Creole genesis and the acquisition of grammar

The case of Haitian creole

This study focuses on the cognitive processes involved in creole genesis
relexification, reanalysis and dialect levelling processes which the author
demonstrates play a significant role in language genesis and change in general.
Dr Lefebvre argues that the creators of pidgins/creoles use the parametric values
of their native languages in establishing those of the language that they are creating and the semantic principles of their own grammar in concatenating morphemes and words in the new language. The theory is documented on the basis
of a uniquely detailed comparison of Haitian creole with its contributing French
and West African languages. Summarising more than twenty years of funded
research, the author examines the input of adult, as opposed to child, speakers and
resolves the problems in the three main approaches, universalist, superstratist and
substratist, which have been central to the recent debate on creole development.
Claire Lefebvre is professor of linguistics at the Universit du Qubec Montral.

CAMBRIDGE STUDIES IN LINGUISTICS

General Editors: s. r. anderson, j. bresnan, b. comrie,
w. dressler, c. ewen, r. huddleston, r. lass,
d. lightfoot, j. lyons, p. h. matthews, r. posner,
s. romaine, n. v. smith, n. vincent

In this series
52 michael s. rochemont and peter w. culicover: English focus constructions
and the theory of grammar
53 philip carr: Linguistic realities: an autonomist metatheory for the generative
enterprise
54 eve sweetser: From etymology to pragmatics: metaphorical and cultural aspects
of semantic structure
55 regina blass: Relevance relations in discourse: a study with special reference
to Sissala
56 andrew chesterman: On definiteness: a study with special reference to English
and Finnish
57 alessandra giorgio and giuseppi longobardi: The syntax of noun phrases:
configuration, parameters and empty categories
58 monik charette: Conditions on phonological government
59 m. h. klaiman: Grammatical voice
60 sarah m. b. fagan: The syntax and semantics of middle construction: a study with
special reference to German
61 anjum p. saleemi: Universal Grammar and language learnability
62 stephen r. anderson: A-Morphous morphology
63 lesley stirling: Switch reference and discourse representation
64 henk j. verkuyl: A theory of aspectuality: the interaction between temporal and
atemporal structure
65 eve v. clark: The lexicon in acquisition
66 anthony r. warner: English auxiliaries: structure and history
67 p. h. matthews: Grammatical theory in the United States from Bloomfield to
Chomsky
68 ljiljana progovac: Negative and positive polarity: a binding approach
69 r. m. w. dixon: Ergativity
70 yan huang: The syntax and pragmatics of anaphora
71 knud lambrecht: Information structure and sentence form: topic, focus, and the
mental representations of discourse referents
72 luigi burzio: Principles of English stress
73 john a. hawkins: A performance theory of order and constituency
74 alice c. harris and lyle campbell: Historical syntax in cross-linguistic
perspective
75 liliane haegeman: The syntax of negation
76 paul gorrell: Syntax and parsing
77 guglielmo cinque: Italian syntax and Universal Grammar
78 henry smith: Restrictiveness in case theory
79 d. robert ladd: Intonational phonology

80 andrea moro: The raising of predicates: predicative noun phrases and the theory
of clause structure
81 roger lass: Historical linguistics and language change
82 john m. anderson: A notional theory of syntactic categories
83 bernd heine: Possession: cognitive sources, forces and grammaticalization
84 nomt erteschik-shir: The dynamics of focus structure
85 john coleman: Phonological representations: their names, forms and powers
86 christina y. bethin: Slavic prosody: language change and phonological theory
87 barbara dancygier: Conditionals and prediction
88 claire lefebvre: Creole genesis and the acquisition of grammar: the case of
Haitian creole
Supplementary volumes
liliane haegeman: Theory and description in generative syntax: a case study
in West Flemish
a. e. backhouse: The lexical field of taste: a semantic study of Japanese taste
terms
nikolaus ritt: Quantity adjustment: vowel lengthening and shortening in early
Middle English
Earlier issues not listed are also available

CREOLE GENESIS AND THE

ACQUISITION OF GRAMMAR
THE CASE OF HAITIAN CREOLE

CLAIRE LEFEBVRE

Cambridge
university press

CAMBRIDGE UNIVERSITY PRESS

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
Cambridge University Press
The Edinburgh Building, Cambridge CB2 2RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521593823
Claire Lefebvre 1998
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 1998
This digitally printed first paperback version 2006
A catalogue record for this publication is available from the British Library
Library of Congress Cataloguing in Publication data
Lefebvre, Claire.
Creole genesis and the acquisition of grammar: the case of
Haitian creole / Claire Lefebvre.
p. cm. (Cambridge Studies in Linguistics; 88)
Includes bibliographical references and index.
ISBN 0 521 59382 4 (hardback)
1. Creole dialects. 2. Pidgin languages. 3. Psycholinguistics.
4. Second language acquisition. 5. Creole dialects, French Haiti
Grammar. I. Title. II. Series.
PM7854.H3L44 1998
417.22 dc21
9812930 CIP
ISBN-13 978-0-521-59382-3 hardback
ISBN-10 0-521-59382-4 hardback
ISBN-13 978-0-521-02538-6 paperback
ISBN-10 0-521-02538-9 paperback

To Ken Hale and to Mathieu,

both magicians in their own way

To an extent unparalleled in the study of languages anywhere else in

the world, African language classification has been beset by persistent
hypotheses of language mixture, intermediate or transitional languages,
substrata, pervasive external influence far in excess of what is usually
recognised as normal, and innovative exuberance unmatched in recorded
language history. Perhaps the most dramatic and preposterous example
of speculation in linguistic theory is provided by Sir Harry Johnston
(1919 p. 27): A great jumble of events, and lo! new languages spring
suddenly into existence. (Welmers 1973: 2)
Sapere aude!
[Dare to think by yourself.]
Immanuel Kant

Contents

Tables
Preface
Abbreviations

page xiii
xv
xviii

1
1.1
1.2
1.3
1.4
1.5

The problem of creole genesis and linguistic theory

The complex problem of pidgin and creole genesis
The perspective of this book
Linguistic theory
The hypothesis
The scope and limitations of this book

1
1
5
7
9
13

2
2.1
2.2
2.3
2.4

Cognitive processes involved in creole genesis

The mental process of relexification
The role of relexification in the genesis of mixed languages
The role of relexification in the genesis of pidgin languages
The relationship between the notion of transfer and the
process of relexification in creole genesis and second
language acquisition
Relexification and second language acquisition in creole
genesis
Reanalysis
Dialect levelling
Parameter setting, semantic interpretation and principles of
concatenation
An optimal account of creole genesis
Conclusion

15
16
19
30

The research methodology

The economy, demography and linguistic diversity of early
Haiti: 16591740
The typological features of the source languages of Haitian
The superstratum data the creators of Haitian were exposed to
The linguistic test
What counts as evidence for the hypothesis and how can it
be falsified?

52

2.5
2.6
2.7
2.8
2.9
2.10
3
3.1
3.2
3.3
3.4
3.5

33
35
41
46
47
47
49

52
58
62
65
70
ix

Contents

3.6
3.7

The data
Mode of presentation of the data and analyses

Functional category lexical entries involved in nominal

structure
The [+definite] determiner
The plural marker
The so-called indefinite determiner
The [+deictic] terms
Case markers within the noun phrase
Conclusion

4.1
4.2
4.3
4.4
4.5
4.6
5

5.3
5.4
5.5

The preverbal markers encoding relative Tense, Mood

and Aspect
Overview of the tma systems of Haitian, French and Fongbe
The historical derivation of the Haitian tense, mood and aspect
markers
The temporal interpretation of bare sentences
Dialect levelling
Conclusion

6
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8

Pronouns
Personal pronouns
Possessives
Logophoric pronouns
Pronominal clitics
Expletives
Reflexives
Wh-phrases and Wh-words
Conclusion

Functional category lexical entries involved in the structure

of the clause
Complementisers and complementiser-like forms
Complementisers or resumptives in the context of extracted
subjects?
The nominal operator in relative and factive clauses
Clausal conjunction
The mystery of Haitian se
Negation markers
Yesno question markers
Markers expressing the speakers point of view
Conclusion

5.1
5.2

7.1
7.2
7.3
7.4
7.5
7.6
7.7
7.8
7.9

75
76

78
79
84
87
89
101
110

111
111
115
134
137
139
141
141
143
147
148
157
159
171
182

184
184
193
203
205
206
208
211
213
217

Contents
8
8.1
8.2

xi

8.4
8.5

The determiner and the structure of the clause

The functions of the determiner in clause structure
The clausal determiner with the function of an assertive
marker
The clausal determiner with the function of an event
determiner
The *Det Det Filter in grammar 1 and grammar 2
Conclusion

229
240
246

9
9.1
9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9
9.10
9.11
9.12
9.13

The syntactic properties of verbs

Types of argument structures
body-state verbs
weather verbs
Reflexive verbs
Verbs licensing expletive subjects
Raising verbs
Existential verbs
Control verbs
Light verbs
Inherent object verbs
The Case-assigning properties of verbs
Double-object verbs
Conclusion

248
248
250
251
253
259
262
269
271
278
280
283
287
301

10
10.1
10.2

303
303

10.4
10.5

Are derivational affixes relexified?

Identifying the derivational affixes of Haitian
The derivational affixes of Haitian compared with those
of French
The derivational affixes of Haitian compared with those
of Fongbe
The historical derivation of the Haitian derivational affixes
Conclusion

11
11.1
11.2
11.3
11.4

The concatenation of words into compounds

The semantics of compounds
Establishing word order in compounds
Types of compounds
Conclusion

334
334
339
342
348

12
12.1
12.2
12.3
12.4

Parameters
Is Haitian a null subject language?
Verb raising
Serial verbs
The double-object construction

349
349
351
355
357

8.3

10.3

219
219
221

312
318
323
333

xii

Contents

12.5
12.6
12.7

The interpretation of negative quantifiers

Verb-doubling phenomena in particular grammars
Conclusion

360
363
374

13
13.1
13.2
13.3
13.4
13.5
13.6

Evaluation of the hypothesis

The lexicon
The interpretive component
Parameters
Word order
Further questions
Overall evaluation of the hypothesis

375
375
386
386
388
390
394

14

Theoretical consequences

395

Appendices
Appendix 1 List of available Haitian creole texts (17761936)
397
Appendix 2 Phonemic inventories and orthographic conventions 398
Appendix 3 Sample of non-matching derived words in Haitian
and French
403
Notes
References
Index of authors
Index of languages and language families
Index of subjects

408
424
452
457
459

Tables

3.1
3.2

3.3
3.4
3.5

11.1
13.1
13.2
13.3
13.4
13.5
13.6
13.7

Comparison of Haiti after 22 years (1681) and Martinique

after 25 (1660)
page 53
People of colour as a proportion of the colonys total
population. Comparison of Haiti after 2262 years
(16811721) with Martinique after 2565 years (16601700)
54
Distribution by branch of NigerCongo
55
Breakdown of the Kwa population in the two censuses and
the Remire list
55
The growth in the proportion of the population of colour in
Haiti juxtaposed with the percentage of speakers of Gbe
dialects among the African population in the French Caribbean
and the African slave-export population
56
Types of compounds in Haitian, Fongbe and French
343
Minor category lexical entries involved in noun structure
376
Tense, Mood and Aspect markers
376
Pronominal forms
377
Minor category lexical entries involved in the structure
of the clause
377
Derivational affixes
381
Syntactic properties of verbs
382
Comparison of the parametric options in the three languages
under comparison
387

xiii

HH

Preface

This book focusses on the cognitive processes involved in creole genesis:

relexification, reanalysis, dialect levelling and parameter setting. The role of
these processes in creole genesis is documented on the basis of a detailed comparison of Haitian creole with two of its major source languages: French, its
main lexifier language, and Fongbe, one of its West African substratum languages. The findings reported on in this book are based on twenty years of
research that I have done on the languages involved, alone or in collaboration
with colleagues and students, through various projects carried out at the Universit
du Qubec Montral.
From 1976 to 1980, with Lynn Drapeau, I conducted a project on popular
French (financed by fcar, r-uqam and olf). In parallel (19757), I started a
small project on the syntax of Haitian creole (financed by r-uqam) with some
Haitian students and Hilda Koopman, who was then a visiting student in our
department, writing an M.A. thesis on Haitian. This project led to the publication with Karoma Press of Syntaxe de lhatien (1982), written in collaboration
with Hilda Koopman, Hlne Magloire-Holly and Nanie Piou. Building on the
results of this project, I began comparing the lexicon and grammar of Haitian
creole with the lexicons and grammars of two of its source languages: French
and Fongbe. This work was done in the framework of the project Le crole
hatien: langues africaines relexifies? that I directed together with Jonathan
Kaye from 1985 to 1989 (financed by fcar). This project also involved the
participation of several students: Anne-Marie Brousseau, Jolle Brillon, Rjean
Canac-Marquis, Rose-Marie Dchaine, Sandra Filipovich and Jean-Robert Cadely.
The research continued with the project The morphology and the syntax of
Haitian from 1986 to 1988 (financed by sshrcc), in which Diane Massam and,
later, John Lumsden participated as professional researchers; some of the graduate students from the first project also took part, as well as some new students
who joined the team such as Rollande Gilles, Marie-Denyse Sterlin and Danielle
Dumais. The results of these projects enabled me to formulate a long-term
project to test the hypothesis that the mental processes of relexification and
reanalysis play a major role in creole genesis.
I obtained a major grant from sshrcc (and complementary grants from
r-uqam) for a project designed to test this hypothesis in detail, based on the
case study of Haitian creole. This project, La gense du crole hatien: un cas
particulier dinvestigation sur la forme de la grammaire universelle, which I
xv

xvi

Preface

directed with the collaboration of John Lumsden, lasted five years (198994). It
focussed significant professional and material resources on the lexicon and grammar of Haitian creole and its source languages. Besides the two major researchers,
the team included several other researchers: Elizabeth Ritter, Paul Law, Kinyalolo
Kasangati, Alain Kihm (from the CNRS, France), John Singler (from New York
University) and Anne-Marie Brousseau; research technicians Danielle Dumais,
Monique Poulin, Andre Blanger and Anne-Marie Benoit; our secretary
Lorraine Rainville; several graduate students, most of whom are native speakers
of one of the languages under study: Anne-Marie Brousseau, Aim Avolonto,
Maxime Da Cruz, Joseph Sauveur Joseph, Hrold Mimy, Juvnal Ndayiragije,
Michel Platt and Efoe Wallace; an autonomous researcher, France Martineau;
visiting graduate students: Tonjes Veenstra (University of Amsterdam) and Chris
Collins (mit); a series of visiting scholars: Elizabeth Cowper (University of
Toronto), Kenneth Hale (mit), Richard Larson (mit) and Gillian Sankoff (University of Pennsylvania); and various collaborators: Albert Bienvenu Akoha
(Centre bninois des langues trangres), Hounkpati Capo (Labo Gb, Universit
nationale du Bnin), Pierre Vernet (Universit dtat dHati) and Marc Laurent
Hazoum (cenala, Universit nationale du Bnin). Finally, some thirty native
speakers of Haitian and Fongbe were involved with the project as informants.
Complementary grants in collaboration with John Lumsden entitled Les proprits
lexicales, leur reprsentation dans le lexique et leur projection dans la syntaxe
(fcar 19902) and Lorganisation des lexiques et des entres lexicales (fcar
19935) also contributed to the realisation of this research. To the best of my
knowledge, this is the first time that sufficient resources have been gathered
together to make a detailed and extensive comparison of the grammar and lexicon
of a creole language with those of its superstratum and substratum sources. Furthermore, this project has included original research to document the historical
situation at the time of the creation of Haitian creole. Finally, it should be noted
that recent developments in linguistic theory (e.g. in the theory of parametric
variation, functional category theory, lexical semantic theory, the results of the
mit Lexicon Project, etc.) have provided us with precise tools for the comparative analyses.
The aim of this book is threefold. First, I present the theory of creole genesis
formulated around the major processes involved and the methodology developed
for testing it. Second, I present an extensive comparison of the properties of the
lexicon and the syntax of Haitian with those of its contributing languages. Third,
I evaluate the hypothesis on the basis of the data presented in this book. The
data and analyses presented in this book draw not only on my own work on the
languages involved, but also on data and analyses available in the literature and
research produced by the members of the various teams involved in the projects
mentioned above. Putting it all together in this book and filling the holes required
a lot of additional work. Danielle Dumais assisted me in compiling the enormous
amount of Haitian and Fongbe data. Olivier Tardif assisted me in compiling
seventeenth- and eighteenth-century French data. Andre Blanger formatted the

Preface

xvii

manuscript and Zofia Laubitz copy-edited it. This work was financed by a grant
from sshrcc (19947). Finally, work referred to as Brousseau (1995a and b)
was financed in part by a grant to John Lumsden and in part by an sshrcc grant
to me. The usual disclaimer is in order here: none of the people involved in the
above-mentioned projects are to be held responsible for the views advocated in
this book, nor for the particular analyses I propose to account for the data. This
book was written between April 1995 and February 1997.
I would like to thank my university and my department for supporting this
research and the funding agencies for making it possible. Patricia Dunn and
Elaine Isabel from sshrcc gave me invaluable support to begin and complete
this research. I would like to thank my collaborators mentioned above for their
work, for fruitful discussions and for the good times that we had together. I am
grateful to my colleagues of the local and international community for their numerous questions, comments and even objections related to this research; their reactions contributed to making the claims and analyses more precise. Many thanks
to Anne-Marie Brousseau, Mark Durie, Ken Hale, Rich Larson, John Lumsden
and Lisa Travis for most insightful discussions on several theoretical issues
raised in the course of the research. Special thanks to Jean-Robert Placide for
sharing with me his knowledge of the Haitian lexicon and grammar over all
these years. I am indebted to the following people for their comments on drafts
which became part of some of the chapters in this book: Marthe Faribault, YvesCharles Morin, Elizabeth Ritter, Pierrette Thibault, Lydia White and Raffaella
Zanuttini.
Julie Auger, Anne-Marie Brousseau, Bernard Comrie, Christine Jourdan,
Lisa Travis and an anonymous reader read a first draft of this manuscript; their
questions and comments contributed a great deal to its final form. I owe special
thanks to Andre Blanger, Anne-Marie Brousseau and Danielle Dumais for
their friendship, support and encouragement in the final phase of this research.
Bernard Comrie gave me invaluable support during the time I was working on
this manuscript. I do not know how to thank Ken Hale for supporting this research
throughout. I would also like to express my gratitude to Gillian Sankoff and Paul
Kay, who taught me how to work. Finally, last but not least, I would like to
thank my friends, my parents, my family and my son Mathieu. They were there
all the time.

Abbreviations
acc
adv
ag
agent
an
ant
ap
asp
ass
att
aux
body
case
cl
comp
de+les
def-fut
deic
dem
Det
det
dm
ec
emph
fc
foc
fp
fem
fut
fut-possib
gen
hab
imp
inan
ind-fut
xviii

accusative
adverb
agentive affix
agent
animate
anterior
plural article
aspect
assertive
attributive affix
auxiliary
body-part reflexives
Case
clitic
complementiser
partitive+plural
determiner
definite future
deictic
demonstrative
determiner (clausal)
determiner (head of DP)
discourse marker
Eastern Cushitic
emphatic
functional category
focus
focal pronoun
feminine
future
future-possibility
genitive
habitual
imperfective
inanimate
indefinite future

in
ins
int
irr
lo(c)
log
mme
mo
Neg
neg
nom
Num
obv
op
part
p(a)st
pl
poss
post
prep
pro
prog
prosp
prox
q
ref/r
rel
res
sc
self
sem
srp
sub
surp
to
val

inflection
markers of insistence
interrogative
irrealis
locative
logophoric pronoun
French emphatic mme
mood
negative marker
negation marker
nominaliser
number
obviative
operator
partitive
past
plural
possessive
postposition
preposition
covert pronominal
progressive
prospective
proximate
question marker
reflexive
relative marker
resumptive
Southern Cushitic
-self anaphor
semantic properties
subject-referring pronoun
subjunctive
surprise
topic marker
validator

The problem of creole genesis

and linguistic theory

This book addresses the cognitive processes hypothesised to account for the
properties of creole languages. It presents a theory of creole genesis based on
processes otherwise observed to play a role in language genesis and in language
change in general. It is intended as a contribution to our understanding of the
mechanisms by which the properties of the source languages of a creole manifest themselves in the creole in the way they do. This chapter summarises the
salient features of creole languages which any theory of creole genesis must
be able to account for and situates this book with respect to other approaches
to the problem. Section 1.3 introduces the theoretical framework within which
these problems are addressed. Section 1.4 presents the hypothesis underlying the
research. Section 1.5 discusses the scope and limitations of this book.
1.1

The complex problem of pidgin and creole genesis

The history and structure of pidgin and creole languages are characterised by
the following features.1 First, as was pointed out by Whinnom (1971), these languages are only developed in multilingual communities. Whinnom argues that,
in bilingual communities, the speakers of one group will eventually learn the
language of the other group.2
Second, communities where pidgin and creole languages emerge generally
involve several substratum languages spoken by the majority of the population
and a superstratum language spoken by a relatively small but economically powerful social group. Crucially, the substratum community does not have one common
language. This situation creates the need for a lingua franca (see e.g. Hymes
1971a; Foley 1988), not only to permit communication between the speakers of
the substratum languages and of the superstratum language, but also to permit
the speakers of the substratum languages to communicate among themselves
(see e.g. Foley 1988; Singler 1988: 47; Thomason and Kaufman 1991).
Third, in communities where creole languages emerge, speakers of the substratum languages generally have very little access to the superstratum language
(see Thomason and Kaufman 1991). As Foley (1988: 163) puts it: the language
of the dominant group is not easily made available to the members of the subordinate group(s). In fact, as has been pointed out on several occasions in the
literature, creoles that most resemble their superstratum languages were created
in communities where the speakers of the substratum languages had relatively
1

The problem of creole genesis and linguistic theory

more access to the superstratum community. Creoles that are more radical (i.e. less
like the superstratum language) come from communities where language learners
had very little access to the superstratum community (see e.g. Bickerton 1977: 55;
Baker and Corne 1982; Andersen 1983a; Thomason and Kaufman 1991; Baker
1993; Valdman 1978, 1993). For example, as is argued in Valdman (1993),
Louisiana creole is closer to French than Haitian is because the substratum
speakers had more access to French in Louisiana than the African population
had in Haiti. Baker and Corne (1982) also discuss this issue on the basis of data
from Mauritius and Reunion creoles. On Reunion, French native speakers outnumbered substratum speakers during the formative period of the creole, and
Reunion creole grammar displays a significant number of French grammatical
categories. By contrast, during the formative period of Mauritius creole, the proportion of native French speakers was much lower, and thus the West African
speakers had a much stronger input into the creole.
A fourth point is that, ordinarily, languages change gradually. Within the span
of several generations, speakers of innovative and conservative dialects are able
to communicate, even though, over the course of centuries, a new language may
evolve (see Lightfoot 1979). By contrast, creole languages are created in a short
span of time (see e.g. Hall 1958; Voorhoeve 1973; Alleyne 1966; Chaudenson
1977, 1993; Bickerton 1984). This observation dates back to Van Name (1869
70: 123, cited in Goodman 1964: 135): Under ordinary conditions these changes
proceed at so slow a pace as to be appreciable only at considerable periods of
time, but here two or three generations have sufficed for a complete transformation. Hesseling (1933: xi) further reassesses this point in the following terms:
The genesis of human language is a psychological problem that no single language will ever solve, but from creole one can best learn how a given language
emerges from old data and develops, because here something takes shape at
a high speed, in a past recognisable to us, something which is the product,
in other cases, of many centuries, with a very obscure past in its background.

Thus, in contrast to regular linguistic change, creole languages diverge abruptly

from their source languages (see Thomason and Kaufman 1991), so that, within
one or two generations, a different language is created. Hancock (1987: 265)
claims that: most of the principal characteristics that each creole is now associated with were established during the first twenty-five years or so of the settlement of the region in which it came to be spoken. Hymes (1971a), Mintz (1971)
and Ferraz (1983) suggest that a creole can develop within fifty years or less.
Singler (1996) is of the opinion that it takes sixty to eighty years for a creole to
form. Whatever the outcome of this issue may be, creole languages constitute a
unique case of accelerated linguistic change when compared with regular cases
of linguistic change.
Fifth, creole languages tend to be isolating languages. This observation goes
back to Schuchardt (1979) and Hesseling (1933: xvi). It is also found in Hagge
(1985: 39). But it was Mufwene (1986, 1990, 1991) who clearly established this
property of creole languages and the problem it poses for scholars who work on

1.1 The complex problem of pidgin and creole genesis

creole genesis. Indeed, Mufwene has documented the fact that this tendency
appears to hold even when the contributing languages are not isolating languages. For example, Mufwene (1986) shows that Kituba, a creole language that
has emerged almost exclusively from contact among agglutinative Bantu languages, is an isolating language. Kituba has selected Kikongos seemingly marked
periphrastic alternative over the more common and apparently unmarked agglutinating system (Mufwene 1990: 12).
Sixth, it has long been noted in the literature that creole languages are mixed
languages in that they derive some of their properties from those of the substratum languages and some from those of the superstratum language (see e.g.
Alleyne 1966, 1981; Holm 1988). Moreover, several scholars have noticed that
the type of mix we find in radical creoles is not random. For example, Adam
(1883: 47) states that:
Jose avancer . . . que les soi-disant patois de la Guyane et de la Trinidad constituent des dialectes ngroaryens. Jentends par l que les ngres guinens,
transports dans ces colonies, ont pris au franais ses mots, mais quayant conserv dans la mesure du possible, leur phontique et leur grammaire maternelles
. . . Une telle formation est coup sr hybride . . . La grammaire nest autre que
la grammaire gnrale des langues de la Guine.
[I go so far as to claim . . . that the so-called patois of Guyana and Trinidad
constitute NegroAryan dialects. By that I mean that the Guinean Negroes who
were transported to the colonies adopted the words of French but, as much as
possible, kept the phonetics and grammar of their mother tongues . . . Such a
formation is clearly hybrid . . . The grammar is no different from the general
grammar of the languages of Guinea.]

Speaking of Haitian creole, Sylvain (1936: 178) observes that: Nous sommes en
prsence dun franais coul dans le moule de la syntaxe africaine, ou . . . dune
langue w vocabulaire franais. [We are in the presence of a French that has
been cast in the mould of African syntax or . . . of an Ewe language with a
French vocabulary.] Similarly, in his extensive study of French-based creoles,
Goodman (1964) observes, over and over again, that particular lexical items in
the creoles have a phonological representation similar to a French expression but
that these creole lexical items share properties with corresponding lexical items
in the African substratum languages. On the basis of data drawn from Djuka,
Huttar (1971: 684) also remarks that the use of morphemes borrowed by a
pidgin or a creole language . . . from a European language often diverges from
the use of the source morpheme in the source language and often corresponds
to the use of the corresponding word in the substratum languages. Voorhoeve
(1973) makes a similar remark on the basis of Sranan and Saramaccan data.
These observations suggest that creole languages are not formed by an arbitrary
mixture of the properties of the languages present at the time they are being
created. The general pattern that seems to emerge from the observations reported
above is the following: while the forms of the lexical entries of a radical creole
are derived from the superstratum language, the syntactic and semantic properties

The problem of creole genesis and linguistic theory

of these lexical entries follow the pattern of the substratum languages. This raises
the question of what the process which generates such a division of properties
could be. The answer to this question is the main topic of this book.
Any theory of creole genesis must account for the properties of these languages. Therefore, as has been pointed out in Lefebvre and Lumsden (1989a), an
optimal theory of creole genesis must account for the fact that creole languages
emerge in multilingual contexts where there is a need for a lingua franca and
where the speakers of the substratum languages have little access to the superstratum language. It must account for the fact that creole languages tend to be
isolating languages even when they emerge from contact situations involving
only agglutinative languages. It must also account for the fact that creole languages manifest properties of both their superstratum and substratum languages
and explain why these properties are divided as they are.3 In this book, the problem of creole genesis is addressed on the basis of an in-depth study of the genesis
of Haitian creole, a typical example of a radical creole (see Bickerton 1984).
Pidgins and creoles have long been considered as separate entities on the
basis of the following two sets of criteria. While pidgins have been defined as
reduced codes, creoles have been defined as expanded versions of these reduced
codes (see e.g. Hymes 1971b). Also, while pidgins have been found to always
constitute the second language of the speakers who use them, a creole is often
considered to be a pidgin that has become the first language of a new generation
of speakers (see Kay and Sankoff 1974). In more recent literature, the distinction
between pidgins and creoles has been levelled out in view of the fact that there
are some pidgins (still used as a second language) that have been shown to have
expanded in the same way as languages known as creoles (see e.g. Mhlhusler
1980, 1986a, for an extensive discussion of this point). Hancock (1980a: 64)
states: I prefer not to acknowledge a distinction between pidgin and creole, and
to consider stabilisation more significant than nativisation in creole language
formation. Similarly, Mufwene (1990: 2) uses the term creole to refer to varieties traditionally called creoles but also to those called pidgins that serve as
vernaculars or primary means of communication for at least a portion of their
speakers. Moreover, in recent literature in the field, scholars have started referring to pidgins and creoles as pcs, suggesting that they fall into a single category. Furthermore, as will be seen in chapter 2, pidgin and creole languages
cannot be distinguished on the basis of the processes which play a role in their
formation (see also Woolford 1983, for a general discussion of this point).
Indeed, the processes hypothesised to play a role in the formation and development of human languages apply to both pidgins and creoles. This is a major
drawback to Bickertons (1977, 1981, 1984) Language Bioprogram Hypothesis
of creole genesis, which crucially requires that pidgins and creoles be different
entities formed by different processes. Since this book is about the processes
involved in the genesis of these languages and since these languages cannot be
distinguished on the basis of these processes, I will not make any distinction
between them.

1.2 The perspective of this book

1.2

The perspective of this book

In this book, the problem of pidgin and creole genesis is cast within the framework of the cognitive processes otherwise known to play a role in language
genesis and change in general. This general perspective is akin to Van Names
(186970: 123) claim that the type of changes undergone in creole genesis are
no different in kind from those observed in regular cases of linguistic change:
the changes which they [the creole dialects] have passed through are not essentially different in kind, and hardly greater in extent than those, for instance,
which separate the French from the Latin, but from the greater violence of the
forces at work they have been far more rapid. The major processes hypothesised
to be involved in the genesis of pidgin/creole lexicons are relexification, reanalysis
and dialect levelling. These processes can be argued to play a significant role
in language genesis and language change in general (see chapter 2). It is also
hypothesised that the creators of the pidgin/creole use the parametric values of
their native languages in establishing those of the language that they are creating
and the semantic principles of their own grammar in concatenating morphemes
and words. This approach compares with others as follows.
The presence of substratum features in pidgin and creole lexicons has traditionally been considered to result from calquing (see e.g. Keesing 1988) or transfer (see e.g. Naro 1978; Andersen 1980, 1983b; Mufwene 1990, 1993c; Siegel
1995). In this book, it is argued that such cases constitute examples of relexification when lexical properties are involved. Parametric values are hypothesised
to be set on the basis of those in the substratum languages and to be carried over
into the creole by its creators. The same hypothesis applies to semantic interpretation (see chapter 2).
The problem of pidgin and creole genesis has traditionally been addressed
from the point of view of simplification, or reduction, and expansion (see e.g.
Hymes 1971a, 1971b). Pidgins and creoles have traditionally been viewed as
reduced or simplified codes when compared with their superstratum languages.
Such a view, however, has been challenged by Alleyne (1966: 281), among
other researchers, on the basis of a comparison between a creole language and
its contributing languages.
Dans lhistoire de la morphologie, est-il permis de partir du systme de flexions
franais et de ne voir dans les croles franais quune rduction ou une simplification de ce systme amenant des pertes ou des disparitions des flexions
franaises? Ou bien notre point de dpart devrait-il tre la morpho-syntaxe
ouest-africaine, qui est caractrise par linvariabilit du mot, donc par labsence
de flexions?
[In the history of its morphology, is it permissible to start from the French
inflectional system and to see in the French creoles only a reduction or simplification of this system, resulting in the loss or disappearance of French
inflections? Or should we start with West African morphosyntax, which is
characterised by invariable words, i.e. by the absence of inflections?]

The problem of creole genesis and linguistic theory

The two questions posed in the quotation from Alleyne above stress the fact
that the notion of simplification/reduction arises only when creoles are compared with their superstratum languages. These notions do not have the same
relevance, however, when creoles are compared with their substratum languages
(see also Brousseau, Filipovich and Lefebvre 1989). The approach taken in this
book builds on the second alternative raised by Alleyne. The cognitive process
of relexification will be shown to account for the link that exists between the
morphosyntax of a creole and that of its substratum languages. The notion
of expansion, as referred to in Hymes (1971a), corresponds to the result of the
process of reanalysis, claimed to play a role in the development of pidgins
and creoles.
In the last twenty years, discussions of creole genesis have centred around the
debate over three main approaches to the problem (see Muysken and Smith
1986b): the universalist approach (see e.g. Bickerton 1981, 1984, 1986; Seuren
and Wekker 1986), the superstratist approach (see e.g. Chaudenson 1993) and
the substratist approach (see e.g. Alleyne 1981; Holm 1988).4 The universalist
approach does not account for the fact that creole lexicons manifest the properties of their source languages in the way they do (see section 1.1). Furthermore,
creole languages are not uniform; like other natural languages, they manifest
language-specific features, as is extensively documented in Muysken (1988b).
The universalist approach does not account for the variation that exists between
creoles. The superstratist approach raises a problem best stated by Mufwene
(1996: 166): One of the problems with the superstrate hypothesis is the absence
of any explanation for why creoles lexified by European languages do not correspond to any particular dialect of their lexifiers. Finally, the problem with the
substratist approach has been stated by Hall (1958), who points out that creoles
in general have retained very few, if any, visible features of their substratum
languages. The perspective adopted in this book isolates the discussion of pidgin
and creole genesis from these approaches in addressing the problem from the
point of view of the processes at work in their formation. Furthermore, it resolves
the problems with all three approaches; indeed, the nature of the processes
hypothesised to play a role in the formation of these languages will be shown
to predict the respective contributions to pidgins or creoles of the languages
involved in their formation.
Traditional accounts of creole genesis have generally addressed the problem
using the notion of language. Given the nature of the mix found in creole
languages, however, it is necessary to distinguish between a particular form and
its properties. This requirement has been clearly stated by Alleyne (1966: 282)
on the basis of the status of temporal and aspectual morphemes in Haitian.
Il est ais de voir que ce systme diffre beaucoup de celui des verbes franais,
dans lequel les distinctions temporelles sont beaucoup plus importantes que les
distinctions daspect. Par contre, les langues ouest-africaines font preuve de
systmes verbaux du mme genre que celui du crole franais, et il serait
videmment plus valable dattribuer au systme verbal du crole une origine

1.3 Linguistic theory

africaine plutt quune origine franaise. Mais les particules qui marquent les
aspects et les temps dans le systme verbal crole semblent toutes drives de
mots ou de groupes de mots franais. Ainsi ap()<aprs; t (particule du
pass)<tait; ava, va, a (futur)<avoir (ou peut-tre va ou avant, daprs dautres
hypothses); fk< fait que (mw fk kums je viens de commencer). Donc si
notre point de dpart rside dans les formes et non pas dans le systme, nous
serions amens chercher les moyens par lesquels le sens et la fonction
grammaticale de certaines formes franaises ont pu voluer ou se transformer
dans le crole. Cela peut fausser nos perspectives et nous plonger dans des
hypothses irrationnelles pour exprimer le changement de fonction de aprs en
crole.
[It is easy to see that this system differs greatly from the French verb system, in
which tense distinctions are far more important than distinctions of aspect. On
the other hand, the West African languages have a verb system of the same
kind as that in the French creole, and it would obviously be much more valid to
attribute an African rather than a French origin to the creole verb system. But
the particles that mark aspects and tenses in the creole verb system all seem to
be derived from French words or groups of words. Thus, ap()<aprs;5 t (past
particle)<tait; ava, va, a (future)<avoir (or maybe va or avant, according to
other hypotheses); fk< fait que (mw fk kums I just began). Thus, if our
starting-point resides in the forms, rather than the system, we shall be led to
look for the means whereby the meaning and grammatical function of certain
French forms evolved or were transformed in the creole. This can distort our
point of view and immerse us in irrational hypotheses to account for aprss
change of function in the creole.]

The situation described above calls for a theory of grammar which allows forms
and functions to be manipulated independently. Models developed within the
framework of generative grammar do allow for such analyses, for they provide
a modular approach to the various components that define a grammar. In this
approach, each module is independent from the others. Hence, phonological
representations may be treated independently from the semantic and syntactic
properties that define the functions of particular lexical entries. This general
approach provides a tool to address the problem posed by Alleyne.
1.3

Linguistic theory

The last thirty years have seen a significant shift in the focus of linguistic theory
from E(xternal) language to I(nternal) language. E-language stands for the neogrammarians and structuralists view that a language is a habit system assumed
to be overdetermined by the available evidence. I-language refers to the generativists view that a language is some element of the mind of the person who
knows the language, acquired by the learner, and used by the speakerhearer
(Chomsky 1986: 22).
While traditional accounts of the genesis of creoles have addressed the questions posed by their origin from the point of view of E-language, the account
proposed in this book takes them up from the point of view of I-language. As
Chomsky (1986: 3) puts it:

The problem of creole genesis and linguistic theory

Generative grammar . . . is concerned with those aspects of form and meaning
that are determined by the language faculty, which is understood to be a particular component of the human mind. The nature of this faculty is the subject
matter of a general theory of linguistic structure that aims to discover the framework of principles and elements common to attainable human languages; this
theory is now often called universal grammar (UG) . . . UG may be regarded
as a characterisation of the genetically determined language faculty. One may
think of this faculty as a language acquisition device, an innate component of
the human mind that yields a particular language through interaction with presented experience, a device that converts experience into a system of knowledge
attained: knowledge of one or another language.

Since creole languages are natural languages, it must be the case that the properties of creole languages follow from the more general properties of the cognitive
system which are pertinent to the configuration of natural languages and to the
transmission/acquisition of language in general.
The theory of principles and parameters (see Chomsky 1981, 1986 and related
work) holds that natural languages are basically similar. In this model, those
properties of language that are universal are formulated in terms of universal principles of grammar. The properties that are language-specific are hypothesised to
be located in the lexicon, the syntactic parameters and the interpretive component
of the grammar. This model constitutes a most useful tool for addressing the
problem of creole genesis for it provides us with a principled division between
language universals and language-specific features. Thus, on this approach, universals of language will be manifested in creoles in the same way as in any other
natural languages. What is specific to a particular creole will be found in the
components of the grammar that allow for variation between languages. Thus,
using such a model provides us with a tool to identify areas where the creole can
diverge from or resemble its source languages. A comparison of the languagespecific features of a particular creole with corresponding features in its contributing languages should tell us the source of those features. Likewise, this model
provides us with a tool to address the problem of variation between creoles.
The mentalist approach to grammar and lexicon allows for the manipulation of
semantic and syntactic information independently of phonological representations,
and it provides us with the appropriate tool to discuss transmission/acquisition
in contexts of creole genesis in terms of the transmission/acquisition of grammar
in spite of the fact that, in these cases, a new language has been created.
Finally, the mentalist approach to grammar and the lexicon defines the object
of inquiry (and hence, the methodology) with regard to a creole language as
follows: (1) What does a creole speaker know about the grammar and the lexicon
of his/her language which enables him/her to produce and understand utterances
in this language? (2) Abstracting away from the phonological representations of
the lexical entries of the various languages involved in the genesis of a creole
language, how does this knowledge compare with the knowledge speakers of the
creoles source languages have of their grammars and lexicons? These two questions constitute the central core of the research reported on in this book.

1.4 The hypothesis

1.4

The hypothesis

The general hypothesis tested by the research reported on here is that 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. Creole lexical
entries are mainly created by the process of relexification.6 Two other processes
fed by the output of relexification, dialect levelling and reanalysis, also play a
role in the development of the creole (see Koopman and Lefebvre 1981; Lefebvre
1984, 1993a; Lefebvre and Lumsden 1994b; Lumsden and Lefebvre 1994).
Relexification is a mental process defined as follows by Muysken (1981a: 61):
Given the concept of lexical entry, relexification can be defined as the process
of vocabulary substitution in which the only information adopted from the target
language in the lexical entry is the phonological representation. In testing the
role of relexification in creole genesis, we have adopted the strong position that
all the lexical entries listed in the lexicon could, in principle, undergo relexification
(see e.g. Lefebvre and Kaye 19859 Projects; Brousseau, Filipovich and Lefebvre
1989; Lefebvre and Lumsden 1989a). Thus, based on a theory of the lexicon
which, in addition to listing major category lexical items, lists functional category lexical items, productive derivational affixes and idiosyncratic expressions
such as unpredictable compounds (see e.g. Lieber 1980, 1992), all these lexical
entries7 should, in principle, undergo relexification, within the limits imposed by
the definition of the process (see chapter 2). This book presents evidence that all
these types of lexical entries do, in fact, undergo relexification. The relexification
hypothesis predicts that the lexical entries of the creole will have the semantic
and syntactic properties of the corresponding lexical entries in the substratum
languages and phonological forms derived from phonetic strings found in the
superstratum language. To a great extent, this is exactly what we do find.
The idea that relexification plays a role in pidgin and creole genesis is not a
new one. For example, Stewart (1962), Whinnom (1977), Voorhoeve (1973) and
others have long claimed that this is so. At one point, Muysken (1981a: 77) also
proposed that relexification plays a role in the formation of these languages:
If it is the case that the Caribbean creoles show numerous African survivals in
their syntax and semantics, then I think we can argue that it is not interference
which led to these survivals, but relexification.8
The research presented here has gone further than previous studies in several
ways. First, our research has improved the formal characterisation of how superstratum data are processed in relexification (see chapter 2). Second, our theory
provides a clear statement of how relexification applies in the case of functional
category lexical entries and derivational affixes. Such a theory has never been
proposed in the past, for the general assumption was that functional categories
and derivational affixes do not undergo relexification. For example, Muysken
(1988a: 15) claimed that functional lexical entries do not undergo relexification.
[Functional categories] do not have a meaning outside the linguistic system that

10

The problem of creole genesis and linguistic theory

they are part of, since their meanings are paradigmatically defined within that
linguistic system. So when you relexify a system of function words, automatically the semantic organisation of the target language comes in, and the result is
at best a compromise between source and target language systems. Muysken
(1988c) also claimed that affixes and clitics may not undergo relexification
either. This view is compatible with a theory of the lexicon where functional
category items do not constitute lexical entries. On this approach to the lexicon,
the functional category lexical items of a creole must be hypothesised to have
evolved through reanalysis only, as is extensively discussed in Lefebvre (1984).
In a theory where functional category items and derivational affixes are listed in
the lexicon, however, such lexical entries are, in principle, eligible for relexification. The third difference between this project and previous research is that we
were able to gather the resources to test this hypothesis from a global perspective
(see chapter 3). To the best of my knowledge, this is the first time in the history
of creole studies that such a large enterprise has been undertaken.
It is claimed here that, in creole genesis, the process of relexification is used
by speakers of the substratum languages as the main tool for acquiring a second
language, the superstratum language. The account proposed in this book is a further development of the second language acquisition theory of creole genesis (see
Lefebvre and Lumsden 1989a). For example, Alleyne (1971, 1981), Schumann
(1978), Valdman (1980), Andersen (1980), Mufwene (1990), Thomason and
Kaufman (1991), and Chaudenson (1993) have proposed that pidgin/creole languages constitute a crystallised incomplete stage of second language acquisition.
Without relexification, however, this approach to creole genesis does not explain
why creole languages have crystallised in the way they have (see Lefebvre 1984,
to appear a; Lefebvre and Lumsden 1989a). It is argued that the relexification
hypothesis does explain why creole lexicons reflect the properties of both their
superstratum and substratum source languages in the way they do.
By definition, relexification is a mental process that is available to speakers
who are in possession of a mature lexicon. The relexified lexicons constitute the
first instantiation of a new language: the early creole. Hence, according to the
relexification hypothesis, a creole is not created by children who are deprived of
a model for language, as is advocated by Bickerton (1981, 1984). Rather, it is
created by speakers who already have a mature lexicon. This claim is compatible
with the fact that the lexical entries of the relexified lexicons reproduce the
semantic and syntactic properties of the substratum languages. On the basis
of both historical and linguistic facts involving the genesis of Haitian creole,
it is argued that creole languages must be created by adult speakers with a
mature lexicon.
On the one hand, relexification is a mental process and hence it is an individual activity. On the other hand, situations where creoles are created typically
involve several substratum languages. Consequently, and as has been pointed
out by Lumsden and Lefebvre (1994), the new lexicons may present differences
which, by hypothesis, should reflect the differences between the original lexicons.

1.4 The hypothesis

11

After more than two hundred years of independent evolution, and because of
dialect levelling (see below), the differences between these lexicons might not
be as great as they were in the early creole. Some of these differences, however,
appear to have been maintained, since they can be observed when comparing
modern Haitian with the lexicons of its substratum languages. Examples of such
cases will be presented in several chapters of this book.
By its very nature, relexification cannot be the only process involved in
the formation of a creole, even a radical creole. First, as has been observed in
Lefebvre and Lumsden (1994b), relexification applies in creole genesis when
speakers of the substratum languages are targeting the superstratum language.
When these speakers stop targeting the superstratum language and start targeting
the relexified lexicons, that is, the early creole, they are no longer using this process. Second, as was also pointed out in Lumsden and Lefebvre (1994), since situations where creole languages are formed typically involve several substratum
languages, the lexicons produced by relexification in the context of creole genesis
are not necessarily uniform. Thus, when language learners begin to target the
language of their own community (the early creole), some compromises may be
required to reconcile these variants.
The process of dialect levelling, observed in dialect contact situations (see e.g.
Domingue 1981; Trudgill 1986; Siegel 1995, to appear), is proposed to account
for the compromises that speakers of different relexified lexicons may have to
make in creating a new language (see Lumsden and Lefebvre 1994). The proposal
that dialect levelling plays a role in the development of pidgins and creoles has
existed in the literature for some time (see e.g. Mhlhusler 1980; Mufwene 1990,
1993a; Harris 1991; Siegel to appear). The originality of our proposal (Lumsden
and Lefebvre 1994) lies in the claim that, in this case, dialect levelling operates
on the variation resulting from the relexification of the various substratum lexicons. Several examples of the process will be provided throughout this book.
It is further hypothesised that reanalysis a mental process whereby a particular form which signals one lexical entry becomes the signal of another lexical entry which is observed in cases of regular linguistic change (see Lightfoot
1979), plays a role in the development of a creole (see e.g. Lefebvre 1984;
Koopman and Lefebvre 1981; Lefebvre and Lumsden 1994b). Again, the idea
that reanalysis is involved in creole development is not new (see e.g. Sankoff
and Laberge 1973; Washabaugh 1975; Valdman and Highfield 1980; Mhlhusler
1986a; Rickford 1987; Foley 1988; Romaine 1988; Sankoff 1990, 1991; Baker
and Syea 1996). The originality of the proposal developed in the course of this
research (see Lefebvre and Lumsden 1994b) lies in the claim that, in the early
creole, reanalysis assigns a phonological form to a lexical entry produced by
relexification (see chapter 2). The formal account that is proposed will be shown
to resolve the paradox observed in the literature (see e.g. Mhlhusler 1986a,
1986b; Mufwene 1990; Sankoff 1991) to the effect that substratum languages
may influence a creole even when they are no longer in use in the community
where it is developing.

12

The problem of creole genesis and linguistic theory

The proposal advocated in this book thus argues that three major processes
are involved in the formation of creole lexicons: relexification, dialect levelling
and reanalysis. It is argued that relexification produces the input that feeds the
other two: dialect levelling and reanalysis. As for the syntax, it is hypothesised
that, in creating the creole, speakers of the substratum languages use the parametric values of their own grammars. This predicts that the creole should pair
with the substratum languages when they differ from the superstratum. To a
great extent, this prediction will be shown to be borne out by the data.
In particular grammars, there are facts related to semantic interpretation which
are independent of lexical entries and of the parametric options relevant to the
syntax. The semantic principles governing the concatenation of simplexes into
productive compounds or of affixes with bases, or the semantic interpretation
data associated with specific constructions (e.g. cleft constructions) are examples
in point. It is argued that in these cases, as well, the creole patterns on the model
of the substratum grammars rather than on that of the superstratum, showing that
its creators use the semantic principles of their own grammars in creating the
new language.
It is often assumed that creolisation involves a break in transmission. This is
the position advocated by Bickerton (1981, 1984), who claims that creole languages emerge in situations where children lack sufficient data to acquire the
language that they are exposed to. Thomason and Kaufman (1991: 152) have
claimed that creole languages resulted from a sharp break in transmission and
that they did not arise through any sort of direct transmission. While I fully
agree with these authors that creole genesis involves a break in the transmission of language, my hypothesis is that there is no break in the transmission of
semantic and grammatical properties when these are looked at from the perspective
of a given creoles substratum languages. Relexification produces a new language,
the creole. In this new language, the phonological representations of lexical
entries are derived from superstratum forms, but their semantic and syntactic
properties are derived from the corresponding lexical entries in the substratum
languages. The adult native speakers create a creole on the basis of the properties
of their own lexicon and grammar through relexification, on the one hand, and
by using the parametric values of their own grammar, on the other, and they speak
this new language to their children. The first generation of children exposed to
the incipient creole deduce the properties of the lexicon and grammar they are
exposed to on the basis of the data that they are presented with, just as in any
other case of first language acquisition. What they learn, then, is what they are
exposed to: the properties of the relexified lexicon and the parametric values of
the early creole. In the view advocated in this book, this explains why, after more
than two hundred years of separation between the substratum languages and the
creole, they can still be argued to share semantic and syntactic properties.
Finally, as has been pointed out in Lefebvre and Lumsden (1994b), like any
other language, a new language created as described above may innovate. In this
book, innovations will be referred to as cases of independent development.

1.5 The scope and limitations of this book

1.5

13

The scope and limitations of this book

The aim of this book is to document the role of relexification, reanalysis and
dialect levelling in creole genesis, to show how parametric values are set in this
context, and to document how semantic interpretation is established. This is
done on the basis of Haitian data. Due to space limitations, I cannot cover all the
material available, so I have chosen to limit myself to the most controversial
cases: functional (and functional-like) categories, the syntactic properties of verbs,
derivational affixes, the principles governing the concatenation of morphemes,
and syntactic parameters. The remainder of the available data (e.g. major category lexical entries and phonology) will be discussed elsewhere.
The question of how these processes are implemented in everyday life (given
new arrivals of slaves over several decades, etc.) is not discussed here (see Durie
in progress, for a discussion of this topic). The claim that the incipient creole is
created by adult native speakers of the substratum languages does not entail that
children have no role to play in creole genesis. I assume that, as in other situations of linguistic change, children play an important role in the development of
creoles. I refer the reader to the insightful work of G. Sankoff and her associates
for a thorough discussion of this issue based on case studies of ongoing changes
in Tok Pisin (see e.g. Sankoff 1990, 1991; Sankoff and Laberge 1973). The processes identified above as playing a role in the genesis and development of
pidgin/creole languages take place in communities and, hence, they interact with
the social components which define the features of these communities. Although
social factors are not discussed here, I assume that they interact with linguistic
processes (particularly dialect levelling) in a way similar to that described in
Labovs (and his associates) meticulous work on the interplay of linguistic and
social factors in linguistic communities (for an extensive discussion of this point,
see also Jourdan 1985 and Siegel to appear).
The book is organised as follows. Chapter 2 provides a formal definition of
the processes hypothesised to play a role in creole genesis: relexification, dialect
levelling and reanalysis. It shows how these processes apply in other cases of
language genesis and language change and how they apply in the specific context of creole genesis. It also provides a definition of parameters and how parametric values are hypothesised to be set in creole genesis. Finally, this chapter
ends with a discussion of how the concatenation of morphemes is hypothesised
to apply in creole genesis.
Chapter 3 describes the methodology developed to test the hypothesis using
Haitian creole. Since any account of the genesis of Haitian must be compatible with the external factors that prevailed at the time the creole was formed,
the research included a historical study, designed to establish when Haitian
creole was formed and the characteristics of the Haitian population during that
period. This chapter starts with a summary of the major findings of the historical research. It discusses the typological features of the languages spoken in
Haiti at the time the creole was formed and the issue of the French data the

14

The problem of creole genesis and linguistic theory

African population in Haiti was exposed to. The linguistic test of the hypothesis
rests on a detailed comparison of the lexicons and grammars of a given creole
with those of its contributing languages. This chapter reports on the linguistic
test we were able to make and discusses the database used for the research.
The second part of the book presents the results of an extensive comparison
of the lexicons and grammars of Haitian and its contributing languages. Chapters 4 to 8 discuss the properties of the functional category lexical entries involved
in Haitian nominal and clausal structures. The data presented in these chapters
support the claim that relexification has played a major role in the genesis of
Haitian creole. Furthermore, they show that both functional and lexical items
have been relexified in the process of creating Haitian. Some cases of reanalysis
and dialect levelling will also be discussed. Chapter 9 discusses the syntactic
properties of verbs. It is shown that, in this area of the lexicon as well, relexification can be argued to have played a major role. Chapter 10 addresses the
question of whether derivational affixes undergo relexification and argues that,
like other lexical entries, they do. In chapter 11, the concatenation of words into
compounds is shown to follow the pattern of the substratum languages rather
than that of the superstratum language. Chapter 12 shows that the parametric
values of Haitian pair with those of the substratum languages and contrast with
French. Cases of dialect levelling and reanalysis will be discussed throughout.
Haitian data which do not follow from the relexification hypothesis will be
pointed out, as will innovations. Finally, chapter 13 evaluates the general hypothesis on the basis of the data presented in this book and chapter 14 discusses the
consequences of these findings.

Cognitive processes involved in

creole genesis

This chapter discusses the nature of the processes claimed to be involved in

creole genesis relexification, reanalysis and dialect levelling and shows
how they apply in that context. The mental process of relexification is argued to
play a central role in creole genesis. As is advocated in Lefebvre and Lumsden
(1994b), creole languages are formed when a group of people create new lexicons through relexification and use these relexified lexicons as the primary means
of communication within the creole community. Section 2.1 discusses the formal representation of relexification. It is argued that this mental process is available to human cognition. Indeed, it has been shown to play a role in the formation
of mixed languages (section 2.2) and pidgins (section 2.3) and in some cases of
second language acquisition (section 2.4). The presence of substratal features in
pidgins and creoles is generally attributed to calquing and transfer; however, the
cases of calquing and transfer discussed in this chapter are argued to actually be
the result of relexification. Section 2.5 presents the proposal that relexification
plays a central role in creole genesis and shows how it applies in this context.
But, although relexification is claimed to be the central process involved in creole genesis, it is not the only one. When speakers of a creole community start
targeting the relexified lexicons rather than the superstratum language, two
other processes are put to work in the development of the creole: reanalysis
(section 2.6) and dialect levelling (section 2.7). These two processes, which are
available in other cases of regular change or contact between dialects, are shown
to apply in creole genesis in a particular way, interacting with relexification.
Section 2.8 addresses the problem of how parametric values, semantic interpretation facts and principles of morpheme and word concatenation are established
in creole genesis. Together with the three processes mentioned above, they are
argued to account for how the properties of creole languages emerge and develop
(section 2.9). The account of creole genesis presented in this chapter has been
developed within the framework of the research reported on throughout this
book. The main features of this account are discussed in a preliminary fashion in
Lefebvre (1982a, 1984, 1986, 1993a), the Lefebvre and Kaye Projects (19859),
Brousseau, Filipovich and Lefebvre (1989), Lefebvre and Lumsden (1989a, 1992a,
1994a, 1994b) and Lumsden and Lefebvre (1994).

15

16
2.1

Cognitive processes involved in creole genesis

The mental process of relexification

The first formal definition of relexification was provided by Muysken (1981a: 61):
Given the concept of lexical entry, relexification can be defined as the process
of vocabulary substitution in which the only information adopted from the target
language in the lexical entry is the phonological representation. Muyskens
representation of the process is reproduced in (1).
(1)

source language

target language 1

/phon/i
syni
subi
semi
seli

/phon/j
synj
subj
semj
selj
new language
/phon/j
syni
subi
semi
seli

(=(17) in Muysken 1981a)

Relexification is thus a mental process that builds new lexical entries by copying the lexical entries of an already established lexicon and replacing their
phonological representations with representations derived from another language. Following the terminology in Lefebvre and Lumsden (1994a, 1994b),
I will refer to this second phase of relexification as relabelling. According to
Muyskens (1981a: 62) proposal, relexification is semantically driven. For relexification to occur, the semantic representations of source and target language
entries must partially overlap; otherwise, the two entries would never be associated with each other. Other features of the two entries may, but need not, be
associated with each other.
Lefebvre and Lumsden (1994a, 1994b) propose a slightly different representation of the process.
(2)

original lexical entry

lexifier language
[phonetic string]j used in
specific semantic and
pragmatic contexts

[phonology]i
[semantic feature]k
[syntactic feature]n
new lexical entry
[phonology]j or []
[semantic feature]k
[syntactic feature]n

(=(1) in Lefebvre and Lumsden 1994a, 1994b)

2.1 The mental process of relexification

17

In the above representation, relabelling proceeds on the basis of phonetic strings

found in the superstratum language rather than the phonological representations
of the superstratum lexical entries. There are numerous examples in the literature showing that a phonetic string used to relabel a copied lexical entry does not
necessarily correspond to a word in the lexifier language. Instead, it may correspond to a sequence of words; for example, the Tok Pisin lexical entry baimbai>
bai is derived from the English expression by and by (see Sankoff and Laberge
1973). Cases like this are known as freezing in the literature on languages in
contact. Furthermore, as is argued by Brousseau (in preparation), the phonetic
strings of the lexifier language are interpreted by the relexifiers on the basis of
their own phonological system such that the phonological form of the new
lexical entry is often quite different from the superstratum form. Although the
phonological system of the creole appears to be historically derivable from that
of its substratum languages, the resulting system is still distinct from the substratum systems. The lexical entry created by relexification in (2) thus has a phonological representation which differs from those of both of its source languages.
A second point of difference from Muyskens representation is that the lexifier
language lexical entry in (2) is deprived of features (compare (1) and (2) ). This
is because, as we will see below, relexifiers either do not have access to this
information or, if they do, they do not use it in creating the new lexical entry.
For example, a verb may be relabelled on the basis of a noun in the superstratum
language (e.g. the Haitian verb bezwen to need which takes its phonological
representation from the French noun besoin need). Furthermore, the representation in (2) allows for a functional category lexical entry to be relabelled on the
basis of a major category lexical entry in the superstratum language. Several
examples of this type will be provided in the data chapters. Muyskens insistence on partial semantic overlap between the source and target lexical entries is
preserved in the representation in (2) by specifying that the meaning of the
phonetic string selected to relabel a copied lexical entry is deduced from its use
in specific semantic and pragmatic contexts.
Relexification as represented in (1) and (2) is a process which consists in
copying the properties of a lexical entry and relabelling it. Throughout this
book, I will assume that copying applies to all lexical entries and that it is relabelling which is semantically driven. Thus, only those functional categories
which have some semantic content (e.g. determiners, demonstrative terms, etc.)
may be assigned a new label during relexification. Those functional categories
which have no semantic content (e.g. Case markers, operators, etc.) are copied
but not relabelled. They are assigned a null form at relabelling (see (2) ). Practically speaking, this means that these lexical entries are not pronounced. The
claim that functional categories may be assigned a null form at relabelling is
independently motivated by the fact that, in natural languages, functional categories required by Universal Grammar are not always spelled out. The category
Case is an example in point. While some languages, such as Latin, Turkish
or Quechua, do exhibit overt morphological cases, other languages, such as

18

Cognitive processes involved in creole genesis

English, do not. Noun phrases are universally required to bear Case, however,
by virtue of the Case Filter (see Chomsky 1981). It has been argued that, in
languages which do not have overt Case markers, the category Case is projected
in the syntax, as in other languages (as K(ase)P, in Travis and Lamontagnes
1992 proposal), but that Case is phonologically null in these languages. The
claim that functional category lexical entries may be relabelled by a phonologically
null form thus finds independent motivation in the fact that, in some languages
not known to be creoles, functional category lexical entries may indeed be
phonologically null. Furthermore, in section 2.6, it will be argued that copied
lexical entries which were assigned a phonologically null form during relabelling are required to account for the properties of the creole lexical entries in
the further development of the language. The fact that copied lexical entries
may receive a phonologically null representation during relexification makes
the early creole lexicons look simpler than the original ones (see Lumsden
1995, for a discussion of this point). Throughout this book, I will assume the
definition of relexification in (2). By definition, lexical entries produced by
relexification have the semantic and syntactic properties of those in the original
lexicons; they differ from the original entries only in their phonological representations. It will be argued that the properties of the phonologically null forms
can be deduced by the language learners on the basis of the data they are presented with.
Since relexification is a mental process that applies to lexical entries, by
hypothesis, it should apply to all types of lexical entries. Current theories distinguish between major category lexical items (i.e. nouns, verbs, adjectives,
prepositions, adverbs and derivational affixes identified for major categorial
features) and minor or functional category lexical items (i.e. determiners, complementisers, Tense markers, etc.) (see Chomsky 1972, 1986, 1989; Jackendoff
1977, and related work). Muysken (1988a: 428 and 1988c) has claimed, however, that functional category lexical items do not undergo relexification: the
only African features that could have been transmitted more or less intact
through relexification are those dependent on properties of content words. This
means: lexically determined semantic distinctions and subcategorisation features, but no syntactic properties related to function words. A different view is
argued for in this book. It is claimed that the mental process of relexification
may, in principle, apply to minor as well as major category lexical entries. It is
argued that, in creole genesis, both minor and major category lexical entries
can undergo relexification as specified above (see Lefebvre and Lumsden 1989a,
1994a; Lefebvre 1994a). In my view, the fact that functional category lexical
entries have not been relexified in specific mixed languages (see below) reflects
a limitation imposed not on the process itself but on the situation in which
it applies.
Relexification is a mental process that is available to human cognition. The
following sections document this claim.

2.2 Relexification in the genesis of mixed languages

2.2

19

The role of relexification in the genesis of mixed languages

This section discusses three cases of mixed languages: Media Lengua, Michif
and Inner Mbugu or Maa. It will be shown that the same cognitive process has
played a major role in the formation of all three languages, namely relexification.
These three cases are complementary in several ways. First, these mixed languages
emerged in different geographical areas: the Andean area of South America,
Western Canada (North America) and Tanzania (Africa), respectively. Second,
the languages involved in their formation belong to quite different families:
Quechua and Spanish (in the first case), Cree and French (in the second), and
Mbugu, Bantu, Cushitic and other languages (in the third). Third, although the
types of mix we find in these three cases share some basic properties, on the
surface they present differences, showing that one cognitive process, namely
relexification, can yield superficially different results. The differences in the
output are claimed to be independent of the cognitive process itself. Rather, they
are argued to follow from the differential components defining the situation in
which relexification applies. The section ends with a discussion of the features
that distinguish mixed languages and pidgin/creole languages.
2.2.1

Media Lengua

Relexification was an important tool in the creation of Media Lengua, a mixed

language spoken in Ecuador. Muysken (1981a, 1988c) shows that Media Lengua
has a lexicon where the phonological forms of lexical categories (i.e. nouns, verbs
and adjectives) are almost entirely derived from Spanish, while the forms of the
affixes and the functional category system are derived from Quechua. The examples
in (3) and (4) illustrate this division. (In the Media Lengua examples below, the
Spanish forms are in regular characters and the Quechua forms are in italics.)
(3)

a. No s.
not know-1st
I do not know.

spanish

b. Mana yacha-ni-chu.
No
sabi-ni-chu.
not
know-1st-val
I do not know.

quechua
media lengua
(=(3) in Muysken 1981a)

(4)

a. Si llueve
dems,
no voy
if rain-3rd too-much, not go-1st
If it rains too much, I will not go.
b.

Yalli-da
tamia-pi-ga, mana
Dimas-ta
llubi-pi-ga, no
if too-much-acc rain-lo-to not
If it rains too much, I will not go.

a ir.
to go

spanish

ri-sha-chu.
i-sha-chu.
go-asp-val

quechua
media lengua

(=(1) in Muysken 1981a)

20

Cognitive processes involved in creole genesis

Muysken (1981a) shows that, although the lexical categories of Media Lengua
derive their phonological representations from Spanish, their semantic content
is derived from Quechua. For example, the phonetic form of the Spanish verb
sentarse sit down was used to relexify the Quechua verb tiya-ri sit, live,
locative be, there is yielding the Media Lengua verb sinta-ri sit, live,
locative be, there is. As Muysken (1981a: 56) points out, a single Media
Lengua word is substituted for the Quechua word, preserving the various meanings of the latter even when, in Spanish, each of these meanings would be
expressed by a separate lexical item: estar sentado sit, vivir live, estar locative
be, hay there is. Similarly, the phonetic form of the Spanish verb tener to
have, to hold was used to relexify the Quechua verb chari- to have yielding the Media Lengua verb tini- to have (see Muysken 1988c). The above
examples show that, although there must be some overlap in the semantics
of the verbs that are matched in the process of relexification, the semantics of
the new Media Lengua verbs follows the details of Quechua rather than the
Spanish semantics.
Muysken (1979, 1981a) argues that Media Lengua cannot be considered to
represent a stage in learning Spanish as a second language since many Media
Lengua speakers also speak fluent Spanish, and Media Lengua is very different
from QuechuaSpanish interlanguage. The fact that the Media Lengua lexical
entries have the semantic and syntactic properties of Quechua rather than Spanish
is thus not attributable to a lack of exposure to the lexifier language. Nonetheless, the facts show that the creators of Media Lengua did not use the properties
of the Spanish lexical entries which provided the phonetic matrices for the new
lexical entries. This is compatible with the representation of the lexifier phonetic
string in (2), which shows no specific features.
Muysken (1981a) also reports on several cases of freezing, that is, cases of
morphologically separate Spanish forms which appear as a single lexical item in
Media Lengua. Some examples are given in (5).
(5)

spanish
a. no ha habido
there has been no . . .

media lengua
nuwabishka

b. no hay
there is no . . .

nway

c. an no
not yet

ano

d. a mi
me (non-nominative pronoun)

ami
(=(6) in Muysken 1981a)

These data show that some Media Lengua lexical entries have been relabelled on
the basis of complex expressions involving more than one word in the lexifier
language. This argues in favour of the claim that, in the process of relexification,

2.2 Relexification in the genesis of mixed languages

21

the new lexical entry derives its phonological representation from a lexifier
language phonetic string (see (2) ) rather than a phonological form (see (1) ).
Furthermore, as is pointed out in Muysken (1981a), Media Lengua shows the
systematic adaptation of Spanish vocabulary to Quechua phonology. This phenomenon is typical of situations where relexification takes place.
According to Muyskens analysis, then, Media Lengua is a typical case of
a mixed language produced by the relexification of the major category lexical
items of the substratum lexicon. It is an intra-group language, not known outside
the communities where it is spoken. As mentioned above, Media Lengua cannot
be considered to represent a stage in learning Spanish as a second language.
Muysken (1981a: 75) further argues that it was not communicative needs that
led to it, but rather expressive needs.
2.2.2

Michif

Michif is a CreeFrench mixed language developed by the Mtis buffalo hunters

of Canada and the Northern United States (see Papen 1988; Bakker 1989, 1992,
1994). Cree is an Amerindian language of the Algonquian family, and thus a
typical agglutinative language, whereas French is a Romance language. According to the description of Michif in Bakker (1994), the lexicon of this mixed
language is divided between its source languages as follows. While the forms of
nouns are approximately 90 per cent from French, the forms of verbs are almost
all from Cree. Whereas the forms of possessive pronouns are almost always
from French, the forms of personal pronouns and demonstrative pronouns are
from Cree. Numerals and adjectives are always from French. As for adpositions,
the forms of prepositions in Michif tend to be derived from French and those of
postpositions from Cree. Adverbial particles, negative elements and conjunctions appear to be drawn from both of the source languages. Articles are from
French and verbal morphology is from Cree. Given this description, it appears
that French-derived lexical elements are predominant in noun phrases, while
Cree lexical elements predominate in verb phrases. Nonetheless, there are Cree
lexical items in noun phrases, such as demonstratives, and French lexical items
in verb phrases, such as nouns derived from French forms which are incorporated into the verb as in Cree.
As is the case with Media Lengua, where all affixes are from Quechua,
in Michif all affixes are from Cree. These affixes appear on both nouns and
verbs regardless of whether they are of French or Cree origin. For example, in
(6) the obviative affix occurs on the object noun. (In the Michif examples below,
words of Cree origin are in italics; words in regular characters are of French
origin.)
(6)

La jument l
talon wa
otin w.
michif
the mare
the stallion obv take (s)he / him / her
The mare takes the stallion.
(=(3) in Bakker 1994: 21)

22

Cognitive processes involved in creole genesis

In (7), the Cree affix -ipan deceased is suffixed to a noun of French origin.
(7)

mq vj
ipan
my husband deceased
my deceased husband

michif
(=(49) in Bakker 1992: 169)

Similarly, in (8) the Cree plural suffix appears on a noun which derives its
phonological form from French.
(8)

John tahkuht am
li: fiy
John bite
3 4 ap girl
Iwa:w.
3.pl
John bites the girls ears.

anIhl
si:
dem-obv 3.poss.pl

zar=j
ear

michif

(=(56) in Bakker 1992: 171)

Cree affixes also occur on verbs which are of French origin as in (9).
(9)

gi: l ga:n: n
1.pst the bet
in.1
I bet on the brown one.

syr
on

l
the

brt.
brown

michif
(=(74) in Bakker 1992: 174)

The above data illustrate the kind of mix we find in Michif, showing that, as is
observed by Bakker (1994: 167): Michif structure is the result of the combination of Cree grammar with French lexicon . . . The grammatical bound elements
are Cree and the lexical-free elements are French. As is pointed out by Bakker,
the verb generally has a Cree form because it consists of grammatical bound
elements (see below).
Bakker (1992: 146, 1994: 20) refers to the process yielding such results as
language intertwining, a term which he claims is an alternative to relexification.
In my view, the term language intertwining does not refer to a mental process
any more than language mixing, pidginisation or creolisation do. Rather, it refers
to the result of a process, and this process, I claim, is relexification. Indeed, the
data presented by Bakker show the division of properties predicted by the process of relexification as depicted in (2) above: while the semantic and syntactic
properties of the new lexical entries are derived from the original lexical entries
(in this case Cree), the phonological representations of the new lexical entries
are derived from the phonetic strings of another language (in this case French).
A striking fact supporting this claim is found in the syntactic features of
Michif nouns which derive their phonological representations from French. The
syntactic features of these nouns appear to be determined by the corresponding
Cree nouns. This can be seen in the fact that Michif demonstratives, which are
of Cree origin, agree in number and animacy with the nouns they modify,
showing that the Michif nouns are identified for the feature [animate] rather
than for the feature [feminine], as is the case with French nouns. The Michif
agreement facts are illustrated in (10).

2.2 Relexification in the genesis of mixed languages

(10)

a. anima
l
that.inan the
that book

23

li:v
b. ana
lum
michif
book
this.an man
this man
(=(38), (39) in Bakker 1992: 1718)

These agreement facts are to be expected if Cree nouns have been relexified on
the basis of French phonetic strings; according to the relexification hypothesis,
the Michif nouns should have the syntactic properties of the original lexical
entries that they were relexified from.
A second example showing that the Michif lexical entries must have been
created through relexification resides in the semantic properties of the locative
preposition da/d meaning in, on, at. This preposition derives its phonological representation from the French preposition dans in. As noted by Bakker
(1992: 171), however, the range of meanings covered by the Michif preposition
is much wider than that covered by its French source but parallels that of the
corresponding Cree lexical entry. Cree has a locative postposition -ihk (with
an allomorph -ohk) meaning in, on, at. As Bakker (1992: 171) observes: The
preposition da/d apparently copied the wider functions of the Cree locative. In
the light of the following, it does not seem possible to simply say that there
has been a process of generalisation of meaning. The da preposition in Michif
is sometimes combined with other Cree or French adpositions in Michif which
is not possible in French. Examples showing that the semantics of Michif da
follows that of the corresponding Cree lexical entry are given in (11) and (12).
(11)

a. narj=r d
l
ma:r
in-back loc the car
behind the car
b. otahk
otpnask-ohk
behind car
loc
behind the car

michif

cree
(=(57), (59) in Bakker 1992: 171)

(12)

a. d
l
frdn uhDi
loc the fridge from
out of the fridge
b. tahkascikan-ihk ohci
fridge
loc from
out of the fridge

michif

cree
(=(58), (58) in Bakker 1992: 171)

The semantic properties of the Michif lexical entry da/d thus parallel those of
the Cree lexical entry -ihk (-ohk). This follows straightforwardly if the Michif
lexical entry was created through relexification.2 Many more examples of the
same type are provided in Bakker (1989, 1994), showing that relexification was
the process at work in the creation of Michif.
So, like Media Lengua, Michif is a mixed language produced by relexification.
The type of mix we find in Michif, however, is slightly different from the type
of mix in Media Lengua. While almost all major category lexical items in Media

24

Cognitive processes involved in creole genesis

Lengua derive their phonological representation from Spanish, in Michif only a

subset of the major category lexical items have been relexified. For example,
in Michif a mere handful of verbs have a phonological representation derived
from French and the great majority have come from Cree. Bakker (1994: 21)
attributes this fact to the polysynthetic nature of Cree:
For the mixed language of the Mtis, one would expect a Cree grammatical
system with French lexicon. But the problem is that here it is impossible to
combine the two in the same way as in the other cases, due to its polysynthetic structure and often blurred morpheme boundaries in the verb. There is a
continuum between stem-formational, derivational and inflectional morphemes
in the verb, which makes it impossible to separate the stems from the affixes.
French verb stems cannot be combined with Cree verbal morphology without
destroying the whole organisation of the Cree structure (only in some marginal
cases which look like loanwords is this apparently possible, as discussed above).
For that reason, Cree verbs belong as a whole to the grammatical system and
therefore have to remain Cree in a combination with other languages.

Thus, the difference between Media Lengua and Michif lies in the fact that,
in the latter case, only a subset of the major category lexical entries has been
relexified. In both cases, the functional category lexical entries are from the
substratum language, i.e. Quechua and Cree, respectively.
In what context was Michif created? Bakker (1992, 1994) proposes the following scenario. Michif was created by bilingual speakers (French/Cree), probably adolescents whose fathers spoke French and whose mothers spoke Cree.
According to Bakker (1994: 23), Michif shows no characteristics of a second
language acquisition interlanguage. It developed rapidly as an intra-group language among the Mtis, who had mixed backgrounds from the point of view of
both race and culture. Thus, like Media Lengua, Michif is an intra-group language, not known outside the communities where it is spoken. And, like Media
Lengua, it is not mutually intelligible with either of its source languages: French
and Cree.
2.2.3

Inner Mbugu or Maa

The case of Inner Mbugu or Maa, a mixed language spoken in Tanzania, was
first brought to the attention of creole researchers in a short paper by Goodman
(1971) entitled The strange case of Mbugu. Goodmans (1971: 253) paper ends
with the following puzzling questions:
Can Mbugu be considered an instance of pidginisation and subsequent
creolisation? . . . If Mbugu has undergone pidginisation, it is of a very different
type from that undergone by known pidginised varieties of Bantu languages.
Can Mbugu, on the other hand be considered a Bantu language which has
been relexified with a non-Bantu vocabulary? A comparison of the Bantu
and non-Bantu portions of the Mbugu vocabulary shows that the former is, in
general, more likely to have been borrowed than the latter. . . . If Mbugu were

2.2 Relexification in the genesis of mixed languages

25

a relexified Bantu language, one would expect the reverse situation, since
higher numerals are more readily borrowed than lower ones. Thus, the development which Mbugu has undergone defies easy categorisation; it remains a
unique specimen.

The puzzling questions posed by Goodman have received puzzling answers.

For example, Thomason and Kaufman (1991) suggest that Inner Mbugu or Maa
originated from a Cushitic language that underwent massive borrowing from
Pare, a Bantu language. Mhlig (1983) has proposed an opposite view according
to which Maa is a Cushiticised Bantu language.
A more recent study of Inner Mbugu or Maa by Mous (1994, 1995, to
appear), based on extensive fieldwork on the languages involved, demonstrates,
however, that it is a mixed language which has much in common with Media
Lengua and Michif. According to Mouss description, the Mbugu people, surrounded by speakers of Shambaa and Pare, two Bantu languages, speak two
languages: normal Mbugu (nm), a Bantu language, and Inner Mbugu (im), a
mixed language. nm and im (or Maa) share the same grammar, which is basically Pare. As is documented in detail in Mous (1994, to appear), both languages
share the same complex verbal inflectional system. Mous (to appear: 2) asserts
that: The fact that Inner and Normal Mbugu have the same grammar can be
illustrated by the verbal inflectional system, which shows a richness in tense.
Over forty tenses have been found so far. The two varieties share all of these
tenses with exactly the same morphological form, including tone patterns.
Both languages share the same complex system of subject and object concords,
realised by the same morphological forms. They have the same verbal suffixes
and the same tense, mood and aspect prefixes and suffixes (see Mous 1994, to
appear). The other lexical and functional categories in the two languages have
the same grammatical features and semantic properties. Mous (1994) provides
ample evidence that nm and im differ only in the phonological representation of
the lexical entries. A few examples are given below.
im and nm share the same complex system of nominal classes, a typical
feature of Bantu languages. As is pointed out by Mous (1994: 187), equivalent
nouns are members of the same class in both varieties. The difference between
the two languages appears to lie only in the phonological representations of the
lexical entries involved. Examples of noun classes in im and nm are provided in
(13) along with the hypothesised source of the im form.
(13)

Examples of noun classes in Inner and Normal Mbugu

Class
1
2
3
4
14.1
6

nm / Pare
m-nhtu
va-nhtu
m-kno
mi-kno
vu-sh
ma-sh

Subj.
person /
people v / v
arm

arms

face
v
faces
/

Obj.
m
v

im
m-h
va-h
m-hargha
mi-hargha
vu-bas
ma-bas
(=(14)

Source of im
Iraqw (sc): hee
idem
Oromo (ec): harka
idem
Origin unclear
idem
in Mous 1994: 187)

26

Cognitive processes involved in creole genesis

Mous (to appear) emphasises that corresponding lexical entries in im and nm are
complete synonyms, even though their phonological representations are different. A sentence containing the verb to break exemplifies this fact.
(14)

a. a-p
ndat ku.
1:pst-break stick.9 his
He broke his stick.
b. a-baha
ndat
1:pst-break stick.9
He broke his stick.

y-akw.
9-his

inner mbugu

normal mbugu
(=(5a) in Mous to appear)

As Mous (to appear: 4) points out, if the object of the verb to break is a body
part or something belonging to the subject of the verb, no possessive is needed
and a reflexive object pronoun is optional. Again, Inner Mbugu and Normal
Mbugu do not differ in this respect. This is illustrated in (15).
(15)

a. a-(k)-p
kusme.
1:pst-rfl-break leg.15
He broke his leg.
b. a-(k)-baha
kugh.
1:pst-rfl-break leg.15
He broke his leg.

inner mbugu

normal mbugu
(=(5b) in Mous to appear)

A last set of examples showing semantic parallels between im and nm is provided in (16), showing that in both languages parallel lexical items share metaphorical uses.
(16)

im
mxat
hlku
h
gewa

nm
mt
gwisha
chuma
jughulwa

tree, afterbirth
to drop, give birth
to sew, mould
to be opened, to be allowed to get married
(=(6) in Mous to appear)

Based on the above description, im or Maa appears to be similar to the other

cases of mixed languages reported above: it has the grammatical structure of
nm or Pare, and a lexicon with the syntactic and semantic properties of nm, but
with phonological representations which, to a large extent, are derived from
other languages. An extensive discussion of the various sources of the phonological representations of im lexical entries may be found in Mous (to appear).
Suffice it to say here that these phonological representations come from a variety
of neighbouring languages, including Cushitic languages, and from modifications to the corresponding forms in nm. In this respect, im or Maa differs from
the other mixed languages reported on above, where the phonological representations of the relexified lexical entries are derived from only one language.
Mouss (1994, to appear: 10) interpretation of these facts is that im and nm
are two registers of the same language: each lexical entry has one semantic and

2.2 Relexification in the genesis of mixed languages

27

morphological description but two forms, one labelled Inner Mbugu and the
other Normal Mbugu. The difference is one of expressing more or less Mbuguness, more or less emphasis on ethnic identity, on being different from the surrounding people. Both registers are acquired as a first language and both are
spoken within the family. Mous (1994: 176) points out that Mbugu children also
acquire Shambaa, the dominant Bantu language in the area, and Swahili, the
national language, before they start school. Speakers of the neighbouring languages, such as Pare, understand Normal Mbugu but find Inner Mbugu incomprehensible. According to Mous (1994: 199): Inner Mbugu is a lexical register
that was created by speakers of Normal Mbugu. They did this consciously and
on purpose, to set themselves apart from their Bantu neighbours. Furthermore,
IM shows no characteristics of a second language acquisition interlanguage (for
further discussion of this point, see Mous 1994, to appear).
Based on this description, then, im shares many linguistic and sociolinguistic
properties with the other mixed languages discussed above. These facts bring
us to a discussion of the process which generated this mixed language. Since
relexification was shown to play a significant role in the other two cases, one
would expect that the creators of im had recourse to the same type of cognitive process in creating it. Mous (1995: 1), however, proposes that a cognitive
process which he refers to as paralexification was used in the creation of im.
Paralexification is a process which creates a new phonological form for a given
lexical entry. Paralexification is the addition of a word form to a lexical entry.
This added form is on a par with the existent word form of the lexical entry in
question. That is: two word forms share meaning, metaphorical extensions, and
morphological properties such as noun class membership for nouns and predicate frame for verbs. Mouss definition of paralexification can be schematised
as in (17), where a given lexical entry has two phonological representations and
only one set of semantic and syntactic features.
(17)

G [phonology]i / [phonology]j J
K
H
[semantic feature]k
L
I
[syntactic feature]n

In my view, paralexification (as in (17) ) and relexification (as in (2) ) constitute two slightly different ways of representing the same cognitive process. Both
representations describe a process which consists in creating a new phonological
representation for an already established lexical entry. Both representations allow
for the availability of both forms in the competence of speakers over (a certain
period of ) time. Indeed, based on the representation in (2), speakers have two
parallel lexicons (the original one, and the one created by relexification) which they
can use alternately. In the representation in (17), on the other hand, speakers have
a single lexicon wherein each lexical entry has two phonological representations
which can be used alternately. Because the definition in (2) specifies what type
of information from the lexifier language is used in relexification, I will retain this
definition for the discussion that follows. I will reconsider (17) in chapter 13.

28
2.2.4

Cognitive processes involved in creole genesis

Variation in the overall outputs of relexification in mixed languages

The three mixed languages discussed above (Media Lengua, Michif and Inner
Mbugu or Maa) share certain characteristics which are typical of mixed languages
in general.3 First, situations where mixed languages emerge generally involve
only two languages: a substratum language and a lexifier language. While Media
Lengua and Michif conform to this general tendency, im appears to depart from
it since it has more than one lexifier language. Second, as has been emphasised by
several scholars who have done extensive work on the genesis of mixed languages,
the creation of mixed languages does not involve second language acquisition,
since in these situations a relatively large percentage of the community masters
both languages (see e.g. Muysken 1981a; Bakker 1992; Bakker and Mous 1994a).
The three mixed languages discussed here are claimed not to involve second
language acquisition. Third, mixed languages are argued to constitute intra-group
languages motivated by a will for in-group identity vis--vis neighbouring linguistic groups. Media Lengua, Michif and im were all shown to possess this
characteristic. Fourth, mixed languages are claimed to derive their grammatical
properties from one language (the substratum language). The phonological representations of the lexical entries are, however, drawn from both languages: the
phonological representations of major category lexical entries tend to be derived
from the superstratum or lexifier language, and those of functional category
lexical entries from the substratum language. This reflects the fact that, in the
formation of mixed languages, major category lexical entries are relexified but
minor category lexical entries generally are not. Although the mixed languages
discussed above conform to this general tendency, they depart from it in two ways.
In the first place, we have seen that im has more than one lexifier language
and thus the phonological representations of im lexical entries are derived from
more than two languages, in contrast to Media Lengua and Michif. On the surface, then, the mixed language im looks slightly different from Media Lengua
and Michif. Moreover, while in Media Lengua and im all major category lexical
entries have been relexified, in Michif, verbs have not been relexified due to a
structural constraint imposed by the substratum language. In this respect, Michif
also looks different from the other two cases, since lexical entries related to
noun phrases appear to have been relexified whereas those related to verb phrases
have not. Thus, although the same process of relexification can be argued to
have been a primary tool in the creation of all these languages, the overall output
of this process varies. As we have seen, however, the variation in the output is
not due to the process as such but to components which define the situations in
which it applies, such as the structural make-up of the substratum language,
whether more than one lexifier language is involved, etc. As will be shown
below, the genesis of pidgin and creole languages presents yet another case
where relexification plays a role. And we will see that the output of the process
of relexification in these cases differs from that found in mixed languages due to
the nature of the situations in which it applies.

2.2 Relexification in the genesis of mixed languages

2.2.5

29

Mixed languages versus pidgin/creole languages

The characteristics of the mixed languages discussed above compare with those
of creole languages as follows.4 First, while situations where mixed languages
emerge generally involve only two languages, situations where pidgin and creole
languages are formed involve several languages (see chapter 1). In the latter situation, there is a need for a lingua franca, whereas in the former there is no such
need. Mixed languages constitute intra-group languages motivated by a desire
for in-group identity vis--vis neighbouring linguistic groups. Pidgins and creoles
are formed for a different purpose, that is, to ease communication between
groups that have no language in common (see chapter 1).
Second, mixed languages derive their grammatical properties from one language (the substratum language); however, the phonological representations of
the lexical entries are drawn from two languages: major category lexical items
tend to be derived from the superstratum language and functional category lexical items from the substratum language. As will be seen throughout this book,
creole languages, like mixed languages, derive their grammatical properties from
their substratum languages; in contrast to mixed languages, however, both major
and minor category lexical items derive their phonological representations from
a single source, the superstratum language. This reflects the fact that while, in
the formation of mixed languages, minor category lexical items are generally not
relexified, in creole languages they generally are. The claim that minor category
lexical items with some semantic content undergo relexification in creole genesis will be extensively argued for in several chapters of this book. The reason
for this difference between mixed and creole languages is, I believe, the following. On the one hand, since there is only one substratum language involved in
the creation of mixed languages, and since that language is shared by the creators of the new language, the retention of minor category lexical items from
that shared language does not prevent communication between speakers. On the
other hand, since several substratum languages are involved in the creation of
pidgin and creole languages, the retention of minor category lexical items by
speakers of these various languages would hinder communication between them.
Since, as we saw above, the purpose of creating a creole is to ease communication between groups without a common language, it follows that, in creole genesis, functional category lexical entries must be relexified if the speakers of the
various substratum languages want to share a common vocabulary in this area of
the lexicon as well.
Third, as has been emphasised by several scholars who have done extensive
work on the genesis of mixed languages, the creation of mixed languages does
not involve second language acquisition, since in these situations a relatively
large proportion of the community already masters both languages involved (see
e.g. Muysken 1981a; Bakker 1992; Bakker and Mous 1994a). In contrast, creole
genesis is a particular case of second language acquisition where there is reduced
access to the superstratum language (see chapter 1). As will be seen below

30

Cognitive processes involved in creole genesis

(following Lefebvre and Lumsden 1994a), creole genesis is a particular case of

second language acquisition (see e.g. Thomason and Kaufman 1991) where the
mental process of relexification must be used as the main tool for creating a
new lexicon.
In spite of the differences discussed so far, both mixed languages and pidgin/
creole languages share the property of being created in a relatively short time
span. Since relexification is proposed to be a process available to human cognition to serve, among other things, as a tool for creating new lexicons, and since
this process has been shown to play a significant role in the genesis of mixed
languages, it can be hypothesised to play a similar role in pidgin and creole
genesis (see Lefebvre 1986).
2.3

The role of relexification in the genesis of pidgin languages

This section documents the claim that relexification has played a major role in
the formation of pidgin languages on the basis of meticulous work by Keesing
(1988) on Solomons Pidgin. The second part of this section discusses situations
where expanded pidgins develop with those where abrupt creolisation is hypothesised to occur.
2.3.1

Calquing as relexification in the formation of Solomons Pidgin

Keesing (1988: 1, 2) writes:

Sitting on a Solomon Islands mountain in 1977, reading Derek Bickertons
review article on Pidgin and Creole Studies (1976), I was led to think more
seriously than I ever had about the history and structure of Solomon Islands
Pidgin. I had earlier been struck, when I had learned Solomon Pidgin in the
1960s through the medium of Kwaio, an indigenous language I already spoke
fluently, that this learning task mainly required learning Pidgin equivalents of
Kwaio morphemes. The syntax of Solomon Pidgin was essentially the same as
the syntax of Kwaio, although somewhat simpler and lacking some of the surface
marking; in most constructions, there was a virtual morpheme-by-morpheme
correspondence between Kwaio and Pidgin. (This was not just an odd local
process of calquing: the Pidgin I was learning in terms of Kwaio was spoken
with only minor variations throughout the southeastern and central Solomons,
although it was everywhere adapted to local phonologies.) Although most of
the Pidgin lexical forms were ultimately derived from English, I found this
largely irrelevant to my language-learning task. The semantic categories they
labeled corresponded to Kwaio ones, not English ones; grammatical morphemes
corresponded to Kwaio ones, not English ones. Thus semantically Pidgin dae
corresponded directly to Kwaio mae be dead, die, be comatose, be extinguished,
not to English die. Pidgin baebae corresponded to the Kwaio marker of future/
nonaccomplished mode, ta-, not to English by and by.

Keesing accounts for the linguistic situation he describes in terms of calquing.

That is, the substratum speakers of Solomons Pidgin calque the properties of
their native languages (e.g. Kwaio) when speaking the pidgin.

2.3 Relexification in the genesis of pidgin languages

31

Keesing (1988) documents the fact that calquing of the substratum properties
can be observed throughout the lexicon of Solomons Pidgin. He shows that the
pronominal system of this pidgin is quite similar to that of the complex system
of the substratum languages in distinguishing singular, dual and plural, inclusive
and exclusive first-person plural, etc. He argues that the Tense, Mood, Aspect
system of Solomons Pidgin reproduces the idiosyncrasies of the system of the
substratum languages. As Keesing (1988: 215) puts it: In fact, the entire set of
Kwaio particles marking the time-frame of the verb, some of which are preverbal
and some postverbal, correspond in their Solomons Pidgin usage to a set of particles derived from English but carrying exactly the same import as the Kwaio
particles, and placed in exactly the same slots. Keesing further shows that, as is
the case in the substratum languages, Solomons Pidgin has a predicate marker.
The same pattern is also found in interrogative constructions, relative clauses,
etc. In short, Keesing provides extensive evidence that, while the phonological
representations of Solomons Pidgin lexical entries are derived from English phonetic matrices, the properties of these lexical entries do not correspond entirely
to those of English lexical entries; he convincingly demonstrates that the properties of the Solomons Pidgin lexical entries do, however, correspond to those
of its substratum languages, including functional category lexical entries. The
following example illustrates this situation.
(18)

a. Gila
ta-la
leka.
fp (them) fut-srp (they) go
They will go.
b. Olketa
bae-i
go.
fp (them) fut-srp (3pl) go
They will go.

kwaio

solomons pidgin
(from Keesing 1988: 214)

While olketa in the pidgin derives its phonological representation from the English expression all together, it has the meaning and uses of the substratum
strong personal pronoun gila them. While bae in the pidgin derives its phonological representation from a reduced form of the English expression by and by,
its meaning and uses correspond to the substratum lexical entry ta-, a future
marker. As in the substratum language, the future marker of the pidgin is marked
for a third-person pronominal form. This pidgin form is derived from the English he, but it does not share the uses of the form it is phonologically derived
from; it does, however, share the properties of the substratum forms, as can be
observed by comparing the (a) and (b) sentences in (18).
In the conclusion to his book, Keesing (1988: 227) addresses the following
question to theoretical linguists: How could a pidgin have evolved that, despite
its almost total lexification from English as superstrate language, has a structure
so close to that of Southeast Solomonic Oceanic languages? My answer to this
question is that the massive calquing identified by Keesing in Solomons Pidgin
is, in fact, the product of the process of relexification.5 As shown above, the

32

Cognitive processes involved in creole genesis

Solomons Pidgin lexical entries present the division of properties we expect of

lexical entries produced by relexification (see (2) ). This being the case, Keesing
was the first scholar to document in detail the central role of relexification in the
genesis of a pidgin language.
The output of relexification in this case is, however, different from that
observed in mixed languages. Both functional and major category lexical items
have phonological representations derived from English forms, suggesting that
functional category lexical items have also been derived through relexification.
This should not come as any surprise given that Solomons Pidgin is an expanded
pidgin in the sense discussed in chapter 1.
2.3.2

Expanded pidgins versus cases of abrupt creolisation

According to the description in Keesing (1988), Solomons Pidgin is used as a

lingua franca by speakers of its numerous (but genetically related) substratum
languages. This pidgin is thus the second language of all the speakers who use it.
It permits communication between speakers whose native languages have different phonological representations of their lexical entries but similar semantic and
syntactic properties. Hence, Solomons Pidgin coexists with its substratum languages such that its speakers still have access to the grammars and lexicons of
their native languages.
This situation appears to be different from what Thomason and Kaufman
(1991: 148) refer to as cases of abrupt creolisation (e.g. radical creoles). The
situation is different when we turn our attention to creoles, especially those
creoles for which no fully crystallised pidgin stage is attested namely, primarily, creoles that arose in the context of the European slave trade in Africa, the
Caribbean area, and several islands in the Indian Ocean. It is assumed that, in
abrupt cases of creolisation, the substratum languages did not coexist with the
newly developed language for as long as in the case of Solomons Pidgin. As
Singler (1988: 47) puts it: The dichotomy between extended pidgin and creole
correlates with differences in social history: the creole scenario . . . leads to a
break with the substrate, but the extended pidgin scenario need not.
It could be hypothesised that it is the bilingual context in which Solomons
Pidgin is spoken which makes the role of calquing or relexification so important
in the genesis of this pidgin. Likewise, it could be hypothesised that, in abrupt
creolisation, the break with the substrate languages does not favour calquing or
relexification. This is, in fact, the position taken by Thomason and Kaufman
(1991: 152), who claim that cases of abrupt creolisation result from a sharp
break in transmission. They further claim that:
all early creolised creoles also have a non genetic origin. That is, the African languages were abandoned, but the European languages were not acquired as whole
languages by the slave populations. This removes all these creoles from consideration for genetic classification: they are not changed later forms of any parent
language. They did not, in our view, arise through any sort of direct transmission

2.4 The relationship between transfer and relexification

33

from one speaker group to another, either by generation-to-generation/peer-toyounger-peer enculturation or by a shift involving acquisition (perhaps with
modification) of the grammar of a TL [= Target Language]. Instead, they were
created in various multilingual communities by the first generations of slaves.

This is not the position taken in this book, however, for several reasons. First
of all, this view is not compatible with the numerous examples in the literature
of the presence of substratal features in radical creoles (see references cited in
chapter 1, section 1.1.1; see also Holm 1993; Alleyne 1981; Smith, Robertson
and Williamson 1987; McWhorter 1996). As we will see below, these cases are
generally referred to as cases of transfer from the substratum languages into the
creole. Second, as will be seen in chapter 3 (section 3.1), Haitian creole, a radical
creole, coexisted in Haiti with its substratum languages for a period of maybe a
hundred years. Third, as will be extensively discussed in this book, the genesis
of Haitian creole can be argued to have involved relexification as a major process
in its formation. Thus, even though the substratum languages disappeared from
the scene rather quickly, compared to the Solomons Pidgin case, relexification
can be argued to have played a major role in the formation of this radical creole.
2.4

The relationship between the notion of transfer and the process of

relexification in creole genesis and second language acquisition

The notion of transfer (which goes back to Weinreich 1953: 1) refers to the use
of features of their own language by speakers who are acquiring a second language. Substratal features in radical creoles have long been considered as cases
of transfer.6 For example, Andersen (1983a: 7) writes: Transfer from a learners
previously acquired language . . . is assumed to interact with the normal acquisitional process by causing the learner to perceive input in terms of certain aspects
of the structure of the previously acquired language. And, according to Mufwene
(1990: 2):
For the purpose of this article the connection between transfer and substrate
influence is assumed in terms of cause and effect. Transfers apply putatively in
the speech of multilingual speakers and/or at the stage of SLA [= Second Language Acquisition]; substrate influence is observed in a language as a relatively
crystallised system. Once transfers have been replicated by different speakers,
repeated by most of them, and established in the contact situations new linguistic system (even as variable features), they may be characterised genetically
as substrate influence. The latter need not be associated synchronically with
multilingual speakers and/or SLA. Some of the best evidence for substrate
influence has actually been collected from monolingual or monolectal speakers,
especially in creolistics.

Siegel (1995, to appear: 11) makes similar remarks: Here I am defining substrate
influence as the evidence of transfer (or interference) at an earlier stage of development. Transfer refers to speakers unconsciously carrying over features from
one language (usually their first) when speaking (or trying to speak) another language. Allsopp (1980: 95) refers to cases of transfer from substratum languages

34

Cognitive processes involved in creole genesis

as cases of misascription: By this term I mean not only the semantic shifting,
but the ascribing of the wrong sense or the wrong usage or the wrong grammatical function to a word or phrase of the target language. The reader will have
understood that what Allsopp means by wrong grammatical function corresponds to the grammatical function of the speakers native languages. Thus, the
notion of transfer, originally used in second language acquisition, is also used in
the literature on creole genesis to refer to the presence of substratal features in a
radical creole.
In the view advocated in this book, the type of data claimed to be associated
with the notion of transfer in creole genesis corresponds to the result of the
process of relexification, as will be shown throughout. That is, it is claimed that
substratal features are transferred into the creole by means of relexification. Furthermore, relexification accounts for the type of linguistic transfer observed in
creole genesis. Indeed, although not all types of transfer are related to relexification
(see Meisel 1983a, 1983b; Muysken 1981a), the type observed in creole genesis
corresponds to the definition of relexification in (2) (see section 2.1). The claim
that there is a relationship between the notion of transfer as it applies in creole
genesis and the process of relexification goes back to Naro (1978: 337), who
mentions A transfer process of this general sort, termed relexication . . .
So far, we have established that the notion of transfer is pertinent to both
second language acquisition and creole genesis (for a thorough discussion of this
issue, see Meisel 1983b). Given that the genesis of creole languages is claimed
to constitute a particular case of second language acquisition, the fact that transfer is claimed to apply to both situations should not come as a surprise. The
claim has been made, however, that the type of transfer observed in creole
genesis is best characterised as being the product of relexification. This raises
the question of whether this process also plays a role in ordinary cases of second
language acquisition. Mous (1995), Siegel (to appear) and Lumsden (in press a)
argue that, in some cases of second language acquisition, relexification (or paralexification, in Mouss term) may play a role. For example, Lumsden (in press a)
illustrates this claim on the basis of contrastive data from French and English
drawn from Adjemian (1983). While some French verbs require the pronominal
reflexive (e.g. sasseoir to sit, se prparer to prepare, se retirer to withdraw,
se battre to fight, etc.), the (more or less) equivalent English verbs do not
require a reflexive. Lumsden points out that Anglophone Canadian university
students acquiring French as a second language typically make errors of the type
in (19). (Correct forms are given in parentheses.)
(19)

a. Tu assieds sur une chaise. (tassieds)

you sit
on a
chair
You are sitting on a chair.
b. Je vais prparer pour la fte. (me prparer)
I
go
prepare for
the party
Im going to get ready for the party.
(=(12) in Lumsden in press a)

2.5 Relexification and L2 acquisition in creole genesis

35

On the other hand, Francophone Canadian university students acquiring English

typically make errors such as those in (20).
(20)

a. At sixty-five years old they must retire themselves because . . .

b. They want to fight themselves against this (tuition increase).
(=(13) in Lumsden in press a)

The data in (19) and (20) show that the native language lexicon has influenced
the acquisition process. Lumsden proposes that the influence of the native language that is evidenced by these data can be explained in terms of relexification.
In his view, the learner makes a copy of the lexical entry of his native language
and replaces the original phonological form with a phonological form derived
from the second language data. The new lexical entry retains the semantic and
syntactic properties of the original lexical entry that it was copied from. As is
pointed out by Lumsden (in press a), the result of relexification is open to further
revision in the normal course of second language acquisition, so that relexification is only one of the many cognitive strategies and processes second language
learners use to acquire a new language (for discussion of the strategies and
processes involved in second language acquisition, see White 1996; Ritchie and
Bathia 1996b). Because of the social contexts where creoles arise, however,
relexification appears to be a major tool used by creole creators.
2.5

Relexification and second language acquisition in creole genesis

The theory of creole genesis outlined in this section makes two basic claims.
First, the mental process of relexification plays a central role in creole genesis.
Second, in creole genesis, this mental process applies in a situation which involves
second language acquisition. This section begins with a general statement of
how relexification applies in this context. Section 2.5.2 presents a hypothesis
concerning how this mental process applies in the case of minor/functional
category lexical entries. Section 2.5.3 addresses the question of how word order
is established in creole genesis.
2.5.1

Relexification: a means of creating a common lexicon

Relexification applies in creole genesis in the following way (see Lefebvre and
Lumsden 1994a, 1994b). Native speakers of various substratum languages are
exposed to a superstratum language, the language of the colonisers. Crucially,
the speakers of the substratum community do not have a common language, a
situation which creates the need for a lingua franca. Speakers of the substratum
languages relexify the lexical entries of their respective lexicons on the basis of
phonetic strings found in the superstratum language as described in (2). The
relexification of various lexicons on the basis of a single superstratum language
provides the speakers of the substratum languages with a common vocabulary.

36

Cognitive processes involved in creole genesis

It has been claimed that creole genesis is a function of second language

acquisition in the specific context where substratum speakers have reduced access
to the superstratum language. The fact that, as we saw in chapter 1, the amount
of exposure to the superstratum language plays a role in the radicalness of the
creole argues in favour of this claim. As was mentioned in chapter 1, however,
in situations where radical creoles are formed, speakers of the substratum community typically have only limited access to the superstratum language. Lefebvre
and Lumsden (1994a) emphasise that it is this limited direct access to the superstratum language that makes relexification so important in the formation of
these creoles.
As we saw in section 2.2, the formation of mixed languages basically involves
relexification of major category lexical items. Minor category lexical items tend
to be retained from the substratum language. It is uncommon, however, to find
minor category lexical items carried over from a substratum language into a
creole. Such exceptional cases can be found in Saramaccan7 (see Smith 1995,
1996; McWhorter 1996), which has retained the functional lexical item wY8
from Fongbe, one of its African substratum languages, and in Berbice Dutch,
spoken in Guyana (see Kouwenberg 1987; Smith et al. 1987; Robertson 1983,
1993), which has retained a few forms such as the plural marker -apu from
Eastern Ijo (Kalabari), its major West African substratum language. However,
such cases are far from being the general rule in the languages known as
creoles.
It is a well-known fact that, in contrast to mixed languages, in creole languages, both minor and major category lexical entries have phonological representations which are similar to some phonetic string of the superstratum language
(see e.g. Lefebvre 1984). It has long been noted, however, that these functional
items do not have the same properties as the corresponding superstratum
words. Indeed, because speakers of the substratum languages have very limited
access to the superstratum data, they typically fail to identify the functional
categories of the superstratum language (see e.g. Lefebvre 1984; Carden and
Stewart 1988; Mufwene 1991; Lefebvre and Lumsden 1989a, 1992a). Throughout this book, it will be shown that, although the creole phonological forms of
the functional category lexical entries are derived from superstratum phonetic
strings, the properties of the lexical entries so labelled do not correspond to
those of minor category lexical entries in the superstratum language but rather
to those of the lexical entries in the substratum languages. This situation is
claimed to result from relexification. It is thus argued that, in creole genesis,
both minor and major category lexical entries undergo relexification, and that
the creators of a creole relexify the functional categories of their own lexicon on
the basis of phonetic strings from the superstratum language (see Lefebvre and
Lumsden 1989a, 1994a). This provides the speakers of the various substratum
languages with a common vocabulary in this area of the lexicon as well. But
how do the relexifiers go about relexifying their own functional category lexical entries?

2.5 Relexification and L2 acquisition in creole genesis

2.5.2

37

Relexifying functional categories

From the point of view of the relexifiers, relexification consists in identifying, in

the superstratum language, an appropriate phonetic string to provide a phonological representation for a copied lexical entry (see (2) ). For major category
lexical entries, an appropriate form in the superstratum language is a phonetic
string which shares some meaning with the corresponding lexical entry in the
copied lexicon. The meaning of the superstratum form is assumed to be deduced
from its occurrence in semantic and pragmatic contexts (see (2) ). For functional
categories, the process is slightly more complex.
Recall that the creators of a radical creole cannot identify the functional
categories of the superstratum language because they do not have enough exposure to the language. Consequently, relexifiers do not relexify the functional
category lexical entries of their own lexicon on the basis of those of the superstratum language. In Lefebvre and Lumsden (1994a), it is proposed that the functional category lexical entries of the copied lexicon are relabelled on the basis
of major category lexical items (e.g. nouns, adjectives, verbs, adverbs and prepositions) of the superstratum language. For example, in Haitian creole the
copied lexical entry of a substratum language determiner has been relabelled on
the basis of a French adverb (see chapter 4). Furthermore, Lefebvre and Lumsden
(1994b) propose that relabelling of a functional category lexical entry responds
to three types of clues. First, as is the case for major category lexical entries,
there must be some semantic overlap between the lexical entry copied from the
substratum lexicon and the superstratum form (see (2) ). Hence, only those functional category lexical entries which have some semantic content can be assigned
a new phonological form at relabelling. Functional category lexical entries without semantic content are assigned a null form at relabelling (see above). Second,
the distributional properties of the superstratum form must be similar to those of
the copied lexical entry. For example, a postnominal determiner may be relabelled
on the basis of a superstratum form which also occurs postnominally. Relabelling
of functional category lexical entries is thus constrained by what the superstratum
language has to offer in terms of major category lexical items whose semantics
and distribution are appropriate to provide a phonetic matrix for a copied functional category lexical entry. The limits imposed by the superstratum lexicon
predict that a given subset of substratum lexicons relexified using data from
different superstratum languages may end up with slightly different inventories
of overt functional categories. Such cases will be discussed in the data chapters
of this book. Third, relabelling of a functional category lexical entry may also
respond to a phonological similarity between the substratum and superstratum
lexical entries. This is the phenomenon of phonological conflation discussed
by Kihm (1989, 1994). In other words, phonetic similarity of the superstratum
string and the substratum lexical entry may trigger relabelling. For example,
as is shown in Kihm (1989), where Manjaku, a West African language, has the
forms dika, an unaccomplished negation marker, and kats, a negative auxiliary

38

Cognitive processes involved in creole genesis

meaning no longer, and Portuguese has nunca never, Kriol a West African
Portuguese creole has ka, a sentential negative marker (see Kihm 1989). (See
also Mhlhusler 1986a, for several examples of conflated forms in Tok Pisin.)
What happens if speakers of the substratum languages do not find any appropriate string in the superstratum language to relabel a lexical entry copied from
their own lexicon? One possible option is to abandon the lexical entry. Another
option discussed in Lefebvre and Lumsden (1992a, 1994b) is to assign the
copied functional category lexical entry a phonologically null form (represented
as in the schema in (2) ). That is, when this item is used in a sentence, it is not
pronounced. The difference between the two options is visible in the syntax of
the creole as compared with that of its substratum languages. As will be seen
in the data chapters, a substratum lexical entry is considered not to have been
reproduced if nothing signals its presence in the syntax of the creole. In contrast,
a substratum lexical entry is considered to have been assigned a null form in
relexification if it is visible in the syntax of the creole. An example of such
a case will be discussed in chapter 5. Furthermore, as has been pointed out in
Lefebvre and Lumsden (1992a, 1994b), a functional category lexical entry that
has been assigned a null form at relabelling may be signalled by a periphrastic
expression. For example, a lexical entry having a temporal/aspectual meaning
but a null phonological representation may be signalled by the use of an adverb
with a similar meaning. As will be discussed in section 2.6, the periphrastic
expression may later become the phonological representation of the lexical entry
initially assigned a null form, through the process of reanalysis. The possibility
of assigning a null form to a copied functional lexical item will be shown to be
most relevant for the discussion, in section 2.6, of how the substratum lexicons
may have an input into the creole even after they are no longer in use in the early
creole community.
2.5.3

Establishing word order

The problem of how word order is established in creole genesis has been raised
several times in the literature. For example, in an article appropriately entitled
Bonnet blanc and blanc bonnet, Mhlhusler (1986b: 47) addresses the problem of adjectivenoun order in Tok Pisin: Whereas colour adjectives follow the
noun categorically in Tolai,9 at no point in the history of Tok Pisin has this been
the case. In Tok Pisin, colour adjectives have always preceded the noun.
Mhlhuslers conclusion is thus that the word order adjectivenoun observed in
Tok Pisin was not established on the basis of substratum language data. Similarly,
Mufwene (1990: 5) addresses the word order problem on the basis of Haitian
data: the Haitian creole question phrase ki mun who (lit. what person) is
not an exact match to the Fon10 m^ tX (lit. person what); the constituent order
is not the same and calls for an explanation. In both examples, the word order
observed in the pidgin/creole is opposite to that found in at least one of the
substratum languages. It should be noted, however, that both these examples

2.5 Relexification and L2 acquisition in creole genesis

39

involve major category lexical items (adjectives and nouns). Examples involving
minor category lexical items such as determiners show a different pattern. For
example, the deictic determiner ia in Solomons Pidgin (and Tok Pisin) is postnominal, as shown in (21).
(21)

Blad ia
i gud.
blood dem it be good
This blood is good.

solomons pidgin
(from Keesing 1988: 102)

By contrast, in English, the lexifier language of Solomons Pidgin, there is a

prenominal deictic determiner, as can be seen in the translation of (21), showing
that in this case the word order in the pidgin does not follow that of the superstratum language. Given the above data on word order and lexical categories,
why should the word order in (21) be as it is? The data examined so far suggest
that, in pidgin and creole genesis, word order may be established differently
for major and minor category lexical items. It will be shown below that this is
indeed the case.
The theory of word order underlying this research is that presented in Travis
(1984), Koopman (1984) and Lieber (1992). According to this theory, heads in
particular languages are marked for the directionality properties of their specifier,
modifier(s) and complement.11 Building on this general theoretical approach to
word order, Lefebvre and Lumsden (1992a) make a twofold proposal as to how
word order is established in creole genesis. First, because the relexifiers intend
to reproduce the phonetic strings of the superstratum language an assumption
that follows logically from the claim that creole genesis is a function of second
language acquisition and because they identify major category lexical items of
the superstratum language, they acquire the directionality properties of the superstratum major category lexical entries. Thus, the word order of major category
lexical entries in the creole is predicted to follow the word order of lexical categories in the superstratum language. Consequently, if the superstratum language
has prenominal adjectives, the creole will have prenominal adjectives. In this
view, colour adjectives occur prenominally in Tok Pisin because this is where
they occur in English, its superstratum language. Similarly, the adjective ki
which in the Haitian expression above occurs prenominally because the corresponding French adjective occurs in this position (see chapter 6 for an extensive
discussion of these data). Likewise, creoles whose lexifier languages have prepositions but no postpositions are predicted to have only prepositions. This is the
case, for example, in English-based creoles of the Caribbean, such as Jamaican
creole, which have prepositions in spite of the fact that their West African substratum languages also have postpositions. By contrast, creoles whose lexifier
languages have postpositions are expected to have postpositions. This is the case,
for example, with Dutch-based creoles. Dutch has both pre- and postpositions
(see Van Riemsdijk 1978), and so does Berbice Dutch (see Smith et al. 1987;
Kouwenberg 1992; Robertson 1993).

40

Cognitive processes involved in creole genesis

On the other hand, as pointed out earlier, the creators of a radical creole
do not identify the functional categories of the superstratum language because
they do not have enough access to the superstratum language to identify these
lexical entries as such. Lefebvre and Lumsden (1992a) hypothesise that the
creators of the creole retain the directionality properties of the functional category lexical entries of their own lexicon in relexification. Consequently, the
creole functional categories will have the same word order as the substratum
entries that they were copied from. According to this proposal, the position of
the deictic determiner in Solomons Pidgin is expected to follow the order of its
substratum languages. This prediction is borne out by the data (see (21) ). For
example, Toabaita, a language of Malaita (Solomon Islands) and one of the
Solomons Pidgin substratum languages, has a postnominal deictic determiner as
shown in (22).12
(22)

Koro kasi-a ai
we
cut
tree
We cut this tree.

baa.
dem

toabaita
(from Lichtenberk 1984)

Thus, assuming that the position of specifiers, modifiers and complements

with respect to the head (left or right) is determined by the directionality properties of heads in particular languages (see Travis 1984; Koopman 1984; Lieber
1992), in creating a creole, speakers of the substratum languages will adopt the
directionality properties of the superstratum lexical heads and retain the directionality properties of the functional heads of their native lexicons. The proposal
made in this section will be shown to be borne out by the data throughout
this book.

2.5.4

Summary

The account of creole genesis developed in the course of this research advocates
that relexification plays a central role in this process. In the theoretical framework adopted here (see chapter 1), most of the information that distinguishes
particular grammars is registered in the lexicon. Consequently, as has been
pointed out in Lefebvre and Lumsden (1994b), relexification is central to creole
genesis because most properties of specific languages are stored in the lexicon.
Furthermore, as will be seen below, it is the output of relexification that feeds
the processes of reanalysis and dialect levelling.
The lexicons created by relexification become the basis of a lingua franca
within the creole community. As is pointed out in Lefebvre and Lumsden (1994b),
at this point, the speakers are no longer targeting the superstratum language.
They are now targeting the common language that they have developed through
relexification. At this stage, two other processes come into play: reanalysis and
dialect levelling.

2.6 Reanalysis
2.6

41

Reanalysis

This section discusses the role of reanalysis a mental process by which the
phonological form of one lexical entry becomes the phonological form of
another lexical entry in linguistic change in general and the development
of pidgin and creole languages in particular. I start with a discussion of the
process of reanalysis, sometimes referred to in the literature as desemanticisation or grammaticalisation. I survey the evidence showing that reanalysis is a
major process involved in regular cases of linguistic change. Cases of reanalysis
drawn from West African languages and from various creole languages will
further illustrate the types of changes produced and show that cases of reanalysis
reported on in the literature on pidgins and creoles are of the same type as those
observed in other languages. These case studies show that reanalysis is an
important process in the evolution of pidgin and creole languages and that, in
this context, it operates similarly to the way it does in other languages. A major
point addressed in this section is the paradox often alluded to in the literature
on pidgins and creole languages concerning the observation that changes occurring in these languages make them more similar to their substratum languages,
even in communities where the latter are no longer spoken and the creole has
become the sole means of communication. It is proposed that we can resolve
this paradox based on the interaction between relexification and reanalysis in
the development of pidgins and creoles.
2.6.1

Reanalysis and related phenomena

Reanalysis is a mental process by which a particular form which signals one

lexical entry becomes the signal of another lexical entry (Lightfoot 1979). A
classic example documented by Lightfoot (1979) is the reanalysis of English
verbs such as must and may as modals. Another example is the reanalysis of
verbs as adverbs. For example, the Yoruba verbal expression sa ere run race
has been reanalysed as sere quickly (see B0gb- 1974; Lord 1976).
Through reanalysis, a phonological form which signals a major category
lexical entry may become the signal of a minor category lexical entry. For
example, the preposition of in English has been reanalysed as a Case marker
(see Chomsky 1981). In the literature, such cases of reanalysis are sometimes
referred to as cases of desemanticisation or grammaticalisation. For example,
the term desemanticisation is defined as follows in Heine and Reh (1984: 36):
By this process, a lexical item receives a second, non-lexical function,
which may ultimately become its own function. Thus, in addition to its lexical
meaning, a word receives a grammatical function and can eventually develop
into a grammatical morpheme. In many African languages, for example, a
verb like finish was desemanticised to an aspect, i.e. perfect marker, or a
verb meaning say became an object clause complementiser . . . as a result of
Desemanticisation.

42

Cognitive processes involved in creole genesis

The term grammaticalisation has been given several related definitions. For
example, Hopper and Traugott (1993: 2) define it as follows: Grammaticalisation
is usually thought of as that subset of linguistic changes through which a lexical
item in certain uses becomes a grammatical item, or through which a grammatical item becomes more grammatical. Heine and Reh (1984: 96) provide the
following definition: Grammaticalisation is associated with specific evolutional
processes involving in particular a decrease in semantic and phonetic complexity, pragmatic significance, and syntagmatic variability. In fact, desemanticisation and grammaticalisation both refer to a situation where a content word
has come to assume the grammatical properties of a function word. The authors
cited above agree that grammaticalisation and reanalysis are related. Hopper
and Traugott (1993: 32) assert that reanalysis is the most important mechanism
for grammaticalisation, as for all change. Heine and Reh (1984: 97) point out
that grammaticalisation and reanalysis show some striking similarities. Both
involve a change in the syntactic category of a given lexical item. The difference between the two lies in the fact that not all cases of reanalysis lead to
grammaticalisation. For example, the reanalysis of a verb as a preposition entails
a categorial switch from [n, +v] to [n, v]. It does not entail a switch from
a major to a minor category such as tense, aspect, etc. On the definition of
reanalysis adopted above, however, the process is neutral with respect to output.
Reanalysis may assign a particular form to a new lexical entry which is either a
major or a minor category entry. Whatever the output, the process is the same: it
consists in assigning the phonological form of a given lexical entry to another
lexical entry.
Thus, in all these cases, reanalysis is the relevant mental process. It may or
may not involve desemanticisation; it may or may not involve grammaticalisation.
But there is a single mental process at work, reassigning a particular form from
one lexical entry to another. In the discussion below, cases referred to in the
literature as desemanticisation or as grammaticalisation will thus be considered
as regular cases of reanalysis.
2.6.2

Reanalysis: a major process in linguistic change

Reanalysis has been shown to play a major role in linguistic change. Lightfoot
(1979) discusses several cases in languages from various families. The literature on West African languages abounds in examples of reanalysis (sometimes referred to as desemanticisation), as is extensively discussed in Heine
and Reh (1984 and the references therein). The literature on pidgin and creole
languages also reports on a number of cases of reanalysis (sometimes referred
to as grammaticalisation) (see e.g. Sankoff and Laberge 1973; Washabaugh
1975; Valdman and Highfield 1980; Koopman and Lefebvre 1981; Lefebvre
1984; Mhlhusler 1986a; Rickford 1987; Bickerton 1988; Foley 1988; Muysken
1988b; Romaine 1988; Sankoff 1990, 1991; Plag 1993; Baker and Syea 1996, to
name but a few). Cases of reanalysis reported on in the literature on pidgins

2.6 Reanalysis

43

and creoles are of the same type as those observed in other languages. I will
illustrate this point by comparing a few such examples with cases from West
African languages.
One example involves the reanalysis of verbs as prepositions. Such a case is
argued for in Lord (1976: 182) on the basis of data from Ewe, where the verb
n give has been reanalysed as a dative benefactive preposition meaning for.
A similar case of reanalysis is reported on by Dchaine (1988: 40), who argues
that, in modern Haitian creole, the lexical item ba(y) has the double status of a
verb meaning give and a benefactive preposition meaning for. The second
use of ba(y) is claimed to be the result of reanalysis.
A second example consists in the reanalysis of a verb as an aspectual marker.
For example, Westerman (1907: 139) reports that, in the Dahomeyan dialect of
Ewe which I take to be a Gbe dialect the verb nW remain, stay was reanalysed
as a habitual marker, as shown in (23).
(23)

N nw
s.
I hab sell
I habitually sell.

ewe
(from Westerman 1907: 139)

Similarly, in Tok Pisin, the verb save know has been reanalysed as an iterativehabitual preverbal marker as shown in Sankoff (1991: 70). The data in (24) illustrate this use.
(24)

Na taim san em i
no
save go
and time sun he pr neg hab go
And in the daytime he wouldnt go eat.

kaikai.
eat

tok pisin

(=(25a) in Sankoff 1991)

Another example concerns the reanalysis of a sentence-initial adverb of posteriority as a preverbal marker. For example, Heine and Reh (1984: 120) report
that, in Bari, a West African language, the adverb d then, afterwards became
a future or irrealis preverbal marker. As has been observed by Spagnolo (1933:
1056), in its adverbial function, d occurs sentence-initially; as a future marker,
it occurs between the subject and the verb, as shown in (25).
(25)

a. D
nn kwn . . .
then I
do
Then I do . . .

b. Nn d
kwn . . .
bari
I
fut do
I shall do . . .
(from Heine and Reh 1984: 120)

Similar cases are reported for Sango (see Samarin 1967: 154), Klao, a Kru
language of Liberia (see Singler 1979: 25). In Tok Pisin, the sentence-initial
adverbial baimbai later, afterwards has been reanalysed as an aspectual marker
(see Sankoff and Laberge 1973; Sankoff 1991). In its adverbial function, baimbai
occurs sentence-initially; as an aspectual marker, it occurs in its reduced form
between the subject and the verb, as shown in (26).

44
(26)

Cognitive processes involved in creole genesis

a. Baimbai ologeta wok bilong yumi, i
later
all
work of
us
pr
kamap.
grow
Later, all our work will (be able to) flourish.

ken
fut-possib

tok pisin

(=(2) in Sankoff 1991)

b. Mi bai kisim.
I
fut get
Ill get it.

tok pisin
(=(9) in Sankoff 1991)

The Tok Pisin example in (26) parallels the Bari example in (25).
These examples show that the changes resulting from reanalysis in the development of pidgins and creoles are of the same type as those observed in other
languages. There is, however, something distinctive about the lexical entries to
which reanalysis is hypothesised to apply in the development of pidgins and
creoles. This is the topic of the next section.
2.6.3

Lexical entries to which reanalysis applies in the development of

pidgins/creoles

In the recent literature on changes that occur during the development of pidgin/
creole languages, it has been observed that these changes often yield the paradoxical situation described by Sankoff (1991: 73): The changes ongoing in the
Tok Pisin tense and aspect system are making it more like many of the substrate languages. Similarly, Mhlhusler (1986a, 1986b) points out in several
instances that substratum languages may influence a creole after its formative
stage, even in situations where there is no more contact between the creole and
its substratum languages. For example, he asserts that similarities in the grammars of a pidgin/creole and its alleged substratum language could have arisen at
a time when there were no linguistic contacts (1986a: 41). Mufwene (1990),
Chaudenson (1994) and others make similar observations. This paradoxical situation calls for an explanation.
The paradox can be accounted for straightforwardly by the proposal that,
in the early creole, reanalysis applies to lexical entries that have been created
through relexification. Lefebvre and Lumsden (1994b) propose the following
relationship between relexification and reanalysis in the development of a pidgin
or creole. A substratum functional category lexical entry is copied and assigned
a phonologically null form at relabelling because the creators of the creole did
not find an appropriate form in the superstratum language to relabel the copied
lexical entry or because it had no semantic content and thus was assigned a null
form at relabelling. Practically speaking, this means that, when speakers of the
early creole have to use this copied lexical entry in particular utterances, they do
not pronounce it. It is hypothesised that speakers of the early creole use a major
category lexical item (e.g. an adverb) to signal the covert functional category

2.6 Reanalysis

45

lexical entry. Speakers may eventually assign the form of this major category
lexical item as the phonological form of the previously covert functional category
through the process of reanalysis. In this view, reanalysis provides a phonological form for a creole lexical entry that was generated by copying but was not
assigned a phonological form at relabelling. When this type of reanalysis takes
place, the creole lexical entry has the properties of the substratum lexical entry
that it was copied from.
In this view, then, a relexified lexical entry that has been assigned a null form
at relabelling may acquire a phonological form through reanalysis. This interaction between relexification and reanalysis accounts for the fact that, in the early
stages of the creole,13 substratum languages may influence the developing creole
even when they are no longer spoken in the creole community. Furthermore, it
also accounts for the fact that reanalysis on the basis of similar lexical items
across languages may yield different results. This can be observed from two
cases both involving an adverb of posteriority reanalysed as the phonological
form of a functional category lexical entry in two different languages. As is
documented in detail in Sankoff (1991: 72), the Tok Pisin adverb of posteriority
bai(mbai) has undergone a shift in meaning from future through irrealis to
iterative and punctual. As was pointed out above, she claims that the ongoing
changes in the Tense and Aspect system of Tok Pisin are making it more like the
substratum languages, in which tense is quite subsidiary to aspect. Sankoff
(1991: 73) explains this paradoxical situation as follows:
Despite the fact that particles like bai and bin were originally more temporal
than aspectual, they were never used with anything like the regularity that
would present to learners the hypothesis that tense was being marked. There
was no evidence for tense anywhere in the system as it was used by several
generations of second-language Tok Pisin speakers over the past hundred
years. What they did mark was aspect, a category more important than tense in
the grammars of the languages they spoke natively. There always has been a
stronger relationship between the use of these markers for aspect than for tense
in the speech of adults, and this is, I propose, what the native speakers are
building on in carrying forth more sweeping changes than their parents were
able to do.

By contrast, in the Bari example above (see (25) ), the adverb of posteriority d
has been reanalysed as a future/irrealis, i.e. a Mood preverbal marker, following the general pattern in West African languages. As we will see in chapter 5,
the Haitian adverb of posteriority has been reanalysed as a future/irrealis preverbal marker, following the pattern of the substratum languages of Haitian.
These facts argue for the claim in Lefebvre and Lumsden (1994b) that, in creole
genesis, reanalysis applies to a lexical entry that has been produced by copying
but assigned a null form at relabelling, such that, when, for example, an adverb
of posteriority is reanalysed as the phonological form of a preverbal marker, that
preverbal marker has the properties of the substratum lexical entry that it was
copied from.

46
2.7

Cognitive processes involved in creole genesis

Dialect levelling

Dialect levelling, as discussed in the literature on dialects in contact (see e.g.

Domingue 1980, 1981; Trudgill 1986; Siegel 1995, to appear) refers to the
reduction of variation between dialects of the same language in situations where
these dialects are brought together. As Siegel (to appear: 21) puts it, dialect
differences are reduced as speakers acquire features from other varieties as well
as avoid features from their own variety that are somehow different. This may
occur over several generations until a stable compromise dialect develops. Welldocumented cases of dialect levelling include Bhojpuri as spoken in Mauritius
(Domingue 1980, 1981) and English as spoken outside of England (Trudgill
1986; Siegel to appear).
In the literature on pidgins and creoles, it has been suggested that dialect
levelling plays a role in the development of these languages (see e.g. Mhlhusler
1980: 34; Mufwene 1990: 1389, 1996: 22, 1994; Harris 1991: 199). For example,
Siegel (to appear: 26) asserts that: Mixing and levelling may . . . be important in
the development of stable pidgin and creole languages . . . When the various
versions of the superstrate are then used as the main means of communication
among speakers of different substrate languages (in other words, when vernacularisation occurs), and when these speakers begin to view themselves somehow as a community, then levelling begins. The scenario of creole genesis
developed in the course of this research also includes dialect levelling (see
Lumsden and Lefebvre 1994). The originality of our proposal, however, lies in
the claim that, in this case, dialect levelling reduces the variation between the
lexicons produced by the relexification of the various substratum lexicons. It is
thus hypothesised that, in the context of creole genesis, dialect levelling operates
on the output of relexification.
Typically, situations where pidgins and creoles emerge involve several substratum languages (see chapter 1). Each individual relexifies his or her own
lexicon. Hence, as is pointed out in Lumsden and Lefebvre (1994), speakers
of various substratum languages reproduce the idiosyncratic semantic and syntactic properties of their own lexicons in relexification and thus the product of
relexification is not uniform across the creole community. The features that are
common to all the relexified lexicons (that is, to all substratum languages) will
most probably be maintained in the creole.14 The idiosyncratic features, however, are those that are subject to levelling.15 It is hypothesised that, when the
speakers of the creole community stop targeting the lexifier language and start
targeting the relexified lexicons, that is, the early creole, they begin levelling out
the differences between the relexified lexicons. As is pointed out in Lumsden
and Lefebvre (1994), different creole communities may arrive at different compromises. This may be observed both in geographical dialects of a single creole
and in differences among creoles formed on the basis of a similar pool of languages. For example, as will be shown in chapter 3 and in the data chapters,
the Northern dialect of Haiti is the product of different compromises from the

2.8 Parameter setting, semantics, principles of concatenation

47

Central and Southern ones. Similarly, both Haitian and Martinican creoles arose
in historical situations involving basically the same languages (see Singler 1993b).
Nonetheless, these two languages present lexical differences (see e.g. Goodman
1964), showing that dialect levelling did not produce exactly the same results in
these two communities.
Cases of dialect levelling operating on the basis of relexified lexicons will be
presented throughout this book. However, it will also be shown that idiosyncratic
features need not result in compromise within a given creole community, for
some cases of variation within a creole can be argued to correspond to variation
between its substratum languages, and even between dialects of a single substratum language (see e.g. chapters 6 and 8).
2.8

Parameter setting, semantic interpretation and principles of

concatenation

Relexification and reanalysis, as discussed above, are processes which apply

to the lexical component of the grammar. This component is not, however, the
only one in the model of grammar adopted for this research (see chapter 1). For
example, parameters are also part of this model. Furthermore, there are semantic
interpretation data associated with particular constructions (see e.g. cleft constructions), which are language-specific. And, finally, there are principles of
concatenation of morphemes and words which are also language-specific.
The general view advocated in this book is that the creators of a creole use
not only the lexical properties of their native lexicons, but also the other grammatical and semantic properties of their native grammars in establishing those
of the creole. It is thus hypothesised that, in setting the parametric values of the
new language, the creators of the creole use those of their own grammars. As
will be seen in chapter 12, this claim is, to a great extent, borne out by the data.
Likewise, it is hypothesised that the creators of the creole will use the semantic
interpretation rules of their own grammars to establish those of the creole. This
is also borne out by the data, as will be seen in various chapters. Finally, it is
hypothesised that the creators of the creole use the principles of morpheme and
word concatenation of their own grammars in developing the creole. This claim
is borne out by the data in chapter 10 on derivational affixes and chapter 11 on
compounds. This explains why, in these areas of the grammar as well, creole
languages follow the patterns of their substratum languages, rather than those of
their superstratum languages. These aspects of the grammar are also subject to
dialect levelling, as will be discussed in several chapters.
2.9

An optimal account of creole genesis

The hypothesis of creole genesis outlined in this chapter provides a straightforward and optimal account of all the properties of creole languages discussed

48

Cognitive processes involved in creole genesis

in chapter 1. The following discussion builds on a preliminary one in Lefebvre

and Lumsden (1989a, 1994a).
First, by virtue of the definition of the process, creole lexical entries are
predicted to have the same semantic and syntactic properties as the corresponding lexical entries in the substratum languages, but phonological representations
derived from the phonetic strings of the superstratum language. The hypothesis
thus explains why creoles reflect the properties of both their superstratum and
their substratum source languages in the way they do (see e.g. Sylvain 1936;
Goodman 1964; Huttar 1971; Keesing 1988). As was pointed out in Lefebvre
and Lumsden (1989a), the hypothesis that relexification plays a central role in
creole genesis is falsifiable. If a comparison of the lexical properties of a radical
creole with the lexical properties of its source languages were to show that the
syntax and semantics of the creole are not systematically parallel to the syntax and semantics of the substratum languages, then the hypothesis would be
falsified. The test of this hypothesis, however, must involve quantity as well as
quality of data (see Lefebvre and Lumsden 1989a, 1994a). One or two examples
either way are not enough to falsify or support the hypothesis. The real test
involves a global comparison of the lexicons of these languages (see chapter 3
for an extensive discussion of this point).
Likewise, the hypothesis that the creators of the creole use the parametric
values, semantic interpretation rules and principles of concatenation of their own
grammars in creating the creole explains why creoles pair with their substratum
languages in these areas of the grammar as well. Again, this hypothesis is falsifiable. If a comparison of the grammatical and semantic properties of a radical
creole with those of its source languages were to show that the properties of the
creole are not systematically parallel to those of the substratum languages, then
the hypothesis would be falsified. Again, the test must be based on quantity as
well as quality. A true test must involve the global comparison of the parametric
values, semantic interpretation rules and principles of concatenation of the languages involved.
The second main point is that, given the multilingual situation prevailing
in contexts where creoles emerge (see Whinnom 1971), and given the urgent
need for a lingua franca in such a situation (see e.g. Hymes 1971a; Foley 1988;
Thomason and Kaufman 1991) and, furthermore, given the limited access that
the bulk of the population has to the superstratum language (see e.g. Thomason
and Kaufman 1991; Foley 1988), only one generation of speakers is required to
form a new language by means of relexification and the use of the parametric
values and other principles of the native grammars. The claim that these processes are at work in creole genesis accounts for the fact that creole languages
can be created relatively quickly as compared with regular cases of linguistic
change (see e.g. Voorhoeve 1973; Hancock 1987).
Finally, as is observed in Lefebvre and Lumsden (1994a), the fact that creoles
are generally isolating languages also follows from the above proposal. Since
the minor category lexical entries of creole languages derive their phonological

2.10 Conclusion

49

forms from major category lexical items in the superstratum language (see section 2.5), or from reanalysis (see section 2.6) and, since these categories are
typically free morphemes, it follows that creoles will tend to be isolating languages. Mufwene (1989: 124) accounts for the isolating character of creole languages by appealing to the notion of salience: With regard to the issue made
here, viz., explaining why periphrasis is generally preferred to inflections in
pcs, I submit that salience should do. The proposal advocated in this research
is somewhat similar to Mufwenes, for major categories may be viewed as
salient when compared with minor categories. The proposal made here is,
however, more specific, since it links the observed facts to the processes that
generate them, namely relexification and reanalysis, and to the linguistic material to which they apply in creole genesis (see section 2.5.3).

