Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole (Cambridge Studies in Linguistics)

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 Université du Québec à Montréal.

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, São 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 98–12930 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: 1659–1740 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

x

Contents

3.6 3.7

The data Mode of presentation of the data and analyses

4

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

7

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 Yes–no question markers Markers expressing the speaker’s 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 (1776–1936) 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 colony’s total population. Comparison of Haiti after 22–62 years (1681–1721) with Martinique after 25–65 years (1660–1700) 54 Distribution by branch of Niger–Congo 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 Université du Québec à Montréal. From 1976 to 1980, with Lynn Drapeau, I conducted a project on popular French (financed by fcar, fir-uqam and olf). In parallel (1975–7), I started a small project on the syntax of Haitian creole (financed by fir-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 l’haïtien (1982), written in collaboration with Hilda Koopman, Hélène 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 créole haïtien: langues africaines relexifiées? 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, Joëlle Brillon, Réjean Canac-Marquis, Rose-Marie Déchaine, 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 fir-uqam) for a project designed to test this hypothesis in detail, based on the case study of Haitian creole. This project, La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle, which I xv

xvi

Preface

directed with the collaboration of John Lumsden, lasted five years (1989–94). 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, Andrée Bélanger 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, Aimé Avolonto, Maxime Da Cruz, Joseph Sauveur Joseph, Hérold Mimy, Juvénal 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 béninois des langues étrangères), Hounkpati Capo (Labo Gbè, Université nationale du Bénin), Pierre Vernet (Université d’État d’Haïti) and Marc Laurent Hazoumé (cenala, Université nationale du Bénin). 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 propriétés lexicales, leur représentation dans le lexique et leur projection dans la syntaxe (fcar 1990–2) and L’organisation des lexiques et des entrées lexicales (fcar 1993–5) 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. Andrée Bélanger formatted the

Preface

xvii

manuscript and Zofia Laubitz copy-edited it. This work was financed by a grant from sshrcc (1994–7). 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 Andrée Bélanger, 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

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

inflection markers of insistence interrogative irrealis locative logophoric pronoun French emphatic même 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

1

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

2

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 Hagège (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

3

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 Kikongo’s 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: J’ose avancer . . . que les soi-disant patois de la Guyane et de la Trinidad constituent des dialectes négro–aryens. J’entends par là que les nègres guinéens, transportés dans ces colonies, ont pris au français ses mots, mais qu’ayant conservé dans la mesure du possible, leur phonétique et leur grammaire maternelles . . . Une telle formation est à coup sûr hybride . . . La grammaire n’est autre que la grammaire générale des langues de la Guinée. [I go so far as to claim . . . that the so-called patois of Guyana and Trinidad constitute Negro–Aryan 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 présence d’un français coulé dans le moule de la syntaxe africaine, ou . . . d’une langue éwé à vocabulaire français.’ [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

4

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. Mühlhäusler 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 Bickerton’s (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

5

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 Name’s (1869–70: 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 l’histoire de la morphologie, est-il permis de partir du système de flexions français et de ne voir dans les créoles français qu’une réduction ou une simplification de ce système amenant des ‘pertes’ ou des ‘disparitions’ des flexions françaises? Ou bien notre point de départ devrait-il être la morpho-syntaxe ouest-africaine, qui est caractérisée par l’invariabilité du mot, donc par l’absence 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?]

6

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 aisé de voir que ce système diffère beaucoup de celui des verbes français, dans lequel les distinctions temporelles sont beaucoup plus importantes que les distinctions d’aspect. Par contre, les langues ouest-africaines font preuve de systèmes verbaux du même genre que celui du créole français, et il serait évidemment plus valable d’attribuer au système verbal du créole une origine

1.3 Linguistic theory

7

africaine plutôt qu’une origine française. Mais les particules qui marquent les aspects et les temps dans le système verbal créole semblent toutes dérivées de mots ou de groupes de mots français. Ainsi ap(é)
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 speaker–hearer’ (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:

8

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 creole’s 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

9

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 1985–9 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. Mühlhäusler 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; Mühlhäusler 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. Mühlhäusler 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 creole’s 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 Labov’s (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.

2

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 (1985–9), 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.’ Muysken’s 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 Muysken’s (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 [ø] [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 Muysken’s 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. Muysken’s 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 Lamontagne’s 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 Ma’a. 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 sé. 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 demás, 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 Quechua–Spanish 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 . . .’

núway

c. aún no ‘not yet’

aúno

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 Muysken’s 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 Cree–French mixed language developed by the Métis 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’ étalon – wa otin – êw. 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 vjø – 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: – lι – ga:n: – n 1.pst the bet infl.1 ‘I bet on the brown one.’

syr on

lι 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 [αanimate] rather than for the feature [αfeminine], 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 lι that.inan the ‘that book’

23

li:v b. ana lum michif book this.an man ‘this man’ (=(38), (39) in Bakker 1992: 17–18)

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/dã 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/dã 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. ãnarj=r dã lι ma:r in-back loc the car ‘behind the car’ b. otahk otãpãnask-ohk behind car loc ‘behind the car’

michif

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

(12)

a. dã lι frιdn 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/dã 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 Métis, 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 Métis, 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 Ma’a

The case of Inner Mbugu or Ma’a, 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’. Goodman’s (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 Ma’a originated from a Cushitic language that underwent massive borrowing from Pare, a Bantu language. Möhlig (1983) has proposed an opposite view according to which Ma’a is a Cushiticised Bantu language. A more recent study of Inner Mbugu or Ma’a 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 Mous’s 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 Ma’a) 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-kóno mi-kóno vu-shó ma-shó

Subj. ‘person’ é / á ‘people’ vé / vá ‘arm’ ú ‘arms’ í ‘face’ vú ‘faces’ é/á

Obj. mù vá ú í vú á

im m-hé va-hé m-harégha mi-harégha vu-basá ma-basá (=(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. Áa-pú ndaté kú’u. 1:pst-break stick.9 his ‘He broke his stick.’ b. Áa-baha ndatá 1:pst-break stick.9 ‘He broke his stick.’

y-akwé. 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. Áa-(kú)-pú kusáme. 1:pst-rfl-break leg.15 ‘He broke his leg.’ b. Áa-(kú)-baha kughú. 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 mxatú hlúku hí gewa

nm mtí 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 Ma’a 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 Ma’a differs from the other mixed languages reported on above, where the phonological representations of the relexified lexical entries are derived from only one language. Mous’s (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.’ Mous’s 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 Ma’a) 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-à-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-à-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 Bickerton’s 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 learner’s 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 situation’s 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 relexification . . .’ 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 Mous’s 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. s’asseoir ‘to sit’, se préparer ‘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. (t’assieds) you sit on a chair ‘You are sitting on a chair.’ b. Je vais préparer pour la fête. (me préparer) I go prepare for the party ‘I’m 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 Mühlhäusler 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 ø 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’, Mühlhäusler (1986b: 47) addresses the problem of adjective–noun 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. Mühlhäusler’s conclusion is thus that the word order adjective–noun 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, To’aba’ita, 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.’

ba’a. dem

to’aba’ita (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 Bá0gbó-é 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; Mühlhäusler 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 ná ‘give’ has been reanalysed as a dative benefactive preposition meaning ‘for’. A similar case of reanalysis is reported on by Déchaine (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 sà. 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 wouldn’t 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 dé ‘then, afterwards’ became a future or irrealis preverbal marker. As has been observed by Spagnolo (1933: 105–6), in its adverbial function, dé occurs sentence-initially; as a future marker, it occurs between the subject and the verb, as shown in (25). (25)

a. Dé nàn kwn . . . then I do ‘Then I do . . .’

b. Nàn dé 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 ‘I’ll 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, Mühlhäusler (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 dé 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. Mühlhäusler 1980: 34; Mufwene 1990: 138–9, 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 Mufwene’s, 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 Mühlhäusler (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; Mühlhäusler 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

3

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: 1659–1740

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 don’t have the same interests and don’t 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 engagés among engagés + 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

a

Figures for women and engagés 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: 1–37); 1664, ansom g1 470, nos. 1–8; 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, engagés3 and slaves.’ In this type of economy, the Haitian population was made up of the colonists, people of colour, engagés and people of mixed race. As can be seen in Table 3.1 (=Table 2 in Singler 1996), the people of colour and the engagés 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 engagés 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 colony’s total population. Comparison of Haiti after 22–62 years (1681–1721) with Martinique after 25–65 years (1660–1700) 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 population’s exposure to French. Indeed, under the first type of economy, the slave population was in frequent contact with French through the landowners, engagés and people of mixed race. In the shift to a sugar economy, the engagés 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: 237–8) 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 Niger–Congo (percentage)

Atlantic Mande Kwa Gur Nigerian Benue–Congo 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? Singler’s research shows that the African languages spoken by the slave population in Haiti at the crucial period were all from the Niger–Congo 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 1675–9 1680 1680–9 1681 1690 1690–9 1700 1700–9 1700–19 1710–19 1713 1717 1720–9 1721 1730 1730–9 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 Haiti’s 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 Curaçao 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 Haiti’s 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 créole de la deuxième génération, parlait le fongbe de ses ancêtres dahoméens. L’occasion lui fut donc offerte d’apprendre cette langue en Haïti même. Goman maintiendra une tradition kikongo avec son groupe de marrons jusqu’en 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: ‘À l’époque de la Révolution (1791), la position du créole sur le terrain était déjà suffisamment forte pour servir d’instrument fédérateur et d’identification du pouvoir des masses noires. L’érosion se poursuivrait aux dépens 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 Singler’s (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 Niger–Congo 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, Xwèdagbe, Kogbe. In the Atlas sociolinguistique du Bénin (1983: 55), we find the following statement: ‘Entre les langues de ce sous-groupe ou continuum existe une intercompréhension si frappante parfois que l’on se demande si telles ou telles langues forment vraiment des unités 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 . . . Ewe–Gen–Aja–Fon 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, Xwèdagbe and Xlwagbe by Hazoumê (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 Ifè, etc. Again, according to the Atlas sociolinguistique du Bénin (1983: 55), ‘Comme pour les langues du continuum gbe, il existe entre celles du continuum ede une intercompréhension parfois étroite.’ [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 Ewe–Gen–Aja–Fon 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 Bryan’s (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 Bryan’s (1970: 90–5) 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 subject–verb–object. 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, Givón (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, Givón (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 Givón’s 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 infl, 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 subject–verb–object 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, Île-de-France (Paris), Poitou, Saintonge, Brittany). The dialects spoken in this area of France constitute the langue d’/oïl/ (as opposed to langue d’oc). These dialects were not necessarily mutually intelligible among themselves or with the variety of French spoken in Île-de-France (Paris). (For various views on this matter, see the papers in Mougeon and Beniak 1994.) But, as noted by MeyerLübke (1909: 123), these regional dialects were influenced by Parisian French. ‘On sait très bien que dans les provinces de l’Ouest, les dialectes ont subi, plus profondément que partout ailleurs en France, et même avant l’époque de la colonisation de l’Amérique, l’influence 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’/oïl/. 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: 191–3), the variety of French spoken by the colonists was close to the variety of French spoken in Île-de-France in the seventeenth and eighteenth centuries: ‘Les traits syntaxiques sont facilement rattachables à des tendances de la syntaxe de l’époque. Il en est de même 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 Île-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), Göbl-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. Valdman’s 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 larivyè ‘river’ lari ‘street’ olye ‘instead’

french expression la rivière ‘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’à ce que ‘until’ tandis que ‘while’

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

haitian lexical entry afòs ‘by means of’ apati ‘starting from’ apen ‘hardly’ apeprè ‘more or less’

french expression à force de ‘by means of’ à partir de ‘starting from’ à peine ‘hardly’ à peu près ‘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 1985–7 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: ‘À Haïti . . . les diverses cultures ne sont plus que des éléments intégrés et subordonnés de la culture fon . . . À Haïti . . . c’est la culture dahoméenne, plus particulièrement fon qui l’emporte.’ [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: 24–5) 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): ‘L’hypothèse prédit un parallèle significatif entre les propriétés syntaxiques et sémantiques du créole et celles des langues du substrat considérées dans leur ensemble. Le fait que notre comparaison n’implique qu’une seule langue du substrat minimise la possibilité de trouver un tel parallèle et maximise la possibilité de falsifier l’hypothèse.’ [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: 273–4), reproduced in Appendix 1. Among the earlier sources, we find Moreau de Saint-Mery (1797), Ducœurjoly (1802), a text entitled Un habitant d’Hayti (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: 26–7), these texts present a problem of interpretation. For example, they write that Ducœurjoly and Moreau de Saint-Méry 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 Dejean’s 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: Capo’s (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. ‘Puisqu’il existe, pour deux langues données, de nombreuses façons d’être différentes mais une seule façon d’être identiques, un changement qui résulterait en un parallèle accidentel est moins probable qu’un changement qui résulterait en une différenciation 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 1989–94 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. Dumais’s 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 à 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 à 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 Université de Montréal and the Lefebvre and Drapeau Project at the Université du Québec à Montréal. 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 fún ‘feather’ ‘hair’

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

làn ‘meat’ ‘edible animals’ (complement of the verb ‘to kill’)

dife ‘fire’ ‘brand’

(du) feu ‘fire’

mywn ‘fire’ ‘brand’

tèt ‘head’ ‘roof’

tête ‘head’

tà

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 1985–9; 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 Bickerton’s (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 Smith’s claim, in my view, a theory cannot be proven; rather, one presents evidence supporting or falsifying it. With regard to Thomason’s 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 Singler’s 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 Koopman’s 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 Bickerton’s 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 Bickerton’s 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 Singler’s (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). Mufwene’s 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, kè a mwen ‘my heart’ in the North contrasts with kè 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

4

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 ø] àsvn [nyy twn] crab me gen ‘these/those crabs of

sa a yo élv 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, ‘l’article annonce une relative ou un génitif sans qu’aucune mention antérieure 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) bòn pou lasante. bread det good for health ‘Bread is good for one’s health.’

(6)

*(Le) pain est bon pour la (det) bread is good for det ‘Bread is good for one’s health.’

haitian (=(19) in Lefebvre 1994a) santé 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 mangé 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 [ ø 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 determiner’s 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 larivyè ‘river’ lakay ‘home’ listwa ‘history’ latè ‘world’

< < < <

Corresponding la rivière la case l’histoire 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 l’eau 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. larivyè 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 là 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. wémà 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 Wwxúxú (*v) nyvn nú lànmyyén. bread det good for health ‘Bread is good for one’s 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)

Kwkú 3ù blé3ì. 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 mwlíkú 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)

Súnù [3è é yì] 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 determiner’s 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 là 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 là 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 là shown in (20) has been attested in French since at least the seventeenth century (see Flûtre 1970: 141, 146; Hull 1975: 2). (19)

Cet homme -là vient d’arriver. dem man [+deic] come prep arrive ‘This/that man just arrived.’

french

(20)

a. L’ homme là vient d’arriver. det man dm come prep arrive ‘The man [there] just arrived.’

popular french

b. Un homme là vient d’arriver. det man dm come prep arrive ‘A man [there] just arrived.’

popular french

c. Un / l’ homme là, qui vient d’arriver là . . . 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 là’s 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 determiner’s interpretation. Thus, là, 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 determiner’s. The French là 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) ), là has no semantics in common with the Haitian determiner. Thus, the French form là 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 là, yielding the Haitian determiner la.

84 (21)

Functional category lexical entries involved in nominal structure fongbe lexical entry

french

/ú/ [+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 á or lá and it has semantic and syntactic properties very similar to those of the Fongbe determiner. In Kihm’s (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 àsvn 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 àsvn. 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 àsvn 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 àsvn 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 mangé 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 mangé 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. àsvn lz crab pl ‘the crabs’

b. Yi yì. 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

d

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 #é (a reduced form of the numeral ó#é ‘one’) which serves as an indefinite determiner as in àsVn #é ‘a crab’. The form #é is in complementary distribution with the definite determiner V but it may co-occur with the plural marker lX, as in àsVn #é lX ‘some crabs’. Furthermore, some speakers (but not all) distinguish between the form #é 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, #é 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 [αspecific] 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 ún(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 lòt ‘the other’, lòt 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 #é/#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 ó#é ‘one’ was relexified as younn on the basis of the French form úne. Brousseau (in preparation) argues that the interpretation of úne 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 wè, 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 [αproximate], 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 [αproximate].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 [αanimate]. 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’ wè – 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 ø ki vini one [+deic] op res come ‘This/that one who came.’ b. – sila ø 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 ø ki vini one [+deic] pl op res come ‘These/those ones who came.’ b. – sila yo ø 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 [αanimate]. 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 [αproximate].

92 (40)

Functional category lexical entries involved in nominal structure a. ! ce # @ ces $

garçon(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 [αanimate] 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 ça, 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]. Ça 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 [αproximate]. By contrast, ceci is used only to pick out objects close to the speaker. Consequently, it must be specified for the feature [+proximate]. Because ça, 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)

*ça / 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 ça // 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)

*Ça/cela/ceci que j’ai vu ‘What I saw.’

french (=(15) in Lefebvre 1997)

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

Ce que j’ai vu. ‘What I saw.’

french (=(16) in Lefebvre 1997)

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

94 (48)

Functional category lexical entries involved in nominal structure a. ça:

/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 [αanimate]. 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, ça, 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 [αanimate]. 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 [αproximate]. 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:

/sø/ [+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 [αanimate]. 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 [αproximate], 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 là 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.’

là. [+deic]

french (=(21) in Lefebvre 1997)

(53)

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

b. cette bague-là 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, là is a general deictic form, not specified for any value of the feature [αproximate]. 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 là and ci are represented in (54). (54)

a. là:

/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, là and ci are adverbial. In French, the positive value of the feature [αproximate] is lexically encoded (e.g. ci), whereas in Haitian it is the negative value of the feature [αproximate] 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 ça, 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 . . . là, 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 élV and nX or énX. According to that author, these forms are generally pronounced lV and nX, respectively, where emphatic é- is not realised. For the Fongbe speakers whose grammar is discussed in this section, while énX is used only to designate objects that are far from the speaker, élV 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 [αproximate].10 In the examples below, the meanings of énX and élV 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 [αanimate]. The Fongbe demonstrative terms élV and énX have the following distributional properties. They are [+deictic] determiners and, as shown in (55), they occur postnominally. (55)

a. àlwkx élv // énx ring [+deic] // [−prox] ‘this/that // that ring’

fongbe

b. súnù élv // énx 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)

àlwkx gé élv // énx 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 élV and énX 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, élV and énX 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 – élv // énx. 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. – élv 3ê é wá one [+deic] op res come ‘This/that one who came’

v. det

fongbe

b. – énx 3ê é wá 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. – élv lz 3ê é wá one [+deic] pl op res come ‘These/those ones who came’

v. det

fongbe

3ê é wá op res come

v. det

fongbe

b. – énx lz one [−prox] pl ‘Those who came’

(=(28) in Lefebvre 1997)

If élV and énX 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 jló – kíkló élv v. I want one big [+deic] det ‘I want this/that big one.’

fongbe

v. det

fongbe

b. N jló – kíkló énx I want one big [−prox] ‘I want that big one.’

(=(29) in Lefebvre 1997)

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

a. /élv/ [+deictic]

‘this/that’

fongbe

b. /énx/ [+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 [αproximate]. Sila in Haitian and énX in Fongbe are both [−proximate], while sa and élV are [αproximate] (see (39) and (61) ). In contrast to French, both languages lack a form that lexically encodes the positive value of the feature [αproximate] (see (48c) ). The lexical entries of the demonstrative terms in the two languages are represented in (62). (62)

haitian a. /sa/ [+deictic]

fongbe /élv/ [+deictic]

b. /sila/ [+deictic] [–proximate]

/énx/ [+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 ça 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-là ‘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 ça and either cela or celui-là. 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

/(é)lú/ [+deictic]

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

b. fongbe lexical entry

french

/(é)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: sín ‘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 mímVn ‘denial’ is not affected and hence it cannot occur as the complement of tWn although it can occur as the complement of sín. (64)

a.

nùgbó v sín mímvn v truth det of denial det ‘the denial of the truth’

b. *mímvn denial

nùgbó 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 sín but not with tWn, as shown in (65). (65)

a.

[càkpálô sín] gò corn-beer of bottle ‘the bottle of beer’

b. *gk bottle

[càkpálô 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 sín, as shown in (66). (66)

a.

Àwà [vì v twn] v arm child det gen det ‘The child’s arm has a wound.’

b. ?*[vì child

v det

sín] àwà of arm

dó have

àkpà. wound

fongbe

v dó àkpà 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 sín, it is not with tWn, as shown by the contrast in grammaticality between (67a and b). (67)

a.

[Kwkú sín] [Aristote sín] 3ì3è lz fongbe Koku of Aristotle of sketch pl [Lit.: ‘the sketches of Koku of Aristotle’] ‘Koku’s sketches of Aristotle’

[Aristote twn] b. *3ì3è sketch Aristotle gen

[Kwkú 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 sín 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)

Báyí sín 3í3é Kwkú twn v Bayi case sketch Koku case the ‘(the) Koku’s sketch of Bayi’

fongbe

dp d′ Det

fp f′

kp

ú

f 0 Kwkúi twn ø

np n′

kp

ti n0 3í3e

Báyí sín

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

the king’s portrait of the queen

english

fp kp

f′

the king’s f 0 ø

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 à ‘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 d’Aristote de Rembrandt du Louvre ‘the portrait of Aristotle by Rembrandt in the Louvre’

french

b. la réponse du professeur à la question de l’étudiant french ‘the answer of the professor to the question of the student’ (=(17) in Lumsden 1991)

Thus, the French forms à 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 d’Aristote ‘his picture of Aristotle’

french

fp f′

np son

np

f0 ø

n′ n0 portrait

kp d’Aristote

(=(19) in Lumsden 1991)

On Lumsden’s 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). 3ì3é Kwkú twn sketch Koku gen ‘Koku’s sketch’

(73)

a.

(74)

a. *Marie Mary

b.

portrait sketch

àsvn 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)

àsvn nyy twn élv 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. pòtre [pèche] a portrait fisherman the ‘the portrait of the fisherman’

haitian

b. pòtre [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 [pwofèse] 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 sín in Fongbe or the properties of à/de in French? Lumsden’s (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. *pòtre portrait

pèche fisherman

pent sa a painter this the

b. *pòtre portrait

pent painter

c. *repons answer

kesyon mwen an question my the

sa pèche this fisherman

a the

pwofèse 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. pòtre pèche 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 pwofèse a sou kesyon answer professor the on question ‘the professor’s answer to my question’

haitian

mwen my

an the

c. repons kesyon mwen pwofèse te fèt 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) Mary’s child’

haitian

dp d′ fp f′

kp

a

f 0 Marii ø ø

np n′

d

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 ø an àsvn Kwkú twn v crab J/K gen det ‘J/K’s crab (in question/that we know of )’

haitian fongbe

b. krab mwen ø an àsvn 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 ø àsvn nyy twn crab me gen ‘these/those crabs of

sa a yo haitian élv 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. respè 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 boutèy byè a ‘the bottle of beer’. In this example, byè 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 pòtre Wòbè Mari a portrait Robert Mary det ‘Mary’s portrait of Robert’

haitian

Assuming that Mari in (83) is part of the Genitive phrase following Lumsden’s analysis above, Wòbè 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 potrè Mari Wòbè a (compare with (83) ), Wòbè 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 d’Aristote de Rembrandt and le portrait de Rembrandt d’Aristote 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, Lumsden’s 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 sín 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 ‘John’s book’

northern haitian

b. liv Jan ø book John gen ‘John’s 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)

ìwé e Ayò book part Ayo ‘Ayo’s 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 à occurring in expressions such as le livre à 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 à, 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.

5

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: j’aimai ‘I loved’, j’aime ‘I love’, j’aimerai ‘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 subject–verb 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 kò

irrealis • Definite future h f ap ná • Indefinite future h f a-va ná-wá • Subjunctive h f pou ní

non-complete • Habitual • Imperfective h f h f – nw ap 3ò . . . 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 speaker’s 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 speaker’s 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 ní 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 sòti. 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 M’ap sòti 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 après with a progressive meaning (après is used in some of the same contexts as en train de or à; see Grevisse 1975: 646; Féraud 1768). Finally, the example in (d) shows the use of été, a form of the auxiliary être ‘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 après/à/en train de manger. ‘John is eating.’

d. Jean a été 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 après in (3c), and Haitian te is phonologically similar to the French form été 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 kò ant would would

ap prepare pat. ná 3à 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 être ‘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: j’ai fini ‘I have finished’, je suis venu ‘I have come’. The formation of the future perfect also follows this pattern: j’aurai fini ‘I will have finished’, je serai venu ‘I will have come’. The past conditional is also formed in this way: j’aurais fini ‘I would have finished’, je serais venu ‘I would have come’, and so on. In Haitian creole, there are no auxiliaries corresponding to être 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 kònnèn 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 tùn 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 wè ‘to catch sight of’, jwenn ‘to find’, etc. Finally, stative expressions do not refer to a process at all. These involve verbs like kònnèn ‘to know’, bezwèn ‘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 Lè 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 wè volè 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 kònnèn 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 été, a form of the French auxiliary verb être ‘to be’. In French, été 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 été malade. ‘Mary had been sick.’

b. Marie a été malade. ‘Mary was sick.’

french

c. Marie aura été malade. ‘Mary will have been sick.’

d. Marie aurait été malade. ‘Mary would have been sick.’

french

While Haitian te occurs with predicates of all aspectual classes (see (7)–(9) ), French été 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, kò.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 kò and a dynamic verb are always interpreted as pluperfect. (11)

Dynamic verb Mari kò 3à wh. Mary ant prepare dough ‘Mary had prepared dough.’

fongbe (=(60) in Lefebvre 1996b)

Sentences containing kò 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 kò mw àjvtv 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 kò tùn 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 kò, 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 kò 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 kò, 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 kò encoding anteriority. They copied this lexical entry, and looked for an appropriate phonetic string in French to relabel it. The French form été (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 kò as te, a reduced form of été.7 This derivation accounts straightforwardly for the fact that the properties of te in the Haitian lexicon parallel those of kò in Fongbe. 5.2.2

The irrealis mood marker pou

Mood expresses the speaker’s 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 sòlda pou vini laplas kunyè 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 protège ‘May God protect you’; Que Marie prépare la pâte ‘Mary should prepare dough’; Que tous les soldats viennent à 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 (être) pour (see (3b) ) in French is described by Grevisse (1975:

5.2 Historical derivation of Haitian

TMA

markers

119

646) in the following terms: ‘Être pour peut servir à indiquer un fait prochain, à présenter une action comme convenue, préparée, ou encore à marquer la nuance qu’exprimeraient “être de nature à, être disposé à, destiné à”.’ [Être 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 speaker’s attitude that the event denoted by the clause may, should or must take place. Although French être 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 ní 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, ní, like pou, has been glossed as ‘subjunctive’. (15)

a. Mfwl ní cv wè. God sub protect you ‘May God protect you.’

fongbe (Anonymous 1983: v, 4)

b. Mari ní 3f wh. Mary sub prepare dough ‘Mary should prepare dough.’

fongbe

c. Mì ní 3ù. 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 ní 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 ní 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 wè ‘to catch sight of’, or a stative verb, such as kònnèn ‘to know’, cannot be interpreted as progressive, as shown in (17). (17)

a. #Jan ap wè volè a John imp catch-sight-of thief det [Lit.: ‘John is catching sight of the thief.’]

haitian

b. #Jan ap kònnèn 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 jwènn 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: être après, être à and être en train de, as is shown in (20). (20)

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

french

b. Marie est à 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 après, as it occurs in (20a) (see Goodman 1964).8 In addition to being assigned a progressive interpretation, both Haitian ap and French après share semantic properties. Like ap (see (19) ), après 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 après 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 après cannot, as shown by the ungrammaticality of (22). (22)

*Nous trouverons Marie après 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 être après. Based on the above discussion, it appears that, although the creators of Haitian used the French phonetic string après, 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 #ò ‘at’. The preposition #ò 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)

Kwkú [3ò [ [àsvn 3ù] 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 ‘know’11), a sentence containing the imperfective aspect is assigned a habitual interpretation, as shown in (24). (24)

Sìká 3ò àzwn-jy wy. Cica at sick post ‘Cica is habitually sick.’

fongbe (=(75) in Lefebvre 1996b)

This should come as no surprise, given Comrie’s (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 #ò . . . 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 Sìká 3ò wh 3à 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 #ò 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 #ò in the copied lexicon was relabelled as apre>apr>ap, derived from the phonetic representation of the French preposition après in such sentences as Marie est après manger ‘Mary is eating’ (see (3c) ). The French form après shares some semantics with #ò, 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 #ò. While #ò 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 ná, respectively. The properties of these markers will be discussed in turn. The definite future marker ap is used to convey the speaker’s 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 speaker’s 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 Comrie’s (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 wè Jan. Mary def-fut catch-sight-of John ‘Mary will catch sight of John.’ c. Stative verb Mari ap kònnèn 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 wè 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 kònnèn 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 ná.

