(Cambridge Approaches to Language Contact) Enoch Oladé Aboh - The Emergence of Hybrid Grammars_ Language Contact and Change (Cambridge Approaches to Language Contact)-Cambridge University Press (2015)

The Emergence of Hybrid Grammars
Children are extremely gifted in acquiring their native languages, but

languages nevertheless change over time. Why does this paradox exist? In
this study of creole languages, Enoch Aboh addresses this question, arguing
that language acquisition requires contact between different linguistic sub-
systems that feed into the hybrid grammars that learners develop. There is no
qualitative difference between a child learning their language in a multilingual
environment and a child raised in a monolingual environment. In both
situations, children learn to master multiple linguistic subsystems that are in
contact and may be combined to produce new variants. These new variants are
part of the inputs for subsequent learners. Contributing to the debate on
language acquisition and change, Aboh shows that language learning is
always imperfect: learners’ motivation is not to replicate the target language
faithfully but to develop a system close enough to the target that guarantees
successful communication and group membership.
enoch o. aboh is Professor of Linguistics at the University of Amsterdam.

His publications include The Morphosyntax of Complement-head Sequences
(2004). In 2012, he was awarded the renowned 1-year NIAS fellowship
and, in 2003, he obtained the prestigious Dutch Science Foundation (NWO)
5-year vidi grant to study the relation between information structure and
syntax.
Downloaded from https://www.cambridge.org/core. University of Cambridge, on 02 Feb 2020 at 21:41:16, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167
Cambridge Approaches to Language Contact
General Editor
SALIKOKO S. MUFWENE, University of Chicago
Editorial Board
robert chaudenson, Université d’Aix-en-Provence
braj kachru, University of Illinois at Urbana
raj mesthrie, University of Cape Town
lesley milroy, University of Michigan
shana poplack, University of Ottawa
michael silverstein, University of Chicago
Cambridge Approaches to Language Contact is an interdisciplinary series

bringing together work on language contact from a diverse range of research
areas. The series focuses on key topics in the study of contact between languages
or dialects, including the development of pidgins and creoles, language evolution
and change, world Englishes, code-switching and code-mixing, bilingualism
and second language acquisition, borrowing, interference, and convergence
phenomena.
Published titles
Salikoko Mufwene, The Ecology of Language Evolution
Michael Clyne, Dynamics of Language Contact
Bernd Heine and Tania Kuteva, Language Contact and Grammatical Change
Edgar W. Schneider, Postcolonial English
Virginia Yip and Stephen Matthews, The Bilingual Child
Bernd Heine and Derek Nurse (eds.), A Linguistic Geography of Africa
J. Clancy Clements, The Linguistic Legacy of Spanish and Portuguese
Umberto Ansaldo, Contact Languages
Jan Blommaert, The Sociolinguistics of Globalization
Carmen Silva-Corvalán, Bilingual Language Acquisition
Emanuel J. Drechsel, Language Contact in the Early Colonial Pacific
Lotfi Sayahi, Diglossia and Language Contact
Further titles planned for the series

Rakesh Bhatt, Language Contact and Diaspora
Gregory D. S. Anderson, Language Extinction
Bridget Drinka, Language Contact in Europe
Fredric Field, Spanish and English in the United States
Bao Zhiming, The Making of Vernacular Singapore English
Ralph Ludwig, Steve Pagel, and Peter Mühlhäusler, Linguistic Ecology and Language
Contact
Kingsley Bolton, Samuli Kaislaniemi, and Anna Winterbottom, Language Contact and
the East India Company
The Emergence of Hybrid
Grammars
Language Contact and Change
Enoch Oladé Aboh
University Printing House, Cambridge CB2 8BS, United Kingdom
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9780521769983
© Enoch Oladé Aboh 2015
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2015
A catalog record for this publication is available from the British Library
Library of Congress Cataloging in Publication data
Aboh, Enoch Oladé, author.
The emergence of hybrid grammars : language contact and
change / Enoch Oladé Aboh.
pages cm. – (Cambridge approaches to language contact)
ISBN 978-0-521-76998-3 (hardback)
1. Interlanguage (Language learning) 2. Sublanguage. 3. Native
language. 4. Language acquisition – Age factors. I. Title.
P118.23.A36 2015
4170 .22–dc23
2014047368
ISBN 978-0-521-76998-3 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents
List of figures page vi

List of maps vii
Foreword by Salikoko S. Mufwene ix
Acknowledgments xv
List of abbreviations xvii
1 Introduction 1
2 The agents of creole formation: geopolitics and cultural aspects
of the Slave Coast 16
3 The emergence of creoles: a review of some current hypotheses 60
4 Competition and selection 113
5 The role of vulnerable interfaces in language change: the case
of the D-system 171
6 The emergence of the clause left periphery 222
7 The emergence of serial verb constructions 269
8 Conclusions: some final remarks on hybrid grammars, the creole
prototype, and language acquisition and change 304
References 317
Author index 337
Language index 340
Subject index 343
Downloaded from https://www.cambridge.org/core. University of Texas Libraries, on 09 Jan 2020 at 13:07:19, subject to the Cambridge Core
Figures
2.1 Portrait of Dom Matheolopes, Ambassador Roy d’Arda.

Prints on the History of France (Volume 50), Parts 4532–4594
(1670–1671). Source: gallica.bnf.fr / Bibliothèque nationale de
France page 25
2.2 Linguistic distribution in the French Caribbean in the
seventeenth century (adapted from Singler 1996: 205) 54
2.3 Ethnic groupings in Saint-Domingue 1757–1797 (adapted
from Debien 1974) 56
2.4 Contrasting the Congo and the Kwa in eighteenth-century
Saint-Domingue (adapted from Debien 1974) 57
vi
Maps
2.1 The Slave Coast and its major ports page 17

2.2 Map of migrations within and around the Allada Kingdom 21
2.3 Approximate area of control of the Allada Kingdom 22
vii
Foreword
salikoko s. mufwene
The present book represents the outcome of research that Enoch Oladé Aboh has
conducted on structures of especially his mother tongue Gungbe and on Haitian,
Sranan, and Saramaccan, over the past fifteen years, counting from the time
when I first met him at a conference on African linguistics at the University of
Cape Town, in January 2000. He has since then contributed, in paper after paper,
an enlightening comparative perspective that sheds informative light on how far
similarities obtain between the Gbe languages and these three creoles. This
subject matter has become central in the literature on the emergence of creoles
thanks to some relexificationists who claim that grammars of creoles are
patterned on those of their most important substrate languages.
In the case of these specific creoles, the field has been blessed to see some of
the relevant issues addressed by a linguist who is a native speaker of one of the
Gbe languages, who is impeccably well trained in the generative framework
used by prominent relexificationists such as Claire Lefebvre and John
Lumsden, and who has relentlessly investigated aspects of the grammars of
the relevant creoles based on not only the literature but also work with their
native speakers, as well as collaboration with other experts on the relevant
languages. So, this book brings to its readers the fruits of the application of the
comparative approach (not to be confused with the comparative method of
historical linguistics) to the letter. The reader will find in the following pages
plenty of data that raise issues about the Relexification Hypothesis applied to
Haitian, Sranan, and Saramaccan.
Contrary to claims that creoles form a typological class of their own, Aboh also
shows that the grammatical structures of these particular creoles are not identical,
not even in some respects that can be associated with their apparently most
conspicuous Gbe substrates. Some of the structures that distinguish them from
their lexifiers are not shared identically in the Gbe group either. Although this
does not mean that we should deny substrate influence on creoles’ structures, we
nevertheless should not continue thinking of structural similarities between
creoles and their dominant susbstrates in the traditional way, viz., that
particular meanings or structures from a particular language or group thereof
were transferred intact into the system of the emergent language (variety).
ix
x Foreword
The same comparative approach, once promoted earlier in the 1990s by

Robert Chaudenson but executed in detail by Aboh, also shows that, while we
cannot deny the legacy of the lexifiers on their metamorphic offspring (so to
speak), it is important to pay attention not only to what has been retained but
also to how those particular retentions have been integrated in the overall
emergent grammar. How has the initial koinéized system been restructured?
Does this prove exceptional evolution on the part of creoles? Arguing like
Michel DeGraff and myself against “creole exceptionalism,” Aboh’s answer is
a resounding “no.”
His answer is also negative to the question of whether there is some hope of
salvaging any version of Derek Bickerton’s Language Bioprogram Hypothesis,
which presupposes pidgin ancestors for the emergence of creoles and claims
that the latter were created by children. A most innovative part of Aboh’s
arguments, developed in Chapter 2, lies in showing that the slave trade on the
western coast of Africa, especially in the Kingdom of Allada, in today’s Benin,
did not proceed as traditionally assumed in the literature. No incipient pidgin is
reported to have emerged out of the initial contacts between Europeans and
African traders. Instead, the trade was organized like today’s international, if
not so globalized, partnerships, involving powerful companies and institutions,
with brokers/intermediaries conducting business for them in languages that
they had some acceptable command of.
The study is very consistent with Mufwene (2014), which argues that the
history of trade between Europeans and non-Europeans from the fifteenth
century to the nineteenth century suggests that this relied heavily on
interpreters, also identified as intermediaries and linguists. Genetic creolists
should not have ignored them and should have investigated how they
developed. The satisficing answer to date lies in the fact that, like other
important traders in Asia and the Pacific, the African kings and chiefs sent
princes and children of their entourages to Europe, to learn the language of their
prospective trade partners, while the European explorers left some members of
their crews behind as gages of good faith. On both sides, the concerned
individuals learned the languages of their hosts by immersion; those who
survived these exchanges would serve as interpreters when trade started in
earnest years after the first encounters. Immersed in European families in the
host countries, the Africans had no reason to develop pidgins, no more than did
the European crew members who were left behind in Africa or the lançados and
future factors who settled there by choice. There were no open slave bazaars
where just anybody could bring captives or disavowed relatives to sell to
Europeans. Nor were the slave markets as business institutions like open
fields where just any European can walk with his gun and money and
purchase slaves. Rather, this ignominious trade was well organized and
constrained in its patterns of interactions, contrary to the traditional
Foreword xi
assumption that sporadic encounters between European traders and societally

multilingual Africans unavoidably triggered the emergence of pidgins. It is in
fact noteworthy that the merchants of (Asian) bazaars, the antecedents of
modern shopping malls, have typically been polyglots, ready to speak the
buyer’s language. The contact history summarized in Mufwene (2014) shows
that the European trade and exploitation colonization of Africa, Asia, and the
Pacific relied heavily on interpreters, precluding extensive direct interactions
between the Natives and Europeans.
As a matter of fact, European trade on the African and Asian coasts relied
heavily on Portuguese as the lingua franca till the late eighteenth century, if not the
early nineteenth century. Pidgins appear to have emerged later than traditionally
assumed, perhaps later than creoles of the Caribbean and Indian Ocean. I
conjecture that their emergence, so late indeed, was the consequence of trade
expansion, which then produced a shortage of good interpreters. The
pidginization process must thus have been similar to the basilectalization
hypothesized by Robert Chaudenson for the emergence of creoles, with the
non-native productions of the lexifier becoming more and more divergent from
the closer approximations produced earlier by those who had learned the language
under contact conditions more favorable to “faithful” learning. Thus, in addition
to all the sound arguments based on structural features of Haitian, Sranan, and
Saramaccan marshalled by Aboh against the Language Bioprogram Hypothesis,
one cannot overlook the fact that the history of the slave trade does not support it.
For those who still thought that Ingo Plag’s Interlanguage Hypothesis, which
suggests that creoles reflect convergent interlanguages in arrested stage, is the
alternative to the Language Bioprogram Hypothesis, Aboh also adduces a
variety of structural evidence against it. He likewise shows how misinformed
it is to claim, like John McWhorter, that creoles are among the world’s simplest
languages. Both fail to realize that the plantation settlement colonies left no
room for the break in the transmission of the lexifiers traditionally associated
with the ecology of the emergence of creoles. Once race segregation (not to be
confused with race-based discrimination) was instituted during the transition
from the homestead phase to the plantation phase, there were already cohorts of
Black Creoles who spoke the same colonial koinés as the White Creoles.
Language transmission in these colonies must therefore be approached in
terms of who spoke which variety and when, which underscores the
significance of population structure and periodization of the peopling of the
colonies, rather than with the simplistic equation “Race Segregation = Break in
Transmission.” Otherwise, it would be like assuming, inaccurately, that English
is spreading around the world today through the dispersal of primarily the Brits
and Americans, whereas reality shows that a very large proportion of English
speakers today have learned it from non-native speakers outside the United
Kingdom, the Anglophone North America, Australia, and New Zealand.
xii Foreword
Aboh likewise mines the alleged simplicity of creoles not only by providing
counterevidence but also by asking whether a linguistic system should be
reduced to its morphosyntax, detached of its semantics. Is it accurate to
reduce morphosyntactic complexity/simplicity to what Michel DeGraff has
identified as “bit complexity,” which amounts to how many units (phonemic,
morphemic, and syntactic rules) a linguistic system is populated with? How
about the constraints that govern the usage of the relevant units and the ways in
which rules of their use interact with each other, either in feeding or bleeding
some of them? How about the impact of an overall simplicity of a language on
its expected capacity to convey information successfully, allowing easy
discrimination between different contents?
A great number of John McWhorter’s claims has also depended on his
assumption that creoles are young languages. Assuming that there is genetic
discontinuity between the lexifier and the relevant creoles, is complexity really
a function of how old a language is? Did modern “young languages” start from
scratch in the same ways that the hominine species evolved from having no
linguistic communication to producing languages, going through an embryonic
protolinguistic stage that involved short lists of denoting terms and minimal
grammar? Did the minds of the inventors of modern language varieties such as
creoles or even pidgins regress to the state of the mind of Homo erectus or early
Homo sapiens? There is no evidence in these language varieties or the
ecologies of their emergence suggesting that they may illustrate the
disputable position that “ontogeny recapitulates phylogeny.”
As a matter of fact, one should also ask whether, for instance, Gullah is
younger than other American English varieties and whether it is justified to
deny its ultimate ancestry in Old English, or perhaps the ancestors thereof.
From a genetic linguistics point of view, wouldn’t a more adequate account of
the differential evolution of English be one that does not too hastily invoke
overall simplification for some varieties simply because they have been
disfranchised as creoles for reasons that appear to be fundamentally social?
Besides, the morphological simplification of English does not appear to be an
exclusive peculiarity of the formation of creoles; it appears to have been the
general evolutionary trajectory of the language since the transition to Middle
English. Even if the producers of creoles have overextended the simplification
of the lexifier’s morphology, haven’t they also complexified some other aspects
of the grammars of the emergent language varieties? In the final analysis, don’t
things boil down to the typological realignment of some parameters of the
creoles’ grammars compared to their lexifier?
It is in the context of the above considerations that Aboh submits the
hypothesis of hybrid grammars. This notion is so much the more interesting
as it addresses the emergence of new language varieties from both the
ontogenetic and the communal perspectives, focusing respectively on
Foreword xiii
language development in individual speakers and on the convergence of

emergent idiolects into new sociolects, dialects, or varieties such as creoles
that have been stipulated by linguists to be separate languages altogether. Aboh
presents data that question both the uniparental conception of language
“transmission” in genetic linguistics and the idea that native speakers of a
particular language “acquire” it intact from the population of speakers they
learn it from. Idiolects are as hybrid in kind as creoles have been claimed to be
because they integrate selectively elements from different sources. An idiolect
is as different from those of the model speakers as a creole is from the
languages out of whose contact it emerged. As I see it, both are shaped
gradually, constantly influenced by the increasing and sometimes competing
inputs and adapting themselves to communicative pressures. The main
differences lie in the number of generations, the size of the population (one
versus many speakers from the point of view of the learners), and of course in
the richness/diversity of variants in the feature pool, where the competition and
selection that shape the new varieties arise and take place, respectively.
Overlooking differences in learning skills, the nature of the feature pool
(consisting of variants from the target language and the other languages that it
may be in contact with) plays a critical role in determining the outcome of
language “transmission,” bearing in mind that the pool is naturally constrained
by the relevant population structure and by the particular period in which one
learns a variety. One must always remember that, as economic ventures, the
plantations were not produced overnight; they typically developed
incrementally, with their populations increasing depending on when need
arose and/or when their owners could afford them. Thus, the presence and/or
numerical strength of particular ethnolinguistic groups varied according to
what the market provided at specific times, notwithstanding the speed of
population replacement, especially at the time when the slave population
grew more by importation than by birth. To make matters more complex,
these ecological factors may have varied from one plantation to another,
though there were general trends in particular colonies, on which Aboh, like
other creolists working on the subject matter, capitalizes to shed light on the
grammatical hybridity of the emergent creoles.
Aboh takes us a step further than the feature-recombination that I invoked in
the work that he cites, thus making the Complementary Hypothesis, as I
conceived of it, more explicit. Like Richard Lewontin in 1970, in the context
of biological evolution, he raises implicitly the issue of what the units of
selection are in both language “acquisition” and the restructuring processes
that produced creoles as communal varieties. We are reminded that linguistic
systems consist of forms, structures, rules, and meanings and that each one of
these kinds of elements can be complex. Shedding light on the often-invoked
notion of congruence as a factor favoring some of the competing variants, Aboh
xiv Foreword
shows that even a rule from a particular source language can be selected
partially; substrate grammatical patterns were not necessarily faithfully
reproduced in the emergent creoles and certainly not in identical ways from
one creole to another. This is as obvious with serial verb constructions as with
determiners and predicate fronting. Substratists should especially remember
that even the semantics was not faithfully reproduced, aside from the fact that
internal dynamics of the emergent systems generated some peculiarities that
can be traced to neither the lexifier nor the particular substrate languages
singled out in the present study.
This book reveals how complex the subject matter of the emergence of
creoles is and how germane it is to that of the formation of idiolects and
communal language creation in general. It is a demonstration of the useful
contribution that genetic creolistics can make to the study of language
acquisition and change, as well as to theoretical linguistics. Transcending the
traditionally narrow confines of discussions in creolistics, the book is both a
strong argument against creole exceptionalism and an invitation to keep away
from simplistic accounts. For those who have been complaining about lack of
data that support the contention that creoles have evolved like other natural
languages (bearing in mind that every case of language evolution is local,
constrained by ecological factors specific to the setting and time of its
emergence), there is no scarcity of data in this book. One is instead struck by
the modesty of the author and the soft tone of his claims. As I said in Mufwene
(2001), creolistics should not be a consumer discipline content with borrowing
ideas from other areas of linguistics (and other disciplines); it should also
contribute to understanding issues relevant to linguistics at large. Being more
detailed in the analyses of a wide range of constructions than is typical in
creolistics, The emergence of hybrid grammars has passed another milestone
toward that goal. Non-creolists that are typologically inclined will find it
compelling, while theoretically inclined creolists will appreciate the ways in
which the author makes the data relevant to formal syntax and historical
linguistics. I am so proud that Aboh chose Cambridge Approaches to
Language Contact to defend his positions and flattered that he invited me to
write a foreword for it.
s a l i k o k o s . mu f w e n e , University of Chicago
Acknowledgments
I became familiar with some of the literature in creolistics in the mid 1990s
when I was writing my thesis. Back then, part of the generative literature on the
Gbe (and more generally Kwa) languages stemmed from work by Claire
Lefebvre and her team at UQAM. It also happened that most of this study
was meant to show substrate influence in creole genesis: the role of Gbe (Kwa)
languages in the emergence of the Caribbean creoles. While some of the
features or patterns identified in the creoles as Gbe/Kwa substrate influence
looked reasonable to me (e.g., predicate cleft), some others (e.g., the
directionality of certain functional items) appeared doubtful. In addition,
even the features that could be ascribed to the substrate languages never
seemed replicated identically in the creole. This observation led me to think
that the creoles I was reading about then might be telling us something more
fundamental about language acquisition in general: how learners weave
together different aspects of the heterogeneous inputs they are exposed to.
The chapters in this book present some of my thinking on this question over the
past decade.
Since the mid 1990s, I’ve had opportunities to talk to many people who
helped me think harder and reformulate the conclusions presented in this book.
I probably don’t recall everyone who took part in what turned out to be a
fantastic journey for me, but none of this could have been possible without the
friendship and support of Norval Smith, Pieter Muysken, Michel DeGraff, and
Salikoko Mufwene.
My thoughts actually started taking form when in 2001 Norval and Pieter
offered me a postdoc position in their NWO project: The TransAtlantic
Sprachbund. During the following ten years or so, I’ve have daily
discussions with Norval both in our offices over a coffee and on the train on
our way back home. Needless to say I learned a lot during these discussions, but
most importantly, I realized that history matters. Thanks, Norval. And thank
you, Pieter, for being always supportive and for always pointing me in the right
direction. My gratitude also goes to other members of the TransAtlantic
research team: Felix Ameka, Adrienne Bruyn, James Essegbey, and Margot
van den Berg.
xv
xvi Acknowledgments
It was also during this time that I met Michel DeGraff and Salikoko
Mufwene. I’m immensely indebted to both of them for their friendship and
support. Michel and Sali commented on almost all my writings related to
creolistics. Their numerous constructive comments, suggestions, and positive
criticisms throughout these years helped me refine the theory developed in this
book, broaden my research, and think harder about the big picture. I owe a great
deal of the literature cited in this work to them, and both gave me several
opportunities to present my work to broader audiences: Michel by inviting me
to teach a semester at MIT in spring 2008 and Sali by giving me the chance to
publish this book in this series. Sali’s role as a friend and editor of the
Cambridge Approaches to Language Contact series cannot be overestimated.
He has been supportive of this project from the very early phases. Despite his
busy schedule, he has always managed to find the time to critically evaluate the
manuscripts I sent him, starting from the book proposal up to the final product
now in the hands of the reader. It is not an overstatement to say that this book
could not have existed without his contribution. Kouzen and Grand-Frère, I
alone assume full responsibility for all remaining shortcomings and hope that
the end result does not let you down: ùn dó kpέ ná mì.
I’m immensely grateful to my sister Frieda Sessi Aboh for her numerous and
very constructive comments on earlier versions of Chapter 2. My gratitude also
goes to Umberto Ansaldo and Anne Zribi-Hertz for their comments and
suggestions on various aspects of the work presented here and for being very
supportive to me throughout these years. I feel equally indebted to my students,
my colleagues of the Amsterdam Center for Language and Communication,
and particularly to Kees Hengeveld for his friendship, for always being
supportive, and for making it possible for me to carry out my research in the
best of working conditions. I also thank Ekaterina Bobyleva for the numerous
constructive discussions we’ve had while she was writing her thesis under my
supervision, and Roland Pfau for the past gezellig moments and those to come.
Parts of Chapters 4, 5, and 6 were previously published as self-contained
essays, but much of this book took form in 2011–2012 when I was offered a
Netherlands Institute for Advanced Study in the Humanities and Social
Sciences (NIAS) fellowship for a year. I’m grateful to NIAS for believing in
this project and for giving me the chance to complete it. I’m particularly
indebted to my colleague Aafke Hulk, then the rector of NIAS. It is clear to
my mind that I could never have written this book if I had not been given the
opportunity to spend a whole year at NIAS, writing, meeting with people of
different backgrounds in the humanities, and contemplating the wonders of
nature in Wassenaar. Aafke, thanks for giving me this opportunity.
Finally, I owe the greatest debt to Anne, my wife, our son Fèmi and
daughters Aniola and Orê for always being there for me. Thanks for your
love. This book is dedicated to you.
Abbreviations
a:dcl affirmative declarative

abl ablative
abs absolutive
adv adverb/adverbializer
cl clitique
cmp comparative
co cooperative object
dcl declarative
def definite
dem demonstrative
det determiner
dim diminutive
dir direction
ds different subject
emph emphatic
f feminine
foc focus
fut future
gen genitive
hab habitual
imp imperative
inc incompletive
indef indefinite
inf infinitive
instr instrument
inter interrogative
intl intentional
io indirect object
loc locative
m masculine
mid middle voice
mom momentaneous
xvii
Downloaded from https://www.cambridge.org/core. The University of Iowa, on 18 Jan 2020 at 13:04:36, subject to the Cambridge Core terms
of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167
xviii List of abbreviations
mood modality marker

neg negation
num number
nvr non-veridical
perm permissive
pcl particle
pf perfective aspect
pl plural
pol polite form
pos possessive
prn pronoun
pst past
rel relative
s subject
sbj subjective
sg singular
sg.f singular feminine
sg.m singular masculine
top topic
wh question word
Downloaded from https://www.cambridge.org/core. The University of Iowa, on 18 Jan 2020 at 13:04:36, subject to the Cambridge Core terms
of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167
1 Introduction
An obvious empirical fact about languages is that they evolve constantly in

core domains of their grammars. This evolution (or change), though not
obvious in the present, becomes evident when a language is investigated
diachronically. For instance, students of Modern English know about the
Great Vowel Shift (Otto Jespersen 1860–1943), which distinguishes
phonological and phonetic properties of Middle English from those of
Modern English. Similarly it is not uncommon that words start with a precise
semantics but acquire new meanings in the course of time, sometimes losing
their original semantics. A simple illustration is that of Waterloo, the name of a
village in present-day Belgium where “the allied pursuit caused Napoleon’s
army to disintegrate entirely,” and which is commonly used in current English
to mean “a decisive defeat or failure” (The New Oxford Dictionary of English:
2001: 2087). With regard to syntax, Old English displayed V-final patterns
comparable to those of Modern German as illustrated in (1), cited from Kroch,
Taylor, and Ringe (1997, their example 3a):
(1) ðeah hit ær upahæfen wære
although it before up-raised was
While Middle English had lost the V-final pattern, it maintained V2
properties similar to those encountered in modern Dutch (Kroch 1994: 2):
(2) Quene Ester looked never with swich an eye.
As can be seen from this example, the verb is placed to the left of the negative
adverbial never. Such verb placement is impossible in Modern English as can
be illustrated by the pair of sentences in (3), where the sequencing in (2) is
excluded (3b), in contrast with the construction where negation precedes the
verb (3a):
(3) a. Queen Esther never looked with such an eye
b. *Queen Esther looked never with such an eye
What we see here is that over the centuries, English has changed in various
modules of its grammar (e.g., phonology, semantics, syntax). For instance, the
Downloaded from https:/www.cambridge.org/core. Columbia University Libraries, on 07 Jul 2017 at 03:17:25, subject to the Cambridge
Core terms of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167.002
2 Introduction
syntax has gradually moved away from a proto-typical West Germanic

language with robust V-final and V2 properties, to an SVO language in
which the lexical verb necessarily occurs to the right of negative adverbs
such as never, with the object immediately to its right (e.g., van Kemenade
1987; Roberts 1993, 2007; Haeberli 1999; Kroch 1989a, 1989b, 2001; Kroch
and Taylor 1997; Pintzuk 1999; Lightfoot 2006). The only legacy of English
past history lies in direct questions, where the verb or the auxiliary verb
necessarily occurs in the second position as is in V2 languages.
(4) a. What will John buy?
b. Wat zal Jan kopen? [Dutch]
Another domain of English syntax that has changed is the use of determiners
or articles. Present-day English has a complex set of rules regarding usage and
omission of articles in ways that contrast with Old English, in which bare nouns
could be used in syntactic contexts not acceptable today. About the
development of articles in English, Gardnier (1932: 47) cited in McColl
Millar (2000: 300) notes:
It is sometimes said that such relatively insignificant words [such as the articles] are
grammatical tools. But the function of tools is to achieve some specific end. That is
precisely what, in many cases, the article does not do, or at all events does only in a very
slight and uncertain degree. Often it is a mere useless ballast, a habit or mannerism
accepted by an entire speaking community . . . The accumulation of old rubbish is so
easy.
What Gardnier refers to as “old rubbish” is now part of the mental grammar that
every native speaker of English develops.
One notices the same drastic change in the Romance languages as well.
While Latin lacks articles and complementizers, most modern Romance
languages include such grammatical items. Comparable lists of notable
changes could be drawn for every single human language. Speakers often
notice linguistic changes in the lexicon (including borrowings from other
languages); sometimes they resist them (often in vain). However, language
users are typically less aware of more subtle changes involving grammatical
notions, such as, the loss of V2 in English or its correlated change from OV to
VO order. These are far more obscure changes which, once they are created,
may take a very long time before they spread within the population of speakers
and become a stable feature of the language that is passed on to later
generations of speakers. While the English example above indicates that
certain syntactic changes may sometimes take several centuries before they
spread through the community or population of speakers, it is obvious that it
does not take speakers centuries to create these changes. Building on
Chomsky’s (1986) notions of I(nternal)-language and E(xternal)-language,
Introduction 3
DeGraff (1999: 9) characterizes the time-lapse between the birth or “creation”

of a change (i.e., the development of an I-language) and the moment when this
change becomes noticeable on the population level (i.e., the E-language) by
distinguishing between two levels of analysis when it comes to language
change.1 As DeGraff puts it:
We need to draw a sharp conceptual line between, on the one hand, the diachronic
accretion of the “elements of a creole’s grammar” and the dating of a creole in terms of
establishment of a new community language with such elements (i.e., an E-creole) and,
on the other hand, the genesis of a creole as the emergence of an I-language of a
particular sort (i.e., the development in individual speakers’ minds/brains of a grammar
that shows a certain typological distance from the grammars of the languages in contact –
an I-creole).
Applied to language change in general, this would mean that studies of

language evolution must address the topic of language change on two
independent, though related, levels: (i) the population level where one can
observe how a particular new feature is being used in a community (see also
Mufwene 1986, 1994, 2001, 2002, 2003, 2005a, 2005b, 2008, 2009; Croft
2000), and (ii) the individual level where we may try to observe how a new
form is created by the speaker (e.g., Aboh 2006a, 2009a).
This book is about language change at the individual level (as characterized
by DeGraff (1999). Within the context of the Minimalism framework
(e.g., Chomsky 1995, 2001, 2008), which I’m adopting here, many questions
arise including the following: How does change come about? What are the
ingredients of change? How can we access such ingredients? These questions
drive the discussion in this book.
It has traditionally been assumed that contact is a major factor in language
change. In the literature on language change, various proposals have been made
to explain how grammatical changes arise gradually in a contact situation.
A common hypothesis is that such changes evolve from imperfect second
language acquisition (SLA) by adults. In the case of English, for instance,
Kroch, Taylor, and Ringe (2000) argue that Northern Middle English and
Southern Middle English display distinct V2 properties. In the Northern
varieties, the verb moves to the complementizer domain (CP), making these
varieties CP-V2 languages, similarly to Mainland Scandinavian, German, or
Dutch. In the Southern varieties, however, the verb moves to the inflectional
domain (INFL), as it does in Yiddish and Icelandic. These variants therefore
display the IP-V2 phenomenon. The authors further indicate that the emergence
1
According to Chomsky (1986), an Internal-language, which is the object of linguistic study, is the
mental representation of the linguistic knowledge of a speaker. Thus, I-language is a mental
object, which contrasts with External language roughly characterized as the knowledge of
language use in a community, which includes performance.
4 Introduction
of CP-V2 in the Northern dialects is contingent to the loss of verb movement to

INFL to support verbal inflectional morphology. The latter is a consequence of
the collapse of the verbal inflectional morphology, itself resulting from
imperfect SLA by the Vikings who invaded Northern and Eastern England.
Given that the Northern and Southern varieties of Middle English had divergent
V2 properties (viz, CP-V2 versus IP-V2), the authors hypothesize that the
eventual loss of V2 in Modern English, which made it an SVO Germanic
language, could have resulted from extensive contact between speakers of the
Northern and Southern varieties. This language contact situation would have
led to competition between the two variants, which eventually led to the loss of
the V2 linguistic feature in the changing language. Under this scenario
therefore, the main agents of variation within English dialects, and the
resulting change from OV to VO, were L2 learners.
According to this scenario, language change is primarily a consequence of
imperfect SLA (or imperfect replication): new features emerge in a language
because adult L2 learners fail to replicate the features of the target faithfully
or inject features of their native language in the target (L1 transfer). Though
this view is compatible with various sociolinguistic aspects of language
change (e.g., code-switching, borrowing of vocabulary items, or calques), it
leaves unanswered the fundamental question that this book tries to answer:
Which properties of grammar allow the combination of syntactic features
into a new system? In the case at hand, how can we account for the emergence
of CP-V2 in Northern Middle English, attributed to the linguistic influence
of Scandinavian invaders, in a principled manner? As is well known in
Germanic syntax, V2 phenomena correlate with verbal morphology, verb
movement as well as very subtle syntactic operations that relate to general
principles of UG (e.g., economy, head movement, probe–goal relation,
subject–auxiliary inversion) that go beyond borrowing of a lexical category
or a simple misanalysis of a grammatical category. If we make the reasonable
hypothesis that the same general UG principles are at work in the syntax of
Middle English, where they may have different effects from V2, we may want
to ask how Scandinavian CP-V2 syntax combined with the syntax of the verb
in Middle English to produce the Northern Middle English variant with CP-
V2 (instead of, say, a completely unspeakable or un-learnable language).
Indeed, if languages (as we know them) are extrapolations from idiolects
(which in turn are expressions of speakers’ I-language), what principles
govern the combination of the linguistic features from different languages
into a speaker’s I-language? Assuming such a combination is possible (as
suggested by the data) then the newly created I-language presumably
involves a ‘hybrid’ system.
The term hybrid has been used in various social constructs with a pejorative
meaning. I use this term here in a strictly neutral sense to refer to a stable
Introduction 5
linguistic system that emerges from the contact of (typologically and/or

genetically) different linguistic varieties. As I argue throughout the book,
natural languages involve hybrid systems as a rule because every I-language
derives from the mix of features that are expressed in the Primary Linguistic
Data (PLD), the latter being fed by expressions of mutually distinct I-languages
(cf. DeGraff 1999: chapter 1; Mufwene 2001). In this regard, linguistic
hybridization as argued for here is fundamentally different from Whinom’s
(1971) views on hybridization and how it could apply to pidgins, creoles, and
languages in general.
If languages are indeed the collective expressions of individual hybrid
I-languages, how do such hybrid I-languages evolve into stable systems
such that some features, replicated by other speakers, spread across a
community to eventually become part of a stable E-language identified as a
new language?
These questions are rarely directly addressed in the literature. In an effort to
answer them, and contribute to the debate on language evolution in general, this
book takes a closer look at some linguistic features in grammars and examines
how, in a situation of contact, syntactic and semantic features of different
language types may recombine into a new form as part of a new emergent
language. In the context of this book, the term contact is taken to mean the
coexistence and competition between linguistic systems (viz., languages,
dialects, or idiolects) in the mind of the same speaker. The contact is thus
between two (or more) different lexica and between typologically different
linguistic systems, though the differences may vary from minor to very
significant ones. Under this view therefore, both L1 acquisition and L2
acquisition involve language contact, though the two processes differ
qualitatively. Issues of age aside, one such difference is that L1 acquisition
necessarily involves contact of idiolects and/or dialects, both being related
variants of the same language. In this case we are dealing with variants that are
genetically and typologically related. Simultaneous bilingualism (2L1) and L2
acquisition, on the other hand, must involve (dialects or varieties of) two
distinct languages that may not be related genetically or typologically. These
differences notwithstanding, this study assumes that any learner, including
L1 learners and L2 learners, finds himself/herself in a situation of linguistic
contact, often having to select from among competing variants in his/her
feature pool (Mufwene 2001, 2005a, 2008). In L1 contexts, the learner
develops a system from competing inputs and produces a new grammatical
system, though not completely different from the input systems. In 2L1 and
L2 contexts, however, the learner operates on different languages that may
influence each other, thus creating a new linguistic system that may be
significantly different from the source languages. This view explains why
learners systematically develop hybrid mental grammars.
6 Introduction
Once we adopt this perspective, it becomes clear that the study of the
creation of change informs us about the evolution of language in general but
it also bears on issues of language acquisition. In understanding how learners
(whether L1/2L1 or L2) create new idiolectal varieties based on their individual
linguistic experiences, we hope to understand how learners access abstract
linguistic properties during the acquisition process. A crucial assumption that
this book builds on is that new varieties do not generally emerge as the result of
misanalysis (or imperfect replication), as is often assumed in theories relying
on imperfect SLA, but from ‘a certain analysis of abstract features of linguistic
entities’ (see, for instance, Klein and Perdue 1997). Though this linguistic
analysis may be deviant from the one generally accepted in the community or
the one a school teacher may expect learners to make in order to acquire
command of the target language (i.e., the normative form), it remains one of
the possible alternatives the learner has access to, based on UG and linguistic
experience. This in turn implies that in order to understand the driving forces of
language change we have to try to probe into the speaker’s knowledge at the
moment of the creation. That is, we have to probe into I-language. Obviously,
such endeavor can only be indirect, and in most cases speculative because we
often do not know when the change came into existence and diachronic studies
only inform us about the distances between two communal systems (which are
collective abstractions), for instance, between Middle English and Modern
English. While the enterprise may look completely hopeless, creole
languages might offer us an extraordinary opportunity to investigate the
creation of change, largely because of the shallow history and because we
have more information about the communal systems in contact and about the
ethnographies of the contacts than in traditional historical linguistics.
1.1 Creoles as a test-bed

In order to study the production of change, this book focuses on the genesis
of the creoles of Suriname and Haiti. The choice of these creoles as case studies
is not guided by the supposedly exceptional nature of creole languages (e.g.,
Bickerton 1981, 1984; Lefebvre 1998; McWhorter 2001, 2011; Bakker et al.
2011), but by the following factors:
(i) Creoles developed recently (within about four centuries) and have not
lived long enough for their original ‘ingredients’ to fade out due to a
long history of linguistic change.
(ii) Creoles result from the extensive contact between languages that are not
genetically related (and exhibit typological differences in certain domains
of their grammars; e.g., Romance/Germanic vs. Niger-Congo). Thus,
creole languages differ from contact languages which developed from
1.1 Creoles as a test-bed 7
the contact between languages that are genetically and typologically

related (Indo-European languages).
In the literature on contact languages and language change, creoles have been
considered by many to be extreme cases of SLA, allowing extensive influence
from the substrate languages spoken by the enslaved Africans. The general
assumption has been that the Africans whose descendants became native
speakers of creoles had very limited access to the target language spoken by
the European colonial power, typically identified in English as the lexifier. In
Lefebvre’s (1998) Relexification Hypothesis, for instance, creoles are assumed
to be radically divergent L2 varieties involving massive L1 transfer. Therefore,
creoles combine syntactic and semantic properties of the substrate languages
with PF properties of the target language: Haitian Creole = Fongbe (Kwa)
grammar + French PF.
On the other hand, Bickerton (1981, 1984, 2008) claims that radical SLA
failure, coupled with break in transmission, led to a situation where adult
learners developed a macaronic pidgin that is inadequate for L1 acquisition.
Bickerton contends that in such a situation of language degeneration, children
as L1 learners are faced with an inadequate input. Accordingly, they rely on
their innate language capacity, the Language Bioprogram, which allows them
to create a full-fledged language out of the degenerate inputs provided by their
parents or other adult speakers of the local pidgin. According to him, the
Language Bioprogram explains the fundamental similarities observed across
creoles, which appears to have led McWhorter (1998, 2001, 2011) to posit
creole prototypes, whose structures are putatively the closest to that produced
by the bioprogram.
Under the same premises of SLA failure, theories that reject both the
Relexification and the Language Bioprogram hypotheses but focus on the
role of the superstratum assume that creole languages are new varieties of
European languages spoken in the colonies. For instance, Chaudenson (2003:
448) suggests that French creoles result from “the unguided appropriation of
approximate varieties of French koiné” (see also his earlier work since 1979).
According to him, creole languages emerged as a consequence of imperfectly
acquired koinés of European languages spoken by the European colonists, with
incremental divergences from the original colonial variety, hence the idea of
‘approximations of approximations.’ This theory presupposes that even though
creoles may differ from the target European varieties in various respects, they
share basic and fundamental morphosyntactic features (identified by
Chaudenson as matériaux de construction) with the non-standard dialects of
these languages.
More recently, Plag (2008a, 2008b) has proposed a theory of creolization
(and therefore language change) that interprets creoles as ‘conventionalized
8 Introduction
interlanguages of an early stage.’ What this means is that creoles are instances
of interlanguages that got frozen before their inventors had the chance to reach
a more advanced stage in acquiring the target language. Here again, SLA
failure is assumed to be one of the driving engines of creolization.
1.2 Speakers create variants not imperfect replicas

I return to a critique of these theories in Chapter 3. In the current discussion,
suffice it to note that these theories uniformly presuppose that creoles and, by
implication, contact languages in general are (in)direct consequences of
imperfect replications of the relevant target languages. As this book will
show, however, the linguistic output of language contact, which we
commonly refer to as contact (mixed or creole) languages, depending on
their socio-historical context, is systematically a complex object made of
morpho-phonological and semantico-syntactic features of the source
languages. This position raises some serious conceptual and empirical issues
about the above-mentioned theories, because they fail to properly address the
question of language change and language creation as general phenomena in
language evolution, that is, the combination of distinct linguistic features in a
developing I-language.
Following Mufwene (1996, 2001, 2005a, 2008), DeGraff (1997, 1999,
2001a, 2001b, 2002) and much related work, this book shows that creoles
represent a normal instance of language change resulting from the contact
between typologically different and genetically unrelated languages (e.g.,
Romance/Germanic vs. Kwa/Bantu [Niger-Congo]) that had been
geographically far apart (Europe/America vs. Africa). These languages were
brought together in extensive, multiple contacts in a very limited space (e.g., a
plantation). As already noted in the previous section, a creole differs in this
respect from the case of English cited above in that Middle English evolved out
of contacts between genetically and, to a large extent, typologically related
languages. Thus creoles represent an empirical domain where, focusing on
particular syntactic and semantic aspects of the emergent languages, we can
isolate distinctive syntactic and semantic features contributed by particular
languages or language groups. Under the uniformitarian view of language
change adopted in this book (shared also with DeGraff and Mufwene), the
same principles underlying linguistic variation apply to creoles and non-
creoles. The only difference is that changes may be more contrastive in
creoles due to their diverse typological origins (Aboh and Ansaldo 2007).
This way, creoles like any contact language, provide us with a window into
the general principles of language evolution.
Adapting Mufwene’s (2001, 2002, 2003, 2005a, 2005b, 2008, 2009)
ecological approach to the evolution of syntactic features/patterns, as well as
1.2 Speakers create variants not imperfect replicas 9
work on the competition of grammars (e.g., Kroch 1989a, 1989b, 2001;

Lightfoot 2006) it is shown that new languages emerge from a process of
competition and selection that leads to a recombination of the syntactic
features of the languages in contact. Following current minimalist
assumptions, I assume that syntactic features are the properties of functional
categories, the latter being the locus of parameters and language variation (e.g.,
Chomsky 1995; Kayne 1994; Muysken 2008). I further hypothesize that
syntactic features (e.g., tense, definiteness, and interrogative) have
specific semantic properties (Chomsky 1995: 381, footnote 14). Therefore, a
functional category involves three aspects minimally: phonology, which I’m
not discussing here, semantics, and syntax. Semantics has to do with issues of
interpretation, while syntax relates to the formal licensing of the category.
Building on this characterization, I argue that language contact (during the
development of an I-language) may lead to fission of the functional category
such that its syntax and semantics are affected differently under the pressure of
the languages in competition. I demonstrate that the ecology of language
contact allows competing components of functional categories to be
recombined into a new functional category that intersects with the same
category in the source languages.
According to this theory, two major possibilities arise in a contact situation:
The emergent language may retain both the semantic and syntactic properties
of a functional category from one of the competing languages. This situation is
described in (5a) and is referred to as pattern transmission. On the other hand,
the emergent language may exhibit a functional category that results from the
recombination of a feature on the basis of its semantics (e.g., discourse
function) in a competing language, while its syntax may be determined under
pressure from other competing languages and/or based on the principles of UG,
the ultimate filter for combinatory possibilities in syntax. The latter possibility,
represented in (5b) and referred to as feature transmission, illustrates linguistic
hybridity as discussed in this book.
(5) a. Fx [Function (semantics) = Lx; Syntax = Lx] ➜ Pattern transmission
b. Fy [Function (semantics) = Lx; Syntax = . . .] ➜ Feature transmission
Within the Minimalist Program, such a split between syntax and semantics is
reasonable if we assume that a feature (F) is associated with a unique general
semantic representation cross-linguistically (e.g., the notion of definiteness,
past, negation), while its syntax (i.e., its licensing properties) is subject to
variation which itself is related to parameter-setting. This would mean that the
same syntactic feature may be valued differently cross-linguistically, even
though its semantics remains the same (Aboh 2006a).
As I show in Chapter 5, the patterns in (5) can be illustrated by noun phrases
in Haitian and Sranan, two creoles that have the same substrate languages (Gbe;
10 Introduction
see Smith 1987, Arends 1989), but differ with regard to their major superstrate
languages: French and English, respectively. According to Aboh (2006a), these
two creoles display specificity markers which function in a way very similar to
the Gbe languages. Therefore, both Haitian and Sranan display bare nouns
similarly to Gungbe as shown by the bracketed nouns in (6a–c):
(6) a. [Wosiyòl] manje kowosòl. [Haitian; DeGraff 1992: 105]
nightingale eat soursop
‘Nightingales eat soursop.’
b. Kofi, go na wowoyo go bai [bana] tya kon gi mi. [Sranan]
Kofi go LOC market go buy banana carry come give 1SG
‘Kofi, go to the market to buy me (plantain) banana(s).’
c. [Àxwèlé] nɔ̀ ɖù [gbàdó]. [Gungbe]
turtle-dove HAB Eat corn
‘A turtle-dove habitually eats corn.’
‘Turtle-doves habitually eat corn.’
Such nouns can be interpreted as definite, indefinite or generic
depending on context. Yet, Gungbe, Haitian, and Sranan display discourse
specificity markers (7):
(7) a. Wosiyòl manje [kowosòl a]. [Haitian; DeGraff 1992: 105]
Nightingale eat Soursop DET
‘Nightingales ate the soursop (in question).’
b. Kofi, teki [a bana] tya gi mi. [Sranan]
Kofi take DET banana carry give 1SG
‘Kofi, give me the (plantain) banana (in question).’
c. Àxwlé ɖù [gbàdó lɔ́ ] [Gungbe]
turtle-dove eat Corn DET
‘A turtle-dove ate the corn (in question).’
‘Turtle-doves ate the corn (in question).’
Aboh (2006a: 224) defines the combination of specificity and
definiteness in these languages as in (8) (see Chapter 5 for discussion):
(8) a. A specific definite noun phrase is strongly D(iscourse)-linked and represents a
unique referent assumed to be known to both speaker and hearer, to which the
speaker intends to refer.
b. A specific indefinite noun phrase need not be D-linked. It represents an
existing referent that the hearer may not know about, but which the speaker
has in mind and to which he/she intends to refer.
As is obvious from these examples, Haitian and Gungbe display the same
pattern in that the noun phrase precedes the specificity marker (viz, NP-det). In
Sranan, however, this marker precedes the noun phrase (viz., det-NP) even
1.2 Speakers create variants not imperfect replicas 11
though the semantics of such marked expressions is very similar to their

equivalents in Gungbe and Haitian Creole. In accounting for these structural
differences, Aboh (2006a) hypothesizes that Haitian has adopted the
morphosyntax of the D-system in Gbe, while Sranan has adopted the
morphosyntax of English. What matters in this discussion is that in both
cases, the newly created forms show features from the competing language
types (e.g., Romance, Germanic, and Gbe). Therefore, even though Haitian
noun phrase sequence is of the type NP-det, the language shows French-like
syntax in many aspects of the noun phrase. For instance, adjectives may
precede or follow the noun (as in French) while in Gbe they can only follow
the noun. Likewise, even though Sranan has det-NP order of the English type,
the language displays bare nouns that have the same semantics and distribution
as in the Gbe languages.
I argue in this book that an important aspect of the theory adopted here lies
in the hypothesis that the emergent languages display recombinations of
features that replicate none of the donor languages faithfully. The book
investigates various aspects of the noun phrase and the clause in these three
creoles (i.e., Haitian, Sranan, and Saramaccan) and contrasts them with
corresponding data from French, English, and Gungbe. The discussion
shows that, though the creoles have lexifiers that are different genetically
and typologically (i.e., Romance and Germanic), this difference alone cannot
explain the typological variation among these creoles. Going back to previous
examples, it appears that both French and English display det-NP order.
Accordingly, the NP-det versus det-NP contrast we observed in examples
(7) cannot be traced back to any obvious difference between French and
English. Instead, it will be shown for each case study that the creoles often
display syntactic and semantic properties that combine those of their substrate
and lexifier languages in non-trivial ways. As observed above, though the
Gbe languages acted as a trigger for the emergence of the feature specificity
in the noun phrase of these creoles, only Haitian Creole adopted Gbe syntax to
any degree, while English seems to have provided the basis for the syntax of
the determiner in Sranan and Saramaccan. Similar observations hold for
various aspects of the creoles such as, adpositions, relative clauses,
modality as related to the development of the comp-system, and serial verb
constructions. It is argued that the emergent language is never a replica of any
of the source systems; rather it involves hybrid recombinations of competing
features.
One question that arises in this respect is what features can be retained in the
formation of the I-language as the outcome of language creation and why? An
assumption made in this book is that features most likely to be selected are
those associated with interfaces, for instance, syntax and morphology or syntax
and discourse/semantics. Thus, some analyses of the complementizer system
12 Introduction
(the interface between sentence and discourse) assume that it is a vulnerable

domain for language transfer (Haznedar and Schwartz 1997; Lardiere 2000;
Prévost and White 2000; Platzack 2001; Goad and White 2004; Sorace 2005;
Tsimpli and Sorace 2006; Sorace and Serratrice 2009). Similarly, Aboh (2006a)
argues that the Determiner system, the interface between the nominal structure
and the discourse, represents a potentially vulnerable domain. Under these
analyses, certain domains of the Faculty of Language (e.g., the syntax–
discourse interface) are more sensitive (or vulnerable) to language contact
effects than others. This, in turn, suggests that certain aspects of core syntax
(e.g., order of merge and predicate structure) might be immune, or less
sensitive, to language transfer. For example, the rigid SVO or head–
complement pattern found across creoles appears to be a consequence of the
merging order imposed by UG. Heads merge first with their complement, and
the set formed in that way merges with the specifier (Kayne 1994).
This analysis, where domains of grammar show different sensitivity to
language creation, due to interface properties, makes it possible to change
perspectives on issues of inflection loss in SLA, as in the case of Middle
English. Indeed, the loss of inflections in language contact would follow
naturally if we assume that inflectional morphology (and agreement in
general) is a by-product of core syntactic configurations, such as specifier–
head vs. complement–head relations. As such, inflectional morphology does
not play a crucial role at the syntax–discourse/semantics interface, and will
therefore not be readily selected in language transfer. This in turn would mean
that the loss of inflections in a language contact situation must be dissociated
from issues related to imperfect SLA. This analysis distinguishes between
inflectional morphology and derivational morphology in a principled way:
the latter, often retained, has denotational content, unlike the former (which
is often lost).
Central to this book is its empirical coverage and how the comparative
syntactic analysis adopted feeds into the proposed theory of language contact
and language evolution in general. Though the core empirical data come from
the above-mentioned creoles as well as their primary substrate and their lexifier
languages, many of the linguistic recombinations to be discussed in this book
also happen in non-creole languages. For instance, the absence in creole
languages of articles that express definiteness only can be contrasted with the
emergence of these grammatical elements in most Romance and Germanic
languages to help understand why only some languages but not others develop
articles. Crucially, this cannot be easily associated with imperfect replication of
a putative target language. Similarly, the study of serial verb constructions in
creoles and the substrate languages sheds new light on the analysis of clause
union phenomena as studied in Germanic and Romance (Cinque 2004). In this
regard, the present book is the first that tries to unravel the genesis of new
1.3 The structure of the book 13
languages by providing a detailed morphosyntactic analysis that addresses the

issue of the emergence of creoles specifically, while simultaneously shedding
new light on the morphosyntax of the parent languages.
1.3 The structure of the book

Given that creoles are discussed in this book as a case study for general
principles of language change, Chapter 2 introduces the socio-historical
settings in which creoles emerged. As Alleyne (1979: 96) notes:
Since we agree that pidgins and creoles demonstrate so forcefully the influence of social
context on language change, language structure, and language usage, we should be
concerned, if we wish to reconstruct the early language and language processes of
pidgin and creole speakers, with reconstruction of the sociolinguistic situation at the
time of its formation, with the structure of the communicative network, and the com-
municative needs of different sectors of the population involved.
This chapter addresses Alleyne’s concern. The discussion primarily focuses on

geo-political factors on the Slave Coast which help identify more precisely the
origins and therefore linguistic profiles of the enslaved Africans deported to Haiti
(then Saint-Domingue) and Suriname. In the context of this book, in which I
investigate the (re)combination of syntactic features in a contact situation, this
discussion is of great importance as it provides us with the necessary background
for identifying which ‘types’ of linguistic patterns or features to consider.
The findings in Chapter 2 feed in the critical evaluation of certain approaches
to creole genesis, which analyze these new languages as exceptional outcomes
primarily due to imperfect second language acquisition. The critical evaluation
in Chapter 3 shows that these theories fail to account for various aspects of
creoles. The main point there appears that no linguistic feature qualitatively
distinguishes creoles from ‘normal’ languages that developed from a more
familiar diachronic change (e.g., Old English to Middle English to Modern
English). This conclusion leads to the proposal that the descriptive tools that
help account for the formation of creoles should also apply to more familiar
diachronic changes (or vice versa).
In this regard, Chapter 4 argues that the competition and selection model
offered by Mufwene (2001, 2005a, 2008) and already adopted in Aboh (2006a,
2007a, 2009a; Aboh and Ansaldo 2007; Ansaldo 2009) offers such a tool when
applied at the individual speaker’s level. Assuming that formal features have
semantic properties (Chomsky 1995: 381, footnote 14), it is shown that, in a
situation of language contact, functional categories (the landmarks of clause
structure) are affected differently, depending on their semantics and licensing
conditions (i.e., their syntax). Accordingly, an emergent contact language may
recombine the syntactic and semantic features differently from the source (or
parent) languages.
14 Introduction
Chapters 5, 6, and 7 put to test the conclusions reached in Chapter 4 on

competition and selection as a model for language change. In order to make this
enterprise possible, various aspects of the selected creoles are studied in detail
in order to: (i) identify the features of the source languages that were in
competition, and (ii) explain how these features have recombined in the new
language. Chapter 5 builds on the results in Aboh (2006a) and deals with the
nominal domain. The discussion here focuses on the emergence of determiners
and their functions (e.g., the expression of topicality as related to
specificity and definiteness, as well as number) in the creole languages.
The analysis further indicates that interfaces play a special role in the
recombination of syntactic features.
Chapter 6 builds on Aboh (2006b, 2007a, 2010e) and Aboh and Dyakonova
(2009) to investigate the structural properties of the clausal left periphery. The
discussion focuses on the expression of modality and the emergence of discourse
markers and their relations to the comp-system. In so doing, this chapter offers a
comparative analysis of the comp-system of Saramaccan and Gungbe. It is
shown that these languages display a comp-system that provides room for
discrete functional projections, the heads of which are realized as markers
which express discourse-related properties such as interrogative force,
topic, or focus, and whose specifiers host distinct fronted elements. Adopting
Rizzi’s (1997) left periphery analysis as implemented in Aboh (2004a, 2006b,
2007a, 2010c, 2010e), I further demonstrate that these projections occur between
force and finiteness, which delimit the complementizer system and may be
realized by various modal particles in Saramaccan and Gungbe. According to the
proposed analysis, the Saramaccan form fu represents two complementizer-
types: fu1, which encodes irrealis mood under force, and fu2, which realizes
deontic modality under finiteness. The same holds for the Gungbe
conditional complementizer ní1 that merges under force and ní2 that
expresses deontic modality under fin. This would mean that both force
and fin express mood specifications, contrary to what is commonly assumed
in the literature. These complementizers delimitate the space where one finds
interrogative, topic, and focus projections. The discussion further confirms the
role of the interfaces in syntactic recombination.
Chapter 7 discusses the emergence of serial verb constructions (SVC) in
Haitian. Building on Aboh’s (2003a, 2009b) analysis of serialization in Kwa,
this chapter takes a step further in showing that the so-called SVCs are widely
represented in languages of the world (including Indo-European), where they
take the form of clause union or restructuring as defined in generative grammar
(Wurmbrand 2001, 2004; Cinque 2004). Under this finding, it appears that the
creoles display (new) recombinations of linguistic features from their source
languages and, consequently, they display two distinct types of clause union
phenomena.
1.3 The structure of the book 15
While the discussion shows that we can account for creole formation by
investigating the recombination of syntactic features, it also indicates that such
recreated forms or structures often arise with local complexities: a
reorganization of the relevant linguistic module, such as when the creole
displays patterns which not only are absent in the source languages but also
exhibit ‘hybrid’ properties that could not have emerged independently from
these languages. For instance, Haitian does not exhibit consecutive serial verb
constructions of the Gbe type (Súrù ɖà làn ɖù. Lit. Suru cooked rice ate) even
though it displays serial verb constructions of transfer of possession that are
very similar to those found in Gbe and Kwa in general. These two types of
series are absent in French. Accordingly, Haitian appears to be more similar to
Gbe than it is to French in this respect. However, Haitian also displays
causative series involving the verb fèt (< faire) as V1. Such series are absent
in Gbe, contrary to French, where they are comparable to causatives involving
the verb faire ‘to make.’ In sum, Haitian developed verb serialization patterns
that are not isomorphic with the Gbe or French structures that contributed to
their development. In addition, because serialization in Haitian comes with its
own irregularities, this domain of Haitian grammar involves local complexity.
Given this observation and assuming that similar developments affect other
domains and sub-modules of Haitian grammar, one may reasonably wonder on
which ground various recent studies claim creoles to form a uniform language
type with simple grammars or even to involve the ‘simplest grammar of all’
(McWhorter 2001).
Chapter 8 concludes the book indicating that natural languages are hybrid
systems that emerge in the course of acquisition.
2 The agents of creole formation: geopolitics
and cultural aspects of the Slave Coast
I mentioned in the Introduction (Chapter 1) that this book adopts a

uniformitarian approach to language creation and language change whereby
learners create new variants, using the same ingredients made available to
them by UG, subject to the specific ecologies of their linguistic experience.
Notwithstanding biological differences between children and adults, which may
underlie their different learning capacities, I assume that language learners
(L1 or L2) produce qualitatively different outcomes, as a consequence of the
ecology of the contact situation (Mufwene 2001, 2003, 2005a, 2008). If the
same principles underlie language creation and language change, we can
conclude that the same general linguistic processes that drive diachronic
change of the type observed, for instance, in English (see Chapter 1) also
governed the formation of creoles. Since I discuss Haitian and the Suriname
creoles as a case study for these general principles of language change, this
chapter presents the socio-historical and socio-political factors that
contributed to the development of these languages. The facts presented here
relate to the type of language that the enslaved Africans deported to the
Americas are likely to have spoken as their L1’s. The main goal of this
chapter is therefore to infer from historical records the linguistic profiles of
the Africans sent to Haiti and Suriname. Section 2.1 deals with the history, the
geopolitics, and the linguascape of the Slave Coast (which currently includes
Western Nigeria, Bénin, Togo, and Ghana) and its major ports Apa/Badagri,
Jakin/Offra, Glehue/Whydah, Xwlagan/Great Popo, Keta, and Accra as
shown on Map 1 (see also Robin Law 1991: 18).
The discussion here focuses on the Kingdom of Allada during a period that
includes the slave trade, that is, from the fourteenth to the eighteenth centuries.
The choice of this region of West Africa is motivated by the fact that we now have
enough historical information suggesting that a significant proportion of the
creators of Haitian Creole and the Suriname creoles in the seventeenth and
early eighteenth centuries originated in this region (Postma 1970, 1975; Price
1973, 1975, 1976, 1983, 2002; Debien 1956, 1974; Fouchard 1972). In addition,
various linguistic studies (e.g., Smith 1987, 1996, 1999, 2009; Smith, and Hugo
Gardoso 2004; Arends 1989, 1995a, 1995b, 1998, 2009; Bruyn 1995a; Migge
16
Downloaded from http:/www.cambridge.org/core. University of Florida, on 03 Jan 2017 at 17:26:45, subject to the Cambridge Core terms of use,
available at http:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/CBO9781139024167.003
2.1 Geopolitics on the Slave Coast: the case of Allada 17
GHANA TOGO
Rep. of BENIN NIGERIA
TADO
EY
BOM
OYO
AG
ALLADA
BADAGRI/APA
PORTO-NOVO
KUTONOU/COTONOU
JAKIN/OFFRA
GLEHUE/WHYDAH
XWLAGAN/GREAT POPO
LITTLE POPO
LOME
KETA
ACCRA
Map 2.1 The Slave Coast and its major ports
1998, 2003; van den Berg 2007 for Suriname creoles, and Sylvain 1936; Singler
1996; Anglade 1998; Lefebvre 1998; Aboh 2006a; Aboh and DeGraff 2014,
forthcoming; Aboh and Smith 2015a) have shown that speakers of the Gbe
languages of the Kwa family contributed to the formation of these creoles.1
Section 2.2 briefly discusses population factors on the plantations in Suriname
and Haiti. Because there have been many well-informed studies on this topic, this
section only summarizes the historical and linguistic findings that are relevant for
the discussion in this book. Section 2.3 concludes the chapter.2
2.1 Geopolitics on the Slave Coast: the case of Allada

In creole studies, it is commonly assumed that creole languages were
exclusively created by enslaved Africans who did not have sufficient access
1
This obvious limitation of the geographical area does not mean overlooking the important role
played by Central African languages in shaping these creoles, as some of the languages (such as
Kikongo in the case of Suriname) were also spoken by large proportions of slaves. Indeed,
including them in the discussion complexifies the competition and selection scenario articulated
in this book. This realistic narrowing of language coverage is intended to keep the discussion
more focused and manageable, hoping that later scholarship with similar detail of analysis of
Bantu languages with the same creoles can help us develop a more comprehensive and richer
evolutionary perspective, especially in highlighting the competition and selection that obtained
among the various substrate patterns (Mufwene 2001, 2005a, 2008).
2
I’m immensely grateful to my dear sister Frieda Sessi Aboh for her numerous and very
constructive comments on earlier versions of this chapter.
18 The agents of creole formation
to the European target language. As I argued in Chapter 3, a growing number of

theories of creolization hypothesize that the European languages spoken in the
colonies must have evolved quickly into local koinés (e.g., Chaudenson 1992,
2001, 2003; Mufwene 2001, 2005a, 2008; Siegel 2008). Though such a koiné
might not have been a very stable one, it was probably a fairly focused system
that the Bozal, African-born slaves, could have targeted as a second language
not long after the colony was founded. This was probably the case during
the homestead period where most enslaved Africans interacted closely with
their masters and could speak the ‘standard’ language on the colony or
plantation (see, for instance, Chaudenson 1992, 2001, 2003 for discussion).
The situation changed drastically during the plantation phase where the
enslaved Africans significantly outnumbered the Europeans, race
segregation was institutionalized, and the slaves had to learn from one
another. Without getting into the details of the theories of creolization (see
Chapter 3), I contend that the creoles emerged at the moment when the master
was no longer the linguistic model, but the enslaved African was. Why then
did these enslaved Africans who now constituted the numerically dominant
group continue to target (though indirectly) the linguistic model of the
Europeans instead of falling back onto their own native languages? As
Mufwene (2001, 2005a, 2008) reminds us, one has to wonder why the
colonial language and its related creole systematically displaced African
languages in the Americas. The common understanding is that:
(i) the enslaved Africans were not allowed to speak their languages on the
plantations;
(ii) the enslaved Africans came from too varied areas of Africa to be able to
create a lingua franca or maintain an ethnically dominant language;
(iii) the Bozals had to pattern their adaptive behaviors largely on those who had
come before them who had survived the atrocities they suffered; and/or
(iv) the phenomenon could be attributed to a combination of all the above
explanations.
Under such assumptions, one could surmise that the displaced Africans created
the creoles for various reasons, including but not limited to:
– Inter-class communication (i.e., communication with the ruling class)
– Intra-class communication (i.e., communication within the slave group)
– Expressive motivation (creation of an identity, see Muysken 1981a).
Yet, the answer to the question of the emergence of the creoles and the assumed
loss of the African languages cannot be this simple (see also Mufwene 2001,
2008 and references therein for discussion). Let us consider the socio-historical
context on the Slave Coast prior to and during the transatlantic slave trade or the
Atlantic ‘triangular trade.’
Though studies on the Atlantic ‘triangular trade’ have demonstrated the

importance of this trade in the global economy of the world from the
sixteenth to nineteenth centuries, its impact on the African countries is not
often taken into consideration in creolistics. The implicit assumption seems to
be that Africa was a rather chaotic, defenseless continent where isolated tribes
suffered from slave raiders who organized slave caravans across West and
Central Africa with the help of a few local smugglers. The history of the Slave
Coast indicates a more complex reality.
Slavery was unquestionably part of a global economic system later to be
known as globalization. In the context of Africa, one often finds in historical
documents that trade routes were described in terms of the relevant coastal
areas where merchants could land and what they could trade. Accordingly,
earlier maps of West Africa include sections such as the Slave Coast, the Gold
Coast, or the Grain Coast. This is different from topographic names in the
Americas where European settlers adopted local names (e.g., Jamaica;
Mexico), imported European topographical names (e.g., New France, New
Spain), and/or names inspired by their religion (e.g., Trinidad, Saint-
Domingue, Saint Martin).3 It is therefore important to understand the geo-
politics of the areas where Africans were traded or more precisely, captured and
sold. In a case study like this one, where we are trying to probe into the mind of
the speaker in the process of language creation, it is equally important that we
understand their linguistic background. We cannot limit ourselves to a few
broad generalizations about West Africa. In this book, I exclusively focus on
the Aja communities from the Bight of Benin (aka the Slave Coast). It is my
contention that the Aja people were instrumental in the development of the
Haitian and Surinamese creoles.
2.1.1 The Aja communities on the Slave Coast: the case of Allada
Referring to the Slave Coast, Law (1994: 59) remarks:
This section of the West African coast east of the River Volta, comprising the western
half of the Bight of Benin, was so prominent in the supply of slaves for the Atlantic trade
that it became known to Europeans as the “Slave Coast”. The most powerful state (and
principal supplier of slaves for the Atlantic trade) in the region during the sixteenth and
seventeenth centuries was the kingdom of Allada.
3
This difference in the naming patterns of regions on the continents on both sides of the Atlantic
may be correlated with Mufwene’s (2001, 2005a) distinction between trade and settlement
colonies, the former in Africa and the latter in the Americas. They reflect differences not only
in the kinds of investments that Europeans had in the colonies but also in the attitudes they
developed toward them.
Origins of Allada According to Pazzi (1979), Adande (1984), and Law

(1997), the various ethnic groups which relate to the Aja tradition and are
now found on the coastal area stretching from Western Nigeria to Ghana,
including Central and Southern Benin, as well as Togo, once belonged to the
larger community of Aja-Tado. This community was founded by migrants from
Oyo (present-day Nigeria). It is believed that this group of migrants (led by
Togbui Anyi) eventually settled in Tado (today’s Republic of Togo) on the
Mono River no later than 1000 AD, (Pazzi 1979: 148).4 Pazzi (1979: 152)
further suggested that the area of Tado was probably already inhabited by other
groups originating from Ife, also in Nigeria (e.g., the Ana who worship
Na-buruku). The relevant point for our discussion is that the Aja originated
from Oyo, that is, a Yoruba culture. We can infer from this that they presumably
first migrated to an area displaying a culture familiar to their own.
After settling in Tado, the Aja soon seized power and founded the Kingdom
of Aja-Tado which would later become the dispersal point for the Aja across
and along the Mono River. Indeed, oral history retains Tado as the starting
point of successive migration waves that characterized what we can now refer
to as the Aja expansion (Pazzi 1979, 1984). This expansion led to the creation
of several kingdoms, among which: Ƞɔtse (present-day Togo) to the west
(c. fourteenth to fifteenth centuries), Allada to the southeast (early fifteenth
century) and Dànxòmὲ further inland to the North (early seventeenth
century). Both Allada and Dànxòmὲ are part of present day Republic of
Benin. The emergence of these various kingdoms of course coincided with
the dispersion of various clans or sub-clans along the coast and further inland
concomitantly with the decline of the ancestral city of Tado. According to
Pazzi (1979: 147), five major groups emerged from these successive
migrations: Ewe, Xwla, Aja, Ayizo, and Fon. Of these five groups, we can
observe that four correspond to older languages as reconstructed by Capo
(1991: 15): Ewe, Xwla, Gen, Aja, and Fon. Map 2.2, adapted from Pazzi
(1984:10/12), illustrates these earlier migrations.
Politics and institutions It is difficult to determine exactly when Allada was

founded but historical documents suggest that this kingdom was well
established before it came to be known to Europeans. For instance, in a letter
written in Beni (present-day Nigeria) on August 30, 1539, three Portuguese
missionaries describe the King of Beni in the following terms: “his way of
mistreating and arresting all the ambassadors of the kings who sent him
messages, as he just did with those of Labadi and Allada and many others”
(my translation). As stated by Pazzi (1979: 148) this letter informs us not only
4
Pazzi (1979: 156) established this approximate date based on the fact that the Aja people were
never converted to Islam unlike most kingdoms in the environs of the Niger in early 1000 AD.
GHANA TOGO Rep. of BENIN NIGERIA
EY
TADO
BOM
OYO
AG
ADA
ALL
N tse
PORTO-NOVO
JAKIN/OFFRA
BADAGRI/APA
KUTONOU/COTONOU
GLEHUE/WHYDAH
XWLAGAN/GREAT POPO
LITTLE POPO
LOME
KETA
Map 2.2 Map of migrations within and around the Allada Kingdom
about the rise of Allada but also about its institutions. Given that Allada was
developed enough to have political relations with the King of Beni, its
founders, that is, the people who left Tado (i.e., the Agàsú royal family),
must have arrived there several decades, or maybe a century, earlier. This
would set the foundation of this kingdom between the late fourteenth century
and the fifteenth century. That the King of Allada was able to establish regional
connections with other kings in the sixteenth century (e.g., the King of Beni)
and was then known to Europeans indicates that it was already a prosperous and
powerful kingdom. For instance, Father Jose de Najara, who visited this area in
1672, reports on an event that Pazzi (1979: 175) dated from around 1500
saying: “within the great number of wives that the King of Popo had there
was even a sister of the King of Allada, who for some reason refused to pay the
yearly tribute.” It is further mentioned in this report that the King of Allada
eventually defeated the King of Popo. The expansion of Allada, partially
illustrated by this report, continued throughout the sixteenth century. A note
by du Casse (1687) cited in Pazzi (1979: 176), reports that “the kingdoms of
Great Popo and Little Popo were both satellites of Allada . . . The same holds of
Whydah which depended on Allada” (my translation). These different reports
suggest that the Kingdom of Allada was organized around a central power that
ruled over a number of satellite or smaller kingdoms that had been annexed.
The exact size of Allada is difficult to estimate but some reports in the
seventeenth century (e.g., Olfert Dapper, a Dutch geographer) suggest that it
TADO
GHANA TOGO Rep. of BENIN NIGERIA
Y
ME
GBO
OYO
A
ALLADA
PORTO-NOVO
BADAGRI/APA
KUTONOU/COTONOU
JAKIN/OFFRA
GLEHUE/WHYDAH
XWLAGAN/GREAT POPO
LITTLE POPO
LOME
KETA
Map 2.3 Approximate area of control of the Allada Kingdom
stretched from Dànxòmὲ , to the north up to the coast on the Atlantic Ocean. It
shared a border with Oyo to the east and northeast (as indicated by the fact that
the port of Apa, in present day Badagri, in Nigeria, was under its control) and it
included the territories up to the defeated kingdoms of Great Popo to the west
(Pazzi 1979; Law 1997).5
A point shown on this map that is immediately relevant to this book is that
Allada once controlled all the major ports on the Slave Coast where the
Europeans were trading and sometimes settled: Apa/Badagri, Jakin, Offra,
Glehue/Whydah, and Xwlagan/Great Popo. While Jekin (also referred to as
Jakin, Jakkein, Jaquin) and Offra (sometimes referred to as Offer) seemed to
have been the primary ports of Allada, sources also occasionally mention
Whydah and Apa as controlled by Allada to whom the local governors
(Tògán ‘chief of country’) had to pay a yearly tribute. That such governors
5
Though the kingdom of Dànxòmὲ eventually became independent and even conquered Allada in
1724, I’m assuming here that it once belonged to the network of satellite kingdoms that were
under the rule of Allada and to which it had to pay tributes. For instance, a report dating from
1716 describes the ruler of Dànxòmὲ as follows “The Viceroy of Agoeme [i.e., Dahomey][in the]
province of Foin [Fon] in the dependent territories of the King of Ardres has thrown off the yoke,
he has made himself independent” (Law 1997: 110). This actually corresponds to oral tradition
where it is believed that Dànxòmὲ (Dahomey) was part of Allada until it became independent.
Oral history also suggests that prior to independence the Viceroy of Dànxòmὲ was appointed in
Allada where the rituals were held (see Pazzi 1979; Adande 1984; Law 1991, 1997 for
discussion).
were under the rule of the King of Allada can be inferred from the fact that the
latter is referred to as “the Great King of Allada” in some historical documents.
This is presumably a translation of the local terminology such as the Gungbe
title Àxɔ́ sú àxɔ́ sú lέ tɔ̀ n ‘the King of kings’ (Pazzi 1979; Adande 1984; Law
1991, 1997). That Allada could have control over the coast can also be inferred
from the size of its army which amounted to 40,000–50,000 (Law 1997: 20).6
We can infer from these figures that Allada was a densely populated
kingdom. According to Law (1997: 20):
the Allada kingdom was generally regarded as relatively densely populated: Dapper in
the mid-seventeenth century, for example, describes it as having ‘towns’ [vleken, lit.
markets] and villages in great number; d’Elbée, traveling from Offra to the capital in
1670 found the land ‘peopled with many towns and villages’. The Dutch in the late
seventeenth century, it may be noted, thought Allada more densely populated than Benin
to the east.
The capital of this kingdom referred to as “Great Allada” situated in Togudo in

present-day Republic of Benin, was about 50 km inland and was well known to
Europeans who found it impressive both in terms of its architecture and the
trading possibilities it offered. The population of the city was estimated to
about 30,000. According to Olfert Dapper, the Dutch geographer, the kingdom
of Allada
is a country full of villages and well situated for trade because one finds there all year
long and in great abundance, millet, palm wine, fruits and vegetables such as yams,
potatoes, oranges, limes, coconuts, etc. It is a country of plains and beautiful valleys
including great rivers and the roads that lead there are well maintained. People there
produce a lot of salt. (Adande 1984: 239, my translation).7
The reports then continued with great details regarding the customs of the
inhabitants, their diets, cooking traditions, kitchen utensils, clothing, hygiene,
and funeral rites. The King is said to have two royal palaces in the capital,
which, according to Olfert Dapper as cited in Law (1997: 15), “were enclosed
by thick earth walls, and each was as large as the town of Monnikendam in
Holland” (see also Pazzi 1979 and Adande 1984). Likewise, Captain François
d’Elbée, who visited the city in 1670 as a representative of the French
Compagnie des Indes Occidentales, stayed in the royal palace, where he was
put up in an apartment reserved for foreigners. He reported that:
6
Other sources, e.g., d’Elbée (p. 381), estimate the King’s army at 500,000.
7
“C’est un païs plein de villages et bien situé pour le commerce, parce que c’est un terroir fecund
en millet, en vin de palme, en plantes et en fruits qui durent toutes l’année et qui leur sont d’un
grand secours, comme les injames, les batatates, les oranges, les citrons, les noix de coco etc.
C’est un païs de plaines et de vallées entrecoupé de grandes rivières et les chemins qui y mènent
sont grands et bien battus. On fait beaucoup de sel sur la côte d’Arder.” (The spelling is as in the
source document; see Adande 1984: 239.)
this palace is very large and well constructed. It is divided into large courts and large
gardens. The building is all surrounded by porches which form several galleries that lead
to the various courts or gardens. It is a two-storey-house composed of big halls and
rooms including various furniture.8
As noted by Law (1997: 15), the existence of a storey-house in the Allada of the
seventeenth century where the King can put up some of his visitors is
particularly noteworthy as this is indicative not only of the wealth of the
Kingdom but also of its socio-cultural aspects and more generally of African
civilization prior to colonization. The description that Captain François d’Elbée
made of the royal palace also gives us indications about how to understand
Allada within the geopolitics and economies of the seventeenth century. Allada
was participating in international trade and the King had political and
commercial agreements with other kings in West Africa and overseas. We
know from the discussion in previous paragraphs that the King of Allada sent
ambassadors to Beni back in 1539. In addition, of the three kings that are
known to have ruled Allada, the first, Toxonu, sent an ambassador to the King
of Spain Philippe IV in 1658. Diplomatic relations between Allada and Spain
are also illustrated by the publication in 1658 of La Doctrina Christina by
Spanish missionaries who had settled there. This catechism represents one of
the oldest documents on a Gbe language that exists to this day and on which I
comment below.
The second King, Tézinon, received Captain François d’Elbée in the year of
1669–1670, as noted above. One objective of this visit was to re-establish and
strengthen commercial and diplomatic relations between the King of Allada
and the King of France Louis XIV. During Captain d’Elbée’s stay in Allada, an
incident between the newly arrived French and the Dutch factors who were
settled in Jekin for 25 years was brought to the King. In settling these matters,
the King Tézinon decided to send an ambassador, Mattheo Lopes (see
Figure 2.1), to the King of France Louis XIV. Mattheo Lopes who was one of
the King’s most appreciated translators embarked on La Concorde, the second
French ship sent to Allada the same year. Mattheo traveled to France with three
of his sons, three of his wives and six slaves. He is reported to have arrived in
France on December 3 of the same year and met with Louis XIV on December
19 (d’Elbée 1670: 308–315; Adande 1984).
These facts indicate that Allada was probably one of the most powerful
kingdoms in West Africa in the sixteenth and seventeenth centuries both in
terms of its army and institutions. As mentioned in previous paragraphs, the
8
“Ce lieu est grand et spacieux et assez bien bâti, il est divisé en grandes cours et grands jardins,
et le bâtiment est tout entouré de pilliers qui forment des galeries et porches pour se promener tant
du côté des cours que des jardins : il est élevé de deux étages où il y a divers appartements
fort grands et spacieux composés de salles, chambres et autres accompagnements.” (D’Elbée
1671: 418).
Figure 2.1 Portrait of Dom Matheolopes, Ambassador Roy d’Arda. Prints on

the History of France (Volume 50), Parts 4532–4594 (1670–1671) Source:
gallica.bnf.fr / Bibliothèque nationale de France
kingdom consisted of a central nucleus, Allada, where the ‘King of Kings’

(àxɔ́ sú àxɔ́ sú lέ tɔ̀ n) lived, and of satellite regions (e.g., Dànxòmὲ , Jeken,
Whiddah) that were ruled by chiefs (tò gán) or subordinated state kings
(àxɔ́ sú), according to Law (1997: 20). This suggests that the Kingdom of
Allada was relatively decentralized, a point that will become relevant when
we discuss geopolitics and how it affected the slave trade on the Slave Coast
during the eighteenth century.
Economy Allada was a rich kingdom with a booming economy integrated in

the world’s global economy of the sixteenth to early eighteenth centuries. The
kingdom was at the crossroads of the trade routes between Beni and Ƞɔtse
(present-day Togo) from east to west and from the south (i.e., the coast) to the
north. This strategic location allowed Allada to participate in regional trade
(e.g., the so-called ‘pearl route’) and trans-Atlantic trade (see Pazzi 1979 for
discussion). During this period, Allada’s economy revolved around two main
axes: local economy and slave economy.
Local economy
The local economy was based on locally produced goods that were traded in the
kingdom and were also exported. Among such products, we find farm products
mentioned above by Olfert Drapper (e.g., fruits, millet, vegetables, etc.). While
part of these products is for local consumption, others such as salt and palm oil
were also used for exports. For instance, Pazzi (1979: 154) refers to a letter
dated from 1574 by G. M. Branco, a Portuguese merchant from São Tomé,
where he reports that “with our friend the King of Allada, who is close to Mina
[i.e., El Mina in present Rep. of Ghana] we trade for slaves, ivory, cotton cloth,
palm oil, and many vegetables such as yam and other foods. Each year we
import from that port one or two ships loaded of what is mentioned above” (my
translation).9
Clearly this text illustrates the two economic axes of Allada: slaves and
locally produced goods whether agricultural or manufactured. In addition, it
gives us an estimate of the quantity of such goods that is consistent with Olfert
Dapper’s description above that one finds a great number of vegetables and
other products there all year long. With regard to manufactured goods, Pazzi
(1979) further indicates that the economic strength of Allada also relates to its
craft industry involving artisans who were probably hired from neighboring
areas or kingdoms. A case in point is cotton cloths. Indeed, that Allada was
selling cotton cloths in the sixteenth century indicates that it was now
9
“Avec notre ami le roi d’Arda /Alada/ qui est près de Mina, nous envoyons traiter des esclaves
nègres, de l’ivoire, des pagnes de coton, de l’huile de palme et beaucoup de legumes comme
l’igname et autres aliments. Tous les ans il sort de ce port un ou deux navires de ce qui est indiqué
ci-dessus” (Pazzi 1979: 154, footnote 3). See also Law (1997: 92 and footnote 526).
competing with Beni, which was then well known to Europeans for producing
cotton fabrics. One could think that the clothes bought in Allada might not have
been produced locally but were instead imported for subsequent export. While
this view may not be disregarded outright, some reports based on ships’
manifests in the seventeenth century cited in Pazzi (1979: 117) indicate the
following quantities: 588 pieces on a Dutch boat in 1645 and 5,000 pieces on an
English boat the following year. With regard to these cotton clothes, the English
Captain Philips remarks that, in Whydah another port further to the west, the
cotton clothes are made of strips of about 25 cm wide that are stitched together
in larger bands of three strips. The strips found in Allada were narrower. In
addition, Allada’s cotton strips also differed from those of Beni, which were
stitched together alternating between white and blue strips. Because the cotton
strips found in Allada were distinct from both those found in Whydah and
Benin, we can conclude that they were locally made (i.e., in the interior of the
Kingdom). We can infer from the Capitain’s remarks that Allada had local
weaving technology.10 Law (1997: 92) suggests that the cloth export from
Allada must have been somehow limited compared to that of Beni. However,
the relevant point here is the diversity of products and goods that could be
found on the coast of Allada which in turn could be indicative of the King’s
efforts to make Allada an attractive stop for European traders. Other
commodities sold to Europeans are cowries which served as exchange money
in West Africa during this period.
Given that these exports were mainly documented by Europeans trading on
the coast and involved in the slave trade, it is reasonable to assume that part of
these commodities were provisions for the slaves and the crews during the
middle passage. This would suggest that a significant part of Allada’s non-
human export actually fed into the slave trade. Thus, local economy, as defined
above, partially depended on the slave economy. This assumption is critical for
the argument put forth in this book as it relates to the origin of the Africans who
were exported to Haiti and Suriname in the seventeenth century.
Slave economy
As suggested by the quote from Law (1994) above, Allada was one of the main
suppliers of slaves for the Atlantic trade during the seventeenth and early
eighteenth centuries. We saw from the Portuguese merchant G. M. Branco’s
letter in previous paragraphs that the King of Allada was already trading slaves
back in 1574. Until 1635, the Portuguese monopolized the trade with Allada.
Comparing two Portuguese reports from 1607 and 1620 that showed estimates
of the value or volume of trade between São Tomé and ports in the Bight of
10
Actually, this has become a tradition in Benin, where one still finds traditional weavers in the
Royal Pallace of Agbómὲ, now a museum.
Biafra, Law (1994: 65) evaluates the number of slaves from Allada as about
667 slaves a year out of a total number of 1,167 slaves from the Slave Coast as a
whole. Though these numbers would grow dramatically when the trade was at
its peak (i.e., late seventeenth and eighteenth centuries), the Africans traded by
the Portuguese from Allada of the early seventeenth century represented
slightly more than fifty percent of the total numbers of slaves. As Law (1994)
remarks, this could be due to the fact that “although a small part of this trade
was in commodities other than slaves this is offset by the likelihood that (as was
generally the case later) slaves on the Slave Coast were cheaper than in
Angola.” This is indeed supported by the following extract of a letter by a
Dutch factor, I. Van Hoolwerf who arrived in Jeken (Allada’s port) in 1686.
According to him, “the slaves here are in good shape and cheap. As for our
living conditions here, we are in a period of peace with a relative abundance of
slaves” (Pazzi 1979: 64–65, my translation).11
The Portuguese monopoly over trade in the region was soon replaced by a
Dutch monopoly (1636–1653). Historical documents indicate that the Dutch
authorities from Brazil “envisaged imports of 1,500–1,600 slaves annually
from Allada, and a further 500–600 from Popo” (Law 1994: 66). While these
numbers were thought of as optimistic (Law 1994), they are nevertheless
important if we take into consideration the fact that Popo represented one of
the satellites of Allada. Adding these figures up, the estimates of slave out of the
Allada kingdom were about 2,000 to 2,200 annually. This estimate clearly
shows a sharp increase in the trade, which is four times the numbers reported in
previous paragraphs with regard to the Portuguese trade of the early
seventeenth century. Whether these objectives were achieved is unclear but
Law (1994: 66) indicates that the
WIC trade by van den Boogaart and Emmer suggests that Dutch purchases of slaves
from “Guinea” . . . averaged only about 1600 per year in 1637–44 and only once in
(1642) reached over 2000 in a single year. . . . Van den Boogaart and Emmer were able to
establish the provenance of around three-quarters of the total, of whom almost half
(48%) were from Allada.
According to Law (1994), the trade in Allada will then decline between 1649
and 1658 but peaked again in 1659–1667, when the Dutch had to compete with
other European nations also trading with Allada. The following dates show the
flow of different European trade settlements in Allada (Law 1997: 5–9).
– 1574–1636: The Portuguese monopoly.
– 1639: The Dutch West India Company (WIC) constructed a trade post in
Allada.
11
“Les esclaves ici sont bons et de bas prix. Quand aux conditions de vie dans cette region, nous
nous trouvons dans la paix avec une discrète abundance d’esclaves . . . ” (Pazzi 1979: 64–65).
– 1658: Trade contacts between the King of Allada and the King of Spain.
– 1664: James Duke of York (future King James II) then Governor of the
Company of Royal Adventurers (CRA) wrote a letter to the King of Allada.
– 1670: The French West Indies Companies established a trade post in Allada.
The Kind of Allada sent an ambassador, Mattheo Lopez, to King of France
Louis XIV.
Another European company that is reported to have operated in Allada
during the same period is the Danish African Company. It is now clear that
not only were Allada and its main ports part of the world economy, it also seems
to have been a location favorable for trade where European nations competed.
This competition led to a dramatic increase in the number of slaves exported
from Allada. Of the 5,000 to 6,000 slaves the Dutch delivered to Spanish
America between 1668 and 1672, Law (1994: 73) reports that “it is clear that
a large proportion of these slaves were obtained from Allada.”12 This appears to
be consistent with the claims of François d’Elbée who visited Allada in 1670
and evaluated the exports from Allada at an average of 3,000 slaves per year.
Similarly, a Dutch source of 1670 indicates that “over and above the English
and French trade, the Dutch in Allada were purchasing every year . . . 2,500 to
3,000, sometimes even more slaves” (Law 1994: 75). In addition to the Dutch
trade, two French ships are reported to have commerced with Allada in 1670:
La Justice (captained by François d’Elbée) embarked 434 slaves, while La
Concorde, which arrived a few days before d’Elbée’s departure, embarked 563
slaves. Thus, Allada was exporting about 4,000 slaves a year. This incredibly
large number is consistent with another French source reported in Law (1994:
76) in which estimates of exports form Allada in 1671 were between 4,000 and
5,000 people.
Because these numbers sometimes include import estimates made as part of
the commercial plans of the European trading companies, they probably do not
reflect reality and should therefore be taken with caution (Curtin 1969). In an
effort to evaluate the overall slave export from Allada in the 1710s and its
impact on the local economy, Law (1997: 102) suggests that Allada’s highest
slave exports could be estimated at about 15,000 a year. Eltis (2011: 275) gives
even higher figures suggesting that the trade in Allada and its subordinated
ports (e.g., Whydah, Jakin) rose from about “1,000 a year in the 1650s to 8,000
in the 1680s, to a peak over 19,000 a year in the 1720s.”13 These impressive
numbers certainly fluctuated and varied according to various factors, including
12
Emmer (2005: 35), on the other hand, suggested a figure of 4,000 slaves delivered to Brazil per
year.
13
These figures can be put in the context of more recent figures by Eltis (2011: 272), which
indicate that in the years 1766 to 1776 the total volume of captives exported from Africa to the
Americas increased from about 12,000 to 80,000 a year: a clearly astonishing figure.
but not limited to war, climate, rain/drought, and epidemics. For instance, it is a
known fact that while wars may yield war prisoners to be sold as slaves, they
also often led to roadblocks whose primary impact was to prevent slave
caravans access to the coast. Similarly, several reports suggest that local
slave owners tended to sell their slaves in periods of droughts, when the costs
of maintaining a slave were high, rather than during advantageous seasons,
when the owners could use their slaves for their own purposes (e.g., farming),
to increase their income. A case in point is the following extract from a letter by
I. Van Hoolwerf dating from February 1688:
Since I arrived here, slave trade has been in steady progress, but these days, the trade is
slowing down because of the lack of wars in the interior and the abundance of grains this
year, which does not lead people to sell their slaves, contrary to last year during the
droughts. . . . The arrogance of the Fidalgo . . . A couple of months ago, he started a
war . . . as a consequence, the roads were completely blocked and we could no longer
receive slaves from the interior. (Pazzi 1979: 65–66, my translation).14
The fluctuations described by Hoolwerf affected not only the number of slaves
sold in Allada but also prices. In this regard, Law (1994: 78) reports that Dutch
accounts for slaves bought in Allada between 1638 and 1680 suggest an
average price of about 32 fl per head. English records of the 1660–1681
period, based on “prime costs i.e., cost in Europe of the goods exchanged for
slaves,” indicate an average price between £2.71 and £2.88. However, Law
(1997: 102) evaluates the price of the slave to be at 48.000 cowries
“conventionally equivalent to £15 in local value.” Assuming that Allada
slave export ranged between 15,000 and 19,000 slaves per year (Law 1997:
102 and Eltis 2011: 275, respectively), we can estimate the yearly gross income
of Allada to vary between £225,000 and £285,000, that is, between 720 and 912
million cowries.
Though these figures are speculative, we can infer from their magnitude that
various social classes in Allada must have benefited from the trade. Because of
its institutions and how the trade was organized locally, it appears that a large
portion of the income generated by trade in general and the slave trade in
particular was to the benefit of the King. This can be seen from his numerous
luxurious palaces described above. Other locals who benefited from this trade
were probably other state authorities (e.g., Ministers and Captains, see d’Elbée
1671: 438–440), the subordinated chiefs (i.e., the tò-gán) in the different ports
(e.g., Jakin, Whydah) and in the interior (e.g., Agbomey, capital of Dànxòmὲ )
14
“Jusqu’ici la traite a été en constant progress depuis mon arrive, mais en ces jours elle semble
ralentir un peu, à cause du manque de guerres à l’intérieur et de l’abondance des céréaux en cette
année, ce qui ne pousse pas, comme faisait la famine l’année passée, à vendre les esclaves . . .
L’insolence du Fidalgo . . . Il y a quelques mois, il se chargea d’une guerre sur lui seul . . . Les
chemins furent alors tellement bloqués que nous ne pouvions plus avoir d’esclaves” (Pazzi
1979: 64–65).
as well as independent traders and notables of both African and European

descent (e.g., the famous Francisco Felix de Sousa in Whydah during the
eighteenth century). It seems that this trade mainly benefited a local elite in
the cities at the expense of the more rural population inland.15 Considerations
about this social class are important in the context of our discussion on
language contact because people belonging to this elite are also the ones who
had access to ‘formal’ education, learned the languages of the Europeans or
made use of interpreters, who, as noted by Mufwene (2005a, 2008) in
particular, were very important in mediating communication between
Europeans and the locals and for often serving as linguistic models to other
local L2 learners (see Law 2011 for discussion). In 1670, François d’Elbée
reports his meeting with the Prince of Allada stating: “Though this Prince
understands Portuguese, or more precisely a corrupt version of Spanish used
here as lingua franca, he nevertheless has everything translated to him by his
two interpreters: Mattheo Lopez and Fransisco” (my translation, p. 405).16 As
explained above, Mattheo Lopez was appointed later that year as ambassador in
France, where it turned out that he spoke Portuguese perfectly and understood
Spanish (d’Elbée 1670: 538–544).
Likewise, le Chevalier Marchais’ (1725) report on Allada, cited in Labouret
and Rivet (1929: 20), indicates that “a form of corrupt Portuguese is preserved
there and developed into a jargon or lingua franca that almost all the people
understand and speak such that those who know Portuguese do not need
interpreters in that State” (my translation).17
Put together, Captain d’Elbée’s and Chevalier Marchais’ reports indicate that
within 55 years, a variety of Portuguese emerged in Allada where it was
commonly used as lingua franca.
2.1.2 Competition between Allada and its satellites and the fall of Allada
It is not difficult to imagine that Allada’s revenues from trade furthered
competition between the centralized power in the capital city of Allada and
the subordinated chieftaincies of the kingdom, including Dànxòmὲ and
15
What was happening then in Allada is quite similar to current situations in many developing
countries where a central power in alliance with some multinationals monopolize the resources
of the country at the expense of local populations. In the fifteenth to eighteenth centuries, the
resources then were manual labor: the slaves.
16
“Quoi que ce Prince entende fort bien le Portugais, ou pour mieux dire, l’Espagnol corrompu,
tenant de langue Franque, il ne laisse pas de se faire tout interpreter par ses deux interprètes,
Mattheo et Fransisco” (d’Elbée 1670: 405).
17
“La langue Portugaise corrompue s’y est conservée jusqu’à present, et produit un jargon ou
langue franque que Presque tout le peuple entend, et parle de sorte que ceux qui savent le
Portugais n’ont pas besoin d’interprète dans cet État” (Labouret and Rivet 1929: 20).
Whydah, and neighboring kingdoms such as Beni. Indeed, between the late
seventeenth and early eighteenth centuries, the Kingdom of Allada had to
contend with internal pressures from its various chieftaincies demanding
independence and strife from the ever-expanding Fon Kingdom of Dànxòmὲ
north to the Ko (see Maps 2.2 and 2.3).
According to Pazzi (1979: 195) the Kingdom of Dànxòmὲ emerged in the
early 1600s as an offshoot of the Allada royal family, after a family dispute over
the succession of the late King Kɔkpon. The royal family split into three
different groups: one remained in Allada representing the central power,
while the other two migrated (south)east and north, respectively. The north
bound group founded Dànxòmὲ. Given such circumstances, it is easy to assume
that it has always been a challenger to the central power in Allada, which it
eventually vanquished and conquered on March 30, 1724 (Law 1997: 114; see
also Pazzi 1979; Adande 1984). The decline and fall of Allada have been
largely attributed to internal divisions, presumably motivated by the need of
the King to maintain military power and control over an economy fuelled
largely by the triangular slave trade. Allada suffered a severe defeat against
Oyo in 1698, but the reasons for this war seemed to be “an appeal to the Oyo by
the subject of the King of Allada for protection against the misgovernment of
his viceroys” (Law 1997: 111). This happened shortly after the Fon from
Dànxòmὲ asserted their independence (see note 5) “The Viceroy of Agoeme
[i.e., Agbómὲ ] [in the] province of Foin [Fon] in the dependent territories of the
King of Ardres [Allada] has thrown off the yoke, he has made himself
independent” (Law 1997: 110). The following two reports from different
visitors cited in Law (1997) illustrate this period of instability:
Arda [Allada] was once a powerful and populous nation . . . but having afterwards been
weakened by the revolts of the Quitta [Keta] . . ., and the Popoe, Whydah, Appee [Ekpe],
Bidagry, and Lagos, it became incapable of opposing arms of Trudo. (Law 1997: 111,
emphasis is mine)
Trudo, commonly known as Agaja was then the king of Dànxòmὲ. Quite
remarkably, all the names in italics in this quote referred to slave ports once
controlled by Allada and around which big cities emerged (see Map 1 and
Pazzi 1979, Law 2011). This gives further credence to the idea that control
over the economy played an important role in the crisis that Allada went
through during the late seventeenth and early eighteenth centuries. As a
result, the fall of the Allada capital in 1724 is the epilogue to a long period
of tension involving successive wars which considerably weakened the
Kingdom. As Law states:
The Kingdom of Arda [Allada] was reduced, both in extent and power, by the revolt and
separation of several provinces, before it was attacked by the Dahomans, under Guadja
Trudo. (Law 1997: 112)
Table 2.1 Slave import from the Bight of Benin
1680 1690 1700 1710 1720 1730 1740 1750 1660 1670 1780 1790
1689 1699 1709 1719 1729 1739 1749 1759 1669 1779 1789 1803
SC % 39.4 42.8 70.6 69.0 28.1 31.8 2.2 0 0.9 0 0 0

GC% 3.3 0 8.3 0 65.3 53.7 7.4 20.0 16.9 19.4 26.6 23.3
SC = Slave Coast, GC = Gold Coast (Source: Arends 1995b: 243).
It is not clear to what extent this critical period and the subsequent fall of
Allada affected the slave trade generally. For instance, the English factor
Bulfinch Lamb, who was made prisoner during the conquest of Allada by the
Fon of Dànxòmὲ, reports that “there were more than eight thousand enslaved
prisoners of war” (cited in Pazzi 1979: 246). There is little doubt that these
prisoners were sold to European factors trading on the coast. These facts appear
compatible with the figures in Table 2.1 taken from Arends’ (1995b: 243)
estimations of Dutch slave imports to Suriname. This table suggests that while
the trade apparently flourished until 1740 with a peak period between 1700 and
1719, almost no slaves were bought on the Slave Coast (SC) until 1740 when
the Gold Coast (GC) became the major source of slaves.
The peak in slave numbers between 1700 and 1719 is important because it is
during this period that Allada had to face internal divisions often leading to
wars and roadblocks. In addition, the political crisis within Allada could have
forced the trade to shift to the Gold Coast, thus allowing some of the slaves of
Allada to be sold on the Gold Coast, which accounts for the flourishing of the
trade there by 1729 (see Pazzi 1979; Law 1994, 1997). With this in mind, we
can now ask: How structured was the supply of slaves and where did they
originate?
2.1.3 Supply of slaves and the slaves’ origins