2.10

Conclusion

The account of creole genesis presented in this chapter reduces to three processes otherwise observed to play a role in language genesis and language
change in general: relexification, reanalysis and dialect levelling. Along with the
retention of the parametric values, semantic interpretation rules and principles
of concatenation of the original grammars, these three processes are claimed
to account for the internal and external properties of creole languages. Several
researchers have challenged the proponents of theories advocating significant
input by the substratum languages into a creole to explicitly state how this
happens. For example, Bickerton (1992: 314) writes: In order to make a case,
they have to describe exactly and explicitly how, in creolisation, syntactic structures got from substratum languages into creole languages. Our proposal does
just that. Relexification, together with parameter setting, accounts for the fact
that substratum features are found in all components of the grammar of a creole,
contrary to Mhlhusler (1980: 36), who claims that the syntax and derivational
morphology of pidgins and creoles are relatively independent of substratum (or
superstratum) influence. More specifically, relexification accounts for the fact
that a creole lexicon reflects the properties of its substratum lexicons not only
for major but also for minor category lexical entries (see chapters 4, 5, 6, 7, 8),
derivational affixes (see chapter 10), idiosyncratic compounds (see chapter 11),
etc. Furthermore, the claim that the creators of a creole use the parametric values
of their own grammars accounts for the similarity between the creole and its
substratum languages with respect to parametric choices (see chapter 12).
The interaction between reanalysis and relexification discussed in section 2.6
accounts for the fact that the substratum languages may have input into the
creole even after they are no longer spoken in a community where the creole has
become the sole means of communication. This provides an explicit answer to
the question posed by Mufwene (1990: 9): Is there a particular stage in the
development of a creole when substrate influence may (or may not) occur?

50

Cognitive processes involved in creole genesis

According to the view advocated in this book, substratum languages may influence the creole in both its initial and developmental stages.
Finally, the proposal concerning how dialect levelling operates in the development of creole languages allows for a sound explanation of the facts referred
to in the literature on creole studies as the cafeteria principle a term used first
by Dillard (1970) and later by Bickerton. As Bickerton (1981: 49) puts it: As
things stand, we are asked to believe that different African languages contributed different rules and features to particular creoles . . . it is . . . absurd to suppose that a creole could mix fragments of Yoruba, Akan, Igbo, Mandinka, and
Wolof. The proposal that dialect levelling operates on the output of the various
relexified lexicons involved in creole formation provides a principled explanation of the observation that several different substratum languages may contribute features to a given creole. Indeed, in the competition among different creole
dialects (created by the relexification of different substratum lexicons), there are
winners and losers. As will be seen throughout this book, the competition is not
always won by speakers of the same relexified lexicon. This provides a straightforward explanation of why and how features of various substratum languages
end up in a given creole.
The three processes claimed to play a role in creole genesis account simply
and concisely for the long-standing claim that pidgin and creole formation
involves mixing and reduction followed by expansion (see chapter 1). First,
relexification, as schematised in (2), accounts for the type of mix we find in
pidgins and creoles with respect to the contribution of the substratum and superstratum languages. Dialect levelling, hypothesised to apply in creole genesis as
described in section 2.7, accounts for the contributions of the various substratum
languages to a given creole. The possibility of assigning a null form to a copied
functional category lexical entry (see section 2.5.2) may be related to what has
traditionally been referred to as simplification. That is, if a functional category
lexical item is not pronounced in an early creole sentence, the utterance looks
simpler than equivalent utterances in the contributing languages (see Lumsden
1995). Furthermore, this proposal accounts for the directionality of the simplification involved in pidgin and creole formation. As was observed by Brousseau,
Filipovich and Lefebvre (1989), based on the relexification account of creole
genesis, the notion of simplification ought to be viewed from the point of
view of the substratum languages rather than of the superstratum language.
Finally, the process of reanalysis may be related to the notion of expansion
(see e.g. Hymes 1971a, 1971b; Hancock 1980b; Mhlhusler 1986a). On the
analysis provided in section 2.6, this process accounts for the visibility of previously covert lexical entries. Indeed, reanalysis, as it is hypothesised to apply in
the early creole, provides a phonological form for lexical entries that were not
assigned one at relabelling. Hence, functional category lexical items which are
null at the time of relexification acquire a phonological representation in the
period when the creole develops. On the surface, this looks as though the creole
has expanded.

2.10 Conclusion

51

When the relexified lexicons become the target of the creole community, a
new language is born. It is assumed that this new language evolves like other
natural languages. Although innovation was not discussed as a separate section
in this chapter because it is not a mental process in itself it is assumed that,
once a creole is formed, its speakers can innovate in developing it, using the
same processes that are observed in the development of other natural languages.
Such cases will be pointed out in the various data chapters of this book.

52

The research methodology

The research methodology

This research was designed to test how the processes hypothesised to play a role
in creole genesis work on the basis of Haitian creole (see Lefebvre 1986, 1993a),
the classic example of a radical creole (see Bickerton 1984). Since any account
of the genesis of Haitian creole must be compatible with the history of Haiti, the
research included a historical study. The first section of this chapter is dedicated
to the major findings of this study, designed to establish when Haitian creole
was formed and the characteristics of the Haitian population during that period.
Section 3.2 discusses the typological features of the languages spoken in Haiti
during the period when the creole was formed, i.e. West African languages and
French. Section 3.3 considers what French data the creators of Haitian were
exposed to. The linguistic test of the hypothesis rests on a detailed comparison
of the lexical entries of the creole with those of its contributing languages. Section 3.4 describes the linguistic test we were able to do within the limits of the
project research and section 3.5 discusses how the theory can be falsified. Section 3.6 provides information on the data on which the analyses presented in this
book are based. Section 3.7 introduces how the data and analyses are presented.
3.1

The economy, demography and linguistic diversity of early Haiti:

16591740

The careful and abundant work by historians on the slave trade and the ethnic
composition of particular colonies does not provide precise enough information on the external factors that prevailed in the early history of Haiti for linguists to determine when Haitian creole was formed and who was there at the
relevant time (see Filipovich 1986; Lefebvre 1993a). As Singler (1993b: 171)
has remarked, linguists and historians dont have the same interests and dont
ask the same questions. For example, in history books, the ethnic and linguistic
origins of the African slaves are often assumed on the basis of the African ports
from which they were sent to the Caribbean rather than their actual origins (see
Singler 1986, for an extensive discussion of this point). For linguists seeking
to construct a theory of creole genesis which is compatible with the context
in which the creole was formed, who need to study the relevant substratum
languages for the relevant period, this gap in the historical research represents a
major problem. It therefore proved necessary for our research project to carry
out its own historical study.
52

3.1 Demography and linguistic diversity of early Haiti

53

Table 3.1. Comparison of Haiti after 22 years (1681) and Martinique

after 25 years (1660)

%
%
%
%
%

of
of
of
of
of

people of colour
women among adult Africans
children among Africans
engags among engags + pop. of colour
mixed race in population of colour

Haiti

Martinique

34.8
40.5
14.9
40.4
c. 6.0

51.5
45.7a
24.5
12.0a
c. 1.0b

Figures for women and engags in Martinique come from 29 years (1664)
and 34 years (1669), respectively.
b
Figures for mixed race are extrapolated from the combined category of
mixed race and Caribs.
Sources: Haiti: ansom = Archives Nationales, Section Outre-Mer (Aix-enProvence) g1 509, no. 2. Martinique: 1660, Petit Jean Roget (1980: 1376)
and Petit Jean Roget and Bruneau-Latouche (1983: 137); 1664, ansom g1
470, nos. 18; and 1669, ansom g1 499, no. 2.
(Table 2 in Singler 1996)

The historical research was designed to answer the following questions.1

When was Haitian creole formed? What were the salient demographic characteristics of the Haitian population during that period? Who were the people present
at the relevant time? What was their linguistic background (see Lefebvre 1993a)?
The historical research in the colonial archives of France was carried out by
John Singler (see Singler 1993a, 1993b, 1996).
Following Heinl and Heinl (1978: 17), Singler assumes 1659 to be the starting date of the French colony in Haiti, and he provides evidence showing that
Haitian creole was probably created between this date and 1740.2 The beginning
of this period was marked by a shift from a tobacco and cotton economy to a
sugar economy. This shift created a favourable context for a creole language
to emerge.
As indicated above, early Caribbean society was characterised by a tobacco
and cotton economy. As Singler (1996: 193) puts it: The initial crops of Caribbean colonies were ones like tobacco and cotton. Where these crops were grown,
there were large and small agricultural units alike, large and small slave holdings and a labour force that was a mixture of colonists, engags3 and slaves. In
this type of economy, the Haitian population was made up of the colonists,
people of colour, engags and people of mixed race. As can be seen in Table 3.1
(=Table 2 in Singler 1996), the people of colour and the engags were almost
equal in numbers.
Drawing on Butel (1982: 98), Singler estimates that the shift from tobacco
and cotton to a sugar economy in Haiti took place around 1690. This shift
provoked a radical change in the configuration of the early Haitian population:
the large landowners drove out the small ones, the practice of importing engags
ceased, and the number of slaves exploded (Singler 1996: 193). The figures

54

The research methodology

Table 3.2. People of colour as a proportion of the colonys total
population. Comparison of Haiti after 2262 years (16811721)
with Martinique after 2565 years (16601700)
Years after colonisation

% of Africans, Caribs, and people of

mixed race in the total population
Haiti

22
25
34
41
47
54
62
65

Martinique

34.8
51.5
61.1
66.7
68.3
81.5
87.8
69.7

Sources: Haiti: 1681, ansom g1 509, no. 2; 1700, ansom g1 509, no. 8; 1713,
ansom g1 509, no. 12; 1721, ansom g1 509, no. 19. Martinique: 1660, Petit
Jean Roget (1980: 1376); 1669, ansom g1 499, no. 2; 1682, ansom g1 499,
(Table 3 in Singler 1996)
no. 10; 1700, ansom g1 470 bis, no. 1.

provided in Table 3.2 (=Table 3 in Singler 1996) show the increase in the population of Africans and people of mixed race, thus illustrating the dramatic change
in the distribution of the Haitian population in comparison with the figures in
Table 3.1.
As is observed by Singler (1996), this shift in the composition of the early
Haitian population had the effect of modifying the slave populations exposure
to French. Indeed, under the first type of economy, the slave population was
in frequent contact with French through the landowners, engags and people of
mixed race. In the shift to a sugar economy, the engags disappeared from the
Haitian population and the enslaved population increased dramatically such that
day-to-day contacts between French speakers and the bulk of the Haitian African
population were greatly reduced. This historical situation created a favourable
context for a creole language to emerge. Singler therefore concludes that the
period when Haitian creole was formed was probably between 1680 and 1740.
Singler (1993a, 1995, 1996) further documents the fact that the nativisation
of plantation societies was an extremely slow process. First, as is also pointed
out by Curtin (1976), the slave traders imported twice as many men as women.
Second, the birth rate during that period was very low (see also Kiple 1984)
and furthermore there was a high rate of infant mortality. Third, the lifespan of
Africans in the Caribbean was short. Singler (1993a: 2378) points out the dramatic consequences of this cluster of facts: This combination of factors yielded
societies unable to reverse the natural population decrease. They were societies
marked by both a disproportionately small number of children and an ongoing

3.1 Demography and linguistic diversity of early Haiti

55

Table 3.3. Distribution by branch of NigerCongo (percentage)

Atlantic
Mande
Kwa
Gur
Nigerian BenueCongo
Ijoid
Bantu
Mallet
Arabic

1664

1680

1690

11
2
22
4
10
4
45
1
1
n = 82

21
39

1
8
54

18
2
19

6
5
17

<1
n = 277

n = 65

(Table 3 in Singler 1993b)

Table 3.4. Breakdown of the Kwa population in the two censuses and
the Remire list (percentage)

Akan
Gbe
Other

1664

1680

1690

11
89

65
35
1
n = 110

9
91

n = 18

n = 65

(Table 4 in Singler 1993b)

stream of recently arrived slaves from Africa. Since the bulk of the Haitian
population at the time the creole was formed was adult, Singler (1996: 199)
concludes that the principal agents of creole genesis must have been adults.
What languages were these adults speaking? Singlers research shows that
the African languages spoken by the slave population in Haiti at the crucial
period were all from the NigerCongo group and more particularly from the
Kwa (Gbe and Akan) and Bantu language families. In a detailed analysis of
seventeenth-century censuses,4 Singler (1993b) documents the linguistic origin
of the slaves sent to French colonies. Speakers of Mande, Atlantic, Gur, Ijoid,
Bantu and Kwa (Akan, Gbe, etc.) languages are identified. The distribution
of the slaves by linguistic group is shown in Table 3.3 (=Table 3 in Singler
1993b: 179). The figures are based on the identifications made from the two
censuses and the Remire inventory.5
The figures in Table 3.3 show that, between 1664 and 1690, while the percentage of Bantu speakers decreased, the percentage of Kwa speakers increased.
The breakdown of the Kwa population for the same period is given in Table 3.4

56

The research methodology

Table 3.5. The growth in the proportion of the population of colour
in Haiti juxtaposed with the percentage of speakers of Gbe dialects
among the African population in the French Caribbean and the
African slave-export population
Year
16759
1680
16809
1681
1690
16909
1700
17009
170019
171019
1713
1717
17209
1721
1730
17309
1739

(1)

2,312
9,082

(2)

(3)
18.8
14
64.6

(D)
(M)
(D)

49
60.4

(R)
(D)

38.9
62.8
65.8

(D)
(F)
(D)

62.6

(F)

50.2

(F)

34.8
66.7

25,273
38,809

81.5
84.2

43,586
81,910

87.8

120,592

91.5

(date of Haitis founding: approximately 1659)

(1) Population of colour in Haiti
(2) Population of colour in Haiti as a percentage of the total population
(3) Estimated percentage of speakers of Gbe dialects in the African
population of the French Caribbean or the slave-export population.
D = Dutch African slave exports to Curaao
F = French African slave exports to the Caribbean
M = 1680 Martinique census
R = 1690 Remire inventory
Sources of Haitian censuses not previously cited: 1717, ansom g1 509,
no. 14; 1730, ansom g1 509, no. 20; 1739, ansom g1 509, no. 21.
(Table 12 in Singler 1996)

(from Singler 1993b: 179), showing that by 1690 Gbe speakers were dominant
among the Kwa.
Consider also the data in Table 3.5 (=Table 12 in Singler 1996) showing the
growth of the population of colour in Haiti, between 1675 and 1739, together
with the percentage of speakers of Gbe languages among the African population
in the French Caribbean. As is pointed out by Singler (1996: 215), the most
important finding set forth in Table 12 is the fact that the first fifty years of
Haitis sugar boom coincided with Gbe predominance in the African population
of the French Caribbean.
The figures in Table 3.5 show that, during the period hypothesised to be the
formative period of Haitian creole, Gbe speakers made up more than 50 per cent
of the French Caribbean slave-export population.

3.1 Demography and linguistic diversity of early Haiti

57

As is pointed out in Lefebvre and Lumsden (1994b), the situation that prevailed in Haiti between 1680 and 1740 presented the two basic prerequisites for
the emergence of a creole language. First, the community was a multilingual
community speaking languages that are not mutually intelligible (although not
entirely disparate, see section 3.3), and given the number of languages present, it
was obviously in need of a lingua franca. Second, as was convincingly shown by
Singler, given the composition of the population (see Table 3.2), the African people
could not have had much direct access to native speakers of French during that
period. The historical evidence thus shows that between 1659 and 1740 the social
conditions in Haiti were such as to permit the development of a creole language.
In view of some remarks in the literature, I would like to emphasise that the historical situation described in this section does not constitute any kind of argument for or against the linguistic processes hypothesised to play a role in creole
genesis. The historical evidence only shows that, at a certain time in Haiti, the
conditions were there for a creole to develop. Again, the external factors do not
argue for or against any theory of how the creole might have developed.
A final point is whether the African languages continued to be spoken in Haiti
and, if so, for how long. Hilaire (1993: 247) writes: On affirme que Toussaint
Louverture [1743?1803], esclave crole de la deuxime gnration, parlait le
fongbe de ses anctres dahomens. Loccasion lui fut donc offerte dapprendre
cette langue en Hati mme. Goman maintiendra une tradition kikongo avec son
groupe de marrons jusquen 1820. [It is stated that Toussaint Louverture [1743?
1803], a second generation creole slave, spoke the Fongbe of his Dahomeyan
ancestors. Therefore, he must have had the opportunity to learn this language
in Haiti. Goman kept up a Kikongo tradition with his group of maroons until
1820.] According to Hilaire (1993), African languages were spoken in Haiti up
until the beginning of the nineteenth century. At the time of the Revolution
(1791), the now well-established creole took over. As Hilaire (1993: 255) puts
it: lpoque de la Rvolution (1791), la position du crole sur le terrain tait
dj suffisamment forte pour servir dinstrument fdrateur et didentification du
pouvoir des masses noires. Lrosion se poursuivrait aux dpens des langues
africaines. [At the time of the Revolution (1791), the position of the creole in
the region was sufficiently strong for it to serve as an instrument of unification
and identification of the power of the black masses. This was to result in the
continued erosion of the African languages.]
These observations are in line with Singlers (1988: 28) observation: The
slow nativisation of plantation societies argues for the slow nativisation of the
creoles that developed there. This in turn suggests a prolonged period during
which the incipient creole would have co-existed with African languages, and
would have been influenced by them.
In short, Haitian creole is hypothesised to have been created by adult speakers
between 1680 and 1740 at the beginning of the sugar economy. These adults were
native speakers of languages of the NigerCongo group and more specifically
the Kwa languages, with a majority of Gbe speakers. At the time Haitian creole

58

The research methodology

developed, African languages were still in use in Haiti and remained so for a
period of approximately a hundred years.
3.2

The typological features of the source languages of Haitian

This section presents a short discussion of the typological features of the source
languages of Haitian. Both the substratum languages and the superstratum language will be examined.
3.2.1

The typological features of the African substratum languages

of Haitian

To what extent are the substratum languages of Haitian homogeneous? This

question is relevant from two points of view. First, since, as will be seen below,
the methodology of the research is based on an in-depth study of only one substratum language, it is important to know whether the other languages involved
shared a substantial number of properties with that one. Second, Singler (1988: 27,
45) has argued that the degree of homogeneity in substratal input has an impact
on the extent of substratal influence on a pidgin or a creole. I provide evidence
showing that the substratum languages of Haitian are quite homogeneous from a
typological point of view.
I take the notion of homogeneity to bear on the typological properties shared
between different dialects/languages. It may (but need not) involve mutual intelligibility. Of course, with typologically similar languages, mutual intelligibility
is a function of the similarity of the phonological representations of corresponding lexical entries. In light of this clarification, I shall discuss the salient features
of the language groups/families represented in Tables 3.3 and 3.4.
I begin with the Gbe languages of the Kwa family, which have been shown to
be predominant in Haiti at the time Haitian creole was formed (see Table 3.4).
As is often the case in African dialectology, the number of Gbe languages/
dialects varies among authors. Man and Dalby (1987: 62) list the following
Gbe languages, with the specification that there are several dialects of each:
Ewegbe, Gengbe, Ajagbe, Fongbe, Gungbe, Wem=gbe, Phlaphera, Cigbe, Xwdagbe,
Kogbe. In the Atlas sociolinguistique du Bnin (1983: 55), we find the following
statement: Entre les langues de ce sous-groupe ou continuum existe une intercomprhension si frappante parfois que lon se demande si telles ou telles langues
forment vraiment des units linguistiques autonomes. [Among the languages of
this subgroup or continuum, there is a degree of mutual intelligibility so striking
that one wonders at times whether particular languages are really autonomous
linguistic units.] Capo (1984: 167) further asserts that from a linguistic point
of view . . . EweGenAjaFon constitute a dialect cluster. Several linguists who
work on the area where these languages are spoken (see e.g. Westerman 1930;
Bertho 1946; Ansre 1961) refer to them as one language. According to Capo
(1984: 168),

3.2 Typological features of source languages of Haitian

59

there is mutual intelligibility between dialects that are contiguous, e.g. Ewe and
Gen, Gen and Aja, Aja and Fon, etc.; but the degree of mutual intelligibility is
related to geographical distance, e.g. although there is some mutual intelligibility between Ewe and Aja, it is less than between Ewe and Gen, and there
seems to be none between Ewe and Fon which are the extreme ends of the
dialect cluster.

Pazzi (1976) shows that this dialect cluster shares a large number of lexical
entries with the same phonological representations. Moreover, when the phonological representations of lexical entries differ, they often share semantic and
syntactic properties. For example, the comparison of Gungbe, Fongbe, Ajagbe,
Twfingbe, Xwdagbe and Xlwagbe by Hazoum (1990) shows that the functional category lexical entries of these Gbe languages are very similar (although
not identical) with respect to inventories, as well as semantic and syntactic properties. Finally, Capo (1991) argues extensively that, from a phonological point of
view, these language varieties have a great deal in common in terms of inventories
and processes.
The neighbouring languages to the Gbe cluster constitute the Ede cluster. This
group of dialects includes Ede Nago, Ede Yoruba, Ede If, etc. Again, according
to the Atlas sociolinguistique du Bnin (1983: 55), Comme pour les langues
du continuum gbe, il existe entre celles du continuum ede une intercomprhension
parfois troite. [As with the languages in the Gbe continuum, there is a sometimes close mutual intelligibility among those of the Ede continuum.] As shown
by Capo (1984), however, the dialect cluster EweGenAjaFon is clearly distinct from Yoruba (and, I assume, from the Ede cluster considered as a whole).
Indeed, the Gun variety of the Gbe cluster is not mutually intelligible with
Yoruba, although, as is pointed out by Capo (1984: 168), both languages are
natively spoken in Porto Novo (Benin). The Gbe cluster is also not mutually
intelligible with its western neighbouring language Ga-Dangme. Capo (1984:
168) further points out that the Gbe cluster is also distinct from Akan, Adele
and Aguma spoken in the area north of it. However, although the Gbe and
Ede clusters of dialects are analysed as separate clusters on the basis of the
lack of mutual intelligibility, they are not very different in their semantics and
syntax. For example, both dialect clusters have serial verbs, predicate cleft,
markers encoding tense, mood and aspect that occur between the subject and
the verb, etc.
In Westerman and Bryans (1970) classification, the Gbe and Ede clusters
are part of the Kwa family of languages. According to them, the Kwa family
includes the following languages or dialect clusters: Lagoon (including Abbey);
Akan, which includes the three dialect clusters Twi-Fante, Anyi-Baule and Guang;
the Gan-Adangme dialect cluster; the Yoruba cluster (identified above as part of
the Ede cluster); the Ewe cluster (identified above as part of the Gbe cluster); the
Nupe languages, which divide up into several dialect clusters; the Bini languages;
and the Igbo dialect cluster.6 Westerman and Bryans (1970: 905) review of the
salient typological features of the Kwa languages shows that these languages

60

The research methodology

share a number of significant properties. For example, tones are very important
in all of them. Verbal and nominal roots are mainly monosyllabic, consisting in
cv. Reduplication of a verbal base is used to form adjectives from verbs. Compound nouns and verbs are very frequent. These languages exhibit no true noun
classes and no concord phenomena. There is no grammatical gender in any
of these languages. The verb root is invariant. Tense, Mood and Aspect are
expressed by particles occurring between the subject and the verb. Word order in
simple sentences is subjectverbobject. Salient differences between these languages involve the formation of plural, the word order in Genitive constructions
(where the possessor can either precede or follow the possessed noun) and the
variable availability of an overt affix converting verbs into nouns. As will be
seen in the data chapters, some of these differences are reflected in the creole.
Westerman and Bryan (1970) also review the properties of the West Atlantic,
Kru, Mande (which includes Malinke, Bambara and Dyula) and Gur languages,
showing the typological features of each of these language families. From their
description, it is clear that, in spite of the lack of mutual intelligibility between
the languages in these different families, they share a large number of typological
features. In fact, these languages have semantic and syntactic properties that are
so similar as to lead Koopman (1986: 233) to treat them as a coherent linguistic
group which she refers to as West African.7 Considering a sample of Kru and
Kwa languages, as well as a Mande and a Gur language, Koopman writes: It
will be shown that these languages share a number of characteristics, which are
also characteristic of many other West African languages, such as Yoruba for
instance. The presence of a number of common properties will allow us to refer
to general West African properties.
Although the Bantu languages were most important in Haiti before and after
the formative period of Haitian creole, these languages were still represented at
the time the creole was formed (see section 3.1). Taken as a group, these languages differ from the other West African languages discussed above in two
major ways. First, while the languages discussed above do not present class prefixes or concord systems,8 the Bantu languages are overwhelmingly characterised
by class prefixes and a complex system of agreement (see e.g. Welmers 1973;
Lipou 1983). Second, whereas the languages discussed so far are, to a large
extent, isolating languages, the Bantu languages are typically agglutinative (see
e.g. Greenberg 1966a; Welmers 1973; Mufwene 1986). However, while the former characteristic appears to be a real property distinguishing between the two
types of languages, the latter appears to be a rather superficial one. For example,
Givn (1971) presented evidence that the modality prefix in Bantu languages
arose historically from main verbs dominating sentential complements, and the
condensation of main modal verbs into modality prefixes in Bantu is a relatively
recent process. Furthermore, Givn (1971) argued that Bantu and Kwa have
similar underlying structures and the difference between them lies in the fact
that predicate raising takes place in Bantu but not in Kwa. Building on Givns
proposal, Baker (1988, 1991) argues, within a more recent framework, that the

3.2 Typological features of source languages of Haitian

61

contrast between Bantu (agglutinating) and Kwa (isolating) languages is correlated with the verb raising parameter. In the former languages, verbal heads
raise to in, in the latter they do not. In Bantu, the verb collects the affixes
as it moves up in the syntactic tree whereas, in Kwa, the verb stays in its basic
position. Consequently, whereas the Kwa construction is realised as a series
of free morphemes, the Bantu construction is realised as a collection of bound
morphemes agglutinated onto the verb.
In spite of these differences, Kwa (and other West African languages of the
same type) and Bantu languages have been considered to share a strong degree
of homogeneity (see e.g. Koelle 1854; Greenberg 1966a, 1983) on the basis of
the fact that they all share a number of typological features (phonological, morphological and semantic). Furthermore, Africanists who have worked on both
types of languages (e.g. Maarten Mous and Hilda Koopman, p.c.) assert that
they share many properties at all levels of the grammar (phonology, lexicon,
syntax). Consequently, although the African languages spoken in Haiti at the
time Haitian creole was formed were numerous, they shared a significant number
of typological properties. Thus, it can be claimed that they constitute a relatively
homogeneous group.
As will be seen throughout this book, Haitian is very similar to the nonBantu substratum languages. Why should this be so? What happened to the
Bantu speakers in Haitian creole genesis? These questions will be addressed in
chapter 13.
3.2.2

The typological features of the superstratum language of Haitian

French is the superstratum language of Haitian.9 It is a Romance language, a

member of a language family typologically quite different from the West African languages discussed above. The typological features of French compare with
those of the Kwa languages enumerated above as follows. In contrast to the Kwa
languages, French has no tones; rather, it has a very simple stress system. Unlike
the Kwa languages, the verbal and nominal roots are not generally monosyllabic. French does not exhibit reduplication of verbal bases to form adjectives.
In contrast to the Kwa languages, French has two grammatical genders: masculine and feminine. It manifests agreement phenomena: the tensed verb agrees
in person and number with its subject; in specific contexts, the past participle
agrees in gender and number with its direct object; determiners and modifiers
agree in number and gender with the noun that they determine or modify. Tensed
verbs must be marked minimally for mood and tense, a typological feature which
contrasts with Kwa, where the verb is invariant. Like Kwa languages, word order
in simple clauses is subjectverbobject but, unlike Kwa, French has subject
verb inversion in specific contexts. Finally, French has no serial verbs and no
constructions involving verb-doubling phenomena such as predicate cleft. These
two types of constructions are salient ones in the Kwa languages. The differences between the two types of languages will be extensively discussed.

62
3.2.3

The research methodology

Summary

In spite of their differences, the West African substratum languages of Haitian

have been shown to constitute a rather homogeneous group of languages, typologically speaking. French, the superstratum language of Haitian, has been shown
to be typologically quite different. The creators of Haitian, who spoke typologically similar languages and who had similar grammars and lexicons, were thus
faced with a language that was typologically quite foreign to their own, in terms
of both lexical and grammatical properties. It will be shown how, given these
conditions, they processed the superstratum data in creating Haitian creole.
3.3

The superstratum data the creators of Haitian were exposed to

What variety of French did the French colonists speak in Haiti? There is a consensus in the literature (see e.g. Faine 1937; Hull 1979; Valdman 1978, 1979)
that the bulk of the French colonists who went to the Caribbean in the seventeenth
and eighteenth centuries were from western and central France (i.e. Normandy,
le-de-France (Paris), Poitou, Saintonge, Brittany). The dialects spoken in this
area of France constitute the langue d/ol/ (as opposed to langue doc). These
dialects were not necessarily mutually intelligible among themselves or with the
variety of French spoken in le-de-France (Paris). (For various views on this
matter, see the papers in Mougeon and Beniak 1994.) But, as noted by MeyerLbke (1909: 123), these regional dialects were influenced by Parisian French.
On sait trs bien que dans les provinces de lOuest, les dialectes ont subi,
plus profondment que partout ailleurs en France, et mme avant lpoque de la
colonisation de lAmrique, linfluence de la langue officielle. [It is very well
known that, in the Western provinces, more than anywhere else in France, the
dialects were influenced by the official language, even before the colonisation
of the Americas.] Furthermore, there appears to be quite a widespread consensus
in the literature on the characteristics of the French colonists of the seventeenth
and eighteenth centuries (see e.g. Rivard 1914; Mougeon and Beniak 1994;
Chaudenson 1992; Poirier 1994; Wolf 1994). First, they were mainly from the
domaine d/ol/. Second, even though they were born in the provinces, they had
lived in cities before leaving France. Third, most of them spoke French. Finally,
about half of them were literate (that is, they knew how to read and write).
According to Juneau and Poirier (1973: 1913), the variety of French spoken by
the colonists was close to the variety of French spoken in le-de-France in the
seventeenth and eighteenth centuries: Les traits syntaxiques sont facilement
rattachables des tendances de la syntaxe de lpoque. Il en est de mme pour
les traits morphologiques. [The syntactic features can very easily be linked to
syntactic tendencies of the period. The same is true of morphological features.]
Wolf (1991) also draws a similar conclusion.10 I will therefore assume that the
French colonists in Haiti spoke a variety of French similar to that referred to as
seventeenth- or eighteenth-century French in dictionaries and grammars.

3.3 Superstratum data

63

It is perhaps worth noting at this point that some scholars who have worked
on French-based creoles have adopted a different methodology which consists in
looking for traces of French regional dialects in the creoles. For example, the
French dialect spoken in Picardy, Picard, is often cited as the source of a creole
form or feature. According to Juneau (1972), however, only 2 per cent of the
French colonists sent to the Americas were from Picardie. Consequently, it is
unlikely that the bulk of the African population in the French Caribbeans were
exposed to this dialect of French in a way that would have had a significant
impact on the French-based creoles. Furthermore, it is likely that, in the colonies,
the French colonists who spoke regional dialects natively used the variety of
seventeenth-century French that they had in common, as attested in the historical
work cited above. Given these considerations, and although I will occasionally
refer to regional French dialects in the data chapters, I will assume that the creators of Haitian were mainly exposed to the variety of French spoken in le-deFrance in the seventeenth and eighteenth centuries.
There is also a consensus in the literature that the French colonists who went
to North America in the same period came from the same pool as those who
went to the Caribbean (see e.g. Juneau 1972; Valdman 1978, 1979). Various
authors hypothesise that the seventeenth- and eighteenth-century French spoken
by the French colonists was similar regardless of whether they went to North
America, the Caribbean or Africa. Furthermore, modern French spoken in North
America (e.g. the province of Quebec, Acadia) has kept several of the particularities of the seventeenth- and eighteenth-century French spoken in Western
and Central France (see Juneau 1972). Hence, I will refer to these dialectal
varieties of modern French, a common practice in creole studies.
Having established that the French colonists spoke a variety of seventeenthand eighteenth-century French, we can still ask the following question: What kind
of French data were the speakers of West African languages in Haiti exposed to?
The question is not a trivial one for, in the literature on pidgins and creoles,
it has been proposed on several occasions that speakers of the substratum languages were presented with a reduced (baby-talk or foreigner-talk) version of
the superstratum language. In this view, the plantation owners were voluntarily
speaking a reduced version of their own language in order to maximise communication with the slave population. Schuchardt (as translated by Goodman
1964: 124), Jespersen (1922: 233), Bloomfield (1933), Gbl-Galdi (1934) and
Hall (1966) all hold some version of this view. Similar proposals have also been
made more recently. For example, Ferguson (1971: 147) advocates the view
that the foreigner talk of a speech community may serve as an incipient pidgin.
This view asserts that the initial source of the grammatical structure of a pidgin
is the more or less systematic simplification of the lexical source language which
occurs in the foreigner talk registers of its speakers, rather than the grammatical
structure of the language(s) of the other users of the pidgin. (See Naro 1978,
for an extensive discussion of this issue.) Similarly, Foley (1988: 166) writes: I
suggest that a pidgin is a version of a foreigner talk of a superstratum community

64

The research methodology

that has been conventionalised and accepted, most importantly by speakers of

the substrate language(s). According to some of the proponents of this approach,
creole languages lack the functional category lexical entries of their superstratum
languages because their creators were not presented with them.
While it could very well be the case that French speakers did not use a
very elaborate style of French while talking to the African population of Haiti,
there is plenty of evidence from the Haitian lexicon that the speakers of the
substratum languages were exposed to the functional category lexical items of
French. Valdmans et al. (1981) dictionary abounds in examples where a Haitian
word corresponds to a French expression which includes a French functional
item. For example, a simple Haitian lexical entry may contain an agglutinated
French determiner, as shown in (1).
(1)

haitian lexical entry

larivy river
lari
street
olye
instead

french expression
la rivire the river
la rue
the street
au lieu
instead

A simple Haitian lexical entry may also contain an agglutinated French partitive
(e.g. du), as shown in (2).
(2)

haitian lexical entry

diri rice
dife fire

french expression
du riz rice
du feu fire

A Haitian simple word may also contain an agglutinated French complementiser

(que that), as shown in (3).
(3)

haitian
eske
fok
pas(k)e
jiskaske
tandiske

lexical entry
question marker
complementiser
because
until
while

french expression
est-ce-que
is it the case that . . .
(il) faut que
there must be
parce que
because
jusqu ce que until
tandis que
while

A simple Haitian lexical entry may include the agglutinated French functional
item , as shown in (4).
(4)

haitian lexical entry

afs
by means of
apati
starting from
apen
hardly
apepr more or less

french expression
force de by means of
partir de starting from
peine
hardly
peu prs more or less

Finally, a Haitian lexical entry may include an agglutinated French conjunction,

as in (5).
(5)

haitian lexical entry

epi
and
oubyen or

french expression
et puis and then
ou bien or

3.4 The linguistic test

65

As for the French category gender, it is reflected in the Haitian lexicon in

examples such as the following. The Haitian form of the verb meaning to cook
is kwit. This phonological form does not correspond to the basic form of the
corresponding French verb cuire but to the past participle form inflected for
feminine gender: cuite cooked (fem.). The Haitian form of the noun meaning
cat is chat. This phonological form is derived from the French form meaning
cat inflected for feminine gender: chatte cat (fem.).
These examples, and many more, show that the Haitian African population
must have been exposed to forms containing French functional items. This conclusion is in line with Alleyne (1971: 170), who states that the African population of the Caribbean area was exposed to European languages in their full
morphological and syntactic forms. The above data constitute a major drawback to the claim that a creole lacks the functional category lexical entries of its
superstratum language because the substratum speakers were simply not exposed
to them. Instead, they show that the creators of Haitian creole were exposed to
French functional items but did not identify them as such because they did not
have enough exposure to French. These data thus support the claim in chapter 2
that language learners in the context of creole genesis have very limited direct
access to the superstratum language and, consequently, they fail to identify the
functional categories of that language. Furthermore, the data provide additional
support for the claim made in chapter 2 (see (2) ) that relabelling proceeds on the
basis of phonetic strings found in the superstratum language rather than on the
basis of the phonological representation of the target language lexical entries.
So, even though we do not know from direct observation the exact nature of
the French data the creators of Haitian were exposed to, we can at least make
some inferences about them on the basis of the phonological representations of
the Haitian lexical entries. Many more examples will be presented throughout
this book showing that the creators of Haitian were presented with adequate
French data, but that they did not have enough exposure to French to appropriately analyse them.
3.4

The linguistic test

The test of the hypothesis that speakers of different languages use the properties
of their own lexicons and grammars in creating a creole rests on a three-way
comparison between the properties of the creole and those of its contributing
languages: the superstratum language, on the one hand, and the substratum
languages, on the other hand. Next to studying the making of a creole language
firsthand, the ideal comparison would be based on the following data. First, it
would require a comparison of the creole with all the languages that were
spoken in the area during the period when the creole was formed. In the case
of Haitian, these include French, the superstratum language, and the various
African substratum languages identified in section 3.1. Furthermore, this comparison should ideally be done on the basis of the grammars and lexicons of

66

The research methodology

the languages as they were spoken in Haiti at that time. Given the number of
African languages involved at the time the creole was formed (see section 3.1)
and the lack of seventeenth- and eighteenth-century data on most of them (see
below), the comparison that we have already been able to do differs from the
ideal in two ways.11
First, the decision was made to study only one substratum language of
Haitian for the detailed comparison (see Lefebvre and Kaye 19857 Project).
This decision was a consequence of another methodological choice, based on the
conviction that an in-depth study of one West African language would provide
a better database for a detailed comparison than a general survey of the properties of all the West African languages involved. Furthermore, a substratum language had to be chosen prior to doing the historical research.12 In view of the
lack of precise information on the origin of African slaves in the early history
of Haiti, Jonathan Kaye and I made our choice on the basis of sociological and
anthropological factors. Many anthropological studies assert the prominence of
the Fon culture in Haiti (see Lefebvre 1986). For example, Bastide (1967: 17
18) writes: Hati . . . les diverses cultures ne sont plus que des lments intgrs
et subordonns de la culture fon . . . Hati . . . cest la culture dahomenne,
plus particulirement fon qui lemporte. [In Haiti . . . the various cultures are no
more than elements integrated into and subordinated to Fon culture . . . In Haiti . . .
Dahomeyan culture, and more specifically, Fon culture, is dominant.] Herskovits
(1975: 245) points out that in Haiti,
The majority of African traits . . . derive from the geographical centre of the
slaving area, especially the region of Dahomey. Even the famous word voodoo is Dahomean, as all Haitian writers and some Europeans and Americans
have recognised. It is the vodun of the Fon-speaking people a word which is
best translated as god. Names of gods, and social and economic institutions,
point their origin to this same Dahomean source.

So, because of the prominence of the Fon people in the history and popular
culture of Haiti, Fongbe, one of the Gbe languages in the Kwa family, spoken
mainly in Benin (the former Dahomey), was chosen as the substratum language
for detailed comparison.
This methodological choice meets the challenge set forth by Bickerton (1984,
1986) to scholars who advocate that the grammars and lexicons of the substratum
languages constitute an important input into the grammars and lexicons of creole
languages. According to Bickerton, these scholars must show point-for-point
similarity between the grammar of a given creole and the grammar of some
specific substratum language, rather than some vague generalisation across substratum languages, and they must show that speakers of this substratum language
were in the right place at the right time to make its input in the genesis of that
creole plausible. Evidence supporting the first point will be presented throughout
this book. As for the second point, the historical evidence (presented in section
3.1) shows that Fongbe speakers were among the right people, in the right place,
at the right time.

3.4 The linguistic test

67

I would like to emphasise that the decision to conduct an in-depth study of

one of the substratum languages rather than a general survey of the properties of
all the West African languages involved is the result of a methodological choice
forced by time and resource limitations. This methodological choice, however,
does not imply that Fongbe was the only substratum language of Haitian creole
nor that Haitian is a relexified version of Fongbe. Unfortunately, I have been
misinterpreted on this point in several instances, namely in Thomason (1993:
281) and in Arends, Muysken and Smith (1995). Nor have I ever stated that
it is useless to study the other African substratum languages of Haitian, as
Chaudenson (1994: 8) claims I did. Recall from chapter 2 that relexification is
a mental process and that, therefore, it is an individual activity. This being the
case, Fongbe speakers will use the properties of their own lexicon in relexification.
So will the Yoruba, Ewe, Bantu, etc., speakers.
As pointed out in Lefebvre and Lumsden (1994a), the fact that only one
substratum language has been considered in detail is less significant than it first
appears. We have seen that the substratum languages of Haitian constitute a
fairly homogeneous group (see section 3.2.1). Furthermore, as has been pointed
out in Lefebvre and Lumsden (1994a: 56): Lhypothse prdit un parallle significatif entre les proprits syntaxiques et smantiques du crole et celles des
langues du substrat considres dans leur ensemble. Le fait que notre comparaison
nimplique quune seule langue du substrat minimise la possibilit de trouver un
tel parallle et maximise la possibilit de falsifier lhypothse. [The hypothesis
predicts significant parallelism between the syntactic and semantic properties of
the creole and those of the substratum languages as a group. The fact that our
comparison only involves a single substratum language minimises the possibility of finding such parallelism and maximises the possibility of falsifying the
hypothesis.] Thus, from a methodological point of view, a detailed and in-depth
comparison of Haitian creole, French and Fongbe does provide a valid test of the
processes hypothesised to play a role in creole genesis, in spite of the fact that
there were other substratum languages present at the time the creole was formed.
Fongbe will thus be used as the representative substratum language of Haitian.
References to other substratum languages will be made, however, whenever
necessary, and particularly for the discussion of cases of dialect levelling.
In fact, the three-way comparison used to test the hypothesis has proved to be
a most useful methodological tool. Recall from chapter 2 that the hypothesis that
relexification plays a central role in creole genesis predicts that the creole lexical
entries should, to a great extent, have the same semantic and syntactic properties
as the corresponding lexical entries in the substratum languages but phonological
representations derived from the phonetic strings of the superstratum language.
On the one hand, the semantics of the lexical categories of natural languages
have a great deal in common but different languages may differ in the way they
divide up semantic representations in the lexicon. On the other hand, there is a
finite set of functional categories in languages of the world, and languages may
vary with respect to the fraction of this finite set they encode. In this view, the

68

The research methodology

comparison of Haitian lexical entries with those of its source languages makes
it possible to find cases where both the superstratum and substratum lexical
entries have the same semantic and syntactic properties. The data corresponding to this possibility do not constitute evidence either way. The pertinent cases
for the test of the hypothesis advocated in this book are those involving lexical
entries which have different properties in Fongbe and in French. In these cases,
Haitian lexical entries which conform to the prediction made by the relexification hypothesis constitute data supporting that hypothesis. Haitian lexical entries
which do not conform to this prediction require further investigation. In the
latter case, there are three possibilities. First, the properties of an apparently
non-conforming Haitian lexical entry may be traced to another West African
language. In this case, we have identified data which are relevant to the process
of dialect levelling. Second, the properties of such a lexical entry in Haitian may
be traced to French. In this case, we have identified data which show the contribution of the superstratum language. A third possibility involves cases where
the properties of such a lexical entry cannot be traced to any of the source languages. In this case, we have identified an independent development. The same
reasoning applies to parameter setting. Suppose that the value for a given parameter is either positive or negative. In this view, the hypothesis is supported by
cases where Haitian and Fongbe have the same value for a given parameter and
French does not. Other cases require further investigation along the same lines
as indicated above.
A second major point is that the ideal test would consist in comparing the
grammars and lexicons of the languages as they were spoken in Haiti during the
period when the creole was created. The actual test we have been able to do
departs from the ideal on this point, as well. The reason is that the material on
seventeenth- and eighteenth-century Haitian and Fongbe is rather scarce.
A list of available written sources on Haitian between 1776 and 1936 is
provided in Baker and Corne (1982: 2734), reproduced in Appendix 1. Among
the earlier sources, we find Moreau de Saint-Mery (1797), Ducurjoly (1802),
a text entitled Un habitant dHayti (1811) and Anonymous (1983). The question
here is whether these texts can be considered as data representing the speech
of the majority of Haitian speakers at the time they were written. Chaudenson
(1977: 259) makes the following general remark about these historical documents: Documents describing the early states of these languages are rare and,
when available, are so fragmentary as to be relatively untrustworthy. According
to Carden and Stewart (1988: 267), these texts present a problem of interpretation. For example, they write that Ducurjoly and Moreau de Saint-Mry
were whites who fled Haiti at the time of the revolution. They were presumably native speakers, or at least very fluent speakers, of the version of
HC spoken by the whites and the blacks in the higher positions. (Note that D
makes no obvious differences between the speech of the whites and the blacks
in his Conversations, but the blacks are commandeurs and sucriers, not field
hands.)

3.4 The linguistic test

69

DeGraff (1993a: 90, note 56) reports on Dejeans opinion: Yves Dejean advises
caution in using early Haitian Creole texts for historical purposes. These texts
were written by non-native speakers that were often disdainful of the creole.
Moreover the language is systematically betrayed by the then-prevalent Frenchbased orthography. Finally, Lefebvre and Lumsden (1994a: 56) are also of the
opinion that the above-mentioned written sources most probably do not reflect
the speech of the majority of the African slave population in Haiti at the time
they were written. So, although the data contained in the above-mentioned sources
may be of some interest (see Carden and Stewart 1988), there appears to be a
consensus that they cannot be taken as representing the speech of the majority of
the contemporary African Haitian population. Furthermore, to the best of my
knowledge, historical reconstructions of Haitian subsystems are not available in
the literature. Hence, the Haitian data used for the three-way comparison presented in this book are from Haitian as it is spoken today.
As for Fongbe, there are no texts available to show us earlier stages of this
language.13 I know of only one historical reconstruction study: Capos (1991)
masterpiece on proto-Gbe phonology based on a comparison of a large number
of Gbe languages. This work will be most useful to students who seek to study
the phonological system of Gbe languages. So again, the Fongbe data used for
our comparison are from modern Fongbe.
The effect of comparing Haitian and Fongbe as they are now, rather than as
they were spoken 250 years ago, is less significant than it appears at first glance.
As has been pointed out in Lefebvre and Lumsden (1994a: 56), if one of the two
languages had undergone significant changes since Haitian creole was formed,
and if these changes had increased the significant parallels predicted by the hypothesis, that would be an extraordinary coincidence. Puisquil existe, pour deux
langues donnes, de nombreuses faons dtre diffrentes mais une seule faon
dtre identiques, un changement qui rsulterait en un parallle accidentel est moins
probable quun changement qui rsulterait en une diffrenciation accidentelle.
[Since, for any two languages, there are many ways of being different but only
one way of being identical, a change that resulted in accidental parallelism is
less probable than a change that resulted in accidental differentiation.]
As for French, some material (texts, grammars and dictionaries) is available
on seventeenth- and eighteenth-century French. However, there are major gaps
in the linguistic research concerning this period in the evolution of the language.
While the history of French is rather well documented up until the end of the
sixteenth century, it seems that seventeenth- and eighteenth-century French has
not attracted specialists in this field. One reason for this might be that, as has
been suggested to me by colleagues who specialise in the history of French,
from a global perspective, late seventeenth-century French does not appear to
present significant differences from modern French, in that the basic properties
of the language are generally considered to have been established by the end of
the seventeenth century. In this respect, Danielle Dumais, one of the research
technicians for the 198994 Project, has read more than 700 pages of documents

70

The research methodology

brought back by John Singler from the Colonial Archives. These documents
consist of correspondence between the French colonists/captains of ships and
the King of France. They report on the situation in the French colonies of West
Africa and the Caribbean. The letters were written between 1664 and 1766.
Dumaiss task was to identify, in these texts, lexical or syntactic uses which
differ from current ones. She found only very minor differences. The most salient
difference she encountered was the selection of or de by verbs which assign
Dative Case to their object. This finding is congruent with similar observations
found in grammars of seventeenth-century French. For example, Spillebout (1985)
remarks that the use of and de in seventeenth-century French differs in some
contexts from the modern use of these forms. Given this fact, and due to limited
time and resources, this research project has not carried out any original research
on seventeenth- and eighteenth-century French. Prior to this book, most of the
comparative work accomplished within the framework of this research was based
almost exclusively on French as it is spoken today.
For the purpose of this book, however, I surveyed the available literature on
seventeenth- and eighteenth-century French.14 Grammars and dictionaries were
consulted.15 French plays written during the seventeenth and eighteenth centuries were also considered as a source of data for particular constructions. These
will be mentioned in the text. I found minor syntactic differences from modern
French including the position of some adjectives with respect to the noun, the
position of negation in infinitival clauses and clitic climbing phenomena (still
available in early seventeenth-century French but not in modern French).16
Although the aforementioned sources are reliable in terms of the standard language, they sometimes fail to provide information about the colloquial variety of
French spoken at the time. Recourse to conservative dialects, such as Montreal
French, proved to be most useful in this respect. As mentioned above, it has
been shown that the North American varieties of French share properties with
the French of the seventeenth and eighteenth centuries, in contrast to standard
international French. Montreal French has been extensively described in the context of the Sankoff and Cedergren Project at the Universit de Montral and the
Lefebvre and Drapeau Project at the Universit du Qubec Montral. Both
projects have given rise to a large number of publications. The publications of
the Sankoff and Cedergen Project are listed in Thibault and Vincent (1990).
The core results of the Lefebvre and Drapeau Project are published in Lefebvre
(1982c). Work by individual researchers has also been consulted.
In conclusion, although the test carried out by this project differs from the
ideal test as discussed above, I believe that the scientific methodology set up to
test the hypothesis is valid.
3.5

What counts as evidence for the hypothesis and how can it

be falsified?

Recall from chapter 1 that, in the theoretical framework adopted for this research,
individual languages show properties that are universal and other properties that

3.5 How can the hypothesis be falsified?

71

are language-specific. Since they are natural languages, creole languages are
assumed to exhibit both universal and language-specific properties. Recall also
from chapter 1 that, in the model of grammar adopted for this research, languagespecific features are hypothesised to be manifested in the lexicon, the syntactic
parameters, and the interpretive component of the grammar. The languagespecific features of creoles are expected to be found in these components of the
grammar, and they are hypothesised to parallel those of the substratum languages
when the latter contrast with those of the superstratum language. In this view,
the data that count as evidence in favour of the hypothesis advocated in this
book are cases where the properties of Fongbe and French differ and where the
properties of Haitian match those of Fongbe and contrast with those of French.
As we have seen, the relexification hypothesis predicts that the creole lexical
entries will derive their phonological representations from phonetic strings found
in the superstratum language and their semantic and syntactic properties from
the substratum lexical entries. In the examples below the Haitian lexical entries
have a phonological representation derived from French. In each case, however,
there is an extra meaning associated with the Haitian lexical entry which is also
associated with the corresponding entry in Fongbe.
(6)

haitian
plim
feather
hair

french
plume
feather

fongbe
fn
feather
hair

vyann
viande
meat
meat
edible animals
(complement of the verb
to kill)

ln
meat
edible animals
(complement of the verb
to kill)

dife
fire
brand

(du) feu
fire

mywn
fire
brand

tt
head
roof

tte
head

van
wind
air

vent
wind

jwhwn
wind
air

head
roof

The type of data illustrated above follows from the relexification hypothesis and
constitutes evidence supporting it.
Likewise, the syntactic properties of verbs are not uniform across languages.
The data that count as evidence are those where the Haitian and Fongbe verbs
share syntactic properties that differ from those of the corresponding French
verbs. Similarly, there is a finite inventory of functional categories; each language presents a subset of this inventory. If this subset is the same in the creole
and the substratum language and contrasts with that shown in the superstratum

72

The research methodology

language, this is evidence supporting the hypothesis. Languages of the world

vary in terms of their inventory of derivational affixes. These inventories may
vary in size; for example, while French has over a hundred derivational affixes,
Vietnamese has none at all. The inventories may also vary in terms of the particular subset of derivational affixes manifested in a given language. If a creole
has an inventory of derivational affixes which corresponds to that of its substratum
language in contrast with that of its superstratum language, this is also evidence
supporting the hypothesis advocated in this book.
The account of creole genesis presented in chapter 2 is formulated in terms
that are precise enough to be falsifiable. How, then, can it be falsified? The
hypothesis that relexification plays a central role in creole genesis predicts that
Haitian lexical entries will derive their phonological representations from phonetic matrices found in the superstratum language and their semantic and syntactic properties from the substratum lexical entries. As has been pointed out by
Lefebvre and Lumsden (1989a: 257, 1994a: 51), if we were to find an arbitrary
division of properties between the Haitian lexical entries and those of the source
languages, the hypothesis that relexification plays a central role in the genesis
of creole languages would be falsified. Similarly, the hypothesis that the creators
of a creole use the parametric values of their own grammar in creating a creole
predicts that, where substratum and superstratum grammars differ in the value
assigned for a given parameter, the value found in the creole should be that
of the substratum grammar. In the contrary case, the hypothesis would also
be falsified.
Testing the hypothesis that the creators of the creole used the properties of
their own lexicons and grammars in creating the new language involves the
global comparison of the lexicons and grammars of the languages involved (see
Lefebvre and Kaye Projects 19859; Lefebvre 1993a; Lefebvre and Lumsden
1989a, 1994a). As emphasised by Lefebvre and Lumsden (1994a), one or two
examples either way neither support nor falsify the hypothesis. Indeed, any
example taken individually outside of the perspective of the global comparison
may be interpreted as a fact of Universal Grammar (see e.g. Bickerton 1977,
1981, 1984) or as being related to recurrent patterns observed in ordinary cases
of second language acquisition. The results of a global comparison, however,
cannot be so interpreted. Either the bulk of the properties of the Haitian lexicon
and parametric values pattern on the model of the substratum languages when
they contrast with French, or they do not. Systematic parallels between Haitian
and its substratum languages, where the latter differ from French, would be an
extraordinary coincidence for theories which claim that the properties of creole
languages are to be attributed to linguistic universals or second language acquisition strategies.
Ever since Bickertons (1984) claim that creole languages present the unmarked options of Universal Grammar, markedness has been an issue in the
literature with respect to how substratal influence in creole genesis can be
demonstrated. For example, Muysken and Smith (1986a: 4) make the following

3.5 How can the hypothesis be falsified?

73

statement: In order to prove substrate influence we have to look for marked

structures appearing in both languages the languages potentially subject to
substrate influence and the potential substrate language. Thomason (1993: 287)
also claims that most shared features found everywhere in West Africa are
likely to be universally unmarked and thus difficult or impossible to establish as
Africanisms in Afro-American languages. Similarly, Singler (1988: 29) asserts
that: for an exclusively substratal explanation to prevail, the phenomenon in
question would have to be highly marked; otherwise, an explanation based on
linguistic universals could be invoked. Before addressing the issue of markedness itself, some comments on the above quotations are in order. With regard to
Muysken and Smiths claim, in my view, a theory cannot be proven; rather, one
presents evidence supporting or falsifying it. With regard to Thomasons claim,
I wonder on what basis she can claim the shared features of West African languages to be universally unmarked. For example, how can predicate cleft, a rare
phenomenon among the languages of the world but common in West African
languages, be argued to be an unmarked feature of Universal Grammar? As for
Singlers assertion, I do not know of any theory of creole genesis advocating an
exclusively substratum explanation, for any theory must account for the contribution of at least two groups of languages: the substratum and superstratum languages. That being said, I now turn to a discussion of the markedness issue itself.
With the exception of Muysken (1981c), the above-mentioned authors do
not present us with a precise theory of markedness. Taking it for granted that
they have in mind a definition of markedness as provided in the Generative
Linguistics work produced over the last twenty years, I assume that by marked
feature they mean a feature that is rare among the languages of the world and
that has to be learned by language learners on the basis of the primary data
they are exposed to (see e.g. Van Riemsdijk 1978). By contrast, an unmarked
feature would be one that is widespread among the languages of the world and
does not need to be learned on the basis of primary data. According to current
theories, unmarked features are considered to constitute the core of Universal
Grammar whereas marked features are assumed to constitute the periphery (see
Chomsky 1981).
For the sake of discussion, let us assume with Bickerton (1984) that creole
languages represent the unmarked case, as defined above. In this view, predicate cleft, which is a widespread feature of Caribbean creoles, would have to
be considered an unmarked feature of Universal Grammar. But, as Koopman
(1986: 249) remarks: If it is an unmarked property to admit this construction,
as Bickerton claims, one would expect the French or English child to produce
instances of it spontaneously . . . this does not seem to happen. Why not? This
may be simply explained if primary data are needed to deduce its existence . . .
as well as to determine other properties of this construction.17 According to
the definition of markedness provided above, and given Koopmans observation,
predicate cleft would appear to constitute a marked option in Universal Grammar.
Nonetheless, it is a salient feature of Haitian as well as of other Caribbean

74

The research methodology

creoles (see chapter 12). Therefore, on the one hand, the presence of this construction in Caribbean creoles constitutes a counterexample to Bickertons claim
that creole languages represent the unmarked case. On the other hand, with
respect to the claims in Muysken and Smith (1986a), Thomason (1993) and
Singler (1988) cited above, the presence of predicate cleft in Caribbean creoles
would constitute an argument in favour of substratum influence on these creoles,
since this construction is widespread among their substratum languages. In my
view, the simplest explanation of the fact that predicate cleft is found in Caribbean creoles is that the creators of these creoles had this option in their own
grammars and used it in creating the new language (see chapter 12). Whether the
construction happens to be marked or not is not crucial for the view advocated
in this book.
Let us now consider the opposite case of an unmarked feature found in both
the substratum languages and the creole itself. For the purposes of this discussion, I will use pied-piping of prepositions, which contrasts with preposition
stranding in particular grammars. Pied-piping of prepositions is considered to be
an unmarked option of Universal Grammar because it need not be learned by the
language learner. By contrast, preposition stranding, as in What did you do this
for?, is considered to be a marked option because it needs to be learned on the
basis of primary data (see Van Riemsdijk 1978, for an extensive discussion of
this construction in relation to markedness). So, assume a situation where the
substratum languages of a given creole do not allow for preposition stranding;
assume also that the superstratum of that creole does allow preposition stranding
and that, furthermore, the creole emerging from these two types of contributing
languages does not. This hypothetical case is, in fact, found in some Englishbased creoles of the Caribbean area. Saramaccan (as discussed in Veenstra and
Den Besten 1995: 315) appears to exemplify this situation: Gbe languages do
not allow for preposition stranding, English does, but Saramaccan does not. As
is the case in Gbe languages, Saramaccan has a resumptive pronoun strategy for
extraction out of a pp in relative clauses. According to Bickertons view, the
creole data would be considered to constitute the unmarked case, and Universal
Grammar would be called upon as the explanation for the linguistic facts. In
Muysken and Smith (1986a), Thomason (1993) and Singlers (1988) view, as
quoted above, the option chosen by the creole could not be used to argue for
substratal influence on the creole because pied-piping is an unmarked option of
Universal Grammar. Again, a much simpler explanation is available, namely,
that the creole ended up the way it is because this is the option that its creators
had in their native grammars. Of course, here again, quantity is important. If
parallels between a creole and its substratum languages can be shown in case
after case, we have a strong argument, and whether these parallel cases are
marked or not is not crucial for our hypothesis.
The above statement is akin to one made by Mufwene (1990, 1991), who
advocates quite a different definition of markedness from that discussed above.
Mufwene has suggested that, regardless of whether a feature is marked or not in

3.6 The data

75

the technical sense, if it is common to the substrate languages, it is the unmarked

case for the speakers of this language group. In his view, a feature shared by the
substratum languages is likely to be found in the creole (see also Singler 1988).
Mufwenes views on this question, which I believe to be right, make the markedness issue, as set out by the above-mentioned authors, less significant to the
debate on creole genesis than has been claimed. If it is undoubtedly a coincidence
that the Mande languages of West Africa and the Mayan languages of Central
America have strikingly similar distinctions between alienably and inalienably
possessed nouns (Welmers 1973: 1), the systematic parallels found between a
creole language and its substratum languages cannot be considered coincidental.
In light of this discussion, I conclude that the test of the hypothesis that the
creators of a creole construct it using the properties of their own lexicons and
grammars need not depend on the degree of markedness (as defined by Bickerton
and the other authors discussed above) of the features involved. Either the vast
majority of creole lexical entries follow from the relexification hypothesis or they
do not. Either the parametric values of the creole are those of the substratum
languages or they are not. Again, according to the view adopted in this book,
the real test of the hypothesis involves a global comparison of the lexicons and
grammars of the languages involved.
3.6

The data

The literature on Haitian makes reference to three geographical dialects: Northern,

Central and Southern. Romain (1959: 112), however, has proposed that there are
in fact only two dialects: the Northern one and the rest. Orjala (1970: 130) comments: This observation comes close to being true. If the number and extent of
dialect differences alone is taken into consideration, one might conclude that
there are only two areas. In his extensive study of Haitian dialectology, Orjala
(1970) outlines the dialectal differences between the Northern and Central/
Southern dialects. The most often cited difference between the Northern dialect
and the others involves the expression of possession in noun phrases. For example,
while Northerners use the preposition a to introduce the possessor phrase, the
speakers of other dialects do not; hence, k a mwen my heart in the North contrasts with k mwen my heart elsewhere (see Orjala 1970: 36; see also Goodman
1964). Orjala reports that the Northerners perceive their way of speaking as
being much closer to French than the other Haitian dialects. All regional dialects
are claimed to be mutually intelligible.
Fongbe is spoken mainly in Benin (the former Dahomey). Three regional
dialects have been studied: the dialects of Abomey, of Ouidah and of Cotonou/
Porto Novo. The Fon people constitute the largest ethnic group of the Kingdom
of Abomey and the Fongbe speakers of Abomey claim that they speak true
Fongbe. Ouidah is one of the ports from which African slaves were shipped
to the French colonies. Cotonou is the capital city of the country, and Porto
Novo the governmental administrative city. There are differences between these

76

The research methodology

geographical dialects (as will be pointed out in the various chapters of this book)
but they are all mutually intelligible.
Work done within the framework of this research is based on original data
collected by the members of the various projects (see Preface) during field trips
to Haiti and Benin, as well as during informant sessions with native speakers in
Montreal. The Haitian informants were all from the Centre or South of Haiti.
The Haitian speakers we worked with in Montreal were all born in Haiti and
had come to Montreal as adult immigrants. According to Labelle et al. (1993),
there are more than 50,000 Haitians living in Montreal. The majority of these
people immigrated to Montreal between 1970 and 1980. They speak Haitian at
home and in their community. The Fongbe informants come from three different regions of Benin: Abomey, Cotonou/Porto Novo and Ouidah. The Fongbe
speakers we worked with in Montreal were all born in Benin. The majority of
them were foreign students who spent a few years studying in Montreal. Some
15 native speakers of Haitian and 15 native speakers of Fongbe were involved
as informants. Furthermore, several of the students involved in the various projects underlying this research were native speakers of Haitian creole or of Fongbe.
One student was a native speaker of Ewe.
The research reported on in this book is based on the grammar of those native
speakers with whom I have done intensive work over the last twenty years, the
grammar of the speakers reported on by the various researchers who participated
in this research, and data already established in the literature. Dictionaries and
grammars of Haitian18 and Fongbe19 were also consulted. Variations in the data
will be taken into account, whenever such data are available. No attempt has
been made, however, to account for all the possible variations that I suspect
exist in both of these languages. The sources for the French data were identified
in section 3.4.
3.7

Mode of presentation of the data and analyses

As can be seen from the table of contents, the historical derivation of the Haitian
lexicon and grammar will be presented in subsets of data which constitute coherent subsystems of the lexicon and grammar of Haitian and of its contributing
languages. The properties of the lexical entries, paradigms and parameters will
be discussed in detail for the three languages being compared.
In some cases, I will begin by presenting the properties of a set of Haitian
facts, show the discrepancy between these properties and those of French, in
spite of similar phonological representations, and resolve the issue by presenting the Fongbe data. In other cases, I will start by showing the discrepancies
between the Fongbe and French data, presenting the facts as a problem to be
resolved by the creators of Haitian. The resolution of the problem turns out to
be what we see in Haitian as it is spoken today. In still other cases, Fongbe and
Haitian will be contrasted with French, showing that French differs from both
other languages in the same way.

3.7 Mode of presentation of the data and analyses

77

In view of the discussion in section 3.5, I will not argue whether each case of
parallelism between Haitian and Fongbe could be interpreted in terms of Universal Grammar or of second language acquisition strategies. Rather, I will study
each case for its own sake within the global comparative perspective advocated
in section 3.4.
Given the hypothesis put forward in this book, mismatches between the properties of Haitian and French are expected. As mentioned in section 3.4, however,
mismatches between Haitian and Fongbe constitute cases which require further
investigation. These cases will be discussed within each chapter of this book and
identified as revealing one of the three possibilities discussed in section 3.4:
dialect levelling, contribution from French or innovation from within the creole.
Similarly, in order to preserve coherence in the presentation of the various subsets of data, cases of reanalysis will be discussed within the context of each subset.
Finally, I will identify topics and questions for further research as I discuss the
various subsets of data.
The analyses are presented within the model of Principles and Parameters
(see Chomsky 1981, 1986, 1989 and related work). For one thing, the research
was carried out within this theoretical framework. Moreover, it is available to a
wider range of readers than the more recent minimalist model. Technical terms
are explained in such a way that the analyses are accessible to general linguists
and to linguists working within the framework of other models.
The phonemic inventories and orthographic conventions of the three languages under comparison are provided in Appendix 2.

78

Functional category lexical entries involved in nominal structure

Functional category lexical entries

involved in nominal structure

This chapter examines the origin of functional category lexical entries involved
in the Haitian creole nominal structure. In Haitian, as in Fongbe, a possessor
phrase, a demonstrative term, the definite determiner and the plural marker
may all be manifested within a single nominal structure. All are postnominal.
This is shown in (1).
(1)

krab [mwen ]
svn [nyy
twn]
crab me
gen
these/those crabs of

sa
a
yo
lv
v
lz
dem det pl
mine (in question/that we know of)

haitian
fongbe

This distribution contrasts with that found in French. As shown in (2), French
determiners occur before the noun and a French noun phrase may contain only
one determiner.
(2)

a. *le
mon
det poss
b.

ce
crabe
dem crab

french

1 le 5
2 mon 6
3 ce 7

crabe

french

1 the 5
2 my 6
3 this 7

crab

I argue that the Haitian lexical entries involving the [+definite] determiner, the
plural marker and the [+deictic] terms have been created through relexification.
The substratum languages so-called indefinite determiner was abandoned. Haitian developed its own indefinite form. Finally, the Case markers involved in the
nominal structure of Fongbe were assigned a null form at relabelling. Since Case
markers have no semantic content, they cannot be assigned a phonological form
in relexification. Evidence will be presented, however, showing that the Haitian
null forms have the same Case specifications as the overt Case markers of the
substratum language. This argues in favour of the claim that the substratum Case
markers were copied and assigned a null form during relexification.
78

4.1 The [+ definite] determiner

4.1

79

The [+
+ definite] determiner

This section argues that the lexical entry of the Haitian determiner has been
created through relexification. The data and analysis reported on in this section
are drawn from Lefebvre (1994a), based on a series of papers on the various
facets of this Haitian lexical entry (e.g. Lefebvre 1982b, 1992b; Lefebvre and
Massam 1988).
Haitian creole has a postnominal determiner la (with phonologically conditioned allomorphs a, an, nan and lan), as illustrated in (3). The presence of this
determiner indicates that the information conveyed by the noun phrase is part of
the shared knowledge of the participants in the conversation (see Fournier 1977;
Lefebvre 1982b; Lefebvre and Massam 1988). The Haitian determiner is not
marked for gender.
(3)

a. timounn nan
child
det
the child (in question/that we know of)

haitian

b. liv
la
book det
the book (in question/that we know of)

haitian

By contrast, the French determiner appears before the noun, as shown in (4), and
is specified for gender and number. Le is masculine singular, la is feminine
singular, les is plural, and l is a phonologically conditioned allomorph in the
singular.
(4)

a. l
enfant
det child
the child

b. le
livre
det book
the book

french

c. la
table
det table
the table

d. les livres / tables

det books / tables
the books/tables

french

In contrast with the Haitian determiner, the French determiner does not necessarily identify old/known information. According to Milner (1978: 23), the
definite determiner is either anaphoric, identifying an object that already has
been mentioned, or cataphoric. In the latter case, larticle annonce une relative
ou un gnitif sans quaucune mention antrieure ne soit requise. [The article
indicates a relative or genitive without any previous mention being necessary.]
According to Haase (1965) and Brunot (1905), the French determiner was optional
in Old and Middle French but it had become obligatory by the end of the sixteenth century.
The Haitian determiner cannot appear with nouns that have a generic or mass
interpretation; thus, Haitian allows for bare nps. By contrast, the French determiner must appear with such nouns (see Milner 1978: 25; Haase 1965; Brunot
1905). These facts are illustrated in (5) and (6), respectively.

80

Functional category lexical entries involved in nominal structure

(5)

Pen
(*an) bn pou lasante.
bread det good for health
Bread is good for ones health.

(6)

*(Le) pain est bon

pour la
(det) bread is good for
det
Bread is good for ones health.

haitian
(=(19) in Lefebvre 1994a)
sant
health

french
(=(24) in Lefebvre 1994a)

Furthermore, French has a partitive determiner de+la or du (a contracted form of

de + le), which appears with mass nouns, as in (7).
(7)

Jean a
mang du
John aux eat
de+le
John ate bread.

pain.
bread

french
(=(25) in Lefebvre 1994a)

According to Haase (1965), this partitive determiner has been attested in French
since the fifteenth century. Milner (1978: 24) points out the exceptional character of French in this respect, and he notes that in most languages there is no
determiner in contexts where the French partitive determiner occurs. Haitian
does not have a partitive determiner, as shown in (8), the Haitian counterpart of
the French sentence in (7).
(8)

Jan
manje pen.
John eat
bread
John ate bread.

haitian

On the basis of the above discussion, we can conclude that availability of bare
nps in Haitian cannot be derivable from French.
Finally, in Haitian, the head noun and the determiner may be separated by a
relative clause, as in (9).
(9)

Mounn [
ki
pati]
man
op res leave
The man who left.

a
det

haitian
(=(20) in Lefebvre 1994a)

By contrast, in French, the head noun and the determiner may not be separated
by a relative clause, as shown by the ungrammaticality of the sentence in (10).
(10)

*Le
[qui
est
parti] homme
det [who aux leave] man
[Lit.: The [who left] man]

french
(=(26) in Lefebvre 1994a)

The Haitian creole and French determiners thus have quite different semantic
and syntactic properties, which indicates that the Haitian determiners properties
are not derived from French. Moreover, the French determiner does not appear
to have been the source of its phonological representation either. The French
determiner is often found as part of Haitian simple nouns, as shown in (11)
(see also Baker 1984).

4.1 The [+definite] determiner

(11)

Haitian nouns
larivy
river
lakay
home
listwa
history
lat
world

<
<
<
<

Corresponding
la rivire
la case
lhistoire
la terre

81

French dps
the river
the house
history
the earth

The French partitive determiner illustrated in (7) above is also found as part of
Haitian simple nouns, as shown in (12).
(12)

Haitian nouns
dlo
water
dife
fire
diri
rice

<
<
<

Corresponding
de leau
du feu
du riz

French dps
water
fire
rice

The data in (11) and (12) show that the creators of Haitian did not identify the
French determiners that they were presented with as independent morphemes
and often perceived them as part of the phonetic strings of the nouns with which
they appear. Further support for this claim comes from the fact that Haitian
nouns which contain an agglutinated French determiner may occur with the
postnominal Haitian determiner, as shown in (13).
(13)

a. larivy a
river
det
the river

b. diri a
haitian
rice det
the type of rice (in question)
(=(29) in Lefebvre 1994a)

Consequently, like Sylvain (1936) and Fournier (1977), I conclude that the French
determiner could not have been the phonetic source of the Haitian determiner.
Where, then, do the properties of the Haitian determiner come from? I argue
below that, while its semantic and syntactic properties are derived from the substratum languages corresponding lexical entry, its phonological representation
is derived from the superstratum lexical item l which occurs at the end of
constituents.
Fongbe has a postnominal determiner V (with a phonologically determined
allomorph Vn1), as illustrated below.
(14)

a. vj
v
child det
the child (in question/that we know of)

fongbe

b. wm vn
book det
the book (in question/that we know of)

fongbe

This determiner, like the Haitian one, indicates that the information conveyed
by the noun phrase is part of the shared knowledge of the participants in the conversation (see Lefebvre 1992b). Thus, like the Haitian determiner (see (3) ), the
Fongbe determiner in (14) is obligatorily anaphoric. Furthermore, the Fongbe
determiner cannot appear in noun phrases that have a generic or a mass interpretation, as shown in (15), which parallels the Haitian data in (5).

82
(15)

Functional category lexical entries involved in nominal structure

Wwxx (*v) nyvn n lnmyyn.
bread
det good for health
Bread is good for ones health.

fongbe
(=(19) in Lefebvre 1994a)

As is the case in Haitian (see (8) ), Fongbe has no partitive determiner and bare
nouns are interpreted as mass nouns.
(16)

Kwk 3 bl3.
Koku eat bread
Koku ate bread.

fongbe

Thus, like Haitian, Fongbe allows for bare nps.

In both languages, a determiner that co-occurs with a mass noun yields the
interpretation type of x.
(17)

diri
a
mwlk v
rice
det
the type of rice (in question)

haitian
fongbe

In Fongbe, like Haitian (see (9) ), the head noun and the determiner may be
separated by a relative clause, as in (18).
(18)

Sn [3
y]
v.
man
op res leave det
The man who left.

fongbe

Furthermore, in both Haitian and Fongbe, but not in French, the determiner
is required in the syntactic structure of relative clauses (see Koopman 1982a;
Lefebvre 1986), and conditional and factive clauses (see Lefebvre and Massam
1988; Kinyalolo 1993a; Collins 1994). Finally, as will be discussed in detail in
chapter 8, in both Haitian and Fongbe, but not in French, the determiner plays
a central role in clause structure (see Lefebvre 1992b, 1996a, to appear b) and,
for some speakers, it may appear in constructions involving verb-doubling phenomena (see Lefebvre and Ritter 1993; Lefebvre 1994b) (see chapter 12).
The above data show that the semantic and syntactic properties of the Haitian
creole determiner are strikingly similar to those of the Fongbe determiner. Both
languages contrast with French in the same way with respect to the properties of
this functional category lexical entry. It thus seems that substratum languages of
the type of Fongbe were the source of the Haitian determiners properties.
Following the analysis in Lefebvre and Massam (1988) and Lumsden (1989,
1991) for Haitian, and the analysis in Brousseau and Lumsden (1992) for Fongbe,
I assume that the forms la and V (and their allomorphs) are the head of the
functional category phrase dp. These determiners are specified for the features
[+definite] and [+anaphoric], and their complement occurs to the left.

4.1 The [+definite] determiner

83

What, then, is the source of the phonological form of the Haitian creole
determiner? It has been proposed that the French form l is a good candidate for
the phonetic source of Haitian la (see e.g. Faine 1937; Goodman 1964; Fournier
1977; Valdman 1978; Lefebvre 1993b). In addition to its use as a locative
adverb meaning there, French l may appear following the noun in a noun
phrase as an emphatic deictic marker, as in (19), or as a discourse marker (see
Vincent 1984) after a noun phrase or a clause, as in (20). Note that the use of l
shown in (20) has been attested in French since at least the seventeenth century
(see Fltre 1970: 141, 146; Hull 1975: 2).
(19)

Cet homme -l
vient darriver.
dem man [+deic] come prep arrive
This/that man just arrived.

french

(20)

a. L
homme l
vient darriver.
det man
dm come prep arrive
The man [there] just arrived.

popular french

b. Un homme l
vient darriver.
det man
dm come prep arrive
A man [there] just arrived.

popular french

c. Un / l homme l, qui vient darriver

l . . .
popular
det
man
dm who come prep arrive dm
french
A/the man [there] who just arrived [there] . . .
(=(30) in Lefebvre 1994a)

These examples show that the French form ls distribution is somewhat parallel
to that of the Fongbe determiner and, furthermore, the deictic interpretation of
this French form (see (19) ) overlaps with the Fongbe determiners interpretation. Thus, l, pronounced [lu] or [l8] in popular French, was probably the
source of the phonological form of the Haitian determiner la. As is pointed out
in Lefebvre (1993b, 1994a, 1996a), the syntactic and semantic properties of the
French form are not the same as the Haitian determiners. The French l can
appear with definite or indefinite noun phrases (see (20) ), whereas the Fongbe
and Haitian determiners are incompatible with indefinite noun phrases. Furthermore, when it is used in the context of a clause (see (20c) ), l has no semantics
in common with the Haitian determiner. Thus, the French form l may have
been the source of the phonological representation of the Haitian creole determiner but it did not contribute its syntactic and semantic properties.
The creation of the lexical entry for the Haitian determiner thus proceeded as
depicted in (21), which should be read in light of the general schema in chapter 2 (see (2) ). The copied lexical entry corresponding to the Fongbe determiner
V was relabelled on the basis of the phonetic matrix of the French form l,
yielding the Haitian determiner la.

84
(21)

Functional category lexical entries involved in nominal structure

fongbe lexical entry

french

//
[+definite]
[+anaphoric]
complement / head

[l8 / lu]
used in semantic and
pragmatic contexts

haitian creole lexical entry

/la/
[+definite]
[+anaphoric]
complement / head

Interestingly enough, in Ewe (see Wallace 1995a) the postnominal determiner

has the forms or l and it has semantic and syntactic properties very similar to
those of the Fongbe determiner. In Kihms (1994) view, the phonological form of
the Haitian determiner constitutes a case of phonological conflation. The Haitian
phonological form la was then subject to phonological processes discussed in
Brousseau (in preparation) which produced the various allomorphs that we find
in present-day Haitian.
4.2

The plural marker

This section argues that the lexical entry of the Haitian plural marker yo has
been created by relexification. The data involving this Haitian lexical entry,
however, are of further interest for yo is also the manifestation of the thirdperson plural personal pronoun.
As shown in Lefebvre (1994a), the Haitian plural marker yo shares a number
of properties with the Fongbe plural marker lZ. Both plural markers occur postnominally, as shown below.
(22)

krab yo
svn lz
crab pl
the crabs
*(some) crabs

haitian
fongbe

(=(31) in Lefebvre 1994a)

A noun followed by the plural marker alone is always interpreted as [+definite].

It can never be interpreted as [definite], as shown in (22). The indefinite plural
is not signalled, as shown in (23).
(23)

M achte krab.
N xw
svn.
I
buy
crab
I bought (some) crabs.

haitian
fongbe
(=(32) in Lefebvre 1994a)

In both Haitian and Fongbe, for some speakers, the plural marker may occur
within the same noun phrase as the determiner.2 In this case, at S-structure, the
plural marker follows the determiner, as shown in (24).

4.2 The plural marker

(24)

krab la
yo / *yo a
svn v
lz / *lz v
crab det pl
the crabs (in question)

85

haitian
fongbe
(=(33) in Lefebvre 1994a)

As shown in (25), in these languages, the singular is not indicated by a specific

marker.
(25)

krab la
svn v
crab det
the crab (in question)

haitian
fongbe
(=(34) in Lefebvre 1994a)

The properties of the Haitian plural marker discussed so far are similar to those
of the corresponding marker in Fongbe.
These properties contrast systematically with the expression of number in
French. First, in French, the feature [+/plural] is an obligatory feature of the
prenominal determiners and is expressed within them. While le and la are the
singular forms of the definite determiner, les is the plural form.
(26)

Jean a
mang la
pomme /
John aux eat
det apple
John ate the apple(s).

les
pommes.
det apple

french

(=(36) in Lefebvre 1994a)

Second, in contrast to Haitian (see (22) ), in French, the expression of plural is

obligatory even when a noun is indefinite. This is shown in (27) where des is the
contracted form of de+les.3
(27)

Jean a
mang des
pommes.
John aux eat
de+les apple
John ate (some of the) apples.

french
(=(37) in Lefebvre 1994a)

Thus, so far there is a systematic parallel between Haitian yo and Fongbe lZ,
which both contrast with the expression of number in French. Haitian and Fongbe
differ, however, in that the Haitian morpheme yo is also used as a third-person
plural personal pronoun. This is shown in (28).
(28)

a. krab yo
crab pl
the crabs

b. Yo
pati.
3rd-pl leave
They left.

haitian

In Fongbe, the third-person plural personal pronoun is expressed by a different

morpheme, as shown in (29).
(29)

a. svn lz
crab pl
the crabs

b. Yi
y.
3rd-pl leave
They left.

fongbe

86

Functional category lexical entries involved in nominal structure

As Mufwene (1986) asks, how can the difference between the two languages be
accounted for within the framework of the relexification hypothesis?
Sylvain (1936) and Goodman (1964) have suggested that the use of yo as a
plural marker is in fact an extended usage of the third-person plural pronoun.
In their view, the phonological representation of yo would be derived from
the strong form of the third-person personal pronoun eux. The pronoun eux in
French occurs as an emphatic form before a clitic, as in (30).
(30)

Eux, ils
mangent du
riz
them, they eat
part rice
Them, they eat rice every day.

chaque jour.
every
day

french

(=(40) in Lefebvre 1994a)

Furthermore, in colloquial French, the third-person pronoun eux may also occur
as an emphatic pronoun at the end of a nominal constituent, as shown in (31)
(A.-M. Brousseau, p.c.). Note the neutralisation in gender shown in (31b), also
noted by Gougenheim (1973).
(31)

a. Les gars, eux,

ils . . .
det guys them, they . . .
The guys, them, they . . .
b. Les filles, eux,
ils . . .
det girl
them, they
The girls, them, they . . .

french

popular french
(=(41) in Lefebvre 1994a)

The French pronoun eux thus has a similar surface distribution to the Haitian
form yo. It is clear, however, that it shares only a few semantic features with yo
since eux is only pronominal; unlike yo, it is never used as a plural marker.
These facts suggest that, if the Haitian form yo derives its phonological representation from the phonetic matrix of the French form eux (a predictable phonological derivation, as is argued in Brousseau in preparation),4 that is not where it
derives its semantic properties. How did Haitian yo come to be used as both a
third-person plural pronoun and a plural marker, while the corresponding form
in French is only used as a pronoun and the corresponding form in Fongbe is
only used as a plural marker?
Interestingly enough, several West African languages present cases of
homophony between the third-person plural pronoun and the plural marker. For
example, in Ewe, wK is the form for both the third-person plural pronoun and the
plural marker (see Westerman 1930: 45, 57). According to Goodman (1964: 46,
47), Yoruba, Mandingo and other languages also show this type of homophony.
According to Mufwene (1986: 138), in Vai, a Mande language, the third-person
plural pronoun and the plural marker are also homophonous. The proposal in
Ritter (1992) provides a theoretically motivated account of this homophony.
Since the determiner and the plural marker can co-occur in a single noun phrase,
they cannot head the same projection. While the determiner is the head of dp,
the plural marker is the head of Num(ber)P. Ritter argues that, while first- and

4.3 The so-called indefinite determiner

87

second-person pronouns are of the category D(eterminer), third-person pronouns

are of the category Num(ber). According to her analysis, third-person pronouns
are derived by raising Num to D, as is schematically represented in (32).
(32)

1st/2nd-person pronoun

3rd-person pronoun

dp

dp

NumP
Num
ti

d
[+definite]

yoi
(from Ritter 1992)

On this account then, both the plural marker and the third-person plural pronoun
belong to the category Num. Whereas in Haitian creole, Yoruba, Mandingo etc.,
there is only one lexical entry fulfilling these two functions, in Fongbe the
category Num is realised by two different morphemes: lZ and yI, respectively
(see (29) ). It thus appears that, in this case, Haitian follows the pattern provided
by substratum languages other than Fongbe.
Fongbe lZ cannot be claimed to have been relexified by yo, for yo, unlike lZ,
is also the third-person personal pronoun. Following the proposal in Sylvain
(1936) and Goodman (1964), let us suppose that the third-person plural pronoun
(yI in Fongbe, wK in Ewe, etc.) was relabelled as yo on the basis of French eux.
(See chapter 6, where it is argued that the strong personal pronouns of the substratum languages were relexified.) Let us suppose further that, having relexified
this pronominal form, speakers of the substratum languages for whom this form
also encoded plurality in nps extended the use of early Haitian yo to encode
plurality in Haitian nps. In this view, Fongbe speakers, who had two different
morphemes to encode the third-person personal pronoun and plurality in the
noun phrase, would have abandoned the plural morpheme of their own language
and, in line with other substratum languages, adopted the extended use of the
relexified third-person pronoun. As has been pointed out in Lumsden and Lefebvre
(1994), this constitutes a simple case of dialect levelling where other substratum
languages than Fongbe imposed the properties of their lexicons upon the new
language. The above discussion illustrates how dialect levelling operates on the
basis of the early creole lexicons relexified from various substratum languages.
This process role in creole genesis accounts for the fact that the properties of
some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages.
4.3

The so-called indefinite determiner

Much more work needs to be done on the so-called indefinite determiners in the
three languages under comparison. In this section, I outline the basic facts and
raise some questions for further research.

88

Functional category lexical entries involved in nominal structure

Along with the definite determiner, discussed in section 4.1, Fongbe has a
form # (a reduced form of the numeral # one) which serves as an indefinite
determiner as in sVn # a crab. The form # is in complementary distribution
with the definite determiner V but it may co-occur with the plural marker lX, as in
sVn # lX some crabs. Furthermore, some speakers (but not all) distinguish
between the form # and the form #I. For these speakers, #I is assigned the
meaning unique and it is only used in contexts where the object referred to
is specific. Other speakers do not distinguish between these two forms. For
example, Segurola (1963) presents only one lexical entry, which has both meanings discussed above. At first glance, # would appear to be in a paradigmatic
relationship with V and #I would appear to have the status of an adjective.
These forms give rise to the following questions for future research. First, do
they constitute two lexical entries? If so, does the feature [specific] distinguish
them? What is their precise meaning? What is their syntactic status? Second, #I
interacts with other morphemes. For example, m^-#I V means the other, m^-#I
lX means the others, etc. An analysis of these morphemes should be based on a
complete documentation of their distribution.
French has a prenominal form un(e), homophonous with the numeral n(e)
(where the accent signals stress) (see Milner 1978), which serves as an indefinite
determiner as in un crabe a crab. Some of the properties of un(e) are discussed
by Milner (1978), Valois (1991), and Bernstein (1993) but these authors do not
provide clear analyses of the kinds of objects (specific or otherwise) this French
form takes, nor do they propose any definite syntactic analysis.
Finally, Haitian has a prenominal form yon, phonologically similar to the
numeral younn one, which serves as an indefinite determiner as in yon crab a
crab. The properties of this form have yet to be studied in detail. For example,
does it take only specific objects? How does it interact with other morphemes
(e.g. yon lt the other, lt yo the others, etc.)? What is the syntactic status of
yon? As is pointed out in Lefebvre (1994a), yon cannot be the head of dp, for
if it were, we would expect it to occur in the same position as the [+definite]
determiner, i.e. postnominally. Yon cannot be the head of Num(ber)P either, as
is claimed by DeGraff (1992b), for if it were, we would expect it to occur in the
same position as the plural marker yo, again postnominally.
In a preliminary comparison of the properties of the indefinite forms in the
three languages in light of the relexification hypothesis, Lefebvre (1994a) establishes the following points. First, the postnominal Fongbe form #/#I was not
relexified. Second, the French form indefinite un a was not identified as such
by the creators of Haitian since Haitian yon cannot possibly have been phonologically derived from un /r/. According to Brousseau (in preparation), /r/ is
not part of the phonemic inventory of Haitian. The phoneme /r/ in French words
is realised as /s/ in the corresponding words in Haitian (see Appendix 2). If yon
had been derived from un [r], it would be the only case of this phonetic correspondence in the language. Lefebvre (1994a) proposes that the Fongbe numeral
# one was relexified as younn on the basis of the French form ne. Brousseau
(in preparation) argues that the interpretation of ne as younn by the creators of

4.4 The [+ deictic] terms

89

Haitian is of the same type as the interpretation of eux as yo. Lefebvre (1994a)
further proposes that the Haitian form yon evolved within the creole as the weak
form of younn through a process of resyllabification discussed in Brousseau (in
preparation). In this view, yon occurs prenominally because it evolved from the
numeral younn and numerals are prenominal in Haitian (e.g. de ti-mounn two
children). The Haitian word order follows the French word order (e.g. deux
enfants two children) and contrasts with the Fongbe (e.g. vJ w, literally: child
two, two children). This is predicted by the hypothesis set out in chapter 2,
according to which the word order of lexical categories in the creole follows the
superstratum word order, unlike that of functional categories, which follows the
word order of the substratum languages.
Since there is no functional category projection in the nominal structure to
host the so-called Haitian indefinite determiner yon, could it occupy the same
position as numerals? A decision as to whether yon in modern Haitian has the
properties of the corresponding lexical entries in Fongbe or in French or whether
its properties are independent of both sources will have to await further research.
It is clear, however, that any conclusion must be based on careful documentation
and analysis of the data in the three languages, a difficult task given the scope of
the problem posed by indefinites, as shown in the recent literature on the topic.
4.4

+ deictic] terms
The [+

The deictic terms of Haitian provide yet another example of functional category
lexical entries which have been created through relexification. The data and
analysis presented in this section are drawn from Lefebvre (1997). In the languages of the world, the function of demonstrative terms is to point at objects;
demonstrative terms are thus universally [+deictic]. Other properties of demonstrative terms may, however, vary among languages. For example, in some
languages, certain demonstrative terms may only be used to point at objects that
are close to the speaker. Such demonstrative terms are specified for the feature
[+proximate]. Conversely, some demonstrative terms may be used only to point
at objects that are far from the speaker. These terms are specified for the feature
[proximate]. Finally, some demonstrative terms are not marked for any value
of the feature [proximate], and may be used for objects that are either close to
or far from the speaker. Demonstrative terms may vary in their specification for
categorial features. Demonstratives may have the syntactic properties of nouns;
in this case, they are specified for the categorial features [+n, v] (categorial
features which also identify nouns). Alternatively, they may occur as part of the
determiner system in nominal structures; in this case, they are not specified for
the major features [+n, v]. Demonstrative terms may also belong to the category adverb, like here and there in English. Furthermore, demonstratives may
optionally be marked for animacy and other features such as gender and number.
Since not all the properties of demonstrative terms are universal, these lexical
items offer an interesting subset of data with which to test the role of relexification
in creole genesis.

90

Functional category lexical entries involved in nominal structure

In Haitian, there is a paradigm of demonstrative terms which consists of two

forms, sa and sila (see also e.g. Goodman 1964: 50; Sylvain 1936). In the
grammar of the Haitian speakers who provided the data on which the following
analysis is based, sila is used only to point at objects that are far from the speaker,
whereas sa is used as a general deictic term to designate objects that are either
close to or far from the speaker. So, while sila is specified as [proximate], sa is
not specified for any value of the feature [proximate].5 In the examples below,
the meaning of sila is rendered as that and sa is translated as this/that.
Furthermore, as is pointed out in Valdman (1978), these Haitian demonstratives
may be used to point at objects which are either animate or inanimate, as illustrated in (33), so they are not specified for the feature [animate].
The [+deictic] determiners sa and sila appear postnominally, as shown in
(33) below.
(33)

a. bag sa //
sila
ring [+deic] [prox]
this/that // that ring

b. mounn sa //
sila
haitian
man
[+deic] [prox]
this/that // that man
(=(2) in Lefebvre 1997)

Furthermore, sa and sila may occur within the same noun phrase as the possessive marker, the [+definite] determiner, and the plural marker. This is shown in
(34).
(34)

bag mwen sa //
sila
a
yo
haitian
ring poss
[+deic] [prox] det pl
these/those // those rings of mine (in question/that we know of )
(=(3) in Lefebvre 1997)

Given that sa and sila occur within nominal structures as in (33) and (34), we
can safely hypothesise that they constitute a class of determiners defined by the
feature [+deictic]. As such, they are not identified for the categorial features
[+n, v].
Sa and sila may, however, also appear in contexts where we expect an np.
In (35), they appear as part of an argument of the verb.
(35)

M w
sa
// sila.
I
see one [+deic] // [prox]
I saw this/that // that one.

haitian
(=(4) in Lefebvre 1997)

In (36), they appear as part of the head of a relative clause (see also Valdman
1978: 207).
(36)

a.
sa

ki
vini
one [+deic] op res come
This/that one who came.
b.
sila

ki
vini
one [prox] op res come
That one who came.

an.
det
an.
det

haitian

haitian
(=(5) in Lefebvre 1997)

4.4 The [+deictic] terms

91

Furthermore, the data in (37) show that demonstrative terms that occur as part of
the head of a relative clause may be followed by the plural marker yo.
(37)

a.
sa
yo ki
vini
one [+deic] pl op res come
These/those ones who came.
b.
sila
yo
ki
vini
one [prox] pl op res come
Those ones who came.

an.
det
an.
det

haitian

haitian
(=(6) in Lefebvre 1997)

How are the data in (35), (36) and (37) compatible with the earlier claim that
sa and sila are [+deictic] determiners? In Lefebvre (1997), it is argued that,
when sa and sila occur in contexts such as (35)(37), they are part of a nominal
structure headed by a phonologically null head noun corresponding to one in
English. This analysis is supported by the fact that the null head may also be
modified by an adjective, as shown in (38).6
(38)

a. M vle
gwo
sa
I
want big one [+deic]
I want this/that big one.

a.
det

haitian

b. M vle
gwo
sila
I
want big one [prox]
I want that big one.

a.
det

haitian
(=(7) in Lefebvre 1997)

The semantic and syntactic properties of the lexical entries sa and sila are
presented in (39).
(39)

a. /sa/
[+deictic]

this/that

haitian

b. /sila/
[+deictic]
[proximate]

that

haitian
(=(8) in Lefebvre 1997)

French has a series of demonstrative terms which can be divided into three
major groups on the basis of their categorial status: they are either determiner,
pronominal or adverbial.
The forms ce(t), cette and ces constitute the first group of French deictic
terms. These forms are nominal determiners. They agree in gender and number
with the noun they determine, so they bear gender and number features: ce(t)
(ms. sg.), cette (fem. sg.), ces (pl.). As is shown in (40), these forms occur
prenominally. They can be used with animate (40a) or inanimate (40b) objects.
Therefore, they are not specified for any value of the feature [animate]. Furthermore, they can be used to designate an object that is either close to or far
from the speaker; they are therefore not specified for any value of the feature
[proximate].

92
(40)

Functional category lexical entries involved in nominal structure

a. ! ce #
@ ces $

garon(s)

french

! this/that
# boy(s)
@ these/those $
b. ! cette #
@ ces $

bague(s)

french

! this/that
# ring(s)
@ these/those $
(=(9) in Lefebvre 1997)

The demonstrative terms in (40) cannot occur with other determiners, as shown
in (41) (order irrelevant).
(41)

a. *la
ma
cette
bague
the poss [+deic] ring

french

b. *les mes ces

bagues
the poss [+deic] ring

french
(=(10) in Lefebvre 1997)

The ungrammaticality of the data in (42) shows that these [+deictic] determiners
cannot modify an np headed by a phonologically null head.
(42)

*J
I

ai
vu
aux see

ce / cette / ces
[+deic]

french
(=(11) in Lefebvre 1997)

The semantic and syntactic properties of the lexical entries ce(t), cette and ces
are presented in (43).
(43)

a. ce(t): /s1/ ~ /s=t/

[+deictic]
[fem]
[pl]

this/that

french

b. cette: /s=t/
[+deictic]
[+fem]
[pl]

this/that

french

c. ces:

these/those

french

/s=/ ~ /se/
[+deictic]
[+pl]

(=(12) in Lefebvre 1997)

The French forms in (43) share certain properties with the Haitian forms in
(39): the deictic terms are determiners, and they are not specified for a value of
the feature [animate] in either language. However, the French and Haitian
forms differ with respect to other properties: the Haitian forms can co-occur
with other determiners (see (34) ), but the French forms cannot (see (41) ); the

4.4 The [+deictic] terms

93

French forms cannot determine a phonologically null head (see (42) ), whereas
the Haitian forms can (see (35), (36), (37), (38) ); and the French lexical entries
bear gender and number features, while the Haitian forms do not. The semantic
and syntactic properties of the Haitian deictic terms sa and sila do not correspond to the properties of the French demonstrative determiners ce(t), cette and
ces; it is unlikely, then, that the Haitian terms properties could be derived from
these French determiners.
The second class of French demonstrative terms contains two sets of pronominal forms, which are distinguishable on the basis of animacy. The three
pronominal forms a, cela and ceci make up the first set. Since these forms
may only refer to inanimate objects, they must be specified for the feature
[animate]. a and cela are general deictic terms which may be used to point at
objects that are either close to or far from the speaker;7 they are not specified for
any value of the feature [proximate]. By contrast, ceci is used only to pick out
objects close to the speaker. Consequently, it must be specified for the feature
[+proximate].
Because a, cela and ceci are strong pronouns8 and, as such, marked for the
categorial features [+n, v], they do not occur as nominal determiners, as shown
in (44).
(44)

*a / cela / ceci livre

dem
book
[Lit.: this/that book]

french
(=(13) in Lefebvre 1997)

Since they are specified as [+n, v], we expect these deictic forms to be able to
occur in an argument position of the verb. They can, as shown in (45).
(45)

J ai
vu a
// cela
I aux see [+deic] // [+deic]
I saw this/that // this/that // this.

//
//

ceci.
[+prox]

french
(=(14) in Lefebvre 1997)

They do not occur as the head of a relative clause, however, as shown by the
ungrammaticality of the sentence in (46).
(46)

*a/cela/ceci que jai vu

What I saw.

french
(=(15) in Lefebvre 1997)

In this context, the form ce is used, as shown in (47).

(47)

Ce que jai vu.

What I saw.

french
(=(16) in Lefebvre 1997)

The semantic and syntactic properties of the lexical entries a, cela and ceci
are summarised in (48).

94
(48)

Functional category lexical entries involved in nominal structure

a. a:

/sa/
[+pro]
[+deictic]
[+n, v]
[animate]

this/that

french

b. cela:

/s1la/
[+pro]
[+deictic]
[+n, v]
[animate]

this/that

french

c. ceci:

/s1si/
[+pro]
[+deictic]
[+proximate]
[+n, v]
[animate]

this

french

(=(17) in Lefebvre 1997)

The properties of the French forms in (48) are rather different from those of
the Haitian forms in (39). The French forms are pronominal and are specified as
[animate], whereas the Haitian forms are not pronominal and are not specified
for any value of the feature [animate]. Being pronominal, the French forms
cannot be modified by an adjective, in contrast to the Haitian forms (see (38) ).
Furthermore, while the paradigm of French pronouns in (48) includes one which
lexically encodes the feature [+proximate], the paradigm of Haitian forms in
(39) contains a form which lexically encodes the feature [proximate]. Thus, a,
ceci and cela are unlikely to be the source of the semantic and syntactic properties of the Haitian demonstrative terms.
The forms celui, ceux and celle(s) constitute the second set of demonstrative
pronouns in French. These pronouns are used to point at objects that are either
animate or inanimate and therefore are not specified for any value of the feature
[animate]. They are specified for gender and number: celui (ms. sg.), ceux
(ms. pl.), celle(s) (fem. sg./pl.). They are neutral with respect to the feature
[proximate]. The data in (49) show that celui, ceux, celle(s) do not occur as
nominal determiners.
(49)

*1 celui 5
2 ceux 6
3 celle(s) 7
[+deic]

livre(s)

french

book
(=(18) in Lefebvre 1997)

They do, however, occur as the head of a relative clause, as shown in (50).
(50)

Celui
qui est
Ceux
qui sont
Celle(s) qui est / sont
[+deic] who aux
This/that/these/those who

venu
venus
venue(s)
come
came

french

(=(19) in Lefebvre 1997)

4.4 The [+deictic] terms

95

The semantic and syntactic properties of celui, ceux and celle(s) are shown in
(51).
(51)

a. celui:

/s1l2i/
[+pro]
[+deictic]
[+n, v]
[fem]
[pl]

this/that

french

b. ceux:

/s/
[+pro]
[+deictic]
[+n, v]
[fem]
[+pl]

these/those

french

c. celle(s):

/s=l/
[+pro]
[+deictic]
[+n, v]
[+fem]

this/that; these/those

french

(=(20) in Lefebvre 1997)

The French forms in (51) share only one property with the Haitian forms
in (39): none are specified for a value of the feature [animate]. Apart from
this, the other properties of the French forms contrast with those of the Haitian
forms: the French forms are specified as [+n, v], the Haitian forms are not;
the French lexical entries bear gender and number features, the Haitian forms
do not; none of the French forms in (51) are marked for a value of the feature
[proximate], but one of the Haitian lexical entries in (39), namely sila, is
specified as [proximate]. Thus, it seems that the semantic and syntactic properties of the Haitian deictic terms sa and sila are not derived from those of celui,
ceux, celle(s).
Finally, there is a third group of deictic terms in French: the adverbials l and
ci. These two forms combine with the deictic pronominal forms of the second
group, as in (52), or with the deictic determiners occurring in nominal structures,
as in (53).
(52)

a. J ai
vu celuiI aux see [+deic]
I saw this one.

ci.
[+prox]

french

b. J ai
vu celuiI aux see [+deic]
I saw this/that one.

l.
[+deic]

french
(=(21) in Lefebvre 1997)

(53)

a. cette
bague-ci
[+deic] ring [+prox]
this ring

b. cette
bague-l
french
[+deic] ring [+deic]
that/this ring
(=(22) in Lefebvre 1997)

96

Functional category lexical entries involved in nominal structure

As shown in the above examples, l is a general deictic form, not specified for
any value of the feature [proximate]. By contrast, ci is used to pick out objects
that are close to the speaker; it is specified for the feature [+proximate]. The
properties of the lexical entries l and ci are represented in (54).
(54)

a. l:

/lu/ ~ /l8/
[+deictic]
adv

there/here

french

b. ci:

/si/
[+deictic]
[+proximate]
adv

here

french

(=(23) in Lefebvre 1997)

Like the Haitian demonstrative terms, these two French lexical items occur
postnominally (see (52) and (53) ). They differ, however, with respect to categorial
features. While the Haitian demonstrative terms are determinative, l and ci are
adverbial. In French, the positive value of the feature [proximate] is lexically
encoded (e.g. ci), whereas in Haitian it is the negative value of the feature
[proximate] that is so encoded (e.g. sila). Thus, it appears that these French
lexical items are not the source of the properties of the Haitian deictic terms
in (39).
Chaudenson (1993) claims that the properties of the Haitian demonstrative
terms are generally derived from French demonstratives. The detailed comparison presented in this section shows that this cannot be correct. There is no
French form with exactly the same properties as the Haitian forms.
Could it be that French demonstrative terms had different properties at the
time the creole was formed? Historical evidence argues against such a possibility. Brunot (1926), Rosset (1911) and Dees (1971) show that a, cela and ceci
have had the properties described above since the seventeenth century. Dees
(1971) shows that celui, celle(s) and ceux have been attested in French since the
middle of the fourteenth century and that, by the seventeenth century, they had
acquired the properties they have in modern French. As for ce(t), cette and
ces . . . l, several authors (see Lommatzsch 1925; Guiraud 1966; Yvon 1951)
are of the opinion that they have been used as general deictic terms (to point at
objects that are either far from or close to the speaker) since the twelfth century.
According to Dees (1971), this confusion started only at the end of the fourteenth century.9 Whatever the outcome of this discussion of historical facts
might be, it is clear that the properties of the French demonstrative terms were
already established by the seventeenth century. Since Singler argues that Haitian
creole appears to have been created between 1680 and 1740 (see chapter 3), the
French data described in this section were those that the creators of Haitian were
exposed to. Therefore, the discrepancy between the properties of demonstrative
terms in Haitian and in French cannot be attributed to the fact that the properties
of French demonstrative terms were different at the time the creole was formed.

4.4 The [+deictic] terms

97

Why, then, did Haitian end up with the system described in (39)? The properties
of demonstrative terms of the substratum languages provide a clear answer to
this question.
In Fongbe, there is a paradigm of two demonstrative terms. Segurola (1963)
lists them as lV and nX. Anonymous (1983) lists them as lV or lV and nX or nX.
According to that author, these forms are generally pronounced lV and nX,
respectively, where emphatic - is not realised. For the Fongbe speakers whose
grammar is discussed in this section, while nX is used only to designate objects
that are far from the speaker, lV is used as a general deictic term for objects that
are either close to or far from the speaker. So, while the first form is specified as
[proximate], the second form is not specified for any value of the feature
[proximate].10 In the examples below, the meanings of nX and lV are rendered
as that and this/that respectively.11 Furthermore, the Fongbe demonstrative
terms may be used with objects which are either animate or inanimate (see
(55) ). Thus, they are not specified for the feature [animate].
The Fongbe demonstrative terms lV and nX have the following distributional properties. They are [+deictic] determiners and, as shown in (55), they
occur postnominally.
(55)

a. lwkx lv
// nx
ring
[+deic] // [prox]
this/that // that ring

fongbe

b. sn lv
// nx
man [+deic] // [prox]
this/that // that man

fongbe
(=(24) in Lefebvre 1997)

The [+deictic] determiners may occur within the same noun phrase as the possessive, the [+definite] determiner and the plural marker. This is shown in (56).
(56)

lwkx g
lv // nx
v
lz
fongbe
ring
poss [+deic] [prox] det pl
these/those // those rings of mine (in question/that we know of)
(=(25) in Lefebvre 1997)

Given that lV and nX occur within nominal structures as in (55) and (56), we
can safely hypothesise that they constitute a class of determiners defined by the
feature [+deictic]. As such, then, they are not specified for the categorial features
[+n, v].
However, lV and nX may also be found in contexts where we expect an np.
In (57), they appear as part of an argument of the verb.
(57)

N mw
lv
// nx.
I
see one [+deic] // [prox]
I saw this/that // that one.

fongbe
(=(26) in Lefebvre 1997)

98

Functional category lexical entries involved in nominal structure

In (58), they appear as part of the head of a relative clause.

(58)

a.
lv
3
w
one [+deic] op res come
This/that one who came

v.
det

fongbe

b.
nx
3
w
one [prox] op res come
That one who came

v.
det

fongbe
(=(27) in Lefebvre 1997)

The data in (59) show that demonstrative terms occurring as part of the head of
a relative clause may be followed by the plural marker lZ, as in (59).
(59)

a.
lv
lz 3
w
one [+deic] pl op res come
These/those ones who came

v.
det

fongbe

3
w
op res come

v.
det

fongbe

b.
nx
lz
one [prox] pl
Those who came

(=(28) in Lefebvre 1997)

If lV and nX are determiners in (55) and (56), the simplest assumption is that
they are determiners in (57), (58) and (59) above. Therefore, it must be the case
that, in those contexts, they determine an np headed by a phonologically null
head corresponding to one in English. This analysis is supported by the fact
that the null head may be modified by an adjective, as shown in (60).
(60)

a. N jl

kkl lv
v.
I want one big
[+deic] det
I want this/that big one.

fongbe

v.
det

fongbe

b. N jl

kkl nx
I want one big
[prox]
I want that big one.

(=(29) in Lefebvre 1997)

The semantic and syntactic properties of the lexical entries lV and nX are
presented in (61).
(61)

a. /lv/
[+deictic]

this/that

fongbe

b. /nx/
[+deictic]
[proximate]

that

fongbe
(=(30) in Lefebvre 1997)

When we compare the Fongbe and Haitian data, we find a systematic parallel
between the properties of lexical entries in the two languages. In both languages,

4.4 The [+deictic] terms

99

there is a paradigm of demonstrative terms with two forms, and these forms are
determiners (see (34) and (56) ). The distribution of the two sets of forms is
parallel compare (34) and (56), (35) and (57), (36) and (58), (37) and (59). In
both languages, the terms may be used for either animate or inanimate objects,
and in both languages, one term is specified [proximate] while the other is
unspecified for the feature [proximate]. Sila in Haitian and nX in Fongbe are
both [proximate], while sa and lV are [proximate] (see (39) and (61) ). In
contrast to French, both languages lack a form that lexically encodes the positive
value of the feature [proximate] (see (48c) ). The lexical entries of the demonstrative terms in the two languages are represented in (62).
(62)

haitian
a. /sa/
[+deictic]

fongbe
/lv/
[+deictic]

b. /sila/
[+deictic]
[proximate]

/nx/
[+deictic]
[proximate]

this/that

that
(=(31) in Lefebvre 1997)

As can be seen in (62), the semantic and syntactic properties of the Haitian and
Fongbe deictic terms are parallel: the Haitian lexical entries differ from the
Fongbe ones only in their phonological representations.
The discrepancy between the inventory and properties of the demonstrative
terms in French and Haitian creole clearly shows that the creators of Haitian did
not acquire the semantic and syntactic properties of the French system of demonstratives. Rather, they ended up with a system of demonstrative terms that has
the semantic and syntactic properties of their existing system, with phonological
representations derived from French phonetic matrices, as discussed below; this
is what the relexification hypothesis would predict. According to this view, the
creators of Haitian copied the lexical entries of their own lexicon and relabelled
them on the basis of phonetic matrices found in the superstratum language. In
other words, the semantic and syntactic properties of the Haitian creole demonstrative terms are identical to those of the corresponding lexical entries in the
substratum language because they were copied from an already established lexicon; the phonological representations of the Haitian lexical entries are different
from those of the original lexical entries, and similar to French phonetic strings,
because the copied lexical entries were relabelled on the basis of French phonetic matrices.
How did the creators of Haitian establish phonological forms for the copied
lexical entries? Recall from chapter 2 that the choice of the pertinent phonetic strings in the superstratum language is mainly based on their use in specific semantic and pragmatic contexts. This choice is further constrained by the
requirement that there be some semantic overlap between the semantics of the
superstratum strings and of the copied lexical entries to be relabelled. Given this

100

Functional category lexical entries involved in nominal structure

constraint, we can safely hypothesise that the creators of Haitian were looking for French forms that they could identify as [+deictic]. French has eleven
such forms. On the basis of the phonetic correspondences between French and
Haitian creole (see Appendix 2), there are three French forms that could have
been chosen to relabel the copied lexical entries. The French form a is pronounced [sa], which is identical to the form sa in Haitian creole. The French
form cela is pronounced [s1la] or [sl7] yielding sila in Haitian creole. It is a
well-documented fact that the French central vowel /1/ has been reinterpreted as
/i/ in Haitian creole (e.g. the French word /p1ti/ small is /piti/ in Haitian). It is
also a well-documented fact that the epenthetic vowel /i/ is used to break illicit
consonant clusters in Haitian (e.g. French [pti] is also realised as /piti/ in Haitian) (see Brousseau 1994b). Finally, the complex French deictic term celui-l
this/that one, pronounced [s2ila], could also have been the phonetic source
of Haitian sila (see Sylvain 1936: 60). While the French phonetic sequence [2i]
has been reinterpreted as [wi] in some contexts (e.g. French [2it] eight is /wit/
in Haitian creole), it has been reinterpreted as [i] in others (e.g. French [l2i]
him is /li/ in Haitian12). Hence, we can safely assume that the French phonetic
matrices that were the source of the phonological representations of the Haitian
demonstrative forms sa and sila were a and either cela or celui-l. All three of
these forms are strong pronouns in French, and thus are lexical forms. But, as
we saw earlier, these French forms did not contribute the semantic and syntactic
properties of the Haitian lexical entries.
The creation of the lexical entries for Haitian demonstrative terms thus proceeded as depicted in (63), which should be read in light of the general schema
in chapter 2.13
(63)

a. fongbe lexical entry

french

/()l/
[+deictic]

[sa] used in specific

semantic and pragmatic
contexts
haitian creole lexical entry
/sa/
[+deictic]

b. fongbe lexical entry

french

/()nx/
[+deictic]
[proximate]

[s1la] / [s2ila] used in

specific semantic and
pragmatic contexts
haitian creole lexical entry
/sila/
[+deictic]
[proximate]

4.5 Case markers within the noun phrase

4.5

101

Case markers within the noun phrase

Case markers14 have no semantic content and hence they cannot be relabelled.
This section argues, however, that the covert Case markers of Haitian have the
same Case specifications as those of the substratum language. The facts presented
here further illustrate the constraint imposed by the superstratum language on
the constituent order of the creole. The content of this section is based on work
by Brousseau and Lumsden (1992) on the nominal structure of Fongbe, work
by Lumsden (1989) on the nominal structure of Haitian, comparative work by
Lumsden (1991) and additional work that I did on this construction. I begin with
the Fongbe data, which present the richest overt Case system of the three languages under comparison.
The Fongbe nominal structure exhibits two Case markers: sn of and tWn
s (i.e. Genitive). As is extensively argued in Brousseau and Lumsden (1992),
these two Case markers have the following distinguishing properties. First,
when the argument is a Theme, it can appear with the tWn Case marker only if it
is an affected Theme. In example (64), the argument of mmVn denial is not
affected and hence it cannot occur as the complement of tWn although it can
occur as the complement of sn.
(64)

a.

ngb v
sn mmvn v
truth
det of denial det
the denial of the truth

b. *mmvn
denial

ngb
truth

fongbe

v
twn
v
fongbe
det gen det
(=(13a and b) in Brousseau and Lumsden 1992)

Second, when the noun phrase has a partitive meaning, it can occur with sn but
not with tWn, as shown in (65).
(65)

a.

[ckpl sn] g
corn-beer of
bottle
the bottle of beer

b. *gk
bottle

[ckpl
corn-beer

twn]
gen

v
det

fongbe

v
fongbe
det
(=(15a and b) in Brousseau and Lumsden 1992)

Third, when the argument is inalienably possessed by the head noun, it must
appear with tWn and not with sn, as shown in (66).
(66)

a.

w [v
v
twn] v
arm child det gen det
The childs arm has a wound.

b. ?*[v
child

v
det

sn] w
of
arm

d
have

kp.
wound

fongbe

v
d
kp
fongbe
det have wound
(=(17a and b) in Brousseau and Lumsden 1992)

102

Functional category lexical entries involved in nominal structure

Finally, and most importantly, while iteration of arguments is possible with sn, it
is not with tWn, as shown by the contrast in grammaticality between (67a and b).
(67)

a.

[Kwk sn] [Aristote sn] 33

lz
fongbe
Koku of
Aristotle of
sketch pl
[Lit.: the sketches of Koku of Aristotle] Kokus sketches of Aristotle

[Aristote twn]
b. *33
sketch Aristotle gen

[Kwk twn] lz
fongbe
Koku gen pl
(=(8a and b) in Brousseau and Lumsden 1992)

The impossibility of iterating arguments with the Genitive Case marker (=s in
English) has been proposed in the literature (e.g. Fukui and Speas 1986) as a
diagnostic test to identify Genitive Case. Brousseau and Lumsden (1992) argue
that, since in Fongbe the iteration of arguments is not possible with tWn (see
(67) ), tWn must express Genitive Case (equivalent to s in English) in this language. By contrast, the possibility of iterating arguments is a property of Case
markers like of in English. On the basis of (67a), Brousseau and Lumsden (1992)
thus conclude that sn is the Fongbe counterpart of of in English.
The Fongbe nominal structure in (68) is adapted from Brousseau and Lumsden
(1992) (see Lefebvre 1994c).
(68)

By sn
33
Kwk twn
v
Bayi case sketch Koku case the
(the) Kokus sketch of Bayi

fongbe

dp
d
Det

fp
f

kp

f 0 Kwki twn

np
n

kp

ti
n0
33e

By sn

As can be seen in (68), the complement of the head noun occurs to the left of
the head and is marked for the Objective Case sn (equivalent to of in English)
assigned by the head noun. With Hoekstra (1992), I assume that thematic roles
are assigned within lexical category projections. Hence, the possessor phrase
originates in Spec of np, where it is assigned a thematic role. From this position,

4.5 Case markers within the noun phrase

103

it moves to Spec of fp (a functional category projection). On Brousseau and

Lumsdens (1992) analysis, Genitive Case is assigned to the possessor phrase by
the null functional head projecting to fp, under SpecHead agreement.15 Thus,
on their account, the Fongbe nominal structure is the mirror image of the English nominal structure in (69).
(69)

the kings portrait of the queen

english

fp
kp

the kings f 0

np
n
n0
portrait

kp
of the queen

(=(9) in Lumsden 1991)

Lumsden (1991) notes that the structure in (69) corresponds to current analyses
of the English Genitive constructions (e.g. Szabolcsi 1987; Abney 1987), where
the possessor is realised outside of the np, in the specifier position of a functional category phrase (represented as fp) headed by a phonologically null form.
In his comparative study of Fongbe, Haitian and French nominal structures,
Lumsden (1991) points out that, in contrast to Fongbe, in French noun phrases,
the nominal arguments are typically realised after the head noun. The complements are introduced by the forms de of (see Milner 1978) or of (see
Tremblay 1991).16 More than one argument can be realised in a single phrase
(see Milner 1978), as shown in (70).
(70)

a. le portrait dAristote de Rembrandt du Louvre

the portrait of Aristotle by Rembrandt in the Louvre

french

b. la rponse du professeur la question de ltudiant

french
the answer of the professor to the question of the student
(=(17) in Lumsden 1991)

Thus, the French forms and de have a function similar to that of of in English.
As noted in Tremblay (1990) and Lumsden (1991), an argument of a noun phrase,
realised as a special pronominal form (referred to as a possessive adjective in
traditional grammars) must precede the noun. As Lumsden (1991) shows, however, this position allows the expression of only one argument. This is exemplified
in (71).
(71)

a. son portrait
his/her portrait

b. ses enfants
his/her children

c. ta table
french
your table
(=(18) in Lumsden 1991)

104

Functional category lexical entries involved in nominal structure

On the basis of this distribution, Lumsden (1991) proposes that the French
nominal structure is as in (72).
(72)

son portrait dAristote

his picture of Aristotle

french

fp
f

np
son

np

f0

n
n0
portrait

kp
dAristote

(=(19) in Lumsden 1991)

On Lumsdens analysis, the French facts contrast in several ways with the
Fongbe facts. First, while the complement occurs to the right of the noun in
French (see (72) ), it occurs to the left of the noun in Fongbe (see (68) ). Second,
the Genitive phrase occurs to the left of the noun in French and to the right
of the noun in Fongbe. Third, while the Genitive phrase may be realised as a
noun or a strong personal pronoun in Fongbe, it cannot be either in French.
This contrast between the two languages is exemplified in (73) and (74).
33
Kwk twn
sketch Koku gen
Kokus sketch

(73)

a.

(74)

a. *Marie
Mary

b.

portrait
sketch

svn nyy
crab me
my crab

twn
gen

b. *elle portrait
her sketch

fongbe

french

Furthermore, as is pointed out in Lefebvre (1994a), in Fongbe, the Genitive

phrase is compatible with the other determiners which can appear in the nominal
structure. Hence, it can co-occur with a demonstrative term, a determiner and a
plural marker, as is shown in (75).
(75)

svn nyy twn

lv
v
lz
crab me gen dem det pl
these/those crabs of mine

fongbe
(=(100) in Lefebvre 1994a)

In French, on the other hand, the Genitive phrase is not compatible with any of
the determiners that might otherwise occur in the nominal structure.
(76)

a. *la
the

ma
cette
poss [+deic]

bague
ring

french

b. *les
the

mes ces
poss [+deic]

bagues
ring

french
(=(10) in Lefebvre 1997)

4.5 Case markers within the noun phrase

105

The contrasting properties of the substratum and superstratum languages with

respect to the data discussed above present an interesting challenge to the creators of the creole. How did they manage, given the properties of their native
grammar, on the one hand, and the French data they were faced with, on the
other hand? This question is not a trivial one for, as was proposed in chapter 2,
constituent order is constrained by the superstratum language. Moreover, although
Case markers can be copied, they cannot be assigned a new label because they
have no semantic content.17
Lumsden (1991) shows that Haitian creole nouns may have a complex argument structure, as is illustrated in (77).
(77)

a. ptre
[pche]
a
portrait fisherman the
the portrait of the fisherman

haitian

b. ptre
[pent
sa] a
portrait painter this the
the portrait by this painter

haitian

c. repons [kesyon mwen] an

answer question my
the
the answer to my question

haitian

d. repons [pwofse] a
answer professor the
the answer of the professor

haitian
(=(4) in Lumsden 1991)

In the above examples, no Case marker is overtly manifested in the phrase

containing the argument of the noun. Lumsden (1991) assumes, however, that
the Haitian creole expression is assigned Case, since this is required by the
Principles of Universal Grammar (see e.g. Chomsky 1981). Lumsden (1991)
further assumes that, in all natural languages, Case is realised in a functional
category (see e.g. Travis and Lamontagne 1992; Lumsden 1987), and that therefore there is a functional category realising Case in the Haitian expressions in
(77). Presumably, then, the phrase containing the argument of the noun must
bear a phonologically null Case. But what Case is it? Does it have the properties
of tWn or of sn in Fongbe or the properties of /de in French?
Lumsdens (1991) data show that Haitian noun phrases permit only one np
argument. This is shown by the ungrammaticality of the noun phrases in (78)
which contain more than one argument.
(78)

a. *ptre
portrait

pche
fisherman

pent
sa
a
painter this the

b. *ptre
portrait

pent
painter

c. *repons
answer

kesyon
mwen an
question my
the

sa
pche
this fisherman

a
the

pwofse
professor

haitian
haitian
a
haitian
the
(=(5a, b, c) in Lumsden 1991)

106

Functional category lexical entries involved in nominal structure

Lumsden (1991) remarks that, if a second argument is required in a noun phrase,

it must be expressed as a pp or in a relative clause, as in (79).
(79)

a. ptre
pche
a
pent
sa
a
te
pentire
portrait fisherman the painter this he past paint
the portrait of the fisherman which this painter painted

haitian

b. repons pwofse a
sou kesyon
answer professor the on question
the professors answer to my question

haitian

mwen
my

an
the

c. repons kesyon
mwen pwofse
te
ft
haitian
answer question my
professor past make
the answer to my question which the professor gave
(=(5d, e, f) in Lumsden 1991)

Since iteration of arguments is not permitted with the phonologically null

Case, Lumsden (1991) argues that this Case must be of the same type as tWn in
Fongbe and thus must be the Genitive Case (see also Gilles 1988, for a similar
conclusion). This means that the Haitian argument that follows the noun is being
assigned the same syntactic analysis as the postnominal argument in Fongbe
rather than that in French. The nominal structure of Haitian is as in (80), adapted
from Lumsden (1989, 1991).
(80)

timounn Mari a
child
Mary the
(the) Marys child

haitian

dp
d
fp
f

kp

f 0 Marii

np
n

ti

n
timounn

The fact that the Haitian possessor phrase shares properties with the Fongbe
possessor phrase further supports the proposal in Lumsden (1991). First, in Haitian,
as in Fongbe, the possessor may be realised as a noun or as a strong pronoun
(see chapter 6), as is shown in (81).

4.5 Case markers within the noun phrase

(81)

107

a. krab Jan

an
svn Kwk twn
v
crab J/K
gen det
J/Ks crab (in question/that we know of )

haitian
fongbe

b. krab mwen
an
svn nyy
twn
v
crab me
gen det
my crab (in question/that we know of)

haitian
fongbe

The Haitian and Fongbe data in (81) are parallel. They contrast with the French
data in (74). Moreover, in both Haitian and Fongbe the Genitive phrase is compatible with the other determiners which can appear in the nominal structure.
Hence, in both languages, the Genitive phrase may co-occur with a demonstrative
term, a determiner and a plural marker, as is shown in (82).
(82)

krab mwen
svn nyy
twn
crab me
gen
these/those crabs of

sa
a
yo
haitian
lv
v
lz
fongbe
dem det pl
mine (in question/that we know of)
(=(100) in Lefebvre 1994a)

The Haitian and Fongbe data in (82) contrast with the French data in (76).
Given that the possessor phrase in Haitian occurs to the right of the noun, as
in Fongbe (see (77) and (66a) ), why is it the case that, unlike Fongbe, Haitian
has no complement preceding the noun (see the Haitian data (77), (78) )? In
French nominal structures, the argument follows the noun. Lumsden (1991)
assumes that the creators of Haitian identified that order as a possible constituent
order in French. As shown above, however, they assigned the French constituent
following the noun an analysis which was compatible with their own grammar.
Lumsden (1991) suggests that they did not see any prenominal complement in
French, and therefore they abandoned the complement position of their own
grammar.
On the basis of additional data, however, I shall argue that the Haitian
nominal structure in (80) should include a complement position on the right of
the head noun. Haitian speakers who do not allow for iteration of arguments
(see (78) ) do present data which, according to the analysis in Brousseau and
Lumsden (1992), require one to posit a complement position. For example, a
non-affected argument must be realised in a complement position as shown in
(64) (for Fongbe). Haitian speakers allow for non-affected arguments in nominal
structures and these arguments follow the noun (e.g. resp bondye respect of
God). Furthermore, when a noun phrase has a partitive meaning it must occur
in a complement position (see (65) ). Haitian speakers allow a noun phrase with
a partitive meaning to occur to the right of the head noun as in bouty by a the
bottle of beer. In this example, by cannot possibly be in a position where it
will be assigned Genitive Case. Moreover, there are Haitian speakers who allow
for two arguments in the nominal structure. For a subset of speakers, (83) is
grammatical; Mari is interpreted as either the possessor or the agent.

108
(83)

Functional category lexical entries involved in nominal structure

ptre
Wb Mari a
portrait Robert Mary det
Marys portrait of Robert

haitian

Assuming that Mari in (83) is part of the Genitive phrase following Lumsdens
analysis above, Wb must occur in a complement position of the head noun.
The three sets of facts above suggest that the nominal structure of Haitian must
include a complement position in addition to the Genitive phrase. This means
that the creators of Haitian who had a grammar of the type of Fongbe reproduced in the creole the nominal structure of their native language, except for the
order of constituents. As Lumsden (1991) has argued, they kept the Genitive
phrase position because it had the same constituent order as the French complement. But they did not abandon the complement position. Rather, it appears that
they kept it but assigned it the same constituent order as in French. These facts
constitute a dramatic illustration of the constraint imposed by the superstratum
language in establishing constituent order in creolisation.
Now, is there any evidence supporting the claim that, for speakers who allow
two arguments in the nominal structure, one is the complement and the other is
part of the Genitive phrase? How do we know that this is not just iteration of
complements such as we find in French (e.g. le portrait de Robert de/par Marie
the portrait of Robert of/by Mary)? One argument supports the analysis that
Mari in (83) is part of the Genitive phrase. For speakers who allow two arguments following the head noun, the second one is always interpreted as the agent
or the possessor. We expect noun phrases bearing an agent or possessor Thetarole to be part of a Genitive phrase. Moreover, if the order of arguments is
reversed as in potr Mari Wb a (compare with (83) ), Wb is now interpreted
as the agent/possessor. The Haitian data contrast with French, where a series of
arguments following a noun can appear in a relatively free order (e.g. le portrait
dAristote de Rembrandt and le portrait de Rembrandt dAristote both meaning
the portrait of Aristotle made by Rembrandt. This cluster of facts shows that
the two Haitian arguments following the head noun do not constitute iterating
complements, in contrast to French, and that the second argument is not in a
complement position but in a Genitive phrase. Thus, Lumsdens analysis whereby
there is a Genitive phrase in Haitian holds even for speakers who accept two
arguments after the head noun.
The scenario I propose to account for the historical derivation of the Haitian
facts is the following. Speakers of a language of the type of Fongbe who were
creating the creole had a complement and a Genitive phrase in their nominal
structure. They reproduced these two positions in the creole, assigning them a
surface position compatible with the surface order of constituents in French, that
is, to the right of the head noun, as discussed above. Consequently, in Haitian,
both the complement and the Genitive phrase occur to the right of the head
noun. Since Case markers are not relabelled because they have no semantic
content, the lexical entries copied from sn and tWn were assigned a null form at
relabelling. This means that when speakers of the early creole had to use Case

4.5 Case markers within the noun phrase

109

markers in Haitian nominal structures, they refrained from pronouncing the Case
markers of their lexicons. In modern Haitian, Case markers are still phonologically
null. We have to assume null cases (Objective and Genitive) since, as mentioned
above, Case is required by the Principles of Universal Grammar. But the creators of Haitian used these arguments according to the principles of their own
grammar. Assuming this scenario to be correct, how did the first generation of
Haitian native speakers deduce the properties of the null Objective and Genitive
Case markers? I believe that they were able to do this based on the data that they
were exposed to and general principles of Universal Grammar. They deduced
the existence of Genitive Case since the second argument of the noun is always
associated with an agent/possessor Theta-role. Objective Case was revealed by
the fact that the first argument of the noun can bear the thematic roles associated
with this position (in the case of deverbal nouns) and by the semantic relationship between the complement position and the head noun (in other cases).18
The Haitian data discussed so far are from the Central and Southern regions
of Haiti. As was mentioned in chapter 3 (and the references therein), however,
the Northern part of Haiti distinguishes itself from the rest of the country in
using a prenominal possessive particle, as shown in (84).
(84)

a. liv
a
Jan
book part John
Johns book

northern haitian

b. liv
Jan

book John gen

Johns book

central and southern haitian

(=(1) in Lumsden and Lefebvre 1994)

Lumsden and Lefebvre (1994) show that this regional difference in Haitian
creole corresponds to differences between the various substratum source languages of the creole. Goodman (1964: 54) observes that several West African
languages such as Hausa, Wolof, Igbo, Mande and Mandingo make use of a prenominal particle in possessive constructions. Furthermore, Manfredi (1992: 207)
reports that Yoruba also introduces possessed nps with a prenominal possessive
particle, as shown in (85).
(85)

w
e
Ay
book part Ayo
Ayos book

yoruba
(from Manfredi 1992: 207)

Lumsden and Lefebvre (1994) hypothesise that speakers of these West African
languages who were relexifying their lexicon with data from French used the
French preposition occurring in expressions such as le livre Jean the book
of John to relabel the prenominal connective particle of their native grammar.
As is hypothesised in Lumsden and Lefebvre (1994), in early Haitian, then,
there would be two dialects: one where the prenominal possessive connective
particle of lexicons such as Yoruba has been relabelled as a (see (84a) ) on the

110

Functional category lexical entries involved in nominal structure

basis of French , and one where the postnominal Case markers (from languages
like Fongbe) have been assigned a phonologically null form. In the dialect
levelling process, the Northern dialect of Haitian arrived at a consensus which
follows the pattern of one subset of West African languages (i.e. Yoruba-type
languages), while speakers of the Central and Southern dialects arrived at a
different consensus following the pattern of another subset of languages (i.e.
Fongbe-type languages19). The above data reflect the fact that dialect levelling is
a process that operates at the level of speech communities, and that different
areas did not necessarily arrive at the same consensus concerning the representation of a given lexical entry.20
4.6

Conclusion

The data discussed in this chapter show that the creators of Haitian creole did
not perceive the functional categories involved in French nominal structure as
such. They relabelled the determiner and the demonstrative terms of their own
lexicon with phonetic strings corresponding to major lexical category items in
French. The plural marker was shown to have come into the language through
the relabelling of the third-person plural pronoun. Case markers were assigned
a phonologically null form. The fact that the Case specifications of the noun
phrases occurring in Haitian nominal structure reproduce those of Fongbe
argues for this claim. During the dialect levelling period, they levelled out some
of the differences between the early Haitian dialects. The history of the so-called
indefinite marker yon does not follow this general pattern, however, as this
determiner was shown to have developed from within the creole.