5.2 Historical derivation of Haitian

TMA

markers

125

The Fongbe preverbal marker ná, like the Haitian preverbal marker ap, is used to convey the speaker’s 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)

É ná he def-fut ‘He will die.’

kù. 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 ná 3à wh. Mary def-fut prepare dough ‘Mary will prepare dough.’ b. Resultative verb Mari ná mw Jan. Mary def-fut catch-sight-of John ‘Mary will catch sight of John.’ c. Stative verb Mari ná tùn Jan. Mary def-fut know John ‘Mary will know John.’

fongbe

fongbe

fongbe (=(66) in Lefebvre 1996b)

A Fongbe clause containing ná may also be assigned a future perfect interpretation. This is shown in (34), where ná has scope over the adverb kò ‘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 ná (kò) 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 kò with the definite future marker ná 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 kò ná 3à 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 kò ná 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 kò Mary ant ‘Mary would ‘Mary would

ná tùn Kwkú. 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 ná 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 ná was relexified as ap from the French form après occurring in expressions such as Marie est après 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 être après with an immediate future interpretation. All the examples of être après 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 être après is progressive. This suggests that the creators of the creoles perceived être après 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 être après was the phonetic string used to relabel the lexical entry copied from Fongbe ná. 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 après. In addition to occurring in the periphrastic expression of the progressive (as in (3c) ), après is a preposition meaning ‘after’.

5.2 Historical derivation of Haitian

TMA

markers

127

French grammars report that in sentence-initial position, as in Après avoir mangé, nous partirons ‘After eating, we will leave’, après is a ‘preposition indicating posteriority’ (see d’Alembert, Mouchon and Diderot 1776–7: 560). Féraud (1768) notes that ‘when placed in front of the verb, it only serves as a temporal preposition: e.g. Après dîner “After (eating) lunch”.’ It is clear that the preposition après in these contexts shares with the Fongbe marker ná 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 ná. Although Fongbe ná shares some elements of meaning with the French preposition après, the two lexical items do not occur in the same surface position. Ná occurs between the subject and the verb, and après 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 ná 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 wè-l yo rakònte . . . 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 language’s 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 Lumsden’s (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 ná, 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 ná, as illustrated in (39a). (39)

a. subject . . . apr

xp

b. subject . . .

xp

xp /ø/ [+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 Hazoumê 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 après 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 après ‘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 speaker’s 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 yòn 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 wè 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 wè 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 speaker’s 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 à l’école aujourd’hui ‘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, ná-wá is used to convey the speaker’s attitude that the event referred to in the clause may eventually take place (see Anonymous 1983: v, 3). (47)

É ná-wá kù. 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 ná-wá 3à wh. Mary ind-fut prepare dough ‘Mary will eventually prepare dough.’ b. Resultative verb Mari ná-wá mw Jan. Mary ind-fut see John ‘Mary will eventually catch sight of John.’ c. Stative verb Mari ná-wá tùn Kwkú. Mary ind-fut know Koku ‘Mary will eventually know Koku.’

fongbe

fongbe

fongbe (=(89)–(91) in Lefebvre 1996b)

A Fongbe clause containing ná-wá 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 é ná-wá 3w àdigbán v yi ná-wá xwé. 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 kò ná-wá 3à 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 kò ná-wá 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 kò Mary ant ‘Mary might ‘Mary might

ná-wá tùn Kwkú. 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 má ná-wá 3à 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 ná-wá is morphologically complex. It is formed by the combination of the definite future marker ná and wá, 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 ná and wá, the clause is

5.2 Historical derivation of Haitian

TMA

markers

133

grammatical but ná and wá 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)

Sìká Cica a.

ná nw wá 3à 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 ná-wá, 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 á, the indefinite future marker, or by vá, which in isolation means ‘to go’, or by a combination of both forms: á-vá. 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 ná-wá 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; Déchaine 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

wè catch-sight-of

vòlè 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; Déchaine 1991; Lumsden in press b). (60)

Mari kònnèn 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 3à 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 3à 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-kún v ná. harvest det give ‘The harvest has brought us money.’ [Lit.: ‘The harvest has given.’]

fongbe

b. Jj-nl-kún ná. 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

àjótv 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 tùn 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 3à 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 3à 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 lá

non-complete • Habitual nw . . . my

• Imperfective ly . . . my

• Indefinite future á (vá) • Subjunctive né

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. Hazoumê (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. ná, A, lá). 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 3ò vpn wy lè vpn v lè vpn v / ø tw vpn ø 3ó vpn ø nw vp

fongbe ajagbe gengbe gungbe tufingbe xwedagbe (from Hazoumê 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, Hazoumê’s (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 Bernabé (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 Bickerton’s (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.

6

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 é(yy) mí yé

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 low–high tone) and the form meaning ‘you (pl.)’ as mC (bearing a mid tone). Mid and low–high 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 mí 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 mí ‘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 l’ancien français et souvent encore au XVIe siècle, on employait pour marquer la possession, à la place du pronom possessif atone, le pronom personnel tonique avec de. Cet emploi n’existe plus pour ainsi dire au XVIIe siècle. Cependant on construit quelquefois de avec le pronom personnel de la 3e pers. pour éviter l’équivoque 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 frères ‘his/her brothers’), the construction les frères d’elle ‘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 à, as in ses frères à 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: cè ‘1st person sg.’ and tòwè ‘2nd person sg.’. As can be seen in (6), these forms occur postnominally. (6)

xwé cè / tòwè 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. [xwé [Bàyí twn] house Bayi gen ‘Bayi’s houses’

lz] pl

b. [xwé [nyy / hwy / é / mí / yé 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)

Hazoumê’s (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 ø] yo] house John gen pl ‘John’s houses’ b. [kay [mwen / u / house me / you / ‘my/ your houses’

145

haitian

. . . ø] 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 frères d’elle mentioned above. Indeed, as was argued in chapter 4, French de is equivalent to of in English, not to ’s, so d’elle occurs in the complement position of frères rather than in a specifier position. The Haitian data also contrast in Case with the emphatic pronominal phrase in ses frères à lui also mentioned above. Again, as was extensively argued in chapter 4, French à 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 frères d’elle ‘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 frères à 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 cè and tòwè 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. nôtre nôtres 2nd pl. vôtre vôtres 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 Hazoumê 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)

Kòkú sv [– [Àsíbá / nyy Koku take Asiba me ‘Koku took Asiba’s/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 Mary’s/mine.’

mwen me

ø] ]. 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. Hagège 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 émì 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)

Sìká tùn Cica know hwlá Àsíbá hide Asiba

3w say sìn of

yè they gbh. goat

3w à say you

flín 3w remember say

émì log

fongbe

a. ‘Cica knows that they said that youi remember that youi hid A’s goat.’ b. ‘Cica knows that theyi said that you remember that theyi hid A’s goat.’ c. ‘Cicai knows that they said that you remember that shei hid A’s 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 émì has no independent reference.3 (14)

*Émì hwlá Àsíbá sín gbh log hide Asiba of goat ‘S/he hid Asiba’s 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. *C’est je/C’est 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 mì hwy à wè é(yy) é è4 mí mí yé yé (=(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. Àsíbá 3ò Bàyí / hwy kpvn wy. Asiba at Bayi / you look post ‘Asiba is looking at Bayi/ you.’

fongbe

b. Àsíbá 3ò kpíkpvn wè 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 è 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. Kwkú sv àsv ná è. b. Kwkú sv è àsv ná. 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 annoncé la nouvelle à Mary aux announce det news to ‘Mary announced the news to John.’ b. Marie la lui a annoncée. 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 à ‘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 à 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 à 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 pòte 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 infl.’ 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

remèt 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 kòb 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 Cadely’s (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 ‘It’s not us he saw.’

→

kaye’l dechire a kisa’l fe M we le’l vini Se pa nou’l we

haitian

154

Pronouns Le mwen tande ‘When I hear.’

→

Le’m tande

Le li tande ‘When he hears.’

→

Le’l tande

Le nou tande ‘When we hear.’

→

Le’n 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 infl 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. Kwkú lìn Àsíbá. É lìn è. Koku think Asiba 3rd think 3rd ‘Koku thinks of Asiba. He thinks of her.’ b. À ná yì Pari. À ná 2nd sg. def-fut go Paris 2nd sg. def-fut ‘You will go to Paris. You will go (there).’

fongbe

yì –. 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 . . . Yé 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]

mì

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 [αargument] and [αnominative], 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 [αargument] 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 é 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. É nywn 3w Kwkú ní yì. 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 daprè 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)

(É) 3ì 3w Kwkú j’àzwn. 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)

3ì nú mì / hwy / é . . . ‘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; Déchaine and Manfredi 1994; Kinyalolo 1994; Brousseau 1995a). Déchaine 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 tèt ‘head’ and kò ‘body’ followed by a possessor phrase containing a personal pronoun may also be assigned a reflexive interpretation. (44)

a. Mweni wè mweni nan glas I see me in mirror ‘I saw myself in the mirror.’

la. det

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

touye tèt mweni . b. Mi ap I def-fut kill head me ‘I will kill myself.’

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

lii . c. Lii blese kò 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 wè mi . I see me ‘I saw myself.’

(46)

a. Oui wè wi . you see you ‘You saw yourself.’

b. Jaki wè m j . haitian James see me ‘James saw me.’ (=(2a) in Déchaine and Manfredi 1994) b. Jaki wè wj . haitian James see you ‘James saw you.’ (=(2b) in Déchaine and Manfredi 1994)

As has been repeatedly observed in the literature (see e.g. Carden and Stewart 1988; Déchaine 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 wè l i /j . he see him /her/it ‘S/he saw her(self)/him(self )/it.’

haitian (=(1a) in Déchaine 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 si’aime (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 Corne’s 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 wè mweni / Mweni wè ‘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 d’un plus petit que soi ‘Sometimes one needs someone smaller than oneself’. According to Morin (1977: 75), ‘Soi est nécessairement indéfini, et comme le clitique se, est toujours coréférentiel 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 Smith’s (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 Hazoumê (1990), all Gbe languages encode reflexivisation by means of a personal pronoun +self. In Fongbe, this lexical item is -#éè. 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 -#éè, as is illustrated in (56). (56)

a. Ni ná xù 1sg. def-fut kill ‘I will kill myself.’

nyy-3éèi . me-ref

b.

Bàyíi mw é-3éèi . 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

n

Pronoun

-self

Kinyalolo (1994) proposes to account for the Fongbe pronoun+#éè in the same way. In his analysis, the Fongbe reflexive expression would have the structure in (58). np

(58)

Pronoun

fongbe

n -#éè

There is no overt form in Haitian corresponding to Fongbe -#éè, which might lead one to conclude that the Fongbe lexical entry -#éè has been abandoned and the Haitian facts constitute an independent development. There is, however, another way of looking at the Haitian data given the language’s 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 -#éè 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

n ø

In terms of the relexification process, this would mean that the lexical entry copied from Fongbe -#éè 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 -#éè. A potential French candidate for this task would perhaps be même,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 - même. Peter aux proud of him même ‘Peter is proud of himself.’

french

This lexical item is a potential candidate because Haitian actually has a lexical item mèm which is phonologically derived from French même and because, as is the case in French, Haitian mèm may occur with pronouns, as shown in (61). (61)

mwen mèm / ‘myself mèm /

ou mèm / li mèm, etc. yourself mèm / himself mèm’

haitian

This would be a likely hypothesis under two conditions. First, pronoun+même in French and pronoun+mèm in Haitian should have the same distribution. This is not the case, however. In the limited contexts where même is required in French for a reflexive interpretation to obtain (see (62a) ), mèm does not occur in Haitian. As can be seen in (62b), in this case, a pronoun may be assigned a reflexive interpretation. Tèt-li may also be used and my informants consider that, in this context, pronoun+mèm would be a gallicism. (62)

a. Jeani est en contradiction avec luii John be in contradiction with him ‘John is in contradiction with himself.’

-même. -même

french

(=(54) in Morin 1977) li i/j / tèt -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-même. ‘John speaks with himself.’

french

b. Jani pale ak lii /j / tèt -lii . ‘John speaks with him/himself.’

haitian

These data, together with the fact that the -#éè anaphor occurs in the Fongbe sentences corresponding to (64b) and (65b), shed doubt on the hypothesis that French même would have been used to relabel the lexical entry copied from -#éè. Second, French même and -#éè should have properties that are similar enough to make même an eligible candidate to relabel a lexical entry copied from -#éè. Does même have the properties of an anaphor? There are competing views on this matter. On the one hand, Morin (1977: 77) claims that pronoun+même in French shares some distributional properties with pronoun+self in English. Like many grammarians (see e.g. Grevisse 1975), he concludes that pronoun+même 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 même is optional in some contexts and excluded from others, as shown in (64a and b), respectively. (64)

a. Je leuri ai parlé d’ euxi (-mêmes). 1st 3rd pl. aux speak of them même ‘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

(*-même)

(=(53) in Morin 1977)

On the other hand, Zribi-Hertz (1995) discusses at length the properties that distinguish pronoun+même 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 même-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. 349–50). The analysis that pronoun+même is not an anaphor is not in contradiction with the fact that même is optional in some contexts and excluded from others (see (62) ). Assuming that pronoun+même in French does not have anaphoric status, it contrasts with Gbe pronoun+self, which does (see Kinyalolo 1994), I conclude that French même did not have the appropriate distributional and semantic properties to be used to relabel the lexical entry copied from Fongbe -#éè. 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+mèm (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 wè -m i , mweni -mèm. I see me me mèm ‘I saw myself (emphatic).’

haitian

, li -mèm. b. Jani wè lii John see him him mèm ‘John saw himself (emphatic).’

haitian

The above examples further show that pronoun+mèm 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 -#éè 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

n

n

ø

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; Déchaine and Manfredi 1994). (67)

Yo wè 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)

Yé mw yé – 3éè. 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 -#éè 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-mêmes [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, tèt + pronoun and kò + 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.’

tèt mwèni . head me

lii . b. Jani touye tèt John kill head him ‘John killed himself.’

haitian

haitian (=(36) in Brousseau 1995a)

168 (71)

Pronouns a. Jani blayi kò lii atè. John lay body him on the floor ‘John lay down on the ground.’

haitian

lii . b. Jani chape kò 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 tèt + pronoun and kò + pronoun are not grammatical reflexives (see e.g. Goodman 1964; Déchaine 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’, tèt + pronoun may be assigned either a reflexive or a free interpretation, as shown in (72). (72)

a. Mwen blese tèt mwen. ‘I hurt myself.’ or ‘I hurt my head.’

haitian

b. Ou blese tèt ou. ‘You hurt yourself.’ or ‘You hurt your head.’

haitian

c. Li blese tèt li. ‘He hurt himself.’ or ‘He hurt his head.’

haitian (=(1) in Carden and Stewart 1988)

Similarly, in some contexts kò + pronoun may be interpreted literally, as shown in (73). (73)

a. . . . tout kò m fè 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 tèt is derived from French tête ‘head’ and that of kò 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 tèt + pronoun in Haitian cannot be traced to Fongbe. Furthermore, they also show that the Fongbe word meaning ‘body’ wù 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 gblé ‘spoil’ with the noun wù ‘body’ forms a complex predicate meaning ‘hurt’. Similarly, the verb yá ‘be quick’ combined with wù forms a complex predicate meaning ‘hurry’, the verb lY ‘bathe’ together with wù forms a complex predicate meaning ‘wash (oneself )’, and the verb klú ‘scratch something’ with wù means ‘scratch (oneself )’. As is pointed out by both Kinyalolo (1994) and Brousseau (1995a), in these complex predicates, wù cannot be analysed as a reflexive. First, a complex predicate containing the noun wù ‘body’ may take a direct object which is not coreferential with the subject, as in (75), which shows that wù does not have a reflexive meaning. (75)

É gblé -wù Bàyí. 1sg. spoil body Bayi ‘He hurt Bayi.’

fongbe (=(46b) in Brousseau 1995a)

Second, when a reflexive meaning is intended for a complex predicate containing wù, reflexivisation is achieved by using the -#éè anaphor discussed above, as is shown in (76). According to Brousseau (1995a), -#éè is optional in this context. (76)

Kwkú gblé wù é Koku spoil body him ‘Koku hurt himself.’

-3éè. self

fongbe (=(68a) in Kinyalolo 1994)

If wù were a reflexive, we would not expect it to co-occur with -#éè, and yet it does. Thus, like Kinyalolo and Brousseau, we have to conclude that Fongbe wù does not in itself induce a reflexive meaning. Consequently, the reflexive usage of the Haitian possessive expression involving kò ‘body’ cannot be attributed to Fongbe either.11 Since the reflexive use of kò and tèt 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 tête ‘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 collègue astu rencontré? which colleague aux you meet ‘Which colleague did you meet?’ b. De quelle façon astu réparé of which manner aux you repair ‘How did you repair the car?’

(78)

french

la voiture? the car

a. Qui astu rencontré? who aux you meet ‘Who did you meet?’ b. Comment astu réparé 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 distingué collègue astu rencontré? which distinguished colleague aux you meet ‘Which distinguished colleague did you meet?’

french

b. De quelle ingénieuse façon astu réparé 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 (bèl) 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 (
a. Ki-sa ou achte? what you buy ‘What did you buy?’

b. *Ki which

bèl 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 bèl 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-lès (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 fè? ‘What did you do?’ (see Valdman et al. 1981; Koopman 1982b; Lefebvre 1986; Brousseau 1995a). A specific property of the form ki-lès ‘which one’ is that it occurs with the plural marker yo when a plural meaning is intended. (83)

Ki-lès 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 bèl bagay ‘which nice thing’; ki bèl kote ‘which nice place’; ki bòn jan ‘which good manner’; ki bèl kalite ‘which nice kind’; ki bòn lè ‘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 fè 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 bò ‘which place/where’ ki jan ‘which manner/how’ ki kalite ‘which kind/how’ ki lè ‘which time, moment/when’14

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; Furetière 1984).

174 (86)

Pronouns lequel / laquelle / lesquels / lesquelles qui que / quoi où 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), Hirschbühler (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. C’est 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. À 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 Furetière 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 où ‘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 côté / bord (de) quelle manière quelle sorte (de) quel moment

‘which ‘which ‘which ‘which ‘which ‘which

person’ thing’15 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-lès

‘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 / bò jan

ki kalite ki lè pou ki (-sa)

semantically closest french form lequel, laquelle, lesquel(le)s que / quoi comment combien

quelle personne / qui ?quelle chose quel côté / quel bord / où ‘which manner/ how’ quelle manière / 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-lès

(le)quel est-ce [keles] *quel + ça comment [kumã] 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 côté / quel bord quel genre quelle qualité quelle heure

‘which ‘which thing’ ‘which ‘which ‘which ‘which

*pour quel + ça

‘for which this’

kote / ki bò jan kalite lè

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 qualité ‘which quality’ does not mean ‘how’ as it does in Haitian, and so on. The Haitian form ki-lès ‘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 ça and *pour quel ça, 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-lès ‘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; Hirschbühler 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:

3è-tx (é)-tx / àní n6gbwn nàbí

b. Wh-phrases:

my tx nl tx fí (tx) àlw tx àlwkpà tx hwènù tx (é)tx ú(tú) / àní ú(tú)

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 àní ‘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^ à #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 ká gbWn bW à wá [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, hwènù tX ‘when’ may also be realised as hwètXnù where tX is incorporated between hwè ‘sun’ and nù ‘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’: (é-)tX and àní. Brousseau (1995a) points out that the two forms do not seem to have any distinguishing semantic or syntactic properties except for the fact that àní is less acceptable than (é-)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 ú(tú) ‘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-lès (ki-)sa kouman konbyen

fongbe 3è-tx (é-)tx / àní n6gbwn nàbí

b. Wh-phrases: ki mounn my tx ki bagay nl tx (ki) kote / ki bò fí (tx) ki jan àlwtx ki kalite àlwkpà tx ki lè hwènù tx pu ki(-sa) (é)tx ú(tú) / àní ú(tú)

‘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 àní ‘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 #è-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)

3Dè-tx

lz à 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 nàbí ‘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. Nàbí à xw? how-much / many you buy ‘How much/many did you buy?’

fongbe

b. Hwèví nàbí à 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 (é)tX ú(tú) 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 ú(tú) 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’, où ‘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 ú(tú), 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-lès ‘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-œr] (quelle heure ‘which hour’) was assigned the phonological representation /ki lè/ ‘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 à 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 Furetière 1984). The lexical entry corresponding to Fongbe fí ‘place’ was relabelled as kote or bò ‘place, side’ on the basis of the French words côté or bord ‘side’. Fongbe àlW ‘manner’ was relabelled as jan ‘manner’ on the basis of French genre ‘style, type’. Fongbe àlWkpà ‘kind, manner’ was relabelled as kalite ‘kind, manner’ on the basis of French qualité ‘quality’. The Fongbe derived noun #è-tX ‘which one’ has a Haitian counterpart ki-lès ‘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 nàbí ‘how many/much’ was relabelled as konbyen ‘how many/much’ on the basis of French combien ‘how many/much’, and that corresponding to (é-)tX ‘what’ was relabelled as (ki-)sa, where ki corresponds to tX and sa to é. As is pointed out in Brousseau (1995b), both sa and é are deictic forms which can be translated as ‘that’. The Haitian form sa takes its phonological representation from the French deictic form ça ‘that’, as shown in chapter 4. As for the Fongbe form àní ‘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 ànì 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 àní, 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 lè presents the following particularity. This lexical entry takes its phonological representation from French l’heure which means ‘the hour’. According to Valdman et al. (1981), however, the Haitian noun lè means ‘hour, time, clock and watch’. The latter meaning of the Haitian noun lè is exemplified in (96).

182 (96)

Pronouns Lè m lan rete. watch me det stop ‘My watch has stopped.’

haitian (from Valdman et al. 1981)

In Fongbe, the noun gàn shares with the Haitian noun lè the meanings ‘hour, clock, and watch’ (see Segurola 1963). This argues that Fongbe gàn was relexified by lè on the basis of French l’heure. Interestingly enough, however, while Haitian lè also means ‘time/moment’, Fongbe gàn does not. In Fongbe, this notion is rendered by hwènù ‘time, moment’. I can think of two scenarios to explain this situation. In the first one, hwènù was not relexified and the semantics of lè, relexified from gàn, was extended so as to cover that of hwènù. This could not be attributable to the influence of French since French l’heure means ‘the hour’. A more plausible scenario, however, is the following. Suppose the creators of Haitian relexified gàn on the basis of French l’heure, yielding lè. Suppose that they also relexified hwènù based on French l’heure. This would yield two homophonous Haitian lexical entries, each of which corresponded to one original lexical entry, as shown in (97). (97)

fongbe a. gàn

haitian lè ‘hour, clock, watch’

b. hwènù

lè

‘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 lè 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

7

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. Yes–no question markers are discussed in section 7.7. Finally, the markers expressing the speaker’s 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 kwè ‘believe’, di ‘say’, panse ‘think’, etc., is phonologically null, as shown in (1). (1)

Jan kwè / di / panse [CP ø 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 kwè / panse lii /j refè. 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 Sterlin’s proposal. Sterlin (1988) also shows that the subject of the complement clause may be extracted, as illustrated below. (3)

Ki mounni li kwè / panse kii refè? 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 language’s 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)

Kwkú 3ì / 3w / lìn [3w Bàyí wá]. 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. Kwkú Koku

3ì believe

b. Kwkú Koku

3ì [3w [Bàyí wá] ]. believe comp Bayi come

[3w [ø say comp

[Bàyí Bayi

wá] ] ]. 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 táa ‘that’, a reduced form of the verb tákì ‘say’. Táa introduces the sentential complement of verbs of the say-class, as illustrated in (7). (7)

A tákì táa dì mujèe . . . he say that the woman . . . ‘He said that the woman . . .’

saramaccan (=(85b) in Byrne 1987)

The above data suggest that, in Saramaccan, the verb tákì ‘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 . . . ]] tákì ø saramaccan di ø haitian

I further hypothesise that, in the development of Saramaccan, the verb tákì 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 Jakmèl]. Jacmel ‘They wanted me to join Jacmel’s troops.’ [Lit.: ‘They wanted that I joined Jacmel’s troops.’] (=(10) in Koopman and Lefebvre 1982)

(10)

Li bòn [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)

Pòte sa pou mwen. bring this for me ‘Bring this for me.’

haitian (=(4) in Koopman and Lefebvre 1982)

(12)

M te bezwen èskont 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 être 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 ça pour do this for ‘Do this for me.’

moi. me

french

7.1 Complementisers and complementiser-like forms (16)

J’ ai besoin d’argent 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 j’entre dans les troupes de Jacmel french ‘They wanted me to join Jacmel’s 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 j’entre dans les troupes de Jacmel. ‘They wanted me to join Jacmel’s 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; Hirschbühler 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 être 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 fè]? which thing we ant sub do ‘What were we supposed to do?’ b. Ki which

a. det

te ti ant

fè]? 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

fè

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 nú, as shown in (25) and (26), respectively. (25)

Ùn jló nú à ní wá. 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)

É nyw nú ùn ní yì. 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 ní, as shown in (27) and (28). (27)

Ùn jló ní à ní wá. I want comp 2nd sub come ‘I want you to come.’ [Lit.: ‘I want that you come.’]

fongbe

(28)

É nyw ní ùn ní yì. 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)

Nú and ní 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 nú in (25) and (26) is homophonous with the preposition nú ‘for’ which selects either np complements, as in (29), or purposive clausal complements, as in (30). (29)

Kwkú ná àsvn nú Àsíbá. Koku give crab for Asiba ‘Koku gave a crab for Asiba.’

fongbe (=(ib) in Lefebvre 1994c: 98)

(30)

Ùn 3ò fòngbè kplvn wy nú mì ná fongbe 1st at Fongbe learning part prep 1pl def-fut sìxú 3’àlísá. 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 ní is homophonous with the mood marker ní discussed in chapter 5, as shown in (31) (=(15b) in chapter 5). (31)

Mari ní 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 nú and ní selected by the corresponding verbs and adjectives (see (25), (26) and (27), (28) ) are homophonous

192

Functional category lexical entries

with the preposition nú (see (29) and (30) ) and the preverbal mood marker ní (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 ní and nú. The morpheme ní 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, nú 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 ní was minimally specified for the feature [−realis] and it was relabelled on the basis of French pour occurring in être 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 nú 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 ní was probably first manifest as a mood marker in the incipient creole. Likewise, the Haitian lexical entry relexified from Fongbe nú 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 à qui Marie man comp John believe to whom Mary parlé 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 être arrivé? 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)

(C’est) 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 wè 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

wè ti . see

di Jan wè 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

wè 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 wè Mari. it-is John James say ? see Mary ‘It is John that James said saw Mary.’

haitian

b. (Se) ki mounn, Jak di ki wè 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)

Koopman’s 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 Koopman’s account, Haitian ki is like French qui, that is, a tensed complementiser which has the property of properly governing the empty subject position. Koopman’s analysis can be challenged from both a theoretical and an empirical point of view. First, a major drawback to Koopman’s 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, Koopman’s 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 entèlijan. it seem John intelligent ‘It seems that John is intelligent.’

haitian

b. Jan sanble *(li) entèlijan. John seem he intelligent ‘John seems to be intelligent.’

haitian (=(1) in Law 1992)

c. Yo sanble *(yo) entèlijan. 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 entèlijan? 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. Kwkú wy Àsíbá mw. Koku it-is Asiba catch-sight-of ‘It is Koku that Asiba caught sight of.’

fongbe

b. È tx (wy) Àsíbá 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 é 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) Kwkú 3ì 3w éi person which it-is Koku think that ? ‘Who is it that Koku thinks saw Asiba?’

a.