The question of the supply of slaves and the slaves’ origins is one of the most
difficult to address in creolistics, because the records do not always explicitly
mention where the slaves originated. Most records mention the place of
embarkation of the slaves (Chaudenson 1992, 2001), but this often coincides
with the place where Europeans bought their human cargos. Usually, we have
no indication of the place where the enslaved Africans came from or where they
were captured and/or bought inland.
In this book, I inferred the origin of the enslaved Africans sold on the Slave
Coast through the ports under the control of Allada by looking at the slave
supply of this Kingdom. As noted above, it has been commonly assumed that
the slave trade on the African coast was the result of uncontrolled private
ventures. Indeed, there must have been slave raiders as well as smugglers
selling slaves through the different ports of Allada. The following letter from
W. Bosman, a trader of the Dutch West India Company (WIC) cited in Pazzi
(1979: 67, my translation) is illustrative of such slave raiders.
A description of Little Popo, it is an infertile land: its inhabitants, some of whom were
chased away from Accra, and are good soldiers, their war with those of Offra /Jaken/ and
Fida /Whydah/ and what happened afterward. They too live on loot and slave
trafficking.18
Given the ‘illegal’ practices of the Dutch WIC on the West African coast
(Pazzi 1979; Law 1991, 1994, 1997), we can infer from this quote that a
fraction of the overall exported Africans from Allada and its related ports
was provided by slave raiders who were not necessarily under the control of
the central power in Allada. However, historical documents on this kingdom
suggest that much of the trade was strictly controlled (if not organized) by the
State. Because of the State central role in the trade, there exist some indications
about the origins of the slaves. Given the tensed political situation described
above and the Fon expansion that led to the fall of Allada, it is obvious that
these troubled times must have yielded a lot of enslaved prisoners. For instance,
Law (1994: 83) asserts that “in general, the evidence suggests that the principal
source of slaves for export was capture in warfare.” Because Allada was often
at war with its own subordinate states, many of which derived from the
ancestral royal town of Tado, we can assume the following:
– Many of the slaves sold on the coast came from further inland, as a con-
sequence of internal warfare.
– The slaves must have been primarily from the same Aja-Tado culture.
Earlier documents cited by Pazzi (1979) and Law (1994, 1997) suggest that
the slaves traded in Allada came from two zones: Yoruba lands (probably Oyo)
and Dànxòmὲ . As one reads in Law (1994: 83),
In the early Dutch trade, in the 1630s or early 1640s, it was reported that many slaves
were supplied for sale at Allada from the kingdom of “Lukumi” or Yoruba or even
further in the interior.
Olfert Dapper, who reported about Allada of the mid seventeenth century,
also suggests that “the slaves sold in Allada were brought from the kingdom of
Ulkami, i.e., Lukumi” (Law 1997: 101). Other documents indicate that the
slaves came from, or through, Dànxòmὲ. Indeed, a document dating from
March 1680 and cited by Law (1994: 83) suggests that “the greatest part of
18
“Une description de Petit Popo, c’est un pays infertile. De ses habitants, dont certains ont été
chassés d’Accra, et sont de bons soldats. Leur guerre avec ceux d’Offra et Fida, ce qui s’est
passé après. Eux aussi vivent de butin et de traffic des esclaves” (Pazzi 1979: 64).
the slaves sold in Allada came from Dahomey.” Whether Dànxòmὲ , which was
subordinated to Allada until 1724, directly supplied the slaves to the coast or
allowed slave caravans from the north to cross its lands and reach the coast is
not clear. This is so even though Law (1997: 101–102) reports that “an increase
in the price of slaves at Whydah in 1688 was explained by reference to certain
differences with the King of Fouin [Fon], which had interrupted the supply
from the interior.” The question also arises with regard to other satellites of
Allada such as Whydah and Jakin for which it is not clear whether they were
direct providers of slaves. Though more research is needed here, the geopolitics
of the time which favored internal competition between the satellite states and
allowed various alliances supports the hypothesis that many of the slave
cohorts must have come from the interior of Allada, within the frontiers of
the kingdom (see Map 2.3), as a result of warfare. Various documents (and
citations) suggest that selling people from the interior must have been a
common practice throughout the seventeenth and the eighteenth centuries.
Law (1997: 104) reports that according to William Bosman, in the 1690s,
“most of the slaves that are offered to us are prisoners of war, which are sold
by the victors as their booty.” The author further mentions a document dating
from the 1720s where it is reported that:
The slaves exported through Whydah were captured in wars close to the coast, alleging
that, if the king of Whydah could not get enough slaves by “fair agreements with his
country neighbours”, he marches an army, and depopulates, and observing He and the
King of Ardra adjoining, commit great depredations inland. (Law 1997: 104–105).
Such practices must have continued after the Fon conquered Allada in 1724.
Bulfinch Lamb, the English trader who was in Allada when it was conquered by
the Fon in 1724 indicates that there were about 8000 prisoners (Pazzi 1979:
247). Clearly, these were soldiers of the defeated kingdom or inhabitants of the
capital city of Allada. As Pazzi (1979: 257) further explains, Agaja (i.e.,
Trudo), who conquered Allada, suffered a severe defeat by the Oyo in the
1730s and had to pay a yearly tribute to Oyo. Pazzi argues that such a payment
arrangement was possible thanks to the slave trade in Whydah (then under the
control of Dànxòmὲ). Raids on the Maxi people living north of Agbomey, the
capital city of Dànxòmὲ, were instrumental in keeping the trade going.
All these facts suggest that a significant portion of the slaves sold through the
ports of Allada came from within the Kingdom. With regard to the Slave Coast,
Arends (1995b: 249) argues that the maximum distance between the areas
where the enslaved Africans were captured (or bought) and the location where
they embarked is about 200 km. This puts the zone of ransack within the
frontiers of Dànxòmὲ, a former satellite of Allada.
The geopolitics of the Allada Kingdom therefore suggests that a significant
part of the slaves sold through its ports were of Aja origin. Various factors
support this view. For instance, Captain François d’Elbée who visited Allada in
1670, reported that:
These slaves are of several sorts, some are prisoners of war, others contributed by
neighbouring kingdoms, and these are foreigners; in addition to which there are the
slaves born in his country, and those who are so through having incurred legal
punishments . . . there are also some of them who are condemned to slavery for failing
to pay their debts, and whom their creditors have sold for their payments. (cited in Law
1997: 104)
Though not explicitly stated in this quote, it appears from previous discussion
and the history of Allada that the prisoners of war must be of Aja descent owing
either to the Fon expansion or to internal conflicts between the different
chieftaincies or between the chieftaincies and the central power. Likewise,
Captain François d’Elbée reported that the Queen of Allada sold him eight of
her co-wives as slaves. These women were presumably of Aja descent. The
idea that a significant number of slaves sold on the markets of Allada must have
been of Aja descent is compatible with the following document cited in Pazzi
(1979: 66, my translation) which mentions the ethnicities of the slaves:
Van Hoolwerff had wounded three blacks, but not to death. Then those of Arda, after
surveilling him for seven or eight days attacked him and took him to the Fidalgo who
had his hands and feet tied and removed all the Arda slaves and Mina slaves from the
factory.19
While the Arda (Allada) people directly relate to the Aja-Tado migration, the
Mina include both Fanti-Anε from El Mina, and the Ga from Accra. These
people migrated to the Aja-Tado area where they merged into a single ethnic
group with its own language currently referred to as Gengbe. It is noteworthy
that the author of this report was aware of the ethnicity of the slaves in his
factory. Contrary to common beliefs, this is partial evidence that European
factors on the coast had good knowledge of local customs and the different
ethnicities they dealt with. Assuming that language is a reliable ethnicity
marker, we can infer that some of these factors also knew the local
languages, a fact that I will return to later.
Focusing on the ethnicity of the slaves, the structure of the trade as imposed
by the King of Allada represents an important factor which indirectly restricted
the slave supply to what we could refer to as a ‘local reservoir’ (i.e., mainly
people of Aja descent). Indeed, one reason why the Slave Coast provided a far
greater number of slaves (i.e., through Allada and its ports) than the Gold
Coast, for instance, has to do with its institutions and the authority of the State,
19
“Van Hoolwerff avait blesse trois noirs, mais non à mort. Alors ceux d’Arda, après l’avoir
surveillé pendant sept ou huit jours, l’ont attaqué et emmené chez le Fidalgo, qui le fit lier mains
et pieds et fit enlever tous les esclaves Arda et Myna qui étaient à la factorerie” (Pazzi 1979: 66).
which Eltis (2011: 274) argues, is “necessary not so much for enslavement as
for providing orderly markets and debt-collection institutions at the point
where the enslaved Africans became the property of Europeans.” In the case
of Allada, the slave transactions were organized on the ports only after proper
authorization from the King. Law (1997) provides ample evidence of the royal
control over this trade. As discussed there, historical documents indicate that
trade in Allada was possible on the sole condition that the Europeans factors
first traded with the King or his courtiers. In this regard, a Dutch report
mentioned that “you cannot conduct trade with anyone else before first
meeting the King” (Law 1997: 95). Other European reports (e.g., Captain
François d’Elbée) suggest that traders were required to first purchase their
slaves from the King and his entourage (including local chiefs) before trading
freely on the coast. This is clearly illustrated by the following report cited in
Law (1997: 100):
The King of Ardrah has obliged the Portugueze at Jacquin to pay him 35 sletias per slave
for custom, whereas it used to be but 17, he has likewise made Captain Heartsease buy
20 slaves off him (which are good for nothing) before he can mark one off anybody else,
and demands all the booges [i.e., cowries] for himself. But what’s worse talks of
obligeing him and all Captains that shall come there for the future to go to Ardrah and
there stay and make their trade.
It appears from this document that not only did the King require to be the first
contact of the Europeans, but he also levied taxes on the slaves. It is likely that a
similar tax applied to the African sellers as well (Law 1997: 97). Given that the
capital city of Allada is about 50 km inland, the requirement that Europeans had
to trade first with the King also implies that they first had to travel inland before
coming back on the coast to trade. This is possible because the King and his
entourage controlled the roads to the interior and could effectively block slave
supplies from the interior in period of crisis. (See e.g., Hoolwerf’s letter cited
above.) We can therefore conclude that the State of Allada monitored both the
trade on the coast and the supply of slaves. Consequently, this Kingdom more
than any other on the Slave Coast “provided an ordered environment and a set
of rules for carrying on business” (Eltis 2011: 275). A report from the 1720s
describes the subjects of the King of Allada in the following terms: “all of them
obsequious slaves, who dare not sell or buy anything without license” (Law
1997: 95).
I submit that an indirect consequence of this “ordered environment” is that
slave traders, Africans and Europeans alike, could not afford to fetch their
slaves beyond the limits of the Kingdom. Recall that the Kingdom of Allada is
sandwiched between Oyo and its slave ports to the east and Ƞɔtse to the west,
where the Gold Coast lies. Both zones are heavily involved in slave trade.
North of Allada was the competitor and later conqueror Kingdom of Dànxòmὲ.
Beyond Dànxòmὲ there were various Muslim kingdoms, which were less
favorable to slave trade (Law 2011). Given this geo-political situation,
organizing slave caravans without a “license” from outside the limits set by
the Kingdom exposed traders to severe sanctions from the King,
notwithstanding any other perils they might have faced en route. But even if
the traders could possibly still organize slave caravans, such an economic
enterprise would not have been appealing. It is unlikely that Allada was the
only kingdom in the region which laid taxes on slave trade and controlled
the slave routes. Accordingly, bringing back slaves from beyond the borders of
the Kingdom would have unnecessarily increased the risks and costs of
transporting the human cargoes. Prices would have become too high and the
commerce less lucrative.
According to Law (1994, 1997), the slave trade flourished in Allada partly
because prices were lower there than in competing neighboring kingdoms. My
contention is therefore that a significant portion of the people sold in Allada
were of Aja descent. In addition to warfare and political instability in the
interior, economic factors (e.g., the cost of a slave) contributed to making
unfortunate locals the prime targets and victims of the trade.
2.1.4 The origins of the slaves: some notes on their cultures and languages
It is important to remember that, compared to other regions in West Africa
involved in the slave trade (e.g., Gold Coast), “more people left the Bight of
Benin than any other region except West Central Africa” (Eltis 2011: 275).
Given the prominent role of Allada in this trade, this would mean that a
significant number of the slaves taken from the Bight of Benin were of Aja
descent, as noted above. For instance, Eltis (2011: 276) claims that “the
majority of slaves before 1725 would have been Gbe speakers.” The author
further suggests that the slaves embarked in Allada after this period were
probably “a mix of Gbe-speaking and Yoruba people” (Etis 2011: 276). In
the context of our discussion, this is an important finding. Even though West
Central Africa appears to have been the primary supplier of slaves throughout
the slave trade period (viz., during the seventeenth and eighteenth centuries),
the burst of Allada in this trade, the astonishing growth of slave exports (e.g.,
from 1,000 to 19,000 according to Eltis 2011), and the structure of its slave
economy suggests that large numbers of Aja people must have been shipped
together to the Americas. This conclusion is contrary to the common
assumption that the slaves came from diverse places in Africa and were
dispatched to diverse places in the Americas. A closer look at the political
and economic structures underlying the slave trade in different West African
locations (viz., Upper Guinea Coast, Gold Coast, Bight of Benin/Slave Coast,
Bight of Biafra, and West Central Africa) suggests otherwise. In the case at
hand, it is highly probable that significant numbers of people of Aja descent

shipped from Allada eventually ended up in the same colonies and apparently
also on the same plantations, in some cases. Clearly, this will be an important
factor bearing on the nature of language contacts and the socio-historical
ecologies that produced the so-called creoles.
In the context of this book, for instance, because people of Aja descent were
numerically not necessarily dominant throughout the colonization of Haiti and
Suriname (Debien 1974; Singler 1996; Smith 1987), the question often arises
of why they would have had a more significant impact on the development of
the cultures and languages of these colonies rather than the Bantu people who
were also present at about the same time? Our findings on Allada’s slave trade
help us provide a reasonable answer to this question. Slaves of Aja descent
came in more compact groups and during a short period of time, roughly from
the mid seventeenth to the mid eighteenth centuries. In addition, because of the
way the slave trade was organized in Allada, I agree with Arends (1995b) that
these groups were probably more uniform ethnolinguistically (see, e.g., Diouf
2007 for some discussion). This is different from the Bantu slaves coming from
West Central Africa and speaking many more different languages. Contrary to
Allada, the Portuguese slave traders controlled the slave routes there and could
fetch slaves inland up to 700 miles from the coast (Eltis 2011: 280). These long
distances may have generated relative ethnolinguistic diversity: the slaves
came in large numbers but in fragmented ethnic groups.20 Such a diversity
coupled with different periods of arrival may deeply impact the ecology of the
founder population thus affecting the dynamics of contact that led to the
emergence of creoles. I therefore consider differences in time of arrival and
in critical masses at the time of the emergence of creoles as more significant
factors in the competition between the different languages families (Indo-
European, Kwa, Bantu) that came in contact on plantation colonies of the
Americas.
Once we take this perspective, we understand better why in the case of Haiti
and Suriname the slaves of Aja descent, speakers of Gbe languages (genetically
and typologically related), had such a significant impact on the emergent
cultures and languages of Haiti and Suriname. Yet, before we make any
definite conclusion, we need to see how homogeneous the Aja-Tado culture
is and how people of Aja descent relate to other Kwa language-speaking
communities of West Africa. In this regard, I will list a few cultural traits that
illustrate some ethnic ties between people of Aja descent and other
20
This of course need not mean that under other ecological circumstances, the enslaved Africans
from West Central Bantu would not have a greater impact than the Kwa as well as the Aja
descent people (see, for instance, Thornton and Heywood 2007). It is precisely because of such
variations depending on the ecology of the colony that more work is needed in order to
determine the contributions of the different ethnic groups that came in contact in the Americas.
communities of the Slave Coast such as the Yoruba also commonly referred to
as the Nagos.
Some cultural traits of the Aja In previous sections, I argue that people of Aja
descent originally migrated from Oyo, a Yoruba land (Pazzi 1979, 1984, and
much related work). We may assume that there has been sustained contact
between these two ethnic groups. Indeed, various cultural traits are shared by
these communities, for instance, the Aja cult of Afa, which is a variant of the Ifa
cult among the Yoruba. As reported by Pazzi (1979: 50), the bokɔ-jìsá tradition
considers Ifε to be the spiritual center of the Afa as well as the place where
spirits will return after death, thus directly relating the Aja to Oyo. The Aja and
the Yoruba share a weekly calendar of four days for rituals to the ancestors as
well as for agricultural purposes, although the Aja have also adopted an
additional seven-day calendar for celebrating birth days. Finally, both the Aja
and the Yoruba practice circumcision (Pazzi 1979). All these traits point to a
common cultural heritage from Oyo, which probably also involves a common
linguistic heritage.
Other cultural traits of the Aja include the cults of Vòdú, of the royal
python Dàngbúi and of the related python Dàn, and of lightning Sò. These
cults (and variants thereof) are found across communities of Aja descent,
many of which have distinctive facial scarifications often identifying them
with their religion. Yet, another cultural aspect that is shared by people of
this ethnicity (and that was maintained in the Caribbean) is their common
naming system described below. As is well known in creolistics, children
born in this cultural area systematically get a day name, which varies
according to gender.
Female Male
Monday Àjó/Ajówá Kwàjó, Kɔ̀ jó

Tuesday Àblá Kɔ̀ mlán
Wednesday Àkú, Àkúwá Kwàkú, Kɔ̀ kú
Thursday Àyábá Yàwó
Friday Àfí, Àfíwá Kɔ̀ jí
Saturday Àmà Kwàmì, Kɔ̀ mì
Sunday Àsíbá, Àwísí, Kɔ̀ síwá Kwàsí, Kɔ̀ sí
It is clear from this brief description that although the Aja share some core
cultural practices that make their ethnic group relatively homogeneous and
distinct from other groups, they also share significant cultural practices with
other Kwa people (e.g., Yoruba). In the context of this book, it seems to be
justified to assume that such cultural homogeneity within the Aja group and
kinship with other Kwa people must have influenced significantly the
emergent creole cultures and languages. Such cultural and linguistic affinities
must have motivated these enslaved Africans to interact whenever possible
with one another and learn from one another while adjusting to their new
societies. An aspect of such interaction that is important in our study regards
language.
Some Aspects of Aja languages People of Aja descent speak the Gbe
languages. Speakers of these languages currently live in West Africa, in the
southern part of the Volta region in Ghana, the southern part of Togo, the
southern part of Benin, and in different localities of Ogun State and Lagos State
in Nigeria. Capo’s (1991) study on comparative phonology suggests that the
Gbe languages form a continuum from Nigeria to Ghana and involve five main
branches: Ewe, Gen, Aja, Fon, and Phla-Phéra, as illustrated below (Capo
1991: 15):
Kluge (2005), who conducted a lexical comparative analysis of 49 Gbe

varieties, suggests a slightly different classification. According to her, there
appears to be three major Gbe clusters: (1) Western Gbe, comprising the
Ewe and Gen sub-clusters; (2) Central Gbe, including the Aja varieties; and
(3) Eastern Gbe, consisting of the Fon and the Western and Eastern Phla-
Phera sub-clusters (see also Kluge 2000). It appears from Kluge’s (2000)
classification that even though one may establish relatively coherent sub-
groups across Gbe, these clusters overlap significantly. Indeed, although the
Gbe languages show subtle morphosyntactic differences (see Aboh 2004a;
Aboh and Essegbey 2010), they share many lexical roots, as shown on the
Table 2.2 A random wordlist in fifteen Gbe varieties selected according to

Capo’s (1991) five major Gbe clusters
Clusters Varieties Eye Three Red Eat Drink Push Pull
EWE Adan ŋku etɔ dzi ɖu no tutu he

Be ŋku etɔ jiε ɖu nɔ-nu tutu dɔ
Waci ŋku-vi εtɔ ja ɖu nu tutu dɔ
GEN Agoi (Glidji) ŋku-vi etɔ ja ɖu nu tutu dɔ
Anexo ŋku etɔ jε ɖu nu tutu dɔ
Gen ŋku-vi etɔ ja ɖu nu tutu dɔ
AJA Dogbo nku-vi ama-tɔ eji ɖu nu cuicui dui
Hwe (Azovè) ŋku-vi ama-tɔ ju ɖu nu cucu dɔ
Sikpi nku ɔtɔ ovε ɖu nu cuicui dui
PHLA-PHERA Alada nuku atɔ vεvε ɖu nu sisε dɔ
Ayizo ŋku-vi atɔ vεε ɖu nu sisε dɔ
Tofin nuku ɔtɔ fɔfɔ ɖu nɔ-nu sisε dɔ
FON Gun nuku atɔ vε ɖu nu sisε dɔ
Fon nuku atɔ vɔvɔ ɖu nu sisε dɔ
Maxi nuku otɔ vɔvɔ ɖu nu tu ɖɔ
short random list (adapted from Kluge 2000) below. For the purpose of
clarity, tone specifications as well as nasality were not included.21
Because the Gbe languages are isolating, they lack the typical noun class
systems found in Bantu. As a result, the roots listed in Table 2.2 are usually found
as is (i.e., with no other morphological modification) in sentences. While nouns
may involve more than one syllable, verbs are generally monosyllabic, except
when they derive from reduplication. In this regard, we observe that contrary to
the verbs ‘eat,’ ‘drink,’ and ‘pull,’ which are all monosyllabic and usually involve
the same roots in these languages, the verb ‘push’ is disyllabic: it derived from
reduplication of the verb roots tu/cui or sε which can be roughly translated as
‘displace.’ That all the varieties resort to the same word-formation process (viz.,
reduplication) in very specific cases like this one is indicative of the tight
relationship between these languages. In her lexical comparison, Kluge (2005:
22) concluded that lexical similarity of the entire Gbe cluster ranges between 64
percent and 73 percent, depending on the method of calculation.
The same pattern is observed in syntax. Here again, the Gbe languages show
very similar patterns summarized below using word order patterns in both the
noun phrase and the clause. Other aspects are detailed in Aboh (2004a) and
Aboh and Essegbey (2010). The data presented here are primarily based on
Gungbe, but when necessary I draw on data from other Gbe languages as well.
21
For a more representative list, the reader is referred to Kluge (2000, 2005). Kluge (2000)
provides a list of a hundred words based on the Swadesh list.
Gbe languages generally display the order of noun phrases in (1a)

schematized in (1b) where the symbol “>” indicates precedence:
(1) a. kòkló ɖàxó àwè énὲ lɔ́ lέ [Gungbe]
chicken big two DEM DET PL
‘those two big chickens’
b. Noun>adjective>numeral>demonstrative>determiner>Number marker
In most Gbe, a relative clause can precede or follow the demonstrative as
illustrated in (2a–b). This leads to the general schema in (2c–d) for the noun
phrase.
(2) a. Kòkló ɖàxó àwè énὲ [ɖě mí xɔ̀ ] lɔ́ lέ. [Gungbe]
chicken big two DEM REL 1PL buy DET PL
‘Those two big chickens that we bought.’
b. Kòkló ɖàxó àwè [ɖě mí xɔ̀ ] énὲ lɔ́ lέ.
chicken big two REL 1PL BUY DEM DET PL
‘Those two big chickens that we bought.’
c. Noun>adjective>numeral>demonstrative>relative clause>determiner>Number
marker
d. Noun>adjective>numeral>relative clause>demonstrative >determiner>Number
marker.
As we can see from these examples, these languages do not typically exhibit
inflectional morphology, hence the absence of agreement between the noun and
its modifiers. Finally, all Gbe languages allow the use of bare nouns in all
contexts (see Aboh 2004a, 2010a; Aboh and DeGraff 2014). Accordingly,
sentence (3) is perfect in Gungbe. In such a sentence, the bare noun can be
interpreted as definite, indefinite or generic, depending on context.
(3) Mí xɔ̀ Kòkló [Gungbe]
1PL buy chicken
a. ‘We bought a chicken.’

b. ‘We bought the chicken.’
c. ‘We bought chickens.’
Possessive constructions generally come in two types: Possessor>Possessed
sequences as in (4a) versus Possessed>Possessor sequences as in (4b).
(4) a. Áyɔ̀ sín kòkló [Gungbe]
Ayo POSS chicken
‘Ayo’s chicken’
b. kòkló Áyɔ̀ tɔ̀ n
chicken Ayo POSS
‘Ayo’s chicken’
While these two options are found across Gbe, the languages differ regarding
whether the construction is sensitive to features such as [human], [animate],
[alienable], or whether the possessive markers are overtly realized or not.
Finally, most Gbe languages involve both a set of prepositions and
postnominal locative elements. The combination of these two leads to
complex locative expressions such as in (5), discussed in Ameka (2003) and
Aboh (2005a, 2010a, 2010b):
(5) Áyɔ̀ zé sìn sɔ́ n távò jí. [Gungbe]
Ayo take water PREP table ON/ABOVE
‘Ayo took (the) water from the table.’
With regard to the clause, the Gbe languages are generally assumed to be of
the SVO type. Indeed a simple declarative sentence in Gungbe can be
illustrated as in (6a), in which the aspect marker precedes the verb. Note,
however, from example (6b) that the equivalent sentence in Gengbe displays
a post-verbal aspect marker. Put together, we get the variation in (6c) which
again derives from language specificities as to which aspect marker may/must
precede and which may/must follow, as discussed in Aboh (2003a, 2004a,
2009b) and Aboh and Essegbey (2010):
(6) a. Mí nɔ̀ xɔ̀ kòkló. [Gungbe]
1PL HAB buy chicken
‘We usually bought/used to buy chicken(s).’
b. Mí plé-nà kòkló. [Gengbe]
1PL buy-HAB chicken
‘We usually bought/ used to buy chicken(s).’
c. Subject>(Aspect)>Verb>(Aspect)>Object>Adjunct
Another context where one finds variation across Gbe is the expression of
negation. As discussed in Aboh (2004a, 2010c), one finds three basic patterns
across Gbe: preverbal negation as in Gungbe (7a), sentence-final negation as in
Fongbe (7b), and simultaneous preverbal and sentence-final negation as in
Gengbe (7c).
(7) a. Mí má xɔ̀ kòkló. [Gungbe]
1PL NEG buy chicken
‘We did not buy any chicken.’
b. Mí xɔ̀ kòkló á. [Fongbe]
1PL buy chicken NEG
c. Mí mú xɔ̀ kòkló ò. [Gengbe]
1PL NEG buy chicken NEG
With regard to word order variation, it is important to observe that all Gbe
languages also display SOV order in very specific contexts. A case in point is
the expression of progressive aspect which requires the object to precede the
main verb. In addition, such OV constructions typically include a post-verbal
particle (Aboh 2003a, 2004a, 2009b; Aboh and Essegbey 2010).
(8) Mí ɖò kòkló xɔ̀ wὲ [Fongbe]
1PL ASP chicken buy PCL
‘We are buying (a) chicken.’
Such OV orders can be related to serial verb constructions (Aboh 2003a,
2009b), which are found across Gbe. An example of the instrument serials is
given in (9).
(9) Mí sà kòkló ɖòkpó ɖù sɔ̀ [Gungbe]
1PL sell chicken one eat yesterday
‘We sold one chicken yesterday.’
Word order variation also arises in interrogative (10a), focalized (10b), and
topicalized (10b) sentences:
(10) a. Étέ wὲ mí xɔ̀ ? [Gungbe]
What FOC 1PL buy
‘What did we buy?’
b. Kòkló wὲ mí xɔ̀ .
chicken FOC 1PL buy
‘We bought a chicken.’ (It’s a chicken that we bought.)
c. Kòkló lɔ́ yà mí xɔ̀ ὲ.
chicken DET TOP 1PL buy 3SG
‘As for the chicken, we bought it.’ (The chicken, we bought it)
In all these cases, the relevant category must front to the left edge of the
sentence, where it is typically bounded with a specific interrogative, focus, or
topic marker. As Aboh (2004a, 2004b, forthcoming) shows, these markers have
a very specific syntax that impacts discourse semantics in Gbe. For instance,
most Gbe languages exhibit verb focus constructions, known in creolistics and
generative syntax as ‘Verb/Predicate Clefting,’ in which the verb in the focus
position is repeated in its regular position in the rest of the sentence, as
illustrated by the Gungbe example in (11a). The Ewegbe example in (11b)
indicates that in some of these languages, the fronted verb must be reduplicated
(Ameka 1992, 2010).
(11) a. Xò wὲ yè xò è. [Gungbe]
beat FOC 3PL beat 3SG
‘They beat him/her.’
b. ɸo-ɸo é wò ɸo é. [Ewegbe]
RED-beat FOC 3SG beat 3SG
‘S/he beat him/her.’ [Adapted from Ewegbe (Kwa); Ameka 1992: 12]
The properties presented here are summarized in the following table
showing word order patterns in Gbe.
Two questions arise from this brief description of the Gbe languages: (i) How
did these languages relate to the larger family of the Kwa languages of West
Africa? And (ii) How do the synchronic aspects of these languages as described
in Table 2.3 relate to Gbe varieties of the seventeenth and eighteenth centuries?
Gbe in the larger context of Kwa languages The first question is easy to
answer because the Gbe languages display morphosyntactic patterns that are
largely shared by other Kwa languages. Most Kwa languages have isolating
morphosyntax, although some languages show some degree of agglutination.
In any case, almost all the syntactic patterns in Table 2.3 are found across Kwa
languages (Aboh and Essegbey 2010). Kwa languages typically display focus
constructions of the type described in Gungbe, in which non-verbal elements
must be fronted to some clause peripheral position leaving a gap inside the
Table 2.3 General word patterns in Gbe
Word order patterns in Gbe
Noun phrases a. Noun>adjective>numeral>demonstrative>relative

clause>determiner>Number marker
b. Noun>adjective>numeral>relative
clause>demonstrative>determiner>Number marker
Locative expressions Preposition>Noun>Postnominal element
Tense and aspect a. Subject>Tense>(Aspect)>Verb>(Aspect)>Object>Adjunct
specifications in clauses b. Subject>Tense>(Aspect)>(Aspect)
>Object>Adjunct>Verb>Particle
Negation in clauses a. Subject>Negation>Verb>Object
b. Subject>Verb> Object>Negation
c. Subject>Negation>Verb>Object>Negation
Interrogation WHPHRASE>Focus marker>[IP . . . . . . <gap> . . . ], content
questions
Subject>(TMA)>Verb>(XP)>Qparticle, yes–no questions
Focus a. XP>Focus marker>[IP . . . . . . <gap> . . . ]➔ non-verbal focus
b. Verb>(Focus marker)>[IP . . . . . . Verb . . . ]➔ verbal focus
c. Verbred>(Focus marker)>[IP . . . . . . Verb . . . ]➔ verbal focus
Topic XP>(Topic marker)>[IP . . . . . . <gap> . . . ]
Serial verb construction Subject>TMA>V1>Object>V2>Adjunct
clause (as in 10b). Such focus constructions play a role in the formation of wh-
questions which show the same pattern (10a). On the other hand, verbal focus
typically results in a doubling strategy where the focused verb is fronted but
another token is found in its canonical position (11). These properties are not
unique to Kwa (see Koopman 1984; Mufwene 1987; Manfredi 1993; Aboh
2006c, 2007a, forthcoming; Aboh, Hartmann, and Zimmermann 2007, Aboh
and Dyakonova 2009), but they are found robustly in these languages.
Serial verb constructions are yet another salient peculiarity of this language
family. The same holds true of the periphrastic TMA marking and its
implications for word order (Aboh and Essegbey 2010). Indeed, a significant
number of Kwa languages of the Bight of Benin display a VO versus OV
alternation that is sensitive to aspect licensing (Déchaine and Manfredi 1997).
All these points lead to the conclusion that, morphosyntactically, Kwa
languages form a typologically homogeneous group. This means that an
enslaved African of Aja descent speaks a language that is typologically very
close to the languages of the enslaved people of Nago or Edo descent.
Accordingly, though it is often assumed that the slaves were of too diverse
origins, this might not be true when considering their linguistic profile and the
typological relations between the languages. Based on this, and taking into
account the geopolitics of the Slave Coast, it appears that while it might be
difficult for an enslaved person of Aja, Nago, or Edo descent to acquire a
Central Bantu language readily, this must not have been the case when such
enslaved people were confronted with a neighboring Kwa language.
Indeed, traditional considerations of linguistic diversity based on the vast
and diverse area from which the slaves were imported do not do justice to the
typological homogeneity of some subareas (and to the spread of some areal
features in sub-Saharan Africa). Such structural kinship suggests that in a
number of ways there were larger proportions of Africans sharing structural
properties than suggested by geography-based demographics alone. Also, it is
not excluded that the Kwa languages shared some of the same properties with
Bantu languages, such as the position of demonstratives cum determiners and
some nominal modifiers, verb compounding or light verb constructions which
appear to be similar to Kwa serial verb constructions. This, is an important
factor to keep in mind when it comes to the intra group communicative
strategies that the enslaved Africans might have adopted as well as the
learning strategies they might have developed in their attempts to acquire
European languages.
Indeed, the above specifics about how and where the slave trade was
conducted in West Africa suggest that the impact of speakers of Kwa
languages on the emergence of Haitian, Sranan, and Saramaccan could be
explained by the fact that the enslaved Kwa speakers might have entertained
the same learning hypotheses while acquiring the European colonial
languages. Such learning hypotheses could have been enhanced and

maintained by successive waves of enslaved Kwa speakers, which, as noted
above, came in more coherent groups during the early stages of the formation
of these creoles.
Modern Gbe versus ‘Old Gbe’ The high rate of lexical similarities among
current Gbe languages and their typological closeness to other Kwa languages
helps us address the vexing issue of the relation between what we can refer to as
a Modern Gbe language, say Gungbe, and its ancestor spoken in Allada in the
seventeenth century. Indeed, we can infer from the current strong similarities
among Gbe and between Gbe and other Kwa languages that whatever changes
they may have undergone over the past three hundred years or so are probably
not significant enough to dispute the assumption that they were just as related
during the time of the slave trade. In this regard, we have access to two old
documents only.
Using La Doctrina Christina (1658) and La Grammaire Abrégée (1730)
as references and framework, let’s consider certain aspects of seventeenth-
and eighteenth-century Gbe. La Doctrina Christina is a catechism published
by Spanish priests sent to Allada in the context of cooperation between the
King of Allada and the King of Spain (see Section 2.1). Because of the
obscure transcription system used by the Spanish monks to transcribe this
Gbe language, it is not yet possible to reconstruct all the lexical forms that
appear in this document. Yet, a cursory look reveals strong similarities
between that variety and Modern Gbe languages. For instance, three
words in Table 2.2 are transcribed in the document. In my presentation
below, the first line is the Spanish text followed by the Gbe translation as
written by the missionaries including the relevant lexical item in boldface.
The third line provides the same item in Gungbe preceding its translation in
parentheses.
(12) a. Buelue tus ojos misericordiosos [La Salve]
Dio nucu-bo-mi nunc-mito
nùkún (eye: Gungbe)
b. Los tres primeros pertenecen al honor de Dios [Los Mandamientos de la Ley
aton juegue teegue nudajoguana Vodu de Dios]
àtɔ̀ n (three: Gungbe)
c. La primera dar de comer al hãbrieto [Las obras de Misericordia]
Uguegue na-nu-ie-ru jobono
ɖù (eat: Gungbe)
Other lexical items found in this text include dasi ‘ɖàsìn: tears’, suru ‘súnù:
male person’; nubo ‘núgbó: true/truth’ as shown in E Vodu suru nubo the
translation of the Spanish text Dios y hombre verdadero.22 We can see from the
Gungbe equivalents of these words that they haven’t changed much. In a
related vein, a close look at the translations in (12) provides us with
significant information about the language. In the translation of (12a) we see
the italicized words bo, which corresponds to Gungbe verb bɔ̀ meaning ‘to
help’ or ‘to forgive.’ Next we see the first-person plural weak pronoun mi (mí in
Gungbe) as well as its possessive equivalent mito (mítɔ̀ n in Gungbe). The latter
is even more informative as it shows that in both seventeenth-century and
Modern Gbe, possessive pronouns are formed by combining a weak pronoun
and the possessive marker illustrated in (4b).
The translation of example (12c) is also interesting. Here we see the lexical
verb na (ná ‘to give’ in Gungbe) in a serial verb construction including the verb
ru (ɖù ‘eat’ in Gungbe). In this construction, the so-called ‘shared object’ is
nu-ie (nú-è, /thing this/ ‘this/the thing’). As we can see in these sequences the
noun precedes the determiner, just as it does in Modern Gbe. Finally, another
aspect of Gbe that stands out in the manuscript involves the use of a focus
marker in (13a–b). Though the transcription is not always systematic, the fact
that the Spanish authors transcribed the focus marker è with an accent indicates
that they understood its function as marker of emphasis.
(13) a. El tercero creer que es Hijo [Los Articulos de la Fè]
b. atoe un-dique Vi-to-è
three 1SG-believe child-POSS-FOC
‘the third is THE CHILD’
The examples in (14a–b) indicate that the focus marker attaches to wh-
phrases in wh-questions:
(14) a. Quien es Dios? [Sobre la Dotrina Christiana]
Menu-e Vodu?
who-FOC Vodu
‘Who is Vodu?’
b. Como es Dios?
Anuniqu-e Vodu?
how-FOC Vodu
‘How is Vodu?’23
22
It is interesting to realize that in these early texts, the term vodu ‘voodoo’ was used as translation
for God. This is evidence that this term acquired its negative connotations later only when
missionaries on the Slave Coast or in Haiti decided to make a sharp distinction between their
‘God,’ the warrant of their “mission civilisatrice,” and the ‘god’ of the “primitive” Africans to
be brought to light.
23
This wh-phrase here could be alternatively segmented as ‘anuni-que’, where ‘que’ represents
the focus marker.
Further work is needed to establish precisely which Gbe variety served for
the translation of La Doctrina Christina, but the facts discussed here show
that significant grammatical aspects of Gbe have not changed since the
seventeenth century. The second historical document at our disposal, La
Grammaire Abrégée (GA), further confirms this conclusion. This annex to
Voyage du Chevalier des Marchais, reproduced in 1730, represents a
pedagogical grammar. This can be determined from the description
provided by the author and translated here:24
Grammaire abrégée, an interview in French and in the languages of the Negroes of
Whydah, it is very useful for those who trade Negroes in that Kingdom, and for ship
surgeons who can communicate with the Negroes when they are sick. It includes what
can serve as a small dictionary. (Cited in Labouret and Rivet 1929: 31)
This document displays a better transcription and provides us with significant

facts about Fongbe as spoken in Whydah in the early eighteenth century. Of the
seven words in Table 2.2, I found six common ones, listed below:
(15) Grammaire abrégée Gungbe Gloss
noucou nùkun eye
doü ɖù eat
nou nù drink
sa sὲ push
dé ɖè pull/shoot/take away
otton àtɔ̀ n Three
Again, the similarities between eighteenth-century Fongbe and twenty-first-
century Fongbe/Gungbe are obvious in this example. The document also
involves various examples of serial verb constructions and wh-questions.
Below are some illustrations. Sentence (16) exemplifies a serial verb
construction. The first line is the French sentence; the second gives the
proposed translation. The third line provides my Gungbe translation, which
in turn is translated in English on the fourth line.
(16) Aporte moi de l’eau.
Soƭi ou anam.
Sɔ́ sìn wá ná mì.25 [Fongbe/Gungbe]
take water come give 1SG
‘Bring me some water.’
24
“Grammaire abrégée, on entretien en Langue Françoise et celles des Negres de Juda, très-utile à
ceux qui font le commerce des Noirs dans ce Royaume, et pour les Chirurgiens des Vaisseaux
pour interroger les Noirs lorsqu’ils sont malades. Ce qui peut servir pour composer un petit
Dictionnaire” (Labouret and Rivet 1929: 31).
25
Note that anam packs together the sequence wá ná mì in (modern) Fongbe/Gungbe. In normal
speech, however, the -i of the pronoun mi is dropped leading to the transcription found in the
Grammaire abrégée.
Example (17) illustrates a wh-question. Here again we see the combination

between the wh-word and the focus marker.
(17) a. Quel eƭt cet homme?
Mênou-a?
Mέnù-wὲ?
Who-FOC
‘Who is it?’
b. Quelle eƭt cette femme?
Nignone te ouè?
Nyɔ̀ nù tέ wὲ?
girl Q FOC
‘Which girl/woman is it?’
It appears from these two documents that major aspects of both the lexicon
and the grammar of the Gbe languages have been relatively stable since the
seventeenth century. This is a striking continuity given pre- and post-colonial
political developments and related migrations in West Africa. This new finding
is very relevant to the discussion below on the emergence of the creoles.
Indeed, recall that the documents just described were published during the
period when slave exports from Allada peaked to a maximum of 19,000 slaves
a year according to Eltis (2011). Thanks to these documents we now have a
reasonable representation of the type of Gbe language that could have been
used by the enslaved Aja people in Suriname and Haiti in the seventeenth
century. In this context we can reasonably conclude the following:
1. Enslaved Gbe people on Haitian and Surinamese plantations of the seven-
teenth and eighteenth centuries were likely to communicate among them-
selves using varieties of their languages.
2. In the context of language acquisition and its relation to language change the
high rate of lexical and morphosyntactic similarities between the Gbe
varieties must have particularly converged in favoring substrate influence
either in the selection of particular options from within the lexifier (thus by
congruence) or in modifying particular target structures. The effect must
have been stronger in those cases where the substrate structures were shared
by other slaves’ languages (e.g., Bantu languages) (Mufwene 1986; Singler
1988).
3. Because the Gbe languages show close morphosyntactic kinship to other
Kwa languages, we can assume that speakers of the Kwa languages in
general must have produced similar approximations of the target language.
If we hypothesize that such approximations are the result of syntactic
recombination not to be regarded as ‘imperfections’ or ‘instances of L2
failure’ then it appears that Mufwene’s (2001) Founder Principle reflects to
some extent the effect of the particular approximations of the target lan-
guage by speakers of typologically related languages at a significant early
stage of the emergence of the new vernacular.
2.2 Plantation settlement colonies: Suriname and Haiti

There has been considerable work on plantation settlements in the colonies and
particularly in Suriname and Haiti. In what follows, I will limit myself to a
couple of points that are relevant to the analysis of language change that I am
submitting. Because this book focuses on the recombination of syntactic
features into a stable grammatical system, it is important that we know who
might be the carriers of such syntactic features in the first place. With regard to
Suriname and Haiti, this means that we want to know whether enslaved
Africans of Aja descent brought to these colonies could have played some
role in the emerging cultures and languages.
2.2.1 Aja Nengre in Suriname

Anecdotes are not scientific proof of any phenomena but they lead us to use
scientific methodologies to attempt to address various questions that they raise.
Several cultural aspects struck me in Suriname in 2003 when I first met with the
Saramaka of Semoisi. Throughout the journey to Paramaribo, I was extremely
excited by the idea of traveling to the Amazon Rainforest, a very exotic place I
thought, to meet with the Saramaka (also a very exotic people, I presumed). As
we departed from Paramaribo to the river bank to catch a boat to Semoisi, my
excitement turned into a strange feeling of ‘déjà vu.’ The reddish soil, the flora
and fauna reminded me of towns where I grew up: Agbomey (capital of
Dànxòmὲ) and Xògbónù (or Porto-Novo) capital of present-day Republic of
Benin. I felt even more at home when we reached Semoisi and I noticed the
palm fronds at the entrance used here and back home as protection against evil
spirits. Next to the entrance was a small Voodoo (called Tò Lέgbà in Gungbe
and Fongbe) followed by another bigger one in the center of the village. At this
point, I really felt at home. Anyone who has visited Gbe communities in Benin
will find this setting familiar. The next morning, before we met with anyone
else in the village, we were welcomed by the “Captain” i.e., the chief of the
clan, accompanied by two of his ministers: of defense and of communication.
The next thing that struck me was their greeting practice, particularly the
expression un weki no, in the form of a question, which literally means “are
you awake?” This expression of course sounds tautological in any European
language that I know, but not in my father’s and mother’s tongues, Gungbe and
Gengbe, respectively, in which the translations à f ɔ̀ n in Gungbe and ò f ɔ̀ n à in
2.2 Plantation settlement colonies: Suriname and Haiti 53
Gengbe (literally, ‘you awake?’) do not have this literal meaning but rather
‘good morning/day.’ Finally, when we asked our consultants about the
genealogy of their clan, my fellow researcher, James Essegbey (from Ghana)
and I were astonished to hear that they identified themselves with three
different groups: Ji-vi-Allada ‘born-child-Allada’; Ji-vi-f ɔ̀ n ‘born-child-fɔ̀ n’;
and Ji-vi-xwèɖà ‘born-child-xwèɖà’, that is, three Gbe ethnic groups: Àlàdá,
Fɔ̀ n, and Xwèɖà related to three Kingdoms or chieftancies discussed in this
book: Allada, Dànxòmὲ , and Xwèɖà (Whyddah), respectively. Another related
clan we were told, lived in a nearby village called Dànxòmὲ . As we recounted
what we knew from oral history about our own origins as Aja people, their
expansion, and how the communities they mentioned related to each other, we
were thereafter referred to as Aja-nengre.
While all these observations aroused in me the feeling of a deep (sometime
painful) ancestral link with the people, there were also many aspects of their
culture which were completely foreign to me. A case in point is Saramaccan
woodcarving representing astonishingly geometrical and highly abstract
forms from a Gbe perspective. Indeed, Gbe traditional designs, paintings,
woodcarvings, and copper sculptures are generally naturalistic. So I wasn’t at
home after all! I’m not sure how to evaluate this anecdote from a historical
perspective, but two things seem clear to me:
(i) The Saramaccan people were not just New World’s cousins of the Gbe or
Aja. Though there are similarities, the Saramaccan certainly developed a
new culture integrating elements of their new ecology.
(ii) The Africans enslaved in Suriname included people of Aja origin. Some of
these contributed to the maroon society of the Saramaccan; and they seem
to have been numerically significant enough to have contributed a large
numbers of traits to the new Saramaccan culture.
Point (i) has been documented extensively by Price (1973, 1975, 1976, 1983,
1996, 2002, 2008), Price and Price (1991, 1999, 2003), and other scholars. I
will not discuss it here. Likewise, point (ii) has been documented in work by
Postma as reported in Arends (1989, 1995b), as well as by Smith (1987, 1996,
1999, 2009), who has focused on both demographic and linguistic factors. For
instance, according to Smith, Saramaccan has not only retained Gbe lexical
items but also selected functional items such as question words and a focus
marker from a variety of Fongbe. This finding has been corroborated by Aboh
(2006a, 2006b, 2007a), Migge (1998, 2003), Migge and Winford (2007, 2009),
and Winford (2007), among others.
Here, I am particularly interested in whether a Gbe (or any other African
language e.g., Kikongo) could have been retained within the Saramaccan
society? If so, why were the African languages eventually displaced by the
emerging creole (Mufwene (2005a, 2005b, 2008)? I will come back to these
questions throughout the book.
2.2.2 Voodoo Chile in Haiti

At the moment of writing this section, I have never been to Haiti and I had no
anecdote to tell except that I once met with a man on a bus in the United States,
who I was convinced was a Fon from Benin, only to find out after I addressed
him in French that he was Haitian. From the practices of Voodoo to the alleged
origins of Toussaint L’Ouverture as Arada (i.e., Allada), for instance by de
Cauna (2004), it is a truism to say that Haiti shares cultural heritage with the
area identified in history as the Slave Coast. Various studies of demographic
factors on Haiti indeed report the presence of people of Allada descent in the
colony and discuss their contribution to Haitian culture (e.g., Debien 1956,
1974; Fouchard 1972). Similarly, Singler (1996) shows that though the
speakers of Gbe languages might not have been the majority group
throughout the colonization of Haiti, they seem to have constituted a
significant component of the population at the time of the emerging creole.
For instance, the following chart adapted from Singler (1996: 205) shows the
progression of Kwa (including Gbe) enslaved Africans on the French
Caribbean during the second half of the seventeenth century.
As Figure 2.2 shows, there is a significant drop of the Bantu population in the
late seventeenth century (1680–1690). This drop corresponds to a sharp rise in
the Kwa population at about the same time. More important is the fact that this
60
50
40
Kwa
30
Bantu
20
10
0
1664 1680 1690
Figure 2.2 Linguistic distribution in the French Caribbean in the seventeenth

century (adapted from Singler 1996: 205)
change happens when the French colonies shifted from “crops such as tobacco
and cotton to sugar” (Singler 1996: 206), that is, a time when we can assume a
transition from small-scale plantations, called habitations in colonial French
and translated as homesteads in English, to large plantations, which required
much larger slave populations and with which the emergence of creoles has
been associated. This appears to have been a crucial factor still consistent with
Mufwene’s (1996, 2001) Founder Principle, according to which the founder
populations during the emergence of creoles were more likely to influence their
structural properties than late comers.
Debien (1974) provides specific figures on the population of different
regions of Saint-Domingue in the eighteenth century which indicate that the
colony included a significant number of both Congo and Kwa people. Below, I
report the figures as presented in Debien (1974) for the ethnic groups
originating from West Central Africa (i.e., Congo) and those from the Slave
Coast (i.e., Nago, Arada, Ibo, Mines, Adia/Adja, and Gold Coast). The figures
under the label “Other” include ethnic groups from other geographical areas.26
With regard to the South of Saint-Domingue, Debien (1974: 59) writes:
the Congos were not particularly numerous on the lists we consulted as of 1767. The
Ibos were then of about the same proportion, while there were a few Nagos and Aradas.
But soon after, the Congos were almost as many as the Creoles, while the Ibos were less
but the number of Aradas and Nagos increased.
The author further indicates that of the slaves whose origin was known, “the
Congos were 488, the Ibos 144, the Aradas 216 and the Nagos 203.” At this
stage, it is important to remember that the Aradas and Nagos share several
ethnolinguistic traits that justifies grouping them together (see Section 2.1.4).
This would mean that the community described by Debien in the South of
Saint-Domingue included two major groups: the Congos and the Kwa (i.e.,
Arada-Nago).
Regarding the northern part of the island, Debien (1974: 65) reports the
following figures dating from 1760 to 1785. The whole slave population
included 3,568 people. Of the 1,399 known origins, Debien provides the
following figures: Congos 633; Nagos 168; Arada 120; Ibos 55; Mines 43;
Adias/Aja 15; Gold Coast 14; Other 351.
Finally, Debien (1974: 67) presents the following figures dating from 1796
and 1797 for the whole island. The slave population included 14,167 people
including 7,445 Creoles, 391 Mixed and 63 whose origin was unknown. Of the
remaining 6,268 slaves of African origin, the author provides us with the
26
The records consulted by Debien listed such groups as: Aguias, Alemonans, Aras, Bambaras,
Bandias, Barbas, Bibis, Bobos, Cangas, Cap-Laous, Cotocolis, Coromantis/Cramentis,
Dambouans, Dias, Foëdas, Haoussas, Misérables, Mallés, Mandingues, Mocos,
Mondongues, Mozambiques, Poulars, Sénégals, Sosos, Tacouas, Tapas, Thiambas, Timbouts.
Congo
1800
1600 Nago
1400 Arada
1200
1000 Ibos
800
600 Mines
400 Aja
200
0 Gold Coast
South 1757 North 1760 Saint-
Other
1791 1785 Domingue
1796 and
1797
Figure 2.3 Ethnic groupings in Saint-Domingue 1757–1797 (adapted from

Debien 1974)
following figures for the ‘major’ ethnic groups including 4,792 people: Congos
1,651; Nagos 736; Aradas 544; Ibos 519; Mines 44; Adias/Aja 25; Other 1273.
Though the ethnic groups labeled “other” represent a sizable number, it appears
that these enslaved Africans have rather diverse origins (see note 26, which
gives details about these ethnic groups). Focusing therefore on Congos, Nagos,
Arada, Ibos, Mines, and Aja, the figure presented here can be put together as in
Figure 2.3.
What these figures show is that the Congo group remained dominant through
the second half of the eighteenth century. Similarly, the group labeled as
“Other” increased significantly. However, it is remarkable that the Kwa
group including the Nagos, Aradas, Ibos, Mines, Aja, and Gold Coast
represented the most significant group. Merging the figures of these
communities together, we obtain the following results: 573 Kwa in the south
(1757–1791), 415 Kwa in the north (1760–1785), and 1868 Kwa for the whole
island (1796–1797). Put together, we get the graph in Figure 2.4 which gives us
the contrast between the Congo, the Kwa, and Other ethnic groups during this
period.
This regrouping shows that the Kwa group was not only significant
throughout the second half of the eighteenth century but this population went
through an extraordinary growth to the point where Kwa people were more
numerous than the Congo in 1797. According to Debien (1974), the creole (i.e.,
the locally born) slaves were always in the majority. Put together, these
demographic factors point to three main groups which might have played a
significant role in structuring the emerging Haitian culture: the Creoles, the
Congo – who could have been more ethnically diverse than it looks on the
surface as they were captured further inland than on the Slave Coast (see
Section 2.1) – and the Kwa who came in compact and homogeneous
2000
1800
1600
1400 Congo
1200
1000 Kwa
800 Other
600
400
200
0
South 1757 1791 North 1760 1785 Saint-Domingue
1796 and 1797
Figure 2.4 Contrasting the Congo and the Kwa in eighteenth-century Saint-
Domingue (adapted from Debien 1974)
groups.27 The group labeled here as “Other” includes twenty-six different

ethnicities listed in (Debien (1974: 65, 67), see note 26. This group appears
to have been so diverse that it could have dissolved under the pressure from the
dominant groups. Given the names listed by Debien, many of which are
unidentifiable, it is also not unlikely that this group may have included
unidentified Kwa and Congo slaves.
We do not know much about the dynamics of the enslaved populations,
especially about the processes of acculturation and adaptation to the culture of
other African nations as well as to the emerging Creole culture. These issues are
extremely important for our understanding of the socio-historical pressures
which led to the emergence of these new cultures and languages. What we do
know, however, is that Haitian Creole, like other French-based Caribbean
creoles, has retained many lexical items from the African cultures that
contributed to its emergence. Though the overall amount of substrate
vocabulary is relatively small, it is nevertheless important to try to evaluate
specific vocabulary items in terms of their language family. This in turn can
give us an idea of the role of the speakers of those languages during the
formation of the creole.
Since this book is on structural properties, it is important to emphasize here
already that I’m assuming that vocabulary retention is not a trivial process
where a word from one language is simply added to the lexicon of another
27
This need not mean that the Congo were speakers of typologically different languages. To the
contrary, the languages spoken by these slaves must have been relatively close typologically. As
noted earlier, aspects of these languages (e.g., positioning of nominal modifiers, verb com-
pounds) are also similar to certain Kwa patterns. Put together, these groups are more likely to
converge in their approximations of the target language. Where these groups differ however, is
their internal dynamics and to what extent that dynamics affect the emerging creole culture. The
fact that the Kwa arrived in more compact groups seems to have favored them in establishing
some of their cultural and linguistic traits.
language. As will be explained in Chapter 4, a fundamental hypothesis in this

book is that lexical items consist of a triplet involving morpho-phonological,
morphosyntactic, and semantic properties that learners must access to acquire a
language. From this perspective, vocabulary borrowing or retention happens
with transfer of (part of) related morphosyntactic and semantic properties
which in turn may affect the host grammar.
With this in mind, it is noteworthy that Anglade’s (1998) lexical inventory
includes 1,117 lexical items which the author relates to various African
languages.28 As is always the case with this type of work, there may be
misanalyses or mistakes here and there, but what is remarkable for our
discussion is that 992 of these lexical items originated from Niger Congo.
581 of these items are related to Kwa and are distributed as follows: 356
Fongbe, 85 Yoruba, 44 Ewegbe, 30 Nago, 17 Ashanti, 11 Sefwi/Sehwi, 11
Gungbe, 8 Mina/Gengbe, 4 Agni, 4 Mahi/Maxi, 3 Baoule, 2 Igbo, 2 Fanti, 1
Pédah/Xweɖa, 1 Adja, 1 Nzima. Quite surprisingly, the Gbe languages
provided 73 percent of the items on the Kwa list: 356 Fongbe + 44 Ewegbe +
11 Gungbe + 8 Gengbe + 4 Maxi + 1 Adja + 1 Xweɖa = 425. Weighed against
the sample of 1,117 items investigated in this work, we reach the conclusion
that the Gbe represents 38 percent of the sample. This would mean that the Gbe
languages alone provided more than a third of the African lexemes found in
these French-based creoles of the Caribbean. I am not aware of any other study
that came to this conclusion; further work is needed to confirm the figures
shown here and evaluate the impact of these languages on Caribbean French-
based creoles. However, these figures are compatible with both Singler’s
(1996) observations about population factors in the emergence of French-
based Caribbean creoles and, more specifically, Debien’s (1974) discussion
on the African population in Haiti. These figures are also compatible with my
own conclusions based on the geopolitics on the Slave Coast, which shows that
most enslaved Africans who embarked from the ports of Allada or those of its
satellite chieftaincies were of Aja descent.
2.3 Conclusion
This chapter shows that, during the seventeenth and eighteenth centuries, the
Kingdom of Allada, on the Slave Coast, played an important role in the
triangular trade. In this period, the slave export in Allada peaked to an
average ranging between 15,000 and 19,000 slaves a year (e.g., Law 1991,
1994, 1997; Eltis 2011). Our findings indicate that socio-political and
economic factors in the Bight of Benin in the seventeenth and eighteenth
28
This dictionary was kindly brought to my attention by Michel DeGraff to whom I’m very
grateful.
2.3 Conclusion 59
centuries contributed to making locals, namely people of Aja descent, victims

of a globalized world economy, in which Africans slaves were considered as
livestock among other goods (for instance, see the first edition of Le code noir
in the seventeenth century).
The conclusions drawn on the basis of geopolitics as well as socio-economic
factors corroborate with work on population factors in the Caribbean and in
Suriname which shows that the Kwa people represented a significant part of
slave cohorts exported to the colonies during this period. While these Kwa
people, including significant numbers of Gbe speakers, probably did not always
constitute the most numerous ethnic group in the Caribbean during the slavery
period, they seemed to have been instrumental in shaping the culture of the
emerging colonies. Two related factors underlie this: (i) more and more Kwa/
Gbe people were enslaved during the transition of the colonies from the
homestead to the plantation phase (Chaudenson 1992, 2001), when they
needed larger numbers of slaves (usually more than the infrastructure on the
plantation would allow, see Smith 1999, 2009); (ii) the Gbe/Kwa people
originated from a homogeneous cultural background, a factor that probably
allowed them to form tight sub-communities. The structure of such
communities could have helped them occupy leading functions in the slave
population, in domains such as, religious practices, traditional medical care,
fishing traditions, and warfare (e.g., in the case of maroonage).
Put together, these two factors seem to have favored Kwa/Gbe impact on the
emerging Haitian, Sranan, and Saramaccan. While this scenario makes sense in
view of Mufwene’s (2001ff.) competition and selection model adopted here, it
also suggests that his (1996, 2001) Founder Principle should not necessarily be
understood in terms of the numerically dominant ‘original/primary founding
population.’ Rather, we should understand this principle as related to which
(ethnic) group or which population of speakers could have exerted a more
cohesive pressure on the emerging linguistic patterns during the early stages of
the nascent language. In the case of Haiti and Suriname, it seems that the
ecology of the plantation communities favored speakers of Gbe/Kwa
languages. Given this conclusion, the question we can now ask is how the
creoles emerged from the contacts between enslaved Africans and their
European masters and among the enslaved Africans themselves.
3 The emergence of creoles: a review of some
current hypotheses
Kwa speakers are generally assumed to have played a central role in the
formation of various creoles in the Caribbean. The conclusions of Chapter 2
suggest that in the case of Suriname and Haiti we can even be more specific and
single out slaves of Aja descent (i.e., speakers of Gbe languages), as critical
agents in the emergence of their respective creoles. While the speakers of Gbe
languages were numerically certainly not the dominant group throughout the
period of slavery, their contribution to the emerging culture in these colonies
may have been boosted by the fact that they shared a fairly homogeneous
cultural background. Chapter 2 shows, indeed, that the geopolitics of the
Slave Coast probably favored the enslavement of closely related ethnic
groups. The question then is how to integrate these findings in a theory of
language acquisition, language creation, and language change that can both
account for the emergence of the creoles in Haiti and in Suriname and provide
new insights into common phenomena of diachronic change.
The study of creole languages over the past thirty years or so has revolved
around this question, about which various genetic hypotheses have been
proposed. Three schools of thought are often referred to in the literature:
superstrativism, substrativism, and universalism. These schools differ
fundamentally regarding the particular mechanisms that produced a
particular Creole or Creole-type, though they all build on the central idea
that unsuccessful or incomplete L2 learning is the key to understanding the
so-called creolization process. While all theories of language change
presuppose language acquisition (either L1 acquisition or L2 acquisition),
noteworthy here is the fact that these genetic hypotheses presuppose failure
or interruption of acquisition: for one of these reasons, the enslaved
Africans could not acquire the target language completely. Proponents of
these hypotheses have evoked various reasons to explain this failure or
interruption (e.g., resistance to the oppressor or expressive need) but one
motive that is recurrent in the literature is the large discrepancy between the
European population providing the linguistic model and the enslaved
African population targeting the colonial language. The received wisdom
therefore is that in the case of creolization there was a break in language
60
The emergence of creoles 61
transmission (e.g., Bickerton 1988a, 1988b). As the argument goes, a

consequence of this discrepancy is that cohorts of enslaved Africans had
to learn approximations of the target language from one another. This
process eventually led to the emergence of a variety that rapidly diverged
structurally from the target language to such an extent that it had to be
identified as a creole. A recent model that adopts a rather extreme
interpretation of this idea is found in Plag (2008a: 115), who claims that
“creoles are conventionalized interlanguages of an early stage.” That is,
creoles are fossilized interlanguages of the type often observed in studies on
second language acquisition (SLA).
It seems obvious to me that any model that accounts for the development
of creole languages in terms of failure of acquisition, eventually leads to a
theory of language change that is ‘exceptional’ and thus applies to these
languages only but not to other situations of language change in the world. In
this vein, Bickerton (1988a: 268), for instance, distinguishes between gradual
change and catastrophic change. The former is the normal case and “appears
to obey no particular laws and to be largely at the mercy of historical
accident.” The latter, however, represents an exceptional situation where
“new languages are produced ab ovo.” Acording to Bickerton, there is no
principled way in which we could extend explanations of catastrophic change
to gradual change. Thus, I see no reasonable way of extending Plag’s (2008a,
2008b) position on the emergence of creoles to diachronic changes that
occurred in English and have been documented, for example, by Kroch
(1989a, 1989b, 2001).
In this book, I align myself with creole studies such as Muysken
(1981a, 1987, 1988), Muysken and Smith (1986), Mufwene (1986, 1987,
1994, 1999, 2001, 2008), DeGraff (1995, 1997, 1999, 2001a, 2001b, 2003,
2005), Ansaldo, Matthews, and Lim (2007), Ansaldo (2009), Aboh (2009a),
Aboh and Smith (2009), and Aboh and DeGraff (forthcoming). I submit that
explanations of the development of creoles should contribute to
understanding issues of language acquisition proper and issues of
language change tout court, with no need for labels such as ‘normal’ or
‘catastrophic’ changes. More precisely, I adopt Mufwene’s (2001, 2002,
2003, 2005a, 2005b, 2008) competition and selection model, focusing
especially on the emergence of individual morphosyntactic strategies.
Thus, I address the question of how competition and selection operates
when it comes to syntactic features in the mind of the speaker. This
question is further discussed in Chapter 4 which lays out the technical
aspects of competition and selection and how it relates to language
acquisition and syntactic change.
It is however important that I first answer the natural question of why such a
model is needed, against the old exceptionalist tradition in creole studies: What
62 The emergence of creoles
does this model contribute to our understanding of issues of language acquisition

and language change that other competing models do not (viz., the interlanguage,
superstrate, substrate, and universal hypotheses)? Clearly, this question can be
answered only after a critical evaluation of these hypotheses, as I do below.
The following sections show that theories of the development of creoles that
argue for a unique driving force (e.g., superstrate influence, substrate influence,
universal grammar, or fossilized interlanguage) fail to answer the following set
of questions which any work focused on language acquisition and diachronic
change should address:
(i) Why do languages change the way they do?
(ii) How does a new language emerge out of the contact of many?
(iii) How does our understanding of principles underlying change shed light on
issues of first and second language acquisition?
(iv) How does our understanding of change inform us on human cognition?
There exist in the literature several detailed critical evaluations of the
creole genesis theories identified above. I will not repeat them here; the
reader is referred to the cited references. In this chapter, I limit myself to
specific points that suffice to show the shortcomings of specific theories and
underscore the need for more adequate and comprehensive accounts of
language acquisition, language creation, and language change. Section 3.1
discusses the superstrativist approach as developed in Chaudenson (2001,
2003), while Section 3.2 focuses on the substratist approach as formulated
by Lefebvre (1998), more specifically the Relexification Hypothesis.
Section 3.3 deals with Bickerton’s (1987, 1988a, 1988b, 1999) bioprogram
hypothesis, while Section 3.4 is devoted to the interlanguage hypothesis
as proposed by Plag (2008a, 2008b, 2009a, 2009b, 2011). Section 3.5
concludes the chapter.
3.1 Chaudenson (2001, 2003): creoles are distant varieties

of their lexifier languages
As briefly mentioned in the introduction, a common point between
supertrativism, substrativism, universalism, and the interlanguage hypothesis,
is that creoles developed as a consequence of incomplete or inadequate second
language acquisition. The social aspects of the plantations (e.g., demographic
composition, rigidly stratified society, interaction between masters and slaves)
are often mentioned as the sources of the discrepancy between what is
considered ‘normal’ SLA and what certain students of creoles assumed
happened on the plantations: exceptional SLA.
In this regard, theories that focus on the role of the superstratum (or lexifier)
suggest that creole languages are distant varieties of European languages as
3.1 Creoles as distant varieties of their lexifiers 63
they were spoken on the colonies. Chaudenson (2003: 448), for instance,
argues that French creoles result from “the unguided appropriation of
approximate varieties of [the colonial] French koiné.” Creole languages
emerged as a consequence of imperfectly acquired varieties of European
languages, hence the idea of ‘approximation.’ This theory presupposes that
even though creoles may differ from the target European varieties in various
respects, they share basic and fundamental similarities with these languages.
Positing a “français zéro,” which includes the total range of variation in
the French varieties spoken by native French speakers, Chaudenson (2003)
argues that French creoles developed from successive restructuring stages
corresponding to ‘approximations of approximations,’ which crystalize
variation in tendencies that are actually observable in varieties of non-
standard French. In order to accommodate the fact that French creoles are not
isomorphic with non-standard French varieties, Chaudenson (2003: 185)
further suggests that a possible distinction between non-standard French and
a French-based creole could be that the latter includes restructured patterns that
do not belong to “français zéro.” That is, French-based creoles developed new
patterns outside of the range of variation observable in “français zéro.” Such
exogenous patterns could be due to “self-regulating processes” that apply as the
language evolves or, when possible, to substrate transfer. Substrate transfer,
however, is subject to the following two principles:
(1) The principle of superstrate as gatekeeper
The superstrate (i.e., the target language) must provide favorable conditions
or contexts for substrate patterns to be integrated.
According to Chaudenson (2003: 199), this would mean that “les substrats
ne peuvent pas passer en force,” meaning that the substrate patterns cannot
force their way into the creole unless the grammar of the superstrate language
allows it. This view is compatible with the second principle: congruence.
(2) The principle of congruence
Those morphosyntactic patterns that are (partly) shared by the superstrate
and the substrate languages are more likely to be selected into the emerging
creole than conflicting patterns.
According to Chaudenson, the combination of these two principles can
explain the variation observed within the nominal domain in French-based
creoles and summarized in Table 3.1.
As this table shows, definiteness markers and demonstratives do not exhibit
the same distribution across French-based creoles. In Mauritian for instance,
the demonstrative precedes the noun while the definite marker follows it.
In Antillean creoles, the demonstrative follows the noun but precedes the
definite marker. Guadeloupian displays an alternative order, in which the
Table 3.1 Distribution of definite, demonstrative and plural markers across

French-based creoles (adapted from Déprez 2001: 52)1
Creoles Singular Plural
Seychelle sa N Dem>N sa ban N Dem>Pl>N

Mauritian sa N la Dem>N>Def sa ban N la Dem>Pl>N>Def
Antillean N sa (l)a N>Dem>Def se N sa (l)a Pl>N>Dem>Def
N ta (l)a se N ta (l)a
Guadeloupian N la sa N>Def>Dem se N la sa Pl>N>Def>Dem
New Louisiana (la) N sa la (Def)>N>Dem>(Def) le N sa la (Def/Pl)>N>Dem>(Def)
Old Louisiana N la N>Def N (sila) ye N>sila>Pl
Guyanese sa N la Dem>N>Def sa N y(e l)a Dem>N>Pl>Def
Haitian N sa a N>Dem>Def N sa yo N>Dem>Pl
1
See also Bobyleva (2013) for a comprehensive study of the nominal domain and the different
kinds of variation across creoles.
Table 3.2 Distribution of definite and demonstrative articles in standard

French, non-standard French, and French-based creoles (adapted from
Chaudenson 2003: 282)
French Non-standard French Creoles
ce . . . (là), ce . . . (ci) ste . . . (la) . . . la (Quebec French and sa; sa . . . la; . . . la (Seychelle
Missouri French) and Mauritian)
cet . . . (là); cet . . . (ci) le/la . . . la; . . . la (Quebec French sa . . . a (Guyanese)
and Missouri French)
ce . . . (là); ce . . . (ci) ce . . . là, ce . . . ça (Louisiana . . . sa-la (New Louisiana)
French)
ces . . . (là); ces . . . (ci) ces . . . la; ces . . . ça (Louisiana . . . sa-a, sila-a (Haitian)
Acadian French)
ces . . . (là); ces . . . (ci) ces . . . ça là (Louisiana Acadian . . . la (most early Caribbean
French) creoles)
definite marker precedes the demonstrative, while both occur in post-nominal

position. The relative order of these markers may further vary in the context of
the plural marker.
In discussing this variation, Chaudenson (2003) contrasts the patterns
observed in French-based creoles with the order of the same elements in both
standard French and non-standard French varieties.
Thus, the pre-nominal and post-nominal patterns observed in French-based
creoles can be said to derive from the range of variations already present in
non-standard French varieties. While such a superstrate-oriented theory may
appear convincing in this case, it cannot account alone for all the relevant
morphosyntactic properties found in the creoles. Let’s consider the following
sentences in which the verb is focused by doubling. Such constructions are
found in many Caribbean creoles, including Saramaccan and Haitian.
(3) a. Si Kofi *(si) di mujee. [Saramaccan]
see Kofi see DET woman
‘Kofi SAW the woman.’
b. Se mache Bouki te mache, li pa te kouri [Haitian]
FOC walk Bouki TNS Walk 3SG NEG TNS run
‘Bouki had WALKED, not run.’ [DeGraff 2007: 113]
These examples, to which I return in Chapter 6, correspond to what,
according to Chaudenson (2003: 222), should be considered as cases of
substrate transfer, because they display patterns that are absent from the
lexifiers (e.g., French and English) but are attested in the substrates. As
shown in (4a) and (4b) verb focus with raising is impossible in French and
English (Glaude and Zribi-Hertz 2012), but not in Gungbe (Gbe), where such
constructions are grammatical (4c).1
(4) a. *C’est volé Bokassa a volé l’argent de l’Etat.
b. *It is stole Bokassa stole the State money.
c. Fìn Sàgbó fìn kwέ mítɔ̀ n. [Gungbe]
Steal Sagbo Steal money 1PL.POSS
‘Sagbo STOLE our money.’
Because Chaudenson does not discuss verb focus with doubling specifically,
it is not clear to me how one would account for the development of such
innovative patterns in his framework. Indeed, constructions such as in (3)
violate the two principles articulated in (1) and (2), because the superstrate
1
Note, however, that verb focus constructions with doubling are attested in Romance. Below are
some examples from Spanish and Portuguese:
(i) a. Comprar, Juan ha comprado un libro (aunque luego no lo ha leido)

buy.INF John has bought a book but later not CL has read
‘As for buying, Juan has bought a book, although he didn’t read it later.’
b. [Comprar un libro], Juan lo ha comprado [Spanish]
buy. INF a book John CL has bought
‘As for buying a book, Juan has bought it’ [Vicente 2005: 44]
(ii) a. Temperar o cozinheiro temperou o peixe.

to-season the cook seasoned the fish
‘As for seasoning, the cook seasoned the fish.’
b. [Temperar o peixe] o cozinheiro temperou. [Brazilian Portuguese]
to-season the fish the cook seasoned
‘As for seasoning the fish, the cook seasoned it.’ [Cable 2004a: 21]
See also Glaude and Zribi-Hertz (2012) for a study of verb focus with doubling in Haitian Creole
compared to French.
language in this case does not display verb focus with doubling. Accordingly,
they cannot provide a favorable context for such constructions to be
restructured, nor can they act as reinforcers in a context of congruence. Yet,
we can see from the Haitian example that verb raising with doubling is
combined with the French clefting device c’est ‘it is’ which is realized in the
creole as se. Comparing the examples in (3), (4), and (5), what we see is a
construction that clearly combines properties of verb focus in the Gbe (or Kwa)
languages with ingredients of focus in French (Romance), where the
highlighter c’est, realized as se, plays an important role (Glaude and Zribi-
Hertz 2012).
It appears that a strict superstratist hypothesis is inadequate. Indeed, while
the two language types are congruent regarding the semantics of focus
constructions, the examples in (3), (4), and (5) indicate that they display
divergent syntactic properties. French requires c’est ‘it is’, typical of a cleft
structure but this device is absent in Gbe. On the other hand, Gbe requires verb
fronting with doubling, a strategy excluded by French syntax.
Things get worse when we take a closer look at cases, such as illustrated in
Tables 3.1 and 3.2, which have been claimed to support the superstrate
influence hypothesis. As I argued above, the position of the definite
determiner (viz., pre-nominal versus post-nominal), can be accounted for on
the basis of the range of variation attested in “français zéro.” However, this
view does not extend easily to number markers. The latter consistently occur
pre-nominally in (non-standard) French varieties, but surface both pre-
nominally and post-nominally in French-based creoles. We must explain
where this asymmetry comes from.
A possible answer is that the number markers developed subsequent to the
definite determiner, which indirectly determines its position. But this does not
appear to be the right explanation since these two elements do not always
cluster on the same side of the head noun and may actually occur separately on
both sides. In addition, the number marker appears to precede or follow the
demonstrative and the definite marker across the creoles. This is the case, for
instance, in Mauritian and Antillean as indicated by the two rows in grey in
Table 3.1. The relevant patterns are repeated in (5).
(5) Definite marker vs. Demonstrative Number maker vs. Demonstrative
Mauritian sa N la sa ban N la
DEM N DEF DEM PL N DEF
Antillean N sa (l)a se N sa (l)a
N DEM DEF PL N DEM DEF
In both languages, the number marker, here in boldface, is pre-nominal,

while the definite determiner is post-nominal. However, the number marker
does not occur in the same position relatively to the demonstratives in these
languages. It follows the demonstrative in Mauritian, but precedes the noun

which in turn precedes the demonstrative in Antillean creoles.
Given this non-uniform distribution of the markers, we should ask ourselves
the following questions:
(i) Where does post-nominal number marking come from?
(ii) To what extent does French syntax allow it or reinforce it, as stated in
(2) and (3)?
In discussing the Haitian post-nominal number marker yo, which arguably
derives from French third person plural pronoun eux, Chaudenson (2003: 277)
explores an interesting possibility, though ultimately it appears quite
improbable. According to him, the fact that the number marker is sensitive to
definiteness effect and appeared mainly in subject DPs in the eighteenth-
century document La passion selon Saint-Jean en langage nègre is reminiscent
of the usage of resumptive pronouns in non-standard French topic
constructions. Therefore, the Haitian noun phrase in (6a) is comparable to the
bracketed sequence in the Martiniquais example in (6b). Both examples
seem to relate to sentence (6c) taken from La passion, a document that
seems to have been largely distributed in the French Caribbean of the
eighteenth century. As shown in (6d), the creole and non-standard French
appear similar.2
(6) a. sé mésié-là yo [Haitian Creole of 18th]
DEM man-DET PL
‘these men’
b. [Cé bèf La yo] mo. [Martiniquais Creole of 19th]
DEM cow DET PL die
‘Those cows are dead.’
c. Tous pères jouifs la ïo tous faire complot. [Caribbean French creole of 18th]
all father Jews DEM PL ALL make plot
‘All those Jew priests, they plotted together.’
d. Tous ces prêtres juifs-là, eux complotèrent . . . [non-standard French]
all DEM priest jews-there 3PL plot
Under Chaudenson’s approach, the development of the French third person
tonic (or strong) pronoun eux into the Haitian (and other French-based creoles)
post-nominal number marker ïo/yo could be an extension of non-standard
French constructions such as (6d). The demonstration is quite appealing
because it allows us to account for two related grammatical properties of
Haitian noun phrase without additional speculations:
2
These examples are taken from Chaudenson (2003: 277), but the gloss and English translation
are mine.
(i) The Haitian number marker is specified for definiteness because of its
pronominal origin, as well as its usage in topic constructions where it
resumes a definite topic antecedent.
(ii) The Haitian definite determiner is specified for discourse-anaphora
because it developed from the French nominal demonstrative reinforcer
là, which encodes discourse deixis.
However, this hypothesis faces a number of empirical problems and
has theoretical implications which are inconsistent with theories of clause-
periphery constructions involving topicalization and focalization in particular.
An empirical problem with Chaudenson’s view is that the DPs including ïo/yo
are restricted to the subject position as in (7a), though the same constructions
can also occur in object position, as in (7b):3
(7) a. Zozo la yo va bientôt volé. [Ducœurjoly 1803: 336]
bird DET PL will soon fly
‘These/the birds will soon fly away.’
b. Mo voir zhomme la yo.
1SG see man DET PL
‘I saw the/these men.’
It is not clear to me how Chaudenson’s proposal can account for cases like
(7b) where a noun phrase marked by yo is in object position. One could argue
that the object position is a generalization from the pattern in the subject
position. Indeed, Chaudenson’s main point is that creoles did not literally
replicate the lexifier’s system but have extended the application of some of
the strategies that were allowed in “français zéro.”
However, the analysis proposed for the N-eux sequences in subject positions
is not without problems. Indeed, Chaudenson’s scenario according to which the
French resumptive strong pronoun eux in (6d) develops into a DP-internal
number marker ïo/yo in Haitian Creole (6a, 7a) implies that the Africans
learning French did not figure out that in a context like (6d) the resumptive
pronoun eux and the preceding noun phrase tous ces prêtres juifs-là do not form
a constituent and therefore should not necessarily be adjacent, as shown below
in (10). Under current syntactic theories as discussed in Chapter 6 and
references cited therein, a topic sentence like (6d) can be alternatively
represented as in (8a) or (8b) where the topic and the resumptive pronoun
belong to two different syntactic domains.
(8) a. . . . [TOPICP tous ces prêtres juifs-là [FOCP . . . . [IP eux complotèrent]]]
b. . . . [TOPICP tous ces prêtres juifs-là [FOCP eux [IP teux complotèrent]]]
3
I am immensely grateful to Michel DeGraff for sharing all these references with me.
In (8a) the resumptive subject pronoun is emphasized in situ and therefore

occurs in the canonical subject position of the proposition which is represented
here by IP, the projection heading finiteness. Such a representation is
compatible with French topicalizations where the resumptive pronoun is a
weak pronoun ils, as in tous ces prêtes juifs-là, ils complotèrent ‘all those
Jew priests, they plotted together.’ In representation (8b), however, the
resumptive pronoun has been moved to the focus position in [spec FocP].
This latter example is comparable to example (9a) where the resumptive
pronoun is focused by clefting. Example (9b) shows that atonic pronouns
cannot enter cleft constructions (Kayne 1975).
(9) a. Tous ces prêtres juifs-là, c’est eux qui complotèrent.
all these priest Jew-there, it-is them who plotted
b. *Tous ces prêtres juifs-là, c’est ils qui complotèrent.
all these priest Jew-there, it-is they who plotted
Deciding between analyses (8a) and (8b) is not relevant to the present
discussion. What is crucial and ignored by Chaudenson’s analysis, is that
the two alternatives in (8) imply that the resumptive subject pronoun and
the preceding DP do not form a constituent and they are not in a local
relation. This is already shown by example (9a) where we see that these two
constituents can be separated by intervening elements (such as focus). Example
(10a) further shows that a modifying expression can scope over the pronoun
and therefore intervene between the fronted noun phrase “tous ces prêtres juifs-
là” and the resumptive pronoun “eux.” An alternative to this example is (10b)
where the modifier follows the pronoun. Finally, sentence (10c) shows that the
topic and the resumptive pronoun can be separated by a whole sentence.
(10) a. Tous ces prêtres juifs-là, seuls eux complotèrent.
all those priests jew-there, only they plotted
b. Tous ces prêtres juifs-là, eux-seuls complotèrent.
all those priests jew-there, they alone plotted
c. Tous ces prêtres juifs-là, tu sais quoi? seuls eux complotèrent . . .
all those priests jew-there, you know what? only they plotted
Sentences (9) and (10) therefore indicate that the tonic pronoun eux
represents a full noun phrase on its own that can be focused or modified
internally, similarly to any noun phrase. Under current views on the structural
make-up of pronouns, these findings indicate that the strong pronoun eux
qualifies as a determiner phrase labeled here as DP (Cardinaletti and Starke
1999; Aboh 2004a, 2006b). According to this characterization, a sentence like
(6d) can be schematized as in (11), where the first DP (tous ces prêtres juifs-là)
acts as the antecedent of the pronoun eux representing the second DP.
(11) . . . [TOPICP DP[tous ces prêtres juifs-là] [FOCP . . . [IP DP[eux] complotèrent]]]
Accordingly, the development scenario proposed by Chaudenson (2003)

implies that the pronoun inside the DP[eux] develops into a marker (i.e.,
typically a functional head) within a preceding distant DP. This scenario
could be visually represented as in (12).
(12) TopP
spec Top¢
DP1
Top IP
Tous ces prêtres
juifs-là spec
DP2 complotèrent
eux
The hypothesis that the creators of Haitian have conflated DP1 and DP2 as
illustrated in (12) suggests that they failed to properly analyze topic
constructions as involving a topic (i.e., the antecedent) that is being resumed
by a tonic pronoun which itself can sometimes be focused. Such a conclusion
contradicts Chaudenson’s own view that the creole speakers did acquire the
koinés they were exposed to. Indeed, Haitian Creole does distinguish between
topic and focus constituents, as shown by (13):4
(13) [Tout prèt jwif sa yo] (ki la a), se [yo] ki te fè konplo
all priest Jew DEM PL who there DET FOC 3PL COMP TNS make plot
‘All these Jew priests (who are there), it’s them who plotted.’
In this example, the topic Tout prèt jwif sa yo in sentence-initial position and
the focused resumptive subject pronoun yo are separated by the intervening
relative clause ki la a ‘who are there.’
In addition, Chaudenson’s scenario implies that these speakers also failed to
make the right generalization about strong pronouns as a constituent with
internal structure similar to that of a noun phrase. Instead, the pronoun eux is
here wrongly analyzed as a phrasal clitic (probably a head of some sort) that can
attach to any preceding phonological phrase. One can defend such a scenario,
because it recalls certain Haitian forms such as lapolis ‘the-police,’ in which
the determiner is fused to the noun. It is often assumed or implied that such
forms emerged from an incorrect segmentation of the French noun phrase
involving a determiner followed by the noun. Accordingly, the representation
in (12) could be just another example of this confusion.
This hypothesis is however highly improbable. First, examples such as lapolis
sometimes form minimal pairs. In the case at hand, lapolis ‘police station’ is
4
This Haitian translation of sentence (9a) was kindly suggested to me by Michel DeGraff
(p.c. July 5, 2012).
opposed to polis ‘police officer’, as explained by, for instance, Tagète and Urciolo
(1993). Thus, what at first sight looks like a mere confusion by the creators of
Haitian turns out to be a more subtle property of the lexicon where the presence of
the fused determiner is related to meaning distinction. The phenomenon is not
general and some fused forms such as nonm (< un homme) and zanfan (les enfants)
in Lesser Antillean Creoles do not involve minimal pairs. Given this variation, one
cannot immediately conclude that the creators of these forms were all confused by
the French determiner system. Actually, looking at the minimal pairs closely, it
appears that the creators of these forms acquired their syntax as well as (part of)
their semantics, though the process seems to have involved reanalysis. There is not
much work on the contrast lapolis versus polis exemplified in Haitian Creole.
However, if it turns out that there is a general tendency for words embedding the
determiner to denote generic referent (e.g., an institution, a kind) while words
lacking the determiner are primarily referential (e.g., pointing to individuals or
specific entities in discourse), then this contrast may relate to:
(i) structural properties of the noun phrase (i.e., DP): the two forms are not just
contrastive lexemes of the Haitian lexicon but rather involve different
structural make-up;
(ii) subtle semantics of French definite determiners.
Indeed, example (14) shows that French determiners can encode both
definiteness and generic reference. In this sentence, the noun phrase le vélo
‘the bicycle’ can receive generic reading (i.e., cycling), as in (14a), or a definite
reading (i.e., the bicycle being discussed), as in (14b):
(14) Tu sais quoi? Mon fils aime bien le vélo.
you know what? My son loves well the bicycle.
a. You know what? My boy loves bicycles in general (i.e., he likes biking).
b. You know what? My boy loves this bicycle (i.e., the one we are talking about
now).
What emerges clearly here is that the definite determiner allows usages
where it denotes generic entities. This function, it seems to me could have
well triggered the Haitian contrast between pairs like lapolis versus polis. More
study is needed before we confirm this analysis. But if we allow ourselves to
entertain this hypothesis, then we are led to conclude that the Haitian speakers
did have access to abstract properties of the noun phrase in French.
This conclusion is actually compatible with example (6c), repeated here for
convenience:
(6) c. Tous pères jouifs la ïo tous faire complot.
all father Jews DEM PL ALL make plot
‘All those Jew priests, they plotted together . . .’
A close look at this example shows that speakers of eighteenth-century

Haitian were indeed aware of the syntax of the quantifier tous ‘all’ which
occurs in prenominal position as opposed to its pronominal equivalent which
occurs in subject position. These speakers were also aware of the fact that the
quantifier and the determiner do not have the same semantic function which is
why they can co-occur, as they do in this example. Clearly these speakers had
some knowledge of the determiner system in French and its relation to the
pronominal system. Under Chaudenson’s (2003) scenario described in (6) and
(12), it is not clear to me how such ‘well informed’ speakers could have failed
to deduce the pronominal properties of eux ‘them.’
Second, even if we accept the scenario described in (6) and (12), which
builds on the role of non-standard French, we still have to assume, in a
uniformitarian way, that the overgeneralization of the subject resumptive
tonic pronoun eux into a number marker is part of a learning strategy that one
expects to arise in context of second language acquisition. As far as I know,
no study on L2 learners of French reports similar facts. Nor am I aware of any
study of French L1 acquisition that mentions such a possibility. Actually,
certain studies on the L2 acquisition of French determiners indicate that the
use of determiner does not seem to represent a domain of great difficulty for
learners. For instance, Granfeldt (2000: 277), who compared Swedish
bilingual L1 learners of French to L2 learners, concludes that:5
The use of determiners in French is not a major problem for Swedish adult learners.
The gradually emerging use of determiners that we saw in the bilingual children
is not observed, but instead, already in the early recordings, we see productive use
of different determiners and very few omissions. More importantly, there is no
evidence of restrictions on phrase structure at any point, since the determiner
omission rate is not higher in simple Det±N contexts than in complex Det±Adj±N
sequences. There is, thus, no evidence that a Swedish adult learner starts over with
only a lexical NP.
Third, it is not clear to me how the proposed scenario in (6) sheds light on the
relative order of the number marker within the French-based creoles. As noted
above, this marker occurs pre-nominally in some creoles and post-nominally in
others. More importantly, however, while Haitian and Guyanese Creoles have a
post-nominal number marker that also corresponds to the third person plural
pronoun, other creoles display prenominal number markers that evolved from
diverse etyma. In Mauritian, the plural marker ban is assumed to have
developed from French bande ‘group’ (see Guillemin 2011 and references
5
See also work by Deprez, Sleeman, and Guella (2011), who studied the acquisition of French
determiners by L2 Dutch and Arabic learners and showed that even though these learners
produce non-target-like sequences, they seemed to have acquired the basic semantic properties
of these determiners (e.g., their obligatoriness in argument position).
3.2 Substrate influence as the main factor 73
cited there) and it does not function as a third person plural pronoun (the typical
source for nominal plural markers in several Atlantic creoles). Similarly,
Guadeloupian se arguably developed from the French demonstrative plural
ces ‘these/those’ or possibly the possessive pronoun ses (his, her plural) and
does not function as third person plural pronoun either (Déprez 2004, 2007).
Quite remarkably, the plural marker in Guadeloupian precedes the noun but the
latter precedes the definite marker which precedes the demonstrative (se N la
sa): an order that is completely alien to non-standard French varieties. The
plural markers in French-based creoles show varying syntactic properties
which do not always correlate with the syntactic properties of their etyma in
non-standard French. Thus, even though the morphological form of these
markers clearly shows that they evolved from French it is not always
possible to explain their syntactic behavior only on the basis of the syntax of
their French etyma.
I conclude that a theory of the development of creoles that only argues for
a successive restructuring process that takes the creoles away from their
lexifier without any other complementary explanation cannot account for all
the relevant facets of the evolution. For instance, our observations on number
marking in French creoles indicate that we need a more articulated theory
that takes into account the language contact ecology of the plantation, by
investigating the (possible) contribution of competing languages or varieties
present in the colony during the critical period of the emergence of the
creole. In addition, it appears that, even though principles (1) and (2)
might operate in certain cases, they do not seem to be major driving forces
in the development of the creole. Our conclusions about superstrativism
extend to their opposing alternatives, that is, theories of creolization that
only put forward the influence of the substrate languages spoken by the
slaves. In the following section, I discuss one extreme example of such
theories.
3.2 Lefebvre (1998): substrate influence as the main factor

in creole genesis
Substrate influence as formalized in Lefebvre (1998), a synthesis of research
she conducted on the subject matter with several collaborators for over fifteen
years, appears to be the mirror image of Chaudenson’s theory. According to
Lefebvre, the grammars of the creoles are primarily determined by the
grammars of the African languages spoken by the slaves. This is so even
though creoles generally derive their lexicon from European lexifiers and
learners are not categorically indifferent to the syntactic behaviors of the
different lexical items. This view is often caricatured with a quote from
Sylvain (1936: 178), who, quite inconsistent with the contents of her book,
concluded her discussion on Haitian morphosyntax saying “nous sommes en

presence . . . d’une langue éwé à vocabulaire français” (‘we face here an Ewe
language with a French vocabulary’). Her thesis is that Haitian shares its
core grammatical properties with Gbe languages, though it derives its
vocabulary from French. According to Lefebvre (1998: 9), “the creators of a
creole language, adult native speakers of the substratum languages, use the
properties of their native lexicons, the parametric values and semantic
interpretation rules of their native grammars in creating the creole.”
Ironically, Sylvain provides numerous examples throughout her book that
trace various grammatical markers in Haitian Creole to various non-standard
French varieties.
Like other theories of creole genesis, Lefebvre’s Relexification Hypothesis
presupposes unsuccessful second language acquisition. However, unlike in
Chaudenson’s scenario, the ‘relexifier’ speakers resort irremediably to their
native tongues. In a sense, these speakers continued to speak out their native
grammars only substituting the lexifier’s vocabulary for lexical items of their
native languages.
Lefebvre’s (1998) theory is actually a modification of Muysken’s (1981a)
original hypothesis on the emergence of Media Lingua. Muysken (1981a)
then argued that this new language developed from adaptations of Spanish
vocabulary to Quechua grammar through relexification and translexification.
Muysken (1981a: 55) defines relexification as a “process of vocabulary
borrowing in which the borrowed element adopts the meaning and use
of the element in the receptor language for which it is substituted.”
Accordingly, only the phonological representation is maintained in the
receptive language. On the other hand, translexification is defined as “a
process of vocabulary substitution in which, in addition to the
phonological representation, all other levels of information are adopted
from the target language as well” (p. 61).
Muysken (1981a) adopts a traditional view of the lexicon whereby a
lexical item embeds a bundle of features determining (i) phonological, (ii)
syntactic, and (iii) semantic representations, (iv) subcategorization frames,
and (v) selectional properties. Given this, relexification and translexification
can be technically described as in (15), adapted from Muysken (1981a: 61).
In this representation, I have substituted the neutral terms LA and LB for
Muysken’s original terms “target language” and “source language,”
respectively. These stand for two different languages, regardless of which
of the two is the native tongue of the speaker. The choice of these neutral
terms will become clear when I evaluate Lefebvre’s (1998) version of the
Relexification Hypothesis. In this representation, the broken lines stand
for translexification (explained below), while the solid lines represent
relexification.
(15) Vocabulary substitution:
LA LB
/PHON/A /PHON/B
SYNA SYNB
SUBA SUBB
SEMA SEMB
SELA SELB
/PHON/A¢ /PHON/B¢
SYNA SYNB¢
RELEXIFICATION SUBA SUBB¢ TRANSLEXIFICATION
SEMA SEMB¢
SELA SELB¢
Receptor language
The important point about this diagram is that relexification is a partial

process of vocabulary integration whereby the only information retained in
the receptor language is the phonological representation. Translexification on
the other hand can be seen as total integration of all abstract features of the
borrowed lexical item into the receptor language. In this regard, Muysken
(1981a) builds in his model the important observation that speakers have
access to abstract representations of lexical items in the two languages (here
LA and LB) and can draw on this knowledge to various degrees. Indeed,
Muysken (1981a: 62) remarks that “for relexification to occur, the semantic
representations of the source and target entries must partially overlap [i.e., be
partly congruent]; otherwise, the two entries would never be associated with
each other” (emphasis is mine). This means that relexification and
translexification as defined here primarily operate on the basis of abstract
lexical properties involving semantics and syntax. Accordingly, the fact that
speakers may sometimes retain the phonological form only (e.g., in
relexification) should not be taken to mean that they could not access abstract
properties of the adopted vocabulary item or that they “relexify” across the
board.
This is a significant difference between Muysken’s (1981a) model and that
proposed by Lefebvre (1998). For the latter, relexification applied to creole
languages is a process whereby the phonetic strings are retained from the target
language (p. 17). Accordingly, the adopted lexical item is stripped of abstract
properties such as those listed in (15). As the author argues, “this is because . . .
relexifiers either do not have access to this information or, if they do, they do
not use it in creating the new lexical entry (p. 17).” Thus, Lefebvre (1998: 16)
proposes the following description for relexification:
(16) ORIGINAL LEXICAL ENTRY LEXIFIER LANGUAGE
PHONOLOGYi PHONETIC STRINGj used in specific

SEMANTIC FEATUREk semantic and pragmatic context
SYNTACTIC FEATUREn
NEW LEXICAL ENTRY
[phonology]j’or [Ø]
[semantic feature]k
[syntactic feature]n
Looking at the new lexical entry, we realize that this lexical item retains all
abstract properties in the “original lexical entry” (i.e., in the speaker’s native
tongue). Even the phonetic strings are parsed on the model of the learner’s
native grammar. In this representation, the newly created item is fundamentally
different from its source in the lexifier language because it is strictly
constrained by the native grammar of the speaker. Accordingly, the noun
phrases in (17), which contrast Haitian Creole with Gungbe, could represent
a perfect illustration of relexification.
(17) a. Krab mwen sa a yo [Haitian, Lefebvre 1998: 78]
b. àgásá cè éhè lɔ́ lέ [Gungbe]
crab 1SG-POSS DEM DET PL
‘these crabs of mine’
c. N>Poss>Dem>Det>Pl
As the sequence in (17c) indicates, Haitian Creole and Gungbe appear to
manifest the same word order in the noun phrase: the possessive pronoun, the
demonstrative, the determiner, and the number marker all occur post-
nominally, in this very order. Note also from these examples that both
Gungbe and Haitian Creole lack number inflections and gender specifications
on the head noun and the demonstrative.
In French, the determiner, the demonstrative, and the possessive pronoun are
pre-nominal and they are specified for gender and number (18a–c; 18a’–c’).
(18) a. le(s) crabe(s) a’ La table
DET.MASC.PL crab DET.FEM
table
b. ce(s) crabe(s) b’ Cette table
DEM.PL crab DEM table
c. mon/mes crabe(s) c’ Ma table

MY.SG/MY.PL crab MY.FEM table
The ungrammatical example in (19) further indicates that in French (unlike

in Gungbe and Haitian) these elements are mutually exclusive.
(19) *Le(s) ce(s) mes crabes
DET(PL) DEM(PL) MY(PL) crab(PL)
The post-nominal definite determiners are strongly discourse-anaphoric in

Haitian Creole and Gungbe, because they identify shared/known information
only. This is not necessarily so in French, in which, as shown in example (13),
the definite determiner may be ambiguous. The semantics of the Haitian and
Gungbe determiners is tentatively rendered here by the expression “the X in
question.”6
(20) a. pè-a [Haitian; Sylvain 1936: 55]
priest-DET
‘the priest (in question)’
b. mɔ̀ pὲ lɔ́ [Gungbe]
priest DET
‘the priest (in question)’
Both Haitian Creole and Gungbe use bare noun phrases in a way that French
does not, though Old French displayed bare nouns. On the other hand, the
patterns of usage of bare nouns are not identical in Haitian Creole and Gungbe,
as the following examples illustrate:
(21) a. Wosiyòl manje kowosòl. [Haitian; DeGraff 1992: 105]
b. Àlúέ nɔ̀ ɖù gbàdó.
magpie HAB eat corn
‘Magpie(s) habitually eat(s) corn.’
b’. *Pie mange maïs
magpie eat corn
These contrasts between Haitian Creole, Gungbe, and French require some
explanation in both the superstratist and the relexificationist approaches. For
instance, one needs to know more about the distribution of bare nouns in Old
French and how this compares with Haitian Creole. In addition, one would
need to determine which variant of Old French may have survived in which
non-standard French varieties and whether the Africans were exposed to it in
Haiti in the eighteenth century.
6
For a detailed discussion of this topic, see Aboh and DeGraff (2014).
On the other hand, the parallels between Haitian Creole and the Gbe
languages as opposed to French appear to be readily compatible with the
Relexification Hypothesis. Acording to Lefebvre (1998: 78) the Haitian
“[+definite] determiner, the plural marker and the [+deictic] terms have
been created through relexification” whereby relexification is understood as
“a mental process that builds new lexical entries by copying the lexical entries
of an established lexicon and replacing their phonological representations with
representations derived from another language.” While this analysis might look
convincing at first glance, it comes with a number of drawbacks, which point to
both internal theoretical inconsistencies and empirical inadequacies. I discuss a
few here, though the reader should consult DeGraff (2002) for a very detailed
critical evaluation.
Regarding the conceptual inadequacy, an important point that is often
overlooked in the application of the Relexification Hypothesis to creole
languages is the fact that Muysken’s (1981a) model was designed to account
for the emergence of a new language, Media Lengua, within a bilingual
community, in which only two major languages were in contact. As Muysken
(1981a, 1997) makes clear, this contact language is not used for communication
outside the Quechua community (i.e., with other ethnic groups). The author
then argues that this variety came into existence probably due to a need within
the community to create a new identity rather than the sudden need to create a
makeshift language that would allow speakers of different linguistic
background to communicate (but see Shappeck 2011 for an alternative).
Muysken (1997: 376) also explains that “many Media Lengua speakers also
speak fluent Spanish.” The author further demonstrates that none of the social,
cultural, and political factors that are often evoked to account for the emergence
of pidgins and creoles applies to the emergence of Media Lengua.
So, this is a language variety that resulted from the contact of only two
languages, Quechua and Spanish, both of which the speakers spoke fluently.
This sociolinguistic setting makes perfect sense when one considers Muysken’s
(1981a) relexification model as a process that primarily involves abstract
lexical properties in the languages in contact. Therefore it is by definition a
mental process that is available to bilinguals or at least to speakers who have
acquired pieces of knowledge of abstract properties of a language that they are
learning or that they are confronted with in their daily lives. Crucially,
relexification so defined is not a theoretical model that can be applied to the
African second language learners of French in Haiti, who are assumed by
Lefebvre (1998) and (Plag 2011), to have no intuition whatsoever about the
target language.
In this regard, let us recall that most theories of creolization heavily rely on
the idea that these languages emerged as a result of restricted quantitative and
qualitative access to the target language. In the plantation context as envisaged
by these studies it cannot be assumed that the Africans were bilinguals in

their language and in the colonial language (viz., English, Dutch, French,
Portuguese, or Spanish).7 One possible interpretation of this situation would
be to say that the enslaved Africans were not competent enough in the target
language to have access to abstract properties of words. This is obvious in
Lefebvre’s (1998) formulation that the relexifier slaves did not have access to
the abstract properties of lexical items in the lexifier language. All they could
parse was the phonetic string.
But if this is so, how could relexification occur in such a context? How
could the relexifier identify relevant phonetic strings in the lexifier language
that could possibly feed (or match with) the syntax of comparable elements
in their native languages? More specifically, how could the representation in
(16) generate a new lexicon, if all that the Africans had access to was the
phonetic string in the lexifier language coupled with the “specific semantic
and pragmatic contexts” where such strings occur? And even if we grant
this possibility, how could these speakers know the “specific semantic and
pragmatic contexts” in which certain lexical items are used without accessing
their abstract properties? In fact the speakers’ ability to identify the semantico-
pragmatic and syntactic contexts in which some lexical items are used is a
direct consequence of their capacity to manipulate abstract concepts about
these lexical items. DeGraff (2002) already exposed this major contradiction,
arguing that it represents the most serious conceptual drawback for Lefebvre’s
(1998) theory of relexification as represented in (16).
In addition to such conceptual inconsistencies, this version of relexification
also faces empirical problems. Returning to our previous comparison of
determiners in Haitian Creole and Gungbe, a distinction that emerges is that
Gungbe displays an indefinite counterpart of the definite discourse-anaphoric
determiner lɔ́ in (20b), viz. ɖé. It occurs in post-nominal position (22a), where it
can be combined with the number marker as in (22b). Example (22c) indicates
that the definite marker lɔ́ and the indefinite marker ɖé are in complementary
distribution and can be said to compete for the same position. Finally, example
(22d) shows that ɖé is formally distinct from the numeral one: ɖòkpó.
7
This view, which is shared widely within the creolist community, is actually misleading. A
cursory search in the following database www.marronnage.info/en/index.html reveals an
astonishing number of advertisements describing (runaway) slaves as speaking the colonial
language fluently. It is also important in the context of this discussion to realize that a
significant number of these slaves were multilingual. A case in point is the following
advertisement in Haiti “1767-02-25 – Francisque, étampé sur la joue droite V, parlant
français, espagnole & anglois, marqué de petite vérole, ayant le nez écrasé, trapu & de
moyenne taille, maron depuis la fin de janvier dernier. Ceux qui le reconnoîtront, sont priés
de le faire arrêter & d’en donner avis à M. Durse aîné, Capitaine de Navire, au Port-au-Prince,
ou à M. Laville, Négociant au Cap. Il y aura récompenses.” I thank Michel DeGraff for
making this gold mine available to me.
(22) a. mɔ̀ pὲ (ɖé)

priest DET
‘a certain priest’
b. mɔ̀ pὲ (ɖé) (lέ)
priest DET PL
‘certain/some priests’
c. *mɔ̀ pὲ ɖé lɔ́
priest DET DET
d. mɔ̀ pὲ ɖòkpó
priest one
‘one priest’
Haitian, on the other hand, does not have a proper indefinite marker in post-
nominal position. Instead, this language uses the numeral yon ‘one,’ which in
Romance and Germanic languages can also function as a specificity marker. In
addition, the Haitian numeral occurs in pre-nominal position, like other
numerals in this language.
(23) yon jwèt ti-moun lan [DeGraff 1992: 111]
One toy child DET
‘a toy of the child’
The contrast between Haitian and Gungbe suggests Table 3.3, which shows
the determiner-like elements in Gungbe and Haitian.
It appears from this table that where Gungbe has two dedicated morphemes
for encoding specificity and definiteness (in addition to numerals), Haitian has
only one. Clearly this is a problem because it suggests that the Gbe speakers
ended up not relexifying part of their grammar. In response to this issue,
Lefebvre (1998: 78) claims that “the substratum languages’ so-called
indefinite determiner was abandoned. Haitian developed its own indefinite
form.”
While one may want to hold this view, the question is why this strategy was
abandoned. Why, according to the relexificationist hypothesis, would speakers
sometimes fail to relexify a significant property of their native grammar? And if
Table 3.3 Certain determiner-like elements that encode

definiteness in Gungbe and Haitian
Anaphoric determiner Plural
Haitian La/lan/a [definite] yo