The preverbal markers encoding

relative Tense, Mood and Aspect

Ever since the famous paper by Bickerton (1984) on the Language Bioprogram
Hypothesis, the Tense, Mood and Aspect systems (henceforth the tma systems)
of creole languages have been a target for competing theories of creole genesis.
This chapter examines the origin of the Haitian tma system within the framework of the relexification hypothesis. First, it is shown that the general features
of the Haitian tma system pattern on the model of Fongbe rather than French. It
is then argued that, while most of the semantic and syntactic properties of the
lexical items involved in the tma system of Haitian are derived from the corresponding lexical entries in the substratum language, the phonological representations of these markers appear to be derived from the phonetic representations
of French periphrastic, and thus lexical, expressions. Section 5.3 discusses the
temporal interpretation of bare sentences. The chapter ends with a discussion of
dialect levelling. The data and analysis presented in this chapter rely heavily on
an extensive and detailed comparative study of the Haitian, French and Fongbe
tma systems by Lefebvre (1996b).
5.1

Overview of the tma systems of Haitian, French and Fongbe

In Haitian creole, the verb in a tensed clause is invariant; it always occurs in its
simple form. In French, however, the verb in a tensed clause obligatorily bears
inflectional morphology. Tense, mood and aspect are encoded by means of inflectional affixes on the verb. In the following examples, the different endings on
the verb distinguish between simple past, present and future tense, respectively:
jaimai I loved, jaime I love, jaimerai I will love. Similarly, the different
endings on the verbs in the following examples distinguish between indicative
and subjunctive mood: je finis I finish, . . . que je finisse that I finish. Aspect
may also be encoded by an inflectional affix on the verb. The imperfect form
is an example in point: je finissais I was finishing. Furthermore, French has
obligatory subjectverb agreement for person and number, which is reflected in
the inflectional morphology of a verb occurring in a tensed clause. Hence, je
finir-ai I will finish, nous finir-ons we will finish, etc. French allows no bare
verbs: each verb in a tensed clause must bear inflectional morphology encoding
tense, mood, person and number.
None of the verbal morphology found in French has made its way into Haitian. In this respect, Haitian follows the pattern of its West African (non-Bantu)
111

112

Preverbal markers encoding tense, mood and aspect

substratum languages. For example, in Haitian, as in Fongbe, the verb of a tensed

clause always occurs in its simple form; there are no subject/verb agreement
markers for person or number, and no affixes encoding tense, mood, or aspect.
In both Haitian and Fongbe, temporal relationships, mood and aspect are encoded
by means of markers occurring between the subject and the verb.
The inventory of the tma markers of Haitian is quite parallel to that found in
Fongbe.1 This is shown in (1).2
(1)

The inventory of tma markers in Haitian and in Fongbe

anterior
Past /Past perfect
h
f
te
k

irrealis
Definite future
h
f
ap
n
Indefinite future
h
f
a-va n-w
Subjunctive
h
f
pou
n

non-complete
Habitual
Imperfective
h
f
h
f

nw
ap 3 . . . wy

(=(115) in Lefebvre 1996b)

Both languages have a marker which encodes anteriority. Both lexically distinguish between definite and indefinite future. The definite future markers are
used to convey the speakers attitude that the event referred to by the clause will
definitely take place in the near future. By contrast, the indefinite future markers
are used to convey the speakers opinion that the event referred to by the clause
might eventually or potentially take place at an undetermined point in the future.
The fact that speakers of Haitian distinguish between definite and indefinite future
is widely documented in the literature (see Valdman 1970, 1978; Spears 1990, and
the references therein). For Fongbe, this distinction is pointed out in Anonymous
(1983: v, 3). Both languages have a marker glossed as subjunctive for convenience. This term subsumes the three meanings of pou and n respectively: both may
be interpreted as must, should or may. Both languages have a form which
encodes imperfective aspect. As can be seen in the table in (1), there is a one-toone correspondence between the preverbal markers in the two languages, except
that Fongbe has one encoding the habitual aspect, and Haitian does not.
The Haitian definite future marker ap and imperfective marker ap are homophonous. In Lefebvre (1996b), it is argued that there are two separate lexical
entries signalled by ap in the Haitian lexicon: one encoding imperfective and
one encoding future.3 The first argument supporting this claim comes from data
drawn from a subset of speakers who allow two aps within a single clause, as
shown in (2).
(2)

M ap
ap sti.
I
def-fut imp go-out
I will be going out.

haitian
(=(20) in Lefebvre 1996b)

5.1 Overview of the

TMA

systems of Haitian, French and Fongbe

113

A second argument is that speakers who do not accept two co-occurring aps still
have the pertinent interpretations. For example, for this second group of speakers,
a sentence containing ap such as Map sti may be assigned three interpretations:
(a) I am going out, where ap is assigned an imperfective reading; (b) I will go
out, where ap is assigned a future interpretation; or (c) I will be going out.
This latter interpretation shows that the second group of speakers (those who do
not pronounce two aps in a row) still have the interpretation corresponding to
(2), which does contain two aps. Thus, the two groups of speakers present
similar interpretive data regardless of whether they allow the co-occurrence of
two aps at surface structure. Lefebvre (1996b) proposes linking the difference
between the two groups of speakers to a more general constraint in Haitian
concerning the utterance of two adjacent similar forms. The details of this constraint will be discussed in chapter 8. Suffice it to say for now that speakers in
the first group have a relaxed version of this constraint and thus can utter two
aps in a row (see (2) ), whereas speakers in the second group manifest a strict
version of this constraint and thus cannot say two aps in a row, although they do
manifest the pertinent interpretations. In this view, then, the difference between
the two groups of speakers is not attributable to a difference in the properties of
the lexical entries, but rather to variation among speakers with respect to how
they apply a more general constraint. Furthermore, it will be shown below that
the two aps have made their way into the Haitian tma system via different
paths. These sets of facts argue that there are two lexical entries for ap in the
Haitian lexicon. This should not come as a surprise in view of the fact that, as
we will see below, a large sample of substratum languages have different words
to signal definite future and imperfective.
The French expressions which are closest to the Haitian tma markers are the
periphrastic expressions which can be used to encode tense, mood or aspect,
as illustrated in (3). The (a) sentence is an example of the periphrastic future;
the (b) sentence shows a popular usage of the preposition pour for/to with a
prospective meaning (see also Grevisse 1975: 646); (3c) illustrates the use of the
preposition aprs with a progressive meaning (aprs is used in some of the same
contexts as en train de or ; see Grevisse 1975: 646; Fraud 1768). Finally, the
example in (d) shows the use of t, a form of the auxiliary tre to be, to
conjugate stative predicates.
(3)

a. Jean va manger.
John will eat (in the near future).

b. Jean est pour partir.

John is about to go.

french

c. Jean est aprs//en train de manger.

John is eating.

d. Jean a t malade.
french
John has been sick.
(=(52) in Lefebvre 1996b)

The French periphrastic expressions listed in (3) are similar to the Haitian
tma markers in two ways. First, like the Haitian tma markers, they occur
between the subject and the verb. Second, in several cases, the phonological
representations of the periphrastic expressions are similar to those of the Haitian

114

Preverbal markers encoding tense, mood and aspect

tma markers. For example, Haitian a-va is phonologically similar to French va

in (3a). Haitian pou is phonologically similar to French pour in (3b). Haitian ap
(imperfective) is phonologically similar to French aprs in (3c), and Haitian te is
phonologically similar to the French form t as it occurs in (3d). As has been
extensively discussed in the literature (see e.g. Sylvain 1936; Goodman 1964),
the phonological representations of the Haitian tma markers are probably derived
from these French periphrastic expressions. As will be seen below, however,
when we compare the semantics of these pairs of lexical items, we find that the
members of each pair have strikingly different properties. On the other hand,
when we compare the semantics of the Haitian/Fongbe pairs in (1), we find that
they have strikingly similar properties (see section 5.2).
In both Haitian and Fongbe, complex tenses are formed by combining the
preverbal markers. For example, a clause containing both the marker of anteriority and the definite future marker is interpreted as conditional. This is shown
in (4).
(4)

Mari
Mari
Mary
Mary
Mary

te
k
ant
would
would

ap
prepare pat.
n
3
wh.
def-fut prepare dough
prepare dough.
have prepared dough.

haitian
fongbe

(=(123) in Lefebvre 1996b)

Note from the translation that this combination of markers yields either a present or past conditional interpretation, depending on the context. An exhaustive
inventory of the possible combinations of tma preverbal markers in Haitian
and Fongbe is provided in Lefebvre (1996b). It is shown that both languages
present exactly the same inventory of complex tenses.
The process involved in the formation of complex tenses in Haitian contrasts
with French as follows. In French, complex tenses are derived by combining a
form of the auxiliary tre to be or avoir to have with the past participle of the
verb. In these cases, the auxiliary bears the tense, mood, aspect, person and
number morphology discussed above. For example, the perfect is formed in this
way: jai fini I have finished, je suis venu I have come. The formation of the
future perfect also follows this pattern: jaurai fini I will have finished, je serai
venu I will have come. The past conditional is also formed in this way: jaurais
fini I would have finished, je serais venu I would have come, and so on. In
Haitian creole, there are no auxiliaries corresponding to tre and avoir in French.
Such auxiliaries are not found in Fongbe either.
Haitian allows for bare sentences, that is sentences which contain no overt
expression of tense, mood or aspect. An example of such a sentence is given
in (5).
(5)

Mari knnn Jan.

Mary know
John
Mary knows John.

haitian

5.2 Historical derivation of Haitian

TMA

markers

115

The possibility of bare sentences in Haitian contrasts with their impossibility

in French. In French, each clause must minimally be overtly marked for mood
(indicative, subjunctive, imperative or infinitive), and each tensed clause must
bear tense morphology (present, past or future). However, the possibility of bare
sentences in Haitian finds its parallel in the substratum languages. For example,
Fongbe also allows them, as shown in (6).
(6)

Mari tn
Jan.
Mary know John
Mary knows John.

fongbe

The question of how bare sentences are assigned a temporal interpretation will
be addressed in section 5.3.
While absolute tense relates the time of the event denoted by the clause to the
moment of speech, relative tense relates the time of the event to a reference
point (see Comrie 1976: 2). French has at least three simple absolute tenses:
past, present and future. By contrast, Haitian has no absolute tense. As is the
case in Fongbe, the markers involved in the temporal interpretation of the clause
are assigned a relative tense interpretation. This fact will be illustrated in the
discussion of the individual markers in section 5.2.
The overview of the basic features of the three tma systems described above
suggests that, from a global perspective, Haitian follows the Fongbe model
rather than the French model. In Haitian and Fongbe, unlike French, the verb in
a tensed clause always occurs in its bare form; temporal relationships, mood and
aspect are encoded by markers occurring between the subject and the verb; the
inventories of these markers are quite similar in the two languages; complex
tenses are formed by a combination of preverbal markers rather than with auxiliary verbs as in French; both Haitian and Fongbe allow for bare sentences in
contrast to French, which does not; both languages exclusively express relative
tense in contrast to French, which has a range of absolute tenses. Even though
the phonological representations of the Haitian preverbal markers appear to have
been derived from French phonetic matrices, the other properties of the Haitian
tma system appear to be derived from those of the substratum language, as will
be seen in the three-way comparison below.
5.2

The historical derivation of the Haitian tense, mood and

aspect markers

This section provides a comparison of the semantic and syntactic properties of

the Haitian tma markers identified in (1) with those of the French periphrastic
expressions shown in (3) and with the Fongbe tma markers also listed in (1).
It is shown that, while the tma markers of Haitian derive most of their phonological representations from the French periphrastic expressions, they derive
most of their semantic and syntactic properties from the corresponding lexical
entries in Fongbe.

116
5.2.1

Preverbal markers encoding tense, mood and aspect

The marker of anteriority te

Sentences containing the marker of anteriority te are interpreted as past or pluperfect depending on the aspectual class of the verbal expression. Damoiseau
(1988) distinguishes three aspectual classes of verbs in Haitian creole: dynamic,
resultative and stative.4 Dynamic expressions describe a situation involving a
process which can be perceived as ongoing. These involve verbs like manje to
eat, plante to plant, etc. Resultative expressions describe a situation that is the
result of some process. These involve verbs like w to catch sight of, jwenn to
find, etc. Finally, stative expressions do not refer to a process at all. These
involve verbs like knnn to know, bezwn to need, etc.
Sentences containing te with a dynamic verb are always interpreted as pluperfect. The pluperfect situates an event prior to a reference point that is itself in
the past with respect to the moment of speech (see Comrie 1985).
(7)

Dynamic verb
L
m rive,
Mari te
prepare
When I
arrived Mary ant prepare
When I arrived Mary had prepared dough.

pat.
dough

haitian

(=(4) in Lefebvre 1996b)

Sentences containing te and a resultative verb are ambiguous. They may be interpreted as pluperfect or past, depending on the context. Past indicates that the
event described by the verb is simultaneous with a reference point which is past
with respect to the moment of speech (Hornstein 1977; Comrie 1985).
(8)

Resultative verb
Mari te
w
vol a.
Mary ant catch sight of thief det
Mary caught sight of the thief.
Mary had caught sight of the thief.

haitian

(=(5) in Lefebvre 1996b)

Sentences containing te and a stative verb also have a past or a pluperfect

interpretation, depending on the context.
(9)

Stative verb
Mari te
knnn Jan.
Mary ant know
John
Mary knew John. or Mary had known John.

haitian
(=(6) in Lefebvre 1996b)

The above examples show that te always situates an event in the past with
respect to the moment of speech. In some cases, the time of the event coincides
with the reference point (i.e. past); in others, the time of the event is itself prior
to the reference point (i.e. pluperfect). The fact that a sentence containing te may
be assigned different past readings (i.e. past or pluperfect), which are defined
with respect to the reference point in different ways, suggests that te is best
analysed as a relative tense marker, rather than an absolute tense marker.

5.2 Historical derivation of Haitian

TMA

markers

117

As mentioned in section 5.1, French does not have a preverbal marker of

anteriority. According to Goodman (1964), the source of the phonological form
of the Haitian preverbal marker te would be the past participle t, a form of the
French auxiliary verb tre to be. In French, t is used to form various complex
tenses in the environment of non-dynamic predicates. The French pluperfect,
compound past, future perfect and past conditional tenses are illustrated in (10).
(10)

a. Marie avait t malade.

Mary had been sick.

b. Marie a t malade.
Mary was sick.

french

c. Marie aura t malade.

Mary will have been sick.

d. Marie aurait t malade.

Mary would have been sick.

french

While Haitian te occurs with predicates of all aspectual classes (see (7)(9) ),
French t occurs only with non-dynamic predicates. Thus, although the two
forms are phonologically similar and do have an element of meaning in common
(namely anteriority), their semantic and syntactic properties are not parallel.
What is the source of the semantic and syntactic properties of te?
Fongbe has a preverbal marker expressing anteriority, k.5 Sentences containing this marker are interpreted as past or as pluperfect depending on the aspectual
class of the verb. Fongbe verbal expressions also fall into three aspectual classes:
dynamic, resultative and stative.6 As is the case in Haitian (see (7) ), sentences
containing k and a dynamic verb are always interpreted as pluperfect.
(11)

Dynamic verb
Mari k
3
wh.
Mary ant prepare dough
Mary had prepared dough.

fongbe
(=(60) in Lefebvre 1996b)

Sentences containing k with resultative verbs or stative verbs are ambiguous.

They may be interpreted as past or as pluperfect depending on their context.
(12)

Resultative verb
Mari k
mw
jvtv v.
Mary ant catch-sight-of thief det
Mary caught sight of the thief.
Mary had caught sight of the thief.

fongbe

(=(61) in Lefebvre 1996b)

(13)

Stative verb
Mari k
tn
Jan.
Mary ant know John
Mary knew John. or Mary had known John.

fongbe
(=(62) in Lefebvre 1996b)

Thus, in both Fongbe and Haitian, there is a preverbal marker of anteriority that
can appear with verbs of all aspectual classes. The above examples show that
Fongbe k, like Haitian te, always situates an event at a time that is past with
respect to the moment of speech. In some cases, the time of the event coincides
with the reference point (i.e. past); in other cases, the time of the event is itself

118

Preverbal markers encoding tense, mood and aspect

prior to the reference point (i.e. pluperfect). The fact that a sentence containing
k may be assigned different past readings (i.e. simple past or pluperfect), which
relate to the reference point in different ways, depending on the context, suggests that Fongbe k, like Haitian te, encodes relative rather than absolute tense.
The marker of anteriority thus induces the same temporal interpretation of the
clause in both Fongbe and Haitian. Compare (7) and (11), (8) and (12), and (9)
and (13).
The division of properties which characterises the Haitian lexical entry te
follows from relexification. Speakers of a language like Fongbe had a lexical
entry k encoding anteriority. They copied this lexical entry, and looked for
an appropriate phonetic string in French to relabel it. The French form t (see
(10) ) is appropriate: it shares some semantics with the original lexical entry
(namely, anteriority), and it occurs between the subject and the predicate. The
relexifiers thus relabelled the lexical entry copied from k as te, a reduced form
of t.7 This derivation accounts straightforwardly for the fact that the properties
of te in the Haitian lexicon parallel those of k in Fongbe.
5.2.2

The irrealis mood marker pou

Mood expresses the speakers attitude towards the content of an utterance. The
mood marker pou is used to express a wish (exhortative), as in (14a), an obligation, as in (14b), or an order (injunctive), as in (14c). This marker has been
glossed as subjunctive in order to capture the range of the meanings it covers.
(14)

a. Dye pou proteje u.

God sub protect you
May God protect you.

haitian

b. Mari pou prepare pat.

Mary sub prepare dough
Mary should prepare dough.

haitian

c. Tut slda
pou vini
laplas kuny a.
all soldier sub come square now
det
All soldiers must come to the square now.

haitian
(Sylvain 1936: 90)

In the contexts where pou occurs in Haitian, French uses the subjunctive
mood encoded by an affix on the verb: Que Dieu te protge May God protect
you; Que Marie prpare la pte Mary should prepare dough; Que tous les
soldats viennent la place, maintenant All soldiers must come to the square
now. Obviously, modal morphology on the verb was not perceived by the
creators of Haitian. It has been proposed in the literature on Haitian that pou is
derived from the French preposition pour occurring in periphrastic expressions
such as those in (3b) (see e.g. Fournier 1987). The semantic properties of pour
in the context of (3b), however, are not parallel to those of Haitian pou. The
semantics of (tre) pour (see (3b) ) in French is described by Grevisse (1975:

5.2 Historical derivation of Haitian

TMA

markers

119

646) in the following terms: tre pour peut servir indiquer un fait prochain,
prsenter une action comme convenue, prpare, ou encore marquer la nuance
quexprimeraient tre de nature , tre dispos , destin . [tre pour may
indicate that an event is about to happen, or that an action is agreed on, or
prepared or it may express the nuance to be likely to, to be disposed to or to
be destined to.] In contrast, Haitian pou is used to express the speakers attitude
that the event denoted by the clause may, should or must take place. Although
French tre pour and Haitian pou have a semantic element in common, namely
their irrealis meaning, they are not exactly parallel. Suppose, then, that pou
derives its phonological representation from the French phonetic matrix pour
occurring in contexts such as (3b). Where does pou get its semantics from?
Again, the substratum languages provide an answer to this question.
For example, the Fongbe mood marker n may be used to express a wish
(exhortative), as in (15a), an obligation, as in (15b), or an order (injunctive), as
in (15c). In order to capture its range of possible meanings, n, like pou, has
been glossed as subjunctive.
(15)

a. Mfwl n
cv
w.
God
sub protect you
May God protect you.

fongbe
(Anonymous 1983: v, 4)

b. Mari n
3f
wh.
Mary sub prepare dough
Mary should prepare dough.

fongbe

c. M
n
3.
you (pl.) sub eat
You must eat.

fongbe
(=(74) in Avolonto 1992: 55)

Thus, in both Haitian creole (see (14) ) and Fongbe (see (15) ), there is a
preverbal marker which can be interpreted as may, should or must. Given
the relexification hypothesis, the Haitian data are predictable. A speaker of a
language like Fongbe who was relexifying his vocabulary on the basis of data
from French would make a copy of the lexical entry of n and relabel it with a
phonetic string from French. Apparently, the closest parallel that the relexifiers
found was pour in sentences like (3b). Furthermore, as will be seen in chapter 7,
both Haitian pou and Fongbe n may also occur in the complementiser position
unlike French pour, which does not. Thus, although the phonological form of
pou is derived from French, the details of its semantic and syntactic properties
appear to be derived from the substratum language lexical entry, as predicted by
the relexification hypothesis.
5.2.3

The imperfective aspect marker ap

Aspects are different ways of viewing the internal temporal constituency of a

situation. In Haitian, ap is the imperfective marker. The imperfective aspect

120

Preverbal markers encoding tense, mood and aspect

describes a situation that has already begun, but that is not complete at the time
of the moment of speech or at the time of a reference point (see Comrie 1976).
This marker is illustrated in (16).
(16)

Mari ap manje krab la.

Mary imp eat
crab det
Mary is eating the crab.

haitian
(=(18) in Lefebvre 1996b)

Sentences containing ap can only be interpreted as progressive if they contain a

dynamic verb, as in (16). Sentences containing a resultative verb, such as w to
catch sight of, or a stative verb, such as knnn to know, cannot be interpreted
as progressive, as shown in (17).
(17)

a. #Jan
ap w
vol a
John imp catch-sight-of thief det
[Lit.: John is catching sight of the thief.]

haitian

b. #Jan
ap knnn Mari
John imp know
Mary
[Lit.: John is knowing Mary.]

haitian
(=(19) in Lefebvre 1996b)

Note, however, that the sentences in (17) are not ungrammatical. Both can be
interpreted as future. This is because, as noted above in section 5.1, the phonological representations of the definite future marker and the imperfective marker
are homophonous (see (1) and (2) ), a matter which will be taken up in section 5.2.4.
Two additional facts about the interpretation and distribution of the imperfective marker ap deserve attention. For one thing, ap is used in contexts where
English would require a gerund, as in (18).
(18)

Nou ap
jwnn Mari ap prepare
we
def-fut find
Mary imp prepare
We will find Mary preparing the dough.

pat
dough

la.
det

haitian

(=(21) in Lefebvre 1996b)

Furthermore, a sentence containing ap may be assigned a habitual interpretation.

The presence of the adverb toutan all the time in (19) prevents a progressive
interpretation of the clause.
(19)

Mari ap joure
toutan.
Mary imp swear all-the-time
Mary swears all the time.

haitian
(=(22) in Lefebvre 1996b)

This is interesting for several reasons. First, Comrie (1976) reports that the use
of the same form to encode both progressive and habitual is common, and is
found in languages belonging to various genetic and geographical groupings.
Thus, it is not a property of creole languages. Comrie further shows that this form
is best characterised as imperfective, a gloss which covers both its progressive

5.2 Historical derivation of Haitian

TMA

markers

121

and habitual meanings. This suggests that ap is best characterised as an imperfective marker. Also, according to Comrie (1976), there are languages in which
the progressive is expressed using a locative form. In these languages, the progressive has the meaning be in a state of doing x rather than be in the process
of doing x. Comrie (1976: 103) also remarks that the locative expression of
progressive meaning is basic, and only if a language has this possibility can it
further extend the same form to habitual meaning. The fact that ap may be
assigned a habitual interpretation suggests that it is a locative expression of the
progressive. In this view, sentence (16) is best characterised as meaning Mary
is at eating the crab, rather than Mary is in the process of eating the crab.
In spoken French, there are three periphrastic expressions which can be used
to express the fact that an event is not complete: tre aprs, tre and tre en
train de, as is shown in (20).
(20)

a. Marie est
aprs manger.
Marie aux after eat
Mary is (in the state of) eating.

french

b. Marie est
manger.
Marie aux at eat
Mary is (in the state of) eating.

french

c. Marie est
en train de manger.
Marie aux prog
eat
Mary is (in the process of) eating.

french
(=(54) in Lefebvre 1996b)

The phonological representation of the Haitian imperfective ap is probably

derived from the French locative preposition aprs, as it occurs in (20a) (see
Goodman 1964).8 In addition to being assigned a progressive interpretation, both
Haitian ap and French aprs share semantic properties. Like ap (see (19) ), aprs
may be assigned a habitual interpretation, as shown in (21), where the presence
of the adverb toujours all the time prevents a progressive interpretation.
(21)

Marie est
toujours aprs
Mary aux always
imp
Mary swears all the time.

sacrer.
swear

french

Both occur between the subject and the main verb of the clause. Their distribution, however, is not exactly the same: while Haitian ap can appear in contexts
where English would require a gerund (see (18) ), the French aprs cannot, as
shown by the ungrammaticality of (22).
(22)

*Nous trouverons Marie aprs manger

[Lit.: We will find Mary eating.]

french
(=(56) in Lefebvre 1996b)

In order to convey the same meaning as the Haitian sentence in (18), speakers of
French require another periphrastic expression Nous trouverons Marie en train
de manger we will find Mary eating or a present participial form Nous

122

Preverbal markers encoding tense, mood and aspect

trouverons Marie mangeant We will find Mary eating. So again the Haitian
form ap shares some, but not all, of the properties of the French tre aprs.
Based on the above discussion, it appears that, although the creators of Haitian used the French phonetic string aprs, as in (20a), to provide a phonological
representation for the lexical entry of the imperfective preverbal marker, the two
lexical entries do not have exactly the same distribution. Where does the use of
Haitian ap in (18) come from?
In Fongbe, the imperfective aspect makes use of the locative preposition #
at. The preposition # selects a complement headed by the postposition wY;9
this postposition selects a nominalised vp. In Fongbe nominalisations, the object
precedes the nominalised verb. The construction is illustrated in (23).10
(23)

Kwk [3 [ [svn 3]
wy.] ]
Koku at
crab
eating post
Koku is eating crab.
[Lit.: Koku (is) at crab-eating.]

fongbe

(=(74) in Lefebvre 1996b)

As is the case in Haitian (see (16), (17) ), a sentence containing the imperfective
construction can only be interpreted as progressive in the context of a dynamic
verb, as in (23). Furthermore, the semantic interpretation data in (23) suggest
that, like Haitian, Fongbe has a locative expression of progressive meaning,
which is rendered by be in a state of x rather than be in the process of x
(see Welmers 1973, for a discussion of similar data in a wide range of West
African languages). This is further supported by the fact that, in the context of
a stative verb such as be sick (but not know11), a sentence containing the
imperfective aspect is assigned a habitual interpretation, as shown in (24).
(24)

Sk 3 zwn-jy wy.
Cica at sick
post
Cica is habitually sick.

fongbe
(=(75) in Lefebvre 1996b)

This should come as no surprise, given Comries (1976: 103) observation that languages that have the locative expression of a progressive meaning may extend
the same form to the habitual meaning. This suggests that the form # . . . wY
is best glossed as imperfective, which subsumes both its progressive and its
habitual meanings, which parallels the Haitian data discussed above. Finally, the
imperfective aspect is used in Fongbe in contexts where we find a gerund in
English, as shown in (25).
(25)

N mwn Sk 3 wh
3
I
see Cica at dough prepare
I saw Cica preparing dough.

wy.
post

fongbe
(=(77) in Lefebvre 1996b)

The Fongbe data in (25) parallel the Haitian data in (18).

The semantic properties of the imperfective aspect are extremely similar
in Fongbe and in Haitian. In both languages, the imperfective is rendered by

5.2 Historical derivation of Haitian

TMA

markers

123

a locative expression which may be assigned a progressive (see (16), (23) ) or

habitual interpretation (see (19), (24) ). In both languages, the construction may
be used in contexts where we find a gerund in English (see (18), (25) ). However, the lexical items involved in the expression of imperfective aspect differ in
their selectional properties. While Haitian ap selects a vp, Fongbe # selects the
postpositional item wY which itself selects a nominalised vp. Can these facts be
accounted for within the relexification hypothesis?
Lefebvre (1996b) proposes the following historical derivation. The lexical
entry corresponding to # in the copied lexicon was relabelled as apre>apr>ap,
derived from the phonetic representation of the French preposition aprs in such
sentences as Marie est aprs manger Mary is eating (see (3c) ). The French
form aprs shares some semantics with #, and it occurs between the subject
and the verb. Thus, it has the appropriate properties for relabelling the Fongbe
lexical entry. It is further assumed that no phonetic string was found in French
to relabel the copied lexical entry corresponding to the postposition wY. In chapter 7, it is shown that this substratum lexical entry has no counterpart in Haitian
and no effect on the syntax of Haitian, and that, consequently, it must have been
abandoned during the formation of the creole. In early Haitian creole, then, the
only visible form of the imperfective was ap. Note, however, that the distribution of Haitian ap is more restricted than that of Fongbe #. While # may occur
as a Locative preposition in contexts that do not involve the imperfective, ap
does not. The fact that the correspondence between the two lexical entries is not
perfect in this case suggests that another factor was involved. This is indeed the
case, as we will see in section 5.4.

5.2.4

The definite future marker

Both Haitian and Fongbe have a definite future marker: ap and n, respectively.
The properties of these markers will be discussed in turn.
The definite future marker ap is used to convey the speakers attitude that the
event referred to by the clause will definitely take place in the near future, as
shown in (26) (see also Spears 1990).
(26)

M ap
vini.
I
def-fut come
I will definitely come.

haitian
(=(7) in Lefebvre 1996b)

Since ap conveys the speakers attitude towards an event yet to occur, it is best
analysed as an irrealis mood marker (see (1) ) rather than a tense marker. This is
in line with Comries (1985: 44) observation that future is often a difference of
mood, rather than tense.
The definite future marker ap can occur with verbs of all three aspectual
classes, as shown in (27).

124
(27)

Preverbal markers encoding tense, mood and aspect

a. Dynamic verb
Mari ap
prepare pat.
Mary def-fut prepare dough
Mary will prepare dough.
b. Resultative verb
Mari ap
w
Jan.
Mary def-fut catch-sight-of John
Mary will catch sight of John.
c. Stative verb
Mari ap
knnn
Mary def-fut know
Mary will know John.

Jan.
John

haitian

haitian

haitian
(=(9) in Lefebvre 1996b)

Given the appropriate context, a clause containing the definite future marker
may be assigned a future perfect interpretation. This is shown in (28) where ap
has scope over deja already, forcing a future perfect interpretation of the clause.
(28)

Mari ap
(deja)
prepare pat.
Mary def-fut already prepare dough
Mary will (already) have prepared dough.

haitian
(=(13) in Lefebvre 1996b)

In Haitian, there is no other way to express the future perfect.12

The combination of the anteriority marker te and the definite future marker
ap yields a present or past conditional interpretation of the clause, as shown in
(29), (30) and (31).
(29)

Dynamic verb
Mari t
ap
prepare pat.
Mary ant def-fut prepare dough
Mary would prepare dough.
Mary would have prepared dough.

haitian

(=(29) in Lefebvre 1996b)

(30)

Resultative verb
Mari t
ap
w
Jan.
Mary ant def-fut catch-sight-of John
Mary would catch sight of John.
Mary would have caught sight of John.

haitian

(=(30) in Lefebvre 1996b)

(31)

Stative verb
Mari t
Mary ant
Mary would
Mary would

ap
knnn
def-fut know
know John.
have known John.

Jan.
John

haitian

(=(31) in Lefebvre 1996b)

Where do the semantic and syntactic properties of the definite future marker in
Haitian come from? Again, the properties of ap are strikingly parallel to those of
the Fongbe definite future marker n.

5.2 Historical derivation of Haitian

TMA

markers

125

The Fongbe preverbal marker n, like the Haitian preverbal marker ap, is
used to convey the speakers feeling that the event referred to by the clause will
definitely take place in the near future (see Anonymous 1983: v, 3). This irrealis
mood marker is illustrated in (32).
(32)

n
he def-fut
He will die.

k.
die

fongbe
(Anonymous 1983: v, 3)

This definite future marker may occur with verbs of all aspectual classes, as
shown in the examples in (33), which parallel the Haitian data in (27).
(33)

a. Dynamic verb
Mari n
3
wh.
Mary def-fut prepare dough
Mary will prepare dough.
b. Resultative verb
Mari n
mw
Jan.
Mary def-fut catch-sight-of John
Mary will catch sight of John.
c. Stative verb
Mari n
tn
Jan.
Mary def-fut know John
Mary will know John.

fongbe

fongbe

fongbe
(=(66) in Lefebvre 1996b)

A Fongbe clause containing n may also be assigned a future perfect interpretation. This is shown in (34), where n has scope over the adverb k already,
forcing a future perfect interpretation of the clause (see also Anonymous 1983:
v, 5, for similar data). These interpretive data parallel the Haitian data in (28).
(34)

Mari n
(k)
3f
wh.
Mary def-fut already prepare dough
Mary will (already) have prepared dough.

fongbe
(=(68) in Lefebvre 1996b)

In Fongbe, as in Haitian, there is no other way to express the future perfect.

The combination of the anteriority marker k with the definite future marker
n yields a present or past conditional interpretation of the clause, as shown in
(35)(37), which parallel the Haitian data in (29)(31).
(35)

Dynamic verb
Mari k
n
3
wh.13
Mary ant def-fut prepare dough
Mary would prepare dough.
Mary would have prepared dough.

fongbe

(=(86) in Lefebvre 1996b)

126
(36)

Preverbal markers encoding tense, mood and aspect

Resultative verb
Mari k
n
mw
Jan.
Mary ant def-fut catch-sight-of John
Mary would catch sight of John.
Mary would have caught sight of John.

fongbe

(=(87) in Lefebvre 1996b)

(37)

Stative verb
Mari k
Mary ant
Mary would
Mary would

n
tn
Kwk.
def-fut know Koku
know Koku.
have known Koku.

fongbe

(=(88) in Lefebvre 1996b)

The semantic and syntactic properties of the Haitian definite future marker ap
and the Fongbe definite future marker n are parallel. Again, this is predicted by
the relexification hypothesis. But where does the phonological representation of
the Haitian definite future marker come from?
There are two possible historical derivations.14 A first possibility would be
that Fongbe n was relexified as ap from the French form aprs occurring in
expressions such as Marie est aprs manger. This appears to be an unlikely historical derivation in this case, however, on two grounds. First, an extensive survey
of the literature (grammars and dictionaries15) on the pertinent French dialects
spoken in the seventeenth century (see chapter 3) shows no example of tre aprs
with an immediate future interpretation. All the examples of tre aprs in these
sources are given a progressive interpretation. This conclusion is further supported
by data drawn from several French-based creoles provided by Goodman (1964).
Goodman shows that, in Louisiana creole, Haitian creole and Indian Ocean creoles, the meaning of the forms phonologically derived from French tre aprs is
progressive. This suggests that the creators of the creoles perceived tre aprs as
a form encoding the progressive but not as one encoding the immediate future.
Given the requirement that the copied lexical entry and the superstratum phonetic string hypothesised to provide a label for it must share some element of meaning (see chapter 2), it is unlikely that tre aprs was the phonetic string used to
relabel the lexical entry copied from Fongbe n. I therefore conclude that, although
the historical derivation hypothesised above is plausible, it is not likely.
Recall from chapter 2 (section 2.6.2) that in both West African and creole
languages, future markers have often been shown to have evolved from presentential lexical items through reanalysis. This provides us with a second possible
historical derivation for ap. The Haitian future marker could have evolved from
within the creole as a result of the reanalysis of a sentence-initial lexical item
encoding posteriority in much the same way as the preverbal markers discussed
in chapter 2. In the following paragraphs, I further explore this possibility.
Since the phonological representation of the Haitian definite future is ap, we
can hypothesise that, like the imperfective marker ap, it also comes from some
usage of the French form aprs. In addition to occurring in the periphrastic
expression of the progressive (as in (3c) ), aprs is a preposition meaning after.

5.2 Historical derivation of Haitian

TMA

markers

127

French grammars report that in sentence-initial position, as in Aprs avoir mang,

nous partirons After eating, we will leave, aprs is a preposition indicating
posteriority (see dAlembert, Mouchon and Diderot 17767: 560). Fraud (1768)
notes that when placed in front of the verb, it only serves as a temporal preposition: e.g. Aprs dner After (eating) lunch. It is clear that the preposition aprs in these contexts shares with the Fongbe marker n the meaning of
posteriority. In this case, however, ap could not have become the phonological
representation of the copied lexical entry through the process of relabelling the
lexical entry copied from n. Although Fongbe n shares some elements of meaning with the French preposition aprs, the two lexical items do not occur in the
same surface position. N occurs between the subject and the verb, and aprs in
sentence-initial position. In Lefebvre (1996b), it is proposed that ap may have
become the phonological representation of the copied lexical entry corresponding to Fongbe n through reanalysis of a Haitian preposition. Haitian creole, in
fact, has a sentence-initial preposition of posteriority apre after which derives
its phonological representation from the French preposition discussed above. An
example of this Haitian preposition is provided in (38).
(38)

Apre yo
fin
w-l
yo
raknte . . .
after they finish see-him they told
After they had seen him, they told . . .

haitian
(from Hall 1953: 221)

Haitian apre after, as it occurs in (38), would have made its way into the
languages tma system in much the same way as the sentence-initial adverb
baimbai after in Tok Pisin.
Sankoff and Laberge (1980) argue that the sentence-initial adverb baimbai
was grammaticalised as a preverbal marker. First the full form was reduced
to bai. The reduced form then became unstressed and realised as [b1]. These
reduced forms can occur within the same clauses as adverbs with a future meaning. Finally, the reduced forms have a tendency to occur between the subject and
the verb, rather than at the edge of the clause. In a more recent article, Sankoff
(1990: 72) documents the semantic shift in the meaning of bai from future
through irrealis to iterative and punctual. She points out that if adults, who
speak Tok Pisin as a second language, and children, who are native speakers of
Tok Pisin, have moved together in the word order change and in the change
from the disyllabic to monosyllabic form . . . a close examination of the semantics
of bai shows the native speakers role in remodeling the Tok Pisin tense and
aspect system into a strictly aspectual system. Sankoff (1990: 73) further comments: The changes ongoing in the Tok Pisin tense and aspect system are making it more like many of the substrate languages, in which tense is quite subsidiary
to aspect. Sankoff (1990: 73) accounts for this apparent paradox as follows:
There always has been a stronger relationship between the use of these markers
for aspect than for tense in the speech of adults, and this is, I propose, what the
native speakers are building on in carrying forth more sweeping changes than

128

Preverbal markers encoding tense, mood and aspect

their parents were able to do. The facts and analysis provided in Sankoff and
Laberge (1980) and Sankoff (1990) may easily be recast within the framework
adopted in this book. Recall from chapter 2 (section 2.6.3) that, on Lefebvre and
Lumsdens (1994b) approach to reanalysis in the early creole, this process applies
to a lexical entry that has been copied and assigned a null form at relabelling.
In light of this theoretical background, I now return to the historical derivation of the future marker ap in Haitian creole (following the proposal in Lefebvre
1996b). As was mentioned above, the creators of Haitian did not find an appropriate phonetic string in the superstratum language to relabel the copied lexical
entry corresponding to n, the Fongbe future marker. They therefore assigned it
a null phonological form. They used the sentence-initial preposition of posteriority
apre after, as it occurs in (38), to clarify the information which was not phonologically signalled. I assume that sentence-initial apre made its way between the
subject and the verb in a way similar to baimbai in Tok Pisin and that it was
then reduced to apr and eventually to ap. Suppose that the reduced form occurred
between the subject and the verb as an adjunct to the phrase headed by the
phonologically null mood marker copied from n, as illustrated in (39a).
(39)

a. subject . . .
apr

xp

b. subject . . .

xp

xp
//
[+irrealis]

/ap/
[+irrealis]

At some point, the Haitian speakers copied the phonological form of this preposition of posteriority onto the lexical entry of the previously hidden functional
category, as in (39b). In this way, reanalysis provided a phonological form for a
creole lexical entry that was generated by relexification but not assigned a phonological representation upon relabelling. When this lexical entry was assigned a
phonological form, it had the properties of the corresponding substratum lexical
entry. Recall from chapter 2 that, while copying and relabelling apply in creole
genesis when the creators of the creole are targeting the superstratum language,
reanalysis applies when speakers start targeting the relexified lexicons. The above
historical derivation thus illustrates how the data produced by relexification feed
the process of reanalysis in creole genesis.
At this point, we might ask whether there is independent evidence for the
claim made earlier that the Haitian lexicon has two different lexical entries
signalled by ap. Of course, as we have seen, in Fongbe the two lexical entries
(imperfective and future) are signalled by phonologically distinct forms. But
what about the other substratum languages of Haitian? As is the case in Fongbe,
the other Gbe languages render the imperfective and future by two different
forms (see Hazoum 1990). Other substratum languages of Haitian also encode
the two meanings with different forms. As is extensively documented by Wallace
(1995b), in Igbo (see Ward 1936), Yoruba (see Ogunbewale 1970), Akan (see
Dolphyne 1971) and Wolof (see Sauvageot and Manessy 1963), the imperfective

5.2 Historical derivation of Haitian

TMA

markers

129

and the future are rendered by different forms. On the basis of this sample of
substratum languages, we can safely assume that the creators of Haitian had
two different lexical entries to relexify, one for the imperfective and one for the
future. So the claim that there are two different lexical entries for the imperfective and the future in the Haitian lexicon finds additional motivation when a
broader range of substratum languages is considered.
Furthermore, the historical derivations proposed above for the definite future
marker and the imperfective aspect marker (section 5.2.3) provide an explanation for the homophony of the phonological forms. The creators of Haitian
had two different lexical entries to relexify, the definite future marker and the
imperfective marker. The copied lexical entry corresponding to the imperfective
was relabelled as ap, from the French form aprs in the periphrastic expression
of the progressive (section 5.2.3). The copied lexical entry corresponding to the
future was assigned the phonological representation ap through reanalysis of the
Haitian preposition of posteriority apre, which derives its phonological form from
the French preposition aprs after. Historically, then, the two Haitian forms ap
made their way into the Haitian tma system by different paths. The fact that the
creators of Haitian managed to create two lexical entries despite the limitations
of the superstratum language is predicted by the relexification hypothesis. The
relexifiers had two lexical entries to relexify and they did just that.
5.2.5

The indefinite future marker

The indefinite future marker a-va (and its allomorphs va, av, a (see Sylvain
1936; Hall 1953; Valdman 1970; Valdman et al. 1981) ) is used to convey the
speakers attitude that the event referred to by the clause might eventually or
potentially take place at some undetermined point in the future.
(40)

M a
vini
yn jou.
I
ind-fut come one day
I will/might eventually come one day.

haitian
(Dumais 1988: 247)

As shown in (41), a-va can occur with verbs of all aspectual classes.
(41)

a. Dynamic verb
Mari a-va
prepare pat.
Mary ind-fut prepare dough
Mary will eventually prepare dough.
b. Resultative verb
Mari a-va
w
Jan.
Mary ind-fut catch-sight of John
Mary will eventually catch sight of John.
c. Stative verb
Mari a-va
malad.
Mary ind-fut sick
Mary will eventually be sick.

haitian

haitian

haitian
(=(10) in Lefebvre 1996b)

130

Preverbal markers encoding tense, mood and aspect

Given the appropriate context, sentences containing a-va may also be assigned a
future perfect interpretation, as shown in (42).
(42)

Mounn ki
a
manti, yo
a-va
neye.
haitian
person who ind-fut lie
they ind-fut drown
[Lit.: The person who will have lied, they will drown (him).]
They will drown the person who has lied.
(Sylvain 1936: 87)

Finally, the combination of the anteriority marker te with a-va (or its allomorphs)
yields a conditional or potential interpretation of the clause, as shown in (43)
(45).
(43)

Dynamic verb
Mari t
a
prepare pat.
Mary ant ind-fut prepare dough
Mary might prepare dough.
Mary might have prepared dough.

haitian

(=(32) in Lefebvre 1996b)

(44)

Resultative verb
Mari t
a-va
w
Jan.
Mary ant ind-fut catch-sight-of John
Mary might catch sight of John.
Mary might have caught sight of John.

haitian

(=(33) in Lefebvre 1996b)

(45)

Stative verb
Mari t
a-va
malad.
Mary ant ind-fut sick
Mary might be sick.

haitian
(=(34) in Lefebvre 1996b)

The phonological representation of the Haitian indefinite future marker a-va

(va, av, a) is similar to the French va, a suppletive form of the verb aller to go,
used as a periphrastic future in French (see (3a) ). The properties of the French
morpheme va and the Haitian morpheme a-va are, however, quite distinct in
some interesting ways.
French grammars report that the periphrastic future is used to express a near
future, in contrast with the future form encoded by an affix on the verb, which
expresses a distant future (see Grevisse 1975: 731). Unlike the French periphrastic future, Haitian a-va is used to convey the speakers attitude that the
event referred to by the clause may eventually or potentially take place, at some
undetermined time (see (40) ). So, although the two forms share the semantic
property irrealis, they have opposite values with respect to the property far/near
future. Furthermore, we saw that a-va combines with te to form the conditional
or potential (see (43)(45) ). French va does not have this property. Finally, as
shown in (46), Haitian a-va is not compatible with negation, a fact also noted by
Sylvain (1936: 87), Spears (1990) and Hall (1953: 33).

5.2 Historical derivation of Haitian

(46)

*N p
a / a-va / va kabicha
we neg ind-fut
doze off
[Lit.: We might not doze off.]

TMA

markers

131

haitian
(Sylvain 1936: 87)

Although negation appears not to be a favourable environment for the use of the
periphrastic future in Montreal colloquial French (see Emirkanian and Sankoff
1984) or in colloquial French spoken in France (see Jeanjean 1988: 253), a sentence containing both negation and the periphrastic future is grammatical. For
example, a sentence like Jean (ne) va pas aller lcole aujourdhui John will
not go to school today is perfectly grammatical. So, although the Haitian indefinite future form a-va is phonologically similar to the French periphrastic future
form va, and in spite of the fact that these lexical entries both share the meaning
irrealis, the details of their semantics and distribution are not parallel. Where does
the semantics of the Haitian indefinite future marker a-va come from? Again, an
examination of the substratum languages provides a clear answer.
For example, the properties of Haitian a-va are remarkably similar to those of
the corresponding marker in Fongbe. In Fongbe, n-w is used to convey the
speakers attitude that the event referred to in the clause may eventually take
place (see Anonymous 1983: v, 3).
(47)

n-w
k.
he ind-fut die
He will eventually die.

fongbe
(Anonymous 1983: v, 3)

As in Haitian (see (41) ), the indefinite future marker may occur with verbs of all
aspectual classes, as shown in (48).
(48)

a. Dynamic verb
Mari n-w
3
wh.
Mary ind-fut prepare dough
Mary will eventually prepare dough.
b. Resultative verb
Mari n-w
mw Jan.
Mary ind-fut see John
Mary will eventually catch sight of John.
c. Stative verb
Mari n-w
tn
Kwk.
Mary ind-fut know Koku
Mary will eventually know Koku.

fongbe

fongbe

fongbe
(=(89)(91) in Lefebvre 1996b)

A Fongbe clause containing n-w may be assigned a future perfect interpretation, given the appropriate context. The Fongbe example in (49) parallels the
Haitian data in (42).

132
(49)

Preverbal markers encoding tense, mood and aspect

M6
n-w
3w digbn v
yi
n-w
xw. fongbe
person he ind-fut say lie
det they ind-fut beat
[Lit.: The person who will have lied, they will beat.]
They will beat the person who has lied.

As is also the case in Haitian (see (43)(45) ), the combination of the marker
of anteriority with the indefinite future marker yields a conditional or potential
interpretation of the clause, as shown in (50)(52).
(50)

Dynamic verb
Mari k
n-w
3
wh.
Mary ant ind-fut prepare dough
Mary might prepare dough.
Mary might have prepared dough.

fongbe

(=(89) in Lefebvre 1996b)

(51)

Resultative verb
Mari k
n-w
mw
Jan.
Mary ant ind-fut catch-sight-of John
Mary might catch sight of John.
Mary might have caught sight of John.

fongbe

(=(90) in Lefebvre 1996b)

(52)

Stative verb
Mari k
Mary ant
Mary might
Mary might

n-w
tn
Kwk.
ind-fut know Koku
know Koku.
have known Koku.

fongbe

(=(91) in Lefebvre 1996b)

Finally, as is the case in Haitian (see (46) ), the indefinite future marker in
Fongbe is not compatible with negation; this is shown by the ungrammaticality
of the sentences in (53).
(53)

*Jan
*Jan
John
[Lit.:

pa
a-va
prepare pat
m n-w
3
wh
neg ind-fut prepare dough
John will not eventually prepare dough.]

haitian
fongbe
(=(128) in Lefebvre 1996b)

Whatever the formal account of this incompatibility may be, the fact that it
exists in these two languages, but not in French (see the discussion above),
reveals the shared semantic properties of the lexical items involved.
The semantic and syntactic properties of the Haitian preverbal marker encoding indefinite future are thus extremely similar to those of the corresponding
Fongbe preverbal marker. Furthermore, the morphological makeup of the two
markers is strikingly parallel. The marker n-w is morphologically complex.
It is formed by the combination of the definite future marker n and w, which,
in isolation, means to come/to go. Nothing can intervene between these two
morphemes when they are used together as an indefinite future marker. For
example, in (54), where nW usually is inserted between n and w, the clause is

5.2 Historical derivation of Haitian

TMA

markers

133

grammatical but n and w are interpreted as independent words (see interpretation a). In this case, the clause cannot be assigned an indefinite or potential
future interpretation (see interpretation b).
(54)

Sk
Cica
a.

n
nw
w
3
def-fut usually come prepare

wh.
dough

fongbe

Cica will usually come to prepare dough.

b. #Cica will usually eventually prepare dough.

(=(65) in Lefebvre 1996b)

Similarly, the full form of the Haitian indefinite future marker a-va comprises a-,
which Lefebvre (1996b) hypothesises to be a reduced form of the definite future
marker ap, and va. As is the case for Fongbe n-w, the two morphemes forming the indefinite future in Haitian, a- and va, cannot be dissociated. Moreover,
in Haitian, the morpheme va is not used in isolation, as is shown by the
ungrammaticality of (55b). The lexical form of the verb to go is ale, as shown
in (55c).
(55)

a. Jan
a-va
ale.
b. *Jan
John ind-fut go
John
John will perhaps/eventually go.
c. Jan
ap
John def-fut
John will go.

ale.
go

ap
def-fut

va
go

haitian

haitian
(=(129) in Lefebvre 1996b)

It is interesting to note at this point that, in Ewe (see Wallace 1995a), the
indefinite future is rendered either by the form , the indefinite future marker, or
by v, which in isolation means to go, or by a combination of both forms: -v.
From the point of view of an Ewe speaker, this is an interesting case of phonological conflation (see Kihm 1994).
The comparison of the properties of a-va in Haitian and n-w in Fongbe
shows that the semantic, syntactic and morphological properties of these concatenated forms are quite parallel. I assume that subsequent morphological reduction of a-va yielded the numerous variants reported in the literature on Haitian:
va, av, and a.
5.2.6

Summary

The three-way comparison of the preverbal markers of Haitian with those of its
superstratum and substratum language shows once again that the Haitian lexical
entries derive most of their semantic and syntactic properties from those of the
corresponding lexical entries in the substratum language and their phonological
representations from phonetic strings found in the superstratum language. This
is exactly what we expect according to the relexification hypothesis.

134

Preverbal markers encoding tense, mood and aspect

5.3

The temporal interpretation of bare sentences

Haitian allows for bare sentences, that is, sentences in which there is no preverbal
marker. The temporal interpretation of such sentences is computed from the various components that participate in establishing the aspectual properties of a clause
(see Bentolila 1987; Damoiseau 1988; Dchaine 1991; Lefebvre 1996b; Lumsden
in press b). These components are the aspectual class of the verb (see Lumsden in
press b), the definiteness of the direct object of the verb (see Damoiseau 1988)
and the definiteness of the subject (see Bentolila 1987). A noun phrase is definite if
it has the postnominal determiner la or its allomorphs a, an, nan (see chapter 4).
For example, a bare sentence containing a dynamic verb with a definite object can
only be interpreted as present perfect, as shown in (56). Present perfect expresses
the present result of a past event; it indicates that the past situation has current
relevance (Comrie 1985).
(56)

Mari prepare pat

la.
Mary prepare dough det
Mary has prepared the dough.

haitian
(=(47) in Lefebvre 1996b)

A sentence containing a dynamic verb with an indefinite object is ambiguous

between a general present and a past interpretation. (A general past interpretation is not available.)
(57)

Mari
Mary

prepare
prepare

pat.
dough

haitian

a. Mary (generally) prepares dough.

b. Mary prepared dough.
(=(48) in Lefebvre 1996b)

In (56), the definiteness of the object imposes an end-point on the event denoted
by the verb. This property of the object forces the event of the bare sentence
to be interpreted as accomplished. In (57), the object is not determined. Nondetermined nps in Haitian are assigned a generic reading (see chapter 4, section 4.1). Generic objects do not impose an end-point on the event denoted by
the verb (see Tenny 1988). Hence, the event of the bare sentence may be interpreted either as accomplished (i.e. past) or as non-accomplished (i.e. general
present).
Bentolila (1987) shows that the definiteness of the subject may also have an
effect on the interpretation of a bare sentence. Compare the temporal interpretation of (58a), which contains a definite subject, with that of (58b), which contains a generic subject.
(58)

a. Mayi a
ba
nou bon garanti.
corn det give us
good profit
The corn has brought us good money.

haitian
(Bentolila 1987: 107)

5.3 The temporal interpretation of bare sentences

b. Mayi ba
nou bon garanti.
corn give us
good profit
Corn (generally) brings us good money.

135

haitian
(Bentolila 1987: 107)

In (58a), the definiteness of the subject induces a present perfect reading of the
bare sentence, while in (58b) the generic interpretation of the subject induces a
generic interpretation of the sentence (i.e. general present).
A bare sentence containing a resultative verb is ambiguous between a present and a present perfect interpretation, as shown in (59) (see also Lumsden in
press b).
(59)

Mari
Mary

w
catch-sight-of

vl
thief

(a).
det

haitian

a. Mary catches sight of the thief.

b. Mary has caught sight of the thief.
(=(50) in Lefebvre 1996b)

Finally, a bare sentence containing a stative verb is always interpreted as present

(see also Damoiseau 1988; Dchaine 1991; Lumsden in press b).
(60)

Mari knnn Jan.

Mary know
John
Mary knows John.

haitian
(=(51) in Lefebvre 1996b)

The data presented above show that the temporal interpretation of a Haitian
bare sentence is computed from the various components defining the aspectual
properties of the clause. The temporality of such a sentence thus ranges from
past through present perfect to general present, depending on the other aspectual
components present.
The possibility of bare sentences in Haitian contrasts with the impossibility
of such sentences in French. In French, each clause must minimally be overtly
marked for mood (indicative, subjunctive, imperative or infinitive), and each
tensed clause must bear tense morphology (present, past or future). Once again,
the grammaticality of bare sentences in Haitian finds its parallel in the substratum
languages.
For example, in Fongbe, the temporal interpretation of bare sentences also
proceeds on the basis of the various components that participate in establishing
the aspectual properties of a clause: the aspectual class of the verb (see Avolonto
1992), and the definiteness of its arguments. In Fongbe, a noun phrase is definite
if it has the determiner V (see chapter 4). As is the case in Haitian (see (56) ), a
bare sentence containing a dynamic verb with a definite object is interpreted as
present perfect, as shown in (61).
(61)

Mari 3
wh
v.
Mary prepare dough det
Mary has prepared the dough.

fongbe
(=(110) in Lefebvre 1996b)

136

Preverbal markers encoding tense, mood and aspect

A sentence containing a dynamic verb with a non-definite object is interpreted as

past, as shown in (62). This parallels the Haitian data in (57b).
(62)

Mari 3
wh.
Mary prepare dough
Mary prepared dough.

fongbe
(=(111) in Lefebvre 1996b)

As is the case in Haitian, the definiteness of the subject plays a role in the temporal interpretation of a clause. Consider the examples in (63), which parallel
the Haitian data in (58).
(63)

a. Jj-nl-kn v
n.
harvest
det give
The harvest has brought us money.
[Lit.: The harvest has given.]

fongbe

b. Jj-nl-kn n.
harvest
give
Harvest (generally) brings us money.
[Lit.: The harvest generally gives.]

fongbe

(=(112) in Lefebvre 1996b)

In (63a), the subject is definite and the clause is assigned a present perfect
interpretation. In (63b), the subject is generic and the clause has a general
present interpretation.
Similarly, a bare sentence containing a resultative verb is ambiguous between
a present and a present perfect interpretation. The Fongbe data in (64) parallel
the Haitian data in (59).
(64)

Mari
Mary

mw
catch-sight-of

jtv
thief

(v).
det

fongbe

a. Mary catches sight of the/a thief.

b. Mary has caught sight of the/a thief.
(=(113) in Lefebvre 1996b)

Finally, a bare sentence containing a stative verb is always interpreted as present

(see also Avolonto 1992: 7). Sentence (65) parallels the Haitian sentence in (60).
(65)

Mari tn
Jan.
Mary know John
Mary knows John.

fongbe
(=(114) in Lefebvre 1996b)

Thus, in Fongbe as in Haitian, the temporal interpretation of a bare sentence

ranges from simple past through present perfect to general present, depending on
the various components that define the aspectual properties of the clause. Once
again, there is a close parallel between the semantic interpretation facts in the
two languages.
There is one exception to this general pattern, which Lefebvre (1996b) proposes to link to the availability of an overt preverbal marker of habitual aspect in

5.4 Dialect levelling

137

Fongbe (nW) and the absence of a corresponding overt preverbal marker in Haitian
(see (1) ). In both Haitian and Fongbe, a clause containing a dynamic verb and
an indefinite object may be assigned a past interpretation, as shown in (66).
(66)

Mari
Mari
Mary
Mary

prepare
3
prepare
prepared

pat.
wh.
dough
dough.

haitian
fongbe
(=(146) in Lefebvre 1996b)

Recall from (57) that the Haitian sentence in (66) may also be assigned a general
present interpretation (i.e. Mary (generally/habitually) prepares dough). This
interpretation does not obtain for the Fongbe sentence in (66). In Fongbe the
preverbal habitual marker nW16 is used to convey this meaning, as shown in (67).
(67)

Mari nw
3
wh.
fongbe
Mary hab prepare dough
Mary prepares dough. [i.e. she habitually does this, or this is
what she does for a living]
(=(147) in Lefebvre 1996b)

This difference between the two languages follows from the fact that, while
Fongbe has an overt preverbal marker to encode habituality, Haitian does not.
The question of why Haitian lacks a habitual marker will be taken up in section 5.4 on dialect levelling.
Thus, the temporal interpretation of bare sentences in both Haitian and Fongbe
proceeds similarly. The one exception to this general pattern follows from the
fact that there is an overt preverbal marker of habitual aspect in Fongbe, which
is lacking in Haitian (see section 5.4). The parallel observed in the temporal
interpretation of bare sentences in Haitian and Fongbe cannot be attributed to
relexification since, in this case, there is no lexical entry involved. A formal
account of what triggers the semantic interpretation data discussed in this section will have to await further research. The conclusion that one can draw from
the data reported on here, however, is that the creators of Haitian used the properties of their own grammar in assigning bare clauses a temporal interpretation in
the new language that they were creating.
5.4

Dialect levelling

Based on the above description and analysis, the Haitian tma system appears to
be derivable quite straightforwardly from the relexification hypothesis and the
claim that, in the early creole, reanalysis applies to a lexical entry that has been
created through relexification but that has been assigned a phonologically null
form at the relabelling stage (see the discussion of the definite future marker
in section 5.2.4.). In two cases, however, the predictions of the hypothesis are
not met: the properties of the Haitian imperfective marker only partially match
those of Fongbe, and furthermore, while Fongbe has a preverbal marker encoding habitual, Haitian does not. This section argues that these two cases, as well

138

Preverbal markers encoding tense, mood and aspect

as the additional case of the Haitian marker of anteriority (see below), can be
straightforwardly accounted for by the process of dialect levelling hypothesised
to apply in the early creole, as described in chapter 2.
West African languages encode Tense, Mood and Aspect rather similarly, as
can be seen from the literature. For example, Ewe, a neighbouring language of
Fongbe, has the inventory in (68).
(68)

The inventory of tma markers in Ewe (from Wallace 1995a)

anterior

irrealis
Definite future
l

non-complete
Habitual
nw . . . my

Imperfective
ly . . . my

Indefinite future
(v)
Subjunctive
n

The inventory is strikingly similar to the Fongbe inventory in (1), with two
differences. The first one involves the selectional properties of the habitual
marker. In Ewe, but not in Fongbe, this marker selects a postposition. A more
important difference between the two languages lies in the fact that Ewe lacks a
marker of anteriority. However, other substratum languages of Haitian do, like
Fongbe, have such a marker. For example, Yoruba has ti (see e.g. Welmers 1973;
Ogunbewale 1970; Holm 1986), and Bambara has tun (see e.g. Holm 1986).
Given this situation, the relexified lexicons of a Fongbe, Yoruba or Bambara
speaker would include a lexical entry for a marker of anteriority, whereas the
relexified lexicon of an Ewe speaker would not. The Haitian lexical entry te, however, appears to be widespread in the present-day lexicon. Indeed, it is attested in
all grammars and dictionaries of Haitian, and during the course of this research
we have met no Haitian speaker who does not have it. This suggests that West
African speakers who relexified a lexicon lacking a marker of anteriority must
have acquired this lexical entry, probably during the dialect levelling period.
Hazoum (1990) provides an overview of the morphemes encoding the
imperfective, the habitual and the definite future in six Gbe languages: Gungbe,
Fongbe, Ajagbe, Gengbe, Tufingbe and Xwedagbe. The general picture that
emerges from this work is as follows. First, all the Gbe languages have a lexical
entry encoding definite future which is phonologically distinct from the imperfective marker. The phonological representation of this lexical entry may differ
from language to language (e.g. n, A, l). Second, all these languages lexically
encode the imperfective aspect by means of a preposition meaning at. These
prepositions vary, however, in their selectional properties. As can be seen in
(69), in some languages the preposition selects a postposition which in turn
selects a nominalised vp, as is the case in Fongbe (see (23) ). In other substratum
languages, the construction does not involve a postposition; in this case, the vp
complement may or may not be nominalised.

5.5 Conclusion
(69)

139

The imperfective aspect in Gbe languages

3 vpn wy
l vpn v
l vpn v /
tw vpn
3 vpn
nw vp

fongbe
ajagbe
gengbe
gungbe
tufingbe
xwedagbe
(from Hazoum 1990)

Haitian encodes the imperfective in the same way as Xwedagbe. In other substratum languages of Haitian, we also find similar facts. For example, in Bambara,
a Mande language, the imperfective marker, b+, precedes the vp and no postposition is involved (see Holm 1986: 263). On the basis of these facts, it is hypothesised that, when the speakers of the relexified lexicons started targeting the
speech of their community, they levelled out the differences between the relexified
lexicons and settled on a single lexical entry that occurs preverbally and selects
a vp to encode the imperfective aspect.
Finally, Hazoums (1990) survey of tma markers in Gbe languages reveals
that there are two groups of languages with respect to the availability of a habitual marker. Gungbe, Ajagbe and Gengbe, like Fongbe, do have such a marker,
while languages such as Tufingbe and Xwedagbe do not. Presumably then, in
early Haitian, there were two dialects of Haitian: one with a habitual marker and
one without. Apparently, the speakers of the Haitian community eventually settled
on the dialect that did not have a habitual marker since this is what we see in
modern Haitian. They could have chosen otherwise, however, since Martinican
creole, which emerged from the same pool of languages as Haitian, appears to
have chosen the other option. According to Bernab (1983: 1053), Martinican
creole has a habitual marker ka which derives its phonological form from a
reduced form of the French word capable capable [kapab>kap>ka]. The difference between Haitian and Martinican creole with respect to the presence of
a habitual marker reflects the difference observed between the substratum languages. It illustrates the claim made in chapter 2 that different creole communities may have settled on different options. This represents yet another dimension
of the process of dialect levelling hypothesised to play a role in the development
of a creole.
5.5

Conclusion

The data presented in this chapter show that the tma markers of Haitian creole
derive their properties from their contributing languages in a systematic way:
while the phonological forms are derived from superstratum phonetic strings,
the semantic, and most of the syntactic, properties of the Haitian lexical entries
are derived from those of the corresponding lexical entries in the substratum languages. This is exactly what one would expect according to the relexification hypothesis. Furthermore, the facts pertaining to the interpretation of bare sentences in

140

Preverbal markers encoding tense, mood and aspect

Haitian and Fongbe show that the creators of the creole used the principles of
their own grammar in assigning bare clauses a temporal interpretation in the
new language that they were creating. This constitutes a major piece of evidence
against Bickertons (1984) Language Bioprogram Hypothesis (see also Lefebvre
1996b). Finally, the Haitian tma system provides interesting cases of reanalysis
and dialect levelling and of how these processes apply in the early creole on the
basis of the relexified lexicons.

Pronouns

This chapter compares the pronominal (and pronoun-like) forms of Haitian with
those of its contributing languages. The paradigms of personal pronouns, possessives, logophoric pronouns, clitics, expletives, reflexives and Wh-words and phrases
are discussed. The semantic and syntactic properties of the paradigms of Haitian
forms discussed below are shown to contrast with those of French and to resemble
those of the substratum languages. The data presented in this chapter follow from
the hypothesis outlined in chapter 2. The data on which the contents of this chapter
are based are drawn from various sources, which will be identified in each section.
6.1

Personal pronouns

In current theories, strong or tonic pronominal forms have the same distribution
as nouns in that they occur in argument positions. The three-way comparison
between Haitian, French and Fongbe shows that the paradigm of Haitian strong
pronouns and the Fongbe paradigm share striking similarities which set both
languages apart from French.
The Haitian paradigm of personal pronouns is as in (1).
(1)

mwen
ou / [wu]
li
nou
yo

1st-person singular
I, me
haitian
2nd-person singular
you (sg.)
3rd-person singular
he/she/it /him/her
1st- and 2nd-person plural we/us/you (pl.)
3rd-person plural
they/them
(from Valdman et al. 1981)

The most striking fact about this paradigm is that the same form is used for both
first- and second-person plural.
The paradigm of modern French tonic pronouns is as in (2) (see Grevisse
1975). According to Haase (1965), this paradigm had already been established
by the seventeenth century.
(2)

moi
toi
lui / elle
nous
vous
eux / elles

1st-person singular
2nd-person singular
3rd-person singular
1st-person plural
2nd-person plural
3rd-person plural

I, me
you (sg.)
he/she/it
we/us
you (pl.)
they/them

french

141

142

Pronouns

In contrast to Haitian, the French paradigm distinguishes between first- and

second-person plural pronouns.1 Furthermore, in French, third-person pronouns
are marked for gender (lui is masculine; elle(s) is feminine; eux they is masculine or neuter depending on the dialect).2 In the Haitian paradigm in (1), none of
the pronominal forms is marked for gender.
The differences between the two paradigms of personal pronouns are all the
more surprising given that the phonological representations of the Haitian forms
appear to be derivable from the phonetic matrices of the French forms. Hence,
Haitian mwen me takes its phonological representation from the French phonetic matrix [mwa]/[mwe] me; Haitian ou you (sg.) is derived from the French
phonetic matrix [vu] you (pl.); Haitian li he is derived from the French
phonetic matrix [l2i] him (or [li] in some Northern and Northwestern dialects
of France). Similarly, Haitian nou we, you (pl.) clearly takes its phonological
representation from the French phonetic matrix [nu] we. As we saw in chapter 4, the phonological representation of Haitian yo they comes from the strong
form of the third-person personal pronoun eux. Assuming the phonological derivations described above, why did the Haitian personal pronoun paradigm end up
the way it did? The paradigm of pronouns in the substratum languages provides
an answer to this question.
For example, the Fongbe personal pronoun paradigm is as shown in (3).
(3)

nyy
hwy
(yy)
m
y

1st-person singular
I, me
fongbe
2nd-person singular
you (sg.)
3rd-person singular
he/she/it/him/her
1st- and 2nd-person plural we/us/you (pl.)
3rd-person plural
they/them
(=(18) in Brousseau 1995a)

Interestingly enough, in Fongbe, like Haitian, the same lexical entry is used
for both first- and second-person plural. Traditional Fongbe grammars (e.g.
Anonymous 1983) sometimes represent the form meaning we as mJ (bearing
a complex lowhigh tone) and the form meaning you (pl.) as mC (bearing a
mid tone). Mid and lowhigh tones are phonetic variants (see Capo 1991); the
distinction between the two is thus phonemically irrelevant. Furthermore, extensive fieldwork by Brousseau (1995a) on the tonological specification of the form
m reveals that this form always bears a phonological high tone. As in Haitian,
the Fongbe paradigm of personal pronouns presents no gender distinctions. Thus,
the properties of the Fongbe lexical entries in (3) correspond to those of the
Haitian lexical entries in (1).
These correspondences follow directly from the relexification hypothesis. In
this view, speakers of a language like Fongbe who were relexifying their native
lexicon with data from French would have looked for a pronominal form to
relabel the copied lexical entry for m we/you (pl.). They relabelled it as nou
we/you (pl.), on the basis of the French form nous we. They would not
search for another form since their own lexicon had only one lexical entry for
both first- and second-person plural. Similarly, there would be no search for

6.2 Possessives

143

alternative forms of the third person to encode gender because there were no
such forms in their native lexicon. Consequently, in contrast to French, and like
Fongbe, the Haitian paradigm of personal pronouns has only two forms to refer
to third person: li, singular, and yo, plural. The fact that in Haitian, but not in
Fongbe, the form of the third-person personal pronoun also serves as a plural
marker in noun phrases has been accounted for in chapter 4.
Typically, strong personal pronouns may be clefted, topicalised and conjoined.
Tonic pronouns in Haitian, French and Fongbe are no exception (see Brousseau
1995a). Nevertheless, when the syntactic distribution of French and Fongbe personal pronouns differs, Haitian follows the Fongbe rather than the French pattern,
as we will see in the section on possessives.
6.2

Possessives

French has paradigms of possessive adjectives and pronouns. Haitian does not
have any corresponding lexical entries. It will be shown that this situation also
follows from the relexification hypothesis: the creators of Haitian (including the
Gbe speakers) simply had no such forms to relexify.
In French noun phrases, the possessor phrase is expressed by means of what
traditional grammars refer to as possessive adjectives (see Grevisse 1975; Haase
1965). The paradigms of possessive adjectives are shown in (4). These forms
encode the person and number of the possessor and, in some cases, the gender
and number features of the possessed.
(4)
1st sg.
2nd sg.
3rd sg.
1st pl.
2nd pl.
3rd pl.

Possessed singular
masculine
feminine
mon
ma
ton
ta
son
sa
notre
votre
leur

Possessed plural
mes
tes
ses
nos
vos
leurs

french
my
your
his/her
our
your
their

As shown in (5), French possessive adjectives occur prenominally. Number and

gender features must match those of the possessed np.
(5)

a. mon / notre enfant

my / our child

b. mes / nos enfants

my / our children

french

As is pointed out in Haase (1965), the paradigm in (4) had already been established by the seventeenth century.
Dans lancien franais et souvent encore au XVIe sicle, on employait pour
marquer la possession, la place du pronom possessif atone, le pronom personnel tonique avec de. Cet emploi nexiste plus pour ainsi dire au XVIIe sicle.
Cependant on construit quelquefois de avec le pronom personnel de la 3e pers.
pour viter lquivoque et aussi pour accentuer le pronom.

144

Pronouns
[In Old French and often even in the sixteenth century, possession was marked
by the tonic personal pronoun with de, instead of the atonic possessive pronoun. For all practical purposes, this usage no longer existed in the seventeenth
century. However, sometimes de was used with the third-person personal pronoun to avoid ambiguity and also to emphasise the pronoun.]

Hence, although possession was generally encoded by means of possessive

adjectives in seventeenth-century French (e.g. ses frres his/her brothers), the
construction les frres delle her brothers was still in use with third-person
personal pronouns. As is pointed out in Chaudenson (1993), in popular French,
the possessive adjective may also be emphasised by means of a personal pronoun introduced by , as in ses frres lui (Lit.: his brothers to him). How do
these French facts compare with Fongbe?
As Brousseau (1995a) shows, Fongbe has possessive adjectives, but the paradigm is defective in that it comprises only two forms: c 1st person sg. and
tw 2nd person sg.. As can be seen in (6), these forms occur postnominally.
(6)

xw
c / tw
house my / your
my/ your houses

lz
pl

fongbe
(=(31) in Brousseau 1995a)

However, the regular way of encoding possession in Fongbe is by means of the

Genitive construction discussed in chapter 4. The possessor phrase is headed by
the Genitive Case marker tWn, which takes a noun or a strong pronominal form
as its complement, as shown in (7).
(7)

a. [xw [By twn]

house Bayi gen
Bayis houses

lz]
pl

b. [xw [nyy / hwy / / m / y

house me / you / (s)he / we, you / they
my/your/his, her/our, your/their houses

fongbe

twn]
gen

lz]
pl

fongbe

(=(30) in Brousseau 1995a)

Hazoums (1990) description of possessive constructions in several Gbe dialects reveals the following similarities and differences. First, no Gbe dialect has
a complete paradigm of possessive adjectives. Like Fongbe, Gungbe only has
two forms corresponding to those in (6). Ajagbe and Gengbe only have a form
for first-person singular, and Tufingbe only for second-person singular. Other
Gbe dialects such as Xwedagbe have no such forms at all. Second, in all Gbe
dialects possession may be expressed by means of the Genitive construction
illustrated in (7b) for Fongbe.
Haitian compares with its contributing languages as follows. First, it has no
possessive adjectives of the type we find in French (see (4) ) (see also Mufwene
1993b). It has no forms corresponding to the Fongbe ones in (6) either. The only
way of expressing possession in Haitian is by means of a personal pronoun used
in the Genitive construction, as in (8). Recall from chapter 4 that Genitive Case
is phonologically null in Haitian.

6.2 Possessives
(8)

a. [kay
[Jan ]
yo]
house John gen pl
Johns houses
b. [kay
[mwen / u
/
house me
/ you /
my/ your houses

145

haitian

. . . ]
yo]
. . . gen pl

haitian

The Haitian facts in (8b) parallel the Fongbe facts in (7b). In both, the possessor
phrase is headed by a Genitive Case marker (=s in English) and, as was argued
in chapter 4, the possessor phrase occurs in the specifier position of the functional category assigning Genitive Case. This contrasts with the Case assigned to
the possessive phrase in the French expression les frres delle mentioned above.
Indeed, as was argued in chapter 4, French de is equivalent to of in English,
not to s, so delle occurs in the complement position of frres rather than in
a specifier position. The Haitian data also contrast in Case with the emphatic
pronominal phrase in ses frres lui also mentioned above. Again, as was
extensively argued in chapter 4, French is not a Genitive Case marker. Haitian
also parallels Fongbe in allowing the full paradigm of pronominal forms in the
Genitive construction (see (7b), (8b) ). This contrasts with French which, according to Haase (1965), only allows third-person pronouns to occur in the construction les frres delle the brothers of her. I therefore conclude that Haitian
follows the pattern of the substratum languages in expressing possession by
means of a pronoun in a Genitive construction, which contrasts with all the
French constructions used to encode possession. This conclusion runs counter to
that of Chaudenson (1983), who claims that the Haitian expression of possession
in (8b) follows the French pattern in such expressions as ses frres lui (Lit.:
his brothers to him).
The lack of possessive adjectives in Haitian follows straightforwardly from
the relexification hypothesis. The French forms in (4) which have no Fongbe
counterpart (see (6) ) were not incorporated into Haitian because there were no
such entries in the native lexicon to be relexified. Brousseau (1995a) suggests
that the Fongbe forms c and tw in (6) were not relexified for the following
reasons. First, the native grammar already offered another regular option for
encoding the same relationships, namely personal pronouns used in the Genitive
construction (see (7b) ). Second, the fact that the Fongbe paradigm of possessive
adjectives is defective probably played a role in leading the creators of Haitian
to abandon these two lexical entries in creating the new lexicon. Furthermore, as
we saw above, the paradigm of possessive adjectives is not only defective but, in
some Gbe languages, it is lacking altogether.
A similar case involves the French possessive pronouns. These pronouns are
used to refer to a possessive np as a whole, as shown in (9).
(9)

Les tiens sont rouges. Les miens sont bleus.

det yours are red
det mine are blue
Yours are red. Mine are blue.

french

146

Pronouns

The French paradigm of possessive pronouns is presented in (10) (from Grevisse

1975; Haase 1965). These pronominal forms encode person, number and gender
features. Number and gender features must match those of the antecedent np
(see (9) ).
(10)

Possessed singular
Possessed plural
masculine
feminine masculine
feminine
1st sg. mien
mienne
miens
miennes
2nd sg. tien
tienne
tiens
tiennes
3rd sg. sien
sienne
siens
siennes
1st pl.
ntre
ntres
2nd pl.
vtre
vtres
3rd pl.
leur
leurs

french
mine
yours
his/hers/its
ours
yours
theirs

Gbe languages do not have possessive pronouns corresponding to the French

ones in (10) (see Hazoum 1990). For example, in Fongbe, this type of possessive construction is rendered in the same way as in English. The Fongbe possessed np is phonologically null. At S-structure, this null head is followed by the
possessor phrase, which can be headed by either a noun or a personal pronoun
marked for Genitive Case realised as tWn. This is shown in (11), where the
antecedent of the possessed may be anything one can possess.
(11)

Kk sv
[ [sb / nyy
Koku take
Asiba
me
Koku took Asibas/mine.

tw] ].
gen

fongbe

Haitian lacks the French paradigm of possessive pronouns in (10), sharing

this feature with Gbe languages. Furthermore, it encodes reference to a whole
possessive np by means of a Genitive construction involving personal pronouns
followed by the null Genitive Case marker.
(12)

Jan
pran [pa
[Mari /
John take thing Mary
John took Marys/mine.

mwen
me

] ].
gen

haitian

The Haitian data in (12) have the same structure as the Fongbe ones in (11).
Note, however, that the possessed np in Haitian is realised as pa, a head filler.
Haitian pa is phonologically derived from the French form part share (see
Goodman 1964). However, it is most appropriately glossed as thing since it
may refer to any possible antecedent that may be possessed. This contrasts with
Fongbe, where the possessed np is null (see (11) ). The difference between the
two languages thus appears to be that, while Fongbe allows the possessed to
be null, Haitian does not (see Kinyalolo 1994; Brousseau 1995a). This may be
attributable to the fact that, in Fongbe, Genitive Case is phonologically spelled
out, whereas in Haitian it is not, as is shown in (11) and (12), respectively (see
also chapter 4). A more precise explanation of this discrepancy between the two
languages will have to await further research.
Again, the fact that the French possessive pronoun paradigm in (10) did not
make its way into Haitian follows directly from the relexification hypothesis.

6.3 Logophoric pronouns

147

The native speakers of a language like Fongbe who were relexifying their lexicon with data from French would not search for possessive pronominal forms
because they had no such lexical entries to relexify. Mufwene (1993b) shows
that, while some creole languages such as Haitian and Jamaican do not seem to
have possessive pronouns/adjectives, others do (e.g. Gullah, Papiamentu and the
Portuguese creoles of the African Atlantic coast). In view of the account proposed
for Haitian, a comparison of the latter group of creoles with their substratum languages might explain this discrepancy. The two sets of data discussed in this
section constitute further evidence that the syntactic distribution of Haitian personal pronouns follows that of the substratum languages, exactly as one would
expect under the hypothesis that Haitian pronouns are historically derived by
relexification.
6.3

Logophoric pronouns

Some West African languages have a form of personal pronoun which, unlike
other pronominal forms, has no independent reference (see e.g. Hagge 1974;
Clements 1975; Hyman and Comrie 1981). This kind of pronoun is referred to as
logophoric. In addition to the personal pronouns discussed in section 6.1, Fongbe
has the logophoric pronoun m illustrated in (13). As is extensively discussed
in Kinyalolo (1993b, 1993c), a logophoric pronoun is morphologically distinct
from the other pronouns of a particular language. In Fongbe, its antecedent may
be second person, as in (13a), or third person, as in (13b) and (13c), but not first
person.
(13)

Sk tn
Cica know
hwl sb
hide Asiba

3w
say
sn
of

y
they
gbh.
goat

3w
say you

fln
3w
remember say

m
log

fongbe

a. Cica knows that they said that youi remember that youi hid As goat.
b. Cica knows that theyi said that you remember that theyi hid As goat.
c. Cicai knows that they said that you remember that shei hid As goat.
(=(13) in Kinyalolo 1993c)

Furthermore, the distribution of such a pronoun is associated with a class of

verbs known as logophoric verbs, that is, verbs of the class of say (see Koopman
and Sportiche 1989). The ungrammaticality of (14) shows that m has no independent reference.3
(14)

*m hwl sb sn gbh
log hide Asiba of goat
S/he hid Asibas goat.

fongbe
(=(6) in Kinyalolo 1993c)

As is pointed out by Kinyalolo (1994), Haitian has no form with the properties of a logophoric pronoun. In my view, this is because logophoric pronouns

148

Pronouns

cannot be relabelled since they are not semantically independent. As we saw in

(13), they are interpreted on the basis of the context in which they appear. Presumably, logophoric pronouns were abandoned in the course of Haitian genesis,
since they are not found in modern Haitian.
6.4

Pronominal clitics

The three languages under comparison distinguish between strong or tonic pronominal forms and weak or atonic pronominal forms, henceforth referred to as
(pronominal) clitics. Diagnostic tests distinguishing clitics from pronouns are
well established in the literature (see e.g. Zwicky 1976; Kayne 1975). While
pronouns may be separated from the tensed verb, clitics cannot, except by
another clitic. For example, in French, Je le lui ai dit I told him that the clitic
le is separated from the tensed auxiliary ai by the clitic lui. While pronouns can
be conjoined (e.g. moi et toi me and you), clitics cannot (e.g. *Je et tu I and
you). While pronouns can be focussed or topicalised, clitics cannot (e.g. *Cest
je/Cest moi It is me).
There are two types of clitics: syntactic and phonological. Syntactic clitics
constitute lexical entries which are distinct from pronouns. In contrast to pronouns, which are major category lexical items, syntactic clitics head functional
category projections in the vicinity of the vp and bind an argument position
which is phonologically empty at S-structure (see Di Sciullo 1990; Laka Mugarza
1990; Ouhalla 1991; Roberts 1993; Rizzi 1986a; Sportiche 1994). Phonological
clitics, on the other hand, do not constitute separate lexical entries; they are
phonologically reduced forms of strong forms, base-generated in an np position.
At pf (the level of Phonological Form), they reduce and cliticise onto the verb.
The following tests distinguish syntactic clitics from phonological clitics.
First, in languages which have syntactic clitics, the paradigm of such clitics
includes at least some forms which are phonologically distinct from the strong
forms. By contrast, the forms of phonological clitics are always phonologically
derived from the strong pronominal forms. Second, while the distribution of
pronominal forms generally follows that of nouns, the surface distribution of
syntactic clitics often differs from that of nouns and pronouns. For example, in
French, while a nominal object follows the verb (e.g. Je veux ce livre I want
this book), a clitic object obligatorily precedes it (e.g. Je le veux I want it).
However, phonological clitics have the same surface distribution as nouns and
pronouns (see below). Third, syntactic clitics may (but need not) undergo clitic
climbing in particular languages. Clitic climbing is the phenomenon whereby a
clitic which binds an argument of an embedded verb is found, at S-structure, in
the domain of the main verb of its clause. Most varieties of Spanish have this
option. For example, in Lo quiero matar I want to kill him, the clitic lo binds
an argument of the verb matar kill, but in the syntax it occurs in the domain of
the verb quiero I want, the main verb of the clause. Unlike syntactic clitics,
phonological clitics are not subject to clitic climbing. Fourth, while syntactic

6.4 Pronominal clitics

149

clitics are syntactically conditioned, phonological clitics are phonologically conditioned as discussed below. Finally, while syntactic clitics are excluded from
nominal structures and from pps (since, in the syntax, they cliticise onto the
verb), phonological clitics may appear in these contexts because they are strong
forms reduced in the phonological component of the grammar. The clitics of the
three languages under comparison will be discussed in light of this theoretical
background. It will be shown that, while both Fongbe and French have syntactic
clitics, Haitian has only phonological clitics. I shall begin this comparison with
Fongbe.
Syntactic clitics in Fongbe are distinguished on the basis of Case: [+/nominative] (see Lefebvre 1991a; Brousseau 1995a). The two paradigms are presented
in (15) together with the pronominal forms discussed in section 6.1.
(15)
1st sg.
2nd sg.
3rd sg.
1st, 2nd pl.
3rd pl.

Strong Weak [+nominative] Weak [nominative]

fongbe
nyy
un
m
hwy

w
(yy)

4
m
m
y
y
(=(18) in Brousseau 1995a)

The Fongbe clitics are marked for person and number. As shown in (15), they
are distinguished for Case only in the singular; plural forms are neutral with
respect to this feature. The same form encodes both first- and second-person
plural, and the clitic plural forms are the same as the corresponding strong forms.
Finally, Fongbe clitics are not identified for gender.
Clitics in Fongbe present all the characteristics of syntactic clitics. First,
some of the clitic forms are distinct from the strong forms: the first- and secondperson singular clitics are clearly phonologically distinct from the corresponding
strong forms and the third-person singular clitics are partially distinct from the
corresponding strong form. Second, in some instances, Fongbe clitics occupy a
surface position which is different from that of nouns and pronouns. For example,
in syntactic nominalisations, the direct object of a verb typically precedes the
verb when it is realised as a noun or pronoun. However, the direct object follows
the verb when it is realised as a clitic. This contrast in word order is illustrated
in (16).
(16)

a. sb 3 By / hwy kpvn wy.

Asiba at Bayi / you look post
Asiba is looking at Bayi/ you.

fongbe

b. sb 3 kpkpvn w
Asiba at look
2nd
Asiba is looking at you.

fongbe

wy.
post

(=(19) in Brousseau 1995a)

Third, for some but not all Fongbe speakers, Fongbe clitics may undergo clitic
climbing (see Lefebvre 1991b).5 This phenomenon, as instantiated in the grammar

150

Pronouns

of a subgroup of Fongbe speakers, is illustrated in (17) with serial verbs. In the

example below, the clitic object binds the direct argument of the second verb,
as shown in (17a). It may optionally occur in the domain of the first verb, as
shown in (17b).
(17)

a. Kwk sv
sv n
.
b. Kwk sv

sv n. fongbe
Koku take crab give 3rd
Koku take 3rd crab give
Koku gave a crab to him.
Koku gave a crab to him.
(=(13) in Lefebvre 1991b)

Finally, Fongbe clitic forms are excluded from nominal structures and from pps,
where only strong pronominal forms are permitted. All these facts argue that
Fongbe clitics are syntactic.
French has two paradigms of clitics also distinguished on the basis of Case:
[+/nominative]. These paradigms, as reconstructed by Brousseau (1995a) on the
basis of French grammars, are presented in (18) together with the pronominal
forms discussed in section 6.1.
(18)
1st sg.
2nd sg.
3rd sg.
1st pl.
2nd pl.
3rd pl.

Strong
moi
toi
lui / elle
nous
vous
eux / elles

Weak [+nominative] Weak [nominative]

je
me
tu
te
il / elle
le / la / lui
nous
vous
ils / elles
les / leur

french

The clitic forms above are marked for person and number. Some of the forms
encoding third person are also marked for gender: il(s) and le are masculine and
elle(s) and la are feminine. The [+/nominative] distinction is lexically encoded
throughout the paradigm of clitics except for first- and second-person plural. The
clitics lui and leur are Dative.
French clitics have the characteristics of syntactic clitics (see e.g. Kayne
1975; Roberge 1990). First, as can be seen in (18), some clitic forms are phonologically distinct from the pronominal forms (e.g. the forms encoding first- and
second-person singular). Second, in contrast to nominal objects, which follow
the verb, object clitics always precede the verb, as is shown in (19).
(19)

a. Marie a
annonc
la
nouvelle
Mary aux announce det news
to
Mary announced the news to John.
b. Marie la
lui a
annonce.
Mary 3rd 3rd aux announce
Mary announced it to him.

Jean.
John

french

french

Third, according to the literature, clitic climbing was still in use in seventeenthcentury French (see Oudin 1640; Vaugelas 1647; Bary 1665; Galet 1971). And,
finally, French clitics do not occur in nominal structures or pps, where only
strong pronouns are allowed. The above facts argue for the syntactic status of
clitics in both seventeenth-century and modern French.

6.4 Pronominal clitics

151

In addition to the forms listed in (18), French also has the clitics en and y.
The clitic en replaces a phrase headed by de of, and y replaces a phrase headed
by to (see e.g. Kayne 1975). These two forms are exemplified in (20).
(20)

a. Tu
as
peur de Marie. Tu
en as
peur.
2nd sg. aux afraid of Mary 2nd sg. cl aux afraid
You are afraid of Mary. You are afraid of her.

french

b. Tu
penses Marie. Tu
y
penses.
2nd sg. think
of Mary
2nd sg. cl think
You think of Mary. You think of her.

french

c. Tu
iras
Paris. Tu
y
2nd sg. go-fut to Paris
2nd sg. cl
You will go to Paris. You will go (there).

french

iras.
go-fut

Finally, French has the clitic on which can be assigned an indefinite interpretation, as in On dit que . . . One/people/they say(s) that . . . (see Gauthier 1995)
or which can be used as a first-person plural, as in Nous, on dit que . . . We, we
say that . . . The available literature on seventeenth-century French shows that
the system of clitics discussed above was already in place at the time Haitian
creole was formed (see e.g. Haase 1965: 13). Furthermore, the literature on
Western dialects of France shows that the clitic system in these dialects is quite
similar to that described above (see Brasseur 1995; Gauthier 1995). Julie Auger
(p.c.) points out, however, that in French regional dialects, such as Picard, firstand second-person plural clitics are not overtly distinguished. Interestingly enough,
this dialect is like Fongbe in this respect (see (15) ).
Both the substratum and superstratum languages of Haitian have syntactic
clitics. Do the clitic forms in Haitian present the characteristics of syntactic
clitics? In addition to the strong pronominal forms discussed in section 6.1,
Haitian has a paradigm of weak pronominal forms. The strong and weak forms
are shown in (21) (see Cadely 1994; Brousseau 1995a). Note that, like the strong
forms, the weak forms are specified for person and number, but not for gender.
The Haitian weak forms show no Case distinctions.
(21)
1st sg.
2nd sg.
3rd sg.
1st, 2nd pl.
3rd pl.

Strong forms
mwen
ou
li
nou
yo

Weak forms
m
u/w
l
n
y

haitian

Deprez (1992a), DeGraff (1992a, 1993b) and Cadely (1994) show convincingly
that the paradigm of reduced forms in (21) exhibits the general characteristics of
clitics. For example, unlike the strong forms, which can be conjoined (e.g. mwen
ak li me and him), the weak forms cannot (e.g. *m ak l ). The strong forms can
also be clefted and topicalised, the weak forms cannot.
On the basis of these facts, DeGraff (1992a, 1993b) concludes that Haitian
weak forms are syntactic clitics. Deprez (1992a) and Cadely (1994), however,

152

Pronouns

argue that they do not have the characteristics of syntactic clitics. First of all,
none of the Haitian weak forms in (21) has a phonological representation which
is independent of the corresponding strong form: all of them appear to be phonologically derived from the strong forms. Moreover, the weak forms have exactly
the same surface distribution as the strong forms. Deprez (1992a) points out that
the subject weak form occurs in the same position as the lexical subject, and
thus it can be separated from the verb or the aspectual markers by an adverb,
as is shown in (22).
(22)

Jan / m poko
ap rakonte yon
John / I not-yet imp tell
a
John/I is/am not yet telling a story.

istwa.
story

haitian
(=(25b) in Deprez 1992a)

According to Deprez (1992a), this shows that, unlike a syntactic clitic, the
Haitian weak form does not necessarily form a unit with the verb or aspectual
markers. As well, the direct object always follows the verb, whether it is realised
as a clitic, a pronoun or a noun. This is exemplified in (23).
(23)

Mari pte l
/ li
/ ti-mounn
Mary carry him / him / child
Mary carried him/the child.

nan.
det

haitian

Clitic climbing does not appear to be an option available to Haitian speakers, as

shown by the ungrammaticality of the sentence in (24b), the Haitian counterpart
of the grammatical Fongbe sentence in (17b).
(24)

a.

Jan
prann krab bay li.
John take
crab give him
John gave him crab.

b. *Jan
prann
John take

-ni krab bay

him crab give

haitian

haitian

Deprez (1992a) also shows that weak forms may occur in nominal structures, as
in (25).
(25)

kaye m
book me
notebook of mine

haitian
(=(31) in Deprez 1992a)

Furthermore, a weak form may occur as a complement of a preposition, as

shown in (26), from DeGraff (1992a).
(26)

Ou-mem
ave m ale nan mache.
you-emph with me go in
market
You and I went to the market.

haitian
(=(12b) in DeGraff 1992a)

Based on these facts, Deprez (1992a: 32) concludes: It is thus clear that the
appearance of the reduced forms is not conditioned by the potential location of

6.4 Pronominal clitics

153

pronouns in in. The Haitian weak forms in (21) present none of the characteristics of syntactic clitics. With Deprez, I therefore conclude that they are not
syntactic clitics.
Cadely (1994) convincingly argues that the Haitian weak forms are phonological clitics. He shows that the distribution of strong and weak forms is
phonologically conditioned. While strong forms precede verbs that begin with
a consonant and follow verbs that end in a consonant, as in (27), weak forms
occur adjacent to vowels, as in (28).
(27)

a. Jan
John
yo
they
John

remt
mwen / ou
/ li
give-back me
/ you / him
liv
la.
book det
gave me / . . . / back the book.

b. Mwen / ou
me
/ you
I / . . . / came.

/ li
/ him

/
/

/ nou
/
/ we, you /

nou
/ yo
vini.
we, you / they come

haitian

haitian

The example in (28a) illustrates regressive cliticisation, while (28b) shows progressive cliticisation.
(28)

a. Jan
mande m / w
/ l
/ n
/ y
John ask
me / you / him / us, you / them
kb
la.
money det
John asked me / . . . / for the money.

haitian

b. M / w
/ l
me / you / him
I / . . . / went home.

haitian

/ n
/ y
ale lakay.
/ we, you / they go home

According to Cadelys (1994) analysis, strong forms are phonologically reduced

in the environment of a vowel. This reduction process yields the weak forms in
(21). The segment that remains after reduction becomes part of the adjacent
syllable of the verb (initial or final) and the new phonological word is resyllabified
according to the principles of syllabic structure discussed in Brousseau (in preparation). Furthermore, when the word (be it a noun, a Wh-word, an adverb or a
pronoun) following a pronominal form starts with a consonant, as in (27b), the
weak form may cliticise leftward, onto a word ending in a vowel. This is shown
in (29) (from Deprez 1992a, based on Cadely 1990).
(29)

kaye li dechire a

the notebook he has torn

kisa li fe
What is he doing?

M we le li vini

I saw when he came.

Se pa nou li we
Its not us he saw.

kayel dechire a
kisal fe
M we lel vini
Se pa noul we

haitian

154

Pronouns
Le mwen tande
When I hear.

Lem tande

Le li tande
When he hears.

Lel tande

Le nou tande
When we hear.

Len tande
(=(32) in Deprez 1992a)

Deprez (1992a: 33) concludes that the above facts

pose a serious challenge to an analysis of cliticisation as a syntactic process in
Haitian. Clearly, indeed, if as proposed by DeGraff, reduced pronominal forms
are agreement markers which occur in in to license an empty pronominal
subject, one does not expect them to be able to cliticise to the previous
words . . . The facts in [ (29) ], on the other hand, pose no particular problems, if
hc pronouns are assumed to be phonological clitics.

With Deprez (1992a) and Cadely (1994), I therefore conclude that Haitian weak
forms are phonological clitics.
Haitian creole thus differs from both of its source languages in not having
syntactic clitics. Furthermore, while both Fongbe and French each have two
paradigms of clitics distinguished on the basis of Case, Haitian has only one
paradigm and the forms of this paradigm are used indiscriminately for all syntactic functions. In addition, while French clitic forms show gender features in
some cases, none of the Haitian forms do. There are no forms corresponding to
French en and y (see (20) ) in Haitian. The corresponding forms in Haitian must
be strong personal pronouns, as shown in (30a) (=(20b) in French), or a null
form as in (30b) (=(20c) in French).
(30)

a. Ou ap panse ak
Mari. Ou ap panse ak
li. haitian
you imp think with Mary you imp think with her
You are thinking of Mary. You are thinking of her.
b. Ou ap
ale Pari. Ou ap
you def-fut go Paris you def-fut
You will go to Paris. You will go (there).

ale
go

haitian

In this respect, Haitian shows the same pattern as Fongbe, which does not have
clitics corresponding to French en/y. The Fongbe data in (31) correspond to the
Haitian data in (30).
(31)

a. Kwk ln
sb.
ln
.
Koku think Asiba 3rd think 3rd
Koku thinks of Asiba. He thinks of her.
b.
n
y Pari.
n
2nd sg. def-fut go Paris 2nd sg. def-fut
You will go to Paris. You will go (there).

fongbe

y .
go

fongbe

Finally, Haitian has no form corresponding to the French indefinite clitic on. In
this respect, as well, however, Haitian follows the same pattern as Fongbe. Both

6.4 Pronominal clitics

155

languages use third-person plural strong personal pronouns in order to convey

an indefinite reference, as shown in (32).
(32)

Yo di . . .
Y 3w . . .
They (indefinite) say . . .

haitian
fongbe

But no matter how much we push the comparison between the three languages,
the fact remains that, in contrast to both of its source languages, Haitian does not
have syntactic clitics. How is this compatible with the theory of creole genesis
outlined in chapter 2?
The contrast between the forms and properties of Haitian and French clitics
clearly shows that the creators of Haitian did not acquire the French system.
This is compatible with the claim in chapter 2 that the creators of a creole do not
identify the functional items of the superstratum language because of their limited access to the data. What is surprising about the Haitian clitics, however, is
the fact that they do not reflect the properties of the substratum language clitics
either, as the relexification hypothesis predicts that they should. Below, I discuss
various possible reasons why the substratum syntactic clitics have not made
their way into Haitian.
A first possibility is that there were no syntactic clitics in Gbe dialects of the
seventeenth century so that the creators of Haitian did not have any clitics to
relexify. This would straightforwardly explain the problematic data. This possibility is not unlikely for, as has been pointed out to me by Yves-Charles Morin
(p.c.), clitics are subject to rapid change in the languages of the world. In line
with the general methodology adopted for the discussion of the data in this book
(see chapter 3), however, I will assume that the substratum lexicons did include
pronominal clitic forms.
Assuming that this was the case, could it be that the nature of clitics prevents
them from being relexified, as is advocated in Muysken (1988c)? Other creoles
show phonological clitics rather than syntactic clitics. For example, Kouwenberg
(1993) argues that Berbice Dutch, like Haitian creole, has a paradigm of strong
personal pronouns which can phonologically cliticise onto the verb. Although
Eastern Ijo, the main African substratum language of Berbice Dutch, does have
syntactic clitics which are arguably distinct from the strong pronominal forms,
Berbice Dutch does not. Haitian creole is thus not the only creole that lacks syntactic clitics when compared to its substratum languages. Given these considerations, it is possible that clitics do not undergo relexification. But since other
functional category lexical entries which have some semantic content can be
argued to undergo relabelling, we cannot exclude the possibility that syntactic
clitics can also be relabelled. Assuming that syntactic clitics may theoretically
undergo relabelling, could it be that they do not enter the creole because of the
way relabelling applies to functional items?
Recall from chapter 2 that functional category lexical items are relabelled with
major category superstratum lexical items. Suppose then, following Brousseau

156

Pronouns

(1995a), that the creators of Haitian relexified the clitics of their own lexicon
using French strong personal pronouns. We saw in section 6.1 that these forms
were also used to relabel the lexical entries copied from the strong pronouns. For
example, on this hypothesis, the copied lexical entries of all three Fongbe firstperson singular pronominal forms were relabelled on the basis of French moi,
yielding mwen in Haitian, as shown in (33) (where [+/argument] stands for
strong and clitic forms, respectively).
(33)
a. [1st], [plural], [+argument]

fongbe
nyy

haitian
mwen

b. [1st], [plural], [argument], [+nominative]

un

mwen

c. [1st], [plural], [argument], [nominative]

mwen

Consequently, in the incipient creole, there would be three homophonous forms

for the first-person singular pronominal lexical entries. The above scenario is not
unlikely for the following reasons. First, there are other cases where two substratum lexical entries have been relabelled on the basis of a single French form
(see e.g. section 6.7). Second, the possibility for a given pronominal form to
appear as both strong and weak is attested in other languages; the French forms
nous and vous in (18) constitute such examples. The availability, in the incipient creole, of the lexical entries in (33) would enable the creators of Haitian who
had both strong and weak pronominal forms in their original lexicons to reproduce these forms in the creole. However, using the same superstratum string to
relabel several lexical entries copied from the substratum language(s) yields
redundancy in the newly created lexicon (see (33) ). In view of this situation,
Brousseau (1995a) further hypothesises that the three homophonous lexical
entries in (33) were reduced to one, with their common features yielding a single
Haitian lexical entry unspecified for the features [argument] and [nominative],
as shown in (34).
(34)

/mwen/: [1st], [plural]

haitian

This is a likely hypothesis since this is what seems to have happened in other
cases where two substratum language lexical entries have been relabelled using
the same superstratum form (see e.g. section 6.7). The fact that the lexical entry
in (34) is underspecified for the feature [argument] also enables the creators of
Haitian who had both strong and weak pronominal forms in their original lexicons to produce these forms while speaking the creole. This cannot be the end
of the story, however, for there are no syntactic clitics in modern Haitian.
How is the claim that syntactic clitics may undergo relabelling, as described
above, reconcilable with the observation that there are no such clitics in modern
Haitian? On the basis of other cases of regular linguistic change (see Lightfoot
1979), the answer to this question should reside in the analysis that the first generation of Haitian native speakers assigned to the data they were presented with.
The question at stake, then, is whether the first generation of Haitian native speakers

6.5 Expletives

157

perceived the distinguishing properties of the forms that they were exposed to.
The answer to this question should be negative on the following grounds.
Whether the first generation of Haitian native speakers were exposed to the
data in (33) or (34), they had no clue for distinguishing between strong and
weak forms on the basis of that data. Presumably, they observed the same form
in all contexts where a pronominal was used by the adult population. Furthermore,
the context par excellence where the clitic and the strong forms were distinguished
in terms of word order in the original grammar (that is, in nominalisations, see
(16) ) had been abandoned in the early creole (see chapter 5). It is thus reasonable to conclude that the first generation of Haitian native speakers could not
deduce the availability of syntactic clitics on the basis of the data that they were
exposed to. Presumably, they interpreted these data as in (35), and thus in
modern Haitian there are no syntactic clitics.
(35)

/mwen/: [1st], [plural], [+argument]

haitian

On this approach, the fact that syntactic clitics did not enter the creole is
linked to how relabelling is hypothesised to proceed in the case of functional
category lexical entries in creole genesis. In this case, relexification creates
redundancy in the incipient creole lexicon (see (33) ). Even if this redundancy is
reduced (see (34) ), the lexical entries produced by relexification are opaque to
the language learners, who cannot deduce their distinguishing properties. The
language learners make an assumption about the opaque data that they are presented with, which results in the loss of syntactic clitics in the creole (see (35) ).
Presumably, phonological clitics develop in creoles through the reduction of full
forms. This is a likely hypothesis since this is what we see in modern Haitian
and in the other creoles mentioned earlier.6
6.5

Expletives

Expletive subjects that have the same form as third-person personal pronouns
(e.g. it in English) are a common phenomenon among the languages of the
world. The three languages being compared here are no exception. In all three,
the form of the expletive is the same as that of the third-person singular personal
pronoun discussed in section 6.1: li in Haitian, il in French and in Fongbe, as
is shown in (36).
(36)

a. Li bon
pou
Jan
pati.
it good comp John leave
It is good that John will leave.

haitian

b. Il est
bon que
Jean parte.
it aux good comp John leave
It is good that John will leave.

french

c. nywn 3w
Kwk n
y.
it good comp Koku irr leave
It is good that Koku will leave.

fongbe

158

Pronouns

As has been noted in several instances in the literature, the Haitian expletive
is not always overtly realised in the subject position of a verb which takes an
expletive subject (see e.g. Koopman 1986; Massam 1989; DeGraff 1992a, 1992d,
1993b, 1994; Deprez 1992a; Vinet 1991; Law 1992). The data in (37) exemplify
this fact.
(37)

(Li) sanble Jan

te
malad.
it
seem
John ant sick
It seems that John has been sick.

haitian

Koopman (1986), Massam (1989), Vinet (1991) and Deprez (1992a) have proposed that, in addition to an overt expletive subject, Haitian has a phonologically
null expletive subject. Not all languages have this option. For example, as is
shown in (38), French does not.
(38)

*semble que
Jean soit malade
seem
comp John be sick

french

Auger (1994) points out, however, that there are a few idiosyncratic cases in
colloquial French where the expletive subject may be missing, in spite of the
fact that this language generally does not allow phonologically null expletives.
The data in (39) are examples of such cases.
(39)

a. (Il) faut
y
aller. b.
it
requires there to-go
We must go there.

(Il) me semble que . . .

french
it
me seem
that . . .
It seems to me that . . .
(=(17a, b) in Auger 1994)

The French data in (39) constitute idiosyncrasies for it is not possible to delete
the expletive throughout a given paradigm. While (39b) is grammatical with or
without the expletive, (40) is not grammatical without the expletive.
(40)

*te / lui / nous / . . .

you / him / we / . . .

semble que . . .
appear that

french

For this reason, Auger argues that cases such as those in (39) are lexically listed
impersonal expressions and she proposes to account for them in terms of morphological rules. These lexical idiosyncrasies in French are different from the
phonologically null expletive subject proposed for Haitian. In Haitian, the possibility of overt or covert expletives is determined for each verb, as will be seen
in chapter 9. When a verb allows a null expletive, it allows it throughout the
paradigm, as shown in (41).
(41)

sanble
dapr mwen / ou / li . . .
It seems to me/ you/him . . .

haitian

The Haitian data in (41) contrast with the French data in (40), showing that
missing expletives in the two languages do not constitute an instantiation of the
same phenomenon.
Interestingly enough, however, Fongbe, like Haitian, does have a phonologically null expletive subject, as is illustrated in (42).

6.6 Reflexives
(42)

() 3
3w
Kwk jzwn.
it
seem comp Koku sick
It seems that Koku is sick.

159

fongbe

In Fongbe as in Haitian, the possibility of overt or covert expletives is determined for each verb (see chapter 9) and, when a verb allows a null expletive, it
allows it throughout the paradigm, as shown in (43).
(43)

3 n
m / hwy / . . .
It seems to me/ you/him . . .

fongbe

The Fongbe data in (43) parallel the Haitian data in (41) and both languages
contrast with French (40) in the same way.
So, in both Haitian and Fongbe, but not in French, a phonologically null
expletive is available. On the analysis that a phonologically null expletive constitutes a lexical entry, the creators of Haitian would have copied it in the creole lexicon. No relabelling was necessary in this case. If a phonologically null
expletive does not constitute a lexical entry, the creators of Haitian would have
carried over into the creole the principle allowing for null expletives in their
original grammar.
6.6

Reflexives

Reflexivity has been a very popular topic in the recent literature on creoles (e.g.
Taylor 1971; Carden and Stewart 1988; Corne 1988, 1989; Muysken and Smith
1995; Dchaine and Manfredi 1994; Kinyalolo 1994; Brousseau 1995a). Dchaine
and Manfredi (1994) have pointed out that Haitian lacks an overt morphological
reflexive form of the type of self in English. Some nouns and pronouns, however, are involved in the interpretation of reflexivity. In fact, Haitian offers three
possibilities for expressing this concept, as is illustrated below. The sentence in
(44a) shows that a personal pronoun may be assigned a reflexive interpretation.
The sentences in (44b) and (44c) show that nouns such as tt head and k
body followed by a possessor phrase containing a personal pronoun may also
be assigned a reflexive interpretation.
(44)

a. Mweni w mweni nan glas

I
see me
in
mirror
I saw myself in the mirror.

la.
det

haitian
(=(1a) in Brousseau 1995b)

touye tt
mweni .
b. Mi ap
I
def-fut kill
head me
I will kill myself.

haitian
(=(2a) in Brousseau 1995b)

lii .
c. Lii blese k
he hurt body him
He hurt himself.

haitian
(=(6d) in Brousseau 1995b)

160

Pronouns

The data in (44) raise the following questions. First, is there anything special
about Haitian pronouns which allows them to have a reflexive interpretation (see
(44a) )? In section 6.6.1, I argue that, when we look at the Haitian facts in light
of the Fongbe data, no special proviso needs to be made to account for the
interpretation of pronouns in Haitian. Second, body-part reflexives, as in (44b)
and (44c), appear to be a widespread phenomenon in creole languages (see
Muysken and Smith 1995). They are not a property of these languages alone,
however. Faltz (1985) reports that in Basque reflexivity is encoded by means of
the word for head plus a possessive pronoun, and in Yuman languages (spoken
in the southwestern United States) reflexivity is encoded by means of the word
for body, either with a possessive pronoun or with a pronoun attached to the
verb in the syntactic position of an incorporated N-stem. On the basis of these
facts, Carden and Stewart (1988: 32) suggest that body-part reflexives in Haitian
creole might constitute an independent development in creoles. As we will see
below, however, body-part reflexives are widespread in Kwa languages (see
Awoyale 1986), although not in the Gbe languages, and the historical derivation
of such reflexives in Haitian is argued to follow directly from the relexification
hypothesis (section 6.6.2). Finally, one might wonder why Haitian (as well as
other creole languages, see Muysken and Smith 1995) offers so many possibilities
for expressing reflexivity, when a language like English, for example, has only
one, that is, the reflexive morpheme -self. The numerous options for expressing
reflexivity in Haitian will be shown to follow from the relexification hypothesis.
This section concentrates on the forms used to encode reflexivity in the three
languages under comparison. The input of several substratum languages will be
shown to be crucial in this case. The selectional properties of verbs in relation
to these forms will be taken up in chapter 9.
6.6.1

The reflexive interpretation of personal pronouns

In Haitian, personal pronouns occurring in object position may be interpreted as

reflexives. This is shown in (45) and (46), for first- and second-person personal
pronouns.
(45)

a. Mi w mi .
I
see me
I saw myself.

(46)

a. Oui w wi .
you see you
You saw yourself.

b. Jaki
w m j .
haitian
James see me
James saw me.
(=(2a) in Dchaine and Manfredi 1994)
b. Jaki
w wj .
haitian
James see you
James saw you.
(=(2b) in Dchaine and Manfredi 1994)

As has been repeatedly observed in the literature (see e.g. Carden and Stewart
1988; Dchaine and Manfredi 1994; Muysken and Smith 1995; Kinyalolo 1994),
third-person personal pronouns occurring in object position with a number of

6.6 Reflexives

161

Haitian verbs are ambiguous between reflexive and pronominal readings. This is
shown in (47).
(47)

Lii w l i /j .
he see him /her/it
S/he saw her(self)/him(self )/it.

haitian
(=(1a) in Dchaine and Manfredi 1994)

Where does the reflexive interpretation of the pronominal forms in (45a), (46b)
and (47) come from?
In the context of the Haitian sentences in (45), (46) and (47), French does not
allow strong pronominal forms, as is shown in (48).
(48)

a. *Jei vois moii

[Lit.: I see me.]

b. *Tui vois toii

[Lit.: You see you.]

c. *Ili voit luii

[Lit.: He sees him.]

french
french

The fact is that, in French, these contexts call for syntactic clitics. The first- and
second-person object clitics discussed in section 6.4 may be assigned either a
reflexive or a free interpretation, as shown in (49) and (50).
(49)

a. Jei vais mei tuer.

1st go
1st kill
I am going to kill myself.

b. Ilj
va mei tuer.
3rd go 1st kill
He is going to kill me.

(50)

tuer.
b. Tuj vas lei tuer.
a. Tui vas tei
2nd go 2nd kill
2nd go 3rd kill
You are going to kill yourself. You are going to kill him.

french

french

French third-person clitics (le, la, les, lui, leur) cannot, however, be assigned a
reflexive interpretation, as is shown in (51a). They are in complementary distribution with the reflexive clitic se, marked for third person. This clitic is not
specified for number or gender, as is shown in (51b).7
(51)

a. *Ilsi lesi lavent

#[They wash themselves]

french

b. Il(s)
/ elle(s)
se
lave(nt).
3rd masc. sg. (pl.) / 3rd fem. sg. (pl.) ref wash (pl.)
He/she/they wash(es) himself/herself/themselves.

french

Recall from section 6.4 that Haitian does not have syntactic clitics. Thus, there
are no Haitian data corresponding to the French data in (49) and (50). Furthermore, Haitian has no form corresponding to French se. Thus, so far, the French
and Haitian data show no parallels at all. There are, however, a few contexts in
French which allow for a reflexive interpretation of a strong pronoun. In the first
place, second-person personal pronouns may be assigned a reflexive interpretation in imperative constructions, as shown in (52).

162
(52)

Pronouns
Lave-toi.
wash you
Wash yourself.

french

Second, as is pointed out in Morin (1977), a strong pronoun may (but need not)
be assigned a reflexive interpretation in contexts where cliticisation is blocked in
French, as shown in (53) and (54).
(53)

a.

Jeani n
aime que luii .
John neg like that him
John likes only himself.

french
(=(16) in Morin 1977)

b. *Il ne siaime
(54)

que

ti

french

Jeani avait un petit chien avec

John have a
small dog
with
John had a small dog with him.

luii .
him

french
(=(53) in Morin 1977)

Other such examples are also provided in Zribi-Hertz (1995).

On the basis of French facts such as those in (52)(54), Corne (1988, 1989)
suggests that the reflexive interpretation of pronouns in French based creoles is
attributable to French. In Cornes view, interpretive facts such as the Haitian data
in (45)(47) constitute an extension to all contexts of the reflexive interpretation
of strong pronominal forms. Carden and Stewart (1988) challenge his claim on
the basis of the fact that French personal pronouns can only be interpreted as
reflexives in limited contexts and crucially not in the context illustrated in (48).
Furthermore, in contrast to French, (phonological) cliticisation need not be blocked
for a pronoun to be assigned a reflexive interpretation in Haitian, as shown in
(55).
(55)

Mweni w mweni / Mweni w

I see myself.

mi .

haitian

Thus, like Carden and Stewart, I conclude that French cannot be the source of
the reflexive interpretation of the Haitian pronouns in (45)(47).
French also has the indefinite non-clitic reflexive form soi oneself occurring
in expressions such as On a parfois besoin dun plus petit que soi Sometimes
one needs someone smaller than oneself. According to Morin (1977: 75), Soi
est ncessairement indfini, et comme le clitique se, est toujours corfrentiel au
sujet de la proposition dans laquelle il se trouve. [Soi is necessarily indefinite
and, like the clitic se, is always coreferential with the subject of the clause in
which it occurs.] According to Morin, the distribution of this French form is
very limited. There is no Haitian form corresponding to this form either.
The discrepancy between the French and Haitian facts thus suggests that
French has played a minor role, if any, in determining the reflexive interpretation of the Haitian personal pronouns in (45)(47). This conclusion is in line
with Muysken and Smiths (1995) observation that the lexifier language of a

6.6 Reflexives

163

creole can only play a limited role in the historical derivation of reflexives in
creole languages in general. In view of this situation, consider data from the
substratum languages of Haitian.
According to Hazoum (1990), all Gbe languages encode reflexivisation by
means of a personal pronoun +self. In Fongbe, this lexical item is -#. Kinyalolo
(1994) states that the semantic and distributional properties of this Fongbe lexical
item are similar to those of the English -self. For example, in English, pronouns
combine with -self (e.g. He washes him-self; we wash our-selves, etc.). Similarly,
in Fongbe, the strong pronominal forms discussed in section 6.1 (but not the
syntactic clitics, as is emphasised by both Kinyalolo 1994 and Brousseau 1995a)
combine with -#, as is illustrated in (56).
(56)

a. Ni
n
x
1sg. def-fut kill
I will kill myself.

nyy-3i .
me-ref

b.

Byi mw -3i .
fongbe
Bayi see she-ref
Bayi saw herself.
(=(45) in Brousseau 1995a)

According to current analyses (see Chomsky 1981) the expression pronoun+self

is assigned an analysis such as that in (57), where -self is the head of a np and
the pronoun occurs in the specifier position.
np

(57)

english

Pronoun

-self

Kinyalolo (1994) proposes to account for the Fongbe pronoun+# in the same
way. In his analysis, the Fongbe reflexive expression would have the structure in
(58).
np

(58)

Pronoun

fongbe

n
-#

There is no overt form in Haitian corresponding to Fongbe -#, which might

lead one to conclude that the Fongbe lexical entry -# has been abandoned and
the Haitian facts constitute an independent development. There is, however,
another way of looking at the Haitian data given the languages historical relationship with its substratum languages. Haitian is similar to Fongbe in that the
sentences being assigned a reflexive interpretation in (45)(47) make use of
strong personal pronouns. In spite of the fact that Fongbe has syntactic clitics,
they are excluded from this construction, as mentioned above. Abstracting away
from the fact that there is no overt form corresponding to Fongbe -# in Haitian, the distribution of Haitian pronouns in (45)(47) parallels that of the Fongbe

164

Pronouns

ones in (56). Suppose, then, that the representation of the Haitian reflexive
phrase is as in Fongbe with the difference that, in Haitian, the head noun of the
construction is phonologically null. This possibility is represented in (59), which
parallels the Fongbe structure in (58).
np

(59)

Pronoun

haitian

In terms of the relexification process, this would mean that the lexical entry
copied from Fongbe -# was assigned a null form at the relabelling stage.8 Recall
that this possibility is allowed (see schema (2) in chapter 2) in cases where the
superstratum language does not offer a phonetic matrix with an appropriate meaning and a suitable distribution to provide a label for the copied lexical entry. The
next question is whether French has a form which could possibly have been used
to relabel a copied lexical entry with the properties of Fongbe -#.
A potential French candidate for this task would perhaps be mme,9 which
occurs with pronouns in some contexts involving a reflexive interpretation. An
example is provided in (60).
(60)

Pierrei est
fier
de lui i - mme.
Peter
aux proud of him mme
Peter is proud of himself.

french

This lexical item is a potential candidate because Haitian actually has a lexical
item mm which is phonologically derived from French mme and because, as is
the case in French, Haitian mm may occur with pronouns, as shown in (61).
(61)

mwen mm
/
myself mm /

ou mm
/ li mm, etc.
yourself mm / himself mm

haitian

This would be a likely hypothesis under two conditions. First, pronoun+mme in

French and pronoun+mm in Haitian should have the same distribution. This is
not the case, however. In the limited contexts where mme is required in French
for a reflexive interpretation to obtain (see (62a) ), mm does not occur in Haitian. As can be seen in (62b), in this case, a pronoun may be assigned a reflexive
interpretation. Tt-li may also be used and my informants consider that, in this
context, pronoun+mm would be a gallicism.
(62)

a. Jeani est en contradiction avec luii

John be in contradiction with him
John is in contradiction with himself.

-mme.
-mme

french

(=(54) in Morin 1977)

li i/j
/ tt
-lii .
b. Jani an kontradiksyon ak
John in contradiction with him / himself / head his
John is in contradiction with him/himself.

haitian

6.6 Reflexives

165

The data in (63) show the same pattern.

(63)

a. Jeani parle avec luii-mme.

John speaks with himself.

french

b. Jani pale ak lii /j / tt -lii .

John speaks with him/himself.

haitian

These data, together with the fact that the -# anaphor occurs in the Fongbe
sentences corresponding to (64b) and (65b), shed doubt on the hypothesis that
French mme would have been used to relabel the lexical entry copied from
-#.
Second, French mme and -# should have properties that are similar enough
to make mme an eligible candidate to relabel a lexical entry copied from -#.
Does mme have the properties of an anaphor? There are competing views on
this matter. On the one hand, Morin (1977: 77) claims that pronoun+mme in
French shares some distributional properties with pronoun+self in English. Like
many grammarians (see e.g. Grevisse 1975), he concludes that pronoun+mme
in French constitutes a compound pronominal reflexive. This analysis does not
account for the fact that the form is obligatory in very specific and limited
contexts such as in (60) above, nor does it not account for the fact that mme is
optional in some contexts and excluded from others, as shown in (64a and b),
respectively.
(64)

a. Je leuri
ai
parl d euxi (-mmes).
1st 3rd pl. aux speak of them mme
I spoke to them about themselves.

french

b. Jeani avait un petit chien avec

John have a
small dog
with
John had a small dog with him.

french

luii
him

(*-mme)

(=(53) in Morin 1977)

On the other hand, Zribi-Hertz (1995) discusses at length the properties that distinguish pronoun+mme in French from pronoun+self in English. She concludes
that, while the latter constitutes an anaphor, the former does not. She argues that
the function of mme-adjunction in (60) is to bring out an unexpected index
(see Edmonson and Plank 1978; Faltz 1985; Levinson 1991), rather than a prominent one (pp. 34950). The analysis that pronoun+mme is not an anaphor is
not in contradiction with the fact that mme is optional in some contexts and
excluded from others (see (62) ). Assuming that pronoun+mme in French does
not have anaphoric status, it contrasts with Gbe pronoun+self, which does (see
Kinyalolo 1994), I conclude that French mme did not have the appropriate
distributional and semantic properties to be used to relabel the lexical entry
copied from Fongbe -#. Since no other French form can be hypothesised to
fulfil this function, it appears that French did not present any appropriate form to
relabel that lexical entry.

166

Pronouns

What, then, is the function of pronoun+mm (see (61) ), in Haitian? It appears

that this expression is used to convey emphasis (see also DeGraff 1992b). As
shown in (65), it may appear after a pronoun which is assigned a reflexive
interpretation.10
(65)

a. Mweni w -m i , mweni -mm.

I
see me
me
mm
I saw myself (emphatic).

haitian

, li
-mm.
b. Jani w lii
John see him
him mm
John saw himself (emphatic).

haitian

The above examples further show that pronoun+mm in Haitian is not in itself a
reflexive expression.
So far, we have seen that Haitian is like Fongbe in having a reflexive anaphor
pronoun+self. In Haitian, however, -self is covert because the superstratum
language did not have any form with the appropriate properties to relabel the
copied substratum lexical entry. Independent evidence for this claim comes from
Berbice Dutch and Gullah whose lexifier languages, Dutch and English, respectively, do have a lexical entry meaning self. According to Robertson (1993: 307),
Berbice Dutch may express reflexivity by means of a pronoun+self. The Berbice
Dutch form corresponding to English -self is -selfu, which probably derives its
phonological form from Dutch -zelv. Similarly, according to Mufwene (1992:
169), Gullah may express reflexivity by means of a pronoun+self. The Gullah
form corresponding to English -self is -self self which obviously derives its
phonological form from English. The above data confirm the claim made in
chapter 2 that relexifiers are constrained by what is available in the superstratum
language when they search for a phonetic form to relabel a lexical entry copied
from their own lexicon. While the creators of Berbice Dutch and Gullah found
appropriate forms in Dutch and English, respectively, to relabel the copied lexical entry meaning -self, the creators of Haitian did not find an appropriate form
in French because French does not have such a form. Consequently, they had to
assign this lexical entry a phonologically null form.
In this view, then, the Haitian lexicon would have a lexical entry corresponding to Fongbe -# with a phonologically null representation, and the ambiguity
between reflexive and pronominal readings in (45)(47) has a straightforward
explanation. A personal pronoun is assigned a reflexive interpretation when it
appears in the specifier position of an np headed by a phonologically null reflexive, as in (66a), and a free interpretation when it appears as the head of an np, as
in (66b).
(66)

a.

np
Pronoun

b.

np

pronoun

6.6 Reflexives

167

This means that Haitian pronouns need not have any special feature to allow the
ambiguous interpretations in (45)(47).
Further support for this analysis comes from the reciprocal interpretation that
may also obtain with pronouns+self in both Haitian and Fongbe. The data in
(67) illustrate this situation for Haitian (see also Koopman 1986; Dchaine and
Manfredi 1994).
(67)

Yo w yo.
they see they
They saw each other/themselves.

haitian

The data in (68) illustrate the same phenomenon in Fongbe (see also Kinyalolo
1994, for further discussion of the Fongbe facts, and Awoyale 1986, for similar
semantic interpretation facts drawn from a sample of Kwa languages).
(68)

Y mw y
3.
3pl. see they self
They saw each other/themselves.

fongbe

The Haitian interpretive data in (67) must follow from the fact that Haitian has
a covert form corresponding to -# in Fongbe (and -self in English). In this
respect, both languages contrast with French, which requires the clitic se in this
case, as shown in (69).
(69)

a. *Ils voient eux / eux-mmes

[Lit.: They see them/themselves.]

french

b. Ils se voient.
They see themselves/each other.

french

On the above analysis, no special proviso needs to be made to account for the
interpretation of personal pronouns in Haitian. Once again, Haitian simply follows
the pattern of the Gbe languages discussed above, as predicted by the relexification
hypothesis. The fact that there is no Haitian reflexive pronoun corresponding to
the French soi can also be explained from this perspective. In Gbe languages,
there is no indefinite pronominal reflexive form corresponding to soi. The creators
of Haitian thus did not have any such lexical entry to relexify.
6.6.2

Body-part reflexives

In Haitian, tt + pronoun and k + pronoun may be assigned a reflexive interpretation, as shown in (70) and in (71), respectively.
(70)

a. Mi ap
touye
I
def-fut kill
I will kill myself.

tt
mwni .
head me

lii .
b. Jani touye tt
John kill
head him
John killed himself.

haitian

haitian
(=(36) in Brousseau 1995a)

168
(71)

Pronouns
a. Jani blayi k
lii
at.
John lay
body him on the floor
John lay down on the ground.

haitian

lii .
b. Jani chape k
John escape body him
John escaped.

haitian
(=(39) in Brousseau 1995a)

With the exception of Carden and Stewart (1988), there is a consensus in the
literature that tt + pronoun and k + pronoun are not grammatical reflexives (see
e.g. Goodman 1964; Dchaine and Manfredi 1994; Kinyalolo 1994; Brousseau
1995a). A major piece of evidence supporting this claim is that neither expression must obligatorily be assigned a reflexive interpretation. With some verbs,
both expressions may be assigned a literal interpretation. For example, with the
verb blese hurt, tt + pronoun may be assigned either a reflexive or a free interpretation, as shown in (72).
(72)

a. Mwen blese tt mwen.

I hurt myself. or I hurt my head.

haitian

b. Ou blese tt ou.
You hurt yourself. or You hurt your head.

haitian

c. Li blese tt li.
He hurt himself. or He hurt his head.

haitian
(=(1) in Carden and Stewart 1988)

Similarly, in some contexts k + pronoun may be interpreted literally, as shown

in (73).
(73)

a. . . . tout
k
m f
whole body me make
My whole body hurts.

m
me

mal.
hurt

haitian

(from Valdman et al. 1981: 290)

The various authors cited above agree that, in the syntax, body-part reflexives
have the structure of a possessive construction, as in (74), where body is the
head noun and the pronoun is part of the genitive phrase. Recall from chapter 4
that the Haitian Genitive Case is phonologically covert.
(74)

[ NP body [pronoun

gen] ]

Where does the possibility of assigning a reflexive interpretation to body-part

expressions like those shown above come from?
The phonological representation of Haitian tt is derived from French tte
head and that of k from French corps body. In French, these two words
cannot be assigned a reflexive interpretation. As has been pointed out by Sylvain
(1936: 65), Goodman (1964: 57), Carden and Stewart (1988: 11) and Muysken
and Smith (1995: 276), among others, there are instances in Old French of the
word corps body being used as a reflexive (e.g. Por lor cors deporter to
amuse themselves, see Muysken and Smith 1995: 276), but there is no evidence

6.6 Reflexives

169

showing that this use of the French word corps was available to the creators of
Haitian. In fact, grammarians report that the reflexive use of corps had disappeared by the mid-sixteenth century (see e.g. Brunot 1905). Therefore, it is
unlikely that the creators of Haitian were exposed to a variety of French which
used body-part reflexives, as is advocated by Chaudenson (1973). In view of this
situation, let us consider the Fongbe data.
Both Kinyalolo (1994) and Brousseau (1995a) report that in Fongbe the word
meaning head does not participate in constructions of the type in (74), and that,
furthermore, this word is never assigned a reflexive interpretation. They therefore conclude that the reflexive interpretation of tt + pronoun in Haitian cannot
be traced to Fongbe. Furthermore, they also show that the Fongbe word meaning
body w does not participate in such constructions either, although it may combine with certain verbs to form complex predicates. Brousseau (1995a) documents
the fact that the meaning of the complex predicate is different from the combined
meaning of the individual lexical items which participate in it. For example, the
combination of the verb gbl spoil with the noun w body forms a complex
predicate meaning hurt. Similarly, the verb y be quick combined with w
forms a complex predicate meaning hurry, the verb lY bathe together with w
forms a complex predicate meaning wash (oneself ), and the verb kl scratch
something with w means scratch (oneself ). As is pointed out by both Kinyalolo
(1994) and Brousseau (1995a), in these complex predicates, w cannot be analysed
as a reflexive. First, a complex predicate containing the noun w body may
take a direct object which is not coreferential with the subject, as in (75), which
shows that w does not have a reflexive meaning.
(75)

gbl -w By.
1sg. spoil body Bayi
He hurt Bayi.

fongbe
(=(46b) in Brousseau 1995a)

Second, when a reflexive meaning is intended for a complex predicate containing

w, reflexivisation is achieved by using the -# anaphor discussed above, as is
shown in (76). According to Brousseau (1995a), -# is optional in this context.
(76)

Kwk gbl w

Koku spoil body him

Koku hurt himself.

-3.
self

fongbe
(=(68a) in Kinyalolo 1994)

If w were a reflexive, we would not expect it to co-occur with -#, and yet it
does. Thus, like Kinyalolo and Brousseau, we have to conclude that Fongbe w
does not in itself induce a reflexive meaning. Consequently, the reflexive usage
of the Haitian possessive expression involving k body cannot be attributed to
Fongbe either.11
Since the reflexive use of k and tt in Haitian cannot be traced to Gbe
languages, as we saw above, could it be traced to other substratum languages?
As is extensively documented in Awoyale (1986), body-part reflexives of the

170

Pronouns

type we find in Haitian (see (70)(71) ) constitute a widespread phenomenon in

Kwa languages. For example, in Yoruba, the word for body is ara; this word
may appear with personal pronouns in possessive constructions of the type in
(74), which are assigned a reflexive interpretation (e.g. ara mi body + 1sg.
pronoun = myself). Awoyale provides similar examples from Igbo, Urhobo,
Ebira and Bassa-Nge. Similar data in Efik and Akan are reported on by Faltz
(1985). Sylvain (1936) reports a Wolof reflexive formed on bob head and Faltz
(1985) mentions a Fula reflexive formed on hooremum (lit.: his head) himself.
Recall from chapter 2 that relexification is an individual activity. Let us assume,
then, that speakers of the above-mentioned languages relexified the body-part
reflexives of their own lexicons and relabelled them on the basis of the French
phonetic matrices corps body and tte head, respectively. In this view, bodypart reflexives would have come into the creole from the substratum languages
that had such reflexives. According to the relexification hypothesis, there is thus
no need to appeal to independent development of the Haitian body-part forms,
as is claimed by Carden and Stewart (1988: 32). According to the view advocated in this book, the argument for independent development would necessitate
a creole lexical entry which had no corresponding form in any of the substratum
languages. This is obviously not the case with body-part reflexives in Haitian.
6.6.3

The multiple substratum sources of the Haitian reflexive forms

The reflexive forms of Haitian constitute a clear example of how relexification

proceeds when various substratum lexicons offer different forms for encoding a
particular notion. The relexification hypothesis thus provides a straightforward
account of the fact that we find several reflexive forms in Haitian creole (and
potentially in other creoles). These forms reflect the differences among the substratum lexicons. This suggests that, in the early creole, there were different
Haitian dialects reflecting the differences among the substratum languages. Hence,
speakers who had relexified Gbe lexicons would use pronominal forms and a
phonologically null anaphor. Speakers who had relexified their lexicons based
on other languages such as those mentioned above would use body-part reflexives.
In communicating with each other, speakers of the early Haitian dialects would
presumably learn the forms that were originally foreign to them, with the result
that it is not rare to find speakers of modern Haitian who use all three reflexive
forms.
The availability of several forms to encode the same notion constitutes an
ideal situation for dialect levelling to operate. Recall from chapter 2 that the
process of dialect levelling refers to the reduction of variation between dialects
in situations where these dialects are brought together. Hence, it is likely that
dialect levelling would operate to reduce the number of forms encoding reflexivity in the various Haitian communities. Evidence that this is so is provided by
Carden and Stewart (1988), who show that dialect differentiation is now taking
place between the North and the South with respect to the lexical entries encoding

6.7 Wh-phrases and Wh-words

171

reflexivity in Haitian. According to them, the beginning of this process goes back
to 1790. This suggests that, while some cases of dialect levelling are completed in
the early stages of the creole (see e.g. the case of the plural marker in chapter 4),
other cases take much longer to be resolved.
Finally, the data discussed in this section are interesting from another point
of view. Singler (1988: 29) advocates the view that when universals, substrate,
and lexifier converge with regard to a given phenomenon, such a phenomenon is
more likely to enter the creole grammar than when the sources compete. The
data discussed in this section constitute a clear case of the sources competing in
different ways. First, as was extensively documented above, the superstratum
and substratum languages do not have much in common in terms of how they
encode reflexivity. Second, the substratum languages divide into at least two
major groups with respect to this area of the lexicon. Nonetheless, the idiosyncratic properties of the substratum languages have made their way into the
creole, as was argued above, showing that creole languages cannot just be a
product of universals.
6.7

Wh-phrases and Wh-words

Wh-expressions may be generated by syntactic or morphological rules. In the

first case, the Wh-phrase is headed by a noun that is modified by a Wh-adjective.
In the second case, the Wh-phrase is realised as a Wh-word. The two options are
illustrated in (77) and (78) using French data (from Brousseau 1995b).
(77)

a. Quel
collgue astu
rencontr?
which colleague aux you meet
Which colleague did you meet?
b. De quelle faon
astu
rpar
of which manner aux you repair
How did you repair the car?