Àsíbá? Asiba

fongbe

wy Kwkú 3ì 3w ti mw Àsíbá

fongbe

tx (wy) Kwkú 3ì 3w Bàyí mw ti ? Èi 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. *Èi tx wy

Kwkú 3ì

3w

Bàyí mw èi

fongbe

The ungrammaticality of (54b) contrasts with the grammaticality of (55a), revealing a subject/object asymmetry. The form é 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 é. This is shown in (56). (56)

[M6 lx]i (wy) Kwkú 3ì 3w yii mw Àsíbá? 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 #ì ‘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 #ì, as in (57a). The embedded subject may be raised to the subject position of #ì, 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. (É) 3ì m6 3è Kwkú j’àzwn v. 3rd seem person op Koku sick det ‘It seems that Koku is sick.’

fongbe

b. Kwkú 3ì 3w é j’àzwn. Koku seem that res sick ‘Koku seems to be sick.’

fongbe

c. Yé 3ì m6 3è yé 3rd pl seem person op res ‘They seem to be sick.’

j’àzwn. 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 yì? person it-is leave ‘Who is it that left?’

b. *M6i wy éi yì ‘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 #Iè which shows up in relative and factive clauses, as shown in (59) and (60), respectively. (59)

Xw 3iè my un d’àmlò 3è house op in I sleep loc ‘The house in which I slept.’

v. det

fongbe (=(14d) in Kinyalolo 1993a)

(60)

Wá 3iè Jan wá v víví nú 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 #Iè can pied-pipe postpositions ( just as which can pied-pipe prepositions in English). In (59) #Iè pied-pipes the postposition mY ‘in’. In (61) (based on Collins 1994), it piedpipes the postposition jí ‘on’ (for an analysis of this type of factive construction, see Collins 1994). (61)

Távò 3iè jí Bàyí súsú v víví nú 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 #Iè does, #Iè 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) #Iè is co-indexed with the noun távò ‘table’. It could not be co-indexed with a postpositional phrase, as is shown by the ungrammaticality of (62) (based on Collins 1994). (62)

*Távò table

jí 3iè on op

Bàyí Bayi

súsú wipe

v víví det please

nú mi for me

fongbe

The fact that #Iè 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 øi m sòti 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 Wá 3iè Rive ø arrive op ‘The fact that

Jan wá 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

sòti 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 [øi 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 arrivée.

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 (kòl)ak (
a. *Jan (e)pi Mari

b. Jan (kòl-)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 yì bw Mari wá. ‘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 kpó#ó (Lit.: ‘with-at’) to the first noun. (71)

Jan kpó3ó Mari (kpó) John with Mary with ‘John and Mary’

fongbe (=(72) in Lefebvre 1993b)

The properties of Haitian (e)pi and (kòl-)ak are derivable straightforwardly according to the relexification hypothesis. BW was relexified as (e)pi on the basis of French et puis and kpó#ó was relexified as (kòl-)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 c’est ‘it is’. Both occur in cleft constructions, as shown in (72). (72)

a. Se

Jan

Mari

wè.

b. C’est Jean que Marie a aperçu. ‘It is John that Mary caught sight of.’

haitian french

In spite of their apparent similarity, however, se and c’est 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 c’est cannot appear in this context (see (73b) ), unless the subject has been topicalised (see (73c) ). (73)

a.

Bouki

b. *Bouki c.

se

yon

c’est un

Bouki, c’est un ‘Bouki is a doctor.’

dòktè.

haitian

médecin

french

médecin.

french

Furthermore, the analyses proposed for Haitian se reveal the differential properties of se and c’est. C’est is made up of a demonstrative pronoun, ce/c’ ‘it’, which occurs in subject position, and the copula être ‘to be’. Haitian se, however, is an unanalysable word. According to Lumsden’s (1990) analysis, se heads infl in (73a) and occurs in subject position in (72a). By Deprez and Vinet’s (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 c’est. It thus appears that, although c’est 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 wè. 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 nyì, as in (75b).14 (75)

a. *Kwkú b.

wy

dótó

Kwkú nyì dótó. ‘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)

Andí wy i bói? 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 Pollock’s (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)

t′

np

NegP

t

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 allé au marché. 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 cinéma. John (n’) go-fut pas to-the movies ‘John will not go to the movies.’

209

french (=(2c) in DeGraff 1993a)

On DeGraff’s analysis, this contrast in word order follows straightforwardly from the analysis that, in French, the tensed verb raises to infl (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 s’ennuient. french Bouki makes the clown for pas that 3pl bore+ref ‘Bouki is clowning around so that they don’t get bored.’ (=(11a) in DeGraff 1993a) *Bouki Bouki

ap fè 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)

Voilà un type pas bête. there a fellow pas stupid ‘There goes a man who is not stupid.’

french (=(12a) in DeGraff 1993a)

(84)

*Men yon mounn pa sòt 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 DeGraff’s 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 Hazoumê 1990). In Fongbe, this marker is mà. As is the case with Haitian pa, this marker generally precedes the tma markers. Compare (85) with (79a). (85)

Kwkú mà ní wá àxì 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/mà, 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. mà ní 3à wh. neg sub prepare dough does not have to prepare dough.’

haitian fongbe

b. Mari Mari Mary ‘Mary

pou ní sub should

haitian fongbe

pa prepare pat. mà 3à 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 infl 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 mà cannot occur at the periphery of the clause that it modifies, as shown in (87). (87)

*À 2nd yàví bore [Lit.:

3ò yvnú-tvm= bló wy bónú mà vj l= at clown do part in-order neg child pl ó ins ‘You are clowning so that the children don’t get bored.’]

fongbe

Finally, according to current analyses of markers like mà in the literature on Kwa languages (e.g. Da Cruz 1994, for Fongbe; Collins 1993 and Àgbèdvr 1995, for Ewe), such markers are the head of NegP. If mà 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 mà- in participial forms modifying a noun, as shown in (88). (88)

Kwkú ná làn màbí-màbí Kòfí. 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 mà

7.7 Yes–no question markers

211

is homophonous with the inversive/negative derivational prefix mà-. As will be discussed at length in chapter 10, mà- is a prefix that attaches to verbal bases, as shown in the following pair of words: #ì ‘to believe’, mà-#ì ‘to doubt’ (see Brousseau 1990). The derived word may be reduplicated to form a participial adjective, as in (88). Whether mà and mà- 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 mé and the inversive/negative prefix is mà- as shown in (89). (89)

a. Jan mé yì ò. John neg go ins ‘John did not go.’

ewe (=(2) in Wallace 1995c)

b. màbl-màbl ‘disrespect’

ewe (=(6b) in Wallace 1995c)

The negation marker mé is the head of NegP in the clause (see Àgbèdvr 1995; Collins 1993) and does not occur in nps, where only the inversive/negative prefix can occur (Wallace, p.c.). Thus, in Ewe, mé and mà- can be argued to constitute two distinct lexical entries with different distributional properties. This analysis can be extended to Fongbe. In this view, mà is the head of NegP and does not occur in nps, and the prefix mà- constitutes a separate lexical entry. Therefore, the conclusion that Haitian pa and Fongbe mà share the same properties can be maintained. The parallel distribution of mà and pa and the fact that they are both analysed as the head of NegP follow from the relexification hypothesis: Fongbe mà 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 kún (see Segurola 1963) which occurs in complementary distribution with mà. 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 kún was also relexified on the basis of French pas, such that, in modern Haitian, there is only one lexical entry pa. 7.7

Yes–no question markers

Fongbe has a sentence-final yes–no question marker à. (90)

Kwkú xw àsvn lx à? Koku buy crab pl q ‘Did Koku buy the crabs?’

fongbe

212

Functional category lexical entries

According to Hazoumê (1990), not all Gbe dialects have such a particle. Gungbe and Xwedagbe encode yes–no 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 yes–no question marker and direct questions are encoded by means of rising intonation. In Haitian, there is no sentence-final yes–no question marker. Questions are encoded in one of two ways: rising intonation, as in (91a), or èske at the beginning of the clause, as in (91b). (91)

a. Jan

vini?

haitian

b. Èske 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 arrivé? ‘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 language’s 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 yes–no questions, as shown in (93). (93)

Est-ce que Jean est arrivé? 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 èske (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 speaker’s point of view

213

In standard French, yes–no questions may also be encoded by raising the tensed verb to comp, as shown in (94). (94)

Est i -il ti arrivé? ‘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 infl 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, yes–no questions in French may be encoded by rising intonation, as in (96). (96)

Jean est arrivé? ‘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 speaker’s point of view

Da Cruz (1994) documents the fact that, in Fongbe, there are markers which express the speaker’s point of view on the proposition: F is a negative marker (see (97) ) and ó is a marker that Da Cruz glosses as ins for ‘insistence’ (see (98) ). (97)

Kwkú xw àsvn lx f. Koku buy crab pl Neg ‘Koku did not buy the crabs.’

fongbe (=(2) in Da Cruz 1994)

(98)

Kwkú ní xw àsvn lx Koku sub buy crab pl ‘Koku must buy the crabs (insistence).’

ó. 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 mà, discussed in section 7.6, except in conditional clauses, as is exemplified in (99a and b), respectively.

214 (99)

Functional category lexical entries a. *Kwkú Koku

mà xw neg buy

àsvn lx crab pl

f Neg

fongbe (=(6) in Da Cruz 1994)

b. (Ní) Kwkú mà xw àsvn lx f, é ná yì. 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, ó usually occurs in negative clauses with the negation marker mà, as in (100). (100)

Mà xw àsvn lx ó. neg buy crab pl ins ‘Don’t buy the crabs (insistence).’

fongbe (=(27) in Da Cruz 1994)

However, Da Cruz points out that ó 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 ó 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. Àgbèdvr 1995, for Ewe). Rather, these markers express the speaker’s point of view: F denies and ó insists on the content of the proposition. Furthermore, the presence of ó 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 Σp. Based on this proposal, Da Cruz (1994) proposes that F and ó are the head of Σp in Fongbe. This is illustrated in (101). (101)

Σp Σ′

vp

Σ

f ó

Since the markers that head Σp in Fongbe may occur in the same clause as the negation marker (see (100) ), Da Cruz proposes that, in this language, Σp is a projection that is different from negp. Are there any functional items available in Haitian to head a projection Σp? To the best of my knowledge, no Haitian form corresponds to the Fongbe

7.8 Markers expressing the speaker’s 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 ó. 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. Mà yì ó! neg go ins ‘Don’t 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 ò (see Wallace 1995c), which corresponds to the Fongbe marker ó. 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 ó (based on my own fieldnotes). Like Fongbe ó, 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 ó, 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. *É wá ó [Lit.: ‘He went.’]

fongbe

a.

Ale non!

haitian

b.

Yì ó! 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 ó is (as we saw above). Except for their occurrence in imperative clauses, non in Haitian and ó in Fongbe always occur in clauses containing the respective negation markers, pa and mà (or kún). 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. É kún yì ó. ‘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’ jló [ní é mà ‘I want him not to go.’

ale

non.]

haitian

yì

ó.]

fongbe

216

Functional category lexical entries

In both languages, this marker is mutually exclusive with the yes–no question marker, as shown in (107). (107)

a. *Èske li

pa

ale

non

b. *

mà yì neg go

ó ins

Q

É he

haitian à Q

fongbe

This is expected on the analysis that non and ó in (107) are markers of insistence. It is possible to process a negative-interrogative question, as in (108). (108)

a. Èske b.

li

pa

ale?

É mà yì ‘Did he not go?’

haitian à?

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 ó (and as Ewé ò, see Joseph 1995). On Joseph’s (1995) analysis, Ewe ò 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 T’as fait ça, 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, *T’as pas fait ça, 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 ó 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 Σp. Both languages also have another marker of insistence, bó in Fongbe and wi (
3Dù-nú bó. Manje wi. eat ins

fongbe haitian

7.9 Conclusion

217

Unlike the markers discussed above, bó and wi are excluded from negative, interrogative and embedded clauses. This suggests that they do not fill the same position as ó 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 speaker’s point of view on the proposition: MoodP1 (=cp) and MoodP2 (=Σp), a lower position. MoodP1 is headed by markers which do not occur in embedded clauses. For example, the Fongbe interrogative marker à heads this position. Since they do not occur in embedded clauses, Fongbe bó and Haitian wi must also head this position. The Fongbe negative marker F and the markers of insistence, ó 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 #éè of Fongbe was not relabelled because it has no semantic content. It was assigned a null form in the Haitian lexicon. The interrogative marker à 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 ó (and Ewe ò) and Fongbe bó, respectively. These data show that, as in the substratum languages, and in contrast to French, Haitian has grammatical markers which express the speaker’s point of view on the proposition. The Haitian form èske 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.

8

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 Súnù v gbà man det destroy

manchinn nan an. mvtò 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 (=Σp)

t

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

speaker’s 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 Σp 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, ó in Fongbe and non in Haitian. With Da Cruz (1994), I will assume that MoodP2 (=Σp) dominates negp, headed by the negation markers mà 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 Σp 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 Spec–Head 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 speaker’s 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 speaker’s 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 speaker’s 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 à: question marker used in yes/no questions lá: marker used to express surprise b. Class B markers F: negative marker ó: marker of insistence

These markers are illustrated in (4)–(6). (4)

Kwkú yì à? Koku leave q ‘Did Koku leave?’

(5)

Kwkú yì Koku leave ‘Koku left!’

fongbe (=(6) in Lefebvre to appear b)

lá! surp

fongbe (=(7) in Lefebvre to appear b)

(6)

Kwkú yì 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)

*Kwkú Koku

yì lá à 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)

Kwkú xw àsvn lx f Koku buy crab pl Neg ‘Didn’t Koku buy the crabs?’

wy à? 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)

*Kwkú Koku

xw àsvn lx buy crab pl

à 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 kánbyw 3w I ask

[Kwkú yì Koku leave

à] 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 [Kwkú má yì 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 (=Σp), respectively (see (2) ). Mood expresses the speaker’s 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 speaker’s 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 Kwkú kán-wèzún Koku run ‘Did Koku run?’

à? q

Resultative verbs Kwkú mw àjòtv v à? 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 Kwkú sè flànsxgbé à? 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 #é, as in (16). (15)

Súnù v wá à / lá ?/! man Det arrive q / surp ‘The man arrived ?/!’

fongbe

(16)

M6 3é wá à / lá ?/! 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 Pòtoprens la. John arrive Port-au-Prince Det ‘Actually, John arrived in Port-au-Prince.’

haitian1 (=(12) in Lefebvre 1996a)

b. Kwkú wá kùtvnú v. Koku arrive Cotonou Det ‘Actually, Koku arrived in Cotonou.’

fongbe1 (=(11) in Lefebvre 1996a)

(18)

Resultative verb a. Jan wè vòlè a. John catch-sight-of thief Det ‘Actually, John caught sight of a thief.’

haitian1 (=(15b) in Lefebvre 1996a)

b. Kwkú mw àjòtv 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 kònnèn fransè a. Kwkú sè flánsygbè v. J / K know French Det ‘Actually, John/Koku knows French.’

225

haitian1 fongbe1 (=(15c) in Lefebvre 1996a)

b. Ròb la blan Àvw v wé 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 yes–no question marker.6 (20)

*Èske [Jan wè vòlè] a. haitian1 q John see thief Det [Lit.: ‘Has John actually caught sight of a thief?’] (=(34) in Lefebvre to appear b)

(21)

*My 3é wá v à 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)

Súnù v wá 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)

Súnù v wá 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 kònnèn [ròb 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 sé [àvv v wé 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 Súnù v man det ‘Actually, the

rive a. wá v. arrive Det man arrived.’

haitian1 fongbe1 (=(38) in Lefebvre to appear b)

(28)

Yon

mounn rive a. Súnù 3é wá 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 speaker’s 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 Spec–Head 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 wá 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 wá 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. Súnù 3é wá 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. Súnù 3é wá 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

t′

dpi

Mood′

[e]i vp

v

Mood

Mood′ vp

Mood Assertive marker v′

[e]i AspP

t

MoodP2 [e]i

v′

[e]i

t′

dpi t

MoodP2

Mood

tp

v

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 Spec–Head 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 yes–no 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 Súnù v man det ‘The man has

kraze yon manchinn nan. haitian1 gbà mvtò 3é 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 Spec–Head 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 clause’s 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 #é 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 Súnù 3é gbà mvtò 3é 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 Súnù v man det ‘The man has

(37)

Yon

kraze yon manchinn nan. haitian1 gbà mvtò 3é v. fongbe1 destroy a car a Det destroyed a car, as we knew.’ (=(33d) in Lefebvre 1992b)

mounn kraze manchinn nan an. haitian1 Súnù 3é gbà mvtò 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 wè manchinn nan an haitian1 Súnù 3é mw mvtò 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 fransè a haitian1 Súnù 3é sè flánsygbè 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 Spec–Head. 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 Súnù v man det ‘The man has

kraze yon manchinn nan. gbà mvtò 3é 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 Súnù 3é gbà mvtò 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)

Súnù v f ìn mvtò 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)

Súnù v f ìn mvtò 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. *Súnù 3é f ìn mvtò 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. Súnù v f ìn mvtò 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. *Súnù 3é f ìn mvtò 3é 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. Súnù 3é f ìn mvtò 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)

*Súnù 3é mw mvtò 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)

Nèg man

la pa vòle 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 nèg pa a man not

vòle 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)

Nèg la pa vòle 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, wè 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 kònnèn in Haitian and tún 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 wè *Súnù v mw man det see [Lit.: ‘The man saw

vòlè a àjòtv v thief Det thieves, as expected.’]

haitian1 fongbe1 (=(56) in Lefebvre to appear b)

(50)

*Jan *Jan John [Lit.:

kònnèn fransè tùn flánsygbè 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 wé 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. wé. 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. wé 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 3ò wíwá 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 3ò ná wíwá 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 [ Súnù 3é jló [ná man a want irr ‘A man wanted to destroy

kraze manchinn nan gbà mvtò 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 Súnù 3é jló a man a want manchinn nan an]. mvtò v v]. car det Det

[pou Jan [ní Kwkú comp J / K

kraze ní gbà 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 vòle manchinn nan *Súnù v fìn mvtò 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 John’s 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 3à 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. 3à 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 3à 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. yì. 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. yì 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 ná 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 ná 3à 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 ná 3à 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 wá v f. John arrive Det Neg ‘John has not arrived, as it was expected he would.’

fongbe1

b. Jan wá 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

wá f v arrive Neg Det

fongbe2

b. Jan John

wá 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 Súnù v man det ‘Actually, the

b. *Mounn *Súnù man

wè manchinn nan an. haitian1 mv mvtò v v. fongbe1 see car det Det man saw the car.’ (=(81) in Lefebvre to appear b)

nan wè manchinn v mv mvtò 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 nùgbó in Fongbe7 to assert the content of the proposition, as in (69).8 (69)

Mounn nan Súnù v man det ‘Actually, the

wè manchinn nan vre. mv mvtò v nùgbó. 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 Súnù v gbà man det destroy ‘The man destroyed the

241

manchinn nan an. haitian1 mvtò v v. fongbe1 car det Det car, as we knew.’ (=(84) in Lefebvre to appear b)

b. *Mounn nan kraze manchinn *Súnù v gbà mvtò 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 à 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.

Kwkú wá sv f v à? fongbe Koku arrive yesterday Neg Det q ‘Didn’t Koku arrive yesterday, as expected?’ (=(86a) in Lefebvre to appear b)

b. *Kwkú wá sw f à Koku arrive yesterday Neg q [Lit.: ‘Didn’t Koku arrive yesterday?’]

fongbe (=(15b) in Da Cruz 1994)

The way to rescue (71b) is to use à(G)é as the form of the negative interrogative, as shown in (72). (72)

Kwkú wá sw à(g)é? Koku arrive yesterday Neg int ‘Didn’t Koku arrive yesterday?’

fongbe (=(15a) in Da Cruz 1994)

242

The determiner and the structure of the clause

The use of à(G)é may be seen as a strategy to circumvent the problem of having to pronounce the sequence *F à 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 men’s 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 à 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 [ø ki kraze [manchinn *Súnù [3è é gbà [mvtò man op res destroy car [Lit.: ‘The man who destroyed the car.’]

nan] ] an v] ] v det det

haitian2 fongbe2

b. Mounn [ø ki kraze Súnù [3è é gbà man op res destroy ‘The man who destroyed the

nan] ] – v] ] – det det

haitian2 fongbe2

[manchinn [mvtò 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 /é/ instead of as /v/. /é/ is otherwise the weak form of the third-person personal pronoun. (80)

[Súnù [ [3è é xw àsvn v] man op res buy crab det ‘The man who bought the crab.’

é] ] 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 [ø Súnù [3ê man op ‘The man that

Jan wè] 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 [ø ki kraze Súnù [3è é gbà man op res destroy ‘The man who destroyed the

[manchinn [mvtò 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 lá 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)

Kwkú mw àjòtv v lá. Koku see thief Det ass ‘Actually, Koku saw the thief.’

245

fongbe3 (=(98) in Lefebvre to appear b)

(84)

Súnù 3é gbà mvtò 3é lá. 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

9

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. É yì / wá. ‘He left/arrived.’

haitian fongbe

b. Il est parti / arrivé. ‘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 à ‘at’. This contrast is shown in (2). (2)

a. Li ale / rive Pòtoprens. É yì / wá Kùtvnú. ‘He went to/arrived in Port-au-Prince/Cotonou.’

haitian fongbe

b. Il est allé / arrivé à 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. É hk / lhn. ‘He barked/jumped.’

haitian fongbe

b. Il a aboyé / sauté. ‘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 fè m mal. haitian stomach me imp do me hurt ‘I have a stomach-ache.’ [Lit.: ‘My stomach hurts me.’] (=(29) in Koopman 1986) b. Tèt mwen ap fè 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 L’estomac me brûle [Lit.: ‘The stomach burns me’], La tête 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 à la tête. 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. Xómy wílí mi. stomach hold me ‘I have a stomach-ache.’ [Lit.: ‘My stomach is holding me.’]

fongbe

b. Tà 3l mi. head eat me ‘I have a headache.’ [Lit.: ‘My head is eating me.’]

fongbe

c. À3ú 3w 3ú3ú 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. Lòraj 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 grêle. ‘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’, L’orage 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. Jí jà. rain falls ‘It is raining.’ [Lit.: ‘Rain falls.’]

fongbe (=(1) in Dumais 1988)

b. Láglásì jà. ice fall ‘It is hailing.’ [Lit.: ‘Ice falls.’]

fongbe (=(2) in Dumais 1988)

c. Jòhwn nyì. wind blow ‘It is windy.’

fongbe (=(3) in Dumais 1988)

d. Hxbyósú dk-gbè. thunder growl ‘It is thundering.’

fongbe

e. Jí 3è-àtán. 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) fè cho ‘It is hot’ and (Li) fè frèt ‘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): yózó gbX (Lit.: ‘heat beexcessive’) (see Dumais 1988) and jòhWn 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) fè cho and (Li) fè frèt 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), tèt-+pronoun (Lit.: ‘X’s head’), or kò-+pronoun (Lit.: ‘X’s 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 tèt-lii . John kill head-his / him ‘John killed himself.’

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

c. Jani chape kò-lii John escape body-his ‘John escaped.’

/ /

*tèt-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 tèt-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 prèse

‘to ‘to ‘to ‘to

mess one’s hair’ hang or hang onto’ comb one’s hair’ hurry’

b. mande raple trène

‘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 – – – (# tèt-li) 2 (# tèt-li) (# tèt-li)

haitian

tèt-li haitian tèt-li tèt-li (=(7) in Brousseau 1995b)

The verbs in (12) are those which select tèt-li as the most prominent form. Some of these verbs may also take either a pronominal form or kò-li as a reflexive. (12)

Verbs which typically take tèt-li a. bèse pann touye trène

‘to ‘to ‘to ‘to

drop down’ hang oneself (suicide)’ kill oneself’ train’

b. gade wè

‘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

kò-li haitian kò-li kò-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 tèt-li or kò-li. (13)

Verbs for which no reflexive form is prominent a. enskri meliore

‘to register’ ‘to improve’

b. blayi debouye depèche fòse layite

‘to ‘to ‘to ‘to ‘to

lie down/stretch out’ manage/cope’ hurry’ try hard’ lie down/stretch out’

other reflexive forms tèt-li tèt-li

haitian

tèt-li / kò-li haitian tèt-li / kò-li tèt-li / kò-li tèt-li / kò-li tèt-li / kò-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 kò-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’ bèse ‘to bend down’ b. chape sove

‘to escape’ ‘to escape/run away’

c. blèse

‘to hurt (oneself)’

kò-li kò-li

haitian

haitian

li / tèt-li / kò-li haitian (=(5) in Brousseau 1995b)

The verb blèse ‘to get hurt’ may select any of the four forms, as is illustrated in (15). (15)

Blèse ‘to a. Jani John ‘Johni

hurt (oneself)’ blèse (nan pye hurt in foot hurt hisi foot.’

lii). his

haitian

b. Jani blèse li i /j (nan pye li i /j). John hurt him in foot his ‘Johni hurt hisi /j foot.’ c. Jani blèse tèt-lii John hurt head-his ‘Johni hurt hisi foot.’

(nan pye in foot

d. Jani blèse kò-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. Brousseau’s (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. Bàyíi mwn é-3éèi . Bayi see her-self ‘Bayi saw herself.’

b. É lx. fongbe he bath ‘He bathed.’ ( (16a)=(11b) in Brousseau 1995b)

Brousseau’s (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 prèse mande raple trène c.

fongbe ‘to ‘to ‘to ‘to ‘to ‘to ‘to

mess one’s hair’ hang or hang onto’ comb one’s hair’ hurry’ wonder’ remember’ drag oneself’

– 3ò kàn kw – hwlx kán bíyv fílín dwn

reflexive form in fongbe – – é3éè – é3éè

Haitian verbs which also select tèt-li haitian abitye amize fiye

fongbe ‘to get used to something’ má ‘to have fun’ – ‘to trust’ –

reflexive form in fongbe –

(=(17) in Brousseau 1995b)

Similarly, the comparison of verbs which typically select tèt-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 tèt-li, most Fongbe verbs select a pronominal form.