Gungbe lɔ́ [definite] lέ
ɖé [indefinite]
Haitian could independently develop an indefinite form, why can’t this

assumption extend to the analysis of the definite determiner?
I don’t know what form a convincing answer to this question could take in
the relexificationist approach. Whatever that answer could be, it would have to
address the following complication as well. In addition to numerals, illustrated
in (23), other nominal modifiers also occur pre-nominally in Haitian Creole,
including adjectives and non-numeral quantifiers (24a–b). Example (24c)
further shows that Haitian also involves post-nominal adjectives.
(24) a. nũ gro gato [Sylvain 1936: 40]
one big cake
‘a/one big cake’
b. Tut gro nèg pu šita [Sylvain 1936: 40]
All big people pu sit-down
‘All the important people should sit down.’
c. yon fanm entelijan [DeGraff 2007: 118]
IND woman intelligent
‘an intelligent woman’
By contrast, such modifiers occur only post-nominally in Gungbe, as shown
in (25a–b).
(25) a. gàtó ɖàxó àwè
cake big two
‘two big cakes’
b. Gán ɖàxó àwó lέ kpó ná sín nùkɔ̀ n.
chief big ten PL all FUT sit front
‘All the ten big chiefs will sit in front.’
The dissimilarity between (24) and (25) shows that the Gbe languages cannot
be taken as the source of the relative ordering of nominal modifiers in Haitian
Creole. In accounting for the contrast between Gbe versus Haitian Creole and
French, Lefebvre (1998: 39) claims that
the relexifiers . . . acquire the directionality properties of the superstratum major cate-
gory lexical entries. Thus, the word order of major category lexical entries in the creole
is predicted to follow the word order of the lexical categories in the superstratum
language. Consequently, if the superstratum language has prenominal adjectives, the
creole will have prenominal adjectives.
This explanation leads to a contradiction. Indeed, we now face a situation
where in developing the pattern in (17), the relexifiers were able to impose the
directionality properties of the Gungbe deictic functional items on Haitian
Creole, ignoring French. But the same cohort of relexifiers failed to apply the
same principle (i.e., ignore French syntax) when it comes to adjectives and
other nominal modifiers. The paradox is even more puzzling because in the
case of (17) what is retained is a grammatical item, while regarding (24) we
are dealing with lexical items which, as we know from studies of language
contact, can be borrowed in a language without necessarily affecting the
grammar of the recipient language. For example, the French adjective petit
‘small’ occurs pre-nominally when used attributively as in j’ai appelé un petit
enfant ‘I have called a small child.’ Yet, in Gungbe, one must say ùn yrɔ́ ví
petit ɖé ‘I call child small ind: I called a small child,’ with the borrowed
adjective petit following the noun. The alternative order *ùn yrɔ́ petit ví ɖé
‘I call small child ind: I called a small child’ is unacceptable. If relexification,
as defined by Lefebvre (1998), were the only process at work in the formation
of Haitian Creole, one would expect the Gbe speakers learning French in
seventeenth-century Haiti to produce the Gbe-type Noun-Adjective order
systematically, which they did not.
More generally, these facts show that Lefebvre’s (1998) style of analysis
must assume that the relexifiers were able to identify the relevant directionality
properties in both the superstrate and the substrate languages. But, as DeGraff
(2002: 359) already pointed out, “the determination of the relevant substrate-
superstrate overlap presupposes, at the very least, the knowledge necessary to
segment the relevant superstrate forms from spoken utterances, plus a fair
amount of abstract analysis.” That is, the Haitian relexifiers could not have
identified the directionality parameter associated with these nominal modifiers
(e.g., adjective, numeral, and quantifier) unless they were able to
identify their functions and their combinatorial properties according to the
morphosyntax of the noun phrase in French. This brings us back to criticisms
already formulated in the preceding discussion, viz., that Lefebvre’s (1998)
version of relexification contains a serious contradiction that cannot be
resolved in any principled way.
Because my criticisms here have focused on the version of relexification
proposed by Lefebvre (1998), the interested reader may want to know whether
Muysken’s (1981a) original formulation could offer a viable alternative to
Lefebvre (1998). The answer, it seems to me, is yes provided we accept that
some Africans were bilinguals and could speak the lexifier language fluently
(as I explain in Chapter 4). These speakers may unconsciously engage in
relexification, but the question we ought to ask is what impact such relexified
forms may have on the population level where other probably less proficient
African speakers produced equally competitive variants. There are probably
traces of relexification or similar mental processes in contact languages in
general (e.g., Schwartz and Sprouse 1996), and in creoles as well. However,
what matters for our discussion is that we cannot maintain (as is done in
Lefebvre 1998 and much related work) that relexification is the sole driving
force behind the development of creoles. Not only is such a view incompatible
3.3 The Language Bioprogram 83
with socio-historical factors of plantation settlement colonies as we know

them, but it also fails to account for all the relevant empirical facts.
3.3 The Language Bioprogram as the main factor in creole genesis

Unlike superstratist or substratist theories, universalist approaches to the
development of creoles claim that these languages express linguistic
universal properties. In the generative framework, this would mean that
creoles are the expression of the human language capacity as specified by
UG. While DeGraff (1999ff.) emphasizes that such theories have to take into
account the input that speakers were exposed to as well as the interaction
between adult L2 learners and children L1 learners (see also Mufwene 2008),
other more radical theories attribute creolization to children only.
According to Bickerton’s theory of creolization, creole languages are
seen as an exceptional case of language change: they result from a break in
transmission, which placed the creole children in a linguistic and cultural
chaos. The underlying assumption here is that, as the number of the African
slaves grew and outnumbered that of European colonists during the plantation
phase, the newcomers (i.e., the bozals) had less access to the target language
and had to learn “useful fragments of one another’s languages . . . What this
gave rise to was a macaronic jargon, a form of communication that employed
words from several languages . . . that lacked any kind of syntactic structure”
(Bickerton 1999: 52). He further concluded that “children with no prior
language experience but with their native language capacity to guide them
will take that same input and make good any deficit between it and a natural
language” (Bickerton 1988a: 273).
Bickerton’s (1981, 1988a, 1988b, 1999, 2008) view of creolization as the
formation of a new language in the space of a single generation (Bickerton
1999: 49) tries to conceptualize population factors in creole societies and how
they could affect the emergence of a new variety. Following Chaudenson (2001,
2003), Bickerton acknowledges that plantation colonies went through two phases:
(i) The homestead phase (i.e., early stage): European colonists form the
majority of the population and represented the linguistic model for a
small population of enslaved Africans.
(ii) The plantation phase (i.e., ‘industrial’ stage), enslaved Africans mas-
sively outnumbered European colonists (e.g., 10 to 1, see Singler 1996,
Chaudenson 2001, 2003, Mufwene 2001, 2002, 2003). Access to the
colonial linguistic model was limited and the enslaved Africans had to
learn from one another.
For Bickerton, however, the latter phase is the determining one because
it resulted in the dilution of the target language characterized by a loss of
grammatical structure then yielding the macaronic jargon (viz., the pidgin).
Accordingly, there was a break of transmission between the homestead phase
and the plantation phase which in turn involves two interrelated stages:
(i) The pidgin phase: Inaccessibility to L2 and absence of a common lingua
franca within the slave community led to an extreme disintegration of the
target producing a macaronic jargon.
(ii) The creole phase: The pidgin is the ground zero of language degeneration.
It cannot serve as input for L1 acquisition. Consequently, children born on
the plantation during this period developed a new language relying exclu-
sively on their language instinct: the language bioprogram.
This theory implies that Hall’s (1962) pidgin-to-creole life cycle results
from two interrelated processes: language disintegration which leads to
pidginization, and linguistic reconstruction which is characterized by the
creation ab ovo of a new language by children. This makes clear why Bickerton
(1988a, 1988b, 1999) concludes that creoles are instances of catastrophic language
change with fundamentally different properties from older human languages that
are ‘normally’ transmitted from generation to generation. After analyzing the
language of trained apes, children under two, adults who were deprived of
language in their early years, and pidgins, Bickerton (1990a: 122) concludes that
the evidence just surveyed gives grounds for supposing that there is a mode of linguistic
expression that is quite separate from normal human language and is shared by four
classes of speakers: trained apes, children under two, adults who have been deprived of
language in early years, and speakers of pidgin. Since this mode emerged spontaneously
in the three human classes; since the second class includes all members of our species in
their earliest years; and since the fourth class potentially includes any person at any time,
we may regard the mode as a species characteristic.
Bickerton’s argument is grounded in the hypothesis that creole children,

faced with a pidgin, found themselves in a situation comparable to hominids
deprived of modern language in the early stages of our phylogeny, a situation
that putatively was comparable to that of trained apes. Faced with the
macaronic pidgin their parents were speaking and under pressure to develop
a normal modern language, the creole children created the new full-fledged
languages known as creoles, which differ from the antecedent pidgins in being
semantically, morphosyntactically, and lexically more complex. According to
Bickerton (1988a: 274–278), that “creole grammars are in some elusive sense
simpler than the grammars of older languages” can be explained by the fact that
dilution implies the loss of morphemes of two types (p. 278):8
8
This longstanding (but unmotivated) view is embraced by McWhorter (2001) in his article “The
world’s simplest grammars are creole grammars” and in related studies, where the label creole
stands for a cluster of linguistic traits defining a typological class (of simple languages),
(i) Non-basic morphemes: if lost, will not be reconstituted (e.g., most pre-
positions, derivational morphology).
(ii) Basic morphemes: if lost, must be reconstituted (e.g., TMA, locative
adpositions, articles).
Because creoles emerged from the conjunction of (i) and (ii), what appears to
be cross-creole resemblance is considered to be direct expressions of the human
capacity for language, what he calls the “Language Bioprogram” (Bickerton
1984). Features that are cited the most to illustrate cross-creole similarities
include, among others:
– Rigid SVO word order.
– Lack of inflectional morphology, which correlates with the use of free
morphemes to encode tense, mood, aspect specifications in the fixed order
Tense>Mood>Aspect (TMA).
– Near-absence of prepositions which is compensated by the use of serial
verbs.
– Absence of non-transparent wh-words: Many creoles display bi-partite wh-
phrases consisting of a wh- or question-word and a second term encoding
location, time, cause, person.
– The use of numeral ‘one’ as indefinite determiner and the development of the
definite determiner from a distal demonstrative derived from the superstrate.
– The use of third person plural pronoun as a nominal number marker.
As Bickerton makes clear, the fact that these properties sometimes cluster
together in creole languages does not necessarily imply that they are all direct
expressions of the Language Bioprogram. Some, for instance the usage of bi-
partite wh-words and the fixed order of TMA markers, may be specific to the
Language Bioprogram. Others, however, (e.g., the use of third person plural
pronoun to mark number) could relate to the individual history of the language
(e.g., the types of languages in contact).
If creoles were indeed the direct expression of the Language Bioprogram,
then these languages would present us with a special case of first language
acquisition in a situation of inadequate input. This way, the study of these
languages would inform us on the internal structure of what the Language
Bioprogram might be and how it might relate to other human cognitive
capacities. There have been several critical evaluations of Bickerton’s
Language Bioprogram, and the creole typology that it suggests. I will not
discuss these issues here and the reader is referred to Muysken and Smith
(1986), Newmeyer (1988), Thomason and Kaufman (1988), DeGraff (1999,
2003), Mufwene (1991, 1999, 2001, 2008), and the references cited therein.
presumably a cluster that is closer to UG (but see Muysken 1988; DeGraff 1999, 2001a, 2001b,
2003; Mufwene 2001, 2008 for a critique of this position).
For the sake of this discussion, I only draw the reader’s attention to a few
theoretical and empirical inconsistencies of the Language Bioprogram
hypothesis, which suffice to show that it is inadequate in many respects.
These inadequacies in turn help better contextualize the proposal I develop in
this book.
Recall from previous discussion that Bickerton (1988a) hypothesizes that
creoles developed backward from the conjunction of language disintegration
(i.e., loss of grammatical properties leading to pidginization) and language
creation (i.e., reconstitution of a full-fledged grammar). Bickerton (1988a: 278)
further concludes that morphemes lost in the pidgin phase are of two types,
which are listed under (26) and (27):
(26) Non-basic morphemes: Morphemes that, if lost, will not be reconstituted
a. Grammatical gender
b. Number agreement
c. Bound verbal morphology, (e.g., conjugation forms)
d. Derivational morphology
e. Pronominal case and gender forms
f. Most prepositions
(27) Basic morphemes: Morphemes that, if lost, must be reconstituted
a. Articles
b. Tense/aspect/modality forms
c. Question words
d. A pluralizer
e. Pronouns for all persons and number
f. A general locative preposition
h. An irrealis complementizer
i. A relativizing particle
j. reflexives and reciprocals
A question that immediately comes to mind is what underlies the
distinction between (26) and (27). One explanation that Bickerton
provides is that while the properties in (27) are found across creole
languages, those in (26) are usually absent in the creoles he studied even if
the same properties are present in the source languages. According to
Bickerton, the properties in (27) represent the core (or unmarked)
properties of syntax. As he puts it:
presumably, the single universal syntax requires some minimal set of grammatical
morphemes for the discharge of the functions that it stipulates. If these are lost,
replacements are recruited from a limited set of lexical items. The fact that, across
creoles, similar lexical items with similar properties are recruited suggests that
there must be markedness in the inventory of (possible) lexical properties and that
creoles select unmarked options – but this possibility requires more extensive study.
(1988a: 279)
That creoles display the properties in (27) can be taken as evidence of

fundamental aspects of the Language Bioprogram. The latter requires that
humanly possible languages discharge the functions in (27). This leads to the
conclusion that (27) represents the absolute minimum properties a grammar
must have to fulfill the function of communication as necessary in human
interactions.
One problem with this argumentation is that the properties in (27) are
not found in all languages and one may wonder why that is the case. Why
do creoles end up exhibiting only certain subsystems of those made
available by the human language capacity (see Mufwene 1999 for
discussion)? Part of the problem is methodological. Because the elements
under (26) are generally absent from creoles and because Bickerton
assumes creoles to have been created ab ovo by children with no
linguistic experience but endowed with the Language Bioprogram, he
is led to conclude that all the grammatical properties which superficially
look similar across creoles must have emerged from the same human
language device: the Language Bioprogram. As the reader can see, the
argumentation is circular.
Another problem that arises with the list of morphemes in (27) relates to how
these morphemes distribute cross-linguistically. As Bickerton (1988a: 279)
makes clear, “while all languages have tense/aspect/modality forms,
pluralizers, question words, pronouns, etc., not all languages have articles or
irrealis complementizers; it is thus unclear why universal syntax would require
creoles to have them.” Indeed, Mandarin Chinese does not have articles, yet it
has morphosyntactic means to express definiteness (Cheng and Sybesma
1999). Similarly Gbe languages do not have an article system of the Indo-
European type and therefore allow bare nouns to be interpreted as (in)definite
or generic (Aboh 2004a, 2004b, 2010a; Aboh and DeGraff 2014). This point
takes us back to creoles. According to Bickerton, creoles develop articles
in order to encode definiteness. Accordingly, “creoles have articles (which
always have to be reconstituted) either because these are simpler to
reconstitute than allowable alternatives, or because they are more unmarked”
(p. 279). The problem here is that many creoles do not need articles to express
definiteness. Indeed, these languages are notorious for using bare nouns which
can be interpreted as generic or (in)definite depending on the context.
Accordingly, definiteness is not a defining property of articles in creoles (see
Baptista and Guéron 2007; Guillemin 2011; Bobyleva 2013 for discussion).
Aboh and DeGraff (2014), for instance, discuss the following examples from
Haitian Creole. In these sentences, a specific definite noun phrase that is
marked by a determiner in the main sentence is resumed by a bare noun in
the subordinate clause.
(28) a. Bouki achte yon bèl machin. Men, machin tonbe bay Bouki pwoblèm.
B. buy a nice car but car fall give Bouki trouble
‘Bouki bought a nice car, but the car started to give trouble to Bouki.’
b. Chwal la te lage. Asefi Te al chache l toupatou . . .
horse DET ANT escape Asefi ANT go look.for 3SG everywhere
Men, chwal gen tan tounen nan poto pandan Asefi te pati.
but horse have time return in pole while Asefi ANT leave
‘The horse had escaped. Asefi had gone to look for it all over, but the horse
already returned to the pole while Asefi was gone.’
In comparison to articles in Romance and Germanic one expects definite
articles in creoles to occur upon second mention of a newly introduced referent
(as in 28a). Example (28b) further shows that anaphoric use of bare nouns is not
limited to newly introduced referents. In this sentence, the second mention of
the referent chwal is a bare noun even though the previously mentioned antecedent
is already a specific definite involving the determiner la comparable to articles in
Romance and Germanic. In these examples both the bare nouns machin (28a) and
chwal (28b) denote specific definite referents, yet they do not include a determiner.
Such bare nouns are not found in French, the lexifier of Haitian Creole.
Given that these languages, like many others (e.g., Kwa) do not necessarily
need articles to encode definiteness, the expected unmarked option for these
languages should be not to develop articles at all (contrary to evidence).
Therefore, one wonders why creoles, which, according to Bickerton, selected
unmarked specifications of the Language Bioprogram, would develop
redundant mechanisms for encoding definiteness: by means of bare nouns
and articles. Instead, it seems that the creoles develop these article-like
elements because they are found in the source languages where they perform
various discourse functions including marking topicality (cf. Aboh 2004c;
Aboh and DeGraff 2014).
A similar question can be raised with regard to (27c), (27e), (27f), (27i), and
(27j) for which it is not clear why they should be basic to language. With regard
to (27c), Aboh and Pfau (2011), for instance, show on the basis of signed and
spoken languages that question words are not necessary for asking questions.
This study shows that not all languages have question words of the type found
in Romance, Germanic, or Creole languages (i.e., the so-called wh-phrases).
A case in point is Indian Sign Language where content questions involve a
clause-final question particle that scopes over an associate phrase inside the
clause. In appropriate context, the associate may be left unpronounced leaving
the question particle only in clause-final position (see also Aboh, Pfau, and
Zeshan 2005). It appears from this study that what is required in Indian Sign
Language content questions is a sentential question particle that encodes the
interrogative force.
Haitian provides us with a very interesting case against (27e), because it does
not display distinct forms for all persons and numbers. Indeed, the Haitian form
nou stands for both first and second-person plural. Notice that this case appears
more complex than the situation with the English form you which can be
considered a unique morphological form for second person, lacking a number
distinction, or the French forms il/ils versus elle/elles, for which the singular
and plural are homophonous except in contexts where liaison applies. This can
be illustrated with il/elle a ‘he/she has’ vs. ils/elles ont ‘they have’ in contrast
with il/elle mange ‘he/she eats’ vs. ils/elles mangent ‘they eat,’ which are all
pronounced the same way, the gender distinction notwithstanding. In Haitian
what gets blurred is the person distinction, a potentially more basic feature, if
one agrees with Bickerton.
Stipulation (27f) also appears to be unevenly distributed across languages. In
the Gbe languages, for instance, some locative expressions as in Súrù yì [xwé
gúdò] (lit. ‘Suru went house back’) only involve a complex noun phrase
consisting of the ground and its parts. No other general preposition is needed.
Examples like this clearly suggest that the overt grammatical marking of
location is not necessary for the expression of location. The same point arises
with property (27i). We can indeed imagine languages without relative clauses,
and therefore without relativizers, which is apparently true of some
Austronesian languages (e.g., Comrie 1989; Lyle Campbell 1996).
Finally, the Gbe languages present yet another counterevidence to (27j),
because they do not formally distinguish between reciprocals and reflexives in
all contexts. Accordingly, the following sentence is ambiguous between the
two readings in (29a) and (29b).
(29) Súrù kpó Àsíbá kpó mɔ̀ n yé-ɖè
Suru COORD Asiba COORD See 3PL-ɖè
a. ‘Suru and Asiba saw themselves (e.g., in a mirror).’
b. ‘Suru and Asiba saw each other (e.g., at a wedding).’
So, all in all, the only property that seems to hold across languages is
that in (27b): all languages display expressions of tense, aspect, and
modality. In creole studies, certain authors (including Bickerton) insist on
the fact that creole languages might be exceptional in exhibiting the order
[Tense–Mood–Aspect]. Unfortunately, such claims are made in total
ignorance of the typological literature. Indeed, it has been shown since the
1980s in creolistics (Muysken 1981b) and both in functionalist and
generativist literature (e.g., Foley and Van Valin 1984; Bybee 1985;
Hengeveld 1989; Pollock 1989; Cinque 1999) that the more complete
order Mood[proposition-oriented]–Tense–Mood[participant-oriented]–Aspect is
a universal tendency. So it can’t help to single this sequence out as a
typical property of creole languages as imposed by the Language
Bioprogram. Instead, the TMA sequence seems to be a general property of

human languages.
If we were to maintain the distinction between lists (26) and (27), it is
important to realize that (27), the minimum required by universal syntax in
terms of Bickerton, presents us with a fairly rich grammatical system. Given
this, we may ask ourselves why creoles must be expected to express these
properties in order to be considered normal languages while others seem able
to move away from it, adopting more complex and opaque structures. What
causes ‘old’ languages to diverge from this stipulation? This question becomes
even more important when we consider the issue of acquisition. If indeed, the
properties in (27) were in some sense more unmarked and related to direct
specifications of the Language Bioprogram, they would be attested in natural
situations of acquisition. According to Bickerton’s model, one would expect
children to make generalizations based on primary settings (or unmarked
options) of the Language Bioprogram. For example, we would expect
children to go through acquisition phases with articles emerging first and
then disappearing in languages with none (e.g., Mandarin), and analytic
TMA markers showing up before being replaced by inflections or serial verb
constructions preceding the emergence of prepositions in some languages.
Similarly, we would expect children to reanalyze questions such as ‘who did
you see?’ as ‘what person did you see?’ in which the wh-phrase is of the type
Q-XP and involves a question word combined with a noun. The same
reanalysis would thus generate [what-person] for ‘who,’ [what-time] for
‘when,’ [what-thing] for ‘what,’ [what-place] for ‘where,’ etc. To the best of
my knowledge, such developmental paths have not been shown to be typical of
child acquisition. We may therefore ask ourselves what prevents children from
entertaining such learning hypotheses even though the Language Bioprogram
should putatively favor them.
Mufwene (1999) shows that while some hypotheses made on the basis of the
Language Bioprogram are confirmed by early child acquisition data, there are
significant differences between child language and creoles. Bickerton (1999)
addresses these issues suggesting that while the process of acquisition is the
same between first language acquisition and creolization, the two situations
differ with regard to the context of acquisition and the input (i.e., exposure
to a target). In ‘normal’ acquisition children receive an input with enough
grammatical material. They can make correct generalizations about the target
in a way that sometimes leads them to move away from the default options such
as those in (27). On the other hand, creole children putatively had to face inputs
which lacked significant portions of grammatical items. The scarcity of such
morphemes would have compelled Creole children to adopt default options
made available by the Language Bioprogram, thus explaining the difference
between ‘normal acquisition’ and creolization. As Bickerton (1999: 57) puts it,
Table 3.4 Adpositions in early Sranan (1707–1798), adapted from

van den Berg 2007: 153
Preposition Meaning Prepositions Meaning
Bifo Before (Na) agter At the back, After

Bilo Below (Na) baka At the back, after,
behind
Fossi Before (Na) abra Over, across
Fu From, of, for, about (Na) fesi Over, across
Leki Like (Na) ini (In)side
Sondro Without (Na) tapu Upper side, On
Na To, at, on (Na) midri In the middle
Te Until (Na) sey Beside
(Na) ondro Under(side)
if the driving force behind creolization is simply the extreme rarity of available
grammatical morphemes, and if the characteristic phenomena of creoles arise largely
as a result of compensatory mechanisms, we would expect to find that in normal
acquisition many creole-like features would make only sporadic and fleeting appear-
ances, if indeed they appeared at all.
Once again, the ultimate explanation falls back on restricted input, but this
leads to a contradiction, given that many subtle features of the target languages
found their ways into the creoles. For instance, the discussion on Haitian
adjectives in previous paragraphs shows clearly that Haitian speakers
acquired the morphosyntax and semantics of French adjectives. Accordingly,
it cannot be the case that these learners found themselves in situations of
“extreme rarity of available grammatical morphemes.”9
I will return to the issue of the input in subsequent chapters. With regard to
the present discussion, it seems to me important to show that the thesis, central
to the Language Bioprogram hypothesis, viz., that creole languages display
systematic similarities because they express unmarked (or default) strategies
made available by the human language capacity, is disputed by many empirical
facts. These inadequacies add to the conceptual flaws discussed in previous
paragraphs. As already observed by Muysken (1988), adpositions represent
obvious counterevidence to the disintegration–reconstitution view developed
in Bickerton’s work. Indeed, Van den Berg (2007) presents ample data
suggesting that adpositions never got lost in Sranan even though they were
reanalyzed. As is evident from Table 3.4, eighteenth-century Sranan already
9
Alleyne (1971) also pointed out that the presence in creoles of forms corresponding to inflected
items in the lexifiers (e.g., ded ‘die, dead’ and broko ‘break’ in English creoles) suggests that the
varieties to which the producers of creoles were exposed were not morphologically as impover-
ished as has traditionally been claimed.
involved matches for most of the (etymologically complex) prepositions found

in English.
As is obvious from this list, the forms corresponding to complex prepositions
in English (e.g., inside, outside) are reanalyzed as two separate grammatical
items (i.e., Na + a nominal-like) that arguably realize two distinct syntactic
positions. What seems to happen here is some sort of reanalysis according to
which syntactic components that are fused together in English are separated in
a systematic way in the creole (see Aboh 2010b). It is obvious that the creators
of these complex forms in Sranan had access to formal properties of complex
prepositions in English. Because these forms are present in early Sranan, they
cannot be said to result from language disintegration or pidginization as
proposed by Bickerton. Similarly, these data cannot be interpreted as ab ovo
creations of children faced with a macaronic input. Thus, they represent a
serious challenge to Bickerton’s Language Bioprogram Hypothesis and the
like on the emergence of creoles that are based on the pidgin-to-creole life
cycle.
Finally, a serious empirical problem for the Language Bioprogram
Hypothesis and the “creole prototype” claimed by McWhorter (1998) is that
it completely fails to predict the variations observed in the nominal domain and
partially discussed in Section 3.2 with regard to French-based creoles
(Bobyleva 2013). Table 3.5 illustrates the relative ordering of demonstratives,
adjectives, and number expression across creoles with different lexifiers and
substrates.
It appears from this table that a strict interpretation of the Language
Bioprogram Hypothesis cannot account for the variations in the nominal
domain across creole languages. Since Abney (1987), Szabolcsi (1987), it
has been assumed in the generative framework that clauses and noun phrases
have parallel structures. Aboh (forthcoming) further argues that clauses and
noun phrases involve the same structural make-up. Put in this context,
Bickerton’s theory actually predicts that creoles should exhibit the same
structural make-up in the nominal domain as well. Accordingly, one expects
to find the same ordering of grammatical elements across creole noun phrases.
Table 3.5 clearly shows that this prediction is not borne out. Bickerton’s theory
does not offer any principled way to account for the discrepancy between
clausal properties that he assumed to be uniform across creoles and the
varying structures that derive from the nominal sequences in Table 3.5.10 The
same can be said of theories which claim creole typology and which, one way
or the other, assume a Bickertonian pidgin-to-creole lifecycle. I conclude from
this that the Language Bioprogram hypothesis cannot be an adequate
explanation for the development of creole languages.
10
To the best of my knowledge, Bickerton never discusses this asymmetry in his work.
3.4 Plag’s interlanguage hypothesis 93
Table 3.5 Relative ordering of demonstratives, adjectives, and

number across creoles with different lexifiers and substrates
Ordering of Dem, N, Adj, Num Creoles
Dem-(PL)-Adj-N-(PL) Bahamian
(PL)-Dem-(PL)-(Adj)-N-(Adj) Kriyol, Tok Pisin, Seychelles, Mauritian
Dem/PL-Adj-N-PL-Dem Berbice Dutch, Sranan
(PL)-(Adj)-N-(PL)-(Adj)-(PL)-Dem Fa d’Ambu, Nubi
(Adj)-N-(Adj)-(Dem)-(PL)-(Dem) Haitian, Papiamentu
(PL)-Adj-N-Dem Sango
(Dem)-PL-N-Adj-(Dem) Lingala
3.4 Plag’s interlanguage hypothesis

The discussion in previous sections shows that most authors assume second
language acquisition to have played a role in the emergence of creoles. For
Chaudenson, creoles emerged as a consequence of generations of ‘unguided
second language acquisition’ during which the productions of the untutored
learners slowly drifted away from the target language. Lefebvre (1998), on the
other hand, assumes that second language acquisition failure left the learners
with one option only: to relexify their native tongues with phonetic strings from
the target language. For Bickerton, chaotic second language acquisition led to
pidginization, which provided an inadequate input for first language acquisition.
Consequently, creole children had to create a new language relying only on their
innate language capacity, the Language Bioprogram.
Plag (2008ff.) makes a stronger claim: “creoles are conventionalized
interlanguages of an early stage” (2008a: 115). Interlanguages are transitional
varieties that learners develop in the course of second language acquisition.
Because interlanguages are developmental phenomena they involve various
stages, e.g., early vs. late, which correspond inversely to the varying degrees to
which the learner approximates the target language. While studies in second
language acquisition indicate that learners generally end up acquiring the target
language, it is in principle possible that acquisition ends at any interlanguage
phase for some learners. A case in point, often cited in the creolistics literature,
is that of Alberto, a thirty-three year old Costa Rican studied by John Schumann
(1978) over a period of ten months of untutored acquisition of English. The
study showed that compared to five other subjects involved in the study,
Alberto shows almost no progress in his acquisition of the target but instead
developed a “reduced and simplified form” that could serve his communicative
needs. Such simplification of the target language to a pidgin level, Schumann
argues, could be universal in second language acquisition, though in normal

circumstances, learners, mostly motivated by social, cultural, and economic
integration, rapidly move past this phase to become more proficient.
Applied to creoles, this would mean that, given their social and cultural
conditions on the plantation as well as their restricted access to the target
European language, most African learners developed rudimentary
communicative systems using some basic ingredients from it. These
rudimentary systems were subsequently conventionalized into a coherent
communal variety. This view is not new and similar arguments can be
found in Mather (2006). Plag’s (2008a, 2008b, 2009a, 2009b)
interlanguage hypothesis builds on processability theory as developed in
certain psycholinguistic models of speech production and acquisition.
According to this theory,
there is a universal, implicational hierarchy of processing procedures derived from
the general architecture of the language processor. In addition, and related to that,
there are specific procedural skills needed for the production of utterances in the
language to be learned. (Plag 2008a: 120)
Given this view, the same processes involved in speech production and
processing are also active in language acquisition. As the author puts it,
the central claim of Processability Theory (. . .) is that these processing procedures
not only reflect their sequence of activation in language generation but also that
the acquisition of these procedures will follow this implicational hierarchy. (Plag
2008a: 122)
Five processing procedures are defined as follows (Plag 2008a: 121–122):
Stage 1, lemma access: learner produces pre-categorial one word utterances.
Stage 2, category procedure: learner can form simple sentences of the type
NVN or NNV.
Stage 3, phrasal procedure: shows evidence of intra-phrasal information exchange
such as NP-internal agreement but not subject–verb agreement.
Stage 4, the sentence-procedure (S-procedure): learner develops a full sentence
procedure and may exhibit subject–verb agreement.
Stage 5, the subordinate clause procedure: sentence embedding is possible.
The diagram in (30), adapted from Plag (2008a: 122), further illustrates the
so-called phrasal procedure and S-procedure. As indicated by the circles, a
phrasal procedure is assumed to happen inside a constituent and may trigger
intra-phrasal agreement or matching, e.g., matching between a noun phrase and
a determiner. On the other hand, an S-procedure is assumed to happen at the
clausal level where one observes inter-phrasal agreement, such as between a
subject and a verb.
Table 3.6 Developmental stages in English interlanguage

(Plag 2008a: 123)
Development Structure Example
Initial state One-word utterances ball

!
Canonical word order Bob kick ball (SVO)
Neg + V He no like coffee
Adverb fronting Then Bob kick ball
Topicalization That I didn’t like
Do-Fronting Do you like it? Do she like it?
Yes–no inversion Has he seen you?
Copula inversion Where is John?
Particle verbs take the hat of
Do/Aux 2nd Why did he sell that car?
Where has he gone?
Target Cancel Inversion I wonder why he sold that car
(30) S
NPsubj VP
Peter owns NPobj
Agreement between Det N

Peter.3sg and V reflects a dog Matching between the article and the noun
S-procedure: Inter-phrasal reflects phrasal procedure, that is, Intra-
Information exchange phrasal information exchange.
According to Plag, these hierarchical procedures help in understanding the

developmental stages of both L1 and L2 acquisition. In this regard, Table 3.6
below illustrates the hypothesized implicational hierarchy as observed in
learners of English.
The developmental path in Table 3.6 can be analyzed based on the
procedural hierarchy given in Table 3.7, where one observes that “starting
with one-word utterances, learners gradually acquire more complex structures
in a specific order, with at least some learners ending up with the most complex
structure” (Plag 2008a: 124).
Focusing on agreement morphology, two observations are in order here:
First, plural marking occurs on nouns already at stage 2, while agreement inside
the noun phrase is realized only at stage 3. Second, subject–verb agreement,
which instantiates inter-phrasal information exchange, occurs relatively late: at
stage 5. Put together, these two observations indicate that the so-called inherent
inflection may appear relatively early (from stage 2 onward), while “the
Table 3.7 Processing procedures for English (Plag 2008a: 124)
Processing
Stage procedure L2 processing Morphology Syntax
1 word/lemma words invariant forms single constituent

2 category procedure lexical morphemes plural on nouns canonical order
possessive pronouns
3 phrasal procedure intra-phrasal NP-agreement ADV, do-fronting
information Neg+V topicalization
exchange
4 S-procedure/Word inter-phrasal Y/N inversion
order rules information copula inversion
exchange
5 S-procedure/Word inter-phrasal SV agreement Aux/do 2nd
order rules information 3sg –s
exchange
6 subordinate clause main and subordinate cancel inversion
Procedure clauses
instantiation of agreement procedures or structural case assignment requires

the most advanced processing procedures and occurs therefore only at later
stages” (Plag 2008a: 124).
Applied to creole languages, this view suggests that the often claimed
absence of inflectional morphology in these languages could be explained by
the fact that they are interlanguages that have been conventionalized at an early
stage, probably not later than stage 2 or 3. Plag (2008b, 2009a, 2009b) proposes
to extend this view to account for syntactic and phonological properties of
creoles in general. For instance, Plag (2008b) observes that creoles tend to
manifest the SVO order, pre-verbal negation, and wh-questions without
subject-auxiliary inversion of the Germanic/Romance type because of their
fossilized interlanguage nature.
As Plag (2008a: 115) acknowledges, the interlanguage hypothesis comes
with a number of caveats, which ultimately indicate that this view cannot be
maintained as a general characterization of the emergence of creoles. One such
caveat relates to the development of interlanguages as a communal variety. It is
indeed important to realize from Schumann’s observation about L2 learners
that what we can now refer to as the Alberto syndrome is an individual process,
as noted by Mufwene (2010). Plag’s (2008a, 2008b, 2009a, 2009b) claim that
creoles are conventionalized interlanguages suggests that the Africans all
engaged in the acquisition of the target the same way and developed the
same interlanguages at the same time. This scenario is quite unlikely. Given
the general characterization of plantation colonies of the eighteenth century as
involving African learners of different ages, cultures, linguistic backgrounds,

and geographical origins (e.g., West Africa vs. West Central Africa; see
Chapter 2), as well as locally born slaves, the Creoles, we can assume that
not all of these learners engaged in the successive stages described above (and
recapitulated in Tables 6 and 7) similarly to a military regiment marching
toward a target. (See DeGraff 2009 and Mufwene 2010 for similar critiques
of Plag’s hypothesis.)
In such a setting, where Africans of different learning skills or different
social groups on the plantation had qualitatively varying access to the target
language (e.g., skilled slaves, plantation slaves, domestic slaves, as noted by
Alleyne 1980, Arends 1995a, among others), we can easily imagine that not all
of them stopped at any of the early interlanguage stages posited by Plag or
follow exactly the second language acquisition order he hypothesized. Because
the plantation population evolved rapidly (owing to various factors associated
with death, cf. Arends 1995a; Mufwene 2001; Emmer 2005), common sense
suggests that the situation is more complex than the linear development
suggested in Tables 3.6 and 3.7.
Instead, the only logical possibility in such a context is that L2 learners (e.g.,
the newly arrived bozals) who might have started at stage 1 – where they are
“only able to produce one-word utterances with unclear category status of the
lemmas retrieved from the lexicon” (Plag 2008a: 123) – would quickly move
through this phase to reach more advanced levels.11 Accordingly, stage 1
appears to be a relatively unstable and short phase that is unlikely to exert
any significant impact on the emerging creole. Instead, it appears that most
plantation learners would develop various learning hypotheses that could be
characterized in terms of Plag’s different stages between stages 2 and 6 in
Table 3.7. Since most Africans are minimally bilingual (as one could infer from
the discussion in Chapter 2), we can further imagine that their learning paths
might not match the linear developmental path suggested in Tables 3.5 and 3.6.
It is also unlikely that all these learners engaged in these learning trajectories at
the same pace. Accordingly, whatever interlanguages these learners might have
developed would have been too unstable and varied to conventionalize into a
communal norm.
Though Plag (2008a) did not discuss the scenario under which such varied
interlanguages could conventionalize into a creole, it seems to me that the only
logical possibility for this view to hold would be to assume the implausible
situation where a great majority of the population would suffer from the Alberto
syndrome and fossilize at a very early stage of acquisition. Then selection
11
I was born and raised in a multilingual context. I’ve never met with a young or adult L2 learner
who continued to live with the fluent-speaking population but whose language acquisition
process stopped at the one-word stage or any of the early-interlanguage stages assumed by
Plag in his demonstration.
should proceed in such a way as to force the remaining portion of the

population who could have escaped this syndrome to eventually adopt the
putative fossilized variety as means of communication. In a sense, this would
be a situation where more advanced learners will have to unlearn to fall back to
the fossilized level.
Put together, all these factors make it difficult to envisage how such a
heterogeneous population of interlanguages (i.e., produced by learners at
different stages of acquisition) could ever conventionalize into a coherent
communal system with all the properties of a fully fledged language: the
creole. The only answer Plag (2008a: 115) offers to this paradox is that “in
addition to the individual mental processes that generate quite variable outputs,
i.e., a pool of variants, we need to acknowledge effects of selection and
establishment of particular features from these outputs across speakers, i.e.,
in the (emergent) speech community.”
Plag (2008a: 115) further refers the reader to work by Mufwene (2001, 2002)
on how selection could proceed. This leads to a contradiction since Mufwene’s
competition and selection model is a general theory of language change that
clearly indicates that no exceptional mechanism of change or creation underlies
the emergence of creoles (see also Aboh and Ansaldo 2007; Aboh 2009a).
Besides this contradiction, it appears that Plag’s stipulation that creole
languages are conventionalized interlanguages must be embedded in an
existing broader framework in order to stand. This obviously shows the
inadequacy of his model. Indeed, Plag acknowledges the shortcomings of his
theory when he declares:
my reason for making the interlanguage hypothesis the central theme of my JPCL
columns in spite of these caveats is not so much that I think the hypothesis is entirely
true. Quite to the contrary, I will, show that there are a number of creole properties that
cannot be explained under this hypothesis. (Plag 2008a: 115)
This quote not only shows the conceptual flaws of this theory but also its
methodological shortcomings. Because I discussed some of the conceptual
drawbacks in previous paragraphs, I now point the reader only to the fact
that while the title of Plag (2008a, 2008b, 2009a, 2009b), “creole as
interlanguages,” suggests that the interlanguage hypothesis accounts for all
aspects of creoles, what the author actually means is that it accounts for some
aspects only. The question then is how to account for those aspects of creoles
which do not fall under the interlanguage generalization (e.g., the emergence of
TMA systems, see Plag 2008a, footnote 2). And concerning this, one should
logically ask whether the solutions or analyses that account for these other
aspects cannot also extend to those that the interlanguage hypothesis is claimed
to cover, thus making the whole interlanguage hypothesis redundant and
therefore undesirable.
These questions bring me to the methodological point. The view the author
adopts in his series of columns in the Journal of Pidgin and Creole Languages
is that creoles arose as a consequence of radical SLA failure. This is so even
though there exists no clear evaluation of what that failure consists of and how
it relates to the target. This in turn raises the question of the target on the
plantation and how to evaluate such target a posteriori (see Chaudenson 2001,
2003, and much related work). Yet, the rationale entertained by this author is
that because we know that creoles emerged from SLA failure and because SLA
generally involves the development of interlanguages, we can deduce current
creole properties from those putative interlanguages that each African could
have developed on the plantation while trying desperately to acquire an
inaccessible target. But if the target language is inaccessible, as claimed by
some popular theories of creolization, and assumed by Plag, then it is pointless
to speak of an interlanguage, because there is no point of reference. Yet, Plag
compares modern creoles, outcomes of the interlanguages developed by the
learning populations on the plantations, to various instances of interlanguages
described in the SLA literature. This is not unproblematic though.
First, interlanguages are not full-fledged systems that can be taken wholesale
as entities of linguistic comparison. Interlanguages are transitional stages that
learners go through in the course of acquisition. So when we refer to
interlanguages, we actually refer to different stages of acquisition. These
stages are not necessarily arranged linearly (contrary to what is presented in
Table 3.6) and, more importantly, not all L2 learners necessarily go through
exactly the same stages. Because of this, comparing a creole to a learner’s
interlanguage amounts to comparing what DeGraff (1999) refers to as an
E-creole (the creole spoken at a community level) to a developing I-language
(i.e., the interlanguage that a learner X may produce at a given point in time
during L2 acquisition). The problem here is that interlanguages are unstable
systems by definition and, as such, they cannot be compared in a holistic way to
a communal linguistic system, which is variable inter-individually. This is so
even though it is perfectly possible to compare sub-modules of the developing
interlanguage to those of the target in order to evaluate or describe the learner’s
acquisition path and the possible learning hypotheses she might be
entertaining.
Second, studies of SLA have shown that target-deviant forms that learners
may develop in the course of acquisition are allowed by UG in the sense that
they are often found in some languages of the world, though these are (or might
be) absent in the target.12 Given this, that some supposedly creole structure may
resemble some interlanguage which in turn is found in some language of the
12
This should not be surprising given that these target-deviant structures are human linguistic
creations, which, just as natural languages, stem from the human language capacity.
world tells us nothing as to the emergence of that particular structure in the

creole, just as it does not inform us about how that particular structure could
have arisen in any language of the world. To give an example, creoles do not
generally show case marking on full noun phrases and lack subject–verb
agreement. With regard to Atlantic creoles, for instance, Romance-based
creoles such as Haitian and Cape Verdian do not mark case on full noun
phrases. The lexifier languages, French and Portuguese, also fail to mark case
on full noun phrases, though Latin their ancestor did. Obviously, we cannot
conclude from the absence of such case marking that the Romance languages
are interlanguages deviating from Latin, no more than we can rush to the
conclusion that, because creoles diverge structurally from their lexifiers, they
are conventionalized interlanguages! In addition to not marking case on noun
phrases, Atlantic creoles also do not exhibit subject–verb agreement. In this
respect, they are like many other languages of the world, including the Gbe
languages of West Africa and the Sinitic languages. Here again, the fact that the
interlanguages of some L2 learners of a language X lack case morphology on
full noun phrases and subject–verb agreement does not make Gbe and Sinitic
conventionalized interlanguages.
Third, a comparison of creoles structures with interlanguages presupposes a
good knowledge of the system of interlanguages in general. More precisely, the
interlanguage hypothesis implies that there exists a fundamental structural
make-up that underlies interlanguages and that it can serve as a yardstick
against which creoles can be measured. To the best of my knowledge, no
such structural characterization of interlanguages exists; an important reason
is that interlanguages are not fully-fledged grammars and they are assessed as
such only relative to the relevant target language. Therefore, the linear
developmental path in Table 3.7 is a mere description of the routes that some
L2 learners may take and a crude indication of possible learning hypotheses
they could entertain in acquiring English, but it is by no means the description
of a coherent existing grammar.
This latter point takes me to the empirical side of the hypothesis represented
in (30) and reproduced here as (31), for convenience:
(31) S
NPsubj VP
Peter owns NPobj
Det N
Agreement between
Peter.3sg and V reflects a dog Matching between the article and the noun
S-procedure: Inter-phrasal reflects phrasal procedure, that is, Intra-
Information exchange phrasal information exchange.
According to Plag (2008a: 125), creole languages “display almost

exclusively structures for which no information exchange between
constituents is necessary.” This, the author argues, can explain why creoles
generally exhibit inherent inflectional morphology (e.g., plural on nouns, and
tense and aspect morphemes on verbs) as opposed to contextual inflections
similar to case assignment or agreement. This conclusion is controversial, as
there have been several studies on Romance-based creoles which argue for
the existence of inflectional morphology in these creoles (e.g., Clements
1996; DeGraff 2001a, 2001b, 2009; Henri 2012). In addition most recent
generative analyses of number expression (e.g., Borer 2005) including work
on creole languages (e.g., Ritter 1991, 1992, 1995; Depréz 2001, 2004, 2007;
Aboh 2004a, 2006a; Stewart 2007; and several chapters in Baptista and
Guéron 2007; Bobyleva 2013) assume that number is determined
structurally during the derivation of the noun phrase. Under such views
plural morphology is determined contextually, and only the noun stem is
selected from the lexicon. According to such analyses, there is no formal
distinction between contextual inflection and inherent inflection when it
comes to structure building.
Setting this aside, however, the notions of intra-phrasal versus inter-
phrasal information exchange, as developed in Plag (2008aff.), are too
obsolete to apply to agreement phenomena as they relate to morphology
(see DeGraff 2009: 949 and later works). According to structure (31), the
matching between a noun and its determiner results from intra-phrasal
information exchange (i.e., agreement inside a constituent) as opposed to
subject–verb agreement, which results from inter-phrasal information
exchange (i.e., agreement across phrases). The characterization therefore
suggests that agreement works two ways: inside and across constituents with
possibly different morphologies associated with them. This is of course
contrary to what is generally accepted in the literature where agreement is
seen as a configurational process between a head and its modifier or
complement (see, for instance, Booij’s 2005 discussion of the role of
inflection). To illustrate this let us consider the French noun phrase in (32):
(32) les très grands hôpitaux
the.PL very big-PL hospital.PL
‘the very big hospitals’
What we observe here is that number marking on the determiner les is
resumed on both the adjective and the head noun. In the case at hand, the
presence of number marking on the adjective is made audible thanks to liaison.
According to Booij (2005: 110), this is an instance of contextual inflection, that
is, “a case of government in which the number of nouns (normally a case of
inherent inflection) plays a role in contextual inflection.” Interestingly, creole
languages show very many instances of such agreement. The Saramaccan

example in (33), for instance, shows that there is systematic contextual
agreement between the quantifier, the determiner, and the numeral.
(33) a. híi dee mii u 7 jaa [Rountree and Glock 1992: Chart 1]
all the.PL child of seven year
‘all those/the seven year old children’
b. dee dii womi
the.PL three person
‘those/the three men’
c. di wan kodo womi
the.SG one single man
‘the one single man/the only man’
It appears from these examples that the plural determiner dee can co-occur
with híi ‘all’ and numerals above ‘one.’ On the other hand, the singular
counterpart di only co-occurs with the numeral ‘one.’ This is precisely what
Booij (2005: 110) refers to as governed inflection, a clear case of contextual
inflection in his framework. As Booij (2005: 109) explains, “agreement is not
always marked by means of morphology but may also be marked through the
choice of a specific lexical item” (e.g., the choice of the relevant determiner).13
It is therefore misleading to claim that creole languages generally lack
contextual inflections. In the case of the examples cited above it is perfectly
arguable that the relevant inflection is a null morpheme that is conditioned by
contextual agreement. If we were to make any generalization about creoles, all
we could say is that they tend to lack concord, according to which agreement is
morphologically marked both on the head and on its dependents. In other
words, creoles languages tend to lack redundant morphology, which is not a
characteristic unique to them. The same pattern is found in Gbe (Kwa) and in
Sinitic, to name just two language families.
Going back to the French example (32), note that the adjective represents a
phrase that is internally modified by a degree element. Under common
syntactic analysis, the phrase can be minimally represented as in (34), where
the determiner heads the Determiner Phrase (DP) which contains the head noun
and the adjectival phrase. The two are contained in a phrase labeled here as a
Functional Projection (FP) for expository purposes but the interested reader is
referred to Abney (1987), Szabolcsi (1987, 1994), Longobardi (1994), Cinque
(1994, 2010), Aboh (2004a, 2004b, 2006a), among others for a detailed
analysis.
13
Unfortunately, Plag’s columns include misrepresentations of Booij’s (2005) careful and detailed
discussion of the role of inflection and the distinction between contextual and inherent
morphologies. Booij’s analysis contradicts Plag’s claims on numerous counts.
(34) DPA
DA FPA
les
APA NPA
tres hopitaux
grands
As the index ‘a’ (agreement) shows, agreement in (34) arises between the
NP that combines with the AP to form FP. The latter agrees with the D it
combines with to form DP. In current generative terms, adjectival phrases are
assumed to realize a specifier position. With this in mind, we can further say
that the agreement in the noun phrase in (32) arises as a consequence of a local
relation between the noun head and the adjectival modifier in its specifier, while
agreement between D and FP arises from a local relation between a head and its
complement. Given this very simple description, what first looks like a simple
matching process (as assumed in Plag 2008a, contrary to Booij 2005) is
actually a succession of locally construed agreement relations that involve a
head and a phrase in its specifier or a head and a phrase in its complement
domain. But because the relation is always between the head of a phrase and
that of another phrase, it necessarily involves inter-phrasal information
exchange regardless of whether these phrases belong together in a larger
phrase that could be a noun phrase, a relative clause, or a sentence. As a
matter of fact, note for instance that French (35a) and other Romance
languages display subject–verb agreement instantiating a local relation
between the auxiliary verb under T, the head of a tense phrase, and the noun
phrase in [spec TP] as illustrated in (35b):
(35) a. Les travailleurs ont manifesté dans la ville
the-PL worker-PL have.3PL demonstrated in the city
‘Workers have demonstrated in the city.’
b. TP
spec T’
DP
T VP
les travailleurs ont
manifeste dans la ville
On the other hand, languages like Edo display verb–object agreement as

illustrated in (36a–b) and schematized in (36c). In (36b), the verbal extension
(ve) -lé ̣, is only compatible with plural objects (Osamuyimen T. Stewart 1997
and Yuka and Omoregbe 2010).
(36) a. Úyì dé ̣ éwù [Yuka and Omoregbe 2010: 13]

Uyi buy.PST shirt
‘Uyi bought a shirt.’
b. Úyì dè ̣-lé ̣ avbé é ̣wù.
Uyi buy.PST-VE PL Shirt
‘Uyi bought shirts.’
c. TP
spec T¢
DP
T VP
Spec V¢
V DP
Once we allow ourselves to consider this relatively simple description

of how agreement works, we immediately see that Plag’s (2008a) analysis
in (31) is far too simplistic and cannot hold. Both the structures in (35b)
and (36c) involve contextual agreement with inflectional morphology on
the verb. While the representation in (35b) instantiates agreement between
a head and its specifier, that in (36c) illustrates agreement between a head
and its complement. Another fact that we learn from the Edo example in
(36b) is that even though there is agreement between the plural verb dè ̣-lé ̣
‘buy-pl’ and the plural noun phrase avbé é ̣wù ‘shirts,’ number morphology
does not occur on the noun head but is realized on the determiner instead.
This is additional evidence that agreement is calculated at the phrasal
level. It is a configurational phenomenon that systematically involves
inter-phrasal information exchange. Under this view there is no possible
distinction between what Plag refers to as ‘intra’ versus ‘inter’ information
exchange.
This leads to a conclusion often overlooked in creolistics that contextual
agreement need not necessarily lead to the presence of an overt affix on the
relevant dependent (Booij 2005). This fact disputes the claim that creoles
lack contextual inflection (or morphology) of the type observed in some
agreement phenomena in Romance and Germanic, as pointed out by DeGraff
(2009). A case in point is Saramaccan which displays a relativizer (different
from wh-pro forms) that agrees with the head noun. This interesting property
is illustrated by the contrast in (37a–b) taken from Huttar, Aboh, and Ameka
(2013):
(37) a. Di fisi di mi tata kisi bigi

DEF.SG fish REL.PL 1SG father catch big
‘The fish that my father caught is big.’
b. Dee fisi dee mi tata kisi bigi
DEF.PL fish REL.PL 1SG father catch big
‘The fish that my father caught are big.’
In (37a), di fisi ‘the fish’ is singular and combines with the singular
relativizer di, while in (37b), dee fisi ‘the fish/es’ combines with the
plural form dee. These Saramaccan facts are reminiscent of the Edo facts
discussed above, according to which agreement between the noun phrase
and the verb head triggers agreement morphology on the verb and on the
determiner associated with the noun but not on the noun itself. Here
agreement morphology shows up on the relativizer heading the relative
clause and the determiner inside the relativized noun phrase, but not on
the noun itself.
Since Chomsky (1977), relative clauses have been analyzed as involving a
link between the relativized noun phrase, the complementizer hosting the
relative wh-pronoun or the relativizer, and the gap associated with the base
position where the relativized noun phrase is interpreted. This is informally
schematized in (38) for sentence (37b), in which the position of the gap
contains [dee fisi].
(38) [dee fisi] dee mi tata kisi [dee fisi] bigi
Chomsky’s position is supported by data from Akan, in which the base

position of the relativized noun phrase is obligatorily realized by a
resumptive pronoun, represented here in boldface (Saah 2010: 92).
(39) Me-hu-u ɔbáá áà Kofi wáré-e no nó
1SG-see-PST woman REL Kofi marry-PST 3SG CD
‘I saw the woman whom Kofi married.’
Setting aside the question of whether the link between the relativized noun
phrase, the relativizer, and the base position of the noun phrase arises as a
consequence of syntactic movement, we can assume that, in the Saramaccan
example in (37b), the relativized noun phrase dee fisi and the relativizer-
complementizer dee do not start out as part of the same NP constituent *[dee
fisi dee]. Instead, the right analysis is that the relativized noun phrase and the
relativizer belong to two separate constituents that together form a larger
relative clause as indicated by DP2 in (40).
(40) [DP2 [DP1 dee fisi] [CP dee mi tata kisi [DP1 dee fisi] bigi]]
Under this partial representation, the agreement between the relativized DP1
dee fisi and the relativizer dee inside the clause CP is a clear example of
contextual agreement signaled by contextual morphology (or inflection) on
the relativizer.14 Taking the description a step further, we can informally
describe the di/dee alternation in Saramaccan as in (41), with the change
being sensitive to the context in which the relativizer occurs, that is, in the
context of singular versus plural relativized noun head.
(41) REL[SG]= d + INFL[SG] ➔ d-i
REL[PL] = d+ INFL[PL] ➔d-ee
In a sense, the Saramaccan di/dee variation reminds us of French que/qui
alternation as in (42a–b), where the complementizer qui indicates subject
relatives as opposed to que, which marks object relatives. According to Rizzi
(1996), the que/qui alternation involves agreement. Note that in Standard
English the agreement occurs in the wh-phrase.
(42) a. L’homme que j’ai invité <l’homme>
the.man that.OBJ 1SG.have invited
‘The man whom I invited.’
b. L’homme qui a invité Jean
the.man that.SUBJ has invited John
‘The man who invited John.’
Complementizer agreement is rather common and one finds the direct
equivalents of the Saramaccan di/dee alternation in Gengbe and Ewegbe
(Western Gbe), where the relativizer is sensitive to number agreement.
Examples (43a-b) are from Gengbe:
(43) a. [[Axɔ́ má] [CP kè lè kplɔ̌ á jí á]] nyɔ́ . [Kangni 1989: 26]
Book REL BE.located table DET on DET be.good
‘The book that is on the table is good.’
b. [[Axɔ́ má] [CP kè-wó lè kplɔ̌ á jí á]] nyɔ́ .
book REL-PL BE.located table DET on DET good
‘The books that are on the table are good.’
Similar examples are also found in South Hollandic Dutch (Germanic),
Zwart (2006: 68):
(44) . . . Jonge-s die-e werk-e wil-le. [South Hollandic Dutch]
boy-PL REL-PL work-INF want-PL
‘ . . . Boys that want to work.’
14
See DeGraff (2009: 950–952) for a discussion of other examples of inter- and cross-clausal
dependencies that further illustrate the inconsistency of Plag’s analysis.
These facts clearly show that, like non-creole languages, creoles exhibit
contextual inflectional morphemes that participate in agreement phenomena.
The fact that similar contextual inflectional morphemes are not typically found
in the verbal domain in creole languages need not be attributed to a putative
interlanguage origin, no more than the same cannot be claimed to account for
the extreme paucity of English verbal inflections (compared to other Germanic
languages) or for the total lack of verbal inflections in Gbe and Sinitic. Not only
do the facts presented here contradict Plag’s (2008a) claims about creoles in
general, they also show that Saramaccan speakers can select the right noun
phrase that agrees in number with the equivalent relativizer. This indicates that
their language does not reflect fossilization of some early-stage interlanguage.
If anything, these constructions would correspond to advanced stages, 3 or 4 in
Plag’s (2008a) scheme.15
A final empirical fact against the interlanguage view comes from
work on SLA which Plag (2008a, 2008b, 2009a, 2009b) ignored in his
columns, though this study is very relevant to his hypothesis. Klein and
Perdue (1997) conducted a longitudinal cross-linguistic study of second
language acquisition outside the classroom. The study involved a total of
40 Punjabi, Italian, Turkish Arabic, Spanish, and Finnish adult learners
of (varieties of) Dutch, English, French, German, and Swedish and lasted
30 months, three times longer than John Schumann’s (1978) study on
interlanguages. They observed that all the learners developed a
communicative system that:
– seemed to be determined by the interaction of a small number of organiza-
tional principles,
– was largely (though not totally) independent of the specifics of source and
target language organization,
– was simple, versatile and highly efficient for most communicative purposes.
(p. 303)
The authors refer to this system as the Basic Variety (BV). They further
observed that for about one third of the learners investigated, acquisition ended
on this structural level, some minor variation aside, they only increased their
lexical repertoire and learnt to make more fluent use of the BV (Klein and
Perdue 1997: 303).
With regard to the structural properties of the BV Klein and Perdue (1997:
332) give the following description:
15
One cannot resort to plain substrate influence or full access to account for this phenomenon
since the Saramaccan relative clause in (36) is structurally different from the Gbe relative
clauses in (42). See Aboh (2006a) for a discussion of the structure of the DP and relative clauses
in Gbe and Saramaccan.
1. Adult second language learners (outside the classroom) regularly develop a

particular form of language, the ‘BV.’ Some of them fossilize at this level,
that is, they keep its structural properties and only enrich the lexical
repertoire, whereas others complexify the variety to a greater or lesser
extent.16
2. The lexicon of the BV is essentially taken from the target language (with
some borrowings from other sources). It mainly consists of (uninflected and
often phonologically distorted) open-class items; closed-class items appear
but are rare. Formation of new words is limited to noun-noun compounds
(my emphasis).
3. Structurally, the BV is characterized by a small set of organizational prin-
ciples. It is the interaction of these principles which determines, for exam-
ple, the concrete form of utterances or the way in which time and space are
encoded. These principles seem to be the same for all learners, irrespective
of source and target language. What varies to some extent is their
interaction.
4. Strikingly absent from the BV are (a) free or bound morphemes with purely
grammatical function and (b) complex hierarchical structure, in particular
subordination. (my emphasis)
The observations in 1–4 are relatively close to Plag’s (2008a) stages 1
and 2 of interlanguages. Because of its cross-linguistic nature and because it
is structurally close to interlanguages of an early stage as described in Plag,
BV appears to be a very good metric for evaluating the interlanguage
hypothesis.
Significant properties of creoles are absent from BV. As is obvious from
points 2 and 4, it exhibits relatively few grammatical or functional items, and it
does not involve TMA markers, unlike creoles. According to Klein and Perdue
(1997: 312, my emphasis):
There are a few quantifiers, a word for negation, a few prepositions with overgeneralized
lexical meaning, but no complementizer and, . . . no inflectional morphology, hence no
markers of agreement, tense, aspect or case. In other words, the repertoire consists
mainly of ‘open-class’ and a small list of ‘closed-class’ items with lexical meaning.
There are some determiners (in particular demonstratives) but hardly ever a determiner
system.
16
Klein and Perdue further argue that
One way the learner has of improving his or her expressive power is simply to enrich
vocabulary, especially by adding temporal adverbials, and to perfect technique on this instru-
ment. And about one-third of the 40 learners whose acquisition was investigated do exactly this:
they do not go beyond the BV, but they steadily improve it in these two respects.. . . Speakers of
the BV can say what they want to say about temporal relations – not what the structure of the
language forces them to say. (p. 323)
In addition, BV displays a reduced adpositional system as well as a limited

pronominal system that only involves the speaker, hearer, and third person, but
there is no anaphoric pronominal reference to inanimates. Because BV lacks
essential grammatical ingredients such as complementizers, it also does not
exhibit subordination.
It appears clearly from these differences that BV (or similar
interlanguages) cannot be equated with creoles, contrary to Plag’s (2008a,
2008b, 2009a, 2009b) claims, as pointed out earlier by DeGraff (2001a,
2001b). In this regard, the absence of TMA markers in BV as well as in
interlanguages represents a serious challenge to Plag’s hypothesis. For the
past thirty years, an important focus of the study of creole genesis has been
to account for the fact that the creoles’ creators did not retain the verbal
inflectional morphology of the target languages (viz., French, English,
Spanish, Portuguese, or Dutch) but developed a series of free morphemes
that express tense, aspect, and modality and occur in a fixed order captured
in the abbreviation TMA. In creolistics periphrastic TMA makers constitute
the most central property of these new languages that requires explanation
(e.g., Bickerton 1981, 1988a, 1988b, 1999; Chaudenson 2001, 2003;
Lefebvre 1998). With regard to the interlanguage hypothesis, Plag (2008a,
footnote 2) claims that
While . . . the loss of superstrate TMA morphology can be nicely accounted for by the
interlanguage hypothesis, the emergence of new preverbal markers cannot be straight-
forwardly explained by this hypothesis, since such markers are not prevalent in
interlanguages.
It is not clear what the author means by “prevalent,” but recall from Klein
and Perdue’s (1997) work that TMA markers are simply absent from BV. We
can therefore assume that they are generally absent from interlanguages.
I showed in the previous paragraphs, that the premises of the interlanguage
hypothesis are flawed and that the theory incorrectly predicts that creole
languages lack contextual inflection (contrary to the Saramaccan facts
discussed above). Accordingly, this theory fails on two counts: First it
provides no explanation for why verbal inflection is lost, while other
contextual inflectional affixes are retained. Second, more so than any other
theory of creole genesis discussed in this chapter, it fails to explain why the loss
of verbal morphology led to the emergence of an extraordinarily subtle and
complex grammatical system of TMA markers that has fascinated all students
of creoles for the past thirty years or so.
Regarding the question of whether creole languages can ever be compared
to interlanguages or BV, Klein and Perdue (1997: 340) correctly point out
that, aside from the fact that so-called pidgins are often assumed to display
rigid SVO order and lack inflectional morphology, “a precise comparison,
however, is quite problematic.” According to these authors, it is difficult to

assume that there is a unique underlying structure for all the languages that
are referred to as pidgins or creoles given that these languages show
considerable variation among themselves, regardless of whether or not
they share the same lexifier. The logical conclusion is therefore that these
languages can only be characterized in terms of similarities but not as
involving consistent isomorphic grammatical structures. The variation we
observe across creoles (especially within the nominal domain or within the
clausal left periphery) constitutes strong empirical evidence for the position
that these languages cannot be assumed to have all emerged from a unique
basic system comparable to BV or a fossilized interlanguage of an early
stage.
3.5 Conclusion
This chapter has shown that monolithic theories that account for creole genesis
exclusively by invoking inheritance from the superstrate, substrate influence,
language universals, or fossilization of some early interlanguage stage all fail
both to explain the relevant properties of creole languages and to relate the
emergence of creoles to more relevant aspects of language acquisition. In a
way, what the theories discussed here aim at is a description of the type of
languages that learners may develop when they fail to acquire the target
language owing putatively to a break in its transmission, a restricted access
to it, a disrupted unguided learning, or, worse, possible psychological
impairment (as may be inferred from the interlanguage hypothesis). Because
these theories are formulated in holistic terms, they try to account for the so-
called creole system as a whole and therefore miss the important subtle
variations that may be observed across or within the modules of a language,
e.g., between morphology and syntax, or between DP and CP. Such variations
may be more indicative of how the different ingredients of the languages
in contact were combined into a coherent grammatical system: the speaker’s
I-language (or I-creole).
A point that is particularly damaging to the SLA failure hypothesis or any
variant thereof is that it hypothesizes that a stable contact language like a creole
can emerge only if:
(i) the population of learners is highly heterogeneous (as is supposed to have
been the case with the enslaved Africans), and
(ii) these learners have limited or no access to an adequate linguistic model
representing the target language.
A problem with this view is that not all contact languages that are regarded
as creoles or share superficial structural similarities with them arose in such
3.5 Conclusion 111
a context. A notable exception is Pitkern. After a mutiny, the creators of

Pitkern settled on Tubai (Tahiti) in 1790. The founder population then
consisted of nine male mutineers and nineteen Polynesians including
six men servants, twelve women and a child, thus a total of twenty eight
people (Thomason and Kaufman 1988; Sebba 1997). With regard to their
linguistic profile, the mutineers included four English men, two Scottish, one
American, one person from St. Kitts in the Caribbean and one other from
Guernsey. We might therefore conclude that the mutineers who were
speakers of different varieties of English probably developed a koiné that
served as means of intra-group communication. The Polynesians were
all speakers of Tubaian and knew very little English (Sebba 1997: 136).
The English-speaking mutineers probably occupied a dominant position in
this new society, though they must have been very close to the Polynesians.
Indeed, the island is only about 5 km2 and the daily subsistence activities
required by settlement in a new territory must have created tight links
within the community and given the Polynesians ample opportunities to
practice and learn the emergent English koiné. However, this is not what
happened; a creole emerged instead. This language must have arisen
immediately after settlement, given that within ten years, by 1800, all but
one of the men of the island had died, thus reducing the population to one man
(John Adams, probably the only surviving fluent English speaker), the six
Polynesian women and twenty three children. The surprising fact here is that
at this stage, the Polynesians were the majority and could have decided to
switch back to Tubaian. They did not and only the creole survived on the
island.
It is obvious that none of the socio-historical reasons put forward
to support the SLA failure hypothesis in the context of the plantation
settlement colonies can hold for Pitkern. It is true that Pitkern is
exceptional in its own way, having arisen in a situation of extreme
physical-ecological adversity but probably within a strongly cohesive
community. Yet, what this example teaches us is that while population
structure may affect the input that learners are exposed to, it does not seem
to predetermine how acquisition proceeds and it certainly tells us nothing
about what variants speakers will develop. All that remains is that creoles
and contact languages emerge in situations where speakers develop new
linguistic variants that serve certain functions, e.g., intra- vs. inter-group
communication, secret code, and language of the work place. If an
ecological context is particularly favorable, such variants may become
established and propagate across a whole speech community. The
development of such variants, I claim, only necessitates human linguistic
capacity as deployed in language acquisition. More precisely, though
such contact languages usually develop grammatical properties that
diverge from their source languages, they do not require a different

explanation from what we ought to give for more common language-
learning experiences (which are the driving force of linguistic change over
millennia). As I show in the next chapter, one way to address this issue is to
adopt Mufwene’s (2001ff.) ecological perspective on language evolution.
4 Competition and selection
A question we were left with in Chapter 3 is why the mutineers and the
Polynesians stranded on Pitcairn Island developed Pitkern? Given the size of
this island and the structure of the clan it would have been possible for the
Polynesians to acquire the English vernacular, probably a koiné, spoken by the
mutineers, just as the latter could have shifted to Tubaian. Even more puzzling
is the fact that even though almost all the English speakers (including other
males on the island) had died within the first ten years of settlement – leaving
aging Sir John Adams as the only surviving linguistic model for English – the
remaining six women, mothering the twenty-three children, and who then
represented the absolute majority on the island, did not shift back to Tubaian.
Instead, this apparently tight community developed a new language from the
mutineers’ English vernacular. None of the sociological scenarios often evoked
in creolistics to uphold the exceptionalist view of creole development can
account for this case. While some authors tend to ignore the development of
Pitkern and similar cases, Thomason and Kaufman (1988: 149) argue on the
basis of Pitkern that “the evidence for the traditional view that a full-fledged
pidgin language, relexified or not, was the major direct source for every
plantation creole (not to mention Pitcairnese) strikes us, as it struck others,
as thin.”
Indeed, it appears that the development of Pitkern is comparable to similar
situations in the Caribbean where the Africans maintained the emerging creole
even though geopolitics and population factors allowed them to make a
different choice: for instance, to switch (back) to an African language of their
choice. A case in point is Saramaccan which continued to develop after
the Africans had escaped the plantations. As explained in Chapter 2, the
Saramaccans kept very many aspects of Gbe culture (e.g., the structure of the
clan and its chieftaincy, religious practices, and culinary practices), as well as
many basic vocabulary items. These are clear evidence that there must have
been enough Gbe speakers around to pass on this legacy to the Saramaccan
This chapter builds on Aboh (2009a), the results of which are presented again here and updated
according to new findings.
113
language and society. The same could be said of Haiti after Independence. The
discussion in Chapter 2 shows that the plantation colonies of Suriname and
Haiti hosted significant groups of people of Aja descent who could have
developed an intra-group Gbe vernacular, which could have enabled them to
maintain a more obvious cultural link with home. In both cases, the leaders
of these communities clearly rebelled against the colonial power and one
could imagine that this rebellion could have translated into the adoption or
development of a new language different from the language of their oppressors.
But this apparently did not happen. Just as in the case of Pitkern, the runaway
slaves, and thus independent colonies, maintained (and further developed) a
language born in the early stages of the settlement.
Because of the important role the founder populations play in the development
and maintenance of the emerging community language, it appears that the
properties of this language will be determined fundamentally by the properties
of the languages in contact at the time of settlement. Assuming that people do not
hold meetings around a fire to plan for the creation of a new language (except
for criminal, war, or other secret activities), the only reasonable hypothesis we
can develop in this respect would be to say that the structural properties of the
emerging language will be shaped by learning hypotheses that speakers
entertained while trying to learn from one another during their interactions.
Mufwene (2001: 28) formalizes this hypothesis in terms of the Founder
Principle in order “to explain how structural features of creoles have been
predetermined to a large extent (though not exclusively) by characteristics of
the vernaculars spoken by the populations that founded the colonies in which
they developed.” According to Mufwene’s biological approach to language
evolution, the founders’ languages, which actually represent an ensemble of
idiolects, provide language learners with a pool of linguistic features on which
acquisition feeds. Going back to the case of Pitkern, the feature pool then
presumably consisted of the English vernaculars of the mutineers (which may
have led to an English koiné) as well as the indigenous language of the
Polynesians (viz., Tubaian). Adapting Mufwene’s (2001: 4) description of the
feature pool, this situation can be represented as in (1) (see also Ansaldo 2009:
100ff. for a similar discussion).
(1) English Scotts American St. Kitts Guernsey Tubaian
Feature Pool: linguistic features on which

learning hypotheses are based
Nascent idiolects: Pitkern
Competition and selection 115
In this representation, I’m assuming, along with Mufwene, that contact

first happens at the individual speakers’ level, in the mind of the learner, who
develops a specific idiolect. Cumulative selection of similar properties across
idiolects leads, by convergence, to a communal language, that is, an
E-language.
In the case at hand, we do not know what motivations the Pitcairnese had to
develop a creole or a new contact language. However, we can already see from
the feature pool description in (1) why such a language could arise in this
specific context (arguably independently of the political intentions of the
learner). To see this, it is important to contrast the Pitcairnese situation to that
of a monolingual context: for instance, a child or early L2 learner acquiring
French in a tight monolingual French-speaking community). This situation can
be described as in (2):
(2) Standard French Home French Regional/local French
Feature Pool: linguistic features on which

learning hypotheses are based
Nascent idiolect: French
As we can see from these two representations, the only difference between
(1), which applies to a learner who ended up speaking Pitkern, and (2), which
applies to a learner of French, is that (1) is more likely to allow a much wider
range of variation than (2) (Aboh and Ansaldo 2007; Ansaldo 2009 and
references cited therein). While the range of variation in (1) is set between
closely related varieties of English and the typologically distant Tubaian, that in
(2) lies within very closely related varieties of French only, notwithstanding
inter-idiolectal variation among speakers of the same dialect. This leads me to
conclude that the variation that is observed across speakers/learners is partly
determined by the feature pool they were exposed to. According to Mufwene
(2001: 5), the output (that is the developing idiolect) “represents variation in the
ways particular (combinations of) features were selected into the emergent
varieties.” What this means is that new varieties are not necessarily different
from their source languages with regard to their fundamental linguistic
ingredients (i.e., features) but rather with regard to which specific features
have been selected and how they have been recombined and/or modified by
learners. In this regard, Mufwene (2001: 5) concludes that “what makes the
new varieties restructured is not only the particular combinations of features
selected, often from different sources, into the new language varieties, but also
the way in which the features themselves have been modified, ‘exapted’, to fit
into the new systems.”
A particularly interesting aspect of Mufwene’s approach to language contact
and language change that is evident from this very simple and brief description
is that language acquisition (whether first or second) implies contact of some
sort. This ‘contact’ happens in the mind of the speakers via the feature pool
which presents the learner with the input on which learning hypotheses are
made. Because of this, the output, that is, the developing idiolect, is partly
determined by the input in the feature pool and partly by hypotheses that the
learner entertained about that input. Clearly, this view (more than any other
theory of creolistics and language creation) provides us with an overarching
framework in which we can investigate variation within and across speakers,
learning hypotheses that speakers make, and how individual selections made
by speakers can propagate within a community as part of the communal
language. Most important of all is that this framework is universalist in that it
requires no exceptional hypothesis or device but the human language capacity
(as generally understood in modern linguistics) and how the ecology of the
speaker/learner may induce variation.
While Mufwene’s work draws a lot on the development of a language at the
population level (i.e., E-language), a major task that I undertake in this book
is to understand variation at birth. That is how learners/speakers select
linguistic features from the input they are exposed to (i.e., the feature pool
accessible to individual learners) and how in so doing they develop new
variants that in turn enter the feature pool. In order to conduct this research
in a reasonable way, we first need to know (or imagine) what linguistic
input an African learner/speaker could be exposed to on a plantation in
seventeenth- and eighteenth-century Suriname or Haiti. Based on the
findings on population factors in Chapter 2 and on discussion in previous
paragraphs, Section 4.1 proposes a scenario for the development of creoles
that is inspired by Mufwene’s approach. In Section 4.2, I further introduce the
reader to aspects of Mufwene (2002, 2005a, 2005b) that are relevant for the
theory of language contact and language acquisition that I put forward in this
book. In this regard, Section 4.3 presents the version of the competition and
selection model adopted here with regard to syntactic structures. Section 4.4
anticipates subsequent chapters in evaluating the proposed framework
on empirical data from the Suriname creoles, Saramaccan and Sranan.
Section 4.5 concludes the discussion.
4.1 Competition and selection: setting the stage

As discussed in Chapter 3 and in the introduction above, theories of
genesis assume various language contact situations that might have led
4.1 Competition and selection: setting the stage 117
to the emergence of a creole. According to Mufwene (2001), creoles are

restructured varieties or approximations of approximations of their
lexifier, which itself is a koiné that developed or was developing from
the contact of diverse metropolitan dialects (see also Chaudenson 1992,
2001, 2003). A crucial notion to Mufwene’s work, however, is that such
new languages result from a process of competition and selection in the
mind of the speaker during the development of idiolects that cumulatively
converge to the emerging E-language. This process was of course
affected by the patterns and structures of the native languages of the
enslaved Africans.
Before engaging in the discussion of the notion of competition and
selection from a syntactic point of view, this section first presents a new
language contact scenario that is compatible with the population factors and
the linguistic aspects discussed in Chapter 3. The discussion there showed
that, in the case of Haiti and Suriname, Africans of Aja descent formed an
ethnolinguistically homogeneous critical mass during the founding period.
According to the Founder Principle, we therefore expect such communities
to have exerted an important impact on the nascent creole society. Given
this view, I argue that the process of competition and selection that operated
in the formation of a ‘lexifier-koiné’ also operated simultaneously among
the African languages (e.g., varieties of Gbe and related Kwa) to form an
African koiné, call it substrate koiné when speakers of related or the same
languages found themselves on the same plantations. The contact between
lexifier and substrate koinés may have led to a new communal language that
will develop into a creole. It is conceivable that the developing creole went
through different stages that involved qualitatively different koiné-types.
This view has never been systematically investigated in the creolist
literature because the implicit (but misleading) assumption is that the native
languages of the Africans ceased to exist from day one on the plantation,
where they had to shift to the master’s language. The received wisdom is
that the Africans came from too varied origins to be able to communicate
with each other using an African language, hence the myth of mutual
unintelligibility.
4.1.1 On the myth of mutual unintelligibility

The hypothesis that African languages disappeared from the plantations soon
after the slaves arrived presupposes the scenario described in (3) where L1 to
L(n+1) represent the lexifier varieties on the plantation.
(3) Colonial settlement

L1……L2……..L3………L(n+1)
Colonial
White population
L(koine) = Linguistic
model Cohorts of Africans
Restructuring
Simplification
Disintegration
Pidgin
Local standard variety Creole
As we can see from this representation, contact between the settling

communities led to two linguistic paths. The European colonists only
contributed to the emergence of a lexifier koiné (or linguistic model) that
eventually developed into a local standard variant. This standard variety
arises from the competition between the different variants spoken by the
colonists. The term standard, as used here, refers to a locally restructured
variety of the metropolitan standard, such as Jamaican Standard English or
Quebec French. In trying to acquire this lexifier koiné, the only accessible
target, the Africans ended up developing a creole. According to this scenario,
the creole is ultimately an Afro-African creation. Some authors will even argue
that the two developmental paths in (3) led to two mutually unintelligible
languages: the local standard and the creole (but see Chaudenson 2003 for a
critique). We can further imagine that such unintelligibility emerged very early
in the development of the creole, possibly prior to the pidgin stage. Assuming
all this is fine, a question we ought to ask ourselves is: How could such a
scenario apply to the orderly functioning of the plantations that sugar or
tobacco business requires?
Because most theories of the development of creoles appear to assume the
structuring dichotomy in (3), the issue is never discussed in any detail. A
crucial point about the representation in (3) and similar scenarios is that they
lead to the very simplistic assumption that there was only one relevant target on
the plantation to which the Africans owe their survival. Under such scenarios
therefore the enslaved Africans immediately switched to the lexifier koiné (or
to a communal pidgin) on their arrival. As the argument goes (cf. Chapter 3),
this shift is made necessary by the fact that the Africans shared no common
language and must rely on the koiné or pidgin for survival. Yet, several facts
about plantation colonies go against this common view in creolistics.
It is indeed true that, after experiencing the middle passage, the Africans had
to acquire the local koiné as quickly as possible for daily survival. However, the
discussion in Chapter 2 teaches us that because of the way slavery was
structured on the Slave Coast, very many enslaved people of Aja descent
ended up on the plantations of especially Suriname and Haiti, where they
formed relatively homogeneous ethnolinguistic critical masses at different
stages of the development of these colonies. Another conclusion we reached
in Chapter 2 is that many of such enslaved Africans were probably
multilinguals, capable of speaking another language in addition to their
mother tongue. Put together, these considerations lead me to conclude that
typological linguistic similarity, related ethnicity, and cultural practices will
rather favor efforts from the Africans to learn each other’s languages rather
than switching immediately to the language of the colonist. Put more
concretely, I’m assuming that daily survival dictates that the enslaved
Africans not only learn the master’s language available to them as quickly as
possible but also to try, whenever possible, to immediately relate to and
communicate with other enslaved mates who they could identify as
belonging to the same regional ethnicity (e.g., the Aja descents from Tado to
ŋɔ́ tse, cf. Chapter 2). As I once put it during a creole meeting in Leipzig in 2005,
if an Eastern Gbe speaker meets with a Western Gbe speaker on a plantation, it
will probably take them less time to figure out how to say “à fɔ́ n gànjí
(Gungbe)?” or “ò fɔ́ n nywέn ɖé à (Ewegbe)?”/you wake.up well?/ that is
‘good morning’ in each other’s language rather than wondering about how
to say “how do you do,” or “comment allez-vous” in French (Aboh 2005a).
Note that the verb ‘wake.up’ is identical in these two Gbe languages.
Multilingualism was common in western Africa before the European
colonization and there is no good reason to assume that, after the middle
passage, the Africans suddenly lost this basic human ability. Indeed, the
closeness of the Gbe languages as discussed in Chapter 2 can only favor this
mutual learning. If we agree that some slaves (say some Eastern Gbe speakers)
may have tried to learn or accommodate with other Gbe varieties (e.g., Western
Gbe like Ewegbe), then there is no principled reason to assume that as normal
human beings endowed with language capacity they could not have tried the
same with less related languages such as Yoruba or even typologically different
languages such as Kikongo.
If we can allow ourselves to give a little bit of humanity back to these
enslaved Africans so as to entertain the situation just described (be it for a
tiny fraction of the population) then we are irremediably led to conclude that
the language contact situation on the plantations was probably more complex
than usually assumed. Actually, the commonly assumed founder scenario in
(3) appears very improbable. Various other scenarios come to mind, but the
basic hypothesis that I will be assuming in this book is described in (4) below.
4.1.2 An apparently different story

Consider the representation in (4):
(4) Colonist community Contact situations Slave community
La…….La+1……….Lb Sa……Sa+1…..Sb
La; a+1……..Lb Sa; a+1 ……..Sb
Stage 0 ..La; b…………Plantation contact variety…..……Sa; b…….
Synchronic Colonial language New language (creole) Survivals of once spoken

continuum African languages
In this representation, the series (La, La+1 . . . Lb) in the upper box to the left
represent the different colonial varieties or languages spoken by the colonists.
In the box to the right, the series (Sa, Sa+1 . . ..Sb) represent the African substrate
languages spoken on the plantation. The middle boxes represent koiné-like
varieties or vernaculars that must have arisen from language contact within
the colonist community and the enslaved Africans, respectively. As indicated
by the labels La vs. Lb and Sa vs. Sb, different sub-communities may have
developed different vernaculars. Assuming a process of competition and
selection, where some features are selected over others, this development
continues until stage 0 which involves levelled varieties or vernaculars in
both the colonist community and the enslaved African community.1 For
convenience sake, I refer to these varieties as plantation varieties.
As also indicated by the bold, vertical broken line labeled “contact
situations,” I assume that contacts between the colonist community and the
enslaved African community triggered mutual language learning which gave
rise to a plantation contact variety at stage 0. What this means is that
competition and selection between the lexifier and substrate varieties was
often mediated in the nascent plantation contact variety, which would evolve
into the creole as a communal vernacular. In synchronic terms therefore, the
developmental scenario sketched in (4) corresponds to a potentially universal
situation in the Caribbean where a local acrolectal variety evolved hand in hand
with the creole (typically associated with the basilect, though it is a continuum),
with the African languages becoming either extinct or relegated to very specific
symbolic functions in religion, ancestral practices, plant names, etc.
1
As extensively discussed in DeGraff (2002, 2009), children and young learners are likely to be
the crucial agents in the homogenization process of idiolects.
An important aspect of the scenario represented in (4) is that it allows for

various target languages including the nascent creole to compete with each
other. This new perspective gives room for virtually all the vernaculars spoken
on the plantation to compete under local ecological constraints still to be better
understood. In addition, it suggests that the African-born slaves who arrived
later, the bozals, would have completely shifted to a particular plantation
variety (thus abandoning their own native tongues) only when such a variety
was stable and functional enough to enable full integration in the plantation
society. This would have been possible only if there was absolutely nobody else
that they shared their ancestral language with. However, it must have taken a
few generations before such a full-fledged plantation variety evolved out of the
plantation contact language and prevailed as the only vernacular. This latter
point is compatible with the gradualist view of the development of creoles,
though the question of the time span within which such a gradual process took
place is still a matter of investigation.
The evolutionary scenario that I put forward here raises several questions
that merit further discussion. An important point about this representation is
that colonial settlements were fundamentally segregated, at least since the
beginnings of the plantation phase. During the homestead phase, there were
intense daily contacts between the Africans and their masters and/or indentured
servants. But such (intimate) contacts which allowed the enslaved Africans to
acquire the master’s language (often in a context of unconfessed interbreeding)
should not hide the fact that the slaves and the colonists lived in different
households and were subject to different laws. Such a founding segregation
structure allowed the Africans to interact more often among themselves than
with the European colonists and to develop regional African interlanguages,
just as intimate interactions among the colonists lead to the formation of a
colonial koiné.
We should not forget, however, that during the homestead phase, the
strongest driving force against African languages obtained in the Creole
population (i.e., locally born slaves), which may have included several
monolinguals that spoke only the emergent colonial vernacular. This Creole
population was instrumental in acculturating the bozal slaves (Mufwene 2008:
chapter 11).
During the plantation phase, there was less direct contact with the colonists
and I assume that this may have enhanced Afro-African interactions and hence
the survival of more viable African vernaculars whenever possible. For
instance in the case of Suriname, it is arguable that the Gbe and the Kikongo
kept speaking such varieties in addition to learning the colonial language from
each other. For instance, Smith (2009) shows that certain Kikongo words that
are retained in Saramaccan display a deviant nominal class. Assuming that the
Kikongo speakers were natives, it is improbable that they were the source of
such ‘mistakes’ (notwithstanding individual cases of language attrition).