(78)

french

la voiture?
the car

a. Qui astu
rencontr?
who aux you meet
Who did you meet?
b. Comment astu
rpar
how
aux you repair
How did you repair the car?

french

french

la
the

voiture?
car

french

As illustrated above, individual languages may offer both options. Thus, any
account of the historical derivation of Wh-expressions in creole languages must
consider whether particular Wh-expressions are generated by syntactic or morphological rules (see Muysken and Smith 1990, for a discussion of this general
issue). A simple test developed by Brousseau (1995a) permits one to distinguish
between the two types of Wh-expressions. When a Wh-expression is generated
by syntactic rules, it is possible to insert a qualitative adjective between the Whelement and the noun, as in (79).

172
(79)

Pronouns
a. Quel
distingu
collgue astu
rencontr?
which distinguished colleague aux you meet
Which distinguished colleague did you meet?

french

b. De quelle ingnieuse faon

astu
rpar
of which ingenious manner aux you repair
la
voiture?
the car
In what ingenious manner did you repair the car?

french

If the Wh-expression is a word, it constitutes an atomic unit and cannot be

broken up. This is the case with qui who and comment how in (78). All the
Wh-expressions in the three languages under comparison have been submitted to
this test (see Brousseau 1995a).
On the basis of this test, the Wh-expressions in Haitian creole can be shown
to divide into two groups, generated by either syntactic or morphological rules.
For example, the Wh-expression ki mounn (<ki which and mounn person)
can be broken up by an adjective, as is shown in (80). This argues that ki mounn
is a Wh-phrase made up of two words: a Wh-adjective ki and a noun mounn
person.
(80)

Ki
(bl) mounn ou
rankontre?
which nice person you meet
Which (nice) person did you meet?

haitian
(=(3) in Brousseau 1995a)

It is not possible, however, to break up ki-sa (<ki which and sa thing), as

shown in (81).
(81)

a. Ki-sa ou
achte?
what you buy
What did you buy?

b. *Ki
which

bl sa
ou
nice thing you

achte
buy

haitian

(=(4) in Brousseau 1995a)

The sentence in (81b) is ungrammatical in spite of the fact that it is possible

to use the deictic form sa with an adjective in contexts such as Ou achte bl
sa a You bought this/that nice one. This argues that the form ki-sa what is
a complex word which is morphologically derived. It is made up of the Whmorphological element ki and the deictic form sa (discussed in chapter 4).
On the basis of this test, Brousseau (1995a) shows that the inventory of Haitian
Wh-words reduces to the four forms listed in (82) (phonetic variants are not considered here). Of these four forms, the first two contain the Wh-morphological
element ki-. The other two forms are unanalysable simplexes.12
(82)

ki-ls
(ki-)sa
kouman
konbyen

which one
what
how
how much, how many

haitian

(=(5) in Brousseau 1995a)

6.7 Wh-phrases and Wh-words

173

As is shown in (82), the form ki-sa what may simply surface as sa as in Sa ou

f? What did you do? (see Valdman et al. 1981; Koopman 1982b; Lefebvre
1986; Brousseau 1995a). A specific property of the form ki-ls which one is
that it occurs with the plural marker yo when a plural meaning is intended.
(83)

Ki-ls
yo ou
achte?
which-one pl you buy
Which ones did you buy?

haitian
(=(6) in Brousseau 1995a)

The form konbyen how much/how many may occur as the sole element of the
Wh-phrase, as in (84a), or it may co-occur with a noun, as in (84b).
(84)

a. Konbyen
ou
achte?
how-much / many you buy
How much/many did you buy?
b. Konbyen
pwason ou
how-much / many fish
you
How much fish did you buy? or
How many fishes did you buy?

haitian

achte?
buy

haitian

(=(7) in Brousseau 1995a)

The Wh-expressions in (85a) below all allow an adjective to occur between

the ki and the head noun of the construction (see Brousseau 1995a) (e.g. ki bl
bagay which nice thing; ki bl kote which nice place; ki bn jan which good
manner; ki bl kalite which nice kind; ki bn l which good time). Hence,
these Wh-expressions are syntactic phrases made up of the Wh-adjective ki and
a noun. Finally, the Haitian expression meaning why in (85b) is made up of the
preposition pou for and the word ki-sa what. In this case, ki-sa may simply
be realised as ki as in Pou ki ou f sa? Why did you do that? (see Lefebvre
1986; Brousseau 1995a).13 The inventory of syntactically derived Haitian Whexpressions is given in (85) (see Koopman 1982b; Lefebvre 1986; Brousseau
1995a).
(85)

a. ki mounn
which person/who
ki bagay
which thing/what
(ki) kote / ki b which place/where
ki jan
which manner/how
ki kalite
which kind/how
ki l
which time, moment/when14

haitian

b. pou ki(-sa)

haitian

for what/why

Why do Haitian Wh-expressions divide up as they do between Wh-words, as

in (82), and syntactically derived Wh-expressions, as in (85)? Let us first consider the data from the superstratum language.
Like Haitian, French has both Wh-words and syntactically derived Wh-phrases.
The inventory of French Wh-words used in questions is given in (86). These
forms can be found in any dictionary of modern French. They are all attested in
seventeenth-century French as well (see e.g. Haase 1965; Furetire 1984).

174
(86)

Pronouns
lequel / laquelle / lesquels / lesquelles
qui
que / quoi
o
quand
comment
combien
pourquoi

which one(s)
who
what
where
when
how
how much/how many
why

french

The first series of French Wh-words is marked for gender and number: lequel
which one is masculine and singular; laquelle which one is feminine and
singular; lesquel(le)s which ones is plural. The other forms of the paradigm are
invariant in gender and number. The selection between the forms que and quoi,
which both mean what, is determined by the syntactic context in which the
Wh-word occurs. The distribution of que/quoi what in French is extensively
discussed in Goldsmith (1978), Hirschbhler (1978), Koopman (1982c), Lefebvre
(1981, 1982d), Obenauer (1976, 1977). The form quoi occurs in situ, in focus
position, or as the complement of a preposition as shown in (87).
(87)

a. Tu fais quoi?
you do what
What are you doing?

b. Cest quoi qu il fait?

it-is
what that he do
What is it that he is doing?

french

c. En quoi est- il
bon? d. quoi arrive-t-il?
in what be 3rd good
at what arrive 3rd
[Lit.: In what is he good?]
What is he getting at?
What is he good at?
(=(2) in Lefebvre 1982d)

The form que what in (88) is the head of cp.

(88)

Que fait Jean?

what do John
What is John doing?

french

Finally, pourquoi why is a complex word made up of the preposition pour

for and the Wh-word quoi what. The status of pourquoi as a lexical entry is
attested in the literature of French as far back as the eleventh century (see Rey
1992; Catach 1995). The fact that it can be used as a noun (e.g. le pourquoi [Lit.:
the why], see Furetire 1984), argues for its status as a word.
Setting aside the morphological variants of a single form, the inventory of
French Wh-words in (86) comprises eight lexical entries. In contrast, Haitian has
only four Wh-words (see (82) ). Consequently, in several cases where French
has a Wh-word, Haitian has recourse to a syntactically derived Wh-phrase. For
example, where French has the lexeme o where, Haitian has the syntactic
phrase ki kote which place.
Like Haitian, French also has Wh-phrases made up of a Wh-adjective and a
noun. These are listed in (89a). Again, the fact that an adjective may intervene between the Wh-adjective and the noun in (89a) argues that these Wh-expressions
are syntactically derived (e.g. quelle belle personne which nice person). The

6.7 Wh-phrases and Wh-words

175

Wh-expression in (89b) is made up of the preposition pour for and a Whphrase meaning which reason.
(89)

a. quelle personne
?quelle chose
(de) quel ct / bord
(de) quelle manire
quelle sorte (de)
quel moment

which
which
which
which
which
which

person
thing15
side
manner
kind
time

b. (pour) quelle raison

(for) which reason

french

french

The structure of the French Wh-phrases in (89a) parallels that of the Haitian Whphrases in (85a). In both languages, the Wh-phrase consists of a Wh-adjective
meaning which, quel and ki, respectively, and a noun. The structure of the
French Wh-expression in (89b), however, is not parallel to that of the Haitian
one in (85b). While French has an expression meaning for what reason, Haitian
has an expression which literally means for what.
Let us now compare the Haitian forms with their closest semantic equivalents
in French, following the methodology adopted in Brousseau (1995a).
(90)

haitian
form

gloss of
haitian form

ki-ls

which one

(ki-) sa
kouman
konbyen

what
how
how many/much

ki
ki
ki
ki
ki

which person/who
which thing/what
which place/where

mounn
bagay
kote /
b
jan

ki kalite
ki l
pou ki
(-sa)

semantically
closest french
form
lequel, laquelle,
lesquel(le)s
que / quoi
comment
combien

quelle personne / qui

?quelle chose
quel ct / quel
bord / o
which manner/ how quelle manire /
comment
which kind/how
quelle sorte
which time/when
quel moment / quand

for what/why

pour quelle raison /

pourquoi

gloss of
french form
which one(s)
what
how
how many/much
which person/who
which thing
which side/where
which manner/
how
which kind
which time/when
for what reason/
why

The data in (90) reveal two striking facts about the Haitian Wh-expressions.
First, in most cases, the semantically closest French form to a Haitian form is
not phonologically similar to it. For example, Haitian ki-sa what is not phonologically similar to French que/quoi. Similarly, Haitian mounn person is not
phonologically similar to French personne person, and so on. Second, while in
most cases French has both a Wh-word and a syntactically derived Wh-expression
for a given semantic notion, Haitian generally offers only a syntactically derived
Wh-expression. For example, in French, the notion who may be rendered by
either quelle personne or qui. In Haitian, it can only be expressed by ki mounn.

176

Pronouns

These two sets of facts are quite striking if one assumes that the Haitian Whexpressions should parallel those of French (see e.g. Chaudenson 1996).
Furthermore, consider the French expressions hypothesised to have been the
source of the phonological representations of the Haitian Wh-expressions. These
are listed in (91).
(91)
a. Wh-words:

haitian

french expression

ki-ls

(le)quel est-ce
[keles]
*quel + a
comment
[kum]
combien
[kpby9]

(ki-) sa
kouman
konbyen
b. Wh-phrases:

gloss of the
french
expression
which one is it
which this
how
how many/much

ki mounn
ki bagay

quel monde
quel bagage

ki
ki
ki
ki

quel ct / quel bord

quel genre
quelle qualit
quelle heure

which
which
thing
which
which
which
which

*pour quel + a

for which this

kote / ki b
jan
kalite
l

pou ki (-sa)

people
luggage/
side
type/style
quality
hour

The French forms hypothesised to have been the source of the phonological
representations of the Haitian forms do not necessarily have the same meaning
as the Haitian forms. For example, monde does not mean person in French but
people. Quelle heure in French literally means which hour/time; it does not
mean when as it does in Haitian. Similarly, quelle qualit which quality does
not mean how as it does in Haitian, and so on. The Haitian form ki-ls which
one derives its phonological representation from (le)quel est-ce which one is it
where est-ce means is it (see Brousseau 1995a). Furthermore, *quel a and
*pour quel a, corresponding to Haitian ki-sa and pou ki-sa, respectively, are
not grammatical expressions in French. It thus appears that the French phonetic
matrices from which the phonological representations of the Haitian lexical
entries are derived did not contribute their semantics.
The Haitian forms also differ from the phonetically closest French forms
in their syntactic features. For example, the Haitian form ki-ls which one is
invariant in number and gender while the corresponding French form is specified for number and gender (see (86) ). The distribution of the Wh-expressions in
Haitian and French is also quite different. In Haitian, the Wh-expressions listed
in (82) and (85) only occur in questions (see Koopman 1982b). They do not
occur in relative clauses (see Koopman 1982a). By contrast, in French, most of
the Wh-words occurring in questions can also be used as relative pronouns.
Furthermore, in Haitian, questions are typically cleft constructions and the only

6.7 Wh-phrases and Wh-words

177

position for the Wh-expressions is the clefted constituent (see Koopman 1982b).
Wh in situ is not allowed in Haitian (see Koopman 1982b). On the other hand,
French (both standard and popular) allows for Wh in situ (see Goldsmith 1978;
Hirschbhler 1978; Koopman 1982c; Lefebvre 1982d; Obenauer 1976, 1977).16
Furthermore, French questions may be rendered by cleft constructions but they do
not have to be (see Lefebvre 1981). Given these discrepancies, it appears that French
did not contribute the syntactic properties of the Haitian Wh-expressions either.
Why does the inventory of Haitian Wh-expressions divide up as it does
between Wh-words and Wh-phrases (see (82) and (85) )? Why does the semantics
of these forms not correspond exactly to that of the French phonetic strings from
which they are derived? Why did the creators of Haitian not simply adopt the
inventory of French Wh-words in (86)? Again, a comparison with Fongbe provides a clear answer to these questions.
On the basis of the test discussed at the beginning of this section, the Whexpressions in Fongbe can be shown to be either words or syntactic phrases (see
Brousseau 1995a). These two types of Wh-expressions are shown in (92), which
combines the data in Anonymous (1983), Lefebvre (1986), Brousseau (1995a)
and further data that I collected with Fongbe informants.
(92)

fongbe
a. Wh-words:

3-tx
()-tx / n
n6gbwn
nb

b. Wh-phrases:

my tx
nl tx
f (tx)
lw tx
lwkp tx
hwn tx
()tx (t) /
n (t)

literal glosses
(from Segurola 1963)
one-which
that-which

person which
thing which
place which
manner which
kind which
moment / time which
what cause

meaning
which one
what
how
how much/many
who
what
where
how
what kind
when
why

The above paradigms require the following clarifications. The form n what,
reported in Anonymous (1983) and in Segurola (1963), was never produced
spontaneously by any of the informants we worked with. The form n^gbWn
appears to be a frozen lexicalised form meaning how, originally derived from
n^, which, according to Anonymous (1983), is a contraction of nL tX [Lit.: thingwhich] what, and gbWn which, according to Segurola (1963), is a postposition
meaning through. When informants are questioned as to how one says how in
Fongbe, the form that is produced in isolation is n^gbWn, and nothing can intervene between the two morphemes, a fact which argues in favour of its analysis
as a frozen lexicalised form. The Wh-phrase mY tX [Lit.: person which] can be
contracted, as m^. As is pointed out in Anonymous (1983), this contraction is
visible in the modification of the tone on the vowel of the word meaning person:

178

Pronouns

/mY tX/ /m^/. The Wh-phrase nL tX [Lit.: thing+which] constitutes a widespread format for questioning the object in West African languages (see Koopman
1986). According to Anonymous (1983: VI, 6), this expression can be contracted,
yielding n^ as in N^ #W? What did you say?. Our informants, however, reject
this contracted form as it occurs in the aforementioned example. They accept it
only in combination with gbWn as in n^gbWn how in (92a) and in Wh-phrases
like the following, where gbWn is a verb meaning to happen (see Segurola 1963):
N^ wY k gbWn bW w [Lit.: what-is-it adv happen that you come] What
happened that you came? or How come you came? (sometimes translated as
Why did you come? in grammars and dictionaries, see e.g. Anonymous 1983).
Finally, hwn tX when may also be realised as hwtXn where tX is incorporated between hw sun and n mouth, which form a compound meaning
time, moment. In this case, informants judgments coincide entirely with the
data in Anonymous (1983: VI, 6) and Rassinoux (1987: 297).
Fongbe thus has five Wh-words, two of which are made up of a noun/
pronoun and the Wh-affix -tX. It has two words meaning what: (-)tX and n.
Brousseau (1995a) points out that the two forms do not seem to have any distinguishing semantic or syntactic properties except for the fact that n is less
acceptable than (-)tX in echo questions.17 The other two Wh-words do not contain the Wh-affix -tX. The remaining Wh-expressions are syntactic phrases comprising a noun plus the Wh-adjective tX which. The Wh-phrase meaning why
is made up of the Wh-word meaning what and the postposition (t) cause
(see Anonymous 1983).
The Fongbe Wh-expressions in (92) are compared with the Haitian Whexpressions (from (82) and (85) ) in (93).
(93)
a. Wh-words:

haitian
ki-ls
(ki-)sa
kouman
konbyen

fongbe
3-tx
(-)tx / n
n6gbwn
nb

b. Wh-phrases: ki mounn
my tx
ki bagay
nl tx
(ki) kote / ki b f (tx)
ki jan
lwtx
ki kalite
lwkp tx
ki l
hwn tx
pu ki(-sa)
()tx (t) /
n (t)

which one
what
how
how many/much
which person/who
which thing/what
which place/where
which manner/how
which kind/how
which moment/time/when
what, cause/why

Haitian and Fongbe have inventories of only four and five Wh-words, respectively. Except for n what, which has no corresponding form in Haitian, each
of the Fongbe forms has a Haitian equivalent. Furthermore, the morphological
makeup of these forms is similar in the two languages. The Fongbe forms which
include the Wh-affix -tX correspond to the Haitian forms with the Wh-affix ki-.
In both languages, the other two Wh-words do not contain this Wh-morpheme.

6.7 Wh-phrases and Wh-words

179

Like its Haitian counterpart, the Fongbe form #-tX which one is unmarked for
gender and requires the plural marker lZ when a plural meaning is intended. The
Fongbe data in (94) correspond to the Haitian data in (83).
(94)

3D-tx

lz
xw?
which one pl you buy
Which ones did you buy?

fongbe
(=(11) in Brousseau 1995a)

Like the Haitian form konbyen how much/many, the Fongbe form nb how
much/many may be the sole lexical element in its projection, or it may occur
with a noun as in (95), which parallels the Haitian data in (84).
(95)

a. Nb

xw?
how-much / many you buy
How much/many did you buy?

fongbe

b. Hwv nb

xw?
fish
how-much / many you buy
How many fishes did you buy?
or How much fish did you buy?

fongbe

(=(12) in Brousseau 1995a)

In both languages, as well, all the other Wh-expressions are syntactic expressions involving the Wh-adjectives ki and tX, respectively, plus a noun. Furthermore, in both languages, the Wh-expressions have the same meaning. Finally,
the forms ()tX (t) and pou ki-(sa) why parallel each other in an interesting
way: both involve a Wh-word meaning what and a lexical element of the category P: the postposition (t) in Fongbe meaning cause and the preposition
pou in Haitian meaning for.
Thus, the reason why the Haitian Wh-expressions divide up as they do
between Wh-words and Wh-phrases appears to be explained by the relexification
hypothesis. Haitian has Wh-words which correspond to lexical entries in the
substratum lexicon. This argues that the creators of Haitian copied their own
lexical entries and relabelled them using French phonetic matrices. Relexifiers
who spoke a language like Fongbe did not look in the superstratum language for
forms that they did not have in their own lexicon. This explains why the French
simplexes qui who, que what, o where, quand when and pourquoi why
did not enter Haitian. The relexification hypothesis also explains why the morphological makeup of the Haitian Wh-words is so similar to that of the corresponding Fongbe words (see (93) ). Finally, Haitian has Wh-phrases exactly where the
substratum language has Wh-phrases, showing that the creators of Haitian used
their own grammar and lexicon in creating the creole.
Furthermore, the syntactic properties of the Haitian Wh-expressions follow
the Fongbe pattern. Indeed, the distribution of Wh-expressions in the two languages is exactly the same. As is the case in Haitian, Wh-expressions in Fongbe
only occur in questions, and not in relative clauses. Also as in Haitian, questions
in Fongbe are typically cleft constructions where the only possible position for

180

Pronouns

the Wh-expressions is the clefted constituent. Wh in situ is not allowed in Fongbe

any more than it is in Haitian (see Koopman 1982b; Ndayiragije 1993) except
with an echo interpretation. Moreover, as will be shown in chapter 7, in both
Haitian and Fongbe, but not in French, extraction out of subject position requires
a resumptive pronoun in the extraction site. It thus appears that Fongbe has also
contributed the syntactic properties of the Haitian Wh-expressions, a fact which
is predicted by the relexification hypothesis.
In spite of these similarities, the Haitian and Fongbe Wh-expressions differ
with regard to the relative positions of the Wh-element and the head of the
construction in which it appears. While ki precedes the head of the construction
in Haitian, tX follows it in Fongbe. This discrepancy, however, is exactly what
is expected under the general hypothesis of creole genesis formulated earlier.
Recall from chapter 2 that, for lexical categories, the word order of the creole
is predicted to follow that of the superstratum language. In French, the Whadjective quel which precedes the noun. The position of Haitian ki follows the
French pattern. Likewise, while the Fongbe Wh-phrase meaning why makes
use of the postposition (t), the corresponding Haitian expression makes use of
a preposition, in accordance with the French word order.
The three-way comparison presented above shows that the semantic and
syntactic properties of the Haitian Wh-expressions are derived from those of
the corresponding lexical entries in the substratum language. Their phonological
representations are derived from phonetic strings found in the superstratum language. This division of properties is exactly what is predicted by the relexification
hypothesis.
What is the phonological source of the Haitian Wh-element ki? Brousseau
(1995a) considers two hypotheses. According to the first one, Haitian ki would
derive its phonological representation from the French pronoun qui [ki] who,
phonetically identical to the Haitian form. As was observed earlier, however, the
Haitian data show that the creators of the creole did not identify any French Whwords that did not correspond to lexical entries in their own lexicon, which ki
does not. Consequently, this derivation is unlikely. On the second hypothesis,
Haitian ki derives its phonological representation from the corresponding French
adjective quel(le) which.18 On the basis of the syllabic structure of the Haitian
word ki-ls which one, Brousseau (1995a) argues that the creators of Haitian
resyllabified the French Wh-expressions they were exposed to. In this view, the
French phonetic string [k+l +s] ((le)quel est-ce) would have been reinterpreted
as /ki-l+s/. Similarly, the French phonetic string [k+l-r] (quelle heure which
hour) was assigned the phonological representation /ki l/ which time/when.
The assumption is that the copied lexical entry from Fongbe tX which was
relabelled as ki which. Haitian ki was then used either as an adjective or as an
affix, based on the Fongbe pattern (see (93) ), but following the word/affix order
of the superstratum language.
The historical derivation of the other lexical entries involved in the Haitian
Wh-expressions was as follows. The copied lexical entry corresponding to Fongbe

6.7 Wh-phrases and Wh-words

181

mY person was probably relabelled as mounn person on the basis of the French
word monde people. In the literature, there are two conflicting views regarding
the source of the phonological representation of the Haitian lexical entry for
person. Goodman (1964) and Chaudenson (1973: 347) claim that it is the French
word monde people. Taylor (1963: 408) claims that mounn is phonologically
derived from West African languages, which have words such as mun-thu, mundhu, mun-tu meaning person. Both views may turn out to be right if we look at
the data as a case of phonological conflation la Kihm (1994), as discussed in
chapter 2. The copied lexical entry from nL thing was relabelled as bagay
thing on the basis of French bagage, which in seventeenth-century French was
used to cover a wide range of things such as clothes, furniture, utensil, war
or travel equipment (see Furetire 1984). The lexical entry corresponding to
Fongbe f place was relabelled as kote or b place, side on the basis of the
French words ct or bord side. Fongbe lW manner was relabelled as jan
manner on the basis of French genre style, type. Fongbe lWkp kind,
manner was relabelled as kalite kind, manner on the basis of French qualit
quality. The Fongbe derived noun #-tX which one has a Haitian counterpart
ki-ls which one, which is morphologically derived as discussed earlier. The
copied lexical entry corresponding to Fongbe n^gbWn how was relabelled as
kouman how from French comment how. Finally, the copied lexical entry
corresponding to Fongbe nb how many/much was relabelled as konbyen
how many/much on the basis of French combien how many/much, and that
corresponding to (-)tX what was relabelled as (ki-)sa, where ki corresponds to
tX and sa to . As is pointed out in Brousseau (1995b), both sa and are deictic
forms which can be translated as that. The Haitian form sa takes its phonological representation from the French deictic form a that, as shown in chapter 4.
As for the Fongbe form n what, it was not relexified since Haitian presents
no form corresponding to it. The closest French form which could have been
used to relabel a substratum lexical entry copied from n would have been
quoi what. It is not clear, however, that quoi as used in seventeenth-century
French (see Haase 1965) had the appropriate distribution to make it an eligible
candidate. For example, unlike n, in seventeenth-century sources, quoi does
not appear at the beginning of the clause. In popular French, however, quoi
may occur in clause-initial position (see Lefebvre 1982d). Whether the creators
of Haitian were exposed to these data will remain a matter of conjecture. The
existence of the form ki-sa with the same meaning and built on the same model
as corresponding forms in many substratum lexicons probably led the substratum
speakers to abandon this lexical entry when they were relexifying their own
lexicon.
Finally, the derivation of the Haitian lexical entry l presents the following
particularity. This lexical entry takes its phonological representation from French
lheure which means the hour. According to Valdman et al. (1981), however,
the Haitian noun l means hour, time, clock and watch. The latter meaning of
the Haitian noun l is exemplified in (96).

182
(96)

Pronouns
L
m lan rete.
watch me det stop
My watch has stopped.

haitian
(from Valdman et al. 1981)

In Fongbe, the noun gn shares with the Haitian noun l the meanings hour,
clock, and watch (see Segurola 1963). This argues that Fongbe gn was relexified
by l on the basis of French lheure. Interestingly enough, however, while Haitian l also means time/moment, Fongbe gn does not. In Fongbe, this notion
is rendered by hwn time, moment. I can think of two scenarios to explain
this situation. In the first one, hwn was not relexified and the semantics of l,
relexified from gn, was extended so as to cover that of hwn. This could not
be attributable to the influence of French since French lheure means the hour.
A more plausible scenario, however, is the following. Suppose the creators of
Haitian relexified gn on the basis of French lheure, yielding l. Suppose that
they also relexified hwn based on French lheure. This would yield two homophonous Haitian lexical entries, each of which corresponded to one original
lexical entry, as shown in (97).
(97)

fongbe
a. gn

haitian
l
hour, clock, watch

b. hwn

time, moment

In this case, the two early Haitian lexical entries would be homophonous because
they were both relabelled using a single French phonetic string. It is plausible to
hypothesise that, over time, the two homophonous and semantically related lexical entries became a single lexical entry, as witnessed by the meanings of l
in Valdman et al. (1981).
The relexification hypothesis thus accounts straightforwardly for the properties of the Haitian Wh-words and phrases in (82) and (85). While the lexical
items involved derive their phonological representations from French phonetic
matrices, they draw their other properties from the substratum language. This
conclusion runs counter to that in Muysken and Smith (1990) based on a sample
of pidgin and creole languages.
6.8

Conclusion

Logophoric pronouns were not relabelled because they do not have independent
semantic content and they were abandoned by the creators of Haitian. Pronominal
syntactic clitics did not make their way into Haitian; it was hypothesised that, in
this case, this is because of the way relabelling proceeds in the case of functional
category lexical items. Aside from these two exceptions, the pronominal (and
pronominal-like) forms discussed in this chapter illustrate the division of properties which characterises the Haitian lexical entries in general. While the phonological representations of Haitian pronouns are derived from French phonetic

6.8 Conclusion

183

matrices, their semantic and syntactic properties often contrast with those of
French and systematically parallel those of Fongbe. This is exactly what is
expected on the hypothesis that these lexical entries were created through
relexification. The fact that the possessive adjectives/pronouns and the reflexive
strong pronominal form of French did not enter Haitian also follows from the
relexification hypothesis. There were simply no such forms to be relexified
in the original lexicons. The various options offered by the creole to encode
reflexivity were shown to be an interesting reflection of the differences in the
lexical entries used to express reflexivity in the various substratum languages of
Haitian.

184

Functional category lexical entries

Functional category lexical entries

involved in the structure of the
clause

This chapter and the next compare the functional category lexical entries involved
in the structure of the clause in Haitian, French and Fongbe. Complementisers,
resumptives in the context of extracted subjects, the operator that occurs in
relative and factive clauses, and clause conjunction are discussed in sections 7.1
to 7.4. The lexical entry se, difficult to gloss, is discussed in section 7.5 and
negation markers in section 7.6. Yesno question markers are discussed in section 7.7. Finally, the markers expressing the speakers point of view with respect
to the proposition are discussed in section 7.8. Clause structure and the central
role of the determiner therein will be addressed in chapter 8. The data discussed
in this chapter concisely illustrate how the processes hypothesised to play a role
in creole genesis apply in the case of functional categories (see chapter 2).
7.1

Complementisers and complementiser-like forms

In this section, I examine the properties of complementisers and complementiserlike forms in Haitian, French and Fongbe. Forms introducing the tensed complements of verbs of the say- and want-classes will be discussed in turn.1 Tenseless
complements will be discussed in chapter 9.
7.1.1

Forms introducing the tensed complements of verbs of the say-class

In Haitian creole, the complementiser introducing sentential complements of

verbs such as kw believe, di say, panse think, etc., is phonologically null,
as shown in (1).
(1)

Jan
kw / di / panse
[CP Mari vini].
John believe / say / think comp Mary come
John believed/said/thought that Mary came.

haitian
(=(54) in Lefebvre 1993b)

Sterlin (1988, 1989) convincingly argues that a complementiser is involved in

the sentential complements of the verbs in (1). First, she shows that the embedded subject is not in the binding domain of the main verb. This is evidenced by
the fact that the embedded subject is referentially free, as shown in (2).
184

7.1 Complementisers and complementiser-like forms

(2)

Lii kw / panse
lii /j ref.
he believe / think he cure
Hei believes/thinks that hei /j is cured.

185

haitian
(=(25) in Sterlin 1988)

If the embedded subject pronoun were in the binding domain of the main verb,
it would have to be referentially disjoint from the subject of that verb. Since the
embedded subject pronoun is referentially free, it follows that there must be a
null complementiser introducing the complement clause and creating a separate binding domain.2 The fact that the embedded pronoun cannot be assigned a
reflexive interpretation (see chapter 6, section 6) further supports Sterlins proposal. Sterlin (1988) also shows that the subject of the complement clause may
be extracted, as illustrated below.
(3)

Ki
mounni li kw / panse
kii ref?
which person he believe / think res cure
Who does he believe/think is cured?

haitian
(=(26) in Sterlin 1988)

As can be seen in (3), the form ki appears when the subject has been extracted.
As will be argued in section 7.2, in (3), ki is in the embedded subject position and
is marked for the Case feature [+nominative]. If there were no null complementiser
to introduce the embedded clause, ki would be in the Case-assigning domain
of the main verb and, being in Nominative Case, it would not be licit in this
position. Consequently, there must be a null complementiser introducing the
embedded clause.
The Haitian data contrast with French, where the tensed complement of verbs
such as croire believe, dire say, penser think, etc., is introduced by the
[+tense] complementiser que that (see e.g. Kayne 1976; Milner 1978), as shown
below. Note that this que requires that the verb in the embedded clause be
marked for indicative mode.
(4)

Jean croit / dit / pense

[ CP que
Marie est partie].
french
John believe / say / think
comp Mary left
John believes/says/thinks that Mary left.
(=(55) in Lefebvre 1993b)

The fact that Haitian has no overt form corresponding to French que tells us that
the creators of Haitian did not identify que as a [+tense] complementiser,3 which
is consistent with the claim in chapter 2 that they did not have sufficient access
to French to identify the languages functional items as such. If the [+tense] complementiser of Haitian is not like that of French, does it have the properties of
the corresponding complementiser in the substratum languages?
In Fongbe, for example, the tensed complement of verbs of the say-class is
introduced by #W (literally say), as shown in (5).
(5)

Kwk 3 / 3w / ln
[3w By w].
Koku believe / say / think say Bayi come
Koku believed/said/thought that Bayi came.

fongbe
(=(30a) in Kinyalolo 1993b)

186

Functional category lexical entries

There are two analyses of the form #W in (5). On the one hand, Kinyalolo
(1993b) proposes that Fongbe #W is a serial verb which selects a cp complement
headed by a phonologically null complementiser. Hence, on his analysis, the
that-type complementiser of Fongbe is null. On the other hand, Lord (1976)
proposes that, in some Gbe languages including Fongbe, the serial verb meaning
say has been reanalysed as the phonological representation of a previously
phonologically null complementiser introducing complements of verbs of the
say-class. On her analysis then, the form #W introducing a sentential complement
in (5) is a that-type complementiser. The two analyses may be schematised as in
(6a) and (6b), respectively.
(6)

a. Kwk
Koku

3
believe

b. Kwk
Koku

3
[3w
[By w] ].
believe comp Bayi come

[3w [
say comp

[By
Bayi

w] ] ].
come

fongbe
fongbe

These two analyses are not necessarily incompatible as they could be viewed as
reflecting two competing dialects in the synchronic lexicon of Fongbe. In this
view, the representation in (6a) would correspond to a conservative dialect and
that in (6b) to a more innovative dialect, where the serial verb #W say has been
reanalysed as the phonological form of the previously null complementiser.
The Haitian data compare with the Fongbe data as follows. First, unlike the
innovative dialect of Fongbe and like the conservative one, Haitian has a phonologically null complementiser. However, unlike the conservative dialect, it does
not have a serial verb meaning say in the construction under discussion. It thus
appears that the difference between Haitian and the most conservative dialect
of Fongbe resides not in the properties of the complementiser itself but rather
in the availability of such a serial verb in the complement of say-class verbs.
Facts from Saramaccan, a creole language to which Fongbe has been shown
to contribute considerable input (see Smith 1987), provide more evidence for an
analysis along these lines.
According to the description in Byrne (1987), Saramaccan can be claimed to
have a form comparable to the [+tense] complementiser #W in Fongbe (see (6b) ).
This form is ta that, a reduced form of the verb tk say. Ta introduces the
sentential complement of verbs of the say-class, as illustrated in (7).
(7)

A tk ta d muje . . .
he say that the woman . . .
He said that the woman . . .

saramaccan
(=(85b) in Byrne 1987)

The above data suggest that, in Saramaccan, the verb tk say (presumably
initially used as a serial verb in the context of say-class verbs) has been reanalysed
as the phonological representation of an initially null [+tense] complementiser.

7.1 Complementisers and complementiser-like forms

187

In this view, the historical derivation of the Saramaccan complementiser would

parallel that proposed by Lord (1976) for Fongbe #W.
The difference between Saramaccan and Haitian with respect to the lexical
item under discussion in this section can be explained straightforwardly if we
assume that the two languages differed at some point of their evolution with
respect to the status of the serial verb meaning say that introduced the [+tense]
complement of verbs of the say-class. I hypothesise the following scenario.
In the incipient creole, both Saramaccan and Haitian introduce complements
of say-class verbs with a serial verb meaning say and a null complementiser,
following the model of Fongbe in (6a). Presumably the Haitian serial verb
would have been di say (from French dire say). This hypothesis is schematised
in (8).
(8)

Hypothesised early creoles

. . . verb of say-class [serial verb say [null complementiser . . . ]]
tk

saramaccan
di

haitian

I further hypothesise that, in the development of Saramaccan, the verb tk

was reanalysed as the phonological form of the previously null complementiser,
similar to Fongbe #W in (6b). Haitian underwent a different development whereby
the use of the serial verb di, hypothesised to have introduced complements of
verbs of the say-class in the early creole, was abandoned. This prevented it
from being reanalysed as the phonological form of the complementiser. In this
view, modern Saramaccan is like the more innovative dialect of Fongbe (see
(6b) ) and modern Haitian is like the more conservative dialect of Fongbe (see
(6a) ) except that it has lost the use of a serial verb meaning say to introduce
tensed complements of say-class verbs.
Assuming this analysis to be correct, the Haitian complementiser introducing
the tensed complements of verbs of the say-class follows the conservative pattern of the substratum language and not that of the superstratum language: both
have a null complementiser. The data and analysis presented in this section
further illustrate the fact that two creoles which share the same important substratum language may choose different options in the course of their further
development.

7.1.2

Forms introducing the tensed complements of verbs and adjectives of

the want-class

As is extensively discussed in Koopman and Lefebvre (1981, 1982), Sterlin

(1988, 1989) and Lefebvre (1993b), in Haitian creole, complements of verbs of
the want-class and of a small class of adjectives such as good are introduced
by pou, as illustrated below.

188

Functional category lexical entries

(9)

Yo
te
vle
[pou m
te
antre nan troup
haitian
they ant want comp me ant join in
troops
Jakml].
Jacmel
They wanted me to join Jacmels troops.
[Lit.: They wanted that I joined Jacmels troops.]
(=(10) in Koopman and Lefebvre 1982)

(10)

Li bn
[pou m t
a
pati].
it good comp I ant ind-fut leave
It is good for me to leave.
[Lit.: It is good that I leave.]

haitian

(=(65) in Lefebvre 1993b)

Koopman and Lefebvre (1981, 1982) show that the complementiser pou is homophonous with the preposition pou which selects np complements, as in (11), or
purposive clauses as in (12).4
(11)

Pte sa
pou mwen.
bring this for me
Bring this for me.

haitian
(=(4) in Koopman and Lefebvre 1982)

(12)

M te
bezwen skont sa
a
pou m te
haitian
I
ant need
money this det for I ant
repati.
start-again
I needed this money for a new start.
[Lit.: I needed this money so that I could start again.]
(=(6) in Koopman and Lefebvre 1982)

As is shown in Koopman and Lefebvre (1981, 1982), the complementiser pou is

also homophonous with the mood marker of obligation pou discussed in chapter
5, as shown in (13) (=(14b) in chapter 5).
(13)

Mari pou prepare pat.

Mary sub prepare dough
Mary should prepare dough.

haitian

As we saw in chapter 5, the Haitian mood marker pou derives its phonological
form from the French periphrastic expression tre pour as shown in (14) (=(3b)
in chapter 5).
(14)

Jean est pour partir.

John is about to go.

french

The preposition pou derives its phonological form from the French preposition
pour for which selects np complements, as in (15), as well as purposive clausal
complements, as in (16).
(15)

Fais a
pour
do
this for
Do this for me.

moi.
me

french

7.1 Complementisers and complementiser-like forms

(16)

J ai
besoin dargent pour que
je puisse repartir.
I aux need
money
for comp I be-able start-again
I need money in order to start again.

189

french

However, as is pointed out in Lefebvre (1993b), in contrast to Haitian, pour in

French does not introduce complements of verbs of the want-class or adjectives
of the good-class, as is shown by the ungrammaticality of the sentences in (17)
and (18), which correspond to the Haitian sentences in (9) and (10), respectively.
(17)

*Ils voulaient pour jentre dans les troupes de Jacmel

french
They wanted me to join Jacmels troops.
(=(67a) in Lefebvre 1993b)

(18)

*Il est bon pour je parte

It is good for me to leave.

french
(=(67b) in Lefebvre 1993b)

In French, the tensed complements of verbs and adjectives of the class in question are introduced by the complementiser que. This is illustrated in (19) and
(20).
(19)

Ils voulaient que jentre dans les troupes de Jacmel.

They wanted me to join Jacmels troops.

french

(20)

Il est bon
que
je
it is good comp I
It is good that I leave.

french

parte.
leave

(=(69) in Lefebvre 1993b)

Note that the complementiser que in (19) and (20) is distinct from the complementiser que that occurs with verbs of the say-class in French (see section 7.1.1).
As has been extensively discussed in the literature on French syntax, the complementiser selected by verbs of the want-class and adjectives of the good-class
bears a special feature that Kayne (1976) represents as [+F], standing for subjunctive mood. While the que selected by verbs of the say-class requires that
the verb of the embedded clause be in the indicative mood, the que selected
by predicates of the want- and good-classes requires the verb of the embedded
clause to be in the subjunctive mood. (For extensive discussions on the properties of the two French complementisers que, see Goldsmith 1978; Hirschbhler
1978; Kayne 1976; Kayne and Pollock 1978; etc.)
So, while the phonological form of the Haitian preverbal mood marker pou
is derived from the phonetic matrix of French tre pour, and the phonological
form of the Haitian preposition pou is derived from the phonetic matrix of the
French preposition pour, the phonological form of the Haitian complementiser
pou is clearly not derived from the phonetic matrix of the corresponding French
complementiser que [+F]. On the one hand, this shows once again that the creators of Haitian did not identify as such the functional items of French. On the
other hand, the above data lead one to ask about the source of the Haitian complementiser pou.

190

Functional category lexical entries

Koopman and Lefebvre (1981, 1982) hypothesised that the complementiser

pou entered the Haitian lexicon through a process of reanalysis involving both
the preposition pou and the mood marker pou. First, the preposition pou introducing purposive complements, as in (12), was reanalysed as a complementiser.
Next, the mood marker pou was also reanalysed as a complementiser due to movement from its position between the subject and the verb to a clause-initial position. This claim is supported by data showing two surface positions for the mood
marker pou. In the (a) sentences below, pou is in the mood marker position, i.e.
between the subject and the verb. In the (b) sentences, it occurs before the subject,
that is, in the position where we normally find complementisers.
(21)

a. M te
achte liv
[m te
I
ant buy
book I
ant
I bought a book that I had to read.
b. M
I

(22)

te
ant

liv
[poui
book comp

a. Se
vini
[nou te
pou
it-is come we
ant sub
We had to come.
b. Se
it-is

(23)

achte
buy

vini
come

[poui
comp

nou
we

pou li]
sub read

sa
thing

[poui
comp

nou
we

haitian

m te
ti li]
a.
haitian
I ant
read det
(=(33) in Koopman and Lefebvre 1981)
vini].
come

haitian

te ti vini].
haitian
ant come
(=(35) in Koopman and Lefebvre 1981)

a. Ki
sa
[nou te
pou f]?
which thing we
ant sub do
What were we supposed to do?
b. Ki
which

a.
det

te ti
ant

f]?
do

haitian

haitian

On this analysis, the fact that pou cannot occur in both positions within the same
sentence, as shown in (24), constitutes an argument for mouvement of pou from
its basic position (in the (a) sentences) to its derived position (in the (b) sentences).
(24)

*Ki

sa

pou

nou

te pou

haitian

On the above account then, pou became a complementiser through the reanalysis
of both the preposition pou introducing purposive complements and the mood
marker pou as a complementiser. When the Haitian facts are considered in light
of the substratum data, however, the reanalysis account becomes less plausible.
For example, in Fongbe, verbs of the want-class and a small class of adjectives
such as good are introduced by n, as shown in (25) and (26), respectively.
(25)

n jl
n

n
w.
I
want comp 2nd sub come
I want you to come.
[Lit.: I want that you come.]

fongbe

(Anonymous 1983: X,2)

7.1 Complementisers and complementiser-like forms

(26)

nyw
n
n n
y.
it good comp 1st sub leave
It is good for me to leave.
[Lit.: It is good that I leave.]

191

fongbe

(Anonymous 1983: X,2)

The complement of these predicates may also be introduced by n, as shown in

(27) and (28).
(27)

n jl
n

n
w.
I
want comp 2nd sub come
I want you to come.
[Lit.: I want that you come.]

fongbe

(28)

nyw
n
n n
y.
it good comp 1st sub leave
It is good for me to leave.
[Lit.: It is good that I leave.]

fongbe

(Anonymous 1983: X,2)

N and n are mutually interchangeable in this context. According to the Fongbe

speakers with whom I did fieldwork, the selection of one or other of these forms
does not change the meaning of a sentence (compare (25) and (27), and (26) and
(28) ). Interestingly enough, as in Haitian, but not in French, the complementiser
n in (25) and (26) is homophonous with the preposition n for which selects
either np complements, as in (29), or purposive clausal complements, as in (30).
(29)

Kwk n
svn n sb.
Koku give crab for Asiba
Koku gave a crab for Asiba.

fongbe
(=(ib) in Lefebvre 1994c: 98)

(30)

n 3 fngb kplvn
wy
n
m n
fongbe
1st at Fongbe learning part prep 1pl def-fut
sx 3ls.
can have-a-conversation.
I am learning Fongbe so that we can have a conversation.
[Lit.: I am learning Fongbe for we will have a conversation.]
(Anonymous 1983: 1X, 6)

Furthermore, as is the case in Haitian, the complementiser n is homophonous

with the mood marker n discussed in chapter 5, as shown in (31) (=(15b) in
chapter 5).
(31)

Mari n
3f
wh.
Mary sub prepare dough
Mary should prepare dough.

fongbe

To sum up, the Haitian complementiser pou selected by verbs of the wantclass and adjectives of the good-class (see (9) and (10) ) is homophonous with
the preposition pou (see (11) and (12) ) and the preverbal mood marker pou (see
(13) ). Similarly, the Fongbe complementisers n and n selected by the corresponding verbs and adjectives (see (25), (26) and (27), (28) ) are homophonous

192

Functional category lexical entries

with the preposition n (see (29) and (30) ) and the preverbal mood marker n
(see (31) ), respectively. The similarity between the Haitian and Fongbe data is
striking. Both languages differ from French in a similar way. In French, complements of verbs of the want-class and adjectives of the good-class are introduced
by que [+F] (see (19) and (20) ), which is not homophonous with the preposition
pour for and which requires that the verb of the subordinate clause be in the
subjunctive mood. In view of these facts, I propose a relexification account of
the Haitian data.
Given a conception of the lexicon whereby monosemy is preferred over polysemy (see e.g. Nida 1948; Ruhl 1989; Cowper 1989, 1995; Johns 1992; Ghomeshi
and Massam 1994; Bouchard 1995), there would be only one lexical entry each
for Fongbe n and n. The morpheme n would be minimally specified for the
semantic feature [realis] allowing it to occur either as a mood marker (between
the subject and the verb) or as head of cp when selected by predicates of the
want-/good-class. Likewise, n would be underspecified for categorial features
allowing it to occur as the head of either pp or cp (as is the case with English for,
which can head cps as well as pps). By hypothesis, during the relexification
period of Haitian creole genesis, these two substratum lexical entries were
copied and relabelled on the basis of French phonetic strings. The lexical entry
copied from n was minimally specified for the feature [realis] and it was
relabelled on the basis of French pour occurring in tre pour (as we saw before)
yielding Haitian pou. The fact that the creole lexical entry is minimally specified
for [realis] allows it to occur as a mood marker between the subject and the
verb or as head of cps selected by predicates of the want-/good-class, just like
the corresponding lexical entry in the substratum language. Likewise, the lexical
entry copied from n was underspecified for categorial features and was relabelled
on the basis of the French preposition pour, yielding another lexical entry pou in
the creole. Since the creole lexical entry is underspecified for categorial features,
it can occur as head of cp or head of pp, as can the corresponding lexical entry
in the substratum language. The two Haitian lexical entries pou are accidentally
homophonous due to the superstratum forms that they were relabelled from
( pour in both cases). This analysis accounts straightforwardly for the striking
similarity between the properties of the substratum lexical entries and those of
the corresponding ones in the creole.
On the above scenario, only relexification is required to account for the
Haitian data. There is no need for recourse to reanalysis as in the first scenario.
This result may seem controversial at first glance, for it is widely assumed that,
in creole languages, complementisers evolve through the process of reanalysis
(see e.g. Washabaugh 1975; Koopman and Lefebvre 1981; Plag 1993). The reanalysis account of complementisers has been proposed, I believe, to capture the
fact that simple clauses are prominent in the incipient creole and embedded
clauses (and hence, complementisers) appear later in the development of the
creole. The second scenario proposed above, based exclusively on relexification,
accounts for this situation as follows. The Haitian lexical entry relexified from

7.2 Complementisers or resumptives?

193

Fongbe n was probably first manifest as a mood marker in the incipient creole.
Likewise, the Haitian lexical entry relexified from Fongbe n probably first
occurred as head of pp in the incipient creole. When subordinate clauses started
to be used, both lexical entries labelled as pou in the Haitian lexicon were available to head cp since, like their Fongbe counterparts, they were underspecified.
7.2

Complementisers or resumptives in the context of

extracted subjects?

It is a well-documented fact that languages present subject/object asymmetries

and that, for example, they offer different strategies to rescue a sentence whose
subject has been extracted.5 While French has qui, a special form of the complementiser, Fongbe has a resumptive pronoun in subject position. In such contexts, Haitian has ki, obviously phonologically derived from French qui. Does
Haitian ki have the properties of a complementiser or of a resumptive pronoun?
In standard French, the form qui appears in a clause whose subject has been
extracted. This form contrasts with que, which appears when a non-subject argument has been extracted. This contrast is illustrated in (32).
(32)

a. Quii penses-tu *que / qui ti est

who think-2nd comp
aux
Who do you think came?

venu?
come

french

vu ti ?
b. Qui crois-tu
quei / *qui Marie a
who believe-2nd comp
Mary aux see
Who do you believe Mary saw?

french

In the literature on French syntax, it has been argued that, while the first occurrence of qui in the sentences in (32) is a Wh-pronoun, the second is not because
it also shows up in relative clauses whose subject has been relativised, as shown
in (33). Since (33) does not involve a question, qui in this context cannot possibly be an interrogative pronoun (see Moreau 1971).
(33)

L
homme qui
est
det man
comp aux
The man that came.

venu.
come

french

Kayne (1976) argues that qui in (32) is not a relative pronoun either, showing
that genuine relative pronouns do not occur in this context. Compare the (a) and
(b) sentences in (34).
(34)

a.

L
homme que
Jean croit
qui
est
det man
comp John believe comp aux
The man that John believes came.

b. *L
det
a
aux
[Lit.:

venu.
come

homme que
Jean croit
qui
Marie
man
comp John believe to whom Mary
parl
speak
The man that John believes Mary spoke to.]

french

french

194

Functional category lexical entries

Since genuine relative pronouns cannot occur in the position where qui occurs in
(34a), qui cannot be a relative pronoun. On the basis of its distributional properties, it has been proposed that qui is a special form of the [+tense] complementiser
(see e.g. Moreau 1971; Kayne 1976). For one thing, it occurs only in tensed
clauses, as shown by the ungrammaticality of (35).
(35)

*Qui crois-tu
qui
tre arriv?
who believe-2nd comp be
come
[Lit.: Who do you believe to have come?]

french

Second, as shown in (32), qui is in complementary distribution with the [+tense]

complementiser que, which requires the embedded clause to be in the indicative
mood. That is, when the subject has been fronted, qui appears; when a nonsubject has been fronted, que appears. Third, as has been pointed out by Law
(1994b), qui is also in complementary distribution with the [+tense] complementiser que marked for the feature [+F], which requires that the embedded
clause be in the subjunctive mood. As shown in (36), in this context as well, que
introduces the embedded clause when a non-subject has been fronted but qui
introduces it when a subject has been fronted.
(36)

a. Qui veux-tu
que
Jean voie?
who want-2nd comp John see-sub
Who do you want John to see?

french

b. Qui veux-tu
qui
vienne?
who want-2nd comp come-sub
Who do you want to come?

french

So qui has the same distribution as que except that it occurs when a subject
has been extracted and que occurs elsewhere. Thus, que and qui are allomorphs
of the [+tense] complementiser que (see Moreau 1971; Kayne 1976). Qui is
selected in environments where the subject position is empty since, unlike que,
it can properly govern the empty subject position, thus allowing this position to
be extracted out of.
In simple questions involving the subject in standard French, no complementiser appears, as shown in (37).
(37)

a. Qui est
venu?
who aux come
Who came?

b. *Qui qui
who that

est
venu standard french
aux come

Standard analyses have it that (37b) is excluded by some version of the doubly
filled comp filter, whereas (37a) is licit because the Wh-pronoun qui in Spec cp
properly governs the empty subject position. By contrast, in popular French, the
complementiser qui shows up in contexts where the subject has been extracted,
as shown in (38).
(38)

(Cest) qui
qui est
venu?
it-is
who that aux come
Who is it that came?

popular french
(from Lefebvre 1982e)

7.2 Complementisers or resumptives?

195

As is extensively argued in Lefebvre (1982e, 1989), the contrast in grammaticality

between (38) and (37b) is due to the following differences in the structure of
questions in the two dialects of French. In standard French, Wh-pronouns are in
Spec cp at S-structure, whereas, in popular French, questions have the properties of clefts in that by S-structure the Wh-pronoun is outside the cp from which
it has been extracted. In the latter case, the Wh-pronoun cannot properly govern
the empty subject position and the complementiser qui must show up as head
of cp to rescue a structure which would otherwise be illicit by virtue of some
version of the that-trace filter. Since the second qui in (38) has the same distributional properties as the complementiser qui in standard French, there is no
reason to assign it a different analysis. In light of this discussion of the interaction between the French complementiser system and subject/object asymmetries,
I now turn to a discussion of the Haitian data.
In her pioneering study of the syntax of questions in Haitian, Koopman
(1982b) points out the similarity between the structure of questions and of clefts
in this language. Both are introduced by se it-is (see (39) ), both exhibit longdistance movement (see (40) ) and both must contain the form ki when the subject has been moved out of a tensed clause (see (41) ). Note that se is optional
in questions.
(39)

a. Se
Marii Jan
w ti .
it-is Mary John see
It is Mary that John saw.
b. (Se) ki
mounni Jan
it-is which person John
Who is it that John saw?

(40)

haitian

w ti .
see

di
Jan
w ti .
a. Se
Marii Jak
it-is Mary James say John see
It is Mary that James said that John saw.
di
Jan
b. (Se) ki
mounni Jak
it-is which person James say John
Who is it that James said that John saw?

(41)

haitian

a. Se
Jani kii vini.
it-is John ?
come
It is John who came.
b. (Se) ki
mounni kii
it-is which person ?
Who is it that came?

haitian

w ti ?
see

haitian

haitian

vini?
come

haitian

Lumsden (1990) presents four arguments based on the distribution of negation

and tense, the properties of se, and the distribution of ki in question formation,
showing that questions in Haitian are generally cleft constructions and cleft constructions are bi-clausal.
As shown above, Haitian manifests subject/object asymmetries. The sentences in (41) show that, when a subject has been extracted, the morpheme ki

196

Functional category lexical entries

must appear in the clause from which it was extracted. The sentences in (39)
and (40) show that, in contrast, when the direct object has been extracted, this
morpheme does not show up. Law (1994b) provides further evidence showing
that extraction out of other non-subject positions does not trigger the appearance
of ki. Thus, ki appears only in contexts where the subject has been extracted.
Furthermore, while the sentences in (41) show that ki must appear in cases of
local subject extraction, the sentences in (42) show that it must also appear in
cases of long-distance extraction.
(42)

a. Se
Jan
Jak
di
ki w Mari.
it-is John James say ? see Mary
It is John that James said saw Mary.

haitian

b. (Se) ki
mounn, Jak
di
ki w
it-is which person James say ? see
Who is it that James said saw Mary?

Mari?
Mary

haitian

One way or another, all authors (see e.g. Koopman 1982b; Koopman and Lefebvre
1982; Law 1994b) but one (see DeGraff 1992a) link the presence of ki in the
context of subject extraction to that-trace effects, restated as ecp effects in
Chomsky (1981).6
The Haitian form ki most probably derives its phonological representation
from the French form qui that occurs in questions (see (37), (38) ). But the
question at stake here is whether it occurs in the head of cp, as in French, or in
another position. There are two competing views in the literature. On the one
hand, Koopman (1982a, 1982b) argues that ki is in comp. On the other hand,
Lumsden (1990), Manfredi (1993) and Law (1994b) all defend the position that,
at S-structure, ki occurs in the extraction site of the fronted subject. I shall
review both proposals.
Consider the data in (43), from Koopman (1982b).
(43)

a.

Ki
mounn ki pou te
which person ? sub ant
Who was supposed to come?

vini
come

an?
det

haitian

(=(53a) in Koopman 1982b)

b. *Ki
mounn
which person
c.

pou
sub

ki te
vini
? ant come

an
haitian
det
(=(53b) in Koopman 1982b)

Ki
mounn ki pou pa
te
vini
an?
haitian
which person ? sub neg ant come det
Who was not supposed to come?
(adapted from (31a) in Koopman and Lefebvre 1982)

Koopmans analysis that Haitian ki is in comp is based on the following argumentation. In (43a), ki precedes pou, which it must do, as is shown by the
ungrammaticality of (43b), where pou precedes ki. In both (43a) and (43c), pou
must be in comp since it precedes the marker of anteriority te and the negation

7.2 Complementisers or resumptives?

197

marker pa, respectively. Since the canonical surface order of these morphemes
is pa te pou, as we saw in chapter 5, whereas in (43c) the order is pou pa te,
Koopman concludes that pou is in comp in the sentences in (43). Since ki must
precede pou, ki must be in comp as well and, at S-structure, the subject position
is empty. The structure she proposes for the data in (43) is shown in (44), which
makes use of two positions in comp.
(44)

[S Ki

mounni [S ki (pou)

[S [e]i te

vini a] ] ]
haitian
(=(55) in Koopman 1982b)

Koopman further proposes that, from its position in comp, ki properly governs
the empty subject position, thus allowing the structure in (44), which would
otherwise be illicit by virtue of the that-trace filter or the ecp. On Koopmans
account, Haitian ki is like French qui, that is, a tensed complementiser which has
the property of properly governing the empty subject position.
Koopmans analysis can be challenged from both a theoretical and an empirical point of view. First, a major drawback to Koopmans account in (44) is that
it allows for a doubly filled comp. This account cannot be maintained in light
of current analyses of doubly filled comp phenomena (see e.g. Chomsky 1981,
and the references therein). Second, the Haitian data presented by Koopman
(see (43) ) clearly show that Haitian ki is compatible with pou, hypothesised to
be in comp in (44). This contrasts with French qui, which, as we saw above, is
mutually exclusive with the two complementisers que. Hence, while the que/
qui alternation in French argues for an analysis of qui as a complementiser, the
same conclusion cannot be drawn for Haitian ki.7 Third, on the basis of a sample
of Haitian speakers, Lefebvre (1996b) shows that the surface word order of
tma markers is subject to variation among speakers. While the order te pou is
accepted by all speakers, the surface order pou te is also acceptable to some
speakers. These two possibilities are illustrated in (45), showing a difference in
scope between the two markers.
(45)

a. Mari
Mary
Mary
Mary

te
ant
had to
had to

pou prepare pat.

sub prepare dough
prepare dough.
have prepared dough.

haitian

(=(39) in Lefebvre 1996b)

b. Mari pou te
prepare pat.
Mary sub ant prepare dough
Mary has to have prepared dough.

haitian
(=(40) in Lefebvre 1996b)

Similarly, the surface word order of pou with respect to pa is also subject to
variation among speakers. Once again, the order pa pou is accepted by all
speakers, whereas the surface order pou pa is also acceptable to some. These
two possibilities are exemplified in (46), which also shows a difference in scope
between the two word orders.

198
(46)

Functional category lexical entries

a. Mari pou pa
prepare pat.
Mary sub neg prepare dough
Mary should not prepare dough.

haitian

b. Mari pa
pou prepare pat.
Mary neg sub prepare dough
Mary does not have to prepare dough.

haitian
(=(132) in Lefebvre 1996b)

The above data show that at S-structure pou may precede te (see (45) ) or even
pa (see (46) ) in a simple clause. In both cases, pou occurs between the subject
and the verb, showing that there is a surface position for pou following the
subject even when it precedes te or pa. Hence, pou need not be in comp in (43a)
or (43c) and the simplest assumption is that it is in the same surface position as
in (46), that is, after the subject. In view of this new fact, Koopmans argument
that ki is in comp because it precedes pou, which must be in comp, is no longer
valid. Furthermore, there is a subset of Haitian speakers for whom (43b) is
grammatical; for these speakers, the word order pou ki is perfectly acceptable
(see Law 1994b, fn. 3, and my own fieldnotes). Given this new set of data, a
plausible structure for the sentences in (43) could be as in (47), where ki occupies the subject position left empty by subject extraction and pou is either in S,
as in (47a), or in comp, as in (47b), depending on the speaker.
(47)

a. [Ki

mounn

[COMP

[S ki

b. [Ki

mounn

[COMP pou

pou

te vini

[S ki te vini

an] ]
an] ]

haitian
haitian

This brings me to the second proposal advocated in the literature, whereby ki

is a nominative resumptive which occupies the position of the extracted subject.
There are three major arguments showing that Haitian ki occurs in subject position. First, consider the data in (48) involving complements of the verb vle
want. Law (1992) points out that ki in the (b) sentence is in the same surface
position as the lexical subject in the (a) sentence and that therefore both ki and
Mari must be in the same surface position, that is, in the subject position of the
embedded clause.
(48)

a. Jan
vle
pou
Mari vini.
John want comp Mary come
John wants Mary to come.

haitian
(=(28) in Law 1992)

b. Ki
mounn Jan
vle
pou
ki
vini?
which person John want comp res come
Who does John want to come?

haitian
(=(29) in Law 1992)

Second, recall from section 7.1.2 that verbs of the want-class subcategorise for
the complementiser pou; thus, pou in (48) is the complementiser of the embedded clause. Given that, in this context, pou must be in comp and ki follows pou,
Law (1992) argues that ki is in the subject position of the extracted subject, as
in (49).

7.2 Complementisers or resumptives?

(49)

Ki mounn

Jan vle

[CP pou

[IP ki

[VP vini] ] ]

199

haitian
(=(29) in Law 1992)

Third, facts involving raising to subject position constitute yet another argument
showing that ki is in the extraction site of a fronted subject. These facts are
extensively discussed in Law (1992). I shall briefly summarise his argument. For
at least a subset of Haitian speakers, the verb sanble seem shares properties
with the English class of raising verbs. As in English, this verb takes a sentential
complement whose subject may be realised in situ. In this case, an expletive
may fill the surface subject position of sanble, as in (50a). However, the embedded subject may also be raised to the subject position of sanble,8 in which case
a resumptive pronoun must occur in its extraction site, as in (50b). When the
raised subject is plural, the resumptive pronoun is also plural, as in (50c).
(50)

a. (Li) sanble Jan

entlijan.
it
seem
John intelligent
It seems that John is intelligent.

haitian

b. Jan
sanble *(li) entlijan.
John seem
he intelligent
John seems to be intelligent.

haitian
(=(1) in Law 1992)

c. Yo
sanble *(yo) entlijan.
they seem
intelligent
They seem to be intelligent.

haitian

Law (1992) shows that, in raising contexts, when the embedded subject is questioned, ki appears in both the extraction site and the subject position of sanble,
as shown in (51).9
(51)

Ki
mounn ki sanble ki entlijan?
which person
seem
intelligent
Who seems to be intelligent?

haitian
(=(8) in Law 1992)

The most embedded ki in (51) is in the same position as Jan in the embedded
clause in (50a), and the ki before sanble in (51) is in the same position as Jan
in (50b). Thus we see that, in raising-to-subject constructions, the embedded
subject moves to the subject position of the matrix verb; moreover, when the
embedded subject is questioned, ki appears in both subject positions. This argues
that ki is in subject position, that is, in the specifier position of ip.10 These facts
support the analysis that, at S-structure, ki is a resumptive occurring in the position of the extracted subject.
Haitian ki does not occur in the subject position of infinitival complements, as
shown in (52). As will be argued in chapter 9, the clausal complement of vle to
want in (52a) is infinitival and, in this case, the subject of the embedded clause
is assigned Accusative Case by vle. When the subject of the embedded clause is
questioned, ki cannot surface in the position of the extracted subject, as is shown
by the contrast in grammaticality between (52b and c).

200
(52)

Functional category lexical entries

a.

Jan
vle
[Mari vini].
John want Mary come
John wants Mary to come.

haitian

b.

vle [ ti vini]?
Ki
mounni Jan
which person John want
come
Who does John want to come?

haitian

c. *Ki mounni Jan vle [kii vini]

haitian

The fact that ki cannot surface in the position of a subject extracted out of a
tenseless clause, as shown in (52b and c), while it must occur in the position of
a subject extracted out of a tensed clause (see (48b) ), argues that ki bears the
Case feature [+nominative]. Indeed, Nominative Case is only assigned to subjects of tensed clauses (see Chomsky 1981). Hence, ki is excluded from (52b)
because it is not compatible with the Accusative Case assigned to the embedded
subject position by vle. With Law (1992), I therefore conclude that Haitian ki is
not a complementiser, occurs in the subject position of an extracted subject and
bears the Case feature [+nominative].
While questions in Haitian can only be rendered by a cleft construction
(see (39)(41) ), in French, they may take the form of a cleft construction (see
(38) ), but they need not (see (37) ). While French qui is a complementiser which
properly governs an empty subject position, Haitian ki is a resumptive occurring
in the position of an extracted subject. The fact that Haitian ki does not have the
properties of French qui provides more support for the claim that the creators of
Haitian did not acquire these properties because they did not have enough exposure to French. Why does Haitian ki have the properties that it has? Once again,
facts from the substratum languages will provide an answer to this question.
As is the case in Haitian, Fongbe questions are rendered by a cleft construction. The fronted constituent is followed by the morpheme wY (literally it is),
used in clefting all types of constituents including clauses. Examples are given
in (53).
(53)

a. Kwk wy sb mw.
Koku it-is Asiba catch-sight-of
It is Koku that Asiba caught sight of.

fongbe

b.
tx
(wy) sb mw?
that which it-is Asiba catch-sight-of
What is it that Asiba caught sight of?

fongbe

Note that, just like Haitian se (see (39)(41) ), Fongbe wY is optional in questions
but not in clefts involving other types of constituents. Law (1994a) and Law and
Lefebvre (1995) argue that clefts are bi-clausal in Fongbe, as they are in Haitian.
As is extensively discussed in Law (1994b), Fongbe exhibits subject/object
asymmetries. The sentences in (54) show that, when the subject of an embedded
clause has been extracted, the form must appear in that clause. The sentences in
(55) show that, when the direct object has been extracted, no such form appears.

7.2 Complementisers or resumptives?

(54)

a.

Myi
tx
(wy) Kwk 3
3w i
person which it-is Koku think that ?
Who is it that Koku thinks saw Asiba?

a.

sb?
Asiba

fongbe

wy Kwk 3 3w ti mw sb

fongbe

tx
(wy) Kwk 3
3w By mw ti ?
i
that which it-is Koku think that Bayi see
What is it that Koku thinks Bayi saw?

fongbe

b. *Myi
(55)

mw
see

201

tx

b. *i tx wy

Kwk 3

3w

By mw i

fongbe

The ungrammaticality of (54b) contrasts with the grammaticality of (55a), revealing a subject/object asymmetry.
The form in (54a) has the same phonological representation as the thirdperson subject form (see chapter 6). Law (1994b) argues that it actually is the
third-person pronoun for, when the extracted subject is plural, the third-person
plural form yI shows up instead of . This is shown in (56).
(56)

[M6
lx]i (wy) Kwk 3
3w
yii mw sb?
person pl it-is Koku think that res see Asiba
Who (pl.) is it that Koku thinks saw Asiba?

fongbe

Since these forms are pronominal, they must occur in a position where they
are allowed to occur. Head of cp is not a licit position for pronominal forms.
Furthermore, on the analysis that #W in (54) and (56) is a complementiser (see
section 7.1.2), the head of cp is not available for any other form. Law (1994a,
1994b) therefore proposes that these forms are pronominal resumptives occurring in the position of the extracted subject. The analysis he proposes is that #W,
like that in English, is not a proper governor for the empty subject position. The
third-person pronominal forms have to be inserted as resumptives in order to
rescue the structure, which would otherwise violate the ecp.
Resumptive forms also occur in the context of subject raising in Fongbe.
This is shown in (57) involving the raising verb # to seem. This verb takes
a sentential complement whose subject may be realised in situ. In this case, an
expletive may fill the subject position of #, as in (57a). The embedded subject
may be raised to the subject position of #, in which case a resumptive pronoun
must occur in the extraction site of the embedded subject, as in (57b). When the
raised subject is plural, the resumptive pronoun is also plural, as in (57c).
(57)

a. () 3
m6
3 Kwk jzwn v.
3rd seem person op Koku sick
det
It seems that Koku is sick.

fongbe

b. Kwk 3
3w

jzwn.
Koku seem that res sick
Koku seems to be sick.

fongbe

c. Y
3
m6
3 y
3rd pl seem person op res
They seem to be sick.

jzwn.
sick

fongbe

202

Functional category lexical entries

The Fongbe data in (57) parallel the Haitian data in (50). Both languages require
resumptives in the base position of a raised subject. This is because, as will be
seen in chapter 9, subject raising in Haitian and Fongbe (but not in French)
operates out of tensed clauses.
According to the above analyses, Fongbe and Haitian share the property that
a subject position left empty by subject extraction must be lexically filled at Sstructure. This strategy allows potential ecp violations to be repaired in these
languages. On this point, both languages contrast with French where potential
ecp violations are repaired by means of a special form of the complementiser
which acts as a proper governor for the empty subject position. Fongbe and
Haitian differ, however, on two points. First, as is pointed out by Law (1994b),
while in Haitian the form ki must appear in the extraction site of both local and
long-distance subject extraction (see (41), (42) ), in Fongbe the pronominal form
in subject position occurs only in long-distance extraction (see (54) ). It does not
occur in short-distance extraction, as is shown in (58).
(58)

a. M6i
wy ti y?
person it-is
leave
Who is it that left?

b. *M6i wy i y
Who is it that left?

fongbe

As Law (1994b) discusses in detail, the fact that (58a) is unexpectedly licit,
combined with the fact that (58b) is unexpectedly illicit, poses a problem for any
theory of empty categories, and a syntactic solution to this problem still requires
further research.
The second difference between the two languages is that, in Fongbe, the
resumptive elements are pronominal forms bearing the person and number features of the extracted subject in all contexts (see (56) and (57) ), whereas in
Haitian the resumptive element is a pronominal form in subject raising contexts
(see (50) ) but the invariant form ki in contexts involving Wh-movement (see
(41) ).
Since Fongbe has pronominal resumptives in the base position of both raised
subjects and Wh-moved subjects, we would expect to find the same distribution
of resumptive elements in Haitian. This expectation is only partially borne out,
however, for, while we find pronominal resumptives in the base position of raised
subjects, we find ki in the base position of Wh-moved subjects. In my view, this
discrepancy between the Fongbe and Haitian data is not relevant for the relexification hypothesis because the distribution of resumptives is syntactically
driven. Hence, what is important here is the fact that a resumptive form occurs
in the position of an extracted subject in both Haitian and Fongbe. As for why ki
appears in this context, Law (1994b) discusses a scenario along the following
lines. The creators of Haitian most probably heard the French form qui in questions involving subjects (see (38) ). Since they were aiming at reproducing what
they heard, they adopted this form (ki in Haitian). However, they did not have
enough exposure to French to acquire its properties and so they assigned it a
function parallel to that of resumptives in their own grammar.

7.3 The nominal operator in relative and factive clauses

7.3

203

The nominal operator in relative and factive clauses

Like other Gbe languages, Fongbe has a lexical operator #I which shows up in
relative and factive clauses, as shown in (59) and (60), respectively.
(59)

Xw
3i my un dml 3
house op in I
sleep
loc
The house in which I slept.

v.
det

fongbe
(=(14d) in Kinyalolo 1993a)

(60)

W
3i Jan
w
v
vv
n
fongbe
arrive op John arrive det make-happy for
nw
twn.
mother his
The fact that John arrived made his mother happy.
(=(3) in Lefebvre 1994b)

Kinyalolo (1993a) and Collins (1994) argue that this morpheme is an operator
rather than a complementiser. The basis for their claim is that #I can pied-pipe
postpositions ( just as which can pied-pipe prepositions in English). In (59) #I
pied-pipes the postposition mY in. In (61) (based on Collins 1994), it piedpipes the postposition j on (for an analysis of this type of factive construction,
see Collins 1994).
(61)

Tv 3i j By ss v
vv
n mi.
table op on Bayi wipe det please for me
The fact that Bayi wiped the table pleases me.

fongbe

Since complementisers cannot pied-pipe material whereas #I does, #I cannot be

a complementiser. The alternative is that it is an operator in Spec of cp. Collins
further argues that it is a nominal operator. In (61) #I is co-indexed with the
noun tv table. It could not be co-indexed with a postpositional phrase, as is
shown by the ungrammaticality of (62) (based on Collins 1994).
(62)

*Tv
table

j 3i
on op

By
Bayi

ss
wipe

v
vv
det please

n mi
for me

fongbe

The fact that #I can be co-indexed with a noun (as in (61) ) but not a postpositional phrase (as in (62) ) shows that it is a nominal operator.
Operators have no semantic content. Since relabelling is semantically driven,
we would expect them to be assigned a null form. What kind of operator does
Haitian have in these kinds of constructions? In Haitian, the operator that occurs
in relative and factive clauses is phonologically null, as shown in (63) and (64).
(The structure of factive constructions will be discussed in chapter 12.)
(63)

Fiyi i m sti
ak
girl op I go-out with
The girl I went out with.

lii
her

a.
det

haitian
(=(44) in Koopman 1982a)

204
(64)

Functional category lexical entries

W
3i
Rive

arrive op
The fact that

Jan
w
v
...
Jan
rive
a
...
John arrive det
John arrived . . .

fongbe
haitian
(=(3) in Lefebvre 1994b)

Since the operator is null in this language, pied-piping phenomena of the type
observed in Fongbe are not attested in Haitian. Crucially, however, this null operator is a nominal operator. Koopman (1982a) provides extensive evidence that pps
cannot occur in the Spec of cp in Haitian, as indicated by the ungrammaticality
of (65).
(65)

*Fiy
girl

ak
lii u
with her you

sti ti a
go-out det

haitian
(=(42) in Koopman 1982a)

In relative clauses involving a pp, the operator in the Spec of cp is co-indexed with
a resumptive pronoun in the complement position of a preposition within the
relative clause, as shown in (63). These facts follow if the null operator is nominal.
How does Haitian compare with French? In French, there is no overt operator.
However, there appears to be a distinction between the properties of the operator
in standard and popular French. This is illustrated for relative clauses in (66). In
(66a), the pp occurs in Spec of cp (in contrast to Haitian (see (65) ), showing that
in standard French the relative operator is not nominal. In (66b), the pp occurs
in situ; the complementiser que introduces the relative clause and there is a null
operator in Spec of cp. The phonologically null operator in this case must be
nominal since it cannot be co-indexed with a pp.
(66)

a. La
fillei [avec quii je suis sorti ti ]
det girl with who I aux go-out
The girl I went out with.
b. La
det
avec
with
The

je
fillei [i que
girl op comp I
(elle)i ]
her
girl I went out with.

suis sorti
aux go-out

standard french

popular french

Moreover, there are no factive clauses of the type in (64). Factive clauses in French
are constructed with the complementiser que as in le fait que . . . the fact that . . .
and contain no overt operator. In the following discussion, I will assume that the
creators of Haitian were exposed to the structure of French relative clauses illustrated in (66b). As is extensively discussed in Bouchard (1982), relative clauses
built on the model of (66b) were common in the variety of French spoken in the
seventeenth century.11 Assuming an analysis where relative clauses like (66b) and
factive clauses require a phonologically null nominal operator in order to be interpreted, all three languages would be similar in having a nominal operator. While
this operator is overt in Fongbe, it is covert in both popular French and Haitian.
What is the historical derivation of the Haitian nominal operator? Given
what we have seen so far, it is unlikely that the creators of Haitian acquired

7.4 Clausal conjunction

205

the properties of the French operator. By hypothesis, they did not pronounce a
nominal operator when uttering relative and factive clauses in Haitian (see (63),
(64) ). This means that they had assigned a null form to this lexical entry. Since
the nominal operator was phonologically null, it could not pied-pipe lexical
material, hence the discrepancy between Fongbe (59) and Haitian (63). The first
generation of Haitian native speakers would have deduced the nominal character
of the null operator on the basis of the fact that they were not exposed to pp
fronting (see (65) ).
7.4

Clausal conjunction

Haitian has a conjunction (e)pi used to conjoin clauses, as shown in (67).

(67)

Jan pati (e)pi Mari rive.

John left and Mary arrived.

haitian
(=(70) in Lefebvre 1993b)

This conjunction derives its phonological representation from the French sequence
of words et puis [Lit.: and then], pronounced [(e)pi], which is used in complementary distribution with et and to conjoin clauses and noun phrases, as shown
in (68).
(68)

a. Jean est parti et / (e)pi Marie

John left and Mary arrived.

est arrive.

french

(=(73) in Lefebvre 1993b)

b. Jean et / (e)pi Marie
John and Mary

french
(=(74) in Lefebvre 1993b)

In contrast to French, the Haitian conjunction (e)pi cannot be used to conjoin

noun phrases (see (69a) ). Conjunction of noun phrases is achieved by adjoining
a prepositional phrase to the first noun. This phrase is headed by the preposition
ak or (kl)ak (<kle-ak close with) (see Gilles 1988) (see (69b) ).
(69)

a. *Jan (e)pi Mari

b. Jan (kl-)ak Mari

John and Mary

haitian

Once again, the above distribution finds a straightforward explanation when we

examine comparable data from the substratum languages.
Koopman (1986) notes that in West African languages different lexical items
are used to coordinate clauses and nps. In Fongbe, for example, the conjunction
bW12 is used to co-ordinate clauses but not nps, as shown in (70).
(70)

a.

Jan y bw Mari w.
John left and Mary arrived.

fongbe
(=(70) in Lefebvre 1993b)

b. *Jan bw Mari
[Lit.: John and Mary]

fongbe
(=(71) in Lefebvre 1993b)

206

Functional category lexical entries

As is the case in Haitian, nps are co-ordinated by adjoining a prepositional

phrase headed by kp# (Lit.: with-at) to the first noun.
(71)

Jan
kp3 Mari (kp)
John with
Mary with
John and Mary

fongbe
(=(72) in Lefebvre 1993b)

The properties of Haitian (e)pi and (kl-)ak are derivable straightforwardly

according to the relexification hypothesis. BW was relexified as (e)pi on the basis
of French et puis and kp# was relexified as (kl-)ak on the basis of French
coller to be close to and avec with.13
7.5

The mystery of Haitian se

The phonological representation of Haitian se in the cleft constructions discussed above is derived from French cest it is. Both occur in cleft constructions, as shown in (72).
(72)

a. Se

Jan

Mari

w.

b. Cest Jean que Marie a aperu.

It is John that Mary caught sight of.

haitian
french

In spite of their apparent similarity, however, se and cest have quite different
properties, as is revealed by the literature on the topic (see e.g. Lumsden 1990;
Deprez and Vinet 1991; DeGraff 1992b, 1992c). For example, while Haitian se
must appear in the environment of a nominal predicate (see (73a) ) (see DeGraff
1992b: 89), French cest cannot appear in this context (see (73b) ), unless the
subject has been topicalised (see (73c) ).
(73)

a.

Bouki

b. *Bouki
c.

se

yon

cest un

Bouki, cest un
Bouki is a doctor.

dkt.

haitian

mdecin

french

mdecin.

french

Furthermore, the analyses proposed for Haitian se reveal the differential properties
of se and cest. Cest is made up of a demonstrative pronoun, ce/c it, which
occurs in subject position, and the copula tre to be. Haitian se, however, is
an unanalysable word. According to Lumsdens (1990) analysis, se heads in
in (73a) and occurs in subject position in (72a). By Deprez and Vinets (1991)
analysis, se is an assertive marker occurring as a functional head in (73a) and is
in subject position in other contexts. For DeGraff (1992b), se is a resumptive
nominal element. Whatever the analysis, the properties of se are different from
those of French cest. It thus appears that, although cest is the source of the
phonological representation of se, it did not provide its other properties.
The closest Fongbe form to Haitian se is wY, which also occurs in clefts, as
shown in (74).

7.5 The mystery of Haitian se

(74)

Se

Mari
Jan
w.
Mari wy Jan
mw.
it-is Mary it-is John catch-sight-of
It is Mary that John caught sight of.

207

haitian
fongbe

Se and wY differ in their distributional properties: se occurs at the beginning of

the clefted constituent, whereas wY occurs at the end of it. On the aforementioned
analyses of Haitian se, it surfaces in a specifier position in (74). In contrast,
Fongbe wY in (74) surfaces in a head position (see Lefebvre 1992a). Furthermore,
unlike Haitian se, which must occur in the environment of a nominal predicate
(see (73a) ), wY does not occur in this context, as shown by the ungrammaticality
of (75a). In this case, Fongbe uses the verb ny, as in (75b).14
(75)

a. *Kwk
b.

wy

dt

Kwk ny dt.
Koku is a doctor.

fongbe
fongbe

It thus appears that, in this case, the properties of se were not provided by the
substratum language. Presumably, the creators of Haitian who had a lexical entry
like Fongbe wY did not find an appropriate form with a suitable distribution in
the substratum language to relabel it. Since this lexical entry is not signalled
by anything in modern Haitian, it is likely that the creators of Haitian abandoned
this form. Unless another substratum language can be argued to have provided
the properties of se, and given the methodology adopted in chapter 3, we must
conclude that they constitute an independent development (see also DeGraff
1992b, for a similar claim).
Interestingly enough, however, Smith (1996) reports that Saramaccan has
what he refers to as a contrastive focus marker. As is shown in (76), this marker
has the form wY and occurs at the end of the clefted constituent.
(76)

And wy i
bi?
what foc 2nd cook
What did you cook?

saramaccan
(from Smith 1996)

Smith (1996) documents the fact that, in addition to being homophonous with
Fongbe wY, this form also has the same tone, semantics and syntactic and distributional properties. Further, he presents data from the early eighteenth century
attesting to the presence of this form in the Saramaccan lexicon of the time.
Since a large number of Fongbe speakers were present at the time Saramaccan
was formed (see Smith 1987; Arends 1995a), Smith concludes that they simply
retained wY (see also McWhorter 1996, for a similar proposal). Presumably this
form spread out in the Saramaccan community. Recall from chapter 2 that the
retention of substratum functional category forms is rather rare in creoles (as
compared with mixed languages). This example, however, shows that it is a
possibility, albeit one that was obviously not chosen by the creators of Haitian.

208
7.6

Functional category lexical entries

Negation markers

In French, clauses are negated with ne . . . pas, as shown in (77), where the
parentheses indicate that ne is not always pronounced/present.
(77)

Jean (ne) mange pas.

John ne
eat
pas
John does not eat.

french

According to Pollocks (1989) analysis, ne is the head of NegP and pas is an

adverb of negation base-generated in the Specifier position of NegP, as in (78).
tp

(78)

np

NegP

Neg

Spec Neg
pas

agrp

Neg
ne

agr

vp

(=(127) in Pollock 1989)

Word order is accounted for as follows: the verb moves to the head of tp;
because ne is a clitic, it must move to t, yielding the word order ne-tensed verbpas in (77).
The Haitian negation marker pa is homophonous with French pas not.
However, although French pas obviously supplied the form of the Haitian negation marker, it did not contribute its other properties. One contrast noted by
DeGraff (1993a) is that, while Haitian pa generally precedes the tense, mood
and aspect markers, in French pas always occurs after the finite verb. This
contrast is illustrated in (79) and (80).
(79)

a.

Jan
pa
t
avale nan mache.
haitian
John neg ant ind-fut go in
market
John would not have gone to the market.
(=(1a) in DeGraff 1993a)

b. *Jan t

av

ale pa nan

mache
(=(1d) in DeGraff 1993a)

(80)

a.

Jean (ne) serait

pas all
au
march.
french
John (ne) would-be pas gone to-the market
John would not have gone to the market.
(=(2a) in DeGraff 1993a)

7.6 Negation markers

b.

Jean n
ira
pas au
cinma.
John (n) go-fut pas to-the movies
John will not go to the movies.

209

french
(=(2c) in DeGraff 1993a)

On DeGraffs analysis, this contrast in word order follows straightforwardly

from the analysis that, in French, the tensed verb raises to in (see Pollock
1989) while, in Haitian, it does not (DeGraff 1993a). This is compatible with the
presence of inflectional verbal morphology in French and the absence of such
morphology in Haitian (see chapter 5).
A second distributional difference noted by DeGraff (1993a) is that, whereas
Haitian pa must occur between the subject and the verb, French pas may, in
some contexts, occur at the periphery of the clause that it modifies. Compare
(81) and (82).
(81)

(82)

Bouki fait
le
clown pour pas qu ils sennuient. french
Bouki makes the clown for
pas that 3pl bore+ref
Bouki is clowning around so that they dont get bored.
(=(11a) in DeGraff 1993a)
*Bouki
Bouki

ap f
imp make

komik
clown

pou
for

pa
pa

yo
anniye
haitian
they bore
(=(11b) in DeGraff 1993a)

Third, French pas, but not Haitian pa, may occur in nps, as shown in (83) and
(84).15
(83)

Voil un type
pas bte.
there a
fellow pas stupid
There goes a man who is not stupid.

french
(=(12a) in DeGraff 1993a)

(84)

*Men
yon mounn pa st
here / there-is a
fellow pa stupid
There goes a man who is not stupid.

haitian
(=(12b) in DeGraff 1993a)

Finally, while French pas is an adverb of negation occurring in the specifier

position of NegP (see Pollock 1989), Haitian pa has been analysed as the head
of NegP (see DeGraff 1993a). These analyses account for the fact that, while
French pas can occur in nps, Haitian pa cannot. These facts and analyses argue
that, although French pas contributed the form of Haitian pa, it did not contribute
its other properties. DeGraff (1993a) further claims that, as the head of NegP,
Haitian pa actually shares properties with French ne (see (77) ), the head of NegP.
But, while French ne is a clitic which must move in order to attach to the tensed
verb, as we saw in (78), there is no evidence at all that Haitian pa is a clitic or
that it must move up in the syntactic tree to attach to the tensed verb. As we saw
above, on DeGraffs analysis, the verb in Haitian always stays in its basic position
within the vp. Where, then, do the properties of Haitian pa come from?

210

Functional category lexical entries

All Gbe languages have a negation marker that occurs between the subject
and the verb (see Hazoum 1990). In Fongbe, this marker is m. As is the case
with Haitian pa, this marker generally precedes the tma markers. Compare (85)
with (79a).
(85)

Kwk m n
w x
my.
Koku neg sub go market in
Koku does not have to go to the market.

fongbe

In Haitian (for a subset of speakers) and in Fongbe, the mood markers can also
precede pa/m, as shown in (86). Note the effect of word order on the interpretation of the sentence.
(86)

a. Mari
Mari
Mary
Mary

pa
pou prepare pat.
m
n
3
wh.
neg sub prepare dough
does not have to prepare dough.

haitian
fongbe

b. Mari
Mari
Mary
Mary

pou
n
sub
should

haitian
fongbe

pa
prepare pat.
m 3
wh.
neg prepare dough
not prepare dough.

(=(132) in Lefebvre 1996b)

As in Haitian, the respective order of the negation marker and the verb in
Fongbe follows from the analysis that the verb does not raise to in in this
language (see Avolonto 1992). This is compatible with the absence of inflectional verbal morphology in Fongbe (see chapter 5). Furthermore, like Haitian pa
(see (82) ), Fongbe m cannot occur at the periphery of the clause that it modifies,
as shown in (87).
(87)

*
2nd
yv
bore
[Lit.:

3 yvn-tvm= bl wy
bn
m vj
l=
at clown
do part in-order neg child pl

ins
You are clowning so that the children dont get bored.]

fongbe

Finally, according to current analyses of markers like m in the literature on

Kwa languages (e.g. Da Cruz 1994, for Fongbe; Collins 1993 and gbdvr
1995, for Ewe), such markers are the head of NegP.
If m in Fongbe is the head of NegP, it should not be allowed to appear in
nps. As is noted in Da Cruz (1994), however, we find the prefix m- in participial
forms modifying a noun, as shown in (88).
(88)

Kwk n
ln
mb-mb
Kf.
Koku give meat uncook-uncook Kofi
Koku gave Kofi uncooked meat.

fongbe
(=(34) in Da Cruz 1994)

Faced with (88), one might jump to the conclusion that the negation marker in
Fongbe can occur in nps, and thus contrasts with Haitian pa in this respect (see
DeGraff 1993a). The fact is, however, that, in Fongbe, the negation marker m

7.7 Yesno question markers

211

is homophonous with the inversive/negative derivational prefix m-. As will be

discussed at length in chapter 10, m- is a prefix that attaches to verbal bases,
as shown in the following pair of words: # to believe, m-# to doubt (see
Brousseau 1990). The derived word may be reduplicated to form a participial
adjective, as in (88). Whether m and m- constitute one or two lexical entries
is hard to tell on the basis of Fongbe data. If one could find a Gbe language
where the negation marker and the inversive/negative prefix were phonologically
distinct, however, their distribution would be clear. Ewe is such a language.
In Ewe, the negation marker is m and the inversive/negative prefix is m- as
shown in (89).
(89)

a. Jan
m y .
John neg go ins
John did not go.

ewe
(=(2) in Wallace 1995c)

b. mbl-mbl
disrespect

ewe
(=(6b) in Wallace 1995c)

The negation marker m is the head of NegP in the clause (see gbdvr 1995;
Collins 1993) and does not occur in nps, where only the inversive/negative
prefix can occur (Wallace, p.c.). Thus, in Ewe, m and m- can be argued to
constitute two distinct lexical entries with different distributional properties.
This analysis can be extended to Fongbe. In this view, m is the head of NegP
and does not occur in nps, and the prefix m- constitutes a separate lexical entry.
Therefore, the conclusion that Haitian pa and Fongbe m share the same properties can be maintained.
The parallel distribution of m and pa and the fact that they are both analysed
as the head of NegP follow from the relexification hypothesis: Fongbe m was
relexified as pa on the basis of the French negation adverb pas. The fact that
French ne did not enter Haitian also follows from the claim in chapter 2 that the
functional items of the superstratum language were not identified as such by the
creators of the creole.
Fongbe also has another negation marker kn (see Segurola 1963) which
occurs in complementary distribution with m. The distribution of this marker is
very limited (see Segurola 1963; Da Cruz 1994). Since there is no other form
than pa in Haitian to encode sentential negation, I assume that kn was also
relexified on the basis of French pas, such that, in modern Haitian, there is only
one lexical entry pa.
7.7

Yesno question markers

Fongbe has a sentence-final yesno question marker .

(90)

Kwk xw svn lx ?
Koku buy crab pl q
Did Koku buy the crabs?

fongbe

212

Functional category lexical entries

According to Hazoum (1990), not all Gbe dialects have such a particle. Gungbe
and Xwedagbe encode yesno questions by lengthening the vowel of the last
word of the clause and assigning it a low tone. According to Wallace (1995c),
Ewe does not have a yesno question marker and direct questions are encoded
by means of rising intonation.
In Haitian, there is no sentence-final yesno question marker. Questions are
encoded in one of two ways: rising intonation, as in (91a), or ske at the beginning of the clause, as in (91b).
(91)

a. Jan

vini?

haitian

b. ske Jan
vini?
q
John come
Has John come?

haitian

Given the substratum data, how did Haitian end up like this? The first question to ask is whether French had a form with the appropriate semantics and a
suitable distribution to relexify the question marker in the Fongbe lexicon. Quebec
French has a question particle tu and regional dialects of France have a similar
particle ti, which occurs after the tensed verb, as shown in (92).
(92)

a. Il est-ti / tu arriv?
Has he arrived?

b. Il vient-ti / tu?
Is he coming?

popular french

As shown in (92), this particle does not systematically occur at the end of the
clause since it attaches to the tensed verb. Its distribution is thus not the same
as that of the substratum languages question marker. Furthermore, as we have
seen over and over again, the creators of Haitian did not identify as such the
functional categories of French because they did not have enough exposure to
the language. Apparently, they did not identify this particle either since it is not
part of the Haitian lexicon. Since there was no other form in French that was
appropriate to relabel the copied lexical entry from the substratum language, it
could not be assigned a label. Presumably it was abandoned by the creators of
Haitian since there is no evidence for it in modern Haitian.
What the creators of Haitian did find in French, however, was the phonetic
string [=sk1] < est-ce que, a complex sequence formed by the verb to be, the
pronoun ce and the complementiser que and used to encode yesno questions, as
shown in (93).
(93)

Est-ce que Jean est

arriv?
q
John aux arrive
Has John arrived?

french

They apparently perceived this sequence of words as a single unit since it

entered the Haitian lexicon as the word ske (see Valdman et al. 1981). Having
been incorporated into Haitian as a major category lexical item, its position in
the clause follows the word order of the superstratum rather than the substratum
language, as is predicted by the hypothesis in chapter 2.

7.8 Markers expressing the speakers point of view

213

In standard French, yesno questions may also be encoded by raising the

tensed verb to comp, as shown in (94).
(94)

Est i -il ti arriv?

Has he arrived?

french

This possibility is not available in Haitian, as shown by the ungrammaticality of

(95).
(95)

*Rive li
arrive 3rd

haitian

This is because verb raising to in is not available in Haitian (see DeGraff

1992b) any more than it is in Fongbe (see Avolonto 1992), as we saw in section 7.6.
Finally, yesno questions in French may be encoded by rising intonation, as
in (96).
(96)

Jean est arriv?

John has arrived?

french

This is also found in Haitian (see (91a) ) and in Ewe, as mentioned above. The
fact that this strategy was available in both a subset of substratum languages and
the superstratum language probably contributed to the fact that it is available in
Haitian. The Haitian data discussed above thus follow straightforwardly from
the hypothesis of creole genesis outlined in chapter 2.
7.8

Markers expressing the speakers point of view

Da Cruz (1994) documents the fact that, in Fongbe, there are markers which
express the speakers point of view on the proposition: F is a negative marker
(see (97) ) and is a marker that Da Cruz glosses as ins for insistence (see
(98) ).
(97)

Kwk xw
svn lx f.
Koku buy crab pl Neg
Koku did not buy the crabs.

fongbe
(=(2) in Da Cruz 1994)

(98)

Kwk n
xw
svn
lx
Koku sub
buy
crab
pl
Koku must buy the crabs (insistence).

.
ins

fongbe
(=(8a) in Da Cruz 1994)

These two markers are mutually exclusive and, hence, are in a paradigmatic
relationship.
As Da Cruz (1994) shows in detail, F cannot appear in clauses containing the
negation marker m, discussed in section 7.6, except in conditional clauses, as is
exemplified in (99a and b), respectively.

214
(99)

Functional category lexical entries

a. *Kwk
Koku

m xw
neg buy

svn lx
crab pl

f
Neg

fongbe
(=(6) in Da Cruz 1994)

b. (N) Kwk m xw
svn lx f,

n
y.
fongbe
if
Koku neg buy crab pl Neg 3rd fut go
If Koku does not buy the crabs, (s)he will leave.
(=(7) in Da Cruz 1994)

By contrast, usually occurs in negative clauses with the negation marker m,

as in (100).
(100)

M xw
svn lx .
neg buy crab pl ins
Dont buy the crabs (insistence).

fongbe
(=(27) in Da Cruz 1994)

However, Da Cruz points out that may occur in affirmative clauses but only
when they are imperative, as illustrated in (98) above. Since it can occur in affirmative clauses, Da Cruz concludes that is not a negation marker in itself, as is
often assumed in the literature on Fongbe (see e.g. Segurola 1963; Hounkpatin
1985) and in the literature on other Kwa languages for comparable markers (see
e.g. gbdvr 1995, for Ewe). Rather, these markers express the speakers point
of view: F denies and insists on the content of the proposition. Furthermore, the
presence of in a clause presupposes that the speaker and the listener disagree.
For example, the sentence in (100) can only be uttered in a context where the
listener was about to buy the crabs.
What projection do these markers head in the structure of the clause? On the
basis of data from Basque and other languages, Laka Mugarza (1990) proposes
that there is a functional projection available in some languages to host such
markers. Since this projection may host both affirmative and negative markers, it
must be different from negp. She therefore labels it as p. Based on this proposal, Da Cruz (1994) proposes that F and are the head of p in Fongbe. This
is illustrated in (101).
(101)

vp

Since the markers that head p in Fongbe may occur in the same clause as the
negation marker (see (100) ), Da Cruz proposes that, in this language, p is a
projection that is different from negp.
Are there any functional items available in Haitian to head a projection p?
To the best of my knowledge, no Haitian form corresponds to the Fongbe

7.8 Markers expressing the speakers point of view

215

negative marker F. However, Joseph (1995) presents ample evidence showing

that Haitian does have a form equivalent to the Fongbe marker . The form of
this marker is non, as shown in (102a) (=(36a) in Joseph 1995), which parallels
the Fongbe sentence in (102b).
(102)

a. Pa

ale non!

haitian

b. M y !
neg go ins
Dont go!

fongbe

Note that there can be no pause between the marker of insistence and the verb
in either language. Joseph (1995) shows that Haitian non in (102) has the same
meaning and distribution as the Ewe marker (see Wallace 1995c), which corresponds to the Fongbe marker . For reasons of consistency, I will compare the
properties of Haitian non (based on Joseph 1995 and on additional work that we
did together on this topic) with those of Fongbe (based on my own fieldnotes).
Like Fongbe , Haitian non can occur in affirmative clauses only when they
are imperative, as shown by the contrast in grammaticality between (103) and
(104). Furthermore, as is the case with Fongbe , Haitian non presupposes that
the speaker and the listener disagree. The sentences in (104) can only be uttered
in the context where the listener does not want to go.16
(103)

(104)

a. *Li

ale

non

haitian

b. * w
[Lit.: He went.]

fongbe

a.

Ale non!

haitian

b.

Y !
go ins
Go!

fongbe

The fact that Haitian non may appear in affirmative clauses argues that it is
not a sentential negation marker any more than Fongbe is (as we saw above).
Except for their occurrence in imperative clauses, non in Haitian and in Fongbe
always occur in clauses containing the respective negation markers, pa and m
(or kn). The co-occurrence of the clausal negation marker and the marker of
insistence in simple sentences is shown in (105).
(105)

a. Li

pa

ale

non.

haitian

b. kn y .
He did not go!

fongbe

In both languages, for some speakers (but not all,17 see below) the marker of
insistence may occur in embedded clauses, as shown in (106).
(106)

a. M

vle

[pou

li

pa

b. N jl [n
m
I want him not to go.

ale

non.]

haitian

.]

fongbe

216

Functional category lexical entries

In both languages, this marker is mutually exclusive with the yesno question
marker, as shown in (107).
(107)

a. *ske li

pa

ale

non

b. *

m y
neg go

ins

he

haitian

fongbe

This is expected on the analysis that non and in (107) are markers of insistence. It is possible to process a negative-interrogative question, as in (108).
(108)

a. ske
b.

li

pa

ale?

m y
Did he not go?

haitian
?

fongbe

It is not possible to question and insist on a proposition at the same time, hence
the ungrammaticality of (107). The contrast between (107a) and (108a) further
shows that Haitian non cannot be analysed as a negative-interrogative marker in
this function. On the basis of the above data, I conclude that Haitian non has the
same function as Fongbe (and as Ew , see Joseph 1995).
On Josephs (1995) analysis, Ewe was relexified as non on the basis of
French non. This is a likely hypothesis since, in French, non may occur at the
end of clauses such as in Tas fait a, non? You did that, no? where the presence of non indicates that the speaker is asking for a confirmation of the truth
value of the proposition. This use of non in French is, however, restricted to
affirmative clauses since, according to the speakers that I consulted on this topic,
*Tas pas fait a, non? You did not do this, no? is not acceptable. This is in
direct contrast with Haitian non which, as we saw above, occurs almost exclusively in negative clauses, as does in the substratum languages. It is worth noting that Haitian is not the only creole to have such a marker. For example,
Schwegler (1991) reports very similar data from Palenquero (spoken in Colombia). The fact that the negative marker F was not relexified follows from the fact
that there was no form available in French with the appropriate semantic properties and a suitable distribution to do so. The fact that one member of the pair was
relexified, however, argues that Haitian, like its substratum languages, has a
projection p.
Both languages also have another marker of insistence, b in Fongbe and wi
(<French oui yes) in Haitian. These are mutually exclusive with the markers
discussed above. In contrast to the former markers, they presuppose that the
speaker and the listener agree on the content of the proposition. For example, the
sentences in (109) can only be uttered in the context where the listener is about
to eat and the speaker is, in fact, inviting him to do so.
(109)

3D-n b.
Manje wi.
eat
ins

fongbe
haitian

7.9 Conclusion

217

Unlike the markers discussed above, b and wi are excluded from negative,
interrogative and embedded clauses. This suggests that they do not fill the same
position as and non.
In the next chapter, I shall develop an analysis whereby there are two positions within the Haitian and the Fongbe clause to host the markers which express
the speakers point of view on the proposition: MoodP1 (=cp) and MoodP2
(=p), a lower position. MoodP1 is headed by markers which do not occur in
embedded clauses. For example, the Fongbe interrogative marker heads this
position. Since they do not occur in embedded clauses, Fongbe b and Haitian
wi must also head this position. The Fongbe negative marker F and the markers
of insistence, in Fongbe and non in Haitian, which can occur in embedded
clauses, head the lower position MoodP2. Now, recall that for some Haitian
speakers non cannot occur in embedded clauses; hence for these speakers, (106a)
is not grammatical. The above proposal can easily accommodate their judgments. I suggest that, in the grammars of these speakers, Haitian non heads
MoodP1 rather than MoodP2. This is compatible with the fact that these markers
are mutually exclusive with those heading MoodP1. Similar cases of variation
among speakers will be discussed in chapter 8.
7.9

Conclusion

The nature of the functional items involved in clause structure follows rather
straightforwardly from the theory of creole genesis outlined in chapter 2. The
null complementiser introducing complements of verbs of the say-class in Fongbe
has a null Haitian counterpart. In both Haitian and Fongbe, there are two underspecified lexical entries: one which can be used as a mood marker and a complementiser, and another which serves as a preposition and a complementiser. Haitian
and Fongbe are also similar in having a resumptive form in the basic position of
extracted subjects, unlike French, which has a special form of the complementiser
that licenses the empty subject position in this context. In both Haitian and
Fongbe, the conjunction used to conjoin clauses cannot be used to conjoin nps,
in contrast to French where the same lexical item can conjoin both clauses and
nps. The negation marker in Haitian was argued to have the same semantic and
distributional properties as the negation markers of the substratum languages but
not the French adverbial form from which it was phonologically derived. The
lexical operator # of Fongbe was not relabelled because it has no semantic
content. It was assigned a null form in the Haitian lexicon. The interrogative
marker and the negative marker F were not relexified because there were no
available forms in the superstratum language to provide them with a phonetic
matrix. The Haitian markers non and wi were shown to have the same properties
as Fongbe (and Ewe ) and Fongbe b, respectively. These data show that, as
in the substratum languages, and in contrast to French, Haitian has grammatical
markers which express the speakers point of view on the proposition. The
Haitian form ske constitutes a case of lexical borrowing from French and the

218

Functional category lexical entries

form se is hypothesised to have developed from within the creole since its
properties correspond to those of neither the substratum nor the superstratum
language. According to the analysis in section 7.1, the properties of the serial
verb #W say (which introduces complements of verbs of the say-class) were not
maintained in the creole.

The determiner and the structure

of the clause

The definite determiner found in the nominal structure of Haitian and Fongbe
(see chapter 4) also plays a central role in the structure of the Haitian and
Fongbe clause (see Lefebvre 1982b, 1991c, 1992b, 1996a, to appear b). In fact,
the semantics of the determiner and the various functions that it may perform
provide information concerning the structure of the clause in these two languages. This finds no parallel in French, for, in this language, the determiner
plays no role at all in clause structure. This chapter discusses the distribution of
the determiner as it occurs in simple Haitian and Fongbe clauses.1 It is argued
that the determiner occurring in the clause may appear as the head of several
functional category projections, namely, MoodP1, MoodP2, tp and AspP. The
data presented in this chapter are drawn from a sample of speakers of both
languages (see Lefebvre 1992b, to appear b). Two clear patterns emerge which
will be referred to as grammar 1 (which includes speakers of both Haitian and
Fongbe) and grammar 2 (which also includes speakers of both Haitian and
Fongbe). The striking fact about these data is that the same cluster of properties
distinguish grammar 1 from grammar 2 in both languages.2
8.1

The functions of the determiner in clause structure

As shown in (1), when the determiner occurs in the context of a clause, it may
be assigned three slightly different interpretations. Throughout this chapter, the
clausal determiner will be glossed as Det rather than as det since it is argued not
to project to dp.
(1)

Mounn nan kraze

Sn
v
gb
man
det destroy

manchinn nan an.

mvt
v
v.
car
det Det

haitian1
fongbe1

a. Actually, the man destroyed the car.

b. The man has destroyed the car, as we knew he would.
c. The man has destroyed the car, as we knew it would be destroyed.
(=(1) in Lefebvre to appear b)

In (1a), the determiner asserts the content of the proposition, relating it to

something that has been said earlier in the conversation. This interpretation is
expressed by actually in the translation. In (1b) and (1c), the determiner identifies
219

220

The determiner and the structure of the clause

an event that is already part of the shared knowledge of the participants. It

literally means this event in question/this event that we know of. The determiner with this meaning has been referred to in earlier work as the event determiner (e.g. Lefebvre 1992b). As such, it may trigger an interpretation which is
subject-oriented, as in (1b), or object-oriented, as in (1c). Lefebvre (to appear b)
proposes the following account of these ambiguities. In the context of the clause,
the determiner may head one of four functional category projections which are
argued to be MoodP1, MoodP2, tp and Asp(ect)P, as shown in (2). The scope
and specific interpretations of the determiner are derived by the position it occupies in the structure of the clause.
MoodP1

(2)

(=cp)

Mood
tp

Mood
t

dpj

MoodP2 (=p)

Mood
Mood

negp
neg
neg

vp
v

[e]j
v

AspP
dp i

Asp
vp

Asp
v

[e]i
v

(=(14) in Lefebvre to appear b)

In the structure above, there are two Mood projections. I argue that these two
projections are MoodPs, because they are headed by markers that express the

8.2 Clausal determiner as assertive marker

221

speakers point of view concerning the proposition. There are two classes of
these markers; their distribution within the clause motivates the two positions
MoodP1 and MoodP2. On my analysis, MoodP1 is to root clauses what cp is to
embedded clauses. MoodP1 and cp are mutually exclusive. In the structure in
(2), MoodP2 corresponds to the position labelled p in Laka Mugarza (1990)
and Da Cruz (1994). As we saw in chapter 7, this position may be headed by
either the negative marker F or the markers of insistence, in Fongbe and non in
Haitian. With Da Cruz (1994), I will assume that MoodP2 (=p) dominates negp,
headed by the negation markers m in Fongbe and pa in Haitian. As we saw in
chapter 7 (section 7.8), the same relationship holds in Haitian, for, in both
languages, the markers that head p were shown to have scope over the negation
markers. Additional evidence for this claim will be presented in this chapter.
The structure in (2) makes use of the complex vp originally proposed by Larson
(1988). Finally, in (2), the functional category AspP intervenes within the complex vp. The proposal that there is such a functional category projection within
the complex vp has been argued for in recent literature on the basis of data from
various languages (see e.g. Travis 1991a, 1991b, 1992; Lefebvre 1991d; Collins
1993). As shown in (2), by S-structure, the subject of the clause occupies the
Specifier position of tp, and the object occupies the Specifier position of AspP.
Finally, following Chomsky (1989, and related work), I assume that agreement
consists in verifying that the features of two elements that are in a SpecHead
relationship are compatible.
This chapter is organised as follows. Section 8.2 discusses the clausal determiner that has the function of an assertive marker. Section 8.3 discusses the
clausal determiner that has the function of an event determiner. Section 8.4
discusses the forms of the constraint which restricts the co-occurrence of two
contiguous determiners and the variation among speakers with respect to this
constraint.
8.2

The clausal determiner with the function of an assertive marker

In one of its functions, the determiner in Haitian and Fongbe may be used to
assert the content of the proposition, as shown in (1a). In this case, the interpretation of the determiner is discourse-oriented; it relates the content of the
proposition to something that has been said earlier in the conversation. In the
translation, the meaning of the determiner occurring in this function is rendered
by actually. It is argued that, when it has this function, the determiner is a
member of one of the two paradigms of markers that express the speakers point
of view of the content of the proposition. These markers may be seen as mood
markers projecting to MoodP, and thus the determiner is the head of MoodP. On
the basis of Fongbe data, I shall establish the fact that, in this language, there
are two classes of markers indicating the speakers point of view. I will refer to
these two classes of markers as class A and B, respectively. I shall then show
that, as an assertive marker, the determiner in Fongbe grammar 1 has the same

222

The determiner and the structure of the clause

properties as the class A markers. Finally, I shall show that, in one of its functions,
the determiner in Haitian grammar 1 has the same properties as the Fongbe determiner used as an assertive marker. Thus, in both languages, the determiner may
serve as an assertive marker within the clause. As such, it is the head of MoodP1.
The Fongbe markers which indicate the speakers point of view of the proposition (see chapter 7) constitute two classes which I will refer to as classes A
and B.
(3)

a. Class A markers
: question marker used in yes/no questions
l: marker used to express surprise
b. Class B markers
F: negative marker
: marker of insistence

These markers are illustrated in (4)(6).

(4)

Kwk y
?
Koku leave q
Did Koku leave?

(5)

Kwk y
Koku leave
Koku left!

fongbe
(=(6) in Lefebvre to appear b)

l!
surp

fongbe
(=(7) in Lefebvre to appear b)

(6)

Kwk y
f.
Koku left
Neg
Koku did not leave.

fongbe
(=(8) in Lefebvre to appear b)

The class A markers are mutually exclusive. For example, the question marker
cannot appear within the same clause as the surprise marker. The content of a
proposition cannot be the object of questioning at the same time as it provokes
surprise.
(7)

*Kwk
Koku

y
l

leave surp q

fongbe
(=(9) in Lefebvre to appear b)

A class A marker may, however, occur within the same clause as a class B
marker. This is illustrated in (8) where the question marker occurs in the same
clause as the negative marker.
(8)

Kwk xw
svn lx f
Koku buy crab pl Neg
Didnt Koku buy the crabs?

wy
?
Det3 q

fongbe
(=(12a) in Da Cruz 1994)

The fact that a class A marker and a class B marker can appear within the same
clause argues that these markers head two different positions.

8.2 Clausal determiner as assertive marker

223

As is noted in Da Cruz (1994), when markers of both classes appear within

the same clause, the class A marker always has scope over the class B marker.
This is shown in (8), in which the question marker occurs to the right of the
negative marker. Since functional category projections are head-final in Fongbe,
the rightmost functional head is highest in the syntactic tree. Sentence (9), in
which the question marker precedes the negative marker, is ungrammatical.
(9)

*Kwk
Koku

xw
svn lx
buy crab pl

wy f
q fc Neg

fongbe
(=(13b) in Da Cruz 1994)

Class A markers can only occur in root clauses (see (4)(6) ). They do not
occur in embedded clauses, as shown by the ungrammaticality of (10).
(10)

*N knbyw 3w
I ask

[Kwk y
Koku leave

]
q

fongbe
(=(12) in Lefebvre to appear b)

By contrast, class B markers can occur in embedded clauses, as shown in (11).4

(11)

N 3w [Kwk m y
f].
I say Koku not leave Neg
I said that Koku did not leave.

fongbe
(=(13) in Lefebvre to appear b)

The fact that the class A and class B markers have different distributional
properties (see (7) versus (8) and (10) versus (11) ) supports the claim that these
markers divide into two classes. The fact that class A markers can occur within
a clause containing a class B marker (see (8) ) shows that the two classes head
two different positions. These two positions are MoodP1 (=cp) and MoodP2 (=p),
respectively (see (2) ). Mood expresses the speakers attitude towards the content
of an utterance. For example, the proposition may be considered as something
to be wished for, or something to be done. Mood markers are linguistic forms
used to express these attitudes (see Roberts 1954; Crystal 1991). The class A
and B markers can be seen as Mood markers, since they convey the speakers
attitude about the proposition. Since the class A markers must have scope over
the negative marker (see (8) and (9) ), they must head the highest MoodP in the
tree. These relationships are represented in (2).
The class A markers have two additional major properties. First, they may
occur with verbs of all aspectual classes, as is exemplified in (12)(14).
(12)

(13)

Dynamic verbs
Kwk kn-wzn
Koku run
Did Koku run?

?
q

Resultative verbs
Kwk mw
jtv v
?
Koku catch-sight-of thief Det q
Has Koku caught sight of the thief?

fongbe

fongbe

224
(14)

The determiner and the structure of the clause

Stative verbs
Kwk s
flnsxgb ?
Koku know French
q
Does Koku know French?

fongbe

Second, in clauses containing a class A marker, the subject of the clause may
contain the [+definite] determiner V, as in (15), or the indefinite marker #, as
in (16).
(15)

Sn v
w
/ l ?/!
man Det arrive q / surp
The man arrived ?/!

fongbe

(16)

M6 3 w
/ l ?/!
someone arrive q / surp
Someone arrived ?/!

fongbe

The fact that class A markers can only occur in root sentences and there can
be only one per sentence suggests an analysis along the line of that proposed in
Lefebvre and Muysken (1988) on the basis of similar data in Quechua: at Logical
Form, the class A markers function as performative verbs.5 Independent evidence
for this proposal comes from the fact that Fongbe does not have simple performative verbs meaning to question or to be surprised at. The semantics of these
markers is compatible with verbs of all aspectual classes (see (12)(14) ).
In both Fongbe and Haitian grammar 1, the determiner used as an assertive
marker has the same distribution as class A markers. This argues that the determiner with this function is a class A marker.
First, the assertive marker determiner may occur with verbs of all aspectual
classes (see (17)(19) ), just as the other class A markers do (see (12)(14) ).
(17)

Dynamic verb
a. Jan
rive
Ptoprens
la.
John arrive Port-au-Prince Det
Actually, John arrived in Port-au-Prince.

haitian1
(=(12) in Lefebvre 1996a)

b. Kwk w
ktvn
v.
Koku arrive Cotonou Det
Actually, Koku arrived in Cotonou.

fongbe1
(=(11) in Lefebvre 1996a)

(18)

Resultative verb
a. Jan
w
vl a.
John catch-sight-of thief Det
Actually, John caught sight of a thief.

haitian1
(=(15b) in Lefebvre 1996a)

b. Kwk mw
jtv lx v.
fongbe1
Koku catch-sight-of thief pl Det
Actually, Koku has caught sight of the thieves.
(=(15b) in Lefebvre 1996a)

8.2 Clausal determiner as assertive marker

(19)

Stative verb
a. Jan
knnn frans
a.
Kwk s
flnsygb v.
J / K know
French
Det
Actually, John/Koku knows French.

225

haitian1
fongbe1
(=(15c) in Lefebvre 1996a)

b. Rb la
blan
vw v
w
dress det white
Actually, the dress

an.
v.
Det
is white.

haitian1
fongbe1
(=(15d) in Lefebvre 1996a)

Second, when used in this way, the determiner is not compatible with the
other class A markers. The following examples show that the determiner cannot
co-occur with the direct yesno question marker.6
(20)

*ske [Jan w
vl] a.
haitian1
q
John see
thief Det
[Lit.: Has John actually caught sight of a thief?]
(=(34) in Lefebvre to appear b)

(21)

*My 3 w
v

someone arrive Det q

[Lit.: Has someone actually arrived?]

fongbe1
(=(23) in Lefebvre to appear b)

We saw that the other markers of class A are also mutually exclusive.
Third, like the other class A markers, the Fongbe determiner may occur within
the same clause as the negative marker.
(22)

Sn v
w
f
v.
man det arrive Neg Det
Actually, the man did not arrive.

fongbe1
(=(24) in Lefebvre to appear b)

As an assertive marker, the determiner must follow the negative marker. When
the determiner precedes the negative marker, it cannot be interpreted as an assertive marker, as shown in (23). (For a discussion of the meaning of the determiner
in the context of (23), see section 8.3.)
(23)

Sn v
w
v
f.
fongbe1
man det arrive Det Neg
#It is not the case that the man actually arrived.
(=(25) in Lefebvre to appear b)

These facts parallel those involving the other class A markers in the context of
the Fongbe negative marker (see (9) ). These data show that, as a member of
class A, the determiner must have scope over the negative marker. As we saw in
chapter 7, Haitian does not have a negative marker corresponding to Fongbe F.

226

The determiner and the structure of the clause

In a negative clause, however, the assertive marker must have scope over the
negation marker pa. This is shown by the semantic interpretation data in (24).
The assertive marker may have scope over the negation marker, as in (24a). The
negation marker may not have scope over the assertive marker, as shown by the
unacceptability of the interpretation in (24b).
(24)

Mounn
man

nan pa
rive
det neg arrive

a.
Det

haitian1

a. Actually, the man did not arrive.

b. #It is not the case that the man actually arrived.
(=(35) in Lefebvre to appear b)

Fourth, like the other markers in class A, the assertive marker determiner
only occurs in root sentences (see (10) ). It is not permitted in embedded clauses,
as shown below.
(25)

*M knnn [rb la
blan
an]
haitian1
I
know
dress det white Det
[Lit.: I know that actually the dress is white.]
(=(37) in Lefebvre to appear b)

(26)

*N s
[vv v
w
v]
fongbe1
I know dress det white Det
[Lit.: I know that actually the dress is white.]
(=(27) in Lefebvre to appear b)

Finally, the subject of a clause containing the assertive marker may contain
either the [+definite] determiner or an indefinite term. This is shown in (27) and
(28), respectively, which parallel the data in (15) and (16).
(27)

Mounn nan
Sn
v
man
det
Actually, the

rive
a.
w
v.
arrive Det
man arrived.

haitian1
fongbe1
(=(38) in Lefebvre to appear b)

(28)

Yon

mounn
rive
a.
Sn
3 w
v.
a
man
a
arrive Det
Actually, a man arrived.

haitian1
fongbe1
(=(39) in Lefebvre to appear b)

The data presented in this section show that the Haitian and Fongbe determiners can function as assertive markers. With this function, in grammar 1, they
have the same properties as the class A markers, which express the speakers
point of view of the proposition. Thus, in one of its functions, the determiner is
a member of the paradigm of class A markers. As such, it heads MoodP1 (see
(2) ). The fact that the assertive marker determiner can only occur in root clauses
follows from the analysis that, at Logical Form, the determiner in this position is

8.2 Clausal determiner as assertive marker

227

interpreted as a performative verb (meaning to assert) just like the other class
A markers. Neither Haitian nor Fongbe has any other performative verb with this
meaning. Assuming SpecHead agreement, the fact that the [+definite] determiner
heading MoodP1 is insensitive to the feature [+/definite] of the subject dp of
the clause (see (27), (28) ) implies that the subject of the clause is not in Spec of
MoodP1 at S-structure. If it were, we would expect the [+definite] determiner in
the head of MoodP1 to allow only [+definite] elements in its specifier.
There is a subset of speakers of both Haitian and Fongbe whose grammar
differs from grammar 1 with respect to the distribution of the determiner and the
restrictions governing its co-occurrence with other markers. Below, I present the
data distinguishing grammar 2 from grammar 1 and provide an account of this
variation. Two properties distinguish grammar 2 from grammar 1 with respect to
the determiner used as an assertive marker. First, we saw that, in grammar 1, the
assertive marker may occur in a negative clause. This is shown in (29).
(29)

Jan
pa
rive
a.
Jan
w
f
v.
John neg arrive Neg Det
Actually, John did not arrive.

haitian1
fongbe1

In grammar 2, the assertive marker cannot occur in a negative sentence, as

shown in (30).
(30)

*Jan
pa
rive
a.
*Jan
w
f
v
John neg arrive Neg Det
[Lit.: Actually, John did not arrive.]

haitian2
fongbe2

Second, grammar 1 allows an indefinite subject dp to occur in a clause containing the assertive marker. This is shown in (31).
(31)

Yon mounn
rive
a.
Sn
3 w
v.
a
man
a
arrive Det
Actually, a man arrived.

haitian1
fongbe1

By contrast, in grammar 2, this structure is not grammatical, as shown in (32).

(32)

*Yon
*
a
[Lit.:

mounn
rive
a.
Sn
3 w
v
man
a arrive Det
Actually, a man arrived.]

haitian2
fongbe2

It has been shown that, in grammar 1, the determiner used as an assertive

marker is the head of MoodP1. To account for the differences between the two
grammars, Lefebvre (to appear b) proposes that, in grammar 2, when the determiner functions as an assertive marker, it is the head of MoodP2, rather than
MoodP1. This is shown in (33) (=(77) in Lefebvre to appear b).

228
(33)

The determiner and the structure of the clause

a. Grammar 1

b. Grammar 2

MoodP1

MoodP1

Mood

Mood

Mood
Assertive marker

tp

dpi

Mood

[e]i
vp

Mood

Mood
vp

Mood
Assertive marker
v

[e]i
AspP

MoodP2
[e]i

[e]i

dpi
t

MoodP2

Mood

tp

AspP

This analysis allows for a straightforward account of the differences between

the two grammars. As head of MoodP2, the assertive marker is a member of the
class B markers. In this position, it is in a paradigmatic relationship with the
negative marker (F in Fongbe, which has no overt counterpart in Haitian). It
follows that, in grammar 2, the determiner used as an assertive marker will not
be able to occur in a clause containing a negative marker (see (30) ).
From the analysis that the assertive marker determiner is the head of MoodP2
in grammar 2, we can also derive the fact that an indefinite subject dp cannot cooccur with it in this grammar, as in (32). By S-structure the subject will have
moved to Spec of tp through Spec of MoodP2, as shown in (33b). A [definite]
subject which would have moved through Spec of MoodP2 is not compatible
with the [+definite] feature of the head of MoodP2 (the determiner). An indefinite subject is thus not allowed to occur in the context of the assertive marker
determiner in the head of MoodP2 (see (32) ). The situation is different in grammar 1, as shown in (33a). In this case, the subject of the clause is never in a
SpecHead relationship with the assertive marker, which is the head of MoodP1.
In this grammar, the subject of the clause containing the determiner used as an
assertive marker is allowed to be indefinite (see (31) ).
The distribution of the assertive marker determiner in grammar 2 has some
properties in common with grammar 1. For example, in both grammar 2 and
grammar 1, the determiner with the function of an assertive marker is mutually
exclusive with the yesno question marker (see (20) and (21) ), and excluded
from subordinate clauses (see (25), (26) ). To account for these parallels, Lefebvre
(to appear b) proposes that in grammar 2 the assertive marker moves to the head

8.3 Clausal determiner as event determiner

229

of MoodP1 at Logical Form. Thus, in grammar 2, the assertive marker determiner occupies both MoodP1 and MoodP2, at some point in the derivation; this
explains its hybrid behaviour.
8.3

The clausal determiner with the function of an event determiner

This section considers the distribution of the determiner in its function as an

event determiner. In this case, the [+definite] determiner means this event in
question/this event that we know of. In translation, the meaning of the determiner
with this function is rendered as as we knew/as expected, as shown in (34).
(34)

Mounn nan
Sn
v
man
det
The man has

kraze
yon manchinn
nan.
haitian1
gb
mvt
3 v.
fongbe1
destroy a
car
a
Det
destroyed a car, as we knew/as expected.
(=(33a) in Lefebvre 1992b)

On the basis of agreement, interpretation and word order facts, it is argued that
the determiner with this function may head two different positions within the
clause: tp and AspP, as shown in (2). As head of tp, the determiner is in a Spec
Head relationship with the subject of the clause. As head of AspP, the determiner is in a SpecHead relationship with the affected object of the clause. I
then discuss the evidence showing that, as head of tp or AspP, the determiner
determines the event denoted by the clause. These two positions reflect in the
syntax the structure of the event denoted by the clause. Third, it will be shown
that the presence of the event determiner within the clause is related to the
clauses temporal and aspectual properties. As head of tp, the determiner is
compatible with [+perfect]. As head of AspP, it is compatible with [+perfective].
Finally, I discuss the differences between grammar 1 and grammar 2 with respect
to the distribution of the event determiner.
Three types of arguments support the claim that there are two positions for
the event determiner within the clause: agreement facts, semantic interpretation
facts and word order facts (see Lefebvre 1992b). One striking property of the
event determiner is that it defines two agreement domains. In a clause where
both subject and object dps contain an indefinite term, yon in Haitian and # in
Fongbe, the event determiner is not permitted, as shown by the ungrammaticality
of (35).
(35)

*Yon
*
a
[Lit.:

mounn
kraze
yon manchinn
nan
haitian1
Sn
3 gb
mvt
3 v
fongbe1
man
a
destroy a
car
a
Det
A man destroyed a car, as we knew.]
(=(33b) in Lefebvre 1992b)

The event determiner is permitted only when either the subject dp or the object
dp (or both) contains the [+definite] determiner, la (or its allomorphs) in Haitian,
and V (or its allomorphs) in Fongbe. The sentences in (36) and (37), as compared

230

The determiner and the structure of the clause

with that in (35), show that, when either the subject or the object dp is [+definite],
the event determiner is permitted.
(36)

Mounn nan
Sn
v
man
det
The man has

(37)

Yon

kraze
yon manchinn
nan.
haitian1
gb
mvt
3 v.
fongbe1
destroy a
car
a
Det
destroyed a car, as we knew.
(=(33d) in Lefebvre 1992b)

mounn
kraze
manchinn nan an.
haitian1
Sn
3 gb
mvt
v
v.
fongbe1
a
man
a
destroy car
det Det
A man has destroyed the car, as we knew.
(=(34a) in Lefebvre 1992b)

One additional fact is pertinent to the distribution of the event determiner in the
context of a [+definite] object dp. The sentences in (38) contain a [+definite]
object dp and yet the event determiner is not permitted, as shown by the ungrammaticality of these sentences.
(38)

a. *Yon
*
a
[Lit.:

mounn
w manchinn nan an
haitian1
Sn
3 mw mvt
v
v
fongbe1
man
a
see car
det Det
A man saw the car, as we knew.]
(=(13a) in Lefebvre 1992b)

b. *Yon
*
a
[Lit.:

mounn
konpran
frans
a
haitian1
Sn
3 s
flnsygb v
fongbe1
man
a understand French
Det
A man understood French, as we knew.]
(=(13b) in Lefebvre 1992b)

Crucially, the object of the verb destroy in (37) is an affected object (in the
sense of Tenny 1987). By contrast, the objects of the verbs see and understand in (38) are not affected objects. The contrast in grammaticality between
(37) and (38) shows that the determiner is permitted only in the context of a
[+definite] affected object.
These facts show that the event determiner agrees in the feature [+definite]
with either the subject or the affected object of the clause. It thus defines two
agreement domains. Agreement proceeds within the configuration SpecHead.
At S-structure the subject is in Spec of tp (see (2) ). This suggests that the head
of tp is a position for the event determiner. Similarly, at S-structure the affected
object is in Spec of AspP (see (2) ). This suggests that the head of AspP is another
position for the event determiner. As head of tp, the determiner agrees in the
feature [+definite] with the subject of the clause. As head of AspP, it agrees in
the feature [+definite] with the affected object of the clause (for a formal account
of these agreement facts, see Lefebvre 1992b).
When the determiner agrees with the subject of the clause, and thus occurs as
head of tp, the whole event is understood as part of shared knowledge, as shown
in (39).

8.3 Clausal determiner as event determiner

(39)

Mounn nan
Sn
v
man
det
The man has

kraze
yon manchinn
nan.
gb
mvt
3 v.
destroy a
car
a
Det
destroyed a car, as we knew/expected he would.

231

haitian1
fongbe1

This follows from the fact that, as head of tp, the determiner can have scope
over the whole event. Furthermore, in this case, the interpretation of the determiner is subject-oriented. The sentence in (39) could also be translated as The
man destroyed a car, as he was expected to. When the determiner agrees with
the affected object of the clause, only a subpart of the event is understood as
being part of shared knowledge, as shown in (40).
(40)

Yon mounn
kraze
manchinn nan an.
haitian1
Sn
3 gb
mvt
v
v.
fongbe1
a
man
a
destroy car
det Det
A man has destroyed the car, as we knew it would be destroyed.

As head of AspP, the determiner has narrower scope than when it heads tp. In
this context, the event determiner is object-oriented, as shown by the interpretation assigned to (40).
Fongbe data involving the distribution of the event determiner in the context of the negative marker provide further evidence that it has two syntactic
positions. Indeed, in this context, the two positions for the event determiner are
directly visible in the distribution of functional lexical items at surface structure.
The determiner may follow the negative marker, as in (41), or precede it, as in
(42).
(41)

Sn v
f n
mvt v
f
v.
fongbe1
man det steal car
det Neg Det
The man has not stolen the car, as it was expected he would not.
(=(17) in Lefebvre 1992b)

(42)

Sn v
f n
mvt v
v
f.
fongbe1
man det steal car
det Det Neg
The man has not stolen the car, as it was expected it would be.
(=(18) in Lefebvre 1992b)

The determiner can only follow the negative marker when the subject dp is
[+definite], as is shown by the contrast in grammaticality between (43a) and (43b)
(=(41) ). Furthermore, in this case, the determiner has scope over the negative
marker, as can be seen in the translation of (43b).
(43)

a. *Sn 3 f n
mvt v
f
v
fongbe1
man a
steal car
det Neg Det
[Lit.: A man has not stolen the car, as expected.]
(=(19a) in Lefebvre 1992b)
b. Sn v
f n
mvt v
f
v.
fongbe1
man det steal car
det Neg Det
The man has not stolen the car, as it was expected he would not.
(=(19b) in Lefebvre 1992b)

232

The determiner and the structure of the clause

The determiner can only precede the negative marker in the context of a [+definite] affected object, as shown by the contrast in grammaticality between (44a)
and (44b) (=(42) ). Furthermore, in this position, the determiner does not have
scope over the negative marker, as can be seen in the translation of (44b).
(44)

a. *Sn 3 f n
mvt 3 v
f
fongbe1
man a
steal car
a Det Neg
[Lit.: A man has not stolen a car, as expected.]
(=(21a) in Lefebvre 1992b)
b. Sn 3 f n
mvt v
v
f.
fongbe1
man a
steal car
det Det Neg
A man has not stolen the car, as it was expected it would be.
(=(21b) in Lefebvre 1992b)

In this position, the determiner cannot occur in the context of a non-affected

object dp, even if the latter is [+definite].
(45)

*Sn 3 mw mvt v
v
f
man a
see car
det Det Neg
[Lit.: A man did not see the car, as we knew.]

fongbe1
(=(22) in Lefebvre 1992b)

When the determiner follows the negative marker, it requires the external
argument of the clause to be [+definite]. Thus, it is the head of tp (see (2) ).
When the determiner precedes the negative marker, the affected argument must
be [+definite]. In this case, the determiner is the head of AspP. The surface
distribution of the determiner with respect to the negative marker thus further
supports the claim that there are two independently motivated positions for the
event determiner.
In Haitian, there is no negative marker like that in Fongbe, so the distribution
discussed above is not directly visible. However, the data presented below show
that, in the context of negation, the same semantic interpretations as in Fongbe
prevail. Consider the negative sentence in (46), which contains a [+definite] subject dp, a [+definite] affected object and an event determiner. In line with the above
discussion, the determiner in (46) can head either tp or AspP. It is permitted in
both positions since the two dps (subject and affected object) that are pertinent
to agreement with the event determiner are both [+definite]. Therefore, (46) should
be ambiguous with respect to what is presupposed as part of shared knowledge.
It should be possible to assign the determiner in (46) either a wide scope interpretation or a narrow scope interpretation. This is exactly what we find. The
clause in (46) may be assigned an interpretation in which the determiner has
scope over negation, (46a), or an interpretation in which it does not, (46b).
(46)

Ng
man

la
pa vle manchinn
det not steal car

nan an.
det Det

haitian1

a. The man has not stolen the car, as it was expected he would not.
b. The man has not stolen the car, as it was expected it would be.
(=(28) in Lefebvre 1992b)

8.3 Clausal determiner as event determiner

233

In the context of an indefinite subject dp, as in (47), the (a) interpretation of (46)
does not obtain.
(47)

Yon ng pa
a
man not

vle
steal

manchinn nan an.

car
det Det

haitian1

#A man did not steal the car, as it was expected he would not.
(=(29) in Lefebvre 1992b)

In the context of an indefinite affected object dp, as in (48), the (b) interpretation
is not possible.
(48)

Ng la
pa vle
man det not steal

yon
a

manchinn nan.
car
Det

haitian1

#The man did not steal a car, as it was expected a car would be.
(=(30) in Lefebvre 1992b)

These facts follow from the analysis that there are two syntactic positions for the
event determiner. The (a) interpretation of (46) arises from the structure where
the event determiner heads tp, and the (b) interpretation from the structure where
it heads AspP.
Thus, agreement facts, semantic interpretation facts, and word order facts all
support the claim that there are two positions for the event determiner. I argue
below that, as head of tp or AspP, the determiner determines the event denoted
by the clause. These two positions and their projections reflect in the syntax the
structure of the event denoted by the clause.
Van Voorst (1988) proposed that the subject of the clause marks the object of
origin of an event and the object of the clause marks the object of termination of
an event. The event determiner focuses on the event denoted by the clause either
from the point of view of the subject, the beginning point of the event (e.g.
agreement with the subject (see (36) ), or from the point of view of the object,
the end-point of the event (e.g. agreement with the object (see (37) ). Moreover,
we saw that, as head of AspP, the event determiner is permitted only in the
context of an affected object (see (37) ), and not a non-affected object (see (38) ).
Affectedness of the object is an eventive aspectual property. Affected objects
delimit the event denoted by the verb, in that they impose an end-point on it (see
Tenny 1987). By contrast, non-affected objects do not delimit the event denoted
by the verb. The possibility of the event determiner occurring as head of AspP
thus correlates with the [+definite] property of the argument that imposes an
end-point on the event denoted by the verb. The event determiner is not permitted in the context of non-affected objects because they do not impose an endpoint on the event. Non-affected objects cannot be in a Spec-Head relationship
with an event determiner. Agreement facts involving the subject/affected object
and the event determiner thus straightforwardly reflect the structure of the event
denoted by the clause.
As was mentioned earlier, both Haitian and Fongbe distinguish among three
aspectual classes of verbs. Dynamic verbs describe a situation corresponding to

234

The determiner and the structure of the clause

a process that can be perceived as either ongoing or terminated, for example,

manje in Haitian and #L in Fongbe meaning to eat. Resultative verbs describe
the result of some process, for example, w in Haitian and mW in Fongbe meaning to catch sight of. Stative verbs do not refer to a process at all; these are
verbs like knnn in Haitian and tn in Fongbe meaning to know. While the
event determiner does occur with dynamic verbs, as shown by the numerous
examples provided above, it does not occur with resultative or stative verbs.
This is shown in (49) and (50), respectively.
(49)

*Mounn nan w
*Sn
v
mw
man
det see
[Lit.: The man saw

vl a
jtv v
thief Det
thieves, as expected.]

haitian1
fongbe1
(=(56) in Lefebvre to appear b)

(50)

*Jan
*Jan
John
[Lit.:

knnn frans
tn
flnsygb
know
French
John knows French,

a
v
Det
as expected.]

haitian1
fongbe1
(=(57) in Lefebvre to appear b)

In both Haitian and Fongbe there are verbs which are ambiguous between a stative
and a dynamic interpretation; for example, blan in Haitian and w in Fongbe.
Indeed, these verbs may be assigned either a stative interpretation, to be white,
or a dynamic interpretation, to become white. This is exemplified in (51).
(51)

Rob
Avw
dress

la
v
det

blan.
w.
white

haitian1
fongbe1

a. The dress is white.

b. The dress has become white.
(=(58) in Lefebvre to appear b)

When the event determiner occurs in a clause containing such verbs, only the
dynamic interpretation is available, as shown in (52).
(52)

Rob la
Avw v
dress det

blan
an.
w
v.
white Det

haitian1
fongbe1

a. #The dress is white, as expected.

b. The dress has become white, as we knew/expected it would.
(=(59) in Lefebvre to appear b)

Why is the event determiner restricted to clauses that contain a dynamic verb?
The determiner requires that its complement be an identifiable object. Events
can be identifiable objects, but only if they are complete; that is, if they have a
beginning and an end-point. Dynamic verbs are the only ones that denote an
event that has a beginning and an end-point. This explains why the event determiner only occurs in clauses with a dynamic verb.

