, Christopher
1993 El desarrollo de las estructuras posesivas en el español temprano.
In Actas del Primer Congreso Anglo-Hispano. Tomo I: Lingüisti-
ca, Ralph Penny (ed.), 215-223. Madrid: Castalia.
Maehlum, Brit
1992 Dialect sociolization in Longyearbyen, Svalbard (Spitsbergen): A
fruitful chaos. In Language Contact: Theoretical and Empirical
Studies, Ernst Hâkon Jahr (ed.), 117-130. Berlin/New York:
Mouton de Gruyter.
Malkiel, Yakov
1963-64 The interlocking of narrow sound change, broad phonological
pattern, level of transmission, areal configuration, sound symbol-
ism: Diachronic studies in Hispano-Latin consonant clusters CL-,
FL-, PL-, Archivum linguisticum 15: 144-173, 16: 1-33.
1976 From falling to rising diphthongs: The case of Old Spanish ió <
*éu (with excursuses on the weak preterite, on the possessives,
and on judío, sandio, and romero). Romance Philology 29: 435-
500.
1983 Alternatives to the classic dichotomy family tree/wave theory?
The Romance evidence. In Language Change, Irmengard Rauch
and Gerald F. Carr (eds.), 192-256. Bloomington: Indiana Uni-
versity Press.
Manessy, Gabriel
1977 Processes of pidginization in African languages. In Pidgin and
Creole Linguistics, Albert Valdman (ed.), 129-154. Blooming-
ton: Indiana University Press.
Martin Martin, José Luis, Luis Miguel Villar García, Florencio Marcos Rodri-
guez, and Marciano Sánchez Rodriguez (eds.)
1977 Documentos de los archivos catedralicio y diocesano de Sala-
manca (Siglos XII-XIII). Salamanca: Universidad de Salamanca.
 References 319

Martínez Alcalde, María José, and Mercedes Quilis Merín

1996 Nuevas observaciones sobre periodización en la historia de la
lengua española. In Actas del III Congreso Internacional de His-
toria de la Lengua Española, A. Alonso González, L. Castro
Ramos, B. Gutiérrez Rodilla, and J. A. Pascual Rodríguez (eds.),
873-886. Madrid: Asociación de Historia de la Lengua Española,
Arco/Libros, Fundación Duques de Soria.
Matute Martínez, Cristina
1999 Estudio comparativo del sistema de pronombres átonos de tercera
persona no reflexivos en tres obras del scriptorium alfonsi: varia-
ción y norma (1254-1280). Memoria de Licenciatura, Universi-
dad Autónoma de Madrid.
McMahon, April M. S.
1994 Understanding language change. Cambridge: Cambridge Uni-
versity Press.
Meillet, Antoine
1975 Aperçu d'une histoire de la langue grecque. 8th ed. Paris:
Klincksieck.
Méndez García de Paredes, Elena
1988 Pronombres posesivos: constitución de sus formas en castellano
medieval. In Actas del I Congreso Internacional de Historia de la
Lengua Española, M. Ariza, A. Salvador, and A. Viudas (eds.),
533-540. Madrid: Arco/Libros.
Menéndez Pidal, Ramón
1954 Cantar de mío Cid. 3rd ed. Madrid: Espasa-Calpe.
1960 Repoblación y tradición en la cuenca del Duero. In Enciclopedia
Lingüísitica Hispánica, XXIX-LVII. Madrid: Consejo Superior
de Investigaciones
1962a El dialecto leonés. Oviedo: Instituto de Estudios Asturianos.
1962b Manual de gramática histórica española. 1 Ith ed. Madrid: Espa-
sa-Calpe.
1962c Sevilla frente a Madrid: algunas precisiones sobre el español de
América. In Estructuralismo e historia: Miscelánea homenaje a
André Martinet, Vol. 3, Diego Catalán (ed.), 99-165. La Laguna:
Universidad de La Laguna.
1964 Orígenes del español: estado lingüístico de la Península Ibérica
hasta el siglo XI. 5th ed. Madrid: Espasa-Calpe.
1966 Documentos lingüísticos de España: Reino de Castilla. Revista
de filología española, Anejo 84. Madrid: Consejo Superior de In-
vestigaciones Científicas.
320 References

Mesthrie, Rajend
1993 Koineization in the Bhojpuri-Hindi diaspora - with special refer-
ence to South Africa. International Journal of the Sociology of
Language 99: 25-44.
1994 Koinés. In The Encyclopedia of Language and Linguistics, R. E.
Asher (ed.), 1864-1867. Oxford: Pergamon Press.
Mestre Campi, Jesús, and Flocel Sabaté
1998 Atlas de la reconquista: la frontera peninsular entre los siglos
VIIIy XV. Barcelona: Península.
Milroy, James
1982 Probing under the tip of the iceberg: Phonological normalisation
and the shape of speech communities. In Sociolinguistic Varia-
tion in Speech Communities, Suzanne Romaine (ed.), 35-47.
London: Arnold.
1992 Language Variation and Change: On the Historical Sociolin-
guistics of English. Oxford: Blackwell.
1993 On the social origins of language change. In Historical Linguis-
tics: Problems and Perspectives, Charles Jones (ed.), 215-236.
London: Longman.
1997 Internal vs external motivations for linguistic change. Multilingua
16(4): 311-323.
Milroy, Lesley
1987 Language and Social Networks. 2nd ed. Oxford: Basil Blackwell.
2002 Introduction: mobility, contact and language change - working
with contemporary speech communities. Journal of Sociolinguis-
tics 6 (1): 3-15.
Milroy, James and Lesley Milroy
1985 Linguistic change, social network, and speaker innovation. Jour-
nal of Sociolinguistics 21 (2): 339-384.
Minguez, José María
1989 La reconquista. Madrid: Historia 16.
Mitchell, J. Clyde
1986 Network procedures. In The Quality of Urban Life, Dieter Frick
(ed.), 73-92. Berlin: Walter de Gruyter.
Mitre Fernández, Emilio
1999 La historiografía sobre la Edad Media. In Historia de la historio-
grafía española, José Andrés-Gallego (ed.), 67-116. Madrid: En-
cuentro.
Moag, Rodney F.
1979 The linguistic adaptations of the Fiji Indians. In Rama's Banish-
ment: A Century Tribute to the Fiji Indians, 1879-1979, V.
Mishra (ed.), 112-138. London: Heinemann.
 References 321

Mohan, Peggy
1978 Trinidad Bhojpuri: A morphological study. Ph. D. diss., Univer-
sity of Michigan.
Mondéjar, José
1991a Dialectología andaluza: estudios. Granada: Don Quijote.
1991b Disquisiciones históricocríticas y metodológicas sobre la interpre-
tación de los datos en el estudio del $e$eo. In Dialectología anda-
luza: estudios, 319-335. Granada: Don Quijote. Original edition,
1985.
1991c Las hablas andaluzas. In Dialectología andaluza: estudios, 213—
234. Granada: Don Quijote. Original edition, 1986.
199Id Edición, léxico y análisis grafemático, fonético y fonológico del
Ordenamiento portuario de Sevilla de 1302. In Dialectología an-
daluza: estudios, 605-629. Granada: Don Quijote. Original edi-
tion, 1989.
Monge, Félix
1983 Notas a una hipótesis sobre el leísmo. In Serta Philologica F.
Lázaro Carreter, 441-453. Madrid: Cátedra.
Montgomery, Michael
1996 Was colonial American English a koiné? In Speech Past and
Present: Studies in English Dialectology in Memory of Ossi Iha-
lainen, Juhani Klemola, Meija Kytö, Matti Rissanen (eds.), 213—
235. Frankfurt/New York: Peter Lang.
Montgomery, Thomas
1975 La apòcope en español antiguo y la î final latina. In Studia his-
pánico in honorem R. Lapesa. Voi. 3, 351-361. Madrid: Gredos.
Moreno Bernal, Jesús
1993 Les conditions de l'apocope dans les anciens textes castillans. In
Le passage à l'écrit des langues romanes, Maria Selig, Barbara
Frank, and Jörg Hartmann (eds.), 193-206. Tübingen: Narr.
Morgan, Michael
1974 A transcription and concordance of the Cancionero de Baena. Ph.
D. diss., University of Wisconsin-Madison.
Μοχό, Salvador de
1979 Repoblación y sociedad en la España cristiana medieval. Ma-
drid: Rialp.
Mufwene, Salikoko
1997 Jargons, pidgins, creóles, and koines: What are they? In The
Structure and Status of Pidgins and Creoles, Arthur K. Spears
and Donald Winford (eds.), 35-70. Amsterdam/Philadelphia:
John Benjamins.
322 References