Similarly, Kouwenberg (2009) reports that certain Ijo items restructured in
Berbice Dutch exhibit deviant properties that cannot be assigned to native
speakers of Ijo. What these two documented examples clearly show is that
there were non-native speakers of these African languages on the plantations.
I’m assuming in this book that these speakers included L2 learners who were
not uniquely of African descent. (I return to this aspect of the scenario below.)
While it is a common practice within creolistics to ignore the continued
presence of African languages on the plantation or downplay their potential to
serve as alternative vernaculars in the contact situation, there are ample
historical records pointing to the contrary. An argument that one often reads
in the literature is that African languages quickly disappeared from the colonies
because they were forbidden. While the argument often focuses on such
prohibitions, the real question is why would there be such prohibitions in the
first place if the African languages were not spoken at all on the plantation?
Note that the prohibition of African languages would have been difficult to
enforce rigorously. Nobody could check what languages some slaves spoke in
their cabins when they were alone (Mufwene 2008: chapter 11).
Perhaps the most notable example of an African language that was spoken on
the colonies is Mina-Jeje as described by the missionary De Ântônio Da Costa
Peixoto (1731, 1741) and reedited by Silveira (1945) and more recently by
Pessoa de Castro (2002). Mina-Jeje is the language spoken by the Africans in
the mining town of Vila Rica (Ouro Petro) in the region of Minas Gerais in the
emerging Brazil of the early eighteenth century (Pessoa de Castro 2002). The
descriptive manuscript reedited by contemporary authors dated from 1731, that
is, around the same date when La grammaire abrégée (which I discussed in
Chapter 2) was published. These two publications are interesting for many
reasons. One is that La grammaire abrégée was written precisely to help traders
on the Slave Coast as well as doctors operating in the ports or embarked on the
slave boats, while De Antonio Da Costa Peixoto’s description Lingoa geral de
Mina was designed to help slave masters manage the slave population in an
orderly manner (a crucial aspect of sugar/tobacco/mining business). According
to this author (as reported in Silveira 1945), “if all slave owners, and even those
who did not own any slaves knew this language there wouldn’t be all those
insults, vandalisms, financial losses, theft, death, and finally many more
atrocious cases” (my translation).2 The period of publication of these two
documents 1730–1741 shows that colonists in the embarking zones as well as
in the New World were very active trying to learn the languages of the Africans.
2
“Si tous les proprietaries d’esclaves, et même ceux qui n’en possèdent pas . . . connaissaient cette
langue, il n’y aurait pas toutes ces insultes, ruhinas, prejudices, vols, morts, et finalement
beaucoup de cas atroces.”
As I showed in Chapter 2, this period corresponds to the time when slave

trade on the Slave Coast was at its peak: late seventeenth century up to middle
eighteenth century. Not only do such documents suggest that some African
languages were spoken on the plantations but they also point to the fact that the
slave owners considered the learning of such languages a fundamental aspect
of business managing. Clearly therefore, it is legitimate to assume that these
languages were part of the linguistic feature pool that both the enslaved
Africans and the colonists on the plantations were exposed to. In this regard,
a final very important piece of information that adds to my conclusion in
Chapter 2, viz., that people of Aja descent formed a critical mass on certain
plantations, is that Mina-Jeje is a Gbe language (as the alert reader could have
guessed). Proving this is not hard thanks to the list of lexical items and short
sentences provided in both Silveira’s (1945) and Pessoa de Castro’s (2002)
descriptions. In the context of this discussion, let us consider again the list of
lexical items from La grammaire abrégée presented under example (15) in
Chapter 2 and repeated here under (5):
(5) Grammaire abrégée Gungbe Gloss
noucou nùkun eye
doü ɖù eat
nou nù drink
sa sὲ push
dé ɖè pull/shoot/take away
otton àtɔ̀ n three
All these words appear in Da Costa Peixoto (1741) as reported in Silveira
(1945) and De Castro (2002). For instance, noucou ‘eye(s)’ and otton ‘three’ in
La grammaire abrégée are listed in Lingoa geral de Mina as anucum and utom,
respectively. Interestingly, the additional a- in a-nucum relates to the initial
vowel that marks nouns in the Gbe languages (Chapter 2; Aboh 2005b, 2010b;
Aboh and Smith 2015a). On the other hand the different transcriptions of ‘three’
otton vs. utom probably relate to the variation observed across Gbe with regard to
the initial vowels a-, o-, ɔ-, ε- and e- (e.g., see Table 2.2 in Chapter 2 on a random
wordlist of fifteen Gbe varieties within the five major Gbe clusters).
The four verbs listed in (5) occur in various contexts as described below. In
these examples, the first line represents Mina-Jeje (as described in Silveira
1945, and De Castro 2002) with the relevant verb in boldface, the second line
provides my segmentation, the third line gives contemporary Gungbe
equivalent, and the last two lines give the gloss and interpretation.
Example (6) illustrates the verb ‘to eat.’3
3
I also refer the reader to De Castro (2002), where one finds Fongbe equivalents of the Mina-Jeje
examples. Actually, this book is a gold mine for those interested in investigating the language
spoken by certain slaves.
(6) mahidenum [Mina-Jeje]

ma-hi-de-num
Má yì ɖù nú. [Gungbe]
1SG go eat thing
‘Let me go eat (something).’
Example (7) represents the verb ‘to drink.’
(7) mahinumasin [Mina-Jeje]
ma-hi-num-asin
Má yì nù (ò)sìn. [Gungbe]
1SG go drink water
‘Let me go drink some water.’
Sentence (8) illustrates the verb ‘to sell.’
(8) mahisalefim [Mina-Jeje]
ma-hi-sa-lefim
Má yì sà lìnfín. [Gungbe]
1SG go sell flour
‘Let me go sell some flour.’
Finally, example (9) represents the verb ‘to pull/shoot/remove’ in the usage
of shooting.
(9) mahidesounim [Mina-Jeje]
ma-hi-de-sou-nim
Má yì ɖè sò énὲ. [Gungbe]
1SG go fire gun DEM
‘Let me go fire this/that gun.’
As examples (6)–(9) show clearly, it does not take high-tech reconstruction
methods or statistical analyses or other phylogenetic trees for one to figure
out that the language described by Costa Peixoto as a slave language on an
eighteenth-century colony in Brazil is indeed a Gbe language. Also remarkable
is the fact that the language spoken on this plantation is very similar to
contemporary Eastern Gbe varieties such as Gungbe and Fongbe. De Castro’s
(2002) study is very informative in this regard.
It appears clearly from this discussion that not only can we establish that at
least some enslaved Africans in the colony spoke their native language in
addition to learning Portuguese but also we can show that this language was
a Gbe language (cf. De Castro 2002). That this is possible in the case of this
community makes me think that it could have been possible in other colonial
communities as well. Indeed, the situation in Minas Gerais should not be
regarded as exceptional or unique (a too common attitude in creolistics). For
instance, the population factors in Haiti and Suriname are compatible with the
view developed here, viz., that the contact situation on some plantations of the
late seventeenth and early eighteenth centuries was much more complex than
usually assumed in creolistics.4 The historical facts reviewed in this book
strongly suggest that such contact situation involves speakers of both lexifier
and substrate vernaculars learning from one another, though for different
motives. While plantation owners’ interests in learning the African
vernaculars had to do with proper management and strict control of an ever-
growing enslaved population, the Africans were probably motivated by
integration or bonding among those originating within the same parts of
Africa. This of course implies that not all Africans were native speakers of
such African vernaculars or related languages and we must be prepared to
admit that not all the slaves engaged in learning the dominant African
languages. Instead, it is likely that several members of the African population
targeted the European standard only.
Under this understanding, it appears logical to conclude that the creole
emerged as a mediation language between the plantation communities where
daily interactions are ‘negotiated’: a contact language. As indicated by DeGraff
(2001a: 251) quoting Schuchardt, “[T]he slaves spoke the creole not only with
the Whites but also among themselves while their mother tongue was still in
existence, the latter being moreover constantly revived to some extent by the
continual immigration from Africa.” From the perspective of the colonists, the
creole therefore represents the language of efficient and orderly management of
the enslaved population that guarantees a flourishing business: sugar and
tobacco production or gold extraction. For (part of) the Africans, however, it
represents the neutral language (just like official languages in modern Africa)
that guarantees successful daily interaction with everyone on the colony. Yet,
this language was probably not used for other survival purposes such as
planning an escape or a rebellion. While this could have been the status quo
in the early years, it appears that as the creole grew in functionality and
expressive power, competition between the lexifier vernaculars, the nascent
creole, and the African vernaculars reduced the latter to symbolic functions
only: religion, magic, plant names, etc.
Once we adopt this perspective, it becomes clear that the question we should
be asking ourselves is not why the African languages eventually disappeared
from the public domain on the colonies, but why the creole was so successful
as to expand in all such domains and become the language of acculturation
(also known in creolistics as ‘seasoning’) and upbringing. My answer to this
question is that the creole won the competition because of its socio-economic
and political power. Consistent with Mufwene’s competition and selection
4
Though there are good reasons to do so, I refrain from overgeneralizing these finding to all
plantation settings because I believe (like Mufwene 2001ff.) that each creole has to be considered
in the local context in which it emerged.
model therefore, the fate of the African languages in the plantation colonies is
strikingly similar to that of certain endangered languages in current sub-
Saharan Africa (Mufwene 2008: chapter 11).
Another advantage of the model put forth here is that there is no need for a
pidgin stage in the evolution of a creole (or a contact language). In this regard, it
is important to realize that the emergence of a ‘macaronic pidgin’ prior to the
creole (as has been postulated by some authors) would simply cause the
relevant socio-economic system to collapse, probably hindering the success
of the plantation industry itself. Recall, for instance, that one motivation put
forward by De Antonio Da Costa Peixoto to promote his description Lingoa
geral de Mina was precisely that knowing this language would prevent “insults,
vandalisms, financial losses, theft, death, and finally many more atrocious
cases.” This indicates that managing the plantation required the existence
of an elaborate and efficient medium of communication that is known to at
least most participants. However, no ‘macaronic pidgin’ can enable elaborate
successful communication, where every single agent expresses her intentions
or thoughts only embryonically and in any conceivable manner (with no
guarantee that the addressee understands the message). Absence of
communication, however, means inefficient organization of labor and low
production, which itself would translate into loss of investment and bankruptcy.
Assuming that the plantations were comparable to military regiments, a
good example that the functioning of such regiments requires efficient
communication is the so-called ‘petit-nègre’ or ‘français tirailleurs’ which,
even though it is inaccurately associated with African soldiers in the French
army, was actually a colonial instrument developed and enhanced by the
French colonial administration hoping to accomplish easy communication
with African soldiers recruited from their various colonies. Any serious
scientific work on this variety can easily show that during the First and
Second World Wars the French military officials did not give the enrolled
Africans the chance to learn the language before they were sent to the front.
Instead, what they thought was needed was an efficient medium of
communication that the African recruits could easily parse. The ‘petit-
nègre’ was designed to this purpose. (See van den Avenne 2005 for an
elaborate discussion.) In the context of colonial settlements in the New
World, I surmise that creole fulfilled a similar function and was used to that
end. The following quote from Pelleprat, cited in Aboh and DeGraff
(forthcoming) illustrates this:
We wait until they learn French before we start evangelizing them. It is French that they
try to learn as soon as they can, in order to communicate with their masters, on whom
they depend for all their needs. We adapt ourselves to their mode of speaking. They
generally use the infinitive form of the verb [instead of the inflected forms] . . . adding a
word to indicate the future or the past. . . . With this way of speaking, we make them
understand all that we teach them. This is the method we use at the beginning of
our teaching . . . Death won’t care to wait until they learn French. (Pelleprat 1655
[1965: 30–31], my translation)
Even though the first two sentences of this quote give the impression that
the instructors expect the Africans to ultimately learn French, this stands in
immediate contradiction with their evangelization mission. Indeed, immediate
and efficient conversion into Christianity forces missionaries to use the creole
as means of instruction/indoctrination. This linguistic choice is not only
assumed to be the best possible option that proves to be very efficient in this
context, but it is further motivated by the then common racist belief that it was
hopeless to try to teach proper French to people they considered to be primitive.
Because the missionaries considered Africans to be inferior humans, they also
assumed that the slaves had not evolved the necessary cognitive capacity to
master an allegedly refined and civilized language such as French. As the
author concludes, “Death won’t care to wait until they learn French.” What is
important for our discussion here, however, is that the creole being the
language of acculturation/indoctrination, must have been used fluently, if not
natively, by the White colonist community as well (more on this below).5
Putting all this together, I argue, based on the model in (4) and on the
historical facts revealed in this book, that creole languages were plantation
(or colony) creations, that is, vernaculars used by both colonists and the
enslaved Africans alike because they were efficient means of communication
in the relevant contact settings. The latter guaranteed economic growth and
political stability for the colonial power while providing the enslaved
population with some form of cohesion. This reason, I submit, explains why
the creole was retained as a vernacular even in Maroon societies, as in the
Saramaccan communities of Suriname where it appears that Gbe and Kikongo
could have vied for this ethnographic function, as they had critical masses of
speakers.
For this ecological view of the development and expansion of the creole due to
its function on the plantation to be complete, I now need to show that colonists
were indeed (native) speakers of the emerging creole. The point just made may
horrify some students of creole languages who adhere to exceptionalist
developmental scenarios that assume a pidgin-like stage in the development of
creoles. But once again, historical facts persistently point in the direction that I’m
proposing in this book. The above quote from Pelleprat already showed that
in Haiti, missionaries used the creole as the acculturation/indoctrination
5
Fattier (1996) discusses La passion de notre seigneur selon St Jean, a document written for
missionary purposes and which she shows to involve ‘hybrid’ lexical properties that are attested
in various French-based creoles in the Caribbean in the seventeenth and eighteenth centuries.
language.6 Another significant example illustrating this point comes from the
following official declarations in eighteenth-century Haiti in creole signed by
French officials Léger-Félicité Sonthonax, Charles Victoire Emmanuel Leclerc,
and Napoléon Bonaparte and reported in Denis (1935). The texts are quoted in
their original form, followed by my translation.7
Au Nom du Gouvernement Français: Liberté, Fraternité.
Proclamation.
Les commissaire v’lé et ordonné que tout ça yo sorti di icy soit affiché tout par tout
dans les communes, dans les habitations, dans les camps de la Colonie. Commissaire
recommendé à tous les officiers, que yo faire exécuté ça yo sorti recommende dans
proclamation ci la la.
Le Priésident de la Commission: Sonthonax
Proclamation là, li tiré dans registre consuls la République
Paris, 17 Brimer, an 10 Répibilique francé, yon et indivisible
Qui ca vous tout yé, qui couleur vous yé, qui côté papa zote vini, nous pas gardé ça:
nous savé tan seleman que zote tout libre, que zote tout égal, douvant bon Dieu et dans
zyé la Répiblique. Dans tan révolution, la France voir tout plein misère, dans la même
que tout monde te fere la guerre contre Français. Français levé les ens contre les otes.
Mes jordi là tout fini, tout fere paix, tout embrassé Français; tout Français zami; tout
hémé gouverneman, tout obéi li . . .
Signé: Bonaparte
Voyez pour zote. Zote à voir que li vélé resté libre. Li pas vélé ôté liberté à yo que yo
gagné en combattant, et que li va mainteni li de tout pouvoir à li. Li va mainteni
commerce et culture, parceque zote doit conné que sans ça, colonie-ci pas cable
prospéré . . .
Général en chef qui vini pour gouverné tout la Colonie: Leclerc
In the name of the French Government: Freedom and fraternity.
Declaration
The commissaries want and order that all that comes out of this commission be
announced and posted everywhere in the communes, plantations and camps of the
Colony. The commissary recommends that all government officials see to it that all
decisions made in this declaration be executed.
The President of the Committee: Sonthonax
The declaration is taken from the governing rules of the Republic.
Paris, 17 Brumaire, year 10 of the French Republic, one and indivisible.
Whoever you are, whatever your skin color, wherever your ancestors are from, that
does not matter to us. We only know that you are all free and all equal before God and
before the Republic. During the Revolution, France experienced a lot of suffering
because every other country fought against the French. The French were fighting each
other. But today, all of that is over. All people have made peace. All people have
6
Various manuals including a list of lexical items and sentence translations (e.g., Ducœurjoly
1803) were also published to inform potential migrants. The existence of such manuals clearly
indicates that Europeans were encouraged to learn the Creole.
7
I’m immensely grateful to Michel DeGraff for sharing all these precious documents with me and
for helping me with the translations.
embraced France. All the French people are friends. They all love and obey the French
government.
Signed: Bonaparte
See for yourself. He wants the Blacks to be free. He does not want to take away the
freedom you gained through battle, a freedom that he wants to protect with all his power.
He will develop commerce and culture because you must know that without these, this
colony cannot become prosperous.
The General who has come to govern the colony: Leclerc
As is clear from the translations, these texts were signed by French officials
of the highest rank (including Napoléon Bonaparte) for all inhabitants of the
colony and not just for the newly freed Africans and their descendants. But
for this to be the case, and for Napoléon Bonaparte to have accepted to have
his declaration translated in the creole, it must have been the case that (i)
government officials were convinced that the creole was far more practical
than standard French, and (ii) that a large proportion of the colony was fluent in
the creole. In such a context, it is not unreasonable to assume that a French
version of the same declaration was also distributed on the plantations. Thus in
addition to showing that the Haitian colony was multilingual, these documents
represent very strong evidence that the emerging creole was used by most
agents in the colony and that it could be used in official contexts involving
public declarations, political statements, and religious instruction or
indoctrination. As such the creole represents a legitimate target for children
and bozal slaves, as well as European newcomers, and an alternative to French
for the inhabitants of the colony.
As the following excerpt from Long Edward (1774: 426) shows, one finds a
similar situation in Jamaica where both White Creoles and Africans spoke the
creole natively or as L2:8
The Africans speak their respective dialects, with some mixture of broken English. The
language of the Creoles [i.e., the local Whites] is bad English, larded with the Guiney
dialect, owing to their adopting the African words, in order to make themselves under-
stood by the imported slaves; which they find much easier than teaching these strangers
to learn English. The better sort are very fond of improving their language, by catching
at any hard word that the Whites happen to let fall in their hearing; and they alter and
misapply it in a strange manner . . . The Negroes seem very fond of reduplications, to
express a greater or less quantity of anything; as walky-walky, talky-talky, washy-
washy . . . and so forth. In their conversations, they confound all the moods, tenses,
cases, and conjugations, without mercy; for example; I surprise (for I am surprised); me
glad for see you (pro, I am glad to see you) . . . This sort of gibberish likewise infects
many of the white Creoles, who learn it from their nurses in infancy, and meet with much
difficulty, as they advance in years, to shake it entirely off, and express themselves with
correctness.
8
I thank Joseph Farquharson for bringing this reference to my attention.
This text strongly supports the scenario advocated in (4). Here we learn that
the Africans could speak their respective languages but also that the Creole
colonists tried to acquire such languages because it was easier for them to use
fragments of these languages to guarantee orderly and efficient management
rather than expecting the slaves to acquire English.9 Yet, the text also shows
that the Africans, though they could speak their own languages, were eager to
learn the target language adequately. This, I think, is a devastating blow to all
exceptionalist theories such as the pidgin-to-creole cycle, the bioprogram and
especially, the interlanguage hypothesis, which we now see can simply not hold
on historical grounds. Finally, the text confirms the conclusion already reached
in previous sections that, notwithstanding segregating rules, some colonists not
only used the emerging creole but also spoke it natively due to their intimate
contacts with the slaves who nursed them as children.
The following text from the unknown author of A short journey in the West
Indies (cf. Robert Charles Dallas 1790) is rather clear:
There was a little urchin, about seven years old [my emphasis], who constantly stunne
my ears with – “me wantee crab, me wantee crab” – to stop his noise, I took up the first
that came to my hand and gave it to him; but he immediately examined it, and finding it
had no eggs, he roared out, “Him no hab egg, him bleu maugre to hell, me no wantee
man crab, me wantee woman crab” – so he had heard his mother and the negroes
distinguish the sexes of crabs. I would have taught the little scoundrel better manners,
but his mother called him to her, kissed him a dozen times, and picked him out the best
woman crab.
Here again is a passage where we see that children pick up the emerging
creole and the plantation vernacular English from birth, a case of bilingual first
language acquisition. Additional evidence for this comes from breast feeding, a
practice that Long (1774: 276) condemns firmly when he writes
While I render all due praise to the Creole ladies for their many amiable qualities,
impartiality forbids me to suppress what is highly to their discredit; I mean, their
disdaining to suckle their own helpless offspring! They give them up to a Negroe or
Mulatto wet nurse.
Clearly there were intimate relations between the White children and the
Africans which necessarily involved linguistic exchange and language
acquisition. This is further supported by the fact that, instead of “teaching the
child better manners” both in language and behavior, as the visitor was hoping
9
A similar observation is made by Christian Oldendorp, cited in Holm (1988: 18): “In the West
Indies, the European languages tend to deviate to an extreme extent. For the most part, only those
people who learned to speak them in Europe can talk the pure European form of the language. On
the other hand, the people who were born here – the Crioles – do not speak the same kind of
language. They change it more or less; they employ words taken from elsewhere, arising from the
collision of the people of many nations.”
for, the mother showed him love. Such an attitude clearly indicates that there is
a priori no repulsive attitude to the creole, just as there does not seem to be any
corrective attitude toward its speakers. This attitude toward the emerging
language clearly reflects the fact that it is the language widely spoken and
accepted on the colony.10
Yet, the colonists were very much aware of the fact that in order to maintain
the segregating rules in the colony, their children had to acquire the standard
European variety in addition to the local vernaculars (including the creole). A
measure often taken to address this problem is the creation of schools for White
children only. In this regard Long (1774: 250, my emphases), arguing for
boarding schools for the Jamaican plantation children, writes:
The utility of a boarding-school for these girls, where their number might admit of
employing the ablest teachers, where they might be weaned from the Negroe dialect,
improved by emulation, and gradually habituated to a modest and polite behaviour,
needs not, I think, any argument to prove it. Young ladies, so far accomplished as, I
think, they might be on a well-conducted plan, would insensibly acquire, on their
emerging into public life, the remaining graces and polish which are to be attained in
gentle company and conversation. They would by these means, become objects of love to
the deserving youths, whether natives or Europeans, and by the force of their pleasing
attractions soon draw them, from a loose attachment to Blacks and Mulattoes, into the
more rational and happy commerce of nuptial union.
This last excerpt further indicates that not only did White Creoles speak the
creole from birth, but it appears to have been their daily language until maturity.
That the creole was the lingua franca in the early years of plantation colonies is
therefore compatible with Mufwene’s (2001) view that these varieties became
disenfranchised later when the segregating rules of plantation societies became
the cornerstone of Creole societies based on slave economy. In the context of
this book, the following points indicate that the historical facts reviewed in
Chapter 2 and in this section have important repercussions on how to consider
the formation of the creole language.
– It is clear from this discussion that commonly used notions such as target
language need to be clarified. The facts presented here suggest that speakers
were exposed to a continuum of varieties of the substrate and lexifier
vernaculars, in addition to the emerging contact language. We therefore
reach a situation where there were potentially several competing targets
with the creole functioning simultaneously as target and playing a useful
function in the chain of information transmission (from the lexifier and
substrate languages).
10
See also Lalla and D’Costa (1991) for discussion.
– Because all available targets contribute to the input from which the
speakers select features into the emerging creole; and because these
speakers varied from each other, the emerging creole ‘inherited’ fea-
tures produced by both (bilingual) L1 and L2 learners of the colony
(i.e., Whites, Creoles, and Blacks alike). Contrary to what is often
assumed in the literature, this point makes it even harder to assign the
creation of the creole exclusively to a specific group of learners (e.g.,
enslaved African L2 learners or children), consistent with DeGraff
(1999: chapter 1, 2002).
– The commonly assumed restricted access to the target also becomes unten-
able. First, there were several targets (including the emerging creole).
Second, even though not every single African was exposed to what Long
(1774) referred to as ‘bad English,’ which we could interpret as ‘non-
standard colonial vernacular’ on a daily basis, there were various channels
through which most inhabitants of the colony could approximate the colonial
vernacular.
– The findings presented in this section indicate that the role of locally born
bilinguals in the development of the creole has been largely underestimated
in previous accounts. As I show in subsequent chapters, Haitian Creole
and the Suriname creoles display very many subtle structural properties of
their lexifiers that cannot be attributed to L2 learners in the ways often
assumed in creolistics (viz., fossilized L2 speech, or L2 learners with poor
access to the target). Many of such features seem to have been injected in
the language by (near) native speakers many of whom were probably child
bilingual speakers (i.e., either bilinguals in the emergent creole and the
local European koiné or in the emergent creole and in some African
language).
– Finally, the notion of substrate transfer (as understood in creolistics) appears
obsolete. As I show in Chapter 5, what one finds instead is a process of
transfer whose direction seems determined by linguistic factors such as the
properties of interfaces rather than the origin of the speaker (e.g., African
ethnicity).
I conclude from this discussion that creoles are only new with regard to
history but the bulk of their morphosyntactic features (i.e., those features which
are selected from the inputs) are to be found in the languages that contributed to
the feature pool. In the discussion that follows, I argue that the model of
competition and selection as described earlier (and in detail in Mufwene
2001, 2005a, 2008) offers an adequate conceptual framework within which
we can approach the evolution of the contact of input languages into a new
language. This discussion bears on issues of language creation and language
change on the level of the speaker.
4.2 Languages as biological species 133
4.2 Languages as biological species

A central theme of Mufwene’s work on the ecology of language evolution is
that “a biological approach to evolution is applicable to languages,” which can
help in understanding the hypothetical life cycle of a language as described in
(10) (Mufwene 2005b: 1).
(10) Phase 1 Phase 2 Phase 3
Language creation/birth…. Language change…. Language death
Language evolution
Given the notion of competition and selection as used in biology, these

different phases are seen as the results of different ecologies. By analogizing
languages to species rather than organisms, Mufwene (2001, 2005a, 2005b)
clearly indicates that the question of why certain linguistic ‘species’
disappear to the benefit of others resides in the linguistic ecologies (e.g., the
socio-historical settings of the colony). This view also implies that not every
single newly born language will go through the three phases described in (10).
It is conceivable that a new language can fade out in a very short period of
time because it cannot survive in a particular linguistic ecology. Work place
pidgins, for instance, are often cited as such examples because they tend to
disappear along with the function they serve. We can infer from this that the
act of language creation itself is contingent on the linguistic ecologies of its
creators.
Setting aside political factors, we can informally say that such an ecology
primarily consists of the interactions of a speaker of a language (Lx) with
speakers of other varieties of the same language (i.e., Lx1. . . . .Lxn+1) or
speakers of other languages (e.g., Ly, Lz). These interactions feed a process
of competition and selection that opposes the grammars in contact and
eventually leads to phases 1, 2, or 3 as described above. How the competition
and selection operate specifically on language structural properties (thus, in the
mind of the speaker) is not clearly established in Mufwene (2001, 2005a,
2005b), who mainly focuses on the population level. Yet, the following
quotes under (12) and (13) give some indications of his thoughts and help me
fine-tune my own suggestions as to how competition and selection may operate
as far as morphosyntax is concerned. According to Mufwene,
(12) A creole is a restructured variety of its lexifier. The latter was primarily a
colonial variety which was spoken by the European colonist and was itself
developing from the contact of diverse metropolitan dialects. It has often been
identified as a koiné (Mufwene 2001: 28).
However, given that creoles do show uncontroversial influence from the
relevant African languages (e.g., regarding predicate fronting with doubling,
topic and focus marking, Aboh 2006a, 2006b, 2007a, Chapter 3 of this book),
Mufwene is careful enough to complement this quote with the following:
(13) It is as necessary to invoke substrate influence from the Celtic languages to
account for the speciation of Latin into the Romance languages as it is to
invoke African substrate influence to account for the evolution of French and
Portuguese into various creoles . . . Both cases are clear instantiations of
Pyrrhic victory – where the prevailing language is so clearly affected by the
displaced ones (Mufwene 2005b: 12).
According to Mufwene’s work therefore, substrate or cross-linguistic
influence, which we should understand as structural interaction between two
L1’s during bilingual acquisition or between L1 and L2 during L2 acquisition is
not only undeniable, but it represents crucial ingredients in change.
4.2.1 Unchaining the competitors

The quotes in (12) and (13) meet two goals in Mufwene’s framework. First,
creole genesis is regarded as just a ‘contrastive’ case of the normal competition
and selection process that is typical of language contact situations. Under this
view, creoles are only special due to their special ecologies, which in turn
derived from the unique conditions under which the creole communities were
formed (viz., slavery, plantation society, segregation). Second, that creoles
share significant similarities with both the colonial lexifiers and the relevant
languages of the enslaved Africans is accounted for. In a sense, the particular
linguistic ecology that led to the development of creoles also provides us with
the raison d’être of their particular phenotype (i.e., the way they look or sound
compared to their source languages).
While I tend to generally agree with this research framework, close scrutiny
reveals a tension between (12) and (13) that is worth discussing. We may infer
from (12) that the Africans switched to their owner’s language and engaged
in acquiring the lexifier koiné to some extent (Chaudenson 2001, 2003;
Mufwene 2001). This learning process eventually led to the death of the
African languages in the colonies, but the battle (or competition) between the
displaced African languages and the lexifier koiné was fierce. Accordingly,
the prevailing language, that is, the approximation of the lexifier koiné which
we now refer to as creole, still shows traces from this formation period. Such
signs are analyzed as consequences of imperfect second language acquisition
sometimes allowing L1 or substrate transfer. Given this description, one has the
feeling that right from the beginning the (changing) ecology favored the
lexifier. Chaudenson’s (2003: 199) claims that “les substrats ne peuvent pas
passer en force [substrate languages cannot forcefully infiltrate the lexifier
language]” (cf. Chapter 3), perfectly illustrates this view and suggests that
the substrate languages can only make their way in the grammar of the lexifier
koiné if the former and the latter show (partly) congruent structures. This, in
turn, means that the linguistic ecology in which the creole developed favored
the lexifier language but constrained the substrate languages. Accordingly,
substrate influence appears limited to areas of congruence.11
This view may look reasonable if one considers political and sociolinguistic
factors (e.g., the oppressing power forcing a linguistic policy on the enslaved
Africans and prohibiting African languages). However, as shown in
Section 4.1, things might not be so clear-cut. The discussion in previous
sections shows that there is ample evidence that some African languages
were spoken for some period in the colonial Caribbean and in colonial Brazil
(e.g., Long 1774; De Castro 2002; DeGraff 2002: footnote 45). This would
mean that linguistic features from the African languages could be selected into
the creole even though they might not converge with the European target
language. Also recall from that discussion that the process of competition
and selection proposed by Mufwene (2001ff.) assumes that linguistic features
of the competing languages form a feature pool on which selection operates.
Under this view, it is not clear to me how the ecology of such a language contact
situation, in which the lexifier represents the target language, constrains
selection inside the feature pool such that non-congruent substrate linguistic
features are strongly disfavored (if not excluded).
If we adopt a minimalist approach to the study of human knowledge of
language, as I do here, it appears that factors that concern speakers’ linguistic
ecology or practice are external to the computational system underlying human
language capacity. The latter, however, is subject to the competition and
selection process, which affects various linguistic modules during the
development of the idiolect in the mind of the speaker.
In order to keep a formal distinction between change at the individual level
and the population level, I adopt DeGraff’s (1999: 9) distinction between
E-creole and I-creole, as already stated in Chapter 1. An E-creole is an
abstraction of the linguistic codes of a creole community, whereas an I-creole
denotes “the development in individual speakers’ minds/brains of a grammar
that shows a certain typological distance from the grammars of the languages in
contact.”
Adopting this distinction, I submit that ecological factors affect I-creole
indirectly. Indeed, ecological factors may affect the learning setting in
various ways thus constraining the feature pool to which the learner is
exposed and which will feed into the development of her I-creole. For
11
As suggested to me by M. DeGraff, Chaudenson’s position can hold only if we have a clear
definition of congruence such that we are able to reconstruct the distinctive features that may
converge in the competing languages. See Aboh (2009b) for some discussion.
example, ecological factors impact learning depending on whether language

acquisition takes place in a multilingual or in a monolingual environment,
whether learning is tutored or untutored or whether learning occurs very
early or late (i.e., L1 vs. L2 acquisition). Similarly, ecological factors may
influence the frequency of certain variants in the input thus favoring their
selection over other alternatives. While all these factors indirectly constrain
the development of an I-language or I-creole, I surmise that they do not
determine how variants selected in the input are combined together into a
coherent grammatical system. More precisely, while ecological factors are
necessary to understand bilingual acquisition, for instance, they cannot
explain the morphosyntactic rules underlying structure building. As I show in
Section 4.3.2 below, early Sranan includes the item aseh-man (witchcraft-
person, i.e., a witch) derived from Gungbe/Fongbe àzé-tɔ́ (witchcraft-person,
i.e., a witch). That the English morpheme -man has been substituted to the Gbe
morpheme -tɔ́ indicates that the creators of the word aseh-man equated these
competing forms at some level of abstraction. Since these forms are associated
with specific grammatical properties, I’m led to conclude that these abstract
grammatical properties also compete in the mind of the speaker. I further argue
that while external ecological factors allow the competition between these two
alternatives in the input, the rules governing the morphosyntactic licensing
properties of these forms are internal to the syntactic component. In other
words, while ecological factors constrain the competition and selection of
variants in the input, they do not directly determine how these variants are
combined together in the syntactic component. Ecological factors therefore do
not define how varying grammatical properties can be recombined into a new
lexical item. I therefore conclude from this that:
(14) The recombination of (morpho)-syntactic features is free (i.e. not subject to
external ecological factors).
Given this view, competition and selection in the syntactic component (i.e.,
the selection of specific syntactic licensing mechanisms) happens freely
between features and patterns of the language in contact. This amounts to
saying that even though the speaker’s linguistic activities may be (dis)
favorable to the competing languages (e.g., workplace language, household
language, ritual language vs. poetic/stylistic language) and therefore to certain
patterns, the competition and selection that applies to alternative linguistic
features and patterns in the mind of the speaker, that is parameter-setting in
the syntactic sense, escapes external ecological factors.12 This amounts to
12
This need not mean that certain external factors (e.g., communication settings) may not favor
certain linguistic patterns, which because of their frequency or discourse prominence may in
turn favor some linguistic features and their associated parameters. To put it more concretely,
external factors such as frequency, or markedness (however defined) may favor certain syntactic
saying that competition and selection happen at two levels: the structuring of
the input, and the combination of variants selected from the input into a
coherent syntactic system. Distinguishing between these two levels of
analysis allows me to focus on syntactic recombination only in this book.
In addition to being perfectly compatible with Mufwene’s general
framework, the proposed distinction enables us to reformulate the question of
some similarities between creoles and their source languages (viz., the lexifier
languages and/or the native languages of the Africans) as well as across creoles
in terms of a more general question related to the emergence of new linguistic
variants: Why are certain linguistic features typically selected over others in
contact situations, and why are these features typically combined in some
particular ways?
The latter question obviously relates to even more difficult questions such as
why the competition and selection process does not affect all components of a
module (e.g., VP, IP, CP, DP in syntax) or all modules of the grammar (e.g.,
syntax, semantics, phonology, morphology, pragmatics) the same way. A
fascinating example discussed in DeGraff (2002, 2005) concerns word order.
Most Kwa languages, the major substrate languages for the Suriname creoles
and Haitian (Arends 1989; Smith 1987; Lefebvre 1998) display VO versus OV
alternation in specific aspectual contexts (e.g., progressive, Aboh 2004a,
2005c). No such aspect-dependent word order variation is found in Haitian,
Saramaccan, or Sranan. This is surprising since the Kwa speakers, being
numerically dominant in this context and having this feature in their
E-language, must have contributed this aspect-dependent word order
variation to the feature pool. Nevertheless, this putatively strong element of
the feature pool was not selected into the emerging creoles.
One could think that the OV order was simply selected against, because the
alternative VO order was congruent with that of the lexifier. But it is worth
noting that in the case of the Suriname creoles, OV was selected against even
in contexts where both the superstrate and the substrate converge (e.g., in
nominalizations: street sweeper, àlìò zà-tɔ́ street/sweep/person vs. figi-strati-
man sweep/street/person ‘street sweeper; van den Berg 2007: 175), see
Section 4.3.2. Similarly, DeGraff (2005) indicates that French has alternative
OVorder with clitics and this was also selected against in Haitian Creole. In this
language, the object must follow the verb.13 These facts underscore the point
made here that the dynamics of competition and selection in the syntactic
module must be more complex and to some extent blind to certain factors
related to the input that may have to do with say, number of speakers, prestige,
properties over others (e.g., periphrastic aspectual expressions with non-finite verbs versus
finite verbs), but these external factors do not determine how syntactic recombination operates.
13
I thank M. DeGraff and S. Mufwene for their suggestions on this issue.
etc. Clearly, there must be some independent reason internal to the syntactic
component why the emergent system selected against OV even when all
external factors would seem to favor it.14
Similarly Aboh (2006a, 2006b, 2007a) showed that while the Gbe
languages have influenced the expression of the complementizer and
determiner systems in the Suriname creoles to some extent, the same does
not hold for other syntactic areas such as the clausal inflectional domain (IP)
or even when it comes to modification inside the noun phrase. More study is
needed before we understand how linguistic features of competing languages
are recombined in the emerging language. This book aims to contribute to a
better understanding of this process and how it applies to language acquisition
in general.
In the context of the present discussion, however, the recombination of
linguistic features in an I-creole, the replication of such I-creoles, and the
emergence of a community of speakers of similar I-creoles leading to the
birth of the E-creole is comparable to the evolution of a population in
biology. The latter may consist of the mutation of a gene and the selection/
reproduction of individuals carrying such gene which eventually leads to a
population change. Other striking parallels between biological species and
languages (e.g., the analogy between successful communication and
biological interbreeding in Mufwene 2005b: 15) lead me to propose that
(15) Languages are species whose phenotypes – the linguistic features on which
typological classifications are based – correspond to genotypes, in this case,
syntactic structures.15
According to (15), syntactic features are expressions of structures (i.e.,
functional projections as defined in generative syntax) that are somehow
comparable to genotypes. The next section discusses this issue.
14
DeGraff (2005: 353, n.32) speculates about certain external factors that may have played a role
in favoring the rise of VO in Haitian Creole in contexts where certain French varieties use VO.
As is usually the case, the problem here too is to identify the right varieties and find out whether
they are relevant for the genesis of Haitian creole. I leave these issues for future research.
15
This claim appears in contradiction to Mufwene (2005b: 15, and footnote 13) who argues that
“Languages are species whose phenotypes – the linguistic features on which typological
classifications are based – correspond to no genotypes . . . Clearly languages and idiolects
cannot have genotypes, because they are not biological systems.” Mufwene’s theory builds on
the fundamental assumption that languages are more comparable to viral species than to either
organisms or to animal species.
I do not discuss these issues in this book, as they do not directly bear on the argumentation. It
is, however, important that the reader keeps in mind that it cannot be assumed that all phenotypic
properties of languages are rooted in the distinctive linguistic features as defined here. What this
book tries to do, instead, is to identify which linguistic features have typical realizations (i.e.,
phenotypic effects) and how such features recombine with other linguistic features in a situation
of language contact.
4.2.2 Syntactic projections are genotypes

Mufwene’s observation that linguistic features correspond to no genotypes is
compatible with his other proposal that languages should be compared to
viral species not to an organism. However, if we take the biological approach
to evolution, we have to admit that language evolution results from the
recombination or mutation of linguistic features that are part of an idiolect
(in this case the I-creole) and the selection of such idiolects. Extended to the
domain of syntax this would mean that the recombination or mutation of
syntactic features observed in the phenotype is a consequence of the
recombination or mutation of syntactic structures that is comparable to the
recombination/mutation of genes.
4.2.2.1 The DNA approach to clause structure and syntactic projections A

question that immediately arises from this discussion is how to define linguistic
features. Indeed, the analogy between genes and linguistic features is not new
(see, for instance, Croft 2000 for a formulation of linguistic features in terms of
lingueme). In addressing this view, I adopt the fairly standard view that lexical
and functional items are expressions of bundle of features that minimally form
a triplet: (i) phonological features, (ii) morphosyntactic features, and (iii)
semantic features. Phonological features represent those that can be assessed
by the phonological component at the interface. Morphosyntactic features
represent the set of principles (properties) that are required for syntactic
licensing in the syntactic component, while semantic features are those that
are necessary for interpretation. Applied to grammar, this definition indicates
that a functional item can be informally represented as in (16):
(16) Functional item
Phonology Morphosyntax Semantics
It is perfectly conceivable (and actually accepted) that the boxes

“Phonology,” “Morphosyntax,” and “Semantics” involve clusters of
properties that interact at different stages of the derivation. Let us consider
the habitual marker -na in the Gengbe example in (17a). As is clear from this
example, this marker is an affix in Gengbe and therefore attaches to the verb: an
instance of verb movement according to Aboh (2003a, 2004a, 2009b), Aboh
and Dyakonova (2009), and subsequent.
(17) Kwésí ɖù-na àkɔ̀ ɖú.
Kwesi eat-HAB banana
‘Kwesi usually eats banana.’
‘Kwesi used to eat banana.’
Taking into account the distribution and semantics of this morpheme, we

reach the description in (18) for the habitual marker -na:
(18) Functional item

/na/ – selects VP comp repetitive occurrence of event E
– triggers V-to-ASP at the time of Reference habitual
MOVEMENT
As we can see from this sketch, the morpheme spelled out as -na displays a
very specific semantics (i.e., habitual) in addition to a cluster of syntactic
properties: (i) it selects a VP as complement and it requires verb-movement.
The latter can be further formalized by suggesting for instance, that -na has a
formal v-feature that needs to be valued (or checked) by V (the verb). Setting
these technical details aside, what matters for the present discussion is that
when it comes to morphosyntax, this morpheme has two properties: a
selectional property that regulates the type of element or category it can
combine with, and a formal feature that governs its licensing and satisfies its
morphological requirements as an affix.
Because the boxes labeled “Phonology,” “Morphosyntax,” and “Semantics”
operate independently, though they interact during the derivation, we can
conclude that functional items are the loci of linguistic features which drive
grammar. This in turn leads me to the conclusion that functional items are the
target and carrier of change. The latter point is compatible with Borer’s (1984)
view that cross-linguistic variation is attributable to differences in abstract
features of lexical/functional items (see also Chomsky 1995, 2001; Baker
2008; and Boeckx 2010 for more elaborate discussion.) Under the traditional
view that functional items head functional projections which represent the
backbone of the clause, the representation in (18) suggests that functional
projections include licensing specifications that govern clause structure
building and determine cross-linguistic variation.
From the point of view of acquisition and change, this description further
indicates that the acquisition of grammar consists of the development of an
algorithm that recombines formal properties of diverging lexical/functional items
into a coherent system. Because acquisition and change go hand in hand, we
further reach the conclusion that in a situation of language contact, any component
(or combination) of the triplets in (16) and (18) (viz., phonology, morphosyntax,
semantics) can be affected. Consequently, we need to understand how lexical/
functional items emerge, in order to understand the emergence of new grammars.
With this discussion in mind, I assume (following Rizzi 1997; Cinque 1999;
and much related work in the cartographic approach) a description of clause
structure as informally illustrated in (19). The distinct articulations in (19)

represent functional projections that have rudimentary semantic content (e.g.,
focus, interrogative, negation, tense, mood, and aspect) and embed
specific parameters (some specifications of which are given in parentheses
together with relevant references).
(19) Left periphery (Clause typing particle, filled vs. unfilled specifier, Aboh and Pfau 2011)
Negation (head or specifier: X° vs. XP, Haegeman 1995, Aboh 2010c)
Tense (free morpheme vs. affix + V-movement, Pollock 1989)
AspP (free morpheme vs. affix + V-movement, Aboh 2004a)
VP (VO vs. OV: aspect dependent word order Aboh 2004a)
Under this view, while the clause structure is similar to the DNA in containing
all the syntactic features as well as their formal licensing properties (i.e., the
parameters), the projections are comparable to genes in that they encode specific
information about specific syntactic features and parameters only. There is,
however, an important evolutionary/developmental difference between the
two: DNA precedes the development of an organism, whereas the clause
structure is a consequence of learners’ experience during acquisition. This is an
important difference between acquisition of a biological trait by genetic
inheritance and the acquisition of a feature by learning. The bearer of a DNA
cannot modify it, while the carrier of a linguistic feature may modify it. These
considerations obviously break down the genotype/phenotype parallelism as
well as that between biological and cultural evolution. It is therefore important
to emphasize that the theory developed in this book builds on an analogy with
biological species and not on isomorphism.
If we grant the description in (19), however, we can further say that
manifestations of linguistic features, which Mufwene correctly treats as
phenotypes, are expressions of a combination of syntactic nodes according
to how the associated parameters have been set in the development of the
I-language, in this case the I-creole.16 In other words, the overt manifestation of
a learner’s mental grammar is nothing but the spell-out of specific
recombination of linguistic genotypes, as illustrated in (20).
16
As Michel DeGraff remarks (p.c. June 21, 2009), it is clear from this description that a certain
amount of “reverse engineering” qua linguistic analysis is necessary to uncover the genotypes that
are reflected in the phenotype (i.e., the overt manifestation of linguistic system). This exercise is
obviously not a simple one given that not all the properties of the phenotype relate to distinctive
linguistic features that are subject to competition and selection. Though further research is needed,
questions of this sort show the complexity of the issue at stake, and cast serious doubts on theories
of creoles that assume a trivial complex versus simple opposition where the presence versus
absence of morpheme types plays an important role (e.g., McWhorter 2001).
(20) E-language
I-Language
Linguistic experience
(e.g., LPD, language contact)
Universal Grammar Genotype vs. Phenotype

Genotypes = syntactic phenotype mapping = spell out of I-creole
features + parameters parameter setting =
development of I-creole
The box with broken lines in this representation is meant to show that what
is referred to as E-language (or E-creole) by DeGraff (1999) is an abstraction
over collective spell-outs of individual I-languages/-creoles. Furthermore,
representation (20) indicates that competition and selection as well as the
recombination of syntactic features happen in the middle box during the
mapping of genotype and phenotype. This actually corresponds to parameter-
setting in the generative framework (see Lardiere 2000, 2005 for some
discussion). Under this view, the languages in contact provide the speaker
with the right triggers (or context) for parameter setting. Taking this line of
thought seriously, the only difference between contact languages or creoles, on
the one hand, and other languages, on the other, is that the input for parameter-
setting is too diverse in the case of creoles. As a result, several possible options
compete for the fixation of one and the same parameter (e.g., VO vs. OV,
affixal TMA morphology vs. free morphemes, specificity vs. definiteness).
Furthermore, the recombined syntactic features that this competition gives
rise to in the I-language can also be very divergent, when compared to the
source languages. This is so because the range of variation set by the
typologically diverse competing languages in a plantation colony confronts
the speakers with a wider range of learning hypotheses than in monolingual
settings. This conclusion is compatible with the position assumed in this book,
viz., that creole languages only differ from older languages with regard to the
socio-historical contexts in which they came into being. This social context
brought together – at a certain point in time and in an isolated geographical
location – typologically different languages which otherwise would not have
met (see also Mufwene 2002, 2003 for some discussion).17
17
On a more general note, the description in (16), (18), and (19) assumes that syntactic features
have a formal component (i.e., licensing properties) and a semantic component (e.g., discourse
function, meaning). Each component can be copied separately or independently during the
recombination process that leads to the new language. According to this hypothesis, language
change boils down to the study of recombination of linguistic features. In addition, if, adopting
this view, we are able to isolate syntactic features to the extent that we can distinguish between
dominant (i.e., semantically prominent) and weak (i.e., semantically empty) syntactic features,
we may have an answer to why certain features are typically found in contact languages as well
as in bilingual acquisition. This predicts that the weak features (under certain ecologies) may
4.2.2.2 Syntactic recombination The idea that syntactic features are

properties of functional projections that are sensitive to the competition and
selection process characterizing the development of the I-language further
suggests that, in a situation of language contact, such functional projections
can be combined in various ways across various modules.18 Anticipating the
discussion in Chapters 5 and 6, the following two tables illustrate a
recombination of syntactic features that led to the emergence of Saramaccan
(see Aboh 2006a, 2006b, 2007a; Aboh and Ansaldo 2007; and references cited
therein). In these tables the symbol “✓” stands for the presence of the relevant
property while “*” indicates its absence.
As discussed in previous work, Table 4.1 indicates that Saramaccan and
Gungbe match in all respects with regard to the morphosyntax of their left
periphery but not with English. I take this to show that the Gbe features of the
complementizer layer have been selected. The overt manifestations of these
features make the Gungbe and Saramaccan clausal left peripheries look similar
(even though not isomorphic). For instance, the two languages have markers
for question, focus, topic, and mood that realize distinct functional heads within
Table 4.1 A comparison of the left periphery of Saramaccan, English, and the
Gbe languages
Languages
Syntactic properties Gungbe Saramaccan English
Irrealis Mood Comp ✓ ✓ *

Deontic Mood Compl ✓ ✓ *
Focus particle ✓ ✓ *
FocP: XPi-wε [IP . . . ti . . .] ✓ ✓ *
Verb focus + V-copy Vi- wε [IP . . . Vi . . .] ✓ ✓ *
Event relativization Vi- Rel [IP . . . Vi . . .] ✓ ✓ *
Focusing the clause [IP . . . . . .]i-wε . . . ti . . . ✓ ✓ *
Analytic Wh-question words ✓ ✓ *
Topic particle ✓ ✓ *
Topicalization of the clause Vi- wε [IP . . . Vi . . .] ✓ ✓ *
Sentence-final yes–no question particle ✓ ✓ *
become dominant under recombinations in new ecologies. I hope to come back to these issues in
future work.
18
This view also implies that the recombination of syntactic features may lead to changes across
syntactic paradigm. For instance, the specification of tense as free morpheme may have
consequences in the development of mood and aspect markers and the blocking of verb move-
ment. This of course strengthens the comparison between languages and biological species, but
I leave the issue for further research.
Table 4.2 Distributive properties of the noun and its modifiers
Languages
Syntactic properties Saramaccan English Gungbe
Bare NPs: generic and (in)definite ✓ %✓ ✓

Definite vs. indefinite * ✓ *
D: specific vs. nonspecific ✓ * ✓
D precedes N ✓ ✓ *
N precedes D * * ✓
N precedes that/this; here/there ✓ ✓ ✓
That/this precedes N ✓ ✓ *
Rel. clause precedes D * * ✓
D precedes Rel. clause ✓ ✓ *
Adjective precedes N ✓ ✓ *
N precedes Adjective * * ✓
the left periphery. Yet, with the exception of the focus marker wὲ , which is
morphologically identical to that of Fongbe and Gungbe, most Saramaccan
markers derive from English (and keep the morphosyntax of English to some
extent). Maintaining the analogy between I-languages and biological
organisms, I claim that this is akin to feature recombination where the Gbe
linguistic features became dominant (Aboh 2006a, 2007a).
Though these languages show strong parallels, when it comes to the
complementizer system, other modules of the grammar may be differently
affected due, to competition and selection. Table 4.2, on the syntax of the
noun phrase, shows this: the determiner in Saramaccan and Gungbe display
semantic parallels but differ with regard to syntax (Aboh 2006a, 2007a).
Table 4.2 therefore suggests that while we can argue that the selected
semantic features are from Gbe, their morphosyntactic realization follows
English rules. By analogy again, this would mean that the features selected
from Gbe became weak under recombination. Alternatively, one could propose
that the competition and selection for the features specific versus nonspecific
implied two choices: one for the semantics of the feature and the other for the
syntax. Accordingly, both competing language types (e.g., Gbe vs. Germanic)
won part of the battle here: Gbe-type languages won on the semantic side and
English won on the syntax side. This is what Mufwene refers to as “Pyrrhic
victory” for the lexifier in (13).
Under the competition and selection process as I interpret it here, these facts
indicate that creoles or contact languages are not (approximate) replicas of
existing linguistic systems. Instead, these new languages are syntactically
4.3 Semantic and syntactic recombination 145
hybrid.19 As such, their emergence is comparable to speciation (e.g., the

emergence of a new species), which arguably evolves from cumulative
recombination of different syntactic features in various modules of the
grammar during the formation of the I-creole.
In some recent critique of this view certain authors (e.g., Plag 2011;
McWhorter 2012; Bakker 2014), misleadingly use terms such as (ad)mixture
to characterize the process of competition and selection as described here. It
occurred to me that these authors misunderstood the point made in Aboh (2006a,
2009a) and Aboh and Ansaldo (2007). They assume that the process of
competition and selection is a mere “mixing” of grammatical properties
of different languages into one. As should be clear by now, the process of
competition and selection adopted and elaborated in the present approach is
constrained not only with regard to the features that it operates on (e.g., syntactic
features) but also with regard to how such features can be recombined into a
coherent system. In order to show this, the following section anticipates the
discussion in Chapters 5, 6, and 7 in showing how competition and selection help
account for semantic and syntactic recombination in the Suriname creoles.
4.3 Semantic and syntactic recombination

As argued for in previous sections, creole languages emerge from the
recombination of various semantic and syntactic features derived from the
competing languages. Keeping this line of thought, I now discuss two
instances of recombination in the verbal and nominal domains. I start with
the verbal domain: the verb ‘eat,’ which translates in Gungbe and Saramaccan
as ɖù and njan, respectively. Recall from the discussion in Chapters 2 and 3 that
Gungbe belongs to the Gbe language family, which, as Smith (1987) and much
related work have shown, played a major role in the emergence of the Suriname
creoles in the late seventeenth and early eighteenth centuries.
4.3.1 Recombination at the syntax–semantics interface: eating poison

In Gungbe, and most languages of the Gbe family, the verb ‘eat’ is comparable
to the so-called inherent complement verbs (ICV). As explained by Avolonto
(1995), Essegbey (1999) and references cited therein, these verbs require an
object in their citation form:
19
As I explained in Chapter 1, the reader should refrain from confusing the process of hybridiza-
tion as developed in this book with the way Whinnom (1971) conceived of it to account for the
emergence of creoles. For instance, the latter spoke of stages of hybridization in ways that do not
apply to language acquisition. Also, he did not discuss feature recombination as a process that
contributes to the restructuring of a language into a new variety, the process that Mufwene
(2001, 2005b, 2008) characterizes as “speciation.”
(21) a. Kòfí ɖó wèzùn.

Kofi plant race
‘Kofi ran.’
b. Kòfí ɖó gbàdó
Kofi plant corn
‘Kofi planted maize.’
c. *Kòf ɖó ...
Kofi plant
‘Kofi planted <something>.’
As the example (21c) indicates, an ICV requires a complement to its right.
More importantly, the contrast between (21a) and (21b) shows that ICV’s
mainly consist of a combination of a light verb and an NP complement, the
semantics of which is compositional in that it builds on the two components of
the VP (see, e.g., Essegbey 1999; Aboh and Dyakonova 2009; Aboh 2010d).
Like ICV’s the Gungbe equivalent of the verb ‘eat’ requires a complement to its
right as shown in (22a) and (22c):
(22) a. Kòfí ɖù nú.
Kofi eat thing
‘Kofi ate.’
b. Kòfí ɖù làn.
Kofi eat meat
‘Kofi ate meat.’
c. *Kòfí ɖù . . .
Kofi eat
‘Kofi ate.’
Considering the contrast between (22a) and (22b), it appears that example (22a)
only denotes the semantics of ‘to eat’ not ‘to eat thing’ while (22b) can only mean
‘to eat meat.’ Compared to the examples in (21), where one has the impression
that it is the following noun phrase which mainly contributes to the semantics of
the ICV, the examples in (22) suggest ɖù does have the basic meaning of ingest/
consume somehow comparable to English ‘eat.’ Yet, this ICV occurs in various
contexts where it is hard to invoke the meaning of ingesting something. Consider
the following examples and their various meanings:
(23) a. Tà ɖù mì.
head eat 1SG
‘I have a headache.’
b. Kòfí ɖù kwέ cè.
Kofi eat money 1SG.POSS
‘Kofi spent my money.’
c. Kòfí ɖù àlè.
Kofi eat profit
‘Kofi made a profit.’
Other expressions include:
(23) d. ɖù àxɔ́
eat debt
‘to go bankrupt’
e. ɖù gbὲ
eat life
‘to enjoy life’
f. ɖù xwè
eat year
‘to celebrate’
g. ɖù àɖì
eat poison
‘to be angry’
h. ɖù wìnyán
eat shame
‘to be ashamed’
i ɖù nù gò
eat mouth PREP[at]
‘to boast’
j. ɖù gán
eat chief
‘be(come) chief’
k. ɖù yà
eat pain
‘to suffer’
While the examples in (23) might look diverse and unrelated at first sight, the
underlying meaning suggests that the logical subject experiences a state or
feeling. In this regard, it should be noted that these examples are not fixed or
idiomatic expressions because they are quite productive and allow further
combinations. Consider, for instance, the sequence (23e) ɖù gbὲ ‘to enjoy life’
as used in sentence (24a). In this example, the inherent object noun phrase gbὲ
‘life’ is further specified to form a possessive noun phrase literally ‘life of Kofi.’
Yet the intended meaning is not one whereby the speaker enjoys Kofi’s life but
rather refers to a state of affairs where the speaker enjoyed or liked an action
performed by Kofi. This sentence can be uttered in various contexts, such as a
situation where Kofi invites his friends to his place or to a restaurant to offer them
dinner.20 Sentence (24b) on the other hand refers to a situation where the speaker
was very happy with the car and liked driving (or being transported in) it.
(24) a. Mí ɖù gbὲ Kòfí tɔ̀ n.
1PL eat life Kofi POSS
‘We enjoyed staying with Kofi.’
b. Mí ɖù gbὲ mótò lɔ́ tɔ̀ n.
1PL eat life car DET POSS
‘We enjoyed the car.’
Finally, the facts in (22), (23), and (24) indicate that the verb ɖù in Gungbe poses
no particular agentivity or animacy restriction on its external and internal
arguments, which could be both theme and experiencer. These data suggest that the
lexical entry corresponding to English ‘eat’ in the Gbe languages covers a wider
range of meanings than English ‘eat,’ whose semantics is much more restricted.
Likewise, that this verb is an ICV in Gbe indicates that it is strongly transitive,
and therefore has a different argument structure than English ‘eat,’ which is either
transitive or intransitive depending on the context. The morphosyntactic
specifications of V-eat in the two language types is represented in (25):
(25) a. Gungbe
vP
VP
V NP
[N-to-V incorporation]
b. English
vP
v VP
V[eat] …
V-to-v movement/incorporation
20
One also finds ɖù mὲ jí: eat someone on => to win over someone, ɖù gú: eat inheritance => to
inherit etc.
Without getting into the intricacies of the formalism, the description in (25a)
suggests that ICV’s in Gbe are genuine light verbs that merge in little v and
select for a VP whose head is a transitive empty V. I submit that such empty V’s,
select for an NP complement, thus creating an incorporation context in which
the head N incorporates into V (Baker 1988; Hale and Keyser 1993; Aboh and
Dyakonova 2009; Aboh 2010d). We therefore reach a situation where it is the
incorporated N that lexicalizes V. Given that little v is a light verb, the
semantics of the ICV derives from the complex v + N (that has incorporated
into V). In the case of the discussed examples, I conjecture that v-ɖù has the
vague meaning of ‘to get’ (by ingestion of X, association with X, or being
affected by X).21 I further conclude that it is the selection of the appropriate
complement that further specifies the meaning of the complex v + N. In case N
is a dummy element, such as nú ‘thing’ in (22a), the verb receives the generic
meaning of ingestion on the basis of the context.
English in (25b), on the other hand, displays a lexical V-eat that is specified
as either transitive or intransitive and merges under V from where it raises to
little v in the current minimalist approach (e.g., Chomsky 1995). Contrary to
the situation in Gungbe, English eat seems to have the basic meaning of
ingestion for the purpose of feeding oneself or someone else.
Thus, let us consider the corresponding verb njan ‘eat’ in Saramaccan. With
regard to syntax, Saramaccan njan exhibits the same behavior as English eat in
that it can be transitive or intransitive. This is illustrated by the following
examples. Sentence (26) indicates that, like English eat, the verb njan may
take an internal argument.
(26) Amato njan di bakuba.
Amato eat DET banana
‘Amato ate banana.’
Example (26) is obviously compatible with the Gungbe example (22b),
in which the light verb v-ɖù selects for a complement. Accordingly, there
seems to be no clear difference between Gungbe, English, and Saramaccan
in this respect. Yet, when we contrast example (26) with those in (27),
taken from Rountree and Glock (1982: 43), we realize that Saramaccan
njan may lack a complement just as English eat (27a–b) and unlike
Gungbe ɖù (22c).
(27) a. I njan kaa no?
2SG eat already Q
‘Have you already eaten?’
21
English examples that illustrate this description include constructions such as ‘I got some food,’
‘I got money,’ or ‘I got malaria.’
b. Ai mi njan (kaa).
yes I eat already
‘Yes I have eaten (already).’
c. Ai mi njan soni.
yes 1SG eat something
* ‘Yes I ate.’
✓‘Yes I ate something.’
In addition, the ungrammaticality of Saramaccan (27c) under the relevant
reading, in contrast with the Gungbe example (22a), indicates that Saramaccan
njan cannot select for a dummy noun phrase. This clearly indicates that this
Saramaccan example has the same argument structure (and therefore syntax) as
the English verb eat. I therefore conclude that the syntax of njan in Saramaccan
maps onto that of English eat as represented in (25b). This would mean that like
English eat, Saramaccan nyan merges under V from where it moves to v, which
it lexicalizes.
Yet, Saramaccan njan occurs in sequences such as in (28), many of which
(e.g., to spend, to suffer, to boast, to have a headache) have literal translations
in Gbe as previously illustrated in (23).
(28) a. Hédi tá njan mí. [Rountree and Glock 1982: 39]
head PROG eat 1SG
‘I’m having a headache.’
b. Nján búka [Donice and Voorhoeve 1963: 80]
eat mouth
‘to boast’
c. Kofi ta nyan suti buka [Haabo, p.c May 22, 2007]22
Kofi PROG eat sweet mouth
‘Kofi is boasting.’
d. Njan moni
eat money
‘to spend money’
e. Njan pena
eat pain
‘to suffer’
f. Njan yai
eat year
‘to celebrate’
22
I thank Vinje Haabo for providing me with these examples.
Assuming that the representation in (25b) holds for the verb njan in
Saramaccan and eat in English, the question arises of how Saramaccan
speakers developed the usages (or semantics) in (28) which are similar to
Gbe constructions. Two options seem possible here. One could propose that
such expressions in Saramaccan are calqued on Gbe and are comparable to
fixed idioms such as kick the bucket in English. But this view is unlikely given
that the sequences in (28) are highly productive and the verb can take additional
arguments as indicated in (29).23
(29) Kofi ta njan Gaamá a baka [Haabo, p.c. May 22, 2007]
Kofi PROG eat Grandman PREP back
‘Kofi is gossiping about the chief.’
Another possibility, worth considering is that the verb njan in Saramaccan
combines semantic properties of both Gungbe ɖù and English eat. More
specifically, I submit that the uses in (28) represent an extension of the
semantic specifications of eat in English, under the influence of Gbe. A look
at the entry eat in The New Oxford Dictionary of English (2001) provides us
with very interesting examples as in (30), which are reminiscent of the Gbe
usages and could have served as the transmission belt for combining English
and Gbe semantics of ‘eat’.
(30) a. Eat one’s heart out [i.e., suffer from excessive longing, especially for some-
one or something unattainable].
b. What is eating you? [i.e., What is worrying or annoying you?]
These examples are similar to the Gbe examples in (23a, 23i) and indicate
that both in Gbe and in English the lexical entry for ‘eat’ may have an
experiencer as internal argument. Under this description, Saramaccan njan
maps the semantic properties of English and Gbe ‘eat’ onto the syntax of
English. In order to see how the argument goes, let us try again to represent
the lexical items ɖù from Gungbe, eat from English, and nyan from
Saramaccan on the basis of the format in (18).
Starting with Gungbe ɖù, it appears that, unlike English ‘eat’, it does not
mean ‘to ingest something for the purpose of nutrition’ but rather has a vague
meaning that is comparable to English ‘to get’, for which emphasis is put on the
experiencer, like in Gungbe. In addition, the morphosyntax of this verb
suggests that it is a light verb. Putting these properties together, we can
describe this verb as in (31):
23
Interestingly, the usage and meaning in (29) do not seem to exist in Gbe, although I may be
wrong.
(31) Lexical item

– ICV CAUSE TO GET: allows
– merges in v productive use with
no V-to-v movement a variety of complements
– v selects a VP with thus providing information
– V= Ø on the experiencer
– V selects for NP and
– N incorporates in V
Cf. representation 25a
English eat on the other hand can be represented as in (32):

(32) Lexical item

it – transitive/intransitive INGEST (mainly) for the
– it merges in V purpose of nutrition.
– V transitive selects Mainly selects for eatable
for NP complement referents as complement
– V-to-v movement
applies.
– No incorporation
Cf. representation 25b
It is clear from these two descriptions that not only do the lexical items
used to express the meaning of ‘eating’ differ regarding their respective
semantics in Gungbe and English but also they show fundamental
differences regarding their morphosyntax. While the Gbe element is
strongly transitive and always requires a complement (even in its citation
form) the English lexical item has usages where it is transitive or
intransitive as in the contrast between the lions ate the visitor versus the
lions already ate. In addition, the Gungbe verb occurs in almost unlimited
semantic contexts with varied meanings, while the English verb has a fixed
meaning of ingesting some substance for the purpose of nutrition. Given
this background, the description of the Saramaccan verb nyan appears to
correspond to representation (33).
(33) Lexical item
Phonology (Morpho)syntax Semantics: English

– transitive/intransitive INGEST (mainly) for
– it merges in V the purpose of
– V selects for NP nutrition. Mainly selects for
complement eatable referents as complement
– V-to-v movement
applies. Semantics: Gungbe
– No incorporation CAUSE TO GET: allows
Cf. representation 25b productive use with a variety of
complements thus providing
information on the experiencer
Several observations are in order here. First, the representation in (33) clearly
indicates that the Saramaccan lexical item njan is not a replica of its counterparts
in English (32) and Gungbe (31). Instead, the newly created item appears to
share some properties of its counterparts in both source languages. Second,
rather than resulting in a mere (ad)mixture as some creolists misleadingly claim,
the process of recombination that results from the competition and selection
appears to be selective. We can indeed conclude on the basis of representation
(33) that the speakers of Saramaccan did acquire the morphosyntax of the lexical
entry corresponding to the verb eat in English concurrently with its different
usages as transitive or intransitive verb. Aside from the phonological form, the
fundamental change (or alteration) that is responsible for a difference between
English (in this case the primary lexifier) and Saramaccan, lies in the semantic
component: the semantics of the English verb has been conflated with that of
Gbe, which accounts for its wider range of usages. These usages in turn come
with their own syntax which may affect sub-domains of Saramaccan syntax.
Such an evolution suggests that language transfer is a selective process.
An apparent issue that is often raised against the feature pool idea and the
recombination of syntactic features proposed here is why one finds some
instances of recombination as opposed to others. For instance, why does
recombination here happen in the semantic component (but apparently not in
the syntactic component)? In the present case, why doesn’t the Saramaccan
verb njan display the syntax of its Gbe cognate ɖù in some well-defined cases
while displaying English syntax only in other contexts? In keeping with the
biological metaphor, it is clear that such a question does not make much sense
as it boils down to asking why mutation happens the way it happens, for
example, why is a mule not half a horse and half a donkey? The relevant
scientific question, it seems to me, is for students of evolution to understand
how recombination happens, producing the particular phenotypes that it yields.
In anticipation of Chapters 5 and 6, I submit that recombination appears to be

particularly sensitive to interfaces. Indeed, the change in (33) happens at the
syntax–semantics interface that regulates the mapping of the form onto the
meaning. Given this observation, we now have a diagnostics as to where to look
for effects of competition and selection: While all linguistic features in the
feature pool can potentially be selected for recombination, those features that
relate to interfaces (e.g., syntax–morphology, syntax–semantics, morphology–
phonology) appear to be more competitive than other formal features
which activate structure building processes only. In the following section, I
discuss derivational morphemes which appear to be sensitive to the syntax–
morphology interface.
4.3.2 Recombination at the morphology–syntax interface: àzé-tɔ́

versus aze-man
Van den Berg and Aboh (2001) show that Gbe languages (e.g., Gungbe) and
Suriname creoles (e.g., Sranan and Saramaccan) display nominal sequences
that look superficially like simple X-Y adjunctions, where X and Y are heads
(e.g., noun). The sequences in (34) provide us with such compounds.
(34) a. Àjàkà dò [Gungbe]
Rat hole
‘rat hole’
b. arátta-oso [Sranan; Schumann 1783: 6]
rat-house
‘rat hole’
These compounds co-exist alongside with sequences involving -man in
Sranan and Saramaccan or -tɔ́ /-nɔ̀ in Gungbe, which one could imagine
involve X-Y adjunction structures too.
(35) a. blέɖì nɔ̀ [Gungbe]
Bread person
‘bread seller/bread man’
b. helpi-man [Sranan; van den Berg 2003: 242]
help person
‘midwife’
c. àzé tɔ́ [Gungbe]
witchcraft person
‘witch’
d. aséh-man [Sranan; Schumann 1783: 8]
witchcraft person
‘witch’
However, a number of facts about these sequences suggest that tɔ́ /nɔ̀ /man are
affixes within a predicate structure. One such fact is that the Sranan and
Saramaccan element -man (derived from English man) and the Gungbe
forms -tɔ́ /-nɔ̀ , which literally mean ‘father’ and ‘mother’ respectively, are
unspecified for gender in compounds, unlike when they occur in isolation.
Therefore, while the examples in (35) can refer to both male and female person,
those in (36) have gender specifications and only refer to either a male or
female person.24
(36) a. (ò)tɔ́ cè
father 1SG.POSS
‘my father’
b. (ò)nɔ̀ cè
mother 1SG.POSS
‘my mother’
c. man-ningre / ningre uman
man black black woman
‘black man/black woman
An immediate conclusion here is that the elements -man (Sranan/
Saramaccan) and -tɔ́ /-nɔ̀ (Gungbe) do not function as lexical items when they
occur in compounds. Instead, they are somehow semantically bleached and
behave like derivational affixes (or grammaticalized items) that attach to the
element on their left. This description finds immediate support in examples
such as those under (37), where these morphemes attach to a verb phrase.
4.3.2.1 VP-tɔ́ /nɔ̀ versus V-man/VP-man In Gungbe, the morphemes -tɔ́

and -nɔ̀ can attach to nominalized verb phrases. In such contexts, the object
must precede the verb as indicated in (37) (see Aboh 2004a, 2005c; Aboh and
Smith 2012 for discussion of such nominalized VPs).
(37) a. Ùn mɔ̀ n [[hún kùn tɔ́ ] lɔ́ ]
1SG see engine drive person DET
‘I saw the engine driver [i.e., the driver].’
b. Ùn mɔ̀ n [[nùkún tɔ́ n nɔ̀ ] lɔ́ ]
1SG see eye pierce person DET
‘I saw the blind person.’
That the noun can be modified as in (38) suggests that it heads a phrase.
(38) a. Ùn mɔ̀ n [[hún gbó kùn tɔ́ ] lɔ́ ]
1SG see engine big drive person DET
‘I saw the big engine or truck driver.’
24
See van den Berg (2003: 244) for similar examples.
b. Ùn mɔ̀ n [[nùkún ɖòkpó tɔ́ n nɔ̀ ] lɔ́ ]

1SG see eye one pierce person DET
‘Lit. I saw the one eyed person.’
It appears that the VP-tɔ́ /-nɔ̀ sequences in (37) and (38) are complex phrases
comparable to noun phrases though they involve a VP. This is indicated by the
fact that the whole VP-tɔ́ /-nɔ̀ sequence can take a determiner and function as
argument of a verb. This latter property leads me to conclude that the whole
VP-tɔ́ /-nɔ̀ sequence is included in a determiner phrase (DP): where D embeds
the nominalized verb phrase.
The Gungbe facts are replicated in Sranan and Saramaccan, where the
morpheme -man attaches to a VP as attested both in early and contemporary
Sranan (Bruyn 1995a, 1995b).25
(39) a. loecke-man [Early Sranan; Court Record 1745]
see-man
‘medicine man’
b. hakisi-man [Early Sranan; Van Dyk 1765]
ask-man
‘inquisitor’
c. figi-strati-man
sweep street man
‘street cleaner’26
At this stage, it is worth noting that Sranan and Saramaccan display VO order
in their nominalized VP while Gungbe exhibits OV order. Setting this
difference aside for the time being, the situation in Sranan and Saramaccan is
very much like that in Gungbe in the sense that man has affixal properties.27
25
Smith and Veenstra (1998) observed similar examples in Saramaccan.
26
Contemporary Sranan displays such sequences with VO order as in the following example:
(i) Sibi-strati-man [Contemporary Sranan]
sweep street man
‘Street sweeper’
27
This issue merits further attention and I return to this discussion in Section 4.3.2.2. As
mentioned in Section 4.2.1, it is remarkable that while nominalized verb phrases display OV
order in both Gbe and English the Saramaccan equivalents maintain VO order. This correlates
with the facts discussed in DeGraff (2002), and Aboh (2006a) where it appears that even though
most Kwa languages exhibit VO versus OV alternation in the context of aspect, none of their
supposedly related creole has the alternation. These facts cannot be accounted for with uni-
versals, since the observed alternations are context-dependent and do not imply a different head
parameter (Aboh 2004a, 2005c). In addition, the compounds discussed here are similar to
English compounds, such as, gingerbread man, house-cleaning man, in which the properties
of man (as a head noun or an affix) are not clear. Assuming that there is surface congruence
between English and the Gbe languages with regard to these compounds, then the decisive
factor for its syntax in the Suriname creoles could have to do with whether or not it is treated as a
derivational affix as in Gbe.
With regard to morphology, for instance, the elements. -tɔ́ /-nɔ̀ and -man
display a reduced form, which supports its analysis as an affix. Empirical
data from, on the one hand, Gungbe and, on the other, Sranan and Saramaccan
support this point. As discussed in Aboh (2005a, 2010b), and in Aboh and
Smith (2015a), Gungbe noun phrases have an initial vowel (a- or o-) as
indicated in (40):
(40) a. òhún ‘drum’
b. àgásá ‘crab’
The prefix o- can be dropped, while the prefix a- cannot.
(41) a. Kɔ̀ kú xɔ̀ òhún/hún ɖé.
Koku buy drum DET
‘Koku bought a drum.’
b. Kɔ̀ kú xɔ̀ àgásá/*gásá ɖé.
Koku buy crab DET
‘Koku bought a crab.’
Yet, o- and a- must drop in compound nouns, when the noun they attach to
occurs as a second term. The examples in (42) illustrate a noun with the initial
vowel o-:
(42) a. (ò)hún kpòtín
drum stick
‘drum stick’ (as stick/mallet to play drums)
b. sìn-(*ò)hún
water drum
‘water drum’ (played at funerals)
Those in (43) on the other hand illustrate a noun with the initial vowel a-:
(43) a. àgásá fὲn
crab foot
‘crab foot/pincers’
b. xùmὲ-(*à)gásá)
sea crab
‘sea crab’
In both cases, the initial vowel must drop when it occurs on the second
conjunct of a compound (42b, 43b). The same constraint applies to the source
nouns òtɔ́ ‘father’ and ònɔ̀ ‘mother’ which fail to retain their initial vowel in
compounds (30).
(44) a. *hún kùn òtɔ́ lɔ́
car engine person DET
‘driver’
b. *nùkún tɔ́ n ònɔ̀

eye pierce person
‘blind person’
Accordingly, the morphemes tɔ́ and nɔ̀ ‘person’ are reduced forms of the
nouns òtɔ́ ‘father’ and ònɔ̀ ‘mother.
The facts in Sranan are similar even though less straightforward. Here, one
finds the alternation man versus ma in some sources, which suggests that the
reduced form ma is an affix, while the full form man is a lexical item.
(45) a. ase-ma(n)
witchcraft-person
‘witch’
b. Joe ben zi hem na da man disi zire boeki.
you PST see him PREP DET man REL sell book
‘You have seen him at the man who sells books.’
[Gy hebt hem by myn Boekverkooper gezien] (Van Dyk c.1765: 31)
Smith and Veenstra (1994, 1998) report a similar alternation in Saramaccan,
where the form -ma represents an affix, while the full form mánu is the noun.
(46) a. édi-ma
head-man
‘boss/headcase’
b. téi-mánu-ma
take-man-man
‘man eater’
In addition to showing that the Africans were aware of the stress difference
in English between fully stressed man (as in this man) and the unstressed
suffix man (as in policeman), these facts suggest that tɔ́ /nɔ̀ and ma(n)
represent affixes that attach to the phrase to their left. In sum, the lexical
elements that correspond to ‘father,’ ‘mother,’ or ‘man’ in the Gbe languages
and the Suriname creoles can be used derivatively to form new phrases. The
next question, to which I now turn, deals with the structure of such phrases.
4.3.2.2 X-tɔ́ /nɔ̀ /man as predicate structure A fact that one immediately
notices about the sequences X-tɔ́ /nɔ̀ /man is that they express a relation such
that X, a noun phrase, denotes an abstract entity, property, or quality that
predicates over tɔ́ /nɔ̀ /man. When X represents a verb phrase, however, it
expresses an event of which tɔ́ /nɔ̀ /man is an agent. The formed X-tɔ́ /nɔ̀ /man
sequences can be used in equative constructions such as in John is X-tɔ́ /nɔ̀ /man
which could mean ‘John is wealth-person (i.e., rich)’ or ‘John is drive-person
(i.e., driver).’
Consistent with Aboh’s (2005a, 2010b) analysis of similar sequences in

Gungbe, I propose that the Gungbe sequence àzé-tɔ́ in (35c) and Sranan
aséh-man in (35d) involve a structure of the type in (47), where the item to
the left is the subject of the nominal predicate expressed by the head -tɔ́ or
-man. The two elements are linked by the inflectional element F° (Kayne 1994;
den Dikken 1998).
(47) FP
spec F¢
aze/ aseh
F° NP
t /man
Under this representation, the head N tɔ́ /nɔ̀ /man incorporates into Fº where it
functions as an affix. Incorporation of N into Fº explains the morphological
change of the full form corresponding to the lexical entry man, òtɔ́ ‘father,’ ònɔ̀
‘mother’ into the reduced form -tɔ́ /nɔ̀ /ma(n), as discussed previously. On this
view, the sequences involving a (nominalized) verb phrase in (37) for Gungbe
and (39) for Sranan can be derived as in (48):
(48) FP
spec F¢
VP[Nom]
F° NP
t /man
Sranan and Gungbe differ in that the Sranan VP realizes the VO order or may
contain the verb only. In Gungbe, however, the nominalization of the VP
requires the OV order. This difference aside, what we see here is that the
Sranan forms aséh-man or sibi-strati-man map on the (Gbe) structures (33)
and (34), where man has been substituted for tɔ́ /nɔ̀ .
The case of aséh-man ‘witch’ is illuminating in this respect because it
raises an interesting issue with regard to language acquisition. Because the
speakers who created this form had spoken Gbe, I surmise that they could
have simply retained the equivalent Gbe word àzé-tɔ́ , just as they retained
aséh alongside with numerous other Gbe lexical items. Yet, this did not
happen. Instead, the form aséh-man represents convincing evidence that
these English learners did entertain learning hypotheses about English
derivational morphemes. This acquisition process, I claim, allows the
learners to relate the Gbe derivational morphemes -tɔ́ /-nɔ̀ to English -man
and -er. For the creators of the Surinamese creoles to be able to match these
English lexemes with Gbe structures, however, they have to make the right
hypothesis that tɔ́ /nɔ̀ corresponds to man in its affixal usage which paved the
way for the development of the derivational morpheme -man. The latter
appears to embed properties of the derivational morpheme -er as a result of
which sequences of the type X-man express a relation such that X may be a
noun phrase or a verb phrase. When X is a noun phrase, it often represents an
abstract entity, property, or quality that predicates over man. On the other
hand, the contexts where X is a verb phrase express events of which man is
the agent.
Yet, morphosyntactic constraints on the development of -man resulted in
a change that is worth discussing. In the previous paragraphs, I mentioned
that while contexts such as (48) require the OV order in Gungbe, they are
realized as VO in Sranan and Saramaccan. Let us consider again the Early
Sranan example in (39c) repeated here as (49a), compared to its Gungbe
equivalent in (49b).
(49) a. figi-strati-man
sweep street man
‘street cleaner/sweeper’
b. àlìò-jí zà tɔ́
road-surface sweep person
c. street sweep-er
If indeed, the Sranan and Saramaccan affix -man were mapped on the Gbe
morphological structure as suggested in (48), one wonders why the two
languages display different word orders within the nominalized verb phrase
in its specifier. The issue appears even more complex when one considers
similar sequences in English. As example (49c) shows, the English equivalent
of (49a) also exhibits the OV order. We are therefore in a situation of strong
congruence where both source languages display OVorder in the same context
but the creole realizes the VO order. Why?
In order to answer this question, it is important to realize that, according to
the structure in (48), the morphemes -tɔ́ /-nɔ̀ and -man/-er have the
selectional requirement that the category in their specifier be a noun
phrase or a nominalized predicate (e.g., a nominalized verb phrase).
Crucially, however, these categories do not act as nominalizers (as one
could imagine from some traditional accounts of English -er).
Accordingly, the variation between, on the one hand, Sranan and
Saramaccan and, on the other, the Gbe languages results from the
structural make-up of nominalized phrases in both language types. The VO
versus OV variation that we observe in (49) is a consequence of how the two
languages nominalize verb phrases (rather than how the elements -tɔ́ /-nɔ̀ ,
and -man /-er combine with their specifier). Here again, we are hitting on an
interface issue, that is, the interface between morphology and syntax.
To make it simple, nominalization is a category changing process whereby
VPs become noun-like constituents in the context of some morphemes. This
can be informally represented in (50), where Fnr is the relevant category
(Marantz 1997):
(50) FPNR
FNR¢
FNR VP
If this is the right characterization, then the word order differences between
Gungbe, on the one hand, and Sranan and Saramaccan, on the other, can be
regarded as consequences of the different licensing properties of Fnr in both
languages. Let us therefore consider how this category is licensed in English
and Gungbe as opposed to Sranan and Saramaccan. Looking at English and
Gungbe, an observation that immediately comes to mind is that such OV
sequences are found in other contexts as well. In English, one also finds OV
order with nominalizations involving the affix -ing as in (51).
(51) a. Street sweep-ing
b. Car steal-ing
In Gungbe, on the other hand, one finds similar OV sequences in VP
nominalizations involving verb reduplication as in (52):
(52) a. àlìò-jí zìzà
road-surface sweep-sweep
‘road sweeping’
b. mótò fìn-fìn
car steal-steal
‘car stealing’
Interestingly, a common denominator for English -ing and Gungbe
reduplicated verbs is that they are also found in progressive constructions as
illustrated in (53a–b) with the verb kú ‘to die’ (Aboh and Smith 2012). Example
(53c) further indicates that when the progressive construction involves a
transitive verb in Gbe, the object must precede the verb. Finally, note that
progressive constructions in Gbe involve a clause-final particle which in
Gungbe is expressed by a floating low tone.
(53) a. John is dy-ing
b. Ján to kú-kû
John PROG die-die.PCL
‘John is dying.’
c. Ján tò mótò kûn

John PROG car drive.PCL
‘John is driving the car.’
Given this observation, let us assume that OV in both English and Gungbe
is contingent on the formation of gerunds in these languages, which itself
brings about nominalization. The latter, I propose, relates to specific
morphological processes which translate into the suffix -ing in English and
verbal reduplication in Gungbe. I have discussed verbal reduplication in Gbe
and its relations to object shift and nominalization in a series of studies to
which the reader is referred for a detailed discussion: Aboh (2003a, 2004a,
2005c, 2007b, 2009b) and Aboh and Smith (2012, 2015b). In the present
demonstration, I limit myself to the relevant aspect of the analysis developed
in these studies.
In order to account for the Gungbe facts in (52b) and (53b), I proposed in, for
instance, Aboh (2004a, 2005c, 2007b, 2009c) that Gbe languages involve a
structure in which the progressive morpheme embeds the structure in (54)
where the projection labeled here as FPnr embeds an inflectional layer on top
of the verb phrase:
(54) FPNR
spec FNR¢
FNR IP
spec I¢
I VP
spec V¢
V DPobject
Because the structure in (54) represents a reduced predicative structure it

involves a subject position [spec IP] that must be filled as required by the
traditional Extended Projection Principle, which stipulates that all predicates
must have a subject. Given this hypothesis, three head positions must be
lexicalized in (54), in addition to the subject position [spec IP]: V, I, and Fnr.
In Gungbe, these positions are lexicalized by the floating low tone under Fnr,
the DP object raised to [spec IP], and the verb raised to I, as illustrated in (55). It
is further assumed that in Gbe, the whole IP raises to [spec FPnr] as indicated
by the arrow in (55). This derivation produces the OV order observed in
example (53c):
(55) AspP
spec Asp¢
Asp FPNR
to
spec FNR¢
FNR IP
¢
spec I¢
DPobject
I VP
V
tV tDPobject
With regard to intransitive verbs, there is no object DP that can raise to the
subject position [spec IP]. I argue that in such circumstances, the Gbe
languages are left with only one option: the licensing of a null expletive in
[spec IP]. This process in turn translates into a morphological requirement on
the verb which reduplicates. Under this view, verb reduplication is therefore
seen as an alternative to inflection which licenses a null subject (see, for
example, Aboh 2007b, 2009b; Aboh and Smith 2012 for a most recent
discussion). Everything else being equal, we reach the derivation in (56)
which stands for example (53b):
(56) AspP
spec Asp¢
Asp FPNR
to
spec FNR¢
FNR IP
¢
spec I¢
Expl
I VP
V-V
tV
Building on this analysis, Aboh (2005c) and Aboh and Smith (2012) further
argue that in OVV sequences such as in (52), where an object precedes a
reduplicated verb, it must be the case that the process in (56) is active in
those environments too. In order for the verb to reduplicate in (52), the object
must not raise to [spec IP] as in (55) but to some other position. The obvious
immediate candidate here is [spec FPnr]. Indeed Aboh (2005c) and Aboh and
Smith (2012) demonstrate that such OVV sequences involve the structure in
(57), in which the object raises to [spec FPnr]. Consequently, an expletive is
inserted in [spec IP], which is licensed by the reduplicated (or say inflected)
verb under I. As the reader can also see from this demonstration, object raising
to [spec FPnr] blocks subsequent movement of IP to the same position (unlike
in representations 55 and 56):
(57) AspP
spec Asp¢
Asp FPNR
spec FNR¢
DPobject
FNR IP
spec I¢
Expl
I VP
V-V
tV tDPobject
As an illustration, the derivation in (57) accounts for examples such as (58)

in Gungbe
(58) Yɔ̀ kpɔ́ lέ gbέ hàn lɔ́ kpíkplɔ́ n
child PL refuse song DET learn-learn
‘The children refused to learn the song/refused learning the song.’
I refer the reader to the cited references for a comprehensive discussion
of the OV and OVV structures in Gbe. What matters for the current
discussion is that we can apply the rationale in (57) to English OV-ing
expressions such as street-sweep-ing and car-steal-ing in (51). More
concretely, I’m assuming that these expressions are derived from the
same structure as (57). An illustration is given in (59), in which I is the
locus of the inflectional affix -ing:
(59) AspP
spec Asp¢
Asp FPNR
spec FNR¢
DP
FNR IP
street
spec I¢
Expl
I VP
sweep-ing
tsweep tDPstreet
This representation leads to two observations: (i) English OV-ing

expressions are systematically licensed in the context of an aspect phrase; (ii)
OV-ing sequences involve movement of the object to [spec FPnr], while the
verb moves to I, where it attaches to the affixal element -ing, whose function is
to license the null expletive in [spec IP].
In the context of this discussion, the latter observation appears very
interesting because it shows that English -ing and verb reduplication in
Gungbe represent two alternative strategies that serve to license a null
expletive in [spec IP]. In addition, it appears that both in English and Gungbe
this device is contingent on the fact that these languages choose to lexicalize
FPnr by means of object raising to [spec FPnr]. But this is obviously not the only
possible logical option. Another option that is made available by UG is for
languages to choose to lexicalize the head Fnr by raising the verb to that position
thus allowing the DP object to raise in [spec IP]. This is illustrated in (60):
(60) AspP
spec Asp¢
Asp FPNR
spec FNR¢
FNR IP
Verb
spec I¢
DPobject
I VP
tVerb
tVerb tDPobject
As we can see from this representation, such a language will not need a null
expletive in this context. Note also that such a language will display a VO order
in contexts comparable to Gungbe OV(V) or English OV-ing structures. Taking
these observations into account and comparing the Gungbe derivations in (55),
(56), (57), the English representation in (59), and the option presented in (60),
we now reach a typology of gerund-like constructions that will help in
understanding the Sranan and Saramaccan pattern illustrated in (49a), where
it appears that these languages display the VO order where their source
languages display an OV order. Indeed, the discussion here shows that in the
context of gerund-like constructions, languages may choose to lexicalize FPnr,
by raising the object DP in [spec FPnr]. But such languages must have a way of
licensing a null expletive in [spec IP]. In this regard, English resorts to the
suffix -ing while Gungbe employs verb reduplication. The resulting surface
order is OV-affix in English or OVV in Gungbe. On the other hand, languages
may choose not to raise the object in [spec FPnr] but in [spec IP] instead,
thus making the need of an expletive and its concomitant licensing device
unnecessary. Such languages, it appears, must lexicalize the head Fnr
otherwise, for instance, by raising the verb to Fnr. Because these different
options are made available by the human language capacity, they are not
necessarily mutually exclusive and it is perfectly conceivable that the same
language may resort to these different strategies in different contexts.
With regard to the development of the Suriname creoles, the model of
competition and selection that I’m arguing for here, which is based on the
feature pool idea, indicates that the Gbe learners of English were confronted
with these different typological options. Returning now to the discussion in
the previous paragraph, an immediate conclusion that emerges is that if a
language lacks (or loses) the relevant device for licensing a null expletive in
[spec IP], that language will not display an OV order. This is because such a
language will have to raise the DP object in [spec IP] while Fnr is lexicalized
otherwise, for instance, thanks to verb raising as shown in (60). This
conclusion sheds new light on the Suriname creoles: the loss of the -ing
affix is linked to the loss of OV pattern. Returning to the example in (49a)
repeated here as (61a), this would mean that the bracketed sequence
representing the nominalized (or gerund-like) phrase is comparable to
English expressions like that in (61b).
(61) a. [figi-strati]-man
sweep street man
b. [house-keep-ing] assistant
Under this characterization, the impossibility of OV sequence in the
Sranan example in (61a) as opposed to the English version (61b) is a direct
consequence of the loss of affixal -ing. Assuming that the bracketed sequence
represents the structure in FPnr, this amounts to saying that (61a) can be
represented as in (62) (ignoring the AspP projection).
(62) FPNR
spec FNR¢
FNR IP
figi
spec I¢
strati
I VP
tfigi
tfigi tstrati
When we merge this sequence with affixal -man as suggested in (48), we

reach the representation in (63), where VP[Nom] is replaced by FPnr:
(63) FP
spec
F¢
FPNR
F° NP
spec FNR¢ -man
FNR IP
figi
spec I¢
strati
I VP
tfigi
tfigi tstrati
A crucial conclusion we now reach is that, in sequences such as figi-strati-

man in Sranan, -man is a phrasal affix. Returning to the question of word order,
this analysis indicates that Sranan and Saramaccan cannot display English-like
OV-ing sequences due to the loss of the affixal -ing. But since English was in
competition with Gungbe which displays the same order thanks to its usage of
verb reduplication as a device for licensing the null expletive in OVV
sequences, one wonders whether the Suriname creoles could have retained
such similar OVV sequences in some contexts.
In this regard, Aboh and Smith (2012) propose a comprehensive analysis of
non-iconic reduplication in the Suriname creoles some examples of which are
illustrated below.
(64) a. ɗá mi [ɗí páu latjá-latjá] [Bakker 1987: 30]

Give 1SG DET wood split-split
‘Give me the wood split.’ (Translation modified by the author!)
b. A njan [ɗí ɓakúɓa lépi-lépi]
3SG eat DET banana ripe-ripe
‘He ate the banana ripe.’
I will not repeat Aboh and Smith’s (2012, 2015b) demonstration (to which I
refer the reader), but we show in that study that these examples can be analyzed
as secondary predicates where the bracketed sequence corresponds to the
structure I have assigned here to FPnr. According to this analysis, the DP
(i.e., ɗí páu, ɗí ɓakúɓa) realize [spec FPnr], unlike in (63). Thus, the language
resorts to the licensing device inherited from the Gbe languages: verbal
reduplication which licenses the null expletive in [spec IP]. This is
represented in (65) adapted from Aboh and Smith (2012: 53).
(65) VP
spec V′
V FPNR
njan
spec FNR′
DP
FNR IP
spec I′
Expl
I VP
lepi-lepi
We now come full circle. During the process of competition and selection
that led to the development of the Suriname creoles, the English inflectional
morpheme -ing was lost. As a consequence the emerging languages
developed another strategy for licensing the relevant functional positions:
the verb moves to Fnr, while the DP object raises to [spec IP] as indicated in
representations (62) and (63). This accounts for the emergence of VO order in
English-like gerund constructions. Yet, because English was in competition
with the Gbe languages which exhibit object raising and non-iconic
reduplication (i.e., OVV) in such contexts, the Suriname creoles did retain
this strategy in a subset of the relevant contexts, such as in secondary
predicates illustrated in (65). The end result of this development is that
the Suriname creoles appear structurally mixed both with regard to
morphosyntax and semantics.
4.4 Conclusion 169
4.4 Conclusion
The conclusion of this chapter is obvious: creole languages are linguistic
hybrids. They emerged from the recombination of linguistic features from
different languages. While this is the normal situation in every instance of
language acquisition, language change, and language evolution (Kroch 1989a,
1989b, 2001; Kroch and Taylor 1997; DeGraff 2001a, 2001b; Mufwene 2001,
2008; Lightfoot 2006), the case of creoles looks striking at first sight simply
because the recombination involves linguistic features from typologically
different languages (Aboh and Ansaldo 2007). This means that creoles are
noticeable only because of their (contrasting) phenotypes. Nothing in their
structures singles them out as a singular type (Muysken 1981b, 1988).
The approach developed here also shows that creole languages develop
opaque syntactic and semantic features (call it local complexity) which could
not have arisen only in the context of their source languages. If opacity is in
some sense related to complexity, we then reach the situation where the creole
languages, far from being ‘simple’ as some creolists claim, actually hide very
complex morphosyntactic and semantic structures (Aboh and Smith 2009). As
I concluded in Aboh (2009a: 340), the comparison between the Suriname
creoles and their source languages shows clearly that “the notion of
simplicity is completely irrelevant to the understanding of the structure and
the genesis of creole languages, and therefore to the study of language change
or language evolution.” To this, McWhorter (2012: 173) replies “but certainly,
so bold a conclusion cannot stand on two morphemes.” The implicit criticism
is that there is not much that we can learn about the development of new
languages and language acquisition by studying just two morphemes. I surmise
that it is probably easier to hop from one creole to the other counting and
contemplating morphs hoping to discover a holistic theory of simplicity that
would single creoles out as a type.
This book does not adhere to such an easy stance. Instead, it addresses the
development of creoles in the context of the general issue of language
acquisition and language change, by engaging in a rigorous, detailed analysis
of aspects of these languages that shed better light on what possible learning
hypotheses learners develop in the course of acquisition, and how different
linguistic features can be recombined into a coherent system representing the
learner’s I-language. Clearly, such an enterprise can only be made step by step.
As the reader must have noticed, there is indeed a lot to be learned in the study
of just two morphemes. One thing we learned in this chapter is that the
recombination of linguistic features is sensitive to interface properties, hence
the different properties of Saramaccan njan. Because the process is sensitive to
interface properties, it appears not to affect all (sub-) domains of grammar the
same way. In this regard, we learned that while njan points to syntax–semantics
interface properties, affixal -man sheds light on morphology–syntax interface

properties.
The latter observation has a direct consequence for theories of simplicity
of the type developed by McWhorter in several essays as well as for other
exceptionalist theories such as Plag’s (2008a, 2008b, 2009a, 2009b)
interlanguage hypothesis. Indeed, a fundamental mistake that these authors
make is to assume that the loss of inflectional morphology (mainly of the Indo-
European type) systematically corresponds to loss of structure. Aboh and
Smith (2009) have already pointed out that such a view makes no sense in
modern linguistics (see also Baptista 2009, Mufwene 2009, Kouwenberg
2010). This is very obvious when we reconsider the discussion about word
order variation in gerund-like OV-ing and OVV constructions in English and
Gbe as opposed to Sranan and Saramaccan VO and NP-VV structures. The
discussion showed that the loss of English affix -ing did not translate into the
loss of the relevant structure in Saramaccan. Instead, the language developed
a new strategy for satisfying the requirements of the relevant functional heads
in the structure: Fnr and I. Accordingly, one cannot talk about structural
simplification unless one can prove that the structure associated with a
missing morph is also missing. No exceptionalist theory has yet shown this.
In the subsequent chapters, I illustrate how the competition and selection
process applies to various domains of grammar such as the noun phrase
(Chapter 5), the complementizer system (Chapter 6) and the use of functional
verbs in serial verb constructions (Chapter 7).
5 The role of vulnerable interfaces in language
change: the case of the D-system
The study of the recombination of linguistic features in Chapter 4 led to the

conclusion that interfaces play an important role in determining which
linguistic features may be selected over which others and how such features
can be recombined into new bundles of features specifying a lexical item. The
discussion there also indicates that, given the varied inputs that the creole
creators were exposed to and given the varied profiles of the learners on the
plantations, traditional notions such as substrate transfer should be refined. The
major motivation for this conclusion is that most learners were bilingual or
multilingual and were exposed to various possible target languages, including
the emerging creole and/or the emergent colonial koiné. Because the creole is
the output of such varied inputs, it displays morphosyntactic variations that
cannot be unilaterally attributed to superstrate legacy, substrate transfer, or
independent Universal Grammar (UG) in any obvious way.
For instance, the determiner systems of Haitian, Sranan, and Saramaccan
display word order sequences that diverge from those found either in the
superstrate languages (viz., French and English) or in the substrate languages
(e.g., Gungbe). Similarly, a universalist account invoking the Language
Bioprogram (e.g., Bickerton 1981, 1999, 2008) faces the problem of the wide
range of variation these languages and other creoles display with regard to their
determiner phrases. Such variation is inconsistent with Bickerton’s analysis,
according to which, as explained in Section 3.3, all creoles are generated by a
predetermined computational system. Finally, Plag’s (2008a, 2008b, 2009a,
2009b) interlanguage hypothesis not only is incompatible with the emergence
of articles in creoles in general but also fails to explain the distribution of
determiner-like elements in these languages, their (covert) agreement patterns,
and how they interact with (in)definite bare nouns.
In discussing the lexical item njan and the functional item -man in Chapter 4,
I showed that the Suriname creoles display a hybrid structure in which these
items exhibit properties that appear to be combinations of semantic and/or
morphosyntactic aspects of the source languages: Gbe and Germanic. As such,
Aspects of this chapter were already presented in Aboh (2006a).
171
172 The role of vulnerable interfaces in language change
creoles present innovative forms compared to their source languages. The

question now arises how to formalize the emergence of such re-created forms.
Following the discussion in Chapter 4 and focusing on functional categories,
I suggest that one way to approach such new variants would be to propose that,
in a situation of language contact, functional categories are affected differently
depending on their semantic and licensing conditions (i.e., their syntax). Under
the assumption that features such as definiteness, specificity, and number
are visible at the syntax–discourse/semantics interface, I argue that what is
commonly referred to as transfer results from a competition among features
that may lead to a split between syntax and semantics. I submit that the
emerging language may retain a feature Fx from a competing language Lx
and adopt its function (or semantics) and formal licensing properties. This
situation is represented in (1a). On the other hand, the emerging language may
select an Fx on the basis of its function in a competing Lx, while leaving its
syntax unfixed, as represented in (1b). I refer to the former situation as “pattern
transmission” and to the latter as “feature transmission.” In this case, I conclude
that the emerging language may develop a syntax of its own under pressure
from other competing languages and/or based on the principles of UG.
(1) a. Fx [Function = Lx; Syntax = Lx] ➜ Pattern transmission
b. Fx [Function = Lx Syntax = . . .] ➜ Feature transmission
The description in (1) indicates that pattern transmission includes retention of a
feature, its function and its syntax, while feature transmission involves the selection
of a feature and its function. What we see here is that while the retention of syntax
depends on the selection of function, the reverse does not necessarily hold. Within
the framework of Minimalism, such a split between syntax and semantics appears
reasonable if we assume that a feature Fx is associated with a unique semantic/
pragmatic representation cross-linguistically, while its syntax is subject to
parametric variation. This would mean that the same feature may be valued
differently in natural languages, even though its semantics remains the same.
For example, let us consider the expression of contrastive focus in English
and in Gungbe. In English, contrastive focus can be expressed by means of a
cleft construction “It is X that . . . ” as illustrated in (2a), the equivalent of which
is represented by the Gungbe sentence (2b). Unlike English, however, Gungbe
uses a fronting construction in which the focused element is marked by a
grammatical element wὲ . In these examples, I represent the focus device in
italics and the focused constituent in boldface.
(2) a. It was a dog that John killed not a goat.
b. Àvún wὲ Kòfí hù é má nyín gbɔ́ .
dog FOC Kofi kill 3SG NEG be goat
‘It was a dog that Kofi killed not a goat.’
The role of vulnerable interfaces in language change 173
It appears that languages deploy different morphosyntactic strategies to

encode contrastive focus, although the latter can be uniformly defined across
languages as “a subset of the set of contextually or situationally given elements
for which the predicate phrase can potentially hold. [Focus] is identified as the
exhaustive subset of this set for which the predicate phrase actually holds”
(Kiss 1998: 245).
Similar morphosyntactic variation can be observed within the nominal
domain as well. In French and English, specificity is encoded by a
demonstrative while it requires fronting of the noun phrase to the left of a
designated specificity marker within the determiner phrase of Gungbe (Aboh
2004a, 2004c). In this chapter, I argue that, in a situation of language contact,
these different patterns (e.g., Gbe versus Germanic/Romance) may compete for
the same function at the syntax–semantics interface, leading to a particular
combination in the emerging language.
But who are the agents acting on pattern or feature transmission as described
in (1)? As once suggested to me by Tonjes Veenstra (p.c., May 29, 2010), the
description in (1) is compatible with a scenario in which pattern transmission
will be typical of adult L2 learners who may transfer properties of their native
language into the target. Feature transmission, on the other hand, would be
typical of children (either L1 learners or bilingual L1 learners). This view is
compatible not only with DeGraff’s (2002) L2 vs. L1 cascade hypothesis as
applied to the development of creoles but also with recent work on bilingual
studies and the effect of bilingualism on diachronic change (Aboh and DeGraff
forthcoming). According to Meisel (2011: 123):
Grammatical change involving reanalyses of syntactic patterns and leading to the
reorganization of mental grammars is most likely to occur in the process of language
transmission. This is in tune with the claim commonly made in generative studies of
diachronic change, according to which first language (L1) learners are the locus of
change . . . My claim is that cross-generation reanalysis of grammars is most likely to
happen in successive acquisition of bilingualism or if second language speakers provide
a substantial amount of input in monolingual or bilingual first language development.
Therefore, “grammatical change necessarily happens in bilingual settings”
(Meisel 2011: 125), where both adult L2 learners and (bilingual) L1 learners can
be agents of change (see also Tracy 1998; Roeper 1999; Kroch 2001; Lightfoot
2006 for similar views). On the other hand, Weerman (2011) argues that while
bilingualism is certainly a necessary condition for change, it is not sufficient.
According to this author, what matters instead is “the opposition between early
versus late acquisition (p. 149), where late acquisition involves both child and
adult L2 learners. Setting aside the details about these studies, what emerges
from these well-documented studies of bilingual acquisition and diachronic
change is that some type of contact of different linguistic systems (involving
different dialects of the same language or typologically and genetically different

languages) is a necessary condition for the reanalysis of syntactic patterns. As
Meisel (2011: 138) remarks,
the human Language Making Capacity is an extraordinarily robust device. Transmission
failure is therefore much less probable and occurs less frequently than is commonly
assumed in diachronic linguistics. The Language Making Capacity can, for example,
cope with a significant reduction of input.
Not only do these findings represent serious counterevidence to most
theories of creolization which invoke failure or break in transmission, but
also they all point to the fact that grammatical change of the type observed in
creoles requires the conjunction of at least two factors:
(i) a multilingual context that makes it possible for learners to entertain
competing learning hypotheses; and
(ii) a profile of bilingual and monolingual first learners that makes them more
successful in regularizing the varying PLD than late child and adult L2
learners.
Clearly, the socio-cultural, economic, and political context of plantation
colonies provides us with an extraordinary context where one finds not only
bilingual and monolingual L1 learners forming the cohort of early learners in
terms of Weerman (2011), but also child and adult L2 learners representing late
learners. Consistent with the contact model described in Section 4.4, this
implies that the agents of change in a creole setting possibly include:
‒ Monolingual L1 learners of the colonial vernacular
‒ Early bilingual L1 learners of the colonial vernacular and the emerging
creole (and/or an African language). This essentially involves the White
children, the Creoles, and the enslaved African children.
‒ Late L2 learners of the colonial vernacular (e.g., the newly arrived bozals)
‒ Late L2 learners of the emerging creole (both Whites and Black).
Given this description, it is reasonable to assume that the interaction between
these different types of learners with their specific I-languages continuously fed
into what would be identified later as a creole. Indeed, what is fascinating about
the development of creoles is that the context of their birth made it possible for
competing acquisition processes to arise simultaneously in relatively isolated
communities under significant demographic changes. The end result of such a
contact situation is that creole languages display robust properties of their
lexifiers alongside with robust substrate influence and internally generated
grammatical innovations. In this chapter, I address these properties with a
special focus on the role of transfer (or cross-linguistic influence) in the
development of these new languages.
5.1 Comparing the creoles with their source languages 175
The empirical domain of this study is the development of the determiner

phrase (DP) in Haitian Creole and in the Surinamese creoles Sranan and
Saramaccan. The chapter is structured as follows: Section 5.1 presents the
general properties of the DP in Haitian, Sranan, and Saramaccan, and contrasts
them with corresponding data from French, English, and Gungbe.1 The
analysis of the nominal structure shows that the creoles display a range of
variation in the nominal structures that cannot be explained only by structural
differences between their related lexifiers.
Instead, it will be shown in Section 5.2 that Haitian Creole, Sranan, and
Saramaccan often exhibit syntactic and semantic properties that combine those
of their substrate and lexifier languages. For instance, I will argue that even
though the Gbe languages acted as a trigger for the emergence of the feature
specificity in these creoles, only Haitian Creole adopted Gbe syntax to any
significant degree, while English seems to have provided the basis for the
syntax of the determiner in Sranan and Saramaccan. In addition, the
discussion of the Haitian data suggests that pattern transmission does not
entail perfect replication of the source system in the emerging language. In
this respect, I argue that transfer results from the vulnerability of interfaces in
general, and more specifically, that of the semantic–syntax interface within the
noun phrase. While compatible with the conclusion reached in Chapter 4, this
view clearly differs from exclusive theories of creole genesis in which the
substrate languages are thought to systematically provide syntactico-semantic
features, while the superstrate languages mainly contribute to the lexico-
phonetic forms, especially Lefebvre’s (1998) Relexification Hypothesis. In a
similar vein, the analysis developed here differs from alternative theories of
genesis such as Bickerton’s Bioprogram hypothesis, Chaudenson’s superstratist
approach, or Plag’s interlanguage hypothesis. Instead, the proposed analysis
presupposes that, in a situation of language contact, any competing language
may affect any module of the grammar of the emerging language, as explained by
Mufwene (2001, 2002, 2008). The main point is that agents of change are
learners, children and adults alike, as they make specific hypotheses about
linguistic paradigms in the input they are exposed to. What matters therefore is
the inputs which feed into the contact setting’s feature pool and how learners act
on them. Section 5.3 concludes the chapter.
5.1 Comparing the creoles with their source languages