8.3 Clausal determiner as event determiner

235

Aspects are different ways of viewing the internal temporal constituency of a

situation. We saw that as head of AspP, the event determiner is only permitted
in the context of an affected object (see (36), (38) ). Affected objects have the
aspectual property of imposing an end-point on the event denoted by the verb,
and hence making it complete. There is yet another way in which the event
determiner is related to aspect: it cannot occur in clauses containing an overt
expression of non-complete aspect. Thus the event determiner is incompatible
with forms expressing the imperfective aspect. This is shown in (53).
(53)

a. *Jan
ap rive
a
John imp arrive Det
[Lit.: John is arriving, as we knew.]

haitian1

b. *Jan
3 ww
wy
v
John at arriving post Det
[Lit.: John is arriving, as we knew.]

fongbe1
(=(60) in Lefebvre to appear b)

The imperfective aspect indicates that an event that has begun has not come
to completion. In a clause containing the imperfective aspect, the event cannot
be perceived as complete, since it does not have an end-point. Consequently, in
this case, the event determiner may not appear as the head of AspP. The event
determiner is also incompatible with forms expressing the prospective aspect
(see Lefebvre 1996b). This is shown in (54).
(54)

a. *Jan
apral rive
a
John prosp arrive Det
[Lit.: John is about to arrive, as we knew.]

haitian1

b. *Jan
3 n
ww
wy
v
fongbe1
John at def-fut arriving post Det
[Lit.: John is about to arrive, as we knew.]
(=(61) in Lefebvre to appear b)

The prospective aspect indicates that an event is about to take place. It therefore
presents the event from the perspective of its beginning point. With this aspect,
an event cannot be perceived as being complete. Consequently, in this context,
the event determiner cannot occur as head of AspP.
The event determiner occurs in tensed clauses, as shown by the numerous
examples presented so far. It does not, however, occur in tenseless/infinitival
clauses. This fact can be observed in its distribution in embedded clauses. In
(55), the sentential complement of the verbs meaning to want is not tensed and
the determiner is not permitted.
(55)

*Yon
*
a
[Lit.:

mounn
vle
[
Sn
3 jl
[n
man
a want irr
A man wanted to destroy

kraze
manchinn nan
gb
mvt
v
destroy car
det
the car, as expected.]
(=(37) in

an] haitian1
v]
fongbe1
Det
Lefebvre 1992b)

236

The determiner and the structure of the clause

By contrast, in (56) the sentential complement of the verbs meaning to want is

tensed and the determiner is permitted. Note that the event determiner in this
example occurs in a clause containing an overt complementiser.
(56)

Yon

mounn
vle
Sn
3 jl
a
man
a
want
manchinn nan an].
mvt
v
v].
car
det Det

[pou Jan
[n
Kwk
comp J / K

kraze
n
gb
sub destroy

haitian1
fongbe1

a. A man wanted J/K to destroy the car, as it was expected he would.

b. A man wanted J/K to destroy the car, as it was expected the car would
be
destroyed.
(=(38) in Lefebvre 1992b)

These data show that the presence of an event determiner within the clause is
correlated with tense. This is not surprising, given the relationship that exists
between the definiteness of a finite clause and the definiteness of the determiner
(see Lefebvre and Massam 1988). Furthermore, there is a relationship between
determinateness and referentiality only referential expressions can be determined.
Finite clauses are temporal R-expressions (see Chomsky 1981). This property
allows them to be determined by an event determiner (see (56) ). Conversely,
non-finite clauses are not referential, so they cannot be so determined (see (55) )
(see Lefebvre 1992b).
Since the event determiner determines the event denoted by the clause, it
follows that there will be only one event determiner per proposition, which is
indeed the case (see Lefebvre and Massam 1988). The sentence in (57) is ungrammatical because the two positions which can host the event determiner are
both lexically filled.
(57)

*Mounn nan pa vle manchinn nan

*Sn
v
fn
mvt
v
man
det not steal car
det
[Lit.: The man did not steal the car, as it was
was expected the car would be stolen.]

an
v
f
Det Neg
expected he

an
haitian1
v
fongbe1
Det
would not/as it

(=(39) in Lefebvre 1992b)

As head of tp or of AspP, the determiner determines the event denoted by the

clause. It can only occur in clauses that denote a complete event; that is, in
tensed clauses containing a dynamic verb and no overt expression of a noncomplete aspect. Since the determiner in head of tp or AspP is an event determiner, it follows that there can be only one per proposition: the event itself is the
object of the determiner with this function. The event determiner heads either tp
or AspP, thus focussing the event either from the perspective of its beginning
point (see agreement with the subject) or of its end-point (see agreement with
the affected object).

8.3 Clausal determiner as event determiner

237

There is a relationship between the position of the event determiner, and

[+perspective] aspect and [+perfect] tense. Following Comrie (1976: 12), I will
assume that perfect refers to a past situation which has present relevance, for
instance the present result of a past event. For example, the sentence John has
broken his arm expresses the present result (John has a broken arm) of the past
event (at some point in the past, John broke his arm). The past tense differs from
the perfect in that it refers to a past situation that does not necessarily have present relevance. The sentence John broke his arm does not entail that Johns arm
is still broken. In the following discussion, I will assume that perfect is a type
of tense. I will further assume that perfective is a type of aspect. This aspect
contrasts with imperfective and denotes a situation viewed in its entirety
(Comrie 1976: 12). There are languages where both of these oppositions, perfect /
non-perfect and perfective/imperfective are grammaticalised; West African languages are among them (see Welmers 1973). As head of AspP, the event determiner is compatible with the [+perfective] aspect. As head of tp, it is compatible
with [+perfect] tense.
We saw that the event determiner cannot occur in clauses containing an overt
expression of non-complete aspect. Thus, it is incompatible with forms expressing the progressive/imperfective (see (53) ) and prospective (see (54) ) aspects.
As head of AspP, the determiner is compatible only with the [+perfective]
aspect.
As head of tp, the event determiner is compatible only with [+perfect]. Facts
supporting this claim come from the temporal interpretation of bare sentences
and of sentences containing an irrealis marker in the context of the event determiner. Recall from chapter 5 that bare sentences are sentences with no preverbal
markers expressing relative Tense, Mood and Aspect. The temporal interpretation of bare sentences has been shown to be computed from the various components that participate in establishing the aspectual properties of a clause, including
the aspectual class of the verb. For example, in both Haitian and Fongbe, a bare
sentence containing a dynamic verb and a generic object is interpreted as past, as
shown in (58).
(58)

Mari
Mari
Mary
Mary

prepare
3
prepare
prepared

pat.
wh.
dough
dough.

haitian
fongbe
(=(65) in Lefebvre to appear b)

However, with a [+definite] affected object, the sentence can only be interpreted
as perfect, as in (59).
(59)

Mari
Mari
Mary
Mary

prepare pat
la.
3
wh
h.
prepare dough det
has prepared the dough.

haitian
fongbe
(=(66) in Lefebvre to appear b)

238

The determiner and the structure of the clause

A clause containing the event determiner as head of tp is always interpreted as

[+perfect], as shown in (60).
(60)

Mari
Mari
Mary
Mary
#Mary

prepare pat
la
a.
haitian1
3
wh
v
v.
fongbe1
prepare dough det Det
has prepared the dough, as we knew/as expected.
prepared the dough, as we knew/as expected.
(=(67) in Lefebvre to appear b)

A clause containing an ergative verb is ambiguous between a past and a perfect

interpretation, as shown in (61).
(61)

Mari
Mari
Mary
Mary

pati.
y.
leave
left. or Mary has left.

haitian1
fongbe1
(=(68) in Lefebvre to appear b)

In this context, the presence of the event determiner as head of tp forces a

perfect interpretation of the clause, as shown in (62).
(62)

Mari
Mari
Mary
Mary
#Mary

pati
a.
y
v.
leave Det
has left, as we knew.
left, as we knew.

haitian1
fongbe1

(=(69) in Lefebvre to appear b)

This shows that, as head of tp, the event determiner is compatible only with the
[+perfect] tense.
The temporal interpretation of a clause containing an irrealis marker in the
context of the event determiner constitutes yet another set of facts showing that,
as head of tp, the event determiner is only compatible with the [+perfect] tense.
As shown in chapter 5, the irrealis preverbal markers ap in Haitian and n in
Fongbe situate, with respect to the moment of speech, an event that is expected
to definitely take place in the near future. The event may coincide with the point
of reference, yielding a simple future, and thus a [perfect] interpretation of the
clause, as in (63a), or it may precede the point of reference, yielding a future
perfect interpretation of the clause, as in (63b).
(63)

Mari ap prepare
Mari n 3
Mary irr prepare

pat
la.
wh
v.
dough Det

haitian1
fongbe1

a. Mary will prepare the dough.

b. Mary will have prepared the dough.
(=(70) in Lefebvre to appear b)

A clause where the event determiner as head of tp co-occurs with an irrealis

marker is obligatorily interpreted as future perfect, as shown in (64).

8.3 Clausal determiner as event determiner

(64)

Mari
Mari
Mary
a. Mary
b. #Mary

239

ap prepare pat
la
a.
haitian1
n 3
wh
v
v.
fongbe1
irr prepare dough det Det
will have prepared the dough, as expected.
will prepare the dough, as expected.
(=(71) in Lefebvre to appear b)

This is further support for the claim that, as head of tp, the event determiner is
only compatible with [+perfect].
As we saw above, in grammar 1, there are two surface positions for the event
determiner: AspP and tp. Agreement facts from both Haitian and Fongbe were
shown to support this claim. Moreover, we saw that, in Fongbe grammar 1,
these two positions are directly reflected in word order, in the context of the
negative marker: as head of AspP, the event determiner precedes the negative
marker, but as head of tp, it follows it, as shown in (65).
(65)

a. Jan
w
v
f.
John arrive Det Neg
John has not arrived, as it was expected he would.

fongbe1

b. Jan
w
f
v.
fongbe1
John arrive Neg Det
John has not arrived, as it was expected he would not.
(=(78) in Lefebvre to appear b)

In Fongbe grammar 2, the event determiner may not follow the negative
marker but it may precede it, as shown in (66).
(66)

a. *Jan
Jan

w
f
v
arrive Neg Det

fongbe2

b. Jan
John

w
v
f.
arrive Det Neg

fongbe2

a.

John has not arrived, as it was expected he would.

b. John has not arrived, as it was expected he would not.

(=(79) in Lefebvre to appear b)

The ungrammaticality of (66a) shows that, in grammar 2, the head of tp is not

an available surface position for the event determiner. However, the data related
to semantic interpretation in this grammar are most interesting. The sentence in
(66b) may be assigned two interpretations. While the (a) interpretation corresponds to the determiner as head of AspP, the (b) interpretation is what we would
expect if the determiner were head of tp. Therefore, Lefebvre (to appear b)
assumes that the event determiner moves to tp at Logical Form. Thus, although
tp is not an available surface position for the event determiner in grammar 2
(see (66a) ), it is available at Logical Form (see (66b) ).
Since Haitian has no overt negative marker, Haitian data corresponding to the
Fongbe data in (66) are not available. However, speakers of Haitian grammar 2
do have two interpretations for the event determiner occurring in the context of
a negative clause, given in (67).

240

The determiner and the structure of the clause

(67)

Jan
John

pa
rive
a.
neg arrive Det

haitian2

a. John has not arrived, as it was expected he would.

b. John has not arrived, as it was expected he would not.
(=(80) in Lefebvre to appear b)

These Haitian interpretive data parallel the Fongbe data in (66b). The difference
between grammar 1 and grammar 2 with respect to the distribution of the event
determiner thus reduces to whether tp is an available surface position for the
determiner.
In this section, it has been argued that, in one of its functions, the determiner in Haitian and Fongbe is an event determiner. Its distribution within the
clause reflects the structure of the event denoted by the clause. On the basis of
agreement, semantic interpretation and word order facts, it has been argued
that the event determiner may fill two syntactic positions within the clause: head
of tp and head of AspP. As head of AspP, the determiner is compatible with
the [+perfective] aspect; as head of tp, it is compatible with [+perfect] tense.
Finally, the major difference between grammar 1 and grammar 2 with respect
to the distribution of the event determiner reduces to whether tp is an available
surface position for it.
8.4

The *Det Det Filter in grammar 1 and grammar 2

Grammar 1 and grammar 2 differ in the fact that they manifest different cooccurrence restrictions. Speakers of grammar 1 allow for two consecutive determiners, speakers of grammar 2 do not. This is shown in (68).
(68)

a.

Mounn nan
Sn
v
man
det
Actually, the

b. *Mounn
*Sn
man

w manchinn nan an.

haitian1
mv mvt
v
v.
fongbe1
see car
det Det
man saw the car.
(=(81) in Lefebvre to appear b)

nan w manchinn
v
mv mvt
det see car

nan an
haitian2
v
v
fongbe2
det Det
(=(82) in Lefebvre to appear c)

In this context, speakers of grammar 2 use the adverbs vre in Haitian and ngb
in Fongbe7 to assert the content of the proposition, as in (69).8
(69)

Mounn nan
Sn
v
man
det
Actually, the

w manchinn nan vre.

mv mvt
v
ngb.
see car
det Adv
man saw the car.

haitian2
fongbe2
(=(83) in Lefebvre to appear b)

A similar pattern is found in the examples in (70).

8.4 The *Det Det Filter in grammar 1 and grammar 2

(70)

a.

Mounn nan kraze

Sn
v
gb
man
det destroy
The man destroyed the

241

manchinn nan an.

haitian1
mvt
v
v.
fongbe1
car
det Det
car, as we knew.
(=(84) in Lefebvre to appear b)

b. *Mounn nan kraze

manchinn
*Sn
v
gb
mvt
man
det destroy car
[Lit.: The man destroyed the car, as

nan an
haitian2
v
v
fongbe2
det Det
we knew.]
(=(85) in Lefebvre to appear b)

While speakers of grammar 1 can utter two determiners in a row, speakers of

grammar 2 cannot.
Lefebvre (to appear b) proposes to link this difference between the two grammars to a more general constraint on the utterance of two adjacent similar forms.
First, I will provide examples showing that this constraint holds independently
in both languages. I will then show that it also applies to a series of determiners
in contexts other than those discussed above. I show that there is variation
among speakers with respect to the strictness of this constraint. I finally show
that the contrast between grammar 1 and grammar 2 illustrated by the data in
(68) and (70) is related to variation among speakers with respect to how they
apply the general constraint on the utterance of two adjacent similar forms.
8.4.1

The general constraint

In both Haitian and Fongbe, there appears to be a constraint preventing the

utterance of two adjacent forms that are phonologically or phonetically similar.
The following examples illustrate this general constraint.
In Fongbe, F and are the forms of the negative marker and the interrogative
marker, respectively. As is pointed out in Da Cruz (1994), these two forms may
appear in the same sentence, provided that they are not adjacent at S-structure.
This is illustrated by the pairs of sentences in (71).
(71)

a.

Kwk w
sv
f
v
?
fongbe
Koku arrive yesterday Neg Det q
Didnt Koku arrive yesterday, as expected?
(=(86a) in Lefebvre to appear b)

b. *Kwk w
sw
f

Koku arrive yesterday Neg q

[Lit.: Didnt Koku arrive yesterday?]

fongbe
(=(15b) in Da Cruz 1994)

The way to rescue (71b) is to use (G) as the form of the negative interrogative,
as shown in (72).
(72)

Kwk w
sw
(g)?
Koku arrive yesterday Neg int
Didnt Koku arrive yesterday?

fongbe
(=(15a) in Da Cruz 1994)

242

The determiner and the structure of the clause

The use of (G) may be seen as a strategy to circumvent the problem of having
to pronounce the sequence *F at surface structure.
This constraint is instantiated in Haitian as well. For example, in Haitian, yo
is the form of both the third-person pronoun and the plural marker, as shown in
(73) (see chapter 4).
(73)

a. Yo
rive.
they arrive
They arrived.

b. liv
yo
book pl
the books

haitian

In nominal structures, personal pronouns may appear in possessive constructions.

(74)

liv
mwen / u
/ li / . . . / yo
book I
/ you / he / . . . / they
my/your/his/ . . . /their book

haitian

Both the possessive marker and the plural marker may occur within the same
nominal structure, as shown in (75).
(75)

liv
li
yo
book (s)he pl
his books

haitian

When the possessor is third-person plural and the noun is plural, we would
expect the sequence yo yo on the model of (75) above. As is shown in Lumsden
(1989), this is not possible, however.
(76)

*liv
yo
yo
book their pl
[Lit.: their books]

haitian
(=(3a) in Lumsden 1989)

Likewise, the sequence yo yo involving two consecutive plural markers is not

possible.
(77)

*liv
[neg yo] yo
book man pl pl
[Lit.: the mens books]

haitian
(=(3a) in Lumsden 1989)

The Fongbe facts discussed in (71)(72) and the Haitian facts in (73)(77)
illustrate a general constraint preventing the utterance of two adjacent phonologically similar morphemes. The Fongbe examples show that this constraint is
a superficial one. If the sequence F can be broken, the two functional categories
(the negative and the question marker) may be realised at surface structure (see
(71a) ); if it cannot, another morpheme is used to convey the same meaning (see
(72) ). Semantic interpretation facts involving the Haitian data discussed above
further support the claim that the impossibility of the sequence *yo yo is a superficial constraint. Indeed, the nominal structure in (78) may be assigned three
interpretations.

8.4 The *Det Det Filter in grammar 1 and grammar 2

(78)

liv
yo
book pl
their book/the books/their books

243

haitian

These interpretive data show that all the pertinent functional categories are being
projected some forms are simply not pronounced.
8.4.2

The *Det Det Filter

When it applies in the environment of two determiners, the constraint is known

in the literature on Haitian and Fongbe as the *Det Det Filter (see Lefebvre
1982b; Lefebvre and Massam 1988; Lumsden 1989). The effect of this constraint may be seen in contexts involving a relative clause. In both Haitian and
Fongbe, the relative clause occurs between the noun and the determiner (see
chapter 4). When the relative clause contains an overt definite direct object,
there are two underlying adjacent determiners, one associated with the direct
object and one with the head of the relative clause. For speakers of grammar 2,
only one of the two adjacent determiners can be uttered, as shown by the contrast between (79a) and (79b).
(79)

a. *Mounn [ ki
kraze
[manchinn
*Sn
[3
gb
[mvt
man
op res destroy car
[Lit.: The man who destroyed the car.]

nan] ] an
v] ]
v
det
det

haitian2
fongbe2

b. Mounn [ ki
kraze
Sn
[3
gb
man
op res destroy
The man who destroyed the

nan] ]
v] ]

det
det

haitian2
fongbe2

[manchinn
[mvt
car
car.

(=(94) in Lefebvre to appear b)

Fongbe speakers of grammar 2 also have the option of assigning the second
occurrence of the determiner a different phonological form. This is shown in
(80), where the second occurrence of the determiner is pronounced as // instead
of as /v/. // is otherwise the weak form of the third-person personal pronoun.
(80)

[Sn [ [3
xw
svn v]
man
op res buy crab det
The man who bought the crab.

] ]
det

fongbe2
(=(95) in Lefebvre to appear b)

I take examples such as (80) to represent a strategy for circumventing the surface constraint that prevents the utterance of two adjacent phonologically similar
forms. None of the Haitian speakers with whom I have been working so far have
manifested this type of strategy.
In relative clauses where the object has been extracted, the determiner associated with the head of the relative clause must be uttered. In that case, the determiner comes right after the verb, so there are no adjacent determiners at surface
structure. This is shown in (81).

244
(81)

The determiner and the structure of the clause

Mounn [
Sn
[3
man
op
The man that

Jan
w] a
Jan
mw] v
John see det
John saw.

haitian2
fongbe2
(=(96) in Lefebvre to appear b)

This supports the claim that the sentences in (79a) are bad for speakers of
grammar 2 because two adjacent determiners are uttered.
For speakers of grammar 1, however, the data are slightly different. These
speakers can utter two determiners in a row, as shown in (82).
(82)

Mounn [ ki
kraze
Sn
[3
gb
man
op res destroy
The man who destroyed the

[manchinn
[mvt
car
car.

nan] ] an
v] ]
v
det
det

haitian1
fongbe1

(=(97) in Lefebvre to appear b)

Speakers of grammar 1 either do not have the *Det Det Filter at all (some Fongbe
speakers from Abomey can utter several determiners in a row) or they have a
less constrained version. For the latter group of speakers, the filter appears to
apply at the phonetic, rather than the phonological, level. In this grammar, if two
adjacent determiners have the same phonetic form, they are ruled out, but if they
are phonetically distinct, they are allowed (see the Haitian example in (82) ).
The facts discussed in this section show that the general constraint on the
production of two adjacent similar forms also applies in contexts involving two
adjacent determiners. In this case, however, speakers divide into two groups:
speakers of grammar 2, who manifest a strict version of this constraint, and
speakers of grammar 1, who have a more relaxed version.
8.4.3

The distribution of the clausal determiner in light of this constraint

The discussion in 8.4.2 allows for a straightforward account of the differences

observed between speakers of grammar 1 and grammar 2 with respect to the
distribution of the clausal determiner (see (68)(70) ). For speakers of grammar
1, who have a relaxed version of the *Det Det Filter, the clausal determiner may
occur in the context of an overt definite affected argument. For speakers of
grammar 2, who have a strict version, the clausal determiner cannot occur in the
context of an overt definite affected argument (see (70) ).
There are two sets of data supporting this interpretation of the facts. First,
speakers of grammar 2 have the pertinent interpretations, even though they do
not utter two adjacent determiners. Second, Fongbe speakers from Ouidah (Fongbe
grammar 3) use the form l instead of the determiner V as an assertive marker.9
Since the form of the determiner is distinct from the form of the assertive marker
in this dialect of Fongbe, we expect the assertive marker to be exempt from the
constraint discussed above. This is exactly what we find. In this dialect, the assertive marker can occur in any context in which it has scope over the sentence.
This is exemplified in (83) and (84).

8.4 The *Det Det Filter in grammar 1 and grammar 2

(83)

Kwk mw jtv v
l.
Koku see thief Det ass
Actually, Koku saw the thief.

245

fongbe3
(=(98) in Lefebvre to appear b)

(84)

Sn 3 gb
mvt 3 l.
man a destroy car
a ass
Actually, a man destroyed a car.

fongbe3
(=(99) in Lefebvre to appear b)

From the data discussed in this section, we see that speakers of grammar 1
have a wide range of possibilities for producing the determiner in the clause,
either as an assertive marker or as an event determiner; speakers of grammar 2,
with their strict version of the constraint, have a much narrower range of possibilities. In fact, these speakers can only produce the determiner in the context of
ergative and intransitive verbs.

8.4.4

Summary of the distribution of the clausal determiner in grammar 1

and grammar 2

In both grammar 1 and grammar 2, the determiner may have the function of both
an assertive marker and an event determiner. In both Haitian and Fongbe, grammar 2 differs systematically from grammar 1. To account for the difference in
the distribution of the assertive marker determiner, it was proposed that, in grammar 1, when the determiner functions as an assertive marker, it is base generated
as the head of MoodP1, whereas, in grammar 2, the determiner with this function
is base generated as the head of MoodP2. In grammar 1, MoodP1 is the surface
position for the assertive marker determiner, whereas, in grammar 2, MoodP1 is a
derived position (at lf) for that marker. The difference between the two grammars
with respect to the distribution of the event determiner was accounted for by
the assumption that, in grammar 1, both the head of tp and the head of AspP
are available surface positions for the event determiner, whereas, in grammar 2,
only the head of AspP is; in this grammar, the head of tp is a derived position
(at lf). Thus, in both grammars, the determiner heads more than one functional
category projection. The difference between the two grammars is whether they
make use of these projections at S-structure or at lf. Finally, it has been shown
that the range of possibilities for the distribution of the clausal determiner (either
as an assertive marker or as an event determiner) depends on the strictness of
the constraint on the utterance of two adjacent similar forms. Speakers of grammar 1 have a relaxed version of the constraint, allowing adjacent forms that
are phonetically similar (e.g. some Fongbe speakers from Abomey) or adjacent
forms that are not phonetically identical (e.g. some Haitian speakers). Speakers
of grammar 2, however, have a stricter version of the constraint, rejecting any
adjacent forms that are phonologically similar. This accounts for the fact that, in

246

The determiner and the structure of the clause

grammar 2, the determiner occurring in the clause has a much narrower distribution at surface structure than it has in grammar 1.
8.5

Conclusion

In chapter 4, we saw that the determiner in nominal structures is the head of dp.
In this chapter, it has been argued that the determiner that occurs in the structure
of the clause may head four functional category projections: namely, MoodP1,
tp, MoodP2 and AspP. One might ask whether the determiner used in the clause
is the same as the one used in nominal structures. There are several arguments
supporting the claim that there is only one determiner in Haitian (see Lefebvre
and Massam 1988) and in Fongbe. In all contexts in which it appears, the determiner shares certain properties. First, in all cases, it is head-final, as can be seen
in the numerous examples provided in this chapter. Second, in all environments,
both clausal and nominal, the determiner undergoes the same phonological processes, yielding numerous variants: la, nan, an and a for Haitian and V and Vn for
Fongbe (see (3) and (14) in chapter 4). Third, in all of its functions, the determiner
is assigned an interpretation that reflects its [+definite] and [+anaphoric] features. In the context of a noun phrase, the presence of the determiner implies
that the referent of the noun is known by the participants (see chapter 4, section 4.1). As an assertive marker, the determiner relates the proposition it is part
of to something that has been said before in the conversation (see (1) ). As an
event determiner, the determiner implies that the event denoted by the clause is
already known to the participants (see (36) ). Fourth, in all contexts in which the
determiner occurs, in both Haitian and Fongbe, there is a superficial constraint
limiting the production of several adjacent determiners. Differences in the strictness of this constraint for speakers of grammar 1 and grammar 2 were discussed
in section 8.4. The determiner shares all these properties regardless of its grammatical function. I therefore conclude that there is only one lexical entry for the
determiner in both Haitian and Fongbe. Since it may head different projections
in these languages, it is a multifunctional head. An account of the multifunctional
character of this lexical item is provided in Lefebvre (to appear b).
This chapter was also intended to show that variation among speakers with
respect to the distribution of the determiner within the clause is not random.
Both languages exhibit three grammars, referred to as grammar 1, grammar 2
and grammar 3. It is a striking fact that the first two grammars postulated for
Haitian parallel the two Fongbe grammars. These grammars differ in similar
ways in both languages. This is no coincidence since Fongbe is one of the
substratum languages of Haitian. A third Haitian grammar shows no use of the
determiner in simple clauses. This parallels a third Fongbe grammar found among
speakers from Ouidah as well as the grammars of speakers of some other substratum languages. Furthermore, the data discussed in this chapter show that
some of the differences between the substratum grammars of Haitian have not
been levelled out and that they were passed on into the creole as properties of

8.5 Conclusion

247

the lexical items involved. The fact that the determiner in both languages has
such parallel functions in the structure of the clause, as well as in the nominal
structure, constitutes a strong argument for relexification. Furthermore, to the
best of my knowledge, only creoles whose substratum languages are West African
have a determiner with such an important role to play in the clause structure.

248

The syntactic properties of verbs

The syntactic properties of verbs

Although the semantic and syntactic properties of verbs are not always easy
to distinguish from each other, this chapter discusses properties of verbs generally considered to be syntactic, such as selectional, raising, control and Caseassigning properties. Verbs syntactic properties are specified in their lexical
entries. The relexification hypothesis predicts that the syntactic properties of
verbal lexical entries in the substratum language will be reproduced in the equivalent lexical entries in the creole. The comparison of the syntactic properties of
Haitian, French and Fongbe verbs shows that, to a great extent, this prediction is
borne out. The selectional properties of body-state verbs, weather verbs and
reflexive verbs are discussed in sections 9.2 to 9.4. The syntactic properties of
verbs that select expletive subjects, verbs that allow for argument raising and
existential verbs are discussed in sections 9.5 to 9.7. Section 9.8 deals with the
syntactic properties of control verbs. Sections 9.9 and 9.10 review light verbs
and inherent object verbs. Section 9.11 discusses the Case-assigning properties
of verbs. The properties of double-object verbs are addressed in section 9.12.
The syntactic properties of most of these verb classes in Haitian contrast with
those of French and are parallel to Fongbe. There are four exceptions to this
generalisation, each attesting to the other processes that play a role in creole
genesis. First, the data pertaining to the selectional properties of reflexive verbs
(and possibly of verbs selecting expletive subjects) show that dialect levelling
has played a role in this case. Second, the properties of the Haitian existential
verb appear to constitute a case of innovation from within the creole. Third, the
Case-assigning properties of Haitian verbs are rather different from either of its
contributing languages. Evidence for this claim will be presented in sections 9.11
and 9.12. In fact, it appears that in some instances the Case properties assigned
to verbs in the creole are quite unrelated to what is found in the contributing
languages, suggesting that this area of the grammar is a locus for independent
development in creole genesis. Finally, in contrast to Fongbe, there are no inherent object verbs in Haitian, showing the influence of French. This chapter begins
with an overview of types of argument structures in the three languages.
9.1

Types of argument structures

Detailed discussions of types of argument structures in Haitian and Fongbe can

be found in Massam (1989) and Lefebvre (1991b), respectively. A preliminary
248

9.1 Types of argument structures

249

comparison of argument structures in Haitian, French and West African languages may be found in Koopman (1986). Building on these findings and on
further work, I shall provide a short overview of the types of argument structures
in Haitian, French and Fongbe, pointing out their similarities and differences.
All three languages have monadic verbs (that is verbs for which the Lexical
Conceptual Structure, henceforth lcs,1 contains only one variable). These include
unaccusative verbs such as to go, to come, etc., as in (1). While such Haitian
and Fongbe verbs occur in their bare form, the corresponding French verbs are
conjugated with the auxiliary to be.
(1)

a. Li ale / rive.
y / w.
He left/arrived.

haitian
fongbe

b. Il est parti / arriv.

He left/arrived.

french

In both Haitian and Fongbe, the locative argument of these verbs need not be
introduced by a Case marker or a pre- or postposition, whereas, in French, the
locative object has to be introduced by at. This contrast is shown in (2).
(2)

a. Li ale / rive Ptoprens.

y / w Ktvn.
He went to/arrived in Port-au-Prince/Cotonou.

haitian
fongbe

b. Il est all / arriv Paris.

He went to/arrived in Paris.

french

Single-variable verbs also include unergative verbs such as to bark and to

jump, as shown in (3). While such Haitian and Fongbe verbs occur in their bare
form, the French equivalents are conjugated with the auxiliary to have.
(3)

a. Li abwaye / sote.
hk
/ lhn.
He barked/jumped.

haitian
fongbe

b. Il a aboy / saut.
He barked/jumped.

french

The range of single-variable verbs is smaller in Fongbe than in French and

Haitian. This is because a concept rendered by a single-variable verb in
Haitian and French is sometimes expressed by a light verb construction or an
inherent object verb in Fongbe (see sections 9.9, 9.10). Single-variable predicates also include weather verbs (section 9.3), some verbs selecting expletive subjects (section 9.5), some raising verbs (section 9.6), and existential verbs
(section 9.7).
The second type of argument structure consists in two-variable (or transitive)
verbs as in He ate bread. All three languages have a large class of these verbs.
Finally, all three languages have three-variable (or ditransitive) verbs. These
include predicates of transfer such as to give. In Haitian and Fongbe, these

250

The syntactic properties of verbs

predicates are rendered by a double-object construction, as in John gave Mary a

book, or a serial verb construction. As will be shown in detail in section 9.12,
French does not have such constructions and three-variable predicates are rendered by a construction of the type John gave a book to Mary. Three-variable
predicates also include some control verbs, as discussed in section 9.8, where it
will be shown that the properties of Haitian and Fongbe verbs are quite similar
and contrast with those of the corresponding French verbs. And they include
verbs participating in the locative alternation such as to load, briefly discussed
in section 9.11.
9.2

body-state verbs

In Haitian, body-state expressions such as I have a headache are built on the

model BODY-PART+VERB+PRONOUN, as shown in the following examples.
(4)

a. Vant
mwen ap f m mal.
haitian
stomach me
imp do me hurt
I have a stomach-ache. [Lit.: My stomach hurts me.]
(=(29) in Koopman 1986)
b. Tt
mwen ap f m
head me
imp do me
I have a headache.

mal.
hurt

haitian
(=(29) in Koopman 1986)

c. Dan ap manje m.
tooth at eat
me
I have a toothache. [Lit.: My tooth is eating me.]

haitian

French also allows body-state expressions on the model of the Haitian ones in
(4). For example, it is possible to say Lestomac me brle [Lit.: The stomach
burns me], La tte me fait mal [Lit.: The head hurts me]. In French, however,
body-state expressions are typically built on the model X HAVE PAIN at BODY-PART.
(5)

a. J ai
mal la
tte.
I have pain at det head
I have a headache.

french

b. J ai
mal au ventre.
I have pain at belly
I have a stomach-ache.

french

c. J ai
mal aux dents.
I have pain at
teeth
I have a toothache.

french

The French expressions in (5) have no counterpart in Haitian. As is pointed out

by Koopman (1986), the structure of the Haitian expressions in (4) is similar to
corresponding expressions in the West African languages, which are also built
on the model BODY-PART+VERB+PRONOUN. Examples from Fongbe are provided
in (6).

9.3
(6)

WEATHER

verbs

251

a. Xmy
wl mi.
stomach hold me
I have a stomach-ache. [Lit.: My stomach is holding me.]

fongbe

b. T
3l mi.
head eat me
I have a headache. [Lit.: My head is eating me.]

fongbe

c. 3 3w 33 mi wy.
tooth at eat
me part
I have a toothache. [Lit.: My tooth is eating me.]

fongbe

Body-state verbs in Haitian (see (4) ) and Fongbe (see (6) ) typically select bodyparts as their subject. This contrasts with French body-state verbs, which typically take a pronominal subject (see (5) ). This should come as no surprise given
the relexification hypothesis.
9.3

weather verbs

As has been pointed out by Koopman (1986) and documented in detail by

Dumais (1988), Haitian expresses various atmospheric phenomena by means of
a construction that uses verbs (which occur in other contexts as well) selecting a
lexical subject referring to a natural element. This is shown in (7).
(7)

a. Lapli tonbe.
b. Laglas tonbe.
haitian
rain
fall
ice
fall
It is raining. [Lit.: Rain falls.] It is hailing. [Lit.: Ice falls.]
(=(1) and (2) in Dumais 1988)
c. Yon tivan
vante.
a
little wind wind
It is windy.
[Lit.: The wind winds.]

d. Lraj gwonde.
haitian
storm growls
It is thundering.
[Lit.: The storm growls.]
(=(3) and (4) in Dumais 1988)

e. Lapli ap farinen.
rain
imp drizzle
It is drizzling. [Lit.: Rain is drizzling.]

haitian
(from Valdman et al. 1981)

Both authors point out that the above Haitan data contrast with French, where
the same concepts are rendered by means of weather verbs selecting an expletive subject, as shown in (8). An expletive subject is not allowed in Haitian in
the context of the weather verbs in (7).
(8)

a. Il pleut.
It is raining.

b. Il grle.
It is hailing.

french

c. Il vente.
It is windy.

d. Il fait un orage. / Il tonne.

It is stormy./It is thundering.

french

e. Il bruine.
It is drizzling.

french

252

The syntactic properties of verbs

Some French expressions built on the model of the Haitian ones in (7) may be
grammatical; for example, we can find La pluie tombe Rain is falling, Lorage
gronde The storm is growling but not *Un petit vent vente A little wind is
blowing nor *La bruine bruine Drizzle is drizzling. As the translations show,
however, these expressions can only be interpreted literally. Furthermore, the
first two expressions can only be used in specific contexts for stylistic effects
(e.g. in poetry). The standard way of using weather verbs in Haitian (see (7) )
thus differs from the standard way of using weather verbs in French (see (8) ).
Again, the properties of the Haitian weather expressions will be shown to
follow the pattern of the substratum languages.
Koopman (1986: 245) points out that weather verbs which select an expletive subject do not exist in West African languages any more than they do in
Haitian. On the basis of examples from Vata and Abe, she shows that West
African languages generally express the various atmospheric phenomena in a
construction involving verbs (that occur in other contexts as well) selecting a
lexical subject referring to a natural element. The Fongbe data in (9), from
Dumais (1988), illustrate this pattern.
(9)

a. J
j.
rain falls
It is raining. [Lit.: Rain falls.]

fongbe
(=(1) in Dumais 1988)

b. Lgls j.
ice
fall
It is hailing. [Lit.: Ice falls.]

fongbe
(=(2) in Dumais 1988)

c. Jhwn ny.
wind blow
It is windy.

fongbe
(=(3) in Dumais 1988)

d. Hxbys dk-gb.
thunder growl
It is thundering.

fongbe

e. J
3-tn.
rain drizzle
It is drizzling.

fongbe

The Haitian expressions in (7) are built on the model of expressions in the West
African languages like those in (9). The data in (7) and (9) reflect the selectional
properties of the verbs involved. In both Haitian and Fongbe, but not in French,
verbs meaning to fall, to wind, to thunder, to drizzle take an argument
which is a natural element such as rain/ice/wind, etc. The Haitian facts are
predicted by the relexification hypothesis.
In light of the above discussion, it is surprising to find, in modern Haitian,
expressions such as (Li) f cho It is hot and (Li) f frt It is cold, modelled on

9.4 Reflexive verbs

253

French (Il) fait chaud and (Il) fait froid, respectively. In Fongbe, and presumably
in other West African languages, the corresponding expressions have the same
form as the weather expressions discussed in (9): yz gbX (Lit.: heat beexcessive) (see Dumais 1988) and jhWn gbX (Lit.: wind be-excessive). The
existence of Haitian data like those in (7), however, suggests the possibility that,
in the early creole, Haitian expressions based on the Fongbe ones might have
existed. Whether (Li) f cho and (Li) f frt were acquired from French by the
substratum speakers at the time Haitian creole was formed or acquired through
contact with French in the further development of the creole, will, I believe,
always remain a matter of conjecture.
9.4

Reflexive verbs

Recall from chapter 6 that, in Haitian, a reflexive interpretation may be induced

by a pronoun+self (where self is phonologically null), tt-+pronoun (Lit.:
Xs head), or k-+pronoun (Lit.: Xs body). A few verbs also allow for a
reflexive interpretation (without any overt reflexive form) when their internal
argument is not realised in the syntax. These four possibilities are illustrated in
(10).
(10)

a. Jani pann lii

sou pye-bwa pou gade match la.
haitian
John hang him on tree
to
watch match det
John hooked himself on a tree in order to watch the match.
(=(1b) in Brousseau 1995b)
b. Jan i touye tt-lii .
John kill
head-his / him
John killed himself.

haitian
(=(2b) in Brousseau 1995b)

c. Jani chape k-lii

John escape body-his
John escaped.

/
/

*tt-lij
head-his

haitian
(=(39b) in Brousseau 1995a)

d. Jan
abiye.
John dress
John dressed himself.

haitian

This section reports on the subcategorisation properties of a sample of some

thirty Haitian verbs studied by Brousseau (1995b). In studying the properties
of Haitian reflexive verbs, Brousseau (1995b) grouped verbs on the basis of the
most prominent (that is, the most common and most natural) reflexive form
that they subcategorise for. Other reflexive forms that a particular verb may take
are listed in the right-hand column of (11). Verbs which most prominently subcategorise for a pronominal form are listed in (11). Some verbs in (11) may also
take tt-li as a reflexive (where li may be replaced by any other pronominal
form).

254
(11)

The syntactic properties of verbs

Verbs which typically take a pronominal form
a. dekwafe
pann
penyen
prse

to
to
to
to

mess ones hair

hang or hang onto
comb ones hair
hurry

b. mande
raple
trne

to wonder
to remember
to drag oneself

c. abitye
amize
fiye

to get used to something

to have fun
to trust

other reflexive forms

haitian

(# tt-li) 2
(# tt-li)
(# tt-li)

haitian

tt-li
haitian
tt-li
tt-li
(=(7) in Brousseau 1995b)

The verbs in (12) are those which select tt-li as the most prominent form.
Some of these verbs may also take either a pronominal form or k-li as a
reflexive.
(12)

Verbs which typically take tt-li

a. bse
pann
touye
trne

to
to
to
to

drop down
hang oneself (suicide)
kill oneself
train

b. gade
w

to look at oneself
to see/look at oneself

c. grate
lave
touye

to scratch (o.s.)
to wash (o.s.)
to kill oneself (by doing x)

other reflexive forms

haitian

li
li

haitian

k-li
haitian
k-li
k-li
(=(9) in Brousseau 1995b)

The verbs in (13) are those for which no reflexive form can be identified
as most prominent. Most of these verbs may select a pronominal form or either
tt-li or k-li.
(13)

Verbs for which no reflexive form is prominent

a. enskri
meliore

to register
to improve

b. blayi
debouye
depche
fse
layite

to
to
to
to
to

lie down/stretch out

manage/cope
hurry
try hard
lie down/stretch out

other reflexive forms

tt-li
tt-li

haitian

tt-li / k-li
haitian
tt-li / k-li
tt-li / k-li
tt-li / k-li
tt-li / k-li
(=(10) in Brousseau 1995b)

9.4 Reflexive verbs

255

The verbs in (14) are those which can be interpreted as reflexive without any
overt reflexive form. These verbs may optionally select a pronoun (interpreted as
reflexive) or k-li.
(14)

Verbs which can be interpreted as reflexive without any overt form

other reflexive form
li
li
li

a. abiye
to get dressed
benyen to have a bath
bse
to bend down
b. chape
sove

to escape
to escape/run away

c. blse

to hurt (oneself)

k-li
k-li

haitian

haitian

li / tt-li / k-li
haitian
(=(5) in Brousseau 1995b)

The verb blse to get hurt may select any of the four forms, as is illustrated
in (15).
(15)

Blse to
a. Jani
John
Johni

hurt (oneself)
blse (nan pye
hurt in
foot
hurt hisi foot.

lii).
his

haitian

b. Jani blse li i /j (nan pye li i /j).

John hurt him in
foot his
Johni hurt hisi /j foot.
c. Jani blse tt-lii
John hurt head-his
Johni hurt hisi foot.

(nan pye
in
foot

d. Jani blse k-lii /j

John hurt body-his
Johni hurt hisi /j foot.

(*nan pye
in
foot

haitian

lii).
his
li)
his

haitian

haitian
(=(6) in Brousseau 1995b)

The data collected by Brousseau (1995b) are in line with those reported on in
previous work on Haitian (see e.g. Sylvain 1936; Faine 1937; Goodman 1964;
Carden and Stewart 1988).
The striking fact about the distribution shown above is that most verbs may
select more than one reflexive form. There are even verbs which may select all
four forms, as in (15). Even though Haitian differs from languages such as
English or French in this respect, interestingly enough, the same situation is
observed in other creole languages (see Muysken and Smith 1995). This is due
to the fact that, as we saw in chapter 6, the different reflexive forms found in a
given creole have been transferred into it from different substratum languages
through relexification. But recall from chapter 6 that each individual substratum
language has a different subset of the total inventory of forms found in the substratum languages as a group. This leads one to predict that the subcategorisation
properties of verbs (considered as a whole) for reflexive forms in a given creole

256

The syntactic properties of verbs

should not match those of the corresponding verbs in any of the substratum
languages taken individually. Brousseaus (1995b) comparison of the subcategorisation properties of verbs for reflexive forms in Haitian and Fongbe shows
that this prediction is borne out. Given the conclusion in chapter 6, section 6.6,
that superstratum languages do not contribute reflexive forms in creole languages,
French is not considered in this comparison.
Recall from chapter 6 that, of the four possibilities for inducing a reflexive
interpretation in the substratum languages, Fongbe has only two. It has one reflexive form, a pronoun + self, and a few verbs allow for a reflexive interpretation
without an overt reflexive form when their internal argument is not realised in
the syntax. These two possibilities are illustrated in (16a and b), respectively.
(16)

a. Byi mwn -3i .

Bayi see her-self
Bayi saw herself.

b. lx.
fongbe
he bath
He bathed.
( (16a)=(11b) in Brousseau 1995b)

Brousseaus (1995b) comparison of verbs which basically take a pronominal

reflexive form in Haitian with the semantically closest Fongbe verbs shows that
the selectional properties of pairs of Haitian and Fongbe verbs are not the same.
Indeed, out of the ten pairs of verbs in (17), in only two cases do both Haitian
and Fongbe verbs select a pronominal form.
(17)

Haitian verbs which typically select a pronominal reflexive form

a./ b. Haitian verbs which select no other form
haitian
dekwafe
pann
penyen
prse
mande
raple
trne
c.

fongbe
to
to
to
to
to
to
to

mess ones hair

hang or hang onto
comb ones hair
hurry
wonder
remember
drag oneself

3 kn kw

hwlx
kn byv
fln
dwn

reflexive form in
fongbe

Haitian verbs which also select tt-li

haitian
abitye
amize
fiye

fongbe
to get used to something m
to have fun

to trust

reflexive form in
fongbe

(=(17) in Brousseau 1995b)

Similarly, the comparison of verbs which typically select tt-li in Haitian with
the semantically closest Fongbe verbs shows important mismatches between
pairs of verbs. As is shown in Brousseau (1995b), where the Haitian verbs select
tt-li, most Fongbe verbs select a pronominal form.

9.4 Reflexive verbs

(18)

257

Haitian verbs which typically select tt-li

a. Haitian verbs which select no other forms
haitian
bse
pann
touye
trne

fongbe
to
to
to
to

drop down
hang oneself (suicide)
kill oneself
train

3 kn kw
h

reflexive form
in fongbe
3
3

b. Haitian verbs which also select a pronoun

haitian
gade
w

fongbe
to look at oneself
to see/look at oneself

kpvn
mw

reflexive form
in fongbe
3
3

c. Haitian verbs which also select k-li

haitian
grate
lave
touye

fongbe

reflexive form
in fongbe
to scratch (o.s.)
kl
w / 3
to wash (o.s.)
lx w / ss w

to kill oneself (by doing x) gb t my

(=(18) in Brousseau 1995b)

Likewise, verbs which select several reflexive forms in Haitian select either a
pronoun+self or no overt form in Fongbe. This is shown in (19).
(19)

Haitian verbs which select several reflexive forms

a. Haitian verbs which select li / tt-li
haitian
enskri
meliore

fongbe
to register
to improve

reflexive form in
fongbe

b. Haitian verbs which select li / tt-li / k-li

haitian
blayi
debouye
depche
fse
layite

fongbe
to
to
to
to
to

lie down/stretch out

manage/cope
hurry
try hard
lie down/stretch out

reflexive form in
fongbe
3

drxn

y w

drxn
3
(=(19) in Brousseau 1995b)

There is a fourth subset of data, however, where pairs of Haitian and Fongbe
verbs are more similar, although not identical, with respect to their selectional
properties. The subset of Haitian verbs which can be assigned a reflexive interpretation without any overt form largely corresponds to the equivalent class of
Fongbe verbs. This is shown in (20).

258
(20)

The syntactic properties of verbs

Haitian verbs interpreted as reflexive without an overt form
a. Haitian verbs which also select li
haitian
abiye
benyen
bse

fongbe
to get dressed
to have a bath
to bend down

3 w
lx
fx

fongbe selectional
properties

b. Haitian verbs which also select k-li

haitian
chape
sove

fongbe
to escape
to escape/run away

hwn
hwlxn gn

fongbe selectional
properties

c. Haitian verbs which also select li / tt-li / k-li

haitian
blse

fongbe
to get hurt

fongbe selectional
properties
gbl-w

(=(15) in Brousseau 1995b)

The fact is that the reflexive interpretation of the verbs in (20) is not driven by
their semantics. Recall from above that these verbs are assigned a reflexive interpretation when their object is not realised. This might explain the congruence
between Haitian and Fongbe. Note, however, that, as is pointed out by Brousseau
(1995b), while the Haitian verbs in (20) may also select an overt reflexive form,
the corresponding Fongbe verbs do not.
Thus, a fair conclusion on the above data is that the selectional properties of
Haitian and Fongbe verbs are not the same as far as reflexive forms are concerned. As paradoxical as it might seem, however, I will argue that these facts
follow straightforwardly from the relexification hypothesis and dialect levelling.
I propose the following scenario.
This scenario assumes, first, that the Haitian reflexive forms were relexified
from various West African languages and, second, that each particular West
African language only presents a subset of these forms (see chapter 6). Imagine,
then, a hypothetical verb in two hypothetical substratum languages A and B.
Suppose that in language A this verb selects a pronoun+self (as is the case in
Gbe languages), and that in language B this verb selects the form head+pronoun
or body+pronoun. Finally, suppose that speakers of these two hypothetical languages relexify their verbal lexicon on the basis of data from French. Speakers
of both substratum languages will end up with a Haitian verb that has a phonological representation derived from a French verb. For speakers of substratum
language A, this lexical entry will select the reflexive form pronoun+self.
For speakers of substratum language B, however, this lexical entry will select
one of the following two reflexive forms: head+pronoun or body+pronoun. In
the early creole community, there would thus be a single Haitian verb, but its
selectional properties would differ across speakers such that there would be two
early Haitian dialects in terms of the verbs subcategorisation properties. In one

9.5 Verbs licensing expletive subjects

259

dialect, the Haitian verb would have the subcategorisation properties of the
hypothetical substratum languages of type A; in the other, the verb would have
the subcategorisation properties of the hypothetical substratum languages of
type B. It is not too farfetched to hypothesise that, as speakers of different early
Haitian dialects communicated with each other, a given Haitian verb might have
ended up selecting any one of the available reflexive forms. It is thus possible
that, at some point in the creoles evolution, the selectional properties of particular reflexive verbs consisted of the combined selectional properties of the corresponding verbs in the various substratum languages. This would explain why
most Haitian verbs in the above sample can take several reflexive forms. These
data illustrate the fact that the idiosyncratic properties of a given verb in different dialects may all become part of this lexical entrys selectional properties at
some point in the creoles development.3
The situation described above is exactly the type of situation where we would
expect dialect levelling to operate in a creole community. Given the number of
reflexive forms that any one Haitian verb can take, we would expect the numerous subcategorisation options of particular verbs to be reduced over time. We
would also expect that different communities would not necessarily settle on the
same option (see work by Carden and Stewart 1988 on this topic).
9.5

Verbs licensing expletive subjects

Recall from chapter 6 that all three languages under comparison have expletive
subjects and that the form of this expletive is the same as that of the third-person
personal pronoun: li in Haitian, il in French and in Fongbe. Recall also from
chapter 6 that Haitian and Fongbe, but not French, also have a null expletive
subject. This section considers the verbs which allow for expletive subjects in
the three languages. The comparison is centered along two dimensions: first,
the classes of verbs and adjectives which license expletive subjects, and second,
which form of the expletive (overt and/or covert) specific Haitian and Fongbe
verbs allow. The Haitian data are drawn from the literature on the topic, complemented by my own field notes. The French data come from my own judgments
and were checked against French grammars. The Fongbe data, unless otherwise
identified, come from my own field notes.
Verbs of the seem-class typically allow for an expletive subject (see Chomsky
1981). Haitian, French and Fongbe are no exceptions to this generalisation, as
shown in (21).
(21)

a. Li sanble Jan
te
malad.
it seem John ant sick
It seems that John has been sick.

haitian

b. Il semble que
Jean soit malade.
it seem
comp John be sick
It seems that John is sick.

french

260

The syntactic properties of verbs

c. 3
3w
Kwk jzwn.
it seem comp Koku sick
It seems that Koku is sick.

fongbe

Verbs such as to remain and to be missing constitute another subclass of verbs

which can take expletive subjects. Haitian, French and Fongbe all allow for an
expletive subject in this context, as shown in (22) and (23).
(22)

a. Li rete
Jan
sinkan goud.
it remain John fifty
gourds
[Lit.: There remain fifty gourds to John.]

haitian
(=(1a) in Lumsden 1993a)

(23)

b. Il reste
cinquante gourdes Jean.
it remain fifty
gourds to John
[Lit.: There remain fifty gourds to John.]

french

c. kp
kpvnw 3 Kwk sn.
it remain fifty
at Koku of
[Lit.: There remain fifty to Koku.]

fongbe

a. Li manke Jan
sinkan goud.
it lack
John fifty
gourds
[Lit.: It is lacking fifty gourds to John.]

haitian
(=(1b) in Lumsden 1993a)

b. Il manque
cinquante gourdes Jean.
it be missing fifty
gourds to John
[Lit.: It is lacking fifty gourds to John.]

french

c. hw kpvnw n Kwk.
it lack fifty
to Koku
[Lit.: It is lacking fifty to Koku.]

fongbe

Some adjectives also allow for expletive subjects, as shown in (24).

(24)

a. Li bon
pou
Jan
pati.
it good comp John leave
It is good that John is leaving.

haitian

b. Il est
bon
que
Jean parte.
it aux good comp John leave
It is good that John is leaving.

french

c. nywn 3w
Kwk n
y.
it good comp Koku irr leave
It is good that Koku is leaving.

fongbe

The data in (25) show the distribution of expletives in the context of the adjective meaning possible for Haitian and French. Fongbe expresses this concept by
a different construction.

9.5 Verbs licensing expletive subjects

(25)

a. Li posib
pou
Jan
vini.
it possible comp John come
It is possible that John will come.

261

haitian
(=(13) in Massam 1989)

b. Il est
possible que
Jean vienne.
it aux possible comp John come
It is possible that John will come.

french

Except for the predicate meaning possible (lacking in Fongbe),4 the predicates
selecting an expletive subject are the same in all three languages. Having established this fact, I shall now discuss which form of the expletive (overt and/or
covert) specific Haitian and Fongbe verbs allow.
The Haitian verb sanble to seem, to resemble allows for either an overt or
a covert expletive, as has been pointed out in Dumais (1988), DeGraff (1992a,
1992b), Deprez (1992a), Law (1992), etc. Likewise, the corresponding Fongbe
verb # to seem, to resemble also allows for both. The Haitian verb genl
which also means to seem, to look like (see Valdman et al. 1981) can only take
a covert expletive (see DeGraff 1992a) but the corresponding Fongbe verb G
to seem, to look like allows for either an overt or a covert expletive.5 The
Haitian verb rete to remain in (22a) selects either an overt or a covert expletive (see Lumsden 1993a), but the corresponding Fongbe verb in (22c) takes only
an overt one. As is pointed out in Massam (1989), the Haitian verb rete may
also occur in a context such as in (26), where no overt expletive is allowed
(see (26b) ).
(26)

a.

Rte

pou

Jan

vini.

b. *Li rte
pou
Jan
vini
it remain comp John come
What is left is for John to come.

haitian
haitian
(=(16b) in Massam 1989)

The Fongbe verb kp may also occur in sentences of the type in (26), as shown
in (27b). But, in this case, it requires an overt expletive, as shown by the contrast
in grammaticality between (27a) and (27b).
(27)

a. *

Kp

3w

Kwk

b. kp
3w
Kwk n
w.
it remain that Koku irr come
What is left is for Koku to come.

fongbe
fongbe

The adjectives meaning good in Haitian and Fongbe (see (24a) and (24c) ) both
require an overt expletive. Finally, the Haitian adjective posib, which has no
Fongbe counterpart, selects either an overt or a covert expletive (see Massam
1989). The above data are summarised in (28).

262
(28)

The syntactic properties of verbs

Selectional properties of verbs for expletive (overt/covert) subjects
A. to seem
(see (21) )
B.

to seem
(see note 5, (ia) )

C.

to remain
(see (22) )

C. specific use of to remain

(see (26), (27) )
D. to be missing/lacking
(see (23) )

haitian
fongbe
overt / covert overt / covert
covert

overt / covert

overt / covert

overt

covert

overt

overt / covert

overt

E.

good
(see (24) )

overt

overt

F.

possible
(see (25) )

overt / covert

Considering that the covert expletive option is a property of Haitian inherited

from the substratum language(s), the distribution in (28) is striking: only two
pairs of Haitian/Fongbe verbs (A and E) appear to have the same selectional
properties. The other pairs show various types of mismatches. Furthermore,
more Haitian verbs allow for a null expletive. I suspect that there is probably
variation across Haitian speakers with respect to these data. I also suspect that
West African languages must vary as well with respect to which kind of expletive subjects (overt or covert) verbs select. This is an area of the lexicon where
dialect levelling is likely to occur. The documentation of the pertinent facts,
however, remains a topic for further research.
9.6

Raising verbs

A subset of verbs which license expletive subjects also allow for argument raising to subject position. Argument raising may proceed from an object position,
as in An apple remains in the basket (<There remains an apple in the basket), or
from an embedded subject position, as in John seems to be sick (<It seems that
John is sick). Some similarities are found between the Haitian and the Fongbe
data discussed in this section. First, in both Haitian and Fongbe, object-tosubject raising shares certain properties which are not found in French. Second,
both Haitian and Fongbe allow for raising of an embedded subject out of a
tensed clause, which is a marked option. This contrasts with French, which
allows raising of an embedded subject only out of an infinitival clause, an
unmarked option.

9.6 Raising verbs

9.6.1

263

Argument raising in the context of verbs meaning to remain and

to be missing

As we saw in section 9.5, the verbs meaning to remain/to be left over and to
be missing/to lack in the three languages under comparison have the common
property of allowing an expletive form in their subject position. Other syntactic
properties of these verbs, such as object-to-subject raising, however, are not
shared by the corresponding lexical entries in the three languages. Haitian and
French differ, with Haitian patterning with Fongbe.
In Haitian, the argument of the verb rete to remain/to be left over may
appear in two surface positions, as shown in (29). In the (a) sentence, the
argument follows the verb and the subject position is optionally filled with the
expletive pronoun li. In the (b) sentence, the argument appears in the subject
position and the object position is empty.6
(29)

a. Li rete
yon span (nan pannye
it remain a
snake in
basket
There remains a snake (in the basket).

an).
det

haitian
(=(14) in Dumais 1988)

b. Yon span rete

nan pannye an.
a
snake be-left-over in
basket det
A snake remains in the basket.

haitian
(=(15) in Dumais 1988)

Dumais (1988) points out that the Haitian data in (29) contrast with the
French data involving the verb rester to remain. As is shown in (30a), the
argument of the French verb rester appears in the position following the verb
and the subject position is obligatorily filled by the expletive form il. The ungrammaticality of the (b) sentence shows that the argument of rester cannot
appear in the subject position when the verb establishes a locative relation.
(30)

a.

Il reste
un serpent dans le
panier.
it remain a
snake
in
det basket
There remains a snake in the basket.

b. *Un
a

serpent reste
snake
remain

dans le
panier
in
det basket

french

french

The ungrammaticality of the French sentence (30b) contrasts with the grammaticality of the Haitian sentence (29b). Thus, although the French verb rester is
the phonetic source of rete, it did not contribute its syntactic properties. As we
are about to see, these properties come from the substratum languages.
For example, in Fongbe, there is a verb kp, which means both to remain
and to be left over (see Segurola 1963). As is the case in Haitian, the argument
of this verb may occupy two surface positions. In (31a), it occurs after the verb
and the subject position is filled with the pleonastic form . In (31b), it occurs in
the verbs subject position.

264
(31)

The syntactic properties of verbs

a. kp
dn
3kp (3 xsn
it remain snake one
(at basket
There remains a snake in the basket.

v
my).
det in)

fongbe

(=(14) in Dumais 1988)

b. Dn
3kp kp
(3 xsn v
my).
snake one
remain (at basket det in)
A snake remains in the basket.

fongbe
(=(15) in Dumais 1988)

As Dumais (1988) shows, in Fongbe, the locative phrase is optional regardless

of the surface position of the argument. The optionality of the locative phrase in
(31b) contrasts with its obligatoriness in the Haitian example in (29b). Except
for this difference, however, both Haitian rete and Fongbe kp allow the argument to occupy two surface positions. It thus appears that the syntactic properties of the Haitian verb rete follow those of the corresponding verb kp in the
substratum language rather than those of the French verb rester which is its
phonetic source.
The properties of the Haitian verb manke to be missing/to lack present a
similar pattern. Manke may take two arguments: one referring to the object that
is missing and one referring to the object which is deprived of the missing
object. Both arguments may follow the verb, in which case the subject position
is optionally filled by the expletive li, as is shown in (32a). The verb manke also
allows for either of its arguments to appear in subject position. In (32b), the
argument referring to the missing object occupies the subject position, and in
(32c), the argument referring to the object which is deprived is in this position.7
(32)

a. Li manke
sl nan soup
it be-missing salt in
soup
There lacks salt in the soup.

la.
det

haitian
(=(16) in Dumais 1988)

b. Sl manke
nan soup la.
salt be-missing in
soup det
Salt is lacking from the soup.

haitian
(=(17) in Dumais 1988)

c. Soup la
manke
soup det be-missing
The soup lacks salt.

sl.
salt

haitian
(=(18) in Dumais 1988)

As is the case in Haitian, the French verb manquer to be missing/to lack

also takes two arguments. Both arguments may follow the verb, in which case
the subject position is obligatorily filled by the expletive il. This is shown in
(33a), which is parallel to the Haitian sentence in (32a). Unlike the Haitian verb
manke, however, French manquer does not allow the argument referring to the
missing object in (33b) to occur in subject position. The ungrammaticality of
(33b) contrasts with the grammaticality of the corresponding Haitian sentence
in (32b).8 Finally, French manquer does allow the argument referring to the
deprived object to occur in subject position, as is shown in (33c), which parallels
the Haitian sentence in (32c).

9.6 Raising verbs

(33)

a.

Il manque
de sel dans
it be-missing of salt in
There lacks salt in the soup.

la
soupe.
det soup

265

french
(=(16) in Dumais 1988)

b. *Du sel
det salt

manque
dans
be-missing in

la
the

soupe
soup

french
(=(17) in Dumais 1988)

c.

La
soupe manque
de
det soup be-missing of
The soup lacks salt.

sel.
salt

french
(=(18) in Dumais 1988)

The data in (32) and (33) show that the syntactic properties of the Haitian verb
manke do not correspond entirely to those of the French verb manquer from
which it is phonologically derived. However, the following data show that the
properties of the Haitian verb parallel those of the corresponding verbs in the
substratum language.
According to Dumais (1988), Fongbe has a verb hwY meaning to be missing/
to lack. As is the case in Haitian and in French, this verb may take two arguments. Both arguments may be realised after the verb, in which case the subject
position is obligatorily filled by the expletive . This is shown in (34a), which
parallels the Haitian and French data in (32a) and (33a), respectively. As is the
case in Haitian but not in French, however, the verb hwY allows for either of its
two arguments to appear in subject position. In (34b), the argument referring to
the missing object occupies the subject position and, in (34c), the argument
referring to the argument deprived of the missing object occupies this surface
position.
(34)

a. hwy
zy
3 nsn v
my.
it be-missing salt at soup
det in
There lacks salt in the soup.

fongbe
(=(16) in Dumais 1988)

b. Zy hwy
3 nsn my.
salt be-missing at soup
in
Salt is lacking from the soup.

fongbe
(=(17) in Dumais 1988)

c. Nsn v
hwy
zy.
soup
det be-missing salt
The soup lacks salt.

fongbe
(=(18) in Dumais 1988)

The Fongbe data in (34) parallel the Haitian data in (32) even in the one context
where Haitian differs from French. Indeed, the Fongbe sentence in (34b) and the
Haitian one in (32b) are congruent and both differ from the French data (see
(33b) ) in the same way. The syntactic properties of this Haitian lexical entry
thus follow directly from the relexification hypothesis.

266
9.6.2

The syntactic properties of verbs

Subject raising in the context of verbs meaning to seem

In English, the verb seem is syntactically classified as a raising verb because its
surface subject appears to have been raised from the subject position of its
complement clause (see Chomsky 1981, and related literature). For example,
compare the English sentences It seems that John is happy and John seems to be
happy. In the first sentence, the subject position of the verb seem is filled by the
expletive pronoun it, but in the second, the subject of the embedded clause
appears as the surface subject of seem. Note, however, that in the former sentence, the embedded clause is tensed, but in the latter, it is infinitival. In English,
it is only possible to raise the subject of an embedded clause to a matrix subject
position if the embedded clause is infinitival. This is illustrated by the contrast in
grammaticality between the following English sentences: *John seems that (he)
is happy versus John seems to be happy. As we will see below, French is like
English in this respect. However, Koopman (1986: 239) shows that, unlike
French, verbs meaning to seem in Haitian and West African languages do not
take infinitival complements and consequently they lack the property of allowing subject raising out of such complements. Nevertheless, these verbs do take
tensed complements and they do allow the embedded subject position to be
coindexed with the matrix subject, therefore allowing subject raising out of
embedded tensed complements.
Sentences containing the Haitian verb sanble to seem, to resemble9 may
have two surface realisations. In the first one, the subject of the embedded
clause is realised in the embedded clause and the subject position of the matrix
verb is optionally filled by the expletive pronoun li. In the second, the subject of
the embedded clause occurs in the subject position of the matrix verb and the
embedded subject position is filled by the resumptive pronoun li (see Massam
1989).
(35)

a. Li sanble Jan
te
malad.
it seem
John ant sick
It seems that John has been sick.
b. Jan
sanble li
te
malad.
John seem
res ant sick
John seems to have been sick.
[Lit.: John seems as if he had been sick.]10

haitian
(=(13) in Dumais 1988)
haitian

(=(12) in Dumais 1988)

There are two basic arguments showing that the complement clause of sanble is
tensed. First, as is pointed out by Dumais (1988), the clause embedded under
sanble may contain the marker of anteriority te, as shown in (35). This marker
can only appear in tensed clauses (see chapter 5). Second, as is extensively discussed in Law (1991, 1992) and Deprez (1992a), when the embedded subject is
questioned, the resumptive form ki must show up in the extraction site, as shown
in (36). This argues that the embedded clause must be tensed for, as we saw in
chapter 7, ki can only occur in tensed clauses.

9.6 Raising verbs

(36)

Ki
mounn ki
sanble
which person res seem
Who seems to have left?

ki
te
pati?
res ant leave

267

haitian
(=(8b) in Law 1992)

Hence, in (35b), the subject of the embedded tensed clause occurs in the position
of the subject of the matrix clause.
Like Haitian, French allows for two surface realisations of sentences containing the verb sembler to seem (see (37a and b) ). In the first, the subject
of the embedded clause is realised in the embedded clause and the subject
position of the matrix verb is obligatorily filled by the expletive il. In the second
one (see (37b) ), the subject of the embedded clause is raised to the subject
position of the matrix verb. Note, however, that, in this case, the embedded
clause is infinitival. The infinitival character of the embedded sentence in (37b)
is revealed by the lack of the [+tensed] complementiser, infinitival morphology
on the verb and the lack of a subject pronoun in the embedded clause. In (37c),
the subject has been raised out of a tensed clause and the resulting sentence is
not grammatical.
(37)

a.

Il semble que Jean soit malade.

it seem
that John be sick
It seems that John is sick.

french

b. Jean semble tre malade.

John seem
be
sick
John seems to be sick.

french

c. *Jean
John

french

semble qu il
seem
that he

soit malade
be sick

The contrast in grammaticality between (37b) and (37c) shows that the option
of raising the subject out of an embedded clause is only available in French
when the complement is infinitival. This is in direct contrast with Haitian, which
allows this option only out of tensed complements, as shown in (35b). The comparison of the syntactic properties of Haitian sanble and French sembler leads
one to conclude that these verbs do not have the same properties.
Dumais (1988) shows that, in Fongbe, sentences containing the verb # to
seem, to resemble may have two surface realisations, as shown in (38). In the
first one, the subject of the embedded verb occurs in its basic position and the
subject position of the matrix verb is optionally filled by the expletive , as
shown in (38a). In the second, the subject of the embedded verb appears in the
subject position of the matrix verb, and the subject position of the embedded
verb is obligatorily filled by the resumptive , as shown in (38b).
(38)

a. () 3
n
3w
Jan
it
seem thing that John
It seems that John is sick.

jzwn.
be-sick

fongbe
(=(13) in Dumais 1988)

268

The syntactic properties of verbs

b. Jan
3
my
(3i)
jzwn.
John resemble person op
res be-sick
John seems to be sick.
[Lit.: John resembles someone who is sick.]11

fongbe

(=(12) in Dumais 1988)

The Fongbe data in (38) resemble the Haitian data in (35) as follows. In both
languages, it is possible to relate the subject of the tensed sentential complement
of the verb meaning seem to the subject position of an embedded tensed clause.
The syntactic properties of the Haitian verb sanble to seem therefore follow
from the relexification hypothesis. The verb derives its phonological representation
from the phonetic form of the French verb sembler to seem. Unlike sembler
(see (37c) ), however, sanble allows the subject of the tensed complement of this
verb to be coindexed with its own subject position. This parallels the Fongbe
data in (38). This property shared by # and sanble is certainly related to the fact
that, in both Fongbe and Haitian (but not in French), the verb meaning to seem
also means to resemble. The facts in (35b) and (38b) reflect this situation.
Another point of interest here is the fact that subject raising out of a tensed
embedded clause of the type we find in Haitian and Fongbe is not a common
phenomenon in the languages of the world. Massams (1989: 105) comment on
the Haitian subject raising construction supports this claim.
this raising is not directly comparable to raising in languages such as English,
since it is seen to occur from a tensed clause with, in some cases, an overt
complementiser, and since it leaves a pronoun copy (obligatorily). Nor is it
comparable to raising in languages such as Niuean, Kipsigis, Moose Cree,
among others, since in these languages raising from finite clauses is possible
from positions other than subject (see Jake and Odden 1979; James 1984; Seiter
1980; and Massam 1985, for discussion of raising in these languages). In hc
only the embedded subject can (must) be coreferential with the matrix subject
. . . Rivero and Sainz (1986) discuss languages such as Modern Greek which
have the same raising characteristics as hc; however their analysis rests on a
morphological richness which does not appear in hc.

Massams comment suggests that the type of subject raising exhibited in

Haitian is rare, which makes it a marked phenomenon. The fact that parallel data
exist in Fongbe shows that a marked feature of the substratum language has
been transferred into the creole. For scholars who believe that the presence of
marked substratum properties in the creole is crucial for a relexification account
of creole genesis (see chapter 3, section 3.5), the raising properties of the Haitian
and Fongbe verbs discussed in this section constitute a case in point. Indeed, on
the assumption that subject raising out of an infinitival clause is the unmarked
option, we are left with the fact that Haitian has retained the marked option from
the substratum languages instead of adopting the unmarked option from the
superstratum language.
All three languages under comparison have a second verb meaning to seem:
genl, avoir lair and G, respectively. These verbs have similar raising properties to the verbs discussed above.

9.7 Existential verbs

9.6.3

269

Summary

The comparison of verbs which license argument raising in the three languages
being examined reveals the following striking similarities between Haitian and
its substratum language. First, in both languages, object-to-subject raising shares
properties which are not found in French. Moreover, both languages allow for
raising of an embedded subject out of a tensed clause, a marked option, in
contrast to French, which allows raising of an embedded subject only out of an
infinitival clause, the unmarked option.
9.7

Existential verbs

In Haitian, existential constructions are introduced by the verb gen to have,

which can also be used to express possession. These two uses of gen are illustrated in (39).
(39)

a. Gen yon pwoblm.

have a
problem
There is a problem.

b. Jan
gen yon liv.
haitian
John have a
book
John has a book.
(from Law 1994a)

In this section, I will be concerned with the properties of gen as used in existential constructions like (39a).
Law (1994a) convincingly argues that gen has the properties of a verb. For
example, it can be preceded by the negation marker pa and the anteriority
marker te. One striking property of gen is that it does not allow a pleonastic
pronoun to occur in its subject position, as shown in (40a). This verb does not
allow its argument to surface in the subject position either, as shown in (40b)
(see e.g. Ritter 1991; Vinet 1991; DeGraff 1992a, 1992d; Deprez 1992a).
(40)

a. Li
gen yon pwoblm.
it / he have a
problem
#There is a problem.
ok He has a problem.

b. *Yon
a

pwoblm
problem

gen
have

haitian

The facts in (40) might suggest at first glance that the subject position of existential gen must obligatorily remain empty. As has been pointed out by Law (1994a),
however, this cannot be true in view of facts involving extraction out of the clause
containing gen. Indeed, Law shows that, when gens argument is extracted out
of its clause, the resumptive ki must appear in the verbs subject position. These
facts are illustrated in (41).
(41)

a. Kisa ki
genyen?
what res have
What is there?

b. Kisa ou
di
ki
genyen?
what you say res have
What do you say there is?

haitian

Recall from chapter 7 that questions are clefts and that clefts are bi-clausal in
this language. Recall also that ki is a nominative resumptive which only occurs
in the extraction site of a subject. Law (1994a) argues that, since ki occurs in the

270

The syntactic properties of verbs

subject position of gen when its argument has been extracted, the extracted constituent must have transited through that position while moving out of the clause
where it was base-generated. On Laws account, this argues against a potential
claim to the effect that the subject position of gen must obligatorily remain
empty.12 When the argument has been extracted, the form of the verb is genyen
rather than gen, (see (41) ), a fact which still remains to be accounted for.13
The Haitian data described above compare to the corresponding French facts
as follows. As is the case in Haitian, existential constructions in French make
use of the verb avoir, which, like Haitian gen, is also used in possession constructions, as shown in (42). Note that (42a) also involves the clitic y which has
no counterpart in Haitian (see chapter 6).
(42)

a. Il
y
a
un problme. b. Jean a
un livre.
there cl have a
problem
John have a
book
There is a problem.
John has a book.

french

The French existential verb avoir does not allow its argument to surface in its
subject position, as is shown in (43). In this respect, the French facts in (43) and
the Haitian facts in (40b) are similar.
(43)

*Un
a

problme y
problem cl

a
have

french

French differs from Haitian, however, in the context of argument extraction:

in this case, the subject of avoir is not a transit site for the extracted argument
for, if it were, we would find the complementiser qui and we do not. Instead,
the subject position is occupied by the pleonastic pronoun il. Compare (41)
and (44).
(44)

a. Quest-ce quil y a?
What is there?

b. *quest-ce qui a

french

Unlike in Haitian, the form of the French verb is not altered in the context of
argument extraction: compare (41) and (44). So the syntactic properties of French
il y a and Haitian gen differ on many points. Furthermore, gen is not phonologically derived from il y a [ya]14 so it seems that the two existential constructions have few features in common.
Haitian gen also differs from the corresponding Fongbe expression. In Fongbe,
the existential construction makes use of a verb meaning to exist, as shown in
(45). Furthermore, the argument must precede the verb. Sentence (45b) is not
grammatical because the argument and the verb are not in the right surface
position.
(45)

a. X
3kp tn.
story one
exist
There is a problem.

b. *()
3rd

tn
x
exist story

3kp
one

fongbe

The Fongbe verb tn used in existential constructions is not used in possessive

constructions. In the latter case, the verb # is used, as shown in (46).

9.8 Control verbs

(46)

Kwk 3 wm.
Koku has a book.

271

fongbe

This contrasts with Haitian, where the same verb occurs in both constructions
(see (39) ).
Given that the properties of gen resemble neither those of French nor those of
Fongbe, we may ask whether they could be derived from an equivalent lexical
entry in some other West African language. Koopman (1986: 248) points out, however, that no West African language has a form with the properties of Haitian
gen. Following the methodology outlined in chapter 3 (section 3.4), we therefore
have to conclude that gen was not created through relexification but is an innovation, as DeGraff (1992b) also concluded.
9.8

Control verbs

This section compares the selectional properties of control verbs in Haitian with
those of its contributing languages. Control verbs are verbs which allow an argument of a matrix sentence to be coindexed with an empty argument position in
the complement clause of the matrix verb. The three-way comparison of Haitian,
French and Fongbe shows striking similarities between Haitian and Fongbe,
which both contrast with French in the same way. Following the methodology
in Koopman (1986), I will distinguish between two-variable and three-variable
control verbs.
9.8.1

Two-variable control verbs

As Koopman (1986) shows, two-variable Haitian control verbs of the want-class

may select either a tensed or an infinitival complement. These two types of complements will be discussed in turn.
9.8.1.1 Tensed complements
The Haitian verb vle to want may select a tensed complement introduced by
the complementiser pou, discussed in chapter 7. The embedded subject may be
either coreferential with the matrix subject or disjoint in reference from it (see
Koopman 1986; Sterlin 1988, 1989). This is shown in (47).
(47)

Lii vle
pou
li /j vini.
he want comp he come
He wants to come. or He wants him to come.

haitian
(=(18b) in Koopman 1986)

Koopman (1986: 240) points out that the two interpretations possible in Haitian
are not available in French. In contrast to Haitian, the tensed complement of
vouloir to want only allows for disjoint reference of the subject, as shown in
(48).

272
(48)

The syntactic properties of verbs

Ili
veut qu
ilj
vienne.
3rd want comp 3rd come
He wants him to come./ #He wants to come.

french
(=(15b) in Koopman 1986)

The Haitian facts in (47) are, however, parallel to the Fongbe facts below. The
Fongbe verb jl to want in (49) selects a tensed complement introduced by
the form n, which corresponds to Haitian pou (see chapter 7). As in Haitian, the
embedded subject can be either coreferential with the matrix subject or disjoint
in reference from it.
(49)

i
jl
n
i /j n
y.
3rd want comp 3rd sub leave
He wants to leave. or He wants him to leave.

fongbe

Koopman (1986: 241) provides similar examples from Vata and Akan, other
substratum languages of Haitian.
Other Haitian verbs of the want-class such as renmen pou to like, pito pou
to prefer, krenn pou to fear, mande pou to request, bezwen pou to need
have the same properties as vle pou in (47) (see Sterlin 1988, 1989, for an
extensive discussion of these data). The corresponding French verbs all have the
same properties as vouloir, which, as we have seen, differ from those of Haitian.
Finally, other Fongbe verbs such as b n to like, byV n to request, etc.,
have the same properties as jl n to want and thus have the same properties as
the corresponding Haitian verbs. Thus, the Haitian verbs of the want-class do
not have the syntactic properties of the French verbs that they were phonologically
derived from. They do, however, have the same properties as the Fongbe verbs
that they were relexified from.
9.8.1.2 Infinitival complements
In all three languages under comparison, verbs of the want-class may take
an infinitival complement in which the matrix subject binds a position in the
embedded clause. In Government and Binding theory (see Chomsky 1981), the
controlled position is represented as pro, a label which stands for an abstract
pronoun that is not pronounced but that has referential properties. This is illustrated in (50) (for Haitian, see Koopman 1986; Sterlin 1988, 1989; for Fongbe,
see Kinyalolo 1992, Lefebvre 1993b).
(50)

a. Jani vle
proi kraze
manchin-nan.
John want
destroy car
det
John wants to destroy the car.

haitian

auto.
b. Jeani veut proi dtruire l
John want
destroy det car
John wants to destroy the car.

french

proi n
gb
mvt
c. Kwki jl
Koku want
def-fut destroy car
Koku wants to destroy the car.

v.
det

fongbe

9.8 Control verbs

273

Verbs of the want-class in all three languages may thus select an infinitival
complement in which the matrix subject controls the subject position of the
tenseless complement clause. There is not much to say about the above data
except for the fact that, in Fongbe, the definite future marker n, discussed in
chapter 5, must appear in the embedded clause,15 whereas, in Haitian, the corresponding marker ap does not appear in this context. This may be a consequence
of the fact that, as proposed in chapter 5, the definite future marker ap became
part of the Haitian lexicon as the result of reanalysis. On this analysis, it is likely
that this marker has lost some of the specifications of the original marker that it
was copied from.
The most interesting facts about the infinitival complements of verbs of the
want-class in the three languages under comparison, however, are shown in (51).
In addition to taking an infinitival complement of the type in (50), both Haitian
and Fongbe, but not French, may select an infinitival complement containing an
overt subject. Unlike the embedded covert subject in (50), however, the embedded
overt subject in (51) must be referentially disjoint from the matrix subject.
(51)

a.

Lii vle lj vini.

He wants him to come.

haitian

b. *Il veut lui venir

[Lit.: He wants him to come.]

french

c.

fongbe

i b j y.
He wants him to come.

In the remainder of this section, I will discuss the Haitian and Fongbe facts
presented above.
Building on a remark in Koopman and Lefebvre (1982), Sterlin (1988, 1989)
shows that complements of verbs of the want-class may, in some cases, appear
without the complementiser pou (discussed in chapter 7). However, Sterlin (1988,
1989) points out the difference in the referential properties of the embedded subjects in (52a and b). In (52a), the subject of the embedded clause may be either
coreferential with, or disjoint in reference from, the matrix subject. In (52b), the
subject of the embedded complement must be referentially disjoint from the
matrix subject.
(52)

a. Lii vle
pou
li /j vini.
he want comp he come
He wants for him to come.

b. Lii vle
lj
vini.
haitian
he want he come
He wants him to come.
(=(8) in Sterlin 1988)

Sterlin argues that, while the embedded clause in (52a) is tensed, the embedded
clause in (52b) is not. The embedded clause in (52a) is tensed because it allows
for the expression of the anterior marker te (disallowed in a tenseless clause)
(see (53a) ); it also allows for ki [+nominative] in the extraction site of the
embedded subject (see (53b) ) (a morpheme which is not allowed in tenseless
clauses, see chapter 7).

274
(53)

The syntactic properties of verbs

a. Lii vle
pou
li /j te
vini.
he want comp he ant come
He wants him to have come.

haitian
(=(22a) in Sterlin 1988)

b. (Se)
ki
mounn li vle
(it-is) which man
he want
Who does he want to come?

pou
ki
vini?
comp res come

haitian

(=(19c) in Sterlin 1988)

By contrast, the embedded clause in (52b) is not tensed because it does not
allow for the expression of the anterior marker te, as shown in (54a), or for ki
[+nominative] in the extraction site of the embedded subject, as shown in (54b).
(54)

a. *Li vle l te vini

[Lit.: He wants him to have come.]

haitian
(=(21a) in Sterlin 1988)

b. *(Se) ki mounn Jan vle ki vini?

[Lit.: Who does John want to come?]

haitian
(=(19d) in Sterlin 1988)

The difference in the finiteness of the two sentences in (52) raises the question
of what Case is assigned to the embedded subjects, for if the embedded clause in
(52b) is non-finite, then its subject cannot be assigned Nominative Case. Sterlin
argues that, in (52a), the subject of the embedded clause is assigned Nominative
Case. Support for this claim comes from the fact that this position allows for a
[+nominative] resumptive pronoun when the subject is extracted (see (53b) ). On
the basis of the fact that a [+nominative] resumptive pronoun is not allowed in
the extraction site of the embedded subject of the non-finite clause in (52b), as
was shown in (54b), Sterlin concludes that the embedded subject in (52b) is not
assigned Nominative Case. Because there is no complementiser in this sentence,
the embedded subject of the non-finite clause is in the Case-assigning domain
of the main verb. She proposes that the embedded subject in (52b) is assigned
Accusative Case by the matrix verb under what Chomsky (1981) refers to as
Exceptional Case marking.
The difference in the referential properties of the embedded subjects in the
sentences in (52) follows directly from the above analysis. In (52a), the embedded clause is introduced by the [+tense] complementiser pou. This complementiser
creates, for the embedded clause, a Binding domain which is distinct from that
of the main clause. Hence, the subject of the embedded clause may be interpreted
as being either co-referential with, or referentially disjoint from, the subject of
the main clause. By contrast, in (52b), the embedded clause is not introduced by
any complementiser, and so the subject of the embedded clause is in the same
binding domain as the subject of the main clause. Consequently, it must be
referentially disjoint from the matrix subject. As we saw in chapter 6, Haitian
pronominal forms are not morphologically distinguished on the basis of Case,
and thus the two pronominal forms in (52) are phonologically identical. The
Fongbe data are more transparent in this respect.

9.8 Control verbs

275

As is shown in (55), Fongbe has data which resemble the Haitian data in (52).
The Fongbe embedded clause in (55a) has the same properties as the Haitian
embedded clause in (52a). First, it is introduced by the complementiser n (or
n) which corresponds to Haitian pou (see chapter 7). Second, the embedded
subject pronoun may be either coreferential with, or referentially disjoint from,
the subject of the main clause. The Fongbe embedded clause in (55b) has the
same properties as the Haitian embedded clause in (52b). It is not introduced by
a complementiser and thus its subject is in the domain of the main verb and must
be disjoint in reference from the matrix subject.16
(55)

a. i
jl
n / n i /j n
y.
3rd want comp 3rd sub leave
Hei wants that hei /j leave.

fongbe

b. i jl j y.
He wants him to leave.

fongbe

Strikingly enough, the embedded subject pronouns in (55a and b) are morphologically distinguished for Case. As we saw in chapter 6, the third-person clitic
bears a low tone when it occurs in a [+objective] context and a high tone elsewhere. In (55b), the embedded subject clitic bears a low tone, showing that it
has been assigned Objective Case by the verb of the main clause, whereas in
(55a) the clitic bears a high tone, the tone it has in the subject position of any
tensed clause. The Fongbe facts can be accounted for straightforwardly along
the lines of the analysis proposed by Sterlin for Haitian. In (55a), the embedded
subject is assigned Nominative Case within the tensed embedded clause, whereas
in (55b) it is assigned Accusative Case by the matrix verb under Exceptional
Case marking.
The striking fact about the data discussed in this section is that, while French
verbs of the want-class can select only one type of infinitival complement (without an overt subject), both Haitian and Fongbe select two types of such complements. In the first type, there is no overt subject, but the subject of the embedded
verb is understood as being coreferential with that of the matrix clause (see (50) ).
In the second type, there is an overt subject and, in both languages, it must be
referentially disjoint from the matrix subject (see (51) ). The selectional properties of Haitian verbs must follow from the relexification hypothesis. Furthermore,
the availability of an overt subject is attributable to the fact that the Haitian and
Fongbe verbs can assign Accusative Case to the subject of the embedded clause,
a property that French verbs do not have. This syntactic property of Haitian
verbs must follow from the relexification hypothesis since it is a property of
substratum but not superstratum verbs. Whether Exceptional Case marking should
be considered as marked is debatable. Clearly, though, it has more than merely
survived in Haitian creole, as demonstrated by the number of Haitian verbs
which have it (see Sterlin 1988, 1989).
The Haitian data presented in this section further challenge the claim in the
literature that all pidgins and creoles with European lexifying languages seem

276

The syntactic properties of verbs

to have obliterated the finite/non-finite distinction and leave no ground for

speaking of infinitival uses of verbs (Mufwene 1986: 139; see also Bickerton
1984), since Haitian verbs, like Fongbe verbs, can be shown to have infinitival
complements.
9.8.2

Three-variable control verbs

On the basis of a survey of the selectional properties of three-variable control

verbs in Haitian, French and West African languages, Koopman (1986: 242)
concludes: The picture that emerges is clear: although the phonetic shape of the
Haitian verbs is clearly derived from French, their selectional properties are
rather different from those of French, and strikingly similar to those observed in
West African languages. The selectional properties of verbs meaning to promise and to ask (request) in Haitian, French and Fongbe illustrate this situation.
In Haitian, the verb pwmt to promise does not select an infinitival complement, as shown in (56a). It does select a tensed complement introduced by
the complementiser pou. In this case the embedded subject may be either corerefential with, or referentially disjoint from, the matrix subject, as is shown in
(56b) and (56c), respectively.
(56)

a. *Mi pwmt Jan

proi vini
I
promise John
come
[Lit.: I promised John to come.]

haitian

pou
mi vini.
b. Mi pwmt Jan
I
promise John comp I
come
I promised John that I will come.

haitian

c.

pou
Marij vini.
Mi pwmt Jan
I
promise John comp Mary come
I promised John that Mary will come.

haitian

In contrast to the Haitian verb pwmt, the French verb promettre to promise
may take an infinitival complement, as in (57a). As is the case in Haitian, however, the French verb may also take a tensed complement and, in this language
as well, the embedded subject may be either coreferential with or disjoint from
the matrix subject, as is shown in (57b) and (57c), respectively.
(57)

a. Ji ai
promis
Jean
I
aux promise case John
I promised John to come.

de
proi venir.
comp
come

promis
Jean que
jei
b. Ji ai
I
aux promise case John comp I
I promised John that I will come.

viendrai.
come-fut

promis
Jean que
Mariej
c. Ji ai
I
aux promise case John comp Mary
I promised John that Mary will come.

viendra.
come-fut

french

french

french

9.8 Control verbs

277

So, although the French (see (57) ) and Haitian (see (56) ) data are similar for (b)
and (c), they differ for (a).
The Fongbe data below, however, conform exactly to the Haitian data in (56).
As is shown in (58), the notion to promise in Fongbe is rendered by the verb
#W which basically means to say. This verb does not select an infinitival complement, but it does select a tensed complement with two referential possibilities
for the embedded subject.
(58)

a. *Ni 3w n
Kwk proi n
I
say comp Koku
def-fut
[Lit.: I promised Koku to come.]

w
come

Kwk 3w
ni n
b. Ni 3w n
I
say comp Koku comp I def-fut
I promised Koku that I will come.
c.

fongbe

w.
come

Kwk 3w
sbj n
Ni 3w n
I
say comp Koku comp Asiba def-fut
I promised Koku that Asiba will come.

w.
come

fongbe

fongbe

Thus, while the subcategorisation properties of the Haitian verb meaning to

promise (see (56) ) do not entirely match those of the French verb from which
it is phonologically derived (see (57) ), they correspond exactly to those of the
equivalent Fongbe expression (see (58) ).
A similar example is provided by the verbs meaning to ask (request) in
the three languages under comparison. Note that, with verbs meaning to ask
(request), the coreferential properties of the embedded subject are a function of
those of the matrix object rather than the matrix subject. While the French verb
selects an infinitival complement, the Haitian and Fongbe verbs do not, as is
shown in (59).
(59)

a. *Yo mande Marii proi pati

[Lit.: They asked Mary to leave.]

haitian
(=(20b) in Koopman 1986)

b. Ils ont demand Mariei de proi

They asked Mary to leave.
c. *Yi byv Marii proi y
[Lit.: They asked Mary to leave.]

partir.

french
fongbe

On the other hand, while the French verb does not select a tensed complement
with the coreferential properties in (60b), the Haitian and Fongbe verbs do.
(60)

a.

Yo mande Marii pou li pati.

They asked Mary to leave.
[Lit.: They asked Mary that she leave.]

haitian
(=(20a) in Koopman 1986)

278

The syntactic properties of verbs

b. *Ils ont demand Mariei quellei parte
[Lit.: They asked Mary that she leave.]
c.

Yi byv Marii 3w i n y.
They asked Mary to leave.
[Lit.: They asked Mary that she leave.]

french
fongbe

Thus, both the Haitian and Fongbe verbs differ from the French verb in the same
way.
The two sets of examples above show that three-place control verbs in
Haitian and Fongbe share the same selectional properties and contrast with the
corresponding French verbs. This follows straightforwardly from the relexification
hypothesis. Koopman (1986: 240) remarks that three-place control verbs corresponding to French verbs such as convaincre to convince, ordonner to order,
and persuader to persuade are difficult or impossible to find in both Haitian
and the West African languages. This gap in the Haitian lexicon also follows
from the relexification hypothesis: the creators of Haitian simply did not have
such lexical entries to relexify.
9.8.3

Summary

While the selectional properties of Haitian control verbs differ from those of the
corresponding French verbs, they are the same as those of the Fongbe verbs. In
both Haitian and Fongbe, but not in French, the subject of the tensed complement of verbs of the want-class may be coreferential with the matrix subject. In
both Haitian and Fongbe, but not in French, verbs of the want-class may select
an infinitival complement with an overt subject which is assigned Accusative
Case by the main verb under Exceptional Case marking. Finally, three-variable
control verbs in Haitian and Fongbe share properties which distinguish them
from French. The syntactic properties of Haitian verbs must have been transferred into the creole through relexification.
9.9

Light verbs

Light verb constructions involve a verb and an object as in take a walk in

English. A major characteristic of these constructions is that the verb contributes
very little to the semantics of the construction; rather, it is the object that is determinative. All three languages under comparison have light verb constructions, as
shown in (61).
(61)

fongbe
fn hwn

french
faire la guerre

haitian
f lag
make war

In Fongbe, however, there are a number of light verb constructions which correspond to simplexes in French and Haitian. Examples are provided in (62). The
Fongbe data are from Brousseau (1988b), the Haitian data from my own field
notes.

9.9 Light verbs

(62)

fongbe
w zv
do work

french
travailler

haitian
travaye

to work

k
tw
die river

se noyer

nwaye

to drown

jy
tgb
fall problem

sinquiter

enkyete

to worry

jy
kp
fall wound

se blesser

blse

to hurt oneself

d
gn
plant metal

enchaner

anchnnen

to chain up

j
hn
give birth song

chanter

chante

to sing

s
x
close speech

crier

kriye

to shout

bl n
tie mouth

jener

jennen

to fast

kn
pick up

courir

kouri

to run

wzn
run

279

The question here is whether the data in (62) constitute evidence for or against
the relexification hypothesis. The answer to this question rests on whether light
verbs and their objects constitute lexical entries or not.
In the recent literature, light verbs have been analysed as complex predicates
listed in the lexicon (see e.g. Cattell 1984; Grimshaw and Meister 1988; Travis
in press). On the basis of phonological and syntactic arguments (e.g. various
types of extraction facts), Brousseau (1988b) extensively argues that the objects
of the light verbs in (62) differ from ordinary objects. She concludes that Fongbe
light verb constructions must therefore be listed in the lexicon. Since they are
listed in the lexicon, they should undergo relexification. How does relexification
proceed in this case?
Substratum speakers who had lexical entries like the Fongbe ones in (62)
searched in the superstratum language for phonetic strings to relabel the lexical
entries copied from their own lexicon. They found the French simplexes in (62)
and used them for this purpose, yielding the Haitian lexical entries in (62). The
fact that simplexes in the superstratum language may be used to relabel light
verb constructions during relexification is in line with the analysis that these
constructions constitute lexical entries. In turn, this is additional evidence for
analyses holding that light verb constructions are complex predicates that are
listed in the lexicon. This conclusion is further reinforced by the fact that, when
both the substratum and superstratum languages encode a notion with a light

280

The syntactic properties of verbs

verb construction, the creole ends up with a light verb construction as well, as
shown in (61).

9.10

Inherent object verbs

Inherent object verbs are verbs which are semantically autonomous. Unlike the
objects of light verbs, their objects do not contribute to the meaning of the verb
itself. These verbs can take different types of objects, but they cannot surface
without an overt object of some kind. When no specific object is intended, these
verbs will appear with the typical object that is appropriate for a given verb, or
an object meaning thing. Fongbe has several such verbs. Their objects may be
cognate to the verb, as in nW nW [Lit.: suck breast] to suckle (see Avolonto
1995), or not, as in kn hn [Lit.: drive vehicle] to drive.
On the basis of data from English, Massam (1990) argues that cognate objects
behave like ordinary objects and consequently cognate object verbs are listed in
the lexicon independently of their object. Based on Fongbe data, Brousseau
(1988b) argues that the objects of inherent object verbs share syntactic characteristics with ordinary direct objects, whether or not they are cognate to the verb.
Her conclusion is thus the same as Massams. These analyses predict that, in
relexification, the inherent object verbs will be relexified independently of their
objects. Since taking an inherent object is a property of verbs, we expect the
Haitian verbs corresponding to inherent object verbs in the substratum to reproduce this property. This prediction is only partially borne out by the data, as we
see in (63). Only three Haitian verbs follow the pattern of the substratum language in requiring an inherent object (see (63a) ). In (63b) the inherent objects of
the substratum verbs are not reproduced in the creole any more than the cognate
objects in (63c).
(63)

fongbe
a. gbw
zwn
calm
disease

french
calmer douleur

haitian
kalme doul

to heal
to hunt

ny
hunt

gb
animals

chasser

chase bt

3n
draw

sn
water

puiser

tire
dlo
to draw (water)
draw water

hn
vehicle

conduire

kndwi

to drive

b. kn
drive
3
eat

n
manger
something

manje

to eat

z
sweep

y
ground

bale

to sweep

balayer

9.10 Inherent object verbs

c. nw
suck

nw
breast

tter

tete

to suckle

3w
pee

3w
piss

pisser

pise

to pee

kpxn
cough

kpxn
cough

tousser

touse

to cough

3
3
believe belief

croire

kw

to believe

kwn
kwn
whistle whistle

siffler

sifle

to whistle

3
eat

gagner

genyen

to win

3
food

281

The difference between the Fongbe verbs in (63a and b) and their corresponding
verbs in French and in Haitian may best be stated in terms of their transitivity
properties: whereas the Fongbe verbs are necessarily transitive, the Haitian and
French verbs may also be used intransitively. How can the discrepancy between
the three languages under comparison be accounted for and what is the relevance of these data for a theory of creole genesis?
It has long been noted in the literature that some transitive verbs may also be
used intransitively in some languages. To eat in English is such a verb: along
with He ate an apple we find He ate. In the three languages under comparison in
this book, the verbs meaning to eat do not behave in the same way in this
respect (see (63b) ). While in Haitian and French the verbs meaning to eat may
be either transitive or intransitive, in Fongbe this verb requires an overt object
and thus it looks as though it cannot be intransitive. This contrast is shown in
(64).
(64)

a. Li manje.
Il a mang.
He ate.

haitian
french

b. 3 nl. / * 3
He ate something. = He ate.

fongbe

In order to characterise the difference between Haitian and French, on the one
hand, and Fongbe, on the other, I will appeal to the analysis of verbs like to eat
in Guerssel (1986).
Verbs of the eat-class are those for which the selectional restrictions that hold
between the verb and its object impose a limitation on the nature of the object
variable. In order to express this fact, Guerssel (1986: 37) proposes that the lcs
of verbs of this type contains a clause identifying the variable. Hence, the lcs of
a verb such as to eat would be: x eat y, where y is typically food. When the
verb is used transitively, the y variable is linked to an argument position in the
syntax. When it is used intransitively, the variable representing the patient role
is not projected into the syntax. Guerssel (1986: 37) assumes that this property

282

The syntactic properties of verbs

holds of those variables that are fully identified in the appended clause in a
verbs lcs. As a result of the optional projection of fully identified variables, the
Predicate Argument Structure of eat will be represented in one of the following two ways, where the appended clause has been omitted. These are represented in (65).
(65)

a.

b.

arg.

x eat y

x eat

(=(60) in Guerssel 1986)

Guerssel points out that, while the structure in (65b) conforms to the Theta
Criterion (see Chomsky 1981), the structure in (65a) does not. In order to remedy
this deficiency, Guerssel (1986: 37) proposes a mechanism whereby a constant
identifying an unlinked variable in lcs is substituted for that variable. Applying
this constant substitution procedure to (65a) yields the predicate argument structure in (66), which is well formed.
(66)

v
v
x eat-food

(=(61) in Guerssel 1986)

The structures in (65b) and (66) derive the transitive and intransitive uses,
respectively, of eat-class verbs.
This analysis provides a simple account of the difference between the
Haitian/French and Fongbe data in (64). The intransitive use of the verbs meaning to eat in Haitian and in French is derived from the structure in (66). The
Fongbe object nL thing in (64b) may be seen as the spell-out of the constant in
the lcs in (66). In this view, Haitian/French and Fongbe verbs in (63b) differ
according to whether the constant in the lcs in (66) must be spelled out or not.
Cognate objects may be seen as another way of spelling out the constant in
the lcs of a verb. In contrast to Fongbe, the Haitian and French verbs in (63c)
surface as simple intransitive verbs, e.g. Li tete and Il a tt He suckled,
respectively, showing that, in this case as well, the constant in the lcs of the
verb need not be spelled out. Here again, Haitian/French and Fongbe may
be claimed to differ based on whether the constant in the verbs lcs needs to
be spelled out. With regard to this property, Haitian follows the option of the
superstratum language rather than that of the substratum language. Presumably,
the creators of the creole perceived that the French verbs above could occur
without an object. On the basis of this evidence, they abandoned the requirement
that objects of verbs of the eat class be projected in the syntax. It is therefore

9.11 The Case-assigning properties of verbs

283

not surprising to find that, as is extensively documented in Dumais (1992),

Haitian has none of the cognate objects that exist in Fongbe.17
9.11

The Case-assigning properties of verbs

In the theory of the lexicon adopted for this research, the Case properties of
verbs are registered in their lexical entries. Individual verbs are thus specified
for whether they have a structural Case to assign to their object. Case-assigning
properties thus constitute another dimension of the syntactic properties of verbs
along which Haitian can be compared with its contributing languages. It will be
shown that the Case-assigning properties of Haitian verbs seem quite free compared to the corresponding verbs in both its superstratum and substratum language. The section will end with a discussion of why this should be so.
The Case-assigning properties of verbs were tested on the basis of a sample
of 93 triplets of verbs. The sample was built as follows. The first five Fongbe
verbs under each letter of the alphabet were selected. The corresponding verbs in
Haitian and French were then listed. The selectional and Case-assigning properties of each verb were then established with native speakers of the three languages. The triplets of verbs were then divided into two major groups: triplets
with the same Case-assigning properties and triplets showing mismatches. Out
of the 93 triplets of verbs, 76 show the same Case-assigning properties in all
three languages. This should come as no surprise for these verbs are typical
transitive or intransitive verbs which are likely to have the same Case-assigning
properties across languages. The interesting cases for the discussion of the issues
at stake in this book are the non-matching cases. There are 17 non-matching
triplets in the sample. These are of various types which will be discussed in turn.
The first group of mismatches consists of five triplets where the Haitian and
Fongbe verbs differ from French in the same way. Two of these involve Haitian
and Fongbe double-object verbs. In (67a), the Haitian and Fongbe verbs meaning
to give assign their structural Case to the Recipient (see section 9.12), whereas
the corresponding French verb in (67b) assigns its structural Case to the Theme.
(67)

a. n Mari svn.
Li bay Mari krab.
He gave Mary some crab.

fongbe
haitian

b. Il a donn du crabe Marie.

He gave some crab to Mary.

french

As will be discussed in detail in section 9.12, while Fongbe and Haitian have the
double-object construction, French does not.
One triplet involves Haitian and Fongbe v+np constructions where the additional argument is realised in a prepositional phrase. This contrasts with the
corresponding French verb, a simplex that assigns Accusative Case to its object.
The pertinent facts are illustrated in (68).

284

The syntactic properties of verbs

(68)

a. x pl
n
Mari.
Li bat bravo pou Mari.
He applauded Mary.

fongbe
haitian

b. Il a applaudi Marie.
He applauded Mary.

french

Another triplet involves verbs participating in the Locative alternation. For

example, in both Fongbe and Haitian, the verb to stuff assigns Accusative Case
to the thing that is being stuffed into something, whereas in French, the verb assigns
Accusative Case to the Locative argument. This contrast is illustrated in (69).
(69)

a. t
vvknf 3 matela v my.
Li boure kotn
nan matela a.
He stuffed cotton in the mattress.

fongbe
haitian

b. Il a bourr le matelas de coton.

He stuffed the mattress with cotton.

french

Finally, there is one triplet of verbs where both Fongbe and Haitian have a
serial verb construction that contrasts with a simple French verb meaning to
send. In French (see (70b) ), the verb assigns Accusative Case to the Theme and
the Goal is introduced by the preposition . In Fongbe and Haitian, the first verb
of the series assigns its Case to the Theme np, and the second assigns its Case to
the Goal np (see Lefebvre 1991b).
(70)

a.
Li
he
He

ty
voye
send
sent a

kn
sy-d
telegram bay
telegram send / give
telegram to Mary.

b. Il a envoy un tlgramme Marie.

He sent a telegram to Mary.

Mari.
Mari.
Mary

fongbe
haitian

french

The above cases exhaust the types of verb triplet mismatches where Haitian contrasts with French and pairs with Fongbe. Given the methodology adopted here, the
Case properties of the Haitian verbs above should be considered as having been
transferred from the substratum language into the creole through relexification.
The second group of mismatches consists of nine triplets in which the Case
properties of the Haitian verbs differ from those of the Fongbe equivalents.
Of these triplets, seven are of the type where the Fongbe verbal expression is
expressed by a light verb construction and the object by a prepositional phrase.
The Haitian and French verbs, however, are simplexes that assign Accusative
Case to their objects (see section 9.9). These contrasting data are illustrated in
(71) using the verbs meaning to like/love.
(71)

a. Li renmen Mari.
Il aime
Marie.
He likes/loves Mary.

haitian
french

b. yjwn
n Mari.
he accept- smell for Mary
He likes/loves Mary.

fongbe

9.11 The Case-assigning properties of verbs

285

As we saw in section 9.9, Fongbe light verb constructions are often relexified by
French simplexes. It appears that in these cases, the Case property of the French
verbs came along with their phonetic matrices.
The reverse situation is found in two triplets where the Fongbe verb is a
simplex that assigns Accusative Case to its object, whereas the corresponding
Haitian and French verbs have no Case to assign to their object, which therefore
has to be rendered by a prepositional phrase. This is shown in (72).
(72)

a. Jan marye
ak
Mari.
Jean sest mari avec Marie.
John got married to Mary.

haitian
french

b. Kwk d Mari.
Koku married Mary.

fongbe

Here also, the Case-assigning properties of the French verbs were carried over
into Haitian. This exhausts the cases where Haitian and French verbs share the
same Case-assigning properties and contrast with Fongbe.
The third group of mismatches comprises three verb triplets where the Case
properties of the Haitian verbs match those of neither French nor Fongbe. The
first triplet involves the verbs meaning to resemble. In Fongbe, this verb assigns
Accusative Case to its object, and in French the complement of this verb is introduced by the Dative preposition , but in Haitian the complement of this verb is
a prepositional phrase headed by ak with.
(73)

a. 3
tw
tvn.
he resemble father his
He resembles his father.

fongbe

b. Il ressemble son pre.

He resembles his father.

french

c. Li sanble
ak
papa l.
he resemble with father his
He resembles his father.

haitian

The second triplet involves the verbs meaning to graze/scrape. In both Fongbe
and French, this verb assigns Accusative Case to its object. Haitian departs from
both of its source languages: the verb has no Case to assign to its object, which
must be rendered in a prepositional phrase.
(74)

a. klxn fv.
He grazed his foot.

fongbe

b. Il sest corch le pied.

He grazed his foot.

french

c. Li kche nan pye

he graze in
foot
He grazed his foot.

l.
his

haitian

286

The syntactic properties of verbs

The third triplet involves the verbs meaning to write. The Fongbe predicate
is rendered by a serial verb construction where each verb assigns Accusative
Case to an internal argument. In French, the verb assigns Accusative Case to
the Theme, and the Goal argument is introduced by the Dative preposition .
In Haitian, the verb is a double-object verb which assigns Accusative Case to
the Recipient. These facts are illustrated in (75).
(75)

a. wln wm sy-3 Sk.

he write letter send Cica
He wrote a letter to Cica.

fongbe

b. Il a crit une lettre Marie.

He wrote a letter to Mary.

french

c. Li ekri Mari yon lt.

He wrote Mary a letter.

haitian

These three triplets show that a Haitian verb may have Case-assigning properties
that differ from both Fongbe and French, even when the Fongbe and French
verbs share the same properties (see (74) ). Following the methodology in chapter 3, the Case properties of the Haitian verbs in (73)(75) would be assumed
either to come from another substratum language (through relexification and
dialect levelling) or to constitute cases of independent development from within
the creole.
To the best of my knowledge, the various types of mismatches discovered
on the basis of the sample of 93 verbs are representative of the types of differences between the Case properties of verbs in the three languages. The results
of this study show, however, that the Case properties of Haitian verbs are less
consistent in terms of the relexification hypothesis than, for example, the subcategorisation properties discussed earlier in this chapter. In only 7 out of 19
cases do the Haitian verbs have the same Case properties as those of the substratum language when the source languages differ. Furthermore, each group
of mismatches could be used to support a different theory of creole genesis. The
first group supports the relexification hypothesis. The second group supports the
view that at least some properties of French verbs were acquired by the creators
of Haitian, and the third group could be used to support the view that creoles are
not like any of their contributing languages. In my view, the overall results of
the examination of the sample of verbs reported on above suggest that the Case
properties of verbs probably constitute the syntactic properties that are most
independent of the source languages in creole genesis. This conclusion is not
surprising, for the following reasons. First, Case is a rather superficial property
of verbs as compared with other syntactic properties (e.g. subcategorisation).
Second, the Case features of verbs constitute an area of grammar which is very
subject to change in the languages of the world.18 Given these facts, it should
come as no surprise if the Case properties of verbs in a given creole show a good
deal of variation compared to its contributing languages. The Case-assigning
properties of double-object verbs will be shown to further support this claim.

9.12 Double-object verbs

9.12

287

Double-object verbs

John sent Mary a letter is an example of the RecipientTheme construction (np

np), whereas John sent a letter to Mary is an example of the ThemeLocative
construction (np pp). These alternations are generally referred to in the literature
as Dative Shift. Like West African languages, Haitian creole has the Recipient
Theme construction. It is a well-documented fact, however, that French does
not. This contrast between the three languages was first noted by Koopman
(1986). In this section, I present evidence illustrating the contrast between Haitian and Fongbe, on the one hand, and French, on the other. I shall begin with an
overview of the formal properties which distinguish np np from np pp constructions (section 9.12.1). Then I will establish the facts concerning the np np construction, showing that both Haitian and Fongbe have this construction, thereby
contrasting with French (section 9.12.2). While the Fongbe np np construction
exhibits two surface word orders, RecipientTheme and ThemeRecipient, the
Haitian np np construction has only one word order: RecipientTheme (section 9.12.3). Fongbe has only a few double-object verbs; Haitian has many more.
In section 9.12.4, I address this problem from the point of view of the Caseassigning properties of the verbs involved in the construction. The discussion
of the Haitian double-object construction partially draws on Veenstra (1992),
Lumsden (1994a), and on further work that I have done; and the Fongbe discussion is based on Lefebvre (1992d, 1993c, 1994c). The analysis of the French
facts is based on the literature on this construction (see e.g. Kayne 1984; Tremblay
1991).
9.12.1

The semantic and syntactic properties of the double-object

construction

In the literature on Dative Shift, it is claimed that the np np and np pp constructions are not semantically equivalent. First of all, it is argued that, in the former
construction, the Goal is interpreted as a Recipient while, in the latter, it is
interpreted as a Locative (see Green 1974; Oehrle 1976; Grimshaw 1989; Pinker
1989; Johnson 1991; Tremblay 1991; Lefebvre 1994c). I will therefore refer to
the np np construction as the RecipientTheme construction and the np pp construction as the ThemeLocative construction. Furthermore, it has been argued
that RecipientTheme constructions are possessional constructions (see e.g. Pinker
1989; Johnson 1991). As Pinker (1989: 48) states, however, possession need
not be literal; . . . verbs of communication are treated as denoting the transfer of
messages or stimuli which the recipient metaphorically possesses. This can be
seen in sentences such as He told her the story, He asked her a question, and
She showed him the answer.
As well, in the RecipientTheme construction, the Recipient is the affected
argument, while in the ThemeLocative construction, it is the Theme that is
affected. In the terminology of current research in Lexical Semantics, the affected

288

The syntactic properties of verbs

argument is the argument corresponding to the variable that undergoes a change

of state or of location in the lcs of a verb (see Hale and Keyser 1987). While the
Theme is generally the affected argument of a verb of change (see e.g. Rappaport
and Levin 1988), in the RecipientTheme construction the Recipient is the
affected object, that is, the participant which is being acted on (see Tenny 1987;
Pinker 1989: 212; Lefebvre 1994c). In this view, the meaning of the Recipient
Theme construction would be something like Somebody causes the Recipient to
become in possession of the Theme, whereas the meaning of the ThemeLocative
construction would be something like Somebody causes the Theme to undergo a
change of location to Goal. (For various more formal discussions of the semantics
of the two constructions, see e.g. Pinker 1989; Jackendoff 1990; Lumsden 1994a;
Lefebvre 1993c, 1994c.)
The RecipientTheme and ThemeLocative constructions are further distinguished on the basis of syntactic tests showing that, in the former, the Recipient
must asymmetrically c-command the Theme (see Barss and Lasnik 1986), whereas
in the latter, the Theme must asymmetrically c-command the Locative (see Larson
1988). In Lefebvre (1993c, 1994c), it is extensively argued that the notion of
c-command, as it applies in these constructions, must entail a relationship of
dominance rather than precedence between the Recipient and the Theme, on the
one hand, and between the Theme and the Locative, on the other hand.19 Syntactic
tests based on six types of phenomena which were developed by Barss and
Lasnik (1986) show that, in the RecipientTheme construction, the Recipient
must c-command the Theme. These tests involve facts related to (1) binding of
a pronoun by a quantifier, (2) reciprocals such as Each . . . the other, (3) superiority effects, (4) weak cross-over phenomena, (5) binding with a reflexive, and
(6) negative polarity items. I refer the reader to the relevant literature for an
exhaustive discussion of these tests. For the purposes of this discussion, I shall
provide one example showing the asymmetrical relationship that holds between
the Recipient and the Theme in the RecipientTheme construction and the
Theme and the Locative in the ThemeLocative construction, on the basis of
data involving binding of a pronoun by a quantifier in English. According to
Binding Theory (as outlined in Chomsky 1981, and related work), a quantifier
must c-command the pronoun it binds. Consider the facts in (76) involving the
RecipientTheme construction (from Barss and Lasnik 1986). In the (a) sentence,
the quantifier is part of the Recipient phrase and the sentence is grammatical. In
the (b) sentence, the quantifier is part of the Theme phrase and the sentence is
not grammatical.
(76)

a.

I showed everyi man hisi picture.

b. *I showed itsi owner everyi picture

english
english

The contrast in grammaticality between the (a) and the (b) sentences shows that,
in the RecipientTheme construction, the Recipient must c-command the Theme.
Now, consider the facts in (77) involving the ThemeLocative construction

9.12 Double-object verbs

289

(from Larson 1988). In the (a) sentence, the quantifier is part of the Theme
phrase, and the sentence is grammatical. In the (b) sentence, the quantifier is part
of the Locative phrase, and the sentence is not grammatical.
(77)

a.

I showed/gave everyi picture to itsi owner.

english

b. *I showed/gave hisi picture to everyi man

english

This contrast in grammaticality shows that, in the ThemeLocative construction,

the Theme must c-command the Locative. These asymmetries, and the others
mentioned above, constitute syntactic tests which permit one to distinguish the
RecipientTheme construction from the ThemeLocative construction. In light
of this theoretical background, I now turn to a discussion of the Recipient
Theme construction in the three languages under comparison.
9.12.2

The RecipientTheme construction in Haitian, French and Fongbe

Veenstra (1992) and Lumsden (1994a) have shown that, in Haitian creole, there
is a subset of verbs which can take two np complements, as is illustrated in (78).
(78)

Mwen bay / montre Pl liv

la.
I
give / show Paul book the
I gave/showed Paul the book.

haitian
(=(19) and (25b) in Lumsden 1994a)

The construction in (78) has the same properties as the English Recipient
Theme construction discussed above. First, certain semantic tests discussed in
Lumsden (1994a) show that, in these sentences, the Goal is interpreted as a
Recipient. Second, Veenstra (1992) shows that the Recipient is the affected
argument of the construction. Recall from chapter 8 that the event determiner
occurring in the context of an internal argument is permitted only for a definite
affected argument. As is shown in Veenstra (1992), in the RecipientTheme
construction, the event determiner is permitted only in the context of a definite
Recipient, as in (79a). By contrast, it is not permitted in the context of a definite
Theme, as is shown by the ungrammaticality of (79b).
(79)

a.

Yon ng montre pitit la

yon foto
a.
haitian
a
man show
child det a
picture Det
A man showed the child a picture, as we knew.
(=(6) in Veenstra 1992)

b. *Yon ng montre
a
man show

yon
a

pitit manchin
child car

nan an
det Det

haitian

The contrast in grammaticality between (79a) and (79b) argues that, in the
Haitian RecipientTheme construction, the affected object of the construction
is the Recipient. Furthermore, the RecipientTheme construction presents the
asymmetries which are characteristic of such constructions in English. Lumsden
(1994a) illustrates this fact with data involving binding of a pronoun by a

290

The syntactic properties of verbs

quantifier, as shown in (80). In (80a), the quantifier is part of the Recipient

phrase and the sentence is grammatical. In (80b), the quantifier is part of the
Theme phrase and the sentence is not grammatical.
(80)

a.

Mwen bay manman chak mounni nan foto

lii
haitian
I
give mother
each person the photo his
I gave each persons mother his photo.
(=(20a) in Lumsden 1994a)

yon foto
chak mounni haitian
b. *Mwen bay manman lii a
I
give mother
his the a
photo each person
(=(20b) in Lumsden 1994a)

The data thus show that, as is the case in English (see (76) ), in Haitian, the
Recipient must c-command the Theme in the RecipientTheme construction.
Veenstra (1992) also demonstrates that the other tests in Barss and Lasnik (1986)
yield the same results in Haitian as they do in English, thus providing further
evidence that the Recipient must c-command the Theme.20 All of these facts thus
argue for the claim that the RecipientTheme construction exists in Haitian.
It has long been noted in the literature on French syntax that French does not
have the RecipientTheme construction (see e.g. Kayne 1984; Tremblay 1991).
Hence, the French strings which are the counterparts of the Haitian sentences in
(78) are not grammatical, as shown in (81).
(81)

*J ai
donn / montr Paul le
livre
I aux give / show
Paul det book
[Lit.: I gave/showed Paul the book.]

french

French only allows a ThemeLocative construction where the Locative is assigned

Dative Case by (see Kayne 1984; Tremblay 1991). This is illustrated in (82).
(82)

J ai
donn / montr le
livre Paul.
I aux give / show
det book to Paul
I gave/showed the book to Paul.

french

Furthermore, on the basis of facts involving binding of a pronoun by a quantifier

(see (83) ), Lumsden (1994a) argues that, in the French sentences in (82), the
Theme must c-command the Locative. In (83a), the quantifier is part of the
Theme phrase and the sentence is grammatical. In (83b), the quantifier is part of
the Locative phrase and the sentence is not grammatical.
(83)

a.

J ai
donn une photo de chacun i sa i mre. french
I aux give
a
photo of each
to his mother
I gave a photo of each person to his mother.
(=(18a) in Lumsden 1994a)

b. *J ai
donn
I aux give
chacuni
each person

sa / leuri
his / their

photo la
photo to the

mre
mother

de
of

french

(=(18b) in Lumsden 1994a)

9.12 Double-object verbs

291

The contrast in grammaticality between the (a) and (b) sentences in (83) shows
that, in French, the Theme must c-command the Locative. In conclusion, then,
both semantic and syntactic facts support the claim that there is a Theme
Locative construction but no RecipientTheme construction in French.
Interestingly enough, the French data in (82) have no parallel in Haitian.
Although the sentence in (84) is grammatical, it is not a ThemeLocative construction. The prepositional phrase can only be interpreted as a Benefactive, not
as a Locative.
(84)

Mwen bay liv

la
pou Pl.
I
give book det for Paul
I gave the book for Paul. #I gave the book to Paul.

haitian

Furthermore, the prepositional phrase introduced by pou for in (84) is a syntactic

adjunct rather than an indirect object of the verb (see Lefebvre in preparation).
Finally, as is extensively argued in Lefebvre (1994c), in this type of language
the ThemeLocative pair of arguments is expressed by a serial verb construction, as is shown in (85).
(85)

Mwen pran liv

bay Pl.
I
take book give Paul
I gave the book to Paul.

haitian

Haitian and French thus differ on two counts. While some Haitian verbs
participate in the RecipientTheme construction (np np) (see (78) ), the corresponding French verbs do not (see (81) ). And while some French verbs participate in the ThemeLocative construction (see (82) ), the corresponding Haitian
verbs do not (see (84) ). While French has obviously contributed the phonetic
matrix of the Haitian verbs involved in the RecipientTheme construction (e.g.
Haitian montre < French montrer to show, Haitian bay < French bailler 21 to
give, etc.), it has equally obviously not contributed their semantic or syntactic
properties. Where do these properties come from? Again, an examination of the
properties of the corresponding verbs in the substratum languages will provide
an answer to this question.
Koopman (1986: 235) points out that, with the exception of Mande languages,
West African languages do manifest RecipientTheme constructions. Lefebvre
(1992d, 1993c, 1994c) documents the fact that Fongbe has a class of Recipient
Theme verbs, as illustrated in (86). The Fongbe examples in (86) are similar to
the Haitian ones in (78).
(86)

Kwk n / xxlx
sb xw
v.
Koku give / show Asiba house det
Koku gave/showed Asiba the house.

fongbe
(=(1a, b) in Lefebvre 1994c)

On the basis of several semantic tests, Lefebvre (1993c, 1994c) argues that in
these sentences the Goal is interpreted as a Recipient. There is also evidence
that the Recipient is the affected argument of the construction. As in Haitian

292

The syntactic properties of verbs

(see (79) ), the event determiner can only occur in the context of an internal
argument if it is a definite affected argument. In the double-object construction,
it is only permitted in the context of a definite Recipient (see (87a) ). It is not
permitted in the context of a definite Theme (see (87b) ).
(87)

a.

Sn 3 n
vj
v
svn 3 v.
fongbe
man a
give child det crab a Det
A man gave the child a crab, as we knew.
(=(38b) in Lefebvre 1994c)

b. *Sn 3 n
vj
3 svn v
v
fongbe
man a
give child a
crab the Det
[Lit.: A man gave a child the crab, as we knew.]
(=(37b) in Lefebvre 1994c)

The contrast in grammaticality between (87a and b) shows that, in the Fongbe
RecipientTheme construction, the affected object must be the Recipient. The
Fongbe facts illustrated in (87) are similar to the Haitian facts in (79). Finally,
Fongbe presents the asymmetries which are characteristic of the RecipentTheme
construction in English (see Lefebvre 1994c). An example is given below on the
basis of data involving binding of a pronoun by a quantifier. In (88a), the quantifier is part of the Recipient phrase and the sentence is grammatical. In (88b), the
quantifier is part of the Theme phrase and the sentence is not grammatical.
(88)

a.

N xxlx / n
my
3vkp3vkpi fvt
I show / give man every
picture
I showed/gave every man his picture.

twni .
his

fongbe

(=(42a) in Lefebvre 1994c)

3vkp3vkpi
fongbe
b. *N xxlx / n
fvttv twni fvt
I show / give owner its picture every
*I showed/gave its owner every picture
(=(43b) in Lefebvre 1994c)

The data show that, as in Haitian (see (80) ), in the Fongbe RecipientTheme
construction, the Recipient must c-command the Theme. In Lefebvre (1993c,
1994c), it is further argued that the other tests used by Barss and Lasnik (1986)
to show that the Recipient must c-command the Theme in the English Recipient
Theme construction apply in Fongbe in exactly the same way.22 These facts
argue in favour of the claim that there is a class of RecipientTheme verbs in
Fongbe.
Furthermore, Lefebvre (1994c) provides extensive evidence that Fongbe verbs
which participate in the double-object construction (np np) do not participate in
a ThemeLocative construction. First, although the sentence in (89) is grammatical, it is not a ThemeLocative construction, for the prepositional phrase
can only be interpreted as a Benefactive. Moreover, Lefebvre (1994b) clearly
shows that the prepositional phrase introduced by n in (89) is a syntactic
adjunct rather than an indirect object of the verb.

9.12 Double-object verbs

(89)

Kwk ny kwx n sb.

Koku loaned money for Asiba.

293

fongbe

Second, the ThemeLocative alternation in Fongbe is rendered in a serial verb

construction, as shown in (90).
(90)

Kwk sv
svn v
n
sb.
Koku take crab det give Asiba
Koku gave the crab to Asiba.

fongbe
(=(71a) in Lefebvre 1994c)

As argued in Lefebvre (1994c), serial verb constructions involving verbs of

transfer in Fongbe, like (90), have the same semantic and syntactic properties
as the ThemeLocative construction with the form np pp in English.
The Haitian and Fongbe facts presented so far are similar in two respects.
First, in both languages, there is a subset of verbs that are syntactically realised
in a RecipientTheme construction. Second, neither language has verbs which
are syntactically realised in a ThemeLocative construction of the type found in
French. In both Haitian (see (84) ) and Fongbe (see (89) ), the pp in the surface
sequence v np pp is a syntactic adjunct, and in both languages ThemeLocative
combinations of arguments are rendered by a serial verb construction (see (85)
and (90) ). It thus appears that substratum language verbs are the source of the
semantic and syntactic properties of the Haitian verbs which participate in the
RecipientTheme construction.
The source of the properties of the Haitian verbs participating in this construction is in accordance with the relexification hypothesis. While the phonological representation of the Haitian verb is derived from the phonetic matrix
of a French verb, its semantic and syntactic properties come from those of the
corresponding verb in the substratum language. This scenario is not complete,
however, since there are two types of differences between the Haitian and the
Fongbe data which need to be accounted for. These differences involve word
order and variations in the range of verbs which participate in the construction.
These differences will be argued to be linked to the Case-assignment requirement and the Case-assigning properties of verbs.

9.12.3

Word order

The word order RecipientTheme shown for the Fongbe data in (86) is only
one of two possible surface orders for the construction. In Fongbe, but not in
Haitian, ThemeRecipient is also a possible surface order, as is shown in (91).
(91)

Kwk n / xxlx
svn sb.
Koku give / show crab Asiba
Koku gave/showed Asiba some crab.

fongbe
(=(2) in Lefebvre 1994c)

294

The syntactic properties of verbs

A natural hypothesis would be to see the data in (91) as an instantiation of the

ThemeLocative construction. Lefebvre (1994c) argues, however, that this is
not the case. First, semantic tests show that, with the order in (91), the Goal is
still interpreted as a Recipient. Second, with this order, the Recipient is still the
affected argument of the construction. This claim is supported by the fact that,
with this word order, it is the Recipient and not the Theme that licenses the
event determiner, as shown in (92).
(92)

Sn 3 n
svn 3 vj
v
v.
man a give crab a
child det Det
A man gave the child a crab (as expected).

fongbe
(=(38a) in Lefebvre 1994c)

The data in (92), as well as those in (87), thus show that the Recipient is the
affected argument of the construction with both word orders. The third point
is that the order ThemeRecipient manifests the same asymmetries as the
order RecipientTheme. For example, with the ThemeRecipient order, when
the quantifier is part of the Recipient phrase, the sentence is grammatical (see
(93a) ). When the quantifier is part of the Theme phrase, the sentence is not
(see (93b) ).
(93)

a.

N xxlx / n
fvt
tvni my
3vkp3vkpi .
fongbe
I show / give picture his man every
I showed/gave every man his picture.
(=(42b) in Lefebvre 1994c)

fongbe
b. *N xxlx / n
fvt
3vkp3vkpi fvttv twni
I show / give picture every
owner its
*I showed/gave its owner every picture.
(=(43a) in Lefebvre 1994c)

The comparison of (93a) with (87a) and (93b) with (87b) shows that, regardless
of the word order, the Recipient must c-command the Theme in the Recipient
Theme construction. As shown in Lefebvre (1993c, 1994c), the same situation
obtains for the other tests in Barss and Lasnik (1986). On the basis of these three
types of arguments, Lefebvre (1994c) concludes that the word order alternation
observed in the Fongbe double-object construction is purely syntactic.
The word order alternation found in Fongbe does not exist in Haitian, which
only shows the order RecipientTheme. How can this difference between the
two languages be accounted for? In order to address this question, we need a
syntactic account of the RecipientTheme construction. A detailed account of
the Fongbe double-object construction may be found in Lefebvre (1994c). Here
I shall only summarise the arguments which are pertinent to the discussion of
the word order facts. First, Lefebvre (1994c) proposes that the Fongbe doubleobject construction consists of a small clause whose inner structure corresponds
to that of dp in the language. The motivation for this view is that, as we will

9.12 Double-object verbs

295

see in chapter 12, the RecipientTheme construction is available in a given

grammar if structural Genitive Case is also available in the nominal structure
of that grammar (see Johnson 1991). The analysis of the double-object construction as having the inner structure of a dp constitutes a natural way of formalising this correlation. As we saw in chapter 4, Genitive Case is available in the
Fongbe nominal structure (see (69) in chapter 4). The inner structure of the
small clause for double-object constructions proposed in Lefebvre (1994c) is
thus as in (94).
xp

(94)

Spec
fp
f
np
n
Theme

x
(Det)
Recipienti

f
ti

n
e
(=(97) in Lefebvre 1994c)

The Recipient is base-generated in the specifier position of np, where it is

compositionally assigned the thematic role Recipient by n. Like the possessor
in nominal structures, the Recipient moves to Spec of fp. Recall from chapter 4
that, in Fongbe, the [+definite] determiner can surface in nominal structures
with a possessor marked for structural Genitive Case (unlike in English). And
recall from chapter 8 that the event determiner is homophonous with the
nominal determiner. The event determiner is involved in the double-object construction (see (87) and (92) ). In this case, the presence of the event determiner
correlates with the property [+definite] of the Recipient of the construction.
Lefebvre (1994c) proposes that, in the context of the double-object construction, the event determiner is the head of the small clause (that is, the head of xp).
This proposal is independently motivated by the fact that, as we saw in chapter
8, the determiner is a multifunctional category that can head different syntactic
projections.
The structure in (94) will generate the surface word order ThemeRecipient
(event determiner) as in (92). Movement of the Recipient to Spec of xp, as in
(95), produces the derived word order RecipientTheme (event determiner) in
(87).