2001 The Ecology of Language Evolution. Cambridge: Cambridge

University Press.
Mühlhäusler, Peter
1974 Pidginisation and Simplification of Language. Canberra: Pacific
Linguistics.
1980 Structural expansion and the process of creolization. In Theoreti-
cal Orientations in Creole Studies, Albert Valdman and Arnold
Highfield (eds.), 19-55. New York: Academic.
1985 Patterns of contact, mixture, creation and nativization: Their
contribution to a general theory of language. In Developmental
Mechanisms of Language, Charles-James N. Bailey and Roy Har-
ris (eds.), 51-87. Oxford: Pergamon.
Navarro Carrasco, Ana I.
1985 Occidentalismos en andaluz. Español actual 43: 69-88.
Navarro Tomás, Tomás
1957 Manual de pronunciación española. 5th ed. Madrid: Consejo
Superior de Investigaciones Científicas.
Nevalainen, Terttu
2000 Mobility, social networks and language change in Early Modern
England. European Journal of English Studies 4: 253-264.
Nida, Eugene Α., and Harold W. Fehderau
1970 Indigenous pidgins and koines. International Journal of Ameri-
can Linguistics 32: 146-155.
Nordenstam, Kerstin
1979 Svenskan i Norge. Gothenberg: University Press.
Ñuño Alvarez, Maria Pilar
1996 Áreas léxicas del andaluz oriental. La Torre 7-8: 245-271.
Ohala, John J.
1993 Coarticulation and phonology. Language and Speech 36 (2):
155-170.
Omdal, Helge
1977 Heyangermâlet - en ny dialect. Spràklig Sämling 18: 7-9.
Oskenberg, L., L. Coleman, and C. Cannell
1986 Interviewer's voices and refusal rates in telephone surveys. Pu-
blic Opinion Quarterly 50: 97-111.
Ostos, Pilar, and Maria Luisa Pardo
1989 Documentos y notarios de Sevilla en el siglo XIII. Madrid: Fun-
dación Matritense del Notariado.
Palmer, Leonard R.
1980 The Greek Language. Atlantic Highlands, N.J.: Humanities Press.
Pascual, José Antonio
1991 Çufrir por sufrir. Voces 2: 103-108.
 References 323

Pastor Díaz de Garayo, Ernesto

1996 Castilla en el trànsito de la antigüedad al feudalismo: Pobla-
miento, poder político y estructura social. Del Arlanza al Duero
(siglos VII-XI). Valladolid: Junta de Castilla y León, Consejería
de Educación y Cultura.
Payne, Arvilla
1976 The acquisition of the phonological system of a second dialect.
Ph. D. diss., University of Pennsylvania.
1980 Factors controlling the acquisition of the Philadelphia dialect by
out-of-state children. In Locating Language in Time and Space,
William Labov (ed.), 143-178. New York: Academic Press.
Pei, Mario
1966 Glossary of linguistic terminology. New York: Anchor.
Pellegrini, Giovanni Battista
1975 Tra lingua e dialetto in Italia. In Saggi di linguistica italiana.
Storia struttura società, 11-35. Turin: Boringhieri.
Penny, Ralph
1969 El habla pasiega: Ensayo de dialectología montañesa. London:
Támesis.
1972 The reemergence of Iii as a phoneme in Castilian. Zeitschrift fur
Romanische Philologie 88: 463-482.
1983 The peninsular expansion of Castilian. Bulletin of Hispanic Stud-
ies 60: 333-338.
1987 Patterns of Language-Change in Spain. London: University of
London, Westfield College.
1990 Labiodental Iii, aspiration and /h/-dropping in Spanish: The
evolving phonemic values of the graphs f and h. In Cultures in
Contact in Medieval Spain: Historical and literary essays pre-
sented to L. P. Harvey, David Hook and Barry Taylor (eds.),
157-182. London: King's College.
1991a A History of the Spanish Language. Cambridge: Cambridge
University Press.
1991b El origen asturleonés de algunos fenómenos andaluces y america-
nos. Lletres asturiones 39: 32-40.
1992 Dialect contact and social networks in Judeo-Spanish. Romance
Philology 46: 125-140.
1993 Neutralization of voice in Spanish and the outcome of the Old
Spanish sibilants: A case of phonological change rooted in mor-
phology? In Hispanic Linguistic Studies in Honour of F. W. Hod-
croft, David Mackenzie and Ian Michael (eds.), 75-88. Llangran-
nog: Dolphin.
324 References

1995 Sobre el concepto del castellano como dialecto 'revolucionario'.

In Actas del I Congreso de Historia de la Lengua en América y
España, María Teresa Echenique, Milagros Aleza, and María Jo-
sé Martínez (eds.), 403-407. Valencia: Universität de València.
2000 Variation and Change in Spanish. Cambridge: Cambridge Uni-
versity Press.
2002 Contacto de variedades y resolución de la variación: aspiración y
pérdida de /h/ en el Madrid del s. XVI. In Actas del V Congreso
Internacional de Historia de la Lengua Española (Valencia, 31
de enero-4 de febrero de 2000), Vol. 1, Juan Sánchez and María
Teresa Echenique Elizondo (eds.), 397-406. Madrid: Gredos.
Pensado, Carmen
1999 El artículo eli y otras formas con -eli final en castellano medie-
val. Boletín de la Real Academia Española 79: 377—406.
Peñarroja Torrejón, Leopoldo
1990 El mozárabe de Valencia: Nuevas cuestiones de fonología mozá-
rabe. Madrid: Gredos.
Pérez de Urbel, Justo
1945 Historia del Condado de Castilla. Madrid: Consejo Superior de
Investigaciones Superiores, Escuela de Estudios Medievales.
Petrini, Dario
1988 La koinè ticinese. Livellamento dialettale e dinamiche innovative.
Berne: Francke.
Piel, Joseph
1931 Da evoluçâo dos grupos consonânticos com / em portugués e
espanhol, a propòsito de duas etimologías cocha e cascho. Biblos
7:512-521.
Posner, Rebecca
1996 The Romance Languages. Cambridge: Cambridge University
Press.
Quilis, Antonio, Margarita Cantarero, María José Albalá, and Rafael Guerra
1985 Los pronombres le, la, lo y sus plurales en la lengua española
hablada en Madrid. Madrid: Consejo Superior de Investigaciones
Científicas.
Ravid, Dorit Diskin
1995 Language Change in Child and Adult Hebrew. Oxford: Oxford
University Press.
Repetti, Lori, and Edward Tuttle
1987 The evolution of Latin PL, BL, FL, and CL, GL in Western Ro-
mance. Studi mediolatini e volgari 33: 52-115.
 References 325

Ridruejo, Emilio
1995 Procesos migratorios y nivelación dialectal en los inicios de la
reconquista castellana. In Estudis de lingüística i filologia oferts
a Antoni M. Badia i Margarit, Vol. I, F. Mohren (ed.), 235-251.
Barcelona: Publicacions de l'Abadia de Montserrat.
Riu Riu, Manuel
1989 Edad Media (711-1500). Madrid: Espasa-Calpe.
Rivarola, José Luis
1996 La base lingüistica del español de América: ¿existió una koiné
primitiva? Lexis 20 (1-2): 577-595.
Salvador Salvador, Francisco
1978 La neutralización 1/r explosivas agrupadas y su área andaluza. Ph.
D. diss., Universidad de Granada.
Samarin, William J.
1971 Salient and substantive pidginization. In Pidginization and Creo-
lization of Languages, Dell Hymes (ed.), 117-140. Cambridge:
Cambridge University Press.
Sánchez Albornoz, Claudio
1966 Despoblación y repoblación del valle del Duero. Buenos Aires:
Instituto de Historia de España.
Sancbis Calvo, María del Carmen
1991 El lenguaje de la Fazienda de Ultramar. Madrid: Anejos del
Boletín de la Real Academia Española.
1992 Sobre leísmo y apócope del pronombre de tercera persona singu-
lar objeto directo. In Actas del II Congreso Internacional de His-
toria de la Lengua Española, Vol. 1, M. Ariza, J. M. Mendoza,
R. Cano, and A. Narbona (eds.), 805-812. Madrid: Pabellón de
España.
Sandve, B. H.
1976 Om talemâlet i industritadene Odda og Tyssedal. Generasjonss-
kilnad og tilnaerming mellom dei to mâlf0ra [On the spoken lan-
guage in the industrial towns Odda and Tyssedal. Generational
differences and convergence between the two dialects]. Ph. D.
diss., Department of Nordic Languages and Literature, University
of Bergen.
Sanga, Glauco (ed.)
1990 Koinè in Italia dalle origini al cinquecento. Bergamo: Lubrina.
Sanz Fuentes, María José
1991 Instituciones de Andalucía: estudio diplomático. In Diplomatario
andaluz de Alfonso X, Manuel González Jiménez (ed.), CLXXV-
CCI. Seville: El Monte. Caja de Huelva y Sevilla.
326 References