9.4 Reflexive verbs (18)

257

Haitian verbs which typically select tèt-li a. Haitian verbs which select no other forms haitian bèse pann touye trène

fongbe ‘to ‘to ‘to ‘to

drop down’ hang oneself (suicide)’ kill oneself’ train’

– 3ò kàn kw hù –

reflexive form in fongbe é3éè é3éè

b. Haitian verbs which also select a pronoun haitian gade wè

fongbe ‘to look at oneself’ ‘to see/look at oneself’

kpvn mw

reflexive form in fongbe é3éè é3éè

c. Haitian verbs which also select kò-li haitian grate lave touye

fongbe

reflexive form in fongbe ‘to scratch (o.s.)’ klú wù / é3éè ‘to wash (o.s.)’ lx wù / súsú wù – ‘to kill oneself (by doing x)’ gbà tà 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 / tèt-li haitian enskri meliore

fongbe ‘to register’ ‘to improve’

reflexive form in fongbe

– –

b. Haitian verbs which select li / tèt-li / kò-li haitian blayi debouye depèche fòse layite

fongbe ‘to ‘to ‘to ‘to ‘to

lie down/stretch out’ manage/cope’ hurry’ try hard’ lie down/stretch out’

reflexive form in fongbe é3éè

drxn – yá wù – – drxn é3éè (=(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 bèse

fongbe ‘to get dressed’ ‘to have a bath’ ‘to bend down’

3ò àwù lx fx

fongbe selectional properties – – –

b. Haitian verbs which also select kò-li haitian chape sove

fongbe ‘to escape’ ‘to escape/run away’

hwn hwlxn gán

fongbe selectional properties – –

c. Haitian verbs which also select li / tèt-li / kò-li haitian blèse

fongbe ‘to get hurt’

fongbe selectional properties gblé-wù – (=(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 verb’s 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 creole’s 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 entry’s selectional properties at some point in the creole’s 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 é 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. É 3ì 3w Kwkú j’àzwn. 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 à Jean. it remain fifty gourds to John [Lit.: ‘There remain fifty gourds to John.’]

french

c. É kpò kpvnwè 3ò Kwkú sín. 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 à Jean. it be missing fifty gourds to John [Lit.: ‘It is lacking fifty gourds to John.’]

french

c. É hwè kpvnwè nú Kwkú. 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. É nywn 3w Kwkú ní yì. 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 #ì ‘to seem, to resemble’ also allows for both. The Haitian verb genlè 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 Gí ‘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.

Rète

pou

Jan

vini.

b. *Li rète 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 kpò 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. *

Kpò

3w

Kwkú

ní

wá

b. É kpò 3w Kwkú ní wá. 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 (
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 sèpan (nan pannye it remain a snake in basket ‘There remains a snake (in the basket).’

an). det

haitian (=(14) in Dumais 1988)

b. Yon sèpan 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 kpò, 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 é. In (31b), it occurs in the verb’s subject position.

264 (31)

The syntactic properties of verbs a. É kpò dàn 3ókpó (3ò xàsún it remain snake one (at basket ‘There remains a snake in the basket.’

v my). det in)

fongbe

(=(14) in Dumais 1988) b. Dàn 3ókpó kpò (3ò xàsún 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 kpò 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 kpò 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 sèl nan soup it be-missing salt in soup ‘There lacks salt in the soup.’

la. det

haitian (=(16) in Dumais 1988)

b. Sèl 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.’

sèl. 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 é. 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. É hwy zy 3ò núsúnú v my. it be-missing salt at soup det in ‘There lacks salt in the soup.’

fongbe (=(16) in Dumais 1988)

b. Zy hwy 3ò núsúnú my. salt be-missing at soup in ‘Salt is lacking from the soup.’

fongbe (=(17) in Dumais 1988)

c. Núsúnú 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 resemble’9 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 être 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 #ì ‘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 é, 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 é, as shown in (38b). (38)

a. (É) 3ì nú 3w Jan it seem thing that John ‘It seems that John is sick.’

j’àzwn. be-sick

fongbe (=(13) in Dumais 1988)

268

The syntactic properties of verbs b. Jan 3ì my (3i) é j’àzwn. 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 #ì 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. Massam’s (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.

Massam’s 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’: genlè, avoir l’air and Gí, 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 pwoblèm. 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 pwoblèm. it / he have a problem #‘There is a problem.’ ok ‘He has a problem.’

b. *Yon a

pwoblèm 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 gen’s argument is extracted out of its clause, the resumptive ki must appear in the verb’s 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 Law’s 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 problème. 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

problème 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. Qu’est-ce qu’il y a? ‘What is there?’

b. *qu’est-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. Xó 3ókpó tìn. story one exist ‘There is a problem.’

b. *(é) 3rd

tìn xó exist story

3ókpó one

fongbe

The Fongbe verb tìn used in existential constructions is not used in possessive constructions. In the latter case, the verb #ó is used, as shown in (46).

9.8 Control verbs (46)

Kwkú 3ó wémà. ‘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 jló ‘to want’ in (49) selects a tensed complement introduced by the form nú, 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)

Éi jló nú éi /j ní yì. 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 bà nú ‘to like’, byV nú ‘to request’, etc., have the same properties as jló nú ‘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 détruire l’ John want destroy det car ‘John wants to destroy the car.’

french

proi ná gbà mvtò c. Kwkúi jló 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 ná, 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

Éi bà èj yì. ‘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 ní (or nú) 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. Éi jló ní / nú éi /j ní yì. 3rd want comp 3rd sub leave ‘Hei wants that hei /j leave.’

fongbe

b. Éi jló èj yì. ‘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 pwòmèt ‘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 pwòmèt Jan proi vini I promise John come [Lit.: ‘I promised John to come.’]

haitian

pou mi vini. b. Mi pwòmèt Jan I promise John comp I come ‘I promised John that I will come.’

haitian

c.

pou Marij vini. Mi pwòmèt Jan I promise John comp Mary come ‘I promised John that Mary will come.’

haitian

In contrast to the Haitian verb pwòmèt, 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. J’i ai promis à Jean I aux promise case John ‘I promised John to come.’

de proi venir. comp come

promis à Jean que jei b. J’i ai I aux promise case John comp I ‘I promised John that I will come.’

viendrai. come-fut

promis à Jean que Mariej c. J’i 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 nú Kwkú proi ná I say comp Koku def-fut [Lit.: ‘I promised Koku to come.’]

wá come

Kwkú 3w ni ná b. Ni 3w nú I say comp Koku comp I def-fut ‘I promised Koku that I will come.’ c.

fongbe

wá. come

Kwkú 3w Àsíbáj ná Ni 3w nú I say comp Koku comp Asiba def-fut ‘I promised Koku that Asiba will come.’

wá. 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 demandé à Mariei de proi ‘They asked Mary to leave.’ c. *Yi byv Marii proi yì [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 l’i 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 demandé à Mariei qu’ellei parte [Lit.: ‘They asked Mary that she leave.’] c.

Yi byv Marii 3w éi ní yì. ‘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 fùn àhwàn

french faire la guerre

haitian fè lagè ‘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 wà àzv ‘do work’

french travailler

haitian travaye

‘to work’

kú tw ‘die river’

se noyer

nwaye

‘to drown’

jy tàgbà ‘fall problem’

s’inquiéter

enkyete

‘to worry’

jy àkpà ‘fall wound’

se blesser

blèse

‘to hurt oneself’

dó gàn ‘plant metal’

enchaîner

anchènnen

‘to chain up’

jì hàn ‘give birth song’

chanter

chante

‘to sing’

sú xò ‘close speech’

crier

kriye

‘to shout’

blá nù ‘tie mouth’

jeûner

jennen

‘to fast’

kàn ‘pick up

courir

kouri

‘to run’

wèzún 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 ànW [Lit.: ‘suck breast’] ‘to suckle’ (see Avolonto 1995), or not, as in kùn hún [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 Massam’s. 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 àzwn ‘calm disease’

french calmer douleur

haitian kalme doulè

‘to heal’ ‘to hunt’

nyà ‘hunt

gbé animals’

chasser

chase bèt

3ùn ‘draw

sín water’

puiser

tire dlo ‘to draw (water)’ ‘draw water’

hún vehicle’

conduire

kòndwi

‘to drive’

b. kùn ‘drive 3ù ‘eat

nú manger something’

manje

‘to eat’

zà ‘sweep

àyí ground’

bale

‘to sweep’

balayer

9.10 Inherent object verbs c. nw ‘suck

ànw breast’

téter

tete

‘to suckle’

3w ‘pee

à3w piss’

pisser

pise

‘to pee’

kpxn ‘cough

àkpxn cough’

tousser

touse

‘to cough’

3ì à3ì ‘believe belief’

croire

kwè

‘to believe’

kwín àkwín ‘whistle whistle’

siffler

sifle

‘to whistle’

3ú ‘eat

gagner

genyen

‘to win’

à3ú 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 mangé. ‘He ate.’

haitian french

b. É 3ú nl. / *É 3ú ‘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 verb’s 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.

v′

v′

b.

v

v

arg.

x eat y

x eat

y

(=(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 tété ‘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 verb’s 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. É ná Mari àsvn. Li bay Mari krab. ‘He gave Mary some crab.’

fongbe haitian

b. Il a donné du crabe à 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. É xò áplò nú 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. É tì àvvkànfù 3ò matela v my. Li boure kotòn nan matela a. ‘He stuffed cotton in the mattress.’

fongbe haitian

b. Il a bourré 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 à. 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. É Li he ‘He

ty voye send sent a

kàn sy-dó telegram bay telegram send / give telegram to Mary.’

b. Il a envoyé un télégramme à 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. É yjwán nú 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 s’est marié avec Marie. ‘John got married to Mary.’

haitian french

b. Kwkú dà 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 à, but in Haitian the complement of this verb is a prepositional phrase headed by ak ‘with’. (73)

a. É 3ì tw tvn. he resemble father his ‘He resembles his father.’

fongbe

b. Il ressemble à son père. ‘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. É klxn àfv. ‘He grazed his foot.’

fongbe

b. Il s’est écorché le pied. ‘He grazed his foot.’

french

c. Li kòche 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 à. In Haitian, the verb is a double-object verb which assigns Accusative Case to the Recipient. These facts are illustrated in (75). (75)

a. É wlàn wémà sy-3ó Sìká. he write letter send Cica ‘He wrote a letter to Cica.’

fongbe

b. Il a écrit une lettre à Marie. ‘He wrote a letter to Mary.’

french

c. Li ekri Mari yon lèt. ‘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 Recipient–Theme construction (np np), whereas John sent a letter to Mary is an example of the Theme–Locative 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, Recipient–Theme and Theme–Recipient, the Haitian np np construction has only one word order: Recipient–Theme (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 Recipient–Theme construction and the np pp construction as the Theme–Locative construction. Furthermore, it has been argued that Recipient–Theme 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 Recipient–Theme construction, the Recipient is the affected argument, while in the Theme–Locative 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 Recipient–Theme 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 Theme–Locative 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 Recipient–Theme and Theme–Locative 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 Recipient–Theme 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 Recipient–Theme construction and the Theme and the Locative in the Theme–Locative 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 Recipient–Theme 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 Recipient–Theme construction, the Recipient must c-command the Theme. Now, consider the facts in (77) involving the Theme–Locative 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 Theme–Locative construction, the Theme must c-command the Locative. These asymmetries, and the others mentioned above, constitute syntactic tests which permit one to distinguish the Recipient–Theme construction from the Theme–Locative 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 Recipient–Theme 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 Pòl 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 Recipient–Theme 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 nèg 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 nèg 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 Recipient–Theme construction, the affected object of the construction is the Recipient. Furthermore, the Recipient–Theme 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 person’s 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 Recipient–Theme 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 Recipient–Theme construction exists in Haitian. It has long been noted in the literature on French syntax that French does not have the Recipient–Theme 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 donné / montré Paul le livre I aux give / show Paul det book [Lit.: ‘I gave/showed Paul the book.’]

french

French only allows a Theme–Locative construction where the Locative is assigned Dative Case by à (see Kayne 1984; Tremblay 1991). This is illustrated in (82). (82)

J’ ai donné / montré le livre à 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 donné une photo de chacun i à sa i mère. 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 donné I aux give chacuni each person

sa / leuri his / their

photo à la photo to the

mère 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 Recipient–Theme 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 Theme–Locative construction. The prepositional phrase can only be interpreted as a Benefactive, not as a Locative. (84)

Mwen bay liv la pou Pòl. 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 Theme–Locative pair of arguments is expressed by a serial verb construction, as is shown in (85). (85)

Mwen pran liv bay Pòl. 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 Recipient–Theme construction (np np) (see (78) ), the corresponding French verbs do not (see (81) ). And while some French verbs participate in the Theme–Locative 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 Recipient–Theme 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 Recipient–Theme 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)

Kwkú ná / xxlx Àsíbá xwé 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.

Súnù 3é ná vj v àsvn 3é 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. *Súnù 3é ná vj 3é àsvn 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 Recipient–Theme 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 Recipent–Theme 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 / ná my 3vkpó3vkpói fvtò I show / give man every picture ‘I showed/gave every man his picture.’

twni . his

fongbe

(=(42a) in Lefebvre 1994c) 3vkpó3vkpói fongbe b. *N xxlx / ná fvtòtv twni fvtò 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 Recipient–Theme 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 Recipient–Theme 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 Theme–Locative construction. First, although the sentence in (89) is grammatical, it is not a Theme–Locative construction, for the prepositional phrase can only be interpreted as a Benefactive. Moreover, Lefebvre (1994b) clearly shows that the prepositional phrase introduced by nú in (89) is a syntactic adjunct rather than an indirect object of the verb.

9.12 Double-object verbs (89)

Kwkú nyà àkwx nú Àsíbá. ‘Koku loaned money for Asiba.’

293

fongbe

Second, the Theme–Locative alternation in Fongbe is rendered in a serial verb construction, as shown in (90). (90)

Kwkú sv àsvn v ná Àsíbá. 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 Theme–Locative 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 Recipient–Theme construction. Second, neither language has verbs which are syntactically realised in a Theme–Locative 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 Theme–Locative 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 Recipient–Theme 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 Recipient–Theme shown for the Fongbe data in (86) is only one of two possible surface orders for the construction. In Fongbe, but not in Haitian, Theme–Recipient is also a possible surface order, as is shown in (91). (91)

Kwkú ná / xxlx àsvn Àsíbá. 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 Theme–Locative 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)

Súnù 3é ná àsvn 3é 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 Theme–Recipient manifests the same asymmetries as the order Recipient–Theme. For example, with the Theme–Recipient 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 / ná fvtò tvni my 3vkpó3vkpói . fongbe I show / give picture his man every ‘I showed/gave every man his picture.’ (=(42b) in Lefebvre 1994c)

fongbe b. *N xxlx / ná fvtò 3vkpó3vkpói fvtòtv 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 Recipient–Theme. How can this difference between the two languages be accounted for? In order to address this question, we need a syntactic account of the Recipient–Theme 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 Recipient–Theme 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)

x′

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 Theme–Recipient (event determiner) as in (92). Movement of the Recipient to Spec of xp, as in (95), produces the derived word order Recipient–Theme (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 Spec–Head 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 Spec–Head agreement (a feature-checking rule) takes place. This approach is independently motivated in Chomsky (1989). On the assumption that structural Case assignment reduces to Spec–Head 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 Recipient–Theme 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 Theme–Recipient 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 Recipient–Theme construction. These verbs are ná ‘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, sú ‘to pay’, which would belong in Gruber’s 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 pwòmèt 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 sèvi 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 Gruber’s 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
‘to ‘to ‘to ‘to ‘to

give, to tell’ lend, to borrow’ give back’ yield, to give, to lend’ show’

ofri sèvi vann paye


‘to ‘to ‘to ‘to

offer’ serve’ sell’ pay’

Verbs of an informational di dimension with an poze animate Goal (e.g. ‘tell’) prèski pwòmèt rakonte swete mande ekri ansènye aprann


‘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’)


‘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 sú ‘to pay’ in (98). For half the informants, sú is a double-object verb, and for the other half, it is not. (98)

ok/*

Kwkú sú xwkwx v Àsíbá / Àsíbá 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)

Kwkú sú xwkwx v nú Àsíbá. 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 sú ‘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 sú 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 ná ‘to give’. (100)

Kwkú sv àsvn v ná Àsíbá. 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)

v′

SpecV′

vp

v

v′

SpecV′

Kwkú

sv

àsvn v

v

Complement

ná

Àsíbá

Assuming the analysis in Lefebvre (1991b), the verb ná 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. Kwkú sv àkwx nyà nú Àsíbá. Koku take money loan to Asiba ‘Koku loaned money to Asiba.’

fongbe (=(110a) in Lefebvre 1994c)

b. Kwkú sv xwénúxò 3w Koku take story tell ‘Koku told a story to Asiba.’

nú Àsíbá. 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 Lefebvre’s (1994c, section 5) analysis, the contrast between (101) and (102) follows from the verb’s 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 sú ‘pay’ (see (98) ). Our proposal predicts two possible realisations for the Goal of sú 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 sú is a double-object verb accept (103a) and reject (103b). Speakers for whom sú is not a double-object verb accept (103b) but reject (103a). (103)

a. ok/* Kwkú sv xwkwx sv-sú Koku take rent take-pay ‘Koku paid the rent to Asiba.’

Àsíbá. Asiba

fongbe

(=(111a) in Lefebvre 1994c) b. */ok

Kwkú sv xwkwx sv-sú Koku take rent take-pay ‘Koku paid the rent to Asiba.’

nú Àsíbá. 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 Recipient–Theme construction. As is the case in Fongbe, many Haitian verbs participate both in the Recipient–Theme 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 Recipient–Theme 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 kòb 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 nyà ‘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 kòb 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 Recipient–Theme construction, a verb must have the property of assigning structural Case to the Recipient. Given that Haitian has a much wider range of Recipient–Theme 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 creole’s 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 djòb ‘job’ (borrowed from English) forming the verb djòb-e ‘to work’. The attributive affix -è is also an affix of this type since it attaches to the noun wanga ‘fetish’ (of West African origin) to derive wanga-t-è ‘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 dé- 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 one’s 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 -è derives langaj-è ‘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 (
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 -è: [v-è]N dechouk-è konsey-è rans-è vant-è

‘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 -è 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 -è: [n-è]N odyans-è ‘joker’ langaj-è ‘chatterbox’ tafya-t-è ‘alcoholic’ wanga-t-è ‘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-rèspèkte de-kale de-mwèle

‘to ‘to ‘to ‘to ‘to ‘to

unwrap’ be impatient’ separate’ insult’ shell’ remove the marrow’

haitian vlope pasyante makone rèspèkte *kale *mwèle

‘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-mwèle, the inversive prefix attaches to a base that is unattested as an actual verb. *Kale and *mwèle 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 mwèl ‘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 *mwèle, 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-wòch ‘pebble’

haitian chat ‘cat’ mounn ‘person’ dlo ‘water’ wòch ‘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 bwòt-ay dechouk-ay kapon-ay fri-t-ay krabin-ay

‘opposition’ ‘moving’ ‘overthrow’ ‘intimidation’ ‘fried things’ ‘crumb’

kontre bwòte 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 pèp 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 pèp 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 pèp 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 (
Nominalising conversion: [v-ø]N ale sòti antre vini rive tonbe desan monte tunen pati pwòche gunmen kuri mache dòmi manti bwòte

‘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 sòti antre vini rive tonbe desan monte tunen pati pwòche gunmen kuri mache dòmi manti bwòte

‘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. bwòte ‘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 ‘John’s leaving’

an b. manti Jan det lie John ‘John’s lie’

c. bwòte Mari a moving Mary det ‘Mary’s moving’

an det

haitian

d. gunmen Jan (ak Wòbè) a haitian fight John (with Robert) det ‘John’s 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. Védrine (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. Bwòte ‘to move’ is one such verb, as shown in (9). (9)

a. Mari bwòte. Mary move ‘Mary moved.’

b. Mari bwòte mèb 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 bwòte ‘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 bwòte Mari a ‘Mary’s moving’. The noun bwòt-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 bwòt-ay mèb 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 *bwòt-ay Mari a, where the Agent is expressed, and of *bwòte mèb 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 Védrine (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-ø]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 Védrine 1992: 167–8)

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 Lieber’s (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 avidèy-man ‘visibly’ alekout-man ‘attentively’ bosal-man ‘wildly’ alèz-man ‘comfortably’ bòzò-man ‘elegantly’

avidèy alekout bosal alèz bòzò

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’ Pétionville’ Saint-Marc’ Sudan’ Chad’

Kap Petyonvil Senmak Soudan Tchad

b. Place of origin/residence suffix -yen: [n-yen]N Jakmèl-yen Tibè-yen Gan-yen Ougand-yen Pakistan-n-yen Ka-yen

‘from ‘from ‘from ‘from ‘from ‘from

Jacmel’ Tibet’ Ghana’ Uganda’ Pakistan’ Les Cailles’

Jakmèl Tibè Gana Ouganda Pakistan Okay

haitian ‘Cape’ ‘Pétionville’ ‘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 Montréal-ais ‘from Montreal’ and not *Montréal-ien but Canad-ien ‘from Canada’ and not *Canadais. Both *Montréal-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 -yèm, and it derives ordinal numbers from cardinal numbers. (14)

Ordinal suffix -yèm san-yèm mil-yèm kat-yèm twa-yèm

‘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 -yèm 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; Védrine 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 (
312

Are derivational affixes relexified?

(15)

The inventory of Haitian affixes a. Agentive suffix base output

haitian -è v n

b. Attributive suffix base output

-è 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

ø v n

h. Adjectival / participial conversion base output

ø 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

-yèm 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 -è 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 -è 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 fêt-er ‘to celebrate’. The Haitian inversive prefix de- is phonologically identical to the French inversive prefix dé- in verbs such as dé-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 [mã]): 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 -yèm derives its phonological representation from the corresponding French suffix -ième. 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 -è v n

french -eur v n

b. Attributive suffix base output

-è n n

-eur n n

(-ard, -ier, -ien)

c. Verbalising suffix base output

-e n v

-er, -é n/a v

(-ir, -ifier, -iser)

d. Inversive prefix base output

dev v

dév/a v

(é-, 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

ø v n

ø v n

(-et, -ot, -on)

(-ion, -ment, -ance, -ure)

314

Are derivational affixes relexified? h. Adjectival / participial ø conversion base v output a

–

(-i, -é, -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

-yèm q a

-ième 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 Furetière 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 Furetière (1984). (17)

haitian ale sòti antre vini rive tonbe desan monte tunen pati pwòche gunmen kuri mache dòmi manti bwòte

‘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: L’ancienne langue employait couramment l’infinitif avec l’acception d’un substantif, mais cet usage devient plus rare au XVIIe siècle . . . à cette époque les substantifs formés d’infinitifs sont sensiblement les mêmes que ceux dont la langue actuelle se sert encore: à côté de ceux qui sont tout à fait courants comme le souvenir, le lever, le coucher, le boire, le manger, etc. quelques-uns qui sont d’un 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 Furetière (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. Furetière’s 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 Furetière 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-respèkte ‘to lack respect/insult’ is made up of the inversive affix de- (=Fr. dé-) and the verb respèkte (
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 -ité, -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, -ée 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, -é, -u h. Adjectival suffixes – adjectival base -âtre, -ot, -ième, -(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 hàn-jí-tv àzw-wà-tv

‘singer’ ‘worker’

fongbe jì-hán (generate-song) ‘sing’ wà-àzw (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-kúkú-nw nù-blá-nw àgy-nw xysì-nw

‘mute’ ‘person who fasts’ ‘criminal’ ‘coward’

fongbe dy-kúkú nù-blá àgy xysì

‘dead tongue’ ‘tied mouth’ ‘crime’ ‘fright’ (=(14) in Brousseau 1990)

The inversive prefix mà- 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 mà-: [mà-v/a]V/A mà-sv ‘not to take’ mà-3ì ‘to doubt’ mà-sísv ‘untakable’ mà-3í3í ‘unbelievable’

fongbe sv 3ì sísv 3í3í

‘to take’ ‘to believe’ ‘takable’ ‘believed, believable’ (=(26), (27) in Brousseau 1990)

The diminutive suffix is -ví. 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 -ví: [n-ví]N àwí-ví kòkló-ví àtín-ví távò-ví

‘kitten’ ‘chick’ ‘bush’ ‘stool’

fongbe àwíì ‘cat’ kòkólò ‘chicken’ àtín ‘tree’ távò ‘table’ (=(9), (10) in Brousseau 1990)

Fongbe has two suffixes of origin: -tV and -nù. 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 -nù 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 Bènxn-tv ‘from Benin’ Bènxn Kùtvnù-tv ‘from Cotonou’ Kùtvnù

319

fongbe ‘Benin’ ‘Cotonou’

fongbe b. Place of origin/residence suffix -nù: [n-nù]N Tìbxy-nù ‘from Tibet’ Tìbxy ‘Tibet’ Glèxwé-nù ‘from Ouidah’ Glèxwé ‘Ouidah’ (=(18) in Brousseau 1990)

The suffix -gVV derives ordinal numbers from cardinal numbers. (25)

Ordinal suffix -gVV wè-gvv àtón-gvv kò-gvv kàn3é-gvv

‘second’ ‘third’ ‘twentieth’ ‘fortieth’

fongbe wè àtón kò kàn3é

‘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. gbìgbá wìwlán dì3à xìxw

‘construction’ ‘writing’ ‘preparation’ ‘buying’

b. xìxw lz ‘the buyings’ xìwlá lz ‘the writings’

fongbe gbá wlán 3à 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 gbìgbá wìwlán xìxw 3ì3à

‘constructed’ ‘written’ ‘bought’ ‘cooked’

fongbe gbá wlán xw 3à

‘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. Làn v 3ò 3ì3à. meat det at cooked ‘The meat is cooked.’

b. Làn v nyì 3ì3à. fongbe meat det be cooked ‘The meat has been cooked.’

c. làn 3ì3à 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 yì ‘go’, tún ‘push’, dYn ‘pull’, nor to verbs of movement such as jY ‘fall’, tVn ‘jump’, or lìn ‘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 #à ‘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 -è v n

fongbe -tV v/n n

b. Attributive suffix base output

-è 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

màv/a v/a

e. Diminutive affix base output

tin n

-ví n n

f. Nominalising suffix base output

-ay v n

copy prefix v n

g. Nominal conversion base output

ø v n

copy prefix v n

h. Adjectival / participial conversion base output

ø 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 / -nù n n

k. Ordinal suffix base output

-yèm 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 -ø 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 -ø, 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-è ‘loud, rowdy’, as we saw earlier, there is a corresponding Fongbe word derived by means of the attributive suffix -nW: zîgí#ì-nW ‘loud, rowdy’. While cases of nominalising conversion (derived by means of the affix -ø) 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 -ø and Fongbe RE-, respectively, with particular bases follows a similar pattern. This is shown in (30). (30)

french – – – – – – – – – – –

haitian tonbe desan tunen pati pwòche gunmen kuri mache dòmi manti bwòte

fongbe jì-jàyí jì-jxté lí-lxkw yì-yì sí-sxwá xì-xòhún kí-kánwèzùn 3ì-3ìzònlìn 3ì-3ìàmlwn 3ì-3ó 3ì-3ìvxwé

‘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 sì-sá wì-wílíbàná

– –

pary-ay kì-kén boukan-t-ay 3ì-3ìv

‘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 mà-. Examples are provided in (32).