In this section, I discuss the general properties of noun phrases in the creoles
as opposed to their lexifiers and substrate languages, based especially on
1
In this work, I will not discuss the influence of Portuguese on Saramaccan syntax; I hope to return
to this issue in future work.
Aboh (2004a, 2004c) and Aboh and DeGraff (2014). Of particular relevance
here are the properties of the determiners related to definiteness and
specificity.
By specificity, I refer to discourse-anaphoric noun phrases that link back
to referents that are pre-established in the discourse, or known or familiar to
the participants (Pesetsky 1987).2 On the other hand, definiteness refers to a
pre-identified set of referents that is not necessarily discourse-linked. These
informal definitions imply that specificity requires a smaller set of referents
than definiteness. For the sake of discussion, I tentatively submit that
specificity identifies a subset of the contextually given or assumed elements
for which a comment holds. Consistent with Kiss’s (1998) account of
exhaustive or identificational focus, this would mean that specificity
expresses exhaustive identification, unlike definiteness, which only selects a
contextually identified set.
As shown in Aboh (2004a, 2004c, 2006a) and Aboh and DeGraff (2014),
specificity and definiteness combine in some languages, such as Gungbe, in
ways that suggest the following characterization:
(i) A specific definite noun phrase is strongly D(iscourse)-linked and repre-
sents a unique referent assumed to be known to both speaker and hearer,
and which the speaker intends to refer to.
(ii) A specific indefinite noun phrase need not be D-linked. It represents an
existing referent that the hearer may not know about, but which the speaker
has in mind and intends to refer to.
More work is needed in order to reach a proper semantic formulation of the
description in (i) and (ii). For the purposes of this discussion, however, the
description of specific definites in (i) is compatible with the notion of
assertion of existence, as discussed in the literature (e.g., Bickerton
1981), while specific indefinites, described in (ii), recall Ionin’s (2006a)
notion of the speaker’s intent to refer. In the Gbe languages, these two
notions are properties of two different determiners, which exclude each other
and which I assume to be expressions of the nominal left periphery (Aboh
2004a, 2004c, forthcoming, and much related work). In these languages, noun
phrases are never ambiguous with respect to specificity because each kind has a
different marker. On the other hand, nonspecific noun phrases, whether
definite, indefinite, or generic, take no marking even though they may be
modified by an adjective, a demonstrative, a numeral, or a relative clause
(see Aboh and Essegbey 2010 for the discussion on Kwa languages in
general, and Aboh and DeGraff 2014 for a comparison of Haitian Creole and
2
This definition is compatible with Prince’s (1981) notion of assumed familiarity for topics and
suggests that topic and specificity interact (Aboh 2004b).
Gungbe). I refer to such noun phrases as “bare noun phrases.”3 Building on this,
I propose that specificity and definiteness are syntactic features anchored in
distinct heads within the nominal left periphery (Aboh 2004a, 2004c,
forthcoming; Szabolcsi 1994).
The following sections describe how the creoles (Haitian, Sranan,
Saramaccan) and their lexifiers encode definiteness, specificity, and number,
as well as the types of determiners used to express these notions. In so doing,
I pay particular attention to the contexts in which the determiners can occur
and the word order variations inside the noun phrase that they may produce.
I also consider the position of the noun relatively to other nominal modifiers
such as relative clauses, adjectives, demonstratives, and genitives. I will start
with Haitian Creole, Gungbe, and French.
5.1.1 Haitian Creole versus Gungbe and French

Studies of the properties and distribution of the determiners in these languages
suggest that determiner phrases (DPs) show strong parallels in Haitian
and Gungbe, as opposed to French.4 For instance, Haitian and Gungbe lack
indefinite or definite singular and plural articles of the type attested in Romance
languages. Instead, these languages include a postnominal marker that strongly
marks specificity in Gungbe, though less so in Haitian Creole (Aboh and
DeGraff 2014). Accordingly, this marker occurs with the noun phrase if and
only if it refers to some entity that is pre-identified in discourse and/or is
assumed by the speaker to be familiar to the addressee (i.e., specific and
definite). In addition, these languages have a separate postnominal number
marker. The number marker also expresses definiteness, though not always
(Aboh 2002, 2004a, 2004c, 2006a, forthcoming; Aboh and DeGraff 2014,
Lefebvre 1998). The following examples illustrate these properties.
The Haitian and Gungbe examples in (3a–b) involve the specificity
determiners, and the noun phrases are interpreted as referring to shared or
known information only. This restriction does not apply to the French example
(3c), where the determiner is ambiguous as indicated by sentence (3d). In this
work, I tentatively translate the meaning of such specificity-marked noun
3
I am therefore assuming, contra Enç (1991) and much related work, that definite noun phrases are
not necessarily specific. In fact, the Gbe languages provide numerous examples of definite noun
phrases that derive their definiteness from the context or from the modifiers that they are
associated with (e.g., relative clause, demonstrative, genitive). For a similar study of the noun
phrase in a creole, Gullah, see Mufwene (1986).
4
See also Brousseau and Lumsden (1992), Bruyn (1995a, 1995b), DeGraff (1992, 1995, 2000),
Déprez (2001), Lefebvre (1998), and Aboh and DeGraff (2014) for a discussion of Haitian
Creole; Aboh (2004a, 2004b) and Lefebvre and Brousseau (2002) for studies of Gbe languages;
and Bernstein (1997, 2001a, 2001b) and Giusti (1996, 1997) for the analysis of Romance and
Germanic languages.
phrases as “The N (in question).” The translations of and emphasis in the

Haitian examples are mine.
(3) a. pè-a [Haitian; Sylvain 1936: 55]
Priest-DET
‘The priest (in question)’
*‘The priest (e.g., in a generic sense)’
b. mɔ̀ pὲ lɔ́ [Gungbe]
Priest DET
‘The priest (in question)’
*‘The priest (e.g., in a generic sense)’
c. le prêtre [French]
‘the priest (e.g., that we know of)/The priest (in question)/The priest (in a
generic sense)’
d. Le prêtre représente l’Église.
the priest represents the.Church
‘The/this priest (i.e., the one of our community) represents the Church (as an
institution).’
‘The priest is the representative of the Church (as an institution).’
These examples indicate that the specificity marker need not occur with
definite or generic nouns in Haitian and Gungbe. Thus, these languages use
bare noun phrases in a wide range of contexts. Modern French, on the other hand,
is more restrictive and does not allow bare nouns in contexts such as those
depicted in (4a) and (4b). Accordingly, the French equivalent example (4c) is
ungrammatical. As one can see from the grammatical French example (4d),
sequences that are most comparable to Haitian and Gungbe bare nouns
necessarily involve conjoined plural nouns (Roodenburg 2004). Such a
constraint does not apply to Gungbe and Haitian bare noun phrases. Note also
that the French coordinated bare nouns do not have the range of meanings (e.g.,
generic) that may be associated with bare nouns in Haitian and Gungbe.
(4) a. Wosiyòl manje kowosòl. [Haitian; DeGraff 1992: 105]
b. Àtòké nɔ̀ ɖù kòròsɔ́ lù. [Gungbe]
bat HAB eat soursop
‘(A) bat(s) habitually eat(s) soursop.’
c. *Chauve-souris mange corossol [French]
d. Parlementaires et politiques de tous bords ont approuvé les mesures d’aide à
l’emploi.
‘Parliamentarians and politicians of all stripes approved the employment
assistance measures.’
As already explained in Aboh and DeGraff (forthcoming), the distribution of

bare nouns in Haitian Creole and in Gungbe can be said to be comparable to that
of bare nouns in Old French, where articles were optional thus allowing the
occurrence of bare nouns in a wide range of positions with varying semantics
(see Mathieu 2009 and references cited therein for discussion). Yet, Haitian
Creole and Gungbe differ from Old French and Modern French in various
ways: First, as noted above, these languages do not display definite articles.
Second, the specificity markers (which some authors regard as definite
determiners) are not optional but necessarily occur with specific-definite
noun phrases (unlike in Old French). Modern French, on the other hand,
requires these determiners in all contexts, except in coordinated plurals
(Roodenburg 2004). Third, the Gungbe and Haitian noun specificity markers
as well as other determiner-like elements such as pronominal possessors and
demonstratives cluster in postnominal position (5a–b). Not all speakers of
Haitian Creole accept Haitian equivalents of Gungbe constructions such as
(5b), in which the number and specificity markers co-occur. Recall, however,
examples (6) and (7) in Chapter 3, repeated here as (5c–e) for convenience,
which indicate that such examples existed in Early Haitian of the eighteenth
century.
(5) a. krab mwen sa a yo [Haitian; Lefebvre 1998: 78]
b. àgásá cè éhè lɔ́ lέ [Gungbe]
crab 1SG-POSS DEM DET PL
‘these crabs of mine’
c. sé mésié-là yo [Haitian Creole of 18th c.; Chaudenson 2003: 277]
DEM man-DET PL
‘these men’
d. Zozo la yo va bientôt volé. [Ducœurjoly 1803: 336]
bird DET PL will soon fly
‘These/the birds will soon fly away.’
e. Mo voir zhomme la yo.
1SG see man DET PL
‘I saw the/these men.’
Furthermore Glaude (2012: 244), a native speaker of Haitian, reports
example (5f) below which suggests that in the context of the number marker,
the specificity marker is optional. These data might be evidence that there is
inter-idiolectal or dialectal variation regarding the co-occurrence of the
specificity and number markers. Setting this variation aside, it appears that
Gungbe and Haitian Creole exhibit the sequence in (5g). Finally, note from
these examples that none of these languages manifests number inflection or
gender specification on the noun. The latter always surfaces in its bare form.
(5) f. M wè chen (an) yo. [Haitian; Glaude 2012: 244]

1SG see dog DET PL
‘I saw the dogs.’
g. N > DOSS > DEM > DET > PL
Returning to examples (5a–b), a contrast that seems to emerge between
Haitian and Gungbe is that the latter, unlike the former, involves a distinct
class of weak possessive pronouns. Indeed, with the exception of first-person
singular (cè ‘my’) and second-person singular (tòwè ‘your’), Gungbe possessive
pronouns derive from a combination of weak personal pronouns plus the genitive
marker tɔ̀ n as in é-tɔ̀ n (3sg+Poss = his), mí-tɔ̀ n (1pl + Poss = ‘our’); mì-tɔ̀ n (2pl
+ Poss = ‘your’); yé-tɔ̀ n (3pl + Poss = ‘their’) (Aboh 2002, 2004a). This set of
weak pronouns must be distinguished from the set of strong pronouns, which
uniformly include the combination of a strong pronoun and the genitive marker:
nyὲ-tɔ̀ n ‘my,’ jὲ-tɔ̀ n ‘your,’ úɔ̀ -tɔ̀ n ‘his/her,’ mílέ-tɔ̀ n ‘our,’ mìlέ-tɔ̀ n ‘your,’ yélέ-
tɔ̀ n ‘their’ (Aboh 2004a). Haitian does not display such combinations in its
possessive pronouns. Instead, Haitian exhibits a [possessee]–[possessor]
juxtaposition. Here, it could be argued that, in sequences such as krab mwen in
(5a), the possessive marker is null. As Lefebvre (1998) argues, this would mean
that the word order and the form of the pronoun are parallel in Haitian and
Gungbe, the only difference being that Haitian lacks an overt possessive marker.
This analysis is compatible with the sequencing in (5d) where the head noun
precedes a cluster of determiner-like elements including the number and
specificity markers and possessive pronouns.
All these facts make Gungbe and Haitian Creole different from French. Indeed,
French determiners and modifiers such as demonstratives and possessive
pronouns occur prenominally and bear gender and number inflection (6a–c;
6a’–c’). Furthermore, demonstratives, possessive pronouns and definite
determiners are mutually exclusive (6d), unlike in Gungbe and Haitian.
(6) a. le(s) crabe(s) a’. la/les table/s [French]
‘the crab(s)’ ‘the table(s)’
b. ce(s) crabe(s) b’. cette/ces table/s
‘this/these crab(s)’ ‘this/these table/s’
c. mon/mes crabe/s c’. ma/mes table/s
‘my crab(s)’ ‘my table/s’
d. *le(s) ce(s) mes crabs
‘the/this/these my crab(s)’
Finally, Haitian and Gungbe display bare nouns that may also function as
head nouns in relative clauses (Aboh and DeGraff 2014, forthcoming):
(7) a. [Moun] ki pa travay p ap touché. [Haitian; DeGraff, p.c.]
people REL NEG work NEG FUT get-paid
‘Those who don’t work won’t get paid.’
b. [Mὲ] ɖê má wà àzɔ́ n màá yí kwέ. [Gungbe]

One REL NEG do work NEG.FUT receive money
‘Anyone who doesn’t work will not get money.’
When these nouns co-occur with the specificity marker, the head noun
precedes the relative clause, which in turn precedes the specificity marker, as
depicted in (8a) and further instantiated in (8b–d):5
(8) a. head noun > [relative clause] > specificity marker > number marker
b. [Fam ki vini wè-u] la. [Haitian; Sylvain 1936: 69]
woman REL come see-2PL DET
‘The woman (in question) who came to visit you.’
c. [Fam ki rete] yo. [Haitian; Sylvain 1936: 62]
woman REL arrest 3PL/PL
‘The women who arrested them.’
or ‘The women who were arrested.’
d. [Náwè ɖê wá kpɔ́ n mì] lɔ́ lέ. [Gungbe]
woman REL come see 2PL DET PL
‘The women (in question) who came to visit you.’
French, on the other hand, does not allow bare nouns in relative clauses,
whence the ungrammatical example (9a), which contrasts with the grammatical
example (9b). As sequenced in (9c), the French definite determiner must
precede the head noun, which in turn precedes the relative clause.
(9) a *Homme que j’ai vu. [French]
man that I.have seen
b. L’homme que j’ai vu.
DET.man that I.have seen
‘The man that I saw.’
c. Determiner > noun > [relative clause]
It appears from this description that Haitian and Gungbe pattern alike in a
way that is different from French. While these similarities point to a close
structural relationship between Haitian and Gungbe, they should not obscure
the fact that Haitian Creole and French do share a number of properties, some
of which distinguish them from Gungbe (and related languages). For example,
Haitian exhibits prenominal and postnominal adjectives, like French (see
Chapter 3). Gungbe, however, has postnominal adjectives only.
5
What matters here is the position of the relative clause vis-à-vis the head noun and the
determiners. The interested reader is referred to Koopman (1982a, 1982b), DeGraff (1992),
Lefebvre (1998), and Aboh (2002, 2005) for the analysis of relative clauses in Haitian and
Gungbe, respectively.
These similarities between Haitian Creole and French as opposed to Gungbe

carry over to the semantic properties of the determiners referred to here as
specificity markers. While these elements occur postnominally in both
languages, their semantic properties point to differences that are discussed in
Aboh and DeGraff (2014) and which I briefly summarize here.
With regard to the Gungbe specificity maker lɔ́ Aboh and DeGraff (2014)
show that it can occur in contexts where the noun phrase is interpreted as
specific. The following two constructed scenarios illustrate this.
Context: The Queen of Holland went to the G20 meeting, but when she set
up to give her speech, her prime minister prevented her from talking (arguing it
was his job not hers).
(10) Òlándù sín àxɔ́ sì wá G20 àmɔ́ n ɖě àxɔ̀ sì lɔ́ jró ná
Holland POSS queen came G20 but when queen DET want prep
ɖɔ̀ xó prèmié mìnìs gbέ.
say word prime minister refuse
‘The Queen of Holland went to G20, but when the Queen (in question/this
Queen) wanted to give a speech, the prime minister refused to let her talk.’
‘La Reine des Pays-Bas est allée au G20, mais lorsque la Reine a voulu
prononcer son discours, le premier ministre a refusé de la laisser parler.’
In this example, “the Queen of Holland” occurs upon first mention in

Gungbe as a bare noun phrase even though it denotes a familiar and unique
referent. This is unlike the English and French translations where such a bare
noun (i.e., denoting unique reference) is not allowed and both ‘queen’ and
‘prime minister’ must be realized with the definite determiner. In the follow-up
sentence, however, àxɔ́ sì ‘queen’ occurs with the post-nominal determiner lɔ́
and the sequence àxɔ́ sì lɔ́ can be understood as ‘the Queen in question’ or ‘this
(very) Queen.’ This is compatible with the observation that Gungbe lɔ́ occurs in
contexts were it marks specificity only.
This observation does not carry over to all the relevant contexts where one
finds Haitian Creole la. For instance, Aboh and DeGraff (2014) show that these
two markers behave differently when it comes to attributive uses of the type
presented in (11) adapted from Heim (1991). Here, Haitian la displays
attributive uses that are similar to those found with the French definite
articles le, la, les. Accordingly, French and Haitian Creole determiners allow
attributive uses that Gungbe lɔ́ disallows.
(11) a. Jounalis la vle kestyonnen candida ki te genyen *(an).
reporter DET want question candidate who TNS win DET
‘The reporter wants to question the candidate who won (i.e., Barack Obama).’
‘The reporter wants to question whichever candidate won (i.e., whoever
that is).’
b Jùnàlís lέ jró ná kàn xó bíɔ́ mὲ xè àxɔ́ ví lέ yí

reporter PL want to inquire word ask person who prince PL take
‘The reporters wanted to question the person who the princes elected
(whoever that is).’
c. Jùnàlís lέ jró ná kàn xó bíɔ́ mὲ xè àxɔ́ ví lέ.
reporter PL want to inquire word ask person who people PL
yí lɔ́ .
take DET
‘The reporters wanted to question that specific person who the princes
elected.’
The Gungbe example in (11b) would be used in a situation where it is
traditionally the case that reporters always want to interview the elected king.
Sentence (11c) on the other hand could be uttered in a context where both the
speaker and the hearer know exactly who the newly elected king is, but did not
want to say his name. What these examples show is that the Haitian Creole la
may occur in contexts where it is nonspecific similarly to French articles.
Gungbe lɔ́ cannot be used in such contexts. This contrast indicates that while
the specificity marker is regulated by discourse factors only in Gungbe (11a),
the distribution of Haitian Creole la is more complex because this marker is
sensitive to both discourse context and purely grammatical requirements such
as in (11b–c). French, on the other hand, seems to be a much simpler case, given
that the determiner is always required.
Another difference between Gungbe lɔ́ and Haitian la is that the former can
co-occur with proper names as illustrated by the following dialogue (12):
(12) A: Aimé Césaire ɖɔ̀ xó kpó Sarkozi kpó.
Aimé Césaire say word COORD Sarkozy COORD
‘Aimé Césaire talked with Sarkozy.’
B: Bé mì kà sè ɖɔ̀ Aimé Césaire lɔ́ kú?
Q 1PL at.least hear that Aimé Césaire DET die
‘Have you been informed that Aimé Césaire died?’
This marker can also occur with personal strong pronouns.
(13) Mì lέ lɔ́ núdònù díndín ná hù mì.
2PL PL DET underneath.of.thing searching FUT kill 2PL
‘As for you (all), curiosity will kill you.’
Haitian la, on the other hand, cannot be used with proper names. Compare
the following Haitian Creole examples to the Gungbe ones in (14):
(14) Aimé Césaire pale ak Sakozi. Men, mwen tande
Aimé Césaire speak with Sarkozy but 1SG hear
Aimé Césaire (*la) mouri.
Aimé Césaire DET die
‘Aimé Césaire spoke with Sarkozy, but I just heard that the Aimé Césaire in
question died.’
Similarly, la cannot co-occur with personal pronouns.
(15) *Yo-menm la, yo te chape nan goudougoudou a.
3PL-EMPH DET, 3sg TNS escape in earthquake DET
‘As for them, they escaped from the earthquake.’
Examples (11–15) indicate that the Haitian marker la displays mixed
properties, some of which make it pattern like Gungbe (Gbe), while others
make it similar to French (Romance). In this regard, the discussion in previous
paragraphs suggests that French does not involve a proper specificity marker.
One might conclude from this that this language does not encode specificity
syntactically. Such a conclusion would be misguided, since French makes use
of demonstrative articles (e.g., ce, cette), in combination with the postnominal
deictic locative là, to encode specificity, as shown in (16a). The ungrammatical
example (16b) shows that the postnominal deictic element cannot occur in
postnominal position by itself, unlike in Gungbe and Haitian.
(16) a. Cette guitare-là me plaît beaucoup. [French]
that guitar there me pleases very much
‘I like that guitar (there) very much.’
b. *Guitare-là me plait beaucoup.
guitar-there me pleases very much
Based on examples similar to (16a), it has been argued that the French deictic
locative là is obviously the source of the Haitian specificity marker la, therefore the
lexifier itself contributed to reinforcing the Gbe pattern in Haitian (see also
Lefebvre 1998 and references cited therein). I will not elaborate on the possible
reinforcing role of French here, however, and the interested reader is referred to
Mufwene (2001), Siegel (2004), and Ansaldo (2005, 2009) for some ideas. Instead,
following the approach developed in Chapter 4, I propose that part of the nominal
structure in Haitian has its source in both Gbe and French with both language types
contributing to the emergence of different aspects of the morphosyntax of the noun
phrase in Haitian. Both Gbe and French provided the creators of Haitian Creole
with the relevant triggers for setting the parameters of the DP.
Anticipating the discussion in Section 5.2, the analysis proposed here
follows Szabolcsi (1987, 1994) and much related work in assuming parallels
between the noun phrase and the sentence. It is argued that the structural make-
up of the noun phrase includes three layers, starting with the core predicate
layer where the (lexical) head noun merges and introduces its arguments (e.g.,
in possessive constructions). This layer further extends to a functional layer,
which consists of distinct functional projections responsible for agreement
features and whose specifiers host noun modifiers such as a numeral and an
adjective. In Cinque’s approach (1994, 1996, 1999, 2004, 2010), this would
mean that nominal modifiers are to the noun phrase what adverbs are to the verb
phrase. Keeping the parallel between the clause and the noun phrase, I refer to
this layer as the nominal inflectional domain, whose head, the nominal
counterpart of clausal I, encodes deixis and may host definite articles
(Szabolcsi 1994).
The nominal inflectional domain is projected under a nominal left periphery
headed by D, which Szabolcsi shows to have properties of a subordinator and to
parallel the clausal left periphery C. Under Rizzi’s (1997) split-C hypothesis,
the clausal periphery involves discrete functional projections responsible for
clause-typing, topicalization, focalization, and finiteness. Adopting this view,
Aboh (2002, 2004a, 2004b, 2004c, 2005c, forthcoming) shows, on the basis of
the Gbe languages, that the nominal left periphery can be split similarly. Aboh
(2004c, forthcoming) further argues that the Gungbe specificity marker
encodes a topic projection (TopP) that projects within the nominal left
periphery. As such, the element lɔ́ links the noun phrase to the discourse. The
number marker lέ , on the other hand expresses a lower projection NumP that
links the nominal left periphery to the lower predicate in a way comparable to
finiteness in the clause. According to Aboh (2004c: 7), the noun phrase in (17a)
can therefore be represented as in (17b), where the nominal left periphery
involves the projections within DP to the left and NumP to the right (see also
Aboh forthcoming for discussion).
(17) a. Távò lɔ́ lέ́
table DET PL
‘the tables (in question)’
b. [DP [D [TopP [FP távò] [Top° lɔ́ [NumP ttávò [Num lέ [FP ttávò]]]]]]] [Aboh 2004c: 7]
I return to the details of this analysis in Section 5.2, where I show that the
variations within Haitian Creole nominal phrase and across Haitian Creole and
the Surinamese creoles can be accounted for by looking more closely into the
nominal left periphery and the properties of this periphery at the discourse–
syntax (or semantics–syntax) interface. But before getting onto this, let us take
a step back and consider the parallels we have observed between Haitian and
Gungbe. What we seem to find here is that the Haitian specificity determiner la
and the number marker yo express the nominal periphery similarly to their
Gungbe counterparts lɔ́ and lέ (Aboh 2004a, 2004c, 2006a).6 According to the
present analysis, such a parallel is regarded as an instance of pattern
transmission, in which a portion of the Gbe-like nominal left periphery has
been transferred to the creole (as explained in Section 5.2.).
6
But see Déprez, Sleeman, and Guella (2011) for a UG-based account of the emergence of
specificity encoding in language acquisition.
The data further suggest that Haitian Creole shows a split in its noun phrase:
the left periphery has a structure from the Gbe languages while the nominal
inflectional structure derives from French syntax. For instance, recall from the
discussion in Chapter 3 that French and Haitian both have prenominal and
postnominal adjectives, which I assume, following Cinque (1994, 1999, 2010)
and much related work, realize the nominal inflectional domain. I submit that
such a hybrid system, involving a left periphery that has been partially influenced
by Gbe and an inflectional domain that has been influenced by French, can be
accounted for straightforwardly only if one adopts the competition and selection
approach elaborated here. In this regard, it is important to note that while the Gbe
languages typically display a limited set of adjectives denoting color, shape, and
size, most adjectival elements are derived from verbs (Aboh 2007b, Aboh and
Smith 2012). This is different from French where different types of adjectives
occupy different positions within the noun phrase (Cinque 2010). The
recombination in Haitian appears to have reproduced the contrast between
French and Gbe. Thus, the properties of the creole are largely to be found in
the inputs that the learner is exposed to; they are just recombined in new ways.
The inputs generated by these languages feed into the learners’ I-creoles, the
convergence of which leads to the E-creole. Further empirical evidence
supporting this view comes from Sranan and Saramaccan, to which I now turn.
5.1.2 Sranan (and Saramaccan) versus Gungbe and English

The facts described for Haitian, Gungbe, and French have parallels in Sranan,
Saramaccan, and English, with the difference that the Surinamese creoles use
specificity determiners in the Haitian and Gungbe way, even though the English
syntax is evident. Data concerning the Surinamese creoles are mainly drawn
from Sranan, but the conclusion reached here also applies to Saramaccan (and
presumably to other Surinamese creoles as well).
The Sranan sentences in (18) indicate that these creoles freely use indefinite,
definite, and generic bare noun phrases. Note in the following examples that the
bare noun bana ‘banana’ is interpreted as generic (18a) or definite, as in the
second instance of bana in (18b). However, the sequence a bana in (18b) is
read as specific.
(18) a. Kofi, go na wowoyo go bai bana tya kon gi mi. [Sranan]
Kofi go LOC market go buy banana carry come give 1SG
‘Kofi, go to the market to buy me bananas.’
b. Na a bana di Ppa tya kon, dati a njan.
COP DET banana REL father carry come that 3SG eat
A bere hati, a njan bana.
3SG stomach hurt 3SG eat banana
‘The banana (in question) that father brought is what he ate. His stomach is
aching because he ate that banana.’
As the sentences in (19) show, the corresponding English examples
involving bare nouns are ungrammatical.
(19) a *John went to the market to buy banana.
b. *Banana that Daddy bought yesterday.
Like Haitian and Gungbe, Sranan has a singular specificity marker that is
realized as (n)a. Note in sentences (18b) and (20a–b) that the presence of this
element to the left of the noun triggers the specific reading. In example (20a),
the first instance of bana is interpreted as specific (i.e., ‘the banana in question’)
as opposed to bana in (20b), which is understood as (in)definite.
(20) a. Kofi, teki a bana tya gi mi. [Sranan]
‘Kofi, give me the banana in question [e.g., the one I brought yesterday].’
b. Kofi, teki bana tya gi mi.
Kofi take banana carry give 1SG
‘Kofi, give me a (or the) banana (s).’
Again, English does not make such a specific-versus-nonspecific distinction
between bare nouns and nouns that occur with the determiner.
(21) a. I like trains very much. [Generic]
b. Every day, I take the train to Amsterdam. [Definite, nonspecific]
In addition, specificity may be encoded in English by various elements
including (in)definite articles, demonstratives, and quantifiers. While Sranan
parallels Haitian and Gungbe in the way it encodes specificity, it differs from
them in lacking a distinct number marker. Hence, the plural counterpart of
example (20a) involves the plural specificity marker, as in (22).
(22) Kofi, teki den bana tya gi mi. [Sranan]
‘Kofi, give me the bananas (in question) [e.g., the ones I brought yesterday].’
Therefore, while Gungbe and Haitian Creole lexicalize specificity and number
by means of two separate and independent morphemes, Sranan and Saramaccan
amalgamate the two features into one morpheme. These variations point to a
possible structural difference between the DP in Sranan, on the one hand, and in
Haitian and Gungbe, on the other. That this is probably the correct description is
further suggested by the fact that the specificity determiner precedes the noun in
Sranan, but follows it in Gungbe and Haitian. This pattern is replicated in relative
clauses, where the specificity determiner precedes the relativized noun, which in
turn precedes the relative clause in Sranan (18b). In Gungbe and Haitian,
however, the relative clause is sandwiched between the head noun and the
determiners, as already illustrated in (7–8). Given that the sequencing of the
noun phrase in Sranan and Saramaccan parallels that of English, I will argue in
what follows that the morphosyntax of the noun phrase in these creoles bears a
strong legacy of English.
5.1.3 Summary
It appears from the above description that Haitian and Sranan use determiners
(and number markers) in a way parallel to Gungbe and other Gbe languages.
These languages mainly distinguish between specific and nonspecific referents,
thus allowing bare nouns to occur in all possible argument positions. In
addition, such bare noun phrases show no obvious distributive differences
from noun phrases marked by the specificity or the number marker (Aboh
and DeGraff 2014). We can therefore conclude that the distribution of
determiner-like elements is regulated by discourse properties in Gbe and the
creole languages. In French and English, on the other hand, the definite and
indefinite articles are governed by purely syntactic constraints. While English
allows bare mass nouns as well as bare plural count nouns, French excludes
bare nouns except in conjoined plural noun phrases. Both Old French (Mathieu
2009) and Old English allow bare nouns, but again the distribution of such bare
nouns differs from that found in the creoles and the Gbe languages. It appears
from this description that the morphosyntax of the noun phrase in Haitian,
Saramaccan, and Sranan is different from that of the noun phrase in the source
languages. Table 5.1 recapitulates the possible feature combinations and
expressions of determiners in all five languages considered:
As the last three rows of this table show, bare nouns occur in Sranan, Haitian,
and in Gungbe whenever the noun phrase has the following feature combinations:
Table 5.1 Feature combinations and determiner expression in creoles and

their source languages
D-features Gungbe Sranan Haitian English French
[+spec +def, +plur] lɔ́ , lέ den la yo the les,

ces . . . ci
[+spec, +def, –plur] lɔ́ na la the le, la
[+spec, –def, +plur] ɖé, lέ Ø (wan tu) – some, certain certains
[+spec, –def, –plur] ɖé wan yon a, some, certain certain
[–spec, +def, +plur] lέ den(?),Ø yo the les
[–spec, +def, –plur] Ø [definite] Ø [definite] Ø [definite] the le, la
[–spec, –def, +plur] Ø [generic] Ø [generic] Ø [generic] Ø, any, [generic] des
[–spec, –def, –plur] Ø [indef] Ø [indef] Ø [indef] a, some, [generic] un
[−specific, +definite, −plural], [−specific, −definite, +plural], or

[−specific, –definite, −plural]. Such bare nouns can thus be interpreted as
definite, indefinite, generic, plural or singular. This is not the case in English and
French, where, when allowed, comparable bare nouns appear to denote an
existential, generic, abstract, or kind reading. Another difference between the
creoles and their source languages, on the other, is that the latter allow for the
morphological expression of plural indefinite specifics as combinations of
the features [+specific, –definite, +plural], while the former lack such
determiners all together. This would mean that with regard to the
morphological expression of the feature combinations in Table 5.1, the creoles
do not express one cell: an instance of irregularity that the learner is faced with.
Indeed noun phrases that can be interpreted as [+specific, −definite,
+plural] have the same morphological shape as nonspecific noun phrases
(viz., [−specific, +definite, −plural], [−specific, −definite, +plural],
or [−specific, −definite, −plural]); they all occur as bare nouns.7
Other differences that we observe between the creoles and their source
languages relate to the distribution of the determiners and the noun phrase. It
appears that the creoles do not always show syntactic patterns that exactly
coincide with those of the languages from which the function was retained. For
instance, Haitian has postnominal determiners like Gungbe but pre- and post-
nominal adjectives like French. Sranan, on the other hand, is just like English,
because it involves prenominal determiners and noun modifiers. This would mean
that Haitian combines properties of the Gbe languages and French in both
semantics and syntax, while Sranan seems to have mapped semantic features
from Gbe onto syntactic patterns from English. These combinations are illustrated
in Table 5.2. Here the sign “+” indicates that the property is present in the relevant
language, while “–” indicates its absence. “%” on the other hand indicates inter-
idiolectal variation, while “?” means that the feature could not be verified.
As is clear from these tables, the creoles are not isomorphic to their source
languages. While Table 5.1 can be regarded as listing the semantic properties of
the syntactic features involved in the nominal left periphery, Table 5.2
recapitulates surface orderings that are contingent to the formal licensing of
these syntactic features. Put together, they suggest that syntactic and their related
semantic patterns are not always uniformly transmitted from a single donor
language, superstrate or substrate, to the emerging languages. Instead, in a
situation of language contact where typologically different languages enter a
competition that may lead to the emergence of a new language, the latter
7
This sort of irregularity is often ignored by those who claim that creoles have simple structures. If
we consider learnability issues, the fact that the creoles assign the same form to specific indefinite
plurals and nonspecific noun phrases not only leads to ambiguity and therefore possible acquisi-
tion difficulties but also implies that the nominal module must have been completely reorganized
to allow different mechanisms of licensing and interpretation for these two types of noun phrases.
Table 5.2 Function and distribution of determiners in creoles and their donor
languages (HC: Haitian Creole; SC: Saramaccan; SR: Sranan; Fr: French;
En: English; Gu: Gungbe)
Creoles Substrate Lexifier
Determiners HC SC SR Gu Fr En
Free use of bare generic NP + + + + – –

Free use of bare NP specific (in) + + + – – –
definite (singular/plural)
Marker of specific definite (not + + + + – –
sensitive to count/mass
distinction)
Marker of specific indefinite – – – + – –
(not sensitive to count/mass
distinction)
Numeral one marks specific + + + – + +
indefinite
Articles mark (in)definite – – – – + +
(sensitive to count/mass
distinction)
Articles and demonstratives – – – – + +
mark specific (in)definites
Marker of specific definite – + + – + +
fused with number: plural
Separate plural marker + – – + – –
(encodes definiteness and is
sensitive to count/mass
distinction)
Specificity marker (singular) + + + – + +
before N
Specificity marker (plural) – + + – + +
before N
N before specificity determiner + – – + – –
(singular)
N before specificity determiner – – – – – –
(plural)
N before number marker + – – + – –
N before specificity marker % – – + – –
preceding number marker
Relative clause before + – – + + –
specificity determiner
singular
Relative clause before number + – – + – –
marker
Relative clause before ? – – + – –
specificity determiner
preceding number marker
5.2 On the vulnerability of interfaces 191
Table 5.2 (cont.)
Creoles Substrate Lexifier
Determiners HC SC SR Gu Fr En
Articles including specificity – + + – + +

determiners (singular/plural)
before relative clause
Adjective before N + + + – + +
N before adjective + + ? + + –
Numeral before N + + + – + +
Numeral after N – – – + – –
Demonstrative before N – + + – + +
Demonstrative after N + + + + – –
Possessive before N – + + – + +
Possessive after N + – – + – –
combines different syntactic and semantic properties selected from the

competing languages in a non-trivial way. More specifically, I argue, consistent
with Chapter 4, that the new languages emerged from feature recombination,
matching syntactic and semantic specifications from the competing languages.
This recombination is possible because, while semantic features of functional
items are comparable across languages (e.g., the expression of specific definite),
their licensing varies cross-linguistically. Thus, in a situation of language contact,
several varying syntactic strategies may compete for the licensing of one
semantic function (i.e., as anchored in a functional item). This means in
principle that the core syntactic properties of the new language and those of its
donors will not match perfectly (as shown by Tables 5.1 and 5.2). I further submit
that the selection of certain semantic and syntactic features can be facilitated by
interface conditions, namely those interfaces that are more vulnerable to
language transfer. Let us now consider the role of the syntax–semantics
interface in language contact on the basis of the described facts in Haitian,
Sranan, Gungbe, French, and English.
5.2 On the vulnerability of interfaces

In order to account for the similarities and variations between Sranan,
Saramaccan, Haitian, and their donor languages, I hypothesize that the way
the determiner system is affected in a situation of language contact depends
on whether the relevant features, their function, and their morphosyntax are
selected from a competing language or on whether just the feature and its
function are selected, leaving it up to the emerging language to develop the
relevant licensing mechanisms, under pressure from other competing
languages, from the interactions of the emergent subsystems, or on the basis
of UG principles. This line of thinking suggests that UG universally determines

the set of possible linguistic features, their semantics, and their related
parameters. These features, however, need appropriate triggers to emerge in a
language. I submit that competing languages provide the learner developing
an I-language with such triggers.
In this book, I argue that Haitian and Sranan show signs of substrate transfer,
because the function of the Haitian and Sranan determiners as the specificity
markers are adapted from the Gbe languages, as a result of D-feature
transmission. However, Haitian and Sranan D-elements require different
formal licensing conditions, which account for why Haitian patterns like Gbe
and partly like French, while Sranan behaves like English and partly like Gbe.8
On the assumption that D-features (and their related parameters) are properties
of UG (Abney 1987; Szabolcsi 1987, 1994; Aboh 2004a, 2004c), I conclude
that the Gbe languages constituted the relevant triggers for fixing the
parameters of the target D-features in Haitian, while English played a similar
role in the case of Sranan. This view accounts for the cross-creole variation in a
straightforward manner. In addition, it presupposes that UG acts as the ultimate
filter for the relevant combinatory possibilities, and only those that are
UG-compatible can converge. For instance, cross-categorial combinations,
say, between V and D, are filtered out by UG. Similarly, if one adopts
Cinque’s (1999) fixed universal functional hierarchy, then an unconstrained
recombination of the features and morphosyntax of different functional
projections will be excluded by UG as being non-converging. Because
competition and selection is anchored in the process of language acquisition,
the recombination of syntactic and semantic features that I argue for here is
constrained by UG similarly to the way UG constrains the range of humanly
possible languages. Under this restrictive view, for instance, no emerging
language would be able to develop an aspectual system in which the specifier
and the head of an aspect phrase are simultaneously or optionally filled by the
corresponding adverb and marker from both the lexifier and the substrate. I
further assume that feature competition is subject to a principle of economy
that guarantees that two identical features cannot be selected in the emerging
language. Accordingly, economy dictates that a creole cannot have two
specificity markers, one derived from the substrate and one from the
superstrate, to fulfill exactly the same syntactic and discourse functions.
The proposed analysis for language transfer therefore implies interaction
between two levels: (i) the retention of syntactic features (including their
associated semantic features); and (ii) the formal requirement that UG sets
for these features to be properly licensed.
8
According to this approach, it is conceivable that Portuguese may have also played a role in the
development of prenominal left peripheral markers in Saramaccan, but I will not discuss this
possibility here.
5.2.1 Feature selection and the syntax–discourse interface

One question that arises at this stage of the discussion is what features can be
retained and why. In other words, which features are contact-sensitive and why
may such features be selected in a situation of language contact?
Given our current limited knowledge of intralinguistic and extralinguistic
factors that interact in a situation of language contact, there seems to be no
obvious answer to the question of why a feature (F1) should be selected over
another feature (F2).9 These complications aside, I suggested in Chapter 4 that
it is reasonable to assume that the features that are most likely to be selected are
the ones that are associated with interface properties. In the context of this
chapter, this includes discourse/semantics features that relate to interfaces:
Discourse-interpretable features which enter the competition are more likely
to be selected in the emerging language. As a consequence, certain domains of
the Faculty of Language (e.g., the syntax-discourse interface) may be more
sensitive to language contact phenomena than others.
This observation, in turn, leads me to conclude that certain aspects of core
syntax (e.g., the order of merge or the predicate structure) might be immune (or
less sensitive) to language transfer or, say, syntactic recombination as explored
here. For example, under Kayne’s (1994) antisymmetry and Sportiche’s (1988)
VP-internal subject hypothesis, the rigid SVO or head–complement pattern
found across creoles appears to be a consequence of the merging order imposed
by UG. Heads merge first with their complements, and the set formed in that
way merges with the specifier. Similarly, if we adopt Cinque’s (1999) universal
hierarchy of the clause structure based on typological findings over the past
decades (e.g., Foley and van Valin 1984; Baker 1985; Bybee 1985; Hengeveld
1989; see also Chapter 3), then the mood–tense–mood–aspect–verb sequence
found in creoles as well as other non-creole languages reduces to a simple
expression of UG.10 The only difference between creoles and superficially
different languages (e.g., Kwa, Romance, Germanic) resides in whether a
language expresses the head position (either with a free morpheme or an
affix) or the specifier position of a particular functional projection within the
clause structure. According to the position advocated here, competing
languages provide the trigger to set such parameters in the emerging language.
Assuming that UG acts as such a filter in language genesis, we may be able to
approach the question of inflection in a completely different way. It is often
argued in the literature that inflectional morphology (e.g., verbal inflection and
9
Social factors, such as prestige, could play a role in favoring one particular feature over another
one, but very little is known about the role of such factors in language contact and language
change in general. In addition, as I argued in Chapter 4, these factors do not directly constrain
the licensing of syntactic features.
10
See Aboh (2006b) for a discussion of mood sequences in Saramaccan and certain other English-
based creoles.
noun inflection) is typically lost in language contact situations, mainly due to

imperfect second language acquisition. Extended to creoles, this has been taken
to mean that these languages lack inflectional morphology due to imperfect
second language acquisition, itself conditioned by restricted access to the target
language (see Chapters 3, 4, and Aboh and DeGraff, forthcoming for a
critique). While various authors ranging from Bickerton (1988a, 1999) to
McWhorter (2001) consider this a diagnostic of the pidgin ancestry of creole
languages in general, others such as Plag (2008a, 2008b, 2009a, 2009b) regard
it as strong evidence for the interlanguage profile of creoles. As I remarked in
Chapter 3, such accounts fail to address the subtle properties of creoles some of
which involve inflectional morphology. Also, they only make sense if one
assumes an exceptional scenario according to which a whole population is
struck by Alberto’s syndrome or systematically fails to acquire aspects of the
target to such an extent that their offspring found themselves in a linguistic
vacuum and had to create a brand new language that literally spells UG out.
However, suppose we were to adhere to such exceptionalist views, what
factors can we invoke to explain why a significant number of the African
learners on the plantation fail to acquire the relevant inflectional paradigm
regardless of their socio-economic position in the community? How can we
account for such a failure in a principled way? And how can we accommodate
both the absence of verbal inflection and the presence of derivational
morphology as well as agreement phenomena in the nominal domain of certain
creoles (e.g., Haitian, Saramaccan) within such an exceptionalist approach?
More specifically, how can we account for the fact that, even though Haitian
lacks verbal endings of the French type, the language does have nominal endings
of the type -(s)yon in words such as dekoupasyon ‘dividing a wall,’ vivasyon
‘conviviality,’ as well as other nominal affixes discussed by DeGraff (2002)?
And how can we account for case distinctions in Saramaccan (Aboh and Ansaldo
2007) and number agreement in the same language as discussed in Chapter 3?
There seems to be no principled way of accounting for such facts by simply
invoking second language acquisition failure.
In the analysis developed in this book, domains of grammar show different
sensitivity to linguistic recombination and language transfer, depending on
whether or not they relate to the syntax–semantics interface. Accordingly, the
systematic loss of inflections in a language contact situation would follow in a
straightforward manner if we assume that inflectional morphology (including
formal agreement in general) is a by-product of core syntactic configurations,
such as specifier–head vs. complement–head relations, which hardly have any
effect at the syntax–semantics interface. Derivational affixes, on the other hand,
are more critical in distinguishing meanings. Accordingly, what Plag (2008a)
incorrectly argued to be an interlanguage property appears to be a general
property of those linguistic features that fail to activate the syntax–semantics
interface, even though they are visible at the syntax–morphology interface.

Thus, a crucial difference between the Sranan and Saramaccan derivational
morpheme -man (discussed in Chapter 3) and, say, the English plural -s which
was lost is that the former was visible at both the morphology–syntax and
syntax–semantics interfaces, while the latter was only accessible at the
morphology–syntax interface (as a configurational agreement morpheme).11
What this discussion suggests is that inflectional morphology is a mere
property of core syntax which does not play a crucial role at the syntax–
discourse/semantics interface, and will therefore not be selected in language
transfer. I submit that inflection is a weak competitor that is hardly visible to
late learners, though it is easily acquired by (bilingual) L1 learners (Weerman
2011). This would mean that the loss of inflections in a language contact
situation should by and large be dissociated from issues related to second
language acquisition proper. The analysis proposed here suggests instead that
late learners do not fail to acquire inflectional morphology, due to its
supposedly complex nature or obscure irregularities, but rather because it is
less visible than other syntactic features in a situation of language contact and is
therefore a weak competitor. Putting it this way, we have shifted the perspective
from what learners ‘fail to do’ to the nature of the object that they are learning.
We can thus distinguish between inflectional morphology and derivational
morphology in a principled way, because the latter, but not the former, often
relates to the semantic component. In addition, we make the prediction that in
cases where inflectional morphology activates the semantic component, it will
be retained. This is what Aboh and Ansaldo (2007) show in the context of case
morphology in Saramaccan and Sri Lanka Malay. (I refer the interested reader
to this study and references therein.) In the next section, I show more precisely
that the syntax–semantics interface is indeed a vulnerable domain for transfer
(i.e., for the recombination of syntactic and semantic features).
5.2.2 C and D as vulnerable interfaces

The discussion above indicates that interfaces are generally vulnerable and
more permeable to language transfer. This idea has already been proposed in
the literature by various authors, including: Haznedar and Schwartz (1997),
Lardiere (2000), Prévost and White (2000), Goad and White (2004), Sorace
(2005), Tsimpli and Sorace (2006), and Sorace and Serratrice (2009). Platzack
11
Note that the competition between affixal -man and -er led to the loss of the latter. We can
account for this on phonological ground: -er is phonologically weaker than -man. For instance
the former cannot be stressed while the latter can in some contexts. Cases like these strongly
support the view put forward here, viz., that the more interfaces an element activates the more
likely it is to be retained. In addition, other factors like phonological saliency and wider
distribution appear to play a role.
(2001), for instance, argues that the complementizer system is a vulnerable

domain for language transfer. This view can be easily illustrated by the
following sentences from Singapore English, in which sentence-final
discourse particles such as in (23a) are typically retained from Hokkien and
Cantonese (23b), which exhibit a robust class of such discourse markers in
sentence-final position (Ler Soon Lay 2005; Wakefield 2012; Lim 2011).
According to Aboh (2006b, 2010c, forthcoming), such discourse particles
represent the complementizer system, though they occur to the right edge (to
be discussed in Chapter 6).
(23) a. He’s quite innocent la hɔ. [Singapore English; Lim 2011]
3SG-be quite innocent PCL PCL
‘He is quite innocent, don’t you agree?’
b. Géi leng a hó? [Cantonese]
quite nice PCL PCL
‘Pretty nice, huh?’
As is evident from these examples, the final particles derived from
Cantonese are fully integrated in the morphosyntax of the Singapore
English complementizer system, because (i) they always occur clause-
finally and tend to follow a fixed order, and (ii) they take scope over the
clause to which they assign a very specific interpretation (a property that
Aboh 2004b, 2010c, 2010e, forthcoming has taken to be typical of elements
that occur within the left periphery of the clause, see Chapter 6). In this
regard, it is interesting to note that, according to Lim (2011), hɔ makes “it
clear that a positive response from the addressee is expected.” Leimgruber
(2009), on the other hand, reports that la “indicates [the] speaker’s mood/
attitude and appeals to [the] addressee to accommodate” to that mood. It is
obvious from the descriptions provided by these authors that these discourse
particles relate to the expression of epistemic modality and are arguably
related to the complementizer system as proposed in Aboh (2006b). If this
view is correct, then the fact that all these particles can be ‘transferred’ from
Cantonese and Hokkien to English is evidence that the complementizer
system is indeed a vulnerable domain.
Related proposals have been made concerning transfer between the bilingual
child’s grammars in L1 acquisition (e.g., Müller and Hulk 2000, 2001), L1
transfer in L2 acquisition (Valenzuela 2005), and L1 attrition under the
influence of the L2 (Sorace 2000). Let us briefly consider some of the
findings in the literature.
With regard to bilingual studies, Müller (1998) investigates transfer in
bilingual first language acquisition of German–French, German–Italian, and
German–English children. The focus of her study is verb placement in German
subordinate clauses. Like many West-Germanic languages, German allows VO
order in main clauses but V-final order in subordinate clauses. An example,

taken from Müller (1998: 151), is given in (24). Here we observe VO order in
the matrix but V-final order in the subordinate.
(24) Ich mag Nebensätze, weil sie so kompliziert sind.
I like subordinate clauses because they that complicated are
‘I like subordinate clauses because they are so complicated.’
Yet, the VO versus V-final asymmetry is not consistent across the board, as
suggested by the following constructions where both matrix and embedded
clauses display VO order (Müller 1998: 151).
(25) Ich mag Nebensätze, weil sie sind so kompliziert.
I like subordinate clauses because they are that complicated
‘I like subordinate clauses because they are so complicated.’
Additional factors adding to this complication (or ambiguity) are as follows:
– Some conjunctions are always used in the same order as in main clauses
(viz., non V-final):
(26) . . . denn heute hat sie Geburtstag.

since today has she birthday
‘ . . . since she has her birthday today.’
– Intonation plays a role in distinguishing between embedding or subordina-
tion and juxtaposition and thus interacts with word order. The following
sentence pronounced with a juxtaposition intonation displays a VO order in
the second clause (Müller 1998: 152).
(27) Ich weiβ, . . . die Nebensätze sind sehr kompliziert.

I know the subordinate clauses are very complicated
‘I know the subordinate clauses are very complicated.’
– Finally, some main clauses (e.g., exclamatives) display verb-final order:
(28) Daβ du deutsche Nebensätze analysieren kannst!

that you German subordinate clause analyse can
‘[I’m surprised] that you can analyse German subordinate clauses.’
In summary, we can say that German syntax displays a VO-vs.-OV

alternation that is sometimes (but not always) sensitive to subordination.
Since den Besten’s (1977) seminal work on verb placement in Germanic, this
alternation has been analyzed as resulting from a V2-rule. Simplifying matters
for the purposes of this discussion, the argumentation is described as in (29) and
(30), which indicate that the underlying order in West-Germanic is OV, but verb
placement in these languages is sensitive to the complementizer system
represented here as compP. This would mean that the VO order in main clauses
derives from the underlying OV order.
(29) Matrix clauses: there is no overt COMP, thus V raises to COMP:
COMPP
Specifier …………COMP ……………..Object………….V

Subject finite/verb or auxiliary
Wh-phrase
Topic
or any fronted constituent
The basic idea of this analysis is that in matrix clauses, comp is not
morphologically realized; consequently the verb must raise there. This
process alters the underlying OV order, giving rise to a surface subject/XP-V-
O order, where XP represents any fronted constituent. In embedded contexts
however verb movement cannot apply as indicated in (30):
(30) Embedded clauses: COMP is realized by a wh-phrase or daβ, V raising is not
possible:
COMPP
Specifier……………COMP……………..Object…………….V
Wh-phrase / dab finite/verb or auxiliary
In embedded contexts, compp may be realized alternatively by a

complementizer (e.g., daβ), which is inserted under comp or by a wh-phrase
that realizes the specifier position [spec compP]. The resulting word order is:
wh-phrase/daβ-Subject-O-V.
While, this analysis accounts for VO versus V-final alternation in general, it
does not extend to the VO order in embedded clauses (e.g., sentences with den),
unless we expand the comp layer as proposed by Rizzi (1997), Zwart (1997a,
1997b), Aboh (2004a, 2006b, and Chapter 6 below), and other related works.
With regard to language learning, various hypotheses are therefore possible,
including the following:
(31) Learning hypotheses about German word order
‒ Main clauses are of the type VO, but some may be OV.
‒ Embedded clauses are of the type OV, but some may be VO. (These can be
identified by paying attention to, for instance, intonation or the
complementizer.)
Accordingly, a learner of German has to entertain various, sometimes
conflicting, hypotheses about verb placement. By comparison, such a
problem might not arise for learners of English, French, and Italian, because
these languages display robust VO order in all contexts, while verb placement
does not involve the comp layer. Given this typological difference, the question
immediately arises of what happens when an L1 learner is confronted with both
OV- and VO-type languages simultaneously. Will she transfer properties of one
language to the other? In case transfer occurs, does it follow a particular
direction or does it target certain specific domains of the architecture of
language (e.g., interfaces)?
The acquisition facts discussed in Müller (1998) shed light on these
questions (see also Müller and Hulk 2000, 2001; Notley, van der Linden, and
Hulk 2007). Starting with monolingual L1 learners of English, French, and
Italian, no specific problem is reported with regard to the acquisition of
subordinate clauses. German children, pass through a phase where they make
some errors (e.g., using non V-final order in embedded clauses) but these errors
are few and children easily overcome this difficulty.
With regard to bilingual children, Müller (1998: 160) reports that:
When children acquire German and a second language, such as, French, Italian or
English, difficulty seems to be asymmetrical, that is, word order in French, Italian,
and English subordinate clauses does not represent a problem space for the children, in
contrast to German, where various error types are attested. This is the case indepen-
dently of whether German is the preferred language or not.
Accordingly, bilingual children overgeneralize (for a longer period than

monolinguals) the V2 order in subordinate clauses, thus producing examples
such as in (32):
(32) a. Wenn da komm andere schiffe . . . [Müller 1998: 158]
when there come other boats
‘When other boats come . . . ’
b. Ich bin so satt daβ ich mag kein blatt. [Müller 1998: 159]
I am so full that I like no leaf
‘I’m so full I don’t want more.’
According to Müller (1998), these ‘mistakes’ are instantiations of transfer
from English, French, and Italian which display VO order systematically to
German, in which this order is variable. In this context, transfer serves to ease
tension. Compared to English, French, and Italian, German appears to have
an opaque syntax for verb placement in embedded clauses. Learners
circumvent such opacity by transferring embedding structures from French,
Italian, or English, which have a more transparent syntax in this domain. This
observation led the author to propose (similarly to Müller and Hulk 2000,
2001) that transfer might follow a directionality pattern such that the
receiving grammatical system appears to be the one for which there is more
than one possible analysis for a given structure (German in the present case).
Because the phenomenon targets the comp domain, we can further conclude
that such interface domains are vulnerable to transfer. Müller and Hulk (2000:
228) argue that cross-linguistic influence is possible if the following two

conditions are met:
(i) Cross-linguistic influence occurs at the interface between two modules of
grammar, and more particularly at the interface between pragmatics and
syntax in the so-called C-domain, since this is an area which has been
claimed to create problems in L1 acquisition also.
(ii) Syntactic cross-linguistic influence occurs only if language A has a syn-
tactic construction which may seem to allow more than one syntactic
analysis and, at the same time, language B contains evidence for just one
of these possible analyses. In other words, there must be a certain overlap
of the two systems at the surface level.
As noted above, I assume in this book that transfer is interface-sensitive
and can therefore occur at any relevant interface.12 Applied to the Singapore
English data in (23), this view implies that the syntax of English is affected by
transfer precisely in a domain where this language shows subtle properties
relying primarily on intonation, while Cantonese uses discourse particles.
Here we observe that when it comes to the clause peripheral domains where
the two languages overlap with regard to the expression of discourse or
pragmatic notions, English exhibits an ambiguous grammar, as opposed to
Cantonese, which displays a transparent system based on distinct discourse
particles.
Additional evidence that compP is vulnerable to transfer (involving the
recombination of syntax and semantic features) comes from work on attrition.
This phenomenon generally occurs when people settle in a foreign
community in which they cannot continue to speak their native language
regularly and progressively lose competence in it, thus incorporating words
and grammatical properties of their L2. Ribbert and Kuiken (2010) studied a
different type of speakers: Germans in The Netherlands, many of whom (e.g.,
students) do not plan to stay in the host country, have ample opportunities to
visit their home country, have ample access to German (e.g., via the media,
Internet), and have a high level of education. The research question Ribbert
and Kuiken started out with was whether the German of such highly
proficient speakers could be affected by their nascent Dutch. In order to
investigate this, the authors studied the use of the complementizer um in
German (33a) in comparison with its Dutch cognate om (33b) (Ribbert and
Kuiken 2010: 42).
12
In this regard, Notely, van Linden, and Hulk (2007: 231) argue that “the only elements that can
be retained from Hulk and Müller’s original hypothesis are the two conditions formulated in the
2000 paper, namely (i) the phenomenon must involve the interface between two modules of
grammar, and (ii) overlap between the two grammars must exist.”
(33) a. Markus treibt Sport, um fit zu bleiben. [German]

Markus plays sports COMP fit to remain
‘Marckus plays sports in order to remain fit.’
b. Hij werkt om geld te hebben. [Dutch]
He works COMP money to have
‘He works in order to have money.’
Table 5.3, adapted from Ribbert and Kuiken, recapitulates the differences
between these two complementizers.
It appears that um/om do not have the same distribution in the two languages
nor the same discourse–semantics impact. According to these authors, German
um is governed by syntactic rules only and is therefore always obligatory.
Dutch om, on the other hand, is obligatory in some contexts but optional in
others. In their definition of optional om, the authors conclude that “speakers
use om (after illocutionary matrix verbs . . . when they expect the contents of the
infinitival construction to be realized”(Ribbert and Kuiken 2010: 43). This led
them to conclude that German um and Dutch om are regulated by syntactic
rules specific to embedded infinitives. In addition, Dutch om encodes epistemic
modality: The syntax of comp om in Dutch has a wider distribution than um in
German.
Ribbert and Kuiken’s (2010) study shows that German L1 speakers who are
in contact with Dutch make more mistakes in German in contexts where um is
disallowed in German but om is allowed in Dutch, albeit optionally. This
amounts to saying that German speakers learning Dutch transfer properties of
the syntax–semantics of om to their use of German um. In this instance of L2
transfer to L1, we face a situation where the transfer of Dutch properties to
German represents the reverse of the situation observed with bilingual L1
learners of German. The transfer proceeds from the language which offers
more than one learning hypotheses (i.e., optional om in Dutch) to that which
offers very strict and transparent rule (i.e., obligatory um in German). Given the
vulnerability of comp as explained above we can account for this situation by
saying that German is being influenced by the nascent Dutch I-language
because Dutch offers more semantic distinctions for om than does German
um. Accordingly, German learners of Dutch are selecting these semantic
distinctions of Dutch comp into their German.
Regarding creoles, these findings involving transfer from L2 into L1
suggest that even speakers of the lexifiers may have contributed to
language change by adopting patterns produced by the African learners.
Furthermore, these facts clearly show that the directionality of transfer (as
discussed in Müller 1998; Müller and Hulk 2000, 2001) is not only
constrained by structural opacity in the receiving language but also by
semantic properties expressed by the relevant grammatical items. This is
Table 5.3 The complementizer um/om in German and Dutch (adapted from Ribbert and Kuiken 2010: 43)
German Dutch Example
Obligatory Obligatory Piet nimmt die Straβenbahn, um nicht zu spat zu kommen [G]
Piet neemt de tram om niet te laat te komen [D]
Piet takes the tram comp not too late to come
‘Pieter takes the tram in order not to be on time.’
Ungrammatical Obligatory *Das ist Ein Aufsatz um zu heulen [G]
Dat is Een werkstuk om te huilen [D]
This is A paper comp to cry
‘This is a paper which could make you cry.’
Ungrammatical Optional *Michael hat versucht um eine Doktorandenstelle zu bekommen
Michael heft gepoogd om een promotieplaats te krijgen. [G]
Michael has tried comp a Ph.D.grant to get [D]
‘Michael has tried to get a Ph.D.grant.’
perfectly compatible with the view I adopt in this book, viz., in a situation of
contact, comp is particularly vulnerable to transfer regardless of whether the
receiving language is the learner’s L1 or L2.
Summing up, transfer seems immediately available when learners are
confronted with competing variants. The phenomenon does not appear to
follow any particular direction: Adult or late L2 learners transfer properties
of L1 to L2 or vice versa, and bilingual L1 learners transfer properties of an
L1 to another L1. Transfer can affect all sub-domains of grammar, but the
process implies a reorganization of the receiving grammar and seems
particularly sensitive to interfaces in general (e.g., comp, representing
syntax–discourse interface, and argument structure, representing syntax–
semantics interface).
I assume, following Rizzi’s (1997) cartographic approach and much
related work, that the structural make-up of comp, is a complex system
involving elements that clause-type the sentence and peripheral
grammatical items expressing finiteness and (speech act) modality. In
between these two classes of elements appear discourse-related items such
as topic and focus. This sequencing (to which I return in Chapter 6) is
represented in (34):
(34) ForceP [clause-typing]
Force TopicP
Discourse–Syntax
interface
Topic FocusP
Focus FinitenessP
Proposition
Finiteness
I have shown in Aboh (2003b, 2004a, 2004b, 2004c, 2006b, 2007a, 2010e,
forthcoming) and Aboh and Pfau (2011) that the Gbe languages support Rizzi’s
hypothesis in a straightforward manner. The following sentence from Gungbe
indicates that all the hypothesized positions, Force, Topic, Focus, and
Finiteness are realized by distinct markers in Gbe:
(35) Ùn kànbiɔ́ ní òsɔ́ éhè yà ògán wὲ.
1SG ask if horse this TOP chief FOC
mí ní zè è yì ná?
1PL MOOD take 3SG go give
‘I asked whether this horse, we should take it to THE CHIEF.’
Abstracting away from details about the derivation, this sequencing can be
represented as follows:
TopicP
Discourse–Syntax
interface
Topic FocusP
ya
Focus FinitenessP
we
Finiteness Proposition
Under such a general view of the properties of interfaces in language contact

situations, it seems reasonable to hypothesize that the domains where new
languages (e.g., creoles) display variation are those domains that relate to
vulnerable interfaces. This idea appears compatible with the fact that, even
though creoles generally have the SVO order and display a rigid hierarchy with
regard to the ordering of tense, mood, and aspect markers, they display a wide
range of variation regarding the position of the determiner in relation to the
noun (Baptista and Guéron 2007; Bobyleva 2013). This is shown in Tables 5.4
and 5.5, which summarize the distribution of the determiner within the noun
phrase in French-based creoles (Table 5.4), and across creoles with different
lexifiers (Table 5.5):
Table 5.4 Distribution of definite, demonstrative and plural markers across

French-based creoles (adapted from Déprez 2001: 52)
Creoles Singular Plural
Seychelles sa N Dem > N sa ban N Dem > Pl > N

Mauritian sa N la Dem > N > Def sa ban N la Dem > Pl > N > Def
Antillean N sa (l)a N > Dem > Def se N sa (l)a Pl > N > Dem > Def
N ta (l)a se N ta (l)a
Guadeloupian N la sa N > Def > Dem se N la sa Pl > N > Def > Dem
New Louisiana (la) N sa la (Def) > N > Dem > (Def) le N sa la (Def/Pl) > N > Dem
> (Def)
Old Louisiana N la N > Def N (sila) ye N > sila > Pl
Guyanese sa N la Dem > N > Def sa N y(e-l)a Dem > N > Pl > Def
Haitian N sa a N > Dem > Def N sa yo N > Dem > Pl
Table 5.5 Relative ordering of demonstratives, adjectives, and number across

creoles with different lexifiers and substrates
Ordering of Dem, N, Adj, Num Creoles
Dem-(PL)-Adj-N-(PL) Bahamian
(PL)-Dem-(PL)-(Adj)-N-(Adj) Kriyol, Tok Pisin, Seychelles, Mauritian
Dem/PL-Adj-N-PL-Dem Berbice Dutch, Sranan
(PL)-(Adj)-N-(PL)-(Adj)-(PL)-Dem Fa d’Ambu, Nubi
(Adj)-N-(Adj)-(Dem)-(PL)-(Dem) Haitian, Papiamentu
(PL)-Adj-N-Dem Sango
(Dem)-PL-N-Adj-(Dem) Lingala
The variation among creoles with the same lexifier (Table 5.4) and across
creoles with different lexifiers (Table 5.5) suggests that such sequencing
cannot simply derive from the different types of lexifiers, but rather must
stem from a more general structural factor such as the properties of the
interfaces.
In this regard, the idea that interfaces are vulnerable to language transfer
offers a new way of looking at such variation. The variation in Tables 5.4 and
5.5 is not surprising if we assume that the clause structure and the nominal
structure include specific syntactic articulations that represent the point of
interaction between syntax and semantics. Extending this idea to the nominal
domain as proposed in Aboh (2004a, 2004c, forthcoming), we reach the
characterization already presented in Section 5.1, viz., that the determiner
system represents a corresponding articulation within the noun phrase.
DP is the highest projection that links the nominal sequence to the
discourse. It is the nominal equivalent of ForceP. NumP, on the other
hand, links the nominal left periphery to the nominal predicate. It is the
nominal equivalent of FinP.
(37) DP
D TopicP
Discourse–Syntax
interface
Topic FocusP
Focus NumberP
Number NP-predicate
For our purposes, I submit, as in Aboh (2004c, forthcoming), that the DP

also involves a topic-focus articulation that hosts the specificity marker and
the DP-internal question particle that must co-occur with wh-questions in
Gungbe.
5.2.3 Transfer at the interfaces: Haitian and Sranan

Consistent with the analysis developed so far, the determiner system,
represented in (37) as the interface between the nominal structure and the
discourse, is a potentially vulnerable domain. In what follows, I will suggest
that the variation observed in Sranan and Haitian (as well as in the other creoles
presented in Tables 5.4 and 5.5) can be understood as a reflection of such
vulnerability. I start with Haitian, Gungbe, and French.
5.2.3.1 Haitian and Gungbe versus French I submit that Haitian and
Gungbe manifest the structure in (38), in which the number markers lέ /yo
and the specificity markers lɔ́ /la merge in Num and Top, respectively (Aboh
2002, 2004a, 2004c, 2005d).
(38) [DP [D [TopP [Top [±specific] lɔ́ /la [NumP [Num[±plur, ±def] lέ/yo]]]]]] [Haitian, Gungbe]
An immediate consequence of this is that the D position, the nominal
equivalent of force, is never overtly realized in these languages. I refer the
reader to Aboh (forthcoming) for detailed discussion but an obvious
empirical fact compatible with this view is that there is no distributive
difference between bare nouns and nouns that are marked with a determiner
in these languages. If we assume, following Szabolcsi (1987, 1994),
Longobardi (1994) and subsequent work, that D is a subordinating
category, then the fact that bare nouns and determined nouns can be used
interchangeably as arguments in these languages suggests that this position
does not need to be overtly realized (Aboh and DeGraff 2014). This further
correlates with the fact that the so-called articles in these languages, do not
seem to play the grammatical role they play in Germanic and Romance.
Building on previous analyses of Gungbe noun phrases, I therefore argue
that Gungbe and Haitian D-elements occur postnominally because number
and specificity are checked in these languages by overt movement of the
nominal predicate to [spec NumP] and [spec TopP], as shown in (39). Note
that this fronting rule is obligatory in Gungbe and Haitian, but not in French,
where specificity and number may be checked by demonstratives, such as ce,
cet, cette, and ces.
(39) DP
spec D¢
D TopP
spec Top¢
Top[±specific] NumP
spec Num¢
Num°[±plural] åP
l /yo
The same analysis carries over to the relative clause.13 According to Kayne
(1994), the relative clause involves a complementation structure similar to that
in (40), where the determiner and the relative complementizer first merge in D
and C, and the relativized noun raises to [spec CP].
(40) [DP [D the [CP mani [C° that [IP Mary will invite ti . . .]]]]]
Recall from examples (7) and (8), repeated below for convenience as (41)
and (42), that Haitian and Gungbe relative clauses differ from French relative
clauses because they display the following general properties:
1. The use of bare noun relatives.
(41) a. [Moun] ki pa travay p ap touché. [Haitian; DeGraff, p.c.]
people REL NEG work NEG FUT get-paid
‘Those who don’t work won’t get paid.’
b. [Mὲ] ɖê má wà àzɔ́ n màá yí kwέ. [Gungbe]
one REL NEG do work NEG.FUT receive money
‘Anyone who doesn’t work will not get money.’
2. The relative clause follows the noun, but precedes the specificity marker
and/or the number marker.
(42) a. [Fam ki vini wè-u] la. [Haitian; Sylvain 1936: 69]
woman REL come see-2PL DET
‘The woman (in question) who came to visit you.’
b. [Fam ki rete] yo. [Haitian; Sylvain 1936: 62]
woman REL arrest 3PL/
PL
‘The women who arrested them/or the women who were arrested.’
13
The fact that the Gungbe and Haitian definite specificity markers display the same syntactic
properties is supported by the fact that they can also function as clausal determiners, unlike
French articles (Aboh 2002, 2004a, 2004c; DeGraff 2007; Lefebvre 1998).
c. [Náwè ɖê wá kpɔ́ n mì] lɔ́ lέ. [Gungbe]

woman REL come see 2PL DET PL
‘The women (in question) who came to visit you.’
Given that relative clauses may precede the specificity marker and a number
marker in these languages, I propose to refine the representation in (16) as in
(22), consistent with Aboh (2002, 2005d):14
(43) [DP [D° [TopP [Top° lɔ́ /la [NumP [Num° lέ/yo [CP [C° ɖĕ/ki [IP. . . ...]]]]]]]]]
Since specificity and number licensing require fronting in Gungbe and
Haitian, I conclude that the relative clause precedes the determiner in these
languages because the relative CP-clause must raise to [spec NumP] and
[spec TopP], as represented in (44a) and (44b), respectively. Representation
(44c) indicates that French does not allow such a fronting rule in its relative
clauses and the article in this language might not qualify as a nominal topic
marker (Aboh 2002, 2004c, 2005d, forthcoming; Aboh and DeGraff 2014,
forthcoming).
(44) a.
b.
c. . . . [DP [D la [NumP [Num° [CP femmei [C° que [IP Marie a invitée ti. . .]]]]]]]
[French]
Under the proposed analysis, the parallels between Haitian and Gungbe
determiner phrases can be regarded as an instance of pattern transmission
because both languages share similar properties with regard to the function
and syntax of the nominal left peripheral elements, such as the specificity
markers lɔ́ /la and the number markers lέ /yo.
At this stage of the discussion, it is worth noting that pattern transmission, as
described here, does not entail a perfect replica of the source system in the
emerging language. Recall from the discussion above that I analyze pattern
transmission as the recombination of syntactic and semantic features of a
source language into a nascent language. Such recombination can only be
partial given that all the properties of the source language cannot be
replicated intact in the emerging language. Accordingly, pattern transmission
need not mean that the source language (whether the substrate or the
superstrate) and the creole have to be isomorphic. Note, for instance, that
14
I agree with Koopman (1982a, 1982b), DeGraff (1992), and Takahashi and Gracanin Yuksek
(2008) that ki is a complementizer, but see Lefebvre (1998) for an alternative analysis.
even though the Gbe determiner encoding the feature specific definite can be
equated with that of Haitian, as far as their syntax is concerned, the same does
not hold true of the Gbe determiner that expresses specific indefinite.
Indeed, all Gbe languages involve a distinct determiner that realizes specific
indefinite, follows the noun phrase, and excludes the specific definite
determiner, as illustrated in (45a–c). Examples (45d–e) indicate that the Gbe
specificity determiners can co-occur with the number marker.
(45) a. Súrù kù mótò lɔ́ wá. [Gungbe]
Suru drive car DET come
‘Suru drove the car back [i.e., he came back with the car in question].’
b. Súrù kù mótò ɖé wá.
Suru drive car DET come
‘Suru drove some car back [i.e., he came back with a certain car]’
c. *Súrù kù mótò lɔ́ ɖé wá.
Suru drive car DET DET come
d. Súrù kù mótò lɔ́ lέ wá.

Suru drive car DET PL come
‘Suru drove the cars back [i.e., he came back with the cars in question].’
e. Súrù kù mótò ɖé lέ wá.
Suru drive car DET PL come
‘Suru drove some cars back [i.e., he came back with certain specific cars].’
Interestingly enough, Haitian lacks a distinct specific indefinite determiner.
Instead, the language resorts to the (specific) indefinite determiner yon, which
is also used as numeral one, thus corresponding to French un (see Chapter 3).
This determiner occurs in prenominal position like numerals and adjectives.
Haitian and French are similar in this regard.
(46) a. Bouki se yon bon doktè. [DeGraff 1995: 242]
b. Bouki est un bon docteur.
Bouki COP DET good doctor
‘Bouki is a good doctor.’
For the time being, it is not clear to me what factors (either external or internal
to language) blocked the development of a Haitian postnominal specific
indefinite determiner modeled on the Gbe pattern. Given my previous
assumption that transfer-sensitive features are those that operate at the syntax–
semantics interface, I hypothesize that the Gbe specific indefinite determiner was
disfavored because it was less visible at the discourse–syntax interface than the
specific definite. This idea is compatible with the observation made previously
that specific definite referents are strongly discourse-anaphoric and must be
established in previous discourse (assertion of existence; Bickerton 1981). No
such constraint, however, holds of specific indefinite noun phrases, which are not
necessarily discourse-anaphoric, but only represent a referent that the speaker
intends to refer to (Ionin 2006a, 2006b). Accordingly, the indefinite specific
determiner may be subject to ambiguity (thus allowing conflicting learning
hypotheses) unlike the definite specific determiner.
The following French examples illustrate this asymmetry. The sequence
cette guitare ‘that guitar’ in (47a) is definite specific because it represents
information that is bound by a discourse antecedent known to both the
speaker and the addressee. In example (47b), however, the sequence une
certaine guitare ‘a certain/particular guitar’ is specific indefinite. It does not
necessitate a discourse-antecedent and need not be known to the addressee,
even though it is interpreted as specific.
(47) a. J’ai acheté cette guitare qu’on a vue dans le magasin ce matin. [French]
I have buy that guitar that 3SG has seen in the shop this morning
‘I bought the guitar that we saw in the shop this morning.’
b. J’ai vue une certaine guitare ce matin, je ne me rappelle

I-have seen a certain/particular guitar this morning, I not 1SG.ACC remember
plus la marque mais elle ressemble beaucoup a la tienne.
anymore the brand but it resembles very.much to DET yours
‘I saw a guitar this morning, I can’t remember what brand but it really looked like
yours.’
Compare these sentences to their Gungbe counterparts. Note from the

translation that in Gungbe (even more than in French) the use of the specific
definite determiner lɔ́ requires that both the speaker and the addressee know
about the referent. No such requirement holds of the specific indefinite marker
ɖé, however.
(48) a. Ùn xɔ̀ gíntá ɖĕ mí mɔ̀ tò cɔ́ fù mὲ égbè àfɔ́ nú lɔ́ . [Gungbe]
1SG buy guitar that 1PL see at shop in today morning DET
‘I bought that guitar that we saw in the shop this morning.’
b. Ùn mɔ̀ gíntá ɖé égbè àfɔ́ nú, má sɔ́ flín

1SG see guitar DET today morning 1SG.NEG again remember
Mákù étɔ̀ n àmɔ́ n é ɖì gíntá tòwè lɔ́ káká.
brand 3SG-POSS but 3SG resemble guitar 2SG-POSS DET a lot
‘I saw a certain guitar this morning, I can’t remember what brand but it really looked
like that guitar of yours.’
Accordingly Gungbe and French behave similarly with regard to the

interpretation of specific indefinite noun phrases, though their syntax is
different.
With regard to the recombination of features, the idea that specific indefinite
referents are less visible than specific definites at the discourse–syntax interface
(47–48), and are therefore disfavored in the competition and selection process
appears compatible with findings in second language acquisition studies. Ionin
(2006a, 2006b), for instance, shows that errors of article misuse in English
(i.e., overuse of the) mainly arise with specific indefinites. On the other hand,
learners seem to use the appropriately for assumed or pre-established referents
(i.e., specific definite in our terms). I interpret this asymmetry as the result of
the weakness of specific indefinite versus the strength (or prominence) of
specific definite at the syntax–discourse/semantics interface.
Recall also from our previous discussion of specific definite marking in
Haitian that the French use of the postnominal deictic locative là to encode
specificity may have reinforced the Gbe pattern. With regard to specific
indefinites, however, French does not seem to provide any congruent
syntactic paradigm that could have reinforced the Gbe pattern. This would
add to the weak discourse properties of specific indefinites in disfavoring the
emergence of a distinct specific indefinite determiner in Haitian modeled on
that in Gbe. This state of affairs further points to the potential role of
typological (di)similarities among the competing languages in favoring or
disfavoring the emergence of a pattern in the creole. Finally, the absence of a
designated indefinite specificity marker in Haitian contributed to block the
development in this creole (and in the Surinamese creoles) of a marker of
indefinite specific plural noun phrases, as I have already remarked in the
discussion related to Table 5.1. More study is needed before we achieve a
better understanding of these congruence phenomena and how they may favor
certain options while disfavoring others, but see Mufwene (2001) and Ansaldo
(2009) for some ideas.
Summing up, the facts presented here support the idea that domains of
transfer in a language contact situation involve those points (or articulations)
that interact with the interfaces. This chapter focuses on instances concerning
the syntax–discourse/semantic interface, but as I suggested in Chapter 4, it is
reasonable to think that other interfaces (e.g., syntax-phonology/morphology)
are relevant as well, as has already been demonstrated for second language
acquisition (e.g., Goad, White, and Steele 2003; Lardiere 2000).
The next section discusses the properties of the determiner phrase in Sranan
and compares them to those of the determiner phrase in English and Gungbe. It
is argued that Sranan retains only the function of the specificity feature as it
appears in the Gbe languages. The syntax of this feature is, however, modeled
on English syntax.
5.2.3.2 Sranan and English versus Gungbe I propose the internal structure
in (28) for the Sranan noun phrase. However, unlike in Haitian and in Gungbe,
in which Top and Num are realized individually by the markers lɔ́ /la and lέ /yo,
respectively, Sranan lexicalizes these two positions with a single grammatical
item. The Sranan determiner merges under Num but raises to Top to encode an
otherwise empty head. This determiner therefore binds two positions: Top
and Num.
(49) DP
spec D¢
D TopP
spec Top¢
spec Num¢
Num°[±plural] åP
na/di/den
Representation (49) accounts for the fact that Sranan involves one
determiner form, na, that expresses the feature combination [+specific,
–plural], and one, den, which encodes the feature combination [+specific,
+plural]. This is clearly different from Haitian and Gungbe, where the
expressions of specificity and number are dissociated (39). In this respect,
Sranan appears similar to Germanic and Romance languages, which conflate
definiteness and number in a single morphological form (e.g., le/la, un/une
versus les, des in French, and the/a[singular] versus the/Ø[plural] in English). The
link between Sranan and the Germanic (and Romance) languages further
translates into Num-to-Top movement within the noun phrase, even though
the semantics of the determiner originates in the Gbe languages (49).
Recall that Sranan clusters with Haitian and the Gbe-type languages in
discriminating between specific and nonspecific referents only, which is why
I assume that the determiner-like elements are topic markers in these languages
as well. For instance, these languages have bare nouns that are interpreted as
(in)definite or generic depending on the context, while noun phrases that are
associated with a specificity marker are necessarily discourse-anaphoric.
According to the syntax–discourse interface hypothesis, this property (i.e.,
the split between specificity and definiteness) clearly sets these creoles and
the Gbe-type languages apart from Romance and Germanic languages, where
the two notions often derive from a single determiner overtly realized in the
noun phrase (Aboh and DeGraff 2014). We therefore face a situation where
Sranan displays a hybrid noun phrase system with the function (or semantics)
of the determiner developing from Gbe languages and the syntax resulting from
English (and possibly Portuguese, in the case of Saramaccan).
However, Sranan and Haitian are parallel (and unlike Gbe) with regard to the
expression of the specific indefinite marker. We conclude in Section 5.2.3.1 that
Haitian lacks a specific indefinite marker due to the weakness of such markers
at the syntax–discourse interface. The same holds for Sranan: the language
does not have a specific indefinite marker. With regard to plural specific
indefinites, some speakers use the combination wan-tu (‘one-two’) with the
meaning of ‘some/certain,’ as in Kofi njan wan tu bana ‘Kofi ate some
bananas.’ These Sranan facts support the analysis in terms of the
vulnerability of interfaces. Indeed, because the specific indefinite marker is
weak in discourse, it allows for ambiguous interpretation at the interfaces and
could not be selected in Haitian or in Sranan.
With this in mind, I conclude that the functional similarities between Sranan
and Gungbe with regard to the specific definite are accounted for by a substrate-
induced feature, where the Gbe languages provided the relevant trigger for the
feature [specificity] to emerge in the creole. The formal licensing of this
feature, however, deviates from the Gbe pattern (or syntax), as a consequence
of the influence of English. The latter can be measured through several
properties that are specific to English and Sranan noun phrases but absent
from the Gbe languages:
1. Modifier–noun order
Both English and Sranan display prenominal modifiers. In Gbe, however, all
modifiers must occur postnominally. Example (50) shows this contrast.
(50) a. den bigi bana [Sranan]
DET big banana
‘the/those big bananas’
b. the big banana(s)
c. àkwékwè dàxó lέ
banana big DET
‘the big bananas’
2. Word order in possessive

Sranan and English display superficially similar possessive constructions
that differ from those found in Gbe. For instance, example (51) indicates that
Sranan and English have possessee–preposition–possessor as well as
possessor–genitive–possessee sequences. In the latter case, Sranan differs
from English only in exhibiting a morphologically null Genitive marker.
Finally, Sranan and English have pronominal possessors in prenominal
position.
(51) a. a oso fu a datra [Sranan; Bruyn 1995b: 266, 267, 269]
DET house of DET doctor
‘the house of the doctor’
b. a datra oso
DET doctor house
‘the doctor’s house’
c. den oso
3PL.POSS house
‘their houses’
In Gungbe, by contrast, possessive constructions manifest the orders
possessor–genitive–possessee and possessee–possessor–genitive.
While sequence (52a) might look similar to the English and Sranan cases in
(51b), it is worth remembering that in the Gbe languages the determiner is
postnominal. In addition, the sequence in (52b) is unavailable in English and
Sranan. Taken together, these facts could be interpreted as evidence that the two
language types (corresponding here to Gungbe versus English) do not use the
same pattern in possessive constructions; Sranan appears to have selected the
English pattern (Aboh 2002, 2004a, 2004c).
(52) a. dàwè lɔ́ sín xwé lɔ́ [Gungbe]
man DET POSS house DET
‘the man’s house [i.e., the particular house of the man in question]’
b. òxwé dàwè lɔ́ tɔ̀ n lɔ́
house man DET POSS DET
‘the man’s house [i.e., the particular house of the man in question]’
3. Demonstrative reinforcer constructions

Another difference between Gbe languages versus Sranan and English is that
the latter display a typical construction in Germanic and Romance languages
that has been referred to as the demonstrative reinforcer construction
(Bernstein 2001a, 2001b). Some examples in Germanic and Romance are
given in (53):
(53) a. ce livre-ci [French]
this book here
b. el libro interessante este [Spanish]
the book interesting this
‘This interesting book here’
c. that book there [English]
Similar constructions are found in Sranan, where the demonstrative element,
which normally occurs prenominally, as in (54a), surfaces postnominally to
encode emphasis, as in (54b) (see Bailey 1966 and Bobyleva 2013 for
discussing similar examples in Jamaican Creole):
(54) a. A kisi wan pikin, dan dati mofo. [Bruyn 1995a: 112, footnote 35]
3SG get DET child, then that mouth
ben langa leki turtur kayman
PST long like turtle caiman
‘She got a child, then that mouth (of her) was long like that of a turtle-caiman.’
b. den pikin disi/dyà [Bruyn 1995b: 265]
DET.PL child this/here
‘these children’
We can therefore conclude that Sranan behaves like Germanic and Romance
languages in making use of both pre- and postnominal positions for nominal
modification (as does Saramaccan). Such word order alternations inside the
noun phrase are simply unavailable in the Gbe languages, where the head noun
always precedes other nominal modifiers, in the rigid order noun-adjective-
numeral-demonstrative, and any word order alteration leads to
ungrammaticality (see Aboh 2004a, 2004c, 2010a).
4. Determiner–noun–[relative] order
Finally, I noted above that Sranan and English display relative clauses in
which the determiner precedes the head noun, which in turn precedes the
relative clause, as illustrated in (55a). The Gungbe example in (55b) indicates
that in the Gbe languages the relative clause must occur between the noun
phrase and the determiner. The relative clause appears to share the same space
as nominal modifiers in Gbe, unlike in Sranan and English (see Bruyn 1995a for
discussion of relative clauses in Sranan).
(55) a. Den uma di mi si na a wowoyo. [Sranan]
DET woman REL 1SG see LOC DET market
‘The women (in question) that I saw at the market.’
b. Náwè [ɖĕ ùn mɔ̀ n tò àxìmὲ] lɔ́ lέ. [Gungbe]
woman REL 1SG see at market DET PL
‘The women (in question) that I saw at the market.’
Even though the Sranan noun phrase is sensitive to the specific versus
nonspecific distinction (like in Gungbe and Haitian), I conclude from these
facts that the syntax of the noun phrase in Sranan differs from that of Gungbe or
Haitian noun phrases. Sranan appears to have inherited English syntax, which
is why noun phrases in both languages share a significant number of syntactic
properties.
Assuming this description, the question arises of how to account for such
asymmetry between two creole groups that are supposed to have the same
substrate Gbe languages. It seems clear to me that the answer to this question
cannot simply be the difference in lexifiers, viz., French versus English.
Indeed, even though these two languages show significant morphosyntactic
differences, they share a number of properties when it comes to the

determiner phrase. Both languages have prenominal determiners that
encode (in)definiteness morphosyntactically. They are alike in having
prenominal modifiers (e.g., adjectives and numerals) and have headed
relatives where the determiner precedes the head noun, which precedes the
relative clause. Finally, both English and French have demonstrative
reinforcer constructions.
All these congruent features of French and English emerged in Sranan and
Saramaccan, but not in Haitian. Instead, the relevant creoles developed various
syntactic patterns that do not perfectly replicate either the Germanic and
Romance patterns or the Gbe alternatives. For instance, Haitian has
postnominal specificity and number markers but prenominal adjectives.
Similarly, Sranan and Saramaccan have specificity markers of the Gbe type,
but these occur prenominally together with nominal modifiers.
In order to account for this non-uniform evolution, it appears justified to
argue that the prominent use of the specificity marker for discourse-anaphoric
(thus, specific) referents attested only in Gbe must have favored the
emergence of this feature in these creoles. That is, discourse prominence
and frequency conspired to make the features [specific, definite] active at
the syntax–discourse interface and so it was selected in the emerging
language. Such a situation must have been made possible by the fact that
the English determiner is ambiguous with regard to definiteness and
specificity and appears to be a weak competitor. From the point of view of
second language acquisition and bilingual language acquisition, this would
mean that the specificity features which are categorically expressed in the
Gbe languages were selected against both Romance and Germanic, which
display ambiguous alternatives.
With regard to morphosyntax, however, English determiner syntax
seems to have prevailed in the competition. Note, for instance, that
unlike Haitian and Gungbe, the licensing of the feature specificity does
not require fronting of the noun phrase complement in Sranan or
Saramaccan. Instead, the Surinamese creoles resort to a null operator in
[spec TopP] that checks the specificity feature under D, as in (35a).
According to Campbell (1996a), such a null operator exists in English
noun phrases, where it binds an empty category in the nominal inflectional
domain, as shown in (35b).15
15
Even though the data discussed here support this scenario, it is still not clear to me what factors
determine the selection of some licensing conditions over others, for instance, first merge of an
operator versus movement of a goal to some designated specifier position. I hope to return to this
issue in future work.
(56) a. DP b. DP
spec D¢ spec D¢
Opi
D TopP D[+specific] NumP
spec Top¢ spec Num¢

Opi ti
Top[+specific] NumP Num°[±plural] åP
the/this/that
spec Num¢ eci NP
ti
Num°[±plural] åP English
na/di/den
eci NP
Sranan
A point that is immediately obvious from these representations is that the

structure of the noun phrase in these creoles seems more complex than that of
the English noun phrase, due to the projection of the topic phrase (TopP). This
may turn out to be a trivial difference (Aboh, forthcoming), but for the purposes of
our discussion here, we can already see that the loss of morphology which
translates into less distinguishing forms in the Suriname creoles, as opposed to
English, does not necessarily correlate with absence of structure (as explained in
Chapter 4). Assuming this analysis is correct, we should not be surprised that the
Surinamese creoles also differ from Haitian and Gungbe with regard to the syntax
of relative clauses, as noted above. I now turn to them for a more detailed account.
5.2.3.3 Agreement in relative clauses in Gbe and the Suriname creoles The
following sentences indicate that, just like English determiners, the Sranan (and
Saramaccan) specificity markers precede the head noun, which precedes the
relative clause.
(57) a. Den uma di mi si na a wowoyo. [Sranan]
DET woman REL 1SG see LOC DET market
‘Those women that I saw at the market.’
b Di womi di mi go kai a kon. [Saramaccan]
DET man REL 1SG go call 3SG come
‘The man that I went out to call came.’
c. The man that Mary called came. [English]
According to the analysis proposed for relative clauses in (56a), Sranan and
Saramaccan are like French and English and unlike Gungbe and Haitian.
Indeed, the Surinamese creoles involve a null operator in [spec TopP] to
license specificity. The presence of this null operator in these languages as well
as in Germanic and Romance languages blocks fronting of the noun phrase
complement or relative clause to [spec NumP] and [spec TopP], unlike in
Haitian and Gungbe. Sranan, Saramaccan, and English relative clauses can
therefore be derived as in (58):
(58) [Sranan]
a. [DP [D° [TopP [Top° Den [NumP [Num° tden [CP uma [C° di [IP mi si tuma na a
wowoyo]]]]]]]]]
[Saramaccan]
b. [DP [D° [TopP [Top° Di [NumP [Num° tdi [CP womi [C° di [IP mi go kai twomi]]]]]]]]] a kon
[English]
c. [DP [D The [NumP [Num° [CP man [C° that [IP Mary called tman . . .]]]]]]] came
It is obvious from these representations that all these languages are
characterized by Germanic/Romance determiner syntax (modulo the
projection of TopP), but Saramaccan displays yet another significant
difference: it exhibits number agreement in relative clauses, a fact that is not
found in English but in Gbe instead. Recall from the Saramaccan example (37)
discussed in Chapter 3 and repeated here as (59) that this language involves two
types of relative markers (different from wh-pro forms) which agree with the
head noun (Huttar, Aboh, and Ameka 2013):
(59) a. Di fisi di mi tata kisi bigi.
‘The fish that my father caught is big.’
b. Dee fisi dee mi tata kisi bigi.
In (59a), di fisi ‘the fish’ is singular and combines with a relativizer in singular
di, while in (59b) dee fisi ‘the fish/es’ combines with the plural form dee. As I
showed in Chapter 3, this sort of agreement is a counter-example to
exceptionalist views of creolization, according to which creoles typically lack
‘contextual agreement.’ On the other hand, we must ask how such agreement
patterns arose in Saramaccan and why they did not in other Surinamese creoles,
including Sranan and Ndyuka. The phenomenon seems so exceptional that it is
often not discussed in the literature. Indeed, the variation among the Surinamese
creoles is reminiscent of that between the Eastern and Western Gbe languages
discussed in Chapter 2. In Gungbe (Eastern Gbe) for instance, the relative marker
has a unique form ɖě, which does not agree in number with the head noun. This is
illustrated by the Gungbe equivalents of the Saramaccan examples.
(60) a. [[Hwèví [ɖě bàbà cè wlé] lɔ́ ]] klò.
fish REL father 1SG.POSS catch DET big
‘The fish that my father caught was big.’
b. [[Hwèví ɖě bàbà cè wlé] lέ]] klò.

fish REL father 1SG.POSS catch PL big
In order to understand these Gungbe examples it is important to remember
that in this language the head noun and its associated determiner are separated
by the intervening relative clause thus leading to the sequence Noun–[Relative
clause]–Determiner. Given this format, the noun hwèví ‘fish’ is singular in
(60a), where it is associated with a singular marker of specificity lɔ́ , but plural
in (60b), where it combines with the number marker lέ . Now consider the
following Western Gbe languages involving Standard Ewegbe and Aŋlɔ
(Huttar, Aboh, and Ameka 2013). Here the plural now selects for a relative
marker that is attached to the number marker wó.
(61) a. aƒé si-wó míe-ƒle lá [Standard]
house REL-PL 1PL-buy TOP
b. aƒé yi-wó míe-ƒle lá [Aŋlɔ]

house REL-PL 1PL-buy TOP
‘the houses which we bought’
In Gengbe, another Western Gbe language, one finds similar constructions. In the
following example, taken from Haddican (2001: 13, citing Kagni 1989), it appears
that the number marker occurs both on the head noun and on the relative marker:
(62) É sà àvɔ̀ xóxó ɖé-wó kè-wó mú gbà lè àsì jí ò à
3SG sell cloth old INDEF-PL REL-PL NEG ADV be market on NEG DET
‘She sold some old loincloths that were outdated.’
As we can see from all these examples, the plural marker wó, which
commonly marks the nominal determiner, attaches to the relative marker and
thus marks agreement between the plural head noun and the relativizer. This is
parallel to the Saramaccan di vs. dee variation which I discussed informally in
Chapter 3 and which I describe below:
(63) REL[SINGULAR]= d + INFL[SINGULAR] ➜ d-i
REL[PLURAL] = d+ INFL[PLURAL] ➜d-ee
This description now finds support in the Gengbe example in (62) in which
we observe the following pattern:
(64) REL[SINGULAR] = kè+ Ø [SINGULAR] ➜ kè
REL[PLURAL] = kè+wó [PLURAL] ➜ kè-wó
Putting (63) and (64) side by side, we can now see the potential trigger for
agreeing relativizers in Saramaccan: the Western Gbe languages. Like above, one
may wonder why Saramaccan retained this particular feature, unlike other
Surinamese creole. Part of the answer to this question lies in the ecological
factors that have to do with population growth and structure, for example,
whether Western Gbe speakers formed part of the founder population and set in
motion the restructuring process, whether at some critical point they were dominant
among the Gbe speakers, and whether they were more concentrated in some parts
of Suriname but not others. More study is needed before we can hope to answer this
question conclusively and understand the dynamics of the emergence of norms in
speech communities, and how population factors interact with the selection of
linguistic features.
Another reason, directly related to the hypothesis proposed in this chapter,
relates to the interfaces. Indeed, if we assume Kayne’s (1994) complementation
analysis of relative clauses, Saramaccan and the Western Gbe languages appear
to have the left periphery (65) in relative clauses.
(65) a. DP [Saramaccan]
spec D¢
D TopP
spec Top¢
spec Num¢
Num°[±plural] CP
di/dee
spec C¢
C
di/dee IP
(65) b. DP [Gengbe]
spec D¢
D TopP
spec Top¢
spec Num¢
Num°[±plural] CP
wo
spec C¢
C
ke-wo IP
5.3 Concluding remarks 221
As I explained in Chapter 3, agreement in this context is a consequence of a

structural relation: the agreeing complementizer is the complement of the Num°
heading NumP within the noun phrase. This is shown by the double arrows. In
addition, a particularly interesting aspect of this analysis is that we are dealing
again with an interface property. Indeed, Kayne’s analysis suggests that
relativization triggers an interaction between the nominal discourse–syntax
interface (i.e., the DP–NumP sequence) and the interface between the discourse
and the proposition (i.e., the ForceP–FinP sequence in 34 and labeled here as CP).
Accordingly, this is yet another piece of evidence that interfaces are vulnerable to
the sort of recombination of syntactic and semantic features commonly referred to
as transfer, and that syntactic features that are active (or prominent) at the
interfaces are the more likely to make it into the emerging grammar.
5.3 Concluding remarks

This chapter suggests that Sranan, Saramaccan, Haitian, and Gungbe
specificity markers differ in syntax even though they have similar functions,
albeit in the family resemblance style. The major distinguishing factor is that
Gungbe-type languages require NP-fronting to license specificity while the
English-type languages (and to some extent the French-type languages) resort
to a (null) operator (Aboh 2004a, 2004c). This difference suggests the
following parametric variation:
(38) Specificity licensing requires:
a. NP-fronting to [spec DP] (via [spec NumP]) (Gungbe, Haitian),
b. Null operator merging in [spec DP] (English, French, Sranan/Saramaccan),
c. Other operator (e.g., demonstrative) plus partial raising (French).
Haitian Creole has adopted (38a), while Sranan and Saramaccan make use of
(38b). Given that these languages have emerged out of language contact, I
argue that the Gbe languages triggered the selection of (38a) in Haitian, while
English (and possibly Portuguese) provided the basis for Sranan and
Saramaccan to select (38b). Haitian does not seem to have constructions of
the type ce NP-ci/là. Accordingly, I submit that the French pattern in (38c) did
not fully contribute to fixing the parameters of the Haitian D-system, even
though the use of the locative deictic là might have reinforced the Gbe pattern.
Similarly, the Western Gbe languages favored the retention of agreement
marking in Saramaccan relative clauses, even though the Surinamese creoles
appear to have adopted the English syntax of noun phrases. Finally, I conclude
that the apparent mismatch between the so-called source languages and the
emerging creoles results from the vulnerability of interfaces, which allows
combinations of semantic and syntactic features from different sources,
provided that the selected combination is UG-compatible.
6 The emergence of the clause left periphery
Chapter 5 shows that the nominal left periphery, the locus of interfaces between
the noun phrase and the discourse, is vulnerable in language contact and
appears to display instances of syntactic recombinations that indicate the
hybrid nature of grammars. This chapter recapitulates the findings in Aboh
(2006b, 2007a) on the emergence of the complementizer system in
Saramaccan. Like in Chapter 5, where I discuss the structural make-up of
comp, I assume here Rizzi’s (1997) cartographic approach, which postulates
structure (34) in the same Chapter 5, repeated below as (1). According to this,
the complementizer system represents the interface between the proposition
and the discourse as illustrated by the boxes to the right. ForceP, the highest
projection, expresses the clause type. Fin(iteness)P, the lowest projection, is
responsible for tense and mood specifications that match those of the
proposition realized within the INFL domain. The space between these two
functional projections hosts various discourse-related projections which
license topic and focus phrases.
Force TopicP
Discourse–Syntax
interface
Topic FocusP
Focus FinitenessP
Finiteness
Proposition
In a series of studies on the left periphery in Gbe and Kwa in general, I have
shown that these languages realize the structure in (1) in a transparent way (e.g.,
Aboh 2003b, 2004a, 2004b, 2006b, 2006c, 2007a, 2009c, 2010c, 2010e, 2010f,
forthcoming; Aboh and Dyakonova 2009). An example is given again in (2a)
222
The emergence of the clause left periphery 223
that instantiates the heads in (1) in Gungbe. A tentative representation is

given in (2b):
(2) a. Ùn kànbíɔ́ wè ní kὲkέ éhè yà ví tòwè wὲ
1SG ask 2SG if bicycle this TOP child your FOC
mí ní zè è ná?
1PL MOD take 3SG give
‘I asked whether this bicycle, we should take it to YOUR CHILD.’
b. ForceP [clause-typing]
Force TopicP
Discourse–Syntax
interface
Topic FocusP
ya
Focus FinitenessP
we
Finiteness
Proposition
I provide the motivation for the representation in (2b) in various sections

throughout this chapter, but we can already realize from this description a
peculiarity of the syntax of the Gbe languages: they display distinct discourse-
markers that cluster within the left periphery of the clause. Consistent with
the conclusion in Chapter 5 that discourse-related properties relate to the
discourse–syntax interface and that interfaces roll the dice in language
contact in determining which features are selected into the emergent
language, we can hypothesize that in the particular case of the creoles we are
concerned with, the peripheral properties discussed here evolved in the same
manner, subject to the relevant ecological factors. In order to verify this
hypothesis, this chapter is devoted to the emergence of complementation in
Saramaccan, focusing on certain properties of the complementizer fu, its
interaction with other complementizers, and how these relate to interrogative,
topic, and focus constructions in the language. While the discussion only draws
on Saramaccan, the conclusion reached here arguably applies to other relevant
Caribbean creoles as well.
Indeed, studies on the distribution of fu/fi in Caribbean English creoles
show that this element occurs in various contexts including subordination.
The following data taken from Winford (1985) indicate that fu/fi can
introduce an argument and therefore function as a case-assigning
preposition (3a–b):
(3) a. Ai me fried fi i sneik. [Providence Island Creole]