Schlieben-Lange, Brigitte
1977 L'origine des langues romanes - Un cas de créolisation? In Lan-
gues en contact: Pidgins, Créoles, Jürgen Meisel (ed.), 81-101.
Tübingen: Narr.
Scholtmeijer, Harm
1992 Het Nederlands van de Ijseelmeerpolders. Ph. D. diss., Univer-
sity of Leiden. Kampen: Mondiss.
1997 Language in the Dutch Polders: Why dialects did not mix. Paper
given at the Conference on the Convergence and Divergence of
Dialects in a Changing Europe, European Science Foundation
Network, October, Heidelberg, Germany.
Schürr, Friedrich
1970 La diphtongaison romane. Tübinger Beiträge zur Linguistik 5.
Tübingen.
Selinker, Larry
1972 Interlanguage. International Review of Applied Linguistics 10:
209-231.
1992 Rediscovering Interlanguage. London: Longman.
Siegel, Jeff
1975 Fiji Hindustani. University of Hawaii Working Papers in Linguis-
tics 7: 127-144.
1985 Koines and koineization. Language in Society 14: 357-378.
1987 Language Contact in a Plantation Environment: A sociolinguis-
tic history of Fiji. Cambridge: Cambridge University Press.
1988a Introduction. In Language Transplanted: The Development of
Overseas Hindi, Richard K. Barz and Jeff Siegel (eds.), 1-22.
Wiesbaden: Otto Harrassowitz.
1988b The development of Fiji Hindustani. In Language Transplanted:
The Development of Overseas Hindi, Richard K. Barz and Jeff
Siegel (eds.), 121-149. Wiesbaden: Otto Harrassowitz.
(ed.)1993a Koines and Koineization International Journal of the Sociology
of Language 99.
1993b Dialect contact and koineization: A review of Dialects in Contact,
by Peter Trudgill. International Journal of the Sociology of Lan-
guage 99: 105-121.
1997 Mixing, levelling and pidgin/creole development. In The Struc-
ture and Status of Pidgins and Creoles, Arthur K. Spears and
Donald Winford (eds.), 111-150. Amsterdam/Philadelphia: John
Benjamins.
2001 Koine formation and creole genesis. In Creolization and Contact,
Norval Smith and Tonjes Veenstra (eds.), 175-197. Amster-
dam/Philadelphia: John Benjamins.
 References 327

Silva-Corvalán, Carmen
1994 Language Contact and Change: Spanish in Los Angeles. Oxford:
Oxford University Press.
Slobin, Dan
1985 Crosslinguistic evidence for the language-making capacity. In
The Crosslinguistic Study of Language Acquisition, Dan Slobin
(ed.), 1157-1256. Hillsdale, NJ: Lawrence Erlbaum.
Smith, John Charles, and Clive Sneddon
2001 The role of koineization in the development of 'preposition +
article' sequences in French. Paper given at the 34th Meeting of
the Societas Linguistica Europaea, 28-31 August 2001, Leuven,
Belgium.
Staaff, Erik
1907 Etude sur l'ancien dialecte léonais. Uppsala: Almqvist and Wik-
sell.
Steinsholt, Anders
1962 Mâlbryting i hedrum. Oslo: Universitetsforlaget.
Tagliavini, Carlo
1949 Le origini delle lingue neolatine. Bologna: Pàtron.
Tarone, Elaine
1977 Conscious communication strategies in Interlanguage: A progress
report. In On TESOL 77. Teaching and Learning English as a
Second Language, H. Brown, C. Yorio, and R. Crymes (eds.),
914-203. Washington, D.C.: TESOL.
Terrado Pablo, Javier
1986 Catalanismos, lusismos y dialectalismos andaluces en un docu-
mento de 1380. Vox Romanica 45: 168-184.
Thomason, Sarah
1997 A typology of contact languages. In The Structure and Status of
Pidgins and Creoles, Arthur K. Spears and Donald Winford
(eds.), 71-88. Amsterdam/Philadelphia: John Benjamins.
Thomason, Sarah Grey, and Terrence Kaufman
1988 Language Contact, Creolization and Genetic Linguistics. Berke-
ley: University of California Press.
Thomson, George
1960 The Greek Language. Cambridge: Heffer.
Tilander, Gunnar
1956 Vidal Mayor: traducción aragonesa de la obra In excelsis dei
thesauris de Vidal de Canellas. Vol. 1. Lund: Hâkan Ohlssons.
Torreblanca, Máximo
1986a Las oclusivas sordas hispanolatinas: El testimonio árabe. Anuario
de letras 24: 5-25.
328 References

1986b Omisión de grafemas en los documentos medievales de Castilla.

Journal of Hispanic Philology 10: 229-236.
1988 La fonología histórica española, los documentos y los dicciona-
rios medievales. Journal of Hispanic Philology 12: 139-149.
1989 Dos observaciones sobre Orígenes del español. Romance Philo-
logy 42 (4): 396-403.
1990 La evolución /kl-, pi-, fi-/ > /λ/ en español. Revista de filología
española 70: 317-327.
Torres Fontes, Juan, Julio González González, Salvador de Moxó, and María
Paz Alonso Romero
1990 Historia de España Menéndez Pidal: La expansión peninsular y
mediterránea (c. 1212 - c. 1350). Vol. I: La Corona de Castilla.
Madrid: Espasa-Calpe.
Trask, R. L.
1997 The History of Basque. London: Routledge.
Trask, R. L., and Roger Wright
1988 El 'vascorrománico'. Verba 15: 361-373.
Traugott, Elizabeth
1977 Pidginization, creolization and language change. In Pidgin and
Creole Linguistics, Albert Valdman (ed.), 70-98. Bloomington:
Indiana University Press.
Trudgill, Peter
1986 Dialects in Contact. Oxford: Blackwell.
1989 Language contact and simplification. Nordlyd 15: 113-121
1992 Dialect typology and social structure. In Language Contact:
Theoretical and empirical studies, Ernst Hâkon Jahr (ed.), 195—
211. Berlin/New York: Mouton de Gruyter.
1994 Language contact and dialect contact in linguistic change.
Dialektkontakt, spràkkontakt och spräkßrändring i Norden:
Föredrag frân ett forskarsymposium, Ulla-Britt Kotsinas and
John Helgander (eds.), 13-22. Stockholm: Stockholm University.
1996 Dialect typology: Isolation, social network and phonological
structure. In Towards a social science of language: Papers in
Honor of William Labov, Vol. 1, Gregory Guy, Crawford Feagin,
Deborah Schiffrin, and John Baugh (eds.), 3-21. Amster-
dam/Philadelphia: John Benjamins.
1998 The chaos before the order: New Zealand English and the second
stage of new-dialect formation. In Advances in Historical Socio-
linguistics, Ernst Hâkon Jahr (ed.), 197-207. Berlin/New York:
Mouton de Gruyter.
 References 329

2002 Linguistic and social typology. In The Handbook of Language

Variation and Change, J. K. Chambers, Peter Trudgill, and Nata-
lie Schilling-Estes (eds.), 707-728. Oxford: Blackwell.
Trudgill, Peter, Elizabeth Gordon, and Gillian Lewis
1999 New-dialect formation and Southern Hemisphere English: The
New Zealand Short Front Vowels. Journal of Sociolinguistics 2:
35-51.
Trudgill, Peter, Elizabeth Gordon, Gillian Lewis, and Margaret Maclagan
2000 Determinism in new-dialect formation and the genesis of New
Zealand English. Journal of Linguistics 36: 299-318
Trudgill, Peter, Gillian Lewis, and Margaret Maclagan
2000 The role of drift in the formation of native-speaker southern
hemisphere Englishes: Some New Zealand evidence. Dia-
chronica 17(1): 111-138.
Tusell, Javier, Carlos Martinez Shaw, and José Luis Martin
1998 Historia de España. Madrid: Taurus.
Tuten, Donald Ν.
2000 Linking social change and linguistic change: Koineization in
early Castile. In New Approaches to Old Problems: Issues in
Romance Historical Linguistics, Steven N. Dworkin and Dieter
Wanner (eds.), 97-105.
Valle, José del
1998 Andalucismo, poligénesis y koineización: dialectología e ideolo-
gía. Hispanic Review 66: 131-149.
Versteegh, Kees
1993 Leveling in the Sudan: From Arabic creole to Arabic dialect.
International Journal of the Sociology of Language 99: 65-80.
Wagner, Max
1920 Amerikanospanisch und Vulgärlatein. Zeitschrift fur Romanische
Philologie 40: 286-312, 385^04.
Wang, William S-Y.
1969 Competing changes as a cause of residue. Language 45: 9-25.
Weinreich, Uriel
1968 Languages in Contact. The Hague: Mouton.
Weinreich Uriel, William Labov, and Marvin Herzog
1968 Empirical foundations for a theory of language change. In Direc-
tions for Historical Linguistics, Winfred P. Lehman and Yakov
Malkiel (eds.), 95-189. Austin: University of Texas Press.
Whinnom, Keith
1980 Creolization in linguistic change. In Theoretical Orientations in
Creole Studies, Albert Valdman and Arnold Highfield (eds.),
203-212. New York: Academic.
330 References