10.4 Historical derivation of the Haitian derivational affixes (32)

french – – – –

haitian de-makòn-n-en de-rèspèkte de-take de-z-apiye

fongbe mà-xìzì mà-sín mà-xwè mà-zín

‘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/-ø 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 dé-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 dé- as sharing with mà- the meaning ‘inversive’ and he would have assigned the lexical entry copied from mà- the phonological form dé-. Similarly, the lexical entry copied from the agentive affix -tV would have been assigned the label -è 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 fè (
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/-ø 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 -ø. Second, whereas Fongbe RE- is also involved in adjectival conversion, as is Haitian -ø, -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, αv]. 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 -ø 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/-ø 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’


b. o-gblo

‘breadth’


c. =-da

‘creature’


‘create’

d. i-bi

‘wickedness’


‘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 -ø 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, αv]

[+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 [αv]. For the sake of the following argument, I will assume that RE- is defined by the syntactic features [+n, αv].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, αv]

b.

Derived adjectives

fongbe

[+n, +v] RE[+n, αv]

[+v, −n]

Brousseau (1993) argues that, in both cases, RE- is the morphological head of the derived word. Its categorial feature [+n] percolates à 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 [αv] in the derived words ( [−v] in derived nouns and [+v] in derived adjectives) in saying that [αv]’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, αv]. As will be argued below, the Haitian null affix must be a suffix. (36)

[+v, −n]

/-ø/

haitian

[+n, αv]

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 [αv] is assigned a positive or a negative value. This is shown in (37). (37)

a. Nominal conversion [+n, −v] [+v, −n]

-ø [+n, αv]

b. Adjectival conversion

haitian

[+n, +v] [+v, −n]

-ø [+n, αv]

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, αv]

target language

early haitian

——

[+v, −n]

dialectA

creole lexical entry [+v, −n]

/-ø/ [+n, αv]

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, αv], 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]

/-ø/ [+n, +v]

The early Haitian dialect of these speakers thus contained the two lexical entries -ay and -ø, 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

-ø [+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 [αv], 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 -ø in forming nouns from verbs. Apparently, -ay came to be used only with two-place predicates and -ø 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 Aronoff’s (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 word’s 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-è n-è v-e de-v ti-n v-ay a-man n-wa / -yen q-yèm

french v-eur n-eur n/a-er / -é dé-v/a n/a-et / -ot / -on v-age a-ment n-ois / -ien q-ième

fongbe v/n-tV n/v-nV – mà-v/a n-vJ RE-v – n-tV / -nù 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 -ø, 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 -ø 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 Furetière’s (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 fléch-er ‘to throw arrows’ > flèche ‘arrow’, flût-er ‘to play the flute’ > flûte ‘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 -ø/-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 -ø/-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. Mühlhäusler 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 lèvre ‘lip’

twou-ne

(trou-nez) ‘hole-nose’

narine

‘nostril’

plim-je

(plume-œil) ‘hair-eye’

cil

‘eyelash’

dèyè-kou

(derrière-cou) ‘back-neck’

nuque

‘nape’

kalbas-tèt2 or tèt-kalbas

(calebasse-tête) ‘calabash-head’

crâne

‘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 dèyè-kou kalbas-tèt or tèt-kalbas

‘skin-mouth’ ‘hole-nose’ ‘hair-eye’ ‘back-neck’ ‘calabash-head’

fongbe nù-fló àwntín-dó wùn-3à kw-gúdó tà-ká

‘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 Chaudenson’s (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 tête 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 tèt-chòv refers to a person who is bald, as shown in (3). (3)

Papa Ti Klod se yon tèt – chòv. 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 tèt-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.’

tèt-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. tèt-chòv (tête-chauve) ‘head-bald’

french chauve

b. tèt-chaje (tête-troublé) ‘head-troubled’

personneproblème

c. je-pete (œil-pété) ‘eye-burst’

aveugle

d. je-chèch (œil-sec) ‘eye-dry’

audacieux

e. je-fò (œil-fort) ‘eye-strong’

prétentieux nùkún-kyn

4

f. tèt-di (tête-dure) ‘head-hard’

fongbe tà-súnsún

‘bald (person)’

‘head-bald’ tà-gbà

‘problematic (person)’

‘head-troubled’ nùkún-tvn-nw3

‘blind (person)’

‘eye-burst-att’ hhn-wùn

‘audacious (person)’

‘clear-eye’ ‘pretentious (person)’

‘eye-strong’ entêté or tête-dure

tà-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. tèt-di (tête-dure) ‘head-hard’

french entêtement

b. tèt-fè-mal (tête-faire-mal) ‘head-hurt’

mal de tête

5

c. mare-tèt (amarrer-tête) ‘attach-head’

fongbe tà-my-tìtì

‘stubbornness’

‘head-in-stuffing’ tà-3ùmy

‘headache’

‘head-hurt’ bandeau or serre-tête

tà-blá

‘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 tèt-fè-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 kwékwé-tín dè-tín

‘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. chêne ‘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 tín ‘tree, wood, stick’ (see Rassinoux 1987), respectively. This is shown in (9). (9)

a. bwa-balè bwa-linèt bwa-pip

(bois-balais) ‘stick-broom’ (bois-lunettes) ‘stick-glasses’ (bois-pipe) ‘stick-pipe’

b. wèmá-tín ‘paper-stick’ wò-tín ‘dough-stick’ wyn-gán-tín ‘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 tín 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 à 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 chwèt-bwa melon-dlo coulèv-dlo poul-dlo soulye-lasèt

french N-n oiseau-mouche chouette de bois melon d’eau serpent d’eau poule d’eau soulier à lacets

b. haitian N-n pwason-zèl aran-sèl fanm-saj

french N-a/a-N poisson ailé (poisson volant) hareng salé 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 nù-fló (‘mouth-skin’). Similarly, in Haitian compounds involving a noun and an adjective, in some cases, the adjective follows the noun, as in tèt-chòv (‘head-bald’) ‘bald’ (see (5) ), and in other cases, it precedes the noun, as in gwo-nèg (‘bigman’) ‘rich, giant’. The latter word order is not permitted in Fongbe, which requires the order noun–adjective in compounds, as in tà-súnsún (‘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-tèt (‘attach-head’) ‘headband’, the verb precedes the noun (see (9) ). By contrast, in Fongbe, the noun precedes the verb as in tà-blá (‘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 noun–noun 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. noun–noun Compound po-bouch skin mouth ‘lip’

b. noun noun in np rob Mari a dress Mary det ‘Mary’s dress’

haitian

In contrast to Haitian, in Fongbe noun–noun 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. noun–noun Compound nù-fló mouth skin ‘lip’

b. noun noun in np Àsíbá sín àvv 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-nèg 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 noun–adjective compounds and adjectives following the noun in noun phrases, as illustrated in (14). (14)

a. noun–adj. 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. noun–adj. Compound àgbàn-kléún baggage small ‘package’

b. noun adj. in np xó kléún word small ‘short word’

fongbe

(from Rassinoux 1987)

Finally, the word order observed in the Haitian verb–noun compounds replicates the verb–complement order found in verb phrases, as is shown in (16). (16)

a. verb–noun Compound mare-tèt attach head ‘headband’

b. verb noun in vp mare li sere tie it tight ‘tie it tight’

haitian

(from Védrine 1992)

Likewise, the word order in the Fongbe noun–verb compounds replicates the complement–verb order found in nominalised verb phrases in this language. (17)

a. noun–verb Compound tà-blà head attach ‘headband’

b. noun verb in nominalised vp fongbe Kwkú 3ò élv blà wy Koku at this attach part ‘Koku is attaching this.’

The three sets of facts discussed above argue in favour of Brousseau’s (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 adjective–noun 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 noun–adjective 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)

n–n compounds pye-bwa (plant-wood) bwat-myèl (box-bee) zo-figi (bone-face) nèg-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

n–n a–n

n–n n–a

n–n a–n n–p–n n–p–v n–a

primary headless compounds

a–n n–a

n–a

synthetic headed compounds

a–n n–a v–n n de n n–n

n–v–n n–vv n–v–tV

According to Brousseau (1989, 1994a), the vast majority of Haitian compounds are of the n–n type. Examples of right-headed compounds made up of an adjective and a noun are provided in (21). (21)

a–n 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. a–n gwo-je gran-gozye

(big-eye) (big-throat)

‘greedy’ ‘pelican’

b. n–a je-fò je-pete

(eye-strong) (eye-burst)

‘pretentious’ ‘blind person’ (=(18) in Brousseau 1994a)

Based on Brousseau’s 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. n–n télévision-couleur pause-café fermeture-éclair assurance-vie

‘colour television’ ‘coffee break’ ‘zipper’ ‘life insurance’

b. n–p–n livre d’images cœur de pomme aide de camp poupée de cire canne à sucre épingle à linge

‘picture book’ ‘apple core’ ‘aide de camp’ ‘wax doll’ ‘sugar cane’ ‘clothes pin’

c. n–p–v poêle à frire machine à laver chambre à coucher métier à tisser

‘frying pan’ ‘washing machine’ ‘bedroom’ ‘loom’

d. n a pouvoir étudiant 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)

a–n 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 v–n compounds in (25a) are extremely productive in French. (25)

Headless compounds

french

a. v–n pèse-personne abat-jour coupe-papier grille-pain

‘scale’ ‘lampshade’ ‘paper cutter’ ‘toaster’

b. n de n pied de biche œil de bœuf

‘nail-claw, crowbar, etc.’ ‘bull’s eye window’

c. a–n rouge-gorge blanc-bec

‘robin’ ‘greenhorn’

11.3 Types of compounds d. n–a blouson-noir cordon-bleu sang-mêlé

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)

n–n auteur-compositeur moissonneuse-batteuse fille-mère professeur-chercheur

french ‘author-composer’ ‘combine-harvester’ ‘unmarried mother’ ‘professor-researcher’ (from Brousseau 1988a)

According to Brousseau’s 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. noun–noun) compounds, as shown in (27). Brousseau (1989) points out that this type of compound is very productive in Fongbe. (27)

n–n compounds tò-gán (country-chief) glé-gán (field-chief) àkwégbá-gán (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 n–a compounds àzwn-vw (sickness-red) tv-gbó (father-big) gàn-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)

n–a 3à-vwvw xòmè-hùnhùn wù-kúkú

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 n–v–n compounds (see (31a) ), n–vv (see (31b) ), and n–v–tV (see (31c) ) compounds.7 (31)

a. sìn-hú-nú nù-tv-nú

(water-spread-thing) ‘watering can’ (thing-sew-thing) ‘needle’

fongbe

b. àgbà-dídà vì-jíjí

(bomb-sending) (child-giving-birth)

‘bombing’ ‘delivery’

fongbe

c. glè-ly-tv nù-sá-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 n–v–tV (see (31c) ) are rendered in Haitian by nouns derived with the suffix -è, as shown in (32b). (32)

fongbe a. sùn-kínkxn àyí-jíjx

haitian pari-ay tonbe-ø

‘bet’ ‘fall’

b. hàn-jítv xò-3vtv

chant-è pal-è

‘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, doki–fisi–ma (dive–fish–man) ‘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. à 3ù *(nú) you eat thing ‘You ate.’

b. *(nù)-3u-xwsá 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 jì hàn dv àmlw zùn zwnlìn

(produce-song) (do-sleep) (walk-walk)

haitian chante ‘sing’ dòmi ‘sleep’ mache ‘walk’ (=(28) in Brousseau 1994a)

In Brousseau’s 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 n–vv 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 n–v–tV 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 -è. 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 α to account for the movement of constituents, what are the values for α 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 regrèt (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 Pollock’s (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)

t′

np

NegP

t

AgrP

Neg

vp

Agr

v′

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). Pollock’s 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.

Boukinèt 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

Kwkú mà nyíwàn nú Àsíbá. Koku not like to Asiba ‘Koku does not like Asiba.’

b. *Kwkú

nyíwàni mà

ti nú

Àsíbá

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. *nyíwàni Kwkú ti Àsíbá

haitian fongbe

Fourth, the verb follows temporal adverbs in both languages. This is illustrated in (9) with the adverb meaning ‘already’. (9)

a.

Li É s/he ‘S/he

deja pase rad yo. kò lì àvv lx. already iron cloth pl already ironed the clothes.’

b. *Li pasei deja *É lì kò

ti rad yo ti àvv 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 Givón 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. Kwkú 3ú nl gàn5í. 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 déjà ‘already’ also precedes the vp. This can be seen in sentences such as Marie a déjà repassé les vêtements ‘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 mangé ‘Mary ate well’. While the position of Haitian deja also coincides with the position of Fongbe kò ‘already’, the position of byen does not coincide with that of Fongbe gàn%í ‘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 DeGraff’s (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 infl 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)

É Li 3rd ‘He

sv àsvn yì àxì 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 marché

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 infl (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 infl and therefore vp can function as a secondary predicate’.5 On the basis of data drawn from Yoruba, Déchaine (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. Déchaine 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 infl 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 infl (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 infl 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 Theme–Goal 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 Kayne’s 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 Kayne’s analysis, French lacks double-object constructions (e.g. *Jean a envoyé Marie une lettre) because it lacks preposition stranding (e.g. *Qui Jean a envoyé une lettre à?). While the French verb envoyer ‘to send’ assigns Objective Case to its direct object, the preposition à ‘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 nú which translates as ‘to/for’ has been stranded and the resulting sentence is not grammatical.7 (14)

a.

Kwkú sà àsvn nú Àsíbá. Koku sell crab to / for Asiba ‘Koku sold crab to/for Asiba.’

fongbe (=(5a) in Lefebvre 1994c)

àsvn nú ti fongbe b. *M6 / Àsíbái wy Kwkú sà 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). Kayne’s 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. John’s 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 à/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 Tremblay’s 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. Mary’s book

english

The motivation for Johnson’s 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 Johnson’s (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, pèsonn 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 n’est pas venu. french nobody ne+is pas come ‘Everybody came.’ [Lit.: ‘Nobody has not come.’] (=(3a) in DeGraff 1993a) b. Ce n’est pas rien. 3sg ne+is pas nothing ‘This is something.’ [Lit.: ‘This is not nothing.’]

french (=(3c) in DeGraff 1993a)

(18)

a. Pèsonn 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 wè pèsonn / anyen. I not ant see no one / nothing ‘I did not see anyone/anything.’

haitian (=(1a) in Deprez in press)

b. Pèsonn *(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 n’ai vu personne. ‘I did not see anyone.’

b. Personne n’est arrivé. ‘No one arrived.’

361

french

In popular French, J’ai 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 arrivé, 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 pèsonn ‘no one’

d –

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 mYtí ‘nobody’ and nLtí ‘nothing’ corresponding to Haitian pèsonn 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.′

Mytí no one

wá f. arrive Neg

a.″ Mytí mà wá nx. no one neg arrive part ‘No one arrived.’

fongbe fongbe

362

Parameters b.′ N’ mw I see

nltí f. nothing Neg

fongbe

b.″ N mà mw nltí 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 mytí / nltí [Lit.: ‘I see no one/nothing.’]

b. *Mytí wá [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 Deprez’s 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)

Kwkú gb6 nltí 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 refusé 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 Deprez’s 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 pèsonn 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 Wá Jan wá (tróló) bw Màrí yì. 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 Wá Jan wá útú Rive Jan rive arrive John arrive cause ‘Because John arrived, Mary

Màrí yì. Mari pati. Mary leave left.’

fongbe haitian (=(2) in Lefebvre 1994b)

(29)

Factive Wá 3iè Jan Rive ø Jan arrive op John twn. li kòntan. his happy ‘The fact that John

wá v, víví nú rive a, fè arrive det make(-happy) for

nw manman mother

fongbe haitian

arrived made his mother happy.’ (=(3) in Lefebvre 1994b)

364 (30)

Parameters Predicate cleft Wá wy Jan wá. 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. Rírà (ìwé) tí Aje ra ìwé. nom buy paper rel Aje buy paper ‘Once Aje bought books.’ ‘The fact that Aje bought books.’

yoruba

(=(47b) in Manfredi 1993) b. Rírà ìwé ni Aje ra nom buy paper comp Aje buy ‘It is book buying that Aje is doing.’

ìwé. 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. Bá0gbó-é 1972; Déchaine 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)

Távòi 3ièi à øi 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 (#Iè in Fongbe and ø 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 #Iè in Fongbe and ø 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 Déchaine 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)

Xò wy Sìká lìn 3w Kòfí 3w Àsíbá xò Kwkú. 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 pòte Mari di m l ap pòte. 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 event’s occurrence in time and/or space. Only Stage-Level predicates may be characterised in terms of their temporal–spatial 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 à lì sàká. eat we ate rice ‘We ate rice.’

vata (=(52a) in Koopman 1984: 38)

b. Zal= x sàly dùùù. 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 vòle Bouki vòle 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 dòmi Jan dòmi (pandan inèdtan). 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

entèlijan Jan intelligent John

entèlijan intelligent

haitian (=(14) in Lefebvre 1990b)

b. *Se it-is

konnèn Jan konnèn 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. Gbà wy súnù 3é gbà mvtò 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. 3Dù wy Kwkú 3ù àsvn 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. Fìn wy súnù 3é fìn mvtò 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)

*Tùn wy súnù 3é tùn Àsíbá 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. Xò wy Àsíbá xò Kwkú. 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. Mvtò v wy súnù 3é gbà. 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 Wá v Jan wá (tróló) Rive a Jan rive arrive det John arrive as-soon-as ‘As soon as John arrived (as we knew he

bw Màrí yì. fongbe1 (epi) Mari pati. haitian1 and Mary leave would), Mary left.’ (=(19) in Lefebvre 1994b)

370 (44)

Parameters Causal adverbial Wá v Jan wá útú Rive a Jan rive arrive det John arrive cause ‘Because John arrived (as we knew

Màrí yì. fongbe1 Mari pati. haitian1 Mary leave he would), Mary left.’ (=(20) in Lefebvre 1994b)

(45)

Factive Wá v 3iè Jan wá v, víví nú fongbe1 Rive a ø Jan rive a, fè 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 Yì v wy Se ale a it-is leave det it-is ‘It is leave (as expected)

Jan yì. 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 ø Jan rive a, fè 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 α, where α=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 α 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.). Oweré (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 Déchaine (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)

Fà ji e-ji. 3pl hold holding ‘They held [something/thing] as expected.’

oweré-igbo (Déchaine 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 Oweré) 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

B

C

X X X

D

E

F

G

H

X X X

X X

Creole lexical entries • the so-called indefinite determiner

X

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

B

C

X

X

D

E

F

G

H

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

H

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 • yes–no question marker • negative marker • two markers of insistence Creole lexical entries • se • èske

B

C

D

E

F

G

H

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 yes–no 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 yes–no question marker is Haitian èske. 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 yes–no 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 yes–no question form èske (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 mà• diminutive -vJ • place of origin / residence -tV / -nù • ordinal -gVV • nominalising affix (not Fongbe) • copy RECreole lexical entries • verbalising -e • adverbial -man

B

X X X X X X X X

C

D

E

F

G

H

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

B

C

D

E

F

G

H

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 Segurola’s (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 Segurola’s (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 -ø (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 -ø 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 Mous’s (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 Mous’s 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 one’s 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 infl (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 Koopman’s (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

bí

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 yes–no question word èske (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 èske, for it is possible to construct an analysis whereby èske, 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 èske 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 creole’s 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 Dutch’s 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 infl 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 Mufwene’s (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 Singler’s 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 Givón (1971), Baker (1991) has proposed that the difference between the two language families is that, in the former, the verb moves up to infl (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 one’s 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 Sproat’s (1985) and Pranka’s (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 “parenté syntaxique” advanced by Sylvain (see pp. 121–2), 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 (1776–1936)

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: 56–7 Hillard d’Auberteuil 1776–7.2: 68 Valdman 1978 Wimpffen 1817: 186 Debien 1962: 138 Gros 1791: 7 Anon. c. 1792: 6–15 Anon. 1796 Moreau de Saint-Méry 1797.1: 37, 65 Descourtilz 1809.2: 129, 303, 352; 3: 113–353 Ducœurjoly 1802.2: 283–393 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: 273–4)

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; Férère 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 ã

Back q p

According to the analysis in Brousseau (1994b), building on Ans (1968), Tinelli (1970), Valdman (1978), Cadely (1988a, 1988b), Anestin (1987), Férère (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 <γ> m n () [ŋ] l w y – [g] <γ> (=(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/ /ã/ /f/ /p/

ou è ò in èn an oun òn

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/ () [ŋ] <γ>

tch dj ch j ny g r

[zepsŋ] zepèng

‘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), Fouché (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 = æ

Front Back rounded ü [Ü] u [U] ö o 1 œ u 7

b. Nasal Front

s h

Front rounded

Back

r

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; Fouché 1952; Gendron 1966; Joly 1995; Laborderie 1994; Pierret 1983; Van Daele 1929; Thurot 1881–3; 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 [ʁ] 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,

Alveo- Palatal Velar palatal k [ky ] g [gy ] [g] [5] m n  (λ) 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] [γepwum] ü [alim=t] [mi] [nis] ü [nus] [suse] [bule] ö [dife] [ve] ö [agyo] œ [av=g] [p=p] [vul=] [5ul] r [brs]

<

1

e < i

<

u < e < o < = <

s <

< < < < < < < < < < < < < < < < < <

french depuis cimetière cerise reproche allumette mur juste juste sucer brûler feu vœu adieu aveugle peuple voleur gueule brun

401

[d(1)p2i] ‘since’ [sim(1)ty=r] ‘cemetery’ [s(1)riz] ‘cherry’ [r(1)prum] ‘reproach’ [alüm=t] ‘match’ [mür] ‘wall’ [jüs(t) ] ‘exactly’ [jüs(t) ] ‘exactly’ [süse] ‘to suck’ [brüle] ‘to burn’ [fö] ‘fire’ [vö] ‘vow’ [adyö] ‘farewell’ [avœgl(1) ] ‘blind’ [pœpl(1) ] ‘people’ [vulœr] ‘thief’ [gœl(1) ] ‘mouth’ [brr] ‘brown’ (=(4) and (5) in Brousseau 1994b)

Correspondences between Haitian and French consonants are as in (vi). (vi)

γ < ø < w < /g< γ

r r 2 


haitian [gγã] [mame] [wit] [kãpã] [kãpãg] [γele]

< < < <

french grand marché huit campagne

[grã] [marme] [2it] [kãpa]

‘big’ ‘market’ ‘eight’ ‘countryside’

<

héler

[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 õ s p ã (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

x γ

xw γw

y [g]

The orthographic convention used in this book follows the official usage in Benin (see Publication of the Centre national de linguistique appliquée (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, [àsb] ‘crab’ is spelled àsVn. By contrast, [mc] is spelled mW ‘to see’. The consonants are written as in (viii) with the following exceptions: /g/ /5/ // /γ/

= = = =

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 low–high or high–low 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: Furetière (1984) and Dubois et al. (1992). The data in (A) exemplify Haitian words derived by means of the agentive affix -è. 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 -è haitian beny-è ‘undertaker’


17th-century french baign-eur ‘swimmer / one who owns bath house / spa’

modern french

djòl-è ‘talker’


–

gueul- eur ‘loud talker’

estat-è ‘starter’

<estat ‘to start’

–

kòmand-è ‘boss’


command-eur ‘commander (of order of knighthood)’

The data in (B) are Haitian words derived by means of the attributive affix -è. French has no counterpart for any of these derived Haitian words. (B)

Words derived by means of the attributive suffix -è haitian bresony-è ‘tippler, lush (drinker)’


17th-century french –

eskandal-è ‘loud, rowdy’

<eskandal ‘scandal, uproar’

–

odyans-è ‘entertaining person’


–

403

404

Appendix 3 rans-è ‘joker, frivolous person’


–

soud-è ‘deaf person’

<soud ‘deaf’

–

tafya-t-è ‘drunkard’


–

tanbouy-è ‘drummer’


–

wanga-t-è ‘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)’


betiz-e ‘to joke, to work in vain, to deceive’


–

–

bourik-e
–

–

bourad-e ‘to shove’


–

–

brak-e ‘to sweeten lightly’


braquer ‘to position a cannon’

braquer ‘to position a cannon’

chyen-t-e ‘to beg, to fawn’


–

–

driv-e ‘to drift, to insult’


–

–

graj-e ‘to grate, grated’


–

–

kalkil-e ‘to reflect’


calculer ‘to count’

calculer ‘to count’

kle-t-e ‘to lock’


–

–

pwòp-t-e ‘to clean’


–

–

Non-matching derived words in Haitian and French rans-e ‘to joke, to dally, to toy, to trifle’


–

–

sabo-t-e ‘to slap’

<sabo ‘slap’

–

saboter ‘to clatter’

salòp-e ‘to dirty’

<salòp ‘slovenly’

–

saloper ‘to botch, to bungle’

tchans-e ‘to speak idly, to delude oneself’


–

–

tèk-e ‘to hit a marble’


–

–

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 dé-gwosi ‘to lose weight, to smooth out, to rough-hew’


17th-century french dégrossir ‘to rough down, to rough-hew, to rough-plane’

de-makonnen ‘to untangle, to disentangle’

<makonnen ‘to tangle’

–

de-respekte ‘to be disrespectful’


–

de-take ‘to unlatch’


–

de-z-apiye ‘to stop leaning’


–

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 –

tibèt ‘insect, bug’

–

ti bourik ‘rude person’

–

406

Appendix 3 ti devan ‘sex, genitals’

–

tifi ‘virgin’

–

tigasòn ‘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)’


17th-century french –

randui-z-ay ‘coating’


–

boukan-t-ay ‘swap’


–

pary-ay ‘bet (wager)’

<pary-e ‘to bet, to wager’

–

plas-ay ‘common-law marriage’


–

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 sòti antre vini rive tonbe desan monte tunen pati pwòche

‘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 dòmi manti bwòte

‘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’


mal-man ‘not so well, rather badly’

<mal ‘badly’

17th-century french – –

Notes

1

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 Whinnom’s 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 après (‘after’) is part of a periphrastic expression used to encode the progressive. Était 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 (1985–9 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 [αn, βv] (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 16–53 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. Mühlhäusler (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/dã 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 Mühlhäusler, 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 Kayne’s 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 1985–7 and 1987–9 Projects with the help of Caroline Fick, a historian at the Centre de Recherches Caraïbes (Université de Montréal), 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 1989–94 Project. For a detailed report on his findings, see Singler (1993b, 1996).

410

Notes to pages 53–69

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 engagés 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 Singler’s (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 Koopman’s (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 1985–7, 1987–9 Projects; Lefebvre 1989–94 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 1985–7 Project. The historical research was conducted during the third year of the 1989–94 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 70–93

411

14 Thanks to Ans de Kok for sharing with me her bibliography on seventeenth- and eighteenth-century French. 15 Alembert et al. (1776–7); Bary (1665); Bonnard (1982); Brunot (1905, 1926); Catach (1995); Darmesteter (1875); Dees (1971); Dubois et al. (1992); Féraud (1768); Fouché (1952); Furetière (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 (1881–3); 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: 248–51). 18 Ans (1968); Faine (1937); Freeman (1988); Goodman (1964); Hall (1950, 1953); Hilaire (1992, 1993); Sylvain (1936); Valdman (1970); Valdman et al. (1981); Védrine (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 ça. Chaudenson (1993: 19) also points out that, in most varieties of French, ça and cela are equivalent in meaning. 8 Auger (1994) points out that there are two ças in French: the strong form discussed in this section and a clitic form, not discussed here. In Ça, ça me plaît ‘This, it pleases me’, the first ça is the strong form and the second one the clitic.