296
(95)

The syntactic properties of verbs

Derived word order
v
v

xp
x

dp i (Recipient)
fp
f
np
n
dp
(Theme)

x
ti

(Det)

f
ti

n
e
(=(99) in Lefebvre 1994c)

Movement of the Recipient to the specifier position of the projection headed by

the event determiner is motivated by the relationship between the Recipient and
the event determiner in this construction. In the structure in (95), the Recipient
of the double-object construction is in a SpecHead relationship with the event
determiner, an appropriate configuration for agreement. When the event determiner (which bears the feature [+definite] ) occurs as the head of xp, the Recipient has to be [+definite] (see (87) ). When the head of xp is not lexically filled,
it is unspecified for the feature [+/definite] and therefore the Recipient can be
[+/definite]. The event determiner may also appear with the word order Theme
Recipient, as in (92). In this case, the Recipient moves to Spec of xp at lf, rather
than at S-structure, and lf is the level of representation at which SpecHead
agreement (a feature-checking rule) takes place. This approach is independently
motivated in Chomsky (1989).
On the assumption that structural Case assignment reduces to SpecHead
agreement (see Chomsky 1986), it is assumed that Case assignment proceeds in
the same way as the agreement rule discussed above. Thus, when the Recipient
has remained in its basic position, as in (94), it moves to Spec of xp at lf. Casechecking then takes place at this level of representation (see Chomsky 1989).
The Haitian RecipientTheme construction can be accounted for along the
same lines as the analysis proposed for Fongbe. As in Fongbe, Genitive Case is
available in Haitian (see chapter 4). So suppose the small clause of the Haitian
double-object construction also has the structure in (94) and the Recipient moves
to Spec of xp, as in (95). This would account for the word order Recipient
Theme (see (78) ). It would also account for the relationship between the Recipient and the event determiner, which is the same as in Fongbe (see (79) ). On this
account, the difference between Haitian and Fongbe lies in the fact that, whereas
movement of the Recipient to Spec of xp is obligatory in Haitian (since there is

9.12 Double-object verbs

297

only one possible surface word order), it is optional in Fongbe (since there are
two surface word orders).
Given this analysis, what is the linguistic account of the difference between
the two grammars? This difference can be explained in terms of the parametric
options which define the conditions under which structural Case is assigned in
particular grammars. Adjacency between the Case assigner (the verb) and the
Case assignee (in this case the Recipient) has been proposed as a condition for
structural Case assignment (see e.g. Chomsky 1981; Stowell 1981). Adjacency
has been proposed to be relevant either at S-structure (see Chomsky 1981: 94)
or at lf (see Chomsky 1989). Suppose, then, that the level at which adjacency
applies (S-structure or lf) is an option of universal grammar. In the analysis
presented above, a grammar of the type of Fongbe would allow the Case assigner
and the Case assignee to be adjacent either at lf or at S-structure. By contrast,
a grammar of the type of Haitian would require adjacency at S-structure, thus
forcing the Recipient to move to Spec of xp in the syntax. In this view, the
speakers of the creole would have settled on the parametric option that we see
today, and thus the word order ThemeRecipient is not available in Haitian.
9.12.4

The range of verbs participating in the double-object construction

The range of verbs that can participate in the double-object construction varies
among languages. For example, English has a wide range of such verbs. Gruber
(1992) analyses the verbs participating in the English double-object construction as belonging to four semantic classes: possessional verbs whose Goal is
animate (e.g. to give), animate control verbs (e.g. to pass), verbs of an informational dimension with an animate Goal (e.g. to tell) and positional verbs with an
animate Goal (e.g. to throw). As is documented in detail in Lefebvre (1994c), in
Fongbe, only three or four verbs (depending on the speaker) participate in the
RecipientTheme construction. These verbs are n to give, which belongs in
the first semantic class identified by Gruber, xXlX to show and kplVn to teach,
from the third semantic class, and for some but not all speakers, s to pay,
which would belong in Grubers second semantic class. By contrast, Haitian
creole has a much wider range of verbs which can participate in this construction
(at least seventeen according to Valdman et al.s 1981 dictionary). A list of
these verbs is provided in (96).
(96)

List of double-object verbs (from Valdman et al. 1981)

bay
ofri
poze
preskri
prete
pwmt
rakonte
rann
sede

to
to
to
to
to
to
to
to
to

give/to tell
offer
ask (a question)
prescribe
lend/to borrow
promise
tell
give back
yield/to give

sede
svi
swete
tire
trete
vann
veye
voye

to
to
to
to
to
to
to
to

lend
serve
wish
deliver (a blow)
treat (like)
sell
sit up with
send

298

The syntactic properties of verbs

My own inventory of double-object verbs in Haitian is even bigger than that

of Valdman et al. (1981). For example, it also includes verbs such as lanse
to throw, mande to ask, ekri to write, di to say, etc.23 A sample of doubleobject verbs drawn from the two inventories mentioned above and organised
according to Grubers semantic classification is given in (97). The French form
which provided the phonological representation of each Haitian verb is also
listed. The glosses in (97) correspond to the meaning of the Haitian verb and
only partially to that of the French verb.
(97)

haitian
bay
prete
rann
sede
montre

french
<bailler
<prter
<rendre
<cder
<montrer

to
to
to
to
to

give, to tell
lend, to borrow
give back
yield, to give, to lend
show

ofri
svi
vann
paye

<offrir
<servir
<vendre
<payer

to
to
to
to

offer
serve
sell
pay

Verbs of an informational di
dimension with an
poze
animate Goal (e.g. tell) prski
pwmt
rakonte
swete
mande
ekri
ansnye
aprann

<dire
<poser
<prescrire
<promettre
<raconter
<souhaiter
<demander
<crire
<enseigner
<apprendre

to
to
to
to
to
to
to
to
to
to

say
ask
prescribe
promise
tell
wish
ask
write
teach
teach

Positional verbs
with an animate
Goal (e.g. throw)

<tirer
<lancer
<envoyer

to deliver (a blow)
to throw
to send

Possessional verbs
with an animate
Goal (e.g. give)

Animate control
verbs (e.g. pass)

tire
lanse
voye

As can be seen in (97), Haitian has double-object verbs of all four semantic
classes identified by Gruber. This contrasts with Fongbe, which only has a
few double-object verbs as specified above. Haitian and Fongbe thus vary with
respect to the range of verbs which can participate in the double-object construction. Why should this be so? An examination of the Case properties of
verbs will provide an answer to this question.
On the basis of some Fongbe data subject to variation among speakers,
Lefebvre (1994c) argues that a verb must have the property of assigning structural
Case to the Recipient in order to be able to occur in this construction. This claim
is supported by the following data. Consider the verb s to pay in (98). For
half the informants, s is a double-object verb, and for the other half, it is not.
(98)

ok/*

Kwk s
xwkwx v
sb / sb
Koku pay rent
det Asiba / Asiba
Koku paid Asiba the rent.

xwkwx v.
rent
det

fongbe

(=(108) in Lefebvre 1994c)

9.12 Double-object verbs

299

For speakers in the second group, the Goal is rendered in a pp as shown in (99).
(99)

Kwk s
xwkwx v
n sb.
fongbe
Koku pay rent
det to Asiba
Koku paid the rent to Asiba (the owner of the house).
(=(106) in Lefebvre 1994c)

Suppose that, for speakers in the first group, the verb s pay is specified as
assigning structural Case to the argument bearing the thematic role Recipient
and, for the second group of speakers, it is not. This would explain why the
first group has s as a double-object verb and the second group does not. Data
involving the serial verb construction support this claim.
As is shown in Lefebvre (1994c, section 5), verbs which occur in the doubleobject construction may also participate in the serial verb construction. This fact
is illustrated in (100) for the verb n to give.
(100)

Kwk sv
svn v
n
sb.
Koku take crab det give Asiba
Koku gave the crab to Asiba.

fongbe
(=(57) in Lefebvre 1994c)

According to the analysis in Lefebvre (1991b), serial verb constructions of the

type in (100) have the structure in (101), which draws on that proposed in
Larson (1988) for English np pp complements.
vp

(101)

SpecV

vp

SpecV

Kwk

sv

svn v

Complement

sb

Assuming the analysis in Lefebvre (1991b), the verb n in (101) assigns structural Case to the Goal np. Now, consider Fongbe verbs which can participate in
the serial verb construction but not the double-object construction, as in (102).
(102)

a. Kwk sv
kwx
ny n sb.
Koku take money loan to Asiba
Koku loaned money to Asiba.

fongbe
(=(110a) in Lefebvre 1994c)

b. Kwk sv
xwnx 3w
Koku take story
tell
Koku told a story to Asiba.

n sb.
to Asiba

fongbe
(=(110b) in Lefebvre 1994c)

300

The syntactic properties of verbs

In contrast to (101), in (102), the Goal is realised as a pp rather than an np.

According to Lefebvres (1994c, section 5) analysis, the contrast between (101)
and (102) follows from the verbs Case feature specification. While, in (101),
the second verb in the serial verb construction assigns structural Case to its
complement, in (102) this verb has no structural Case to assign to the Goal and
hence the Goal must be realised as a pp, in an adjunct position rather than in the
direct object position of the verb. The contrast between the data in (101) and in
(102) thus supports the claim that verbs which participate in the double-object
construction must also be able to assign structural Case to the argument bearing
the thematic role Recipient. With this tool in hand, we have a means of explaining the variation among Fongbe speakers with respect to the verb s pay
(see (98) ). Our proposal predicts two possible realisations for the Goal of s
when the latter is the second verb of a serial verb construction: the Goal can be
realised as an np (as in (101) ) or as a pp (as in (102) ). This prediction is borne
out, as shown in (103). Speakers for whom s is a double-object verb accept
(103a) and reject (103b). Speakers for whom s is not a double-object verb
accept (103b) but reject (103a).
(103)

a. ok/* Kwk sv
xwkwx sv-s
Koku take rent
take-pay
Koku paid the rent to Asiba.

sb.
Asiba

fongbe

(=(111a) in Lefebvre 1994c)

b. */ok

Kwk sv
xwkwx sv-s
Koku take rent
take-pay
Koku paid the rent to Asiba.

n sb.
to Asiba

fongbe

(=(111b) in Lefebvre 1994c)

The data discussed above show that, in order to participate in the doubleobject construction, a verb must have the appropriate Case feature specification,
that is, it must be specified for assigning structural Case to the argument that
bears the thematic role Recipient. In this view, the Case feature specifications of
verbs account for variation among Fongbe speakers with respect to the range of
verbs participating in this construction. In Fongbe, few verbs have this Case
feature specification whereas far more do in English. Hence, while the verbs
meaning to bake or to throw in English can assign structural Case to the
Recipient and thus can participate in the double-object construction, their Fongbe
equivalents cannot.
This proposal provides a straightforward account of the difference between
Haitian and Fongbe with respect to the number of verbs that participate in the
RecipientTheme construction. As is the case in Fongbe, many Haitian verbs participate both in the RecipientTheme construction and in the serial verb construction. This is illustrated in (104), which parallels the Fongbe data in (100).
(104)

Jan
pran krab la
bay
John take crab det give
John gave the crab to Mary.

Mari.
Mary

haitian

9.13 Conclusion

301

In (104), the verb bay give assigns structural Case to the Goal np. Now let us
consider those Haitian verbs which participate in the RecipientTheme construction while their Fongbe counterparts do not. These verbs have different Case
properties from the Fongbe equivalents. This can be seen when they are used in
the serial verb construction, as exemplified in (105) for the verb prete to loan.
(105)

Jan
pran kb
la
prete Mari.
John take money det loan Mary
John loaned the money to Mary.

haitian

As can be seen in (105), the verb prete to loan assigns structural Case to the
Goal of the serial verb construction. This contrasts with the corresponding Fongbe
verb ny to loan, which does not have this property, as we saw in (102a).
Similarly, the Haitian verb paye to pay can also be shown to assign structural
case to the Goal of a serial verb construction, as is shown in (106).24
(106)

Jan
pran kb
la
paye Mari.
John take money det pay Mary
John paid Mary.

haitian

The Case property of this Haitian verb pairs with that of the corresponding
Fongbe verb in the grammar of a subset of speakers (see (103a) ), but differs
from the corresponding verb in the grammar of the other subset of Fongbe
speakers (see (103b) ).
The Haitian data presented above provide additional evidence supporting the
claim in Lefebvre (1994c) that, in order to participate in the RecipientTheme
construction, a verb must have the property of assigning structural Case to the
Recipient. Given that Haitian has a much wider range of RecipientTheme verbs
than Fongbe, one must conclude that there are more verbs in this language
which can assign structural Case to the Recipient than there are in Fongbe.
The logical conclusion is that, even though the possibility of double-object verbs
in Haitian comes from the substratum language, the range of such verbs has
increased in the creole due to a modification in the Case-assigning properties of
individual verbs compared to the corresponding substratum verbs. This change
cannot be attributed to French for, as we saw earlier, French has no verbs that
can assign structural Case to a Goal argument. The above account is in line with
the finding in section 9.11 that the Case-assigning properties of verbs in a creole
language are more independent than other syntactic properties with respect to
the contributing languages. This issue will be further discussed in chapter 13.
9.13

Conclusion

The data reported on in this chapter show that the bulk of the syntactic properties of Haitian verbs follow the properties of the substratum language rather than
those of the superstratum language. Hence, the properties of Haitian body-state,
weather, raising and control verbs contrast with those of French and pair with

302

The syntactic properties of verbs

those of the substratum languages. Furthermore, both Haitian and Fongbe have a
class of double-object verbs in contrast to French, which does not. This situation
follows directly from the relexification hypothesis. The selectional properties of
Haitian reflexive verbs were shown to follow from relexification and dialect levelling. The selectional properties of Haitian verbs that take overt/covert expletives
do not perfectly match those of Fongbe; it is possible that dialect levelling has
played a role in this case as well; but recall that the availability of a null expletive
is a property of both Haitian and Fongbe but not of French. The properties of
the Haitian verb gen were hypothesised to result from an independent development within the creole. The subsets of verbs studied for Case properties show
that Case-assigning properties constitute the syntactic properties that seem to
be the most independent of the creoles source languages. The Case-assigning
properties of double-object verbs were shown to provide further support for
this claim. The lack of inherent object verbs in Haitian was attributed to the
influence of French.

10 Are derivational affixes relexified?

Current linguistic theories maintain that derivational affixes are listed in the
lexicon as individual lexical entries that are minimally specified for categorial
features and selectional and semantic properties (e.g. Lieber 1980, 1992; Di Sciullo
and Williams 1987; Lefebvre and Muysken 1988; Selkirk 1984). Given this
general assumption, and the hypothesis that relexification plays a central role in
creole genesis, lexical entries for derivational affixes are expected to undergo
relexification similarly to simple nouns, verbs and so on. The comparison of
the derivational affixes of Haitian with those of its contributing languages does
indeed show that their properties are quite straightforwardly derivable from the
process of relexification.
Section 10.1 identifies the productive derivational affixes of Haitian. In section 10.2, I compare the Haitian and French affixes. It is shown that, although the
phonological representations of the Haitian affixes are derived from the phonetic
matrices of French affixes, several of their other properties cannot be explained
by reference to French. Section 10.3 compares the derivational affixes of Haitian
with those of Fongbe, showing that the semantics of the Haitian affixes follows
that of the substratum language. Section 10.4 discusses the historical derivation
of the Haitian affixes within the framework of the three processes claimed to
operate in creole genesis: relexification, reanalysis and dialect levelling.
This chapter draws on work by Filipovich (1988) on the morphology of
Haitian, Brousseau (1990, 1993) on the morphology of Fongbe, Brousseau,
Filipovich and Lefebvre (1989) and Brousseau (1994a) on Haitian, French and
Fongbe morphology, and much further work of my own to complete the data,
the specific analyses and the comparative analyses.
10.1

Identifying the derivational affixes of Haitian

On the basis of phonological similarity with the forms that have affixal status
in French, Hall (1953) presents a list of over 70 Haitian derivational affixes.
However, this claim has been challenged by many other authors, who generally
report an inventory of about ten productive affixes in Haitian (see Tinelli 1970;
Valdman 1978; Filipovich 1988; Brousseau, Filipovich and Lefebvre 1989). These
authors all point out the problem of treating the internal structure of Haitian
words on a par with that of French words, since a word which has internal structure in French may not in Haitian. Given that the bulk of the Haitian lexicon is
303

304

Are derivational affixes relexified?

derived from French, some criteria are needed in order to determine whether a
given Haitian word has internal structure and what affixes are productive in, or
native to, this language. Brousseau, Filipovich and Lefebvre (1989) propose five
criteria to identify the native derivational affixes of Haitian creole.
First, a form is a native affix in Haitian creole if it is found affixed to a base
which is foreign to French. For example, the verbalising suffix -e is such an
affix, since it attaches to the noun djb job (borrowed from English) forming
the verb djb-e to work. The attributive affix - is also an affix of this type
since it attaches to the noun wanga fetish (of West African origin) to derive
wanga-t- magician (where -t- is an epenthetic consonant).1
Second, a form is native to Haitian creole if it is used with a base that is
derived from a French word that does not take the equivalent affix. For example,
while the inversive prefix d- in French only attaches to transitive dynamic
verbs, de- in Haitian creole can attach to transitive and intransitive verbs, whether
dynamic or stative. Thus, Haitian has de-pasyante to get impatient derived
from pasyante to be patient whereas French has im-patienter to get impatient.
Similarly, the Haitian nominalising affix -ay attaches to the base kontre to
oppose to form the noun kontr-ay opposition. In French, there is no nominal
derived from the verb contrer. Furthermore, the suffixes -wa and -yen in Haitian,
which derive nouns referring to ones place of origin, are used on bases where
we do not find them in French. For example, while in Haitian we find Soudan-nwa Sudanese and Ougand-yen Ugandan, in French we find Soudan-ais and
Ougand-ais, respectively. Similarly, the attributive suffix - derives langaj-
chatterbox from langaj language. There is no French word derived from
langage language using -eur. Finally, the affix -man can attach to a base that is
not a word in French; for example, it attaches to the adjective alekout (<French
lcoute) attentive to derive the adverb alekout-man attentively, or to avidy
(<French vue dil) visible to derive the adverb avidy-man visibly.
Third, a form is a native affix of Haitian creole if it appears in a different
position from the parallel affix in French. For example, the diminutive affix in
Haitian creole is the prefix ti- (e.g. ti-mounn child), whereas all diminutive
affixes in French are suffixes (e.g. garon-n-et little boy).
Fourth, a form is native to Haitian creole if the semantic and syntactic properties of a word that it derives are different from those of the corresponding word
in French. For example, the Haitian verb bwa-z-e to take to the woods (where
-z- is an epenthetic consonant) is derived from the noun bwa wood(s) by
means of the verbalising suffix -e. The parallel verb in French, derived from bois
wood(s), is bois-er to plant a forest.
Finally, if a morphological process in Haitian has no French counterpart, this
argues that the process is native to Haitian. As will be seen below, the process
of morphological conversion which derives adjectives from verbal bases is one
such example.
On the basis of these criteria, Brousseau, Filipovich and Lefebvre (1989)
identify the following Haitian derivational affixes. First of all, the suffix -

10.1 Identifying the derivational affixes of Haitian

305

derives agentive nouns from dynamic verbs. The output of this concatenation
is a noun with the meaning one who verbs.
(1)

Agentive suffix -: [v-]N

dechouk-
konsey-
rans-
vant-

insurgent
counsellor
joker
braggart

haitian
dechouke to uproot
konseye
to counsel
ranse
to joke
vante
to brag
(=(2) in Brousseau, Filipovich and Lefebvre 1989)

There is another affix - which is an attributive affix. As is pointed out in

Brousseau, Filipovich and Lefebvre (1989), this affix is distinct from the preceding
one not only in meaning but also because it attaches to a nominal rather than a
verbal base. The derived meaning is one who possesses or uses noun.
(2)

Attributive suffix -: [n-]N

odyans- joker
langaj-
chatterbox
tafya-t-
alcoholic
wanga-t- magician

haitian
odyans joke
langaj language
tafya
alcoholic beverage
wanga charm
(=(3) in Brousseau, Filipovich and Lefebvre 1989)

The suffix -e derives verbs from nominal bases, as shown in (3), which
combines data from Brousseau, Filipovich and Lefebvre (1989) and Brousseau
(1994a).
(3)

Verbalising suffix -e: [n-e]V

betiz-e
kle-t-e
madison-n-e
makak-e

to
to
to
to

talk nonsense
lock up
cast a spell
hit with a stick

haitian
betiz
kle
madison
makak

nonsense
key
spell
stick

The prefix de- selects verbs. It derives new verbs which have a meaning
opposite to that of the base form. This is shown in (4), which combines data
from Brousseau, Filipovich and Lefebvre (1989) and Brousseau (1994a).
(4)

Inversive prefix de-: [de-v]V

de-vlope
de-pasyante
de-makone
de-rspkte
de-kale
de-mwle

to
to
to
to
to
to

unwrap
be impatient
separate
insult
shell
remove the marrow

haitian
vlope
pasyante
makone
rspkte
*kale
*mwle

to wrap
to be patient
to bring together
to respect
*to put on a shell
*to put marrow in

Brousseau (1994a) points out that, in some cases, such as de-kale and de-mwle,
the inversive prefix attaches to a base that is unattested as an actual verb. *Kale
and *mwle are not attested verbs in the lexicon of our Haitian informants.2 The
only related forms found in the lexicon of our informants are the nouns kal
shell and mwl marrow. Following Scalise (1984) and the references therein,

306

Are derivational affixes relexified?

Brousseau (1994a) assumes a two-step derivation for these parasynthetic

words. First, the verb is derived from a noun by affixation of the verbalising
suffix -e identified above, yielding *kale and *mwle, respectively. Second, deis prefixed to these possible (though nonexistent) verbal bases. This derivation
yields a privative meaning: to remove or deprive of the entity denoted by the
nominal base.
The diminutive affix is the prefix ti-. The meaning of nouns derived by this
affix is either a small noun or, when the base refers to an animate entity, a
young noun. This affix is illustrated in (5), which combines data from Brousseau,
Filipovich and Lefebvre (1989) and Brousseau (1994a).
(5)

Diminutive prefix ti-: [ti-n]N

ti-chat
kitten
ti-mounn child
ti-dlo
pond, brook
ti-wch
pebble

haitian
chat
cat
mounn person
dlo
water
wch
stone

The suffix -ay derives nouns from two-place verbal predicates. The nouns so
derived refer either to the action or to the result of the action denoted by the
verbs.
(6)

Nominalising suffix -ay: [v-ay]N

kontr-ay
bwt-ay
dechouk-ay
kapon-ay
fri-t-ay
krabin-ay

opposition
moving
overthrow
intimidation
fried things
crumb

kontre
bwte
dechouke
kapone
fri
krabine
(=(6) in Brousseau,

haitian
to oppose
to move
to uproot
to intimidate
to fry
to crumble
Filipovich and Lefebvre 1989)

The specific properties of -ay are not uniform across Haitian speakers. For
example, for the speakers who served as informants for Brousseau, Filipovich
and Lefebvre (1989), the affix -ay has the property of absorbing the Agent of the
verbal base it attaches to. The verb dechouke to uproot, for example, has two
thematic roles: Agent and Theme/Patient. The deverbal noun dechouk-ay overthrow allows only the Theme of the verbal base to be realised in the syntax,
since -ay has absorbed the Agent. Hence, for these speakers, while dechouk-ay
presidan an the overthrow of the president, where the Theme is expressed
in the syntax, is grammatical, *dechoukay pp ayisyen an the overthrow by the
Haitian population, where the Agent is syntactically realised, is not. For other
speakers, -ay does not absorb the Agent (see Brousseau 1994a, and my own field
notes). For these speakers, a noun phrase headed by a deverbal noun containing -ay can either express the Theme (as above) or the Agent as in dechoukay
pp ayisyen an the overthrow by the Haitian population. None of the speakers
consulted, however, accept a nominal phrase where both arguments are realised
in the syntax. Hence *dechoukay presidan pp ayisyen an with the reading the

10.1 Identifying the derivational affixes of Haitian

307

overthrow of the president by the Haitian population is ruled out by both groups
of speakers.3
In Haitian creole, there is yet another way of deriving nouns from verbs,
namely through the process of morphological conversion, which has been extensively discussed in the literature on Haitian (see e.g. Sylvain 1936; Hall 1966;
Tinelli 1970; Fauchois 1983; Filipovich 1988). Lieber (1992) has argued that
cases of morphological conversion fall into two distinct classes. The first involves
new lexical entries which are created by the process of relisting rather than by
a regular derivational process. The relisted entry shares the phonological form,
and some semantics, of an already existing entry. It differs from the latter in
its categorial features. Furthermore, each relisted item needs to be specified for
semantic and morphosyntactic properties because these are not always predictable. The Haitian nouns antre (<Fr. entre) entrance, homophonous with the
verb antre to enter, and manti lie, homophonous with the verb manti to lie,
are among the few nouns which are derived by relisting in this language. The
second class involves the affixation of a morpheme that has all the properties of
an affix, except that it lacks a phonological representation. The null affix determines the syntactic category of the derived words as well as their conceptual
structure. Words derived with a null affix have uniform and predictable properties. In Haitian, the output of this type of conversion refers to either the action
or the result of the action denoted by the verbal base. Examples are provided
in (7) (from Brousseau, Filipovich and Lefebvre 1989).
(7)

Nominalising conversion: [v-]N

ale
sti
antre
vini
rive
tonbe
desan
monte
tunen
pati
pwche
gunmen
kuri
mache
dmi
manti
bwte

action
action
action
action
action
action
action
action
action
action
action
action
action
action
action
action
action

or
or
or
or
or
or
or
or
or
or
or
or
or
or
or
or
or

result
result
result
result
result
result
result
result
result
result
result
result
result
result
result
result
result

of
of
of
of
of
of
of
of
of
of
of
of
of
of
of
of
of

going
going out
entering
coming
arriving
falling
descending
going up
returning
leaving
approaching
fighting
running
walking
sleeping
lying
moving

haitian
ale
sti
antre
vini
rive
tonbe
desan
monte
tunen
pati
pwche
gunmen
kuri
mache
dmi
manti
bwte

to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to
to

go
exit
enter
come
arrive
fall
descend
go up
return
leave
approach
fight
run
walk
sleep
lie
move

The cases of conversion in (7) present all the properties of being derived by an
affix. Like overt affixes, the null affix in (7) changes the category of the base it
attaches to. This affix selects verbal bases which have the common property of
having only one argument. As is shown in Brousseau, Filipovich and Lefebvre
(1989), morphological conversion applies only to unaccusative verbs (e.g. ale

308

Are derivational affixes relexified?

to go), unergative verbs (e.g. krache to spit) and intransitive forms of verbs
which show a transitive/intransitive alternation (e.g. bwte to move). Finally,
the derived noun has the predicate argument structure of the verbal base. In all
cases, the single argument of the verb may be expressed in the syntax, as is
illustrated in (8).
(8)

a. ale
Jan
departure John
Johns leaving

an
b. manti Jan
det
lie
John
Johns lie

c. bwte
Mari a
moving Mary det
Marys moving

an
det

haitian

d. gunmen Jan
(ak Wb)
a haitian
fight
John (with Robert) det
Johns fight (with Robert)

Thus, the cases of morphological conversion in (7) have the same properties as
if they had been derived by an overt affix.
The process of morphological conversion involving a null affix appears to
be rather productive in Haitian. Vdrine (1992) lists sixty verbs which can also
be used as nouns. As Brousseau, Filipovich and Lefebvre (1989) observed, this
process is in complementary distribution with the derivational process involving
the suffix -ay (see (6) ). While -ay selects only verbs which have two arguments,
the null affix involved in morphological conversion selects only verbs with one
argument. Interestingly, some verbs can appear with either one or two arguments. Bwte to move is one such verb, as shown in (9).
(9)

a. Mari bwte.
Mary move
Mary moved.

b. Mari bwte mb
yo.
Mary move furniture det
Mary moved the furniture.

haitian

Brousseau, Filipovich and Lefebvre (1989: 15) show that two nouns can be
derived from this verb. The noun bwte moving, derived by morphological
conversion from the intransitive reading of the verb (see (9a) ), allows for the
expression in the syntax of the sole argument of the base verb (the Agent),
yielding the noun phrase bwte Mari a Marys moving. The noun bwt-ay
moving, derived from the transitive version of the verb (see (9b) ) by means of
the nominalising affix -ay, allows only for the expression of the Theme in the
syntax, yielding the noun phrase bwt-ay mb yo the moving of the furniture.
This is because, as mentioned above, for these speakers -ay absorbs the Agent
thematic role. The ungrammaticality for these speakers of *bwt-ay Mari a,
where the Agent is expressed, and of *bwte mb yo, where the Theme is
expressed, follows from the above analysis.
Brousseau (1994a) argues for another case of morphological conversion in
Haitian creole involving an affix which has no phonological representation. This
phonologically null affix derives adjectival and participial forms from a verbal
base.4 Vdrine (1992) lists 254 verbs which can be used as attributives. A few
examples are provided in (10).

10.1 Identifying the derivational affixes of Haitian

(10)

Adjectival conversion: [v-]A/ PA

abandonnen abandoned
boule
burning hot
chire
torn/ripped
boure
stuffed
kwit
cooked

309

haitian
abandonnen to abandon
boule
to burn
chire
to tear/rip
boure
to stuff
kwit
to cook
(from Vdrine 1992: 1678)

These derived forms are found in verbal and adjectival passive clauses, as illustrated in (11a), or in noun phrases, as in (11b).
(11)

a. Vyann nan kwit.

meat
det cooked
The meat is cooked.
The meat has been cooked.

b. vyann kwit
la
meat cooked det
the cooked meat

haitian

(=(9) in Brousseau 1994a)

This derivational process applies to all verbs involving a change of state (see
e.g. (10) ). It does not apply to locative verbs such as ale go, pouse push, tire
pull, nor to verbs of movement such as tonbe fall, sote jump, naje swim.
Following current theories (see e.g. Levin and Rappaport 1986), I will assume
that the categorial change involved in the conversion of verbs into adjectives/
participles is accompanied by modifications in the argument structure of the
verbal base: (a) the external argument of the verb is not assigned a thematic
role and thus it cannot be projected in the syntax; (b) the deverbal adjective is
deprived of the Case-assigning feature of its verbal base such that the internal
argument of the verbal base must become the external argument of the derived
adjective/participle. For example, the verb kwit to cook takes an external and
an internal argument. In its adjectival/participial form, as in (11a), the external
argument has been absorbed and the internal argument surfaces as the subject.
Brousseau (1994a) argues that cases of conversion of the type in (10) have
properties that are consistent with their being derived by an affix. The null affix
modifies the category of the verbal base it attaches to. In all cases, the predicateargument structure of the verbal base is modified in the same way in the derived
adjectival form, as stated above. Thus, according to Liebers (1992) criteria, the
null affix in (10) is motivated on theoretical grounds. The issue of whether the
two null affixes discussed above (see (7) and (10) ) should be analysed as a
single form will be addressed in section 10.4. For the sake of clarity, I will treat
them as separate forms for now.
Brousseau (p.c.) further identifies the suffix -man, which derives adverbs
from adjectives, as shown in (12).
(12)

Adverbial suffix -man: [a-man]Adv

avidy-man visibly
alekout-man attentively
bosal-man
wildly
alz-man
comfortably
bz-man
elegantly

avidy
alekout
bosal
alz
bz

haitian
visible
attentive
wild
comfortable
elegant
(A.-M. Brousseau p.c.)

310

Are derivational affixes relexified?

She also notes the suffixes -wa and -yen which derive nouns from nouns. The
input noun is the name of a place and the output noun refers to a person who
comes from that place.
(13)

a. Place of origin/residence suffix -wa: [n-wa]N

Kap-wa
Petyonvil-wa
Senmak-wa
Soudan-n-wa
Tchad-wa

from
from
from
from
from

the Cape
Ptionville
Saint-Marc
Sudan
Chad

Kap
Petyonvil
Senmak
Soudan
Tchad

b. Place of origin/residence suffix -yen: [n-yen]N

Jakml-yen
Tib-yen
Gan-yen
Ougand-yen
Pakistan-n-yen
Ka-yen

from
from
from
from
from
from

Jacmel
Tibet
Ghana
Uganda
Pakistan
Les Cailles

Jakml
Tib
Gana
Ouganda
Pakistan
Okay

haitian
Cape
Ptionville
Saint-Marc
Sudan
Chad
(A.-M. Brousseau p.c.)
haitian
Jacmel
Tibet
Ghana
Uganda
Pakistan
Les Cailles
(A.-M. Brousseau p.c.)

The principles governing the choice between the two affixes -wa and -yen
appear to be as obscure as those governing the choice between the French affixes
which perform the same function. For example, French has Montral-ais from
Montreal and not *Montral-ien but Canad-ien from Canada and not *Canadais. Both *Montral-ien and *Canad-ais would be well-formed words in French,
but they are nonexistent. Native speakers of French know, however, which affix
to use with a given base. Similarly, native speakers of Haitian know exactly
which of the two affixes to use when presented with a large list of place names
(Brousseau p.c.).
The affixes discussed above exhaust the list of Haitian affixes identified on
the basis of the five criteria discussed in the beginning of this section. There is
one additional affix, however, which does not conform to these criteria, but
which must be considered as part of the inventory of Haitian affixes on the
basis of its productivity. This affix is -ym, and it derives ordinal numbers from
cardinal numbers.
(14)

Ordinal suffix -ym

san-ym
mil-ym
kat-ym
twa-ym

hundredth
thousandth
fourth
third

haitian
san hundred
mil thousand
kat four
twa three
(A.-M. Brousseau p.c.)

It is unlikely that the creators of Haitian learned the ordinal numbers of French
one by one. If they had, we would expect the Haitian forms to be phonologically
closer to the corresponding French phonetic strings. For example, the Haitian

10.1 Identifying the derivational affixes of Haitian

311

phonemic representation of the lexical entry meaning third is /twa-y=m/. If this

form had been derived from the French phonetic string [trwazy=m], we would
expect it to be /twazy=m/. The latter form is not attested in Haitian, however.
The morphophonemics of the Haitian lexical entries in (14) thus argues for the
claim that the creators of Haitian did not learn the French ordinals individually.
Although -ym does not meet any of the five basic criteria used to identify affixes
that are native to Haitian, it should still be analysed as a productive affix. This
terminates our discussion of the productive affixes of Haitian.
One might believe that most of the Haitian derived words would correspond
to French words derived by means of an equivalent affix. This does not appear
to be the case, however, as is revealed by a close examination of the lexical
entries in Haitian dictionaries of different types (see e.g. Valdman et al. 1981;
Freeman 1988; Vdrine 1992). For example, a careful study of Valdman et al.s
(1981) dictionary provides quite a large number of Haitian derived words which
do not correspond to French derived words, as well as of Haitian derived words
which have a different meaning from their French counterparts. Some examples
are listed in Appendix 3, showing that the affixes identified as native to Haitian
are indeed genuine productive affixes in this language.
Given the size of the inventory of affixes identified as native to Haitian, one
might also wonder about Haitian words which contain a sequence that corresponds to a productive affix in French but was not identified as a productive affix
in Haitian. As is pointed out in Brousseau, Filipovich and Lefebvre (1989),
many productive affixes of French are not productive in Haitian even though
they may be part of Haitian words. A case in point is the sequence -yon (<from
French -ion), which is part of Haitian words like preparasyon preparation. In
Haitian, the sequence -yon meets none of the criteria discussed above. Most of
all, unlike the affixes identified as productive in Haitian, -yon is only found in
Haitian words that correspond exactly to French words (see Brousseau, Filipovich
and Lefebvre 1989). An examination of the list of Haitian nouns ending in -yon
in Freemans (1988) inverse dictionary supports this claim. Furthermore, the
inventory of nouns in Valdman et al.s (1981) dictionary shows that Haitian
words ending in -yon represent only a small fraction of words ending in -ion
in French. Thus, while adorasyon the action of adoring and arstasyon the
action of arresting (somebody) are listed as lexical entries in Valdmans dictionary, words such as *admirasyon admiration (=admirat-ion in French) and,
*adopsyon adoption (=adopt-ion in French), etc., are not. Haitian words which
contain the sequence -yon are thus best analysed as simplexes, that is, as words
without internal structure (see Brousseau, Filipovich and Lefebvre 1989: 15).
These words, and many others which are derived words in French, have entered
Haitian as simplexes and are therefore listed individually in the Haitian lexicon.
The inventory of productive derivational affixes in Haitian thus consists of
the affixes listed in (15). Depending on whether the phonologically null affixes
in (15g and h) constitute one or two lexical entries (see section 10.4), the inventory of productive affixes in Haitian reduces to ten or eleven.

312

Are derivational affixes relexified?

(15)

The inventory of Haitian affixes

a. Agentive suffix
base
output

haitian
-
v
n

b. Attributive suffix
base
output

-
n
n

c. Verbalising suffix
base
output

-e
n
v

d. Inversive prefix
base
output

dev
v

e. Diminutive prefix
base
output

tin
n

f. Nominalising suffix
base
output

-ay
v
n

g. Nominal conversion
base
output

v
n

h. Adjectival / participial conversion

base
output

v
a

i. Adverbial suffix
base
output

-man
a
Adv

j. Place of origin / residence suffixes

base
output

-wa / -yen
n
n

k. Ordinal suffix
base
output

-ym
q
a

Why does Haitian have the particular affixes that it has? Why is the Haitian
inventory of derivational affixes the size it is? The following comparison with
French and Fongbe will shed light on these questions.
10.2

The derivational affixes of Haitian compared with those

of French

The Haitian derivational affixes all derive their phonological representations

from French forms. The Agentive suffix - is derived from the phonetic matrix
of the French agentive suffix -eur as in travaill-eur worker. The attributive

10.2 Derivational affixes of Haitian compared with French

313

suffix - is derived from the French attributive suffix -eur as in parl-eur somebody who talks a lot. The verbalising suffix -e derives its phonological representation from the French suffix -er (pronounced [e]) as in ft-er to celebrate.
The Haitian inversive prefix de- is phonologically identical to the French inversive prefix d- in verbs such as d-placer to displace. The Haitian prefix ti- is
derived from the French prenominal adjective petit (pronounced [p1ti]/[pti]/[ti]).
The nominalising suffix -ay derives its phonological representation from the
French suffix -age occurring in words such as chauff-age heating. The adverbial suffix -man is derived from the French suffix -ment (pronounced [m]):
admirable-ment admirably. Similarly, the Haitian suffixes -wa and -yen derive
their representations from the French suffixes -ois, as in village-ois from the
village, and -ien as in paris-ien from Paris, respectively. Finally, the ordinal
suffix -ym derives its phonological representation from the corresponding French
suffix -ime. When we compare the Haitian affixes with their French counterparts, however, a more complex picture arises.
The productive Haitian affixes with their semantically closest French equivalents are presented in (16) (along the lines of Brousseau 1994a). The French forms
which are not in parentheses are those hypothesised to have provided the phonetic matrices from which the phonological representations of the Haitian affixes
were assigned; those in parentheses are forms which are synonymous (though
they may differ as to subcategorisation and morphophonemic properties).
(16)

Haitian affixes and their closest equivalents in French

a. Agentive suffix
base
output

haitian
-
v
n

french
-eur
v
n

b. Attributive suffix
base
output

-
n
n

-eur
n
n

(-ard, -ier, -ien)

c. Verbalising suffix
base
output

-e
n
v

-er, -
n/a
v

(-ir, -ifier, -iser)

d. Inversive prefix
base
output

dev
v

dv/a
v

(-, in-, ir-)

e. Diminutive prefix
base
output

tin
n

adjective petit [ti]

n/a
n

f. Nominalising suffix
base
output

-ay
v
n

-age
v
n

g. Nominal conversion
base
output

v
n

v
n

(-et, -ot, -on)

(-ion, -ment, -ance, -ure)

314

Are derivational affixes relexified?

h. Adjectival / participial
conversion
base
v
output
a

(-i, -, -ert, -u)

i. Adverbial suffix
base
output

-man
a
Adv

-ment
a
Adv

j. Place of origin
/ residence suffixes
base
output

-wa / -yen

-ois / -ien (-ais, -al, -and, -ain, -an)

n
n

n
n

k. Ordinal suffix
base
output

-ym
q
a

-ime
q
a

The data in (16) show that the derivational affixes of Haitian creole all have
at least one phonetically similar corresponding affix in French, except for the
phonologically null affix in (h) (see below). But the most striking fact about the
distribution in (16) is that, in most cases, except for the agentive suffix in (16a)
and the null affix in (16g), there are several French affixes corresponding to a
single Haitian affix (see also Brousseau, Filipovich and Lefebvre 1989: 18). For
example, while French has several overt affixes converting verbs into nouns,
Haitian has only one (see (16f) ); similarly, while French has several affixes
designating a place of origin, Haitian has only two (see (16j) ). Why did the
other French affixes not make their way into Haitian?
A point of similarity between the Haitian and the French inventories is that
both include cases of nominal conversion. An example of nominal conversion
in French is manger to eat and le manger the food/the action of eating (from
Furetire 1984). Nouns produced by means of morphological conversion in Haitian
(see (7) ) do not always correspond to a French noun so derived, however. This
is shown in (17), where the French data are drawn from Furetire (1984).
(17)

haitian
ale
sti
antre
vini
rive
tonbe
desan
monte
tunen
pati
pwche
gunmen
kuri
mache
dmi
manti
bwte

action
action
action
action
action
action
action
action
action
action
action
action
action
action
action
action
action

or
or
or
or
or
or
or
or
or
or
or
or
or
or
or
or
or

result
result
result
result
result
result
result
result
result
result
result
result
result
result
result
result
result

of
of
of
of
of
of
of
of
of
of
of
of
of
of
of
of
of

going
going out
entering
coming
arriving
falling
descending
going up
returning
leaving
approaching
fighting
running
walking
sleeping
lying
moving

17th / 18th
aller

venir

partir

marcher
dormir

century french
the going
the coming

the leaving

the walking
the sleeping

10.2 Derivational affixes of Haitian compared with French

315

It is worth noting that, although nominal conversion was very productive in

Middle French, its productivity started to decline during the sixteenth and seventeenth centuries in favour of the use of overt affixes. Haase (1965) remarks
that:
Lancienne langue employait couramment linfinitif avec lacception dun
substantif, mais cet usage devient plus rare au XVIIe sicle . . . cette poque
les substantifs forms dinfinitifs sont sensiblement les mmes que ceux dont
la langue actuelle se sert encore: ct de ceux qui sont tout fait courants
comme le souvenir, le lever, le coucher, le boire, le manger, etc. quelques-uns
qui sont dun usage plus rare comme le marcher, le pleurer, le vivre, le vouloir,
etc.
[The old language made frequent use of the infinitive with the sense of a noun,
but this usage became rarer in the seventeenth century . . . in this period the
nouns formed from infinitives were essentially the same as those still used in
the language today: in addition to the very frequently used ones like le souvenir
[remembrance], le lever [rising], le coucher [setting], le boire [drinking],
le manger [food or eating], etc., there are some rarer ones such as le marcher
[walking], le pleurer [crying], le vivre [living], le vouloir [will], etc.]

So, although the productivity of morphological conversion had diminished by

the end of the seventeenth century, it is likely that the varieties of French spoken
in Haiti at the time the creole was formed presented a small sample of deverbal
nouns derived by this process.
The data provided in Furetire (1984) show that the French suffix -age, the
phonetic source of Haitian -ay, was productive in French at the end of the
seventeenth century. This affix was used to derive nouns from verbs. Furetires
data show that this French affix could select either one- or two-place predicates,
as shown in (18).5
(18)

arbitr-age
abord-age
affin-age
badin-age
barbouill-age

arbitration
boarding / grappling
refining
teasing
daubing / smearing

french

(from Furetire 1984)

Consequently, the fact that Haitian -ay selects only two-place predicates cannot
be traced to French -age.
It is notable that, while Haitian has adjectival conversion, French does not
(see (16h) ). Indeed, French derives adjectives from the past participial form of
verbs (see e.g. Levin and Rappaport 1986) rather than from the verbs themselves
(e.g. La viande est cuite the meat is cooked; la viande cuite the cooked meat).
This discrepancy between the two languages raises the question of why and how
adjectival conversion developed in Haitian.
The Haitian and French data may also be compared from the point of view of
the position of the affix with respect to its base. As can be seen in (16), Haitian

316

Are derivational affixes relexified?

prefixes generally correspond to French prefixes and Haitian suffixes to French

suffixes. I assume that this is attributable to the fact that French affixes provided the phonetic matrices for establishing the phonological representation of
the productive Haitian affixes. There is one exception to this general pattern,
however: in contrast to French, where the diminutive affix is a suffix, in Haitian,
it is a prefix. Why did the creators of Haitian not simply use the phonetic matrix
of one of the French diminutive suffixes to provide a phonological form for
this lexical entry?
Haitian and French derived words can also be contrasted from the point of
view of their semantics. Although the individual affixes in the two languages
share a common meaning, and although there are Haitian derived words which
have the same structure as and a similar interpretation to French derived words,
a large number of Haitian derived words have a different meaning from their
French counterparts or simply do not have French counterparts. For example,
the Haitian verb de-respkte to lack respect/insult is made up of the inversive affix de- (=Fr. d-) and the verb respkte (<Fr. respecter to respect).
There is no French verb derived from the concatenation of d- and respecter;
French expresses the notion to lack respect by the locution manquer de
respect. Similarly, the Haitian noun meaning uproar is skandal; this Haitian
noun derives its phonological representation from the French word scandale
scandal. From the Haitian noun it is possible to derive another noun by
means of the attributive affix - yielding skandal- loud, rowdy (see Valdman
et al. 1981). In French, however, it is not possible to derive an attributive
noun from the noun scandale (e.g. *scandal-eur; *scandal-ard; etc.). A sample
of Haitian derived words (drawn from Valdman et al. 1981) in Appendix 3
provides several additional examples of this type. These data show that, in
Haitian, the concatenation of specific affixes with specific bases does not
always conform to the French pattern in spite of the fact that almost all the
Haitian affixes derive their phonological representation from the phonetic
matrices of French affixes. Should these cases be explained in terms of independent developments internal to Haitian or is there another explanation for
these discrepancies?
Finally, while the inventory of Haitian affixes in (16) exhausts the list
of derivational affixes in this language, the inventory of French affixes in (16)
is far from complete. According to Dubois (1962), for example, French has
over 80 derivational suffixes (excluding Greek and Latin affixes involved in
the vocabulary of the natural sciences, as well as all prefixes). These are listed
in (19).
(19)

Derivational suffixes of French (Dubois 1962)

a. Nominal suffixes verbal base
-age, -ement, -tion, -ure, -is, -at, -eur, -euse, -oir(e)
b. Nominal suffixes adjectival base
-it, -ance, -ence, -eur, -esse, -ise

10.2 Derivational affixes of Haitian compared with French

317

c. Nominal suffixes nominal base

-at, -aire, -ien, -ier, -iste, -erie, -ole, -ule, -ade, -aie, -aille, -aine, -e
d. Nominal suffixes adjectival or nominal base
-isme, -itude, -ie, -o, -et, -illon, -ot, -in
e. Nominal/adjectival suffixes nominal base
-ain, -ais, -an, -ien, -ois, -iste
f. Nominal/adjectival suffixes any base
-ard, -asse, -aud, -eux
g. Adjectival suffixes nominal base
-aire, -al, -el, -eux, -ier, -if, -in, -ique, -oire, -esque, -isant, -, -u
h. Adjectival suffixes adjectival base
-tre, -ot, -ime, -(u)ple
i. Adjectival suffixes verbal base
-able, -ible, -uble, -ant, -ent, -escent
j. Verbal suffixes nominal/adjectival base
-er, -ir, -ifier, -ser
k. Verbal suffixes verbal/adjectival base
-ailler, -asser, -eter, -iller, -iner, -ocher, -onner, -oter, -ouiller, -oyer
(=(11) in Brousseau 1994a)

A comparison of the list of Haitian affixes in (16) with the French list in
(19) immediately leads to the conclusion that the majority of the derivational
affixes of French (which has over a hundred affixes) have no Haitian counterpart.
A survey of the literature available on derivational affixes in seventeenth- and
eighteenth-century French (see Rey 1992; Nyrop 1936; Julliand 1965)6 and additional work by Olivier Tardif (a research assistant) reveals that there were more
than 90 productive affixes in seventeenth- and eighteenth-century French and
most of the affixes listed by Dubois were already in use at the time Haitian
creole was formed. According to Tardif, the major difference between classical
and modern French lies in the productivity of specific affixes rather than the size
of the inventory of productive affixes. Thus, whatever the exact number of productive French affixes at the time the creole was formed, we can safely conclude
that they far outnumbered Haitian affixes.
So, although French contributed the phonetic matrices of the Haitian affixes
and, in most cases, their position with respect to the base they attach to, the
comparison between the derivational affixes of Haitian and French leaves us
with a number of questions: Why has Haitian developed this particular set of
affixes? Why is the inventory so small compared with that of French? Why has
Haitian developed adjectival conversion? Should the discrepancies between a
significant number of Haitian and French derived words be considered as independent developments in Haitian? With these questions in mind, I shall now
examine the derivational affixes of Fongbe.

318
10.3

Are derivational affixes relexified?

The derivational affixes of Haitian compared with those
of Fongbe

In her study of the derivational affixes of Fongbe, Brousseau (1990, 1993)

identifies the following productive affixes.
The suffix -tV derives agentive nouns from dynamic verbs.7
(20)

Agentive suffix -tV: [v-tv]N

hn-j-tv
zw-w-tv

singer
worker

fongbe
j-hn
(generate-song) sing
w-zw (do-work)
work
(=(16) in Brousseau 1990)

The attributive suffix -nW selects nominal bases. The derived meaning is one
who possesses or uses noun.
(21)

Attributive suffix -nW: [n-nw]N

dy-kk-nw
n-bl-nw
gy-nw
xys-nw

mute
person who fasts
criminal
coward

fongbe
dy-kk
n-bl
gy
xys

dead tongue
tied mouth
crime
fright
(=(14) in Brousseau 1990)

The inversive prefix m- selects verbal and adjectival bases. It derives new
verbs and adjectives which have a meaning opposite to that of the base form.
(22)

Inversive prefix m-: [m-v/a]V/A

m-sv
not to take
m-3
to doubt
m-ssv untakable
m-33 unbelievable

fongbe
sv
3
ssv
33

to take
to believe
takable
believed, believable
(=(26), (27) in Brousseau 1990)

The diminutive suffix is -v. The meaning of nouns derived by means of

this affix is either a small noun or, when the base refers to an animate entity,
a young noun.
(23)

Diminutive suffix -v: [n-v]N

w-v
kkl-v
tn-v
tv-v

kitten
chick
bush
stool

fongbe
w
cat
kkl chicken
tn
tree
tv
table
(=(9), (10) in Brousseau 1990)

Fongbe has two suffixes of origin: -tV and -n. These suffixes select nominal
bases referring to a city or a country. The output of this derivation yields nouns
designating the people from that location. The principles governing the choice
between -tV and -n are unknown to linguists but native speakers know exactly
when to use one or the other (A.-M. Brousseau p.c.).

10.3 Derivational affixes of Haitian compared with Fongbe

(24)

a. Place of origin/residence suffix -tV: [n-tv]N

Bnxn-tv from Benin
Bnxn
Ktvn-tv from Cotonou
Ktvn

319

fongbe
Benin
Cotonou

fongbe
b. Place of origin/residence suffix -n: [n-n]N
Tbxy-n
from Tibet
Tbxy
Tibet
Glxw-n from Ouidah
Glxw Ouidah
(=(18) in Brousseau 1990)

The suffix -gVV derives ordinal numbers from cardinal numbers.

(25)

Ordinal suffix -gVV

w-gvv
tn-gvv
k-gvv
kn3-gvv

second
third
twentieth
fortieth

fongbe
w
tn
k
kn3

two
three
twenty
forty
(=(20) in Brousseau 1990)

Brousseau (1993) also identifies a prefix which involves reduplication of

the base. Following Marantz (1982), Brousseau (1990, 1991, 1993) analyses
reduplication as a kind of affixation. The affix involved in reduplication is
different from other affixes only in that it is phonologically underspecified, that
is, it is only specified for a skeletal template (it lacks a phonetic representation). The segmental content of the template is determined by the phonological
form of the base. As can be seen in the examples below, the reduplication
process in Fongbe involves a monosyllabic prefix (a cv template) which acquires
its full phonetic content by spreading the segmental (i.e. the consonant) and
tonal content of the base and inserting the epenthetic vowel /i/. This prefix,
henceforth referred to as RE-, bears the feature [+n] (see Brousseau 1993) (see
section 10.4). It converts verbal bases into nouns which refer to the action
denoted by the verb, as in (26a), or to the result of the action denoted by the
verb, as in (26b).
(26)

Nominalising prefix: [re-v]N

a. gbgb
wwln
d3
xxw

construction
writing
preparation
buying

b. xxw lz the buyings

xwl lz the writings

fongbe
gb
wln
3
xw

to
to
to
to

construct
write
prepare
buy
fongbe
(from Brousseau 1994a)

As Brousseau (1993) argues, the prefix re- also converts verbal bases into
adjectival forms and plays a role in the formation of participial forms involved
in the syntactic passive, as shown in (27).8

320
(27)

Are derivational affixes relexified?

Adjectival/participial prefix: [re-v]A/ PA
gbgb
wwln
xxw
33

constructed
written
bought
cooked

fongbe
gb
wln
xw
3

to
to
to
to

construct
write
buy
cook
(from Brousseau 1993)

These derived forms are found in adjectival passive clauses, as in (28a), verbal
passives, as in (28b), and noun phrases, as in (28c).
(28)

a. Ln v
3 33.
meat det at cooked
The meat is cooked.

b. Ln v
ny 33.
fongbe
meat det be cooked
The meat has been cooked.

c. ln
33
v
meat cooked det
the cooked meat

fongbe
(from Brousseau 1993)

As Brousseau (1993: 120) points out, this morphological process may apply to
all verbs which involve a change of state such as those in (27). It does not apply
to locative verbs such as y go, tn push, dYn pull, nor to verbs of movement such as jY fall, tVn jump, or ln swim. As is argued in Brousseau
(1993: 115), the derived adjectival/participial form has the predicate-argument
structure of the verbal base modified in a uniform way: the external argument is
eliminated and the internal argument is externalised. For example, the verb #
to cook takes an external and an internal argument. In its adjectival/participial
form, as in (28a and b), the external argument has been eliminated and the
internal argument surfaces as the subject.
Reduplication is thus involved in the derivation of nouns as well as of
adjectival/participial forms. The issue of whether there are one or two prefixes
involving reduplication will be taken up in section 10.4. For the sake of clarity,
I will treat them as separate forms for now.
The affixes discussed above exhaust the list of productive affixes in Fongbe.
The table in (29) compares the Haitian affixes in (15) with the Fongbe ones
identified above.
(29)

Haitian affixes and their equivalents in Fongbe

a. Agentive suffix
base
output

haitian
-
v
n

fongbe
-tV
v/n
n

b. Attributive suffix
base
output

-
n
n

-nW
n/v
n

c. Verbalising suffix
base
output

-e
n
v

10.3 Derivational affixes of Haitian compared with Fongbe

321

d. Inversive prefix
base
output

dev
v

mv/a
v/a

e. Diminutive affix
base
output

tin
n

-v
n
n

f. Nominalising suffix
base
output

-ay
v
n

copy prefix
v
n

g. Nominal conversion
base
output

v
n

copy prefix
v
n

h. Adjectival / participial conversion

base
output

v
a

copy prefix
v
a

i. Adverbial suffix
base
output

-man
a
Adv

j. Place of origin / residence suffixes

base
output

-wa / -yen
n
n

-tV / -n
n
n

k. Ordinal suffix
base
output

-ym
q
a

-gVV
q
a

The Haitian and Fongbe derivational affixes in (29) differ in their phonological representation and they may also differ in their position with respect to the
base they attach to. As we saw in section 10.2, these properties of the Haitian
affixes appear to have been largely provided by the superstratum language. In
spite of these differences, however, there is a striking resemblance between the
two inventories: in most cases, there is a one-to-one correspondence between the
Haitian and Fongbe affixes. In both languages, there is one agentive, one attributive, one inversive, one diminutive and one ordinal affix. It is a remarkable fact
that, in both Haitian and Fongbe, there are exactly two suffixes referring to a
place of origin/residence (see (29j) ).9 Furthermore, Haitian - is involved in
both nominal and adjectival conversion, two constructions which require the
copy prefix in Fongbe (see (29g and h) ). The correspondence between the Haitian and Fongbe affixes enumerated above contrasts with the Haitian/French
data compared in (16), where it was shown that the majority of the Haitian
affixes correspond to more than one affix in French. How can the correspondences between the Haitian and Fongbe lexical entries be accounted for?
In spite of these similarities between the two inventories, there are also some
differences between them. While the formation of deverbal nouns in Haitian
involves two affixes, -ay and -, which, as we saw earlier, are in complementary
distribution, the formation of deverbal nouns in Fongbe involves only one affix:

322

Are derivational affixes relexified?

the copy prefix. These data constitute a first mismatch between the two inventories. A second mismatch concerns the verbalising affix -e and the adverbial affix
-man, which have no counterparts in Fongbe. These affixes do, however, have
corresponding affixes in French.
Derived words in Haitian and Fongbe can also be contrasted from the point of
view of their semantics. Recall from section 10.2 (and Appendix 3) that there is
quite a large number of Haitian derived words which do not correspond to
derived words in French. Interestingly enough, in most such cases, there is a
Fongbe counterpart to the Haitian word. For example, while there is no French
derived word corresponding to Haitian eskandal- loud, rowdy, as we saw
earlier, there is a corresponding Fongbe word derived by means of the attributive suffix -nW: zg#-nW loud, rowdy. While cases of nominalising conversion
(derived by means of the affix -) in Haitian (see (7) ) do not all have a corresponding lexical item so derived in French (see (17) ), all the Haitian deverbal
nouns in (7) have a corresponding Fongbe lexical item derived with the copy
prefix, showing that the concatenation of Haitian - and Fongbe RE-, respectively, with particular bases follows a similar pattern. This is shown in (30).
(30)

french

haitian
tonbe
desan
tunen
pati
pwche
gunmen
kuri
mache
dmi
manti
bwte

fongbe
j-jy
j-jxt
l-lxkw
y-y
s-sxw
x-xhn
k-knwzn
3-3znln
3-3mlwn
3-3
3-3vxw

action
action
action
action
action
action
action
action
action
action
action

or
or
or
or
or
or
or
or
or
or
or

result
result
result
result
result
result
result
result
result
result
result

of
of
of
of
of
of
of
of
of
of
of

falling
descending
returning
leaving
approaching
fighting
running
walking
sleeping
lying
moving

Similarly, we saw that Haitian nouns derived by means of the nominalising

suffix -ay (see (6) ) often do not have a French equivalent derived in this way
(see (F) in Appendix 3). In most cases, though, there is a Fongbe deverbal noun
derived by means of the copy prefix that corresponds to the Haitian deverbal
noun with -ay. A few examples are provided in (31).
(31)

french

haitian
randui-z-ay
sweny-ay

fongbe
s-s
w-wlbn

pary-ay
k-kn
boukan-t-ay 3-3v

coating or action of coating

good care or action of taking
good care
bet or action of betting
swap or action of swapping

A similar set of examples involves the Haitian inversive prefix de- (see (4) ).
Several Haitian words derived by means of this affix were shown to have no
French counterpart (see (D) in Appendix 3). In these cases, however, we find equivalent derived words in Fongbe involving the inversive prefix m-. Examples
are provided in (32).

10.4 Historical derivation of the Haitian derivational affixes

(32)

french

haitian
de-makn-n-en
de-rspkte
de-take
de-z-apiye

fongbe
m-xz
m-sn
m-xw
m-zn

to
to
to
to

323

untangle
be disrespectful
unlatch
stop leaning

The data in (30)(32) strongly support the claim that the concatenation of particular affixes with particular bases in Haitian follows the pattern of the substratum
language, and that the mismatches between Haitian and French can be explained
by the hypothesis that the creators of Haitian used the semantics of their own
language in concatenating affixes and bases.
Finally, although the correspondences between the two inventories in (29) are
not perfect, the striking fact is that they are of comparable size, in contrast to
that of French (see section 10.2). Indeed, the lists in (29) exhaust the productive
affixes in both Haitian and Fongbe. This strongly suggests that the size of the
Haitian inventory of derivational affixes has, to a great extent, been determined
by the substratum lexicon.
So, on the basis of the comparison between Haitian and Fongbe, it appears
that the substratum language has contributed the semantics of the Haitian affixes,
the principles which govern their concatenation with their bases and the extent
of the inventory. The obvious overall similarity between the Haitian and Fongbe
inventories of derivational affixes leaves us with three major questions, however.
First, how can the mismatch between Haitian -ay/- and Fongbe RE- be accounted
for? Second, how can we derive the fact that the position of the Haitian affixes
with respect to their base does not always match that of either of the source languages? Finally, how did the Haitian inventory of derivational affixes come to
be the way it is?
10.4

The historical derivation of the Haitian derivational affixes

This section addresses the historical derivation of the Haitian derivational affixes
from the point of view of the inventory itself and of the position of morphological heads with respect to their bases. The comparative data from Haitian and
Fongbe in (29) divide up into three groups in terms of their congruence. The first
group contains the affixes that have a one-to-one correspondence in the two
languages, that is, the affixes in (29a, b, d, e, j and k). The second group contains
the affixes involved in nominal and adjectival conversion in (29f, g and h),
where the correspondences between the two languages are not perfect. Finally,
the third group contains the two Haitian affixes in (29c and i) which have no
equivalents in Fongbe. These three sets of data will be discussed in turn.
10.4.1

Establishing the lexical entries of the Haitian affixes that are

congruent with those of Fongbe

In this section, I argue that the Haitian affixes in (29a, b, d, e, j and k) can be
accounted for straightforwardly by the process of relexification. I consider two

324

Are derivational affixes relexified?

possible scenarios, both based on the formal definition of relexification provided

earlier (see (2) in chapter 2). Recall that relexification is a process that builds
new lexical entries by copying established lexical entries and assigning them
either a phonological representation on the basis of phonetic matrices found in
the superstratum language or a null form.
A first scenario for the historical derivation of Haitian affixes would thus
consist in saying that the copied lexical entries were relabelled on the basis of
French phonetic matrices. In this scenario, the creators of Haitian would have
identified phonetic matrices which had an appropriate meaning to relabel the
derivational affixes lexical entries copied from their own lexicons on the basis
of pairs of French words such as faire to do and d-faire to undo, or travailler to work and travaill-eur worker. Hence, a Fongbe speaker relexifying his
lexicon on the basis of data from French would have identified French d- as
sharing with m- the meaning inversive and he would have assigned the lexical
entry copied from m- the phonological form d-. Similarly, the lexical entry
copied from the agentive affix -tV would have been assigned the label - on the
basis of the French agentive affix -eur, and so on. This scenario accounts straightforwardly for the fact that there is a one-to-one correspondence between the
Haitian and Fongbe agentive, attributive, inversive, diminutive, ordinal and place
of origin affixes (see (29a, b, d, e, j and k) ).
A second scenario would suggest that, as a first step, the copied lexical
entries were relabelled with a phonologically null form and then, as a second
step, they were assigned a phonological form through the process of reanalysis.
In this case, the creators of Haitian would have incorporated French words into
the new lexicon as unanalysed forms, that is, regardless of whether they were
simplexes or derived words in French. They would have identified the forms of
the productive affixes on the basis of pairs of words such as f (<Fr. faire) to
do and def (<Fr. dfaire) to undo, and travaye (<Fr. travailler) to work and
travay (<Fr. travailleur) worker, etc., within the incipient Haitian lexicon. Of
course, they would have identified only forms which corresponded to those of
their native lexicons. In this view, the process of assigning a phonological form
to a copied lexical entry would resemble reanalysis, since it would use linguistic
material taken from within the incipient creole. In this second scenario, the
similarity between the Haitian and Fongbe lexical entries would also be attributable to relexification, since reanalysis would have assigned a phonological form
to a copied lexical entry initially relabelled as a null form. Thus, when these lexical entries were assigned a phonological form through reanalysis, their properties
were like those of the substratum rather than the superstratum lexical entries.
There is a strong argument against this second scenario. If the bulk of the
lexicon of a creole language is built through relexification, as we have seen
throughout this book, and given that derived words are not listed in the lexicon (only their constituents are), how could French words such as dfaire to
undo and travailleur worker even enter the incipient creole since their Fongbe
counterparts are not listed in the lexicon? Unless it can be demonstrated that

10.4 Historical derivation of the Haitian derivational affixes

325

relexification is not the main tool in building the lexicon of a creole language,
this second scenario cannot be considered plausible.
As for the first scenario, it entails that, on the basis of a few pairs of words,
the creators of the creole had access to the internal structure of French words. At
first glance, there is no reason to believe that this could not be the case since, for
other major category lexical entries, they probably had to establish the meanings
of words on the basis of pairs such as small/big, come/go, etc. Furthermore, since derivational affixes are lexical categories that are listed in the lexicon, there is no principled reason to rule out the possibility that they undergo
relexification in the same way as other lexical categories, i.e. through copying
and relabelling on the basis of French phonetic matrices. I therefore conclude
that the first scenario is correct.
Relexification accounts for the fact that the properties of the Haitian creole
affixes discussed in this section parallel those of the substratum language. Most
of all, in the case of these affixes, the relexification hypothesis accounts for the
fact that, when French has several affixal forms performing the function of one
Haitian affix (see (16b, d, e and j) ), Haitian has exactly the same number of
forms as the substratum lexicon (see (29a, b, d, e, j and k) ). The extra French
affixes in (16b, d, e and j) and (19) did not make their way into Haitian because
there were no such lexical entries in the original lexicon to be relexified.
10.4.2

The problem of -ay/- and re-

At first glance, the partial mismatch between the Fongbe copy prefix RE- and the
Haitian affixes involved in nominal and adjectival conversion (see (29f, g, h) )
poses a problem for the relexification account. First, whereas Fongbe nominal
conversion involves only the affix RE-, Haitian nominal conversion involves two
affixes: -ay and -. Second, whereas Fongbe RE- is also involved in adjectival
conversion, as is Haitian -, -ay is not. Third, while it is clear from the data in
(31) that the Haitian suffix -ay fulfills some of the functions of RE-, it cannot be
analysed as having relabelled RE-. The two affixes are not categorially identical.
Since -ay only derives nouns from verbal bases, it must be identified for the
categorial features [+n, v]. By contrast, RE- derives both nouns and adjectives
from verbs, and thus, as will be seen below, it is categorially identified for the
features [+n, v]. Furthermore, -ay selects only two-place predicates, whereas
RE- selects either one- or two-place predicates. The fact is that the speakers of
a language like Fongbe had to relexify the affix RE- which converts verbs into
nouns or adjectives. They were faced with French, where deverbal nouns are
created by means of zero affixation or by several overt affixes such as -age, -ion,
-ur, -ment, etc., and where deverbal adjectives are created on the basis of the
past participle of the verb. Haitian ended up with two affixes - and -ay performing the functions of RE- in the substratum language. How did this happen?
Recall from chapter 3 that the properties of Haitian lexical entries which do
not conform to the predictions of the relexification hypothesis may be attributable

326

Are derivational affixes relexified?

to the contribution of other West African languages, and that in such cases we
may have identified data which are pertinent to the process of dialect levelling.
I argue that the partial mismatch between Fongbe RE- and Haitian -ay/- constitutes
such a case.
While reduplication is widely used among Kwa languages to derive adjectives from verbs (see Westerman and Bryan 1970), its use to derive nouns from
verbs is restricted to a subset of the Kwa languages. Gbe languages, such as
Fongbe, are part of this subset. Other Kwa languages such as Twi, Ewe, Yoruba
and Nupe use another affix. This is illustrated in (33).
(33)

a. a-duw
n-na

hoeing
sleeping

<duw
<da

b. o-gblo

breadth

<gblo be broad

c. =-da

creature

<da

create

d. i-bi

wickedness

<bi

be bad
nupe
(from Westerman and Bryan 1970: 91)

hoe
sleep

twi
ewe
yoruba

So, while the Fongbe speakers had only one affix, RE-, to relexify, the speakers
of Ewe, Yoruba, etc., had two: an affix with phonetic content which converted
verbs into nouns, and the affix RE- which converted verbs into adjectives. When
the speakers of these two groups of languages relexified their lexicons on the
basis of French data, there would thus be two dialects in the incipient creole: one
with one affix and one with two affixes. Suppose that, in the latter case, the overt
affix converting verbs into nouns was relabelled as -ay, on the basis of French
-age, and RE- was assigned a null form; this would give us the two affixes that we
find in present-day Haitian. Dialect levelling would then account for the distribution of -ay and - that we find in modern Haitian. This proposal is presented
in more detail below. I begin with a theoretical discussion of the properties of
RE- as manifested in Kwa languages of the type of Fongbe.
Since this affix is involved in both nominal and adjectival conversions (see
(29f, g and h) ), one might ask whether it constitutes one or two lexical entries.
Lieber (1980) has argued extensively against the proliferation of phonetically
null affixes in the lexicon. Null affixes are not distinguishable from each other,
which would cause problems from the point of view of both morphological
derivation and learnability. Drawing on these considerations, Brousseau (1993)
claims that in Fongbe there is only one lexical entry RE-. She proposes that it has
the representation in (34).10
(34)

RE-:

/xx/
[+n, v]

[+v, n]

fongbe
(=(24) in Brousseau 1993)

The phonological representation of this affix is a skeletal template which consists only of two points. Thus, the affix has a phonological representation, but it
is phonetically underspecified (it has no phonetic content). As shown in (34),

10.4 Historical derivation of the Haitian derivational affixes

327

this affix selects verbal bases, that is, bases identified for the features [+v, n].
Since Fongbe RE- converts verbs into either nouns or adjectives, Brousseau (1993)
proposes that the syntactic features associated with it are partially underspecified.
Indeed, while RE- is specified for the feature [+n], it is not specified for any
value of the feature [v]. For the sake of the following argument, I will assume
that RE- is defined by the syntactic features [+n, v].11 This affix attaches to
verbal bases to derive nouns, as in (35a), or adjectives, as in (35b) (adapted from
(25) in Brousseau 1993).
(35)

a.

Derived nominals
[+n, v]
[+v, n]

RE[+n, v]

b.

Derived adjectives

fongbe

[+n, +v]
RE[+n, v]

[+v, n]

Brousseau (1993) argues that, in both cases, RE- is the morphological head of
the derived word. Its categorial feature [+n] percolates la Lieber (1980) and
assigns the derived word the feature [+n], a syntactic feature shared by both
nouns and adjectives. Brousseau (1993) accounts for the differential value of
[v] in the derived words ( [v] in derived nouns and [+v] in derived adjectives) in saying that [v]s value is assigned arbitrarily. If the value is [v], the
derived word is a noun, as in (35a), and if it is [+v], the derived word is an
adjective, as in (35b).
Let us propose a similar account of the Haitian null affix involved in nominal
and adjectival conversion. The lexical entry of this affix would have the representation in (36) where it is neither phonologically nor phonetically specified.
This affix selects verbal bases and is specified for the categorial features [+n,
v]. As will be argued below, the Haitian null affix must be a suffix.
(36)

[+v, n]

/-/

haitian

[+n, v]

Suppose further that the derivation of nouns and adjectives proceeds in Haitian
along the lines proposed for Fongbe in (35), deriving nouns or adjectives depending on whether [v] is assigned a positive or a negative value. This is shown
in (37).
(37)

a. Nominal conversion
[+n, v]
[+v, n]

-
[+n, v]

b. Adjectival conversion

haitian

[+n, +v]
[+v, n]

-
[+n, v]

The Haitian null affix can be seen as the counterpart of the Fongbe copy affix,
an affix which has no phonetic content. I hypothesise the following historical

328

Are derivational affixes relexified?

derivation within the framework of the relexification hypothesis. The creators of

Haitian who had a native language of the type of Fongbe assigned a null form to
the lexical entry copied from RE- yielding the Haitian lexical entry in (36). In this
view, the relexification of RE- would have proceeded as depicted in (38).
(38)

original lexical entry

/xx/
[+n, v]

target language

early haitian

[+v, n]

dialectA

creole lexical entry

[+v, n]

/-/
[+n, v]

The new lexical entry had the syntactic properties of the copied lexical entry
because it was created through relexification. Both the Haitian and Fongbe lexical entries are identified for the features [+n, v], allowing the affix to derive
both nouns and adjectives.
I now turn to the lexicons of speakers who had two lexical entries to relexify,
i.e. those who, in addition to having the RE- affix converting verbs into adjectives, had an overt affix converting verbs into nouns, as in (33). These speakers
had a nominalising prefix identified for the features [+n, v] to relexify. From
all the available French affixes, they chose -age to relabel the lexical entry copied
from their original lexicon. This French form shares with the original lexical
entry the property of converting verbs into nouns, and thus it was appropriate to
relabel the copied lexical entry. Relexification of the nominalising prefix thus
proceeded as in (39). (Affix order will be discussed below.)
(39)

Relexification of the nominalising affix

original lexical entry
/affix/
[+n, v]

target language
[ . . . -an]

[+v, n]

creole lexical entry

[+v, n]

/-ay/
[+n, v]

These speakers also had RE-, converting verbs into adjectives, to relexify. Since,
in the above-mentioned languages, RE- converts verbs only into adjectives, I will
assume that in these languages it bears the syntactic features [+n, +v], as shown
in (40). These speakers also assigned a null form to the new lexical entry.

10.4 Historical derivation of the Haitian derivational affixes

(40)

Relexification of REoriginal lexical entry

/xx/
[+n, +v]

329

target language
[......]

[+v, n]

creole lexical entry

[+v, n]

/-/
[+n, +v]

The early Haitian dialect of these speakers thus contained the two lexical entries
-ay and -, deriving deverbal nouns and deverbal adjectives respectively, as
shown in (41).
(41)

a.

Derived nominals
[+n, v]
[+v, n]

-ay
[+n, v]

b.

Derived adjectives
[+n, +v]
[+v, n]

early haitian
dialect B

-
[+n, +v]

On this view, then, there were two early Haitian dialects with respect to the
affixes discussed in this section: (38) and (41). Some compromises had to be
made in order to accommodate differences between them. I assume that these
accommodations were achieved by the process of dialect levelling. I hypothesise
the following compromises between the two groups of speakers. Speakers of
group A adopted the lexical entry -ay from speakers of group B. Speakers of
group B modified the [+v] value of the null affix as [v], which allowed them to
convert verbs into nouns, as in (37a). Both dialects had to settle on a principle
which would result in the complementary distribution of -ay and - in forming
nouns from verbs. Apparently, -ay came to be used only with two-place predicates and - only with one-place predicates, since this is what we see in modern
Haitian (see section 10.1). This specialisation of the two affixes had to comply
with the constraint referred to as blocking in the literature on morphology (see
e.g. Aronoff 1976; Allen 1978; etc.).12 According to Aronoffs (1976: 41) proposal, Blocking is the non-occurrence of one form due to the simple existence
of another. For example, the existence of the base noun glory blocks a derivation with -ity, hence *gloriosity. Similarly, the derivation of a noun by means of
the suffix -ay in Haitian blocks zero affixation.
One might wonder why the creators of Haitian did not simply keep RE- from
their own lexicons. This option cannot be ruled out a priori for at least two
important reasons. First, reduplication of verbal bases to form nouns and adjectives is a widespread phenomenon in West African languages and it is a salient
feature of the Kwa languages (see Westerman and Bryan 1970). Second, other

330

Are derivational affixes relexified?

Caribbean creole languages did in fact choose this option.13 For example, in
Berbice Dutch, deverbal nouns may be formed by reduplication of the base verb
as in koso koso the cough(ing) (see Kouwenberg 1994: 249). Furthermore, in
Saramaccan, there is a reduplication prefix similar to the one found in Fongbe
which derives adjectives from verbs, as in (42).
(42)

a. di lailai
goni
the load-load gun
the loaded gun

saramaccan
(=(9) in Bakker 1987)

b. De fisi de kuakua.
the fish be fresh-fresh
The fish is fresh.

saramaccan
(=(12) in Bakker 1987)

The Saramaccan data in (42) parallel the Fongbe data in (28). Given these facts,
why was this option not chosen by the creators of Haitian? Brousseau, Filipovich
and Lefebvre (1989) propose that the answer to this question follows from another
option selected by the creators of Haitian, i.e. setting the position of the morphological head as the rightmost position in the word.
10.4.3

Establishing the position of the morphological head in

creole genesis

Di Sciullo and Williams (1987: 26) define the morphological head as follows:
The headF of a word is the rightmost element of the word marked for the feature
F. According to this definition, the head of a word is the rightmost constituent
that determines the categorial features of the word by percolation of its features
to the top node of the word. For example, the head of the English word driver is
the affix -er. This affix bears the categorial features [+n, v], which, by percolation, determine the categorial features of the derived word. This is illustrated in
(43), where the head of the derived word is in bold.
w [+n, v]

(43)
v

af

driv[n, +v]

er
[+n, v]

english

By contrast, in the French word maison-ette small house, the diminutive affix
-ette does not bear any categorial features and consequently it does not contribute to the feature specification of the derived word. The head of the derived
word in this case is the base of the derived word, maison house, because this is
the constituent that determines the derived words categorial features. The word
structure of this derived word is as in (44), where the head is in bold.

10.4 Historical derivation of the Haitian derivational affixes

w [+n, v]

(44)
n

af

maison
[+n, v]

-ette

331

french

While in (43) the head of the word is the rightmost constituent of the word, in
(44) it is not. But in both cases the head is the rightmost constituent that bears
categorial features.
Languages may vary as to whether they allow morphological heads not to be
the rightmost constituent of a word (see Brousseau 1988a; Lieber 1992). In light
of this preliminary discussion, we can compare the position of the morphological heads in Haitian and its contributing languages. The data in (45) show the
position of the morphological head (underlined) in different types of derived
words in the three languages.
(45)
Agentive
Attributive
Verbalising
Inversive
Diminutive
Nominalising
Adverbial
Place of origin
Ordinal

haitian
v-
n-
v-e
de-v
ti-n
v-ay
a-man
n-wa / -yen
q-ym

french
v-eur
n-eur
n/a-er / -
d-v/a
n/a-et / -ot / -on
v-age
a-ment
n-ois / -ien
q-ime

fongbe
v/n-tV
n/v-nV

m-v/a
n-vJ
RE-v

n-tV / -n
q-gVV

As can be seen in (45), Fongbe, and to a lesser extent French, allows nonrightmost elements to be the head of a word. By contrast, in Haitian, the morphological head is always the rightmost constituent of a derived word. In view
of this fact, Brousseau, Filipovich and Lefebvre (1989: 25) propose that, during
Haitian creole genesis, the position of the head was fixed (or parameterised) as
the rightmost position in a derived word. In this view, the alternative possibility
allowing the morphological head sometimes not to be the rightmost constituent of a derived word had to be abandoned. This accounts for the fact that the
Fongbe diminutive suffix, which is not the head of any word it is part of, could
not remain a suffix when it was relexified. Hence, the numerous French diminutive suffixes which could have been used to relabel the lexical entry copied from
-vJ were not identified as such and the creators of the creole used a reduced form
of the prenominal adjective (petit, pronounced [p1ti], [pti] and [ti] small) to
assign a phonological representation to the diminutive affix. The option selected
by the creators of Haitian of setting the position of the head as the rightmost
position in the word thus accounts for the fact that the Haitian diminutive affix is
a prefix even though, in both of the source languages, it is a suffix. Second, and
most importantly, this parametric choice forced the creators of Haitian to abandon reduplication of the base as a means of deriving words. Hence, in relabelling
RE- as -, they had to assign this new affix a position which conformed to the

332

Are derivational affixes relexified?

value of the new parameter they had set. Since - contributes to the categorial
features of the derived words it is part of, it must be a suffix.14
The differences between the position of affixes with respect to their base in
Haitian and Fongbe and the fact that Haitian does not manifest a RE- prefix are
explicable on the hypothesis that the creators of Haitian established the position
of the morphological head as the rightmost position of the word.
10.4.4

The verbalising suffix -e and the adverbial suffix -man

Two of the affixes which can be argued to be native to Haitian (see section 10.1)
have no corresponding lexical entry in Fongbe: the verbalising suffix -e and the
adverbial suffix -man. How did these affixes become part of the inventory of
productive affixes in Haitian (see (15) )?
The verbalising suffix -e deriving verbs from nouns takes its phonological
representation from the French form -er (see (16) ). There is no such affix in
Fongbe and I have seen no indication that any other Kwa languages have such
an affix. Hence, I hypothesise that speakers of these languages did not identify
French -er. Mande languages, however, do have such a suffix. Westerman and
Bryan (1970) provide examples from Malinke and Vai.
(46)

a. foro-ya

to free

foro free

b. tusa-a

to ask

tusa

malinke

question
vai
(from Westerman and Bryan 1970: 44)

Since Mande speakers were present in Haiti at the time Haitian creole was formed
(see chapter 3, section 3.1), we could hypothesise that these speakers relexified
their verbalising suffix as -e on the basis of French -er in a way similar to that
discussed in section 10.4.1 for other affixes. On this assumption, in the incipient
creole, there would be two dialects of Haitian, one without a verbalising affix
and one with a verbalising affix -e. During the period of dialect levelling, speakers
of the first group would have acquired this suffix from those of the second group.
The likelihood of this hypothesis depends, I believe, on the size of the African
Haitian population who, at the time the creole was formed, had a verbalising affix
in their lexicon. Further discussion of this hypothesis will have to await additional
research on the productivity of the verbalising affix in West African languages.
Another possibility is that the creators of the creole simply acquired this
lexical entry directly from French. Support for this scenario might come from
the following facts. The list of French words bearing this affix in Furetires
(1984) seventeenth-century French dictionary is enormous, showing that it was
very productive at the time Haitian creole was formed. Furthermore, as Brousseau
(1994a) points out, in the variety of modern French spoken by native speakers
of Ewe (Togo), the affix -er is used to form denominal verbs such as flch-er
to throw arrows > flche arrow, flt-er to play the flute > flte flute (see
Lafage 1985). Thus, deriving verbs from nouns by means of the affix -er is a
very productive process in Ewe native speakers local French nowadays. Like

10.5 Conclusion

333

Fongbe, Ewe itself has no morphological process for deriving verbs from nouns.
Nevertheless, Lefebvre (1984) argues at length that direct borrowing of affixes is
unlikely because affixes do not present themselves as isolated entities since they
are bound morphemes. This conclusion leads to a third possibility.
Since, at the time Haitian creole was formed, there were many French words
inflected with -er, it could be hypothesised that these derived words were incorporated into the creole lexicon as simplexes (that is, as unanalysable words) in
much the same way as the French words ending with -ion referred to above.
Unlike Haitian words ending with -yon, however, Haitian simplexes ending with
-e would eventually have been reanalysed as having an internal structure: nominal base +e. In this view, the presence of -e as a productive affix of Haitian
would be the result of reanalysis from within the creole. But, in this case, the
presence of this affix in the creole would be attributable to the indirect influence of the lexifier language. This proposal could also account for the existence
of the adverbial suffix -man in the inventory of productive affixes in Haitian.
To the best of my knowledge, deriving adverbs from adjectives is not a property
of any West African languages.15
10.5

Conclusion

The comparison of the derivational affixes of Haitian, French and Fongbe presented in this chapter shows that, while the Haitian derivational affixes obtain
their phonological representation from French phonetic matrices, their semantics
follows that of the substratum language. With two exceptions (-e and -man), and
the partial mismatch between -/-ay and RE-, there is a one-to-one correspondence between Haitian and Fongbe affixes. These facts argue that the creators of
Haitian used the principles of their own grammar in concatenating affixes and
bases in the incipient creole. These facts have been argued to follow from the
relexification hypothesis. Furthermore, it has been extensively argued that the
derivational affixes of Haitian are concatenated with their bases in a way which
patterns on the substratum language rather than the superstratum. The partial
mismatch between -/-ay and RE- was shown to be explained by relexification
from different lexicons followed by dialect levelling. Hence, I conclude that
derivational affixes are indeed relexified. This is in agreement with theories
which advocate that derivational affixes are listed in the lexicon. The Haitian
facts also show that the inventory of derivational affixes in creole genesis is, to
a great extent, determined by that of the substratum language and not by universal principles, as has sometimes been claimed in the literature on creole languages (see e.g. Mhlhusler 1980: 36). Finally, the discrepancy between Haitian
and both of its source languages with respect to the position of affixes and their
bases has been accounted for by the proposal that the creators of Haitian set the
position of the morphological head as the rightmost position in word structure.
This would appear to be an innovation, since Haitian differs from both of its
source languages in this respect.

334

The concatenation of words into compounds

11 The concatenation of words into

compounds

Compound words are syntactic atoms, i.e. they constitute islands to or from
which no lexical material can be inserted, deleted or moved out by syntactic
rules (see e.g. Allen 1978; Di Sciullo and Williams 1987). All the compounds
discussed in this chapter are such syntactic atoms. Furthermore, the literature
on generative morphology generally assumes that the lexicon lists idiosyncratic
information about morphemes (see e.g. Lieber 1980, 1992; Sproat 1985; Di
Sciullo and Williams 1987). Consequently, compounds are listed in the lexicon
only if their properties cannot be predicted from the properties of their constituents and from the word structure rules. Thus, in contrast to idiosyncratic
or listed compounds, productive or regular compounds do not constitute lexical
entries, and, thus, they do not undergo relexification. The productive compounds
of Haitian creole are therefore the result of the concatenation of simple lexical
entries. As will be seen in this chapter, however, the semantics and structure
of many Haitian compounds parallel those of Fongbe compounds and contrast
with French, showing that the creators of Haitian used the principles of their
own grammar in concatenating simplexes into compounds. Furthermore, concepts expressed by compounds in Haitian and Fongbe are often expressed by
simplexes in French. The fact that the French simplexes did not make their way
into Haitian is argued to follow from the relexification hypothesis: the creators
of Haitian had no such lexical entries to relexify. The comparison of Haitian,
French and Fongbe compounds presented below addresses the topic from three
points of view: semantics, word order and types of compounds. The contents of
this chapter draw on work by Brousseau (1988a, 1989, 1994a),1 on compounds
in the three languages under comparison and on additional work that I have done
on the semantics of compounds using data available in dictionaries supplemented
with work with informants.

11.1

The semantics of compounds

Body parts are typically rendered in Haitian creole as compound nouns, even
though French uses simplexes to express these notions (see Brousseau 1988a,
1989). A sample of such compounds is given in (1).
334

11.1 The semantics of compounds

(1)

haitian
po-bouch

(peau-bouche)
skin-mouth

335

french
lvre
lip

twou-ne

(trou-nez)
hole-nose

narine

nostril

plim-je

(plume-il)
hair-eye

cil

eyelash

dy-kou

(derrire-cou)
back-neck

nuque

nape

kalbas-tt2
or tt-kalbas

(calebasse-tte)
calabash-head

crne

skull
(from Brousseau 1989)

As shown in (1), the Haitian nouns participating in the compound expressions

all derive their phonological forms from the phonetic matrices of French words.
However, in French, these words cannot be concatenated to form compounds.
Furthermore, Haitian has none of the French simplex forms in (1). Why is this
so? A comparison with the substratum languages provides a direct explanation.
In Fongbe, for example, these notions are typically realised as compounds.
The data in (2) present the Haitian compounds and their Fongbe counterparts.
(2)

haitian
po-bouch
twou-ne
plim-je
dy-kou
kalbas-tt
or tt-kalbas

skin-mouth
hole-nose
hair-eye
back-neck
calabash-head

fongbe
n-fl
wntn-d
wn-3
kw-gd
t-k

mouth-skin
nose-hole
eye-hair
neck-back
head-calabash

lip
nostril
eyelash
nape
skull

(from Brousseau 1989)

Abstracting away from word order (which will be discussed below), the semantic
parallels between the compounds in the two languages are striking. The Haitian
simplexes (which derive their phonological representation from French phonetic
matrices) are concatenated on the model of the substratum language. A potential
counterexample to this claim might come from Chaudensons (1990) observation that popular French uses the syntactic compound trou de nez hole of nose
instead of the simplex narine nostril and thus it is likely that the creators of
Haitian were exposed to this French syntactic compound. The Haitian expression
twou-ne could then be seen as following the French pattern. To the best of my
knowledge, however, no variety of French presents compounds of the type *poil
des yeux (lit. hair of the eye) for eyelash or *peau de la bouche (lit. skin of
the mouth) for lip or *calebasse de la tte for skull. I therefore conclude that
the structure of Haitian compounds referring to body parts cannot be argued to
be modelled on the structure of French compounds. The Haitian compounds in
(2) can, however, be seen to be modelled on the structure of the substratum

336

The concatenation of words into compounds

language, showing that the creators of the creole used the principles of their own
grammar in compounding simplexes. As was pointed out in Lumsden (1994b)
the relexification hypothesis accounts for the fact that Haitian does not have
simplexes denoting these body parts: there were no such lexical entries for the
creators of Haitian to relexify.
A second set of data supporting the claim that the structure of Haitian compounds follows the pattern of the substratum language rather than that of the
superstratum language comprises compounds which refer to a person having a
property X. For example, the Haitian compound tt-chv refers to a person who
is bald, as shown in (3).
(3)

Papa Ti Klod se
yon tt chv.
father Ti Klod it-is a
head
bald
The father of Ti Klod is bald.

haitian
(from Valdman et al. 1981)

Similarly, the Haitian compound tt-chaje refers to a person who is a problem.

(4)

Tifi sa
a
se
yon
girl dem det it-is a
This girl is a problem.

tt-chaje.
problem

haitian
(from Valdman et al. 1981)

A sample of compounds of this type is given in (5) together with the corresponding expressions in French and Fongbe. The Haitian data are from Valdman et al.
(1981) and Brousseau (1988a); the Fongbe data are from Segurola (1963),
Rassinoux (1987) and Brousseau (1988a).
(5)

haitian
a. tt-chv
(tte-chauve)
head-bald

french
chauve

b. tt-chaje
(tte-troubl)
head-troubled

personneproblme

c. je-pete
(il-pt)
eye-burst

aveugle

d. je-chch
(il-sec)
eye-dry

audacieux

e. je-f
(il-fort)
eye-strong

prtentieux nkn-kyn

f. tt-di
(tte-dure)
head-hard

fongbe
t-snsn

bald (person)

head-bald
t-gb

problematic (person)

head-troubled
nkn-tvn-nw3

blind (person)

eye-burst-att
hhn-wn

audacious (person)

clear-eye
pretentious (person)

eye-strong
entt
or tte-dure

t-my-sixn-tv
head-in-hard-ag

stubborn (person)

11.1 The semantics of compounds

337

As can be seen in (5), the Haitian compounds are made up of words which take
their phonological representations from the phonetic matrices of French words.
With the exception of the last example in (5), these particular French words are
not concatenated as compounds corresponding to the Haitian ones. Furthermore,
the concepts rendered by a compound in Haitian are rendered by a simplex in
French. As shown in (5), both the semantics of the Haitian words involved in the
compounds and the structure of these compounds follow the Fongbe pattern.
The fact that Haitian uses compounds where French has simplexes shows that
the Haitian compounds in (5) are not derivable from French. The fact that these
compounds follow the semantics and structure of the Fongbe equivalents shows
that the creators of Haitian used the principles of their own grammar in concatenating simplex nouns. Finally, the relexification hypothesis accounts once
again for the fact that the creators of Haitian did not identify the French simplexes
in (5): they did not have such lexical entries to relexify.
The examples in (6) all involve the word for head. The Haitian data are
from Valdman et al. (1981) and the Fongbe data from Segurola (1963) and
Rassinoux (1987).
(6)

haitian
a. tt-di
(tte-dure)
head-hard

french
enttement

b. tt-f-mal
(tte-faire-mal)
head-hurt

mal de tte

c. mare-tt
(amarrer-tte)
attach-head

fongbe
t-my-tt

stubbornness

head-in-stuffing
t-3my

headache

head-hurt
bandeau
or serre-tte

t-bl

headband

head-attach

The following example from Valdman et al. (1981) shows that the Haitian
compound in (6b) presents the characteristics of a noun, since it can be preceded
by yon a.
(7)

M gen yon tt-f-mal.

I
have a
headache
I have a headache.

haitian
(from Valdman et al. 1981)

Again, the structure of these compounds follows that of the Fongbe compounds
and not that of the corresponding French words. In (6c), however, the Haitian
compound could also be said to have the same structure as the corresponding
French compound. However, the semantics of the words involved follow the
substratum language, as is also the case in (6a and b).
Another set of very productive compounds involves nouns referring to names
of trees. In both Haitian and Fongbe, these productive compounds all involve the
word for plant and the name of a specific species. Examples are given in (8).

338
(8)

The concatenation of words into compounds

haitian
pye-bannan
pye-palmis
etc.

fongbe
kwkw-tn
d-tn

banana tree
palm tree
(from Brousseau 1989)

Abstracting away from word order, these compounds have the same structure in
both Haitian and Fongbe. By contrast, French uses words derived by means of
the affix -ier to refer to the same notions (e.g. banan-ier banana tree and palmier palm tree) or simplexes (e.g. chne oak). So, again, the data in (8) show
that the creators of Haitian used the principles of their own grammar in compounding. Again, the relexification hypothesis accounts for the fact that the
creators of Haitian did not adopt the French simplexes for tree names: they
simply did not have such lexical entries to relexify. Additional data from French
have to be considered in this case, however. Chaudenson (1990) claims that the
creators of the creole were probably exposed to French noun phrases such as un
pied de bananier, literally a plant of banana tree, or un pied de palmier a plant
of palm tree. Assuming that this was the case, then the Haitian compounds in
(8) may be argued to be superficially similar to the French noun phrases. Thus,
in this case, the superstratum and substratum data could be seen as being superficially similar and the Haitian facts could be claimed to be the way they are
because both source languages provided similar input. Such a scenario is certainly possible but, in my view, it is not the optimal one. Indeed, while the
Haitian and the Fongbe compounds in (8) are semantically equivalent in that
in both languages the compound refers to an X tree, the semantics of the
French expression designates a plant of X tree. On the basis of this semantic
fact, I consider the first scenario to best account for the data.
Another type of evidence that the structure of Haitian compounds follows the
pattern of the substratum grammar comes from the following data. Brousseau
(1989) documents the fact that, in both Haitian creole and Fongbe, there are many
compounds that refer to objects or parts of objects that are long and thin. These
compounds use the Haitian form bwa forest, wood, stick (Valdman et al. 1981)
and the Fongbe form tn tree, wood, stick (see Rassinoux 1987), respectively.
This is shown in (9).
(9)

a. bwa-bal
bwa-lint
bwa-pip

(bois-balais)
stick-broom
(bois-lunettes)
stick-glasses
(bois-pipe)
stick-pipe

b. wm-tn
paper-stick
w-tn
dough-stick
wyn-gn-tn sending-metal-stick

broomstick

haitian

shank of glasses
pipestem
pencil
fongbe
spatula
antenna
(=(43), (44) in Brousseau 1989)

The data in (9) show that the same semantic concepts that are compounded
in the Fongbe words are compounded in the Haitian ones as well. In both

11.2 Establishing word order in compounds

339

languages, a constant term, bwa in Haitian and tn in Fongbe, both meaning

stick, is part of the compounds. This contrasts with French, which does not use
one particular term in the corresponding compounds. For example, *bois de
balais is not licit in French, which has manche balais broomstick. Similarly,
*bois de lunettes and *bois de pipe are not possible French expressions either.
Instead, French has montant de lunettes and tuyau de pipe, respectively. Thus,
while the semantics of the Haitian compounds in (9a) does not follow the French
pattern, it does follow that of the Fongbe equivalents (see (9b) ).
Finally, Brousseau (1989, 1994a) points out that there are relatively few
Haitian compounds that derive their semantics from French compounds. She
remarks that, out of an inventory of over a hundred Haitian and three hundred
French compounds, there are only a dozen pairs with similar meanings. Examples
of such pairs are given in (10).
(10)

a. haitian N-n
zwazo-mouch
chwt-bwa
melon-dlo
coulv-dlo
poul-dlo
soulye-last

french N-n
oiseau-mouche
chouette de bois
melon deau
serpent deau
poule deau
soulier lacets

b. haitian N-n
pwason-zl
aran-sl
fanm-saj

french N-a/a-N
poisson ail (poisson volant)
hareng sal
sage-femme
(=(39) and

hummingbird
owl species
watermelon
watersnake
moorhen
lace-up shoe
flying fish
salt herring
midwife
(40) in Brousseau 1989)

The data presented throughout this section show that Haitian words are often
compounded in a way which follows the pattern of the substratum language.
This argues in favour of the claim that the creators of Haitian used the principles of their own grammar in compounding simplexes in the language that they
were creating. Furthermore, Haitian compounds often correspond to simplexes
in French. The fact that the creators of Haitian did not adopt these French simplexes follows from the relexification hypothesis: their native lexicons did not
provide them with any such lexical entries to relexify.
11.2

Establishing word order in compounds

In spite of the semantic similarity between Haitian and Fongbe compounds,

there is a striking difference between them. The order of elements in the Haitian
compounds is quite often opposite to that of the Fongbe compounds. For example, in (2), whereas Haitian has po-bouch (skin-mouth), Fongbe has n-fl
(mouth-skin). Similarly, in Haitian compounds involving a noun and an adjective, in some cases, the adjective follows the noun, as in tt-chv (head-bald)
bald (see (5) ), and in other cases, it precedes the noun, as in gwo-ng (bigman) rich, giant. The latter word order is not permitted in Fongbe, which
requires the order nounadjective in compounds, as in t-snsn (head-bald)

340

The concatenation of words into compounds

bald (see (5) ). Finally, in Haitian compounds involving a verb and a noun as
in mare-tt (attach-head) headband, the verb precedes the noun (see (9) ).
By contrast, in Fongbe, the noun precedes the verb as in t-bl (head-attach)
headband (see (9) ).
The above data show that Haitian and Fongbe have different directionality
parameters with respect to word order in compounds. It seems unlikely that
French provided the model for the word order in Haitian compounds for, as was
pointed out earlier, relatively few Haitian compounds derive their structure from
French compound words (see (10) ). Furthermore, in the last example in (10), the
word order in the Haitian compound is opposite to the order in the French one.
The fact that neither French compounds nor Fongbe compounds provided the
model for the word order in Haitian compounds raises the question of how this
order was actually established.
Brousseau (1989) argues that word order in compounds corresponds to word
order in syntactic phrases and, in the same way, the position of the head in
compounds corresponds to the position of the head in syntactic phrases. Brousseau
(1989) adopts an intrinsic definition of head as the constituent of which the word
is a hyponym. For example, in doghouse and greenhouse, house is the head of
the compound because a doghouse is a kind of house and so is a greenhouse.
In this view, in Haitian nounnoun compounds (see (2) ), the second noun is
the argument of the first, as is also the case in the syntax, and the first noun is the
head of the compound, as is also the case in the syntax. These relationships are
shown in (11), where the head of the construction is underlined.
(11)

a. nounnoun Compound
po-bouch
skin mouth
lip

b. noun noun in np
rob
Mari a
dress Mary det
Marys dress

haitian

In contrast to Haitian, in Fongbe nounnoun compounds (see (2) ), the first noun
is the argument of the second, following the basic order in syntactic phrases, and
the second noun is the head of the construction, as is also the case in noun
phrases (see Brousseau and Lumsden 1992). These relationships are shown in (12).
(12)

a. nounnoun Compound
n-fl
mouth skin
lip

b. noun noun in np
sb sn vv
v
Asiba of dress det
the dress of Asiba

fongbe

Likewise, the position of adjectives in compounds follows the pattern found

in noun phrases. So Haitian creole has adjective-noun compounds and adjectives
preceding the noun in noun phrases, as shown in (13).
(13)

a. adj.noun Compound
gwo-ng
big man
rich, giant

b. adj. noun in np
haitian
gwo vwa
ou
big voice 2nd
your big voice
(from Valdman et al. 1981)

11.2 Establishing word order in compounds

341

But Haitian also has nounadjective compounds and adjectives following the
noun in noun phrases, as illustrated in (14).
(14)

a. nounadj. Compound
dan-pouri
tooth-rotten
coward

b. noun adj. in np
yon pom pouri
a
apple rotten
a rotten apple

haitian

Unlike Haitian, Fongbe allows only one word order in noun phrases: adjectives
always follow the noun in this language. In compounds involving a noun and an
adjective, the adjective always follows the noun as well. These facts are illustrated in (15).
(15)

a. nounadj. Compound
gbn-kln
baggage small
package

b. noun adj. in np
x
kln
word small
short word

fongbe

(from Rassinoux 1987)

Finally, the word order observed in the Haitian verbnoun compounds replicates the verbcomplement order found in verb phrases, as is shown in (16).
(16)

a. verbnoun Compound
mare-tt
attach head
headband

b. verb noun in vp
mare li sere
tie
it tight
tie it tight

haitian

(from Vdrine 1992)

Likewise, the word order in the Fongbe nounverb compounds replicates the
complementverb order found in nominalised verb phrases in this language.
(17)

a. nounverb Compound
t-bl
head attach
headband

b. noun verb in nominalised vp fongbe

Kwk 3 lv bl
wy
Koku at this attach part
Koku is attaching this.

The three sets of facts discussed above argue in favour of Brousseaus (1989)
claim that the order of elements in compounds is determined by the order of
elements in syntactic phrases and that, similarly, the position of the head in
compounds (left or right) follows the position of the syntactic head in phrases.
In this view, then, the word order in Haitian compounds is opposite to that
in Fongbe compounds because these two languages have opposite word orders
in the syntax.
The fact that Haitian and Fongbe manifest opposite word orders in the syntax
is in line with the hypothesis of creole genesis outlined in chapter 2. According
to the hypothesis developed in Lefebvre and Lumsden (1992a, 1994a), the order
of major category lexical items in the creole is predicted to parallel the order of
such items in the superstratum language. The Haitian word order adjectivenoun
shown in (13b) follows the French word order in (18).

342
(18)

The concatenation of words into compounds

adj. noun in np
ta
grosse voix
your big
voice
your big voice

french

Similarly, the Haitian word order nounadjective shown in (14b) follows the
French pattern in (19).
(19)

noun adj. in np
une pomme pourrie
an apple
rotten
a rotten apple

french

The above data illustrate the more general fact that the word order of major
category lexical items in Haitian syntactic phrases follows the pattern of the
superstratum language. This is exactly what is predicted by our hypothesis.
In short, the word order in Haitian compounds follows the order in Haitian
syntactic phrases, which, in turn, follows the order in French syntactic phrases.
Thus, Haitian and French compounds have the position of the head in common,
since it is determined by the position of the syntactic head in phrases.

11.3

Types of compounds

In discussing the types of compounds in the three languages under comparison,

I will distinguish between synthetic and primary compounds. While the former
are all headed compounds, the latter may further be distinguished on the basis of
whether they are headed (endocentric) or headless (exocentric). Compounds may
be further distinguished on the basis of the types of syntactic categories (nouns,
verbs, adjectives, prepositions) that may be compounded in particular languages.
Table 11.1 provides an overview of the types of compounds found in Haitian,
Fongbe and French based on these distinctions. These types have been established based on the extensive survey of compounds in the three languages made
by Brousseau (1988a, 1989, 1994a). In Table 11.1, the head of the compound (as
discussed in section 11.2) is underlined.
To begin with, Haitian has left-headed compounds made up of two nouns.
These are illustrated in (20).
(20)

nn compounds
pye-bwa
(plant-wood)
bwat-myl
(box-bee)
zo-figi
(bone-face)
ng-lavil
(man-city)
pen-bwat
(bread-box)

haitian
tree
beehive
cheekbone
city dweller
square loaf
(from Brousseau 1988a)

11.3 Types of compounds

343

Table 11.1. Types of compounds in Haitian, Fongbe and French

haitian

fongbe

french

primary
headed
compounds

nn
an

nn
na

nn
an
npn
npv
na

primary
headless
compounds

an
na

na

synthetic
headed
compounds

an
na
vn
n de n
nn

nvn
nvv
nvtV

According to Brousseau (1989, 1994a), the vast majority of Haitian compounds

are of the nn type. Examples of right-headed compounds made up of an adjective
and a noun are provided in (21).
(21)

an compounds
gwo-pus
gran-mounn

haitian
(big-thumb)
(big-person)

thumb
grownup
(from Brousseau 1988a)

Haitian also has headless compounds made up of a noun and an adjective, as

shown in (22).
(22)

Headless compounds

haitian

a. an
gwo-je
gran-gozye

(big-eye)
(big-throat)

greedy
pelican

b. na
je-f
je-pete

(eye-strong)
(eye-burst)

pretentious
blind person
(=(18) in Brousseau 1994a)

Based on Brousseaus survey, these four types exhaust the list of productive
compounds in Haitian.
As can be seen in Table 11.1, French has a much larger inventory of types of
compounds than Haitian. First of all, it has headed compounds. Left-headed
compounds may involve nouns, verbs, adjectives and prepositions. These are
shown in (23).

344
(23)

The concatenation of words into compounds

Left-headed compound

french

a. nn
tlvision-couleur
pause-caf
fermeture-clair
assurance-vie

colour television
coffee break
zipper
life insurance

b. npn
livre dimages
cur de pomme
aide de camp
poupe de cire
canne sucre
pingle linge

picture book
apple core
aide de camp
wax doll
sugar cane
clothes pin

c. npv
pole frire
machine laver
chambre coucher
mtier tisser

frying pan
washing machine
bedroom
loom

d. n a
pouvoir tudiant
arbre fruitier
parc national
maison mobile

student power
fruit tree
national park
mobile home
(from Brousseau 1988a)

Right-headed compounds are made up of an adjective and a noun, as shown

in (24).
(24)

an compounds
basse-cour
court-bouillon
demi-lune

french
farmyard
court-bouillon, poaching liquid
half moon
(from Brousseau 1988a)

Secondly, French has several types of headless compounds, illustrated in (25).

Brousseau (1988a) remarks that the vn compounds in (25a) are extremely
productive in French.
(25)

Headless compounds

french

a. vn
pse-personne
abat-jour
coupe-papier
grille-pain

scale
lampshade
paper cutter
toaster

b. n de n
pied de biche
il de buf

nail-claw, crowbar, etc.

bulls eye window

c. an
rouge-gorge
blanc-bec

robin
greenhorn

11.3 Types of compounds

d. na
blouson-noir
cordon-bleu
sang-ml

345

hoodlum
cordon-bleu
halfbreed
(from Brousseau 1988a)

French also has appositional compounds of the type in (26), which, according to
Brousseau (1989), are very productive.
(26)

nn
auteur-compositeur
moissonneuse-batteuse
fille-mre
professeur-chercheur

french
author-composer
combine-harvester
unmarried mother
professor-researcher
(from Brousseau 1988a)

According to Brousseaus survey of French compounds, the data discussed above

exhaust the types of productive compounds in the language.
A striking similarity between Haitian and French is that neither has synthetic
compounds. But, aside from the fact that they both have right-headed compounds made up of an adjective and a noun (see (21) and (24) ), the types of
compounds in the two languages are rather different. First, while Haitian headed
compounds may involve only nouns and adjectives (see (20) and (21) ), French
headed compounds may involve nouns, verbs, adjectives and prepositions (see
(23) ). Second, as is pointed out in Brousseau (1994a), while headless compounds exist in both Haitian (see (22) ) and French (see (25) ), compounds of this
type do not show any semantic correspondences in the two languages. That is,
Haitian compounds of the type in (22) do not translate as compounds in French
and French compounds of the type in (25) do not translate as compounds in
Haitian. Third, while exocentric vn compounds of the type in (25a) are very
productive in French, this type of compound has not been attested in Haitian
(see Brousseau 1994a). Furthermore, appositional compounds (see (26) ) are also
very productive in French, whereas Haitian presents no such compounds.6 Thus,
although Haitian and French compounds have in common the position of the
head (as discussed above), they differ with respect to a number of other features.
In view of this situation, let us now turn to the types of compounds in Fongbe.
Fongbe has headed binominal (i.e. nounnoun) compounds, as shown in (27).
Brousseau (1989) points out that this type of compound is very productive in
Fongbe.
(27)

nn compounds
t-gn
(country-chief)
gl-gn
(field-chief)
kwgb-gn
(bank-chief)

fongbe
president
agricultural agent
banker
(from Brousseau 1988a)

It also has headed compounds made up of a noun and an adjective, as shown

in (28).

346
(28)

The concatenation of words into compounds

na compounds
zwn-vw
(sickness-red)
tv-gb
(father-big)
gn-vw
(iron-red)

fongbe
leprosy
grandparent, ancestor
copper
(from Brousseau 1988a)

Fongbe also has headless compounds made up of a noun and an adjective. This
is shown in (29).
(29)

na
3-vwvw
xm-hnhn
w-kk

fongbe
(head-red)
(belly-open)
(body-dead)

redhead
joy, pleasure
paralysis
(from Brousseau 1988a)

Considering the types of Fongbe compounds discussed so far, and abstracting

away from the position of the head, they compare with Haitian compounds as
follows. In both Haitian and Fongbe, headed binominals constitute the most
productive type of compounds. Both languages have headed compounds made
up of a noun and an adjective (see (21) and (28) ). Furthermore, like Haitian (see
(22) ), Fongbe has headless compounds made up of a noun and an adjective (see
(29) ). The types of compounds in Haitian and Fongbe are thus quite similar.
In spite of these similarities, however, there are differences between the two
languages with respect to synthetic compounds. Synthetic compounds are characterised by the fact that they involve a deverbal head and their semantic interpretation is derived from the argument structure of this head (see Selkirk 1982).
For example, in English compounds such as in (30), the non-head element is
interpreted as an argument (Theme, Goal or Source) of the head.
(30)

a. snow removal

(Theme)

english

b. church goer

(Goal)

english

c. country born

(Source)

english

Fongbe presents three types of synthetic compounds, all of which have a

verbal constituent. These compounds are the right-headed nvn compounds
(see (31a) ), nvv (see (31b) ), and nvtV (see (31c) ) compounds.7
(31)

a. sn-h-n
n-tv-n

(water-spread-thing) watering can

(thing-sew-thing)
needle

fongbe

b. gb-dd
v-jj

(bomb-sending)
(child-giving-birth)

bombing
delivery

fongbe

c. gl-ly-tv
n-s-tv

(field-grow-agent)
(thing-sell-agent)

farmer
seller

fongbe
(=(24) in Brousseau 1994a)

As Table 11.1 shows, Haitian does not have synthetic compounds of this type.
Brousseau (1989, 1994a) shows that synthetic compounds like those in (31b and
c) above correspond to derived or simplex words in Haitian creole. For example,
the Haitian words corresponding to those in (31b) are nouns derived from verbs

11.3 Types of compounds

347

either by means of the nominalising suffix -ay or by the process of morphological conversion, as shown in (32a). Furthermore, agentive nouns of the type
nvtV (see (31c) ) are rendered in Haitian by nouns derived with the suffix -,
as shown in (32b).
(32)

fongbe
a. sn-knkxn
y-jjx

haitian
pari-ay
tonbe-

bet
fall

b. hn-jtv
x-3vtv

chant-
pal-

singer
speaker
(=(25) in Brousseau 1994a)

Why should Haitian lack synthetic compounds when this type of compound
is so productive in one of its substratum languages (and presumably in others as
well)? Since, throughout this chapter, French has been shown to have contributed little to the semantics and structure of Haitian compounds, I dismiss the
possibility that the lack of synthetic compounds in French had anything to do
with the situation in Haitian. Furthermore, data from Saramaccan show us that
a lack of synthetic compounds cannot be considered to be a characteristic of
creoles, since Saramaccan has synthetic compounds of the type we find in Fongbe.
For example, dokifisima (divefishman) fisherman (from Smith and Veenstra
1994) has the same structure as the Fongbe synthetic compounds in (31c).8 In
view of this situation, could the lack of synthetic compounds in Haitian be
linked to other properties of the creole which contrast with Fongbe?
Brousseau (1989) proposes to link the lack of synthetic compounds in Haitian
to broader differences between Haitian and Fongbe pertaining to the properties of
a subclass of verbs. First, Fongbe requires that the direct object be obligatorily
realised even when it has a generic interpretation (see chapter 9, section 9.10).
This constraint holds in both morphological and syntactic structures, as illustrated
in (33).
(33)

a.
3 *(n)
you eat thing
You ate.

b. *(n)-3u-xws
fongbe
thing-eat-room
restaurant
(=(26) in Brousseau 1994a)

In contrast to Fongbe, in Haitian creole, there is no such constraint (see chapter 9). That is, generic objects need not be realised. This property holds true of
both morphological and syntactic structures, as shown in (34).
(34)

a. Ou te
manje.
you past eat
You ate.

b. sal-manje
haitian
room-eat
dining room
(=(27) in Brousseau 1994a)

Second, while Fongbe has many inherent object verbs, the Haitian counterparts of these verbs are simplexes (see chapter 9, section 9.10). The data in (35)
illustrate these correspondences.

348
(35)

The concatenation of words into compounds

fongbe
j hn
dv mlw
zn zwnln

(produce-song)
(do-sleep)
(walk-walk)

haitian
chante sing
dmi
sleep
mache walk
(=(28) in Brousseau 1994a)

In Brousseaus account, these two differences between Fongbe and Haitian allow
for a straightforward account of the lack of synthetic compounds of the type in
(31b) and (31c) in Haitian. Further, Brousseau (1994a: 22) claims that the Haitian
data are predictable. Indeed, if the direct object is not realised in a Fongbe structure such as nvv in (32a), the resulting structure is a non-compound word: a
deverbal noun formed by reduplication. What we find in Haitian creole is a
deverbal noun derived either by the suffix -ay or by the null affix discussed in
chapter 10. Likewise, if the direct object is not realised in Fongbe structures of
the type nvtV in (32b), the resulting structure is that of a deverbal agentive.
What we find in Haitian is a deverbal agentive noun derived with the suffix -.
This attractive proposal would find independent support if, in Saramaccan, the
subclass of verbs discussed above could be shown to follow the Fongbe pattern
rather than the Haitian one.9
In short, the types of compounds found in Haitian parallel those found in
Fongbe and contrast with those of French. There is one exception to this generalisation: Haitian lacks synthetic compounds, which appear to be very productive
in the substratum grammar considered here. This difference is probably linked,
however, to a difference in the overt realisation of arguments with a specific
subclass of verbs.
11.4

Conclusion

The data and analyses presented in this chapter reveal that Haitian compounds
have the following properties. First, concepts that are rendered as compounds in
Fongbe are often also rendered as compounds in Haitian where French has
simplexes. Furthermore, in Haitian, simplexes are compounded following the
semantics of the substratum language rather than the superstratum language.
These facts argue for the claim that the creators of the creole used the principles
of their own grammar in concatenating simplexes. The relexification hypothesis
accounts for the fact that the French simplexes corresponding to compounds in
the substratum grammar did not enter the creole: there were no such lexical
entries to be relexified. It has also been shown that the order of constituents in
compounds follows the order of constituents in syntactic phrases which, in the
creole, reproduce the word order of the superstratum language. This fact was
shown to be consistent with the general hypothesis outlined in chapter 2. Finally,
with the exception of synthetic compounds, the types of compounds found in
Haitian were shown to parallel the types in Fongbe rather than in French.

12 Parameters
Parameters constitute options provided by Universal Grammar (ug). These options
are sometimes formulated in terms of binary choices having the value +/. For
example, does a particular grammar have verb raising or not? Does a particular
grammar require overt subjects or not? These options can also be formulated in
terms of the values a variable may take. For example, assuming the rule Move
to account for the movement of constituents, what are the values for in a
particular grammar? These options may also be formulated in terms of correlations, for example, property (b) will be available in a given grammar if property
(a) is. As we will see throughout this chapter, parameters account for the relationship between a number of properties which, at first glance, would appear to
be unrelated.1
In the theory advocated in this book, it is hypothesised that, in creole genesis,
the creators of a creole do not have sufficient access to the superstratum language to acquire its parametric values. By hypothesis, the creators of the creole
use the parametric values of their own grammar in assigning a value to the
parameters of the language that they are creating. This hypothesis predicts that,
where the parametric values of the substratum and superstratum differ, the creole
should have the same value as the substratum languages. With one exception,
this prediction is borne out in Haitian creole, as we will see below. The following parameters will be discussed: availability of null subjects, verb raising, the
serial verb construction and the double-object construction, interpretation of
negative quantifiers and availability of verb-doubling phenomena. Theoretical
accounts of the parameters discussed in this chapter have not been worked out in
the same amount of detail within the framework of the theory of parameters. The
discussion below therefore addresses these issues on the basis of available accounts
in the literature. Finally, as before, the discussion is based on a comparison of
Haitian, French and Fongbe (and other Gbe and Kwa languages). The Bantu
languages have parametric values which are sometimes quite different from
those of the Kwa languages. This problem will be addressed in chapter 13.
12.1

Is Haitian a null subject language?

One of the parametric options of ug relates to whether null subjects are available
in particular languages. In point of fact, this parameter represents the remains of
the former pro-drop parameter expressed in Chomsky (1981) (see e.g. Bennis
349

350

Parameters

1982; Safir 1982; Hulk 1986; Law 1992). In recent literature, it has been proposed that languages with syntactic clitics should be considered null subject
languages (see e.g. Jaeggli 1984; Hulk 1986; Roberge 1990). The theory goes
as follows: subject clitics are not generated in np positions but in a functional
category projection (infl(ection) or agr(eement) ) as the spelling-out of person,
number, gender and case features. In languages which have syntactic clitics, the
subject position is thus filled by a small pro, as in Italian (see e.g. Rizzi 1986b),
bound by the clitic. In this view, both French and Fongbe would be null subject
languages, since, as was argued in chapter 6, both languages have syntactic
clitics.2
In the recent literature on Haitian creole, there has been some debate as to
whether Haitian is a null subject language (see DeGraff 1992a, 1992b, 1992d,
1993b, 1996; Deprez 1992a; Law 1992). This debate crucially depends on whether
Haitian has null subjects of the type we find, for example, in Italian. DeGraff
claims that there are empty subjects in Haitian. His proposal is essentially
based on two sets of facts. First, as we saw in chapter 6, Haitian allows for null
expletive subjects. Second, DeGraff provides examples of long distance Whmovement, as in (1), in which the position of the extracted subject is only
optionally filled by the resumptive pronoun ki discussed in chapter 7.
(1)

Ki
mounn ou
regrt (ki) kraze vaz
which person you regret res break vase
Who do you regret has broken the vase?

la?
det

haitian

(=(49) in DeGraff 1996)

On the basis of these facts, DeGraff claims that Haitian clitics are syntactic
clitics spelling out the agreement features of the inflection phrase and identifying pro in subject position. He therefore concludes that Haitian is a null subject
language.
Deprez (1992a) argues against this position on the following grounds. First,
as we saw in chapter 6, an overt expletive is also available in the language and,
in some cases, it is required (see chapter 9). Second, in contrast to null subject
languages like Italian, referential pronouns are obligatory in Haitian creole (a
fact also acknowledged by DeGraff 1992d). Furthermore, Deprez (1992a: 26)
points out the directionality of the correlation between expletive drop and empty
argumental subjects:
Despite permitting empty non-argumental subjects, German, Yiddish and
Icelandic are not regarded as pro-drop languages. In recent years, it has been
recognised that although the existence of empty argumental subjects in a language L implies the existence of empty non-argumental subjects, the reverse is
not true. That is, argument pro-drop implies expletive drop, but expletive drop
does not imply argument drop.

Third, with respect to the data in (1), Deprez (1992a) points out that there
appears to be some variation among speakers in this area of the grammar. For

12.2 Verb raising

351

example, Koopman (1982b), Sterlin (1988, 1989), Deprez (1992a) and Law
(1994b) all insist that the Haitian speakers they worked with require the presence of ki in contexts like (1). On the basis of this fact, Deprez (1992a: 36)
concludes that the syntax of subject extractions thus casts further doubt on the
validity of an analysis of Haitian creole as a pro-drop language. Finally, Deprez
(1992a) argues that Haitian clitics are not syntactic but phonological clitics. The
arguments supporting this analysis were presented in chapter 6. Like Deprez
(1992a), I also conclude that Haitian is not a null subject language.
Assuming the analysis whereby the availability of syntactic clitics defines a
language as a null subject language, we have to conclude that Haitian differs
from both of its source languages. While both French and Fongbe have a positive value for the null subject parameter, Haitian has a negative value. Thus, in
this case, it appears that the creators of Haitian had to reset the value of the
original parameter. This situation is a consequence of the fact that they had
abandoned the syntactic clitics of their original lexicon, as we saw in chapter 6.
Presumably, the first generation of Haitian native speakers assigned this parameter
a negative value since they were exposed to strong subject pronouns. In terms
of the markedness issue, this is extremely interesting. On the basis of work by
Hyams (1986, 1987), DeGraff (1992a) points out that availability of null subjects is the unmarked option of ug. Assuming that this is correct, while both the
substratum and the superstratum languages of Haitian present the unmarked
option for this parameter, Haitian exemplifies the marked one. This is a major
drawback for theories advocating that creole languages systematically present
the unmarked parametric options of ug (see e.g. Bickerton 1984).
12.2

Verb raising

In recent literature, it has been proposed that languages vary based on whether
they allow verb raising (see Chomsky 1981; Pollock 1989, and related literature).
On the basis of facts involving, among other phenomena, negative placement,
question formation and adverb placement, Pollock (1989) argues that, while
French has verb raising, English does not. For example, while the verb may
not precede the negation marker not in English, the tensed verb in French must
precede the negation marker pas not, as is shown in (2).
(2)

a. *John likes not

Mary

b. Jean (n) aime pas Marie.

John neg like not Mary
John does not like Mary.

english
(=(2a) in Pollock 1989)
french
(=(2b) in Pollock 1989)

Similarly, while the verb cannot precede the subject in English questions, it may
do so in French, as shown in (3).

352
(3)

Parameters
a. *Likes he

Mary

english
(=(3a) in Pollock 1989)

b. Aime-t-il Marie?
like he Mary
Does he like Mary?

french
(=(3b) in Pollock 1989)

Likewise, while the verb in English cannot precede a vp adverb, in French it

must precede such an adverb, as shown in (4).
(4)

a. *John kisses often Mary

b. Jean embrasse souvent
John kiss
often
John often kisses Mary.

english
(=(4a) in Pollock 1989)
Marie.
Mary

french
(=(4b) in Pollock 1989)

On Pollocks (1989) analysis, this cluster of differential properties between

English and French can be accounted for by a parametric difference between
the two languages, namely, whether the language allows verb raising. In French,
the verb must raise through agr to Tense, as shown in (5), and even to head of
cp, as in (3b).
tp

(5)

np

NegP

AgrP

Neg

vp

Agr

Adv
v

np

This accounts for the word order in (2b), (3b) and (4b). By contrast, verb raising
is not available in English, and hence the verb stays in its basic position within
the vp, which explains the ungrammaticality of the sentences in (2a), (3a) and
(4a). Pollocks analysis also captures the relationship between the availability of
verb raising and rich verbal morphology in a given language: only languages
with rich verbal morphology, such as French, have verb raising. It is argued that
this inflectional morphology is picked up by the verb as it moves through agr to
Tense, as depicted in (5).

12.2 Verb raising

353

How do Haitian and Fongbe compare with French with respect to this parametric option? First, as we saw in chapter 5, unlike French, neither Haitian nor
Fongbe has inflectional morphology. Second, as has been pointed out in the
literature, the verb never occurs before the negation marker in Haitian (see (6) )
or in Fongbe (see (7) ) (see also chapter 7).
(6)

a.

Boukint pa renmen Bouki.

Boukinet not like
Bouki
Boukinet does not like Bouki.

haitian
(=(33) in DeGraff 1994)

b. *Boukinet renmeni pa
(7)

a.

ti Bouki

Kwk m nywn n sb.

Koku not like
to Asiba
Koku does not like Asiba.

b. *Kwk

nywni m

ti n

sb

haitian
fongbe

fongbe

Third, the verb can never precede the subject in either Haitian or Fongbe (see (8) ).
(8)

a. *renmeni Mari ti Jan

b. *nywni Kwk ti sb

haitian
fongbe

Fourth, the verb follows temporal adverbs in both languages. This is illustrated
in (9) with the adverb meaning already.
(9)

a.

Li

s/he
S/he

deja
pase rad
yo.
k
l
vv
lx.
already iron cloth pl
already ironed the clothes.

b. *Li pasei deja

* l
k

ti rad yo
ti vv lx

haitian
fongbe

haitian
fongbe

The four properties discussed above show that verb raising is not available
in Haitian or Fongbe, a conclusion which is in line with DeGraff (1992b) and
Avolonto (1992). This contrasts with French, which, as argued above, is a verb
raising language par excellence. Thus, for this parameter, Haitian has the same
value as the Gbe languages, and the Kwa languages more generally (see Givn
1971; Baker 1991).
This conclusion has been challenged by DeGraff (1994) on the basis of
examples such as those in (10). DeGraff (1994) shows that the adverbs meaning well in Haitian and Fongbe do not occur in the same surface position. In
Haitian this adverb precedes the vp, whereas in Fongbe it follows the vp.
(10)

a. Jan
byen manje.
John well ate
John ate well.

haitian

b. Kwk 3 nl
gn5.
Koku ate thing well
Koku ate well.

fongbe

354

Parameters

On the basis of this word order discrepancy, DeGraff (1994: 12) concludes
that it seems unwarranted to try to explain the verb-placement facts in Haitian
by evoking substratum influence. In my view, the facts in (10) do not tell
us anything about verb raising; if they did, then in order to explain the word
order in (10b), we would have to say that the whole vp (the verb and its
direct object) has moved over the adverb. On theoretical grounds, this is not
a welcome analysis. Furthermore, even if it were correct, it would involve vp,
not v, raising.
The contrast between the data in (9), on the one hand, and those in (10),
on the other, has something to tell us about the position of different classes of
adverbs. In both Fongbe and Haitian, temporal adverbs such as already occur
before the vp, as in (9). Fongbe manner adverbs, however, systematically occur
after the vp, as in (10b). This contrasts with Haitian, where some adverbs such
as byen well occur before the vp, as in (10a), and others, such as souvan
often, occur after the vp, as in (11).
(11)

Jan
manje pen
souvan.
John eat
bread often
John eats bread often.

haitian

The discrepancy between Haitian and Fongbe adverb placement follows from
the hypothesis outlined in chapter 2, whereby the word order of lexical categories (major syntactic categories) in creole genesis is established on the basis of
the word order of the superstratum language rather than that of the substratum
languages. The position of adverbs in Haitian illustrates this general claim. In
(9a), the Haitian adverb deja already precedes the vp. This follows the French
order where dj already also precedes the vp. This can be seen in sentences
such as Marie a dj repass les vtements Mary has already ironed the clothes
where the auxiliary prevents the lexical verb from raising. Similarly, the position
of the Haitian adverb byen well occupies the same position as the French
adverb bien well as in Marie a bien mang Mary ate well. While the position
of Haitian deja also coincides with the position of Fongbe k already, the
position of byen does not coincide with that of Fongbe gn% well but is the
same as that of the French adverb bien, which occurs before the vp. Thus, data
concerning the position of adverbs in Haitian support the hypothesis that, in
creole genesis, the word order of lexical categories is established on the basis of
the superstratum rather than the substratum languages. I therefore conclude that
the Fongbe facts in (10b) do not constitute a counterexample to the analysis that
Fongbe is not a verb raising language, contrary to DeGraffs (1994) overhasty
conclusion.
Thus, I claim that the value of the verb raising parameter in Haitian is different from that in French and the same as that in Fongbe. This conclusion is in line
with the hypothesis that the creators of Haitian did not have sufficient access to
French to reset the parametric value of their own grammar at the time the creole was being created and therefore used their own value. Based on the primary

12.3 Serial verbs

355

data that they were exposed to, the first generation of Haitian native speakers
identified the properties of in and agr in the language they were exposed to.
On the basis of these properties, they deduced that verb raising is not available
in that language. This is a probable scenario since this is what we see in modern
Haitian.
12.3

Serial verbs

In recent literature, several parameters/correlations have been proposed to

account for the availability of serial verbs in particular languages. Like several
West African languages, Haitian creole has serial verbs, as shown in (12).
(12)

Li
3rd
He

sv
svn y
x
my.
pran crab ale nan mache.
take crab go in
market in
brought the crab to the market.

fongbe
haitian
(=(8) in Lefebvre 1986)

In contrast to Haitian and Fongbe, French does not have this construction.
(13)

*Il prend crabe va au march

french

In this section, I will review some of the many parameters/correlations3 proposed to account for the availability of this construction in light of the Haitian,
French and Fongbe data.
12.3.1

Serial verb languages lack Ps

One proposal is that there is a correlation between the availability of the serial
verb construction in particular languages and the lack of lexical items of the
category P(reposition) (see e.g. Bickerton 1981; Byrne 1987). In this view, French
lacks serial verbs because it has prepositions; Haitian and Fongbe have serial
verbs because they lack lexical items of the category p.
This correlation correctly predicts the French data. French has a large class of
prepositions and it does not have serial verbs. It is incorrect on the Haitian and
Fongbe data, however, for both languages have lexical items of the category p
and serial verbs. Gilles (1988) documents the class of Haitian prepositions in
detail. In Lefebvre (1990a), it is extensively argued that, along with postpositions,
Fongbe has a class of prepositions which are of the category [n,v] and which
have syntactic properties distinguishing them from serial verbs. Crucially, while
the object of the second verb of a series may be extracted, the object of a preposition cannot. Work by Jansen et al. (1978) reports on similar data drawn from
other creole languages. I therefore conclude that this supposed correlation is not
borne out by the data.

356
12.3.2

Parameters
Verb serialisation and lack of derivational verbal morphology

Some researchers have posited a correlation between the availability of verb

serialisation in particular grammars and the lack of derivational verbal morphology. This correlation is best described by Baker (1991: 79): Many people have
observed that notions which are expressed by Serial Verb Constructions . . . in
the Kwa languages of West Africa correspond to a large degree to those which
are expressed by derivational verb morphology in the Bantu languages of East
Africa.4 Muysken (1988d) points out that this correlation certainly holds true
for verbs in the Caribbean creoles and in West African languages, which are
largely mono-morphemic. Data from the three languages under comparison in
this book also support this claim. On the one hand, as we saw in chapter 10,
French has many derivational affixes which modify the meaning of base verbs.
For example, the base verb porter to carry is part of the derived verb ap-porter
to bring. As shown in (12), the latter concept is expressed by a serial verb
construction in both Haitian and Fongbe, which lack derivational affixes of the
type we find in French (see chapter 10). Hence, this correlation is supported by
the three languages being examined here and would account for the fact that,
while French has no serial verbs, Haitian and Fongbe share this construction.
The correlation can be further extended to the lack of inflectional verbal morphology, as we will see below.
12.3.3

Serial verbs and lack of inflectional morphology

In recent literature, it has been proposed that there is a correlation between

the availability of verb serialisation and the lack of inflectional morphology
and hence absence of verb raising to in (see section 12.2). To the best of my
knowledge, this correlation was first proposed by Muysken (1988d) and tentatively stated as follows: In serialising languages, v is separate from in and
therefore vp can function as a secondary predicate.5 On the basis of data drawn
from Yoruba, Dchaine (1993) further explores this proposal. She shows that,
while Yoruba is extremely rich in serial verb constructions, it is extremely poor
in inflectional morphology. The language has no inflectional affixes whatsoever,
and tense and aspect are encoded by means of independent preverbal markers or
not at all. Dchaine observes that this situation contrasts with Igbo, which lacks
serial verb constructions of the type found in Yoruba. Igbo has exceptionally
rich inflectional morphology where every verb must bear an inflectional suffix.
This correlation between the absence of verb raising to in and the availability
of serial verbs in a particular language is further explored in Baker and Stewart
(1996) on the basis of a sample of languages spoken in various areas of the
world.
This correlation accounts for the differential properties of the three languages under comparison in this book. French has inflectional morphology
(see chapter 5) that the verb picks up in raising to in (see section 12.2). By

12.4 The double-object construction

357

contrast, Haitian and Fongbe do not have inflectional morphology (see chapter 5) and verb raising is not available in these languages (see section 12.2).
In French, there are no serial verbs, whereas in Haitian and Fongbe this option
is available.6
12.3.4

Setting the value of the parameter in creole genesis

According to the view advocated in this book, 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 in and hence verb serialisation 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 serialisation in the grammar.
This is a likely hypothesis since it accounts for what we see in modern Haitian.
12.4

The double-object construction

Recall from chapter 9 that John sent Mary a letter is an example of the Recipient
Theme construction (np np), henceforth the double-object construction, whereas
John sent a letter to Mary is an example of the ThemeGoal construction (np
pp). As we saw in chapter 9, unlike French, both Haitian and West African languages have this construction. In this section, I review the various parameters/
correlations proposed in the literature to account for the availability of the construction in particular grammars in light of the Haitian, French and Fongbe data.
The following discussion draws on an earlier one in Lefebvre (1994c) based on
data from Fongbe. As will be shown throughout this section, the Haitian facts
are parallel to Fongbe and they systematically contrast with French.
12.4.1

Availability of the double-object construction and

preposition stranding

The general proposal advocated in Kayne (1984) is that the availability of the
double-object construction in a given language correlates with the availability
of preposition stranding in that language. For example, in English, preposition
stranding is available (e.g. Who did John send a letter to?) and so is the doubleobject construction (e.g. John sent Mary a letter). In Kaynes view, the two
constructions have in common the Case-assigning properties of verbs and prepositions. In English, the preposition to, for example, has the property of assigning Objective Case, the Case normally assigned by a verb to its direct object.
In Kaynes analysis, French lacks double-object constructions (e.g. *Jean a
envoy Marie une lettre) because it lacks preposition stranding (e.g. *Qui Jean a
envoy une lettre ?). While the French verb envoyer to send assigns Objective
Case to its direct object, the preposition to assigns Dative Case to its object.