Willemyns, Michael, Cynthia Gallois, Victor J. Callan, and Jef&ey Pittan

1997 Accent accommodation in the job interview. Impact of inter-
viewer accent and gender. Journal of Language and Social Psy-
chology 16: 3-22.
Wolfram, Walt, and Natalie Schilling-Estes
1998 American English: Dialects and Variation. Oxford: Blackwell.
2003 Dialectology and linguistic diffusion. In The Handbook of His-
torical Linguistics, Brian D. Joseph and Richard D. Janda (eds.),
713-735. Oxford: Blackwell.
Wright, Roger
1980 Linguistic reasons for phonetic archaisms in Romance. In Papers
from the 4th International Conference on Historical Linguistics,
Elizabeth Closs Traugott, Rebecca Labrum, and Susan Shepherd
(eds.), 331-337. Amsterdam/Philadelphia: John Benjamins.
1982 Late Latin and Early Romance in Spain and Carolingian France.
Liverpool: Francis Cairns.
1994a Early Ibero-Romance. Newark, Delaware: Juan de la Cuesta.
1994b How old is the ballad genre? In Early Ibero-Romance, 290-299.
Newark, Delaware: Juan de la Cuesta.
1994c Logographic script and assumptions of literacy in tenth-century
Spain. In Early Ibero-Romance, 165-180. Newark, Delaware:
Juan de la Cuesta.
1994d Metalinguistic change in medieval Iberia. In Early Ibero-
Romance, 31-44. Newark, Delaware: Juan de la Cuesta.
1994e The purpose of the glosses of San Millán and Silos. In Early
Ibero-Romance, 209-219. Newark, Delaware: Juan de la Cuesta.
1994f Sociolinguistics in Spain (8th—11th centuries). In Early Ibero-
Romance, 155-164. Newark, Delaware: Juan de la Cuesta.
1996 Latin in Spain: Early Ibero-Romance. In The Origins and Devel-
opment of Emigrant Languages: Proceedings from the Second
Rasmus Rask Colloquium, Hans Nielsen and Lene Schosler
(eds.), 277-298. Odense: Odense University Press.
1999 Periodization and how to avoid it. In Essays in Hispanic Linguis-
tics Dedicated to Paul M. Lloyd, Robert J. Blake, Diana L. Ran-
son, and Roger Wright (eds.), 25-41. Newark, Delaware: Juan de
la Cuesta.
Yaeger-Dror, Malcah
1993 Linguistic analysis of dialect "correction" and its interaction with
cognitive salience. Language Variation and Change 5: 189-224.
Zamora Vicente, Alonso
1967 Dialectología española. 2nd ed. Madrid: Gredos.
 References 331

Zufferey, François
1987 Recherches linguistiques sur les chansonniers provençaux. Ge-
neva: Droz.
Index

Abad, Francisco, 293 Alcarria, La, 147, 150, 200, 245

abduction, 58-60, 125 Alfonso X, 183, 207, 216, 222, 231,
accentuation, 4,11, 12, 28, 34, 113, 234-237, 294
119, 120, 122, 123, 127, 131, Alonso Romero, Maria Paz, 283
133, 138, 152, 164, 168, 169, Alonso, Amado, 223-225, 255
171, 196, 205-208, 212, 213, Alvar, Manuel, 116, 222, 290, 298
240, 241, 257, 260, 289, 294 Alvar, Manuel, 174
accommodation, 22, 28-33, 35, 37, Álvarez Maurin, Maria del Pilar, 113
38.41.47, 48, 50, 53-55, 57, 59, Álvarez Rodriguez, Adelino, 163,
62, 68-70, 75, 78, 88, 90, 112, 291
119,124, 130, 143, 164, 172, Allen, J. D. H., 168, 169
187, 188, 201,202, 239, 257, allophone, 46, 65, 135, 136, 296
264, 274, 277-279, 281, 282 Amery, Rob, 71
acquisition, 2, 13, 21, 26, 28, 30-34, analogy, 60, 125, 130, 135, 143,
36, 38,40, 43,45, 46, 48, 50, 182, 190, 191, 194, 203,205,
54-66, 68, 70, 74, 80, 85-87, 90, 206, 208, 209,212,213,241,
92, 96, 101, 105, 112, 117, 119, 258, 262, 263, 289
123, 125, 126, 133, 134, 143, Andalusia/Andalusian(s), 5, 7, 17,
153, 155, 157, 178, 179, 183, 68, 85, 121, 143,152, 175, 183,
198,213,219, 231,234, 237, 185, 188, 195,202,215-256,
253, 261, 264, 267, 268, 274, 263, 275, 276, 295, 296, 297,
277-281, 287 298, 299, 300
actuation, 113, 131, 191, 192, 207 Andersen, Henning, 58-60, 80
adolescents, 31, 34, 56, 62, 64, 65, Anderson, James M., 138, 290
67, 68, 71, 75, 90, 164, 257, 280, animacy, 181-184, 186, 187, 190-
282 192
adults, 18, 26, 27, 31, 32, 34, 35,40, apocope, 157, 159-173, 184, 186,
41.47.48, 54, 56-58, 61-71,74, 194-199, 205, 212-214, 232,
75, 77, 90, 93, 118, 119, 124, 233-237, 243-245, 256, 260-
125, 130,213, 231,239, 244, 265, 291, 292, 297, 298, 300,
256, 257, 264, 267, 268, 279- 301
281, 301 Arabic, 15, 18, 22, 70, 120, 147,
Alarcos Llorach, Emilio, 123, 124, 285,291
285, 289 Aragon/Aragonese, 97, 104, 110,
Albalá, María José, 293 114, 116, 117, 120-122, 125-
 Index 333

129, 131, 136, 139, 147, 148, Berruto, Gaetano, 16

150, 152, 154, 158, 164, 173, Bezler, Francis, 161
186,188,195,199,200,209, Bidwell, Charles E., 15
220, 221, 233, 260, 261, 275, Bishko, Charles Julian, 145, 151,
276, 286, 287, 290, 295, 301 152, 295
Ariza, Manuel, 230, 231, 242, 252- Blanc, Haim, 15,18, 22, 70, 89, 277
255, 298-300 Blanchet, Philippe, 276
articles, 94, 113-119,131, 132,138, borrowing, 22, 24, 25, 29, 62, 86,
143, 155, 156, 175, 203, 205, 88,91, 108, 133, 134, 170, 171,
208, 212, 255, 258, 260, 275, 191, 192, 243, 291
285-287, 294, 301 Bortoni, Stella Maris, 3,40
aspiration, 5,47, 110, 132, 133, 135, Britain, David, 43, 44, 46, 71-73,
230, 288, 290, 296 281
assimilation, 35, 118, 131,136,137, Bubenik, Vit, 10-13, 276
171, 254, 277, 286, 287, 289 Buck, Carl Darling, 10, 12
Asturias/Asturian(s), 95-99, 103, Burgos, 4, 5, 7, 93-145, 149, 151-
105, 121, 123, 128, 129,139, 153, 155, 159, 162-164, 167,
140, 142, 143, 147, 148, 151, 170, 177, 179-181, 188, 193,
163,176, 177, 193,195, 206, 195,197-199, 204-207, 210-
221, 276, 287, 288, 290, 294, 212,214, 220, 235-237, 239-
295 242, 245, 258-264, 284-287,
Azevedo Maia, Clarinda de, 234 289, 294, 297, 299, 301
Bybee, Joan, 57, 61, 196,281
Badia Margarit, Antoni, 125, 139
Bailey, Charles-James N., 15 Callan, Victor J., 278
Bamgbose, Ayo, 15 Callary, Robert E., 82, 159
Barrios, Ángel, 99, 142, 200 Canadian Raising, 44, 72
Bartsch, Renate, 49, 50 Cannell, C., 278
Barz, Richard K., 3 Cano Aguilar, Rafael, 301
Basque(s), 87, 93, 95-98, 100, 103, Cantabria/Cantabrian(s), 93, 95, 96,
108, 109, 123, 124, 132-134, 98, 99, 103, 105, 107, 132, 135-
136, 143, 144, 175, 177-179, 137, 139, 140, 142, 143, 148,
181, 182, 191-193, 198,221, 151, 176-180, 186, 191-193,
243, 259, 277, 283-285, 288, 195, 196, 198, 220, 259, 261,
289, 300 285, 287, 290, 293, 295
Beebe, Leslie M., 30 Cantarero, Margarita, 293
Bell, Allen, 278 Cardona, Giorgio Raimondo, 13, 14,
Bemba, 15 16
Berceo, 184, 185, 188
334 Index