412

Notes to pages 96–112

9 In Montreal colloquial French – considered to be a conservative dialect as compared with international modern French – the demonstrative forms with là 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 é/V and énX 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 à 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 Hazoumê’s (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 à 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 112–121

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); Déchaine (1991); Lumsden (in press b). 4 Note that Damoiseau’s (1988) classification of verbs in terms of their aspectual properties differs from other classifications such as the one in Vendler (1967). Dynamic verbs in Damoiseau’s 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 kò is homophonous with an adverb meaning ‘already’ (see Anonymous 1983; Avolonto 1992). As is pointed out in Lefebvre (1996b), the claim that kò has two functions is supported by two sets of facts. First, there can be two occurrences of kò within the same clause. In (i), the first occurrence of kò corresponds to its function as a marker of anteriority, and the second to its function as an adverb. (i)

Sìká kò nw kò 3à wh. Cica ant use-to already prepare dough ‘Cica had usually already prepared dough.’

fongbe

(=(59) in Lefebvre 1996b) Furthermore, as an adverb, kò 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 kò, see Lefebvre (1996b). The function of kò as a marker of anteriority must have been acquired through reanalysis. Westerman (1907: 139) reports that kò 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 été is té. 8 It appears that the creators of Haitian did not select the locative preposition à (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 après have that à 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 122–161

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 ná kò 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 française, the Grand Robert, Littré, Quillet, Logos, Grand dictionnaire de la langue française. Trésor de la langue française, Furetière. 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, eux’s 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 è, 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 164–185

415

8 The idea that -#éè 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 s’est vu lui-même ‘He saw himself (emphatic).’ In Fongbe, the emphatic form #ésu corresponds to mèm in Haitian and même 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 achetées? ‘What things did you buy?’ The complex phrase qu’est-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 Molière 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 Furetière (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, À xW é-t´ε ‘You bought what?’ is more acceptable than ??À xW àní 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

4

5 6

7 8 9 10 11

12

13 14 15

16 17 8

Notes to pages 188–219

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 Koopman’s 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: bó. 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 nyì 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 Lefebvre’s (to appear b) sample who exhibits no clausal determiner.

Notes to pages 222–261

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 èske appears in the specifier position of the projection headed by the yes–no question marker à in Fongbe. 7 According to Segurola (1963: 418), nùgbó 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 lá. 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 tèt-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, tèt-li renforce le rôle agentif, soit en précisant le caractère volitionnel de l’action, en précisant l’autonomie de l’action de l’Agent en opposition à toute autre cause extérieure ou en attirant l’attention sur le caractère exceptionnel ou non attendu d’une telle action.’ [In fact, tèt-li reinforces the agentive role, by specifying the volitional nature of the action, emphasising the autonomy of the Agent’s 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 tèt-li fòk li wè 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 261–283

5 Note that the French expression corresponding to Haitian genlè selects ça rather than il as an expletive subject (see (ib) ). The Haitian verb genlè admits neither li nor sa, as shown in (ia). (i)

a.

Te *Li te *Sa te it ant ‘It seemed

genlè Jan renmen genlè Jan renmen genlè Jan renmen seem John love that John loved Mary.’

b. *Il a l’air que Jean Ça a l’air 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 ná in this context, see Kinyalolo (1992). 16 Verbs such as bà ‘to want’ and byW ‘to ask’ also allow for an infinitival complement with an overt subject, as shown below. (i)

Éi bà èj yì. ‘He wants him to leave.’

(ii) Éi byw èj yì. ‘He asked him to leave.’

fongbe

17 What we do find in Haitian, however, is the type of construction in (i). (i)

Jan dòmi yon ti dòmi. John sleep a little sleep ‘John slept a little.’

haitian

Notes to pages 286–301

18 19

20 21

419

While the phrase yon ti dòmi 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 Recipient–Theme and Theme–Recipient 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 Recipient–Theme 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 Huguet’s (1925) Dictionnaire de la langue française du seizième siècle (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 à 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 kòb paye pou John take money pay for ‘John paid on behalf of Mary.’

Mari. Mary

haitian

420 10

Notes to pages 304–327 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 Védrine’s (1992) dictionaries both list kale with the meaning ‘to shell’. Neither dictionary lists mwèle 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 bétail . . . 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: *frè-r-ay *sè-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 Trésor de la langue française 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)

àzé-tv gbé-tv

‘sorcerer’ ‘hunter’

àzé ‘magic’ gbé ‘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 [αv]. In claiming that RE- is identified for the features [+n, αv], 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 329–346

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 Brousseau’s 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-tèt ‘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 tèt-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: n–vv–tV and n–v–p–n–tV. These types of compounds were not incorporated into Brousseau’s 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 347–365

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 copy’s 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 366–393

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 nyì 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 négro-aryen et maléo-aryen, Paris. Adjemian, Christian 1983, ‘The transferability of lexical properties’, in Gass and Selinker (eds.), pp. 250–68. Àgbèdvr, Paul Kòfí 1995, ‘Negation in Èvè’, 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 1776–7, Encyclopédie ou dictionnaire raisonné des sciences, des arts et des métiers (Microfiches, 105 × 148 mm, microcopy of Paris edition, 1776–7), Paris: France-expansion (1973). Allen, Margaret 1978, ‘Morphological investigations’, Doctoral Dissertation, University of Connecticut, Storrs. Alleyne, Mervyn C. 1966, ‘La nature du changement phonétique à la lumière du créole français d’Haïti’, Revue de linguistique romane 30: 279–303. 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. 89–107. Andersen, Roger W. 1980, ‘Creolisation as the acquisition of a second language as a first language’, in Valdman and Highfield (eds.), pp. 273–95. 1983a, ‘A language acquisition interpretation of pidginisation and creolisation’, in Andersen (ed.), pp. 1–59. Andersen, Roger W. (ed.) 1983b, Pidginisation and Creolisation as Language Acquisition, Rowley, MA: Newbury House. Anestin, Agnès-Lydie 1987, ‘Structure syllabique de l’haïtien et nasalisation’, Master’s Thesis, Université du Québec à Montréal. Anonymous c. 1792, Grand débat entre Duffay et consorts, Polverel et Sonthonax les égorgeurs et les brûleurs 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. 101–2.

424

References

425

Anonymous 1810, Cri des colons contre un ouvrage de M. l’evêque et senateur Grégoire, ayant pour titre de la littérature des nègres . . . , Paris. Anonymous 1821 (1811), Idylles, ou essais de poësie creole, par un colon de St.-Domingue, Cahors. Anonymous 1824, ‘Parabole de l’Enfant Prodigue en créole de Saint-Domingue, Port-auPrince’, in Anonymous (1831), p. 93. Anonymous 1831, Mélanges sur les langues, dialectes et patois . . . , Paris. Anonymous 1983, Éléments de recherche sur la langue Fon, Cotonou. Ans, André-Marcel d’ 1968, Le créole français d’Haïti, 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: 141–72. Arends, Jacques 1995a, ‘Demographic factors in the formation of Sranan’, in Arends (ed.), pp. 233–77. 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 Bénin 1983, Abidjan: Commission nationale linguistique du Bénin, Université d’Abidjan. Audain, M. J. J. 1877, Recueil de proverbes créoles, Port-au-Prince. Auger, Julie 1994, ‘Pronominal clitics in Quebec colloquial French: A morphological analysis’, Doctoral Dissertation, University of Pennsylvania, Philadelphia. Avolonto, Aimé 1992, ‘De l’étude sémantico-syntaxique des marqueurs préverbaux à la structure de la phrase en fwngbè’, Master’s Thesis, Université du Québec à Montréal. 1995, ‘Pour une approche minimaliste des verbes à objets inhérents’, Doctoral Dissertation, Université du Québec à Montréal. 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: 89–129. 1990, ‘Off target?’, Journal of Pidgin and Creole Languages 5: 107–19. 1993, ‘Assessing the African contribution to French-based creoles’, in Mufwene (ed.), pp. 123–55.

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. 17–40. 1989, ‘Relexification in Canada: The case of Métif (French-Cree)’, in Lefebvre and Lumsden (eds.) (1989b), pp. 339–50. 1992, ‘A language of our own: The genesis of Mitchif, the mixed Cree–French language of the Canadian Metis’, Doctoral Dissertation, University of Amsterdam. 1994, ‘Michif, the Cree–French mixed language of the Métis buffalo hunters in Canada’, in Bakker and Mous (eds.), pp. 13–33. Bakker, Peter and Mous, Maarten 1994a, ‘Introduction’, in Bakker and Mous (eds.), pp. 1–11. Bakker, Peter and Mous, Maarten (eds.) 1994b, Mixed Languages (Studies in Language and Language Use 13), Dordrecht: ICG Printing. Bá0gbó-é, Ayo 1972, The Yoruba Verb Phrase, Idàbàn University Press. 1974, ‘On serial verbs and verbal status’, Journal of West African Linguistics 9: 17–48. Barbaud, Philippe 1971, ‘L’ambiguïté structurale du composé binominal’, Cahiers de linguistique 1: 71–116. Barss, Andrew and Lasnik, Howard 1986, ‘A note on anaphora and double objects’, Linguistic Inquiry 17: 347–52. Bary, René 1665, La rhétorique francoise ov povr principale avgmentation l’on trovve les secrets de nostre langve (Archives de la linguistique française 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 Amériques Noires, Paris: Payot. Bauer, Laurie 1980, ‘Deux problèmes au sujet des noms composés comprenant un premier élément verbal en français moderne’, Le français moderne 48: 219–24. 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: 409–27. Bentolila, Alain 1971, ‘Les systèmes verbaux créoles: comparaison avec les langues africaines’, Doctoral Dissertation, Université de Paris V, René Descartes. 1987, ‘Marques aspecto-temporelles en créole haïtien: de l’analyse synchronique à la formulation d’hypothèses diachroniques’, La Linguistique 23: 103–22. Bernabé, Jean 1983, Fondal-Natal (Grammaire basilectale approchée des créoles guadeloupéen et martiniquais 3), Paris: L’Harmattan. 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: 10–12. Bickerton, Derek 1977, ‘Pidginisation and creolisation: Language acquisition and language universals’, in Valdman (ed.), pp. 49–69. 1981, Roots of Language, Ann Arbor, MI: Karoma. 1984, ‘The language bioprogram hypothesis’, The Behavioral and Brain Sciences 7: 173–221. 1986, ‘Creoles and West African languages: A case of mistaken identity?’, in Muysken and Smith (eds.), pp. 25–40.

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: 307–18. 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 à la mémoire de Jean Stefanini, Aix: Publication de l’Université de Provence. Bloomfield, Leonard 1933, Language, New York: Holt, Rinehart, and Winston. Bonnard, Henri 1982, Synopsis de phonétique historique, Paris: Société d’édition d’enseignement supérieur. Bonneau, Alexandre 1856, ‘Les Noirs, les Jaunes et la littérature française en Haïti’, Revue contemporaine 5: 107–55. Boretzky, Norbert, Enninger, Werner and Stolz, Thomas (eds.) 1988, Beiträge zum 4. Essener Kolloquium, Bochum: Brockmeyer. Boretzky, Norbert and Stolz, Thomas (eds.) 1990, Beiträge zum 6. Essener Kolloquium, Bochum: Brockmeyer. Bouchard, Denis 1982, ‘Les constructions relatives en français vernaculaire et en français standard: étude d’un paramètre’, in Lefebvre (ed.), pp. 103–33. 1995, The Semantics of Syntax, University of Chicago Press. Bouvet, Francis and Andler, Pierre (eds.) 1984, Le Littré, Paris: Union générale d’éditions. Brasseur, Patrice 1995, ‘Les pronoms personnels dans l’Atlas linguistique normand’, in Simoni-Aurembou (ed.), pp. 323–51. Brillon, Joëlle and Brousseau, Anne-Marie 1986, ‘Aspects de la phonologie du fon’, in Lefebvre and Kaye (eds.), pp. 361–85. Brousseau, Anne-Marie 1988a, ‘Triptyque sur les composés: les noms composés en français, fongbe et haïtien en regard des notions de tête et de percolation’, Travaux de recherche sur le créole haïtien 2: 1–109 (Université du Québec à Montréal). 1988b, ‘Inherent object verbs in Fongbe’, in Lefebvre (ed.), pp. 371–400. 1989, ‘De “nù-fló” à “po-bouche”: hypothèses sur l’origine des composés en haïtien’, in Lefebvre and Lumsden (eds.) (1989b), pp. 285–312. 1990, ‘Panorama de la morphologie du Fongbe’, Journal of West African Languages 20: 27–45. 1991, ‘Domaines et relations de gouvernement dans les processus tonals du fongbe’, in Kihm (ed.), pp. 93–112. 1993, ‘Représentations sémantiques et projections syntaxiques des verbes en Fungbè’, Doctoral Dissertation, Université du Québec à Montréal. 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 créole haïtien, en français et en fongbé’, in Lefebvre (ed.) (1995a), volume IV, 33 pages. 1995b, ‘Réflexivité et classes de verbes en créole haïtien et en fongbé’, 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: 1–36. Brousseau, Anne-Marie and Lumsden, John S. 1992, ‘Nominal structure in Fongbe’, in Lefebvre (ed.), pp. 5–26. Brunot, Ferdinand 1905, Histoire de la langue française des origines à 1900, Paris: A. Colin. 1926, La pensée et la langue, Paris: Masson et cie. Bruyn, Adrienne 1996, ‘On identifying instances of grammaticalisation in creole languages’, in Baker and Syea (eds.), pp. 29–47. Butel, Paul 1982, ‘Le temps des fondations: Les Antilles avant Colbert’, in Pluchon (ed.), pp. 53–78. 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, ‘Représentations syllabiques et distribution des diphtongues en créole haïtien’, Études créoles 11: 9–40. 1988b, ‘L’opposition r/w en créole haïtien: un paradoxe résolu’, Revue canadienne de linguistique 33: 121–42. 1990, ‘Domaine prosodique d’une règle de sandhi externe et cliticisation des éléments fonctionnels du créole haïtien’, Université du Québec à Montréal MS. 1994, ‘Aspects de la phonologie du créole haïtien’, Doctoral Dissertation, Université du Québec à Montréal, Montreal. Capo, Hounkpati B. C. 1984, ‘Elements of Ewe–Gen–Aja–Fon dialectology’, in Medeiros (ed.), pp. 167–78. 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. 105–18. Carden, Guy and Stewart, William A. 1988, ‘Binding theory, bioprogram, and creolisation: Evidence from Haitian Creole’, Journal of Pidgin and Creole Languages 3: 1–68. Carlson, G. N. 1977, ‘A unified analysis of the English bare plural’, Linguistics and Philosophy 1: 413–57. 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: 49–53. Catach, Nina (ed.) 1995, Dictionnaire historique de l’orthographe française, 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 étude comparée des créoles et parlers français d’outre-mer: survivance et innovation’, Revue de linguistique romane 147–8: 342– 71. 1977, ‘Toward the reconstruction of the social matrix of creole language’, in Valdman (ed.), pp. 259–77. 1983, ‘Où l’on reparle de la genèse et des structures des créoles de l’Océan Indien’, Études créoles 6: 157–224. 1990, ‘Recherche, formation et créolistique’, Revue québécoise de linguistique théorique et appliquée 9: 287–303. 1992, Des îles, des hommes, des langues. Langues créoles/cultures créoles, Paris: L’Harmattan. 1993, ‘De l’hypothèse aux exemples. Un cas de créolisation: selon la formation des systèmes de démonstratifs créoles’, Études créoles 16: 17–38. 1994, ‘Comparatisme et méthode générative transformationnelle: le cas des études sur le créole haïtien’, Créolistique et grammaire générative, Pluralismes 8: 1–20. 1996, ‘Démystification de la relexification’, Études créoles 19: 93–109. 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. 43–74. 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: 141–77. Collins, Chris 1993, ‘Topics in Ewe syntax’, Doctoral Dissertation, MIT, Cambridge, MA. 1994, ‘The factive construction in Kwa’, Travaux de recherche sur le créole haïtien 23: 31–65 (Université du Québec à Montréal). 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: 69–94. 1989, ‘On French influence in the development of creole reflexive patterns’, Journal of Pidgin and Creole Languages 4: 103–16. Cowper, Elizabeth A. 1989, ‘Thematic underspecification: The case of have’, paper read at the Canadian Linguistics Association Annual Meeting, Université Laval, Quebec City. 1995, ‘English participle construction’, Revue canadienne de linguistique 40: 1–38. Crystal, David 1991, A Dictionary of Linguistics and Phonetics, Cambridge: Blackwell.

430

References

Curtin, Philip D. 1976, ‘The Atlantic slave trade, 1600–1800’, in Ajayi and Crowder (eds.), pp. 302–30. Da Cruz, Maxime 1993, ‘Les constructions sérielles du Fungbè: approches sémantique et syntaxique’, Doctoral Dissertation, Université du Québec à Montréal. 1994, ‘Contribution à l’étude de la négation en Fungbè’, in Lefebvre and Lumsden (eds.) (1994e), volume III, pp. 69–111. Da Cruz, Maxime and Kinyalolo, Kasangati 1994, ‘Les classes aspectuelles des verbes en Fongbe’, in Lefebvre and Lumsden (eds.) (1994d), volume III, pp. 1–113. Dalby, David (ed.) 1970, Language and History in Africa, London: Frank Cass & Co. Ltd. Damoiseau, Robert 1988, ‘Éléments pour une classification des verbaux en créole haïtien’, Études créoles 11: 41–64. Darmesteter, Arsène 1875, Traité de la formation des mots composés dans la langue française comparée aux autres langages romanes et au latin (2nd edition), Paris: Champion (1967). Debien, Gabriel 1962, Plantations et esclaves à Saint-Domingue, Dakar. DeCamp, David 1974, ‘Neutralisations, iteratives, and ideophones: The locus of language in Jamaica’, in DeCamp and Hancock (eds.), pp. 42–60. DeCamp, David and Hancock, Ian (eds.) 1974, Pidgins and Creoles. Current Trends and Prospects, Washington, DC: Georgetown University Press. Déchaine, Rose-Marie 1988, ‘Opérations sur les structures d’arguments: Le cas des constructions sérielles en Haïtien’, Master’s Thesis, Université du Québec à Montréal. 1991, ‘Bare sentences’, in Moore and Wyner (eds.), pp. 31–50. 1992, ‘Inflection in Ìgbo and Yorùbá’, in Collins and Manfredi (eds.), pp. 95–119. 1993, ‘Predicates across categories’, Doctoral Dissertation, University of Massachusetts, Amherst. Déchaine, Rose-Marie and Filipovich, Sandra 1986, ‘La construction relative en Fon’, in Lefebvre and Kaye (eds.), pp. 198–221. Déchaine, Rose-Marie and Manfredi, Victor 1994, ‘Binding domains in Haitian’, Natural Language and Linguistic Theory 12: 203–57. Dees, Anthonij 1971, Étude sur l’évolution des démonstratifs en ancien et moyen français, Groningen: Wolters-Noordhoff. DeGraff, Michel A. F. 1992a, ‘Is Haitian Creole a pro-drop language?’, Travaux de recherche sur le créole haïtien 11: 1–22 (Université du Québec à Montréal). 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 créole haïtien 11: 59–74 (Université du Québec à Montréal). 1993a, ‘A riddle on negation in Haitian’, Probus 5: 63–93. 1993b, ‘Is Haitian Creole a pro-drop language?’, in Byrne and Holm (eds.), pp. 71–90. 1994, ‘The morphology-syntax interface in creolisation (and diachrony)’, Studies in the Linguistic Sciences 24: 115–31. 1996, ‘Creole languages and parameter setting: A case study using Haitian Creole and the pro-drop parameter’, in Wekker (ed.), pp. 65–105. DeGraff, Michel A. F. (ed.) In press, Parametric Approaches to Creole Syntax, Cambridge, MA: MIT Press.

References

431

Dejean, Y. 1980, Comment écrire le créole d’Haïti, Montréal: 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. 71–86. Deprez, Viviane 1992a, ‘Is Haitian Creole really a pro-drop language?’, Travaux de recherche sur le créole haïtien 11: 23–40 (Université du Québec à Montréal). 1992b, ‘Raising constructions in Haitian Creole’, Natural Language and Linguistic Theory 10: 191–231. In press, ‘The roots of negative concord in French and French based creoles’, in DeGraff (ed.). Deprez, Viviane and Vinet, Marie-Thérèse 1991, ‘ “Se” in Haitian Creole: an ambiguous X’ category’, MS. Descourtilz, M. E. 1809, Voyages d’un 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: 32–3. 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. 209–32. Di Sciullo, Anne-Marie and Rochette, Anne (eds.) 1990, Binding in Romance: Essays in Honour of Judith McA’Nulty, 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 congrès de la Société linguistique de l’Afrique occidentale, Série H, Abidjan: Annales de l’Université d’Abidjan. Domingue, Nicole 1980, ‘Syntactic innovations in Mauritian Bhojpuri’, McGill University, MS. 1981, ‘Internal change in a transplanted language’, Studies in the Linguistic Sciences 4: 151–9. Dressler, Wolfgang U. and Pfeiffer, Oskar E. (eds.) 1976, Phonologica, Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Dubois, Jean 1962, Étude sur la dérivation suffixale en français moderne et contemporain, Paris: Larousse. Dubois, Jean, Lagane, René and Lerond, Alain 1992, Dictionnaire du français classique: le XVIIe siècle, Paris: Larousse. Ducœurjoly, S. J. 1802, Manuel des habitans de Saint-Domingue, Paris: Lenoir. Dumais, Danielle 1988, ‘INFL en créole haïtien’, in Lefebvre (ed.), pp. 242–65. 1992, ‘Classification sémantique des verbes du fon et du créole haïtien’, in Lefebvre and Lumsden (eds.), volume V, pp. 1–217. 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: 373–413. Emirkanian, Louisette and Sankoff, David L. 1984, ‘Le futur simple et le futur périphrastique’, in Lemieux-Nieger and Cedergren (eds.), pp. 177–91. Fabb, Nigel 1992, ‘The licensing of Fon verbs’, in Lefebvre (ed.), pp. 27–34. Faine, Jules 1937, Philologie créole: études historiques et étymologiques sur la langue créole d’Haïti, Port-au-Prince: Imprimerie de l’État. 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: Réflexions sur le créole haïtien’, Paris V, MS. Féraud, Jean-François 1768, Dictionnaire grammatical de la langue française contenant toutes les règles de l’orthographe, de la prononciation (Microfiches, 105 × 148 mm, microcopy of Paris edition, 1768), Paris: France-expansion, 1973, collection Archives de la linguistique française. Férère, Gérard 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 d’Haïti’, in Lefebvre and Kaye (eds.), pp. 386– 418. 1988, ‘La morphologie de l’haïtien’, Master’s Thesis, Université du Québec à Montréal, Montreal. Flûtre, Louis-Ferdinand 1970, Le moyen picard, d’après les textes littéraires du temps (1560–1660): 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. 162–83. Fouché, Pierre 1952, Introduction (Phonétique historique du français I), Paris: Librairie C. Klincksieck. Fournier, Robert 1977, ‘N ap fè yun ti-kose su la (La grammaire de la particule la en créole haïtien)’, Master’s Thesis, Université du Québec à Montréal. 1987, ‘Pu en créole haïtien’, Revue québécoise de linguistique théorique et appliquée 6: 107–26. Freeman, Bryant C. 1988, Dictionnaire inverse de la langue haïtienne, 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. 128–73. Furetière, Antoine 1984, Le dictionnaire universel d’Antoine Furetière (3 volumes), Paris: Le Robert. Galet, Yvette 1971, L’évolution de l’ordre des mots dans la phrase française de 1600 à 1700, Paris: Presses universitaires de France.