358

Parameters

Hence, French does not have the double-object construction for the same reason
that it lacks preposition stranding.
The correlation proposed by Kayne is not borne out by the Fongbe and
Haitian data. While the double-object construction is available in these languages, preposition stranding is not. (See also Zhang 1990, for similar data from
Chinese.) In (14b), the Fongbe preposition n which translates as to/for has
been stranded and the resulting sentence is not grammatical.7
(14)

a.

Kwk s
svn n
sb.
Koku sell crab to / for Asiba
Koku sold crab to/for Asiba.

fongbe
(=(5a) in Lefebvre 1994c)

svn n ti
fongbe
b. *M6 / sbi wy Kwk s
who / Asiba it-is Koku sell crab to / for
It is who/Asiba that Koku sold crab to/for?
(=(5b) in Lefebvre 1994c)

The situation is similar in Haitian. In (15b), the preposition pou which translates
as to/for has been stranded and the resulting sentence is not grammatical.
(15)

a.

Jan
vann krab pou Mari.
John sell
crab for Mary
John sold crab for Mary.

vann krab
b. *Se
ki
mounni Jan
It-is which person John sell
crab
Who is it that John sold crab for?

haitian

pou
for

ti

haitian

Presumably, in Fongbe and Haitian, both the np governed by the preposition

and the np governed by a verb are assigned structural Accusative Case.8 The
assumption that both the verb and the preposition structurally assign Accusative
Case to their object creates a situation which is much like English (in contrast to
French). Kaynes proposal predicts that Fongbe and Haitian should allow preposition stranding as well as double-object constructions, but they do not.
12.4.2

Availability of the double-object construction and directionality of

theta-role assignment

Tremblay (1991) assumes that the relationship between the Theme and the
Goal of the double-object construction parallels the relationship between the
possessed and the possessor in a noun phrase. In her view, the directionality of
theta-role assignment within the noun phrase determines whether the doubleobject construction will be available in a particular language. In English, the
possessor in a Genitive construction is assigned its theta-role to the left of the
head of the noun phrase, e.g. Johns book, and the double-object construction is
available. In French, the possessor is assigned its theta-role to the right of the
noun phrase, e.g. le livre /de Jean, and the double-object construction is not
available.

12.4 The double-object construction

359

Both the Fongbe and Haitian data constitute a counterexample to this claim.
As was shown in chapter 4, the Genitive phrase follows the head of the noun
phrase in both languages and, hence, the possessor is assigned its theta-role to
the right of the head noun. This creates a situation which is much like French,
and thus Tremblays proposal predicts incorrectly that the double-object construction should not be available in either Haitian or Fongbe, which is not borne
out by the data.
12.4.3

Availability of the double-object construction and of Genitive Case in

nominal structures

Johnson (1991) proposes a direct correlation between the availability of the

double-object construction and the availability of structural Genitive Case (e.g.
s in English) in nominal structures. According to this proposal, the doubleobject construction, as in (16a), is available in a given grammar because structural Genitive Case, exhibited in (16b), is also available in the nominal structure
of that grammar.
(16)

a. John gave Mary a book.

b. Marys book

english

The motivation for Johnsons proposal is the claim that the two nps involved in
the double-object construction are in a possession relationship which parallels
the relationship observed in nominal structures between the possessed and the
possessor marked for Genitive Case. This would mean that the double-object
construction is not available in French because Genitive Case is not available in
French nominal structures. This correlation is borne out by both Haitian and
Fongbe data.
As we saw in chapter 4, in both Fongbe and Haitian nominal structures,
Genitive Case is available. The correlation proposed in Johnson (1991) between
the availability of the double-object construction and the availability of structural
Genitive Case in nominal structures is thus supported by these data. Furthermore,
the relationship between the availability of the double-object construction and
the availability of Genitive Case in nominal structures was shown to be pertinent
to the syntactic account of the double-object construction in chapter 9.
Out of the correlations proposed in the literature to account for the availability of the double-object construction in particular languages, we find that the
Haitian and Fongbe data support Johnsons (1991) proposal that the availability
of this construction is linked to the availability of the Genitive Case.
12.4.4

Setting the value of the parameter in creole genesis

The creators of the creole used their knowledge of their own grammars in setting
the value of the parametric option which allows for double-object constructions.
As was argued in chapter 4, they had a Genitive construction which they reproduced in the creole. This allowed them to have the double-object construction,

360

Parameters

which they also reproduced in the creole (as was argued in chapter 9). The first
generation of Haitian native speakers identified the Genitive Case in the nominal
structure of the language they were presented with. On the basis of this property,
they deduced the availability of the double-object construction in that language.
This is a likely scenario, since modern Haitian has these constructions.
12.5

The interpretation of negative quantifiers

Haitian negative quantifiers derive their phonological form from French phonetic sequences; for example, psonn is phonologically derived from French
personne nobody, and anyen from French rien nothing. As is argued below,
however, the properties of these quantifiers are not derived from the corresponding French forms.
DeGraff (1993a: 67) points out that negative quantifiers interact differently
with Haitian pa than with French pas. Compare (17) and (18), where the truth
condition in (17) is opposite to that of its Haitian counterpart in (18).
(17)

a. Personne nest pas venu.

french
nobody
ne+is pas come
Everybody came. [Lit.: Nobody has not come.]
(=(3a) in DeGraff 1993a)
b. Ce nest pas rien.
3sg ne+is pas nothing
This is something. [Lit.: This is not nothing.]

french
(=(3c) in DeGraff 1993a)

(18)

a. Psonn pa vini.
nobody pa come
Nobody has come.

b. Sa pa anyen.
haitian
3sg pa nothing
This is nothing.
(=(4a, c) in DeGraff 1993a)

DeGraff (1993a: 67) observes that, in French, co-occurring negative elements

cancel each other, giving rise to a net positive statement. This contrasts with
Haitian where the two negative elements are immediately construed as net
negative statements. Deprez (in press) further points out that, in Haitian, negative quantifiers usually require the presence of a negative marker. This is shown
in (19) where pa must occur.
(19)

a. M *(pa) te
w psonn / anyen.
I
not
ant see no one / nothing
I did not see anyone/anything.

haitian
(=(1a) in Deprez in press)

b. Psonn *(pa) rive.

No one arrived.

haitian
(=(2) in Deprez in press)

As noted by Deprez (in press), this contrasts with standard French, where pas
cannot occur in this context, as shown in (20).

12.5 The interpretation of negative quantifiers

(20)

a. Je nai vu personne.
I did not see anyone.

b. Personne nest arriv.

No one arrived.

361

french

In popular French, Jai pas vu personne I did not see anyone, built on the same
model as the Haitian sentence in (19a), is a possible sentence. Crucially, however, *Personne est pas arriv, along the lines of the Haitian sentence in (19b) is
ungrammatical in all varieties of French (Julie Auger p.c.). Furthermore, Oudin
(1640) discusses at length the fact that French pas is excluded from sentences
containing personne no one, rien nothing, etc. The above facts illustrate the
differences between Haitian and French with respect to the interpretation of
negative quantifiers and their interaction with pa and pas, respectively. I refer
the reader to Deprez (in press) for an extensive and thorough discussion of
further differences between the two languages. In view of these differences,
Deprez concludes that the properties of the Haitian negative quantifiers cannot
be attributed to French.
On the basis of very careful and thorough argumentation, Deprez (in press)
proposes accounting for the difference between Haitian and French in terms of
their determiner systems. On her account, French does not have bare nps. (As
was shown in chapter 4, French has a partitive determiner de, du and des.)
French negative quantifiers behave like indefinite determiners (or numerals) with
empty nouns. By contrast, Haitian has bare nps (as we saw in chapter 4), and
negative quantifiers are nouns with empty ds. These two structures are illustrated in (21) (adapted from (93) in Deprez in press).
(21)

a.

b.

french

haitian

dp
d
personne
no one

dp
np

np
psonn
no one

In addition, Deprez (in press) presents extensive arguments showing that

seventeenth-century French is not the source of bare nps in Haitian creole, a
conclusion which is in agreement with my own in chapter 4. In light of this
very insightful proposal, I now turn to a discussion of the substratum data.
In Fongbe, the negative quantifiers are mYt nobody and nLt nothing
corresponding to Haitian psonn and anyen, respectively. The sentences in (22)
show that the negative quantifiers co-occur with negative or negation markers
(discussed in chapter 7) in sentences that are interpreted as negative statements.
This contrasts with the French facts in (17) but parallels the Haitian data in (18).
(22)

a.

Myt
no one

w
f.
arrive Neg

a. Myt
m w
nx.
no one neg arrive part
No one arrived.

fongbe
fongbe

362

Parameters
b. N mw
I
see

nlt
f.
nothing Neg

fongbe

b. N m mw nlt
nx.
I neg see nothing part
I did not see anything.

fongbe

Furthermore, as is the case in Haitian (see (19)), a clause containing a negative

quantifier requires the presence of a negation or negative marker, as is shown by
the ungrammaticality of (23) which contains no such markers.
(23)

a. *N mw myt / nlt
[Lit.: I see no one/nothing.]

b. *Myt w
[Lit.: No one arrive.]

fongbe

It thus appears that the interpretation of negative quantifiers and their interaction
with other negative markers in Fongbe parallel the Haitian facts discussed above.
According to Deprezs general proposal based on Haitian, Fongbe negative quantifiers would thus be like Haitian negative quantifiers, that is, nps rather than determiners. Interestingly enough, and as we saw in chapter 4, Fongbe, like Haitian,
has bare nps. The Fongbe facts thus appear to provide independent support for
the formulation of the parameter proposed by Deprez (in press).
Further facts also support the conclusion that Haitian follows the substratum
pattern. For example, Deprez (in press) points out that, in Haitian, verbs which
have a negative meaning allow for negative quantifiers without the negation
marker, as is shown in (24).
(24)

Jan
refize pou
li
manje anyen.
John refuse comp 3rd eat
nothing
John refused to eat anything.

haitian
(=(3a) in Deprez in press)

These Haitian facts follow the same pattern as in Fongbe, as shown in (25).
(25)

Kwk gb6
nlt
3l.
Koku refuse nothing eat
John refused to eat anything.

fongbe

Both languages contrast with French, which requires a positive quantifier in this
context:
(26)

Jean a
refus de manger quoi que ce soit.
Jean aux refuse to eat
anything
John refused to eat anything.

french

Furthermore, Deprez (in press) documents the fact that Haitian negative quantifiers may violate various constraints. Because of the complexity of the facts, I
shall not discuss these data here. Suffice it to say here that I ran Deprezs Haitian
tests with Fongbe speakers and the same types of violations occurred.
On the basis of the above comparison, I conclude that, although the phonological representations of the negative quantifiers in Haitian are derived from
French, their semantic and syntactic properties are derived from those of the
corresponding lexical entries in substratum languages such as Fongbe. Using the

12.6 Verb-doubling phenomena in particular grammars

363

parameter defined by Deprez (in press), I propose to account for the history of
the Haitian facts as follows. The forms psonn and anyen were incorporated into
the early Haitian lexicon as nouns rather than as determiners since the creators
of Haitian had bare nps. The first generation of Haitian native speakers encountered bare nps in the language they were exposed to and deduced that negative
quantifiers were nps (rather than determiners). This is a probable scenario, given
the facts of modern Haitian creole.
12.6

Verb-doubling phenomena in particular grammars

Verb-doubling phenomena involve four constructions which contain what looks

like an exact copy of the predicate (henceforth the copy). As Koopman (1986)
shows, constructions involving a copy of the verb are attested in Haitian and in
West African languages but not in French. Section 12.6.1 provides a general
overview of the constructions involving verb-doubling phenomena and summarises the related theoretical issues. Section 12.6.2 describes the properties which
distinguish the four constructions involved. Section 12.6.3 discusses certain hypotheses proposed to account for the availability of these phenomena in particular
grammars and the marked character of this parametric option.
12.6.1

Overview

Clauses containing what looks like a copy of the predicate involve four constructions: temporal adverbial, as in (27), causal adverbial, as in (28), factive clauses,
as in (29) and the predicate cleft construction, as in (30). In the examples below,
the first occurrence of the verb is an exact replica of the second one.
(27)

Temporal adverbial
W
Jan
w
(trl)
bw
Mr y.
fongbe
Rive Jan
rive
(epi) Mari pati.
haitian
arrive John arrive as-soon-as and Mary leave
As soon as John arrived, Mary left.
(=(1) in Lefebvre 1994b)

(28)

Causal adverbial
W
Jan
w
t
Rive Jan
rive
arrive John arrive cause
Because John arrived, Mary

Mr y.
Mari pati.
Mary leave
left.

fongbe
haitian
(=(2) in Lefebvre 1994b)

(29)

Factive
W
3i Jan
Rive
Jan
arrive op John
twn.
li
kntan.
his happy
The fact that John

w
v,
vv
n
rive
a,
f
arrive det make(-happy) for

nw
manman
mother

fongbe
haitian

arrived made his mother happy.

(=(3) in Lefebvre 1994b)

364
(30)

Parameters
Predicate cleft
W
wy Jan
w.
Se
rive
Jan
rive.
it-is arrive it-is John arrive
It is arrive that John did (not e.g. leave).

fongbe
haitian
(=(4) in Lefebvre 1994b)

There is a debate in the literature as to which of the two occurrences of the

predicate in the sentences in (27)(30) is the verb and which the copy. Koopman
(1984) and Ndayiragije (1993) argue that the first occurrence of the predicate
is actually the verb of the clause moved to a sentence-initial position and the
second is the spell-out of the trace left by verb movement. Lefebvre (1994b),
Collins (1994) and Law and Lefebvre (1995) argue for a different proposal. In
their view, the second occurrence of the predicate is the verb in the construction,
whereas the first one (which has much in common with cognate objects, see
Lefebvre 1994b) binds the event position in the argument structure of the verb.
A detailed discussion of this issue would go far beyond the scope of this section
and nothing hinges on it. For our purposes, I will assume the first occurrence of
the predicate to be the copy.
In all four constructions above, the copy is an exact replica of the verb. As
has been pointed out on several occasions, the copy cannot bear overt nominalising
morphology9 (see Ndayiragije 1993; Lefebvre 1994b; Law and Lefebvre 1995).
This contrasts with Yoruba, where similar constructions require such morphology, as shown in (31).
(31)

a. Rr
(w) t
Aje ra
w.
nom buy paper rel Aje buy paper
Once Aje bought books.
The fact that Aje bought books.

yoruba

(=(47b) in Manfredi 1993)

b. Rr
w
ni
Aje ra
nom buy paper comp Aje buy
It is book buying that Aje is doing.

w.
paper

yoruba
(=(46b) in Manfredi 1993)

The Fongbe/Haitian facts in (27)(30) also differ from Yoruba in that no complement can occur with the copy, a fact noted several times in the literature
(see e.g. Piou 1982a, 1982b; Lumsden and Lefebvre 1990; Lefebvre 1990b;
Ndayiragije 1993; Lefebvre and Ritter 1993; Collins 1994). By contrast, in
Yoruba the nominalised verb can appear with a complement (see (31) ).
I take the nominalised verb in Yoruba to be an eventive nominal of the same
type as eating the cake in English. This means that it can appear with its complement. On the assumption that the nominalising morphology on the verb binds
the head of dp, it follows that the nominalised verb in Yoruba cannot appear
with a determiner. Following a widespread claim in the literature that the copy
of the verb in verb-doubling phenomena of the type in (27)(30) relates to the

12.6 Verb-doubling phenomena in particular grammars

365

event denoted by the clause (see e.g. B0gb- 1972; Dchaine 1992; Lefebvre
1994b; Law and Lefebvre 1995), I will assume that the copy in (27)(30) binds
the event position in the argument structure of the verb. I will further assume
that the copy belongs to the category n, as argued by Hutchison (1989), Manfredi
(1993), Lefebvre (1994b) and Collins (1994). As is extensively discussed in
Lefebvre (1994b), this noun has the properties of simple nominals: it has no
thematic arguments and hence no complement can occur with it. As we will see
below, in contrast to the Yoruba type of nominalisation, the copy in (27)(30)
can appear with a determiner. This is consistent with the fact that, in this case,
there is no nominalising morphology and hence the head of dp is free and can be
filled by the determiner.
12.6.2

Properties of verb-doubling constructions

The temporal adverbial clause containing the copy in (27) is conjoined (by bW
in Fongbe and epi in Haitian) to the clause interpreted as the main clause. The
causal adverbial clause containing the copy in (28) is analysed as an adjunct to
the main clause (see Lefebvre and Ritter 1993). Unlike these adverbial clauses,
the factive clause containing the copy is a thematic argument of the main verb.
As is pointed out in Lefebvre and Ritter (1993) for Haitian, and Collins (1994)
for Ewe and Fongbe, this type of predicate-doubling construction may appear as
the subject of a causative verb (meaning to make, as in (29) ), or as the object
of a factive verb meaning to regret. The inner structure of the factive clause in
(29) is similar to that of relative clauses. Consider the relative clauses in (32).
(32)

Tvi 3ii
i
u
Tabi
table op
you
The table that you

xw
ti v
achte ti a
buy
det
bought.

fongbe
haitian

The relativised noun in (32) is linked to its basic position within the relative
clause by an operator (#I in Fongbe and in Haitian, as discussed in chapter 7).
The relative clause occurs between the head of the clause and its determiner. Likewise, in the factive construction in (29), the copy is related to the embedded clause
by the operator #I in Fongbe and in Haitian, and the clause interpreted as
factive occurs between the copy and the determiner (for an extensive discussion
of the similarity between factive and relative clauses in Fongbe and Ewe, see
Collins 1994; for a description and analysis of relative clauses in Fongbe, see
Dchaine and Filipovich 1986 and, for Haitian, see Koopman 1982a). Finally, in
the predicate cleft construction in (30), the copy occurs within a clefted constituent which makes use of se it-is in Haitian and wY it-is in Fongbe.10
The four constructions above are further distinguished on the basis of extraction facts. It is well documented that, in the predicate cleft construction, the copy
may undergo long-distance movement similar to long-distance Wh-movement
(see e.g. Piou 1982a; Koopman 1984; Lefebvre 1990b; Lumsden and Lefebvre

366

Parameters

1990; Ndayiragije 1993; Law and Lefebvre 1995). This is illustrated in (33) and
(34) for Fongbe and Haitian, respectively.
(33)

X wy Sk ln
3w Kf 3w sb x Kwk.
fongbe
hit it-is Cica think say Kofi say Asiba hit Koku
It is hit that Cica thinks that Kofi said that Asiba did to Koku.
(=(59c) in Law and Lefebvre 1995)

(34)

Se
pte Mari di
m l
ap
pte.
it-is bring Mary say 1st 3rd def-fut bring
It is bring that Mary told me that she will do.

haitian
(=(53a) in Piou 1982a)

In temporal and causal adverbials, however, the copy must appear within the
same minimal clause as the corresponding verb (see Piou 1982b; Lefebvre and
Ritter 1993; Collins 1994). The fact that the copy cannot be extracted out of a
temporal adverbial clause follows from an independent constraint which prevents extraction out of co-ordinate structures (see Ross 1967). In the case of
causal adverbial clauses, it follows from an independent constraint preventing
extraction out of adjuncts (see Ross 1967). Finally, as we saw above, factive
clauses are complex nps, so the impossibility of extraction out of this construction follows from an independent constraint preventing extraction out of complex nps (see Ross 1967).11
The semantics of the four constructions in (27)(30) imposes a constraint on
the type of predicate that allows verb-doubling phenomena in specific constructions. In discussing this issue, I will appeal to the distinction between Stage- and
Individual-Level predicates discussed in Kratzer (1989) (building on Carlson
1977). A Stage-Level predicate denotes an action or temporary property of the
subject (e.g. eat, speak, sit). By contrast, an Individual-Level predicate denotes
a permanent property of the subject (e.g. altruistic, know, resemble). As is discussed at length in Lefebvre and Ritter (1993), temporal adverbial clauses involving a copy of the predicate are only possible with Stage-Level predicates. The
account of this fact given in Lefebvre and Ritter (1993) is as follows. Since
this construction is used to refer to an event which must be completed at the
moment when the event denoted by the matrix predicate takes place, it must be
possible to locate the events occurrence in time and/or space. Only Stage-Level
predicates may be characterised in terms of their temporalspatial reference (see
Kratzer 1989). Thus, since temporal predicate-doubling constructions must be
completed at a particular point in time, it follows that only Stage-Level predicates will be able to participate in this particular construction. In contrast to the
temporal adverbial verb-doubling construction, the causal and factive constructions allow for both Individual- and Stage-Level predicates (see Lefebvre and
Ritter 1993). Lefebvre and Ritter (1993: 75) point out that a causal adverbial
need not refer to a specific point in time in order to be interpreted as a cause.
Causal adverbials may denote either events or states which affect the matrix event
or state. A similar statement applies to factive clauses. As for the predicate cleft

12.6 Verb-doubling phenomena in particular grammars

367

construction, there appears to be considerable variation among West African

languages and among Haitian speakers.
Koopman (1984) presents examples of the predicate cleft construction in
Vata. Her examples show that, in this language, both Stage- and IndividualLevel predicates may be clefted and, furthermore, in both cases the clefted
phrase is assigned an emphatic reading. This is illustrated in (35).
(35)

a. Lc
l
sk.
eat we ate rice
We ate rice.

vata
(=(52a) in Koopman 1984: 38)

b. Zal= x sly d.
red it red like blood
It is very red.

vata
(=(11f) in Koopman 1984: 157)

DeGraff (1994) presents similar data for Haitian, as shown in (36).

(36)

a. Se
vle Bouki vle lajan
leta.
it-is steal Bouki steal money state
Bouki stole state money.

haitian
(=(17c) in DeGraff 1994)

b. Se
bouke Bouki
it-is tired
Bouki
Bouki is tired.

bouke.
tired

haitian
(=(17a) in DeGraff 1994)

Not all Haitian speakers produce data of the type in (36), however.
On the basis of data drawn from a sample of Haitian speakers, Lefebvre
(1990b) shows that the predicate cleft construction may involve Stage-Level predicates (see (37) ) but not Individual-Level predicates (see (38) ). These examples
also show that the clefted constituent is assigned a contrastive reading (rather
than an emphatic reading as in the Vata examples in (35) and the Haitian examples
in (36) ).
(37)

a. Se
kouri Jan
kouri.
it-is run
John run
It is run that John did (not e.g. walk).

haitian
(=(7) in Lefebvre 1990b)

b. Se
dmi Jan
dmi (pandan indtan).
haitian
it-is sleep John sleep (for an-hour)
It is sleep (not e.g. sit) that John did (for an hour).
(=(8) in Lefebvre 1990b)
c. Se
manje Jan
manje pen.
it-is eat
John eat
bread
It is eat bread that John did (not e.g. bake).

haitian
(=(9) in Lefebvre 1990b)

368

Parameters

The sentences in (38) show that, for the same subset of Haitian speakers,
Individual-Level predicates cannot participate in the predicate cleft construction.
(38)

a. *Se
it-is

entlijan
Jan
intelligent John

entlijan
intelligent

haitian
(=(14) in Lefebvre 1990b)

b. *Se
it-is

konnn Jan
konnn lang
sa
know
John know
language this

c. *Se
it-is

renmen Jan
renmen Mari
love
John love
Mary

a
haitian
det
(=(15) in Lefebvre 1990b)
haitian
(=(16) in Lefebvre 1990b)

d. *Se
it-is

sanble
Jan
sanble
ak
papa -l
haitian
resemble John resemble with father his
(=(17) in Lefebvre 1990b)

Interestingly enough, the Haitian data in (37) and (38) parallel the Fongbe
data in Law and Lefebvre (1995), based on a subset of Fongbe speakers. For this
sample of speakers as well, only Stage-Level predicates (see (39) ) may participate in the predicate cleft construction and the clefted phrase is assigned a contrastive reading.
(39)

a. Gb
wy sn 3 gb
mvt v.
destroy it-is man a
destroy car
det
It-is destroy that a man did to the car (not e.g. fix it).

fongbe

b. 3D wy Kwk 3 svn v.
eat it-is Koku eat crab det
It is eat that Koku did to the crab (not e.g. throw it away).

fongbe

c. Fn
wy sn 3 fn
mvt v.
fongbe
steal it-is man a steal car
det
It is steal that a man did to the car (not e.g. buy it).
(=(11) in Law and Lefebvre 1995)

The sentence in (40) shows that Individual-Level predicates do not participate in

the predicate cleft construction for this subset of Fongbe speakers.
(40)

*Tn
wy sn 3 tn
sb
fongbe
know it-is man a know Asiba
[Lit.: It is know that a man does to Asiba.]
(=(12) in Law and Lefebvre 1995)

It thus appears that subsets of Haitian speakers manifest different patterns which
correspond to the different patterns found in the substratum languages (see (35)
and (36); (37), (38) and (39), (40) ).
Furthermore, the speakers examined in Lefebvre (1990b) and Law and Lefebvre
(1995) allow various contrastive interpretations of the clefted constituents. In the
above examples, the contrastive reading relates to the v alone. In the examples
below, even though the clefted constituent consists only of the copy of the verb,
the contrastive reading affects the whole vp.

12.6 Verb-doubling phenomena in particular grammars

(41)

369

a. Se
manje Jan
manje pen
an.
haitian
it-is eat
John eat
bread det
It is eat the bread that John did (not e.g. drink the water).
(=(44) in Lefebvre 1990b)
b. X wy sb x Kwk.
fongbe
hit it-is Asiba hit Koku
It is hit Koku that Asiba did (not e.g. kill Cica).
(=(66) in Law and Lefebvre 1995)

Likewise, for these speakers, when the delimiting object is clefted, the contrastive reading may bear on this argument alone or on the whole vp, as shown in
(42).
(42)

a. Se
pen
an
Jan
manje.
haitian
it-is bread det John eat
It is the bread that John ate (not e.g. the meat).
or It is eat the bread that John did (not e.g. drink the water).
(=(53) in Lefebvre 1990b)
b. Mvt v
wy sn 3 gb.
fongbe
car
det it-is man a
destroy
It is the car that a man destroyed (not e.g. the bicycle).
or It is destroy the car that a man did (not e.g. build the house).
(=(72) in Law and Lefebvre 1995)

The semantic interpretation facts in (41) and (42) are remarkable and they
show a striking parallelism between the grammars of these subsets of Haitian
and Fongbe speakers. Various accounts of these facts may be found in Lefebvre
(1990b), Larson and Lefebvre (1991), Collins (1994) and Law and Lefebvre
(1995).
There is yet another property which is shared by all four constructions
involving verb-doubling in both Haitian and Fongbe. As is documented in
detail in Lefebvre (1994b), for a subset of Haitian and Fongbe speakers (who
speak a variety I refer to as Haitian1 and Fongbe1, respectively), the copy occurring in (27)(30) may take the determiner V in Fongbe and la in Haitian (or
their allomorphs), as shown in (43)(46). Informants usually render the meaning of the determiner in the context of the copy by as expected/as we knew.
The presence of the determiner in this context induces the interpretation this
event in question/this event that we know of. The meaning of the determiner
in the context of the copy is thus very similar to that of the event determiner
in chapter 8.12
(43)

Temporal adverbial
W
v
Jan
w
(trl)
Rive a
Jan
rive
arrive det John arrive as-soon-as
As soon as John arrived (as we knew he

bw
Mr y.
fongbe1
(epi) Mari pati.
haitian1
and Mary leave
would), Mary left.
(=(19) in Lefebvre 1994b)

370
(44)

Parameters
Causal adverbial
W
v
Jan
w
t
Rive a
Jan
rive
arrive det John arrive cause
Because John arrived (as we knew

Mr y.
fongbe1
Mari pati.
haitian1
Mary leave
he would), Mary left.
(=(20) in Lefebvre 1994b)

(45)

Factive
W
v
3i Jan
w
v,
vv
n
fongbe1
Rive a

Jan
rive
a,
f
haitian1
arrive det op John arrive det make(-happy) for
nw
twn.
manman li
kontan.
mother
his happy
The fact that John arrived (as expected) made his mother happy.
The (very) fact that John arrived made his mother happy.
(=(21) in Lefebvre 1994b)

(46)

Predicate cleft
Y
v
wy
Se
ale
a
it-is leave det it-is
It is leave (as expected)

Jan
y.
fongbe1
Jan
ale.
haitian1
John leave
that John did (not e.g. stay home).
(=(22) in Lefebvre 1994b)

In contrast to speakers of Haitian1, speakers of what I will call Haitian2 accept

the event determiner only at the end of the clause containing the copy. The
judgments of these speakers are illustrated in (47)(50).13
(47)

(48)

Temporal adverbial
Rive Jan
rive
a
(epi) Mari pati.
arrive John arrive Det and Mary leave
As soon as John arrived (as we knew he would), Mary left.
Causal adverbial
Rive Jan
rive
a
Mari pati.
arrive John arrive Det Mary leave
Because John arrived (as we knew he would), Mary left.

haitian2

haitian2

(49)

Factive
Rive
Jan
rive
a,
f
manman
haitian2
arrive op John arrive Det, make(-happy) mother
li
kontan.
his happy
The fact that John arrived (as expected) made his mother happy.
The (very) fact that John arrived made his mother happy.

(50)

Predicate cleft
Se
ale
Jan
ale
a.
it-is leave John leave Det
It is leave (as expected) that John did (not e.g. stay home).

haitian2

Some speakers of the first group (see (43)(46) ) even accept sentences where
the event determiner occurs both with the copy (as in (43)(46) ) and at the end
of the clause containing the copy (as in (47)(50) ). The variation observed

12.6 Verb-doubling phenomena in particular grammars

371

among speakers is akin to the variation observed between West African languages with respect to whether they allow determiners to appear immediately
after the head of a relative clause or a factive construction or at the end of the
clause (for an extensive discussion of these facts, see Collins 1994). A unified
account of these data will require further research. For now, suffice it to say that
the additional data in (43)(46) further illustrate the parallelism between the
grammars of Haitian and Fongbe.
12.6.3

Parameter

To the best of my knowledge, two parameters have been proposed in the literature to account for the possibility of verb-doubling phenomena in particular languages. These two proposals reflect the opposite views discussed in section 12.6.1.
The first proposal, by Koopman (1984), holds that the first occurrence of the
verb in sentences of the type in (30) is the main verb of the clause which has
moved to sentence-initial position and the second is the spell-out of the verbal
trace. In her view, then, verb-doubling phenomena are produced by the rule
Move , where =v. The difference between languages which have the predicate cleft construction and those which do not rests on whether v can be a value
for in particular grammars. The fact that, as we saw in section 12.2, Kwa
languages have a negative value for the verb-raising parameter, however, casts
some doubt on this proposal.
The second proposal, from Lefebvre (1994b), is based on the hypothesis that
the first occurrence of the verb in (27)(30) is a cognate object of the verb.
Hence it is of the category n and is base-generated within the vp in the same
position as cognate objects. From this position, it is moved to sentence-initial
position, yielding the surface structures illustrated in (27)(30). Lefebvre (1993d,
1994b) establishes a correlation between the availability of the copy and the
availability of an event determiner in a given grammar. For example, Fongbe
speakers from Abomey, Cotonou and Porto Novo exhibit the verb-doubling phenomena in (27)(30), the event determiner (see chapter 8 and (43)(46) ) and a
combination of the two as in (43)(46). Likewise, the sample of Haitian speakers
discussed above presents the same pattern, i.e. the copy as in (27)(30), the event
determiner (see chapter 8) and a combination of the two (see (43)(46) ) are all
available.
The hypothesised correlation between availability of the copy and availability
of an event determiner in a given grammar is further supported by facts from
other West African languages. For example, other languages of the Gbe family
such as Gun, Gen and Xula (Maxime Da Cruz p.c.) have both the copy and an
event determiner. Similarly, the Kpele dialect of Ewe shows both the copy and a
functional category which shares properties with the event determiner (Chris
Collins p.c.). Ower (Igbo) presents similar facts. In this language, however, the
determiner of the copy is a bound rather than a free morpheme. According to the
analysis in Dchaine (1992: 107), and as can be seen in (51), the prefix e- that

372

Parameters

appears with the copy induces the same interpretation as the Fongbe/Haitian
determiner in that context.
(51)

F
ji
e-ji.
3pl hold holding
They held [something/thing] as expected.

ower-igbo
(Dchaine 1992: 107)

So the claim that there is a correlation between the availability of the copy and
the availability of an event determiner (whether a free morpheme, as in Fongbe/
Haitian, or a bound morpheme, as in Ower) is further supported by the above
data.14 This claim is also supported by data from grammars which lack both the
event determiner and the bare form of the verb copy. Some Fongbe speakers
from Ouidah, for example, make use of an eventive nominal bearing overt nominal morphology in much the same way as in Yoruba (see (31) ). As we saw in
chapter 8, the event determiner is not available in this grammar.
Based on the above description, two patterns seem to emerge with respect to
the inner constituency of the clefted constituent. These are schematised in (52).
The first one (52a) represents grammars which have a bare copy of the verb; in
this case, the determiner may head the dp. The second one (52b) represents
grammars where the copy bears nominal morphology; in this case, the nominal
morphology on the verb binds the head of dp and hence no determiner can
surface in this position.
(52)

a. Cognate object type (e.g. Fongbe1 / Haitian 1)

dp
copy
of verb

d0
Det

b. Eventive nominal type (e.g. Yoruba / Fongbe2)

dp
nominalised
verbi

d0
ei

These two patterns straightforwardly account for the two types of data discussed
above.
How do speakers of Haitian2 who use a bare verb copy and do not accept the
determiner in this environment (see (47)(50) ) fit into this picture? Suppose that
the grammar of these Haitian speakers is of the type illustrated in (52b). In this
case, however, the nominalising affix is not visible for, as we saw in chapter 10,
this affix is null in Haitian. In this view, the null affix binds the head of dp,
preventing a determiner from occurring in this position.

12.6 Verb-doubling phenomena in particular grammars

373

The fact that the copy may appear without an overt determiner in specific
sentences (see (27)(30) ) (as is the case in ordinary noun phrases, see chapter 4),
and the fact that in some grammars there is no event determiner, however, suggest that the correlation established above should probably be stated in broader
terms. For example, it could be formulated in terms of the properties of the
determiner systems in particular languages. On this analysis, the properties that
distinguish the Fongbe/Haitian definite determiner from the French determiner
would account for the fact that, in the former type of language, the verb copy
is available while, in the latter, it is not. There is plenty of evidence showing
that the French and Haitian/Fongbe determiners do not have identical properties
(see chapter 4 and chapter 8). Whatever the precise characterisation of the parameter accounting for the availability of verb-doubling phenomena in particular
grammars, it is clear from the data presented above that Haitian shares with its
substratum languages the ug option that allows for these phenomena, whereas
French has the opposite value for this parametric option.

12.6.4

Parameter setting in creole genesis

How did verb-doubling phenomena become part of the Haitian creole grammar?
I presume that speakers of the Kwa languages used the parametric value of their
own grammar in setting the value for this parameter in the incipient creole. On
the basis of the primary data that they were exposed to, the first generation of
Haitian native speakers identified the properties of the determiner in the language they were exposed to and deduced the availability of verb-doubling constructions. This is a reasonable hypothesis since, after 200 years of independent
evolution, verb-doubling phenomena remain an important feature of the grammar of Haitian. The claim that this parameter setting was carried over into the
creole by the substratum speakers is further reinforced by the fact that verbdoubling phenomena of the type described here are found only in those creoles
which have substratum languages with this feature. For example, they are not
found in the Pacific creoles. Verb-doubling phenomena thus do not constitute a
general feature of creole languages; this constitutes a strong argument against
the claim that all creole languages are alike (see Bickerton 1984). Furthermore,
given the rarity of verb-doubling phenomena among the languages of the world,
the availability of such phenomena would be a marked option of ug (see Koopman
1986, for an extensive discussion of this issue). This conclusion runs counter to
the claim by Bickerton (1984) and others whereby creole genesis involves setting the parametric options of ug for their unmarked values. As for scholars who
require that only marked options of ug constitute an argument for substratal influence on the creole, the presence of this construction in Haitian creole certainly
constitutes an important argument. Finally, I take issue with researchers (e.g.
Mufwene 1994) who claim that the Haitian verb-doubling phenomena described
above can be explained via the lexifier language to show how this could be the

374

Parameters

case. To the best of my knowledge, no dialect of French presents phenomena of

the type described above with the specific properties enumerated in this section.
12.7

Conclusion

As a result of not having relexified the syntactic clitics of their native languages,
the creators of Haitian had to reset the value of the null subject parameter in
creating the creole. Otherwise, they kept the values of the parameters of their
own grammar. As was shown throughout this chapter, this is largely due to the
fact that they reproduced the properties of the functional categories of their own
lexicons through relexification. As we saw in section 12.2, the absence of verb
raising is linked to the absence of inflectional morphology. The availability of
serial verbs is linked to the lack of derivational and inflectional morphology.
The availability of the double-object construction is linked to the presence of
the Genitive Case. Finally, the interpretation of negative quantifiers is linked
to the existence of bare nps. The availability of verb-doubling phenomena was
hypothesised to be related to the properties of the determiner system. While the
bulk of the parametric options of Haitian correspond to those of the substratum
languages, none correspond to those of French. This fact strongly supports the
view, advocated in chapter 2, whereby the creators of a radical creole do not
have enough access to the superstratum language to learn its parametric values
and must fall back on the values of their native languages in setting the parametric values of the language that they are creating.

13 Evaluation of the hypothesis

In this chapter I evaluate the hypothesis presented in chapter 2 against the data
discussed throughout this book. The discussion is organised in terms of the components of the model of grammar used in this book: the lexicon, the interpretive
component, parameters and word order. The chapter ends with a discussion of
various questions related to the scenario of creole genesis.
13.1

The lexicon

This section evaluates the role of the processes hypothesised to be involved in

creole genesis across the various types of lexical entries: relexification and the
two processes claimed to be fed by its output, reanalysis and dialect levelling.
In evaluating the role played by the processes that created the Haitian lexicon,
I will identify relevant Fongbe lexical entries and show how these processes
generated the corresponding lexical entries in the creole. In order to facilitate
the general discussion, the data will be identified as in (1) in the tables below.
(1)

A. substratum entry relabelled by an overt phonological form;

B. substratum entry assigned a phonologically null form;
C. creole entry reanalysed as the phonological form of an entry produced by B;
D. substratum entry relexified and further subjected to dialect levelling;
E. substratum entry relexified with some reorganisation;
F. substratum entry abandoned;
G. superstratum form/feature acquired by the creators of Haitian;
H. independent development from within the creole.

(A)(E) all constitute various cases of relexification. (F) stands for substratum
entries that have been abandoned. Haitian lexical entries with no corresponding
substratal forms are identified as being either cases of acquisition from French
(G) or cases of independent development from within the creole (H). Functional
lexical entries, derivational affixes and the syntactic properties of verbs will be
discussed in turn.
13.1.1

Minor category lexical entries

Table 13.1 identifies the minor category lexical entries involved in noun structure (see chapter 4) that are affected by the processes identified above. As can
375

376

Evaluation of the hypothesis

Table 13.1. Minor category lexical entries involved in noun structure
A
Substratum lexical entries
definite determiner
plural marker via 3rd person pronoun
the two demonstrative terms
Genitive Case
Objective Case
the so-called indefinite determiner

X
X
X

X
X
X

X
X

Creole lexical entries

the so-called indefinite determiner

Table 13.2. Tense, Mood and Aspect markers

A
Substratum lexical entries
marker of anteriority
definite future marker
indefinite future marker
subjunctive marker
habitual marker
imperfective marker

X
X
X
X
X

be seen in this table, six out of seven minor category lexical entries of the substratum languages were relexified. Four were relabelled on the basis of French
phonetic matrices, two were assigned a phonologically null form (still the case
in modern Haitian) and two relexified lexical entries were also the object of
dialect levelling in the early creole. One creole lexical entry is hypothesised to
constitute an independent development.
Table 13.2 identifies the tma markers affected by these processes (see chapter 5). Five out of six Fongbe lexical entries can be seen to have been relexified.
Three involve relabelling on the basis of a French phonetic matrix. The definite
future marker was relabelled by a null form and later assigned a phonological
representation through reanalysis. The imperfective marker was shown to have
been produced by relexification with some modification of the distributional
properties of the new lexical entry compared to the original one. A few cases of
dialect levelling were also discussed at the end of chapter 5 on the basis of other
substratum languages.
Out of the eight paradigms of pronominal forms listed in Table 13.3 (see
chapter 6), five have been relexified. Four of these involve relabelling on the
basis of French phonetic matrices. The reflexive anaphor was assigned a null

13.1 The lexicon

377

Table 13.3. Pronominal forms

A B C D
Substratum lexical entries
paradigm of personal pronouns
defective paradigm of possessive pronouns
logophoric pronoun
pronominal clitics
availability of a null expletive subject
reflexive anaphor
body-part reflexives (not Fongbe)
Wh-words and phrases

E F G

X
X
X
X
X
X

X
X

X
X

Table 13.4. Minor category lexical entries involved in the structure of

the clause
A
Substratum lexical entries
complementiser introducing complements
of verbs of the say-class
two complementisers introducing
complements of verbs of the want-class
clausal conjunction
focus wY
determiner in the clause
nominal operator
negation marker
yesno question marker
negative marker
two markers of insistence
Creole lexical entries
se
ske

X
X
X
X
X
X
X
X
X
X
X
X

form at relabelling. The null expletive was carried over into the creole (see (B) ).
Dialect levelling was also shown to have been pertinent in two cases.
Table 13.4 shows that, out of the 12 minor category lexical entries involved
in the structure of the Fongbe clause (see chapters 7 and 8), seven were argued
to have been relabelled on the basis of French phonetic matrices. Two were
assigned a phonologically null form. Three Fongbe lexical entries have not been
relexified: the negative marker, the yesno question marker and the focus marker
wY. The semantically closest form to wY in Haitian is se which was shown to
constitute an independent development from within the creole. The semantically
closest form to the Fongbe yesno question marker is Haitian ske. This form,
derived from French est-ce que, was shown to have been incorporated into

378

Evaluation of the hypothesis

Haitian as a major category lexical entry and identified as a case of acquisition

from French since it does not correspond to a lexical entry in Fongbe. Recall,
however, that it has only the semantic properties of French est-ce que, not its
internal structure. In chapter 7, it was also shown that Haitian, like Fongbe (and
unlike French), has a resumptive form in the extraction site of subjects. This
situation does not follow from relexification since the distribution of the resumptive form is syntactically driven. But it constitutes another example showing that
the creole follows the substratum pattern.
On the basis of this summary of the data from chapters 4, 5, 6, 7 and 8, I now
turn to a discussion of some of the issues raised in chapters 1 and 2. Do functional category lexical entries undergo relexification? In Tables 13.1, 13.2, 13.3
and 13.4, we see 24 out of 32 lexical entries or paradigms which were identified
as having been relexified (see columns A and B) and 8 which were not (see column F). Among the entries that have been relexified, only one (the imperfective,
see Table 13.2, column E) entailed a change in the distribution of the lexical
item (for details, see chapter 5). Moreover, in the data chapters, it was shown
that the functional categories of Haitian have retained the directionality properties
of the corresponding lexical entries in the substratum languages. These facts argue
that functional category lexical entries do undergo relexification. Furthermore,
they show that, after more than 200 years of independent evolution, cut off from
its substratum languages, Haitian retains the bulk of the morpho-syntactic categories of these languages. This conclusion is in direct contradiction to the claim
in Muysken (1994b: 25) that the West-African morpho-syntactic categories have
not survived in the Caribbean creoles.
How does relexification proceed in the case of functional category lexical
entries? Recall from chapter 2 that a lexical entry copied from the substratum
lexicon was hypothesised, in this case, to be assigned a label on the basis of a
superstratum phonetic string, perceived as a major category lexical item, that
shares some semantic and distributional properties with the original lexical
entry. If the superstratum did not offer a suitable form, the copied lexical entry
was hypothesised to be assigned a null form. All the lexical entries identified as
A or B in Tables 13.1, 13.2, 13.3 and 13.4 have been individually shown to
conform to this hypothesis (see chapters 4, 5, 6, 7 and 8). Thus, the 16 lexical
entries or paradigms identified as A are cases for which a suitable French form
was available to serve as a label for a copied lexical entry. The lexical entries
identified as B belong to three types. The first one involves the definite future
marker and the reflexive anaphor. These lexical items have some semantic content and therefore they could have been relabelled on the basis of a French
phonetic matrix. As shown in the text, however, in these two cases, no suitable
forms were available in French. It was argued, however, that the null Haitian
forms are visible in contemporary Haitian. The definite future marker in Haitian was shown to have acquired a phonological representation through reanalysis.
The phonologically null reflexive anaphor is visible in the distribution of pronouns. The second type of null form involves the Genitive and Objective Case

13.1 The lexicon

379

markers, and the nominal operator. These have no semantic content and therefore could not undergo relabelling since this process is semantically driven.
Nonetheless, since they are visible in the grammar of Haitian, as was argued
in the data chapters, these forms are assumed to have been copied and assigned
a null representation at relabelling. The third type of null form includes those
which were already null in the substratum language: the complementiser introducing verbs of the say-class and the null expletive. These two lexical entries
were carried over into the creole unchanged. Note that Haitian lexical entries with
no phonological content are distinguishable from substratum lexical entries which
have not been relexified (those identified as F). The former type, although covert,
has visible effects on the syntax of the creole (see e.g. the Genitive Case discussed
in chapter 4), whereas the latter does not.
Substratum entries that were abandoned are of five types. The first type
involves the pronominal clitics, which are claimed not to have been relexified
because of how the process applies to functional items (see chapter 6). The
second type involves the defective paradigm of possessive pronouns. These
were claimed to have been abandoned because they were defective and because
there was another, more regular, way of expressing the same concept (see chapter 6). The third type is the habitual marker. Since Martinican creole has such a
marker, however, it was assumed that a form was available in the superstratum
language to relexify it. Its absence in Haitian creole was hypothesised to result
from dialect levelling. The fourth type involves the focus marker, the yesno
question marker and the negative marker. These could have been relexified since
they have some semantic content but they were not. For these cases (as well as
the null forms of the first type), it was claimed that the creators of Haitian did
not find a form in the superstratum language with a suitable distribution and
appropriate semantics to provide a label for the copied lexical entry. These cases
illustrate the claim in chapter 2 that the superstratum language imposes a constraint on relabelling. The documentation of this constraint would benefit from
comparisons between creoles sharing the same substratum languages but having
different lexifier languages (e.g. Haitian and Saramaccan). Finally, the so-called
indefinite determiner in the substratum language was abandoned.1
Recall from chapter 2 that lexical entries which were assigned a phonologically null form at relabelling can (but need not) be assigned a phonological form
through reanalysis in the later development of the creole. The interaction between
relexification (by a null form) and reanalysis was hypothesised to account for the
paradox raised in the literature whereby substratum languages may continue to
influence a creole even after the creole community has ceased to speak them.
Six copied lexical entries are identified as having been assigned a null form at
relabelling: the Genitive and Objective Case markers (Table 13.1), the definite
future marker (Table 13.2), the reflexive anaphor (Table 13.3), the nominal operator (Table 13.4) and the tensed complementiser introducing complements of
verbs of the say class (Table 13.4). The definite future lexical entry was shown
to have been assigned a label through reanalysis. In chapter 5, it was argued that

380

Evaluation of the hypothesis

when the covert lexical entry received a phonological form through reanalysis,
the new lexical entry had the properties of the original one. This supports the
above hypothesis. Note that this lexical entry has semantic content. Interestingly
enough, the lexical entries which have no semantic content (e.g. the Genitive
Case and the nominal operator) appear not to have acquired phonological content as yet. The question of whether this distinction is significant should be taken
up in light of data from various case studies of the type discussed in this book.
On the above account, only one clear case of reanalysis was identified. This
might seem a surprising result at first glance, since reanalysis is generally claimed
to play an important role in the development of creoles. As we saw in chapter 7
(see complementisers introducing complements of the want-class), if we take
into account the properties of the relevant lexical entries in the substratum languages as a point of departure, we have less need to call upon reanalysis (see
also Keesing 1988; Bruyn 1996).
According to our hypothesis, cases of dialect levelling are to be sought in
areas of the lexicon where the substratum languages differ. Given the relexification hypothesis, these differences are hypothesised to be reproduced in the early
creole lexicons. Dialect levelling is claimed to be the process by which some
of these differences are levelled out as speakers target the relexified lexicons
of their community. Cases identified as D in the above Tables constitute such
cases: the plural marker (whether homophonous or not with the third-person
plural personal pronoun, see chapter 4), a few cases involving tma markers
when examined from the point of view of non-Fongbe substratum speakers (see
chapter 5) and reflexive terms (see chapter 6). No such cases were reported for
functional category lexical entries involved in the structure of the clause. In my
view, this may very well reflect the fact that detailed descriptions of the pertinent data for other substratum languages are not as yet available. It is possible
that some of the cases identified as F (not relexified) in Table 13.4 actually
reflect differences among the substratum languages which were levelled out in
the early creole. In fact, dialect levelling seems to occur rather early in some
cases (e.g. plural marker in chapter 4) whereas it seems to be a long process in
others (e.g. the reflexive terms in chapter 6). The factors determining each type
of case still need to be studied. Finally, the detailed description and analysis of
the meaning and distribution of the determiner in the clause in chapter 8 shows
that not all the differences between substratum lexicons have been levelled out
in the further development of the creole.
Our hypothesis further states that the creators of a radical creole do not have
enough exposure to the superstratum language to identify as such the functional
category lexical entries of that language. This appears to be borne out by the
data, for there are no French functional category lexical entries which can be
identified as being also part of the Haitian lexicon. The one clear case of acquisition from French in the above inventory is the yesno question form ske (see
Table 13.4). But, as mentioned above, this lexical entry was incorporated into
the Haitian lexicon as a major category lexical item.

13.1 The lexicon

381

Finally, out of the 32 lexical entries or paradigms listed in Tables 13.1, 13.2,
13.3 and 13.4, there are only two cases where a Haitian form can be argued to
constitute an independent development from within the creole: the so-called
indefinite determiner yon and the form se. These are the only two cases which
do not conform to the relexification hypothesis. Se cannot be classified as a case
of acquisition from French, for, as we saw in chapter 7, it does not have the
properties of the French form from which it is phonologically derived. The form
yon a was shown to have evolved from younn one, a creole lexical entry.
13.1.2

Derivational affixes

Table 13.5 identifies the productive affixes discussed in chapter 10 in terms of

the processes under consideration. All eight Fongbe affixes have been relexified
(A). The copy prefix RE- was assigned a null form at relabelling. Two affixes
were argued to have involved dialect levelling as well (D). Finally, it was argued
that two productive affixes in Haitian were the product of independent development through reanalysis from within the creole (H). In this case, however, reanalysis was shown to reveal the indirect influence of the superstratum language
for the substratum languages did not have any corresponding affixes.
As can be seen in Table 13.5, all the derivational affixes of Fongbe were
relexified (in Table 13.5, there is no substratum affix identified under F). The
inventory of Fongbe derivational affixes cannot be argued to constitute a universal
core for there are languages, such as Vietnamese, which have no such affixes at
all. The fact that the inventories of Haitian and Fongbe derivational affixes are
quite similar thus cannot simply be attributed to chance. The data summarised
above show that derivational affixes do undergo relexification. Dialect levelling
was also shown to apply in this area of the lexicon.
Table 13.5. Derivational affixes
A
Substratum lexical entries
agentive -tV
attributive -nW
inversive m diminutive -vJ
place of origin / residence -tV / -n
ordinal -gVV
nominalising affix (not Fongbe)
copy RECreole lexical entries
verbalising -e
adverbial -man

X
X
X
X
X
X
X
X

X
X
X
X

382

Evaluation of the hypothesis

Table 13.6. Syntactic properties of verbs
A
Syntactic properties of substratum verbs
body-state verbs
weather verbs
reflexive verbs
expletive subjects
object-to-subject raising
subject-to-subject raising out of
a tensed clause
existential verb
control verbs
light verbs
double-object verbs
Properties of Haitian verbs
existential verb
Case-assigning properties of verbs
constant in lcs not spelled out

13.1.3

X
X
X
X
X

X
?

X
X
X
X
X
X
X
X

The syntactic properties of verbs

Table 13.6 identifies the syntactic properties of verb classes in relation to the
processes hypothesised to play a role in creole genesis (see chapter 9). Most of
the syntactic properties of substratum verb classes can be seen to have been transferred into the creole through relexification (A). In one (possibly two) case(s), dialect levelling has also been shown to play a role (D). The substratum existential
verbs appear not to have been relexified (F). The Haitian existential verb appears
to constitute an independent development (H). The Case-assigning properties of
verbs were also shown to be relatively independent of both source languages
(H). Finally, as we saw in chapter 9 (section 10), Haitian shares with French
the property that the constant in the lcs of a subclass of verbs is not spelled out
(G). Since this property contrasts with the substratum data and matches the
superstratum data, it is claimed to have been acquired from French.
In conclusion, out of the 13 items listed in Table 13.6, 9 can be argued to
have involved relexification (2 of which also involved dialect levelling), 2 constitute independent developments from within the creole, and only 1 can be identified as a case of acquisition from French. It therefore appears that relexification
has played a major role in establishing the syntactic properties of the Haitian
verbal repertoire.
13.1.4

Other major category lexical entries

Due to space limitations, the semantics of verbs and the properties of adjectives
and lexical items of the category P will be discussed elsewhere. The following
paragraphs summarise the major points.

13.1 The lexicon

383

In Lefebvre (in press), it is shown that relexification has played an important

role in establishing the semantic properties of Haitian verbs. When the semantics
of Haitian, French and Fongbe verbs differs, the Haitian lexical entry tends to
follow Fongbe rather than French. Likewise, in several cases where French has
simplex verbs, both Haitian and Fongbe render the same notion by idiomatic
expressions (v+np), and the details of the construction are parallel in both languages. Furthermore, some Fongbe verbs are underspecified for aspectual properties,
as are the corresponding Haitian verbs. This property allows both Haitian and
Fongbe verbs to be used as stative, resultative and dynamic verbs. Both languages contrast with French, where the corresponding predicates are specified for
aspectual properties, thus restricting their use in this respect. Finally, the thematic
properties of Haitian verbs tend to follow those of Fongbe rather than those of
French.2
A major difference between the three lexicons under comparison, however,
resides in the size of the verbal repertoires. While the number of verbal lexical
entries in Segurolas (1963) dictionary of Fongbe is around 900, there are twice
as many in Haitian (see Valdman et al. 1981). I have not counted the number of
verbs in French, but there are undoubtedly more than in Haitian. Although some
of the discrepancy between the sizes of the Fongbe and Haitian verbal lexicons
may be attributable to the different concepts of lexical entries underlying the
authors work (see Lefebvre in press, for a discussion of this point), it is clear
that Haitian has more verbal lexical entries than Fongbe. It therefore seems that,
even though the creators of Haitian did not acquire the entire repertoire of
French verbs, they did acquire some of it since there are Haitian verbal lexical
entries which have no equivalent in Fongbe. For example, Lefebvre (in press)
shows that Haitian has acquired some agentive verbs from French. The introduction of such verbs into the Haitian verbal lexicon is shown to have triggered a
reorganisation in the properties of verbal lexical entries, originally relexified
from Fongbe.
Lexical entries identified as adjectives3 in Segurolas (1963) Fongbe dictionary fall into three categories: deverbal adjectives formed by means of the affix
RE- (see chapter 10), nouns used as adjectives and simple (non-derived) adjectives. Haitian adjectives identified in Valdman et al. (1981) are of the same type:
deverbal adjectives formed by means of the affix - (see chapter 10), nouns used
as adjectives and true adjectives. As mentioned in chapter 10, French does not
form adjectives by means of an affix corresponding to RE- in Fongbe and - in
Haitian. As is the case in Fongbe and in Haitian, however, some nouns may be
used as adjectives and simple adjectives also exist in this language. Thus, with
respect to types of adjectives, Haitian is very much like Fongbe. A major difference between the three lexicons is the number of simple adjectives. Fongbe has
only a few (no more than 15) lexical entries of this type. French has many such
adjectives and, once again, Haitian appears to be in the middle. In this area of
the lexicon as well, Haitian seems to have borrowed a large number of lexical
items from French. It is my contention, however, that many French adjectives
were incorporated into Haitian as stative verbs on the model of the substratum

384

Evaluation of the hypothesis

languages of Haitian (see Lefebvre in press). More work needs to be done in

order to establish the respective contributions of Fongbe and French to this
aspect of the Haitian creole lexicon.
A similar observation holds for lexical entries of the category P. Fongbe has
four prepositions (see Lefebvre 1990a, 1995b) and a larger class of postpositions.
While Saramaccan has maintained the distinction between pre- and postpositions
(see e.g. Byrne 1987; Muysken 1987) in much the same way as Fongbe (based
on my reading of the available data in the literature), Haitian has not, as we saw
earlier. Haitian has a rather large class of prepositions (see Gilles 1988) which
derive their phonological representations from French and which sometimes do
not correspond to lexical entries in Fongbe. It therefore appears that there has
been some acquisition from French in this area of the lexicon as well. However,
there is some evidence that relexification has also played a role here. Recall
from chapter 7 that, in both Haitian and Fongbe but not French, the preposition meaning with serves to conjoin (so to speak) noun phrases. In contrast
to French, Haitian lacks a preposition meaning by. This follows from the
relexification hypothesis: the creators of Haitian had no such lexical entry to
relexify (see Brousseau 1993). Consequently, the Agent must remain implicit in
syntactic passives in Haitian and Fongbe, while it can be expressed in French.
Another point is that the most salient Haitian prepositions only partially share
the semantics of the French forms that they are phonologically derived from (see
Joseph 1994). In my view, this is because the meaning of Haitian prepositions
often corresponds to the conflated meaning of a pre- and a postposition in the
substratum language. This is a fascinating topic for future research.
13.1.5

The representation of relexification

Throughout this book, I have assumed relexification to be a two-step process

involving copying and relabelling on the basis of phonetic matrices from the
superstratum language or by a null form (see (2) in chapter 2). Recall from
chapter 2 that, on Mouss (1995) account, relexification consists in assigning a
parallel label to an existing lexical entry, as shown in (2) (=(17) in chapter 2).
(2)

G [phonology] i / [phonology] j J
K
H
[semantic feature] k
I
[syntactic feature] n
L

The derivations discussed in this chapter can easily be recast within Mouss
framework provided that paralabelling is considered to be semantically driven
and that it allows for phonologically null forms.
Suppose that the hundred years or so when the substratum languages were
still spoken in Haiti constitute the period during which relexification took place.
In the following discussion, I will assume the three types of lexical entries discussed in section 13.1. The first type constitutes those lexical entries which have
some semantic content (major category items, some functional category items and

13.1 The lexicon

385

derivational affixes). A parallel label would be assigned to such lexical entries

on the basis of appropriate French phonetic matrices. If no suitable French form
can be found, the parallel form would be null, that is, would not be pronounced.
However, the representation in (2) provides us with another possibility: using
the form from ones own native lexicon in such cases.4 This possibility should
not be ruled out for several reasons. First, we saw that in mixed languages
speakers do retain the functional categories of their first language. Second, as
mentioned in chapter 3, Berbice Dutch and Saramaccan have retained some
functional category lexical entries from their substratum languages (e.g. Fongbe
wY in Saramaccan, discussed in chapter 7). Third, the cases of retention of major
category lexical entries from the substratum languages discussed in chapter 3
(note 9) can be accounted for in this way.
The second type of lexical entry comprises those which have no semantic content, such as Case markers, operators, etc. In paralabelling, these lexical entries
would be assigned a null form. Practically speaking, this means that they are not
pronounced. But, as we saw in the data chapters, their properties can be deduced
on the basis of linguistic evidence or lack of evidence. The third type of lexical
entries includes those which are phonologically null in the substratum languages
(e.g. the complementiser introducing verbs of the say-class, the null expletive).
These are simply used unchanged when speakers of the substratum languages
speak the new creole.
This approach provides a straightforward explanation of what happens when
the substratum languages cease to be spoken: speakers simply cease to use the
original labels for lexical entries (in most cases) and are left with the new labels
created by relexification. At this stage, reanalysis may apply to fill in the holes
(see the definite future marker discussed in chapter 5). But, as we saw earlier in
this chapter, purely grammatical morphemes still appear to be phonologically
null in modern Haitian. In the data chapters, however, they have been shown to
have the properties of the corresponding overt forms in the substratum languages.
13.1.6

Relexification: a central process in radical creole genesis

The data reported on in this book show that, in the formation of a radical creole,
acquisition of lexical entries from the superstratum language is restricted to
major category lexical entries. The process of relexification applies throughout
the lexicon and across all types of syntactic categories. It is claimed to be a
central process in the formation of radical creoles. First, as was pointed out in
Lefebvre and Lumsden (1994b), most of the information that distinguishes languages is registered in the lexicon. By definition, relexification is a process that
applies to lexical entries. As was demonstrated throughout this book, it applies
throughout the lexicon and across all types of syntactic categories. It is this process which is responsible for the fact that the properties of the Haitian lexical
entries reproduce the idiosyncratic features of the substratum languages rather
than those of the superstratum language. Moreover, relexification is a central

386

Evaluation of the hypothesis

process because, as we have seen over and over again, it feeds the other two
processes that apply to the lexicon: reanalysis and dialect levelling. Indeed, it
has been argued that these two processes operate on the output of relexification.
Another dimension of the central role of relexification in creole genesis lies in
the interaction between the availability of specific functional category lexical
items in particular languages and the parametric choices made by these languages. As we saw in chapter 12, the fact that the parametric values of the creole reproduce those of the substratum languages is linked to the fact that the
functional category lexical entries of the creole were produced by relexification.
For all these reasons, relexification is claimed to be central in the formation of
radical creoles. Of course, recourse to relexification is a function of very limited
access to the superstratum data. In cases where the speakers of the substratum
languages have more access to the superstratum data, acquisition is facilitated
and relexification might be less important than in the case described here.
13.2

The interpretive component

In chapter 2, it was hypothesised that the creators of a creole use the principles
of their own grammar in establishing the semantic interpretation data of the new
language that they are creating. Various types of data presented in this book
show that this is indeed the case.
The temporal interpretation of bare sentences in Haitian was shown to follow the principles of Fongbe (and presumably of other similar substratum languages, see chapter 5, section 3). The interpretive facts involving various types
of cleft constructions (see chapter 12, section 6) also provide strong support for
the above claim. Also pertinent for this issue is the concatenation of morphemes
and words into larger units. Derivational morphology constitutes a case in point.
Recall from chapter 10 that quite a large number of Haitian derived words have
no corresponding derived words in French and that, in most cases, there are such
derived words in Fongbe. The principles underlying the concatenation of simplexes
into productive compounds constitute yet another case in point. As was extensively argued in chapter 11, the principles that govern productive compounding
in Haitian are those of its substratum languages and not of its lexifier language.
Idiomatic expressions discussed in Lefebvre (in press) are yet another example
of this type.
Thus, the data pertaining to the interpretive component of the grammar strongly
support the hypothesis that the creators of a creole use the principles of their
own grammar in establishing semantic interpretation in that creole.
13.3

Parameters

In chapter 2, it was hypothesised that the creators of a creole do not have

sufficient access to the superstratum language to acquire its parametric values.
Instead, they should use the parametric values of their own grammar to assign

13.3 Parameters

387

Table 13.7. Comparison of the parametric options in the three

languages under comparison
Availability of
(A) Verb raising to in (correlates with
inflectional morphology on the verb)
(B) Serial verbs (correlates with lack of
derivational and inflectional
morphology)
(C) Double-object constructions (correlates
with availability of Genitive Case
in nominal structures)
(D) Negative quantifiers as nps (correlates
with availability of bare nps)
(E) Verb-doubling phenomena (correlates
with the properties of the
determiner system)

fongbe

haitian

french

a value to the parameters of the language that they are creating. The hypothesis predicts that, where the parametric values of the substratum and superstratum differ, the creole should have the same parametric value as the substratum
languages.
With one exception, the three-way comparison in chapter 12 supports this
general hypothesis. Since syntactic clitics did not make their way into the creole,
the creators of Haitian had to reset the value of the null subject parameter: thus,
whereas both French and Fongbe are null subject languages, Haitian is not. As
was shown in chapter 12, the other parameters are formulated in terms of correlations between the availability of functional categories and a related syntactic
phenomenon. Hence, we saw that most parametric options set in the creole are
the result of its creators reproducing the properties of the functional categories
of their own lexicons through relexification. The correlations established in chapter 12 are summarised in Table 13.7.
As can be seen in Table 13.7, the parametric options of Haitian systematically contrast with those of French and follow those of substratum languages of
the type of Fongbe.
I therefore conclude that the hypothesis presented in chapter 2 is borne out
by the Haitian data. This conclusion is further supported by Koopmans (1986)
observation that other subsets of data, which can also be formulated in terms
of parametric options, show similar behaviour. For example, Koopman (1986)
remarks that, in Haitian, as in West African languages, headless and infinitival
relative clauses are not available. This contrasts with French, where both constructions are available. Koopman further points out that, in contrast to French,
where the set of phenomena referred to as quantifier float is available, Haitian
and West African languages lack such phenomena.

388
13.4

Evaluation of the hypothesis

Word order

In chapter 2 it was hypothesised that, in creole genesis, word order is established

as follows. Because the creators of the creole are aiming to reproduce the superstratum sequences they are exposed to and, since they are able to identify the
major category lexical entries, the word order of major category lexical items
and major constituents in the creole will follow that of the lexifier language.
However, because the creolisers do not have enough exposure to the superstratum
language, they cannot identify its functional category lexical items; when they
relexify the functional category lexical entries of their native lexicons, they keep
the original directionality properties. Hence, these items are predicted to have
the same word order as the substratum languages. The Haitian data presented in
this book show that this hypothesis is borne out.
For example, the position of modifiers with respect to their heads is predicted
to follow that of the lexifier language. As we saw in chapters 6 (section on Whphrases) and 11 (on compounds), while Fongbe adjectives always follow the
head noun, in French some adjectives precede the noun and others follow it.
As predicted by the hypothesis, in Haitian, the relative position of adjectives
and nouns follows the French pattern. Similarly, while quantifiers systematically
follow the head noun in Fongbe, they systematically precede it in French. In
accordance with the hypothesis, the position of quantifiers in Haitian follows the
French model, as is shown in (3).
(3)

a. vj

lx

fongbe

b. tous

les

enfants

french

c. tout
ti-mounn
q
det child
all the children

yo
pl

haitian
q

Likewise, in chapter 12, the position of adverbs in Haitian was argued to follow
that of French rather than Fongbe. Furthermore, Fongbe constituents whose order
conflicts with that of French were abandoned by the creators of the creole, as
evidenced by the fact that they have no counterparts in modern Haitian. Two
such cases were documented in this book: prenominal complements of nouns in
chapter 4, and prenominal complements of verbs in syntactic nominalisations in
chapter 5. In the same vein, Haitian creole is predicted not to have postpositions
because its lexifier language does not. As we saw, Haitian, like French, has a wide
range of prepositions and, unlike its substratum languages, it has no postpositions.
In contrast to the major category lexical entries of Haitian, which follow the
French word order, the minor category lexical entries have the same order as
the substratum language. Hence, creole determiners follow the head noun, like
Fongbe and unlike French, where determiners precede the noun (see chapter 4).
If the canonical relative order of the tma markers discussed in chapter 5 may be
argued to follow from universal principles of interpretation, along the lines of
the analysis in Woisetschlaeger (1977),5 the non-canonical relative order of these

13.4 Word order

389

markers and their relative order with respect to the negation marker in the creole
follow the details of Fongbe (see chapters 5 and 7). Furthermore, the Haitian
functional category lexical entries discussed in chapters 7 and 8 were shown to
follow the word order of Fongbe rather than of French where the two source
languages differ. This is true, for example, of the negation marker. The negative
marker and the determiner occurring in the Haitian clause have no counterparts
in French. Their positions in the clause are the same as those of the corresponding functional items in Fongbe. Thus, the claim that the directionality properties
of creole functional category lexical entries should be the same as those of the
substratum languages is borne out by the data.
Potential counterexamples to this conclusion include the position of yon, the
so-called indefinite determiner (see chapter 4), the position of se in clefts and in
some contexts of predication and the position of the yesno question word ske
(see chapter 7). In Haitian, these lexical items are constituent-initial, while the
closest functional items in Fongbe are constituent-final. In my view, the positions of these three Haitian words only constitute apparent counterexamples to
the above generalisation. Indeed, it has been argued that these lexical entries
were incorporated into Haitian as major category lexical entries. For example,
yon is a reduced form of younn one. The analyses of se reported on in chapter 7 indicate that, at least in some contexts, it occurs in a specifier position.
Typically, specifiers host major category lexical entries. The same argument goes
for ske, for it is possible to construct an analysis whereby ske, a Wh-word, fills
the specifier position of cp (see Lefebvre and Lumsden 1995). I therefore conclude that the positions of the three Haitian lexical items above do not constitute
counterexamples to the hypothesis concerning the directionality properties of
functional category lexical entries. On the contrary, assuming that they were
incorporated into Haitian as major category lexical entries, their position in the
clause is predicted to follow that of the French phonetic matrices from which
they were phonologically derived. This is the case: se and ske occur at the
beginning of constituents just like the French phonetic sequences that they were
derived from. As for yon, it occurs at the beginning of the nominal constituent
because it is derived from a numeral and numerals precede the noun.
The word order facts summarised above suggest that a creoles word order
constitutes a principled compromise between its source languages. This is not
the end of the story, however, for, as we saw in chapter 10, the position of the
morphological head in Haitian has been set as the rightmost position of the
word. This choice departs from both of the contributing languages, which allow
the non-rightmost morphological constituent to be the head of a word. This type
of innovation was not predicted by the hypothesis. Rather, it was revealed by a
detailed analysis of data pertaining to the morphological component of the grammar. Whether this parametric choice is specific to Haitian or whether other creoles have also made this choice is a question open for future research. Finally,
the variable order of the Theme and Recipient in the Fongbe double-object
construction was not carried over into the creole (see chapter 9). This was

390

Evaluation of the hypothesis

analysed as a consequence of the fact that the adjacency condition on Case

assignment has to operate at S-structure in Haitian but not in Fongbe. This word
order fact in Haitian was shown to be independent of its two source languages.
Aside from the innovations mentioned above, the general proposal in chapter 2, according to which, in creole genesis, word order is established on the basis
of what the substratum speakers perceive in the superstratum language surface
structure, is borne out by the Haitian data. Further evidence for this claim comes
from Berbice Dutch. Kouwenberg (1992) reports that Eastern Ijo, Berbice Dutchs
main substratum language, is underlyingly an ov language. Dutch, the lexifier
language, is also underlyingly an ov language. Berbice Dutch itself is a vo language. Kouwenberg explains this situation as follows. In Dutch simple clauses,
the verb moves to in such that, at surface structure, Dutch simple sentences
exhibit the order svo. According to Kouwenberg, the creators of Berbice Dutch
perceived this word order and hence established the word order svo for the creole.
As noted by Kouwenberg (1992), however, Berbice Dutch has postpositions.
According to the proposal in chapter 2, this should come as no surprise, since
Dutch also has postpositions. The fact that Saramaccan has postpositions (see
Muysken 1987) when its English lexifier language does not, however, constitutes
a counterexample to the claim in chapter 2 and requires further investigation.
13.5

Further questions

This section addresses further questions pertinent to the creole genesis scenario
posited in this book. For example, why are the Kwa features dominant in Haitian
and what about those of the Bantu languages (section 13.5.1)? Section 13.5.2
addresses the question of the homogeneity of the substratum languages. Section 13.5.3 discusses in more detail the role of dialect levelling.
13.5.1

Why are the Kwa features dominant in Haitian and what about
those of the Bantu languages?

This section addresses Mufwenes (1986: 136) observation about the formation
of Haitian creole: why . . . the putative Kwa features have been selected over
their Bantu counterparts.
As we saw in the previous sections, the Haitian and Fongbe lexicons and
grammars show similarities and parallels that go far beyond what chance would
predict. Recall from chapter 3 that Fongbe is one of the Gbe languages, which
in turn are members of the Kwa family. The historical data reported on by
Singler (1996) show that, at the time Haitian creole was formed, the bulk of the
African population in Haiti was of Gbe origin (see Table 3.4). In view of this
constellation of facts, it should come as no surprise that the Gbe languages
had an important input into Haitian creole.
Now, according to the historical data in chapter 3, there were also speakers of
other Kwa languages present at the time Haitian creole was formed. Recall also

13.5 Further questions

391

from chapter 3 that Gbe and other Kwa languages share an important number of
typological features. So the Kwa lexicons and grammars were close enough to
each other that, when these various lexicons were relexified, the output constituted a coherent whole, as we have seen throughout this book. Likewise, since
the parametric values of these languages appear to be similar, it should not be
surprising that the parametric values of Haitian are what they are.
Mande, Gur, West Atlantic languages etc., have also been shown to share
a significant number of typological features with the Kwa languages, and thus
when these lexicons and grammars were reproduced in the early creole they
were not very different from the incipient creole developed on the basis of the
Kwa languages. Furthermore, some of the differences between the substratum
languages, whether Kwa, or Mande, etc., have been reproduced in the creole,
as we saw in several instances (see e.g. reflexive forms, chapter 6; availability
of a nominalising affix, chapter 10). Some differences have been levelled out
(e.g. the plural marker, chapter 4), some have not (e.g. the clausal determiner,
chapter 8; the two types of morphological shape of the copy in verb-doubling
phenomena, chapter 12).
Given these considerations, my answer to the question Why non-Bantu rather
than Bantu features? is twofold: first, non-Bantu languages share a significant
number of typological features, which, abstracting away from the phonological
representations of lexical entries, makes them typologically similar; second,
speakers of these languages outnumbered Bantu speakers (see chapter 3) at the
time Haitian creole was formed. These two factors, internal and external, provide a principled explanation for why Haitian creole is so much like the West
African non-Bantu languages.
According to Singlers figures in chapter 3, Bantu speakers were predominant
in Haiti before and after the formative period of Haitian creole, and there were
some Bantu speakers in Haiti at the time Haitian was formed. I assume that, like
the other members of the African community in Haiti, these speakers were also
relexifying their lexicons on the basis of French. Recall from chapter 3 that
scholars who have worked on both Bantu and West African languages claim that
Bantu and Kwa, for example, share a number of properties that make them
typologically more similar than they appear on the surface. Assuming that this is
a correct characterisation of the situation, the core of the relexified lexicons of
the Bantu speakers would not be so very different from those of speakers of the
non-Bantu languages. Because relabelling proceeds on the basis of free morphemes in the superstratum language, Bantu speakers had, however, to abandon
their complex inflectional system. Furthermore, they had to comply with the lexicon of the majority of the early Haitian creole speakers.
Like other substratum languages, the Bantu languages probably contributed
some features to the creole through relexification. For example, an area of the
lexicon which probably attests to the contribution of Bantu languages is the
range of verbs participating in the double-object construction. Recall from chapter 9 (section 9.11) that Fongbe only has four verbs that can participate in this

392

Evaluation of the hypothesis

construction. According to Maria Polinsky (p.c.), Bantu languages generally use

double-object constructions with the following verbs: offer, sell, pay, ask,
tell, demand, write, send, promise, deny, show, teach. This list is much
closer to the Haitian one in chapter 9 (section 9.11). A detailed comparison of
Haitian, Fongbe and a Bantu language (attested to be present at the time Haitian
creole was formed) is also called for if we want to establish the Bantu contribution to the creole.
The biggest problem the Bantu speakers were faced with probably involved
establishing for themselves the parametric values of the new language that they
were creating together with the African speakers of the non-Bantu languages.
Indeed, recall from chapter 3 that, unlike Kwa languages, Bantu languages are
agglutinative. Building on a proposal by Givn (1971), Baker (1991) has proposed that the difference between the two language families is that, in the
former, the verb moves up to in (collecting affixes as it moves up) while, in
the latter, it does not. In chapter 12, we saw that, in Haitian and Fongbe, verb
raising is not available; this correlates with the lack of inflectional morphology
in these languages. Thus a likely hypothesis is that, during creole genesis, Bantu
speakers who had abandoned their complex inflectional system had to reset the
value of their own grammar for the verb raising parameter in order to conform to
the value set by the majority of speakers. Furthermore, recall from chapter 12
that a correlation was proposed between the non-availability of verb raising and
the availability of the serial verb construction. This correlation was shown to be
supported by Haitian and Fongbe, in which verb raising is not available while the
serial verb construction is. It is also supported by the Bantu languages, where
verb raising is available but the serial verb construction is not. Thus, here again
Bantu speakers who were participating in the formation of Haitian had to reset
the value of their own parameter to conform to the majority grammar. This is a
likely hypothesis given what we see in modern Haitian. Much more work is
needed before a precise characterisation of the Bantu speakers role in Haitian
creole genesis can be provided. I believe, however, that the general view outlined above goes in the right direction.
The types of adaptation that the Bantu speakers had to make in order to comply
with the language of the majority of speakers of the early creole constitute cases
of dialect levelling. It appears that, in the process of creating Haitian creole, Bantu
speakers had to make more compromises than speakers of other languages because
their language family was typologically farthest from the others present at the time
Haitian creole was formed. This brings me to a discussion of the homogeneity of
the substratum languages as a factor in the nature of the relexified lexicons.
13.5.2

The homogeneity of the substratum languages

The substratum languages of Haitian were quite homogeneous from a typological

point of view (see chapter 3). The Bantu languages were shown to present some
significant differences which had to be abandoned as a result of dialect levelling

13.5 Further questions

393

(section 13.5.1). Although I know of no situation where creole languages have

been created from totally disparate substratum languages, one might wonder
what would happen in such a case. On the basis of the theory of creole genesis
developed in this book, I hypothesise that, in such a situation, speakers of the
substratum languages would also relexify the lexicons of their respective languages using data from the superstratum language. This would provide them
with a common vocabulary. The relexified lexicons would reproduce the idiosyncrasies of the substratum languages and thus, in this case, would be more disparate than the early Haitian ones, since they would have been produced on the
basis of very different lexicons. In such a case, I hypothesise that dialect levelling
would play an important role in levelling out differences between the relexified
lexicons which could hinder communication between speakers of the newly created language. In this view, the importance of dialect levelling in the genesis of a
given creole is a function of the lack of typological congruence between the substratum languages. The more congruent the substratum languages, the less burden
there is on dialect levelling for mutual understanding purposes. Conversely, the
less congruent the substratum languages, the more speakers of the early creole
will have to negotiate in order to come to an agreement, through dialect levelling, on disparate lexical entries. Assuming that this scenario is on the right track,
the role of dialect levelling in creole genesis needs to be further characterised.
13.5.3

More on dialect levelling6

While relexification and reanalysis are cognitive, and hence individual, processes,
dialect levelling is clearly a social process. It consists in negotiation between
speakers of different dialects aimed at setting the properties of, for example, a
lexical entry. Dialect levelling is thus different in kind from the two other processes (relexification and reanalysis) hypothesised to play a role in creole genesis.
While the purpose of relexification (and eventually reanalysis) is to provide
speakers of different languages with a common vocabulary, the purpose of
dialect levelling is to reduce variation between the various dialects produced
by relexification. Now, from community studies (see e.g. Labov 1972; Sankoff
1980), we know that dialectal variation is a property of linguistic communities
and that this variation generally does not hinder communication between the
members of a given community. This statement should carry over to creole languages created on the basis of relatively homogeneous substratum languages
(but not to the hypothetical case discussed above).
Since variation is a normal characteristic of linguistic communities, why should
dialect levelling play a role in the development of creoles? Further research is
required to answer this question. Below, I shall merely list a few avenues for
further research on the basis of the data discussed in this book. The lexical
entries identified as having been subjected to dialect levelling in this book (see
Tables 13.1, 13.3, 13.5, 13.6) would probably not have hindered communication between speakers of the early Haitian dialects. If this is correct, could it be

394

Evaluation of the hypothesis

that the function of dialect levelling is related to social cohesion rather than
communicative necessity?
Another question that should be addressed is why some forms are levelled
out (e.g. cases of dialect levelling) and others are not (e.g. the determiner in the
clause, chapter 8). Another dimension of this question is why some substratum
forms appear to have been levelled out (i.e. abandoned) in the early creole (e.g.
the plural marker form distinct from the third-person personal plural discussed
in chapter 4) while others entered the creole and underwent dialect levelling
over a longer period of time (e.g. the reflexive forms discussed in chapter 6).
Another issue has to do with the winners and losers in the competition.
We saw that Fongbe speakers were quite successful with the determiner in the
clause (see chapter 8). The Ewe speakers won on the third-person plural pronoun that also serves as a plural marker in noun phrases. Why is it not always
the same group that comes out a winner? What are the key components of dialect levelling?
Further discussion of these points would go far beyond the scope of this
book. I raise these questions in the hope that they will be addressed in future
research.
13.6

Overall evaluation of the hypothesis

Both historical (external) (see chapter 3) and linguistic (internal) evidence points
to the conclusion that Haitian creole was created by adult native speakers in
possession of mature lexicons and grammars. The data presented throughout
this book massively support the claim that these adult native speakers used the
properties of their lexicons and grammars in creating the creole. The division of
properties found in the creole argues that the genesis of creole languages is a
particular case of second language acquisition in a context where the substratum
speakers have little exposure to the superstratum language. This explains why
substratum speakers rely on relexification to create a new lexicon, and on the
principles and parametric values of their own grammar to establish the grammatical properties of the new language they are creating. As we saw above, this
strategy affects all components of the grammar. Using the properties of ones
own lexicon and grammar in order to create a new language rapidly is the
most economical way of doing it. The claim by Muysken (1994b: 24) that lexical properties of content words survive, properties of function words and other
grammatical patterns do not is definitely not borne out by the data presented in
this book. The data and analyses presented here support the claim that pidgin
and creole genesis can be accounted for in terms of the basic processes already
known to play a role in language change in general: relexification, reanalysis
and dialect levelling. Moreover, these findings do not depend on the grammatical
framework within which they are presented: a systematic comparison of Haitian
with its contributing languages performed within the framework of any other
model of grammar would yield the same results.

14 Theoretical consequences
The fact that it can be demonstrated that relexification plays a role in the formation of various types of languages (e.g. mixed languages, pidgins, creoles) argues
that this is a process available to human cognition. It is a means of creating
new languages in a relatively short time. It may apply to part of the lexicon, as
in mixed languages, or the entire lexicon, as in pidgin and creole languages.
Whether it applies to part of the lexicon or the whole lexicon depends on the
motivation for creating a new language and the overall situation in which the
process takes place. As we saw in chapter 2, mixed languages are created in
bilingual communities with the motivation of creating an in-group language. By
contrast, pidgin and creole languages are created in multilingual communities
with the purpose of easing communication between groups that do not have a
common language, whence the necessity to relexify the whole lexicon. As a
cognitive process, however, relexification is independent from the various contexts in which it applies. The fact that it exists and the very nature of the process support Sproats (1985) and Prankas (1983) proposal that phonological
representations are stored independently in the brain.
The nature of relexification, and the facts that it is available to human cognition and is effectively used in the rapid creation of new languages, have consequences for the theory of the transmission and acquisition of lexicons in
situations where new languages, like creoles, are formed. Indeed, it is in the
nature of this process that lexical entries created in this way have phonological
representations derived from phonetic strings in the lexifier language (thus showing discontinuity) but syntactic and semantic properties derived from the substratum language(s) (thus showing continuity). On the surface, then, it looks
as if a totally new language has been created. In reality, however, the semantic
and syntactic properties of the new lexicon are those of the substratum language
lexicon(s). The properties of the original lexicon(s) are transmitted by adults
and acquired by children even when the latter are presented with a relexified
lexicon. Consequently, although situations where new languages are created by
relexification involve a break in the transmission and acquisition of a language,
there is no such break in the transmission and acquisition of semantic and grammatical properties (see Lefebvre 1993a, 1996b).
As has been pointed out by Hopper and Traugott (1993: 211), the linguistic
changes observed in the creation of pidgins and creoles call into question the
hypothesis that change occurs primarily in the transmission between generations,
395

396

Creole genesis and the acquisition of grammar

and is attributable primarily to children. The very nature of relexification requires

that those who apply it be adult native speakers in possession of a mature lexicon. Hence, the type of change resulting from relexification is initiated by adults.
The very fact that relexification exists as a cognitive process used to form
new languages poses a problem for the genetic classification of the languages
so formed. For example, Hall (1950: 203) classifies Haitian as a French dialect
on the basis of the phonological representations of its lexical entries: Haitian
Creole is to be classified among the Romance languages, and especially among
the northern group of the Gallo-Romance branch, on the basis of its systematic
phonological, morphological, syntactical and lexical correspondences. Goodman
(1964: 136) makes a similar statement: I do feel impelled to restate, however,
that on the basis of no purely linguistic criteria for genetic relationship which
have thus far been advanced, including that of parent syntaxique advanced by
Sylvain (see pp. 1212), can Creole French be classified with any specific language other than French. The data presented in this book, however, cast considerable
doubt on conclusions of this nature. Even though the phonological representations
of Haitian creole lexical entries can be associated with French phonetic strings,
Haitian shares its lexical properties, morpho-syntax, concatenation principles and
salient features of its parametric values with its substratum languages. Scholars
working on the reconstruction of language families should be aware that some
languages (and not necessarily those known as creoles) may have been created by
relexification, thereby straying from the normal course of gradual linguistic change.
Finally, the hypothesis of creole genesis presented here and supported by
Haitian data calls into question the assumption that all creole languages are
alike, as is advocated by Bickerton (1981, 1984). To the best of my knowledge,
this assumption was first challenged by Muysken (1988b) on the basis of a comparison of subsets of data drawn from various creole languages. In view of the
hypothesis presented here, I would like to go one step further and claim that
all radical creoles should show the division of properties between their source
languages argued to exist in Haitian. Therefore, such creoles should have lexical entries with phonological representations derived from phonetic matrices
of their superstratum language; the semantic and syntactic properties of these
lexical entries, as well as the principles of concatenation and parametric values,
should reproduce those of their substratum languages. The different properties
of pronouns in Solomons Pidgin (see chapter 2) and Haitian (see chapter 6)
cannot be attributed to chance. Likewise, while Solomons Pidgin has a predicate marker, Haitian does not. As argued by Lefebvre (1996b), the tma system
of Haitian differs from that of Tok Pisin. Furthermore, while Haitian has several constructions involving verb-doubling phenomena (see chapter 12), pidgins
and creoles of the Pacific do not. Why do we find these differences? Because
Haitian reproduces the properties of its West African substratum languages,
while Neomelanesian reproduces those of its Austronesian substrate (as argued
in Keesing 1988). Systematic comparisons of other creole languages with their
source languages should yield similar results.

Appendix 1 List of available

Haitian creole texts (17761936)

Year
1776a
1776b
1785
1790a
1790b
1791
1792
1796
1797
1802a
1802b
1802c
1802d
1810
1811
1824
1830
1854
1877
1886
1936

Source
Nicolson 1776: 567
Hillard dAuberteuil 17767.2: 68
Valdman 1978
Wimpffen 1817: 186
Debien 1962: 138
Gros 1791: 7
Anon. c. 1792: 615
Anon. 1796
Moreau de Saint-Mry 1797.1: 37, 65
Descourtilz 1809.2: 129, 303, 352; 3: 113353
Ducurjoly 1802.2: 283393
Anon. 1802 (Bonaparte)
Anon. 1802 (Leclerc)
Anon. 1810: 39, 185
Anon. 1821
Anon. 1824: 93
Bonneau 1856
Bigelow 1877
Audain 1877
Janvier 1886
Sylvain 1936

From Baker and Corne (1982: 2734)

397

Appendix 2 Phonemic inventories

and orthographic conventions

An extensive comparison of the phonology of Haitian with that of its source

languages and an analysis of how the Haitian phonological system acquired its
properties are being prepared by Brousseau (in preparation). In this section I
shall only provide basic information concerning phonemic inventories in order
to discuss the orthographic conventions.
Haitian
The phonemic inventory of Haitian vowels consists of seven oral vowels and
five nasal vowels (see Alleyne 1966; Cadely 1988a, 1988b, 1994; Anestin 1987;
Frre 1974; Dejean 1980; Goodman 1964; Brousseau 1994b).
(i)

inventory of vowels in haitian creole

a. Oral
b. Nasal
Front Central Back
Front Central
i
u
d
e
o
=
u
s
a

Back
q
p

According to the analysis in Brousseau (1994b), building on Ans (1968), Tinelli

(1970), Valdman (1978), Cadely (1988a, 1988b), Anestin (1987), Frre (1974),
Dejean (1980) and Goodman (1964), the consonant system of Haitian creole
comprises 20 phonemic consonants and 2 allophones, which appear within square
brackets in the diagram below. The consonant which appears within angle brackets
has been assigned two analyses in the literature and the notation reflects this
fact. The use of parentheses reflects controversial analyses for a given segment.
(ii)

398

inventory of consonants in haitian creole

Bilabial
Labio-dental
Alveo-dental
Alveo-palatal
Palatal
Velar
p
t
k
b
d
g
g
5
f
s
m
v
z
n
<>
m
n
()
[]
l
w
y [g]
<>
(=(6) in Brousseau 1994b)

Phonemic inventories and orthographic conventions

399

The orthographic convention used in this book follows official usage in Haiti.
(For an extensive discussion of various orthographic conventions for Haitian creole,
see Dejean 1980.) The vowels are written as in (i) with the following exceptions:
=
=
=
=
=
=
=
=

/u/
/=/
/u/
/d/
/s/
//
/f/
/p/

ou

in
n
an
oun
n

Note that, in the standard Haitian orthography, vowel nasality is represented

whether it is phonemic or phonetic. For example, /mfn/ person is spelled
mounn; the vowel /u/ is nasal by virtue of occurring in the context of the nasal
consonant [m]. The first [n] in this case indicates the nasality of the vowel and
the second is the final consonant of the syllable.
Orthographic conventions for consonants are the same as in (ii) with the
following exceptions:
=
=
=
=
=
=
=

/g/
/5/
/m/
/n/
()
[]
<>

tch
dj
ch
j
ny
g
r

[zeps] zepng

pin

Haitian has a complex stress system comprising primary and secondary stress
(see Cadely 1988a, 1994). Stress is not represented in the orthography.
French
The phonetic inventory of vowels for seventeenth- and eighteenth-century French
is as in (iii), reconstructed by Anne-Marie Brousseau using data from Laborderie
(1994), Pierret (1983), Van Daele (1929), Zink (1991), Bonnard (1982), Fouch
(1952), Joly (1995), Gendron (1966) and Juneau (1972). It comprises 12 phonemic
oral vowels (with 3 phonetic variants indicated within brackets) and 5 nasal vowels.
(iii)

inventory of vowels in seventeenth- and eighteenth-century

french
a. Oral
Front
i [I]
e
=

Front
Back
rounded
[] u [U]

o
1

u
7

b. Nasal
Front

s
h

Front
rounded

Back

p
\

400

Appendix 2

All authors agree that the bulk of this inventory had already been established by
the thirteenth century and that, by the end of the seventeenth century, the phonemic inventory of French vowels was as in (iii).
The inventory of seventeenth- and eighteenth-century French consonants in
(iv) was established on the basis of the literature (see Bonnard 1982; Fouch
1952; Gendron 1966; Joly 1995; Laborderie 1994; Pierret 1983; Van Daele
1929; Thurot 18813; Zink 1991). Contextually conditioned allophones appear
within square brackets. Consonants in parentheses are those which have a dubious status for the relevant period and dialects. The consonants within angle
brackets constitute alternative realisations of the phoneme /r/. Uvular [] does
not figure in this inventory because it is considered a superficial variant of
velar [r].
(iv)

inventory of consonants in french

Bilabial Labiodental
p
b
f
v
m
w/2

Alveodental
t
d
[ts ]
[dz]
s
z
n
l, <r>

Alveo- Palatal Velar

palatal
k
[ky ]
g
[gy ]
[g]
[5]
m
n

()
<r>
y

Laryngeal

(h)

There is a consensus among the above-mentioned authors on the phonemic/

allophonic status of the segments in (iv).
In this book, French words and sentences are written in current standard
orthography. As for prosodic patterns, French presents a very simple stress
system, and stress is not represented in standard orthography.

Correspondences between French phonetic matrices and Haitian

phonological forms
The hypothesis outlined in chapter 2 states that the phonological forms of
Haitian lexical entries were established on the basis of French phonetic matrices.
In Brousseau (in preparation), it is extensively argued that the creators of Haitian
interpreted the French phonetic matrices they were exposed to on the basis of
their own phonological system. For example, French [1] was interpreted as /i/
or /e/, as shown in (v), which provides a list of the correspondences between
Haitian and French vowels.

Phonemic inventories and orthographic conventions

(v)
i

haitian
[dipi]
[simity=]
1 [seriz]
[epwum]
[alim=t]
[mi]
[nis]
[nus]
[suse]
[bule]
[dife]
[ve]
[agyo]
[av=g]
[p=p]
[vul=]
[5ul]
r [brs]

<

e <
i

<

u <
e <
o <
= <

s <

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<

french
depuis
cimetire
cerise
reproche
allumette
mur
juste
juste
sucer
brler
feu
vu
adieu
aveugle
peuple
voleur
gueule
brun

401

[d(1)p2i]
since
[sim(1)ty=r] cemetery
[s(1)riz]
cherry
[r(1)prum]
reproach
[alm=t]
match
[mr]
wall
[js(t) ]
exactly
[js(t) ]
exactly
[sse]
to suck
[brle]
to burn
[f]
fire
[v]
vow
[ady]
farewell
[avgl(1) ] blind
[ppl(1) ]
people
[vulr]
thief
[gl(1) ]
mouth
[brr]
brown
(=(4) and (5) in Brousseau 1994b)

Correspondences between Haitian and French consonants are as in (vi).

(vi)

<

<
w
<
/g<

r
r
2


<h

haitian
[g]
[mame]
[wit]
[kp]
[kpg]
[ele]

<
<
<
<

french
grand
march
huit
campagne

[gr]
[marme]
[2it]
[kpa]

big
market
eight
countryside

<

hler

[hele]

to call

Fongbe
The phonemic inventory of Fongbe vowels consists of seven oral and seven
nasal vowels, including two high nasal vowels (see Brillon and Brousseau 1986;
Capo 1991).
(vii)

inventory of vowels in fongbe

a. Oral
Front
i
e
=

Central Back
u
o
u
a

b. Nasal
Front Central Back
d
f
9

s
p

(from Capo 1991)

The inventory of consonants in Fongbe is as in (viii), established on the basis

of work by Brillon and Brousseau (1986), Capo (1991) and Brousseau (1994b).
Consonants within square brackets are phonetic variants which are contextually
determined. The inventory of Fongbe segments comprises 22 phonemic consonants and 3 allophones.

402
(viii)

Appendix 2
inventory of consonants in fongbe
Bilabial

Labiodental

f
v
b [m]
w

Alveodental
t
d
s
z
3 [n]
l

Alveopalatal

Palatal

g
5


Velar
k
g

Labiovelar
kp
gb

xw
w

y [g]

The orthographic convention used in this book follows the official usage in
Benin (see Publication of the Centre national de linguistique applique (cenala)
1990). Oral vowels are written as in (viia). Nasal vowels are represented as in
(viib) except that nasalisation is expressed using an n following the vowel rather
than a tilde on the vowel. Note that, in contrast to the Haitian orthography, only
phonemic nasal vowels are represented as such. Hence, [sb] crab is spelled
sVn. By contrast, [mc] is spelled mW to see. The consonants are written as in
(viii) with the following exceptions:
/g/
/5/
//
//

=
=
=
=

c
j
ny
h

Fongbe is a tone language. In the standard orthography, lexical tones are

represented. High tone is represented by an acute accent, low tone by a grave
accent and lowhigh or highlow tone by a combination of the two. An extensive discussion of the processes affecting the quality of lexical tones may be
found in Brousseau (in preparation).

Appendix 3 Sample of nonmatching derived words in

Haitian and French

In the examples below, the first column gives the Haitian derived word with its
meaning. The second column gives the Haitian word from which the word in
the first column is derived. The third column provides the French counterpart
of the Haitian derived word, when available. The Haitian data are from Valdman
et al. (1981). The French data (or absence thereof) come from dictionaries of
seventeenth-century French: Furetire (1984) and Dubois et al. (1992).
The data in (A) exemplify Haitian words derived by means of the agentive
affix -. In most cases, either the corresponding French word has a different
meaning from the Haitian or there is no such form.
(A)

Words derived by means of the agentive suffix -

haitian
beny-
undertaker

<beny-en
to bathe

17th-century french
baign-eur
swimmer / one who owns
bath house / spa

modern french

djl-
talker

<djl-e
to chitchat

gueul- eur
loud talker

estat-
starter

<estat
to start

kmand-
boss

<kmand-e
to order

command-eur
commander (of order
of knighthood)

The data in (B) are Haitian words derived by means of the attributive affix
-. French has no counterpart for any of these derived Haitian words.
(B)

Words derived by means of the attributive suffix -

haitian
bresony-
tippler, lush (drinker)

<breson
alcoholic beverage

17th-century french

eskandal-
loud, rowdy

<eskandal
scandal, uproar

odyans-
entertaining person

<odyans
funny stories,
jokes, banter

403

404

Appendix 3
rans-
joker, frivolous person

<rans
joke

soud-
deaf person

<soud
deaf

tafya-t-
drunkard

<tafya
alcoholic beverage

tanbouy-
drummer

<tanbou
drum

wanga-t-
magician

<wanga
fetish

The data in (C) exemplify Haitian verbs derived by means of the verbalising
affix -e. Again, when French has a counterpart to the derived Haitian word, the
two derived verbs do not have the same meaning.
(C)

Words derived by means of the verbalising affix -e

haitian

17th-century
french

modern
french
entraver
to fetter, to
shackle, to
hobble, to clog

antrav-e
to implicate,
to corner
(in a tight spot)

<antrav
problem

betiz-e
to joke, to work
in vain, to deceive

<betiz
obscenity, nonsense

bourik-e
<bourik
to work like a dog donkey, work horse

bourad-e
to shove

<bourad
push, boost (help)

brak-e
to sweeten
lightly

<brak
slightly sweetened

braquer
to position
a cannon

braquer
to position
a cannon

chyen-t-e
to beg, to fawn

<chyen
dog,
obsequious person

driv-e
to drift, to insult

<aladriv
drifting (idle),
abandoned

graj-e
to grate, grated

<graj
grater

kalkil-e
to reflect

<kalkil
reflexion

calculer
to count

calculer
to count

kle-t-e
to lock

<kle
key

pwp-t-e
to clean

<pwp
clean

Non-matching derived words in Haitian and French

rans-e
to joke, to dally,
to toy, to trifle

<rans
joke,
stupid remark

sabo-t-e
to slap

<sabo
slap

saboter
to clatter

salp-e
to dirty

<salp
slovenly

saloper
to botch,
to bungle

tchans-e
to speak idly,
to delude oneself

<tchans
wishful thinking,
useless talk

tk-e
to hit a marble

<tk
hit (in marbles)

405

The data in (D) provide examples of words derived by means of the inversive
prefix de-. Again, in most cases, there is no corresponding word in French.
When French does have a derived word, its meaning is different from that of the
Haitian word.
(D)

Derived words containing the inversive prefix dehaitian

d-gwosi
to lose weight, to smooth
out, to rough-hew

<gwosi
to put on weight,
to grow larger

17th-century french
dgrossir
to rough down,
to rough-hew,
to rough-plane

de-makonnen
to untangle,
to disentangle

<makonnen
to tangle

de-respekte
to be disrespectful

<respekte
to respect

de-take
to unlatch

<take
to latch, to bolt

de-z-apiye
to stop leaning

<apiye
to lean on

The data in (E) are examples of Haitian words containing the diminutive
prefix ti-. None of these words has a corresponding French word.
(E)

Derived words containing the prefix tihaitian

ti bezwen
genitals of small boy

17th-century french

tibt
insect, bug

ti bourik
rude person

406

Appendix 3
ti devan
sex, genitals

tifi
virgin

tigasn
young male servant

ti granmounn
precocious little girl

timounn
child

ti nouris
new mother

The Haitian words in (F) are derived by means of the suffix -ay. Again, there
is no French word corresponding to these derived words.
(F)

Words derived by means of the nominalising affix

haitian
banbil-ay
bash (party)

<banbile
to carouse

17th-century french

randui-z-ay
coating

<randui
to coat

boukan-t-ay
swap

<boukan-t-e
to swap

pary-ay
bet (wager)

<pary-e
to bet, to wager

plas-ay
common-law marriage

<plas-e
to live together

sweny-ay
good care

<swen
to care for, take care of

The Haitian words in (G) are derived by morphological conversion. Only a

few of them have a corresponding French noun derived by this process.
(G)

Cases of nominal conversion

haitian
ale
sti
antre
vini
rive
tonbe
desan
monte
tunen
pati
pwche

action
action
action
action
action
action
action
action
action
action
action

or
or
or
or
or
or
or
or
or
or
or

result
result
result
result
result
result
result
result
result
result
result

of
of
of
of
of
of
of
of
of
of
of

leaving
going out
entering
coming
arriving
falling
descending
going up
returning
leaving
approaching

17th / 18th-century french

aller
the going

venir
the coming

partir
the leaving

Non-matching derived words in Haitian and French

gunmen
kuri
mache
dmi
manti
bwte

action
action
action
action
action
action

or
or
or
or
or
or

result
result
result
result
result
result

of
of
of
of
of
of

fighting
running
walking
sleeping
lying
moving

marcher
dormir

407

the walking
the sleeping

Finally, (H) shows adverbs derived by means of the affix -man. These
derived adverbs have no counterpart in French.
(H)

Words derived by means of the adverbial suffix

haitian
angran-man
haughtily, imperiously

<angran
arrogant, haughty

mal-man
not so well, rather badly

<mal
badly

17th-century french

Notes

The problem of creole genesis and linguistic theory

1 The contents of this section build on a preliminary discussion in Lefebvre and Lumsden
(1989a, 1994a).
2 There is a general consensus in the literature that multilingualism is a required feature
of communities where creoles may emerge. However, this proposal has recently been
challenged by Smith, Robertson and Williamson (1987), who claim that Berbice Dutch
emerged out of contact between only two languages: Dutch and Eastern Ijo. Assuming
that Berbice Dutch is a true creole (as opposed to a mixed language, see chapter 2) and
that Ijo was the sole African language present at the time this creole was formed, this
case would constitute the first documented evidence against Whinnoms widely accepted
claim. For further discussion of this issue, see also Foley (1988).
3 For a discussion of competing theories of creole genesis with respect to this inventory
of properties, see Lefebvre and Lumsden (1989a).
4 For a discussion of these three approaches, see Arends, Muysken and Smith (1995).
5 French aprs (after) is part of a periphrastic expression used to encode the progressive. tait is the imperfect form of the verb to be. Avoir means to have; va is the
third person of the verb to go and is also part of a periphrastic expression used to
encode the future; avant is an adverb meaning before. The French expression fait que
may be translated as just in the sense referred to in the quotation. The facts discussed
in this quotation are examined in detail in chapter 5.
6 See e.g. Lefebvre (1982a, 1984, 1986, 1993a, 1994a, 1996a, 1996b); Lefebvre and
Kaye (19859 Projects); Brousseau, Filipovich and Lefebvre (1989); Lefebvre and
Lumsden (1989a, 1994a).
7 Major category lexical items are nouns, verbs, adjectives, prepositions and adverbs. They
are defined by the major categorial features [n, v] (see Chomsky 1972; Jackendoff
1976). In current theories (see Lieber 1980, and the references therein), derivational
affixes are also specified for major categorial features. Functional or minor category
lexical entries are defined by minor syntactic features such as Tense, Wh, etc. (see Den
Besten 1978). They include Tense markers, Wh-words, determiners, etc.
8 In a more recent paper written in collaboration with Norval Smith (Muysken and Smith
1990: 884), Muysken amends his earlier claim, allowing for relexification in language
genesis only in bilingual situations: We reject the gradual relexification of believers
in monogenesis (from a West African Portuguese Pidgin) or Afrogenesis, in situations
of communal linguistic confrontation between, e.g. a European planter class and an
African slave class. We do accept the possibility of relexification as a mechanism in
forming a new language in a bilingual situation.

408

Notes to pages 1653

2

409

Cognitive processes involved in creole genesis

1 The term target language might not be the most appropriate term given the various
circumstances in which relexification applies. Mhlhusler (1986a), Thomason and
Kaufman (1991) and Baker (1990) have discussed this issue extensively. In this text,
I will use the term lexifier language instead.
2 Note, however, that, while Michif da/d is a preposition, its Cree counterpart is a
postposition. The question of word order in relexification will be discussed in section
2.5.
3 For an extensive discussion of the properties of mixed languages, see e.g. Thomason
and Kaufman (1991) and Bakker and Mous (1994a). For case studies of a large sample
of mixed languages, see the papers in Bakker and Mous (1994b).
4 For a preliminary discussion of this topic, see Lefebvre and Lumsden (1989a).
5 For a discussion of the relationship between calquing and relexification, see Allsopp
(1980).
6 For competing views on this matter, see the papers in Andersen (1983b).
7 Saramaccan is an English-based creole which also has Fongbe as one of its substratum languages (see Smith 1987) and is claimed to be even closer to Fongbe than
Haitian is (see Muysken 1994a).
8 The functional category lexical entry wY basically introduces new information. A
description of the properties of this lexical entry can be found in Lefebvre (1992a)
and it will be further discussed in chapter 7.
9 According to Mhlhusler, Tolai is one of the substratum languages of Tok Pisin.
10 As will be seen in chapter 3, Fongbe is one of the substratum languages of Haitian.
11 Since then, a new theory of word order has been developed by Kayne (1994) whereby
all languages have the same underlying word order. The data presented in this book
are not discussed within Kaynes framework.
12 I am indebted to Christine Jourdan for finding these data for me.
13 The time span covered by the term early creole remains to be determined.
14 For discussions of this point based on various cases of dialects in contact, see Sankoff
(1984); Trudgill (1986: 98); Gambhir (1988: 77); Singler (1988); Thomason and
Kaufman (1991); Harris (1991: 199); Siegel (to appear).
15 The criteria governing the selection between competing forms require further research
(see Mufwene 1986, 1990). For a preliminary discussion of this point, see Siegel
(to appear).
3

The research methodology

1 The historical research was designed during the 19857 and 19879 Projects with the
help of Caroline Fick, a historian at the Centre de Recherches Carabes (Universit de
Montral), and the collaboration of John Singler during the Linguistic Institute of the
Linguistic Society of America (New York, Summer 1986). An outline of this research
programme was presented at the Conference on Africanisms in Afro-American Language Varieties (University of Georgia, February 1988) and published in the proceedings of the conference (see Lefebvre 1993a). The actual research was carried out by
John Singler during the third year of the 198994 Project. For a detailed report on his
findings, see Singler (1993b, 1996).

410

Notes to pages 5369

2 Carden and Stewart (1988: 17) also state that there was an established creole in Haiti
at least by 1740 or 1750.
3 According to Singler (1996: 189), the engags were indentured servants from France
employed for three years. Together with the population of colour, they formed the
agricultural labour force.
4 The censuses studied by Singler are the Guadeloupe census of 1664 and the Martinique
census of 1680. According to Singlers (1993b: 171) description, the Guadeloupe
census provides a name and/or an ethnic designation for roughly 2,100 of the slaves
being counted, while the Martinique census does so for 3,900. Because of this detail,
they have the potential to provide valuable evidence as to the ethnic composition of
French Caribbean colonies in this earlier period.
5 The Remire inventory is Goupy des Marets 1690 inventory of a plantation at Remire
in French Guyana. For a detailed discussion of this inventory, see Singler (1993b).
Mallet in Table 3.3 refers to a Muslim brought to the Bight of Benin from the
interior. Singler (1996: 205) presumes that the person so designated spoke either a
Mande or a Gur language.
6 The more recent classification of African languages in Bendor-Samuel (1989) groups
these languages differently. I am using the earlier classification rather than the more
recent one only because it is more congruent with the literature of the 1970s and
1980s that I am reviewing, and because it contains more information on morphosyntax.
7 The bulk of Koopmans (1986) paper is dedicated to showing that verbs in West African
languages (and in Haitian), taken as a group, have a large number of common properties.
8 Westerman and Bryan (1970) mention a group of isolated languages spoken in
Togoland which exhibit class affixes and concord systems much like those of the
Bantu languages.
9 Although French is the main lexifier language of Haitian creole, the Haitian lexicon
also contains words of other origins. Amerindian languages spoken on the Island at
the time of the early Haitian colony contributed some 200 nouns designating local
objects such as plants and place names (see Hilaire 1992). A few words of Spanish
origin (see Hilaire 1992) and a few borrowings from English are also part of the
Haitian lexicon. Most importantly, there are about 350 modern Haitian words which
have forms and meanings similar to those of words found in one or several of the
substratum languages of Haitian (see Hilaire 1993). These constitute cases of retention
from the West African languages spoken in Haiti at the time the creole was formed.
10 For extensive discussions of these issues, see Mougeon and Beniak (1994) and the
papers and references therein.
11 The features of an ideal versus a feasible test, due to time and resource limitations,
are discussed in the proposals for grants to support research (see Lefebvre and Kaye
19857, 19879 Projects; Lefebvre 198994 Project) and in papers reporting on the
methodology adopted for the research (see e.g. Lefebvre 1986, 1993a; Lefebvre and
Lumsden 1989a, 1994a).
12 The choice of the substratum language to be studied in detail goes back to joint work
with Jonathan Kaye in the 19857 Project. The historical research was conducted
during the third year of the 198994 Project (see the Preface).
13 Hair (1966, 1970) provides a list of available early West African linguistic material,
mainly vocabularies. He cites one reference on a Gbe language: Labouret and Rivet
(1929). This book contains a short list of Gen words and prayers in Gen written by
Spanish missionaries of the seventeenth century.

Notes to pages 7093

411

14 Thanks to Ans de Kok for sharing with me her bibliography on seventeenth- and
eighteenth-century French.
15 Alembert et al. (17767); Bary (1665); Bonnard (1982); Brunot (1905, 1926); Catach
(1995); Darmesteter (1875); Dees (1971); Dubois et al. (1992); Fraud (1768); Fouch
(1952); Furetire (1984); Galet (1971); Gauthier (1995); Gougenheim (1973); Guiraud
(1966); Haase (1965); Huguet (1925); Imbs (1979); Joly (1995); Juneau (1972);
Laborderie (1994); Nyrop (1936); Oudin (1640); Pierret (1983); Rey (1992); Robert
(1985); Rosset (1911); Spillebout (1985); Thurot (18813); Van Daele (1929); Vaugelas
(1647); Wagner and Pinchon (1962); Wartburg (1971); Zink (1991).
16 The semantics of verbs is the area of the lexicon where I found the most differences
between seventeenth- and eighteenth-century French and modern French. Thus, in
this case, the documentation of seventeenth- and eighteenth-century facts turned out
to be important for the three-way comparison. Due to space limitations, however,
these data will not be discussed here (see Lefebvre in press).
17 For a much more detailed discussion of this issue, see Koopman (1986: 24851).
18 Ans (1968); Faine (1937); Freeman (1988); Goodman (1964); Hall (1950, 1953); Hilaire
(1992, 1993); Sylvain (1936); Valdman (1970); Valdman et al. (1981); Vdrine (1992).
19 Anonymous (1983); Rassinoux (1987); Segurola (1963); Welmers (1973); Westerman
(1907).
4

Functional category lexical entries involved in nominal structure

1 For a discussion of the parallels between the allomorphy of the Haitian and Fongbe
determiners, see Brousseau (in preparation).
2 According to the data in Lumsden (1989), some Haitian speakers do not allow la and
yo to co-occur. Other speakers do accept such sequences, however, as illustrated in
(24).
3 The feature [+plural] may also be visible elsewhere in the French noun phrase due to
agreement in gender and number between the various components of its structure
(see e.g. Bernstein 1993).
4 Goodman (1964) points out that the French form eux is pronounced [yo] in Gascon
and Auvergnat French. However, these dialects were not represented in Haiti at the
time Haitian creole was formed, so it is unlikely that Haitian yo was phonologically
derived from [yo] in Gascon or Auvergnat French.
5 This is an interesting contrast, for in the literature such a contrast has not been
reported on. For example, in Valdman et al (1981), sa and sila are both glossed as
this/that (see also e.g. Joseph (1989), Goodman (1964), etc. for further discussions
of these data).
6 In other frameworks sa and sila would be considered as having pronominal properties. (Thanks to Dominique Estival for pointing this out to me.) If this were the
correct analysis for Haitian, however, the data in (38) could not be explained since
pronouns typically cannot be modified.
7 P. Thibault (p.c.) notes that, in colloquial French, cela is a general deictic term just
like a. Chaudenson (1993: 19) also points out that, in most varieties of French, a
and cela are equivalent in meaning.
8 Auger (1994) points out that there are two as in French: the strong form discussed in
this section and a clitic form, not discussed here. In a, a me plat This, it pleases
me, the first a is the strong form and the second one the clitic.

412

Notes to pages 96112

9 In Montreal colloquial French considered to be a conservative dialect as compared

with international modern French the demonstrative forms with l can be used to
designate objects that are either close to or far from the speaker (P. Thibault, p.c.).
10 This is an interesting contrast because in the literature on Fongbe /V and nX are
glossed as this/that respectively (see Anonymous 1983; Segurola 1963).
11 Wallace (1995d) and C. Collins (p.c.) point out that in Ewe, there are also two
demonstrative forms and that these forms bear the same features as the Fongbe terms.
12 Julie Auger (p.c.) points out that in Picard the form for French lui is also li.
13 For further discussion on this topic, see the forthcoming literature.
14 Case markers are functional lexical items which head the functional category projection K(ase)P (see Travis and Lamontagne 1992). In particular languages, they are
sometimes homophonous with prepositions, which are major category lexical items
(see Van Riemsdijk 1978). For example, of in English may be used either as a Case
marker or as a preposition (see Chomsky 1981). Syntactic tests distinguish between
the sometimes ambiguous syntactic functions of a given form (see e.g. Van Riemsdijk
1978; Rouveret and Vergnaud 1980; Lefebvre and Muysken 1988).
15 The identification of this functional category is a matter of debate in the literature.
For various proposals, see e.g. Abney (1987); Lumsden (1989, 1991); Ritter (1992);
Szabolcsi (1987).
16 Traditional grammars generally consider and de to be prepositions. Lumsden (1991)
considers them as Case markers. For a discussion of this issue, see Kayne (1975,
1981), Tremblay (1991) and the references therein.
17 For a different view on this matter, see Lumsden (1991).
18 I cannot yet explain why the subset of speakers in Lumsden (1991) allow only one
argument to occur after the noun, given that they also appear to have two null Case
markers in their lexicon.
19 According to Hazoums (1990) description of the Gbe cluster of languages, all these
languages encode possession similarly to Fongbe.
20 The data pertaining to the Northern dialect of Haitian call for further research.
For example, the properties of the possessed np introduced by a need to be studied.
Another question that must be addressed is how the details of these properties compare with those of the West African substratum languages which have a prenominal
possessive connector. We must also take into account the amount of exposure to
French that the African speakers in the North had in comparison with those in the
Centre and in the South (see chapter 3 for a preliminary discussion of this issue).
Possibly, French was acquired by the speakers in the North because they had both
more exposure to French and a connector of the type in (85) in their own lexicon. In
short, the Northern facts constitute a topic for further research.
5

The preverbal markers encoding relative Tense, Mood and Aspect

1 For a similar observation, see Bentolila (1971).
2 The inventory of tma markers in Haitian and Fongbe is established in Lefebvre
(1996b) on the basis of syntactic tests which set the preverbal markers apart from
modal and aspectual verbs. First, they all occur between the subject and the verb.
Second, preverbal markers occurring in the same column in (1) are mutually exclusive, showing that they are in a paradigmatic relationship. Third, while modal verbs
do allow for deletion of their vp complement, preverbal markers do not (for Haitian,

Notes to pages 112121

413

see Koopman and Lefebvre 1982; Magloire-Holly 1982; Spears 1990; for Fongbe, see
Lefebvre 1996b). Fourth, most of the preverbal markers in (1) have no meaning
outside of the tma system. Finally, the tma markers may combine to form complex
tenses. There appears to be a general consensus in the literature to the effect that,
although the preverbal markers have properties that distinguish them from verbs/
prepositions, they are not functional heads. Without further discussion, I will thus
assume that they are major category lexical items with restricted uses.
3 For competing views on this matter, see Damoiseau (1988); Dchaine (1991); Lumsden
(in press b).
4 Note that Damoiseaus (1988) classification of verbs in terms of their aspectual properties differs from other classifications such as the one in Vendler (1967). Dynamic
verbs in Damoiseaus account may denote telic and non-telic events. Resultative
verbs correspond to the class of achievement verbs in Vendler.
5 The Fongbe marker of anteriority k is homophonous with an adverb meaning
already (see Anonymous 1983; Avolonto 1992). As is pointed out in Lefebvre (1996b),
the claim that k has two functions is supported by two sets of facts. First, there can
be two occurrences of k within the same clause. In (i), the first occurrence of k
corresponds to its function as a marker of anteriority, and the second to its function
as an adverb.
(i)

Sk k
nw
k
3
wh.
Cica ant use-to already prepare dough
Cica had usually already prepared dough.

fongbe

(=(59) in Lefebvre 1996b)

Furthermore, as an adverb, k must immediately precede the vp, but, as a marker
of anteriority, it occurs as the first of a series of tma markers. For further discussion
on the distribution of k, see Lefebvre (1996b). The function of k as a marker
of anteriority must have been acquired through reanalysis. Westerman (1907: 139)
reports that k started as a (now obsolete) verb meaning be, have finished and that
it has been reanalysed as a past tense marker.
6 Da Cruz and Kinyalolo (1994) have established tests distinguishing between dynamic
and stative verbs in Fongbe. The data provided in Lefebvre (1996b) and in this chapter on the interpretation of the Fongbe tma markers in relation to the aspectual class
of verbs provide evidence for a class of resultative verbs in Fongbe. For example, the
temporal interpretation of a clause containing a resultative verb in the context of a
bare sentence is different from that of a clause containing a dynamic or a stative verb
in the same context (see section 5.3). This is evidence that resultative verbs constitute
an aspectual class distinct from dynamic and stative verbs in the language.
7 Julie Auger informs me that in Picard the form corresponding to French t is t.
8 It appears that the creators of Haitian did not select the locative preposition (see
(20b) ) or the periphrastic expression en train de (see (20c) ) as the French phonetic
form from which to derive an imperfective marker. The French-based creoles discussed in Goodman (1964) have all chosen the same French form as Haitian. This
raises the following question: what does aprs have that and en train de do not from
the point of view of the substratum speakers? This question should be addressed
within the framework of a general theory of second language data processing, a
theory not fully developed in this book (see chapter 2).

414

Notes to pages 122161

9 In Fongbe, wY has the basic and major function of introducing new information (see
chapter 7). The use of this postpositional item in the Fongbe imperfective construction is exceptional with respect to other languages of the same family. As is pointed
out in Fabb (1992), in the other Gbe languages, the imperfective aspect also makes
use of a locative preposition meaning at, but in these languages the preposition
selects a locative postposition meaning in rather than the equivalent of Fongbe wY.
10 For syntactic analyses of the imperfective constructions in Fongbe, see Fabb (1992)
and Kinyalolo (1992).
11 Be sick and know are both stative verbs. The former denotes a transitory property
of the individual, whereas the latter denotes a permanent property (see Diesing 1990).
This explains why be sick may occur with an aspectual marker that triggers a noncomplete interpretation of the event, while know may not.
12 An account of these facts will be provided elsewhere.
13 Anonymous (1983) reports that for some speakers the word order n k may also
yield a conditional interpretation.
14 I would like to thank Dominique Estival for fruitful discussions on this point.
15 The dictionaries consulted are: the Grand Larousse de la langue franaise, the Grand
Robert, Littr, Quillet, Logos, Grand dictionnaire de la langue franaise. Trsor de la
langue franaise, Furetire.
16 Westerman (1907: 139) reports that, in Gbe, the verb nW remain, stay was reanalysed as
a preverbal habitual marker. Evidence for this claim is discussed in Lefebvre (1996b).
6

Pronouns
1 As has been pointed out to me by Julie Auger, even the dialects of French, such as
Picard, which do not distinguish between first- and second-person plural clitics do
distinguish between first- and second-person strong pronouns.
2 While, in standard French, eux must have a masculine antecedent, in popular French,
euxs antecedent may be either masculine or feminine (see e.g. Gougenheim 1973).
For example, in Montreal French, we find les gars, eux the boys, them and les filles,
eux the girls, them (A.-M. Brousseau, p.c.).
3 For an extensive description and account of the properties of this Fongbe lexical item,
see Kinyalolo (1993b, 1993c).
4 Capo (1991) argues that the third-person [nominative] clitic is, in fact, only a low
tone. The epenthetic vowel /i/ is added in order to make this low tone pronounceable.
Fongbe dictionaries and texts use the notation , however. In accordance with the
literature, I will use this notation throughout. Nothing hinges on this issue.
5 As is pointed out in Lefebvre (1991b), Da Cruz (1993) and Brousseau (1993), not all
Fongbe speakers manifest clitic climbing phenomena.
6 This proposal needs to be further documented on the basis of detailed analyses
of clitics in other creoles. For example, Veenstra (1996) claims that Saramaccan
has syntactic clitics. On the basis of the evidence presented, however, it is not
clear whether the Saramaccan clitics are really syntactic rather than phonological. For
example, no mention is made of the position of object clitics with respect to the verb,
or whether clitic climbing is available. More importantly, we need to know whether
clitics are allowed in the environment of a preposition (see (26) ).
7 According to Vasseur (1996: 75), in the Vimeu variety of Picard, the plural form leu
may be used instead of se as in i leu battent they beat themselves.

Notes to pages 164185

415

8 The idea that -# could have been assigned a phonologically null representation at
relabelling is attributable to John Lumsden (research seminar, Fall 1993). The further
development of this idea, as presented in this section, is mine.
9 This possibility was suggested to me by Yves-Charles Morin (p.c.).
10 Note that both source languages of Haitian have similar ways of emphasising a reflexive expression. Consider French Il sest vu lui-mme He saw himself (emphatic). In
Fongbe, the emphatic form #su corresponds to mm in Haitian and mme in French.
11 Malagasy, a substratum language of Mauritian creole, does use the word for body as
a true reflexive (see Keenan 1976). Carden (1993) argues that this lexical entry played
a crucial role in determining the appearance of the body-part reflexive lekor (lit.: the
body) in Mauritian creole.
12 Valdman et al. (1981) also list the form witi where. They mention that this lexical
item is rare. I did not find any Haitian informants who had this form in their lexicon.
For that reason, this lexical entry is not considered here.
13 Valdman et al. (1981) mention sa as a variant of pou ki-sa. Since I could find no
Haitian speaker who had this variant, I do not consider it here.
14 Citing Valdman (1978), Chaudenson (1996) reports the phrase ki kan (lit.: which
when) as an alternate phrase for when. None of the Haitian speakers I consulted
have this Wh-phrase. Furthermore, it has not been retained by Valdman et al. (1981).
For these reasons, I do not consider it in this section.
15 In French, it is grammatical but very unusual to say Quelles choses as-tu achetes?
What things did you buy? The complex phrase quest-ce que (lit.: what is it that)
is used to question the object.
16 Whether the creators of Haitian were exposed to French Wh in situ is a question that
cannot be resolved on the basis of available data from the literature. The seventeenthand eighteenth-century grammars report no such cases (see e.g. Wagner and Pinchon
1962; Haase 1965). Olivier Tardif read three plays by Molire and did not find any
cases of Wh in situ. In the written French texts, it appears that Wh in situ is first
attested in nineteenth-century authors (e.g. Flaubert). It is possible, however, that Wh
in situ was used in oral communication but not in writing at the time Haitian creole
was formed. In Furetire (1984), a seventeenth-century French dictionary, we find
Je ne sais quoi I do not know what. Furthermore, in popular French (such as that
spoken in Montreal, for example), Wh in situ is very productive (see Lefebvre 1982d;
1989). It is thus likely that the creators of Haitian had been exposed to it.
17 For example, xW -t You bought what? is more acceptable than ?? xW n in an
echo question (see Brousseau 1995a).
18 The correspondence between French [=] and Haitian /i/ is rare. For a discussion of
this point, see Brousseau (in preparation).
7

Functional category lexical entries involved in the structure of the clause

1 In the theoretical framework adopted here, heads select complements. In this view,
verbs select the complementisers which introduce their complements. Likewise, complementisers select ips as their complements. Tenseless complementisers select tenseless ips, tensed complementisers select tensed ips, and so on.
2 This argument is based on binding theory as outlined in Chomsky (1981) and related work.
3 Haitian speakers who speak a Gallicised version of Haitian may use the form ke to
introduce the tensed complement of verbs of the say-class. In the above discussion,

416

5
6

7
8
9
10
11

12

13
14
15

16
17
8

Notes to pages 188219

I do not consider these data since they are not part of the lexicon of speakers of the
basilect. If the use of ke were to become generalised among all Haitian speakers, it
would have to be considered a recent development.
Several syntactic tests distinguish the properties of the complementiser pou (see (9),
(10) ) from those of the preposition pou (see (11), (13) ). For a detailed discussion of
this topic, see Koopman and Lefebvre (1981, 1982); Koopman (1982a, 1982b); and
Sterlin (1988, 1989).
See the literature on the that-trace filter and Empty Category Principle (ecp) effects
(e.g. Chomsky 1981, and related work).
To the best of my knowledge, DeGraff (1992a) is the only researcher to have proposed an analysis claiming that Haitian lacks that-trace effects. Given that the facts he
presents to support his analysis are unfamiliar to me as well as to other people who
have written on this topic, I will not discuss his analysis any further here.
For further discussion of Koopmans analysis, see e.g. Lumsden (1990); Law (1992);
Manfredi (1993).
Deprez (1992b) provides several arguments showing that movement is involved in
this construction.
For a discussion of variation among speakers with respect to these facts, see Law (1992).
For further discussion of similar facts involving the verb gen there is, see Law
(1994a).
The structure illustrated in (66a), containing relative pronouns, was not commonly
used in seventeenth- and eighteenth-century French (see Bouchard 1982, and the
references therein).
When the subjects of the two conjoined clauses are coreferential, another form of the
conjunction may show up: b. In this case, the clitic subject of the conjoined clause is
not spelled out.
The question of morpheme order in derived and compound words will be addressed
in chapters 10 and 11.
Segurola (1963) translates ny as to be/to be named.
There is a fourth difference discussed at length by both DeGraff (1993a) and Deprez
(in press) which has to do with the way Haitian pa and French pas interact with
negative quantifiers. This topic will be addressed in chapter 12.
I am indebted to Jean-Robert Cadely and Michel DeGraff for fruitful discussions of
these facts.
Thanks to Michel DeGraff and Jean-Robert Cadely for pointing out this fact.
The determiner and the structure of the clause

1 In both languages, the determiner is also involved in the structure of complex clauses,
such as relative (see Koopman 1982a; Lefebvre and Massam 1988), conditional (see
Kinyalolo 1993a), and factive clauses (see Kinyalolo 1993a; Collins 1994), as well as
in verb-doubling constructions (see Lefebvre and Ritter 1993; Collins 1994; Lefebvre
1994b).
2 Grammar 1 includes Haitian speakers from Jacmel and Port-au-Prince and Fongbe
speakers from Abomey and Cotonou. Grammar 2 also includes speakers from the same
areas. Fongbe speakers from Ouidah present yet a third grammar since they do not use
the determiner in simple clauses. There is one Haitian speaker in Lefebvres (to appear
b) sample who exhibits no clausal determiner.

Notes to pages 222261

417

3 Da Cruz (1994) glosses wY as Det. In (23), wY is in complementary distribution with

the determiner V and it entails that the information contained in the clause is new
information. For a discussion of the semantics of wY, see Lefebvre (1992a). WY appears
to be another multifunctional item in Fongbe. Its various uses are listed in Anonymous
(1983); some of them are discussed in Lefebvre (1992a). For analyses showing the
various positions that can be headed by wY, see Kinyalolo (1992), and Law and Lefebvre
(1995).
4 For a detailed discussion of negative clauses, see Da Cruz (1994).
5 Note that the two classes of markers described for Fongbe have properties which
parallel the two classes of validational markers found in Quechua, a South American
Amerindian language spoken in the Andes (see Lefebvre and Muysken 1988, for a
detailed discussion of these two classes of morphemes in Quechua).
6 On the analysis in Lefebvre and Lumsden (1995), Haitian ske appears in the specifier
position of the projection headed by the yesno question marker in Fongbe.
7 According to Segurola (1963: 418), ngb means both truth and truly: The word is
used to assert [the content of a proposition] with force.
8 Note that speakers of grammar 1 may also use these adverbs after the assertive marker.
9 Note that, in the Fongbe dialect of Ouidah, the assertive marker has the form l. This
form is used to encode surprise in the Fongbe dialect of Abomey (see (3) ).
9

The syntactic properties of verbs

1 The term Lexical Conceptual Structure is used as in Hale and Laughren (1983) and
Hale and Keyser (1986, 1987). It consists in the representation of the concept named
by the verb. In this framework, the participants in the event described by the verb are
represented as variables.
2 As is pointed out in Brousseau (1995b), the Haitian verbs in (11b) may appear with
tt-li but in these cases, it is not simply interpreted as a reflexive, a fact which is
indicated by #. As Brousseau (1995b: 7) puts it: En effet, tt-li renforce le rle
agentif, soit en prcisant le caractre volitionnel de laction, en prcisant lautonomie
de laction de lAgent en opposition toute autre cause extrieure ou en attirant
lattention sur le caractre exceptionnel ou non attendu dune telle action. [In fact,
tt-li reinforces the agentive role, by specifying the volitional nature of the action,
emphasising the autonomy of the Agents action in opposition to any other external
cause or drawing attention to the exceptional or unexpected nature of such an action.]
This is exemplified in (i).
(i)

Jan
rapele
tt-li
fk
li w Jak.
haitian
John remember head-3rd must he see James
John forces himself to remember that he must see James.
(=(8b) in Brousseau 1995b)

3 Given the Haitian facts described above, I would not be surprised to find a lot of
variation among Haitian speakers with respect to the selectional properties of reflexive verbs. I suspect that this area of the grammar would be a gold mine for anyone
interested in variation within speech communities (see Carden and Stewart 1988, for
discussion of this point).
4 This situation suggests that Haitian posib has been acquired from French.

418

Notes to pages 261283

5 Note that the French expression corresponding to Haitian genl selects a rather than
il as an expletive subject (see (ib) ). The Haitian verb genl admits neither li nor sa, as
shown in (ia).
(i)

a.

Te
*Li te
*Sa te
it ant
It seemed

genl Jan
renmen
genl Jan
renmen
genl Jan
renmen
seem John love
that John loved Mary.

b. *Il
a lair que Jean
a a lair que Jean
it
seem that John
It seems that John loves

Mari.
Mari
Mari
Mary

aime Marie
aime Marie.
love Mary
Mary.

haitian

french

6 Note that sentences of the type in (b) are sensitive to definiteness effects (see Massam
1989). For detailed semantic and syntactic analyses of this verb, see e.g. Massam
(1989); Ritter (1991); Lumsden (1993a).
7 For various analyses of manke, see Massam (1989), Ritter (1991) and Lumsden
(1993a).
8 As A.-M. Brousseau has pointed out to me (p.c.), the missing object can appear in
subject position when it is interpreted as part of the Locative as in Trois pages
manquent dans ce document Three pages are missing from this document.
9 For extensive theoretical discussions of the properties of the verb sanble to seem,
see e.g. Dumais (1988); Ritter (1991); Law (1991, 1992); DeGraff (1992a, 1992b);
Deprez (1992a); Lumsden (1993a). For competing views on whether this Haitian
construction involves movement, see e.g. Law (1991, 1992) and Ritter (1991). For the
implications of these data for government theory, see Law (1991, 1992) and Deprez
(1992a).
10 See Lappin (1984) for a discussion of sentences such as John seems as if/like he is
unhappy.
11 Note that the embedded Fongbe clause has the structure of a relative clause headed
by mY person followed by a clause introduced by the operator #I, discussed in
chapter 7.
12 For various accounts of the Haitian facts presented above, see Law (1994a); Ritter
(1991); Deprez (1992a); Vinet (1991); DeGraff (1992a, 1992d).
13 In my view, it is most probable that the affix -yen on gen is the spelled-out trace of its
fronted internal argument. This is a topic for further research.
14 I have not been able to identify the phonological source of the Haitian verb gen.
15 For an analysis of n in this context, see Kinyalolo (1992).
16 Verbs such as b to want and byW to ask also allow for an infinitival complement
with an overt subject, as shown below.
(i)

i b j y.
He wants him to leave.