case, 6, 8, 11, 20, 23, 25-27, 29, 33, contact between stable dialects, 28,
35, 37,40, 42-46, 49, 50, 53-55, 38, 79
58, 59, 66, 67, 69-77, 80, 82, 85, convergence, 29,76, 88
86, 88, 89,91,93,105-107, 110, Cooper, Louis, 199, 295
112, 113, 119, 122, 126, 130, Cooper, Robert L., 155
132, 137, 144, 148, 156-158, Corbett, Noel, 125
164-166, 169, 172-183, 185- Coraminas, Joan, 138, 140, 142, 290
192, 194,195,197-199, 201- count/mass distinction, 157, 175-
203, 220, 221, 228, 232, 243, 177, 179,180, 181, 182, 184,
244, 246, 254-256, 259-264, 186,187, 188, 191, 192, 197,
268, 274, 275, 277, 278, 281, 199, 201,203, 261,280, 293
283,284, 290, 291,293, 294, creole/creolization, 22, 24-27,40,
300 67, 74, 76-78, 87, 109, 267, 282,
Castile, primitive, 96, 98, 99, 105, 285, 301
116, 135, 159, 181, 191-193, Croft, William, 37, 49, 75, 76, 274
195-198, 288, 293 Crystal, David, 17
Catalán, Diego, 122, 167,199, 225, Cuervo, Rufino José, 185, 189, 194
287, 289 cuñalwedge, 127, 131, 153, 154, 301
Catalonia/Catalan(s), 95, 97, 150, Chambers, J. K., 31, 61, 83, 158,
221, 275 159, 280
catastrophic/punctuated change, 14, Cheshire, Jenny, 32, 279
262, 265, 274 children, 25-27, 31, 34, 36, 40, 48,
ceceo, 5, 224, 231, 250, 251, 296, 54, 56-58, 60-72, 75, 78, 82, 86,
298, 299 87, 90,91,93, 118, 119, 125,
Clark, Eve, 119, 198 130, 134, 164, 171, 197, 198,
Clavería Nadal, Gloria, 108 213, 239, 257, 264, 267, 268,
code-mixing, 78, 282 279-282, 293, 301
code-switching, 55, 69, 282 Chomsky, Noam, 49
Coleman, L., 278
consonant clusters, 139, 140, 166, deduction, 58, 60
290 DePaulo, Bella M., 278
consonants, 11,12, 35,44,47,131, Deutscher, Guy, 59, 60
137, 138,140, 141, 157, 160, Díaz Montesinos, Francisco, 253,
161, 163, 165, 166, 168, 169, 299
197, 205, 223, 259, 286, 289, Diaz Roig, Mercedes, 103
290-292 Diaz y Diaz, Manuel C., 161
constraints, 2, 18, 31, 32,48, 50, 78, Diez, Miguel, 154
91, 92, 165-167, 169, 233, 258, difusión (vs. focusing), 27, 39, 67
263, 281,292 Dillard, J. L., 15, 17, 22, 81
 Index 335

dipbthong/diphthongìzation, 11, 38, evidence, 5, 6, 8, 22, 32, 34, 36,43,

44, 69, 105, 113, 120-131, 133, 44, 56, 60, 68, 71-73,79-81, 95,
135,143,156,205,213,258, 98, 99, 106, 107, 111-114, 120-
260, 288, 289, 301 122,125, 127-129, 133, 136-
diphthongizatíon before yod, 126— 138,140, 147, 148,151, 153,
129,131,288, 289 154, 159, 161, 163, 165, 171,
dissimilation, 205, 254, 290 182,183, 186, 188,190, 193,
Dixon, R. M. W., 262 203,207, 209,210,211,214,
Domingue, Nicole, 43, 48 215, 219, 225, 226, 228-234,
Dorian, Nancy, 87 239, 240,243-248, 251,252,
drift, 35, 92,117, 264, 265, 301 253,254, 255, 256, 259, 261,
263, 264, 267, 268, 280, 281,
Eberenz, Rolf, 265, 266 283-289, 291, 294, 295, 297-
Echenique Elizondo, Maria Teresa, 299
184-188, 194, 293 exceptionalism, 104, 283
Elcock, W. D., 16 explanatory limits, 7, 93, 131-143,
Ellis, Rod, 31 255-256
English, Australian, 35,44, 82, 262, external factors, 1, 189,191, 193,
281 266, 274
English, Belfast, 44,45, 50-52, 279 Extremadura, primitive, 147, 148,
English, Canadian, 16, 31,44,72, 200
81, 83, 280
English, English, 4, 15, 31-33, 37, F azienda de Ultramar, 185
63, 65, 67, 69, 72, 80-82, 221, Fehderau, Harold W., 15
264, 279-281, 290, 296 Ferguson, Charles Α., 15, 17, 18, 22,
English, New Zealand, 34, 35, 68, 45
71,81,281 Fernández Ramírez, Salvador, 189,
English, Scottish, 44, 281 201
English, South African, 23, 35, 279 Fernández-Ordóñez, Inés, 151,175-
English, Southern Hemisphere, 34, 195, 199-203, 243, 267, 293
264,279, 296 Ferreira, Manuel, 115
English, U.S., 15, 81, 278 feudalism, 153, 284
Ervin-Tripp, Susan M., 15 Flores Cervantes, Marcela, 190
Estoria de España, 184,186, 211, focusing, 18-20, 22-24, 26-28, 36,
294 39-41,45,48,49, 53, 54, 57, 63,
etymological system, 173-178, 181, 64, 66-74, 92, 100, 103, 105-
182, 186-188, 199, 200, 202, 107, 113, 114, 119, 123, 126,
214, 242, 245, 262, 297 134, 143, 155, 161, 167, 170,
172, 198, 204, 205,207,211,
336 Index

227,259, 260, 264, 279, 288, 136, 139-143, 147, 148, 150,
300, 301 151, 161, 163,200, 220, 221,
Fogny, 15 230,234,245,252,259,276,
Fontanella de Weinberg, María Bea- 285-288
triz, 3, 26,47, 85, 89 Galmés de Fuentes, Alvaro, 120,
Frago Gracia, Juan Antonio, 7, 8, 289, 291
158, 226-231, 234, 246-255, Gallois, Cynthia, 278
263, 290, 296-299 Gambhir, Surendra K., 3, 15, 16, 23,
France/French, 4, 16, 17, 107, 118, 42, 87, 89
119, 146, 158, 170, 171, 173, García de Cortázar, José Ángel, 283
205, 221,233,237, 274, 287, García de Diego, Vicente, 107, 108
290, 292 García González, Francisco, 191
francos, 147, 148, 149, 151, 152, García, Èrica, 190
153, 170,171, 172, 214, 220, gender, 162, 174-179, 192, 197,
221, 261, 265, 290 204-207, 209-211, 213, 238-
frequency, 16-18, 20, 26, 27, 30, 32, 240, 242, 260, 264, 277
33, 35, 37, 41-43, 4 5 ^ 8 , 51, 54, General Estoria, 169, 186, 211, 236,
55, 57-61, 63-65, 67, 68, 70,71, 294
74, 75, 77, 79, 80, 82, 86, 89, 90, generation, 10, 21, 26, 27, 35,40,
96, 100, 104, 105, 111, 112, 114, 50, 53, 60, 62, 64, 67-72, 80, 83,
119, 120, 122,124, 125, 128- 84, 88, 90,91, 104, 125, 129,
131, 133, 134, 136, 138, 139, 137, 229, 234, 239, 259, 262,
142, 143, 148, 149, 152, 155- 264, 267, 282, 296
157, 161, 163, 164, 166-169, German, 15, 107
172, 174, 181, 182, 184, 185, Giddens, Anthony, 282
187, 193, 196-198, 199, 202, Giles, Howard, 29, 30
206, 207,211-213,218, 222, Gimeno Menéndez, Francisco, 110,
230, 234-240, 245, 253, 255, 111
257-261, 275, 277, 279, 281, González Gómez, Antonio, 218, 219,
282, 286, 287, 292, 295, 296, 220, 221, 295
298, 300 González Jiménez, Manuel, 200,
frontier, 96, 102, 126, 145-147, 218-222, 231,232, 295
149-151,202,215,216,218, González, Julio, 150, 201, 218, 219,
222, 284 220, 283
fudging, 36, 38, 279 Gordon, Elizabeth, 3, 33, 35, 42, 68,
Fuero de Madrid, 162, 186 71,81,245, 279,296
Graff, Willem Laurens, 16
Galicia/Galician(s), 95-97, 99-103, grammarian, 13, 201, 202
105, 115, 119, 120, 122, 124, Granda, Germán de, 3, 224, 229, 283
 Index 337