References

433

Gambhir, Surendra K. 1988, ‘Structural development in Guyanese Bhojpuri’, in Barz and Siegel (eds.), pp. 69–94. Gass, Susan M. and Selinker, Larry (eds.) 1983, Language Transfer in Language Learning, Rowley, MA: Newbury House. Gauthier, Pierre 1995, ‘Le système des pronoms dans le Rolea, recueil anonyme poitevin du XVIIe siècle’, in Simoni-Aurembou (ed.), pp. 311–22. Gendron, Jean-Denis 1966, Tendances phonétiques du français parlé au Canada, Paris/ Quebec City: Librairie C. Klincksieck/Presses de l’Université Laval. Ghomeshi, Jila and Massam, Diane 1994, ‘Lexical/syntactic relations without projection’, Linguistic Analysis 24: 175–217. Gilbert, Glen G. (ed.) 1987, Pidgin and Creole Languages: Essays in Memory of John E. Reinecke, Honolulu: University of Hawaii Press. Gilles, Rolande 1988, ‘Réalisation du Cas en créole haïtien dans trois environnements: les prépositions, les marqueurs de Cas et les noms’, Master’s Thesis, Université du Québec à Montréal. Girodet, Jean 1976, Logos/Grand dictionnaire de la langue française, Paris: Bordas. Giurescu, Anca 1975, Les mots composés dans les langues romanes, The Hague/Paris: Mouton. Givón, Talmy 1971, ‘On the verbal origin of the Bantu suffixes’, Studies in African Linguistics 2: 145–63. Göbl-Galdi, László 1934, ‘La structure grammaticale des patois français-créoles’, Zeitschrift für französische Sprache und Literatur 58: 257–95. Goldsmith, John 1978, ‘ “Que” c’est quoi? “Que”, c’est “quoi” ’, Recherches linguistiques à Montréal 11: 1–15. 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. 243–55. Gougenheim, Georges 1973, Grammaire de la langue française au seizième siècle, 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. 73–113. 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. 3–22. Grevisse, Maurice 1975, Le Bon Usage, Paris/Gembloux: Duculot. Grimshaw, Jane 1989, ‘Getting the dative alternation’, in Laka Mugarza and Mahajan (eds.), pp. 113–22. Grimshaw, Jane and Meister, Armin 1988, ‘Light verbs and θ-marking’, Linguistic Inquiry 19: 205–32. Gros 1791, Isle St.-Domingue, Province du Nord. Récit historique sur les évènements qui se sont succédés . . . depuis le 26 octobre 1791, jusqu’au 24 décembre de la même année, n.p. Gruber, Jeff S. 1992, ‘Proper argument projection in Igbo and Yoruba’, in Collins and Manfredi (eds.), pp. 139–65. 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 français, 2nd edition, Paris: Presses universitaires de France, collection Que sais-je? Haase, Albert 1965, Syntaxe française du XVIIe siècle, 5th edition, Paris: Delagrave. Hagège, Claude 1974, ‘Les pronoms logophoriques’, Bulletin de la Société de linguistique de Paris 1: 287–310. 1985, L’homme de parole, Paris: Librairie Arthème Fayard. Hair, P. E. H. 1966, ‘Collections of vocabularies of Western Africa before the polyglotta: A key’, Journal of African Languages 5: 208–17. 1970, ‘The contribution of early linguistic material to the history of West Africa’, in Dalby (ed.), pp. 50–63. 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: 194–203. 1953, Haitian Creole: Grammar, Texts and Vocabulary (American Anthropologist Association 74), Menasha: Georges Banta Publishing Company. 1958, ‘Creolised languages and “genetic relationships” ’, Word 14: 367–73. 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: 17–35. 1980b, ‘Lexical expansion in creole languages’, in Valdman and Highfield (eds.), pp. 63–87. 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. Hazoumê, M. L. 1990, Essai de Classification Synchronique (Étude comparative des parlers Gbe du Sud-Bénin), Cotonou, Bénin: Centre national de linguistique appliquée. 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, 1492–1971, 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: 40–57; 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 amérindiens (L’édifice créole en Haïti 1), Fribourg, Switzerland: Edikreyol. 1993, Soubassements africains (L’édifice créole en Haïti 2), Fribourg, Switzerland: Edikreyol. Hilliard d’Auberteuil, M. R. 1776–7, Considérations sur l’état présent de la colonie française de Saint-Domingue (2 volumes), Paris: Grangé. Hirschbühler, Paul 1978, ‘The Semantics and Syntax of Wh-Constructions’, Doctoral Dissertation, University of Massachusetts, Amherst. Hoekstra, Teun 1992, ‘Subjects inside out’, Revue québécoise de linguistique 22: 45–76. Holm, John 1986, ‘Substrate diffusion’, in Muysken and Smith (eds.), pp. 259–78. 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. 317–27. Hopper, Paul J. and Traugott, Elizabeth C. 1993, Grammaticalisation (Cambridge Textbooks in Linguistics), Cambridge University Press. Horiot, Brigitte (ed.) 1991, Français du Canada – Français de France, Tübingen: Niemeyer (Actes du 2e Colloque international de Cognac (September 27–30, 1988) ). Hornstein, Norbert 1977, ‘Towards a theory of tense’, Linguistic Inquiry 8: 521–57. Hounkpatin, Basile 1985, ‘Le verbal et le syntagme verbal du fon-gbe parlé à Massè’, Doctoral Dissertation, Université de la Sorbonne nouvelle, Paris III. Huguet, Edmond 1925, Dictionnaire de la langue française du seizième siècle, volume I, Paris: M. Didier. Hulk, Aafke 1986, ‘Subject clitics and the PRO-drop parameter’, in Coopmans, Bordelois and Smith (eds.), pp. 107–20. 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, ‘Affinités entre les variétés du français’, in Valdman (ed.), pp. 165–82. 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. 1–22. Hyman, Larry M. and Comrie, Bernard 1981, ‘Logophoric reference in Gokana’, Journal of African Languages and Linguistics 3: 19–37. Hymes, Dell 1971a, ‘Introduction’, in Hymes (ed.), pp. 65–91. 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, Trésor de la langue française, Paris: CNRS. Jackendoff, Ray 1976, ‘Toward an explanatory semantic representation’, Linguistic Inquiry 7: 89–150. 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. 132–53. Jaeggli, Osvaldo A. and Silva-Corvalán, 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: 131–55. James, Deborah 1984, ‘Raising to subject in Moose Cree: A problem for subjacency’, in Cook and Gerdts (eds.), pp. 205–13. Jansen, Bert, Koopman, Hilda and Muysken, Pieter C. 1978, ‘Serial verbs in the creole languages’, Amsterdam Creole Studies 11: 125–59. Janvier, Louis Joseph 1886, ‘Berceuse haïtienne, Zangui (L’Anguille)’, Revue des Traditions Populaires 1: 21–3, 107. Jeanjean, Colette 1988, ‘Le futur simple et le futur périphrastique en français parlé’, in Blanche-Benveniste, Chervel and Gross (eds.), pp. 235–57. Jespersen, Otto 1922, Language: Its Nature, Development and Origin, London: Allen and Unwin. Johns, Alana 1992, ‘Deriving ergativity’, Linguistic Inquiry 17: 587–622. Johnson, Kyle 1991, ‘Object positions’, Natural Language and Linguistic Theory 9: 577– 639. Joly, Geneviève 1995, Précis de phonétique historique du français, Paris: A. Colin. Jones, Charles and Sells, Peter (eds.) 1984, Proceedings of NELS 14, University of Massachusetts at Amherst, GLSA. Joseph, Frantz 1989, ‘La détermination nominale en créole haïtien’, Doctoral Dissertation, Université de Paris VII, Jussieu. Joseph, Joseph Sauveur 1994, ‘De l’étude sémantico-syntaxique de la locution nan du créole haïtien’, Master’s Thesis, Université du Québec à Montréal. 1995, ‘La négation en créole Haïtien: un cas de relexification’, Université du Québec à Montréal 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 française, The Hague: Mouton. Juneau, Marcel 1972, Contribution à l’histoire de la prononciation française au Québec, Paris/Quebec: Librairie C. Klincksieck/Presses de l’Université Laval. Juneau, Marcel and Poirier, Claude 1973, Le livre de comptes d’un meunier québécois (fin XVIIe-début XVIIIe siècle). Édition avec étude linguistique, Quebec City: Les Presses de l’Université Laval. Jungraithmayr, Hermans and Möhlig, 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. 61–72. Kaye, Jonathan, Koopman, Hilda, Sportiche, Dominique and Dugas, André (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. 255–99. 1981, ‘On certain differences between French and English’, Linguistic Inquiry 12: 349–71. 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: 595–621. Keenan, Edward 1976, ‘Remarkable subjects in Malagasy’, in Li (ed.), pp. 247–301. 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. 351–76. Kihm, Alain (ed.) 1991, La Créolisation: Théorie et applications (Recherches linguistiques de Vincennes 20), Paris: Presses universitaires de Vincennes. Kihm, Alain 1994, ‘French–Gbe 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 Fongbè: Études de phonologie, de syntaxe et de sémantique, Paris: Peeters. 1995, Les propriétés lexicales du créole haïtien 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. 35–51. 1993a, ‘Conditionals and factives in Fwn’, in Lefebvre and Lumsden (eds.), volume IV, pp. 161–73. 1993b, ‘On some syntactic properties of #W in Fon’, Lingua 91: 201–33. 1993c, ‘The logophoric pronoun émi as a LF operator-anaphor’, in Schafer (ed.), pp. 223–37. 1994, ‘Pronouns and relexification’, in Lefebvre and Lumsden (eds.) (1994e), volume III, pp. 259–93. Kiple, Kenneth 1984, The Caribbean Slave: A Biological History, Cambridge University Press. Klemperer, Victor and Lerch, Eugen (eds.) 1925, Jahrbuch für Philologie, München: 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. 167–203. 1982b, ‘Les questions’, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 204–33. 1982c, ‘Quelques problèmes concernant que/quoi, ce que et qu’est-ce que’, in Lefebvre (ed.), pp. 135–70. 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. 231–58.

438

References

Koopman, Hilda and Lefebvre, Claire 1981, ‘Haitian Creole “pu” ’, in Muysken (ed.), pp. 201–21. 1982, ‘PU: marqueur de mode, préposition et complémenteur’, in Lefebvre, MagloireHolly and Piou (eds.), pp. 64–91. Koopman, Hilda and Sportiche, Dominique 1989, ‘Pronouns, logical variables and logophoricity in Abe’, Linguistic Inquiry 20: 555–88. 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. 75–102. 1992, ‘From OV to VO. Linguistic negotiation in the development of Berbice Dutch Creole’, Lingua 88: 263–99. 1993, ‘Cliticisation of pronouns in Berbice Dutch and Eastern Ijo’, in Byrne and Holm (eds.), pp. 119–32. 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, Gaétan, Lévy, Joseph and Tardif, Francine 1993, ‘La question nationale dans le discours de leaders d’associations ethniques de la région de Montréal’, Cahiers de recherche sociologique 20: 85–111. Laborderie, Noëlle 1994, Précis de phonétique historique, Paris: Nathan. Labouret, Henri and Rivet, Paul 1929, Le royaume d’Arda et son évangélisation au XVIIe siècle (Travaux et mémoires de l’Institut d’ethnologie), Université de Paris. Labov, William 1972, Sociolinguistic Patterns, Philadelphia: The University of Pennsylvania Press Inc. Lafage, Suzanne 1985, Français écrit et parlé en pays Ewé (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. 236–52. Larson, Richard K. 1988, ‘On the double object construction’, Linguistic Inquiry 19: 335–91. 1990, ‘Double objects revisited: Reply to Jackendoff’, Linguistic Inquiry 21: 589–632. 1991, ‘Some issues in verb serialisation’, in Lefebvre (ed.), pp. 185–211. Larson, Richard K. and Lefebvre, Claire 1991, ‘Predicate clefting in Haitian Creole’, in Sherer (ed.), pp. 247–62. Law, Paul 1991, ‘Verb movement, expletive replacement, and head-government’, The Linguistic Review 8: 253–85. 1992, ‘What is a null subject language?’, Travaux de recherche sur le créole haïtien 11: 41–58 (Université du Québec à Montréal). 1994a, ‘Une note sur la construction existentielle en Haïtien’, Revue québécoise de linguistique 23: 139–67. 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. 7–47.

References

439

Lefebvre, Claire 1981, ‘The double structure of questions in French. A case of syntactic variation’, in Sankoff and Cedergren (eds.), pp. 229–39. 1982a, ‘Introduction’, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 1–20. 1982b, ‘L’expansion d’une catégorie grammaticale: le déterminant la’, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 21–63. Lefebvre, Claire (ed.) 1982c, La syntaxe comparée du français standard et populaire: approches formelle et fonctionnelle (2 volumes), Montreal: Gouvernement du Québec (Collection langues et société). Lefebvre, Claire 1982d, ‘Le répertoire de mots Wh en français vernaculaire et leur insertion dans la grammaire du français’, in Lefebvre (ed.), pp. 73–125. 1982e, ‘Qui qui vient’ ou ‘Qui vient’: Voilà la question’, in Lefebvre (ed.), pp. 47–101. 1984, ‘Grammaires en contact: définition et perspectives de recherche’, Revue québécoise de linguistique 14: 11–49. 1986, ‘Relexification in creole genesis revisited: The case of Haitian Creole’, in Muysken and Smith (eds.), pp. 279–301. Lefebvre, Claire (ed.) 1988, Études syntaxiques, morphologiques et phonologiques. Research report prepared for SSHRCC, FCAR and PAFAC on the project HaitiFon, Université du Québec à Montréal. Lefebvre, Claire 1989, ‘Some problems in defining syntactic variables: The case of questions in Montreal French’, in Fasold and Schiffrin (eds.), pp. 351–66. 1990a, ‘Establishing a syntactic category P in Fon’, Journal of West African Languages 20: 45–63. 1990b, ‘On the interpretation of predicate cleft’, Linguistic Review 6: 169–94. 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. 37–78. 1991c, ‘La distribution du déterminant et des complémenteurs en créole haïtien’, in Kihm (ed.), pp. 21–47. 1991d, ‘Note sur la position de AGRO dans la structure syntaxique’, Revue québécoise de linguistique 20: 281–94. 1992a, ‘On the distribution of clausal wY in Fongbe’, Journal of West African Languages 21: 21–35. 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. 137–56. 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: 113–31. 1993a, ‘The role of relexification and syntactic reanalysis in Haitian Creole: Methodological aspects of a research program’, in Mufwene (ed.), pp. 254–79. 1993b, ‘The role of relexification in creole genesis: The case of functional categories’, Travaux de recherche sur le créole haïtien 14: 23–52 (Université du Québec à Montréal). 1993c, ‘Dominance versus precedence in the double object construction: New facts from Fongbe’, Revue canadienne de linguistique 38: 395–425.

440

References

1993d, ‘Le clivage du prédicat: essai de typologie’, in Kihm and Lefebvre (eds.), pp. 51–60. 1994a, ‘Functional categories and relexification’, in Lefebvre and Lumsden (eds.) (1994c), 74 pages. 1994b, ‘On spelling out E’, Travaux de recherche sur le créole haïtien 23: 1–33 (Université du Québec à Montréal). 1994c, ‘New facts from Fongbe on the double object constructions’, Lingua 94: 69–163. Lefebvre, Claire (ed.) 1995a, Research report prepared for FCAR on the project L’organisation des lexiques et des entrées lexicales (4 volumes), Université du Québec à Montréal. Lefebvre, Claire 1995b, ‘PPs headed by the Fon preposition nú are syntactic adjuncts’, in Kihm and Lefebvre (eds.), pp. 153–83. Lefebvre, Claire (ed.) 1995c, Research report prepared for SSHRCC on the project La moisson de 10 ans de recherche sur la genèse du créole haïtien, Université du Québec à Montréal. Lefebvre, Claire 1996a, ‘The functional category AGR and creole genesis’, in Wekker (ed.), pp. 153–95. 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: 231–313. 1997, ‘Relexification in creole genesis: The case of demonstrative terms in Haitian Creole’, Journal of Pidgin and Creole Languages 12.2: 181–203. 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, Études syntaxiques, morphologiques et phonologiques. Research report prepared for FCAR and PAFAC on the project FonCréole Haïtien, Université du Québec à Montréal. Lefebvre, Claire and Lumsden, John S. 1989a, ‘Les langues créoles et la théorie linguistique’, in Lefebvre and Lumsden (eds.) (1989b), pp. 249–72. Lefebvre, Claire and Lumsden, John S. (eds.) 1989b, La créolisation. Special issue of the Revue canadienne de linguistique 34.3. Lefebvre, Claire and Lumsden, John S. (eds.) 1989c, Le créole haïtien. Special issue of the Revue québécoise de linguistique 18.2. Lefebvre, Claire and Lumsden, John S. 1992a, ‘On word order in relexification’, Travaux de recherche sur le créole haïtien 10: 1–22 Université du Québec à Montréal. Lefebvre, Claire and Lumsden, John S. (eds.) 1992b, Classification sémantique des verbes du fon et du créole haïtien. Research report prepared for SSHRCC, FCAR and PAFAC on the project La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle (5 volumes), Université du Québec à Montréal. Lefebvre, Claire and Lumsden, John S. (eds.) 1993, Études syntaxiques. Research report prepared for SSHRCC, FCAR and PAFAC on the project La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle (4 volumes), Université du Québec à Montréal.

References

441

Lefebvre, Claire and Lumsden, John S. 1994a, ‘Le rôle central de la relexification dans la genèse des langues créoles’, Plurilinguismes 8: 47–93. 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 genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle, Université du Québec à Montréal. Lefebvre, Claire and Lumsden, John S. (eds.) 1994d, Les classes aspectuelles de verbes. Research report prepared for FCAR on the project L’organisation des lexiques et des entrées lexicales (3 volumes), Université du Québec à Montréal. Lefebvre, Claire and Lumsden, John S. (eds.) 1994e, Études syntaxiques. Research report prepared for SSHRCC and PAFAC on the project La genèse du créole haïtien: un cas particulier d’investigation sur la forme de la grammaire universelle (3 volumes), Université du Québec à Montréal. 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, Hélène and Piou, Nanie (eds.) 1982, Syntaxe de l’haïtien, 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: 213–43. 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. 65–91. Lemieux-Nieger, Monique and Cedergren, Henrietta (eds.) 1984, Les tendances dynamiques du français parlé à Montréal, Université du Québec à Montréal. Levin, Beth and Rappaport, Malka 1986, ‘The formation of adjectival passives’, Linguistic Inquiry 17: 623–61. Levinson, Stephen C. 1991, ‘Pragmatic reduction of the binding conditions revisited’, Journal of Linguistics 27: 107–61. Li, Charles N. (ed.) 1976, Subject and Topic, New York: Academic Press. Lichtenberk, Frantisek 1984, To’aba’ita 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. 369–78. Lommatzsch, E. 1925, ‘Deiktische Elemente im Altfranzösischen’, in Klemperer and Lerch (eds.), pp. 202–44. Lord, Carol 1976, ‘Evidence for syntactic reanalysis: From verb to complementiser in Kwa’, in Steever, Walker and Mufwene (eds.), pp. 179–91.

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. 64–93. 1990, ‘The bi-clausal structure of Haitian clefts’, Linguistics 28: 741–60. 1991, ‘On the acquisition of nominal structures in the genesis of Haitian Creole’, Travaux de recherche sur le créole haïtien 4: 37–61 (Université du Québec à Montréal). 1993a, ‘Expletives in double object constructions in Haitian Creole’, in Byrne and Winford (eds.), pp. 269–84. 1993b, ‘Relexification and parameter setting in creole genesis’, Travaux de recherche sur le créole haïtien 14: 52–69 (Université du Québec à Montréal). 1994a, ‘Possession: Substratum semantics in Haitian creole’, Journal of Pidgin and Creole Languages 9: 25–51. 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 aren’t what you might think’, Linguistics 28: 761–83. Lumsden, John S. and Lefebvre, Claire 1994, ‘The genesis of Haitian Creole’, in Lefebvre and Lumsden (eds.) (1994c), 15 pages. Magloire-Holly, Hélène 1982, ‘Les modaux: Auxiliaires ou verbes?’, in Lefebvre, MagloireHolly and Piou (eds.), pp. 92–120. 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 Yorùbá nominalisations’, in Collins and Manfredi (eds.), pp. 205–17. 1993, ‘Verb focus in the typology of Kwa/Kru and Haitian’, in Byrne and Winford (eds.), pp. 3–51. 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: 435–83. 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. 95–130. 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: 461–94. Medeiros, François de (ed.) 1984, Peuples du golfe du Bénin: Aja-Ewé (Colloque de Cotonou), Paris: Karthala. Meisel, Jürgen M. 1983a, ‘Strategies of second language acquisition. More than one kind of simplification’, in Andersen (ed.), pp. 120–57.

References

443

1983b, ‘Transfer as a second-language strategy’, Language and Communication 3: 11– 46. Meyer-Lübke, Wilhelm 1909, ‘Le français au Canada’, Bulletin du français parlé au Canada 8: 121–9. Miller, George A. 1978, ‘Semantic relations among words’, in Halle, Bresnan and Miller (eds.), pp. 60–118. Milner, Jean-Claude 1978, De la syntaxe à l’interprétation, Paris: Éditions du Seuil. Mintz, Sidney W. 1971, ‘The socio-historical background to pidginisation and creolisation’, in Hymes (ed.), pp. 481–98. Möhlig, Wilhelm J. G. 1983, ‘Mbugu’, in Jungraithmayr and Möhlig (eds.), pp. 158–59. 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, ‘L’homme que je crois qui est venu: que, qui relatifs et conjonctions’, Langue française 11: 77–90. Moreau de Saint-Méry, Louis-Élie 1797, Description topographique, physique, civile, politique et historique de la partie française de l’isle de Saint-Domingue (2nd edn, Paris: L. Guérin, 1875), Philadelphia. Morin, Yves-Charles 1977, ‘Interprétation des pronoms et des réfléchis en français’, Recherches linguistiques à Montréal 8: 67–94. Mougeon, Raymond and Beniak, Édouard (eds.) 1994, Les origines du français québécois, Sainte-Foy: Les Presses de l’Université Laval. Mous, Maarten 1994, ‘Ma’a or Mbugu’, in Bakker and Mous (eds.), pp. 175–200. 1995, ‘Language intertwining’, paper read at the Amsterdam Creole Workshop: Creole Genesis and Language Contact, Amsterdam. To appear, ‘Ma’a 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. 129–62. 1989, ‘Some explanations that strike me as incomplete’, Journal of Pidgin and Creole Languages 4: 117–28. 1990, ‘Transfer and the substrate hypothesis in creolistics’, Studies in Second Language Acquisition 12: 1–23. 1991, ‘Pidgins, creoles, typology, and markedness’, in Byrne and Huebner (eds.), pp. 123–44. 1992, ‘Africanisms in Gullah: A re-examination of the issues’, in Hall, Doane and Ringler (eds.), pp. 156–82. 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. 133–44. 1993c, ‘Créole, créolisation, substrat, et autres notions apparentées: quelques réflexions sur Des îles, des hommes, des langues. Langues créoles – cultures créoles de Robert Chaudenson’, Études créoles 16: 117–41. 1994, ‘Genèse de population et genèse de langue’, Créolistique et grammaire générative 8: 95–113.

444

References

1996, ‘Creole genesis: A population genetics perspective’, in Christie (ed.), pp. 163– 96. Mühlhäusler, Peter 1980, ‘Structural expansion and the process of creolisation’, in Valdman and Highfield (eds.), pp. 19–56. 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. 41–55. 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. 52–79. 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. 181–200. 1987, ‘Prepositions and postpositions in Saramaccan’, in Alleyne (ed.), pp. 89–102. 1988a, ‘Lexical restructuring in creole genesis’, in Boretzky, Enninger and Stolz (eds.), pp. 193–210. 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: 409–22. 1988d, ‘Parameters for serial verbs’, in Demuth and Manfredi (eds.). 1994a, ‘Saramaccan and Haitian: A comparison’, Journal of Pidgin and Creole Languages 9: 305–14. 1994b, ‘The search for universals in language genesis: État de la question and research program’, Amsterdam Creole Studies 61: 16–28. Muysken, Pieter C. and Smith, Norval 1986a, ‘Introduction: Problems in the identification of substratum features’, in Muysken and Smith (eds.), pp. 1–14. 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: 883–903. 1995, ‘Reflexives’, in Arends, Muysken and Smith (eds.), pp. 271–87. Naro, Anthony J. 1978, ‘A study of the origins of pidginisation’, Language 54: 314–47. Ndayiragije, Juvénal 1993, ‘Clivage du prédicat en fon’, in Kihm and Lefebvre (eds.), pp. 61–86. 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 l’histoire naturelle de St. Domingue, Paris. Nida, Eugène 1948, ‘The identification of morphemes’, Language 24: 414–41. Niobey, Georges, Lagane, René and Guilbert, Louis (eds.) 1971, Grand Larousse de la langue française (7 volumes), Paris: Larousse. Nyrop, Kristopher 1936, Formation des mots (Grammaire historique de la langue française III), Copenhagen: Gyldendalske Boghandel, Nordisk Forlag.

References

445

Obenauer, Hans-Georg 1976, Études de syntaxe interrogative du français. Quoi, combien et le complémenteur, Tübingen: Max Niemeyer. 1977, ‘Syntaxe et interprétation’, Le français moderne 45: 305–41. 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, rapportée au langage du temps [Archives de la linguistique française 306], Paris: A. de Sommaville. Ouhalla, Jamal 1991, Functional Categories and Parametric Variation, London: Routledge. Papen, Robert 1988, ‘Convergence et divergence en métif’, paper read at the Département de linguistique, Université du Québec à Montréal. Pazzi, R. 1976, L’homme Ewè, Gèn, Aje, Fon et son Univers (Dictionnaire I: Lexique des noms), Lomé (Microfiche). Pierret, Jean-Marie 1983, Phonétique du français. Notions de phonétique générale et phonétique historique du français, 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 prédicat’, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 122–51. 1982b, ‘Le redoublement verbal’, in Lefebvre, Magloire-Holly and Piou (eds.), pp. 152–66. Plag, Ingo 1993, Sentential Complementation in Sranan: On the Formation of an EnglishBased Creole Language, Tübingen: Max Niemeyer. Pluchon, Pierre (ed.) 1982, Histoire des Antilles et de la Guyane, Toulouse: Privat. Poirier, Claude 1994, ‘La langue parlée en Nouvelle-France: vers une convergence des explications’, in Mougeon and Beniak (eds.), pp. 237–74. Pollock, Jean-Yves 1989, ‘Verb movement, Universal Grammar, and the structure of IP’, Linguistic Inquiry 20: 365–424. Pranka, Paula M. 1983, ‘Syntax and word formation’, Doctoral Dissertation, MIT, Cambridge, MA. Quillet, Aristide 1970, Dictionnaire encyclopédique Quillet (8 volumes), Paris: Quillet. Rappaport, Malka and Levin, Beth 1988, ‘What to do with theta-roles’, in Wilkens (ed.), pp. 7–36. Rassinoux, Jean 1987, Dictionnaire français-fon, Cotonou: Imprimerie Dumas. Rey, Alain (ed.) 1992, Dictionnaire historique de la langue française (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. 1–46. Ritchie, William C. and Bathia, Tej K. (eds.) 1996b, Handbook of Second Language Acquisition, San Diego: Academic Press.

446

References

Ritter, Elizabeth 1991, ‘Déplacement de NP en haïtien: oui ou non’, in Kihm (ed.), pp. 65–86. 1992, ‘Cross-linguistic evidence for number phrase’, in Lefebvre, Lumsden and Travis (eds.), pp. 197–219. Rivard, A. 1914, Études 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 SilvaCorvalán (eds.), pp. 391–419. 1986b, ‘Null objects in Italian and the theory of pro’, Linguistic Inquiry 17: 501–57. Roberge, Yves 1990, The Syntactic Recoverability of Null Arguments, Kingston: McGillQueen’s University Press. Robert, Paul 1985, Le Grand Robert de la langue française (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. 211–16. 1993, ‘The Ijo element in Berbice Dutch and the pidginisation/creolisation process’, in Mufwene (ed.), pp. 296–316. Roeper, Thomas and Williams, Edwin (eds.) 1987, Parameter Setting (Studies in Theoretical Psycholinguistics), Dordrecht: Reidel. Romain, Jean-Baptiste 1959, Quelques mœurs et coutumes des paysans haïtiens (Travaux pratiques d’ethnographie sur la région de Milot à l’usage des étudiants), Port-auPrince: Imprimerie de l’État. 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, Théodore 1911, Les origines de la prononciation moderne étudiées au 17ième siècle, Paris: A. Colin. Rouveret, Alain and Vergnaud, Jean-Roger 1980, ‘Specifying reference to the subject: French causatives and conditions on representations’, Linguistic Inquiry 11: 97–202. 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. 104–19. 1990, ‘The grammaticalisation of tense and aspect in Tok Pisin and Sranan’, Language Variation and Change 2: 295–312. 1991, ‘Using the future to explain the past’, in Byrne and Huebner (eds.), pp. 61–74.

References

447

Sankoff, Gillian and Laberge, Suzanne 1973, ‘On the acquisition of native speakers by a language’, Kivung 6: 32–47. 1980, ‘On the acquisition of native speakers by a language’, in Sankoff (ed.), pp. 195– 209. Sauvageot, Serge and Manessy, Gabriel 1963, Wolof et Sérèr: Études de phonétique et de grammaire descriptive, Université 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: 165–214. Sebba, Mark 1981, ‘Derivational regularities in a creole lexicon: The case of Sranan’, Linguistics 19: 101–17. Segurola, R. P. B. 1963, Dictionnaire fon-français, Cotonou: Procure de l’Archidiocèse. 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. 57–71. Sherer, Tim (ed.) 1991, Proceedings of the North East Linguistic Society, Université du Québec à Montréal: 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 littérature du domaine d’oïl occidental. Lexique des plantes: morphosyntaxe, Fontaine-lès-Dijon: A.B.D.O. Singler, John V. 1979, ‘Verb suffixes in Klao’, Cahiers ivoiriens de recherche linguistique 6: 9–33. 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: 27–51. Singler, John V. (ed.) 1990, Pidgin and Creole Tense–Mood–Aspect 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. 235–53. 1993b, ‘The African presence in Caribbean French colonies in the seventeenth century: Documentary evidence’, Travaux de recherche sur le créole haïtien 16–17: 1–236 (Université du Québec à Montréal). 1995, ‘The demographics of creole genesis in the Caribbean: A comparison of Martinique and Haiti’, in Arends (ed.), pp. 203–32.