(ii) i byw j y.
He asked him to leave.

fongbe

17 What we do find in Haitian, however, is the type of construction in (i).

(i)

Jan
dmi yon ti
dmi.
John sleep a
little sleep
John slept a little.

haitian

Notes to pages 286301

18
19

20
21

419

While the phrase yon ti dmi in (i) modifies the event denoted by the verb (as can be
seen by the translation), the Fongbe cognate object discussed in the text does not have
this function. According to Dumais (1992), the construction illustrated in (i) is quite
productive in Haitian. It is not modelled either on Fongbe or on French, for neither
has such a construction. According to the methodology adopted in this book, this
Haitian construction must result from either the contribution of other substratum
languages or an innovation from within the creole.
See for example Lumsden (1987) and the references therein for changes in the Caseassigning properties of verbs in the history of English.
Since the appearance of the paper by Barss and Lasnik (1986), the fact that the
double-object construction exhibits Theme/Goal asymmetries has been the object of
a major debate in the literature. The central question is whether these asymmetries
should be accounted for in terms of dominance (e.g. Larson 1988, 1990; Aoun and Li
1989; Den Dikken 1991) or precedence (e.g. Barss and Lasnik 1986; Jackendoff
1990; Tremblay 1991). In English, the double-object construction allows for only one
word order: the Recipient always precedes the Theme. This makes it difficult to test
the contribution of linear precedence to these asymmetries in the English data. A
language in which the double-object construction allowed for the two surface word
orders RecipientTheme and ThemeRecipient would clarify this issue. Lefebvre
(1993c, 1994c) provides extensive evidence that the Fongbe double-object construction exhibits both word orders and there is no difference in meaning between them.
With both word orders, the Fongbe double-object construction manifests the same
RecipientTheme asymmetries as the double-object construction in English. This argues
for a dominance account of the asymmetries: in the double-object construction, the
Recipient must asymmetrically c-command the Theme.
Veenstra (1992) points out, however, that negative polarity items and superiority
effects cannot be tested for Haitian due to other intervening factors.
Lumsden (1993b: 54) points out that: Although it has fallen from use in modern
French, the verb bailler to give (the probable source of the phonological form of
the Haitian Creole verb bay, to give) was once more common. As the following
example from Huguets (1925) Dictionnaire de la langue franaise du seizime sicle
(Tome I, p. 456) demonstrates, bailler was a Theme/Goal verb.
(i)

Rabelais II,2
Au temps que Phebus bailla le gouvernement de son chariot lucifique son filz
Phaeton [At the time that Phoebus gave the control of his glowing chariot to his
son Phaeton].

22 Lefebvre (1993c, 1994c) contain examples for all the tests except the one involving
negative polarity items, for which the data were not available at the time.
23 The discrepancy between the two inventories may be attributable to the type of
examples provided in Valdman et al. (1981) rather than to a substantial difference
between the two sets of data.
24 Note that the Haitian verb paye to pay may also take a pp complement introduced by
the preposition pou, as shown in (i). In this case, however, the complement is not
assigned the Goal thematic role, as can be seen from the translation.
(i)

Jan
pran kb
paye pou
John take money pay for
John paid on behalf of Mary.

Mari.
Mary

haitian

420
10

Notes to pages 304327

Are derivational affixes relexified?

1 The epenthetic consonants identified in this section are not always predictable. Exactly
how they were established is a topic for future research.
2 Note, however, that Valdman et al.s (1981) and Vdrines (1992) dictionaries both
list kale with the meaning to shell. Neither dictionary lists mwle as an independent
word.
3 Brousseau, Filipovich and Lefebvre (1989: 12) remark that some Haitian words ending in ay are listed as unanalysable simplexes.
The word vwazinay neighborhood, for example, is probably not derived from a
verb, since, for most Haitian speakers, there is no verb vwazine; most likely, this
lexical item has been borrowed from French voisinage as a simplex. Similarly,
vakabonday vagrancy would be the Haitian version of French vagabondage,
briganday tumult of French brigandage, lib+tinay licentiousness of French
libertinage, and betay cattle of French btail . . . In the same fashion, we do
not consider the following to be derived words: zanmiray friendship, kuzinay
collection of cousins, and fanmiray family. First, it should be noted that the
list of such words is extremely small. Moreover, they are not accepted by all
informants. Second, the potential affix -ay found on such words is not productively used to form other words that would have a collective meaning:
*fr-r-ay
*s-r-ay

the collection of brothers

the collection of sisters

Third, the potential affix -ay is not used with bases other than those of French
origin (e.g. Spanish words or words of West African origin). For these reasons,
we consider the words listed above as simplex frozen forms and not as derived
words.
4 Adjectival conversion manifested in languages such as English is assumed to derive
adjectives from the participial form of the verbs (see e.g. Levin and Rappaport 1986).
In this chapter, I adopt the view argued for in Brousseau (1994a) whereby deverbal
adjectives and participial forms in Haitian are both derived from the base form of verbs.
5 See also Trsor de la langue franaise for comparable data on -age and a long
discussion of the evolution of this affix in French.
6 Variation in the pronunciation of some derivational affixes between classical and
modern French is discussed in Juneau (1972).
7 Brousseau (1994a) points out that the agentive affix -tV may also take a nominal base
in a few cases as is exemplified in (i):
(i)

z-tv
gb-tv

sorcerer
hunter

z magic
gb bush

8 For an extensive discussion of the derivation of adjectival and verbal passives in

Fongbe, see Brousseau (1993).
9 I do not know, however, which Haitian affix corresponds to which Fongbe one.
10 For a discussion of the semantic contribution of RE-, see Brousseau (1993).
11 Brousseau (1993) claims that RE- is specified only for the feature [+n]. Her analysis,
however, also makes use of the feature [v]. In claiming that RE- is identified for
the features [+n, v], I depart from her analysis only in notational terms rather than
on theoretical grounds. But see Brousseau (1993) for a discussion of percolation
conventions.

Notes to pages 329346

421

12 In more recent work, blocking is seen as the expression of a tendency towards economy
in the lexicon rather than a general principle of lexical organisation. For a discussion
of this point, see e.g. Scalise (1984).
13 Reduplication of words is a widespread phenomenon in creoles of the Caribbean. It is
generally considered to be a discourse process used to convey intensification. Cases
of this type of reduplication are reported in Jamaican creole (see e.g. Cassidy 1957;
DeCamp 1974), Sranan (see e.g. Sebba 1981), Saramaccan (see e.g. Bakker 1987),
and Berbice Dutch (see Kouwenberg 1990). Haitian is no exception in using this
strategy to convey intensification (see Sylvain 1936; Hall 1953; Valdman 1970). The
following example illustrates this type of reduplication.
(i)

M t
a
tiretire.
I
ant ind-fut shoot shoot
I would have shot a lot.

haitian

(from Sylvain 1936: 83)

Very few Caribbean creoles, however, have an affix which involves either partial or
total reduplication of the base. I know of only two for which a true prefix RE- has been
argued: Berbice Dutch (see Kouwenberg 1994) and Saramaccan (see Bakker 1987;
Smith 1995).
14 This proposal would be independently supported if it could be shown that creoles
which did retain RE- (e.g. Berbice Dutch and Saramaccan) did not make the same
parametric choice with respect to headedness.
15 It should be noted that this type of reanalysis, revealing input from the superstratum
language morphology, has no counterpart in inflectional morphology, as we saw in
chapters 4 to 8.
11

The concatenation of words into compounds

1 Brousseaus data on Haitian and Fongbe compounds are original, as are some of her
French data. Other data are drawn from the literature on French compounds (see e.g.
Bauer 1980; Barbaud 1971; Darmesteter 1875; Giurescu 1975; Miller 1978; Moody 1973).
2 Brousseau (1989) remarks that the use of kalbas-tt skull in Haitian has a somewhat
ironical flavour.
3 The suffix -nV which occurs with this Fongbe compound in (5c) is the attributive suffix
discussed in chapter 10. The suffix -tW which occurs in (5f) is the Agentive suffix
discussed in chapter 10.
4 Valdman et al. (1981) lists tt-di as two different lexical entries. The first is described
as an attributive phrase meaning stubborn, and the second as a compound meaning
stubbornness. The first usage of this compound is discussed together with the set of
examples in (5). The second usage is discussed in connection with (6).
5 The Haitian lexical entry mare tie probably derives its phonological representation
from the phonetic matrix of the French verb amarrer attach, tie, fasten in use in
Norman French (see Moisy 1969).
6 Brousseau (1988a) lists a few Haitian compounds she describes as appositional. She
later reanalyses them as headed compounds (see Brousseau 1994a).
7 Smith and Veenstra (1994) list two additional types of syntactic compounds in Fongbe:
nvvtV and nvpntV. These types of compounds were not incorporated into
Brousseaus inventory because they are not productive. In her view, they would compare to kiss-me-over-the-garden-gate (a type of flower) in English.

422

Notes to pages 347365

8 The word order in the Saramaccan compound differs from that of its Fongbe counterpart. This discrepancy, I believe, can be accounted for along the lines of the analysis
presented in section 11.2 to account for word order in Haitian compounds.
9 The fact that the Fongbe compounds in (31a) correspond to the French vn compounds
in (25a) rules out the possibility that this type of compound is lacking from Haitian
because of the absence of similar compounds in French. An account of the lack of
compounds of type (31a) in Haitian will have to await further research. For a preliminary discussion of this topic, see Brousseau (1989: 308).
12

Parameters

1 For extensive discussions on parameters and parameter setting in first language acquisition, see Roeper and Williams (1987); for second language acquisition, see Ritchie
and Bathia (1996a, 1996b) and the references therein.
2 An evaluation of the issues involved in this proposal may be found in Law (1992).
3 In view of the enormous literature on the topic, I will limit the discussion to a few of
the proposed parameters. For further discussion, see, for example, Muysken (1988d)
and the literature cited therein. Furthermore, since I am discussing the availability of
the serial verb construction in terms of parametric variation, I do not take into account
proposals which link this construction to the lexical properties of verbs, such as their
Case-assigning, thematic or control properties.
4 For more details on this proposal on the basis of Kwa and Bantu languages, see Baker
(1991).
5 For an extensive discussion of serial verbs as secondary predicates, see Larson (1991).
6 As is pointed out by Muysken (1988d) and by Baker and Stewart (1996), a theoretical
account of this correlation still remains to be formulated within the framework of the
theory of principles and parameters.
7 See also Lefebvre (1990a) for a detailed analysis of the full range of prepositions in
Fongbe and the non-availability of preposition stranding in the language. Lefebvre
(1990a) also presents data on the resumptive strategy which is available in some
cases.
8 In most contexts, this is not visible, since nouns and strong pronouns are not overtly
distinguished for Case in these languages (see chapter 6). We know from chapter 6,
however, that in Fongbe clitics are specified for [+/nominative Case]: while the object
clitic bears a Low tone, the subject clitic bears a High tone. Since syntactic clitics do
not occur as complements of ps, this distinction is not visible in the context under
discussion here.
9 This fact is salient in Fongbe, which has overt nominalising morphology, but not in
Haitian since, as shown in chapter 10, Haitian has a phonologically null nominalising
affix.
10 There is a general consensus in the literature to the effect that the copys sentenceinitial position is a derived one (that is, the copy is moved there from an inner clausal
position). (But see Lefebvre and Ritter 1993, for a discussion of this issue.) Both the
position out of which the copy is moved and the positions into which it moves,
however, are a matter for debate. This debate stems from two other debates concerning, first, the categorial status of the copy, i.e. whether it is of the category v (see
Koopman 1984; Ndayiragije 1993) or n (see e.g. Lefebvre 1994b; Collins 1994; Law
and Lefebvre 1995) and, second, whether movement of the copy involves movement

Notes to pages 366393

11

12
13
14

13

423

of an x0 (see Koopman 1984; Ndayiragije 1993) or an xp (see Lefebvre 1994b;

Collins 1994; Law and Lefebvre 1995). Any detailed discussion of these issues would
go far beyond the scope of this section. My position to date, on the basis of what
is known about these constructions, is that the copy projects and that it therefore
involves movement of a maximal projection (see Lefebvre 1994b; Collins 1994).
On the basis of parallelism between the copy in (27)(30) and cognate objects, I take
the copy to be base-generated in the same position as cognate objects (see Lefebvre
1994b). I believe the surface position (that is, the landing site) of the copy to be Spec
cp in temporal and causal clauses. Lefebvre and Ritter (1993) and Collins (1994) have
argued that, in factive constructions, the copy is outside cp. I will assume that, in this
case, it is adjoined to cp, along the lines of the analysis in Collins (1994). In the
predicate cleft construction, the copy occurs in a clefted constituent. Assuming the biclausal structure of Haitian and Fongbe clefts argued for in Lumsden (1990) and Law
and Lefebvre (1995), respectively, in the predicate cleft construction, the copy must
be in a different clause from the verb that it is the copy of.
In sentences involving verb-doubling phenomena, no further extraction is allowed,
that is, no argument of the verb can be either questioned or clefted. For specific
discussions of these facts, see Piou (1982a), Lefebvre (1990b), Lumsden and Lefebvre
(1990), Lefebvre and Ritter (1993) for Haitian; Ndayiragije (1993) and Law and
Lefebvre (1995) for Fongbe; and Collins (1994) for Ewe and Fongbe. See also
Koopman (1984) for a discussion of this topic on the basis of several West African
languages and Haitian.
See Collins (1994) for additional facts on determiners in the context of verb-doubling
in Ewe and Fongbe.
Similar facts are also reported in Lefebvre and Ritter (1993).
A more detailed discussion of this proposal on the basis of Fongbe data can be found
in Law and Lefebvre (1995).
Evaluation of the hypothesis

1 The definite determiner, the plural marker and the demonstrative terms were relexified.
These have the common property of being definite in some sense. In contrast, the
substratum languages so-called indefinite determiner was abandoned. The copula ny
was not relexified nor was the existential verb. Interestingly enough, in these three
areas of the lexicon, Haitian speakers have had to independently develop morphemes
from within the creole: yon, se and gen. What these lexical entries have in common is
that they are all indefinite in some sense, assuming that existential verbs may be seen
as indefinites. Is it possible that, for semantic reasons, indefinite/existential items cannot be relexified? I leave this question open for further research.
2 The results of the study in Lefebvre (in press) are comparable to those in Lumsden
(1994b) based on a different set of data.
3 This discussion is based on an extensive study that I conducted with Danielle Dumais
on adjectives in Fongbe and Haitian based on lists of adjectives in dictionaries and
supplemented by work with informants for syntactic tests.
4 Thanks to Paul Kay for fruitful discussions of this issue.
5 For a discussion of this point on the basis of data drawn from several creole languages,
see Muysken (1981c).
6 Thanks to Christine Jourdan for fruitful discussions of this issue.

424

References

References

Abney, Steven 1987, The English noun phrase in its sentential aspect, Doctoral Dissertation, MIT, Cambridge, MA.
Adam, Lucien 1883, Les idiomes ngro-aryen et malo-aryen, Paris.
Adjemian, Christian 1983, The transferability of lexical properties, in Gass and Selinker
(eds.), pp. 25068.
gbdvr, Paul Kf 1995, Negation in v, in Manfredi and Reynolds (eds.), pp. 121
34.
Ajayi, J. F. Ade and Crowder, Michael (eds.) 1976, History of West Africa, London:
Longman.
Alembert, Jean Le Rond d, Mouchon, Pierre and Diderot, Denis 17767, Encyclopdie
ou dictionnaire raisonn des sciences, des arts et des mtiers (Microfiches, 105 148
mm, microcopy of Paris edition, 17767), Paris: France-expansion (1973).
Allen, Margaret 1978, Morphological investigations, Doctoral Dissertation, University
of Connecticut, Storrs.
Alleyne, Mervyn C. 1966, La nature du changement phontique la lumire du crole
franais dHati, Revue de linguistique romane 30: 279303.
1971, Acculturation and the cultural matrix of creolisation, in Hymes (ed.), pp. 169
87.
1981, Comparative Afro-American: An Historical-Comparative Study of English-Based
Afro-American Dialects of the New World, Ann Arbor, MI: Karoma.
Alleyne, Mervyn C. (ed.) 1987, Studies in Saramaccan Language Structure (Caribbean
Culture Studies 2), Universiteit van Amsterdam.
Allsopp, Richard 1980, How does the creole lexicon expand?, in Valdman and Highfield
(eds.), pp. 89107.
Andersen, Roger W. 1980, Creolisation as the acquisition of a second language as a first
language, in Valdman and Highfield (eds.), pp. 27395.
1983a, A language acquisition interpretation of pidginisation and creolisation, in
Andersen (ed.), pp. 159.
Andersen, Roger W. (ed.) 1983b, Pidginisation and Creolisation as Language Acquisition, Rowley, MA: Newbury House.
Anestin, Agns-Lydie 1987, Structure syllabique de lhatien et nasalisation, Masters
Thesis, Universit du Qubec Montral.
Anonymous c. 1792, Grand dbat entre Duffay et consorts, Polverel et Sonthonax les
gorgeurs et les brleurs de Saint-Domingue, n.p.
Anonymous 1796, Proclamation de Sonthonax, in Valdman (1978), p. 101.
Anonymous 1802, Proclamation de Bonaparte, in Valdman (1978), p. 101.
Anonymous 1802, Proclamation de Leclerc, in Valdman (1978), pp. 1012.

424

References

425

Anonymous 1810, Cri des colons contre un ouvrage de M. levque et senateur Grgoire,
ayant pour titre de la littrature des ngres . . . , Paris.
Anonymous 1821 (1811), Idylles, ou essais de posie creole, par un colon de St.-Domingue,
Cahors.
Anonymous 1824, Parabole de lEnfant Prodigue en crole de Saint-Domingue, Port-auPrince, in Anonymous (1831), p. 93.
Anonymous 1831, Mlanges sur les langues, dialectes et patois . . . , Paris.
Anonymous 1983, lments de recherche sur la langue Fon, Cotonou.
Ans, Andr-Marcel d 1968, Le crole franais dHati, The Hague: Mouton.
Ansre, Gilbert 1961, The Tonal Structure of Ewe, Doctoral Dissertation, Hartford Seminary Foundation.
Aoun, Joseph and Li, Yen-Hui Audrey 1989, Scope and constituency, Linguistic Inquiry
20: 14172.
Arends, Jacques 1995a, Demographic factors in the formation of Sranan, in Arends
(ed.), pp. 23377.
Arends, Jacques (ed.) 1995b, The Early Stages of Creolisation (Creole Language Library
13), Amsterdam/Philadelphia: John Benjamins.
Arends, Jacques, Muysken, Pieter and Smith, Norval (eds.) 1995, Pidgins and Creoles:
An Introduction (Creole Language Library 15), Amsterdam/Philadelphia: John
Benjamins.
Aronoff, Mark 1976, Word Formation in Generative Grammar, Cambridge, MA: MIT
Press.
Atlas sociolinguistique du Bnin 1983, Abidjan: Commission nationale linguistique du
Bnin, Universit dAbidjan.
Audain, M. J. J. 1877, Recueil de proverbes croles, Port-au-Prince.
Auger, Julie 1994, Pronominal clitics in Quebec colloquial French: A morphological
analysis, Doctoral Dissertation, University of Pennsylvania, Philadelphia.
Avolonto, Aim 1992, De ltude smantico-syntaxique des marqueurs prverbaux la
structure de la phrase en fwngb, Masters Thesis, Universit du Qubec Montral.
1995, Pour une approche minimaliste des verbes objets inhrents, Doctoral Dissertation, Universit du Qubec Montral.
Awoyale, Yiwola 1986, Reflexivisation in Kwa languages, in Dimmendaal (ed.), pp. 1
14.
Bach, Emmon, Kratzer, Angelica and Partee, Barbara (eds.) 1989, Papers on Quantification, Amherst: University of Massachusetts (NSF Report).
Baker, Mark C. 1988, Incorporation: A Theory of Grammatical Function Changing,
University of Chicago Press.
1991, On the relation of serialisation to verb extensions, in Lefebvre (ed.), pp. 79
102.
Baker, Mark C. and Stewart, Osamuyimen Thompson 1996, Unaccusativity and the
adjective/verb distinction: Edo evidence, paper read at NELS 27, McGill University, Montreal.
Baker, Philip 1984, Agglutinated French articles in Creole French: Their evolutionary
significance, TeReo 27: 89129.
1990, Off target?, Journal of Pidgin and Creole Languages 5: 10719.
1993, Assessing the African contribution to French-based creoles, in Mufwene (ed.),
pp. 12355.

426

References

Baker, Philip and Corne, Chris 1982, Isle de France Creole, Ann Arbor, MI: Karoma.
Baker, Philip and Syea, Anand (eds.) 1996, Changing Meanings, Changing Functions
(Westminster Creolistics Series 2), London: University of Westminster Press.
Bakker, Peter 1987, Reduplications in Saramaccan, in Alleyne (ed.), pp. 1740.
1989, Relexification in Canada: The case of Mtif (French-Cree), in Lefebvre and
Lumsden (eds.) (1989b), pp. 33950.
1992, A language of our own: The genesis of Mitchif, the mixed CreeFrench language of the Canadian Metis, Doctoral Dissertation, University of Amsterdam.
1994, Michif, the CreeFrench mixed language of the Mtis buffalo hunters in Canada,
in Bakker and Mous (eds.), pp. 1333.
Bakker, Peter and Mous, Maarten 1994a, Introduction, in Bakker and Mous (eds.),
pp. 111.
Bakker, Peter and Mous, Maarten (eds.) 1994b, Mixed Languages (Studies in Language
and Language Use 13), Dordrecht: ICG Printing.
B0gb-, Ayo 1972, The Yoruba Verb Phrase, Idbn University Press.
1974, On serial verbs and verbal status, Journal of West African Linguistics 9: 1748.
Barbaud, Philippe 1971, Lambigut structurale du compos binominal, Cahiers de
linguistique 1: 71116.
Barss, Andrew and Lasnik, Howard 1986, A note on anaphora and double objects,
Linguistic Inquiry 17: 34752.
Bary, Ren 1665, La rhtorique francoise ov povr principale avgmentation lon trovve les
secrets de nostre langve (Archives de la linguistique franaise 023), Paris: P. Le Petit.
Barz, Richard K. and Siegel, Jeff (eds.) 1988, Language Transplanted: The Development
of Overseas Hindi, Wiesbaden: Harrassowitz.
Bastide, Roger 1967, Les Amriques Noires, Paris: Payot.
Bauer, Laurie 1980, Deux problmes au sujet des noms composs comprenant un
premier lment verbal en franais moderne, Le franais moderne 48: 21924.
Bendor-Samuel, John (ed.) 1989, The Niger-Congo Languages, New York: University
Press of America.
Bennis, Hans 1982, De Pro-drop-Parameter en Subjektloze Zinnen in het Nederlands,
Spektador 6: 40927.
Bentolila, Alain 1971, Les systmes verbaux croles: comparaison avec les langues
africaines, Doctoral Dissertation, Universit de Paris V, Ren Descartes.
1987, Marques aspecto-temporelles en crole hatien: de lanalyse synchronique la
formulation dhypothses diachroniques, La Linguistique 23: 10322.
Bernab, Jean 1983, Fondal-Natal (Grammaire basilectale approche des croles
guadeloupen et martiniquais 3), Paris: LHarmattan.
Bernstein, Judy B. 1993, Topics in the syntax of nominal structure across Romance,
Doctoral Dissertation, City University of New York.
Bertho, Jacques 1946, Adja-Tado, races et langues du Bas Dahomay et du Bas Togo,
Grands Lacs 61: 1012.
Bickerton, Derek 1977, Pidginisation and creolisation: Language acquisition and language universals, in Valdman (ed.), pp. 4969.
1981, Roots of Language, Ann Arbor, MI: Karoma.
1984, The language bioprogram hypothesis, The Behavioral and Brain Sciences 7:
173221.
1986, Creoles and West African languages: A case of mistaken identity?, in Muysken
and Smith (eds.), pp. 2540.

References

427

1988, Creole languages and the bioprogram, in Newmeyer (ed.) (1988a), pp. 268
84.
1992, The sociohistorical matrix of creolisation, Journal of Pidgin and Creole Languages 7: 30718.
Bigelow, John 1877, The Wit and Wisdom of the Haytians, New York: Scriber &
Armstrong.
Blanche-Benveniste, C., Chervel, A. and Gross, M. (eds.) 1988, Grammaire et histoire de
la grammaire; Hommage la mmoire de Jean Stefanini, Aix: Publication de
lUniversit de Provence.
Bloomfield, Leonard 1933, Language, New York: Holt, Rinehart, and Winston.
Bonnard, Henri 1982, Synopsis de phontique historique, Paris: Socit ddition
denseignement suprieur.
Bonneau, Alexandre 1856, Les Noirs, les Jaunes et la littrature franaise en Hati,
Revue contemporaine 5: 10755.
Boretzky, Norbert, Enninger, Werner and Stolz, Thomas (eds.) 1988, Beitrge zum 4.
Essener Kolloquium, Bochum: Brockmeyer.
Boretzky, Norbert and Stolz, Thomas (eds.) 1990, Beitrge zum 6. Essener Kolloquium,
Bochum: Brockmeyer.
Bouchard, Denis 1982, Les constructions relatives en franais vernaculaire et en franais
standard: tude dun paramtre, in Lefebvre (ed.), pp. 10333.
1995, The Semantics of Syntax, University of Chicago Press.
Bouvet, Francis and Andler, Pierre (eds.) 1984, Le Littr, Paris: Union gnrale dditions.
Brasseur, Patrice 1995, Les pronoms personnels dans lAtlas linguistique normand, in
Simoni-Aurembou (ed.), pp. 32351.
Brillon, Jolle and Brousseau, Anne-Marie 1986, Aspects de la phonologie du fon, in
Lefebvre and Kaye (eds.), pp. 36185.
Brousseau, Anne-Marie 1988a, Triptyque sur les composs: les noms composs en
franais, fongbe et hatien en regard des notions de tte et de percolation, Travaux
de recherche sur le crole hatien 2: 1109 (Universit du Qubec Montral).
1988b, Inherent object verbs in Fongbe, in Lefebvre (ed.), pp. 371400.
1989, De n-fl po-bouche: hypothses sur lorigine des composs en hatien,
in Lefebvre and Lumsden (eds.) (1989b), pp. 285312.
1990, Panorama de la morphologie du Fongbe, Journal of West African Languages
20: 2745.
1991, Domaines et relations de gouvernement dans les processus tonals du fongbe, in
Kihm (ed.), pp. 93112.
1993, Reprsentations smantiques et projections syntaxiques des verbes en Fungb,
Doctoral Dissertation, Universit du Qubec Montral.
1994a, Morphological structure and relexification, in Lefebvre and Lumsden (eds.)
(1994c), 30 pages.
1994b, The relexification of phonology, in Lefebvre and Lumsden (eds.) (1994c), 44
pages.
1995a, Les pronoms en crole hatien, en franais et en fongb, in Lefebvre (ed.)
(1995a), volume IV, 33 pages.
1995b, Rflexivit et classes de verbes en crole hatien et en fongb, in Lefebvre
(ed.) (1995a), volume IV, 15 pages.
In preparation, The Genesis of the Phonological System of Haitian Creole [provisional
title].

428

References

Brousseau, Anne-Marie, Filipovich, Sandra and Lefebvre, Claire 1989, Morphological

processes in Haitian Creole: The question of substratum and simplification, Journal
of Pidgin and Creole Languages 4: 136.
Brousseau, Anne-Marie and Lumsden, John S. 1992, Nominal structure in Fongbe, in
Lefebvre (ed.), pp. 526.
Brunot, Ferdinand 1905, Histoire de la langue franaise des origines 1900, Paris:
A. Colin.
1926, La pense et la langue, Paris: Masson et cie.
Bruyn, Adrienne 1996, On identifying instances of grammaticalisation in creole languages, in Baker and Syea (eds.), pp. 2947.
Butel, Paul 1982, Le temps des fondations: Les Antilles avant Colbert, in Pluchon (ed.),
pp. 5378.
Byrne, Francis 1987, Grammatical Relations in a Radical Creole. Verb Complementation in Saramaccan (Creole Language Library 3), Amsterdam/Philadelphia: John
Benjamins.
Byrne, Francis and Holm, John (eds.) 1993, Atlantic Meets Pacific (Creole Language
Library 11), Amsterdam / Philadelphia: John Benjamins.
Byrne, Francis and Huebner, Thom (eds.) 1991, Development and Structures of Creole
Languages (Creole Language Library 10), Amsterdam/Philadelphia: John Benjamins.
Byrne, Francis and Winford, Donald (eds.) 1993, Focus and Grammatical Relations in
Creole Languages (Creole Language Library 12), Amsterdam/Philadelphia: John
Benjamins.
Cadely, Jean-Robert 1988a, Reprsentations syllabiques et distribution des diphtongues
en crole hatien, tudes croles 11: 940.
1988b, Lopposition r/w en crole hatien: un paradoxe rsolu, Revue canadienne de
linguistique 33: 12142.
1990, Domaine prosodique dune rgle de sandhi externe et cliticisation des lments
fonctionnels du crole hatien, Universit du Qubec Montral MS.
1994, Aspects de la phonologie du crole hatien, Doctoral Dissertation, Universit
du Qubec Montral, Montreal.
Capo, Hounkpati B. C. 1984, Elements of EweGenAjaFon dialectology, in Medeiros
(ed.), pp. 16778.
1991, A Comparative Phonology of Gbe (Publications in African Languages and Linguistics 14), Dordrecht: Foris.
Carden, Guy 1993, The Maurician Creole lekor reflexive: Substrate influence on the
target-location parameter, in Byrne and Holm (eds.), pp. 10518.
Carden, Guy and Stewart, William A. 1988, Binding theory, bioprogram, and creolisation:
Evidence from Haitian Creole, Journal of Pidgin and Creole Languages 3: 168.
Carlson, G. N. 1977, A unified analysis of the English bare plural, Linguistics and
Philosophy 1: 41357.
Carrington, Lawrence (ed.) 1983, Studies in Caribbean Language, St. Augustine, Trinidad:
Society for Caribbean Linguistics.
Cassidy, Fred 1957, Iteration as a word-forming device in Jamaican folk speech, American Speech 32: 4953.
Catach, Nina (ed.) 1995, Dictionnaire historique de lorthographe franaise, Paris:
Larousse.
Cattell, R. 1984, Composite Predicates in English (Syntax and Semantics 17), North
Ryde, Australia: Academic Press.

References

429

Chaudenson, Robert 1973, Pour une tude compare des croles et parlers franais
doutre-mer: survivance et innovation, Revue de linguistique romane 1478: 342
71.
1977, Toward the reconstruction of the social matrix of creole language, in Valdman
(ed.), pp. 25977.
1983, O lon reparle de la gense et des structures des croles de lOcan Indien,
tudes croles 6: 157224.
1990, Recherche, formation et crolistique, Revue qubcoise de linguistique thorique
et applique 9: 287303.
1992, Des les, des hommes, des langues. Langues croles/cultures croles, Paris:
LHarmattan.
1993, De lhypothse aux exemples. Un cas de crolisation: selon la formation des
systmes de dmonstratifs croles, tudes croles 16: 1738.
1994, Comparatisme et mthode gnrative transformationnelle: le cas des tudes sur
le crole hatien, Crolistique et grammaire gnrative, Pluralismes 8: 120.
1996, Dmystification de la relexification, tudes croles 19: 93109.
Chomsky, Noam 1972, Studies on Semantics in Generative Grammar, The Hague: Mouton.
1981, Lectures on Government and Binding, Dordrecht: Foris.
1986, Knowledge of Language: Its Nature, Origin and Use, New York: Praeger.
1989, Some notes on the economy of derivation and representation, in Laka Mugarza
and Mahajan (eds.), pp. 4374.
Christie, Pauline (ed.) 1996, Caribbean Language Issues: Old and New, Kingston,
Jamaica: University of the West Indies Press.
Clements, Georges N. 1975, The logophoric pronoun in Ewe: Its role in discourse,
Journal of West African Languages 10: 14177.
Collins, Chris 1993, Topics in Ewe syntax, Doctoral Dissertation, MIT, Cambridge,
MA.
1994, The factive construction in Kwa, Travaux de recherche sur le crole hatien
23: 3165 (Universit du Qubec Montral).
Collins, Chris and Manfredi, Victor (eds.) 1992, Proceedings of the Kwa Comparative
Syntax Workshop (MIT Working Papers in Linguistics 17), Cambridge, MA: Department of Linguistics, MIT.
Comrie, Bernard 1976, Aspect (Cambridge Textbooks in Linguistics), Cambridge University Press.
1985, Tense (Cambridge Textbooks in Linguistics), Cambridge University Press.
Cook, Eung-Do and Gerdts, Donna (eds.) 1984, The Syntax of Native American Languages (Syntax and Semantics 16), New York: Academic Press.
Coopmans, Peter, Bordelois, Ivonne and Smith, Bill Dotson (eds.) 1986, Formal Parameters of Generative Grammar, Dordrecht: ICG Printing.
Corne, Chris 1988, Mauritian Creole reflexives, Journal of Pidgin and Creole Languages 3: 6994.
1989, On French influence in the development of creole reflexive patterns, Journal of
Pidgin and Creole Languages 4: 10316.
Cowper, Elizabeth A. 1989, Thematic underspecification: The case of have, paper read
at the Canadian Linguistics Association Annual Meeting, Universit Laval, Quebec
City.
1995, English participle construction, Revue canadienne de linguistique 40: 138.
Crystal, David 1991, A Dictionary of Linguistics and Phonetics, Cambridge: Blackwell.

430

References

Curtin, Philip D. 1976, The Atlantic slave trade, 16001800, in Ajayi and Crowder
(eds.), pp. 30230.
Da Cruz, Maxime 1993, Les constructions srielles du Fungb: approches smantique et
syntaxique, Doctoral Dissertation, Universit du Qubec Montral.
1994, Contribution ltude de la ngation en Fungb, in Lefebvre and Lumsden
(eds.) (1994e), volume III, pp. 69111.
Da Cruz, Maxime and Kinyalolo, Kasangati 1994, Les classes aspectuelles des verbes en
Fongbe, in Lefebvre and Lumsden (eds.) (1994d), volume III, pp. 1113.
Dalby, David (ed.) 1970, Language and History in Africa, London: Frank Cass & Co. Ltd.
Damoiseau, Robert 1988, lments pour une classification des verbaux en crole hatien,
tudes croles 11: 4164.
Darmesteter, Arsne 1875, Trait de la formation des mots composs dans la langue
franaise compare aux autres langages romanes et au latin (2nd edition), Paris:
Champion (1967).
Debien, Gabriel 1962, Plantations et esclaves Saint-Domingue, Dakar.
DeCamp, David 1974, Neutralisations, iteratives, and ideophones: The locus of language
in Jamaica, in DeCamp and Hancock (eds.), pp. 4260.
DeCamp, David and Hancock, Ian (eds.) 1974, Pidgins and Creoles. Current Trends and
Prospects, Washington, DC: Georgetown University Press.
Dchaine, Rose-Marie 1988, Oprations sur les structures darguments: Le cas des constructions srielles en Hatien, Masters Thesis, Universit du Qubec Montral.
1991, Bare sentences, in Moore and Wyner (eds.), pp. 3150.
1992, Inflection in gbo and Yorb, in Collins and Manfredi (eds.), pp. 95119.
1993, Predicates across categories, Doctoral Dissertation, University of Massachusetts, Amherst.
Dchaine, Rose-Marie and Filipovich, Sandra 1986, La construction relative en Fon, in
Lefebvre and Kaye (eds.), pp. 198221.
Dchaine, Rose-Marie and Manfredi, Victor 1994, Binding domains in Haitian, Natural
Language and Linguistic Theory 12: 20357.
Dees, Anthonij 1971, tude sur lvolution des dmonstratifs en ancien et moyen franais,
Groningen: Wolters-Noordhoff.
DeGraff, Michel A. F. 1992a, Is Haitian Creole a pro-drop language?, Travaux de
recherche sur le crole hatien 11: 122 (Universit du Qubec Montral).
1992b, Creole grammars and acquisition of syntax: The case of Haitian, Doctoral
Dissertation, University of Pennsylvania, Philadelphia.
1992c, The syntax of predication in Haitian, Proceedings of the 22nd Annual Meeting
of the Northeastern Linguistics Society, Amherst, MA: University of Massachusetts,
Graduate Linguistics Students Association.
1992d, Haitian null subjects revisited, Travaux de recherche sur le crole hatien 11:
5974 (Universit du Qubec Montral).
1993a, A riddle on negation in Haitian, Probus 5: 6393.
1993b, Is Haitian Creole a pro-drop language?, in Byrne and Holm (eds.), pp. 7190.
1994, The morphology-syntax interface in creolisation (and diachrony), Studies in the
Linguistic Sciences 24: 11531.
1996, Creole languages and parameter setting: A case study using Haitian Creole and
the pro-drop parameter, in Wekker (ed.), pp. 65105.
DeGraff, Michel A. F. (ed.) In press, Parametric Approaches to Creole Syntax, Cambridge, MA: MIT Press.

References

431

Dejean, Y. 1980, Comment crire le crole dHati, Montral: Collectif Paroles.

Demuth, Katherine A. and Manfredi, Victor (eds.) 1988, Niger-Congo Syntax and
Semantics 1, Boston University, MS.
Den Besten, Hans 1978, On the interaction of root transformation and lexical deletive
rules, paper read at GLOW IV, Amsterdam.
Den Dikken, Marcel 1991, Particles and the dative alternation, in Van Lit, Mulder and
Sybesma (eds.), pp. 7186.
Deprez, Viviane 1992a, Is Haitian Creole really a pro-drop language?, Travaux de
recherche sur le crole hatien 11: 2340 (Universit du Qubec Montral).
1992b, Raising constructions in Haitian Creole, Natural Language and Linguistic
Theory 10: 191231.
In press, The roots of negative concord in French and French based creoles, in
DeGraff (ed.).
Deprez, Viviane and Vinet, Marie-Thrse 1991, Se in Haitian Creole: an ambiguous
X category, MS.
Descourtilz, M. E. 1809, Voyages dun naturaliste . . . (3 volumes), Paris.
Diesing, Molly 1990, Bare plural subjects and the stage/individual contrast, University
of Massachusetts, Amherst, MS.
Dihoff, Ivan R. (ed.) 1983, Current Approaches to African Linguistics (Publications in
African Languages and Linguistics 1), Dordrecht: Foris.
Dillard, Joey Lee 1970, Principles in the history of American English: Paradox, virginity, and cafeteria, Florida FL Reporter 8: 323.
Dimmendaal, Gerrit J. (ed.) 1986, Current Approaches to African Linguistics (Publications in African Languages and Linguistics 3), Dordrecht: Foris.
Di Sciullo, Anne-Marie 1990, On the properties of clitics, in Di Sciullo and Rochette
(eds.), pp. 20932.
Di Sciullo, Anne-Marie and Rochette, Anne (eds.) 1990, Binding in Romance: Essays in
Honour of Judith McANulty, Ottawa: Canadian Linguistic Association.
Di Sciullo, Anne-Marie and Williams, Edwin 1987, On the Definition of Word (Linguistic
Inquiry Monograph Series 14), Cambridge, MA: MIT Press.
Dolphyne, Florence 1971, A classification of Akan verb stems, Actes du 8e congrs de
la Socit linguistique de lAfrique occidentale, Srie H, Abidjan: Annales de
lUniversit dAbidjan.
Domingue, Nicole 1980, Syntactic innovations in Mauritian Bhojpuri, McGill University, MS.
1981, Internal change in a transplanted language, Studies in the Linguistic Sciences 4:
1519.
Dressler, Wolfgang U. and Pfeiffer, Oskar E. (eds.) 1976, Phonologica, Innsbruck:
Innsbrucker Beitrge zur Sprachwissenschaft.
Dubois, Jean 1962, tude sur la drivation suffixale en franais moderne et contemporain,
Paris: Larousse.
Dubois, Jean, Lagane, Ren and Lerond, Alain 1992, Dictionnaire du franais classique:
le XVIIe sicle, Paris: Larousse.
Ducurjoly, S. J. 1802, Manuel des habitans de Saint-Domingue, Paris: Lenoir.
Dumais, Danielle 1988, INFL en crole hatien, in Lefebvre (ed.), pp. 24265.
1992, Classification smantique des verbes du fon et du crole hatien, in Lefebvre
and Lumsden (eds.), volume V, pp. 1217.
Durie, Mark In progress, Language, Function and Time, Oxford University Press.

432

References

Edmonson, Jerold A. and Plank, Frans 1978, Great expectations: an intensive self analysis, Linguistics and Philosophy 2: 373413.
Emirkanian, Louisette and Sankoff, David L. 1984, Le futur simple et le futur priphrastique, in Lemieux-Nieger and Cedergren (eds.), pp. 17791.
Fabb, Nigel 1992, The licensing of Fon verbs, in Lefebvre (ed.), pp. 2734.
Faine, Jules 1937, Philologie crole: tudes historiques et tymologiques sur la langue
crole dHati, Port-au-Prince: Imprimerie de ltat.
Faltz, Leonard M. 1985, Reflexivisation: A Study in Universal Syntax, New York and
London: Garland Publishing.
Fasold, Ralph W. and Schiffrin, Deborah (eds.) 1989, Language Change and Variation
(Current Issues in Linguistic Theory 52), Amsterdam/Philadelphia: John Benjamins.
Fauchois, Anne 1983, Opposition verbo-nominale: Rflexions sur le crole hatien,
Paris V, MS.
Fraud, Jean-Franois 1768, Dictionnaire grammatical de la langue franaise contenant
toutes les rgles de lorthographe, de la prononciation (Microfiches, 105 148 mm,
microcopy of Paris edition, 1768), Paris: France-expansion, 1973, collection Archives
de la linguistique franaise.
Frre, Grard A. 1974, Haitian creole sound-system, form-classes, texts, Doctoral
Dissertation, University of Pennsylvania, Philadelphia.
Ferguson, Charles A. 1971, Absence of copula and the notion of simplicity: A study
of normal speech, baby talk, foreigner talk, and pidgins, in Hymes (ed.), pp. 141
50.
Ferraz, Luiz 1983, The origin and development of four creoles in the Gulf of Guinea
in the social context of creolisation, in Woolford and Washabaugh (eds.), pp. 120
5.
Filipovich, Sandra 1986, Le peuplement dHati, in Lefebvre and Kaye (eds.), pp. 386
418.
1988, La morphologie de lhatien, Masters Thesis, Universit du Qubec Montral,
Montreal.
Fltre, Louis-Ferdinand 1970, Le moyen picard, daprs les textes littraires du temps
(15601660): Textes, lexique, grammaire, Paris: Presses du Palais-Royal.
Foley, William A. 1988, Language birth: The processes of pidginisation and creolisation,
in Newmeyer (ed.) (1988b), pp. 16283.
Fouch, Pierre 1952, Introduction (Phontique historique du franais I), Paris: Librairie
C. Klincksieck.
Fournier, Robert 1977, N ap f yun ti-kose su la (La grammaire de la particule la en
crole hatien), Masters Thesis, Universit du Qubec Montral.
1987, Pu en crole hatien, Revue qubcoise de linguistique thorique et applique
6: 10726.
Freeman, Bryant C. 1988, Dictionnaire inverse de la langue hatienne, Lawrence, KS.
Fukui, Naoki, Rapoport, Tova and Sagey, Elizabeth (eds.) 1986, MIT Working Papers in
Linguistics 8, Cambridge, MA: MIT.
Fukui, Naoki and Speas, Margaret 1986, Specifiers and projection, in Fukui, Rapoport
and Sagey (eds.), pp. 12873.
Furetire, Antoine 1984, Le dictionnaire universel dAntoine Furetire (3 volumes), Paris:
Le Robert.
Galet, Yvette 1971, Lvolution de lordre des mots dans la phrase franaise de 1600
1700, Paris: Presses universitaires de France.

References

433

Gambhir, Surendra K. 1988, Structural development in Guyanese Bhojpuri, in Barz and

Siegel (eds.), pp. 6994.
Gass, Susan M. and Selinker, Larry (eds.) 1983, Language Transfer in Language Learning, Rowley, MA: Newbury House.
Gauthier, Pierre 1995, Le systme des pronoms dans le Rolea, recueil anonyme poitevin
du XVIIe sicle, in Simoni-Aurembou (ed.), pp. 31122.
Gendron, Jean-Denis 1966, Tendances phontiques du franais parl au Canada, Paris/
Quebec City: Librairie C. Klincksieck/Presses de lUniversit Laval.
Ghomeshi, Jila and Massam, Diane 1994, Lexical/syntactic relations without projection,
Linguistic Analysis 24: 175217.
Gilbert, Glen G. (ed.) 1987, Pidgin and Creole Languages: Essays in Memory of John E.
Reinecke, Honolulu: University of Hawaii Press.
Gilles, Rolande 1988, Ralisation du Cas en crole hatien dans trois environnements:
les prpositions, les marqueurs de Cas et les noms, Masters Thesis, Universit du
Qubec Montral.
Girodet, Jean 1976, Logos/Grand dictionnaire de la langue franaise, Paris: Bordas.
Giurescu, Anca 1975, Les mots composs dans les langues romanes, The Hague/Paris: Mouton.
Givn, Talmy 1971, On the verbal origin of the Bantu suffixes, Studies in African
Linguistics 2: 14563.
Gbl-Galdi, Lszl 1934, La structure grammaticale des patois franais-croles, Zeitschrift
fr franzsische Sprache und Literatur 58: 25795.
Goldsmith, John 1978, Que cest quoi? Que, cest quoi , Recherches linguistiques
Montral 11: 115.
Goodman, Morris F. 1964, A Comparative Study of Creole French Dialects, The Hague:
Mouton.
1971, The strange case of Mbugu (Tanzania), in Hymes (ed.), pp. 24355.
Gougenheim, Georges 1973, Grammaire de la langue franaise au seizime sicle, Paris:
A. et J. Picard.
Green, Georgia M. 1974, Semantics and Syntactic Regularity, Bloomington: Indiana
University Press.
Greenberg, Joseph H. 1966a, Some universals of grammar with particular reference to
the order of meaningful elements, in Greenberg (ed.), pp. 73113.
Greenberg, Joseph H. (ed.) 1966b, Universals of Language, Cambridge, MA: MIT Press.
Greenberg, Joseph H. 1983, Some areal characteristics of African languages, in Dihoff
(ed.), pp. 322.
Grevisse, Maurice 1975, Le Bon Usage, Paris/Gembloux: Duculot.
Grimshaw, Jane 1989, Getting the dative alternation, in Laka Mugarza and Mahajan
(eds.), pp. 11322.
Grimshaw, Jane and Meister, Armin 1988, Light verbs and -marking, Linguistic Inquiry
19: 20532.
Gros 1791, Isle St.-Domingue, Province du Nord. Rcit historique sur les vnements qui
se sont succds . . . depuis le 26 octobre 1791, jusquau 24 dcembre de la mme
anne, n.p.
Gruber, Jeff S. 1992, Proper argument projection in Igbo and Yoruba, in Collins and
Manfredi (eds.), pp. 13965.
Guerssel, Mohamed 1986, On Berber verbs of change: A study of transitivity alternations, MIT Lexicon Project Working Papers 9, Cambridge, MA: MIT Center for
Cognitive Science.

434

References

Guiraud, Pierre 1966, Le moyen franais, 2nd edition, Paris: Presses universitaires de
France, collection Que sais-je?
Haase, Albert 1965, Syntaxe franaise du XVIIe sicle, 5th edition, Paris: Delagrave.
Hagge, Claude 1974, Les pronoms logophoriques, Bulletin de la Socit de linguistique
de Paris 1: 287310.
1985, Lhomme de parole, Paris: Librairie Arthme Fayard.
Hair, P. E. H. 1966, Collections of vocabularies of Western Africa before the polyglotta:
A key, Journal of African Languages 5: 20817.
1970, The contribution of early linguistic material to the history of West Africa, in
Dalby (ed.), pp. 5063.
Hale, Ken and Keyser, Jay 1986, Some transitivity alternations in English, MIT Lexicon
Project Working Papers 7, Cambridge, MA: MIT Center for Cognitive Science.
1987, A view from the middle, MIT Lexicon Project Working Papers 10, Cambridge,
MA: MIT Center for Cognitive Science.
Hale, Ken and Laughren, Margareth 1983, The structure of verbal entries: Preface to
dictionary entries of verbs, Walpiri Lexicon Project, Cambridge, MA: MIT.
Hall, Joan H., Doane, Nick and Ringler, Dick (eds.) 1992, Old English and New: Studies
in Language and Linguistics in Honor of Frederic G. Cassidy, New York: Garland.
Hall, Robert A. 1950, The genetic relationships of Haitian Creole, Ricerche Linguistiche
1: 194203.
1953, Haitian Creole: Grammar, Texts and Vocabulary (American Anthropologist
Association 74), Menasha: Georges Banta Publishing Company.
1958, Creolised languages and genetic relationships , Word 14: 36773.
1966, Pidgin and Creole Languages, Ithaca, NY: Cornell University Press.
Halle, Morris, Bresnan, Joan and Miller, George (eds.) 1978, Linguistic Theory and
Psychological Reality, Cambridge, MA: MIT Press.
Hancock, Ian F. 1980a, Gullah and Barbadian: Origins and relationships, American
Speech 55: 1735.
1980b, Lexical expansion in creole languages, in Valdman and Highfield (eds.), pp.
6387.
1987, A preliminary classification of the anglophone Atlantic creoles with syntactic
data from thirty-three representative dialects, in Gilbert (ed.), pp. 264 333.
Harris, John W. 1991, Kriol/the creation of a new language, in Romaine (ed.), pp. 195
203.
Hazoum, M. L. 1990, Essai de Classification Synchronique (tude comparative des
parlers Gbe du Sud-Bnin), Cotonou, Bnin: Centre national de linguistique applique.
Heine, Bernd and Reh, Mechthild 1984, Grammaticalisation and Reanalysis in African
Languages, Hamburg: Helmut Buske Verlag.
Heinl, Robert D. and Heinl, Nancy G. 1978, Written in Blood: The Story of the Haitian
People, 14921971, Boston: Houghton Mifflin.
Hensey, Fritz G. and Lujan, Marta (eds.) 1976, Current Studies in Romance Linguistics,
Washington, DC: Georgetown University Press.
Herskovits, Melville J. 1975, Life in a Haitian Valley, New York: Octagon Books.
Hesseling, Dirk C. 1933, Een Spaans boek over het Papiaments, TNTL 52: 4057;
review of Rodolfo Lenz El Papiamento, la lengua criolla de Curazao (la grammatica
mas sencilla), Anales de la Universidad de Chile, Seperatum (Santiago, 1928).
Highfield, Arthur R. and Valdman, Albert (eds.) 1981, Historicity and Variation in Creole Studies, Ann Arbor, MI: Karoma.

References

435

Hilaire, Jeannot 1992, Soubassements amrindiens (Ldifice crole en Hati 1), Fribourg,
Switzerland: Edikreyol.
1993, Soubassements africains (Ldifice crole en Hati 2), Fribourg, Switzerland:
Edikreyol.
Hilliard dAuberteuil, M. R. 17767, Considrations sur ltat prsent de la colonie
franaise de Saint-Domingue (2 volumes), Paris: Grang.
Hirschbhler, Paul 1978, The Semantics and Syntax of Wh-Constructions, Doctoral
Dissertation, University of Massachusetts, Amherst.
Hoekstra, Teun 1992, Subjects inside out, Revue qubcoise de linguistique 22: 4576.
Holm, John 1986, Substrate diffusion, in Muysken and Smith (eds.), pp. 25978.
1988, Pidgins and Creoles (Cambridge Language Surveys 1), Cambridge University
Press.
1993, Phonological features common to some West African and Atlantic creole languages, in Mufwene (ed.), pp. 31727.
Hopper, Paul J. and Traugott, Elizabeth C. 1993, Grammaticalisation (Cambridge Textbooks in Linguistics), Cambridge University Press.
Horiot, Brigitte (ed.) 1991, Franais du Canada Franais de France, Tbingen: Niemeyer
(Actes du 2e Colloque international de Cognac (September 2730, 1988) ).
Hornstein, Norbert 1977, Towards a theory of tense, Linguistic Inquiry 8: 52157.
Hounkpatin, Basile 1985, Le verbal et le syntagme verbal du fon-gbe parl Mass,
Doctoral Dissertation, Universit de la Sorbonne nouvelle, Paris III.
Huguet, Edmond 1925, Dictionnaire de la langue franaise du seizime sicle, volume I,
Paris: M. Didier.
Hulk, Aafke 1986, Subject clitics and the PRO-drop parameter, in Coopmans, Bordelois
and Smith (eds.), pp. 10720.
Hull, Alexander 1975, On the origin and chronology of the French-based creoles, paper
read at the International Conference on Pidgins and Creoles, University of Hawaii.
1979, Affinits entre les varits du franais, in Valdman (ed.), pp. 16582.
Hutchison, John 1989, On interpreting the Haitian predicate cleft construction, Boston
University, MS.
Huttar, Georges L. 1971, Sources of creole semantic structures, Language 51: 684 95.
Hyams, Nina M. 1986, Language Acquisition and the Theory of Parameters, Dordrecht:
Reidel.
1987, Theory of parameters and syntactic development, in Roeper and Williams
(eds.), pp. 122.
Hyman, Larry M. and Comrie, Bernard 1981, Logophoric reference in Gokana, Journal
of African Languages and Linguistics 3: 1937.
Hymes, Dell 1971a, Introduction, in Hymes (ed.), pp. 6591.
Hymes, Dell (ed.) 1971b, Pidginisation and Creolisation of Languages, Cambridge University Press (Proceedings of a conference held at the University of West Indies,
Mona, Jamaica, April 1968).
Ileana, Paul and Phillips, Vivianne (eds.) In press, Formal Issues in Austronesian Linguistics, Dordrecht: Kluwer Academic Publisher.
Imbs, Paul (ed.) 1979, Trsor de la langue franaise, Paris: CNRS.
Jackendoff, Ray 1976, Toward an explanatory semantic representation, Linguistic
Inquiry 7: 89150.
1977, X Syntax: A Study of Phrase Structure (Linguistic Inquiry Monograph 2), Cambridge, MA: MIT Press.

436

References

Jackendoff, Ray (ed.) 1990, Semantic Structures (Current Studies in Linguistics 18),
Cambridge, MA: MIT Press.
Jaeggli, Osvaldo A. 1984, Subject extraction and the null subject parameter, in Jones
and Sells (eds.), pp. 13253.
Jaeggli, Osvaldo A. and Silva-Corvaln, Carmen (eds.) 1986, Studies in Romance Linguistics (Publications in Language Sciences 24), Dordrecht: Foris.
Jake, Janice and Odden, David 1979, Raising in Kipsigis, Studies in the Linguistic
Sciences 9: 13155.
James, Deborah 1984, Raising to subject in Moose Cree: A problem for subjacency, in
Cook and Gerdts (eds.), pp. 20513.
Jansen, Bert, Koopman, Hilda and Muysken, Pieter C. 1978, Serial verbs in the creole
languages, Amsterdam Creole Studies 11: 12559.
Janvier, Louis Joseph 1886, Berceuse hatienne, Zangui (LAnguille), Revue des Traditions Populaires 1: 213, 107.
Jeanjean, Colette 1988, Le futur simple et le futur priphrastique en franais parl, in
Blanche-Benveniste, Chervel and Gross (eds.), pp. 23557.
Jespersen, Otto 1922, Language: Its Nature, Development and Origin, London: Allen and
Unwin.
Johns, Alana 1992, Deriving ergativity, Linguistic Inquiry 17: 587622.
Johnson, Kyle 1991, Object positions, Natural Language and Linguistic Theory 9: 577
639.
Joly, Genevive 1995, Prcis de phontique historique du franais, Paris: A. Colin.
Jones, Charles and Sells, Peter (eds.) 1984, Proceedings of NELS 14, University of
Massachusetts at Amherst, GLSA.
Joseph, Frantz 1989, La dtermination nominale en crole hatien, Doctoral Dissertation, Universit de Paris VII, Jussieu.
Joseph, Joseph Sauveur 1994, De ltude smantico-syntaxique de la locution nan du
crole hatien, Masters Thesis, Universit du Qubec Montral.
1995, La ngation en crole Hatien: un cas de relexification, Universit du Qubec
Montral MS.
Jourdan, Christine 1985, Sapos Yumi Mitim Yumi: Urbanisation and creolisation
of Solomon Islands pigin, Doctoral Dissertation, Australian National University,
Canberra.
Julliand, Alphonse 1965, Dictionnaire inverse de la langue franaise, The Hague: Mouton.
Juneau, Marcel 1972, Contribution lhistoire de la prononciation franaise au Qubec,
Paris/Quebec: Librairie C. Klincksieck/Presses de lUniversit Laval.
Juneau, Marcel and Poirier, Claude 1973, Le livre de comptes dun meunier qubcois
(fin XVIIe-dbut XVIIIe sicle). dition avec tude linguistique, Quebec City: Les
Presses de lUniversit Laval.
Jungraithmayr, Hermans and Mhlig, Wilhelm J. G. (eds.) 1983, Lexicon der Afrikanistik,
Berlin: Dietrich Reimer.
Kant, Immanuel 1949, The Philosophy of Kant (The Modern Library), New York: Random House, p. 132.
Kay, Paul and Sankoff, Gillian 1974, A language-universals approach to pidgins and
creoles, in DeCamp and Hancock (eds.), pp. 6172.
Kaye, Jonathan, Koopman, Hilda, Sportiche, Dominique and Dugas, Andr (eds.) 1983,
Current Approaches to African Linguistics (Publications in African Languages and
Linguistics 2), Dordrecht: Foris.

References

437

Kayne, Richard S. 1975, French Syntax: The Transformational Cycle (Current Studies in
Linguistics 6), Cambridge, MA: MIT Press.
1976, French relative que, in Hensey and Lujan (eds.), pp. 25599.
1981, On certain differences between French and English, Linguistic Inquiry 12:
34971.
1984, Connectedness and Binary Branching, Dordrecht: Foris.
1994, The Antisymmetry of Syntax (Linguistic Inquiry Monograph 25), Cambridge,
MA: MIT Press.
Kayne, Richard and Pollock, Jean-Yves 1978, Stylistic inversion, successive cyclicity
and Move NP in French, Linguistic Inquiry 9: 595621.
Keenan, Edward 1976, Remarkable subjects in Malagasy, in Li (ed.), pp. 247301.
Keesing, Roger M. 1988, Melanesian Pidgin and the Oceanic Substrate, Stanford University Press.
Kenesei, Istvan (ed.) 1987, Approaches to Hungarian, Szeged: JATE.
Kihm, Alain 1989, Lexical conflation as a basis for relexification, in Lefebvre and
Lumsden (eds.) (1989b), pp. 35176.
Kihm, Alain (ed.) 1991, La Crolisation: Thorie et applications (Recherches linguistiques
de Vincennes 20), Paris: Presses universitaires de Vincennes.
Kihm, Alain 1994, FrenchGbe lexical conflations in the genesis of Haitian, in Lefebvre
and Lumsden (eds.) (1994c), 24 pages.
Kihm, Alain and Lefebvre, Claire (eds.) 1993, Aspects de la grammaire du Fongb:
tudes de phonologie, de syntaxe et de smantique, Paris: Peeters.
1995, Les proprits lexicales du crole hatien et du fongbe. Special issue of
Linguistique africaine 14.
Kinyalolo, Kasangati K. W. 1992, A note on word order in the progressive and prospective in Fon, in Lefebvre (ed.), pp. 3551.
1993a, Conditionals and factives in Fwn, in Lefebvre and Lumsden (eds.), volume IV,
pp. 16173.
1993b, On some syntactic properties of #W in Fon, Lingua 91: 20133.
1993c, The logophoric pronoun mi as a LF operator-anaphor, in Schafer (ed.),
pp. 22337.
1994, Pronouns and relexification, in Lefebvre and Lumsden (eds.) (1994e), volume
III, pp. 25993.
Kiple, Kenneth 1984, The Caribbean Slave: A Biological History, Cambridge University
Press.
Klemperer, Victor and Lerch, Eugen (eds.) 1925, Jahrbuch fr Philologie, Mnchen:
Verlag der Hochschulbuchhandlung Max Hueber.
Koelle, S. W. 1854, Polyglotta Africana, London: Church Missionary Society.
Koopman, Hilda 1982a, Les constructions relatives, in Lefebvre, Magloire-Holly and
Piou (eds.), pp. 167203.
1982b, Les questions, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 20433.
1982c, Quelques problmes concernant que/quoi, ce que et quest-ce que, in Lefebvre
(ed.), pp. 13570.
1984, The Syntax of Verbs: From Verb Movement Rules in the Kru Languages to
Universal Grammar, Dordrecht: Foris.
1986, The genesis of Haitian: Implications of a comparison of some features of the
syntax of Haitian, French and West African languages, in Muysken and Smith
(eds.), pp. 23158.

438

References

Koopman, Hilda and Lefebvre, Claire 1981, Haitian Creole pu , in Muysken (ed.),
pp. 20121.
1982, PU: marqueur de mode, prposition et complmenteur, in Lefebvre, MagloireHolly and Piou (eds.), pp. 6491.
Koopman, Hilda and Sportiche, Dominique 1989, Pronouns, logical variables and logophoricity in Abe, Linguistic Inquiry 20: 55588.
Kouwenberg, Silvia 1987, Eastern Ijo-derived words in Berbice-Dutch, University of
the West Indies at Mona, Jamaica, MS.
1990, Reduplication in Berbice Dutch Creole, in Boretzky and Stolz (eds.), pp. 75102.
1992, From OV to VO. Linguistic negotiation in the development of Berbice Dutch
Creole, Lingua 88: 26399.
1993, Cliticisation of pronouns in Berbice Dutch and Eastern Ijo, in Byrne and Holm
(eds.), pp. 11932.
1994, A Grammar of Berbice Dutch Creole, Berlin/New York: Mouton de Gruyter.
Kratzer, Angelika 1989, Stage-level and individual-level predicates in Bach, Kratzer
and Partee (eds.).
Labelle, Micheline, Beaudet, Gatan, Lvy, Joseph and Tardif, Francine 1993, La question nationale dans le discours de leaders dassociations ethniques de la rgion de
Montral, Cahiers de recherche sociologique 20: 85111.
Laborderie, Nolle 1994, Prcis de phontique historique, Paris: Nathan.
Labouret, Henri and Rivet, Paul 1929, Le royaume dArda et son vanglisation au XVIIe
sicle (Travaux et mmoires de lInstitut dethnologie), Universit de Paris.
Labov, William 1972, Sociolinguistic Patterns, Philadelphia: The University of Pennsylvania Press Inc.
Lafage, Suzanne 1985, Franais crit et parl en pays Ew (Sud-Congo), Paris: SELAF.
Laka Mugarza, Miren Itziar 1990, Negation in syntax: On the nature of functional
categories and projections, Doctoral Dissertation, MIT, Cambridge, MA.
Laka Mugarza, Miren Itziar and Mahajan, A. (eds.) 1989, Functional Heads and Clause
Structure (MIT Working Papers in Linguistics 10), Cambridge, MA: Department of
Linguistics and Philosophy, MIT.
Lappin, Shalom 1984, Predication and raising, in Jones and Sells (eds.), pp. 23652.
Larson, Richard K. 1988, On the double object construction, Linguistic Inquiry 19:
33591.
1990, Double objects revisited: Reply to Jackendoff, Linguistic Inquiry 21: 589632.
1991, Some issues in verb serialisation, in Lefebvre (ed.), pp. 185211.
Larson, Richard K. and Lefebvre, Claire 1991, Predicate clefting in Haitian Creole, in
Sherer (ed.), pp. 24762.
Law, Paul 1991, Verb movement, expletive replacement, and head-government, The
Linguistic Review 8: 25385.
1992, What is a null subject language?, Travaux de recherche sur le crole hatien
11: 4158 (Universit du Qubec Montral).
1994a, Une note sur la construction existentielle en Hatien, Revue qubcoise de
linguistique 23: 13967.
1994b, Resumptives in Haitian and relexification, in Lefebvre and Lumsden (eds.)
(1994c), 36 pages.
Law, Paul and Lefebvre, Claire 1995, On the relationship between event determiners and
predicate cleft in the Kwa languages: The case of Fongbe, in Kihm and Lefebvre
(eds.), pp. 747.

References

439

Lefebvre, Claire 1981, The double structure of questions in French. A case of syntactic
variation, in Sankoff and Cedergren (eds.), pp. 22939.
1982a, Introduction, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 120.
1982b, Lexpansion dune catgorie grammaticale: le dterminant la, in Lefebvre,
Magloire-Holly and Piou (eds.), pp. 2163.
Lefebvre, Claire (ed.) 1982c, La syntaxe compare du franais standard et populaire:
approches formelle et fonctionnelle (2 volumes), Montreal: Gouvernement du Qubec
(Collection langues et socit).
Lefebvre, Claire 1982d, Le rpertoire de mots Wh en franais vernaculaire et leur insertion dans la grammaire du franais, in Lefebvre (ed.), pp. 73125.
1982e, Qui qui vient ou Qui vient: Voil la question, in Lefebvre (ed.), pp. 47101.
1984, Grammaires en contact: dfinition et perspectives de recherche, Revue qubcoise
de linguistique 14: 1149.
1986, Relexification in creole genesis revisited: The case of Haitian Creole, in Muysken
and Smith (eds.), pp. 279301.
Lefebvre, Claire (ed.) 1988, tudes syntaxiques, morphologiques et phonologiques.
Research report prepared for SSHRCC, FCAR and PAFAC on the project HaitiFon, Universit du Qubec Montral.
Lefebvre, Claire 1989, Some problems in defining syntactic variables: The case of questions in Montreal French, in Fasold and Schiffrin (eds.), pp. 35166.
1990a, Establishing a syntactic category P in Fon, Journal of West African Languages 20: 4563.
1990b, On the interpretation of predicate cleft, Linguistic Review 6: 16994.
Lefebvre, Claire (ed.) 1991a, Serial Verbs: Grammatical, Comparative and Cognitive
Approaches (Studies in the Sciences of Language Series 8), Amsterdam/Philadelphia: John Benjamins.
Lefebvre, Claire 1991b, Take serial verb constructions in Fon, in Lefebvre (ed.),
pp. 3778.
1991c, La distribution du dterminant et des complmenteurs en crole hatien, in
Kihm (ed.), pp. 2147.
1991d, Note sur la position de AGRO dans la structure syntaxique, Revue qubcoise
de linguistique 20: 28194.
1992a, On the distribution of clausal wY in Fongbe, Journal of West African Languages 21: 2135.
1992b, AGR in languages without person and number agreement: The case of the
clausal determiner in Haitian and Fon, in Lefebvre, Lumsden and Travis (eds.),
pp. 13756.
Lefebvre, Claire (ed.) 1992c, Topics on the Syntax and Semantics of Fongbe. Special
issue of Journal of West African Languages 22.
Lefebvre, Claire 1992d, The double object construction and the DP hypothesis, McGill
Working Papers in Linguistics 7: 11331.
1993a, The role of relexification and syntactic reanalysis in Haitian Creole: Methodological aspects of a research program, in Mufwene (ed.), pp. 25479.
1993b, The role of relexification in creole genesis: The case of functional categories,
Travaux de recherche sur le crole hatien 14: 2352 (Universit du Qubec
Montral).
1993c, Dominance versus precedence in the double object construction: New facts
from Fongbe, Revue canadienne de linguistique 38: 395425.

440

References

1993d, Le clivage du prdicat: essai de typologie, in Kihm and Lefebvre (eds.),

pp. 5160.
1994a, Functional categories and relexification, in Lefebvre and Lumsden (eds.)
(1994c), 74 pages.
1994b, On spelling out E, Travaux de recherche sur le crole hatien 23: 133
(Universit du Qubec Montral).
1994c, New facts from Fongbe on the double object constructions, Lingua 94: 69163.
Lefebvre, Claire (ed.) 1995a, Research report prepared for FCAR on the project
Lorganisation des lexiques et des entres lexicales (4 volumes), Universit du
Qubec Montral.
Lefebvre, Claire 1995b, PPs headed by the Fon preposition n are syntactic adjuncts, in
Kihm and Lefebvre (eds.), pp. 15383.
Lefebvre, Claire (ed.) 1995c, Research report prepared for SSHRCC on the project
La moisson de 10 ans de recherche sur la gense du crole hatien, Universit du
Qubec Montral.
Lefebvre, Claire 1996a, The functional category AGR and creole genesis, in Wekker
(ed.), pp. 15395.
1996b, The tense, mood and aspect system of Haitian Creole and the problem of
transmission of grammar in creole genesis, Journal of Pidgin and Creole Languages 11: 231313.
1997, Relexification in creole genesis: The case of demonstrative terms in Haitian
Creole, Journal of Pidgin and Creole Languages 12.2: 181203.
To appear a, Relexification: The central process in creole genesis, its effects on the
development of the creole, in Smith and Veenstra (eds.).
To appear b, Multifunctionality and variation among grammars: The case of the determiner in Haitian and Fongbe, Journal of Pidgin and Creole Languages.
In preparation, The category P in Haitian.
In press, Substratum semantics in the verbal lexicon of Haitian creole, Studies in
Language.
Lefebvre, Claire and Kaye, Jonathan (eds.) 1986, tudes syntaxiques, morphologiques et
phonologiques. Research report prepared for FCAR and PAFAC on the project FonCrole Hatien, Universit du Qubec Montral.
Lefebvre, Claire and Lumsden, John S. 1989a, Les langues croles et la thorie linguistique, in Lefebvre and Lumsden (eds.) (1989b), pp. 24972.
Lefebvre, Claire and Lumsden, John S. (eds.) 1989b, La crolisation. Special issue of the
Revue canadienne de linguistique 34.3.
Lefebvre, Claire and Lumsden, John S. (eds.) 1989c, Le crole hatien. Special issue of
the Revue qubcoise de linguistique 18.2.
Lefebvre, Claire and Lumsden, John S. 1992a, On word order in relexification, Travaux
de recherche sur le crole hatien 10: 122 Universit du Qubec Montral.
Lefebvre, Claire and Lumsden, John S. (eds.) 1992b, Classification smantique des verbes
du fon et du crole hatien. Research report prepared for SSHRCC, FCAR and
PAFAC on the project La gense du crole hatien: un cas particulier dinvestigation
sur la forme de la grammaire universelle (5 volumes), Universit du Qubec
Montral.
Lefebvre, Claire and Lumsden, John S. (eds.) 1993, tudes syntaxiques. Research report
prepared for SSHRCC, FCAR and PAFAC on the project La gense du crole
hatien: un cas particulier dinvestigation sur la forme de la grammaire universelle
(4 volumes), Universit du Qubec Montral.

References

441

Lefebvre, Claire and Lumsden, John S. 1994a, Le rle central de la relexification dans la
gense des langues croles, Plurilinguismes 8: 4793.
1994b, Relexification in creole genesis, in Lefebvre and Lumsden (eds.) (1994c), 28
pages.
Lefebvre, Claire and Lumsden, John S. (eds.) 1994c, The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole. Research report prepared for
SSHRCC on the project La gense du crole hatien: un cas particulier dinvestigation
sur la forme de la grammaire universelle, Universit du Qubec Montral.
Lefebvre, Claire and Lumsden, John S. (eds.) 1994d, Les classes aspectuelles de verbes.
Research report prepared for FCAR on the project Lorganisation des lexiques et des
entres lexicales (3 volumes), Universit du Qubec Montral.
Lefebvre, Claire and Lumsden, John S. (eds.) 1994e, tudes syntaxiques. Research report
prepared for SSHRCC and PAFAC on the project La gense du crole hatien: un
cas particulier dinvestigation sur la forme de la grammaire universelle (3 volumes),
Universit du Qubec Montral.
Lefebvre, Claire and Lumsden, John S. 1995, Relexification in creole genesis, in Lefebvre
(ed.) (1995c), 79 pages.
Lefebvre, Claire, Lumsden, John S. and Travis, Lisa (eds.) 1992, Functional Categories.
Special issue of the Revue canadienne de linguistique, 37.2.
Lefebvre, Claire, Magloire-Holly, Hlne and Piou, Nanie (eds.) 1982, Syntaxe de lhatien,
Ann Arbor, MI: Karoma.
Lefebvre, Claire and Massam, Diane 1988, Haitian Creole syntax: A case for DET as
head, Journal of Pidgin and Creole Languages 3: 21343.
Lefebvre, Claire and Muysken, Pieter C. 1988, Mixed Categories: Nominalisations in
Quechua (Studies in Natural Language and Linguistic Theory), Dordrecht: Kluwer
Academic Press.
Lefebvre, Claire and Ritter, Elizabeth 1993, Two types of predicate doubling adverbs in
Haitian Creole, in Byrne and Winford (eds.), pp. 6591.
Lemieux-Nieger, Monique and Cedergren, Henrietta (eds.) 1984, Les tendances dynamiques
du franais parl Montral, Universit du Qubec Montral.
Levin, Beth and Rappaport, Malka 1986, The formation of adjectival passives, Linguistic
Inquiry 17: 62361.
Levinson, Stephen C. 1991, Pragmatic reduction of the binding conditions revisited,
Journal of Linguistics 27: 10761.
Li, Charles N. (ed.) 1976, Subject and Topic, New York: Academic Press.
Lichtenberk, Frantisek 1984, Toabaita Language of Malaita: Solomon Islands (Working papers in Anthropology, Archaeology, Linguistics, Maori Studies 65), University of Auckland.
Lieber, Rochelle 1980, On the organisation of the lexicon, Doctoral Dissertation, MIT,
Cambridge, MA.
1992, Deconstructing Morphology, University of Chicago Press.
Lightfoot, David 1979, Principles of Diachronic Syntax (Cambridge Studies in Linguistics
23), Cambridge University Press.
Lipou, A. 1983, Du statut des concordial elements dans les langues bantu, in Kaye,
Koopman, Sportiche and Dugas (eds.), pp. 36978.
Lommatzsch, E. 1925, Deiktische Elemente im Altfranzsischen, in Klemperer and
Lerch (eds.), pp. 20244.
Lord, Carol 1976, Evidence for syntactic reanalysis: From verb to complementiser in
Kwa, in Steever, Walker and Mufwene (eds.), pp. 17991.

442

References

Lumsden, John S. 1987, Syntactic features: Parametric variation in the history of English, Doctoral Dissertation, MIT, Cambridge, MA.
1989, On the distribution of determiners in Haitian Creole, in Lefebvre and Lumsden
(eds.) (1989c), pp. 6493.
1990, The bi-clausal structure of Haitian clefts, Linguistics 28: 74160.
1991, On the acquisition of nominal structures in the genesis of Haitian Creole, Travaux
de recherche sur le crole hatien 4: 3761 (Universit du Qubec Montral).
1993a, Expletives in double object constructions in Haitian Creole, in Byrne and
Winford (eds.), pp. 26984.
1993b, Relexification and parameter setting in creole genesis, Travaux de recherche
sur le crole hatien 14: 5269 (Universit du Qubec Montral).
1994a, Possession: Substratum semantics in Haitian creole, Journal of Pidgin and
Creole Languages 9: 2551.
1994b, Relexification and lexical semantics, in Lefebvre and Lumsden (eds.) (1994c),
36 pages.
1995, On the complex nature of grammatical simplification in pidgins and early creole
languages, paper read at the Conference of the Society for Pidgin and Creole
Linguistics (LSA), New Orleans.
In press a, Language acquisition and creolisation, in De Graff (ed.).
In press b, Aspect and lexical semantics in Haitian Creole, The Linguistic Review.
Lumsden, John S. and Lefebvre, Claire 1990, Predicate cleft constructions and why they
arent what you might think, Linguistics 28: 76183.
Lumsden, John S. and Lefebvre, Claire 1994, The genesis of Haitian Creole, in Lefebvre
and Lumsden (eds.) (1994c), 15 pages.
Magloire-Holly, Hlne 1982, Les modaux: Auxiliaires ou verbes?, in Lefebvre, MagloireHolly and Piou (eds.), pp. 92120.
Man, Michael and Dalby, David 1987, A Thesaurus of African Languages: A Classified
and Annotated Inventory of the Spoken Languages of Africa with an Appendix on
their Written Representation, London/Munich/New York/Paris: Hans Zell.
Manfredi, Victor 1992, A typology of Yorb nominalisations, in Collins and Manfredi
(eds.), pp. 20517.
1993, Verb focus in the typology of Kwa/Kru and Haitian, in Byrne and Winford
(eds.), pp. 351.
Manfredi, Victor and Reynolds, Karl (eds.) 1995, Niger-Congo Syntax and Semantics,
Boston University African Studies Center.
Marantz, Alec 1982, Re reduplication, Linguistic Inquiry 13: 43583.
Massam, Diane 1985, Case theory and the projection principle, Doctoral Dissertation,
MIT, Cambridge, MA.
1989, Predicate argument structure in Haitian Creole, in Lefebvre and Lumsden
(eds.) (1989c), pp. 95130.
1990, Cognate objects as thematic objects, Canadian Journal of Linguistics 35: 161
90.
McWhorter, John 1996, A deep breath and a second wind: The substrate hypothesis
reassessed, Anthropological Linguistics 38: 46194.
Medeiros, Franois de (ed.) 1984, Peuples du golfe du Bnin: Aja-Ew (Colloque de
Cotonou), Paris: Karthala.
Meisel, Jrgen M. 1983a, Strategies of second language acquisition. More than one kind
of simplification, in Andersen (ed.), pp. 12057.

References

443

1983b, Transfer as a second-language strategy, Language and Communication 3: 11

46.
Meyer-Lbke, Wilhelm 1909, Le franais au Canada, Bulletin du franais parl au
Canada 8: 1219.
Miller, George A. 1978, Semantic relations among words, in Halle, Bresnan and Miller
(eds.), pp. 60118.
Milner, Jean-Claude 1978, De la syntaxe linterprtation, Paris: ditions du Seuil.
Mintz, Sidney W. 1971, The socio-historical background to pidginisation and creolisation,
in Hymes (ed.), pp. 48198.
Mhlig, Wilhelm J. G. 1983, Mbugu, in Jungraithmayr and Mhlig (eds.), pp. 15859.
Moisy, H. 1969, Dictionnaire de patois normand (reprint of Caen edition, 1887), Geneva:
Slatkine reprints.
Moody, Marvin D. 1973, A Classification and Analysis of Noun+de+Noun Constructions in French, The Hague/Paris: Mouton.
Moore, Steven and Wyner, Adam (eds.) 1991, Proceedings of SALT I, Cornell University
Working Papers in Linguistics.
Moreau, Marie-Louise 1971, Lhomme que je crois qui est venu: que, qui relatifs et
conjonctions, Langue franaise 11: 7790.
Moreau de Saint-Mry, Louis-lie 1797, Description topographique, physique, civile,
politique et historique de la partie franaise de lisle de Saint-Domingue (2nd edn,
Paris: L. Gurin, 1875), Philadelphia.
Morin, Yves-Charles 1977, Interprtation des pronoms et des rflchis en franais,
Recherches linguistiques Montral 8: 6794.
Mougeon, Raymond and Beniak, douard (eds.) 1994, Les origines du franais qubcois,
Sainte-Foy: Les Presses de lUniversit Laval.
Mous, Maarten 1994, Maa or Mbugu, in Bakker and Mous (eds.), pp. 175200.
1995, Language intertwining, paper read at the Amsterdam Creole Workshop: Creole
Genesis and Language Contact, Amsterdam.
To appear, Maa as an ethno-register of Mbugu, Sprache und Geschichte in Afrika 15.
Mufwene, Salikoko S. 1986, The universalist and substrate hypotheses complement one
another, in Muysken and Smith (eds.), pp. 12962.
1989, Some explanations that strike me as incomplete, Journal of Pidgin and Creole
Languages 4: 11728.
1990, Transfer and the substrate hypothesis in creolistics, Studies in Second Language Acquisition 12: 123.
1991, Pidgins, creoles, typology, and markedness, in Byrne and Huebner (eds.),
pp. 12344.
1992, Africanisms in Gullah: A re-examination of the issues, in Hall, Doane and
Ringler (eds.), pp. 15682.
Mufwene, Salikoko S. (ed.) 1993a, Africanisms in Afro-American Language Varieties,
Athens, GA/London: The University of Georgia Press.
Mufwene, Salikoko S. 1993b, Are there possessive pronouns in Atlantic creoles?, in
Byrne and Holm (eds.), pp. 13344.
1993c, Crole, crolisation, substrat, et autres notions apparentes: quelques rflexions
sur Des les, des hommes, des langues. Langues croles cultures croles de Robert
Chaudenson, tudes croles 16: 11741.
1994, Gense de population et gense de langue, Crolistique et grammaire gnrative
8: 95113.

444

References

1996, Creole genesis: A population genetics perspective, in Christie (ed.), pp. 163
96.
Mhlhusler, Peter 1980, Structural expansion and the process of creolisation, in Valdman
and Highfield (eds.), pp. 1956.
1986a, Pidgin and Creole Linguistics (Language in Society II), Oxford: Blackwell.
1986b, Bonnet blanc and blanc bonnet: Adjective-noun order, substratum and language universals, in Muysken and Smith (eds.), pp. 4155.
Muysken, Pieter C. 1979, Substratum and stratification: The gerund in Ecuadorian Rural
Spanish, paper read at the 43rd Conference of the Americanists, Vancouver.
1981a, Half-way between Quechua and Spanish: The case for relexification, in
Highfield and Valdman (eds.), pp. 5279.
Muysken, Pieter C. (ed.) 1981b, Generative Studies on Creole Languages (Studies in
Generative Grammar 6), Dordrecht: Foris.
Muysken, Pieter C. 1981c, Creole tense/mood/aspect systems: The unmarked case?, in
Muysken (ed.), pp. 181200.
1987, Prepositions and postpositions in Saramaccan, in Alleyne (ed.), pp. 89102.
1988a, Lexical restructuring in creole genesis, in Boretzky, Enninger and Stolz (eds.),
pp. 193210.
1988b, Are creoles a special type of language?, in Newmeyer (ed.) (1988a), pp. 285
301.
1988c, Media Lengua and linguistic theory, The Canadian Journal of Linguistics 33:
40922.
1988d, Parameters for serial verbs, in Demuth and Manfredi (eds.).
1994a, Saramaccan and Haitian: A comparison, Journal of Pidgin and Creole Languages 9: 30514.
1994b, The search for universals in language genesis: tat de la question and research
program, Amsterdam Creole Studies 61: 1628.
Muysken, Pieter C. and Smith, Norval 1986a, Introduction: Problems in the identification of substratum features, in Muysken and Smith (eds.), pp. 114.
Muysken, Pieter C. and Smith, Norval (eds.) 1986b, Substrata Versus Universals in
Creole Genesis (Creole Language Library 1), Amsterdam/Philadelphia: John
Benjamins.
Muysken, Pieter C. and Smith, Norval 1990, Question words in pidgin and creole languages, Linguistics 28: 883903.
1995, Reflexives, in Arends, Muysken and Smith (eds.), pp. 27187.
Naro, Anthony J. 1978, A study of the origins of pidginisation, Language 54: 31447.
Ndayiragije, Juvnal 1993, Clivage du prdicat en fon, in Kihm and Lefebvre (eds.),
pp. 6186.
Newmeyer, Frederick J. (ed.) 1988a, Linguistic Theory: Extensions and Implications
(Linguistics: The Cambridge Survey 2), Cambridge University Press.
1988b, Language: The Socio-Cultural Context (Linguistics: The Cambridge Survey 4),
Cambridge University Press.
Nicolson 1776, Essai sur lhistoire naturelle de St. Domingue, Paris.
Nida, Eugne 1948, The identification of morphemes, Language 24: 41441.
Niobey, Georges, Lagane, Ren and Guilbert, Louis (eds.) 1971, Grand Larousse de la
langue franaise (7 volumes), Paris: Larousse.
Nyrop, Kristopher 1936, Formation des mots (Grammaire historique de la langue franaise
III), Copenhagen: Gyldendalske Boghandel, Nordisk Forlag.

References

445

Obenauer, Hans-Georg 1976, tudes de syntaxe interrogative du franais. Quoi, combien

et le complmenteur, Tbingen: Max Niemeyer.
1977, Syntaxe et interprtation, Le franais moderne 45: 30541.
Oehrle, Richard T. 1976, The grammatical status of the English Dative alternation,
Doctoral Dissertation, MIT, Cambridge, MA.
Ogunbewale, P. O. 1970, The Essentials of the Yoruba Language, Edinburgh: University
of London Press.
Orjala, Paul R. 1970, A dialect survey of Haitian creole, Doctoral Dissertation, Hartford
Seminary Foundation, Ann Arbor, MI.
Oudin, Antoine 1640, Grammaire francoyse, rapporte au langage du temps [Archives
de la linguistique franaise 306], Paris: A. de Sommaville.
Ouhalla, Jamal 1991, Functional Categories and Parametric Variation, London: Routledge.
Papen, Robert 1988, Convergence et divergence en mtif, paper read at the Dpartement
de linguistique, Universit du Qubec Montral.
Pazzi, R. 1976, Lhomme Ew, Gn, Aje, Fon et son Univers (Dictionnaire I: Lexique des
noms), Lom (Microfiche).
Pierret, Jean-Marie 1983, Phontique du franais. Notions de phontique gnrale et
phontique historique du franais, Louvain-la-Neuve: Cabay.
Pinker, Steven 1989, Learnability and Cognition: The Acquisition of Argument Structure,
Cambridge, MA: MIT Press.
Piou, Nanie 1982a, Le clivage du prdicat, in Lefebvre, Magloire-Holly and Piou (eds.),
pp. 12251.
1982b, Le redoublement verbal, in Lefebvre, Magloire-Holly and Piou (eds.),
pp. 15266.
Plag, Ingo 1993, Sentential Complementation in Sranan: On the Formation of an EnglishBased Creole Language, Tbingen: Max Niemeyer.
Pluchon, Pierre (ed.) 1982, Histoire des Antilles et de la Guyane, Toulouse: Privat.
Poirier, Claude 1994, La langue parle en Nouvelle-France: vers une convergence des
explications, in Mougeon and Beniak (eds.), pp. 23774.
Pollock, Jean-Yves 1989, Verb movement, Universal Grammar, and the structure of IP,
Linguistic Inquiry 20: 365424.
Pranka, Paula M. 1983, Syntax and word formation, Doctoral Dissertation, MIT, Cambridge, MA.
Quillet, Aristide 1970, Dictionnaire encyclopdique Quillet (8 volumes), Paris: Quillet.
Rappaport, Malka and Levin, Beth 1988, What to do with theta-roles, in Wilkens (ed.),
pp. 736.
Rassinoux, Jean 1987, Dictionnaire franais-fon, Cotonou: Imprimerie Dumas.
Rey, Alain (ed.) 1992, Dictionnaire historique de la langue franaise (2 volumes), Paris:
Le Robert.
Rice, Frank A. (ed.) 1962, Study of the Role of Second Languages in Asia, Africa and
Latin America, Washington: Center for Applied Linguistics for Modern Language
Association.
Rickford, John R. 1987, Dimensions of a Creole Continuum: History, Texts and Linguistic
Analysis of Guyanese Creole, Stanford University Press.
Ritchie, William C. and Bathia, Tej K. 1996a, Second language acquisition: Introduction, foundations, and overview, in Ritchie and Bathia (eds.), pp. 146.
Ritchie, William C. and Bathia, Tej K. (eds.) 1996b, Handbook of Second Language
Acquisition, San Diego: Academic Press.

446

References

Ritter, Elizabeth 1991, Dplacement de NP en hatien: oui ou non, in Kihm (ed.),

pp. 6586.
1992, Cross-linguistic evidence for number phrase, in Lefebvre, Lumsden and Travis
(eds.), pp. 197219.
Rivard, A. 1914, tudes sur les parlers de France au Canada, Quebec City: Garneau.
Rivero, Maria-Luisa and Sainz, Koldo 1986, Barriers and NP movement, University of
Ottawa, Department of Linguistics, MS.
Rizzi, Luigi 1986a, On the status of subject clitics in Romance, in Jaeggli and SilvaCorvaln (eds.), pp. 391419.
1986b, Null objects in Italian and the theory of pro, Linguistic Inquiry 17: 50157.
Roberge, Yves 1990, The Syntactic Recoverability of Null Arguments, Kingston: McGillQueens University Press.
Robert, Paul 1985, Le Grand Robert de la langue franaise (9 volumes), Paris: Le
Robert.
Roberts, Ian 1993, Verbs and Diachronic Syntax: A Comparative History of English and
French, Dordrecht: Kluwer Academic Press.
Roberts, Paul 1954, Understanding Grammar, New York: Evanston, and London: Harper
and Row.
Robertson, Ian 1983, The significance of Berbice Dutch suffixes, in Carrington (ed.),
pp. 21116.
1993, The Ijo element in Berbice Dutch and the pidginisation/creolisation process, in
Mufwene (ed.), pp. 296316.
Roeper, Thomas and Williams, Edwin (eds.) 1987, Parameter Setting (Studies in Theoretical Psycholinguistics), Dordrecht: Reidel.
Romain, Jean-Baptiste 1959, Quelques murs et coutumes des paysans hatiens (Travaux
pratiques dethnographie sur la rgion de Milot lusage des tudiants), Port-auPrince: Imprimerie de ltat.
Romaine, Suzanne 1988, Pidgin and Creole Languages, London/New York: Longman.
Romaine, Suzanne (ed.) 1991, Language in Australia, Cambridge University Press.
Ross, John Robert 1967, Constraints on variables in syntax, Doctoral Dissertation, MIT,
Cambridge, MA.
Rosset, Thodore 1911, Les origines de la prononciation moderne tudies au 17ime
sicle, Paris: A. Colin.
Rouveret, Alain and Vergnaud, Jean-Roger 1980, Specifying reference to the subject:
French causatives and conditions on representations, Linguistic Inquiry 11: 97202.
Ruhl, Charles 1989, On Monosemy: A Study in Linguistic Semantics, Albany, NY: State
University of New York Press.
Safir, Kenneth 1982, Syntactic chains and the definiteness effect, Doctoral Dissertation,
MIT, Cambridge, MA.
Samarin, William 1967, A Grammar of Sango, The Hague: Mouton.
Sankoff, David and Cedergren, Henrietta (eds.) 1981, Variation Omnibus, Edmonton:
Linguistic Research.
Sankoff, Gillian (ed.) 1980, The Social Life of Language, Philadelphia: University of
Pennsylvania Press.
Sankoff, Gillian 1984, Substrate and universals in the Tok Pisin verb phrase. Meaning,
form and use in context: Linguistic applications, in Schiffrin (ed.), pp. 10419.
1990, The grammaticalisation of tense and aspect in Tok Pisin and Sranan, Language
Variation and Change 2: 295312.
1991, Using the future to explain the past, in Byrne and Huebner (eds.), pp. 6174.

References

447

Sankoff, Gillian and Laberge, Suzanne 1973, On the acquisition of native speakers by a
language, Kivung 6: 3247.
1980, On the acquisition of native speakers by a language, in Sankoff (ed.), pp. 195
209.
Sauvageot, Serge and Manessy, Gabriel 1963, Wolof et Srr: tudes de phontique et de
grammaire descriptive, Universit de Dakar.
Scalise, Sergio 1984, Generative Morphology, Dordrecht: Foris.
Schafer, Amy (ed.) 1993, Proceedings of the North East Linguistic Society 23, Amherst:
GLSA.
Schiffrin, Deborah (ed.) 1984, Georgetown University Round Table on Language and
Linguistics, Washington, DC: Georgetown University Press.
Schuchardt, Hugo 1979, The Ethnography of Variation: Selected Writings on Pidgins and
Creoles, edited and translated by T. L. Markey, Ann Arbor, MI: Karoma.
Schumann, John H. 1978, The Pidginisation Process: A Model for Second Language
Acquisition, Rowley, MA: Newbury House.
Schwegler, Armin 1991, Negation in Palenquero: Synchrony, Journal of Pidgin and
Creole Languages 6: 165214.
Sebba, Mark 1981, Derivational regularities in a creole lexicon: The case of Sranan,
Linguistics 19: 10117.
Segurola, R. P. B. 1963, Dictionnaire fon-franais, Cotonou: Procure de lArchidiocse.
Seiter, William 1980, Studies in Niuean Syntax, New York: Garland Publishing.
Selkirk, Elisabeth 1982, The Syntax of Words, Cambridge, MA: MIT Press.
1984, Phonology and Syntax: The Relation between Sound and Structure, Cambridge,
MA: MIT Press.
Seuren, Pieter and Wekker, Herman 1986, Semantic transparency as a factor in creole
genesis, in Muysken and Smith (eds.), pp. 5771.
Sherer, Tim (ed.) 1991, Proceedings of the North East Linguistic Society, Universit du
Qubec Montral: GLSA.
Siegel, Jeff 1995, Koine formation and creole genesis, paper read at the Amsterdam
Creole Workshop.
To appear, Mixing, levelling and pidgin/creole development, in Spears and Winford
(eds.).
Simoni-Aurembou, Marie-Rose (ed.) 1995, Dialectologie et littrature du domaine dol
occidental. Lexique des plantes: morphosyntaxe, Fontaine-ls-Dijon: A.B.D.O.
Singler, John V. 1979, Verb suffixes in Klao, Cahiers ivoiriens de recherche linguistique
6: 933.
1986, Caribbean creoles and West African languages, City University of New York,
Summer Institute, Linguistic Society of America.
1988, The homogeneity of the substrate as a factor in pidgin/creole genesis, Language 64: 2751.
Singler, John V. (ed.) 1990, Pidgin and Creole TenseMoodAspect Systems (Creole
Language Library 6), Amsterdam/Philadelphia: John Benjamins.
Singler, John V. 1993a, African influence upon Afro-American language varieties: A
consideration of sociohistorical factors, in Mufwene (ed.), pp. 23553.
1993b, The African presence in Caribbean French colonies in the seventeenth century:
Documentary evidence, Travaux de recherche sur le crole hatien 1617: 1236
(Universit du Qubec Montral).
1995, The demographics of creole genesis in the Caribbean: A comparison of Martinique
and Haiti, in Arends (ed.), pp. 20332.

448

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: 185231.
Smith, Norval 1987, The genesis of the creole languages of Surinam, Doctoral Dissertation, University of Amsterdam.
1995, Voodoo chile: Why did they do it differently?, paper read at the Amsterdam
Workshop on Creole Genesis and Language Contact. To appear in Smith and Veenstra
(eds.).
1996, WE-focus in Saramaccan/substrate feature or grammaticalisation?, in Baker
and Syea (eds.), pp. 11329.
Smith, Norval, Robertson, Ian and Williamson, Kay 1987, The Ijo element in Berbice
Dutch, Language in Society 16: 4989.
Smith, Norval and Veenstra, Tonjes 1994, Affixation in a radical creole, paper read at
the SPCL Meeting, January 1994, Boston.
Smith, Norval and Veenstra, Tonjes (eds.) To appear, Proceedings of the Amsterdam
Workshop on Creole Genesis and Languages Contact, Amsterdam: John Benjamins.
Spagnolo, L. M. 1933, Bari Grammar, Verona: Missioni Africane.
Spears, Arthur K. 1990, Tense, mood, and aspect in the Haitian Creole preverbal marker
system, in Singler (ed.), pp. 11942.
Spears, Arthur K. and Winford, Donald (eds.) To appear, Pidgins and Creoles: Structure
and Status, Amsterdam: John Benjamins.
Spillebout, Gabriel 1985, Grammaire de la langue franaise du XVIIe sicle, Paris:
Picard.
Sportiche, Dominique 1994, Subject in French and romance complex inversion and clitic
doubling, UCLA, MS.
Sproat, Richard W. 1985, On deriving the lexicon, Doctoral Dissertation, MIT, Cambridge, MA.
Steever, Sanford B., Walker, Carol A. and Mufwene, Salikoko S. (eds.) 1976, Papers
from the Parasession on Diachronic Syntax, University of Chicago Press.
Sterlin, Marie-Denise 1988, Les diffrentes caractristiques de pou en crole hatien,
Travaux de recherche sur le crole hatien 3: 134 (Universit du Qubec Montral).
1989, Les caractristiques de pou: Un modal en position de complmenteur, in Lefebvre
and Lumsden (eds.) (1989c), pp. 13147.
Stewart, William A. 1962, Creole languages in the Caribbean, in Rice (ed.), pp. 34
53.
Stowell, Timothy 1981, On the origins of phrase structure, Doctoral Dissertation, MIT,
Cambridge, MA.
Sylvain, Suzanne 1936, Le crole hatien: morphologie et syntaxe, Wetteren, Belgium:
Imprimerie De Meester/Port-au-Prince: By the author.
Szabolcsi, Anna 1987, Functional categories in the noun phrase, in Kenesei (ed.),
pp. 16789.
Taylor, Douglas 1963, The origin of West Indian creole languages: Evidence from
grammatical categories, American Anthropologist 65: 80014.
1971, Grammatical and lexical affinities of creoles, in Hymes (ed.), pp. 2938.
Tenny, Carol 1987, Grammaticalising aspect and affectedness, Doctoral Dissertation,
MIT, Cambridge, MA.
1988, Studies in generative approaches to aspect, Lexicon Project Working Papers
24, Cambridge, MA: MIT Center for Cognitive Science.

References

449

Thibault, Pierrette and Vincent, Diane 1990, Un corpus de franais parl, Quebec:
Dpartement de linguistique, Universit Laval.
Thomason, Sarah G. 1993, On identifying the sources of creole structures. A discussion
of Singlers and Lefebvres papers, in Mufwene (ed.), pp. 28095.
Thomason, Sarah G. and Kaufman, Terrence 1991, Language Contact, Creolisation, and
Genetic Linguistics, Berkeley/Los Angeles/Oxford: University of California Press.
Thurot, Charles 18813, De la prononciation franaise depuis le commencement du XVIe
sicle daprs le tmoignage des grammairiens, Geneva: Slatkine Reprints (1966).
Tinelli, Henri 1970, Generative phonology of Haitian creole, Doctoral Dissertation,
University of Michigan, Ann Arbor, MI.
Travis, Lisa 1984, Parameters and effects of word order variation, Doctoral Dissertation, MIT, Cambridge, MA.
1991a Inner aspect and the structure of VP, paper read at the Workshop on Lexical
Syntactic Relations, Universit du Qubec Montral.
1991b Derived objects, inner aspect, and the structure of VP, paper read at NELS,
University of Delaware.
1992, Inner aspect and the structure of VP, Cahiers de linguistique de lUQAM 1:
13249.
In press, L-syntax/S-syntax boundary: Evidence from Austronesian, in Ileana and
Phillips (eds.).
Travis, Lisa and Lamontagne, Greg 1992, The case filter and licensing of empty K, in
Lefebvre, Lumsden and Travis (eds.), pp. 15774.
Tremblay, Mireille 1990, French possessive adjectives as dative clitics, WCCFL 8:
399413, Stanford University Press.
1991, Possession and datives: Binary branching from lexicon to syntax, Doctoral
Dissertation, McGill University, Montreal.
Trudgill, Peter 1986, Dialects in Contact (Language in Society 10), Oxford: Blackwell.
Un habitant dHayti, 1811, Idylles et chansons, ou essais de la posie crole, Philadelphia: J. Edwards.
Valdman, Albert 1970, Basic Course in Haitian Creole, The Hague: Mouton.
Valdman, Albert (ed.) 1977, Pidgin and Creole Linguistics, Bloomington/London: Indiana
University Press.
Valdman, Albert 1978, Le crole: structure, statut et origine, Paris: Librairie C.
Klincksieck.
Valdman, Albert (ed.) 1979, Le franais hors de France (Croles et franais rgionaux),
Paris: Honor Champion.
Valdman, Albert 1980, Creolisation and second language acquisition, in Valdman and
Highfield (eds.), pp. 297311.
1993, Matrice et diffusion culturelle dans le dveloppement des croles franais,
paper read at the Universit du Qubec Montral.
Valdman, Albert and Highfield, Arthur R. (eds.) 1980, Theoretical Orientations in Creole
Studies, New York: Academic Press.
Valdman, Albert et al. 1981, Haitian CreoleEnglishFrench Dictionary (2 volumes),
Bloomington: Indiana University, Creole Institute.
Valois, Daniel 1991, The internal syntax of DP, Doctoral Dissertation, University of
California, Los Angeles.
Van Daele, Hilaire 1929, Phontique historique du franais depuis les origines jusqu
nos jours, Paris: Librairie A. Hatier.

450

References

Van Lit, J., Mulder, R. and Sybesma, R. (eds.) 1991, Proceedings of the Second Leiden
Conference for Junior Linguists, Leiden University Press.
Van Name, Addison 186970, Contribution to creole grammar, Transactions of the
American Philological Association, 186970: 12367.
Van Riemsdijk, H. C. 1978, A Case Study in Syntactic Markedness, Dordrecht: Foris.
Van Voorst, Jan 1988, Event Structure (Current Issues in Linguistic Theory 59),
Amsterdam /Philadelphia: John Benjamins.
Vasseur, Gaston 1996, Grammaire des parlers picards du Vimeu (Somme) avec
considration spciale du dialecte de Nibas, Abbeville: F. Paillart.
Vaugelas, Claude Favre de 1647, Remarques sur la langue francoyse utiles ceux qui
veulent bien parler et bien crire (nouvelle dition 1981), Paris: Champ Libre.
Vdrine, Emmanuel W. 1992, Dictionary of Haitian Creole Verbs (with Phrases and
Idioms), Cambridge, MA: Soup to Nuts Publishers.
Veenstra, Tonjes 1992, Double objects in Haitian creole, in Lefebvre and Lumsden
(eds.), volume 1, pp. 17093.
1996, Serial Verbs in Saramaccan: Predication and Creole Genesis, Dordrecht: Holland Institute of Generative Linguistics.
Veenstra, Tonjes and Den Besten, Hans 1995, Fronting, in Arends, Muysken and Smith
(eds.), pp. 30318.
Vendler, Zeno 1967, Linguistics in Philosophy, Ithaca, NY: Cornell University Press.
Vincent, Diane 1984, Les ponctuants de la langue, Doctoral Dissertation, Universit de
Montral.
Vinet, Marie-Thrse 1991, Observations thoriques sur les expltifs: le cas du crole
hatien, Revue qubcoise de linguistique 20: 195214.
Voorhoeve, Jan 1973, Historical and linguistic evidence in favor of the relexification
theory in the formation of creoles, Language in Society 2: 13345.
Wagner, R. L. and Pinchon, J. 1962, Grammaire du franais classique et moderne, Paris:
Hachette.
Wallace, Efo 1995a, Lw: temps, mode et aspect, in Lefebvre (ed.) (1995a), volume
II, 54 pages.
1995b, Le prsent progressif et le futur: les marqueurs, in Lefebvre (ed.) (1995a),
volume II, 14 pages.
1995c, Le processus de la relexification: Le cas de la ngation en w dans le crole
Hatien, Universit du Qubec Montral, MS.
1995d, Le groupe nominal en w, in Lefebvre (ed.) (1995a), volume II, 15 pages.
Ward, Ida C. 1936, An Introduction to the Ibo Language, Cambridge: W. Refferd and
Sons Ltd.
Wartburg, Walter von 1971, Franzsisches Etymologisches Wrterbuch: Eine Darstellung
des Galloromanischen Sprachschatzes, Basel: Zbinden Druck.
Washabaugh, William 1975, On the development of complementisers in creolisation,
Working Papers on Language Universals 17: 10940.
Weinreich, Uriel 1953, Languages in Contact, The Hague: Mouton.
Wekker, Herman (ed.) 1996, Creole Languages and Language Acquisition (Trends in
Linguistics 86), Berlin/New York: Mouton de Gruyter.
Welmers, William E. 1973, African Languages Structures, Berkeley/Los Angeles/London:
University of California Press.
Westerman, Diedrich 1907, Grammatik der Ewe-Sprache, Berlin: Dietrich Reimer.
1930, A Study of the Ewe Language, London: Oxford University Press.

References

451

Westerman, Diedrich and Bryan, Margaret A. 1970, The Languages of West Africa,
Folkestone: Dawsons, The Gresham Press (New edition).
Whinnom, K. 1971, Linguistic hybridisation and the special case of pidgins and creoles,
in Hymes (ed.), pp. 91115.
1977, Lingua franca: Historical problems, in Valdman (ed.), pp. 295310.
White, Lydia 1996, Universal grammar and second language acquisition: Current trends
and new directions, in Ritchie and Bathia (eds.), pp. 85120.
Wilkens, W. (ed.) 1988, Thematic Relations (Syntax and Semantics 21), New York:
Academic Press.
Wimpffen, Francis Alexander Stanislaus, Baron de 1817, A Voyage to Saint Domingo in
the Years 1788, 1789 and 1790, London.
Woisetschlaeger, Erich 1977, A semantic theory of the English auxiliary system, Doctoral Dissertation, MIT, Cambridge, MA.
Wolf, Lothar 1991, Le langage de la Cour et le franais canadien, in Horiot (ed.),
pp. 11523.
1994, Postface, in Mougeon and Beniak (eds.), pp. 32532.
Woolford, Ellen 1983, Introduction: The social context of creolisation, in Woolford and
Washabaugh (eds.), pp. 19.
Woolford, Ellen and Washabaugh, William (eds.) 1983, The Social Context of Creolisation,
Ann Arbor, MI: Karoma.
Yvon, H. 1951, Cil et Cist, articles dmonstratifs, Romania 72: 14581.
Zhang, Shi 1990, Correlations between the double object construction and preposition
stranding, Linguistic Inquiry 21: 31216.
Zink, Gaston 1991, Phontique historique du franais, Paris: Presses Universitaires de
France.
Zribi-Hertz, Anne 1995, Emphatic or reflexive? On the endophoric character of French
lui-mme and similar complex pronouns, Journal of Linguistics 31: 33374.
Zwicky, Arnold M. 1976, On clitics, in Dressler and Pfeiffer (eds.), pp. 2939.

452

Index of authors

Index of authors

Abney, S. 103, 412

Adam, L. 3
Adjemian, C. 34
gbdvr, P. K. 21011, 214
Alembert, Jean Le Rond d 127, 411
Allen, M. 329, 334
Alleyne, M. C. 23, 56, 10, 33, 65, 398
Allsopp, R. 33, 409
Andersen, R. W. 2, 5, 10, 33, 409
Anestin, A.-L. 398
Anonymous (1983) 68, 97, 112, 119, 125, 131,
142, 1778, 191, 411, 41314, 416
Anonymous c. 397
Ans, A.-M. d 398, 411
Ansre, G. 58
Aoun, J. 419
Arends, J. 67, 207, 408
Aronoff, M. 329
Audain, M. J. J. 397
Auger, J. 158, 411
Avolonto, A. 119, 1356, 210, 213, 280, 353,
413
Awoyale, Y. 160, 167, 16970
Baker, M. C. 60, 356, 392, 422
Baker, P. 2, 11, 42, 68, 80, 397, 409
Bakker, P. 214, 289, 330, 409, 421
B0gb-, A. 41, 365
Barbaud, P. 421
Barss, A. 288, 290, 292, 294, 419
Bary, R. 149, 411
Bastide, R. 66
Bathia, T. K. 35, 422
Bauer, L. 421
Bendor-Samuel, J. 410
Beniak, . 62, 410
Bennis, H. 349
Bentolila, A. 134
Bernab, J. 139
Bernstein, J. B. 88, 411
Bertho, J. 58
Bickerton, D. 2, 4, 6, 10, 12, 42, 4950, 66,
724, 111, 140, 276, 351, 355, 373, 396
Bigelow, J. 397
Bloomfield, L. 63

452

Bonnard, H. 399400, 411

Bonneau, A. 397
Bouchard, D. 192, 204, 416
Brasseur, P. 151
Brillon, J. 401
Brousseau, A.-M. 6, 9, 15, 17, 50, 82, 84, 86,
889, 1003, 107, 1425, 14951, 153,
1556, 159, 163, 1679, 1713, 1757,
17981, 211, 2538, 27880, 30311,
31314, 31720, 3267, 3302, 3346,
33848, 384, 398, 401, 408, 411, 41415,
417, 4202
Brunot, F. 79, 96, 169, 411
Bruyn, A. 380
Bryan, M. A. 5960, 326, 329, 332, 410
Butel, P. 53
Byrne, F. 186, 355, 384
Cadely, J.-R. 151, 1534, 398
Capo, H. B. C. 589, 69, 142, 401, 414
Carden, G. 36, 689, 15960, 162, 168, 170,
255, 259, 410, 415, 417
Carlson, G. N. 366
Cassidy, F. 421
Catach, N. 174, 411
Cattell, R. 279
Cedergren, H. 70
Chaudenson, R. 2, 6, 10, 44, 62, 678, 96,
1445, 169, 176, 181, 335, 338, 411, 415
Chomsky, N. 78, 18, 41, 73, 77, 105, 163,
1967, 200, 221, 237, 259, 266, 272, 274,
282, 288, 2967, 349, 351, 408, 412,
41516
Clements, G. N. 147
Collins, C. 82, 203, 21011, 221, 3646, 369,
371, 416, 4223
Comrie, B. 11516, 1203, 134, 147, 238
Corne, C. 2, 68, 159, 162, 397
Cowper, E. A. 192
Crystal, D. 223
Curtin, P. D. 54
Da Cruz, M. 21011, 21314, 221, 223, 241,
41314, 41617
Dalby, D. 58

Index of authors
Damoiseau, R. 115, 1345, 413
Darmesteter, A. 411, 421
Debien, G. 397
DeCamp, D. 421
Dchaine, R.-M. 43, 1345, 15961, 1678,
356, 365, 3712, 413
Dees, A. 96, 411
DeGraff, M. A. F. 69, 88, 1512, 158, 166,
196, 20610, 213, 261, 269, 271, 3501,
3534, 360, 367, 416, 418
Dejean, Y. 398
Den Besten, H. 74, 408
Den Dikken, M. 419
Deprez, V. 1514, 158, 206, 261, 266, 269,
3501, 3603, 416, 418
Descourtilz, M. E. 397
Diderot, D. 127
Diesing, M. 414
Dillard, J. L. 50
Di Sciullo, A.-M. 148, 303, 330, 334
Dolphyne, F. 128
Domingue, N. 11, 46
Dubois, J. 316, 403, 411
Ducurjoly, S. J. 68, 397
Dumais, D. 129, 2512, 261, 2638, 283, 418
Durie, M. 13
Edmonson, J. A. 165
Emirkanian, L. 131
Fabb, N. 41314
Faine, J. 62, 83, 255, 411
Faltz, L. M. 160, 165, 170
Fauchois, A. 307
Fraud, J.-F. 113, 127, 411
Frre, G. A. 398
Ferguson, C. A. 63
Ferraz, L. 2
Filipovich, S. 6, 9, 15, 50, 52, 3038, 311,
314, 3301, 365, 408, 420
Fltre, L.-F. 83
Foley, W. A. 1, 11, 42, 48, 63, 408
Fouch, P. 399400, 411
Fournier, R. 7880, 83
Freeman, B. C. 311, 411
Fukui, N. 102
Furetire, A. 1734, 181, 31415, 332, 403,
411, 415

Goldsmith, J. 174, 177, 189

Goodman, M. F. 23, 245, 478, 63, 75, 83,
867, 90, 109, 114, 117, 121, 126, 146,
168, 181, 255, 396, 398, 411, 413
Gougenheim, G. 86, 411, 414
Green, G. M. 287
Greenberg, J. H. 601
Grevisse, M. 113, 118, 130, 141, 143, 146, 165
Grimshaw, J. 279, 287
Gros 397
Gruber, J. S. 297
Guerssel, M. 2812
Guiraud, P. 96, 411
Haase, A. 7980, 141, 143, 146, 151, 173,
181, 315, 411
Hagge, C. 2, 147
Hair, P. E. H. 410
Hale, K. 288, 417
Hall, R. A. 2, 6, 63, 127, 12930, 303, 307,
396, 411, 415, 421
Hancock, I. F. 2, 4, 48, 50
Harris, J. W. 11, 46, 409
Hazoum, M. L. 59, 128, 1389, 144, 146,
163, 210, 212, 412
Heine, B. 413
Heinl, N. G. 53
Heinl, R. D. 53
Herskovits, M. 66
Hesseling, D. C. 2
Highfield, A. R. 42
Hilaire, J. 57, 41011
Hilliard dAuberteuil, M. R. 397
Hirschbhler, P. 174, 177, 189
Hoekstra, T. 102
Holm, J. 3, 6, 33, 1389
Hopper, P. J. 42, 395
Hornstein, N. 115
Hounkpatin, B. 214
Huguet, E. 411, 419
Hulk, A. 350
Hull, A. 62, 83
Hutchison, J. 365
Huttar, G. L. 3, 48
Hyams, N. M. 351
Hyman, L. M. 147
Hymes, D. 12, 46, 48, 50
Imbs, P. 411

Galet, Y. 149, 411

Gambhir, S. K. 409
Gauthier, P. 151, 411
Gendron, J.-D. 399400
Ghomeshi, J. 192
Gilles, R. 205, 355, 384
Giurescu, A. 421
Givn, T. 60, 392
Gbl-Galdi, L. 63

453

Jackendoff, R. 18, 288, 408, 419

Jaeggli, O. A. 350
Jake, J. 268
James, D. 268
Jansen, B. 355
Janvier, L. J. 397
Jeanjean, C. 131
Jespersen, O. 63

454

Index of authors

Johns, A. 192
Johnson, K. 287, 295, 359
Joly, G. 399400, 411
Joseph, F. 411
Joseph, J. S. 215, 384
Jourdan, C. 13
Julliand, A. 317
Juneau, M. 623, 399, 411, 420
Kant, I. viii
Kaufman, T. 12, 10, 12, 25, 30, 32, 48, 409
Kay, P. 4, 66
Kaye, J. 9, 15, 66, 72, 408, 410
Kayne, R. S. 1489, 151, 185, 189, 1934,
287, 290, 3578, 409, 412
Keenan, E. 415
Keesing, R. M. 5, 302, 39, 48, 380, 396
Keyser, J. 288, 417
Kihm, A. 378, 84, 133, 181
Kinyalolo, K. K. W. 82, 1467, 15960, 163,
165, 1679, 1856, 203, 272, 41314,
416, 418
Kiple, K. 54
Koelle, S. W. 61
Koopman, H. 9, 11, 3940, 42, 601, 73, 82,
147, 158, 167, 1734, 1768, 180, 190,
192, 1958, 2035, 249, 2512, 266,
2703, 2768, 287, 291, 351, 3635,
367, 371, 373, 387, 41012, 41516,
4223
Kouwenberg, S. 36, 39, 155, 330, 390, 421
Kratzer, A. 366
Labelle, M. 76
Laberge, S. 11, 13, 17, 423, 1278
Laborderie, N. 399400, 411
Labouret, H. 410
Labov, W. 13, 393
Lafage, S. 332
Laka Mugarza, M. I. 148, 214, 221
Lamontagne, G. 18, 105, 412
Lappin, S. 418
Larson, R. K. 2889, 299, 369, 419, 422
Lasnik, H. 288, 290, 292, 294, 419
Laughren, M. 417
Law, P. 158, 198202, 261, 266, 269, 3501,
3646, 3689, 416, 418, 4223
Lefebvre, C. 4, 6, 912, 1516, 18, 30, 3540,
42, 446, 48, 50, 523, 57, 667, 6970,
72, 7985, 8799, 102, 104, 107, 109,
11114, 11617, 12037, 14950, 1734,
177, 181, 1845, 189, 1912, 1948, 200,
2037, 21922, 22441, 2436, 248,
2723, 284, 2878, 291301, 3038, 311,
314, 3301, 333, 341, 355, 3578, 363,
371, 3836, 389, 3956, 40810,
41217, 41920, 4223
Levin, B. 288, 309, 315, 420

Levinson, S. C. 165
Li, Y.-H. A. 419
Lichtenberk, F. 40
Lieber, R. 9, 3940, 303, 307, 309, 326, 331,
334, 408
Lightfoot, D. 2, 11, 412, 156
Lipou, A. 60
Lommatzsch, E. 96
Lord, C. 41, 43, 186
Lumsden, J. S. 4, 912, 1516, 18, 30, 3440,
446, 48, 50, 57, 67, 69, 72, 82, 87,
1019, 128, 1345, 1956, 206, 2423,
2601, 28790, 336, 3401, 3645, 385,
389, 40813, 41619, 423
Magloire-Holly, H. 412
Man, M. 58
Manessy, G. 128
Manfredi, V. 109, 15961, 1678, 196, 3645,
416
Marantz, A. 319
Massam, D. 79, 82, 158, 192, 237, 243, 246,
248, 261, 268, 280, 416, 418
McWhorter, J. 33, 36, 207
Meisel, J. M. 34
Meister, A. 279
Meyer-Lbke, W. 62
Miller, G. A. 421
Milner, J.-C. 7980, 88, 103, 185
Mintz, S. W. 2
Mhlig, W. J. G. 25
Moisy, H. 421
Moody, M. D. 421
Moreau, M.-L. 1934
Moreau de Saint-Mry, L.-. 68
Morin, Y.-C. 162, 1645
Mouchon, P. 127
Mougeon, R. 62, 410
Mous, M. 259, 34, 61, 384, 409
Mufwene, S. S. 26, 1011, 36, 38, 44, 46,
49, 60, 74, 86, 144, 147, 166, 276, 373,
390, 409
Mhlhusler, P. 4, 11, 38, 42, 44, 46, 4950,
333, 409
Muysken, P. C. 6, 910, 1621, 289, 34, 42,
67, 72, 74, 155, 15960, 162, 168, 171,
182, 224, 255, 303, 356, 378, 384, 390,
394, 396, 4089, 412, 417, 4223
Naro, A. J. 5, 34, 63
Ndayiragije, J. 180, 364, 366, 4223
Nicolson 397
Nida, E. 192
Nyrop, K. 317, 411
Obenauer, H.-G. 174, 177
Odden, D. 268
Oehrle, R. T. 287

Index of authors
Ogunbewale, P. O. 128, 138
Orjala, P. R. 75
Oudin, A. 149, 361, 411
Ouhalla, J. 148
Papen, R. 21
Pazzi, R. 59
Pierret, J.-M. 399400, 411
Pinchon, J. 411, 415
Pinker, S. 2878
Piou, N. 3646, 423
Plag, I. 42, 192
Plank, F. 165
Poirier, C. 62
Pollock, J.-Y. 189, 2089, 3512
Pranka, P. M. 395
Rappaport, M. 288, 309, 315, 420
Rassinoux, J. 178, 3368, 341, 411
Reh, M. 413
Rey, A. 174, 317, 411
Rickford, J. R. 11, 42
Ritchie, W. C. 35, 422
Ritter, E. 82, 867, 269, 3646, 412, 416, 418,
4223
Rivard, A. 62
Rivero, M.-L. 268
Rivet, P. 410
Rizzi, L. 148, 350
Roberge, Y. 149, 350
Robert, P. 411
Roberts, I. 148
Roberts, P. 223
Robertson, I. 33, 36, 39, 166, 408
Roeper, T. 422
Romain, J.-B. 75
Romaine, S. 11, 42
Ross, J. R. 366
Rosset, T. 96, 411
Rouveret, A. 412
Ruhl, C. 192
Safir, K. 350
Sainz, K. 268
Samarin, W. 43
Sankoff, G. 4, 11, 13, 17, 425, 1278, 131,
393, 409
Sauvageot, S. 128
Scalise, S. 305, 420
Schuchardt, H. 2, 63
Schumann, J. H. 10
Schwegler, A. 216
Sebba, M. 421
Segurola, R. P. B. 88, 97, 1778, 182, 211,
214, 3367, 383, 411, 41617
Seiter, W. 268
Selkirk, E. 303, 346
Seuren, P. 6

455

Siegel, J. 5, 11, 13, 334, 46, 409

Singler, J. V. 12, 32, 43, 47, 528, 735,
171, 3901, 40910
Smith, N. 6, 33, 36, 39, 67, 72, 74, 15960,
162, 168, 171, 182, 186, 207, 255, 4089,
421
Spagnolo, L. M. 43
Spears, A. K. 112, 123, 130, 412
Speas, M. 102
Spillebout, G. 70, 411
Sportiche, D. 1478
Sproat, R. W. 334, 395
Sterlin, M.-D. 1845, 2715, 351, 415
Stewart, O. T. 356, 422
Stewart, W. A. 9, 689, 15960, 162, 168,
170, 255, 259, 410, 417
Stowell, T. 297
Syea, A. 11, 42
Sylvain, S. 3, 48, 81, 867, 90, 100, 114, 118,
129, 168, 170, 307, 3967, 411, 421
Szabolcsi, A. 103, 412
Taylor, D. 159, 181
Tenny, C. 134, 230, 234, 288
Thibault, P. 70
Thomason, S. G. 12, 10, 12, 25, 30, 32, 48,
67, 734, 409
Thurot, C. 400, 411
Tinelli, H. 303, 307, 398
Traugott, E. C. 42, 395
Travis, L. 18, 3940, 105, 221, 279, 412
Tremblay, M. 103, 287, 290, 358, 412, 419
Trudgill, P. 11, 46, 409
Valdman, A. 2, 1011, 42, 623, 83, 90, 112,
129, 303, 3978, 411, 415, 421
Valdman, A. et al. 64, 129, 141, 168, 173,
1812, 261, 2978, 311, 316, 3368, 340,
383, 403, 411, 415, 419, 421
Valois, D. 88
Van Daele, H. 399400, 411
Van Name, A. 2, 5
Van Riemsdijk, H. C. 39, 734, 412
Van Voorst, J. 234
Vasseur, G. 414
Vaugelas, C. F. de 149, 411
Vdrine, E. W. 30811, 341, 411, 419
Veenstra, T. 74, 287, 28990, 414, 419, 421
Vendler, Z. 413
Vergnaud, J.-R. 412
Vincent, D. 70
Vinet, M.-T. 158, 206, 269, 418
Voorhoeve, J. 23, 9, 48
Wagner, R. L. 411, 415
Wallace, E. 84, 123, 133, 138, 211, 215,
412
Ward, I. C. 128

456

Index of authors

Wartburg, W. von 411

Washabaugh, W. 11, 42, 192
Weinreich, U. 33
Wekker, H. 6
Welmers, W. E. 75, 122, 138, 238, 411
Westerman, D. 43, 5860, 86, 326, 329, 332,
41011, 413
Whinnom, K. 1, 9, 48, 408
White, L. 35
Williams, E. 303, 330, 334, 422
Williamson, K. 33, 408

Wimpffen, F. A. S. Baron de 397

Woisetschlaeger, E. 388
Wolf, L. 62
Woolford, E. 4
Yvon, H. 96
Zhang, S. 358
Zink, G. 399400, 411
Zribi-Hertz, A. 162, 165
Zwicky, A. M. 148

Index of languages and language families

457

Index of languages and

language families

Abe 252
Adele 59
Aguma 59
Ajagbe 589, 1389, 144
Akan cluster 50, 55, 59, 128, 170, 272
Anyi-Baule 59
Bambara 60, 1389
Bantu 19, 25, 27, 55, 601, 3902, 410, 422
Bari 435
Basque 160, 214
Bassa-Nge 170
Berbice Dutch 36, 39, 155, 166, 330, 385, 390,
408, 421
Bhojpuri 46
Bini 59

Gungbe 589, 1389, 144, 212, 371

Gur 55, 60, 391, 410
Hausa 109
If 59
Igbo 50, 109, 128, 170, 3712
Ijo (Eastern) 55, 155, 390, 408
Indian Ocean creoles 126
Jamaican creole 39, 147

Caribbean creoles 9, 39, 734, 330, 356, 378,

420
Cigbe 58
Cree 19, 214, 409
Cushitic 19, 257

Kikongo 3, 57
Kituba 3
Klao 43
Kogbe 58
Kriol 38
Kru 43, 60
Kwa 557, 5961, 66, 160, 167, 170, 210,
214, 326, 329, 332, 353, 3567, 371, 373,
3902, 422
Kwaio 301

Djuka 3
Dutch 166, 390, 408
Dyula 60

Lagoon 59
Latin 5, 17, 316
Louisiana creole 2, 126

Ebira 170
Ede cluster 59
Efik 170
English 1718, 301, 39, 41, 46, 74, 89, 103,
120, 123, 160, 1656, 199, 201, 255, 266,
280, 293, 297, 299, 330, 346, 3512,
3579, 364, 41920
Ewe 3, 43, 589, 76, 84, 86, 133, 138,
21114, 21617, 326, 3323, 365, 371,
412, 423

Maa (see Inner Mbugu)

Malagasy 415
Malinke 60, 332
Mande cluster 55, 60, 75, 86, 109, 332, 391, 410
Mandingo 867, 109
Mandinka 50
Manjaku 37
Martinican creole 47, 139, 379
Mauritius creole 2, 415
Mayan 75
Mbugu
Inner Mbugu 19, 248
Normal Mbugu 19, 257
Media Lengua 1921, 23, 28
Michif 19, 214, 28

Ga-Dangme cluster 59
Gbe cluster 43, 559, 66, 69, 74, 128, 1389,
144, 155, 160, 163, 165, 167, 170, 186,
203, 21012, 258, 326, 353, 371, 3901,
410, 41214
Gengbe 58, 1389, 144, 371, 410
Guang 59
Gullah 147, 166

Nago 59
NigerCongo 55, 57
Nupe 59, 326

457

458

Index of languages and language families

Palenquero 216
Papiamentu 147
Pare 25, 27
Phlaphera 58
Picard 151, 41214
Portuguese 38
Portuguese creole 147, 408

Tok Pisin 13, 17, 389, 435, 1278, 396,

409
Tolai 409
Turkish 17
Twi-Fante 59, 326

Quechua 17, 1921, 417

Vai 332
Vata 252, 272, 367
Vietnamese 381

Reunion creole 2
Sango 43
Saramaccan 3, 36, 74, 1867, 207, 330, 347,
379, 3845, 409, 414, 421
Shambaa 25, 27
Solomons Pidgin 303, 3940, 396
Spanish 1921, 24, 148
Sranan 3
Swahili 27
Toabaita 40
Twfingbe 59, 1389, 144

Urhobo 170

Wemgbe 58
West Atlantic 55, 60, 391
Wolof 50, 109, 128
Xlwagbe 59
Xula 371
Xwdagbe 589, 1389, 144, 212
Yoruba 41, 50, 5960, 867, 10910, 128,
138, 170, 326, 356, 3645, 372
Yuman languages 160

Index of subjects

459

Index of subjects

adverbial clauses
see causal and temporal clauses
anterior markers 112, 114, 11618, 1245,
130, 132, 138
assertive markers 2219
bare sentences (interpretation of ) 11415,
1347, 13940
BODY-state verbs 2501
calquing 5, 15, 30, 32
Case
Case assigning properties of verbs 2836,
297301
Case Filter 18, 10110, 2736
Exceptional Case marking 2736
Genitive Case 10114, 10610, 1446, 168,
35760
Objective Case 102, 10910
causal clauses 36371
clause types
see adverbial clauses; causal clauses;
complement clauses; factive clauses;
purposive clauses; relative clauses;
temporal clauses
cleft constructions 193202, 36373
clitics
clitic climbing 148, 150, 152
pronominal clitics 14857, 182, 34951
cognate object verbs 2803
complement clauses
finite 18493, 2712
non-finite 2716
complementisers
introducing complements of verbs of the
say-class 1847
introducing complements of verbs of the
want-class 18793
introducing infinitival complements in the
context of extracted subjects 2716
compounds
principles of concatenation of words into 47,
334, 348
semantics of 12, 3349
types of 3428
word order in 33942

conflation (phonological) 378, 84, 181

conjunction
of clauses 205
of nps 206
control verbs
three-variable control verbs 2768
two-variable control verbs 2716
creole languages
features of 14, 12, 71
pidgins and creoles as pcs 4
versus mixed languages 2930
creolisation
abrupt 32
and second language acquisition 29, 336,
72
data
Fongbe 656, 69, 756
French
Montreal 70
seventeenth and eighteenth century 69
spoken by the colonists 625, 70
Haitian 689, 756
definite determiner
in nominal structure 7886, 8895, 979,
104, 106, 110, 1345
in the clause 21947, 28797, 36971
definite future markers 11214, 120, 1239,
1323, 138
demonstrative terms 78, 8992, 94, 96100,
104
derivational affixes 30333, 4037
desemanticisation 412
dialect levelling
cases of 47, 87, 110, 1379, 1701, 25362,
32333
definition of 467, 3934
dialects (regional)
Fongbe 589, 75
French 623
Haitian 46, 75
double-object verbs 287301, 35760
Empty Category Principle (ecp) effects
193203
Exceptional Case marking 2745, 278

459

460

Index of subjects

existential verbs 26971

expansion 6, 50
expletives 1579, 25962
factive clauses 82, 2035, 36371
filters (surface) 2406
finite/non-finite issue in creole studies 2718
focus markers 2068
genetic affiliation of creole languages 396
grammaticalisation 42
habitual marker 1369
Haiti
early African population in 537, 656
languages spoken at the time of creole
genesis 558
the colonists 53, 623
Haitian phonological forms and French phonetic
matrices (correspondences of ) 398402
homogeneity of the substratum languages
5862, 67, 3903
imperfective marker 11214, 11923, 1289,
1379
indefinite determiners 78, 879, 915, 979,
104, 106, 110
indefinite future markers 112, 12933
inherent object verbs 2803
innovations in Haitian 12, 51, 879, 2068,
26971, 2837, 32333, 34951
insistence markers 21317
language intertwining 22
light verbs 27880
logical form 2279, 2937
logophoric pronouns 1478, 182
markedness
issue 725
marked/unmarked structures 734, 2629,
36374
mixed languages 1930
morphological head 3302
nativisation 4
negation markers 20811
negative markers 21317
negative quantifiers (interpretation of ) 3603
null subject parameter 34951
operator
in factive clauses 2035
in relative clauses 2035
orthographic conventions
Fongbe 4012
French 399400
Haitian 3989

paralexification 27, 3845

parameters
parameter setting in creole genesis 8, 47, 68,
34975, 3868
see double-object verbs; (interpretation of )
negative quantifiers; null subject
parameter; serial verbs; verb doubling
phenomena; verb raising parameter
personal pronouns 847, 104, 110, 1413, 160,
182
phonological inventories
Fongbe 4012
French 399400
Haitian 3989
pidgin languages
expanded pidgins 4, 32
features of 1, 4, 634
pidgins and creoles as pcs 4
plural markers 78, 848, 901, 98, 104, 110,
173
possessives 26, 78, 90, 103, 106, 109, 1437,
183
predicate cleft construction 59, 61, 734,
36373
pronominal clitics 14857
pronominal forms
see expletives; logophoric pronouns;
personal pronouns; possessives;
pronominal clitics; reflexives; resumptives
purposive clauses 18793
question markers 21113
raising verbs 2626
reanalysis
cases of 434, 87, 1289, 18493, 376,
378
definition of 416, 137
related phenomena 11, 412
reduction 56, 50
reflexive verbs 2539
reflexives
anaphor 1607, 183
body-part 26, 15960, 16770, 183
relabelling
definition of 16
of minor category lexical entries by a null
form 1718, 378, 445, 50, 78, 10810,
128, 164, 37586
of minor category lexical entries by an overt
form 1718, 109, 123, 129, 142, 156,
1801, 37586
relative clauses 80, 82, 90, 93, 98, 179,
193205, 365
relexification
and dialect levelling 467
and reanalysis 416
in creoles 3041

Index of subjects
in mixed languages 1930
in pidgins 303
in second language 3041
representation of 1519, 22, 27, 3846
resumptives
in the context of extracted subjects 193202
in the context of raising verbs 2629
semantic interpretation
see (interpretation of ) bare sentences; cleft
constructions; (principles of concatenation
of words into) compounds; derivational
affixes
serial verbs
serial verb parameter 3557
simplification 56, 50
stabilisation 4
subject raising 2629
subjunctive markers 112, 11819
syntactic categories
major or lexical 18, 21, 234, 289, 367,
44, 49
minor or functional 18, 24, 289, 31, 367,
44, 489, 59, 645, 67
temporal clauses 36371
tma markers
see anterior markers; definite future markers;
habitual marker; imperfective marker;
indefinite future markers; subjunctive
markers
transfer 15, 334

461

transmission and acquisition in creole genesis

8, 12, 34975, 3956
verb doubling phenomena 82, 36374
verb raising parameter 61, 3515
verbs
see body-state verbs; cognate object verbs;
control verbs; double-object verbs;
existential verbs; inherent object verbs;
light verbs; raising verbs; reflexive verbs;
serial verbs; verbs licensing expletive
subjects; weather verbs
types of argument structures 248
verbs licensing expletive subjects 25962
weather verbs 2513
Whmovement in predicate cleft 36373
movement in questions 1769, 193203
phrases 17182
words 17182
word order
affix order 180, 3302
establishing word order in creole genesis of
functional categories 3940, 89
establishing word order in creole genesis of
lexical categories 3840, 89, 180
evaluation of the hypothesis 38890
in compounds 33942
in double-object constructions 2937
in nominal structure 78
in the clause 184246