Granovetter, Mark, 51 identity, 13, 28, 29, 38, 39,45, 49,

Greek, 9, 11, 12, 13, 21, 24, 77, 276 50, 52, 54, 69, 70, 75, 77, 78, 83,
Guaraní, 175 88,92,102-104, 111, 124,126,
Guerra, Rafael, 293 143, 147, 157, 163, 198, 257,
Gumperz, John J., 88 259, 263, 277, 278, 284, 285
Imhoff, Brian, 290, 291
Hall, Robert Α., 17, 276 immigrant koine, 20, 21
Haller, Hermann W„ 15 impoverishment, 25, 87
Hancock, Ian F., 16 Indonesian, 15
Hardwick, Margaret, 83 induction, 58, 60, 80, 90, 130, 258,
Harris, John, 47 261
Harris-Northall, Ray, 6,165,169, innovation, 23, 31, 36, 50, 51, 54,
170,172, 196, 235, 236, 253, 56-61,78, 80, 83, 87, 111, 120,
265, 266, 292, 299 143, 155, 158, 159, 165, 166,
Hartmann, R. R. K., 15 170, 178, 181, 182, 194, 197,
Haugen, Einar, 26, 84 198, 201,205, 207-211,231,
Hebrew, Israeli, 15, 17, 22, 58, 70, 240, 241, 258, 259, 261-263,
89, 277, 281 275, 279, 285, 289, 294, 297
Henríquez Ureña, Pedro, 223 input/output, 18, 25,46, 49, 56, 58,
Henry, Alison, 60 60-62, 63, 70, 74-77, 80, 81, 87,
Hernández, César, 292 90,118,119, 124-126, 130, 134,
Herzog, Marvin, 2 166, 195, 197, 198, 213, 214,
Hidalgo, Margarita, 3, 283 240, 258, 260, 261, 287
Hill, Trevor, 15 intelligibility, mutual, 17, 24, 27, 29,
Hindi-Bhojpuri, 3, 15, 20, 22-25, 75, 77, 79, 90, 106, 143, 257,
27, 38,41-43, 87, 89 268, 276
Historia troyana, 186 interdialect, 28, 36-38,42, 54, 61,
Hock, Hans Heinrich, 10, 11 90, 106, 159, 202
Hofstadter, Richard, 284 interlanguage, 18, 38, 54-56, 70, 90,
homogenization, 80, 82, 83,123, 134, 159,178
155, 262 internal factors, 33, 191, 192, 301
homonymie clash, 32, 192,278, 290 invisible hand, 2, 47-50, 63, 64, 90-
Hopper Paul, 61, 196 92, 257
Horrocks, Geoffrey, 10-13,18, 275, Iradiel, Paulino, 283
276 Italian, 4, 15, 16, 26, 134, 177, 221,
Heyanger, 40, 45, 67-69 276, 290
Huber, Joseph, 115 Izzo, Herbert, 133
hypercorrection, 37, 163, 197, 209
hyperdialectalism, 38, 278 Jakobson, Roman, 14
338 Index

Janda, Richard D., 126, 286 189,194, 197, 225, 227, 233-
jargon, 16,24 237, 243, 246-248, 254, 255,
Jews, 147,149,219 274,291,292,298,299
Jonxis-Henkemans, Wilhelmina 207, Lapidario, 211, 294
211 Lass, Roger, 192,268, 274, 279
Latin, 3, 14, 15,87, 95, 119, 120,
Kannada, 88 122, 131, 132, 133, 135, 136,
Karlgren, Bernhar, 15 138-140, 144,156,165, 174,
Kasa, 15 183, 185, 189, 196, 204-206,
Kasten, Lloyd, 207,211 259,266, 276, 288-290, 294,
Kauffeld, Cynthia J., 299, 300 296
Kaufman, Terrence, 25, 74, 75, 79, least common denominator, 14, 38,
86, 88,277, 282 261
Keller, Rudi, 2, 268 Lee, Everett S., 284
Kerswill, Paul, 3, 7, 19, 26, 27, 31- leísmo, 6, 68, 106, 157, 159, 162,
35,40,46, 53, 61-73, 76, 80, 86, 173-203, 214, 232, 242-245,
89, 92,93, 229, 257, 264, 268, 256, 260-265, 293, 297, 298
274, 277, 279, 281,282 Lenneberg, Eric, 61
Kikongo, 15 Leon/Leonese, 96-105, 110,111,
Kiparsky, Paul, 57, 58, 280 114, 115, 117, 120-123, 126-
Klein-Andreu, Flora, 175, 202 129, 131, 136, 139, 140, 142,
Knecht, Pierre, 17 143,145-151,154-158, 163,
Koine, Greek, 9-14, 16, 18, 20, 24, 164, 166, 167, 181, 187, 188,
275, 276 197, 199, 200, 202, 206, 209,
Krüger, Fritz, 121 216, 217, 220-223, 228, 230-
234, 237, 245, 260, 275, 276,
Labov, William, 2, 6, 38, 112, 126, 283-292, 295, 297, 298, 300,
262, 274, 278, 281 301
Lacarra, José Maria, 152 LePage, Robert B., 18, 39,40, 83
laísmo, 173, 174, 179, 180-183, leveling, 3-5, 28, 29, 41,45-47, 64,
185, 187, 188, 192, 203,293 68,74, 76, 86, 87, 90,91, 112,
language death, 86, 87, 89 114, 138, 160, 207-211,213,
language missionary, 297 223, 229, 238, 242, 258, 264,
language shift, 21, 86, 89, 133, 301 275, 279, 280, 294, 295, 297
language spread, 21, 140, 149, 153, leveling, dialect, 13, 22, 33, 67, 68,
155, 158,219, 222, 260 79-84, 86, 89,107, 126, 137,
Lantolf, James, 300 153-159, 222, 223, 227, 228,
Lapesa, Rafael, 151, 154, 160, 162- 256, 260, 275,277, 282
168, 170,171, 184, 185, 187-
 Index 339

Lewis, Gillian, 3, 33, 35,42, 68,71, Madrid, 4, 5, 147, 149, 179, 182,
81, 245, 264, 279, 296, 301 200, 289
lexical diffusion, 36, 62,129,130, Mahlum, Brit, 282
140, 144, 258, 290 Malkiel, Yakov, 139, 140,142, 241,
lexicon, 26, 29, 32, 74, 75, 77, 88, 290, 294
92, 140, 280, 298 Mancha, La, 147, 150, 202, 245, 263
Liber Regum, 199, 295 Manessy, Gabriel, 15
Libro conplido en los judizios de las Marathi, 88
estrellas, 186, 187 Marcos Rodríguez, Florencio 156
Libro de Buen Amor, 187, 188 markedness, 25, 28, 38, 41, 42,46,
Libro de las cruzes, 186, 211, 294 64, 77, 79, 80, 87, 119, 124, 130,
Libro de las leyes, 211, 294 152, 155, 157, 172, 181, 186,
Lightfoot, David, 57,280 190,192, 193,197, 202, 213-
Lingala, 15 215,234, 237, 241,245, 256,
lingua fianca, 14-21, 109, 285 258-261, 264, 275, 282, 293,
Lipset, Seymour, 284 295, 297, 301
literacy, 55, 57, 84, 85, 118, 123, Martin Martin, José Luis, 100, 156,
173, 281,285,294 283
Lodge, Anthony, 26, 275 Martínez Alcalde, María José, 265
loísmo, 173, 174, 179, 181, 183, Martínez Shaw, Carlos, 100, 283
185,187, 188, 203, 293 Marzys, Zygmunt, 17
López Bobo, María Jesús, 187 Matute Martínez, Cristina, 186
López de Aberasturi Arregui, Igna- McMahon, April M. S., 59
cio, 298 Meillet, Antoine, 13, 14, 111
López García, Ángel, 108, 109, 285 Melanesian Pidgin, 15
Lüdtke, Helmut, 118 Méndez Garcia de Paredes, Elena,
Lunt, Horace G., 15 204, 206, 207, 213, 238, 240,
Luquet, Gilles, 292 241, 297
Lyons, Christopher, 204, 205 Menéndez Pidal, Ramón, 100, 103-
Lleal, Coloma, 118, 242 107, 112-117, 119-123, 127,
Llórente Maldonado, Antonio, 175 128,131, 132, 138, 139, 153,
Lloyd, Paul M., 122, 127, 132,134- 154, 160, 161, 164, 195-197,
136, 142, 143, 241, 288-290, 204-207, 225, 241, 246, 258,
299, 300 259, 274, 283-285, 288, 293-
295, 300
Macedonian, 15 merger, 4, 5, 11, 17, 52, 130, 143,
Maclagan, Margaret, 33, 35,42, 68, 150, 181, 199, 223, 245,246,
71,81,245, 264, 279, 296, 301 251,253, 255,263,264, 281,
291,295, 296, 298, 300
340 Index