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: 185–231. 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. 113–29. Smith, Norval, Robertson, Ian and Williamson, Kay 1987, ‘The Ijo element in Berbice Dutch’, Language in Society 16: 49–89. 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. 119–42. 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 française du XVIIe siècle, 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 différentes caractéristiques de pou en créole haïtien’, Travaux de recherche sur le créole haïtien 3: 1–34 (Université du Québec à Montréal). 1989, ‘Les caractéristiques de pou: Un modal en position de complémenteur’, in Lefebvre and Lumsden (eds.) (1989c), pp. 131–47. 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 créole haïtien: 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. 167–89. Taylor, Douglas 1963, ‘The origin of West Indian creole languages: Evidence from grammatical categories’, American Anthropologist 65: 800–14. 1971, ‘Grammatical and lexical affinities of creoles’, in Hymes (ed.), pp. 293–8. 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 français parlé, Quebec: Département de linguistique, Université Laval. Thomason, Sarah G. 1993, ‘On identifying the sources of creole structures. A discussion of Singler’s and Lefebvre’s papers’, in Mufwene (ed.), pp. 280–95. Thomason, Sarah G. and Kaufman, Terrence 1991, Language Contact, Creolisation, and Genetic Linguistics, Berkeley/Los Angeles/Oxford: University of California Press. Thurot, Charles 1881–3, De la prononciation française depuis le commencement du XVIe siècle d’après le témoignage 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, Université du Québec à Montréal. 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 l’UQAM 1: 132–49. 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. 157–74. Tremblay, Mireille 1990, ‘French possessive adjectives as dative clitics’, WCCFL 8: 399–413, 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 d’Hayti, 1811, Idylles et chansons, ou essais de la poësie créole, 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 créole: structure, statut et origine, Paris: Librairie C. Klincksieck. Valdman, Albert (ed.) 1979, Le français hors de France (Créoles et français régionaux), Paris: Honoré Champion. Valdman, Albert 1980, ‘Creolisation and second language acquisition’, in Valdman and Highfield (eds.), pp. 297–311. 1993, ‘Matrice et diffusion culturelle dans le développement des créoles français’, paper read at the Université du Québec à Montréal. Valdman, Albert and Highfield, Arthur R. (eds.) 1980, Theoretical Orientations in Creole Studies, New York: Academic Press. Valdman, Albert et al. 1981, Haitian Creole–English–French 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, Phonétique historique du français depuis les origines jusqu’à 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 1869–70, ‘Contribution to creole grammar’, Transactions of the American Philological Association, 1869–70: 123–67. 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 considération spéciale du dialecte de Nibas, Abbeville: F. Paillart. Vaugelas, Claude Favre de 1647, Remarques sur la langue francoyse utiles à ceux qui veulent bien parler et bien écrire (nouvelle édition 1981), Paris: Champ Libre. Védrine, 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. 170–93. 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. 303–18. Vendler, Zeno 1967, Linguistics in Philosophy, Ithaca, NY: Cornell University Press. Vincent, Diane 1984, ‘Les ponctuants de la langue’, Doctoral Dissertation, Université de Montréal. Vinet, Marie-Thérèse 1991, ‘Observations théoriques sur les explétifs: le cas du créole haïtien’, Revue québécoise de linguistique 20: 195–214. Voorhoeve, Jan 1973, ‘Historical and linguistic evidence in favor of the relexification theory in the formation of creoles’, Language in Society 2: 133–45. Wagner, R. L. and Pinchon, J. 1962, Grammaire du français classique et moderne, Paris: Hachette. Wallace, Efoé 1995a, ‘L’Éwé: temps, mode et aspect’, in Lefebvre (ed.) (1995a), volume II, 54 pages. 1995b, ‘Le présent progressif et le futur: les marqueurs’, in Lefebvre (ed.) (1995a), volume II, 14 pages. 1995c, ‘Le processus de la relexification: Le cas de la négation en Éwé dans le créole Haïtien’, Université du Québec à Montréal, MS. 1995d, ‘Le groupe nominal en Éwé’, 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, Französisches Etymologisches Wörterbuch: Eine Darstellung des Galloromanischen Sprachschatzes, Basel: Zbinden Druck. Washabaugh, William 1975, ‘On the development of complementisers in creolisation’, Working Papers on Language Universals 17: 109–40. 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. 91–115. 1977, ‘Lingua franca: Historical problems’, in Valdman (ed.), pp. 295–310. White, Lydia 1996, ‘Universal grammar and second language acquisition: Current trends and new directions’, in Ritchie and Bathia (eds.), pp. 85–120. 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 français canadien’, in Horiot (ed.), pp. 115–23. 1994, ‘Postface’, in Mougeon and Beniak (eds.), pp. 325–32. Woolford, Ellen 1983, ‘Introduction: The social context of creolisation’, in Woolford and Washabaugh (eds.), pp. 1–9. Woolford, Ellen and Washabaugh, William (eds.) 1983, The Social Context of Creolisation, Ann Arbor, MI: Karoma. Yvon, H. 1951, ‘Cil et Cist, articles démonstratifs’, Romania 72: 145–81. Zhang, Shi 1990, ‘Correlations between the double object construction and preposition stranding’, Linguistic Inquiry 21: 312–16. Zink, Gaston 1991, Phonétique historique du français, Paris: Presses Universitaires de France. Zribi-Hertz, Anne 1995, ‘Emphatic or reflexive? On the endophoric character of French lui-même and similar complex pronouns’, Journal of Linguistics 31: 333–74. Zwicky, Arnold M. 1976, ‘On clitics’, in Dressler and Pfeiffer (eds.), pp. 29–39.

452

Index of authors

Index of authors

Abney, S. 103, 412 Adam, L. 3 Adjemian, C. 34 Àgbèdvr, P. K. 210–11, 214 Alembert, Jean Le Rond d’ 127, 411 Allen, M. 329, 334 Alleyne, M. C. 2–3, 5–6, 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, 177–8, 191, 411, 413–14, 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, 135–6, 210, 213, 280, 353, 413 Awoyale, Y. 160, 167, 169–70 Baker, M. C. 60, 356, 392, 422 Baker, P. 2, 11, 42, 68, 80, 397, 409 Bakker, P. 21–4, 28–9, 330, 409, 421 Bá0gbó-é, 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, É. 62, 410 Bennis, H. 349 Bentolila, A. 134 Bernabé, J. 139 Bernstein, J. B. 88, 411 Bertho, J. 58 Bickerton, D. 2, 4, 6, 10, 12, 42, 49–50, 66, 72–4, 111, 140, 276, 351, 355, 373, 396 Bigelow, J. 397 Bloomfield, L. 63

452

Bonnard, H. 399–400, 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, 88–9, 100–3, 107, 142–5, 149–51, 153, 155–6, 159, 163, 167–9, 171–3, 175–7, 179–81, 211, 253–8, 278–80, 303–11, 313–14, 317–20, 326–7, 330–2, 334–6, 338–48, 384, 398, 401, 408, 411, 414–15, 417, 420–2 Brunot, F. 79, 96, 169, 411 Bruyn, A. 380 Bryan, M. A. 59–60, 326, 329, 332, 410 Butel, P. 53 Byrne, F. 186, 355, 384 Cadely, J.-R. 151, 153–4, 398 Capo, H. B. C. 58–9, 69, 142, 401, 414 Carden, G. 36, 68–9, 159–60, 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, 67–8, 96, 144–5, 169, 176, 181, 335, 338, 411, 415 Chomsky, N. 7–8, 18, 41, 73, 77, 105, 163, 196–7, 200, 221, 237, 259, 266, 272, 274, 282, 288, 296–7, 349, 351, 408, 412, 415–16 Clements, G. N. 147 Collins, C. 82, 203, 210–11, 221, 364–6, 369, 371, 416, 422–3 Comrie, B. 115–16, 120–3, 134, 147, 238 Corne, C. 2, 68, 159, 162, 397 Cowper, E. A. 192 Crystal, D. 223 Curtin, P. D. 54 Da Cruz, M. 210–11, 213–14, 221, 223, 241, 413–14, 416–17 Dalby, D. 58

Index of authors Damoiseau, R. 115, 134–5, 413 Darmesteter, A. 411, 421 Debien, G. 397 DeCamp, D. 421 Déchaine, R.-M. 43, 134–5, 159–61, 167–8, 356, 365, 371–2, 413 Dees, A. 96, 411 DeGraff, M. A. F. 69, 88, 151–2, 158, 166, 196, 206–10, 213, 261, 269, 271, 350–1, 353–4, 360, 367, 416, 418 Dejean, Y. 398 Den Besten, H. 74, 408 Den Dikken, M. 419 Deprez, V. 151–4, 158, 206, 261, 266, 269, 350–1, 360–3, 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 Ducœurjoly, S. J. 68, 397 Dumais, D. 129, 251–2, 261, 263–8, 283, 418 Durie, M. 13 Edmonson, J. A. 165 Emirkanian, L. 131 Fabb, N. 413–14 Faine, J. 62, 83, 255, 411 Faltz, L. M. 160, 165, 170 Fauchois, A. 307 Féraud, J.-F. 113, 127, 411 Férère, G. A. 398 Ferguson, C. A. 63 Ferraz, L. 2 Filipovich, S. 6, 9, 15, 50, 52, 303–8, 311, 314, 330–1, 365, 408, 420 Flûtre, L.-F. 83 Foley, W. A. 1, 11, 42, 48, 63, 408 Fouché, P. 399–400, 411 Fournier, R. 78–80, 83 Freeman, B. C. 311, 411 Fukui, N. 102 Furetière, A. 173–4, 181, 314–15, 332, 403, 411, 415

Goldsmith, J. 174, 177, 189 Goodman, M. F. 2–3, 24–5, 47–8, 63, 75, 83, 86–7, 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. 60–1 Grevisse, M. 113, 118, 130, 141, 143, 146, 165 Grimshaw, J. 279, 287 Gros 397 Gruber, J. S. 297 Guerssel, M. 281–2 Guiraud, P. 96, 411 Haase, A. 79–80, 141, 143, 146, 151, 173, 181, 315, 411 Hagège, C. 2, 147 Hair, P. E. H. 410 Hale, K. 288, 417 Hall, R. A. 2, 6, 63, 127, 129–30, 303, 307, 396, 411, 415, 421 Hancock, I. F. 2, 4, 48, 50 Harris, J. W. 11, 46, 409 Hazoumê, M. L. 59, 128, 138–9, 144, 146, 163, 210, 212, 412 Heine, B. 41–3 Heinl, N. G. 53 Heinl, R. D. 53 Herskovits, M. 66 Hesseling, D. C. 2 Highfield, A. R. 42 Hilaire, J. 57, 410–11 Hilliard d’Auberteuil, M. R. 397 Hirschbühler, P. 174, 177, 189 Hoekstra, T. 102 Holm, J. 3, 6, 33, 138–9 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. 1–2, 4–6, 48, 50 Imbs, P. 411

Galet, Y. 149, 411 Gambhir, S. K. 409 Gauthier, P. 151, 411 Gendron, J.-D. 399–400 Ghomeshi, J. 192 Gilles, R. 205, 355, 384 Giurescu, A. 421 Givón, T. 60, 392 Göbl-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. 399–400, 411 Joseph, F. 411 Joseph, J. S. 215, 384 Jourdan, C. 13 Julliand, A. 317 Juneau, M. 62–3, 399, 411, 420 Kant, I. viii Kaufman, T. 1–2, 10, 12, 25, 30, 32, 48, 409 Kay, P. 4, 66 Kaye, J. 9, 15, 66, 72, 408, 410 Kayne, R. S. 148–9, 151, 185, 189, 193–4, 287, 290, 357–8, 409, 412 Keenan, E. 415 Keesing, R. M. 5, 30–2, 39, 48, 380, 396 Keyser, J. 288, 417 Kihm, A. 37–8, 84, 133, 181 Kinyalolo, K. K. W. 82, 146–7, 159–60, 163, 165, 167–9, 185–6, 203, 272, 413–14, 416, 418 Kiple, K. 54 Koelle, S. W. 61 Koopman, H. 9, 11, 39–40, 42, 60–1, 73, 82, 147, 158, 167, 173–4, 176–8, 180, 190, 192, 195–8, 203–5, 249, 251–2, 266, 270–3, 276–8, 287, 291, 351, 363–5, 367, 371, 373, 387, 410–12, 415–16, 422–3 Kouwenberg, S. 36, 39, 155, 330, 390, 421 Kratzer, A. 366 Labelle, M. 76 Laberge, S. 11, 13, 17, 42–3, 127–8 Laborderie, N. 399–400, 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. 288–9, 299, 369, 419, 422 Lasnik, H. 288, 290, 292, 294, 419 Laughren, M. 417 Law, P. 158, 198–202, 261, 266, 269, 350–1, 364–6, 368–9, 416, 418, 422–3 Lefebvre, C. 4, 6, 9–12, 15–16, 18, 30, 35–40, 42, 44–6, 48, 50, 52–3, 57, 66–7, 69–70, 72, 79–85, 87–99, 102, 104, 107, 109, 111–14, 116–17, 120–37, 149–50, 173–4, 177, 181, 184–5, 189, 191–2, 194–8, 200, 203–7, 219–22, 224–41, 243–6, 248, 272–3, 284, 287–8, 291–301, 303–8, 311, 314, 330–1, 333, 341, 355, 357–8, 363, 371, 383–6, 389, 395–6, 408–10, 412–17, 419–20, 422–3 Levin, B. 288, 309, 315, 420

Levinson, S. C. 165 Li, Y.-H. A. 419 Lichtenberk, F. 40 Lieber, R. 9, 39–40, 303, 307, 309, 326, 331, 334, 408 Lightfoot, D. 2, 11, 41–2, 156 Lipou, A. 60 Lommatzsch, E. 96 Lord, C. 41, 43, 186 Lumsden, J. S. 4, 9–12, 15–16, 18, 30, 34–40, 44–6, 48, 50, 57, 67, 69, 72, 82, 87, 101–9, 128, 134–5, 195–6, 206, 242–3, 260–1, 287–90, 336, 340–1, 364–5, 385, 389, 408–13, 416–19, 423 Magloire-Holly, H. 412 Man, M. 58 Manessy, G. 128 Manfredi, V. 109, 159–61, 167–8, 196, 364–5, 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-Lübke, W. 62 Miller, G. A. 421 Milner, J.-C. 79–80, 88, 103, 185 Mintz, S. W. 2 Möhlig, W. J. G. 25 Moisy, H. 421 Moody, M. D. 421 Moreau, M.-L. 193–4 Moreau de Saint-Méry, L.-É. 68 Morin, Y.-C. 162, 164–5 Mouchon, P. 127 Mougeon, R. 62, 410 Mous, M. 25–9, 34, 61, 384, 409 Mufwene, S. S. 2–6, 10–11, 36, 38, 44, 46, 49, 60, 74, 86, 144, 147, 166, 276, 373, 390, 409 Mühlhäusler, P. 4, 11, 38, 42, 44, 46, 49–50, 333, 409 Muysken, P. C. 6, 9–10, 16–21, 28–9, 34, 42, 67, 72, 74, 155, 159–60, 162, 168, 171, 182, 224, 255, 303, 356, 378, 384, 390, 394, 396, 408–9, 412, 417, 422–3 Naro, A. J. 5, 34, 63 Ndayiragije, J. 180, 364, 366, 422–3 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. 399–400, 411 Pinchon, J. 411, 415 Pinker, S. 287–8 Piou, N. 364–6, 423 Plag, I. 42, 192 Plank, F. 165 Poirier, C. 62 Pollock, J.-Y. 189, 208–9, 351–2 Pranka, P. M. 395 Rappaport, M. 288, 309, 315, 420 Rassinoux, J. 178, 336–8, 341, 411 Reh, M. 41–3 Rey, A. 174, 317, 411 Rickford, J. R. 11, 42 Ritchie, W. C. 35, 422 Ritter, E. 82, 86–7, 269, 364–6, 412, 416, 418, 422–3 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, 42–5, 127–8, 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, 177–8, 182, 211, 214, 336–7, 383, 411, 416–17 Seiter, W. 268 Selkirk, E. 303, 346 Seuren, P. 6

455

Siegel, J. 5, 11, 13, 33–4, 46, 409 Singler, J. V. 1–2, 32, 43, 47, 52–8, 73–5, 171, 390–1, 409–10 Smith, N. 6, 33, 36, 39, 67, 72, 74, 159–60, 162, 168, 171, 182, 186, 207, 255, 408–9, 421 Spagnolo, L. M. 43 Spears, A. K. 112, 123, 130, 412 Speas, M. 102 Spillebout, G. 70, 411 Sportiche, D. 147–8 Sproat, R. W. 334, 395 Sterlin, M.-D. 184–5, 271–5, 351, 415 Stewart, O. T. 356, 422 Stewart, W. A. 9, 68–9, 159–60, 162, 168, 170, 255, 259, 410, 417 Stowell, T. 297 Syea, A. 11, 42 Sylvain, S. 3, 48, 81, 86–7, 90, 100, 114, 118, 129, 168, 170, 307, 396–7, 411, 421 Szabolcsi, A. 103, 412 Taylor, D. 159, 181 Tenny, C. 134, 230, 234, 288 Thibault, P. 70 Thomason, S. G. 1–2, 10, 12, 25, 30, 32, 48, 67, 73–4, 409 Thurot, C. 400, 411 Tinelli, H. 303, 307, 398 Traugott, E. C. 42, 395 Travis, L. 18, 39–40, 105, 221, 279, 412 Tremblay, M. 103, 287, 290, 358, 412, 419 Trudgill, P. 11, 46, 409 Valdman, A. 2, 10–11, 42, 62–3, 83, 90, 112, 129, 303, 397–8, 411, 415, 421 Valdman, A. et al. 64, 129, 141, 168, 173, 181–2, 261, 297–8, 311, 316, 336–8, 340, 383, 403, 411, 415, 419, 421 Valois, D. 88 Van Daele, H. 399–400, 411 Van Name, A. 2, 5 Van Riemsdijk, H. C. 39, 73–4, 412 Van Voorst, J. 234 Vasseur, G. 414 Vaugelas, C. F. de 149, 411 Védrine, E. W. 308–11, 341, 411, 419 Veenstra, T. 74, 287, 289–90, 414, 419, 421 Vendler, Z. 413 Vergnaud, J.-R. 412 Vincent, D. 70 Vinet, M.-T. 158, 206, 269, 418 Voorhoeve, J. 2–3, 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, 58–60, 86, 326, 329, 332, 410–11, 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. 399–400, 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 58–9, 138–9, 144 Akan cluster 50, 55, 59, 128, 170, 272 Anyi-Baule 59 Bambara 60, 138–9 Bantu 19, 25, 27, 55, 60–1, 390–2, 410, 422 Bari 43–5 Basque 160, 214 Bassa-Nge 170 Berbice Dutch 36, 39, 155, 166, 330, 385, 390, 408, 421 Bhojpuri 46 Bini 59

Gungbe 58–9, 138–9, 144, 212, 371 Gur 55, 60, 391, 410 Hausa 109 Ifé 59 Igbo 50, 109, 128, 170, 371–2 Ijo (Eastern) 55, 155, 390, 408 Indian Ocean creoles 126 Jamaican creole 39, 147

Caribbean creoles 9, 39, 73–4, 330, 356, 378, 420 Cigbe 58 Cree 19, 21–4, 409 Cushitic 19, 25–7

Kikongo 3, 57 Kituba 3 Klao 43 Kogbe 58 Kriol 38 Kru 43, 60 Kwa 55–7, 59–61, 66, 160, 167, 170, 210, 214, 326, 329, 332, 353, 356–7, 371, 373, 390–2, 422 Kwaio 30–1

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 17–18, 30–1, 39, 41, 46, 74, 89, 103, 120, 123, 160, 165–6, 199, 201, 255, 266, 280, 293, 297, 299, 330, 346, 351–2, 357–9, 364, 419–20 Ewe 3, 43, 58–9, 76, 84, 86, 133, 138, 211–14, 216–17, 326, 332–3, 365, 371, 412, 423

Ma’a (see Inner Mbugu) Malagasy 415 Malinke 60, 332 Mande cluster 55, 60, 75, 86, 109, 332, 391, 410 Mandingo 86–7, 109 Mandinka 50 Manjaku 37 Martinican creole 47, 139, 379 Mauritius creole 2, 415 Mayan 75 Mbugu Inner Mbugu 19, 24–8 Normal Mbugu 19, 25–7 Media Lengua 19–21, 23, 28 Michif 19, 21–4, 28

Ga-Dangme cluster 59 Gbe cluster 43, 55–9, 66, 69, 74, 128, 138–9, 144, 155, 160, 163, 165, 167, 170, 186, 203, 210–12, 258, 326, 353, 371, 390–1, 410, 412–14 Gengbe 58, 138–9, 144, 371, 410 Guang 59 Gullah 147, 166

Nago 59 Niger–Congo 55, 57 Nupe 59, 326

457

458

Index of languages and language families

Palenquero 216 Papiamentu 147 Pare 25, 27 Phlaphera 58 Picard 151, 412–14 Portuguese 38 Portuguese creole 147, 408

Tok Pisin 13, 17, 38–9, 43–5, 127–8, 396, 409 Tolai 409 Turkish 17 Twi-Fante 59, 326

Quechua 17, 19–21, 417

Vai 332 Vata 252, 272, 367 Vietnamese 381

Reunion creole 2 Sango 43 Saramaccan 3, 36, 74, 186–7, 207, 330, 347, 379, 384–5, 409, 414, 421 Shambaa 25, 27 Solomons Pidgin 30–3, 39–40, 396 Spanish 19–21, 24, 148 Sranan 3 Swahili 27 To’aba’ita 40 Twfingbe 59, 138–9, 144

Urhobo 170

Wemεgbe 58 West Atlantic 55, 60, 391 Wolof 50, 109, 128 Xlwagbe 59 Xula 371 Xwèdagbe 58–9, 138–9, 144, 212 Yoruba 41, 50, 59–60, 86–7, 109–10, 128, 138, 170, 326, 356, 364–5, 372 Yuman languages 160

Index of subjects

459

Index of subjects

adverbial clauses see causal and temporal clauses anterior markers 112, 114, 116–18, 124–5, 130, 132, 138 assertive markers 221–9 bare sentences (interpretation of ) 114–15, 134–7, 139–40 BODY-state verbs 250–1 calquing 5, 15, 30, 32 Case Case assigning properties of verbs 283–6, 297–301 Case Filter 18, 101–10, 273–6 Exceptional Case marking 273–6 Genitive Case 101–14, 106–10, 144–6, 168, 357–60 Objective Case 102, 109–10 causal clauses 363–71 clause types see adverbial clauses; causal clauses; complement clauses; factive clauses; purposive clauses; relative clauses; temporal clauses cleft constructions 193–202, 363–73 clitics clitic climbing 148, 150, 152 pronominal clitics 148–57, 182, 349–51 cognate object verbs 280–3 complement clauses finite 184–93, 271–2 non-finite 271–6 complementisers introducing complements of verbs of the say-class 184–7 introducing complements of verbs of the want-class 187–93 introducing infinitival complements in the context of extracted subjects 271–6 compounds principles of concatenation of words into 47, 334, 348 semantics of 12, 334–9 types of 342–8 word order in 339–42

conflation (phonological) 37–8, 84, 181 conjunction of clauses 205 of nps 206 control verbs three-variable control verbs 276–8 two-variable control verbs 271–6 creole languages features of 1–4, 12, 71 pidgins and creoles as pcs 4 versus mixed languages 29–30 creolisation abrupt 32 and second language acquisition 29, 33–6, 72 data Fongbe 65–6, 69, 75–6 French Montreal 70 seventeenth and eighteenth century 69 spoken by the colonists 62–5, 70 Haitian 68–9, 75–6 definite determiner in nominal structure 78–86, 88–95, 97–9, 104, 106, 110, 134–5 in the clause 219–47, 287–97, 369–71 definite future markers 112–14, 120, 123–9, 132–3, 138 demonstrative terms 78, 89–92, 94, 96–100, 104 derivational affixes 303–33, 403–7 desemanticisation 41–2 dialect levelling cases of 47, 87, 110, 137–9, 170–1, 253–62, 323–33 definition of 46–7, 393–4 dialects (regional) Fongbe 58–9, 75 French 62–3 Haitian 46, 75 double-object verbs 287–301, 357–60 Empty Category Principle (ecp) effects 193–203 Exceptional Case marking 274–5, 278

459

460

Index of subjects

existential verbs 269–71 expansion 6, 50 expletives 157–9, 259–62 factive clauses 82, 203–5, 363–71 filters (surface) 240–6 finite/non-finite issue in creole studies 271–8 focus markers 206–8 genetic affiliation of creole languages 396 grammaticalisation 42 habitual marker 136–9 Haiti early African population in 53–7, 65–6 languages spoken at the time of creole genesis 55–8 the colonists 53, 62–3 Haitian phonological forms and French phonetic matrices (correspondences of ) 398–402 homogeneity of the substratum languages 58–62, 67, 390–3 imperfective marker 112–14, 119–23, 128–9, 137–9 indefinite determiners 78, 87–9, 91–5, 97–9, 104, 106, 110 indefinite future markers 112, 129–33 inherent object verbs 280–3 innovations in Haitian 12, 51, 87–9, 206–8, 269–71, 283–7, 323–33, 349–51 insistence markers 213–17 language intertwining 22 light verbs 278–80 logical form 227–9, 293–7 logophoric pronouns 147–8, 182 markedness issue 72–5 marked/unmarked structures 73–4, 262–9, 363–74 mixed languages 19–30 morphological head 330–2 nativisation 4 negation markers 208–11 negative markers 213–17 negative quantifiers (interpretation of ) 360–3 null subject parameter 349–51 operator in factive clauses 203–5 in relative clauses 203–5 orthographic conventions Fongbe 401–2 French 399–400 Haitian 398–9

paralexification 27, 384–5 parameters parameter setting in creole genesis 8, 47, 68, 349–75, 386–8 see double-object verbs; (interpretation of ) negative quantifiers; null subject parameter; serial verbs; verb doubling phenomena; verb raising parameter personal pronouns 84–7, 104, 110, 141–3, 160, 182 phonological inventories Fongbe 401–2 French 399–400 Haitian 398–9 pidgin languages expanded pidgins 4, 32 features of 1, 4, 63–4 pidgins and creoles as pcs 4 plural markers 78, 84–8, 90–1, 98, 104, 110, 173 possessives 26, 78, 90, 103, 106, 109, 143–7, 183 predicate cleft construction 59, 61, 73–4, 363–73 pronominal clitics 148–57 pronominal forms see expletives; logophoric pronouns; personal pronouns; possessives; pronominal clitics; reflexives; resumptives purposive clauses 187–93 question markers 211–13 raising verbs 262–6 reanalysis cases of 43–4, 87, 128–9, 184–93, 376, 378 definition of 41–6, 137 related phenomena 11, 41–2 reduction 5–6, 50 reflexive verbs 253–9 reflexives anaphor 160–7, 183 body-part 26, 159–60, 167–70, 183 relabelling definition of 16 of minor category lexical entries by a null form 17–18, 37–8, 44–5, 50, 78, 108–10, 128, 164, 375–86 of minor category lexical entries by an overt form 17–18, 109, 123, 129, 142, 156, 180–1, 375–86 relative clauses 80, 82, 90, 93, 98, 179, 193–205, 365 relexification and dialect levelling 46–7 and reanalysis 41–6 in creoles 30–41

Index of subjects in mixed languages 19–30 in pidgins 30–3 in second language 30–41 representation of 15–19, 22, 27, 384–6 resumptives in the context of extracted subjects 193–202 in the context of raising verbs 262–9 semantic interpretation see (interpretation of ) bare sentences; cleft constructions; (principles of concatenation of words into) compounds; derivational affixes serial verbs serial verb parameter 355–7 simplification 5–6, 50 stabilisation 4 subject raising 262–9 subjunctive markers 112, 118–19 syntactic categories major or lexical 18, 21, 23–4, 28–9, 36–7, 44, 49 minor or functional 18, 24, 28–9, 31, 36–7, 44, 48–9, 59, 64–5, 67 temporal clauses 363–71 tma markers see anterior markers; definite future markers; habitual marker; imperfective marker; indefinite future markers; subjunctive markers transfer 15, 33–4

461

transmission and acquisition in creole genesis 8, 12, 349–75, 395–6 verb doubling phenomena 82, 363–74 verb raising parameter 61, 351–5 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 259–62 weather verbs 251–3 Whmovement in predicate cleft 363–73 movement in questions 176–9, 193–203 phrases 171–82 words 171–82 word order affix order 180, 330–2 establishing word order in creole genesis of functional categories 39–40, 89 establishing word order in creole genesis of lexical categories 38–40, 89, 180 evaluation of the hypothesis 388–90 in compounds 339–42 in double-object constructions 293–7 in nominal structure 78 in the clause 184–246