1SG TNS afraid fi DET snake
‘I was afraid of the snake.’ [Winford 1985: 589]
b. Dat a fi mi buk. [Jamaican Creole; Winford 1985: 589]
DEM COP fi 1SG book
‘That’s my book.’
Aside from its function as argument licenser, fu/fi can also realize the INFL
domain of a clause where it expresses various meanings such as deontic
modality. The examples in (4) indicate that this function of fu/fi is not limited
to root contexts but occurs both in main and embedded clauses.
(4) a. Im fi kom op ya. [Jamaican Creole; Winford 1985: 589]
3SG fi come up here
‘He ought to come up here.’
b. An mi tel im fi stap. [Jamaican Creole; Winford 1985: 604]
and 1SG tell 3SG fi stop
‘And I told him to stop.’
The examples in (5) illustrate fu/fi as a complementizer selected by inceptive
and desiderative verbs (5a–b):
(5) a. Jan stat fi mek moni.
John start fi make money
‘John started to make money.’
b. Jan waan fi mek moni. [Guyanese Creoles; Winford 1985: 589]
John want fi make money
‘John wanted to make money.’
Finally, the sentences in (6) indicate that fu/fi also functions as a
complementizer that introduces purpose clauses (6a–b).1
(6) a. Im drap bred skrumz fi dey fala di trak.
3SG drop bread crumbs fi 3PL follow DET track
‘He dropped bread crumbs so that they could follow the track.’
b. Dem don put no paip fi we get waata. [Providence Island Creoles]
3PL NEG put NEG pipe fi 1PL get water
‘They didn’t put down any pipe for us to get water.’ [Winford 1985: 612]
In his analysis of fu/fi in Caribbean creoles, Winford (1985, 1993) argues that
there are two instances of complementizer fu/fi in these languages, aside from
the prepositional and related uses in (3). In one usage, fu/fi precedes the subject
1
See Winford (1985, 1993) for the discussion on the different classes of verbs that can select fi as
well as other instances of fi in these Caribbean English Creoles.
The emergence of the clause left periphery 225
of the subordinate clause that it marks and expresses irrealis or purpose, as in

(6). In the second usage, fu/fi follows the subject of the clause and encodes
deontic modality, as in (4). According to him, some Caribbean creoles allow
simultaneous occurrence of these two forms as indicated in (7) taken from
Winford (1985: 608):2
(7) a. I wuda en nais fi Jan fi go. [Jamaican Creole]
3SG would-have PST nice fi John fi go
‘It would have been nice for John to go.’
b. I haad fi mieri fi sel dem so.
3SG hard fi Mary fi sell 3PL so
‘It’s hard for Mary to sell them so.’
It appears from this summary that the morphological form fu/fi covers
semantic distinctions that combine properties of English prepositions
for-to. Under this characterization, the first type of complementizer fu/fi
(which will be referred to as fu/fi-type1) displays a syntax that is
comparable to that of for while the second complementizer (referred to
as fu/fi-type2) exhibits a distribution similar to that of English to. Given
these observations and the clear morphological distinction between
English for and to which learners arguably perceive, one wonders why
certain Caribbean English creoles employ the same morphemes to encode
two distinct syntactic positions.
In the context of this book, this question is even more puzzling when one
compares the Jamaican Creole examples in (7) to its Gungbe counterpart in (2a)
and represented in (2b). As the reader can see from these examples, the creole
fi1-fi2 sequence appears comparable to the Gungbe sequence ní-ní, thus
suggesting that there might be a typological relation between the two. Indeed,
the distribution of the morphemes ní/ne/nε in the Gbe languages also indicates
that these morphemes come in two types. The first instance (i.e., ní-type1)
expresses conditional similarly to English if/when (8a) and introduces
embedded interrogative as in (8b). In such constructions, ní-type1 precedes
the subject of the clause that it marks.
(8) a. Ní Súrù wá, mì yrɔ̀ -ὲ ná mì.
ní Suru come 2PL call-3SG PREP 1SG
‘If/when Suru comes, call him for me.’
b. Ùn kànbíɔ́ ní Kòfí wὲ xɔ̀ wémà lɔ́ ?
1SG ask ní Kofi FOC buy book DET
‘I asked if KOFI bought the book?’
2
This is similar to the distribution of Haitian pou derived from French pour ‘for’ as discussed in
Aboh and DeGraff (2014). I return to this in Section 6.4.
In its second usage, ní (i.e., ní-type2) expresses deontic modality and

systematically follows the subject of the clause that it marks.
(9) a. Kòfí ní xɔ̀ wémà lɔ́ .
Kofi ní buy book DET
‘Kofi should buy the book.’
b. Ùn ɖɔ̀ ɖɔ̀ Kòfí ní xɔ̀ wémà lɔ́ .
1SG say that Kofi ní buy book DET
‘I said that KOFI should buy the book.’
Taking things at face value, this description suggests that the sequences fu/fi-
type1/ní-type1 and fu/fi-type2/ní-type2 in the Caribbean creoles and the Gbe
languages as well as their equivalent for-to sequences in English belong to a
universal paradigm of the complementizer system. Yet, the fact that the creoles
seem to have adopted the Gbe, or, more generally, Kwa pattern raises the
question of whether these languages were the triggers of the emergence of
the complementizer in certain Caribbean creoles. If this is the case, one also
expects to find other properties of the Kwa clausal left periphery in the relevant
creoles. In order to address this question, the following sections compare
Saramaccan to its source languages Gungbe and English. When necessary, I
cite data from Haitian which shed light on the phenomenon being discussed.
While the analysis is based on Saramaccan data, it is expected to apply to all
relevant sets of data across creoles and non-creole languages.
Starting with the markers which delimit the frontiers of the complementizer
system, Section 6.1 discusses the relative distribution of fu in Saramaccan
compared to Gungbe and English. Section 6.2 extends the discussion to
topic, focus, and interrogation in Saramaccan. Building on the findings there,
Section 6.3 investigates structures involving predicate doubling and their
relation to clause structure. Section 6.5 briefly discusses the complementizer
pou ‘for’ in Haitian Creole thus pointing to parallels between the creoles and
their source languages. Section 6.5 concludes the chapter showing that most
peculiarities of the Kwa left periphery are retained in creoles together with
properties of the colonial source languages thus confirming the hypothesis that
interfaces are particularly vulnerable to syntactic recombination.
6.1 Complementizer fu in Saramaccan

The discussion here builds on Aboh (2006b, 2007a) which provide a detailed
comparative analysis of the complementizer system in Saramaccan and
Gungbe. In the following sections I only recapitulate the findings in these
studies. As the discussion in previous paragraphs shows, the Caribbean
creoles display various usages where fu can be used to introduce a DP
argument, mark genitive, introduce an embedded clause or express modality
6.1 Complementizer fu in Saramaccan 227
as in (4) above, repeated below for convenience, as (10a). While (10a) is taken
from Jamaican Creole and can be found in many other creoles (Winford 1985,
1993), the Saramaccan equivalent example in (10b) has been under debate,
with authors disagreeing regarding the syntactic category of fu in such
examples.
(10) a. Im fi kom op ya. [Jamaican Creole; Winford 1985: 589]
3SG fi come up here
‘He ought to come up here.’
b. Amato fu bói dí ganía. [Saramaccan]
Amato fu cook DET chicken
‘Amato should cook the chicken.’
c. Amato kĕ fu Ajawa kísi dí ógífou a matu.
Amato want fu Ajawa catch DET owl LOC jungle
‘Amato wants Ajawa to catch the owl in the jungle.’
d. Amato bói dí ganía fu nján.
Amato cook DET chicken fu eat
‘Amato cooked the chicken to eat.’
In the example (10b), fu expresses deontic modality. In this example, fu
follows the subject of the clause. This is different in the embedded example
(10c) expressing irrealis mood where fu precedes the subject. The latter
example of fu seems to also introduce non-finite purpose clauses as in (10d).3
Looking at the surface order, these examples suggest that Saramaccan involves
two types of fu: fu1 precedes the subject (10c–d) contrary to fu2 which follows
the subject (10b). (See Winford 1985 and Mufwene 1989 for related
discussions.) This is the conclusion reached in Aboh (2006b, 2007a), contra
Damonte (2002) and much related work.
6.1.1 Former analyses of fu in Saramaccan

The categorical status of fu in Saramaccan is controversial. Some authors (e.g.,
Wijnen and Alleyne 1987; Damonte 2002) argue that the examples in (10b–d)
instantiate a unique type of complementizer fu. Under this view, example (10b)
represents a subordination structure in which the main verb of the matrix abi
(derived from English have) has been elided. Thus, (10b) is a covert variant of
the examples in (11), where the matrix verb is morphologically realized by ábi
or musu and immediately precedes the complementizer fu.
3
Because the Gbe languages and the creoles discussed here do not show any verbal morphology,
‘finiteness’ is determined relative to the context in which the verb occurs (see Aboh 2004a for
discussion on Gbe; DeGraff 1992, 2007 on Haitian Creole; and Veenstra 1996 on Saramaccan).
(11) a. Amato ábi fu bói dí ganía. [Deontic modality]

Amato have fu cook DET chicken
‘Amato has to/must cook the chicken.’
b. Amato musu fu bói dí ganía.
Amato must fu cook DET chicken
‘Amato must cook the chicken.’
As already discussed in Aboh (2006b, footnote 17), that the element musu
can be considered a lexical verb in such contexts is suggested by the fact that,
unlike common Saramaccan TMA markers, such instances of musu can
undergo predicate focus with doubling. Predicate doubling is indeed a typical
property of lexical verbs both in Saramaccan and Gbe. I return to this issue in
Section 6.3.
(12) Musu a musu gó a wósu. [Wijnen and Alleyne 1987: 44]
Must 3sg must go loc house
‘He must definitely go home.’
Put together, the examples in (11) and (12) support the hypothesis that
example (10b) is an instance of a subordinate clause in which the matrix verb
ábi has been elided. This view is indeed adopted in Damonte (2002), who follows
the cartographic approach to the Saramaccan left periphery and claims that fu
realizes Rizzi’s (1997) finiteness position (FinP); accordingly, it expresses the
irrealis mood. According to Damonte’s (2002) analysis, the subject of the
subordinate clause is expressed by a null pronoun PRO licensed by the subject
of the matrix clause under control. The proposed structure is represented in (13):
(13) [IP Amatoi [VP ábi/musu/Ø [ForceP [Force° [FinP [Fin° fu [IP PROi bói dí
ganía]]]]]]]
As this analysis shows, two elements have gone unpronounced in this
structure: the matrix verb ábi and the subject of the embedded clause. This
way, this analysis appears compatible with the fact that the complementizer fu
precedes the subject in embedded contexts in Saramaccan and that the matrix in
example (10b) does not involve an overt verb. Damonte (2002) further argues
that the complementizer fu must be distinguished from other Saramaccan
complementizers such as táa ‘that’ illustrated below by the contrast between
(14a) and (14b).
(14) a. Amato kĕ fu /(*táa) Ajawa kísi dí ógífou a matu.
Amato want fu/táa Ajawa catch DET owl LOC jungle
b. Amato sábi táa/(*fu) Ajawa kísi dí ógífu a mátu.
Amato know táa/fu1 Ajawa catch DET owl LOC jungle
‘Amato knows that Ajawa caught the owl in the jungle.’
These examples show that the embedded clause introduced by fu must be

interpreted as unrealized (i.e., irrealis), contrary to factive subordinate clauses
introduced by táa. The semantic properties of these two complementizers also
match with their selecting verbs: fu is selected by a volition verb (e.g., ‘want’),
which takes a non-realized complement. Such a verb cannot select táa, as
indicated by example (14a). Instead, the complementizer táa combines with
factive verbs such as ‘know’ (14b).
(15) [IP Amato [VP sábi [ForceP [Force° táa [FinP [Fin° [IP Ajawa kísi dí ógífu a
mátu]]]]]]]
This analysis therefore postulates a unique fu in Saramaccan, which realizes
FinP, a different position than the one expressed by the complementizer táa.
Adapted to the structure in (2), this would mean that the Saramaccan
complementizer system can be represented as in (16):
Force TopicP Discourse–Syntax

taa
interface
Topic FocusP
Focus FinitenessP
Finiteness
fu Proposition
I discuss the topic-focus articulation in Sections 6.3 and 6.4, but the
representation in (16) seems to be compatible with the following example
reported in Veenstra (1996: 116), in which the two positions seem overtly
realized (see also Veenstra 2000).
(17) I táki táa fu a náki dí dágu.
3SG say that fu 3SG hit DET dog
‘You told him to hit the dog.’
Damonte (2002) therefore concludes that Saramaccan unique fu does not
express deontic modality but derives such semantics from the higher selecting
verb (e.g., ábi or musu). Yet, as Aboh (2006b, 2007a) shows, several facts about
Saramaccan suggest that Damonte’s (2002) analysis and related work are
misguided. The following section recapitulates a few shortcomings of this
analysis that are relevant to the discussion.
6.1.2 Against a unified account for Saramaccan fu

A crucial ingredient of the analysis of a unique fu is the assumption that in a
context like (10b) the selecting matrix verb ábi is elided, contrary to example
(11a), in which it is fully realized. Given this view, a simple question that one
must ask is whether ábi ever deletes (or goes unpronounced) in any other
grammatical context in Saramaccan. This question is relevant because the
verb ábi can be contracted in some well-defined contexts. The following two
sentences are taken from the SIL online Saramaccan–English dictionary (the
gloss is mine).
(18) a. A á di fou ta dë a goon.
3SG have det bird PROG be LOC ground
‘There are certain birds that live on the ground.’
b. Mi ku ën á soni.
1SG with 3SG have thing
‘S/he and I have something against each other.’
Because ábi can be contracted to á, one may think that it is (optionally)
elided in contexts such as (10a) and (11b). Yet, as already discussed in Aboh
(2006b) ábi or its contracted form á must always be realized as indicated by the
ungrammatical example in (19a), contrary to (19b):
(19) a. *I – wósu nö?
2SG have house Q
‘Do you have a house?’
b. I á wósu nö? [Adapted from Rountree and Glock 1982: 34]
2SG have house Q
‘Do you have a house?’
Furthermore, Wijnen and Alleyne (1987) reported that ábi contraction in
Saramaccan triggers vowel lengthening of the host element. This would mean
that the third person subject pronoun in example (18a) is realized as long /a:/.
To the best of my knowledge no such lengthening process has been reported for
examples like (10a) nor is such a phenomenon described in the analysis of the
following sentences reported in the literature:
(20) a. Hén fu heépi mi. [Wijnen and Alleyne 1987: 43]
3SG fu help 1SG
‘It is HIM who should help me.’
b. A fu nján dí físi [Bakker et al. 1994: 225]
3SG fu eat DET fish
‘S/he should/must eat the fish.’
c. A fu lóbi dí físi.
3SG fu love DET fish
‘He should love the fish
d. I ku en fu go. [Muysken 1987: 90]
2SG with 3SG fu go
‘You and him should go.’
As these examples show clearly, fu can be immediately preceded by different
types of subjects including weak and strong pronouns, as illustrated by the third
person singular weak form a versus the strong form hen (cf. Veenstra 1996).
Given the ungrammaticality of (19a), it is not clear to me what would motivate
total deletion of ábi only in contexts like (10a), (11a), and (20). There seems to
be no principled explanation for this deletion hypothesis given that the
contracted form á is obligatorily realized even in contexts where it is
immediately preceded by the third person weak pronoun a as in (18a). This
appears to be the most favorable context where one would expect the deletion
of àbi, which could fuse with the subject pronoun and would presumably make
the latter longer and tone-marked as in aá.
Example (18b) offers a similar, though less favorable, context where one
expects deletion to take place, but this does not happen contrary to what tenants
of a unique fu analysis commonly claim (without demonstration). Actually,
absence of ábi deletion in all these contexts conforms to a common observation
in linguistics that elements that usually delete within the verbal domain are
functional items rather than the lexical verb itself. To the best of my knowledge,
no other lexical verb of Saramaccan has been demonstrated to optionally
delete. The only option left for a unique fu analysis would be to suggest that
ábi (or musu for that matter) only deletes before fu, but this hypothesis makes
the argumentation circular and therefore unattractive.
All these facts point to what seems to me the most obvious conclusion:
Saramaccan displays a type of fu that follows the subject and encodes deontic
modality as indicated by all the examples presented thus far. This element,
labeled here as fu2 should be distinguished from another complementizer fu1
which precedes the subject and introduces irrealis and purposive subordinate
clauses (Aboh 2006b, 2007a). Indeed, while the examples involving fu2 in
(10a–b) and (20) express (weak) deontic mood, those involving fu1 in (10c–d),
(14), and (17a) convey irrealis meaning with no particular deontic force.
This observation leads me to another characteristic of fu2. Indeed, if
we assume that the combination ábi fu (and musu fu) in (11) are akin to the
English combinations have to or have got to, then we reach an interesting
characterization already mentioned in Aboh (2006b, 2007a) according to
whom this fu is comparable to English to rather than its correlate for (in
for-to complement clauses). This leads to the conclusion that the different
usages of the morpheme fu in Saramaccan cut across those of both for and to
in English. As I show below, English does involve deontic constructions such
as John is to write a new book, where deontic modality seems to be encoded by
the preposition to.
Aboh (2006b) further shows that while Saramaccan speakers interpret
constructions like (11) as conveying strong deontic modality, those in (20)
express weak deontic modality (Palmer 1987). It seems obvious to me that
Damonte’s (2002) approach, which treats the contrast between (11) and (20) as
a mere PF phenomenon, in which the verb ábi is spelled out in one case but not
in the other, cannot account for this semantic difference. In the context of the
current discussion, however, such subtle semantic variations are compatible
with the hypothesis that the morphological form fu occurs in two distinct
syntactic positions associated with distinct semantic effects in Saramaccan:
fu1 versus fu2. This would mean that examples such as (11) are structurally
different from those in (20). The latter are monoclausal as opposed to the
former which are bi-clausal.
Finally, Damonte’s (2002) analysis, according to which Saramaccan unique
fu realizes FinP, as indicated in (16), suggests that this complementizer must
follow topicalized and focused elements but necessarily precedes the subject.
We have already seen in example (20) that this description cannot be adequate,
given that there is one type of fu which follows the subject. The following
examples complete the picture by showing that there is yet another fu which
precedes topicalized and focused elements, contrary to Damonte’s predictions.
In sentence (21a), fu precedes the focused constituent di míi ‘the child.’ In
(21b), fu precedes both the topicalized constituent and the focalized resumptive
strong pronoun hén:
(21) a. A kĕ fu [di míi]i dέ ai nján dí kúku.
3SG want fu1 DET child TOP 3SG eat DET cookie
‘As for the child, he wants him to eat the cookie.’
b. A kĕ fu [di míi]i dέ [hén]i ti nján dí kúku.
‘As for the child, he wants HIM to eat the cookie.’
Altogether, these facts indicate that a complementizer fu that is immediately
selected by a verb and introduces an embedded clause must precede the
topicalized and focalized constituents in addition to the subject, while a fu
that encodes weak deontic modality must follow the subject (and,
consequently, topicalized and focused constituents). Consistent with Aboh
(2006b, 2007a), therefore, I argue in the following section that these two
functional elements realize the delimiting projections of the complementizer
system, that is, ForceP and FinP.
6.1.3 fu/fi1 and taa versus fu/fi2

According to Aboh (2006b, 2007a), the distribution of the complementizer fu
presented in the previous sections indicates that fu1 (which precedes the topic,
the focus, and the subject) realizes Forceo, the head of the highest projection
within the comp-system, which is responsible for clause typing. On the other
hand, fu2, which necessarily follows the subject, is assumed to realize Fino, the
head of the lowest projection of the comp-system FinP. This leads to the
sequencing in (22a) partially schematized in (22b):
(22) a. fu1 . . . topic . . . focus . . ..subject . . . fu2 . . . verb . . .
b. [ForceP [Force° fu1 /táa [TopP [Top° [FocP [Foc° [FinP . . ..[Fin° fu2 [IP . . . . . .]]]]]]]]]
Setting aside the details of the syntactic analysis fully implemented in
Aboh (2006b), what matters for the current discussion is that the
morphological form fu in Saramaccan plays several functions in the clausal
domain, two of which relate to the expression of finiteness and modality. Fu2
is a deontic mood marker in Saramaccan that merges under Fino while fu1,
which encodes irrealis mood, merges under Forceo. This view is compatible
with the observation already made about the sentences in (14) repeated
below as (23). Note that while fu1 and táa compete for the same position
Forceo, they do not impose the same readings on their complements, nor do
they show the same selectional properties with regard to the verbs they
combine with. Recall from the discussion there that while fu1 requires its
complement to be irrealis, this is not the case with táa which can select for
factive complements.
(23) a. Amato kĕ fu /(*táa) Ajawa kísi dí ógífou a matu.
Amato want fu/táa Ajawa catch DET owl LOC jungle
b. Amato sábi táa/(*fu) Ajawa kísi dí ógífu a mátu.
Amato know táa/fu1 Ajawa catch DET owl LOC jungle
‘Amato knows that Ajawa caught the owl in the jungle.’
Verbs that can select for fu1 include da taanga ‘to give strength, to
encourage,’ duwéngi ‘to force,’ paamúsi ‘to promise,’ da piimísi ‘to allow/
give permission to,’ and bigi ‘to begin’ all of which require that the
propositional content of the embedded clause be unrealized. On the other
hand, verbs that select for táa, including táki ‘to say’ or sábi ‘to know,’
require a factive complement (Damonte 2002).
Because fu1 selects for an irrealis complement, it can also introduce
purposive subordinate clauses unlike the factive complementizer táa. This is
shown by the following contrast:
(24) a. Amato bái fátu fu/*táa boí dí ganía.

Amato buy oil fu1/táa cook DET chicken
‘Amato bought fat in order to cook the chicken/Amato bought oil for cooking
the chicken.’
b. Mi táki táa /(*fu1) Amato nján dí bakúba.
1SG say táa/fu1 Amato eat DET banana
‘I said that Amato ate the banana.’
Aboh (2006b, 2007a) shows that táa can co-occur with fu2 as indicated by
the following example, in which fu2 expresses deontic modality within the
embedded clause. Unlike the complementizer táa, it does not depend on the
selectional properties of the matrix verb táki.
(25) Mi táki táa Amato fu nján dí bakúba.
1SG say táa Amato fu2 eat DET banana
‘I said that Amado should eat the banana.’
Based on these examples, I conclude that táa encodes declarative speech
modality, fu1 encodes irrealis mood, while fu2 expresses weak deontic mood, as
illustrated in (26):
(26) ForceP[táa: declarative; fu1: irrealis] > TopP > FocP > FinP[fu2 : deontic]
The representation in (26) therefore matches with the fu-fu sequences
reported in other Caribbean creoles by Winford (1985: 608) presented in (7)
and repeated here as (27a–b) for convenience, and the Gullah example in (27c)
from Mufwene (1989: 307):
(27) a. I wuda en nais fi Jan fi go. [Jamaican Creole]
3SG would-have PST nice fi John fi go
‘It would have been nice for John to go.’
b. I haad fi mieri fi sel dem so.
3SG hard fi Mary fi sell 3PL so
‘It’s hard for Mary to sell them so.’
c. I wan fǝ him fǝ pay mi mi moni (my spelling)
1SG want fǝ1 3SG fǝ2 pay 1SG-ACC 1SG-POSS money
‘I want him to pay me back.’
Similar Saramaccan examples are not reported in the literature but Glock
(1986: 40) discusses the following examples involving subordination, in which
fu1 introduces the subordinate clause, which contains fu2. As my translations as
well as Glock’s own glosses show, these examples are parallel to English
constructions involving for-to sequences and should receive the same
analysis. This example is reported using the author’s spelling and glosses and
translation, followed by my own translation.
(27) d. Ma nöö hën a ko pakisei taa a bi bai wan köni nöö

But now then he came thought saying he PST buy a wisdom and
di köni taa ja musu abi langahati fu soni u njan tumusi.
DET wisdom say you.NEG must have greed for something to eat too-much
Glock’s translation: ‘ . . . When he thought he had bought a piece of wisdom and the
piece of wisdom said you shouldn’t be too greedy . . . ’
My literal translation: ‘ . . . When he thought he had bought a piece of wisdom and the
piece of wisdom said you shouldn’t have greed for something to eat too much.’
e. Hën a taa an sa disa soni u naki möön.

then 3SG say 3SG.NEG can leave things to hit anymore
Glock’s translation: ‘Then he said he can’t stop knocking on things anymore.’
My literal translation: ‘Then he said he cannot leave things to (be) hit anymore.’
It has become a tradition in the literature to debate over the status of fu in

these Saramaccan constructions. I will not engage in such a debate here: it
seems clear enough to me that the two instances of fu in examples (27d–e) do
not encode the same function (Mufwene 1989). That Saramaccan developed
these two instances of the complementizer fu similarly to other Caribbean
creoles raises the question of what factors in the contact situation of the
Caribbean and, more specifically Suriname, could have triggered the
emergence of this pattern.4
6.1.4 fu-fu and ní-ní: Saramaccan versus Gungbe

Even though the two instances of complementizer fu discussed in this chapter
play the functions of English for and to, a morphological merger of these two
forms into a single form fu occurring in two distinct syntactic positions cannot
have derived from English. Indeed, the sentences in (27) show that English has
for-to sequences where Caribbean English creoles exhibit fu-fu sequences. In
this regard, Caribbean creoles are comparable to the Gbe languages. In these
languages, deontic modality is expressed by a marker consisting of the
consonant [n] combined with a vowel [i], [e], or [ε]. Accordingly, only this
vowel distinguishes between the different Gbe mood markers. In the following
examples, the Gungbe deontic marker is realized as ní, while its Gengbe
counterpart is expressed by nέ.
(28) a. Sέná ní wá fí. [Gungbe]
b. Sέná nέ vá fíè. [Gengbe]
Sena MOOD come here
‘Sena should come here.’
4
Although Suriname is geographically in South America, it has been treated in creolistics as
belonging in the Caribbean, along with Guyana and French Guyana. Politically both Guyana and
Suriname are included in the Caribbean Community, more commonly referred to as CARICOM.
Compared to the examples in (29), we observe that, everything else

being equal, the only element responsible for deontic interpretation in (28)
is ní/nέ:
(29) a. Sέná wá fí. [Gungbe]
b. Sέná vá fíè. [Gengbe]
Sena come here
‘Sena came here.’
Similarly to the Caribbean fu2, the Gbe mood markers also follow the
subject, hence the label ní2. I show in Aboh (2004a) that in Gungbe (and
most Gbe languages) ní2 precedes the negative marker má (30a), which in
turn precedes the future marker ná (30b) (Aboh and Nauze 2008):
(30) a. Àsíbá ní má ɖà làn lɔ́ blô.
Asiba ní2 NEG cook meat DET not
‘Asiba should not cook the meat.’
b. Àsíbá má ná ɖà làn lɔ́
Asiba NEG FUT cook meat DET
‘Asiba will not cook the meat.’
Though the mood marker ní and the future marker ná do not co-occur in
Gungbe, their distribution in (30) shows that they do not realize the same
syntactic category. In addition, the following examples further indicate that
the mood marker ní should not be confused with other mood markers such as
ɖó-ná ‘have to/must’ and sìgán ‘can.’ These occur in a different position, to the
right of negation (Aboh 2004a; Aboh and Nauze 2008):
(31) a. Àsíbá má sìgán ɖà làn lɔ́ .
Asiba NEG can cook meat DET
‘Asiba cannot cook the meat.’
b. Àsíbá má ɖó-ná ɖà làn lɔ́ .
Asiba NEG have-to cook meat DET
‘Asiba must no cook the meat.’
At this stage of the discussion we therefore have the sequencing in (32)
showing that the functional item expressing weak deontic modality precedes
the negative marker which in turn precedes the tense marker and subsequent
mood markers:
(32) ní[weak deontic modality]>má[negation]>ná[future]> . . . ɖó-ná . . ..sìgán
The Gungbe mood marker ní shares some syntactic properties with Saramaccan
fu2. In addition to following the subject, this marker can also introduce
subordinate clauses with irrealis meaning:
(33) Ùn jró ná Àsíbá ní wá.

1SG want COMP Asiba ní come
Lit: I want for Asiba to come
‘I wish Asiba could come.’
Indeed, that ná in (33) is used in a way comparable to English for is readily
demonstrated by the fact that, in addition to being selected by volition verbs as
in (33), ná also introduces purpose clauses as in (34a) as well as a dative
argument as in (34b):
(34) a. É sà xwé étɔ̀ n ná ví lέ ní ɖó yì àzɔ́ mὲ.
3sg sell house 3SG-POSS COMP child PL ní have go school
‘S/he sold his/her house so that his/her children could go to school.’
b. Súrù xò tó-mὲ ná Sέná.
Suru hit ear-in for Sena
‘Suru slapped Sena’s face.’
In this regard, it is interesting to observe that while ní is selected by the
complementizer ná in (33) and (34a), it can also be selected by the declarative
complementizer ɖɔ̀ , the equivalent of Saramaccan táa. It is worth noting here
that both complementizers derive from verbs of saying: Saramaccan
complementizer táa derives from táki ‘say/talk,’ while the Gungbe
complementizer derives from the inherent complement verb (Essegbey 1999)
ɖɔ̀ -xó ‘say/talk’. There is a universal tendency for complementizers introducing
declarative object clauses to derive from verbs of saying, but the historical and
structural relations between Saramaccan and the Gbe languages suggest that
this parallelism might not be a simple coincidence. The pair in (35) seems to
support this view. Example (35a) repeats the Saramaccan sentence under (25)
now contrasted with its Gungbe equivalent.
(35) a. Mi táki táa Amato fu nján dí bakúba. [Saramaccan]
1SG say that Amato fu2 eat DET banana
b. Ùn ɖɔ̀ ɖɔ̀ Àmátó ní ɖù àkwékwè lɔ́ . [Gungbe]
1sg say that Amato ní eat banana DET
These examples clearly show that Gungbe ní2 should be equated with
Saramaccan fu2. The parallelism between these morphemes is further
illustrated by the fact that both must follow the subject in main clauses where
they express weak deontic modality. This is shown by the ungrammatical
examples in (36):
(36) a. *Fu Amato nján dí bakúba [Saramaccan]
fu2 Amato eat DET banana
b. *Ní Àmátó ɖù àkwékwè lɔ́ [Gungbe]

ní2 Amato eat banana DET
‘Amado should eat the banana.’
Given these similarities between the languages, it can be assumed that both
fu2 and ní2 merge under Fin° as proposed in Aboh (2004a, 2006b, 2007a). Thus,
Gungbe also displays another instance of ní, which, just like Saramaccan fu1,
occurs in sentence-initial position from where it precedes the topic-focus
articulation and the subject. I refer to this form as ní1, as indicated in (37). In
such constructions, ní1 introduces conditionals (37a), indirect yes–no questions
(37b) or a subordinated adverbial clause (37c).5
(37) a. Ní ùn ɖó kwέ wὲ, ùn ná xɔ̀ xwé ɖàxó.
ní1 1SG have money FOC 1SG FUT buy house big
‘If I had some money, I would buy a big house.’
b. Ùn kànbiɔ́ ní Kòfí wâ?
1SG ask COMP Kofi came-INTER
‘I wonder whether Kofi came?’
c. Ní Kòfí gɔ̀ wá xwé-gbè, mì yrɔ́ -ὲ ná mì.
ní1 Kofi return come house-at 2PL call-3SG for 1SG
‘When Kofi returns, call him for me.’
The distribution of ní1 suggests that it occupies the highest position in the
complementizer system where it may clause-type the proposition. This view is
immediately supported by the fact that ní1 is in complementary distribution
with the declarative complementizer ɖɔ̀ unlike ní2. The sentences in (38a–b)
indicate that while the complementizer ɖɔ̀ can introduce indirect wh-questions,
ní1 introduces indirect yes–no questions:
(38) a. Ùn kànbiɔ́ ɖɔ̀ mέnù wὲ xɔ̀ wémà lɔ́ ? [Gungbe]
1SG ask that who FOC buy book DET
‘I asked who bought the book?’
b. Ùn kànbiɔ́ ní Súrù wὲ xɔ̀ wémà lɔ́ ? [Gungbe]
1SG ask COMP who FOC buy book DET
‘I asked if/whether SURU bought the book?’
The data in (38) suggest that these two complementizers can be selected by a
verbum dicendi that requires an interrogative complement. Because they share
this selectional property, these complementizers cannot co-occur in Gungbe as
indicated by the ungrammatical example (39):
5
See Aboh (2004a), Aboh and Pfau (2011), and references cited therein for the discussion of
questions in Gbe.
(39) *Ùn kànbiɔ́ ɖɔ̀ ní mὲɖé xɔ̀ wémà lɔ́ ? [Gungbe]

1SG ask that ní1 someone buy book DET
‘I asked whether anyone bought the book?’
As already indicated by example (35b), this is contrary to ní2, which co-
occurs with the declarative complementizer ɖɔ̀ . An additional example is given
in (40):
(40) É jὲ ɖɔ̀ yɔ̀ kpɔ́ lέ ní wá.
3SG convenient that child PL ní2 come
‘It is convenient that the children come over.’
Further examples showing that Gungbe involves two types of the marker ní
come from topic and focus constructions (Aboh 2006b). Below, ní1 precedes
the topic-focus articulation (41a–b), while nì2 follows (41c–d). This is
reminiscent of the distribution of fu1 and fu2:
(41) a. Ùn kànbiɔ́ ní [Kòfí yà] é xɔ̀ wémà lɔ́ ?
1SG ask ní1 Kofi TOP 3SG buy book DET
‘As for Kofi, I asked whether he bought the book.’
b. Ùn kànbiɔ́ ní [Kòfí wὲ] xɔ̀ wémà lɔ́ ?
1SG ask ní1 Kofi FOC buy book DET
‘I asked whether KOFI bought the book?’
c. Ùn ɖɔ̀ ɖɔ̀ Kòfí wὲ ní xɔ̀ wémà lɔ́ .
1SG say that Kofi FOC ní1 buy book DET
‘I said that Kofi should buy the book.’
d. Ùn kànbiɔ́ ní [Kòfí yà] [àmànsì éhè wὲ] Àsíbá ní ɖà ná-ὲ?
1SG ask ní1 Kofi TOP medicine DEM FOC Asiba ní2 prepare for-3SG
‘As for Kofi, I asked whether Asiba should prepare THIS MEDICINE for him?’
In their discussion of the clausal and TMA markers in Gungbe, Aboh, and
Nauze (2008: 225) propose that this language displays the following
sequencing:6
(42) Mood[ní:conditional]>Topic>Focus>[Subject]>Mood[ní:deontic]>Negation[má]
>Tense[ná:future]> Adverb[kà:willingly]>Adverb[tὲ:even]>Adverb[kò:already]
>Adverb[sá:nevertheless]>Adverb[gbɔ̀ :reluctantly]>Adverb[sɔ́:again]>Adverb[gbέ:at least]
>Mood[ɖóná:must]>Mood[sígán:can]>Aspect[nɔ̀ : habitual] > Aspect[tò: progressive]
>Object>Aspect[ná: prospective]>Verb>Nominalizer
According to this formula, the Gungbe example (41d) with ní1-ní2 sequence
is particularly telling because it shows that some Caribbean creoles and the Gbe
languages manifest the same pattern when it comes to the structural make-up of
their complementizer system. Similarly to Saramaccan therefore, Gungbe
6
See Gibson (1986) for TMA sequencing in Guyanese and Durrleman-Tame (2008) for Jamaican
Creole.
displays a complementizer system in which ní1 expresses Force°, while ní2

realizes Fin°. There is to my knowledge no systematic semantic analysis of
these markers in Gbe and in the relevant Caribbean creoles. In the context of the
current discussion, suffice it to say that ní1 and fu1 introduce a non-realized
proposition while ní2 and fu2 express deontic modality.
Putting everything together, we reach the characterization in (43) for both
Gungbe and Saramaccan:
Force TopicP
Discourse−Syntax
interface
Topic FocusP
Focus FinitenessP
Finiteness
fu 2 2 Proposition
It seems obvious from this description that the morpheme fu emerged in

Saramaccan as the result of a recombination of the syntactic and semantic
features of the Gbe morphemes n-V where -V represents the distinguishing
vowels [i], [e], or [ε] across Gbe as mentioned previously. These morphemes
have the property of introducing an irrealis clause, in which case they occur in
clause initial position or may be selected by non-factive verbs. As such, these
morphemes precede the topic-focus articulation as well as the subject of the clause
under their scope. In addition, these morphemes can encode deontic modality, as a
result of which they must follow the subject and participate in the sequencing of
clause-internal TMA markers. Looking at these morphemes from the English
perspective, it is obvious that they cut across usages of the prepositions for and to,
some properties of which are also associated with Caribbean preposition-
complementizer fu (27). In the following section I contrast Saramaccan fu with
these English prepositions (2006b: footnote 48; 2007).
6.1.5 fu1-fu2 and for-to: Saramaccan versus English

The discussion thus far has shown that the morphological form fu in
Saramaccan (and in other Caribbean English creoles) covers two types of
elements that correspond to the various usages of the morphemes n-V[i, e, ε]
in Gbe. Similarly, fu has been shown to share certain properties with English
prepositions ‘for’ and ‘to.’ More precisely, fu1 appears akin to English ‘for’
while fu2 is comparable to the preposition ‘to.’ In this regard, Winford’s (1985)
discussion on decreolization was meant to explain why certain speakers of
Caribbean English creoles substitute the complementizer fu1/fi1 for ‘for,’ while
deontic fu2/fi2 is replaced by ‘to.’ The discussion in this chapter however
suggests that we need not resort to decreolization to account for this property.
Actually, these two usages are reported in early documents suggesting that they
have always been part of the varieties of English-based creoles. Similarly, these
two usages are not typical of Standard English only, but can be found in non-
standard Englishes as well; so there is no need to resort to decreolization to
account for their emergence in Caribbean English creoles in general. From the
point of view of competition and selection, the different properties of the
morphemes fu suggest the characterization in (44) which cuts across Gbe and
English.
(44) a. fu1/fi1 [Saramaccan, Caribbean English creoles] = ní1/n-V1 [Gungbe (Gbe] =
for [English]
b. fu2/fi2 [Saramaccan, Caribbean English creoles] = ní2/n-V2 [Gungbe (Gbe)] =
to [English]
It therefore appears from this description that during the formation of the
creole, the creole forms fu1/fi1 and fu2/fi2 competed with the English
complementizers ‘for’ and ‘to,’ respectively, in addition to competing with
the Gbe n-V1/2 forms. In the context of this discussion it is important to realize
that English displays deontic expressions in which the prepositions for and to
are used in contexts similar to their creole equivalents fu1 and fu2.7
(45) a. John is to finish his book.
b. John is to come here immediately.
c. John wants for Mary to catch the owl in the jungle.
Similarly, one sometimes encounters examples such as (46) in certain non-
standard English varieties. In these constructions, the complex for-to occurs in
positions similar to those occupied by fu-type elements described by Winford
(1985) in the context of Caribbean English creoles (46b).
(46) a. Always the evenings for to get the men for to do it.
[South West English dialect; Wagner 2004: 168]
b. Mi a hop fi Jan fi go. [Jamaican Creole]
1SG PROG hope fi John fi go
‘I’m hoping for John to go.’ [Winford 1985: 602]
Though these examples are arguably formally distinct, the fact that non-
standard English involves for-to sequences may reinforce learners’ analyses of
modal ‘to’ as being associated with preposition-complementizer ‘for.’ Note
that such a surface analysis is compatible with the Gbe data where the two
7
I thank Anne Zribi-Hertz for bringing these cases to my attention.
preposition-complementizers typically have the same morphological shape. It

is obvious from these examples that the forms fu in Caribbean English creoles
are not simple replicas of the English or Gbe corresponding grammatical items.
Instead, what we see here is a complex recombination of certain properties of
the prepositional complementizers in English with properties of comparable
complementizer elements in Gbe (i.e., Kwa) into a unique morphological form
f-V across Caribbean creoles. This form covers subtle but distinct syntactic and
semantic properties suggesting that it realizes two distinct structural positions
in the clause. Given this understanding, the question now is to understand to
what extent this new syntactic form differs from its sources.
6.1.5.1 fu2, ní2 versus to The discussion has shown that the Saramaccan
deontic marker fu2 occurs in finite sentences where it follows the subject. This
would mean that in such Subject- fu2-V constructions, the subject is marked for
nominative case. Though most creoles don’t have overt case morphology,
Saramaccan does. In this language, the third person weak pronoun exhibits
two case-sensitive forms: a for nominative and hεn/εn for accusative
(cf. Veenstra 1996).8 As shown in example (20a–b), repeated here as (47a–b),
both the nominative form a and the tonic form, expressing emphasis, can
precede deontic fu2. Saramaccan fu2 constructions therefore involve finite
clauses.
(47) a. A fu nján dí físi.
3SG fu2 eat DET fish
‘S/he should/must eat the fish.’ [Bakker, Post, and van der Voort 1994: 255]
b. Hén fu heépi mi. [Wijnen and Alleyne 1987: 43]
3SG fu help 1SG
‘It is HIM who should help me.’
This distribution is different from that of the English preposition to, which
cannot encode modality on its own (48a) but must combine with a verb or a
(modal) auxiliary that expresses tense specifications and which also licenses
nominative forms only (48b). Such constructions can be analyzed similarly to
raising constructions (e.g., John seems happy, John seems to like baseball)
where the subject of the non-finite embedded clause raises to the matrix to be
assigned nominative case.9 Finally, in non-finite embedded contexts where to is
8
Building on Veenstra (1996) one can actually propose that Saramaccan exhibits case distinction
in its entire pronominal system since weak pronouns are restricted to the subject position, while
tonic pronouns typically occur in object positions even though they can be used emphatically in
subject positions as is the case in some Romance, Germanic, and Kwa languages.
9
Mufwene (2005c: 233) discusses examples of headlines or caption where “to” seems to encode
irrealis.
(i) President Bush to meet Prime Minister Blair.
immediately preceded by a subject, the latter must be in accusative form. Under

the traditional analysis of exceptionally case-marked (ECM) constructions, it is
assumed that such subjects are case-licensed by the matrix verb, hence the
accusative form.
(48) a. *John to finish his book/*he to finish his book/*him to finish his book.
b. John is to finish his book/He is to finish his book/*him is to finish his book
c. John wants {her/*she} to catch the owl in the jungle.
Saramaccan fu2 therefore differs from English to with regard to its
distribution and the licensing of the subject to its left. Fu2 is not sensitive to
finite vs. non-finite distinction, unlike English to. Accordingly, fu2 can occur in
contexts where the immediately preceding subject has a nominative form,
while to occurs in raising constructions where its subject is licensed within
the matrix clause or in embedded clauses where its subject is assigned
accusative case by the matrix verb.
The following Gungbe examples show that this language patterns like
Saramaccan. In this language, the deontic mood marker is preceded by a
nominative subject (49a), as well as the tonic form when used emphatically
(49b). These constructions are therefore indistinguishable from Gungbe finite
constructions.
(49) a. É ní ɖù hwèví lɔ́ .
3SG ní2 eat fish DET
‘He should eat the fish.’
b. Úɔ̀ ní ɖù hwèví lɔ́ (ná Súrù ní ɖù làn lɔ́ .
3SG ní2 eat fish DET CONJ Suru ní2 eat meat DET
‘He should eat the fish, while Suru should eat the meat.’
The parallels between Saramaccan and Gungbe are further highlighted by
the sentences in (35a–b), repeated here as (50a–b) where it appears that the
markers fu2 and ní2 can occur in embedded finite clauses introduced by the
declarative complementizer ‘that.’ English to being restricted to non-finite
contexts is not allowed in such sentences (50c).
(50) a. Mi táki táa Amato fu nján Dí bakúba. [Saramaccan]
1SG say that Amato fu2 eat DET banana
b. Ùn ɖɔ̀ ɖɔ̀ Àmátó ní ɖù àkwékwè lɔ́ . [Gungbe]
1sg say that Amato ní eat banana DET
c. *John said that Mary to catch the owl in the jungle
6.1.5.2 fu1, ní1 versus for In the previous sections I showed that Saramaccan
fu1 shares certain semantic and distributive properties with the English
preposition for. This is shown again in example (14a), repeated here as (51a)
which we can compare to the English example in (45c) represented again
in (51b).
(51) a. Amato kĕ fu Ajawa kísi Dí ógífou a matu.
Amato want fu Ajawa catch DET owl LOC jungle
b. John would prefer for Mary to catch the owl in the jungle.
Aside from this surface similarity, however, there appear to be structural
differences between these two forms which make Saramaccan and Gungbe
pattern alike. Indeed, an important aspect of fu1 and ní1 is that both precede
the topic and focus articulation unlike English ‘for.’ An illustration was given
in (21b) repeated here as (52a) together with a grammatical Gungbe
comparable sentence (52b). In these two examples, the topicalized and
focused elements follow the subordinator. The ungrammatical English
example (52c), indicates that topicalization is excluded in the
corresponding English constructions (Rizzi 1997; Haegeman and Ürögdi
2010a, 2010b).
(52) a. A kĕ fu [dí míi dέ] [hén] nján dí kúku.
‘As for the child, he wants HIM to eat the cookie.’
b. Ùn kpɔ́ n ní [ví éhè yà] [àklà lɔ́ wὲ] é ɖù.
1SG look ní1 child DEM TOP cookie DET FOC 3SG DET
‘I checked whether as for this child, he ate THE COOKIE.’
c. *I would prefer for, this child, the cookie he ate.
Accordingly, while Saramaccan fu1 can precede topic and focus constituents
in embedded clauses, English apparent equivalent for cannot (51b vs. 52c).
Similarly to fu2 and ní2, which license a finite clause whose subject receives
nominative case, (unlike English to), the irrealis complementizers fu1 and ní1
also license finite clauses, contrary to English for. In these examples, the
embedded clause displays a pronoun in the nominative form, to the exclusion
of the accusative form.
(53) a. A kĕ fu a/*εn nján dí bakúba. [Saramaccan]
3SG want fu1 3SG-NOM/ACC eat DET banana
‘S/he wants her/him to eat the banana.’
b. Ùn kpɔ́ n ní à/*wè ɖù [àklà lɔ́ ].
1SG look ní1 2SG-NOM/ACC eat cookie DET
‘I checked whether you ate the cookie.’
c. It was not possible for {him/*he} to meet the deadline.
Based on these facts, I conclude that while the Saramaccan morphemes fu1
and fu2 exhibit semantic properties that cut across the Gbe languages and
English, the morphosyntax appears to be mainly mapped on the Gbe
languages. Note, however, that the Gungbe ní1 and ní2 and the Saramaccan
fu1, and fu2 do not show identical distribution and semantic properties.
At this stage of the discussion, a non-trivial question that arises is why
Saramaccan (and possibly other English creoles) have the same form for the
two types of complementizers. Given the proposed analysis, the fact that the
Saramaccan complementizers fu1 and fu2 behave similarly to both Gungbe
n-V1/2 complementizers and English for-to sequences does not automatically
imply that the Saramaccan complementizers should exhibit a unique form.
After all, that Gungbe displays ní1 and ní2 and Saramaccan fu1 and fu2 could be
a simple coincidence. Certain facts about Gbe suggest a different view. Indeed,
the Gbe languages appear to display a striking correlation with regard to the
form of ‘dative’ prepositions and certain tense and mood markers.
As shown in example (54), the Gungbe ‘dative’ preposition also used to
introduce purpose clauses has the form ná in this language. Two additional
examples are given below:
(54) a. Súrù xɔ̀ kὲkέ ná ví lέ ní dó yì wéxɔ̀ mὲ.
Suru buy bicycle for child PL ni2 ride go school’
‘Suru bought a bicycle for the children to cycle to school.’
b. Súrù sán tà ná kòkló lɔ́ .
Suru cut head to chicken DET
‘Suru chop the chicken’s head off.’
Accordingly, while the preposition used to introduce purpose constructions
and new arguments is ná in Gungbe, the irrealis/conditional marker is
expressed as ní1 while the deontic marker expressing subjunctive and
injunctive is realized as ní2. In English, these notions can be expressed by
means of various auxiliaries or modals which in subordinate contexts combine
with for and to functioning as complementizers in non-finite embedded clauses.
It therefore appears that there is an asymmetry between English and Gungbe as
to the form of the complementizers and their relation to the preposition. Yet, the
broader picture from Gbe provides us with a new insight into this apparent
paradox. Here, I present data from an Eastern Gbe language, Fongbe which is
closely related to Gungbe, and two closely related Western Gbe languages,
Ewegbe and Gengbe. I start with Fongbe.
Fongbe (Eastern Gbe), has a deontic marker that occurs between the subject
and the verb and displays the form ní (hence ní2) as illustrated in (55a).
However, this example also shows that Fongbe is different from Gungbe in
that the purpose subordinator or conditional marker in (55b) has the form nú,
which is also identical to the dative preposition shown in (55c).
(55) a. Ùn jló nú à ní wá. [Fongbe; Anonymous 1983: X2]

1SG want nú 2SG ní2 come
‘I want that you come.’
b. Nú à kò wá sɔ̀ ɔ́ mí ná wà àzɔ́ n.
nú 2SG ADV come tomorrow DET.CLAUSE 1PL FUT do work
‘When you come tomorrow, we will work.’
c. É ɖɔ̀ xó nú mì kaka.
3SG say word nú 1SG a.lot
‘He talked to me a lot.’
As already discussed in the literature, the Fongbe dative marker displays nú
versus ná alternation whenever the complement of the preposition is displaced
for question or focus purposes (Da Cruz 1992, 1997; Aboh 2003a, 2009b). This
is illustrated by examples (55d–f). In (55d), only the form nú is allowed,
similarly to (55c). In (55e) and (55f), where the complement has been
preposed for wh-question and focusing, only the form ná is allowed.
(55) d. Kɔ̀ kú sà mɔ̆ tò ɔ́ nú/*ná Àsíbá.
Koku sell car DET to/for Asiba
‘Koku sold the car to/for Asiba.’
e. Mε̆ wὲ Kɔ̀ kú sà mɔ̆ tò ɔ́ ná/(*nú)?
who FOC Koku sell car DET to/for
‘To/for whom did Koku sell the car?’
f. Àsíbá wὲ Kɔ̀ kú sà mɔ̆ tò ɔ́ ná/(*nú).
Asiba FOC Koku sell car DET to/for
‘Koku sold the car to ASIBA.’
Accordingly, both Gungbe and Fongbe display ná as a dative preposition,
though these forms do not have the same distribution. With regard to the form
nú, we observe that the Fongbe conditional/irrealis marker and the dative
preposition are homophonous. These forms are different from the deontic
(subjunctive/injunctive) marker ní2.
Ewegbe (Western Gbe) shows a different pattern. In this language, it is the
deontic marker which shares the same form as the dative preposition. In
example (56a), the conditional/irrealis marker is realized as né, which
contrasts with the subjunctive/injunctive marker ná in (56b). The latter
appears identical with the dative preposition in (56c).10
(56) a. Né Kofi vá lá mi-yɔ́ -m.
né Kofi come TOP 2PL-call-1SG
‘If Kofi comes you all (should) call me.’
10
I’m grateful to James Essegbey for providing me with these examples and for sharing with me
his thoughts about aspects of the analysis presented here.
b. É-dze bé Kofi ná-vá.

3SG-be.appropriate that Kofi ná-come
‘It is convenient that Kofi comes.’
c. Kofi ƒo tó-me ná Komi.
Kofi hit ear-inside for Komi
‘Kofi slapped Komi.’
Finally, Gengbe (Western Gbe) completes the picture with the pattern
illustrated in (57), where the preposition and the two mood markers (i.e., the
subjunctive/injunctive and the conditional/irrealis) have the same form né:
(57) a. Né Kòfí vá á mì yrɔ̀ -ὲ.
né Kofi come TOP 2PL call-3SG
‘If/when Kofi comes, call him.’
b. Kɔ̀ jó glɔ̀ n bé Kwésí né vá.
Kojo said that Kwesi né come
‘Kojo said that Kwesi should come.’
c. Kòfí pò tómὲ né Kɔ̀ jó.
Kofi hit ear-in né Kojo
‘Kofi slapped Kojo’s face.’
Table 6.1 recapitulates the data presented thus far. Looking at the solid lines in
contrast to the broken lines, Table 6.1 shows that the Gbe languages display a
function syncretism between the dative prepositions and mood markers. While
this syncretism is not immediately visible on the basis of individual languages
(e.g., Gungbe), it is obvious once we look at the variation between Eastern and
Western Gbe languages. Western Gbe languages exhibit an overlap between the
dative preposition and the expression of injunctive and conditional. Fongbe and
Gungbe have a different form for the dative preposition: ná. But this form is
found in Ewegbe as well where it encodes dative and injunctive. Ewegbe further
uses the form né for conditional. This same form is used in Gengbe for the
Table 6.1 Syncretism of function and form between the dative

preposition and the deontic modality
Conditional/Irrealis Subjunctive/Injunctive Preposition

Gengbe né né né
Fongbe nú ní nú ná
Ewegbe né ná ná
Gungbe ní ní ná
expressions of dative, injunctive, and conditional. Only the Fongbe and Gungbe
form ní seems to be an innovation in this Gbe context.
A plausible explanation for this syncretism is that the mood markers and the
prepositions developed from the same form. It is traditionally assumed that the
dative preposition developed from the verb ‘give’ which is realized as ná
across Gbe. Two additional assumptions are needed. First I adopt without
discussion Kayne’s (2002) proposal that prepositions are merged within the
extended domain of the VP (i.e., VP-externally). Second, I follow Roberts and
Roussou’s (2003) proposal that grammaticalization is the reflex of head
movement where the element being grammaticalized moves from the lexical
domain to the functional domain and subsequently acquires the property of
being directly merged in the functional domain. (See Aboh 2009b and
Chapter 7, this book, for a discussion of serial verb constructions.) Put in the
context of the sequencing in (42), these hypotheses lead to the characterization
in (58), in which the lexical verb evolves into a preposition and then into a
subjunctive/injunctive maker and eventually into a conditional/irrealis marker.
Incidentally, it is worth mentioning that in Gungbe, the prospective aspect
marker and the future marker also have the form ná, which could have emerged
from the same source. As the structure shows, each realization corresponds to a
different syntactic position:
(58) Conditional/Irrealis (Gbe)
n-V[-e, -u, -i]

Subjunctive/Injunctive (Gbe)
n-V[-e, -a, -i]

Negation
Tense (Gungbe)
n-V[ -a]
Prospective aspect (Gungbe)
n-V[ -a]
Dative Preposition (Gbe)
n-V[-e, -a, -u]

Verb (Gbe)
na
Assuming this analysis or some variant thereof is correct, the homophony

observed between Saramaccan fu1 and fu2 cannot be accidental. Instead, it is
reasonable to propose that this sort of homophony is rooted in learning
hypotheses of speakers of languages which exhibit the syncretism of function
6.2 Topic, focus, and interrogation in Saramaccan 249
and form described in Table 6.1, which itself arguably results from the
grammaticalization path described in (58). This view is in support of the
process of recombination of syntactic features developed in this book.
In addition, these facts allow us to understand the role of interfaces in the
development of new linguistic patterns. Our discussion has shown that even
though the Gbe languages show a certain degree of variation with regard to
the forms selected, these languages exhibit a syncretism between dative
prepositions, complementizer-prepositions, and deontic mood markers. In all
these languages, the similar forms are used to express these distinct
grammatical functions. Because the complementizer prepositions and the
deontic mood markers relate to the complementizer domain which has been
shown to be vulnerable during acquisition in a multilingual context, this
contributes to favor cross-linguistic influence between Gbe and English such
that the functions of n-V2 in Gbe were extended to cover the usage of English
to in modal contexts.
6.2 Topic, focus, and interrogation in Saramaccan

A crucial aspect of the discussion in Chapters 4 and 5 is that interfaces are
particularly vulnerable to feature competition. Given such a hypothesis in the
context of the emergence of the complementizer system in Saramaccan, we
predict, on the basis of the properties of fu1 and fu2, that this language should
manifest other significant Gbe properties within its complementizer system
(though not necessarily in other domains of its grammar). More precisely, the
contrast between, on the one hand, the Saramaccan and Gungbe examples in
(52a–b) and, on the other, the English example (52c) suggests that both
Saramaccan and Gungbe display similar properties with regard to topic and
focus constructions. As shown in these examples, Gungbe and Saramaccan
allow the fronting of topic and focus constituents in embedded clauses (52a–b),
contrary to English, which generally disallows such constructions, as indicated
again in example (59):
(59) *I think that, as for John, the car he bought.
The fact that similar constructions are freely available in Gungbe and
Saramaccan (to the exclusion of English) is clearly not accidental but
emerges from the properties of the clausal left periphery in these languages
as described in (43). In order to see this, let us consider topic, focus, and
interrogation in Saramaccan more closely.
As explained in Aboh (2006b, 2007a) and references therein, English,
Saramaccan and Gungbe display topic strategies in main clauses in which the
topic is fronted in clause-initial position and repeated clause-internally by a
resumptive pronoun. This is illustrated again by the examples under (60):
(60) a. As for the chicken, I cooked it. [English]

b. [Dí ganía] dέ mi bói en. [Saramaccan]
DET chicken TOP 1SG cook 3SG-ACC
‘As for the/this chicken, I cooked it.’
c. Kòkló lɔ́ yà ùn ɖà ὲ. [Gungbe]
chicken DET TOP 1SG cook 3SG
‘As for the/this chicken, I cooked it.’
The following difference is noteworthy: While Gungbe and Saramaccan
exhibit a topic marker, yà and dέ, respectively, which follows the
topicalized constituent, English lacks such a marker. Another difference,
already mentioned in the previous paragraph, is that topic constructions
can be embedded in Saramaccan and Gungbe but not in English (52).
However, it is interesting that Saramaccan and its source languages (i.e.,
Gbe and English) display long extraction, as indicated by the following
examples:
(61) a. Mi táki táa dí bakúba dέ Amato bói en.
1SG say that DET banana TOP Amato cook 3SG
‘I said that, as for the banana, Amato cooked it.’
b. dí bakúba dέ mi táki táa Amato bói en.
DET banana TOP 1SG say that Amato cook 3SG
‘As for this banana, I said that Amato cooked it.’
When it comes to wh-questions, Saramaccan displays a notable peculiarity: a
set of its wh-words are derived from the Gbe languages. Smith (1987, 1999)
argues that the Saramaccan wh-words andí ‘what’ and ambὲ ‘who’ derive from
the Fongbe wh-words àní ‘what’ and mέ , the shortened form of mέ -nù ‘who’ in
Fongbe and Gungbe.11 Even more remarkable are the data discussed in Smith
(1996), which show that Saramaccan displays contrastive focus constructions
which are marked by the element wὲ , which serves an identical function in
Fongbe and Gungbe. (See Aboh 2004a; and Aboh and Essegbey 2010 and
references cited therein for focus constructions in Gbe and Kwa in general.)
These facts are illustrated by the Saramaccan wh-question in (62a) followed by
the Fongbe and Gungbe equivalents. While the Fongbe example shows
complete parallels with Saramaccan regarding the focus marker and the
wh-word, Gungbe completes the picture in showing that these three
languages use the same focus marker.
11
According to Smith and Cardoso (2004), the reanalysis of simple nasal consonants into
prenasalized consonants in Saramaccan does not only affect words of Gbe origin but also
Germanic and Romance words. For instance, English words such as make and knife are realized
as mbei and ndefi, while words of Portuguese origin such as anel ‘ring’ and demolir ‘demolish’
are realized as andélu and dimbolí.
6.2 Topic, focus, and interrogation in Saramaccan 251
(62) a. Andí wὲ i bói? [Saramaccan]

what FOC 2SG cook
‘What did you cook?’ [Smith 1996: 117]
b. Étέ wὲ à ɖà? [Gungbe]
what FOC 2SG Cook
‘What did you cook?’
c. Àní wὲ à ɖà? [Fongbe]
what FOC 2SG cook
Just as in the case of Gbe, this focus marker is required in Saramaccan
contrastive focus constructions. As already shown by Aboh (2004a, 2010e,
forthcoming), such constructions are necessarily licensed in Gbe by fronting
the focused constituent immediately to the left of the focus marker.
(63) a. [dĺ mujεε] wὲ mi bi bεl, naa dí wómi. [Saramaccan]
DET woman FOC 1SG PAST call not DET man
‘I phoned the woman, not the man.’ (Smith 1996: 118)
b. Náwè lɔ́ wὲ ùn yrɔ́ mă yrɔ́ dáwè lɔ́ . [Gungbe]
woman DET FOC 1SG call 1SG.NEG call man DET
‘I called the woman, not the man.’
It appears from these examples that both Saramaccan and its Gbe source
languages display ex situ wh-question and focus constructions in which the
wh-phrase or the focused constituent is fronted to a position immediately to the
left of the focus marker wὲ. I therefore conclude that Saramaccan and the Gbe
languages display a similar morphosyntax with regard to the clausal left
periphery, contrary to English, which does not exhibit any of the discourse
markers discussed here.
This conclusion is further supported by data related to direct yes–no questions.
As the examples in (64a–b) show, Saramaccan patterns like the Gbe languages in
displaying a clause-final yes–no question particle, unlike English, in which such
questions involve subject–auxiliary inversion, hence the contrast between the
English translation of the Saramaccan and Fongbe questions and the English
example in (64c) which appears to be possible in some colloquial varieties only.
(64) a. A bói dí ganía nó? [Saramaccan]
1SG cook DET chicken INTER
‘Did you cook the chicken?’
b. À ɖà kòkló ɔ́ à [Fongbe]
2SG cook chicken DET INTER
‘Did you cook the chicken?’
c. %You cooked the chicken?
We now reach a characterization according to which the Gbe languages

and Saramaccan display a series of discourse markers (e.g., topic, focus,
interrogative) with a specific syntax: these scope-taking elements must be
right-adjacent to the element they have scope over. According to Aboh
(2004a, 2004b, 2010e, forthcoming), these markers belong to the clausal left
periphery but can cluster to the left or right periphery depending on the element
within their scope. When these markers have scope over a constituent inside the
proposition, the latter must be fronted next to the relevant marker. On the other
hand, when the marker has scope over the whole proposition, the clause as a
whole must front. This view explains why in the appropriate context, the focus
and topic markers can also occur in the right periphery, similarly to the yes–no
question marker. Consider the following examples taken from Rountree and
Glock (1982).
(65) a. [[Nöö fa I si I ta waka naa köni
And as 2SG see 2SG PROG walk LOC careful
futu] de] so seepi tu wan oto sembe ta
foot TOP thus EMPH also one other person PROG
waka naa köni futu.
walk LOC careful foot
‘And just as you are walking cautiously around, so someone else is doing the
same.’
b. [[Hen mbei te i si soni a matu] wee]
3SG make when 2SG see something LOC jungle FOC
I musu luku bunu bé i sa un mbeti.
2SG must look well to.permit 2SG know which beast
‘That’s the reason why when you see something in the wood, you must
look carefully so that you know which animal it is.’
[Rountree and Glock 1982: 183]
In these Saramaccan examples, the topic and focus markers take scope over
the clause as the result of which they occur sentence-finally similarly to their
Gungbe equivalents yà and wὲ in (66).
(66) a. [[Ní À zɔ̀ n dὲdέ lé] yà] kànlìn lὲ mă hɔ̀ n [Gungbe]
ní1 2SG walk slow this.way TOP animal PL NEG.FUT escape
‘If you walk slowly this way, the animals will not escape.’
b. Context: What happened?
[[Kòfí zɔ̀ n gìdìgìdì] wὲ ] bɔ̀ kànlìn lὲ hɔ̀ n
Kofi walk noisily FOC COORD animal PL escape
‘Kofi walked noisily and the animals escaped.’
It appears from these examples that Saramaccan and the Gbe languages
share C-type discourse markers as well as several properties within the
complementizer system that make them typologically similar to Kwa
6.3 Predicate fronting and doubling 253
languages in general (Aboh and Essegbey 2010), and which distinguish them
from English. The data discussed thus far lead to the characterization in (67) for
the three languages:
Force TopicP
Discourse–Syntax
interface
Topic FocusP
d ; ya
Focus FinitenessP
w
Finiteness
Proposition
This representation shows that while all the languages involve relevant markers
to realize Force (English: that/for; Saramaccan: táa/fu1; Gungbe: ɖɔ̀ /ní1) and Fin
(English: to; Saramaccan: fu2; Gungbe: ní2), only Saramaccan and Gungbe
display topic markers (Saramaccan: dέ ; Gungbe yà) alongside the focus marker
(Saramaccan/Gungbe: wὲ ). It is therefore clear that the backbone of the
complementizer system in Saramaccan follows the Gbe pattern, though the
relevant markers embed semantic (and morphosyntactic) properties of certain
English complementizers (e.g., for, to). While the architecture of the
complementizer system as proposed by Rizzi (1997) is provided by UG, the
discussion here shows that the Gbe languages combined with English to trigger
the emergence of very specific properties of the Saramaccan clausal left periphery.
In the following section, I discuss predicate fronting with doubling which further
enhances the similarities between the Surinamese creoles and the Gbe languages.
6.3 Predicate fronting and doubling

In this section, I focus on constructions which have been referred to as
‘predicate clefts’ in the literature and are illustrated in (68). Aboh (2006c)
argues in detail that this term is inadequate, hence the expression of predicate
(or verb) fronting with doubling, as illustrated below.
(68) a. Nján Amato bí nján dí bakúba. [Saramaccan]
eat Amato PAST eat DET banana
‘Amato ATE the banana.’
b. Đù Asíbá ɖù àkwékwé lɔ́ . [Gungbe]
eat Asiba eat banana DET
‘Asiba ATE the banana.’
Like other Caribbean creoles (e.g., Haitian), Saramaccan also displays

relativized event constructions of the type in (69a) where the fronted verb
(phrase) is nominalized. Comparable constructions are also found in Gbe as
indicated by the Gungbe equivalent in (69b). I return to these constructions in
Section 6.3.2.
(69) a. Dí Nján dí míi bí ta nján dí bakúba. [Saramaccan]
DET eat DET child PAST PROG eat DET banana
a toóbi mi-seei.
3SG annoy 1SG-EMPH
‘The fact that the child ate the banana annoyed me really.’
b. Ðù ɖĕ ví lɔ́ ɖù àkwékwè lɔ́ [Gungbe]
eat REL child DET eat banana DET
vέ ná mì.
hurt for me
‘The fact that the child ate the banana hurt me a lot.’
An interesting aspect of these constructions is that, even though there is
doubling of the predicate, similarly to the examples in (68), the reading
obtained is not that of focus on the verb but rather a factive reading. Aboh
(2005d, 2006c, forthcoming), Aboh and Dyakonova (2009) discuss the
Gungbe constructions in (68b) and (69b), while Glaude and Zribi-Hertz
(2012) provide a detailed analysis of similar constructions in Haitian. In
what follows, I only focus on aspects of the discussion in Aboh (2005d),
Aboh and Dyakonova (2009), and Aboh (forthcoming) that are relevant to
the position being developed in this book regarding competition and
selection. I start with predicate focus constructions in Saramaccan and
Gungbe.
6.3.1 Predicate focus and doubling in Saramaccan and Gungbe

A striking property of predicate focus with doubling in Saramaccan is that
unlike most Caribbean creoles, including Sranan, the fronted verb is not
preceded by a focus marker. This can be seen by contrasting the Saramaccan
example in (68a) to the following sentences in Jamaican Creole (70a), in
Haitian (70b), and in eighteenth-century Sranan (70c).
(70) a. A tiif yu tiif di mango-dem. [Jamaican Creole]
FOC steal 2SG steal the mango-PL
‘You STOLE the mangoes!’ [Durrleman-Tame 2008: 111]
b. Se mache Bouki te mache, li pa te kouri [Haitian Creole]

FOC walk Bouki PAST walk 3SG NEG PAST run
‘Bouki had walked, not run.’ [DeGraff 2007: 113]
c. Da koksi ju koksi mi [Sranan]

FOC taunt 2SG taunt 1SG
‘You are really taunting me.’ [Schumann 1783: 154, cited in Arends 2009: 101]
In all these languages (but not in Saramaccan and Gungbe) a copula-like

element is used to mark the fronted wh-word or focused constituent. This
marker must precede the focus in most Caribbean creoles. In Saramaccan
and Gungbe, however, the focus marker is formally different from a copula
and it must follow the focused phrase. Compare the following focus
expressions involving a noun phrase or a wh-phrase in Jamaican Creole
and Haitian Creole with the Saramaccan and Gungbe examples presented
in (62) and (63).
(71) a. A wa im tek? [Jamaican Creole; Durrleman-Tame 2008: 111]
FOC what 3SG take
‘What did he take?’
b. A di mango Pita tiif.
FOC DET mango Peter steal
‘Peter stole THE MANGO.’
c. Se ki moun ou ye? [Haitian Creole; DeGraff 2007: 115]
FOC which person 2SG be
‘Who are you?’
d. Se Poutin ki prezidan Risi.
FOC Poutin COMP president Russia
‘It is Putin who is the president of the Russia.’
These facts lead to the characterization in (72a) for Gungbe and Saramaccan
as opposed to (72b) for Haitian, Jamaican, and other related Caribbean creoles.
(72) a. [[XPFOCUS CONSTITUENT . . . . .] wὲFOCUS MARKER] [Saramaccan/Gungbe]
b. [se/a FOCUS MARKER [XPFOCUS CONSTITUENT . . . . .]] [Jamaican Creole/Haitian Creole]
A simple surface analysis of this contrast would be to say that the head that
triggers focus reading precedes its focus complement in Jamaican Creole and
Haitian Creole, while it follows in Gungbe and Saramaccan. This in turn would
suggest that, word order aside, there is no fundamental distinction between the
marker wὲ in Saramaccan and Gungbe and elements like a, or se in Jamaican
Creole and Haitian Creole, respectively.
Yet, in a detailed comparative analysis of the marker a in Jamaican Creole,
Atlantic creoles, and Kwa languages such as Yoruba and Gungbe, Durrleman-
Tame (2008: 119) shows that there is a crucial distinction between these two
types of markers: Jamaican a, precedes the focused element because it is an
exhaustive marker that takes the focus constituent as complement. On the other
hand, the Gbe or Kwa-type marker does not require exhaustive reading and can
be used in a wide range of contexts, including presentational focus (Aboh
2006c, forthcoming). According to Durrleman-Tame’s analysis, the different
patterns in (72) translate into two distinct structures, as shown in (73):
(73) Saramaccan/Gungbe Jamaican/Haitian
Exh(austive)P
FocP
spec Exh¢
spec Foc¢
XPFOCUS Exh FocP
Foc FinP se/a
w spec Foc¢
XPFOCUS XPFOCUS
Foc FinP
XPFOCUS
According to Durrleman-Tame’s (2008) analysis, both Kwa and Atlantic

creoles display movement of the focused phrase (here XPfocus) to the
specifier of the focus phrase FocP. But there are two differences. In
Kwa, FocP is headed by a focus marker which can express contrast or
exhaustivity in context, but it is not formally endowed with these features.
As explained by Aboh (forthcoming), this would mean that only the
feature [focus] is operative in these languages. In Jamaican Creole and
other Atlantic creoles, however, FocP does not have an overt head.
Instead, this functional projection is licensed under an Exh(austive)
phrase which is headed by the exhaustive marker. This would mean that
the focused constituent in [spec FocP] acquires exhaustive reading
because it is under the scope of the marker heading Exh(austive)P. That
Jamaican Creole and other similar languages involve exhaustive reading
in focus constructions is in my opinion related to the fact that these
constructions embed Kwa-type Focus Raising (i.e., fronting of the
focused constituent) alongside properties of Germanic and Romance
cleft constructions which, as Kiss (1998) convincingly argues, typically
exhibit exhaustive reading. That this is indeed the case can be easily
demonstrated by the fact that in all these creoles, the exhaustive marker
is homophonous with the equative copula (DeGraff 2007; Durrleman-
Tame 2008). This would mean that focus structures in Caribbean creoles
emerge from a combination of Focus Raising of the Kwa-type and cleft
constructions from Romance and Germanic. In this regard, Glaude and
Zribi-Hertz (2012: 131) conclude their analysis of predicate-doubling in
Haitian Creole observing that:
Among the focus-marking strategies which may combine in the derivation of verb focus
doubling constructions in Haitian, low contrastive reiteration, clefting and adverbial
focus markers are all available in French . . . Focus Raising (i.e., the raising of a focused
constituent to the clause periphery, without clefting) is not, on the other hand a natural
focus-marking strategy in French, where focus is generally phrase-final, especially for
verbs . . . This suggests that the one non-French property involved in the diachronic
development of Haitian verb focus with doubling was Focus Raising, a property
characteristically attested in Gbe.
Assuming this analysis is correct, I conclude that the contrast in (72) does
not illustrate a simple surface variation but hides a structural difference that
sets the Gbe (Kwa) languages and Saramaccan apart from other Caribbean
creoles.
A question that has been extensively discussed in the literature is how to
account for the genesis and the structure of the doubling structures in (68), (69),
and (70). I will not review the literature here but I refer the interested reader to
Glaude and Zribi-Hertz (2012) for a recent state of the art as well as a new
alternative. With regard to the genesis of such structures in Atlantic creoles,
these were considered typical of Atlantic creoles and certain Niger-Congo
languages (e.g., Kwa). As such they were seen as empirical evidence for
substrate influence. Though these constructions are highly productive in
Kwa, it is important to realize that they are also present in typologically
diverse languages such as Germanic, Romance, Semitic, Slavic, and Semitic
(Aboh and Dyakonova 2009). Accordingly, that such structures are attested in
Atlantic creoles cannot simply be claimed to illustrate substrate influence.
Instead, we need to adopt Glaude and Zribi-Hertz’s (2012) more nuanced
approach where only certain properties of these structures can be related to
the relevant African languages. Building on Aboh (2005d) and Aboh and
Dyakonova (2009), I assume that Predicate Fronting (with doubling) for the
purpose of focusing and event-relativization is such a relevant property.
Starting with the structural account for doubling, Aboh and Dyakonova
(2009) argue, on the basis of Gungbe and Russian, that sequences such as in
(68) result from parallel chains. Technical details aside, the argument goes as
follows: In Gbe languages of the Gungbe-type, the verb must be licensed for
aspect, which requires the verb root to move to the immediately relevant aspect
position. This creates a first chain Vasp-V. In the context of verb focus, the verb
root is simultaneously attracted by an aspect head within the INFL domain as
well as the focus head within the clausal left periphery, thus creating a second
chain Vfocus-V. The tail of these two chains is deleted at spell out, thus leading
to the phonetic realization of the two heads Vfocus and Vasp, as in (74)
(Chomsky 1995, 2008; Aboh 2009c).
(74)
FocP
spec Foc¢
Foc FinP
VFOCUS
TP
AspP
spec Asp¢
VASP VP
In addition to accounting for various properties of predicate fronting with

doubling (e.g., the impossibility of long extraction, the incompatibility with
intervening negation), this analysis provides us with a new perspective on
factive events of the type in (69), which Aboh (2010f) regards as a subcase
of (74). The only difference is that the attracting element is the relative head.
(However, see Glaude and Zribi-Hertz 2012 for an alternative analysis.) To see
this, let us consider again the factive constructions in relation to focus
expressions in Gungbe and Saramaccan.
6.3.2 Event-relativization and doubling in Gungbe and Saramaccan

In relation to the examples in (68) and (69), I mentioned that the two sets of
examples have a different readings, viz., focus versus factive, even though
they both involve fronting of the predicate with doubling. There is, however, a
crucial morphological and structural difference between these sets of
examples. Consider again the following Saramaccan and Gungbe examples
of verb focus:
(75) a. Lúku a tà lúku dì míi tá kó a lío.
watch 3SG PROG watch DET child PROG come LOC river
‘He is watching the child coming from the river.’ [Saramaccan; Byrne 1987: 58]
b. Kpɔ́ n é kpɔ́ n ví ɖĕ wá tɔ̀ tó.
look 3sg look child rel come river bank
‘S/he looked at the child who came to the river bank.’
Let us now compare the examples in (75) to the factive sentences in (69)
repeated here.
(76) a. Dí nján dí míi bí ta nján dí bakúba [Saramaccan]

DET eat DET child PAST PROG eat DET banana
a toóbi mi-seei.
3SG annoy 1SG-EMPH
‘The fact that the child ate the banana annoyed me really.’
b. Ðù ɖĕ ví lɔ́ ɖù àkwékwè lɔ́ [Gungbe]
eat REL child DET eat banana DET
vέ ná mì.
hurt for Me
‘The fact that the child ate the banana hurt me a lot.’
In the Gungbe and Saramaccan examples in (75), the fronted verb is bare
suggesting that this element is a root. In such constructions the default
interpretation is that of verb focus. In the Saramaccan example (76a),
however, the fronted verb is nominalized and co-occurs with the determiner
dí. In the Gungbe example (76b), the nominal character of the fronted verb is
indicated by the presence of the relativizer ɖĕ.12 As shown by the interpretation,
these examples require a factive reading similar to French and English
constructions of the type in (77).
(77) a. Le fait que Jean ait vendu la maison m’étonne.
the fact that John have.SUBJUNCTIVE sold the house 1SG.surprise
b. The fact that John sold the house surprised me.’
Both French and English also allow similar constructions where the phrase
“the fact” is unpronounced.
(78) a. Ø Que Jean ait vendu la maison m’étonne.
b. Ø That John sold the house surprised me.
In discussing these examples, Aboh (2005d, 2010f) argues that the examples
in (75) and (76) involve distinct structural properties that feed on doubling
structures. While (75) involves focus constructions along the lines of (74), the
sentences under (76) involve event relativization where the event head (i.e., the
verb) functions as heads of a relative clause. Under this description, the fronted
(or relativized) verb in Gbe and Saramaccan serves the same function as the
expletive phrase “the fact” in French and English. (However, see Glaude and
Zribi-Hertz 2012 for an alternative analysis.) In order to see this, let us first
consider certain properties of event relatives in (76). The demonstration here
12
Mufwene (1987: 80) discusses examples of related predicate focus constructions in Bantu
where the focused VP is marked “as a verbal noun.” Such findings obviously reinforce the
view developed here where in a situation of competition and selection, the Gbe patterns are
reinforced by congruent Bantu patterns (see also Ansaldo 2009).
draws on Gungbe but I assume that comparable facts can be demonstrated for
Saramaccan.
1. As noted above, event relatives involve the relative marker in Gungbe. In
addition, these constructions exclude the focus marker. Examples (79b–c)
are ungrammatical whether the focus marker precedes or follows the rela-
tivizer. Example (79d) indicates that the presence of the focus marker is
excluded even in contexts where the relative marker is not overtly realized.
Put together, these examples show that event relativization is formally
distinct from verb focusing.
(79) a. Wlé *(ɖĕ) mí wlé àgásá ɖàxó lɔ́ lέ vέ ná Kòfí.
catch REL 1PL catch crab big DET PL hurt for Kofi
‘The fact that we caught the big crabs annoyed Kofi.’
b. *Wlé wὲ ɖĕ mí wlé àgásá ɖàxó lɔ́ lέ vέ ná Kòfí
catch FOC REL 1PL catch crab big DET PL hurt for Kofi
c. *Wlé wὲ ɖĕ mí wlé àgásá ɖàxó lɔ́ lέ vέ ná Kòfí
catch FOC REL 1PL catch crab big DET PL hurt for Kofi
d. *Wlé wὲ mí wlé àgásá ɖàxó lɔ́ lέ vέ ná Kòfí
catch FOC 1PL catch crab big DET PL hurt for Kofi
2. A typical property of the Gbe languages is that they involve a clausal
determiner, that is, an article that marks nominalized clauses. In Gungbe
this article is lɔ́ which also marks noun phrases as indicated in (80)
(Lefebvre 1998; Lefebvre and Brousseau 2002; Aboh 2004a).
(80) a. Àgásá ɖàxó lɔ́ nyɔ́ n tàun.
crab big DET good very
‘The big crab is very nice.’
b. Ðě Àgàsú hɔ̀ n yì lɔ́ má nyɔ́ n kpálí.
REL Agasu flee go DET NEG good at.all
‘That Agasu fled is not nice at all.’
Because event relatives are nominal clauses, they typically involve this
determiner. This is not the case with simple verb focus constructions, which
tend to resist this determiner.
(81) a. [[Hɔ̀ n ɖĕ Àgàsú hɔ̀ n yì] lɔ́ ] vέ ná Kòfí.
flee REL Agasu flee go DET annoy for Kofi
‘That Agasu fled annoyed Kofi.’
b. *[[Hɔ̀ n Àgàsú hɔ̀ n yì] lɔ́ ]
flee Agasu flee go DET
‘Agasu FLED.’
These examples further indicate that event relativization is a subcase of

relative clauses illustrated in (82a). In Chapter 5, I proposed that such a Gbe
relative clause can be represented as in (82b), in which the relativized noun is
raised from the clause-internal position to the specifier position of the relative
CP. Subsequently, the whole relative clause CP is raised to [spec NumP] and
[spec DP], respectively (Chapter 5, Aboh 2004a, 2004c, 2006a). A crucial
aspect of this analysis is that the determiners that relate to the relativized noun
head occur to the far right of the relative clause, as schematized in (82c).
(82) a. Kòfí xɔ̀ [àgásá ɖàxó [ɖĕ mí wlé] lɔ́ lέ]
Kofi buy crab big REL 1PL catch DET PL
‘Kofi bought the big crabs that we caught.’
b.
c. [Relativized NP-[CP Relative Clause]- lɔ́ lέ]

As explained in Aboh (2002, 2005d), a fascinating aspect of relative
clauses that applies to event-relativization is that the factive event reading
is obtained by manipulating the surface position of the determiner. Consider
the following variant of the bracketed relative clause in (82a) in which the
relativized head àgásá is immediately followed by the determiner lɔ́ lέ, as
schematized in (82b).
(83) a. [àgásá ɖàxó] lɔ́ lέ ɖĕ mí wlé.
crab big DET PL REL 1PL catch
‘The fact that we caught the crabs.’
*‘The big crabs that we caught.’
b. [[Relativized NP- lɔ́ lέ]-[CP Relative Clause]––]
As is shown by the translation of this example, manipulation of the ordering
of determiners inside the relative clause (82c vs. 83b) yields the loss of the
relative meaning. The resulting sentence is thus interpreted as a factive event.
3. While example (83a) clearly underscores the relation between event relatives and
nominal relative clauses, the two constructions exhibit distinct structural proper-
ties. Consider for instance, the following sentence which involves both the
relative clause and the event relative.
(84) Àgásá lɔ́ [ɖĕ Kòfí wlé] nyɔ́ n, àmɔ́ n àgásá lɔ́ kpàkpà má nyɔ́ n.
crab DET REL Kofi catch good but crab DET itself NEG good
‘The fact that we caught the crab is welcome but the crab’s meat is not good.’
If event relatives were simple relative clauses, (84) should be a contradiction

with both clauses denying each other. Instead, the glosses and translation
indicate that the first event relative refers to the Event of catching the crab,
while the second part of the clause suggests that the meat of the crab was not
delicious.
Another difference between nominal relative clauses and event relatives is
that nominal relatives typically occupy the same syntactic positions as
common noun phrases (viz., as external or internal arguments, or as
adjunct), while event relatives function only as clausal subjects. The
following examples show that while nominal phrases can function as
complements, event relatives cannot. Accordingly, (85b) is ungrammatical,
unlike example (84). For (85b) to be grammatical the factive event has to
become a subject as in (85c).
(85) a. Súrù sà [àgásá ɖĕ Kòfí wlé lɔ́ ].
Suru sell crab REL Kofi catch DET
‘Suru sold the crab that we caught.’
b. *Súrù gbέ [àgásá lɔ́ ] ɖĕ Kòfí wlé]].
Suru refuse crab DET REL Kofi catch
‘Suru did not accept the fact that we caught the crab.’
c. [àgásá lɔ́ ] ɖĕ Kòfí wlé] má yì xέ Súrù kpálí.
crab DET REL Kofi catch NEG go with Suru at.all
‘The fact that we caught the crab did not please Suru at all.’
Furthermore, while nominal relative clauses involve a wide range of
constituents, event relatives involve the verb or the internal argument but
exclude the external argument. This is shown by sentences (86):
(86) a. [[Gán lɔ́ ] ɖĕ náwè lɔ́ kàn wé xlán] vέ ná mì.
chief DET REL woman DET write letter to annoy for me
‘The fact that that woman wrote a letter to THE CHIEF annoyed me.’
b. [[Wé lɔ́ ] ɖĕ náwè lɔ́ kàn xlán ògán lɔ́ ]] vέ ná mì.
letter DET REL woman DET write to chief DET annoy for Me
‘The fact that the woman wrote THAT LETTER to the chief annoyed me.’
c. ??/*[[Náwè lɔ́ ] ɖĕ kàn wé xlán ògán lɔ́ ] vέ ná mì.
woman DET REL write letter to chief DET annoy to Me
‘The fact that that woman wrote a letter to the chief annoyed me.’
Sentences (86) are very illustrative of the properties of event relatives in that
they show that while the factive reading can be obtained by fronting the internal
arguments (including double objects) or the verb expressing the Event head,
fronting of the external argument does not yield the same result.
According to Aboh’s (2005d) detailed analysis, the contrast in (86) is
indicative of the fact that event relativization does not target the verb root
(similarly to focus constructions) but rather targets the event phrase (i.e.,
minimally the verb phrase). Thus, the fronted internal DP argument and the
verb heading the event relate to aktionsart and can therefore be regarded as part
of a functional projection denoting the event. This would mean that the internal
argument or the verb bears an event feature that can be attracted by a head
within the clausal periphery. Consistent with this hypothesis, a factive clause
such as the bracketed sequence in (85c) can be represented as in (87), in which
CP is a shorthand for the articulated structure in (34). In this example, the verb
raises to Asp to be licensed for aspect, while the DP object is attracted within
the CP field for relativization.
(87) CP
spec C¢
DPOBJECT
C TP
AspP
spec Asp¢
VASP VP
V
DPOBJECT
On the other hand, when the event head (viz., the verb) is attracted, I
hypothesize that it is the remnant VP that is fronted (after the DP object has
raised to its licensing position). A possible argument in favor of this hypothesis
is that the fronted relativized event can be combined with a determiner as in the
Saramaccan example (76a). As discussed in Collins (1994), similar examples
are found in Fongbe and other Kwa languages as well. This would mean that a
factive clause with a fronted verb can be represented as in (88), ignoring the
intermediate movement of the object to its licensing position.
(88) CP
spec C¢
VP
C TP
AspP
spec Asp¢
VASP VP
As in the case of focus constructions, doubling here results from the spell out
of two heads of two parallel chains that have a common tail inside VP. The
analysis proposed here indicates that factive clauses involving fronting of the
DP internal argument and the remnant VP fundamentally share the same
structure, though they are felicitous under different pragmatic conditions.
Indeed, factive clauses with verb fronting and doubling generally tend to
imply emphasis and topicality, but I hope to return to these semantic
differences in future work.
What matters for the present discussion is that if we accept to extend the
analysis just proposed to the Saramaccan example in (76a) as well as to the
French and English constructions in (77) and (78), we reach the typological
conclusion according to which Gungbe and Saramaccan involve internal
argument or remnant VP fronting (with doubling) while English and French
display insertion of an expletive factive phrase ‘the fact’ or its covert
equivalent. Following Aboh’s (2005d) analysis of such English and French
constructions, I therefore conclude that the examples in (77) and (78) can be
represented as in (89):
(89) CP
spec C¢
XPTHE FACT/Ø
C TP
that/que
AspP
VP
As is clear from this characterization, no movement operation is involved in

the derivation of such constructions in Romance and Germanic. Instead, these
languages resort to the first merge of an expletive phrase. More concretely, we
are led to the following typological variation between Kwa/Saramaccan and
French/English:
(90) a. Kwa/Saramaccan (Remnant VP-fronting with doubling: focus)
[CP [VP REMNANT VERBAL PREDICATE] [C° RELATIVE MARKER [IP . . .VERB COPY . . .]]]
b. Kwa (Internal argument fronting: factive)
[CP [DP INTERNAL ARGUMENT] [C° RELATIVE MARKER [IP . . . . . . . . . . . . . gap]]]]]]
c. French/English (Expletive the fact/le fait, zero insertion: factive)
[CP [DP EXPLETIVETHE FACT/LE FAIT] [C° RELATIVE PRONOUN [IP . . . . . . . . . . . .]]]]]]
In the context of our discussion on the emergence of creoles from the
perspective of competition and selection of syntactic features, a conclusion
that emerges is that Saramaccan retained Focus Raising of nominal and verbal
constituents. In the latter case, such raising structures involve doubling because
6.4 The HC left periphery 265
the focused verb is attracted by two higher heads simultaneously. While this
could be seen as raising structure typical of focus constructions, the discussion
has shown that a similar strategy arises in event factives as well. Here again
Saramaccan partly patterns like Gbe in allowing event factive clauses to derive
from the raising of the verb phrase to the clausal left periphery in conjunction
with a doublet that occurs internally to the proposition.
If we agree that the typological options in (90) are made available by UG,
the fact that Saramaccan, like many other Caribbean creoles, developed
the strategy in (90a) alongside Focus Raising but excludes (90c) cannot be
accidental. Instead, such a development derives from the sort of recombination
of syntactic features that I have argued for in this book. Indeed, these facts
constitute strong evidence that the relevant syntactic features for Focus Raising
and factive clauses are active in both Saramaccan and Gbe. In addition, that
these features activate the left periphery, which represents an interface between
the proposition and the discourse, further supports the view defended in
this book, viz., that interfaces are particularly vulnerable for syntactic
recombination. Before concluding this chapter, let me briefly address a
question that the alert reader might be wondering about, viz., What about
Haitian Creole? Does it show (part of) the different constructions reviewed
here that could further support the recombination analysis proposed in this
book?
6.4 The HC left periphery

There is a considerable literature on the complementizers ke and ki in Haitian
(e.g., Koopman 1982a, 1982b; DeGraff 1992; Lefebvre 1998; Takahash and
Gračanin-Yuksek 2008). I will not review these studies but will limit myself to
relevant examples which are compatible with the analysis presented here in
showing that the creole recombines properties of its source languages in a
syncretic way.
As Lefebvre (1998), DeGraff (2007), and Aboh and DeGraff (forthcoming)
show, Haitian Creole has a complementizer pou derived from French pour ‘for’
as indicated in the following example taken from DeGraff (2007: 111).
(91) Demokrasi bay pèp la prezidan Yo te vote pou li a.
democracy give people DET president 3SG TNS vote for 3SG DET
‘Democracy gave the people the president they had voted for.’
Like French pour and English for, Haitian pou can also introduce non-finite
purpose clauses (DeGraff 2007: 109):
(92) Kouto sa a pa fèt pou koupe pen. [Haitian]
knife DEM DET NEG make for cut bread
‘This knife is not made for cutting bread.’
Interestingly, however, DeGraff (2007: 109) also shows that HC pou can
introduce full finite clauses as illustrated in (93):
(93) a. Kouto sa a pa fèt pou li Koupe pen. [Haitian]
knife DEM DET NEG make for 3SG Cut bread
‘This knife is not made for cutting bread.’
b. Li te ale nan fèt la pou li te ka fè
3SG TNS go to party DET for 3SG TNS capable do
yon ti danse, men lè li rive, pa te gen mizik.
DET little dance but when 3SG arrive NEG TNS have music
‘S/he went to the party to dance a bit, but when s/he arrived there was no
music.’
Thus, unlike Modern French, Haitian Creole pou can introduce both
subjectless non-finite clauses and full finite clauses (but see Aboh and
DeGraff, forthcoming for discussion). On the other hand, Haitian also shows
evidence of modal pou as illustrated in the three examples in (94):
(94) a. Se Bouki ki pou te vini. [DeGraff 2007: 109]
FOC Bouki COMP MOD TNS come
‘It’s Bouki who had to come’
b. Tout moun pou vin lan fèt La [Sterlin 1989: 143]
all people pou come to party DET
‘Everyone must/should come to the party.’
c. Jan pou danse. [Glaude 2012: 60]
Jean MOD danse
‘John is favorable to dancing.’
‘John can dance.’
The different interpretations provided by these authors suggest that pou can
convey different meanings when used as a modal. More study is needed before
we understand the semantics and morphosyntax of this element. What matters
for the current discussion, however, is that this instance of pou follows the
subject of the clause. A surface comparison therefore suggests that Haitian
Creole, just like Saramaccan and Gungbe, has two types of pou: pou1
introduces purpose or irrealis clauses and precedes the subject as indicated in
(93). pou2, on the other hand, follows the subject and encodes modality. In
addition, the focus example in (94a) suggests that pou2 follows focused
elements, which in Haitian Creole are marked by the element se (in a way
similar to many Caribbean creoles, cf. Section 6.3.1). The Haitian focus marker
is further illustrated in example (95), where it marks the focused verb. As in the
case of Kwa languages, Atlantic creoles and many other languages (Aboh and
Dyakonova 2009) such constructions lead to doubling (Harbour 2008; Glaude
and Zribi-Hertz 2012):
6.5 Conclusion 267
(95) Se kouri Bouki ap kouri. [Harbour 2008: 253]

SE run Bouki PROG run
‘Bouki is running.’
These facts lead to the representation in (43) for Saramaccan, Haitian Creole
and Gungbe:
Force TopicP
Discourse–Syntax
interface
Topic FocusP
Focus FinitenessP
Finiteness
Proposition
Though this description is a first approximation of the facts in Haitian

Creole, it is remarkable that, where English has for-to sequences, and French
displays pour-que combinations, Atlantic creoles exhibit homophonous
complementizers like fu-fu or fi-fi in English-derived creoles, or pou-pou in
Haitian. These sequences recall those found in the Gbe or Kwa languages
which manifest n-V sequences like ní-ní in Gungbe. The syntactic and semantic
correlations between these forms further support the idea developed in this
book that the creoles emerge from a recombination of syntactic properties of
competing languages (e.g., Romance or Germanic versus Kwa) and that such
recombinations are very robust when it comes to interfaces. In the case at hand,
the recombination targets functional projections, viz., ForceP and FinP, which
interface between discourse and syntax, and the left periphery and the
proposition, respectively. That the creoles under study as well as the Kwa
languages exhibit Focus and Topic Raising involving dedicated focus (or
exhaustive) and topic markers further supports this view.
6.5 Conclusion
The facts that I discussed in this chapter show that the left periphery of
Saramaccan and Haitian to some extent displays morphosyntactic properties
that combine specific syntactic features of the Gbe languages with those of the
lexifier European languages. Accordingly, both the Gbe languages and the
creoles display distinct discourse-related markers that occupy topic (TopP)
and focus positions (FocP). These two positions are sandwiched between
ForceP and FinP, which can be realized by prepositional complementizers that

encode irrealis and deontic modality. With the exception of the Saramaccan
focus marker wὲ, which is equivalent to Gungbe/Fongbe focus marker wὲ , all
other discourse markers of the clausal left periphery have evolved from the
European lexifier languages. Saramaccan táa evolved from English talk
while fu1/fu2 developed from for, even though their functions cut across those
of for and to in English. On the other hand, Haitian focus (or exhaustive)
marker evolved from the French expression c’est involving the pronominal
demonstrative ce ‘that’ and est (‘be’.3sg, i.e., ‘is’). The complementizer ke
evolved from French que, while the purpose complementizer pou1 and the
deontic marker pou2 evolved from French dative preposition pour. The
systematic correlation between these forms and their functions across
languages suggests that the structural make-up of the clausal complementizer
is determined by UG as indicated in Rizzi’s (1997) seminal work. Under this
view therefore, the languages in competition during the formation of the creole-
speaking society served as appropriate triggers for the emergence of these
syntactic features in the respective creole languages.
I conclude that in a situation of competition, the selection of a
morphosyntactic pattern from a competing language only serves as an
appropriate trigger for fixing the relevant parameter in the emerging
language. It is in this respect that we should understand the notion of
recombination of syntactic features developed in this book.
7 The emergence of serial verb constructions
The discussion in Chapters 5 and 6 focuses on syntactic recombinations in the

left periphery of the noun phrase and the clause, that is, the D-system and the
comp-system, respectively. It was shown there that interfaces between syntax
and discourse–semantics are vulnerable in language contact situations, thus
allowing the emergence of functional items that combine syntactic and
semantic properties of the competing languages in a non-trivial way.
In this chapter, I examine an instance of syntactic recombination within the
INFL domain, focusing on functional verbs in the creoles and how they
contribute to constructions traditionally referred to as serial verb
constructions (SVC).1 Anyone working on Atlantic creole languages
immediately realizes that they involve constructions in which a series of
finite verbs are stacked together in a single clause. These verbs are typically
not separated by any conjunction of coordination or subordination, appear to
‘share’ an argument, and function as a single predicate, as illustrated in (1):
(1) a. Boukinèt te pran yon flè bay Malis. [Haitian]
Boukinette TNS take DET flower give Malice
‘Boukinèt gave Malis a flower.’ [DeGraff 1992: 58]
b. A tjá sondí kó dá dí Faánsi sèmbè. [Saramaccan]
3SG carry thing come give DET French man
‘He presented something to the Frenchman.’ [Veenstra 1996: 95]
According to some typological studies, the construction is not evenly
distributed across languages of the world. The so-called SVCs are limited to
certain linguistic areas: the Caribbean, West Africa, Southeast Asia, Amazonia,
Oceania, and New Guinea (Aikhenvald 2006).2 Crucially, one observes that the
construction is typically absent from Indo-European languages and therefore
1
I’m very grateful to Michel DeGraff for his insightful comments and suggestions on a previous
version of this chapter as well as for sharing his judgments and intuitions on Haitian Creole with
me. I’m also thankful to Herby Glaude for providing me with additional data on SVC in Haitian.
2
In the context of creole languages, there is a debate as to whether the construction is found in
Indian Ocean Creoles or not (see Bickerton 1989, 1990b versus Seuren 1990a, 1990b, 1991).
269
Downloaded from https:/www.cambridge.org/core. University of Exeter, on 11 Mar 2017 at 01:52:58, subject to the Cambridge Core
terms of use, available at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167.008
the superstrate languages of the Caribbean creoles (i.e., French, Portuguese,

Spanish, English, and Dutch).
Since SVCs are found in most creoles, and are seen as a ‘creole universal’ by
some authors (e.g., Veenstra 1996), one may wonder how these new languages
acquire such an unevenly distributed structure. This question has been at the
center of debates in creolistics throughout the history of the field. The main
question then is whether SVCs emerged in creole languages as the result of
substrate influence (e.g., Lefebvre 1991, 1998) or as innovations due to a
developing grammatical system in an ‘abnormal’ contact situation (e.g.,
Bickerton 1981, 1988a, 1988b, 1990b; Sebba 1987; Byrne 1987).
Bickerton’s view seems in line with observations made in some typological
work that serializing languages lack prepositions (e.g., Lord 1973). As
explained in Chapter 3, Bickerton (1988a) argues that prepositions belong to
categories that, once lost in the creolization process, will not be reconstructed.
Accordingly, creoles circumvent this gap by resorting to SVCs. This is
compatible with sentence (1a), in which the second verb, labeled as V2,
occurs in a surface position corresponding to that of prepositions in
Germanic and Romance. Assuming that V2 and the corresponding
prepositions satisfy the same formal requirements (e.g., licensing dative
case), it does not seem unreasonable to think that the presence of one in a
language implies the absence of the other.
On the other hand, the discussion in Chapter 3 has shown that Bickerton’s
view (and much related work) can be easily discarded on empirical grounds:
most serializing languages also involve a set of adpositions (Aboh 2005b,
2010b). Van den Berg’s (2007) work on eighteenth-century Sranan shows
that this language has a rather extensive class of adpositions presented in
Table 3.4 of Chapter 3. Given that Sranan is assumed to have emerged within
about 30 years of the colonization of Suriname by the Dutch (i.e., roughly one
generation, Smith 1987 and much related work), it is obvious that these
adpositions could not all have been reconstructed in such a short period of
time. This would be impossible even if we were to nuance Bickerton’s claim
by allowing the category P to reconstruct. Instead, the data presented in
Chapter 3, in Muysken’s (1988) reply to Bickerton, and in van den Berg
(2007) strongly suggest that adpositions never got lost during the emergence
of creoles.
Having discarded Bickerton’s position, we are left with the following
scenario: SVCs are assumed to be absent in the superstrate languages,
therefore the only possible explanation for their presence in creoles is
substrate influence. For instance, Lefebvre (1998) invokes relexification to
account for the presence of SVCs in Haitian Creole, assuming that the
creators of creoles transferred these structures from their native languages
to the emerging vernacular. The Relexification Hypothesis comes with its
7.1 The clause in Haitian Creole, Saramaccan, and Gbe 271
own inadequacies and theoretical drawbacks which show that the scenario is
not this simple. I showed in Chapter 3 that the Relexification Hypothesis as
defined by Lefebvre (1998) is untenable, consistent with Singler (1996) and
DeGraff (2002, 2003), among others. Nonetheless, Section 7.1 presents
ample empirical evidence that lends support to the view that SVCs
developed in creole languages as a result of cross-linguistic influence.
Building on this, Section 7.2 argues that cross-linguistic influence must be
defined in a very precise way, such that it is compatible with the theory of
competition and selection adopted in this book. Indeed, a new analysis of
SVCs presented in Sections 7.3 and 7.4 indicates that verb serialization is
widely represented in languages of the world (including Indo-European)
where it takes the form of clause union, also referred to as restructuring
in the generative framework and defined as a process that unites two clauses
into one (Rizzi 1982; Wurmbrand 2001, 2004; Cinque 2004). Under this
finding, serialization appears to be a property of UG for which there does not
seem to be any obvious distinctive parameter that could set serializing
languages apart from non-serializing ones (Aboh 2003a, 2009b). This
means that, in order to demonstrate substrate influence, we can no longer
rely on a superficial one-to-one mapping of SVCs in the creoles onto
apparently corresponding SVCs in the relevant substrate languages. In
Section 7.5, I show that substrate influence can be measured by looking at
both similarities and differences between the emerging creoles and their
source languages. Under a theory of creole formation that assumes
competition and selection, with creoles being normal instances of hybrid
grammars, I account for these facts arguing that the creoles display
recombinations of linguistic features from both the superstrate and the
substrate. Section 7.6 concludes the discussion.
7.1 Some morphosyntactic aspects of the clause in Haitian Creole,