Mestone, Rajend, 3, 18, 20, 21, 23, 155, 158, 203,212,215,227,

26, 36, 279, 280 229, 256-258, 260, 267, 275,
Mestre Campi, Jesús, 98 277,278,281
metaphor, 1, 9, 13, 14, 274 Mohan, Peggy, 15
metathesis, 254 Mondéjar, José, 225, 230, 245,246
methodology, 7, 8, 9, 72, 89, 91, 93, Monge, Félix, 190
259, 263 monophthong, 69, 120, 130
military-religious orders, 146, 150, Montgomery, Michael, 81
153, 202, 222, 232, 245 Montgomery, Thomas, 196
Milroy, James, 1-3, 29, 45,49-52, Moors, 93, 96, 101, 102, 104, 145-
54, 105, 159, 268 147,215-217,219, 283
Milroy, Lesley, 50-52, 54, 117, 124, Morales, Francisco, 154
159, 265, 274, 277 Moreno Bernal, Jesús, 292
Milton Keynes, 36,45, 46, 63-65, Moreta, Salustiano, 283
67-69,78, 82, 89, 99,280,282 Morgan, Michael, 299
Minguez, José Maria, 283 morphosyntax, 6, 12,23, 26, 29, 32,
Mitchell, J. Clyde, 53 33, 35, 37, 46, 49, 58, 60, 62, 68,
Mitre Fernández, Emilio, 104, 283 70, 74-76, 85, 86,88, 111, 129,
mixing, 3,4, 5, 7, 12-14, 16-19, 21, 193, 199, 207,213, 241,242,
23-25, 27, 28, 35, 36, 38, 40,41, 243, 275, 279
44,45, 47, 50,51,72, 74, 79, 83, Μοχό, Salvador de, 97, 98, 101,
86, 88, 90-93, 103, 105-108, 102, 147-153, 200, 217, 219,
112, 113, 119, 128, 129, 133, 220, 221, 283, 290, 295
134, 139, 144-146, 148,149, Mozarab(s), 96, 98-100, 105, 108,
150, 152-154, 156, 158, 159, 120, 129, 143, 147-149, 152,
164, 173, 184, 193, 199-202, 163, 164, 195,200, 201,214,
214,215,219-224, 227-230, 219, 265, 276, 283, 285, 289,
233, 234, 243, 245, 246, 256, 291, 298
258, 259, 261, 263, 265, 274- Mudejars, 147, 148, 152, 214, 216,
276, 279, 281,284, 287, 289, 219, 220, 296
291, 295, 296-298 Mufwene, Salikoko, 76-78, 261,
Moag, Rodney F., 3 267,268, 274, 276, 282, 301
Moamyn, 211, 294 Mühlhäusler, Peter, 16, 24, 25, 45,
mobility, 52, 53, 80, 82, 101, 102, 79, 276
153, 217, 262, 274
model, 3,4, 6-9, 22-28, 31, 38,43, nativization, 15, 16, 18, 23, 24, 26,
47,49, 51, 52, 56-58, 60, 65, 68, 109
71-73, 76, 77, 79, 89-91, 93, naturalness, 2, 21, 32-34,47, 49, 75,
113, 123, 131, 132, 136, 138, 134, 136, 137, 150, 259, 283
 Index 341

Navarre/Navarrese, 95, 97, 98, 100, Ñuño Alvarez, Maria Pilar, 290
102,103, 110, 134,143, 147,
149-152,178,193,195,200, Ohala, John J., 35
220, 221, 289 Omdal, Helge, 3,40, 69
Navarro Carrasco, Ana I., 298 orthography, 12, 85, 113, 120, 122,
Navarro Tomás, Tomás, 206 129, 136, 156, 157, 206, 232,
neutralization, 36, 206, 253, 255, 248,251-253, 259, 288, 291,
264, 296, 300 292, 294, 299
Nevalainen, Terttu, 274 Oskenberg, L., 278
new town, 3, 28,40, 63, 83, 90,153, Ostos, Pilar, 232, 235, 238
282 overgeneralization, 37, 55-57, 59,
Nida, Eugene Α., 15 60,128, 136, 152, 198,210,212,
Nitti, John, 207, 211 223,227, 228, 241,258, 259,
non-prevocalic /-r/, 32, 35, 81 261, 280, 281
Nordenstam, Kerstin, 31, 32
norms, 18,20, 23, 27, 38^0,44, palatalization, 131, 137-140, 289
48-54, 56-63, 65, 67, 68, 71, 72, Palmer, J. M., 278
82, 85, 89, 90, 92,100, 102-106, Palmer, Leonard R., 10,11,12
112, 113, 118,119, 120,123, paraphrase, 11, 54, 55
125, 130, 134, 137, 142, 152, Pardo, Maria Luisa, 232, 235, 238
154-156, 158, 160, 164-166, Pascual, José Antonio, 138, 140,
171-174, 182-184, 187, 193, 142, 255, 290
196, 198, 199, 201, 202, 206- Pastor Díaz de Garayo, Ernesto, 99,
211,213,214, 229, 230, 232- 283, 284
234, 236-238, 241, 243, 252, Payne, Arvilla, 281
257, 261-264, 268, 277, 278, Pei, Mario, 15, 17
281, 282, 284-286, 288, 291, Pellegrini, Giovanni Battista, 16
301 Penny, Ralph, 4-6, 46, 76, 106, 111,
Norway/Norwegian, 31, 37, 40, 282 132, 135, 138, 139, 195,213,
nouns, 11, 98, 168, 175, 177,180, 265, 285-287, 289, 293, 294,
182, 186, 192,199, 203, 206, 297, 301
209, 210, 212, 238, 239, 285, Pensado, Carmen, 287
290, 294, 298 Peñarroja Torrejón, Leopoldo, 121,
number, 6, 11,12, 32, 37,42,45, 51, 291
52, 54,71,73, 75,81,99, 100, Pérez de Urbel, Justo, 283
102, 110, 124,125, 129-131, periodization, 5, 265, 266
136, 143, 149, 165, 166, 171, personal a, 80, 110, 178, 186, 188,
176,210,218, 221,229, 235, 189, 198
236, 240, 251,254,255, 279 Petrini, Dario, 19, 26
342 Index

phoneme, 32, 52, 55, 66, 123-125, 210, 235, 258, 260, 275, 286,
127, 130, 132, 136, 137,144, 287, 301
246,247,253,255,258,259, prestige, 9,10, 52,53, 85,101,104,
298,300 106, 123, 152, 155, 156, 171,
phonologizatìon, 123, 125,126,129, 172,174, 176,182, 189, 201,
133, 137, 138, 142, 144, 288, 230, 245, 261, 276, 279, 285,
301 288, 289, 293
phonology, 6, 23, 26,29, 32, 33, 37, pronouns, demonstrative, 189,190,
43,44,46,47, 62, 69,70,72, 191,198, 203
74-76, 86,91, 112, 119, 120, pronouns, object (see also leísmo), 5,
122, 123, 126, 130, 134,143, 173-203, 232, 233, 242, 294
161, 163, 164, 168, 169, 205, prototype, 9,18, 24, 27, 54, 74, 77,
229, 242,280, 288, 294, 296, 79, 83, 84, 86, 89, 92,102, 144,
298 190, 257, 259, 262, 265, 267,
phonotactics, 32, 165-167, 172, 197, 275, 277, 296
254, 291 Provençal, 16, 276
pidgin/pidginization, 22, 24-27, 67,
74-78, 89, 267, 276, 277, 285, Quechua, 175
301 Quilis Merin, Mercedes, 265
Piel, Joseph, 290 Quilis, Antonio, 293
Pittan, Jeffrey, 278
Portuguese, 40, 118-120, 136, 139, rate of change, 229, 267, 268
140-142, 148, 161, 234, 286, Ravid, Dorit Diskin, 57,70, 281
290 reallocation, 29, 3 6 , 4 3 ^ 5 , 68, 69,
Posner, Rebecca, 111, 130,137, 259, 72, 91, 159-161,172, 173, 204,
261 207-209,211,213,214, 233,
possessives, 43,159,204-214, 232, 238, 240, 241, 260, 265, 282
233, 238-242, 245, 256, 260, reanalysis, 37, 60, 123, 125, 126,
263, 265, 294, 295 130, 132, 135,137, 143, 161,
Pottier, Bernard, 174 163, 164, 166, 173, 183,191,
prekoine linguistic pool, 18, 23, 24, 192, 194,195, 197,199, 202,
26, 27, 29, 35, 38, 39, 43,47, 52, 203, 212, 214, 258-261, 288
55, 56, 60, 65, 76, 80, 85-87, 90, reconstruction, 22, 91,136, 137,
100, 124, 133, 136, 137, 143, 199, 202, 257, 285, 295
172, 193, 198, 231,257, 260, reduction, 3, 4, 14, 16, 19, 21-23,
274, 286, 300 25, 28, 38,41,45,46, 52, 60, 70,
prepositions, 11, 93, 113-117,119, 74, 87,91, 113, 114, 119, 123,
131, 132, 138, 143, 155, 156, 125, 126, 130, 143, 159, 161,
 Index 343