Saramaccan, and Gbe
Various basic syntactic properties of the clause in Haitian and Saramaccan
make them similar to the Gbe substrate languages and unlike their European
lexifiers. Such similarities, in turn, could favor a scenario according to which
SVCs emerged in the creoles as a result of substrate influence. For instance,
Haitian, Saramaccan, and their Gbe substrate languages are typologically
isolating and do not exhibit inflectional affixes. Accordingly, though Haitian
distinguishes between long versus short verb forms, which are arguably
manifestations of verbal inflection (DeGraff 1997, 2002, 2005), both Haitian
and Saramaccan are more isolating than their respective superstrates, French
and English. Table 7.1 summarizes certain aspects of Haitian, Saramaccan, and
Gungbe that distinguish them from English and French.
Table 7.1 Some contrasting properties between the creoles and their source
languages
Haitian Saramaccan Gungbe French English
Inflectional morphology on the verb no no no yes yes

Subject–Verb agreement no no no yes yes
Finite vs. non-finite verb forms no no no yes yes
TMA as free morphemes yes yes yes no No
S-VEVENT-O = perfective aspect yes yes yes no No
S-VSTATE-O = present state yes yes yes no No
TMA particles co-occur yes yes yes no no
It appears that in contexts where the creoles and the substrates share some
morphosyntactic properties, these are absent in the lexifier languages and vice
versa. As a result, Haitian and Saramaccan are more like Gbe languages vis-à-
vis their inflectional morphology and the expression and interpretation of
Aspect than they are like their respective superstrate languages.
Interestingly, this language group often displays aspect markers that evolved
from verbs. A case in point is the habitual marker in Gungbe, which evolved
from the verb nɔ̀ ‘to stay’, as illustrated in (2).
(2) Súrù nɔ̀ nɔ̀ xwégbè.
Suru HAB stay house
‘Suru used to/habitually stay/s at home.’
Similarly, the Saramaccan irrealis/future marker ó is assumed to have
developed from the English verb go, while the progressive aspect marker tá,
which can also be used to mean to wait/to stay, is analyzed as having developed
from English stand (e.g., Veenstra 1996: 13; Winford and Migge 2007: 85, 89;
see also van den Berg and Aboh 2013 on the morphosyntax of kaba/kba in
Sranan). The same could be said of the Haitian mood markers dwe and met,
which are derived from the French verbs devoir ‘must’3 and (per)mettre ‘allow’
(e.g., Glaude p.c. January 24, 2013).
Together, these facts indicate that the Gbe languages, Haitian, and
Saramaccan exhibit a cluster of properties that all appear compatible with the
emergence of verb serialization: They are isolating, they do not exhibit
agreement or inflectional morphology on the verb. They involve TMA
markers which have typically evolved from verbs (sometimes adverbs) and
therefore exhibit some form of auxiliation where both the lexical verb and its
cognate INFL-marker co-occur.
3
In the singular forms, the present tense of devoir was pronounced [dwε(t)] in the seventeenth
century.
7.2 Serial verb constructions: a brief description 273
Given these superficially converging structures, and assuming that the

properties illustrated above are somehow related to the presence of SVCs in
these languages, one is naturally led to believe that the emergence of SVCs in
the Atlantic creoles is a clear case of the structural continuum between these
languages and their substrates. As such, SVCs would appear the strongest
empirical arguments for substrate influence.
Yet, these surface similarities by no means suffice to demonstrate how
speakers of the Gbe languages could have transferred SVC structures to their
L2 versions of French or English that evolved into Haitian Creole and
Saramaccan, respectively. In addition, before we claim exclusive substrate
influence, we have to be certain that the grammar of French and English have
nothing structurally similar to serialization that could favor the retention of
SVC structures from the substrate. I show in the following sections that the
surface (dis)similarities between the superstrates, on the one hand, and the
substrates and the creoles, on the other, might be misleading about the proper
structural make-up of serialization. The discussion shows that serialization
does have its own raison d’être in both French and English. These findings
need not be sad news for strong proponents of substrate influence. Instead, what
we learn from the discussion in Sections 7.4 and 7.5 is that we need to go
beyond surface matching in order to argue for substrate influence or syntactic
recombination in the context of SVCs. In a similar vein, the discussion in this
chapter anticipates Chapter 8 where I show that theories of creole simplicity,
which only search for surface manifestations across these languages, not only
fail to shed light on fundamental structural properties of these languages but are
also particularly misleading regarding linguistic typology. Before getting onto
this, however, let me first return to the issue of serialization proper.
7.2 Serial verb constructions: a brief description

As shown in Section 7.1, Gungbe, Saramaccan, and Haitian resort to the
combination of various verbal elements (whether in the form of TMA or
lexical verb) to encode aspect and modality. Given this possibility, it is not
surprising that SVCs are used in the same way: a combination of lexical verbs is
used to encode result or inception, or to introduce a new argument.
Additional examples are given in (3):
(3) a. Mi wáka gó a dí wósu. [Saramaccan]
1SG walk go LOC DET house
‘I walked to the house.’
b. Yo pote [bon legim sa yo] bay Mari. [Haitian]
3PL take good vegetable DEM PL give Mary
‘They brought those nice vegetables to Mary.’
c. Yé ɖà gbómán lɔ́ ɖù. [Gungbe]

3PL cook spinach DET eat
‘They cooked the spinach and ate (it).’
As explained above, the examples in (1) and (3) show that SVCs can be
defined as sentences involving a series of finite verbs with no intervening
conjunction or subordinator between them and functioning as a single
predicate. Accordingly, the SVCs presented here can be schematized as in
(4), where the parentheses show that the internal arguments or adjuncts are not
always realized.
(4) Subject- Verb- (XP[object/adjunct])- Verb- (YP[object/adjunct])
At first blush, nothing formally distinguishes schemata (4) from the structure
that might underlie the sentences in (5).
(5) a. John made me buy a new car.
b. Jean a fait réparer ma voiture.
John has made repair my car
‘John had my car repaired.’
What then makes the sentences in (5) different from those in (1–3)? A point
often raised in the literature is that the examples in (5) involve a non-finite verb
in the subordinate clause. This point is not conclusive, however. First, because
the creoles do not exhibit finite versus non-finite morphology on the verb,
we can’t rely on such a criterion to distinguish between SVCs and embedded
(non-)finite structures. Second, if the non-finite morphology on the verb in
(5) correlates with embedding of some reduced structure (e.g., AspP, TP, FinP),
then nothing a priori distinguishes the causative structures in (5) from the SVCs
in (3), where a similar embedding structure could be argued for despite the
absence of verbal inflectional morphology (see Section 7.4). Questions of this
sort have led linguists working on SVCs to define a number of criteria that
would help set SVCs apart from instances of subordination as in (5). The next
section reviews some of these criteria and shows their limitations.
7.2.1 On some morphosyntactic properties of SVCs

For ease of exposition, I will illustrate these criteria with examples taken
primarily from Gungbe. (See Lefebvre 1991 and Lefebvre and Brousseau
2002 for discussion of Fongbe.) I call the reader’s attention to the fact that
the properties discussed here do not uniformly cut across all the relevant
languages; there are sometimes startling differences between them. Due to
space limitations, I cannot present these facts in great detail. When necessary,
however, I point to relevant differences between these languages, but the main
aim of this section is to present the reader with very general aspects of SVCs
that bear on the positions defended in this book.
SVCs typically display a single overt structural subject shared by all the
verbs. In example (6), for instance, the same noun phrase is interpreted as
subject of the first and second verbs, that is, the bandits beat and killed the man.
(6) Jàgùdà lέ xò dáwè lɔ́ hù.

bandit PL beat man DET kill
‘The bandits beat the man to death.’
In addition to exhibiting a single subject, SVCs often display a single INFL-

system. In Gungbe, expressions of tense, negation, and aspect occur between
the subject and the first verb only (7a). As the ungrammatical example (7b)
shows, Gungbe excludes sequences in which any of these markers attaches to
the second verb. Similarly, the TMA markers cannot occur on both verbs, as is
evident from the ungrammaticality of (7c):
(7) a. É má ná nɔ̀ nyàn ví lέ yì.
3SG NEG FUT HAB chase child PL go
‘He will not habitually chase the children away.’
b. É nyàn ví lέ (*má) (*ná) (*nɔ̀ ) yì.
3SG chase child PL NEG FUT HAB go
c. *É má ná nɔ̀ nyàn ví lέ má ná nɔ̀ yì.
3SG NEG FUT HAB chase child PL NEG FUT HAB go
The data in (7) indicate that even though SVCs embed a series of ‘finite’
verbs, the latter share the same tense, mood, and aspect system. Accordingly,
TMA markers as well as other INFL-related elements occur before V1 only. In
some Gbe languages, however, the verbs in an SVC may all show the same
aspect marker. This is the case in the Ewegbe example (8), where the habitual
marker attaches to both verbs, unlike in Gungbe (7a).4
(8) E tsɔ-*(na) akɔɖu ɖu-*(na). [Ewegbe]
3SG take-HAB banana eat-HAB
‘S/he habitually eats bananas.’
At this stage of our description, one may wonder whether these SVCs are not
coordinate structures involving a null coordination conjunction. This view
would seem compatible with the fact that certain Gbe languages (e.g.,
Ewegbe) involve non-overt coordination. In such cases, the unpronounced
subject of the second conjunct is necessarily bound by that of the first
conjunct. The variation across Gbe is illustrated in (9):
4
Such INFL spreading is found in many Kwa languages (e.g., Akan, Baule) as well as typologi-
cally different languages (Aikhenvald 2006: 42ff.).
(9) a. Kòfí ná ɖù lέsì *(bò) ná yì xɔ̀ mὲ. [Gungbe]

b. Kòfí á ɖù mɔ́ lú . . . á yì xɔ̀ mὲ. [Ewegbe]
Kofi FUT eat rice (COORD) FUT go room
‘Kofi will eat rice and will go into the room.’
Note that while Ewegbe has a null coordinator in (9b), Gungbe (9a) requires
the coordinator to be overtly realized. In addition, as already discussed by
Collins (1997), an important distinction between this type of coordination and
an SVC is that the sentence in (9b) allows the second conjunct to embed tense
and negation. This is not possible in SVCs, though the second verb may take an
aspect marker. The impossibility of marking tense and negation on the second
verb appears to be the decisive factor for distinguishing between SVCs and
covert coordination or subordination.
Besides the coordination structures in (9), the Gbe languages also involve
another coordinating strategy, according to which the conjunction is
necessarily overt and the subjects of the two conjuncts are realized, as in (10):
(10) Sέsínú ɖà làn bɔ̀ Súrù ɖù nŭsɔ́ nú.
Sesinou cook meat COORD Suru eat soup
‘Sesinou cooked meat and Suru ate soup.’
While these coordinate structures are formally distinct from SVCs because
of the overt conjunction, they also illustrate other differences between SVCs
and coordinate structures. In Aboh (2003a, 2009b), I show that while overt
coordination structures exclude extraction of the arguments (except in across-
the-board fashion), SVCs are transparent to all movement operations. The
examples in (11) indicate that an SVC like (8a) allows extraction of all
arguments. This conclusion further indicates that SVCs should not be seen as
instances of VP or IP coordination.5
(11) a. Mέnù wὲ nɔ̀ zé àkwékwè ɖù? [Gungbe]
who FOC HAB take banana eat
‘Who habitually eats bananas?’
b. Étέ wὲ é nɔ̀ zé ɖù?
what FOC 3SG HAB take eat
‘What does he habitually eat?’
Indeed, unlike SVCs, coordination structures ban extraction out of the
conjuncts, except in an across-the-board fashion. In (12c), for instance,
the two conjuncts contain a variable bound by the same wh-operator (Ross
1967).
5
Michel DeGraff (p.c. June 21, 2009) remarks that this constraint does not hold for Haitian, which
disallows extraction:
(i) *Ki moun Bouki pran flè bay?
wh- person Bouki take flower/s give
‘Who did Bouki gave flowers?’
(12) a. *Étέi wὲ Sέsínú ɖà ti bɔ̀ Súrù ɖù nŭsɔ́ nú?

what FOC Sesinou cook COORD Suru eat soup
‘What did Sesinou cook and Suru ate soup?’
b. *Étέi wὲ Sέsínú ɖà lέsì bɔ̀ Súrù ɖù ti?
what FOC Sesinou cook rice COORD Suru eat soup
‘What did Sesinou cook rice and Suru eat?’
c. Étέi wὲ Sέsínú ɖà ti bɔ̀ Súrù ɖù ti?
what FOC Sesinou cook COORD Suru eat
‘What did Sesinou cook and Suru ate?’
The facts we have discussed thus far indicate that SVCs behave like single
clauses: a property that formally sets them apart from structures involving
coordination (e.g., the Romance and Germanic examples in 5).
An argument that is often used in the literature to sustain this description is
that SVCs typically encode a single event and therefore require all sub-events
to be true at once. A comparison of the SVC examples in (13a–b) with the
purposive ones in (13c–d) illustrates this view.
(13) a. Àyábá zé hì sɔ́ àsú étɔ̀ n

Ayaba take knife stab husband POSS
(*àmɔ́ n é má sɔ́ -ὲ).
but 3SG NEG stab-3SG
‘Àyábá took the knife and stabbed her husband (*but she did not stab him).’
b. Àyábá má zé hì sɔ́ àsú étɔ̀ n
Ayaba NEG take knife stab husband Poss
(*àmɔ́ n é sɔ́ -ὲ).
but 3SG stab-3SG
‘Àyábá did not take the knife and stabbed her husband (*but she stabbed him).’
c. Àyábá zé hì bó ná sɔ́ àsú étɔ̀ n
Ayaba take knife COORD PREP stab husband POSS
(àmɔ́ n é má sɔ́ -ὲ cá ɖó àsú lɔ́ hɔ̀ n).
but 3SG neg stab-3SG pcl cause husband det flee
‘Àyábá took a knife to stab her husband, but she didn’t eventually because he
fled.
d. Àyábá má zé hì bó ná sɔ́ àsú étɔ̀ n
Ayaba NEG take knife COORD PREP stab husband POSS
(àmɔ́ n é sɔ́ -ὲ kpò gà ɖé sɔ́ -ὲ).
but 3SG stab-3SG stick long DET stab-3SG
‘Àyábá did not take a knife to stab her husband, but she did hurt him with a
long stick.’
It appears from these examples that it is not possible to negate only a sub-
event of an SVC (13a–b), but nothing excludes this in a non-serializing
construction (13c–d).
7.2.2 Some semantic constraints on V1 and V2 combinations in SVCs

Finally, a fact not often discussed in the literature is that verb serialization is not
a free process such that any verb type can combine with any other into a
monoclause. In most Gbe languages, there appear to be some semantico-
lexical restrictions on the verbs that enter a series. That is, SVCs are lexically
constrained; verbs in the series belong to distinct ‘semantic’ classes. In
Gungbe, for instance, we observe that while the sequence steal-look is
perfectly accepted in a series, the sequence steal-see is not (14a–b).
(14) a. Sέtù fìn fòtó lɔ́ kpɔ́ n.
Setu steal photo DET look
‘Setu quickly looked at the photos (i.e. without being noticed by other people).’
b. *Sέtù fìn fòtó lɔ́ mɔ̀ n.
Setu steal photo DET see
‘Setu quickly looked at the photos (i.e. without being noticed by other people).’
There seems to be no principled way to distinguish between these two verbs
on a syntactic level. What seems to be at stake here is some kind of control over
the event of looking as opposed to that of seeing, whereby the former involves
some agency from the actor while the latter implies that the actor’s vision is
affected by the seen element. These examples seem to point to an agent versus
experiencer contrast.6 Accordingly, example (14b) must have been ruled out
on the basis of some semantico-lexical constraint on these verbs, which itself
might have structural correlates.7
Similar constraints are found in the creoles under discussion here. In
Saramaccan, for instance, the verbs gó ‘go’ and kó ‘come’ as V2 mostly
combine with verbs of motion as V1 (Veenstra 1996: 93).
(15) A kúle gó.
3sg run go
‘He ran away.’
6
Though this is by no means linguistic evidence, it is worth mentioning that there is a saying in
Gungbe which is compatible with the description provided here. As the reader may see, the
selected verb corresponds to “to see” not “to look.”
(i) Núkún ɖé mɔ̀ n/*kpɔ́ n nú má hù hwὲ.
eye REL see/look thing NEG kill fault
‘An eye that sees (something) is not guilty [i.e., anyone who happens to see something
by chance cannot be judged guilty].’
7
I thank Michel DeGraff for his comments on this issue.
While this may look like a ‘natural’ restriction on V2, there does not seem to be
any syntactic principle that will force V2 to combine with motion verbs only when
it implies direction, transfer, or displacement. Instead, the constraints on such verb
combinations seem to be semantic in nature. Indeed, in Gungbe, for instance, it is
perfectly possible to utter sentence (16) in which the three verbs in the series (in
boldface) include yì ‘go’ following a verb of saying ɖɔ̀ xó ‘say word’, which does
not imply motion. Also notice from the translation that the whole sentence does not
imply movement (though this is possible in a metaphorical sense).
(16) Àzɔ́ nví lέ ɖɔ̀ [xó] yì jὲ gán sín àsì jí
apprentice PL say word go reach chief POSS wife on
bɔ̀ gán sè . . .
COORD chief hear
‘The apprentices talked about the chief’s wife and the chief heard them . . . ’
It therefore appears that the restrictions on the Saramaccan example in (15)
relate to the semantic specifications of the functional projection that hosts V1.
Similarly, DeGraff (1992, 2007) identifies three types of serial verbs in Haitian:
motion (17a), transfer (17b), and comparison (17c).
(17) a. Bouki voye timoun yo ale lekòl. [DeGraff 1992: 57, 58]
Bouki send kids PL go school
‘Bouki sent the kids to school.’
b. Boukinèt te pran yon flè bay Malis.
Boukinet TNS take a flower give Malis
‘Boukinèt gave a flower to Malis.’
c. Boukinèt mache (plis) pase Mari.
Boukinèt walk more pass Mari
‘Boukinèt walked more than Mary.’
With regard to SVCs encoding motion or transfer, DeGraff (1992: 58)
observes that verbs that realize the position V2 are selected from a smaller set
than verbs that encode V1: in motion-type SVCs, V2 is typically taken from (e.g.,
vini, ale, rive, soti), while it is realized as (bay, pote, mete . . .) in transfer-type
SVCs. This would mean that in Haitian serialization, V1[motion/transfer] exhibits
semantic selectional requirements on the lexical verbs V2 it co-occurs with.
Similar semantic factors could be at the source of the impossibility of example
(18a) in Haitian Creole, but not in the corresponding Gungbe example (18b).8
(18) a. *Jan kuit pen an manje.
b. Ján mὲn blέɖì lɔ́ ɖù.
John cook bread DET eat
‘John cooked bread and ate it.’
8
I thank Michel DeGraff for calling my attention to these facts in Haitian.
In this case, V1[kuit: cook] cannot combine with a lexical verb V2 encoding ‘to
eat’. I conclude from these facts that verb combinations in SVCs are subject to
semantic restrictions that govern the choice of verbs that can express V2 in
combination with V1. For the time being, I have no comprehensive analysis to
offer as to what semantic principles govern the various constraints illustrated in
these examples. What matters for our discussion, though, is that SVCs do not
involve a mere syntactic combination of independent lexical verbs into one
clause. Instead, these data suggest that there is a semantic and syntactic
dependence on the verbs in a series. This latter point suggests that previous
analyses of SVCs as involving two independent verbs that discharge their
thematic-roles on shared arguments might not be right. I return to this
discussion below.
7.3 Serial verb constructions: a reanalysis

Besides the contrast we observe in (18), one finds pairs of examples such as
those in (19a–b), versus (19c–d), in which the strong parallelisms between
Gungbe and the creoles contrasts sharply with the inexistence of similar
structures in French and English.
(19) a. Al pran machin lan pote vini ban mwen [Haitian; DeGraff 2007: 116]
b. Yì zé mótò lɔ́ (hὲn) wà ná. mì [Gungbe]
go take car DET bring come give 1SG
‘Go get me the car.’
c. A tjá njanján kó d’ én. [Saramaccan; Aboh, Smith, and Veenstra 2013]
d. É hὲn núɖùɖù wá nὲ-ὲn. [Gungbe]
3SG carry food come give.3SG
‘S/he brought him/her food.’
While the parallel in (19) could be seen as a mere typological accident
supported in fact by a comparison of these creoles to Mandarin Chinese, for
instance, the fact that Haitian Creole, Saramaccan, and the Gbe (Kwa)
languages had a common history at some point in time favors an analysis in
terms of cross-linguistic influence. This would mean that the relevant creoles
developed SVCs because they inherited the serialization parameter from the
substrate languages. As Lefebvre (1998: 357) puts it:
the creators of Haitian, who were native speakers of languages of the Kwa family, would
have kept the parametric value of their original language in creating the creole: they did
not have verb raising to INFL and hence verb serialization was available to them. The
first generation of native speakers of Haitian would have identified the absence of verb
raising in the language that they were presented with. Having identified this parametric
value, they deduced the availability of verb serialization in the grammar.
7.3 Serial verb constructions: a reanalysis 281
While this view somehow holds of example (19), the facts in (18) indicate
that things are not this simple. Indeed, the structural discrepancies between the
creoles and the substrates, such as illustrated in example (18), point to the fact
that the emergence of SVCs in the creoles cannot be attributed exclusively to
substrate transfer. Lefebvre’s (1998) account, provides no principled
explanation for why very productive Gbe structures such as (18b), which
must have been part of the PLD that made speakers “deduce the availability
of verb serialization,” are not retained in Haitian Creole.
In what follows, I report the conclusions in Aboh (2009b), where it appears
that there can be no serializing parameter such that two or more heads of
predicates can share the same internal arguments. This would mean that
theories of SVCs that posit complex predicates cannot be on the right track. In
this particular regard, it also appears that the so-called verb series share similar
properties with structures such as the American English examples “he will go talk
to his advisor today” or “I requested that she come discuss this problem with me
in person,” discussed in Jaeggli and Hyams (1993).9 Accordingly, the
phenomenon of verb sequences is more common than often assumed in the
literature. In order to contextualize the discussion, let us first step back and
consider some approaches to SVCs that have been explored in the literature.
For the past twenty years or so, various parameters have been evoked to
explain serialization: lack of prepositions (Bickerton 1981), lack of inflectional
morphology and hence verb raising (Muysken 1988; Baker 1991, cited in
Lefebvre 1998), and object sharing (Baker 1989). With regard to the
supposed absence of prepositions in serializing languages, the Gungbe
sentences under (20) which involve an SVC or a prepositional phrase to
introduce an instrument clearly indicate that this view is not tenable. Indeed,
most serializing languages also have adpositions (see Chapter 3, as well as
Lefebvre and Brousseau 2002; Ameka 2003; Aboh 2005b, 2010b).
(20) a. Sέtù zé kpò xò kɔ̀ jó.
Setu take stick hit Kojo
‘Setu hit Kojo with the stick.’
b. Sέtù xò Kɔ̀ jó kpó kpò Kpó.
Setu hit Kojo PREP stick Prep
Similarly, the view that serialization occurs in languages which lack verbal
inflections and therefore verb raising (e.g., Muysken 1988; Baker 1991) cannot
be maintained. Aboh (2003a, 2004a, 2005c, 2009b) and Aboh and Dyakonova
(2009) present a very detailed discussion of verb movement in the Gbe
9
As I show in Aboh (2009b) and further discuss here, there appears to be no structural distinction
between these motion verb constructions and SVCs.
languages showing that the verb must raise at least to an aspect position to be
licensed in these languages. The Ewegbe example in (21) illustrates this point.
In this language, the habitual aspect marker is an affix to which the verb must
adjoin. In SVCs therefore, V1 and V2 must move to distinct aspect nodes where
they adjoin to the aspect marker.
(21) E tsɔ-na akɔɖu ɖu-na. [Ewegbe]
3SG take-HAB banana eat-HAB
‘S/he used to/habitually eat/s bananas.’
That serialization is not linked to absence of verb movement is further shown
by comparative data from Igbo (Kwa) and ╪Hoan (Khoisan), illustrated in
(22a) and (22b) respectively:
(22) a. Obi kwa-da-ra Eze. [Igbo, Kwa; Steward 1998: 183]
Obi push-fall-RV Eze
‘Obi pushed Eze down.’
b. Ma a- q║hu │’o djo ki kx’u na. [╪Hoan; Collins 2002: 1]
1SG PROG pour put.in water PART pot in
‘I am pouring water into the pot.’
Osamuyimen T. Stewart (1998) and Collins (2002) show convincingly that
these constructions share strong parallels with SVCs except for the ordering of
the verbs. While in some Kwa languages, the serializing verbs may be
separated by an intervening argument, some other Kwa (e.g., Igbo) and
Khoisan languages display verb compounding where the serialized verbs are
adjacent, with V1 systematically preceding V2. The generalization seems to be
the one in (23) where XP stands for the internal argument.
(23) a. V1 (XP) V2 [Kwa, Gungbe]
b. V1 – V2 (XP) [Kwa, Igbo; Khoisan, ╪Hoan]
Though the analyses they propose are significantly different, both Collins
(2002) and Aboh (2009b) resort to verb movement to account for the variation
in (23). Starting with (23a), for instance, these studies account for the ordering
in (23b) in terms of movement of the verb (i.e., V2) past the object.
If serialization is not a consequence of a lack of prepositions in the
serializing language or of the absence of inflectional morphology and hence
verb raising, the only option we are left with is the object sharing hypothesis
proposed by Baker (1989). According to this author, SVCs are defined by the
fact that the verbs in the series necessarily share a unique internal argument.
There have been several studies (e.g., Campbell 1989, 1992, 1996b; Da Cruz
1993, 1997, among others) indicating that object sharing, as proposed by Baker
(1989) and much related work, cannot be a defining condition on SVC.
Consider the following Gungbe sentences:
(24) a. Àsé lɔ́ lɔ́ n xέ àtín lɔ́ jí.

cat DET jump climb tree DET on
‘The cat climbed on the tree (i.e., by jumping there).’
b. Àlúkú kùn mótò sɔ́ àdó.
Aluku drive car hit wall
‘Aluku crashed a car into the wall (i.e., driving it).’
Example (24a) involves an unergative V1 combined with a transitive V2. V1
and V2 share a unique external argument, but only V2 selects for the locative
DP object. In sentence (24b), V1 and V2 select for their own object and no
sharing arises. Though being SVCs in Gungbe these examples do not involve
object sharing. Similar counterevidence is found in Haitian Creole as
instantiated by the following example:
(25) Bouki achte yon tikè avyon ale Nouyòk. [DeGraff, p.c. June 21, 2009]
Bouki buy DET ticket flight go New York
‘Bouki bought a plane ticket to go to NY (and he did go to NY).’
These examples therefore violate Baker’s object sharing hypothesis.
According to Baker (1989) such examples could be analyzed as hidden
coordination.10 Starting with (24a), the sentences in (26) show that these
examples involve a single expression of tense, mood, aspect, and negation
(26a), and they allow wh-extraction of all arguments (26b–d).
(26) a. Àsé lɔ́ má ná nɔ̀ lɔ́ n (*má *ná *nɔ̀ ) xέ àtín lɔ́ jí.
cat DET NEG FUT HAB jump NEG FUT HAB climb tree DET on
‘The cat will not habitually climb on the tree (i.e., by jumping).’
b. Àsé lɔ́ wὲ . . . lɔ́ n xέ àtín lɔ́ jí.
cat DET FOC jump climb tree DET on
‘THE CAT jumped on the tree.’
c. Àtín lɔ́ jí wὲ àsé lɔ́ lɔ́ n xέ . . .
tree DET on FOC cat DET jump climb
‘The cat jumped on THE TREE.’
10
Indeed, Baker (1989: 547) analyzed examples such as in (ia–b) as covert coordination.
(i) a. Kofi sutu Amba kiri Kwaku. [Sranan]
Kofi shoot Amba kill Kwaku
‘Kofi shot Amba and killed Kwaku.’
b. Kofi naki Amba kiri en.
Kofi hit Amba kill her
‘Kofi struck Amba dead.’
Evidence for treating these examples as covert coordination comes from the fact that these
sentences display island effects typical of coordinate structures (e.g., impossibility to wh-extract
the object DPs), but see note 5.
Given the discussion in Section 7.2 where I showed that extraction for the
purpose of focus or wh-question is possible out of SVCs but impossible out of
coordinate structures, it is clear that the structures in (24) and (26) are bona fide
SVCs, though they do not involve object sharing. These facts represent a
serious argument against analyses that adopt Baker’s object sharing hypothesis.
That said, we are still left with the question of whether object sharing is ever
possible in serialization. Close scrutiny shows that it is never the case that series
of lexical verbs can share the same object (Aboh 2009b). This is because the Gbe
languages involve various instances of SVC where it cannot be claimed that all
the verbs in the sequence have lexical properties. I begin with take-series.
Sentence (27a) shows that zé ‘take,’ in Gungbe, can be used as main predicate
of the sentence, though it cannot select for abstract objects (27b).11
(27) a. Àlúkú zé gbàdó càkì àtɔ̀ n.
Aluku take corn sack three
‘Aluku took three sacks of corn.’
b. *Àlúkú zé àwá-jìjὲ/xó.
Aluku take joy/word
In SVCs, however, it is possible for this verb to be followed by an abstract
object or a DP element which the verb cannot semantically select in isolation.
(28) a. Àlúkú zé àwá-jìjὲ yì jónɔ̀ n lέ.
Aluku take joy receive guest PL
‘Aluku received the guests with joy.’
b. Àlúkú zé xó mítɔ̀ n ɖɔ̀ ná Súrù.
Aluku take word 1 PL talk PREP Suru
‘Aluku told Suru about us.’
It appears from these examples that the DP element that intervenes between V1
and V2 does not occur in that position in order to be jointly theta-marked by these
verbs. Instead, the verb zé ‘take’ in (28) does not theta-mark the following DP,
and therefore cannot be considered a full lexical verb in these series. This
conclusion is supported by the fact that, in certain Gbe languages (e.g.,
Fongbe) take-series involve an instrument DP which is introduced by a
preposition (Da Cruz 1997: 37).12
(29) a. Kɔ̀ kú sɔ́ [jìví ɔ́ ]i sέn wɔ̀ xúxú ɔ́ ná eci
Koku take knife DET cut bread DET with
‘Koku cut the bread with the knife.’
11
These facts are not new and were previously presented in Lefebvre (1991) and Da Cruz (1997)
for Fongbe. However, Gungbe and Fongbe differ in this respect. While Fongbe has two verbs
(zé, sɔ́ ) that can be translated as take, Gungbe has only one form: zé.
12
See Da Cruz (1997) for the discussion on ná/nú alternation in Fongbe.
b. *Kɔ̀ kú sɔ́ sέn wɔ̀ xúxú ɔ́ nú jìví ɔ́ .

Koku take cut bread DET with knife DET
What is remarkable about this example is that even though the bracketed DP jìví
ɔ́ ‘the knife’ is introduced by the preposition ná (which also introduces a dative
argument as I mentioned in Chapter 6), this DP must occur immediately after V1.
Also remarkable is the fact that the form of the preposition here is ná, which, as I
have shown in Chapter 6, with examples (55e) and (55f), is allowed only when the
following DP-complement has been fronted. Aboh (2009b) concludes on the basis
of these facts that the so-called ‘shared object’ must have moved to the position
between V1 and V2 for other reasons than theta-marking and case-licensing. These
facts lead me to conclude that take-series involve a functional verb take.
A possible objection to this view would be to say that SVCs come in two
types in Gbe: (i) a combination of a functional verb and a lexical verb (27–28),
and (ii) a combination of two (or more) lexical verbs (30) (e.g., Baker 1991):
(30) Àlúkú ɖà àbɔ̀ bɔ̀ ɖù.
Aluku cook bean eat
‘Aluku made some beans to eat.’
This view does not seem to be tenable, however. In accounting for take-
series of the type discussed here, Aboh (2009b) demonstrates that V1 expressed
by the verb take merges in the functional domain of the lexical field which is
headed by V2, as schematized in (31):
(31) [CP . . . . . . [TP . . . V1TAKE . . . . . ..[VP . . . [V2 . . ...]]]]
Accordingly, V1 is a functional (or light) verb that has no (internal) θ -role
to assign, but merges within the functional domain (or the extended
projection) of V2. In this framework, the function of V1 take is to encode
the way the event expressed under V2 has been carried out (e.g., Awóyalé
1988; Lefebvre 1991).
Extending this analysis to examples like (30), which seem to involve two
genuine lexical verbs, Aboh (2009b) concluded that in an SVC, elements of the
type V1 are always functional verbs. Partial evidence which led to this
conclusion comes from the fact that in Gbe (unlike in other serializing
languages, e.g., Saramaccan) there is a general restriction on V1 such that
SVCs cannot be introduced by ditransitive verbs (e.g., ná ‘give’) in Gungbe.
(32) a. *Kòfí ná kwέi xɔ̀ xwé proi.
Kofi give money buy house
‘Kofi gave money buy a house [e.g., he gave money to buy a house].’
b. *Kòfí ná wémái Mì wà àzɔ́ n proi.
Kofi give book 1SG-ACC do work
‘Kofi gave me a book to work [e.g., he gave me a book to work with].’
Yet, this restriction does not hold on Inherent Complement Verbs which are
fine with a ditransitive verb, as V1 (33):
(33) Kòfí ná nú ɖù vi lέ.
Kofi give thing eat child PL
‘Kofi gave food to the kids’
As one can see from the interpretation, the restriction in (32a–b) cannot be
semantic since similar combinations are grammatical in the English
translations, and indeed possible in Saramaccan (34a) and in Haitian (34b).
(34) a. Dí míi dá mi dí sópu wási hen. [Veenstra 1996: 79]
DET child give 1SG DET soap wash 3SG
‘The child gave me the soap to wash him with.’
b. Bouki ban m savon lave men m. [DeGraff, p.c. June 21, 2009]
Bouki give 1SG soap wash hand/s 1SG
‘Bouki gave me soap to wash my hands with.’
First, these types of mismatches between the creoles and the substrate
languages indicate that simple substrate influence cannot help account for the
existence of SVCs in the creoles. Second, assuming example (34a) is indeed an
SVC, we can conclude that the so-called single event reading often observed in
SVCs must be a tendency rather than a defining criterion. Indeed, the washing
need not materialize in this case since the event of giving and washing need not
be construed as the same event.
Third, the general ban on ditransitive verbs as V1 in Gungbe SVCs with
intransitive V2 (32a–b) suggests that the elements that realize this V1-position
are subject to a thematic restriction. We can therefore conclude that there is a
structural restriction on V1 in Gungbe. Now, let us recall the data in (14),
repeated here as (35) for convenience. The conclusion we reached there was
that there is a semantic restriction on SVCs such that V1 determines the type of
V2 it may combine with. Accordingly, the sequence steal-look is grammatical,
but not the sequence steal-see is not (14a–b).
(35) a. Sέtù fìn fòtó lɔ́ kpɔ́ n.
Setu steal photo DET look
‘Setu quickly looked at the photos (i.e. without being noticed by other
people).’
b. *Sέtù fìn fòtó lɔ́ mɔ̀ n.
Setu steal photo DET see
‘Setu quickly looked at the photos (i.e. without being noticed by other
people).’
Based on all these facts, we can conclude that elements of the type V1 are
subject to both semantic and structural restrictions. This property reminds us
of auxiliary verbs and auxiliary selection as discussed in Burzio (1986) and

much related work. While certain auxiliaries can be used as main predicates
in some contexts (e.g., have), they show structural and semantic restrictions
when combined with another predicate. In French, for instance, the verb
arriver ‘to arrive’ can only occur with the auxiliary être ‘to be’, while the
verb écrire ‘to write’ must combine with the auxiliary avoir ‘to have’ for the
‘passé composé.’ The behavior of auxiliaries just described is very much like
that of V1 which is subject to structural constraints and appears to impose
certain restrictions on its complement, inside which V2 occurs. Keeping to
this parallel, I conclude that V1 can be analyzed on a par with auxiliary verbs.
I refer to elements of the type V1 as ‘functional verbs’ a term adopted from
Cinque (2004).
In this regard, Gungbe offers several examples, where the verbs in the series
express a meaning that cannot be decomposed into the meaning of the two
verbs taken separately. I take these facts to mean that verbs that have a
functional and a lexical usage display certain syntactic and semantic
restrictions in their functional usage. This in turn suggests a thematic
restriction on V1. Consider, for instance, sentences (36a–b), where the lexical
verb bέ ‘collect’ does not encode this meaning when used functionally. Instead,
V1 bέ fails to assign an internal θ-role to the object to its right because the
intended meaning is that of a modifier of VP2: ‘there has been a lot of eating or
talking nonsense’:
(36) a. Àsíbá bέ lέsì ɖù.
Asiba collect rice eat
‘Asiba ate a lot of rice [not Asiba collected rice and ate].’
b. Àsíbá bέ xó ɖɔ̀ .
Asiba collect word say
‘Asiba said a lot of nonsense [not Asiba collected word and said them].’
Other verbs that show such an alternation between V1 and V2 are kpɔ́ n ‘to
look at/for’ and hὲn ‘to hold’ as illustrated in examples (37).
(37) a. Súrù! kpɔ́ n xɔ́ ɖɔ̀ !
Suru look word say
‘Suru! Be careful about what you say! [not Suru! Look for a word to say!]’
b. Súrù hὲn mótò cè gblé.
Suru hold car my spoil
‘Suru caused my car to break down.’
c. Súrù hὲn mótò wá sɔ́ n yòvótòmὲ.
Suru hold car come from Europe
‘Suru brought a car from Europe.’
It appears from these examples that when the verb kpɔ́ n ‘to look at/for’ occurs
as V1 in an SVC, it does not convey the meaning of English to look or to see (37a).
Instead, the meaning we have here is closer to the English paraphrase watch your
words or speak properly. Taking into account the meaning of the Gungbe SVCs
rather than the morphological form of the verbs it appears that V1 expresses the
manner in which the event expressed by V2 should be performed. The right
generalization is therefore that, in V1-V2 combinations, V1 embeds the manner
component. A similar observation holds for the verb hὲn ‘to hold’, which can be
understood as to ‘cause X to spoil’ (37b), or to ‘cause X to be displaced’ (37c).
Contrary to what is often assumed in the literature, these examples show that
SVCs are not just a random combination of two lexical verbs expressing a
single event. This conclusion is further supported by the fact that even verbs
like ɖín ‘to look for’ or ‘to search for’ show semantic variation (that is tied to a
thematic restriction) when they are used as lexical or as V1 in an SVC. Observe,
for instance, that below this verb has a ‘causative’ meaning in (38a) different
from its lexical meaning in (38b):
(38) Àsíbá, ɖín nú ɖé ɖù tò àdòkɔ̀ n.
Asiba search thing INDEF eat at kitchen
a. ‘Asiba, get yourself something to eat in the kitchen.’
b. *Asiba, look for something to eat in the kitchen.’
Taking all these facts into account, I further submit that, since the meaning of
ɖín in this example is equivalent to English get (or have) in their causative
usage, there is no formal reason for us to assume that V1 and V2 ever share the
same argument in SVCs. This in turn would imply that elements of the type V1
are not lexical verbs. This observation may look like a bold statement, when
one considers example (39):
(39) Mígán ɖà àbɔ̀ bɔ̀ ɖù.
Migan cook bean eat
‘Migan cooked beans and ate.’
Traditional descriptions of similar examples conclude that the verbs in these
series combine two consecutive events (e.g., cooking and eating) into a single
event. Yet, as I demonstrate in Aboh (2009b), the analysis proposed here in
terms of V1 as functional verb extends to examples (39) as well. Observe, for
instance, that Gungbe also displays constructions like (40):
(40) Mìgàn ɖà kpònɔ̀ n lέ sɔ́ yì àhwàn.
Migan cook soldier PL take go war
‘Migan prepared the soldiers to go to war [i.e., by making some magic].’
While it is clear that an event involving cooking and eating can be assumed
in (39), such cannot be the case in (40) unless we assume cannibalism. But for
cannibals too, it makes no sense to literally cook one’s own soldiers before
going to war. Things being so, it must be the case that the intended meaning of
ɖà in (39) and (40) does not imply cooking in the culinary sense but rather ‘to
prepare’ whose extension could include ‘to cook.’ I therefore hypothesize that
the lexical verb ɖà in Gungbe, corresponds to the English verb to prepare (i.e.,
‘to make something ready for use,’ The New Oxford Dictionary of English).
Once we allow ourselves to make this assumption, series such as (39–40) can
be analyzed as hidden causatives that actually mean ‘to make/get/cause beans
(to be) ready to eat’, or ‘to make/get/cause the soldiers (to be) ready for war’
where only the meaning prepare is available.
Now, consider the verb ɖù, which in its lexical use in (39) means ‘to eat’ but
when used as V1 in (41) can be followed by various elements that are not
eatable in the literal sense, including for instance, the DP tò lɔ́ ‘the country’ (see
Section 4.3 for further examples). It is obvious from this example that the DP
complement cannot be said to receive the θ-role Theme from V1 (i.e., the Agent
did not literally eat the country).
(41) Yé ɖù [tò lɔ́ ] vɔ̀ .
3PL eat country DET finish
‘They ate the country finish [they ruined the country misappropriating funds].’
These facts seem to me compelling enough to conclude that, when used

functionally in series, Gungbe verbs are immediately followed by a wide range
of constituents which cannot fulfill the semantic function of Theme. In this view,
the examples discussed here are similar to take-series where we saw that V1
take did not θ-mark the DP-object to its right. The unifying property behind all
these cases is that the lexical meaning corresponds to situations in which the verb
selects for a DP internal argument to which it assigns an internal θ-role. The
functional meaning, however, coincides with situations in which V1 has no
internal θ-role to assign, but selects for a complement inside which the element
to its right is being licensed.
This last point makes SVCs superficially similar to OV sequences in Gbe
which encode various aspect specifications (42), as explained by Aboh (2003a,
2004a, 2005c, 2009) and Aboh and Smith (2012):13
(42) a. Mìgàn yì kpònɔ̀ n lέ yrɔ́ gbé.
Migan go soldier PL call PURPOSE
‘Migan went to call the soldiers.’
b. Mìgàn tò kpònɔ̀ n lέ yrɔ́
̀
Migan go soldier PL call PCL
‘Migan went to call the soldiers.’
13
In Gungbe progressive constructions like in (37b), the sentence-final particle is a floating tone
(Aboh 2004a).
Comparing SVCs and these OV constructions as analyzed in previous work,

Aboh (2009b) argues that the so-called consecutive series with one arguments
are derived as in (43) by movement of the object of V2 to [spec AspP] due to
EPP-licensing.
(43) [TP Àsíbá [AspP tÀsíbá [Asp° ɖà [FP [AspP lέsì [Asp° ɖù [vP[v-ext tÀsíbá [vP[v-appl tɖù
[VP2 tɖù tlέsì]]]]]]]]]]]
According to representation (43), SVCs arise as a combination of a lexical
verb V2 that introduces (all) the argument(s) and a functional verb V1 that
encodes the manner in which the event expressed by the lexical verb is
achieved. This analysis implies that SVCs are types of causative
constructions.
The representation in (43) extends to instrument SVCs in a
straightforward manner. With regard to sentence (44a), for instance, I
argue that V2 merges with the theme to form VP2. Under aspect licensing
and the EPP, V2 raises to Aspo (via v-appl and v-ext) to check its aspect
features, followed by movement of the instrument to [spec AspP].
Accordingly, the conjunction of verb-movement and verb object inversion
gives rise to the sequence V1-(XP)-V2 that is often found in SVCs in these
languages. (44a) is derived as in (44b).
(44) a. Sέtù zé kpò lɔ́ xò Kɔ̀ jó.
Setu take stick DET hit Kojo
b. [TP Sέtù [AspP [Asp° zé [FP [AspP kpò lɔ́ [Asp° xò [vP tSέtù [v-ext txò [vP tkpò lɔ́ [v-appl
txò [VP2 txò Kɔ̀ jó]]]]]]]]]]]
It appears from this analysis that, when the instrument/comitative argument
and the direct object co-occur, the former merges higher where it also checks its
case features. Accordingly, the instrument, being closer, appears to be the
legitimate goal that can check the EPP feature under Aspo. This analysis also
assumes that v-ext (associated with V2) introduces the external argument, but
the latter must raise to [spec TP] to check off case/agreement and the EPP
features under T (see Aboh 2003a, 2009b for discussion). This movement
allows that the subject DP Sέtù is understood as the cause of the hitting (i.e.,
the external argument of xò ‘hit’).
I will not discuss all the details of the analysis developed in Aboh (2009b)
here. What is relevant to the current discussion is that series appear akin to the
French (45a) and Hungarian (45b) causative constructions.
(45) a. Marie a fait traverser la ville à Pierre.
Marie have make cross the city to Pierre
‘Marie made Pierre cross the city.’
7.4 Serialization: functional verbs and lexical verbs 291
b. Péter Elmeséltette Pállal a történetet.

Péter pv-tell-CAUSE.PAST.DEF.3SG Pál-WITH the story-ACC
‘Péter made Paul tell the story.’
These examples show that, in many languages, the causee is a syntactic
dative or comitative argument, even though it is understood as the subject of the
embedded reduced clause. If we ignore morphosyntactic differences between
these languages, the data just discussed in Gungbe appear structurally similar to
the causatives exemplified in (45), where the causee has a dative or comitative
case. The main difference between Gungbe and these languages is that the
Gungbe causee must raise to [spec AspP] due to the EPP.
In sum, the proposed analysis indicates that SVCs are a sub-type of clause
union phenomena (commonly analyzed in generative syntax as restructuring
constructions) which take various forms cross-linguistically (e.g., causatives,
instruments). Morphology type aside, the view developed here suggests that the
combination of verbs in languages like Gungbe to form series cannot be
explained by a peculiar parameter specific to serializing languages that would
allow the structural ‘fusion’ of two lexical verbs. Instead, the combination of
such verbs in SVCs is explained by the ability of certain lexical verbs to directly
merge in a functional position. The phenomenon is not exceptional because it is
made available by UG and can be found in typologically different and
genetically unrelated languages (including Germanic and Romance). This in
turn implies that the so-called serializing languages ‘look’ more serializing
than say English or French, simply because their lexicons allow more
functional verbs.
7.4 Serialization combines a functional verb

with a lexical verb
As Aboh (2003a, 2009b) concludes, the facts just discussed appear even more
striking in languages like Gungbe because the functional verbs have exactly the
same morphological shape as their lexical cognates. Consider again the
following example, where the derived element is in italic and the source in
boldface.
(46) Hwĕnὲnù wὲ ǎ wá wá nɔ̀ jró ná nɔ̀ nɔ̀ dèyè.
at.that.time Suru 1SG.FUT EVENT come HAB want to HAB stay at.my.place
‘AT THAT TIME, you will eventually come to the point that you will want to habitually
stay at my place [i.e., you will eventually agree to the decision to often stay at my
place].’
This example shows that the verb forms wá ‘come’ and nɔ̀ ‘stay’ also have
functional cognates. These verbs are therefore similar to Cinque’s (2004)
functional verbs, which correspond to distinct heads in the clause structure. As

Cinque shows, while this is usually taken to be true of auxiliaries in general, it
also holds of lexical verbs (e.g., motion verbs). Cinque (2004: 12) further
concludes that
only those verbs that happen to match semantically the content of a certain functional
head admit the two distinct possibilities. They are either regular verbs, heading a
VP (in which case they take a full-fledged sentential complement [CP]), or
functional verbs, directly inserted in the head position of the corresponding functional
projection.
The proposed account for SVCs not only complements Cinque’s view on
clause union but also implies that in V1-XP-V2 combinations, only elements of
the type V1 fall in the class of functional verbs sometimes also referred to as
semi-lexical items (see also Da Cruz 1993, 1995, 1997; Hagemeijer 2001;
Cardinaletti and Giusti 2001; Cardinaletti and Shlonsky 2004).
If we extend the proposed analysis to creole languages, we can conclude that
example (47a) from Haitian is derivable as in (47b):
(47) a. Boukinèt te pran yon flè bay Malis.
Bouki TNS take DET flower give Malis
‘Bouki gave a flower to Malis.’
b. [TP Boukinet [T° te [AspP [Asp° pran [FP [AspP yon flè [Asp° bay [vP tBoukinèt [v-ext
tbay [vP tyon flè [v-appl tbay[VP2 tbay Malis]]]]]]]]]]]]
Similarly, the Saramaccan example (48a) can be partially represented as
in (48b):
(48) a. A téi dí fáka kóti dí beée. [Veenstra 1996: 4]
3SG take DET knife cut DET bread
‘He cut the bread with the knife.’
b. [TP A [T° [AspP [Asp° téi [FP [AspP di fáka [Asp° kóti [vP tA [v-ext tkóti [vP tdi fáka
[v-appl tkòti [VP2 tkòti di beée]]]]]]]]]]]]
Assuming this is the right characterization, and in light of Cinque (2004),
the question naturally arises whether we can still consider the simple presence
of SVCs in creoles as proof of substrate transfer. Let us, for instance, consider
the following examples from the superstrate languages, viz., French and
English (49):
(49) a. J’ai accompagné Jean voir ses enfants.
1SG.have accompanied John see POSS. Children
‘I accompanied John to visit his children.’
b. Let me go buy some rice.
These examples can be translated in the substrate languages, e.g., Gungbe,
as in (50):
7.4 Serialization: functional verbs and lexical verbs 293
(50) a. Un kplán Ján yì kpɔ̀ n ví étɔ̀ n lέ.

1SG accompany John go see child POSS PL
‘I accompanied John to visit his children.’
b. Gbɔ̀ má yì xɔ̀ lέsì.
let 1SG.MOOD go buy rice
‘Let me go buy some rice.’
The Saramaccan and Haitian equivalents are in (51).
(51) a. A tja en go a di data. [Saramaccan]
3SG carry 3SG go LOC DET doctor
‘He took her to the doctor.’
b. Li mennen l ale kay-doktè. [Haitian]
3SG take 3SG go house-doctor
‘He took her/him to the doctor.’
If I am right in assuming that SVC structures are just another type of clause
union phenomena (compared to their Romance and Germanic equivalents), it is
not immediately clear whether the creole speakers produced the examples in
(51) on the basis of the structures in (49) or (50), or whether they were using a
combination of both grammars. In what follows, I will distinguish descriptively
between serialization and clause union in Romance and Germanic on the basis
of the point of first merge of functional verbs in the clause spine. As I show
below, ‘serialization’ refers to a sub-case of clause union.
Cinque (1999, 2004) proposes the following hierarchy for aspectual and root
modal heads that may host restructuring verbs cross-linguistically:
(52) MoodPSpeech-act>MoodPEvaluative>MoodPEvidential>MoodPEpistemic>TPPast>
TPFuture>MoodIrrealis>MoodAlethic>AspPHabitual>AspPRepetitive(1)>
AspPFrequentative>MoodPVolitional>AspPCelerative(1)>TPAnterior>
AspPTerminative>AspPContinuative>AspPRetrospective>AspPProximate>
AspPDurative>AspPProgressive>AspPProspective>MoodPObligation>
MoodPPermission>AspPCompletive>[Voice]>Perception>Causative>
AspPCelerative(2)>AspPRepetitive(2)>AspPFrequentative(2)
Comparing my analysis of serialization in Gbe, where the process mainly
builds on causatives and resultatives, to Cinque’s view on clause union in
Romance and Germanic along the hierarchy in (52), it appears that even
though the two phenomena involve first merge of lexical verb in functional
projections, these functional projections differ both in their semantics and
regarding how high they occur in the clause. More precisely, I suggest that
serialization in Gbe occurs when lexical verbs first merge outside the lexical
domain, but not higher than the Voice phrase in (52). Coupled with the absence
of V-to-T movement in these languages, this clause union process yields verb
sequences where the functional verb and the lexical one are in a very local
configuration, sometimes separated by an intervening DP only: V1-(XP)-V2. I

will refer to this case as low restructuring (i.e., a functional verb first
merges in a position below Voice in 52). In this analysis, restructuring
refers to the ability to merge (or base generate) a lexical verb in a functional
position. In Romance and Germanic, clause union seems able to target higher
portions of the clausal spine (either by first merge or movement). I refer to this
latter case as high restructuring (i.e., a functional verb first merges in a
position above Voice in 52).
Given this reinterpretation of the formulas in (52), it appears that Romance
and Germanic languages differ from the Kwa languages in that the former
mainly allow high restructuring, while the latter are very productive in
low restructuring. The two language types also differ because the type of
clause union they exhibit depends directly on the semantics of verbs that enter
this process. Whether high or low, only verbs corresponding to particular
meanings on the clause spine will have the ability to first merge in a
corresponding aspect or modal node (Cinque 1999, 2004).
Bearing this discussion in mind, and assuming that the creoles discussed here
express various recombinations of features from their source languages, it
appears justified to ask how one can reconstruct the distinctive features of
these languages when it comes to SVCs or, more specifically, low
restructuring. How can we identify Gbe features that are retained in
Haitian Creole and Saramaccan, and can we articulate how these recombined
with French and English features respectively?
7.5 The proof is in the pudding

Given what precedes, it appears that clause union is a universal aspect of
languages though it may lead to various structural combinatorics cross-
linguistically. This conclusion would seem to favor claims that SVCs
emerged in creoles as a direct consequence of unmarked features of UG. This
is precisely the hypothesis that this book discourages. Instead, the discussion in
previous chapters builds on Mufwene’s (2001, 2008) ecological approach to
language evolution in assuming that language acquisition, language change,
and the creation of a new language, arise from the competition and selection of
linguistic features from a feature pool determined by the language contact
setting.
In this context, and assuming that formal features have semantic correlates
reflecting semantic properties (Chomsky 1995: 381, footnote 14), I have argued
throughout this book that feature retention (or language transfer) results from a
competition among syntactic features that may lead to fission between their
syntax and semantics. Thus, in a situation of language contact, functional
categories (the landmarks of clause structure) are affected differently
7.5 The proof is in the pudding 295
depending on their semantics and licensing conditions (i.e., their syntax).

Accordingly, an emerging contact language (call it a creole) may recombine
the syntactic and semantic features differently from the source languages.
Within Minimalism, such a split between syntax and semantics is plausible if
we assume that a formal feature (F) is associated with a unique semantic
representation cross-linguistically, while its syntax is subject to parametric
variation. This would mean that the same feature may be valued differently
cross-linguistically, while its semantics remains the same. In a contact situation,
competition arises about how the relevant parameters can be reset, especially
when the settings of the different languages are not the same. Consistent with this
view, the hypothesis that serialization/clause union results from the first merge of
certain functional verbs in semantically matching functional projections provides
us with a nice tool to measure language transfer and feature recombination.
Indeed, retention of clause union can be measured not only by looking at
superficial structural similarities between the source languages and the creoles
but also by (i) determining the semantic properties of verbs that enter the
restructuring process and (ii) identifying the types of recombinations of
features (semantic and syntactic) that are found in the creoles. Some of the
features could not have arisen individually in the source languages, nor in the
emerging creole as a mere accident of UG.
Accordingly, it appears that Romance and Germanic languages allow the
restructuring of modals and aspect verbs less readily than of lexical verbs.
Thus, a sentence like (53a) in French is ungrammatical as opposed to its
Gungbe and Haitian equivalents in (53b–c):
(53) a. *Jean a pris l’ordinateur a donné à un ami.
John has taken the.computer has given to a friend
b. Ján zé ɔ̀ dínátέὲ lɔ́ ná xɔ́ ntɔ̀ étɔ̀ n.
John take computer DET give friend his
‘John has given the computer to his friend.’
c. Jan pran odinatè a bay Zanmi.
Jean take computer DET give friend
‘John has given the computer to a friend’
(53a) is disallowed even though French allows clause union or
restructuring of the type in (54a). On the other hand, the English
structures in (54b) are analyzed by Cable (2004b) as involving
restructuring.14
(54) a. Il faut très bien que tu te comportes. [Cinque 2004: 102]
b. We tried/went eating a duck.
14
See also Jaeggli and Hyams (1993) for a discussion on aspectual go and come, which is
compatible with the analysis in terms of restructuring.
Gbe and most Kwa languages freely allow restructuring of aspectual and
lexical verbs:
(55) a. Àhwàn éhè ná wá vɔ̀ gbèɖòkpó.
war DEM FUT come finish one.day
‘This war will eventually end one day (i.e., sooner or later this war will come
to an end).’
b. Míɔ́ n nyàn kànlìn lέ tɔ́ n sɔ́ n zùgbó mὲ.
fire chase animal PL exit from jungle in
‘The fire drove the animals out of the jungle.’
Given the view of creolization adopted in this book, it is not surprising that
the creoles discussed here, which emerged from the contact of largely Gbe
(Kwa) with Romance and Germanic languages, readily display both types of
restructuring found in the source languages (i.e., low and high
restructuring). Recall from the preceding discussions that Haitian mostly
displays verb series of movement and transfer as illustrated again in (56a) and
(56b), respectively:
(56) a. Bouki voye timoun yo ale lekol. [DeGraff 1992]
Bouki send child PL go school
‘Bouki sent the kids to school.’
b. Boukinèt te pran yon flè bay Malis.
Boukinet TNS take DET flower give Malice
‘Boukinèt gave Malis a flower.’
Compared to their Gungbe counterparts in (57), these examples could be
taken as strong evidence of substrate transfer of an SVC structure from the Gbe
languages into Haitian:
(57) a. Súrù zɔ́ n ví lέ yì wéxɔ̀ mὲ.
Suru order child PL go school
‘Suru sent the kids to school.’
b. Súrù zé vònvó ná Dóná.
Suru take flower give Dona
‘Suru gave Dona flower(s).’
However, some details about serialization (or low restructuring) in
Haitian cast doubt on such a one-to-one mapping. Rather, they indicate that the
Haitian structures display both Kwa and Romance properties, as can be expected
in a theory that evokes recombination of semantic and syntactic features as
argued for here. Indeed, while Gungbe (and other Gbe languages) can use the
verb take in take-series freely (58a), Haitian pran cannot always enter such
combinations. Crucially, some of my informants reject such constructions
altogether, while others reject take-series where the instrument is a full DP
(58b) (see also Lefebvre 1989).15 In the context of this discussion, it is important
to keep in mind that Gbe languages do not show such variation and allow take-
series freely. Once again, we face a situation where the creoles and their relevant
substrates are not isomorphic, though they share many similarities.
(58) a. Súrù zé kpò lɔ́ /ɖé xò Dóná.
Suru take stick DET hit Dona
‘Suru hit Dona using the/a stick.’
b. Boukinèt te pran baton (*na) frappe Malis.
Boukinèt TNS take stick DET hit Malis
‘Boukinet hit Malice using the stick.’
c. Boukinèt te pran (*yon) baton frappe Malis.
Boukinèt TNS take DET stick hit Malis
‘Boukinet hit Malice using a stick.’
In order to realize the equivalent of (58b), some Haitian speakers resort to
instrument complements introduced by ak (59a). Example (59b) shows that this
construction is parallel to French instrument constructions involving avec, the
source of the Haitian preposition.
(59) a. Boukinèt frape Malis ak yon baton. [Haitian]
b. Boukinèt a frappé Malis avec un bâton. [French]
Boukinet has hit Malis with a stick
‘Boukinet hit Malis with a stick.’
Interestingly enough, similar prepositional instrument structures are found in
Gbe along with the series in (60). In this case, Gungbe (as well as other Eastern
Gbe) display adpositions that bracket the noun phrase instrument (a property
absent from the creoles).
(60) Súrù xò Dóná kpó àfɔ̀ kpà kpó.
Suru hit Dona PREP shoe PREP
‘Suru hit Dona with a shoe.’
As already explained at the outset of this chapter, the development of the
series in a language cannot be taken as a consequence of the absence of
prepositional structures in that language (pace Bickerton 1981). In addition,
one should also note that the Haitian instrument example (59a) is akin to the
French example (59b) as opposed to the Gungbe example in (60). Together,
all these facts indicate that serialization is not a default (or unmarked) UG
strategy to compensate for the lack of prepositional phrases. Instead,
serialization, as understood here, appears to be a type of clause union that
15
I thank the Master students (M1 & M2) of the FLA at the Université de Port au Prince for
confirming these findings during my seminar, June 22 and June 29, 2014.
combines a lexical verb and a functional verb. This conclusion in turn

suggests that there may not be ‘a serializing parameter’ as often assumed in
the relevant literature (e.g., Baker 1989 and much related work), as explained
in Aboh (2003a, 2009b).
Continuing our discussion of series in Haitian and Gungbe, note that
while Gungbe can use the verb ‘order’ in (61a), Haitian lacks such a
combination, where the apparently shared object precedes the second
verb (61b). Instead, my informants offered example (61c), where this
object follows the second verb or alternatively example (61d) which is a
bi-clausal structure.16
(61) a. Súrù zɔ́ n lέsì ɖà Dòsà.
Suru order rice cook Dosa
‘Suru ordered Dosa to cook rice.’
b. *Bouki voye diri kuit Malis.
Bouki send rice cook Malis
c. Bouki voye Malis kuit diri.
Bouki send Malis cook rice
‘Bouki sent Malis to cook rice.’
d. Bouki voye Malis pou li kuit diri.
Bouki send Malis PREP 3SG cook rice
‘Bouki sent Malis to cook rice.’
This contrast appears to be related to the fact that Haitian lacks the so-called
consecutive serialization (62a) which is very productive in Gbe (62b).17
(62) a. *Bouki kuit mannyòk manje. [Haitian]
Bouki cook cassava eat
b. Súrù ɖà fέnyέn ɖù.
Suru cook cassava eat
‘Suru cooked cassava and ate.’
These examples show that the types of combinations found in the Gbe
languages are not transferred wholesale to Haitian. Gungbe and Haitian are
not isomorphic (DeGraff 2002, 2005), nor are Gungbe and Saramaccan. Recall
16
The ungrammatical example (56b) should not be confused with examples like the following
which I claim must be analyzed on a par with the English or colloquial French equivalent
restructuring structures in (ii) (Jaeggli and Hyams 1993; Aboh 2009b).
(i) Bouki voye Malis al kuit diri.
Bouki send Malis go cook rice
‘Bouki sent Malis to go cook rice.’
(ii) Bouki a envoyé Malis aller cuire du riz.
17
Example (57a) is still bad with the verb ‘prepare’ (DeGraff, p.c. June 21, 2009)
from our discussion of the Gungbe examples in (38) and (39) that the verb ɖà
does not mean ‘cook’; rather it corresponds to English ‘prepare,’ whose
extension can still include a cooking activity. This appears to explain why
Haitian grammar does not generate sentences like (62). Indeed, it appears that
Haitian, like French and English, has three such lexical verbs with different
meanings: kuit ‘to cook,’ préparé ‘to prepare,’ and paré ‘to prepare in the sense
of to get ready’ (Hall 1953). Given these distinctions in Haitian, we expect this
language to distinguish morphologically between causative ‘prepare’ and the
lexical verb ‘cook.’ This semantic opposition is not possible in Gungbe,
because the two meanings and usages are conflated in a single form ɖà.
Such lexical specifications might also be at the source of the discrepancies we
observed between Haitian and the Gbe languages throughout this chapter. In
addition, Haitian displays new lexico-semantic combinations under the influence
of French (viz., French lexical properties). This is so even though one sometimes
comes across very striking parallels of the type in (63) between the Gungbe
functional verb jὲ , which contributes to the semantic meaning ‘to fall’ (63a), and
its Haitian equivalent tonbe in (63b). Gungbe and Haitian differ because the
Gungbe verb is an inherent complement verb (i.e., it requires an object in its
citation form), while the Haitian verb is similar to French tomber in the sense that
it may or may not be used with a complement (actually a prepositional phrase). In
Haitian, the preposition is incorporated in the noun, unlike in French (63b–c).18
Therefore, French and Haitian contrast with Gungbe because the internal
argument cannot remain unrealized in this language (63a):
(63) a. Súrù jὲ *(àyí).
Suru V ground
‘Suru fell.’
b. Boukinet tonbe (atè).
Boukinet fall ground
‘Boukinet fell.’
c. Boukinet est tombé (*(par) terre).
Boukinet is fall at ground
‘Boukinet fell on the ground.’
However, in both Gungbe and Haitian, these verbal elements can be used
in inceptive constructions of the type illustrated in the pairs (64a–b) and
(64c–d).
(64) a. Súrù jὲ ògán bàí jí. [Gungbe]
Suru V chief make PCL
‘Suru started behaving as a chief.’
18
I thank Herby Glaude for bringing these data to my attention.
b. Jan tonbe joure Prezidan. [Haitian; Glaude p.c. November 10, 2008]
Jean fall insult president
‘John started insulting the president.’
c. Ùn jὲ àví ví jí. [Gungbe]
1SG V cry cry PCL
‘I started crying.’
d. M tonbe kriye, m di “Mama, li pran ti-plat mwen. [Haitian]
I fall cry 1SG say Mama 3SG take little-dish 1SG
‘I started crying: “Mama, he took my little dish.”’
[Adapted from Hall 1953: 75]19
If creoles have hybrid grammars in the sense that they recombine syntactic
and semantic features from both the substrate and superstrate languages, we
expect them to show a ‘family resemblance’ with the source languages, hence
the (dis)similarities observed here (Mufwene 2001, 2008). According to the
biological analogy adopted here, what matters is that the ecology of contact
makes possible new linguistic recombinations which, though absent in the
source languages, may appear in the creole language.
In the case at hand, this would mean that Haitian verb combinations derive
from a recombination of, on the one hand, French processes such as auxiliation,
causative construction, and high restructuring, and, on the other, Gbe low
restructuring (i.e., serialization) patterns. Put together, these combinatorial
possibilities allow Haitian to exhibit verb sequences that are absent from both
French and Gungbe. Such an unexpected recombination appears in the form of
the verb bay in its causative usage.20 I showed in the above discussions that
Haitian can use this verb in transfer series, in which it occurs as V2 (65a). This
usage is clearly parallel to the pattern displayed by the Gungbe equivalent in
(65b). Note, in addition, from the French example (65c), that similar
constructions are excluded in this language.
(65) a. Mari achete liv la bay Jan.
Mary buy book DET give John
‘Mary bought a book for John (i.e., she gave him the book).’
b. Dòsà xò wémà lɔ́ ná Dòsì.
Dosa buy book DET give Dosi
‘Dosa bought a book for Dosi (i.e., he gave her the book).’
19
I thank Michel DeGraff for helping me adapt this example.
20
These data were first brought to my attention by Herby Glaude (November 10, 2008) and then
confirmed by Michel DeGraff (June 21, 2009). I thank both of them for their suggestions on
serialization in Haitian.
c. *Jean a acheté le livre donner à Pierre.

John has bought DET book give to Peter
Instead, French displays (among other causative constructions) structures
such as (66a), in which the dative is the beneficiary, or (66b), in which the
dative is understood as the subject of the verb of the embedded clause:
(66) a. Jean a fait acheter le livre pour Pierre.
John has made buy DET book for Peter
‘John got the book bought for Peter.’
b. Jean a fait acheter le livre à Pierre,
John has made buy DET book to Peter
‘John got Peter to buy the book.’
Unlike the Gbe languages, Haitian too involves faire-like causatives, of the
type shown in (67):
(67) Jan fè Mari pote liv la pou li.
John make Mary carry book DET for/to 3SG
‘John made Mary carry the book for him.’
Yet, unlike both French and Gungbe, Haitian also displays causative
constructions in which the causative head is expressed by the verb bay ‘to
give’ (68):
(68) a. Jean bay Mari pote liv la (pou li).
John give Mary carry book DET PREP 3SG
‘John made Mary carry the book for him (i.e., John caused Mary to carry the
book for him).’
b. Jan bay Mari achte diri pou li.
John give Mary buy rice PREP 3SG
‘John made Mary buy rice for him.’
As DeGraff (p.c. June 21, 2009) remarks, this usage of the verb bay is
parallel to the usage of causative fe or fòse in Haitian, even though these
verbs do not show the same properties. Consider the following contrast:
(69) a. Mari te bay pitit la bwè lèt,
Mary TNS give baby DET drink milk
men pitit la jete lèt la.
but baby DET throw milk DET
‘Mary made the baby to drink the milk, but the baby threw the milk away.’
b. *Mari te fè pitit la bwè lèt,
Mary TNS make baby DET drink milk
men pitit la jete lèt la.
but baby DET throw milk DET
‘Mary made the baby drink the milk, but the baby threw the milk away.’
A causative introduced by bay ‘give’ does not imply ‘single event reading’:
all sub-events need not be true at once (69a). A causative introduced by fè, (or
fòse), on the contrary forces single event reading. All sub-events must be true at
once and cannot be negated individually. Example (69b) is ungrammatical
because it is a contradiction. The first part of the clause implies that the baby
drank the milk, while the second part says that she didn’t. French faire
causatives also imply ‘single event reading’.
Putting these together, the fact that Haitian displays a bay causative that is
formally different from French-type fè/fòse causatives and inexistent in Gbe
would seem quite exceptional, if we were to assume substrate or superstrate
influence only. In the context of this discussion, however, what we have here
appears to be an instance of a new recombination of syntactic and semantic
features from the source languages. Indeed, bay causatives and bay series allow
the same lexical element to first merge in two functional positions which
correspond to the low and high restructuring found in Gungbe and
French, respectively. I therefore conjecture that the fact that bay can be used
in restructuring (similarly to the verb ‘give’ in Gbe series) paved the way
for its usage in high restructuring, similarly to French faire and forcer
causatives. As the reader may see, this new development in Haitian led to the
emergence of local complexity with regard to restructuring.
Quite interestingly, the discussion in the previous paragraphs reminds us of
Veenstra’s (1996) analysis of the Saramaccan verb mbéi, which has the
distributive properties illustrated in (70):
(70) a. A bì tá mbéi témbe. [Veenstra 1996: 158ff.]
3SG TNS ASP make wood
‘He was making wood-carvings.’
b. Dí kái u dí tjúba mbéi hen uwíi munjá tooná kó bè.
DET fall of DET rain make 3SG hair wet turn come red
‘The rainfall made her hair wet and turn red.’
c. Dí tjúba tá kái mbéi hen uwíi munjá tooná kó bè.
DET rain ASP fall make 3SG hair wet turn come red
‘It is raining so that her hair becomes wet and turns red.’
In analyzing the various usages of this verb, Veenstra (1996: 163) concluded
that “Saramaccan has two (distinct) lexical entries for the formative mbéi: one
for a (causative) verb meaning ‘make’, which can also participate in light verb
constructions and one for a complementizer meaning ‘so that’ expressing
causality.” According to Veenstra, the causative verb mbéi corresponds to
cases like (70b), in which the selected complement is a small clause or a
reduced IP. On the other hand, the causative complementizer mbéi
corresponds to example (70c), in which the embedded clause is a full CP.
7.6 Conclusion 303
The two instances of mbéi differ in that the former can combine with an aspect
marker (71a), while the latter cannot (66b) (Veenstra 1996: 162):
(71) a. Dí kái u dí tjúba tá mbéi hen uwíi munjá tooná kó bè.
DET fall of DET rain ASP make 3SG hair wet turn come red
‘The rainfall is making her hair wet and turn.’
b. *Dí tjúba tá kái tá mbéi hen uwíi munjá tooná kó bè.
DET rain ASP fall ASP make 3SG hair wet turn come red
What we see here is therefore another instance in which a verb has two
usages: one lexical and the other functional. Consistent with the analysis
of SVCs as restructuring structures proposed here, the behavior of
Saramaccan mbéi is to some extent similar to that of Haitian bay,
which also appeared to target two different sites in the functional
sequence.
7.6 Conclusion
This chapter has shown that the study of SVCs in certain creoles and their
source languages can help in understanding language birth and the
recombination of syntactic and semantic features into a new idiolect.
Though it could be argued in some specific cases that SVCs reflect
substrate transfer, it is not enough to limit oneself to surface
correspondences between the creoles and their source languages. The
reason is that there appears to be no serializing parameter that could be
isolated as responsible for the emergence of serialization in creoles.
Similarly, it cannot be claimed that the phenomenon is linked to the lack of
V-to-T movement in these languages. Indeed, the Gbe languages display
short V-to-Asp movement as well as serialization. Given that SVCs are
regarded as instances of restructuring constructions, their distributions
across languages appears more general and such constructions are found in
both the lexifiers and the substrate languages, although the congruence is not
total. In such a context, great analytic precision is needed to isolate the
linguistic features (syntactic and semantic) that are recombined into the
creole. In light of the above discussions, it appears that the debate
regarding whether or not SVCs are found in Indian Ocean Creoles (e.g.,
Bickerton 1989, 1990b versus Seuren 1990a, 1990b, 1991) is misguided. The
question should be reformulated as whether Indian Ocean Creoles have
clause union phenomena of the types analyzed here as restructuring,
and, if they do, whether they display low or high restructuring.
8 Conclusions: some final remarks on hybrid
grammars, the creole prototype, and language
acquisition and change
The discussion in this book shows that during the seventeenth and eighteenth
centuries, the Kingdom of Allada on the Slave Coast played an important role in
the Triangular Trade. During this period, socio-political and economic factors
in the Bight of Benin contributed to making the locals, namely people of Aja
descent, victims of a globalized world economy in which enslaved Africans
were treated as livestock and part of the goods exchanged. Our conclusions
drawn on the basis of both geopolitics and socio-economic factors corroborate
work on population factors in the Caribbean and in Suriname according to
which speakers of Kwa languages represented a significant part of slave cohorts
exported to the Caribbean plantation colonies. While these Kwa people,
including significant numbers of Gbe speakers, probably did not always
constitute the most numerous ethnic group in the Caribbean and in Suriname
during the slavery period, they appear to have been instrumental in shaping the
culture of the emerging colonies. Two related factors underlie this view:
(i) more and more Kwa/Gbe people were enslaved during the transition of the
colonies from the homestead to the plantation phase, when there was demand
for larger numbers of slaves; and (ii) the Gbe/Kwa people originated from a
homogeneous cultural background, a factor that probably allowed them to form
tighter sub-communities (compared to other West African or Central African
slaves). The structure of such communities could have helped them exert more
convergent and thus stronger pressures on the emerging linguistic patterns
during the early stages of the nascent language. Accordingly, the ecology of
the plantation communities in Haiti and Suriname favored speakers of Gbe/
Kwa languages and enabled them to contribute a lasting founder effect.
While this perspective is compatible with the competition and selection
model argued for in this book, and in previous studies such as Mufwene
(2001, 2008) and Ansaldo (2009), it also suggests that monolithic theories
that account for creole genesis exclusively by invoking inheritance from the
superstrate, substrate influence, language universals, or fossilization of some
early interlanguage stage are untenable. Instead, the discussion in this book
shows that creole languages emerged from the recombination of linguistic
304
8.2 Evaluating complexity and simplicity 305
features from typologically and genetically unrelated languages that came in

contact during the colonial period. Accordingly, creole languages are linguistic
hybrids.
8.1 Hybridism is the norm

Since this book draws on creoles as test bed, the reader could conclude that only
creoles, the so-called contact languages or mixed languages, develop hybrid
grammars. Such a conclusion would be mistaken. As I argue in this book,
linguistic recombination happens in all contexts of contact and can target all
modules of grammar to different degrees. Because linguistic recombination
generates a new form out of two (or more) independent linguistic units, its
output is hybrid by definition. In addition, I have shown that the process
is selectively sensitive to interface properties: the syntax–discourse interface,
the syntax–morphology interface, or the syntax–semantics interface. The
discussion further indicates that linguistic recombination typically leads to
local complexity because the newly created form displays opaque syntactic
and semantic features which could not have arisen only in the context of their
source languages.
Under this view, linguistic hybridism is the norm in every instance of
language acquisition and change at the individual level, a process that may
subsequently lead to language change at the population level. The case of
creoles looks striking at first sight simply because the recombination involves
linguistic features from typologically and genetically unrelated languages. The
same holds true of contemporary (multi)-ethnolects which developed in the
context of recent migrations of speakers of non-European languages to various
European urban cities. Like creoles, these new varieties are noticeable only due
to their (contrasting) phenotypes. Nothing in their structures singles them out as
a prototype of language creation.
8.2 Evaluating complexity versus simplicity in new languages

Given debates on whether or not creoles have simple grammars and on creole
prototype in creolistics and typological studies during the past fifteen years, as
well as the concomitant discussion on imperfect second language acquisition,
the conclusion that creoles are not simpler languages may seem a paradox.
Indeed, even though the notion of creole simplicity dates back to the colonial
period, the question of creole simplicity and prototype has received
considerable attention recently. An idea central to the notion of creole
simplicity and prototype is the assumption that these contact languages are
too young to develop the kind of opacity and irregularity commonly observed
in old languages. Over the past years, there have been various critiques of
306 Some final remarks on hybrid grammars
studies which postulate creole simplicity and prototype thus making creole
languages an exceptional typological class. These critiques show that ideas
about creole simplicity and prototype usually ignore the socio-historical facts
discussed in Chapter 2 of this book and further rely on mere morphological
counts measuring morphological size and shape (i.e., what DeGraff refers to as
“bit complexity”) as well as trivial surface syntax. For instance, it is not
uncommon to see complexity metrics defined in terms of the size of the class
of marked members of a phonemic system, the number of tonemes, or the
number of distinct grammatical or syntactic rules in a language.
It is obvious that such informal metrics fail to take into account a
fundamental aspect of language: the relation between form and meaning. As
already mentioned in Aboh and Smith (2009) and Aboh (2009a), linguistic
complexity cannot be evaluated without factoring in the relation between form
and meaning, and how such relations compare cross-linguistically. For
instance, it could be argued that a complex form is one that results in
excessive processing cost in evaluating its meaning. Accordingly, in order to
show that an aspect of a language is more complex than that of another, we
would have to show that a given form in one language triggers excessive
processing cost compared to its equivalent in the other language (cf. Hawkins
2004). Thus, by implication, a form can be assumed to be simple if it yields
little (or almost no) processing cost. An evaluation method along these lines
predicts that a simple aspect of a language will typically result in fast
computation of the meaning, while a complex aspect on the contrary will
result in slow computation.
Such an evaluation of complexity and simplicity can be carried out thanks to
new experimental methods in psycholinguistics as well as new computational
models of language. Scholarly exchanges between experts of language
acquisition and evolution, psycholinguists, and computational cognitive
scientists will undoubtedly help us better understand human cognition, and
shed light on how issues of linguistic structural complexity are related to
typological variation. Note, however, that the conclusion I just made about
slow versus fast computation as potentially related to structural complexity
relies on the crucial observation, often ignored in theories of creole simplicity,
that a meaningful notion of complexity can only be one that is relative to a
given module of a grammar, rather than to a grammar as a whole. It seems to
me unlikely that a language will be either overall complex and therefore
allows costly processing and slow computation in all aspects of its grammar,
or alternatively simple and permits easy processing and fast computation in
all components of its grammar. Both excessive complexity and excessive
simplicity would prevent a language from meeting the requirement of
learnability and easy processability, and that of satisfying the often
complex communicative needs of its users, respectively.
Aside from the fact that no serious study as that proposed here, viz., one in
which theory informs experimental work to investigate the relation between
structure, meaning, and complexity, has been conducted on creole languages,
this conclusion further indicates that claims about simplicity as an across-
the-board property of creole languages cannot be correct. Indeed for such
claims to be taken seriously, they would have to show on empirical and
experimental grounds that parsing in creole languages requires almost no
processing and allows fast computation in all modules of these grammars. I
leave it to proponents of creole simplicity and prototype to show that this is
indeed the case, but the task is certainly not a trivial one, as the following
paragraphs show.
The discussion here presents several cases of local complexity that
indicate that claims about creole simplicity and prototype do not typically
provide us with the full picture and therefore fail to answer a fundamental
question in linguistic study: How does structural complexity or simplicity
relate to language acquisition and linguistic variation? By way of
illustration, let us consider a commonly used metric of syntactic
complexity that basically counts the number of distinct syntactic rules in a
module of two languages. In this regard, a common point of comparison
between creoles and their Indo-European source languages (at the exclusion
of their substrate languages) is that of questions. For instance, English
displays yes–no and wh-phrase questions involving subject-auxiliary
inversion. Saramaccan and English-based creoles, in general, lack such a
rule. In comparing the English wh-question (1) to the Saramaccan example
in (2), one could conclude that Saramaccan, which combines properties of
English syntax and Gbe syntax, as shown in Chapter 6, appears to display a
simpler syntax.
(1) What did you cook?
(2) Andí wε i bói?
what FOC 2SG cook
However, as is often the case, such an uninformed comparison would miss
the fact that English resorts to auxiliary-subject inversion in addition to wh-
fronting, while Saramaccan involves a focus marker (absent in English) that
attaches to the fronted wh-phrase. Let us see how these two examples could be
represented structurally. Following the discussion in Chapter 6 (see also Rizzi
1997; Aboh 2004a), I’m assuming that fronted wh-phrases move to a focus
position cross-linguistically. Accordingly, both English and Saramaccan
display wh-movement to [spec FocP]. The arrows illustrate movement
operations in these two languages.
(3) English Saramaccan

FocP FocP
spec Foc′ spec Foc′

what
Foc Foc
did wε
you did cook what i boi
Counting these arrows, as one may, we realize that where English displays
two instances of movement (i.e., wh-movement and auxiliary movement to
Foc), Saramaccan exhibits one arrow only. Instead of raising a tense specifying
pleonastic element (i.e., do) to Foc, Saramaccan developed a focus marker (i.e.,
wε) that fills in this position. We could stop here and conclude that English
(with its two movement operations) has a more complex syntax than
Saramaccan which displays only one movement operation. Actually, most
exceptionalist views on creole languages do just that (e.g., McWhorter 1998,
2001, 2011).
However, for such a conclusion to be firmly established, we must also show
that auxiliary movement of do to Foc (an instance of internal merge in
minimalist syntax) is more complex than base generation of wε in Foc (a case
of external merge in Minimalism). No current theory of creole simplicity takes
this further step. Indeed, even though there are some ideas in the literature
suggesting that internal merge (i.e., movement) is costlier and therefore more
complex than external merge (i.e., base generation) (e.g., Chomsky 1995) no
experimental study has proven this to be generally correct in all the relevant
cases.
Even if we were to grant the view that the syntactic operation underlying
constituent or head movement is generally costlier than base-generation of a
lexical item, the fact that Saramaccan does display wh-movement in content
questions would lead us to a paradox that cannot be handled by theories of
creole simplicity grounded in imperfect second language acquisition. Indeed,
this language would appear to have developed a complex option for part of the
syntax of forming content questions, while the other part would involve a simpler
syntax. How come the creators of the creoles acquired wh-fronting rules even
though they failed to acquire subject–auxiliary inversion? For the picture to be
complete, one would also need to take into account the fact that English subject–
auxiliary inversion derives from the presence of auxiliaries in this language as
opposed to Saramaccan, where verbal conjugations and their related auxiliary
paradigm were replaced by an intricate system of TMA markers. Likewise, the
complementizer system responsible for subject–auxiliary inversion in English
displays a different morphosyntax in Saramaccan, where it involves a series of
discourse markers with their own syntax and semantics. Recall, for instance, from
the discussion in Chapter 6 that one property of such discourse markers is that, in
addition to attracting focused or topicalized constituents, they can also take scope
over the proposition which is therefore fronted. Consequently, some discourse
markers that realize the C-domain in Saramaccan can also occur in sentence-final
position (as shown in Chapter 6).
It is not clear to me how these facts would be accommodated in a holistic
theory of creole simplicity that is only concerned with counting morpheme
types and shapes or movement types without ever considering the relations
between these forms or operations and their semantics. Indeed, all English-
based creoles that I know of developed wh-fronting rules in content questions
even though such rules are not a priori required by human cognition and cannot
be assumed to be default options. Many languages of the world (e.g., Mandarin
Chinese, Japanese, Hindi, and Iraqi Arabic, to name only a few) display wh-in-
situ content questions, where the wh-phrase remains in its base position.
Likewise, the Romance and Germanic languages also allow some content
questions in which it is not immediately clear that the construction involves
movement. A case in point is subject question involving lexical verbs (e.g.,
Who came to the party? Who did what? Qui est venu à la fête ? Qui a fait
quoi ?). Clearly, the input that the creators of creole languages were exposed to
could also have favored the choice of in-situ wh-questions, but this did not
happen. If indeed, movement is costly and complex (and presumably difficult
to acquire), why are English-based creoles not similar to Mandarin Chinese?
What drives the kind of local complexity just described?
One can raise similar questions with regard to relative clauses in creoles.
Most Atlantic creoles involve relative clauses similar to those of their source
languages (Romance, Germanic, and Niger-Congo). In addition, Saramaccan
displays number agreement in relative clauses: the Saramaccan relative marker
agrees in number with the head noun, as discussed in Chapters 3 and 5 (see also
Huttar, Aboh, and Ameka 2013). Given the combination of wh-movement and
number agreement in Saramaccan relative clauses, this language appears to be
more complex in this domain than English is. One is therefore tempted to ask
why Saramaccan and more generally other Atlantic creoles did not resort to an
apparently much simpler or less complex option such as coordination. What
aspects of the input that the slaves were exposed to led them to develop a
structure like example (37b) in Chapter 3 (viz., dee fisi dee mi tata kisi bigi ‘the
fish that my father caught are big’), instead of relying on coordination or
juxtaposition (e.g., ‘my father caught fish, the fish are big’)?
Note that, according to the pidgin-to-creole cycle, one expects creoles to
exhibit this sort of paratactic structures rather than embedding constructions
such as relative clauses. It is not clear to me how these theories would explain
the link between the pidgin phase and the creole phase and how that relates to
the emergence of embedding structures including relative clauses involving

movement operations similarly to wh-questions. Given these facts, one would
expect theories of creole simplicity to explain what properties of the creole
setting led the creators of creole languages to choose simple options (e.g., the
presumed simple morphology) in some cases, while opting for more complex
structures in others? There does not seem to be any principled answer to this
question under creole simplicity.
Likewise, Caribbean creoles typically involve nominal markers of
definiteness and specificity that look on the surface like their Romance and
Germanic equivalents. Here too, we may immediately ask why the creoles
develop such markers even though these do not seem to be necessary for the
expression of specificity and definiteness. For instance, some Sinitic languages
(e.g., Mandarin Chinese) do not involve articles of the Romance and Germanic
type though these languages encode definiteness and specificity by other
grammatical means (e.g., the position of the noun phrase in the clause). The
situation in the creoles is even more puzzling since the morphosyntax and the
distribution of these nominal markers is less predictable than their equivalents
appear to be in the Germanic or Romance source languages. For instance, the
following examples indicate that the English article ‘the’ is necessary for the
expression of definiteness and this article does not inflect for number:
(4) a. the man
b. the three men
In Saramaccan, however, the English example (4a) translates into (5a) and
example (4b) into (5b). In this latter example, the determiner dee ‘the.pl’
agrees in number with the numeral ‘three’ (dii), as explained in Chapter 3:
(5) a. di womi [Rountree and Glock 1992]
DET.SG man
‘the man’
b. dee dii womi
DET.PL three men
‘the three men’
Example (6) also shows that Saramaccan, like most Caribbean creoles,
allows bare nouns in contexts where English disallows them (Aboh and
DeGraff 2014).
(6) a. A ta fan kuma womi [Rountree and Glock 1992]
3SG PROG talk like man
‘He talks like a man.’
b. Kato na Womi
Kato NEG man
‘Kato is not a man.’
8.3 Hybridism drives opacity (and complexity) 311
Compared to English, the Saramaccan noun phrase seems to involve two

complex options: (i) it exhibits number agreement, and (ii) the distribution of
the determiner di/dee is less predictable than that of English determiners (e.g.,
the/this/that). Unlike English, the Saramaccan determiner is not always
required for the expression of definiteness and the grammar does not require
argument noun phrases to be systematically marked by a determiner.
Consequently, the Saramaccan learner must find out the rules underlying the
contexts in which a determiner is obligatory and the contexts where it is not.
This conclusion obviously relates to issues of learnability and how an irregular
system such as that of Saramaccan may affect processing the relation between
form and meaning.
In these concluding remarks, I chose to discuss some of the issues revolving
around the distribution of noun phrases in questions, relative clauses, or simple
clauses in order to show how they bear on questions of simplicity and
complexity that are left unanswered in theories of creole simplicity. Similar
issues can actually be raised regarding other aspects of creole grammars which
I will not discuss here. The point I would like to draw the reader’s attention to is
that the examples just discussed indicate that once one moves beyond the
canonical descriptions that have been ascribed to creole grammars (the
supposedly poor inflectional morphology, fixed word order, reduced
phonemic inventory) and which are recycled continuously in simplicity
theories, one easily comes across constructions which are inconsistent with
these theories and which force us to think harder and propose explanations that
go beyond creole languages.
8.3 Hybridism drives opacity (and complexity)

Indeed, the idea of creole simplicity does not strike me as weak only when one
considers creole empirical data but also when one puts creole languages in the
larger context of new languages, including sign languages that emerged
recently. It is common practice in linguistic theory to classify creole
languages and sign languages such as Nicaraguan and Al-Sayyid Bedouin
Sign Languages as new languages. Since a major argument of creole
prototype is to argue that these languages are young and therefore exhibit a
transparent morphological type (i.e., a one-to-one mapping between form and
meaning), one would expect both groups of languages to share similar
structural properties and score similarly on the various metrics of simplicity
proposed in the literature. This is however not the case.
While creole languages are claimed by some to display almost no inflectional
morphology, most sign languages have been shown to involve a rich verb
agreement system, classifier constructions, and verbal aspects. If indeed, sign
languages and creole languages belong to the class of young languages and if
morphological complexity is associated with how old a language is, as often

claimed, the question arises of why the morphological systems of creoles and
sign languages differ so radically. Proponents of creole simplicity do not
seem particularly interested in this issue even though specialists of sign
languages noticed this paradox and try to explain it. For instance, Aronoff,
Meir, and Sandler (2005) argue that the observed difference between the
morphological type in young creole languages and that in sign languages is
modality-specific: sign languages develop certain complex morphological
properties due to their visual-spatial modality. That is, sign languages adapt
intrinsic properties of the human language capacity to the visual-spatial
mode. An example of such an adaptation is that sign languages involve
both sequential morphological structures (similarly to spoken languages)
and simultaneous morphological structures (typical of sign languages).
While sequential morphological structure is assumed to be intrinsic to the
human language capacity, simultaneous morphological structure is taken to
be specific to the visual-spatial mode of sign languages.
Note, however, that tense, mood, and aspect marking tonemes (e.g., in
various Kwa languages) can be regarded as the equivalent of simultaneous
inflectional morphology in spoken languages. Accordingly, the assumed
spoken versus signed dichotomy might not be as neat as suggested by
Aronoff, Meir, and Sandler’s discussion. Indeed, a point that Aronoff, Meir,
and Sandler (2005) did not take into consideration in their discussion of creole
languages is the role of the substrate languages and how in a contact setting,
competition and selection shape the developing language. Once we take this
perspective, the kind of adaptation claimed by these authors to account for
differences between creoles and sign languages conjures up the variation
illustrated in this book regarding how creoles have syntactic and semantic
features selected from their source languages. I have shown that these
recombinations not only reflect the heterogeneity of the inputs that the
speakers were exposed to but often produced new forms that are not attested in
the context of the source languages. This leads me to conclude that complexity
(however defined) is not to be thought of as a direct consequence of how old a
language is but rather as a process that is contingent on variation in the inputs and
how this variation is reflected in the learning hypotheses entertained by learners
in a situation of competition and selection. The more varied the inputs, the more
likely learners are to produce new, sometimes opaque recombinations that feed
syntactic and semantic complexity. Accordingly, linguistic properties (e.g.,
morphological form, opaque syntactic and semantic rules) that some authors
treat as simple arbitrary accretion have a more dynamic source.
Given this view, what Aronoff, Meir, and Sandler (2005) found to be
modality-specific simply results from linguistic hybridism itself determined
by the ecology in which sign languages evolve. Because linguistic hybridism is
8.4 Beyond creoles 313
universal, we expect to find similar evolution or adaptation in communities

where spoken and signed languages are used interchangeably. A case in point is
code-blending in bimodal bilinguals (i.e., hearing children who use both a
signed and a spoken language natively). For instance, Donati and Branchini
(2013) show that these speakers can produce simultaneous bimodal utterances
where sign language and spoken language fuse to produce a coherent linguistic
expression. Such blending represents new linguistic forms that could not have
been created independently in the spoken or signed modality. These forms
constitute new local complexities of their own since both the speaker and the
hearer must deploy cognitive capacities to process how forms relate to meaning
in two different modalities and how the combinations thereof can be
interpreted.
The discussion in this book therefore shows that regardless of how old they
are, languages in general develop opaque systems which seem to arise from the
combination of linguistic elements derived from the often-competing inputs to
which learners are exposed. While some combinations are selected and passed
onto successive generations, others are eliminated from the pool. Accordingly,
creoles, or contact languages in general, are not exceptional. While studying
these languages may help us understand the dynamics underlying the
development of grammar as a hybrid system, we should not conclude from
this that these languages display a unique linguistic structure that sets them
aside from all other human languages. Instead, what creole languages show us
is that natural languages normally involve hybrid systems, though such
hybridism may vary across modules and across languages, as a consequence
of the differing inputs to which the learner is exposed.
8.4 Beyond creoles

Contrary to the tradition in linguistics that has singled out creoles as the archetype
of language creation out of contact, I have argued in this book that each instance
of acquisition involves language contact of some sort, viz., contact of different
idiolects which sometimes also involves different sociolects, dialects, or even
languages (Chapter 4; see also Roeper 1999; Mufwene 2001, 2005). According
to this position, there is no qualitative difference between a child learning her
language in a multilingual environment (e.g., communities where most
inhabitants of Africa or Asia are raised) and a child raised in a monolingual
environment (as in some Western countries). In both situations, children learn to
master multiple linguistic subsystems that are in contact and may ‘cross-breed’
to produce new variants, which may subsequently serve as inputs for new
learners. This view is construed within the understanding that language
learning is always imperfect: The learners’ motivation is thus not to replicate
the target language faithfully but to develop learning hypotheses that bring them
close enough to the target to guarantee successful communication and

membership in the community. Accordingly, learners do not derive identical
grammars from the pool. This view explains why languages constantly change:
language change is a perpetual phenomenon contingent on learning.
Thus we must assume that change occurs at two levels: (i) the individual
level, and (ii) the population level. Change at the individual level is contingent
on acquisition: each learner develops a grammar that is close enough to the
target to allow communication. In synchrony, communities manage this type of
variation intrinsic to acquisition by developing conventions and norms that
speakers try to converge to. Change at the population or community level,
however, is what diachronic studies are concerned with. It occurs when a
significant number of speakers converge toward a new grammar that
eventually spreads through the whole population (and may become the norm
for subsequent learners). Both levels of analysis (i.e., synchronic and
diachronic) indicate that language change is a perpetual phenomenon related
to learning and population factors: change is always there though it may remain
dormant.
Given these observations, it must now be obvious to the reader why we need
to understand the genesis of change (at the individual level) in order to
understand language acquisition and how the latter relates to change at the
population level. Our hypothesis that lexical items involve three components,
viz., phonology, morphosyntax, and semantics, which can vary independently
of each other suggests that in acquiring every single lexical item, learners can
produce eight possible competing variants. Many of such variants are close
enough to the target not to be immediately noticed at the community level. This
is, for instance, the case of most combinations that involve subtle reanalyses
targeting syntax or semantics, as illustrated in Chapters 4–7. Yet, the fact that
learners typically develop a grammar that converges toward communal norms
indicates that learners must have a way of eliminating unsuccessful or
improbable combinations. A question that we must therefore ask ourselves is
the following: What are the underlying principles of learning that allow
learners to filter out unsuccessful combinations?
We can imagine various solutions in this regard. One factor that certainly
plays an important role is that of the inputs which, to some extent, define the
range of variation that the learner is exposed to. For instance, a learner exposed
to Dutch sociolects or dialects has access to a feature pool with a narrower
range of variation than a speaker exposed to Dutch and Ijo (an Ijoid language
spoken in Nigeria) or to idiolects of Dutch influenced by Ijo, as were, for
instance, the creators of Berbice Dutch in the colonial Caribbean.
Another factor that constrains variation that needs to be investigated in
collaboration with neuroscientists and psycholinguists is that of processing,
already alluded to in previous paragraphs. Let us suppose that learners can
8.4 Beyond creoles 315
evaluate competing alternatives in terms of their processing cost. Suppose further

that combinations with low processing costs are favored and thus retained while
those with excessive processing costs are disfavored and may eventually be
eliminated from the feature pool or be restricted to special contexts of use.
Because such combinations may be disfavored systematically by learners over
generations, they can represent what could be labeled as unlearnable, or just
costly, language outputs, that is, outputs which even though possible in principle,
cannot be learned efficiently because of their processing loads. As I already
suggested in previous paragraphs, a conclusion that can be drawn from this
discussion is that we may better understand linguistic complexity by
investigating the relation between processing of meaning and structure, rather
than by focusing on the size and shape of morphemes. Yet another fact that this
view may help understand is that of typological variation and the distribution
frequency of some patterns, e.g., SVO/SOV vs. OSV/OVS. Indeed, we may find
out that the frequency of certain cross-linguistic patterns correlates with their
processing properties. While informing us on human cognition, such findings
may in turn shed light on issues of competition and selection as to why certain
patterns are systematically selected against others and how selection is affected
by issues of processing or interface properties as discussed in previous chapters.
The notion of competition and selection as argued by Mufwene (2001,
2005a, 2005b, 2008) and in this study suggests that languages can be
analogized to biological species. This view obviously requires close
collaboration with evolutionary biologists and cognitive scientists. Two
points call for further investigation.
The first has to do with feature competition. Recall from the discussion in
previous chapters that lexical items embed bundles of features as part of three
components: phonology, morphosyntax, and semantics. Accordingly, the notion
of competition and selection must be understood in terms of competition and
selection of features that are accessible to the learner. The question then arises
whether there is a universal feature hierarchy according to which some features
or combinations thereof must be preferred to other alternatives and whether that
hierarchy plays a role in constraining learnability and linguistic variation. This
question, I believe must be addressed from a multi-disciplinary perspective
where linguists, cognitive scientists, and computational scientists collaborate to
identify intrinsic aspects of lexical and grammatical items and how these
recombine across populations speaking different languages and/or facing
contacts of different languages.
The second point relates to patterns that appear constant across languages.
Under learnability as emerging from competition and selection, successive
replication of similar combinations across human populations may eventually
leave an imprint on the human brain. This in turn will predispose learners to
systematically select certain combinations over others. As a consequence
recombination of linguistic features might not be as free as it looks on the

surface. An empirical domain that needs to be investigated in this respect is that
of the expression of Tense, Mood, and Aspect which has been shown on the
basis of typological data by functionalists and generativists to systematically
follow a fixed scope hierarchy as discussed in Chapter 7. Such a rigid scope
hierarchy immediately raises the question of the limits on variation: What
aspects of learnability constrain variation, thus yielding apparent cross-
linguistic universals? And how can we model learnability as the driving force
behind change and evolution?
Answering these questions requires an interdisciplinary approach to issues
of language acquisition and change in order to understand what combinations
are likely to be stable across populations and may be transmitted to new
generations. This leads us to the more general question of what learners
inherit from previous generations versus what they create as learners, and
how new norms or conventions emerge accordingly. I hope to return to some
of these questions in future work.
References
Abney, Paul S. 1987. The English noun phrase in its sentential aspect. PhD dissertation,
MIT, Cambridge, MA.
Aboh, Enoch O. 2001. Clitiques phonétiques ou clitiques tout court. In Claude Muller,
Paulo de Carvalho, Laurence Labrune, Frédéric Lambert, and Katja Ploog (eds.),
Clitiques et cliticisation. Paris: Honoré Champion, 461–481.
2002. La morphosyntaxe de la péripherie gauche nominale. Recherches linguistiques
de Vincennes 31: 9–26.
2003a. Les constructions à objet préposé et les series verbales dans les langues kwa.
In Patrick Sauzet and Anne Zribi-Hertz (eds.), Typologie des langues d’Afrique
et universaux de la grammaire. Vol. II: ‘Benue-Kwa, Soninke, Wolof’. Paris:
L’Harmattan, 15–40.
2003b. Focus constructions across Kwa. In Cege Githiora, Heather Littlefield, and
Victor Manfredi (eds.), Kinyira njira! Step firmly on the pathway! Trenton, NJ:
Africa World Press, 7–22.
2004a. The morphosyntax of complement-head sequences: Clause structure and word
order patterns in Kwa. New York: Oxford University Press.
2004b. Left or right? A view from the Kwa peripheral positions. In David Adger,
Cécile De Cat, and George Tsoulas (eds.), Peripheries: Syntactic edges and their
effects. Dordrecht: Kluwer Academic Publishers, 165–189.
2004c. Topic and focus within D. Linguistics in The Netherlands 21: 1–12.
2005a. Pattern and feature competition: Toward a syntactic account to contact-
induced transfer. Conference on Creole language structure between substrates
and superstrates. MPI Leipzig (Germany) June 3–5, 2005.
2005b. The category P: The Kwa paradox. Linguistic Analysis 32: 615–646.
2005c. Object shift, verb movement and verb reduplication. In Guglielmo Cinque and
Richard Kayne (eds.), The Oxford handbook of comparative syntax. New York:
Oxford University Press, 138–177.
2005d. Deriving relative and factive constructions in Kwa. In Laura Brugè,
Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano (eds.),
Contributions to the thirtieth incontro di grammatica generativa. Venice:
Cafoscarina, 265–285.
2006a. The role of the syntax-semantics interface in language transfer. In
Claire Lefebvre, Lydia White, and Christine Jourdan (eds.), L2 acquisition and
creole genesis: Dialogues. Amsterdam: Benjamins, 221–252.
317
318 References
2006b. Complementation in Saramaccan and Gungbe: The case of C-type modal

particles. Natural Language and Linguistic Theory 24: 1–55.
2006c. When verbal predicates go fronting. ZAS Papers in Linguistics 46: 21–48.
2007a. La genèse de la périphérie gauche du saramaka: un cas d’influence du
substrat? In Karl Gadelii and Anne Zribi-Hertz (eds.), Grammaires créoles et
grammaire comparative. Saint-Denis: Presses Universitaires de Paris, 73–97.
2007b. A “mini” relative clause analysis for reduplicated attributive adjectives.
Linguistics in The Netherlands 24: 1–13.
2009a. Competition and selection: That’s all. In Enoch O. Aboh and Norval Smith
(eds.), Complex processes in new languages. Amsterdam: Benjamins, 317–344.
2009b. Clause structure and verb series. Linguistic Inquiry 40: 1–33.
2009c. Delete: A phase-level property. Theoretical Linguistics 35: 229–237.
2010a. The morphosyntax of the noun phrase. In Enoch O. Aboh and James Essegbey
(eds.), Topics in Kwa syntax. Dordrecht: Springer, 11–37.
2010b. The P-route. In Guglielmo Cinque and Luigi Rizzi (eds.), Mapping spatial
PPs: The cartography of syntactic structures. New York: Oxford University Press,
225–260.
2010c. C-type negation markers on the right edge. In Enoch O. Aboh and
James Essegbey (eds.), Topics in Kwa syntax. Dordrecht: Springer, 109–139.
2010d. The syntax of inherent complement verbs in Gungbe. Unpublished
manuscript, University of Amsterdam.
2010e. Information Structuring begins with the numeration. IBERIA 2, 1: 12–42.
www.siff.us.es/iberia/index.php/ij/index
2010f. Event operator movement in factives: Some facts from Gungbe. Theoretical
Linguistics 36: 153–162.
The syntax of edges. Unpublished manuscript, University of Amsterdam.
Aboh, Enoch O. and Umberto Ansaldo. 2007. The role of typology in language creation:
A descriptive take. In Umberto Ansaldo, Steve M. Mathew, and Lisa Lim (eds.),
Deconstructing Creole. Amsterdam: Benjamins, 39–66.
Aboh, Enoch O. and Michel DeGraff. 2014. Some notes on nominal phrases in Haitian
Creole and in Gungbe: A transatlantic Sprachbund perspective. In Tor A. Afarli
and Brit Maehlum (eds.), The sociolinguistics of grammar. Amsterdam:
Benjamins, 203–236).
Forthcoming. A null theory of creolization based on universal grammar. To appear in
Ian Roberts (ed.), Oxford Handbook of Universal Grammar. Oxford University Press.
Aboh, Enoch O. and Marina Dyakonova. 2009. Predicate doubling and parallel chains.
Lingua 119: 1035–1065.
Aboh, Enoch O. and James Essegbey (eds.) 2010. Topics in Kwa syntax. Dordrecht:
Springer.
Aboh, Enoch O., Katharina Hartmann, and Malte Zimmermann (eds.) 2007. Focus
strategies in Niger Congo and Afro-Asiatic: On the interaction between focus and
grammar in some African languages. Berlin: Mouton.
Aboh, Enoch O. and Fabrice Nauze. 2008. Tense, Mood, and Aspect in Gungbe (Kwa).
In Werner Abraham and Elisabeth Leiss (eds.), Modality-aspect interfaces:
implications and typological solutions. Amsterdam: Benjamins, 215–239.
References 319
Aboh Enoch O and Roland Pfau. 2011. What’s a wh-word got to do with it? In
Paula Benincà and Nicola Munaro (eds.), Mapping the left periphery: The
cartography of syntactic structures. New York: Oxford University Press, 91–124.
Aboh, Enoch O. and Norval Smith (eds.) 2009. Complex processes in new languages.
Amsterdam: Benjamins.
2009. Simplicity, simplification, complexity and complexification: Where have the
interfaces gone? In Enoch O. Aboh and Norval Smith (eds.), Complex processes in
new languages. Amsterdam: Benjamins, 1–25.
2012. The morphosyntax of non-iconic reduplications: A case study in Eastern Gbe
and the Surinam creoles. In Enoch O. Aboh, Norval Smith, and Anne Zribi-Hertz
(eds.), The grammar of reiteration in creole and non-creole languages.
Amsterdam: Benjamins, 27–75.
2015a. Migrations, ethno-dynamics and geolinguistics in the Eastern Aja-Tado
cultural area. In Pieter Muysken and Norval Smith (eds.), Surviving the Middle
Passage: The West Africa-Surinam Sprachbund. Berlin: Mouton, 43–66.
2015b. Non-iconic reduplications in Eastern Gbe and Surinam. In Pieter Muysken
and Norval Smith (eds.), Surviving the Middle Passage: The West Africa-Surinam
Sprachbund. Berlin: Mouton, 241–260.
Aboh, Enoch O., Norval Smith, and Tonjes Veenstra. 2013. Saramaccan. In Susanne
Maria Michaelis, Philippe Maurer, Martin Haspelmath, and Magnus Huber (eds.),
The survey of pidgin and creole languages. Vol. I: English-based and Dutch-based
languages. Oxford University Press, 27–38.
Aboh, Enoch O., Roland Pfau, and Ulrike Zeshan. 2005. When a Wh-word is not a Wh-
word: The case of Indian Sign Language. In Tanmoy Bhattacharya (ed.), The
yearbook of South Asian languages and linguistics. Berlin: Mouton de Gruyter,
11–43.
Adande, Alexis B. A. 1984. Tɔgudo-Awute, capitale de l’ancien royaume d’Allada.
Unpublished PhD dissertation, Université de Paris I.
Aikhenvald, Alexandra, and R. M. W. Dixon. (eds.) 2006. Serial verb constructions:
A Cross-linguistic typology. New York: Oxford University Press.
Alleyne, Mervyn C. 1971. Acculturation and the cultural matrix of creolization.
In Dell Hymes (ed.), Pidginization and creolization of languages. Cambridge
University Press, 169–186.
1979. On the genesis of languages. In Kenneth C. Hill (ed.) The genesis of language.
Ann Arbor, MI: Karoma, 89–107.
1980. Comparative Afro-American: An historical-comparative study of English-
Based Afro-American dialects. Ann Arbor, MI: Karoma.
Ameka, Felix. 1992. Focus constructions in Ewe and Akan: A comparative perspective.
MIT Working Papers in Linguistics 17: 1–25.
2003. Prepositions and postpositions in Ewe (Gbe): Empirical and theoretical
considerations. In Parick Sauzet and Anne Zribi-Hertz (eds.), Typologie des
langues d’Afrique et universaux de la grammaire. Paris: L’Harmattan, 41–67.
2010. Information packaging constructions in Kwa: Micro-variation and typology. In
Enoch O. Aboh and James Essegbey (eds.), Topics in Kwa syntax. Studies in
Natural Language and Linguistic Theory 78. Dordrecht: Springer, 141–176.
320 References
Anglade, Pierre. 1998. Inventaire étymologique des termes creoles des caraïbes
d’origine africaine. Paris: L’Harmattan.
Anonymous. 1983. Éléments de recherche sur la langue fon. Cotonou: République du
Bénin.
Ansaldo, Umberto. 2005. Typological convergence and admixture in Sri Lanka
Malay: The case of Kirinda Java. Unpublished manuscript, University of
Amsterdam.
Ansaldo, Umberto. 2009. Contact languages: Ecology and evolution in Asia.
Cambridge University Press.
Arends, Jacques. 1989. Syntactic developments in Sranan. Unpublished PhD dissertation,
University of Nijmegen.
1995a. The socio-historical background of creoles. In Jacques Arends, Pieter Muysken,
and Norval Smith (eds.), Pidgins and creoles: An introduction. Amsterdam:
Benjamins, 15–24.
1995b. Demographic factors in the formation of Sranan. In Jacques Arends (ed.), The
early stages of creolization. Amsterdam: Benjamins, 233–285.
1998. The development of complementation in Saramaccan. In Bernard Caron (ed.),
Proceedings of the eleventh International Congress of Linguistics. Oxford:
Pergamon (CD-ROM).
2009. Syntactic developments in Sranan: Creolization as a gradual process.
www.dbnl.org/tekst/aren012synt01_01/colofon.htm
Aronoff Mark, Irit Meier, and Wendy Sandler. 2005. The paradox of sign language
morphology. Language 81: 301–344.
Avolonto, Aimé. 1995. Pour une approche minimaliste des verbes à objets inhérents en
fongbe. PhD Dissertation, University of Quebec at Montreal.
Awóyalé, Yiwọla. 1988. Complex predicates and verb serialization. MIT Lexicon
Working Papers 28. MIT Center for Cognitive Science. MI, 1–35.
Bailey, Beryl L. 1966. Jamaican Creole syntax: A transformational approach.
Baker, Mark C. 1985. The mirror principle and morphosyntactic explanation. Linguistic
Inquiry 16: 373–415.
1988. Incorporation: A theory of grammatical function changing. The University of
Chicago Press.
1989. Object sharing and projection in serial verb constructions. Linguistic Inquiry
20: 513–553.
1991. On the relation of serialization to verb extensions. In Claire Lefebvre (ed.),
Serial verbs: Grammatical, comparative and cognitive approaches. Amsterdam:
Benjamins, 79–102.
2008. The macroparameter in a microparametric world. In Theresa Biberauer (ed.),
The limits of variation. Amsterdam: Benjamins, 351–374.
Bakker, Peter. 2014. Creolistics: Back to square one? Journal of Pidgin and Creole
Languages 29: 177–194.
Bakker, Peter, Daval-Markussena Aymeric, Parkvall Mikael, and Plag Ingo. 2011.
Creoles are typologically distinct from non-creoles. Journal of Pidgin and
Creole Languages 26: 5–42.
References 321
Bakker, Peter, Marike Post, and Hein van der Voort. 1994. TMA particles and
auxiliaries. In Jacquues Arends, Pieter Muysken, and Norval Smith (eds.),
Pidgins and Creoles: An introduction. Amsterdam: Benjamins, 247–258.
Baptista, Marlyse. 2009. Economy, innovation and degrees of complexity in creole
formation. In Enoch O Aboh and Norval Smith (eds.), Complex processes in new
languages. Amsterdam: Benjamins, 293–315.
Baptista, Marlyse and Jacqueline Guéron (eds.) 2007. Noun phrases in creole
Languages: A multi-faceted approach. Amsterdam: Benjamins.
Bernstein, Judy. 1997. Demonstratives and reinforcers in Romance and Germanic
languages. Lingua 102: 87–113.
2001a. Focusing the “right” way in Romance determiner phrases. Probus 13: 1–29.
2001b. The DP hypothesis: Identifying clausal properties in the nominal domain. In
Mark Baltin and Chris Collins (eds.), The handbook of contemporary syntactic
theory. Oxford: Blackwell, 536–561.
Bickerton, Derek. 1981. Roots of language. Ann Arbor, MI: Karoma.
1984. The language bioprogram hypothesis. Behavioral and Brain Sciences 7: 173–221.
1988a. Creole languages and the bioprogram. In Frederic Newmeyer (ed.),
Linguistics: The Cambridge survey. Vol. II: Linguistics theory: Extensions and
implications. Cambridge University Press, 268–284.
1988b. A dialogue concerning the linguistic status of creole languages. In
Newmeyer, Frederic (ed.), Linguistics: The Cambridge survey. Vol. II: Linguistics
theory: Extensions and implications. Cambridge University Press, 302–306.
1989. Seselwa serialization and its significance. Journal of Pidgin and Creole
1990a. Language and Species. The University of Chicago Press.
1990b. If it quacks like a duck . . . A reply to Seuren. Journal of Piding and Creole
1999. How to acquire language without positive evidence: What acquisitionists can
learn from creoles. In M. DeGraff (ed.), Language creation and language change,
Cambridge, MA: MIT Press, 49–74.
2008. Bastard tongues: A trail-blazing linguist finds clues to our common humanity in
the world’s lowliest languages. New York: Hill and Wang.
Bobyleva, Ekaterina. 2013. The development of the nominal domain in creole
languages: A comparative-typological approach. Utrecht: LOT.
Boeckx, Cedric. 2010. What principles and parameters got wrong. Unpublished
manuscript, Catalan Institute for Advanced Studies.
Booij, Geert. 2005. The grammar of words. New York: Oxford University Press.
Borer, Hagit. 1984. Parametric syntax: Case studies in Semitic and Romance
languages. Dordrecht: Foris.
2005. Structuring sense. Vol. I: In name only. New York: Oxford University Press.
Brousseau, Anne-Marie and John S. Lumsden. 1992. Nominal structure in Fongbe. The
Journal of West African Languages 22: 5–25.
Bruyn, Adrienne. 1995a. Grammaticalization in creoles: The development of determiners
and relative clauses in Sranan. Studies in Language and Language Use 21.
Amsterdam: IFOTT.
322 References
1995b. Noun phrases. In Jacques Arends, Pieter Muysken, and Norval Smith
(eds.), Pidgins and Creoles: An introduction. Amsterdam: Benjamins,
259–270.
Burzio, Luigi. 1986. Italian syntax: A government and binding approach. Dordrecht:
Reidel.
Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form.
Byrne, Francis. 1987. Grammatical relations in a radical creole. Amsterdam:
Benjamins.
Cable, Seth. 2004a. Predicate clefts and base-generation: Evidence from Yiddish and
Brazilian Portuguese. Unpublished manuscript, MIT, Cambridge, MA.
2004b. Restructuring in English. Unpublished manuscript, MIT, Cambridge, MA.
Campbell, Lyle. 1996. Typological and areal issues in reconstruction. In Jacek Fisiak
(ed.), Linguistic reconstruction. Berlin: Mouton, 49–72.
Campbell, Richard. 1989. The grammatical structure of verbal predicates. PhD
dissertation, University of California, Los Angeles.
Campbell, Richard. 1992. Serial verbs and unaccusativity. MIT Working Papers 17:
41–51.
1996a. Specificity operators in SpecDP. Studia Linguistica 2: 161–188.
1996b. Serial verbs and shared arguments. The Linguistic Review 13: 83–118.
Capo, Hounkpati B. C. 1991. A comparative phonology of Gbe. Berlin: Foris.
Cardinaletti, Anna and Giuliana Giusti. 2001. Semi-lexical motion verbs in Romance
and Germanic. In Norbert Corver and Henk van Riemsdijk (eds.), Semi-lexical
categories: On the function of content words and content of function words. Berlin:
Mouton, 371–414.
Cardinaletti, Anna and Ur Shlonsky. 2004. Clitic positions and restructuring in Italian.
Linguistic Inquiry 35: 519–557.
Cardinaletti, Anna and Michal Starke. 1999. The typology of structural deficiency:
A case study of the three classes of pronouns. In Henk Van Riemsdijk (ed.), Clitics
in the languages of Europe. Berlin: Mouton de Gruyter, 145–233.
Chaudenson, Robert. 1992. Des îles, des hommes, des langues: Éssai sur la créolisation
linguistique et culturelle. Paris: L’Harmattan.
2001. Creolization of language and culture. London: Routledge.
2003. La créolisation: Théorie, applications, implications. Paris: L’Harmattan.
Cheng, Lisa L. and Rint Sybesma. 1999. Bare and not-so-bare nouns and the structure of
NP. Linguistic Inquiry 30: 509–542.
Chomsky, Noam. 1977. On wh-movement. In Peter W. Culicover (ed.), Formal syntax.
New York/San Francisco: Academic Press, 71–132.
1986. Knowledge of language: Its nature, origin and use, Westport, CT: Praeger.
1995. The minimalist program. Cambridge, MA: MIT Press.
2001. Derivation by Phase. In Micheal Kenstowixcz (ed.), Ken Hale: A life in
language. Cambridge, MA: MIT Press, 1–52.
2008. On Phases. In Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta
(eds.), Foundational issues in linguistic theory: Essays in honor of Jean-Roger
Vergnaud. Cambridge, MA: MIT Press, 133–166.
References 323
Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance
DP. In Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and
Raffaella Zanuttini (eds.), Paths towards Universal Grammar. Washington, DC:
Georgetown University Press, 85–110.
1996. The “antisymmetric” program: Theoretical and typological implications.
Journal of Linguistics 32: 447–464.
1999. Adverbs and functional heads: A cross-linguistic perspective. New York:
Oxford University Press.
2004. Restructuring and functional heads: The cartography of syntactic structures.
New York: Oxford University Press.
2010. The syntax of adjectives: A comparative study. Cambridge, MA: MIT Press.
Clements, Clancy J. 1996. The genesis of a language: The formation and development of
Korlai Portuguese. Amsterdam: Benjamins.
Collins, Chris. 1994. The factive construction in Kwa. Unpublished manuscript, Cornell
University.
1997. Argument sharing in serial verb constructions. Linguistic Inquiry 28: 461–497.
2002. Multiple verb movement in Hoan. Linguistic Inquiry 33:1–29.
Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago
University Press.
Croft, William. 2000. Explaining language change: An evolutionary approach. London:
Longman.
Curtin, Philip D. 1969. The Atlantic slave trade: A census. Madison, WI: University of
Wisconsin Press.
Da Cruz, Maxime. 1992. ECP et l’alternance nu/na en fongbe. Unpublished manuscript,
University of Quebec at Montreal.
1993. Les constructions sérielles du fongbe: Approches sémantique et syntaxique.
PhD dissertation, University of Quebec at Montreal.
1995. Aspectual verbs fó, v ‘finish’ in Fongbe. The Linguistic Review 12: 361–380.
1997. Serial verb constructions and null arguments in Fon. In Rose-Marie Déchaine
and Victor Manfredi (eds.), Object positions in Benue-Kwa. The Hague: The
Holland Academic Graphics, 31–45.
Damonte, Federico. 2002. The complementizer layer in Saramaccan. In Manuel Leonetti,
Olga Fernàndez Soriano, and Victoria Escandell Vidal (eds.), Current issues in
generative grammar. Tenth Colloquium on Generative Grammar: Selected papers.
Madrid: Universidad Alcalà de Henares, Servicio de Publicaciones, 31–50.
De Castro, Yeda Pessoa. 2002. A língua Mina-Jeje no Brasil. Um falar Africano em
Ouro Preto do Século XVIII. Apoio cultural. Fundação de Amparo à Pesquisa do
Estado de Minas Gerais. Papemig.
De Cauna, Jacques (ed.) 2004. Toussaint Louverture et l’indépendance d’Haïti. Paris:
Karthala-SFHOM.
Déchaine, Rose-Marie and Victor Manfredi (eds.) 1997. Object positions in Benue-
Kwa. The Hague: Institute of Generative Linguistics/Holland Academic Graphics.
Debien, Gabriel. 1956. Études antillaises (XVIIIe siècle). Paris: Association Marc Bloch.
1974. Les esclaves aux Antilles françaises (XVIIe–XVIIIe siècles). Basse-Terre:
Société d’histoire de la Gouadeloupe.
324 References
DeGraff, Michel. 1992. Creole grammars and acquisition of syntax: The case of Haitian
Creole. PhD dissertation, University of Pennsylvania.
1995. On certain differences between Haitian and French predicative constructions.
In J. Amastae, G. Goodall, M. Montalbetti, and M. Phinney (eds.), Contemporary
research in romance linguistics. Amsterdam: Benjamins, 237–256.
1997. Verb syntax in, and beyond creolization. In Liliane Haegeman (ed.), The new
comparative syntax. London: Longman, 64–94.
(ed.) 1999. Language creation and language change. Cambridge, MA: MIT Press.
2001a. On the origin of creoles: A cartesian critique of neo-Darwinian linguistics.
Linguistic Typology 5: 213–310.
2001b. Morphology in creole genesis. In Michael Kenstowicz (ed.), Ken Hale: A life
in language. Cambridge MA: MIT Press, 53–121.
2002. Relexification: A reevaluation. Anthropological Linguistics 44: 321–414.
2003. Against creole exceptionalism. Language 79: 391–410.
2005. Morphology and word order in “creolization” and beyond. In
Guglielmo Cinque and Richard Kayne (eds.), The Oxford handbook of
comparative syntax. New York: Oxford University Press, 293–372.
2007. Kreyòl Ayisyen, or Haitian Creole. In John Holm and Peter Patrick (eds.),
Comparative creole syntax: Parallel outlines of 18 creole grammars. London:
Battlebridge Publications, 101–126.
2009. Language acquisition in creolization (and language change): Some cartesian-
uniformitarian boundary conditions. Language and Linguistic Compass 3/4: 888–971.
D’Elbée, François. 1671. Journal du voyage du Sieur Delbée. In Jean de Clodoré (ed.),
Relation de ce qui s’est passé dans les Isles et Terre-ferme de l’Amerique pendant
la derniere guerre avec l’Angleterre, et depuis en execution du Traitté de Breda ii.
Paris.
Den Besten, Hans. 1977. On the interaction of root transformations and lexical deletive
rules. In Studies in West Germanic syntax. PhD dissertation, University of Tilburg.
Den Dikken, Marcel. 1998. Predicate inversion in DP. In Artemis Alexiadou and
Chris Wilder (eds.), Possessors, predicates and movement in the Determiner
Phrase. Amsterdam: Benjamins, 177–214.
Denis, Serge. 1935. Notre créole. Orleans: G. Luzeray.
Déprez, Viviane. 2001. Determiner architecture and phrasal movement in French
lexifier creoles. In Josep Quer, Jan Schroten, Mauro Scorretti, Petra Sleeman,
and Else Verheugd (eds.), Romance languages and linguistic theory 2001.
2004. Morphological number, semantic number and bare nouns. Lingua 115: 857–883.
2007. Nominal constituents in French lexifier creoles: Probing the structuring role of
grammaticalization. Journal of Pidgin and Creole Languages 22: 263–308.
Déprez Viviane, Petra Sleeman, and Hakima Guella. 2011. Specificity effects in L2
determiner acquisition: UG or pragmatic egocentrism? In Mihaela Pirvulescu,
María Cristina Cuervo, Ana T. Pérez-Leroux, Jeffrey Steele, and Nelleke Strik
(eds.), Selected proceedings of the fourth conference on Generative Approaches to
Language Acquisition – North America (GALANA 2010). Somerville, MA:
Cascadilla Proceedings Project, 27–36.
References 325
Diouf, Sylviane A. 2007. Dreams of Africa in Alabama: The slave ship Clotilda and the
story of the last Africans brought to America. New York: Oxford University Press.
Donati Caterina and Chiara Branchini. 2013. Challenging linearization: Simultaneous
mixing in the production of bimodal bilinguals. In Theresa Biberauer and
Ian Roberts and (eds.), Challenges to linearization. Berlin: Mouton, 93–128.
Donicie, Anton C. R. and Jan Voorhoeve. 1963. De Saramakaanse Woordenschat.
Bureau voor Tallonderzoek in Suriname van de Universiteit van Amsterdam.
Ducœurjoly, S. J. 1803. Manuel des habitants de Saint-Domingue. Paris: Arthus
Bertrand.
Durrleman-Tame, Stephanie. 2008. The syntax of Jamaican Creole: A cartographic
perspective. Amsterdam: Benjamins.
Eltis, David. 2011. Africa, slavery, and the slave trade, mid-seventeenth to mid-
eighteenth centuries. In Nicholas Canny and Philip Morgan (eds.), The Oxford
handbook of the Atlantic world: 1450–1850. New York: Oxford University Press,
271–286.
Emmer, Peter C. 2005. Les Pays-Bas et la traite des noirs. Paris: Karthala.
Enç, Mürvet. 1991. The semantics of specificity. Linguistic Inquiry 22: 1–26.
Essegbey, James. 1999. Inherent complement verbs revisited: Towards an understanding
of argument structure in Ewe. Wageningen: MPI series, Ponsen & Looijen bv.
Fattier, Dominique. 1996. La passion de notre Seigneur selon St. Jean en langage nègre:
Une scripta créole? Études créoles 19: 9–30.
Foley, William A. and Robert D. Van Valin.1984. Functional syntax and universal
grammar. Cambridge University Press.
Fouchard, Jean. 1972. Les marrons de la liberté. Paris: Éditions de l’École.
Gibson, Kean. 1986. The ordering of auxiliary notions in Guyanese Creole. Language
62: 571–586.
Giusti, Giuliana. 1996. Is there a FocusP and a TopicP in the noun phrase structure?
University of Venice Working Papers in Linguistics 6: 105–128.
1997. The categorial status of determiners. In Liliane Haegeman (ed.), The new
comparative syntax. London: Longman Linguistics Library, 95–123.
Glaude, Herby. 2012. Aspects de la syntaxe du l’haïtien. Paris: Anibwe.
Glaude, Herby and Anne Zribi-Hertz. 2012. Verb focus in Haitian: From lexical
reiteration to predicate cleft. In Enoch O. Aboh, Norval Smith, and Anne Zribi-
Hertz (eds.), The morphosyntax of reiteration in creole and non-creole languages.
Glock, Naomi. 1986. The use of reported speech in Saramaccan. In George Huttar and
Kenneth Gregerson (eds.), Pragmatics in non-Western perspective. Publications in
Linguistics. Dallas: Summer Institute of Linguistics, 35–61.
Goad, Heather and Lydia White. 2004. Ultimate attainment of L2 inflection: Effects of
L1 prosodic structure. In Susan Foster-Cohen, Michael Sharwood Smith,
Antonella Sorace, and Mits Ota (eds.), EUROSLA yearbook, Vol.IV. Amsterdam:
Goad, Heather, Lydia White, and Jefferey Steele. 2003. Missing inflection in L2
acquisition: Defective syntax or L1-constrained prosodic representations?
Canadian Journal of Linguistics 48: 243–263.
326 References
Granfeldt, Jonas. 2000. The acquisition of the Determiner Phrase in bilingual and
second language French. Bilingualism: Language and Cognition 3: 263–280.
Guillemin, Diana. 2011. The syntax and semantics of a determiner system: A case study
of Mauritian Creole. Amsterdam: Benjamins.
Haddican, Bill. 2001. Gengbe typing morphemes and determiners. Unpublished
manuscript, New York University.
Haeberli, Eric. 1999. Features, categories and the syntax of A-positions: Synchronic and
diachronic variation in the Germanic languages. PhD dissertation, University of
Geneva.
Haegeman, Liliane. 1995. The syntax of negation. Cambridge University Press.
Haegeman, Liliane and Barbara Ürögdi. 2010a. Referential CPs and DPs: An operator
movement account. Theoretical Linguistics 36: 111–152.
2010b. Operator movement, referentiality and intervention. Theoretical Linguistics
36: 233–246.
Hagemeijer, Tjerk. 2001. Underspecificatoin in serial verb constructions. In
Norbert Corver and Henk van Riemsdijk (eds.), Semi-lexical categories: On the
function of content words and content of function words. Berlin: Mouton, 414–451.
Hale, Ken and Samuel J. Keyser. 1993. On argument structure and the lexical expression
of syntactic relations. In Ken Hale and Samuel. J. Keyser (eds.), The view from
building 20. Cambridge, MA: MIT Press, 53–109.
Hall, Robert A. 1953. Haitian Creole: Grammar, texts, vocabulary. Volume 43.
Philadelphia: American Folklore Society.
1962. The life cycle of pidgin languages. Lingua 11: 151–156.
Harbour, Daniel. 2008. Klivaj predika, or predicate clefts in Haitian. Lingua 118:
853–871
Hartmann, Katharina and Malte Zimmermann. 2007. Exhaustivity marking in Hausa: A
reanalysis of the particle nee/cee. In Enoch O. Aboh, Katharina Hartmann, and
Malte Zimmermann (eds.), Focus strategies in Niger Congo and Afro-Asiatic: On
the interaction between focus and grammar in some African languages. Berlin:
Mouton, 249–272.
Hawkins, J. A. 2004. Efficiency and complexity in grammar. New York: Oxford
University Press.
Haznedar, Belma and Bonnie D. Schwartz. 1997. Are there optional infinitives in child
L2 acquisition? In Elizabeth Hughes, Mary Hughes, and Annabel Greenhill (eds.),
Proceedings of the 21st annual Boston University conference on language
development. Somerville, MA: Cascadilla Press, 257–268.
Heim, Irene. 1991. Articles and definiteness. Published in German as “Artikel und
definitheit,” in Armin von Stechow and Dieter Wunderlich (eds.), Semantics: An
International handbook of contemporary research. Berlin: Mouton, 487–535.
Hengeveld, Kees. 1989. Layers and operators in Functional Grammar. Journal of
Linguistics 25: 127–157.
Henri, Fabiola. 2012. Attenuative verbal reduplication in Mauritian: A morpho-
semantic approach. In Enoch O. Aboh, Norval Smith, and Anne Zribi-Hertz
(eds.), The morphosyntax of retiteration in creole and non-creole languages.
References 327
Holm, John. 1988. Pidgins and creoles. Vol. I: Theory and structure. Cambridge
University Press.
Huttar, George, Enoch O. Aboh, and Felix Ameka. 2013. Relative clauses in Suriname
creoles and Gbe languages. Lingua 129: 96–123.
Ionin, Tania. 2006a. This is definitely specific: Specificity and definiteness in article
systems. Natural Language Semantics 14: 175–234.
2006b. A comparison of article semantics in L2-acquisition and creole languages.
In Claire Lefebvre, Lydia White, and Christine Jourdan (eds.), L2 acquisition and
creole genesis: Dialogues. Amsterdam: Benjamins, 253–273
Jaeggli, Osvaldo, and Nina M. Hyams. 1993. On the independence and interdependence
of syntactic and morphological properties: English aspectual come and go. Natural
Language and Linguistic Theory 11: 313–346.
Kangni, Atah-Ekoué. 1989. La syntaxe du ge. Franckfurt AM: Peter Lang.
Kayne, Richard S. 1975. French syntax. Cambridge, MA: MIT Press.
1994. The antisymmetry of syntax. Cambridge, MA.: MIT Press.
2002. On some prepositions that look DP-internal: English of and French de. Catalan
Journal of Linguistics 1: 71–115.
Kiss, Katalin É. 1998. Identificational focus versus information focus. Language 74:
245–273.
Klein, Wolfgang and Clive Perdue. 1997. The Basic Variety (or: Couldn’t natural
languages be much simpler?). Second Language Research 13: 301–347.
Kluge, Angela. 2000. The Gbe languages varieties of West Africa: A quantitative
analysis of lexical grammar features. MA dissertation, University of Wales,
College of Cardiff.
2005. A synchronic lexical study of Gbe language varieties: The effects of different
similarity judgment criteria. Linguistic Discovery 3: 22–53.
Koopman, Hilda. 1982a. Les constructions relatives. In Claire Lefebvre, Hélène Magloire-
Holly, and Nanie Piou (eds.), Syntaxe de l’Haïtien. Ann Arbor, MI: Karoma, 167–203.
1982b. Les questions. In Claire Lefebvre, Hélène Magloire-Holly, and Nanie Piou
(eds.), Syntaxe de l’Haïtien. Ann Arbor, MI: Karoma, 204–241.
1984. The syntax of verbs: From verb movement rules in the Kru languages to
Universal Grammar. Dordrecht: Foris.
Kouwenberg, Silvia. 2009. The invisible hand in creole genesis: Reanalysis in the
formation of Berbice Dutch. In Enoch O. Aboh and Norval Smith (eds.),
Complex processes in new languages. Amsterdam: Benjamins, 115–158.
2010. Creole studies and linguistic typology: Part 2. Journal of Pidgin and Creole
Kroch, Anthony. 1989a. Reflexes of grammar in patterns of language change. Language
Variation and Change 1, 2: 199–244.
1989b. Function and grammar in the history of English: Periphrastic do. In Ralph W.
Fasold and Deborah Schiffrin (eds.), Language Change and Variation. Current
Issues in Linguistic Theory, vol. 52. Amsterdam: Benjamins, 133–172.
1994. Morphosyntactic variation. In Katharine Beals (ed.), Papers from the 30th
regional meeting of the Chicago Linguistics Society: Parasession on variation and
linguistic theory. Chicago Linguistic Society, 180–201.
328 References
2001. Syntactic change. In Mark Baltin and Chris Collins (eds.), The handbook of
contemporary syntactic theory. Malden, MA: Blackwell, 699–729.
Kroch, Anthony and Ann Taylor. 1997 Verb movement in Old and Middle
English: Dialect variation and language contact. In Ans van Kemenade and
Nigel Vincent (eds.), Parameters of morphosyntactic change. Cambridge
University Press, 297–325.
Anthony Kroch, Ann Taylor, and Donald Ringe. 2000. The Middle English verb-second
constraint: A case study in language contact and language change. In Susan
C. Herring, Pieter van Reenen, and Lene Schøsler (eds.), Textual parameters in
older languages. Amsterdam: Benjamins, 353–392.
Labouret, Henri and Paul Rivet. 1929. Le royaume d’Arda et son évangélisation au
17ème siècle. Paris: Institut d’Ethnologie.
Lardiere, Donna. 2000. Mapping features to forms in second language acquisition.
In John Archibald (ed.), Second language acquisition and linguistic theory.
Malden MA: Blackwell, 102–129.
2005. On morphological competence. Unpublished manuscript, Georgetown University.
Lalla, Barbara and Jean D’Costa. 1991. Language in exile: Three hundred years of
Jamaican Creole. Tuscaloosa, AL: The University of Alabama Press.
Languages of Suriname: Saramaccan-English dictionary. www.sil.org/americas/suri
name/Saramaccan/English/SaramEngDictIndex.html
Law, Robin. 1991. The Slave Coast of West Africa, 1550–1750. New York: Oxford
University Press.
1994. The slave trade in seventeenth-century Allada: A revision. African Economic
Histrory 22: 59–92.
1997. The Kingdom of Allada. Leiden: CNWS Publications, Leiden University.
2011. Africa in the Atlantic world, c.1760–c.1840. In Nicholas Canny and
Philip Morgan (eds.), The Oxford handbook of the Atlantic world: 1450–1850.
New York: Oxford University Press, 585–601.
Lefebvre, Claire. 1989. Instrumental take serial constructions in Haitian and in Fon.
Revue canadienne de linguistique 34: 319–339.
1991. Take serial verb constructions in Fon. In Claire Lefebvre (ed.), Serial verbs:
Grammatical, comparative and cognitive approaches. Amsterdam: Benjamins, 37–79.
1998. Creole genesis and the acquisition of grammar: The case of Haitian creole.
Lefebvre, Claire and Anne Marie Brousseau. 2002. A grammar of Fongbe. Berlin/New
York: Mouton.
Leimgruber, Jakob R. E. 2009. Modelling variation in Singapore English. PhD
dissertation, Pembroke College, University of Oxford.
Ler, Soon Lay Vivien. 2005. An in-depth study of discourse particles in Singapore
English. PhD dissertation, National University of Singapore.
Lim, Lisa. 2011. Tone in Singlish: Substrate features from Sinitic and Malay. In
Claire Lefebvre (ed.), Creoles, their substrates and language typology.
Lightfoot, David. 2006. How new languages emerge. Cambridge University Press.
References 329
Long, Edward. 1774 [2002]. The History of Jamaica. Montreal: McGill-Queen’s

University Press.
Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in
syntax and logical form. Linguistic Inquiry 25: 609–665.
Lord, Carol. 1973. Serial verbs in transition. Studies in African Linguistics 4: 269–295.
Manfredi, Victor. 1993. Verb focus in the typology of Kwa/Kru and Haitian. In
Francis Byrne and Donald Winford (eds.), Focus and grammatical relations in
Creole languages. Amsterdam: Benjamins, 3–51.
Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the
privacy of your own lexicon. In Alexis Dimitriadis, Laura Siegel,
Clarissa Surek-Clark, and Alexander Williams (eds.), Proceedings of the 21st
annual Penn Linguistics Colloquium: Penn Working Papers in Linguistics 4, 2:
201–225.
Mather, Patrick-André 2006. Second language acquisition and creolization: Same (i-)
processes, different (e-)results. Journal of Pidgin and Creole Languages 21:
231–274.
Mathieu, Eric. 2009. From local blocking to Cyclic Agree: The role and meaning of
determiners in the history of French. In Jila Ghomeshi, Ileana Paul, and
Martina Wiltschko (eds.), Determiners: Universals and Variation. Amsterdam:
McColl Millar, Robert. 2000. System. Collapse, rebirth. Weim: Peter Lang.
McWhorter, John. 1998. Identifying the creole prototype: Vindicating a typological
class. Language 74: 788–818.
2001. The world’s simplest grammars are creole grammars. Linguistic Typology 5:
125–165.
2011. Linguistic complexity and simplicity: Why do languages undress? Berlin:
Mouton.
2012. Case closed? Testing the feature pool hypothesis. Journal of Pidgin and Creole
Meisel, Jürgen M. 2011. Bilingual language acquisition and theories of diachronic
change: Bilingualism as cause and effect of grammatical change. Bilingualism:
Language and Cognition 14: 121–145.
Migge, Bettina. 1998. Substrate influence in the formation of the Suriname plantation
creole: A consideration of sociohistorical data and linguistic data from Ndyuka and
Gbe. PhD dissertation, Ohio State University.
2003. Creole formation as language contact: The case of the Suriname creoles.
Migge, Bettina and Don Winford. 2007. Substrate influence on the emergence of the
TMA systems of Surinamese creoles. Journal of Pidgin and Creole Languages 22:
73–99.
2009. The origin and development of possibility in the creoles of Suriname.
In Rachel Selbach, Hugo C. Cardoso, and Margot van den Berg (eds.), Gradual
creolization: Studies celebrating Jacques Arends. Amsterdam: Benjamins,
129–153
Mufwene, Salikoko S. 1986. Number delimitation in Gullah. American Speech 61: 33–60.
330 References
1987. An issue on predicate clefting: Evidence from Atlantic creoles and African
languages. In Philippe Maurer and Thomas Stolz (eds.), Varia Creolica. Bochum:
Brockmeyer, 71–89.
1989. Equivocal structure in some Gullah complex sentences. American Speech 64:
304–326.
1991. Language genesis and human evolution. (Review article on Bickerton 1990.)
Diachronica 8: 239–254.
1994. Genèse de population et genèse de langue. Plurilinguismes 8: 95–113.
1996. The founder principle in creole genesis. Diachronica 13: 83–134.
1999. On the language bioprogram hypothesis: Hints from Tazie. In Michel DeGraff
(ed.), Language creation and language change: Creolization, diachrony, and
development. Cambridge, MA: MIT Press, 95–127.
2001. The ecology of language evolution. Cambridge University Press.
2002 Competition and selection in language evolution. Selection 3: 45–56.
2003. Créoles, écologie sociale, évolution linguistique. Paris: L’Harmattan.
2005a. Créoles, écologie sociale, évolution linguistique: Cours donnés au Collège de
France durant l’automne 2003. Paris: L’Harmattan.
2005b. Language evolution: The population genetics way. In Guenther Hauska (ed.),
Gene, Sprachen, und ihre Evolution. Universitaetsverlag Regensburg, 30–52.
2005c. How many bes are there in English? In Salikoko S. Mufwene, Elaine J.
Francis, and Rebecca S. Wheeler (eds.), Polymorphous linguistics: Jim
McCawley’s legacy. Cambridge, MA: MIT Press, 225–246.
2008. Language evolution: Contact, competition, and change. New York: Continuum
International Publishing Company Group.
2009. Restructuring, hybridization, and complexity in language evolution. In
Enoch O. Aboh and Norval Smith (eds.), Complex processes in new languages.
2010. SLA and the emergence of creoles. Studies in Second Language Acquisition 32:
359–400.
2014. Globalisation économique mondiale des XVIIe-XVIIIe siècles, émergence
des créoles, et vitalité langagière. In Arnaud Carpooran and Yannick Bosquet-
Ballah (eds.), Langues créoles, mondialisation et éducation, Proceedings of the
13th colloquium of the Comité International des Études Créoles, Mauritius
2012. Louis, Mauritius: Creole Speaking Unit, 23–79.
Müller, Natascha. 1998. Transfer in bilingual first language acquisition. Bilingualism:
Language and Cognition 1: 151–171
Müller, Natascha and Aafke Hulk. 2000. Bilingual first language acquisition at the
interface between syntax and pragmatics. Bilingualism: Language and Cognition
3: 227–244.
2001. Cross-linguistic influence in bilingual first language acquisition: Italian
and French as recipient languages. Bilingualism: Language and Cognition 4:
1–21.
Muysken, Pieter. 1981a. Half-way between Quechua and Spanish: The case for
relexification. In Arnold Highfield and Albert Valdaman (eds.), Historicity and
variation in creole studies. Ann Arbor, MI: Karoma, 52–78.
References 331
1981b. Creole tense/mood/aspect system: The unmarked case? In Pieter Muysken

(ed.), Generative studies on creole languages. Dordrecht: Foris, 181–199.
1987. Prepositions and postpositions in Saramaccan. In Mervyn C. Alleyne (ed.),
Studies in Saramaccan language structure, Caribbean Culture Studies 2.
University of Amsterdam, 89–101.
1988. Are creoles a special type of language? In Frederick Newmeyer (ed.),
Linguistics: The Cambridge survey. Vol. II: Linguistic theory: Extensions and
implications. Cambridge University Press, 285–301.
1997. Media Lengua. In Sarah G. Thomason (ed.), Contact languages: A wider
perspective. Amsterdam: Benjamins, 365–426.
2008. Functional categories. Cambridge University Press.
Muysken, Pieter and Norval Smith. 1986. Substrata versus universals in creole genesis.
Notley, Anna, Elisabeth van der Linden, and Aafke Hulk. 2007. Cross-linguistic
influence in bilingual children: The case of dislocation. In Sergio Baauw,
Frank Drijkoningen, and Manuela Pinto (eds.), Romance languages and
linguistic theory 2005. Amsterdam: Benjamins, 229–259.
Palmer, Frank Robert. 1987. Mood and modality. Cambridge University Press.
Pazzi, Roberto. 1979. Introduction à l’histoire de l’aire culturelle Aja-Tado. Institut
National des Sciences de l’Éducation.
1984. Aperçu sur l’implantation actuelle et les migrations anciennes des peuples de
l’aire culturelle Aja-tado. In François de Medeiros (ed.), Peuples du Golfe du
Bénin. Paris: Karthala, 11–20.
Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric Reuland
and Alice ter Meulen (eds.), The representation of (in)definiteness. Cambridge,
MA: MIT Press, 98–129.
Pintzuk, Susan. 1999. Phrase structures in competition: Variation and change in Old
English word order. New York: Garland.
Plag, Ingo. 2008a. Creoles as interlanguages: Inflectional morphology. Journal of
Pidgin and Creole Languages 23: 109–130.
2008b. Creoles as interlanguages: Syntactic structures. Journal of Pidgin and Creole
2009a. Creoles as interlanguages: Phonology. Journal of Pidgin and Creole
2009b. Creoles as interlanguages: Word-formation. Journal of Pidgin and Creole
2011. Creolization and admixture typology, feature pools, and second language
acquisition. Journal of Pidgin and Creole Languages 26: 89–110.
Platzack, Christer. 2001. The vulnerable C-domain. Brain and Language 77: 364–377.
Pollock, Jean-Yves. 1989. Verb movement, universal grammar, and the structure of IP.
Linguistic Inquiry 20: 356–424.
Postma, Johannes. 1970. The Dutch participation in the African slave trade: Slaving on
the Guinea Coast, 1675–1795. PhD dissertation, Michigan State University.
1975. The origin of African slaves: The Dutch activities on the Guinea Coast,
1675–1795. In Stanley L. Engerman and Eugene D. Genovese (eds.), Race and
332 References
slavery in the western hemisphere: Quantitative studies. Princeton University

Press, 33–49.
Prévost, Philippe and Lydia White. 2000. Missing surface inflection or impairment in
second language acquisition? Evidence from tense and agreement. Second
Language Research 16:103–133.
Price, Richard. 1973. Marron societies: Rebel slave communities in the Americas.
Doubleday: Anchor Press.
1975. Kikoongo and Saramaccan: A reappraisal. Bijdragen tot de Taal-, Land- en
Volkenkunde 131: 461–478.
1976. The Guiana Maroons: A historical and bibliographical introduction.
Baltimore: Johns Hopkins University Press.
1983. First-Time: The historical vision of an Afro-American people. Baltimore: Johns
Hopkins University Press.
(ed.) 1996. Maroon societies: Rebel slave communities in the Americas, 3rd edn.
Baltimore: Johns Hopkins University Press.
2002. Maroons in Suriname and Guyane: How many and where. New West Indian
Guide 76: 81–88.
2008. Travels with Tooy: History, memory, and the African American imagination.
University of Chicago Press.
Price, Richard and Sally Price. 1991. Two evenings in Saramaka. University of Chicago
Press.
2003. The root of roots, or, how Afro-American anthropology got its start. Chicago:
Prickly Paradigm Press.
Price, Sally and Richard Price. 1999. Maroon arts: Cultural vitality in the African
diaspora. Boston: Beacon Press.
Prince, F. Ellen. 1981. Toward a taxonomy of given-new information. In Peter Cole
(ed.), Radical pragmatics. New York: Academic Press, 223–255.
Ribbert, Anne and Folkert Kuiken. 2010. L2-induced changes in the L1 of
Germans living in the Netherlands. Bilingualism: Language and Cognition
13: 41–48.
Ritter, Elizabeth. 1991. Two functional categories in Noun Phrases: Evidence from
Modern Hebrew. Syntax and Semantics 25: 37–62.
1992. Cross-linguistic evidence for Number Phrase. Canadian Linguistics Review 37:
197–218.
1995. On syntactic category of pronouns and agreement. Natural Languages and
Linguistics Theory 13: 405–443.
Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris.
1996. Residual verb second and the wh-criterion. In Adriana Belletti and Luigi Rizzi
(eds.), Parameters and functional heads: Essays in comparative syntax. New York:
Oxford University Press, 63–90.
1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of
grammar. Dordrecht: Kluwer, 281–337.
Roberts, Ian. 1993. Verbs and diachronic syntax: A comparative history of English and
French. Dordrecht: Kluwer.
2007. Diachronic syntax. New York: Oxford University Press.
References 333
Roberts, Ian and Ana Roussou. 2003. Syntactic change: A minimalist approach to
grammaticalization. Cambridge University Press.
Roeper, Thomas. 1999. Universal bilingualism. Bilingualism: Language and Cognition
2: 169–186.
Roodenburg, Jasper. 2004. Pour une approche scalaire de la déficience nominale: La
position du français dans une théorie des “noms nus.” LOT PhD series 99.
Utrecht: LOT.
Ross, J. Robert. 1967. Constraints on variables in syntax. PhD dissertation, MIT,
Cambridge, MA.
Rountree, S. Catherine. 1992. Saramaccan grammar sketch. Languages of the
Guianas 8. Paramaribo: Summer Institute of Linguistics.
Rountree, S. Catherine and Naomi Glock. 1982. Saramaccan for beginners: A
pedagogical grammar of the Saramaccan language. Paramaribo: Institut voor
Taalwetenschap and Summer Institute of Linguistics.
Saah, Kofi. 2010. Relative clauses in Akan. In Enoch O. Aboh and James Essegbey
(eds.), Topics in Kwa syntax. Dordrecht: Springer, 91–107.
Schumann, Christian. 1783. Neger-Englishes Wörterbuch. Manuscript published in
André Kramp (1983) Early creole lexicography: A study of C. L. Schumann’s
manuscript dictionary of Sranan, pp. 44–305. Dissertation, University of
Leiden.
Schumann, John H. 1978. The pidginization process: A model for second language
acquisition. Rowley: Newbury House.
Sebba, Mark. 1987. The syntax of serial verbs: An investigation into serialization in
Sranan and other languages. Amsterdam: Benjamins.
1997. Contact languages: Pidgins and creoles. Houndmills: PALGRAVE.
Seuren, Pieter. 1990a. Still no serials in Seselwa: A reply to “Seselwa serialization and
its significance.” Journal of Pidgin and Creole Languages 5: 271–292.
1990b. Verb syncopation and predicate raising in Mauritian Creole. Linguistics 28:
809–844.
1991. The definition of serial verbs. In Francis Byrne and Thom Huebner (eds.),
Development and structures of Creole. Amsterdam: Benjamins, 193–206.
Shappeck, Marco. 2011. Quichua-Spanish language contact in Salcedo, Ecuador:
Revisiting Media Lengua Syncretic Language Practices. PhD dissertation,
University of Illinois, Urbana-Champaign.
Schwartz, Bonnie and Rex Sprouse. 1996. L2 cognitive states and the full transfer/full
access model. Second Language Research 12: 40–77.
Siegel, Jeff. 2004. Morphological elaboration. Journal of Pidgin and Creole Languages
19: 333–362.
2008. The emergence of pidgin and creole languages. Oxford University Press.
Silveira, Luis. 1945. Œuvre nouvelle de langue generale du mina. Manuscript
de la bibliothèque publique de Evora et de la bibliothèque nationale de
Lisbone. Département des publications et bibliothèque Agence générale des
colonies.
Singler, John Victor. 1988. The homogeneity of the substrate as a factor pidgin/creole
genesis. Language 64: 27–51.
334 References
1996. Theories of creole genesis, sociohistorical considerations, and the evaluation of

evidence: The case of Haitian Creole and the Relexification Hypothesis. Journal of
Pidgin and Creole Languages 11: 185–230.
Smith, Norval. 1987. The genesis of the creole languages of Surinam. PhD dissertation,
University of Amsterdam.
1996. W-focus in Saramaccan: Substrate feature or grammaticalization? In
Philip Baker and Anand Syea (eds.), Changing meanings, changing functions:
Papers relating to grammaticalization in contact languages. London: University
of Westminster Press, 113–128.
1999. Pernabuco to Surinam 1654–1665? The Jewish slave controversy. In
Margnus Huber and Mikeal Parkvall (eds.), Spreading the word: Papers on the
issue of diffusion of Atlantic creoles. London. University of Westminster Press,
251–298.
2009. Simplification of a complex part of grammar or not? What happened to
KiKoongo nouns in Saramaccan? In Enoch O Aboh and Norval Smith (eds.),
Complex processes in new languages. Amsterdam: Benjamins, 51–73.
Smith, Norval and Hugo Gardoso. 2004. A new look at the Portuguese element in
Saramaccan. Journal of Portuguese Linguistics 3: 115–147.
Smith, Norval and Tonjes Veenstra. 1994. Ma-nominalizations in Saramaccan.
Handout, University of Amsterdam.
1998. Synthetic compounds in a radical creole: Abrupt versus gradual change. Creole
Conference, Regensburg: June 24–27.
Sorace, Antonella. 2000. Differential effects of attrition in the L1 syntax of near-native
L2 speakers. In S. Catherine Howell, Sarah Fish, and Thea Keith-Lucas (eds.),
Proceedings of the 24th annual Boston University conference on language
development. Somerville, MA: Cascadilla Press, 719–725.
2005. Selective optionality in language development. In Leonie Cornips and Karen
P. Corrigan (eds.), Syntax and variation: Reconciling the biological and the social.
Sorace, Antonella and Ludovica Serratrice. 2009. Internal and external interfaces in
bilingual language development: Beyond structural overlap. International Journal
of Bilingualism 13: 195–210.
Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for
constituent structure. Linguistic Inquiry 19: 425–449.
Sterlin, Marie-Denise. 1989. Les caractéristiques de pou: Un modal en position de
complémenteur. Revue québécoise de linguistique 18: 131–146.
Stewart, Michele Marie. 2007. Quantification in Jamaican creole. The syntax and
semantics of evri (‘every’) in interaction with indefinites. PhD dissertation, The
University of the West Indies.
Stewart, Osamuyimen Thompson. 1997. Object agreement and the serial verb
construction: Some minimalist considerations. In Rose-Marie Déchaine and
Victor Manfredi (eds.), Object position in Benue-Kwa. The Hague: Holland
Academic Graphics, 153–168.
1998. The serial verb construction parameter: PhD dissertation, McGill University,
Montréal.
References 335
Sylvain, Suzanne. 1936. Le créole haïtien: Morphologie et syntaxe. Geneva: Slatkine

Reprints.
Szabolcsi, Anna. 1987. Functional categories in the noun phrase. In István Kenesei
(ed.), Approaches to Hungarian. Szeged: JATE, 167–189.
1994. The noun phrase. In Ferenc Kiefer and Katalin E. Kiss (eds.), Syntax and
semantics 27: The syntactic structure of Hungarian. New York: Academic Press,
179–274.
Tagète, Jean and Raphael G. Urciolo. 1993. Haitian Creole–English Dictionary.
Kensington, MD: Dunwoody Press.
Takahash, Shoichi and Martina Gračanin-Yuksek. 2008. Morphosyntax of movement
dependencies in Haitian Creole. Syntax 11: 223–250.
Thomason, Sarah G. and Terrence Kaufman. 1988. Language contact, creolization and
genetic linguistics. Berkeley: University of California Press.
Thornton, John K. and Linda M. Heywood. 2007. Central Africans, Atlantic Creoles
and the foundation of the Americas 1585–1660. Cambridge University Press.
Tracy, Rosemarie. 1998. Transfer versus coexistent systems. Bilingualism: Language
and Cognition 1: 185–186.
Tsimpli, Ianthi Maria and Sorace Antonella. 2006. Differentiating interfaces: L2
performance in syntax-semantics and syntax-discourse phenomena. In
David Bamman, Tatiana Magnitskaia, and Colleen Zaller (eds.), Proceedings of
the annual Boston University conference on language development 30. Somerville,
MA: Cascadilla Press, 653–664.
Valenzuela, Elena. 2005. L2 ultimate attainment and the syntax-discourse interface: The
acquisition of topic constructions in non-native Spanish and English. PhD
dissertation, McGill University.
Van den Avenne Cécile. 2005. Bambara et français-tirailleur. Une analyse de la politique
linguistique de l’armée colonial française: la Grande Guerre et après. Documents
pour l’histoire du français langue étrangère ou seconde 35: 123–150.
Van den Berg, Margot. 2003. Early 18th century Sranan -man. In Ingo Plag (ed.),
Phonology and morphology of creole languages. Tübingen: Niemeyer, 231–251.
2007. A grammar of early Sranan. Zetten: Drukkerij Manta.
Van den Berg, Margot and Enoch O. Aboh. 2001. Compounding and derivation in two
morphologically “poor” languages. Morfologiedagen Utrecht, December 13–14,
2001.
2013. Done already? A comparison of completive markers in the Gbe languages and
Sranan Tongo. Lingua 129: 150–172.
van Kemenade, Ans. 1987. Syntactic case and morphological case in the history of
English. Dordrecht: Foris; later Berlin: Mouton de Gruyter.
Veenstra, Tonjes. 1996. Serial verbs in Saramaccan: Predication and Creole genesis.
HIL Dissertations 17. Den Haag: Holland Academic Graphics.
2000. Verb serialization and object Position. Linguistics 38: 867–888.
Vicente, Luis. 2005. Toward a unified theory of movement: An argument from Spanish
predicate clefts. Leiden Papers in Linguistics 2: 43–67.
Wagner, Suzanne. 2004. English dialects in the Southwest: Morphology and syntax. In
Bernd Korthmann, Edgar W. Schneider, Kate Burridge, Rajend Mesthrie, and
336 References
Clive Upton (eds.), A handbook of varieties of English. Vol. II: Morphology and
syntax. Berlin: Mouton, 154–174.
Wakefield, John C. 2012. A floating tone discourse morpheme: The English equivalent
of Cantonese lo1. Lingua 122: 1739–1762.
Weerman, Fred. 2011. Diachronic change: Early versus late acquisition. Bilingualism:
Language and Cognition 14: 149–151.
Whinom, Keith. 1971. Linguistic hybridization and the “Special Case” of Pidgins and
Creoles. In Dell H. Hymes (ed.), Pidginization and creolization of languages.
Cambridge University Press, 91–115.
Wijnen, Beppie and Mervyn Alleyne. 1987. A note on fu in Saramaccan. In Mervin
C. Alleyne (ed.), Studies in Saramaccan language structure. Caribbean Culture
Studies 2, University of Amsterdam, 41–51.
Winford, Donald. 1985. The syntax of Fi complements in Caribbean English Creole.
Language 61: 588–624.
1993. Predication in Caribbean English Creoles. Amsterdam: Benjamins.
Winford, Donald and Bettina Migge. 2007. Substrate influence on the emergence of the
TMA systems of the Surinamese creoles. Journal of Pidgin and Creole Languages
22: 73–99.
Wurmbrand, Suzanne. 2001. Infinities. Restructuring and clause structure. Berlin:
Mouton.
2004. Two types of restructuring: Lexical versus functional. Lingua 114: 991–1014.
Yuka, Lendzemo Constantine and Esohe Mercy Omoregbe. 2010. The internal structure
of the Edo verb. California Linguistic Notes 35: 1–19.
Zwart, Jan-Wouter. 1997a. Morphosyntax of verb Movement: A minimalist approach to
the syntax of Dutch. Dordrecht: Kluwer.
1997b. The Germanic SOV languages and the universal base hypothesis. In
Liliane Haegeman (ed.), The new comparative syntax. London: Longman
Linguistics Library, 246–267.
2006. Complementizer agreement and dependency marking typology. In Marjo
van Koppen, Frank Landsbergen, Mika Poss, and Jenneke van der Wal (eds.),
Special issue of Leiden Working Papers in Linguistics 3: 53–72.
Author index
Aboh, i, ii, iii, iv, ix, x, xi, xii, xiii, xiv, xvi, 3, 8, Bailey, 214
9, 10, 11, 12, 13, 14, 17, 41, 42, 43, 44, 45, Baker, 140, 149, 193, 281, 282, 283, 284, 285,
53, 61, 69, 87, 88, 92, 98, 101, 102, 107, 113, 298, 334
119, 123, 134, 135, 137, 138, 139, 141, 143, Bakker, 6, 145, 168, 230, 242
144, 145, 146, 149, 154, 155, 156, 157, 159, Bernstein, 177, 214
162, 163, 171, 173, 176, 180, 181, 185, 186, Bickerton, x, 6, 7, 61, 62, 83, 84, 85, 86, 87,
192, 193, 196, 198, 203, 205, 206, 207, 208, 88, 89, 90, 91, 92, 93, 109, 171, 175, 176,
214, 215, 217, 221, 222, 226, 227, 228, 229, 194, 209, 269, 270, 281, 297, 303, 330
230, 231, 232, 233, 234, 236, 238, 239, 246, Bobyleva, xvi, 64, 87, 92, 101, 204, 214
248, 249, 250, 251, 252, 253, 254, 256, 257, Boeckx, 140
258, 259, 260, 261, 262, 264, 266, 270, 271, Booij, 101, 102, 103, 104
272, 276, 281, 282, 284, 285, 288, 289, 290, Borer, 101, 140
291, 298, 307, 318, 330, 334 Brousseau and Lumsden, 177
Aboh and Ansaldo, 13, 115, 143, 145, 169, Bruyn, xv, 16, 156, 177, 213, 215,
194, 195 Burzio, 287
Aboh and DeGraff, 17, 43, 61, 77, 87, 88, 126, Bybee, 89, 193
173, 176, 177, 179, 180, 182, 188, 194, 206, Byrne, 258, 270, 333
208, 212, 225, 265, 266, 310
Aboh and Dyakonova, 14, 47, 139, 146, 149, Cable, 65, 295
222, 254, 257, 266, 281 Campbell, 216, 282
Aboh and Essegbey, 41, 42, 44, 45, 46, 47, 176, Capo, 20, 41, 42
250, 253 Cardinaletti and Giusti, 292
Aboh and Nauze, 236 Cardinaletti and Shlonsky, 292
Aboh and Pfau, 203, 238 Cardinaletti and Starke, 69
Aboh and Smith, 61, 155, 161, 162, 163, 164, Chaudenson, ii, x, xi, 7, 18, 33, 59, 62, 63, 64,
167, 168, 169, 170, 186, 289, 306 65, 67, 68, 69, 70, 72, 73, 74, 83, 93, 99, 109,
Aboh, Hartmann, and Zimmermann, 47 117, 118, 134, 135, 175, 179
Aboh, Pfau, and Zeshan, 88 Cheng and Sybesma, 87
Aboh, Smith, and Veenstra, 280 Chomsky, 2, 3, 9, 13, 105, 140, 149, 257, 294,
Adande, 20, 22, 23, 24, 32 308, 322
Aikhenvald, 269, 275 Cinque, 12, 14, 89, 102, 140, 185, 186, 192,
Alleyne, 13, 91, 97, 331 193, 271, 287, 291, 292, 293, 294, 295, 317,
Ameka, xv, 44, 45, 46, 281 318, 324
Anglade, 17, 58 Clements, ii, 101
Anonymous, 246 Collins, 263, 276, 282, 321
Ansaldo, ii, xvi, 8, 13, 98, 114, 184, 211, Comrie, 89
259, 304 Croft, 3, 139
Ansaldo, Matthews, and Lim, 61 Curtin, 29, 323
Arends, 10, 16, 33, 35, 39, 53, 97, 137, 255,
322, 329 d’Elbée, 23, 24, 29, 30, 31, 36, 37
Aronoff, Meir, and Sandler, 312 Da Cruz, 246, 282, 284, 292
Avolonto, 145 Damonte, 227, 228, 229, 232, 233
337
Downloaded from https://www.cambridge.org/core. The Librarian-Seeley Historical Library, on 06 Jan 2020 at 22:09:37, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139024167
338 Author index
de Castro, 122, 135 Jaeggli and Hyams, 281, 295, 298

de Cauna, 54
Debien, vi, 16, 39, 54, 55, 56, 57, 58 Kangni, 106
Déchaine and Manfredi, 47 Kayne, 9, 12, 69, 159, 193, 207, 220, 221, 248
DeGraff, x, xii, xv, xvi, 3, 5, 8, 10, 58, 61, 65, 68, Kiss, 173, 176, 256
70, 77, 78, 79, 80, 81, 82, 83, 85, 87, 97, 99, Klein and Perdue, 6, 107, 108, 109
101, 104, 106, 109, 120, 125, 128, 132, 135, Kluge, 41, 42
137, 138, 141, 142, 156, 169, 173, 177, 178, Koopman, 47, 181, 208, 265
180, 181, 194, 207, 208, 209, 227, 254, 255, Kouwenberg, 122, 170, 327
256, 265, 266, 269, 271, 276, 278, 279, 280, Kroch, 1, 2, 3, 9, 61, 169, 173, 327
283, 286, 296, 298, 300, 301, 306, 321, 330 Kroch and Taylor, 2, 169
den Dikken, 159 Kroch, Taylor, and Ringe, 1, 3
Denis, 128
Déprez, 64, 73, 177, 185, 204 Labouret and Rivet, 31, 50
Déprez, Sleeman, and Guella, 185 Lalla and D’Costa, 131
Diouf, 39 Lardiere, 12, 142, 195, 211
Donati and Branchini, 313 Law, 16, 19, 20, 22, 23, 24, 26, 27, 28, 29, 30,
Ducœurjoly, 68, 128, 179 31, 32, 33, 34, 35, 36, 37, 38, 58
Durrleman-Tame, 239, 254, 255, 256 Lefebvre, ix, xv, 6, 7, 17, 62, 73, 74, 75, 76, 78,
79, 80, 81, 82, 93, 109, 137, 175, 177, 179,
Eltis, 29, 30, 37, 38, 39, 51, 58 180, 181, 184, 207, 208, 260, 265, 270, 274,
Emmer, 28, 29, 97 280, 281, 284, 285, 297, 317
Enç, 177 Lefebvre and Brousseau, 177, 260, 274, 281
Essegbey, xv, 53, 145, 146, 237, 246 Lightfoot, 2, 9, 169, 173
Lim, 196,
Fattier, 127 Long, 129, 130, 131, 132, 135
Foley and van Valin, 89, 193 Lyle Campbell, 89
Fouchard, 16, 54
Manfredi, 47, 317, 323
Gibson, 239 Marantz, 161
Giusti, 177 Mather, 94
Glaude, 179, 180, 266, 269, 272, 299, 300 Mathieu, 179, 188
Glaude and Zribi-Hertz, 65, 66, 254, 256, 257, McColl Millar, 2
258, 259, 266 McWhorter, xi, xii, 6, 7, 15, 84, 92, 141, 145,
Glock, 234 169, 170, 194, 308
Goad and White, 12, 195 Meisel, 173
Goad, White, and Steele, 211 Migge, 16, 53, 336
Granfeldt, 72 Migge and Winford, 53
Mufwene, ii, v, x, xiv, xv, xvi, 3, 5, 8, 13, 16, 17,
Haddican, 219 18, 19, 31, 47, 51, 54, 55, 59, 61, 83, 85, 87,
Haeberli, 2, 326 90, 96, 97, 98, 112, 114, 115, 116, 117, 121,
Haegeman, 141 122, 125, 131, 132, 133, 134, 135, 137, 138,
Haegeman and Ürögdi, 244 139, 141, 142, 144, 145, 169, 170, 175, 177,
Hagemeijer, 292
184, 211, 227, 234, 235, 242, 259, 294, 300,
Hale and Keyser, 149
304, 313, 315
Hall, 84, 299, 300 Müller, 196, 197, 199, 200, 201
Harbour, 266 Müller and Hulk, 196, 199, 201
Hawkins, 306 Muysken, xv, 9, 18, 61, 74, 75, 78, 82, 85, 89,
Haznedar and Schwartz, 12, 195 91, 169, 231, 270, 281, 320, 322, 331
Heim, 182 Muysken and Smith, 61, 85
Hengeveld, 89, 193
Henri, 101 Notely, van Linden, and Hulk, 200
Holm, 130 Notley, van der Linden, and Hulk, 199
Huttar, Aboh, and Ameka, 104, 218, 219, 309
Palmer, 232
Ionin, 176, 210 Pazzi, 20, 22, 23, 26, 28, 30, 32, 33, 34, 35, 36, 40
Author index 339
Pesetsky, 176 Smith and Gardoso, 16

Pintzuk, 2, 327 Smith and Veenstra, 158
Plag, xi, 7, 61, 62, 78, 93, 94, 95, 96, 97, 98, 99, Sorace, 12, 195, 196
101, 102, 103, 104, 106, 107, 108, 109, 145, Sorace and Serratrice, 12, 195
170, 171, 175, 194 Sterlin, 266
Platzack, 12, 195 Stewart, Michele, 101
Pollock, 89, 141 Stewart, Osamuyimen, 103, 282
Postma, 16, 53 Sylvain, 17, 73, 77, 81, 178, 181, 207
Price, 16, 53 Szabolcsi, 92, 102, 177, 184, 185, 192, 206
Price and Price, 53
Prince, 176 Tagète and Urciolo, 71
Takahash and Gračanin-Yuksek, 265
Ritter, 101 Thomason and Kaufman, 85, 111, 113
Rizzi, 14, 106, 140, 185, 198, 203, 222, 228, Tracy, 173
244, 253, 268, 271, 307, 318 Tsimpli and Sorace, 12, 195
Roberts, 2, 248, 318, 332
Roberts and Roussou, 248 Valenzuela, 196
Roeper, 173, 313 van den Avenne, 126
Roodenburg, 178, 179 van den Berg, xv, 17, 91, 137, 154, 155, 270,
Ross, 276 272, 329
Rountree and Glock, 102, 149, 150, 230, van Kemenade, 2, 328
252, 310 Veenstra, 156, 173, 227, 229, 231, 242, 269,
270, 272, 278, 286, 292, 302, 334, 335
Schumann, 93, 96, 107
Schumann, Christian, 154, 255 Wagner, 241
Schwartz and Spouse, 82 Wakefield, 196
Sebba, 111, 270 Weerman, 173, 174, 195
Seuren, 269, 303, 333 Whinom, 5
Shappeck, 78 Wijnen and Alleyne, 227, 228, 230, 242
Siegel, 18, 184, 329 Winford, 53, 223, 224, 227, 234, 241, 272, 329
Silveira, 122, 123 Winford and Migge, 272
Singler, vi, 17, 39, 51, 54, 55, 58, 83, Wurmbrand, 14, 271
271, 334
Smith, xv, 10, 16, 39, 53, 59, 121, 137, 145, Yuka and Omoregbe, 103, 104
156, 250, 251, 270, 318, 320, 321, 322, 330,
331 Zwart, 106, 198
Language index
Adja, 55, 58 Fa d’Ambu, 93, 205

Akan, 105, 275 Fanti, 36, 58
American English, xii, 281 Finnish, 107
Angolar, 93, 205 Fongbe, 7, 44, 45, 50, 52, 53, 58, 123, 124, 136,
Ashanti, 58 144, 245, 246, 247, 250, 251, 263, 268,
274, 284, 321, 323, 328
Bahamian, 93, 205 French, 11, 15, 65, 66, 67, 68, 70, 71, 72, 76, 77,
Bantu Languages, 8, 17, 39, 42, 47, 51, 259 81, 82, 88, 89, 91, 100, 101, 102, 103, 106,
Baoule, 58 107, 115, 119, 137, 138, 171, 173, 175,
Baule, 275 177, 178, 180, 181, 182, 183, 184, 186,
Berbice Dutch, 93, 122, 205, 314 188, 189, 190, 191, 192, 194, 196, 198,
Bislama, 93, 205 199, 206, 207, 208, 209, 210, 211, 212,
214, 215, 217, 221, 225, 257, 259, 264,
Cantonese, 196, 200, 336 265, 267, 268, 272, 280, 287, 290, 291,
Cape Verde, 93 292, 295, 297, 298, 299, 300, 301, 302
Cape Verdean, 100, 205 French Caribbean, vi, 54, 67
Caribbean English creoles, 223, 225, 235, 240, French creoles, 7, 63, 73
241, 242 French koiné, 7, 63
Central Gbe, 41 French-based Caribbean creoles, 57, 58
French-based creoles, 58, 63, 64, 66, 67, 72, 73,
Dutch, 1, 2, 3, 106, 107, 109, 200, 201, 202, 92, 127, 204
270, 314
Gbe group, ix
Eastern Gbe, 41, 119, 124, 218, 245, Gbe languages
297, 319 Gbe, ix, 10, 11, 17, 39, 41, 42, 43, 44, 45, 46,
Eastern Phla-Phera, 41 48, 49, 50, 51, 54, 58, 60, 66, 74, 78, 81,
Edo, 103, 104, 105 87, 89, 100, 102, 107, 119, 123, 124, 138,
English, 106, 107, 136, 143, 144, 146, 148, 149, 143, 144, 145, 148, 149, 150, 151, 154,
150, 151, 152, 153, 155, 158, 159, 160, 156, 158, 159, 160, 162, 163, 168, 175,
161, 162, 164, 165, 166, 167, 168, 170, 176, 177, 185, 186, 188, 189, 192, 203,
171, 172, 173, 175, 182, 186, 187, 188, 209, 211, 212, 213, 214, 215, 216, 218,
189, 190, 191, 192, 195, 196, 198, 199, 219, 220, 221, 223, 225, 226, 227, 235,
200, 211, 212, 213, 214, 215, 216, 217, 236, 237, 239, 245, 247, 249, 250, 251,
218, 221, 225, 226, 227, 231, 234, 235, 252, 253, 257, 260, 267, 272, 273, 275,
237, 240, 241, 242, 243, 244, 245, 249, 276, 278, 282, 284, 296, 297, 298, 299,
250, 251, 253, 259, 264, 265, 267, 268, 301, 303, 327, 335
270, 272, 280, 286, 288, 289, 291, 292, Gbe substrates, ix
295, 299, 307, 308, 309, 310 Gengbe, 36, 44, 52, 58, 106, 139, 219, 235, 236,
English-based creoles, 193, 241, 307, 309 245, 247, 326
Ewegbe, 45, 46, 58, 106, 119, 219, 245, 246, German, 1, 3, 107, 196, 197, 198, 199, 200,
247, 275, 276, 282 201, 202, 326
340
Downloaded from https://www.cambridge.org/core. UVA Universiteitsbibliotheek, on 14 Mar 2020 at 09:13:15, subject to the Cambridge Core
Language index 341
Germanic languages Latin, 2, 100, 134

Germanic, 12, 80, 107, 177, 196, 212, 294, Lingala, 93, 205
295, 296, 309, 321, 326 Louisiana, 64, 93, 204, 205
Guadeloupian, 63, 64, 73, 204
Gullah, xii, 177, 234, 329, 330 Mandarin Chinese, 87, 280, 309, 310
Gungbe, 10, 11, 14, 23, 42, 43, 44, 45, 46, 48, Martiniquais, 67
49, 50, 52, 58, 65, 76, 77, 79, 80, 81, 119, Mauritian, 63, 64, 66, 67, 72, 93, 204, 205,
123, 124, 136, 143, 144, 145, 146, 148, 326, 333
149, 150, 151, 152, 153, 154, 155, 156, Maxi, 35, 42, 58
159, 160, 161, 162, 164, 165, 166, 167, Media Lengua, 78, 331, 333
171, 172, 173, 175, 176, 177, 178, 179, Middle English, xii, 1, 3, 4, 6, 8, 12, 13, 328
180, 181, 182, 183, 184, 185, 186, 187, Mina, 26, 36, 58, 122, 123, 124, 126, 323, 324
188, 189, 190, 191, 203, 206, 207, 208, Modern English, 1, 4, 6, 13
209, 210, 211, 212, 213, 214, 215, 216, Modern French, 178, 179, 266
217, 218, 219, 221, 223, 225, 226, 235, Modern Gbe, 48, 49
236, 237, 238, 239, 240, 241, 243, 244,
245, 246, 247, 248, 249, 250, 251, 252, Nago, 47, 55, 58
253, 254, 255, 257, 258, 259, 260, 264, Ndyuka, 218, 329
266, 267, 268, 271, 272, 273, 274, 275, Niger Congo, 58
276, 278, 279, 280, 281, 282, 283, 284, Non-standard French, 63, 64, 67, 72, 73, 74
285, 286, 287, 288, 289, 291, 292, 295, Nubi, 93, 205
296, 297, 298, 299, 300, 301, 302, Nzima, 58
317, 318,
Guyanese, 64, 72, 204, 224, 239, 325 Old English, xii, 1, 2, 13, 188, 331
Old French, 77, 179, 188
Haitian, ix, 7, 9, 10, 11, 14, 15, 16, 19, Old Gbe, 48
47, 51, 54, 56, 57, 59, 64, 65, 66, 67,
68, 70, 71, 72, 74, 76, 77, 78, 79, 80, Papiamentu, 93, 205
81, 82, 87, 88, 89, 91, 93, 100, 129, Pitkern, 111, 113, 114, 115
132, 137, 138, 171, 175, 176, 177, Portuguese, ii, xi, 20, 26, 27, 28, 31, 39, 65, 79,
178, 179, 180, 181, 182, 183, 184, 185, 100, 109, 124, 134, 175, 192, 212, 221,
186, 187, 188, 189, 190, 191, 192, 194, 250, 270, 322, 323, 334
204, 205, 206, 207, 208, 209, 211, 212,
215, 216, 217, 221, 225, 226, 227, 254, Quebec French, 64, 118
255, 256, 257, 265, 266, 267, 269, 270, Quechua, 74, 78, 330
271, 272, 273, 276, 279, 280, 281, 283,
286, 292, 293, 294, 295, 296, 297, 298, Romance languages, 2, 100, 103, 134, 177,
299, 300, 301, 302, 303, 324, 325, 326, 212, 214, 215, 218, 321, 324
328, 329 Romance, 2, 6, 8, 11, 12, 65, 66, 80, 88, 96,
Hokkien, 196 100, 101, 103, 104, 134, 173, 177, 184,
193, 206, 212, 214, 215, 216, 218, 242,
Igbo, 58, 282, 250, 256, 257, 264, 267, 270, 277, 291,
Ijo, 122, 314 293, 294, 295, 296, 309, 310, 321, 322,
Italian, 107, 196, 198, 199, 322, 330, 332 323, 324, 331
Khoisan, 282, St. Lucian, 93, 205

Kikongo, 17, 53, 119, 121, 127 Sango, 93, 205
Kristang, 93 Saramaccan, ix, xi, 11, 14, 47, 53, 59, 65,
Kriyol, 93, 205 102, 104, 105, 106, 107, 109, 113,
Kwa, vi, xv, 7, 8, 14, 15, 17, 39, 40, 46, 47, 48, 116, 121, 127, 137, 143, 144, 145, 149,
51, 54, 55, 56, 57, 58, 59, 60, 66, 88, 102, 150, 151, 152, 153, 154, 155, 156, 158,
117, 137, 156, 176, 193, 222, 226, 242, 160, 161, 166, 167, 169, 170, 171, 175,
250, 252, 255, 256, 257, 263, 264, 266, 177, 186, 187, 188, 190, 191, 192, 193,
267, 275, 280, 282, 294, 296, 304, 312, 194, 195, 212, 215, 216, 217, 218, 219,
317, 318, 323, 329, 334 220, 221, 222, 223, 226, 227, 228, 229,
342 Language index
230, 231, 232, 233, 234, 235, 236, 161, 166, 167, 170, 171, 175, 177, 186,
237, 238, 239, 240, 241, 242, 243, 244, 187, 188, 189, 190, 191, 192, 195, 205,
245, 248, 249, 250, 251, 252, 253, 254, 206, 211, 212, 213, 214, 215, 216, 217,
255, 257, 258, 259, 263, 264, 265, 266, 218, 221, 254, 270, 272, 283, 320, 321,
267, 271, 272, 273, 278, 279, 280, 285, 333, 335, 336
286, 292, 293, 294, 298, 302, 303, 307, Sri Lanka Malay, 195, 320
308, 309, 310, 311, 318, 319, 320, 322, Standard English, 118, 241
323, 325, 328, 331, 332, 333, 334, Swedish, 107
335, 336
Seychelles, 93, 204, 205 Tayo, 93, 205
Singapore English, ii, 196, 200, 328 Tok Pisin, 93, 205
Sinitic, 100, 102, 107, 310, 328 Tubaian, 111, 113, 114, 115
South Hollandic Dutch, 106
Spanish, ii, 24, 29, 31, 48, 49, 65, 74, 78, 79, Western Gbe, 41, 106, 119, 218, 219, 220, 221,
107, 109, 214, 270, 330, 333, 335 245, 246, 247
Sranan, ix, 9, 10, 11, 47, 59, 91, 92, 93, 116,
136, 137, 154, 155, 156, 158, 159, 160, Yoruba, 20, 34, 38, 40, 58, 119, 255
Subject index
acquisition, i, ii, v, xiii, xiv, xv, 3, 5, 6, 7, 13, 15, colonial language, 18, 60, 79, 121
51, 60, 61, 62, 72, 74, 84, 85, 90, 91, 93, 94, Compagnie des Indes Occidentales, 23
95, 96, 97, 98, 99, 107, 108, 110, 111, 114, Company of Royal Adventurers, 29
116, 130, 134, 136, 138, 140, 141, 142, 145, competition and selection, xiii, 9, 13, 14, 17,
159, 169, 173, 174, 185, 192, 194, 195, 196, 59, 61, 98, 116, 117, 120, 125, 132, 133, 134,
199, 200, 210, 211, 216, 249, 294, 304, 305, 135, 136, 137, 141, 142, 143, 144, 145, 153,
306, 307, 308, 313, 314, 316, 317, 324, 325, 154, 166, 168, 170, 186, 192, 210, 241, 254,
326, 327, 328, 329, 330, 331, 332, 333, 335, 259, 264, 271, 294, 304, 312, 315,
336 complementizer system, 11, 14, 144, 170, 196,
Adia/Adja. See Aja 197, 222, 226, 229, 232, 238, 239, 240, 249,
Agni, 58 252, 253, 308
Aja, 47 complexity, xii, 141, 169, 305, 306, 307, 311,
Aja descent, 36, 38, 39, 40, 41, 47, 52, 58, 59, 312, 315, 319, 326, 329, 330
60, 114, 117, 119, 123, 304 compound, 57, 108, 154, 155, 156, 157
Aja Nengre, 52 comp-system, 11, 14, 233, 269
Aja-Tado, 20, 34, 36, 39, 331 Congo, vi, 6, 8, 55, 56, 57, 58, 257, 309,
Allada, vii, x, 16, 17, 19, 20, 21, 22, 23, 24, 26, 318, 326
27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, Congos, 55, 56
39, 48, 51, 53, 54, 58, 304, 319, 328 congruence, xiii, 51, 63, 66, 135, 156, 160,
Arada. See Allada 211, 303
Ashanti, 58 contextual inflection, 101, 102, 104, 109
contrastive focus, 172, 173, 250, 251
Badagri, 16, 22 Costa Peixoto, 122, 123, 124, 126
Bantu, 39, 54 creoles, xi, 55, 56
Baoule, 58
bare nouns, 2, 10, 11, 43, 77, 87, 88, 171, 178, Dahomey, 22, 35
179, 180, 181, 187, 188, 189, 206, 212, 310, Danish African Company, 29
322, 324 definite determiner, 66, 68, 71, 77, 81, 85, 181,
Basic Variety, 107, 327 182, 209, 210
Beni, 20, 24, 26, 27, 32 definiteness, 9, 10, 12, 14, 63, 67, 68, 71, 80,
Bight of Benin, 19, 33, 38, 47, 58, 304 87, 88, 142, 172, 176, 177, 190, 212, 216,
bilingualism, ii, 5, 173, 333 310, 311, 326, 327, 331
bozals, 18 deontic modality, 14, 224, 225, 226, 227, 229,
Brazil, 28, 29, 122, 124, 135 231, 232, 234, 235, 236, 237, 240, 247, 268
break in transmission, 7, 83, 174 derivational morphology, 12, 85, 194, 195
Bulfinch Lamb, 33, 35 determiners, xiv, 2, 14, 47, 71, 72, 77, 79, 108,
176, 177, 179, 180, 181, 182, 186, 188, 189,
Central African languages, 17 190, 191, 192, 207, 209, 216, 217, 261, 311,
Chevalier Marchais, 31 321, 325, 326, 329
clausal left periphery, 14, 110, 185, 226, 249, DNA, 139, 141
251, 252, 253, 257, 265, 268 Dutch, 21, 23, 24, 27, 28, 30, 33, 34, 37
clause union. See restructuring Dutch West India Company, 28, 34
343
Downloaded from https://www.cambridge.org/core. University of Texas Libraries, on 15 Feb 2020 at 07:19:01, subject to the Cambridge Core
344 Subject index
ecological approach, 8, 294 Ibo, 55

E-creole, 3, 99, 135, 138, 142, 186 I-creole, 3, 110, 135, 138, 139, 141, 145
Edo, 47 idiolect, xiii, 115, 116, 135, 139, 303
E-language, 3, 5, 115, 116, 117, 137, 142 Ife, 20
event relativization, 259, 260, 261, 262 Igbo, 58
Ewegbe, 58 I-language, 3, 4, 5, 6, 8, 9, 11, 99, 110, 136, 141,
exceptional SLA, 62 142, 143, 169, 192, 201
indefinite determiner, 80, 85, 209, 211
factive clause, 229, 233, 240, 254, 258, 259, indefinite marker, 79, 80, 210, 212, 213,
261, 262, 263, 264, 265, 317, 323 inflectional morphology, 4, 12, 43, 85, 96, 101,
failure of acquisition, 61 104, 108, 109, 170, 193, 194, 195, 272, 274,
Fanti, 58 281, 282, 311, 312, 331
Father Jose de Najara, 21 inherent complement verbs, 145
feature pool, xiii, 5, 114, 115, 116, 123, 132, inherent inflection, 95, 101
135, 137, 153, 154, 166, 175, 294, 314, 315, interface, 12, 139, 145, 154, 161, 169, 170, 172,
329 173, 175, 185, 191, 193, 194, 195, 199, 200,
feature transmission, 9, 172, 173, 192 203, 206, 209, 210, 211, 212, 213, 216, 221,
focalization, 68, 185 222, 223, 265, 267, 305, 315, 317, 330, 335
focus constructions, 46, 47, 66, 223, 239, 249, interlanguages hypothesis, xi, 8, 61, 93, 96, 97,
250, 251, 254, 256, 259, 262, 264, 265 98, 99, 100, 107, 108, 109, 121, 331,
focus marker, 49, 51, 53, 144, 250, 251, 253, irrealis, 14, 86, 87, 225, 227, 228, 229, 231,
254, 255, 256, 260, 266, 268, 307, 308 233, 234, 236, 240, 242, 244, 245, 246, 247,
Founder Principle, 55, 114, 117 248, 266, 268, 272
Francisco Felix de Sousa, 31
François d’Elbée, 23, 24, 29, 31, 36, 37 Jamaica, 19, 129, 329
French, 24, 29, 50 James Duke of York, 29
French Caribbean, vi, 54, 67 Jekin, 22, 24
French colonies, 55
French West Indies Companies, 29 King of Allada, 21, 23, 24, 27, 32, 37, 48
functional verb, 285, 288, 290, 291, 293, 294, King of Beni, 20
298, 299 King of France Louis XIV, 24, 29
King of Popo, 21
G. M. Branco, 26, 27 King of Spain Philippe IV, 24
Gbe pattern, 184, 209, 211, 213, 221, 253 koiné, 7, 18, 63, 111, 113, 114, 117, 118, 119,
Gbe speakers, 38, 59, 80, 82, 113, 119, 220, 304 120, 121, 132, 133, 134, 135, 171
Gbe-type, 82, 144, 212
Gengbe. See Mina La Concorde, 24, 29
genotype, 138, 139, 141 La Doctrina Christina, 24, 48, 50
geopolitics, 16, 17 La Justice, 29
Ghana, 16, 20, 26, 41, 53 language as biological species, 133, 138, 141,
globalization, 19 143, 315
Gold Coast, 19, 33, 36, 37, 38, 55, 56 Language Bioprogram, x, xi, 7, 83, 85, 87, 88,
Grain Coast, 19 90, 91, 92,
grammaire abrégée, 50 lingua franca, xi, 18, 31, 84, 131
Great Popo, 16, 21, 22 linguistic hybridism, 305, 312
Gungbe, 58 linguistic hybridization, 5
local complexity, 15, 169, 302, 305, 307, 309
Haiti, 6, 13, 16, 17, 27, 39, 49, 51, 52, 54, 58, local economy, 26, 27, 29
59, 60, 77, 78, 79, 82, 114, 116, 117, 119, Long Edward, 129
124, 127, 304 low restructuring, 294, 300
high restructuring, 294, 296, 302, 303 Lukumi, 34
homestead phase, xi, 83, 84, 121
homesteads, 55 Mattheo Lopes, 24
hybrid grammar, i, v, xii, xiv, 271, 300, Mattheo Lopez, 29, 31
304, 305 maxi, 58
Subject index 345
Mina, 58 restructuring, xiii, 14, 63, 73, 145, 220,

Mina Jeje, 122 271, 291, 293, 294, 295, 296, 298, 300, 302,
Mines. See Mina, Gengbe 303, 322, 336
See Gengbe, 55
Mono River, 20 Saint-Domingue, vi, 13, 19, 55, 57, 325
São Tomé, 26, 27
Nago, 40, 47, 55 seasoning, 65, 125
Napoléon Bonaparte, 128, 129 Sefwi, 58
Nigeria, 16, 20, 22, 41, 314 Semoisi, 52
nominal left periphery, 176, 177, 185, 189, serial verb construction, v, xiv, 11, 12, 14, 15,
205, 222 45, 47, 50, 90, 170, 248, 269, 320, 323,
number marker, 66, 67, 68, 72, 76, 79, 85, 177, 326, 328
179, 181, 185, 187, 188, 190, 207, 208, 209, simplicity, xii, 169, 170, 273, 305, 306, 307,
219, 308, 309, 310, 311, 312, 329
Nzima, 58 Slave Coast, v, vii, 13, 16, 17, 18, 19, 22, 26, 28,
33, 35, 36, 37, 38, 40, 47, 49, 54, 55, 56, 58,
object sharing, 281, 282, 283, 284, 60, 119, 122, 304, 328
Olfert Dapper, 21, 23, 26, 34 slave economy, 26, 27, 38, 131
OV constructions, 45, 290 Spanish monks, 48
OVV constructions, 163, 164, 166, 167, specific definite, 10, 87, 88, 176, 190, 191,
168, 170 209, 210, 211, 213
Oyo, 20, 22, 32, 34, 35, 37, 40 specific indefinite, 10, 176, 189, 190, 209,
210, 211, 212
Paramaribo, 52, 333, 336 specificity, 10, 11, 14, 80, 142, 172, 173, 175,
parameter-setting, 9, 136, 142 176, 177, 178, 179, 180, 181, 182, 183, 184,
pattern transmission, 9, 172, 173, 175, 185, 185, 186, 187, 188, 190, 191, 192, 206, 207,
208, 208, 209, 211, 212, 213, 216, 217, 218, 219,
Petit-nègre, 126 221, 310, 325
plantation phase, xi, 18, 59, 83, 84, 121, 304 specificity marker, 10, 80, 173, 178, 179, 181,
plantation settlement, xi, 83, 111 183, 184, 185, 187, 190, 206, 207, 208, 211,
plantation settlement colonies, 52 212, 216
plantation society, 121, 134 substrate influence, ix, xv, 51, 62, 107, 110,
plantation varieties, 120 134, 135, 174, 257, 270, 271, 273, 286, 304,
Polynesians, 111, 113 322
Portuguese, ii, xi, 20, 26, 27, 28, 31, 39, 65, 79, substrate transfer, 63, 65, 132, 134, 171, 192,
100, 109, 124, 134, 175, 192, 212, 221, 250, 281, 292, 296, 303
270, 322, 323, 334 superstrate influence, 62, 66, 302
possessive constructions, 184, 213, 214 Suriname, 6, 13, 16, 17, 27, 33, 39, 51, 52, 53,
predicate clefts, 253, 326, 335 59, 60, 114, 116, 117, 119, 121, 124, 127,
132, 137, 138, 145, 154, 156, 158, 166, 167,
Queen of Allada, 36 168, 169, 171, 217, 220, 235, 270, 304, 320,
325, 327, 328, 329, 332, 333
recombination (of feature), xiii, 9, 14, 15, 52, syntactic recombination, 14, 51, 137, 145, 193,
136, 138, 139, 141, 142, 143, 144, 145, 153, 226, 265, 269, 273
154, 169, 171, 186, 191, 192, 194, 195, 200,
208, 210, 221, 240, 242, 249, 265, 267, 268, Tado, 20, 21, 34, 36, 119
295, 296, 300, 302, 303, 304, 305, 316 target language, i, xiii, 6, 7, 8, 12, 18, 51, 52, 57,
reduplication, 42, 161, 162, 163, 165, 166, 167, 60, 61, 63, 74, 75, 78, 79, 83, 93, 97, 99, 100,
168, 317, 326 107, 108, 110, 130, 131, 135, 194, 313
relative marker, 218, 219, 260, 264, 309 TMA, 46, 47, 85, 90, 98, 108, 109, 142, 228,
relativizer, 104, 105, 106, 107, 218, 219, 239, 240, 272, 273, 275, 308, 321, 329, 336
259, 260 Togo, 16, 20, 26, 41
relexification, 7, 175, 270, 324, 334 topic constructions, 67, 68, 70, 250, 335
Republic of Togo, 20 topic marker, 45, 208, 250
Repuplic of Benin, 20 topicalization, 68, 96, 185, 244
346 Subject index
transatlantic slave trade, 18 verb raising with doubling, 66

transfer, 4, 7, 12, 15, 58, 63, 132, 153, 172, 173, V-final, 1, 2, 197, 198, 199
174, 175, 191, 192, 193, 194, 195, 196, 199, Vòdú, 40
200, 201, 203, 205, 209, 211, 221, 279, 294, vulnerable interfaces, v, 195, 204
295, 296, 300, 317, 333
translators, 24 W. Bosman, 34, 35
triangular trade, 18, 19, 58 weak competitor, 195, 216
West Africa, 16, 19, 24, 27, 38, 39, 41, 46, 47,
UG, 4, 6, 9, 12, 16, 83, 85, 99, 165, 171, 172, 51, 97, 100, 269, 327, 328
185, 192, 193, 221, 253, 265, 268, 271, 291, wh-movement, 307, 308, 309, 322
294, 295, 297, 324 wh-questions, 47, 49, 50, 96, 206, 238, 250,
309, 310
V2 properties, 1, 2, 3 Whydah, 16, 21, 22, 27, 29, 30, 32, 34,
Van Hoolwerf, 28, 30 35, 50
verb focus, 45, 65, 66, 257, 258, 259, world economy, 29, 59, 304
260
verb focus constructions, 45, 65, 260 Xweɖa, 58
verb movement, 4, 139, 143, 198, 281, 282,
317, 323, 327 yes–no question, 143, 251, 252
verb raising, 166, 280, 281 Yoruba, 20, 34, 38, 40, 58, 119, 255

(Cambridge Approaches to Language Contact) Enoch Oladé Aboh - The Emergence of Hybrid Grammars_ Language Contact and Change (Cambridge Approaches to Language Contact)-Cambridge University Press (2015)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Cambridge Approaches to Language Contact) Enoch Oladé Aboh - The Emergence of Hybrid Grammars_ Language Contact and Change (Cambridge Approaches to Language Contact)-Cambridge University Press (2015)

Uploaded by

Copyright:

Available Formats

The Emergence of Hybrid Grammars

Children are extremely gifted in acquiring their native languages, but

enoch o. aboh is Professor of Linguistics at the University of Amsterdam.

Cambridge Approaches to Language Contact is an interdisciplinary series

Further titles planned for the series

Enoch Oladé Aboh

Cambridge University Press is part of the University of Cambridge.

List of ﬁgures page vi

2.1 Portrait of Dom Matheolopes, Ambassador Roy d’Arda.

2.1 The Slave Coast and its major ports page 17

The same comparative approach, once promoted earlier in the 1990s by

assumption that sporadic encounters between European traders and societally

language development in individual speakers and on the convergence of

a:dcl afﬁrmative declarative

mood modality marker

An obvious empirical fact about languages is that they evolve constantly in

syntax has gradually moved away from a proto-typical West Germanic

DeGraff (1999: 9) characterizes the time-lapse between the birth or “creation”

Applied to language change in general, this would mean that studies of

of CP-V2 in the Northern dialects is contingent to the loss of verb movement to

linguistic system that emerges from the contact of (typologically and/or

1.1 Creoles as a test-bed

the contact between languages that are genetically and typologically

1.2 Speakers create variants not imperfect replicas

work on the competition of grammars (e.g., Kroch 1989a, 1989b, 2001;

though the semantics of such marked expressions is very similar to their

(the interface between sentence and discourse) assume that it is a vulnerable

languages by providing a detailed morphosyntactic analysis that addresses the

1.3 The structure of the book

This chapter addresses Alleyne’s concern. The discussion primarily focuses on

Chapters 5, 6, and 7 put to test the conclusions reached in Chapter 4 on

I mentioned in the Introduction (Chapter 1) that this book adopts a

Map 2.1 The Slave Coast and its major ports

2.1 Geopolitics on the Slave Coast: the case of Allada

to the European target language. As I argued in Chapter 3, a growing number of

Though studies on the Atlantic ‘triangular trade’ have demonstrated the

Origins of Allada According to Pazzi (1979), Adande (1984), and Law

Politics and institutions It is difﬁcult to determine exactly when Allada was

GHANA TOGO Rep. of BENIN NIGERIA

Map 2.3 Approximate area of control of the Allada Kingdom

The capital of this kingdom referred to as “Great Allada” situated in Togudo in

Figure 2.1 Portrait of Dom Matheolopes, Ambassador Roy d’Arda. Prints on

kingdom consisted of a central nucleus, Allada, where the ‘King of Kings’

Economy Allada was a rich kingdom with a booming economy integrated in

as well as independent traders and notables of both African and European

Table 2.1 Slave import from the Bight of Benin

SC % 39.4 42.8 70.6 69.0 28.1 31.8 2.2 0 0.9 0 0 0

SC = Slave Coast, GC = Gold Coast (Source: Arends 1995b: 243).

2.1.3 Supply of slaves and the slaves’ origins

hand, it is highly probable that signiﬁcant numbers of people of Aja descent

Monday Àjó/Ajówá Kwàjó, Kɔ̀ jó

Kluge (2005), who conducted a lexical comparative analysis of 49 Gbe

Table 2.2 A random wordlist in ﬁfteen Gbe varieties selected according to

Clusters Varieties Eye Three Red Eat Drink Push Pull

EWE Adan ŋku etɔ dzi ɖu no tutu he

Gbe languages generally display the order of noun phrases in (1a)

a. ‘We bought a chicken.’

Table 2.3 General word patterns in Gbe

Word order patterns in Gbe

Noun phrases a. Noun>adjective>numeral>demonstrative>relative

languages. Such learning hypotheses could have been enhanced and

This document displays a better transcription and provides us with signiﬁcant

Example (17) illustrates a wh-question. Here again we see the combination

2.2 Plantation settlement colonies: Suriname and Haiti