196, 206, 209, 239-241, 245, salience, 28, 29, 32-35,47, 50, 58-
246,256,277,279,286 60, 62, 65, 68, 73, 77,90, 125,
referential system, 175,176,178- 136,157,183,198, 205, 206,
180, 182, 183, 186, 190,199, 257, 260, 263, 264, 277-279,
203, 243 280
regional koine, 16, 19, 82, 84, 276 Salvador Salvador, Francisco, 296
regional standard, 16 Samarin, William J., 15, 22
regularity, 11, 25, 35,44-^7, 57, 58, Sánchez Albornoz, Claudio, 100,
71,76, 90,91, 114, 119, 127- 101, 104, 283, 284
129, 137-142, 157, 160-162, Sánchez Rodríguez, Marciano, 156
167, 172, 173, 176, 181, 188, Sanchis Calvo, María del Carmen,
190, 196, 199, 202-204, 206- 184-186,194
208, 213, 239-241, 257, 258, Sandve, B. H., 40
260, 262, 263, 267, 276, 289, Sanga, Glauco, 17
290,291,299 Sanz Fuentes, María José, 232
rekoineization, 7,28, 161, 214,238, Sarasa, Esteban, 283
256 Saussure, Ferdinand, 1,49
Repetti, Lori, 290 Schilling-Estes, Natalie, 81, 82, 279,
Ridruejo, Emilio, 95, 106, 111, 155 283
Rioja/Riojan(s), 98, 104,108, 114, Schlieben-Lange, Brigitte, 285
116, 118, 139, 143, 148, 182, Scholtmeijer, Harm, 86
184,188, 193,200, 202, 235, Schiirr, Friedrich, 288
236, 284, 287 Selinker, Larry, 38, 54, 55
Riu Riu, Manuel, 283 ser/estar, 109, 131, 285
Rivarola, José Luis, 283 Serbo-Croatian, 15
Roland/rolling distinction, 46, 65 seseo, 5, 8, 215,224-226, 229-233,
Romance, 4, 16, 87, 93, 95, 98,100, 245-256, 263, 296, 298-300
103, 107, 108, 110-112, 114, Seville, 4-7, 121, 195, 204, 207,
118-120, 122-125, 127, 129, 215-256, 264, 265,275, 291,
130, 132-138, 142, 143,147, 296, 298-300
156, 160-162, 172, 173, 177- sibilants, 4, 230, 247, 248, 252-255,
179,185,190-193,195, 196, 263, 295, 296, 299, 300
203-205, 207, 210, 258,259, Siegel, Jeff, 3, 7, 14,15-20, 23-28,
265, 266, 275,284, 285, 288, 33, 34,42-44,47,48, 51, 74, 77,
289, 291 89, 257,276, 277, 280
sigma, 232, 247, 250, 254, 300
Sabaté, Flocel, 98 Silva-Corvalán, Carmen, 87, 301
Sabin, Ángel, 154 simplification, 3, 10, 15, 16, 19, 22,
23, 25,26, 28, 29, 35, 45^18, 52,
344 Index

53, 56, 58, 61, 62, 64, 65, 68,70, stereotype, 32, 33, 50, 178, 257, 263,
73, 74, 80, 85-87, 90, 91, 93, 264, 278
109,112-114,118,119,126, Stork, F. C., 15
130, 132, 138, 155, 158, 159, strategy of neutrality, 69, 275, 282
203, 204, 207, 210, 213, 214, style, 23, 29-31, 36,43, 44,49, 69,
223, 230, 233, 238, 240, 241, 80,91, 153,154, 159, 173,204,
245, 246, 256, 258, 260, 261, 208, 210-214, 233, 238, 240,
263-265, 267, 279, 280,286 260, 263, 264, 278, 280, 284,
Slavish, 15 290
Slobin, Dan, 57, 281 substrate, 18, 21, 70, 86, 87,109,
Smith, John Charles, 274, 287 110, 123, 124, 132, 133, 243,
Sneddon, Clive, 274, 287 274, 276
social networks, 39,41, 45,48-53, Swahili, 15
56, 57, 60, 63, 64, 66, 67, 71, 76, swamping, 279
79, 80, 82, 83, 88, 90,92, 96, syncope, 161, 162,165-168,196,
100-103, 111, 117, 119,129, 240
143, 158, 171, 173,214, 238,
241, 256, 257, 262, 263, 279, Tabouret-Keller, Andrée, 39,40, 83
284, 297 Tagliavini, Carlo, 14
Spanish, American, 3, 6,47, 85, 87, target, 25, 31, 34,42, 54, 56, 67, 74,
92, 223, 224, 246, 283 77, 86, 134, 230, 245, 277, 279
speaker activity, 2, 3, 29,47-50, 91, Tarone, Elaine, 55
92, 99,112, 119, 257, 268, 280 teleology, 2, 192
Staaff, Erik, 115, 120, 129, 234, 285, Terrado Pablo, Javier, 253, 254
291,292, 295, 297 Thomason, Sarah, 25, 74-79, 86, 88,
stages, 4-6,23, 26-28, 54, 81, 100, 277, 282
101, 105, 118, 122, 129, 142, Thomson, George, 10-12
178,192, 206, 213, 222, 234, Tilander, Gunnar, 199
264, 279, 280, 282, 301 time depth, 7, 8,22,42, 89, 257
standard ideology, 105, 267 Tok Pisin, 24, 26,74
standard/standardization, 6, 10, 12, Toledo, 4, 5, 7, 97, 117, 121, 126,
13, 15-18, 20,21,23,26, 27,31, 143, 145-217, 219, 220, 233,
37, 39,41,50, 55, 60, 70, 79, 235, 236,238-243, 245, 247,
83-86, 89, 105-107, 114, 118, 260, 261-265, 283,291,292,
123, 128, 132, 176, 242, 266, 297, 301
267, 275, 276, 279, 281-283, Torreblanca, Máximo, 112, 113, 291
292, 294, 295 Torres Fontes, Juan, 283
Steinsholt, Anders, 297 transfer, 52, 55, 57, 79, 86, 87,151,
178, 189
 Index 345

transitivity, 185, 190 135, 136, 138, 143, 155, 159,

transmission, 2, 51, 67, 75, 76, 83, 161, 163-166, 168-170, 177,
113, 131 178, 195-197, 205, 206,212,
transparency, 11, 33,46,47, 57, 58, 213, 241, 258, 260, 278, 279,
65, 90,91, 114, 118, 123,124, 286-288, 291,292, 294, 295
132, 143, 155, 156, 164, 197,
213,214, 241,257, 258, 260, Wagner, Max, 223
264, 265, 267, 277, 280, 282, Wang, William S-Y., 129
286-288 weakening, 32, 33, 35,47, 52, 85,
Trask, R. L., 109, 110, 132, 133, 87, 100, 157, 171, 179, 205,206,
136, 260, 283, 288 212, 229, 241, 258, 264, 283,
Traugott, Elizabeth, 15 285, 291, 294
Trudgill, Peter, 3, 7, 27, 2 8 ^ 8 , 51, Weinreich, Uriel, 2, 55
54, 56,67, 68,71,76, 80,81, Whinnom, Keith, 285
117, 158, 159, 172, 229, 245, Wilson, Robert, 88
257, 264, 279, 296, 297, 301 Willemyns, Michael, 278
Tusell, Javier, 100, 283 Williams, Ann, 3, 7, 19, 27, 33-35,
Tuten, Donald N., 275 46, 53, 63-73, 76, 80, 86, 89, 92,
Tuttle, Edward, 290 93, 229, 257, 264, 268, 277, 279,
282
uniformity, 11-13, 22,49, 52, 80, Wolfram, Walt, 81, 82, 279, 283
81,225,245, 261,267, 276 Wright, Roger, 3, 5, 95, 106, 109,
Urdu, 88 110, 113, 120, 138, 154, 160,
161,265, 276, 284, 285, 290
Valle, José del, 3
verbs, 4, 5, 11, 37, 57, 74, 129, 131, Yaeger-Dror, Malcah, 34, 35, 70,
168, 175, 185, 189, 190, 191, 277, 278
198, 199, 244, 292-294 yeísmo, 5, 295, 296
Versteegh, Kees, 283 Yoruba, 15
Vidal Mayor, 199
Villar García, Luis Miguel, 156 Zamora Vicente, Alonso, 121, 122,
vowels, 4, 11, 12, 34, 42,44,46, 57, 126, 127, 128, 136, 154, 288,
65, 69, 93, 105, 113, 119, 122, 290, 298
124-127, 129, 130, 131, 133, Zufferey, François, 276