Creolization and Language Change 9783111339801, 9783484303171

Citation preview

Linguistische Arbeiten

317

Herausgegeben von Hans Altmann, Peter Blumenthal, Herbert E. Brekle, Gerhard Heibig, Hans Jürgen Heringer, Heinz Vater und Richard Wiese

Creolization and Language Change Edited by Dany Adone and Ingo Plag

Max Niemeyer Verlag Tübingen 1994

Die Deutsche Bibliothek - CIP-Einheitsaufnahme Creolization and language change / ed. by Dany Adone and Ingo Plag. - Tübingen : Niemeyer.1994 (Linguistische Arbeiten; 317) NE: Adone, Dany [Hrsg.]; GT ISBN 3-484-30317-4

ISSN 0344-6727

© Max Niemeyer Verlag GmbH & Co. KG, Tubingen 1994 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere flir Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspcicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt Einband: Hugo Nadele, Nehren

Contents Preface List of contributors

VII IX

Ingo Plag and Dany Adone Introduction

1

Ingo Plag Creolization and language change: a comparison

3

Dany Adone Creolization and language change in Mauritian Creole

23

Pieter Muysken and Norval Smith Reflexives in the creole languages: an interim report

45

Philip Baker Creativity in creole genesis

65

Anand Syea The development of genitives in Mauritian Creole

85

Tonjes Veenstra The acquisition of functional categories: the creole way

99

Daniel Veronique Naturalistic adult acquisition of French as L2 and French-based creole genesis compared: insights into Creolization and language change?

117

Gilette Staudacher-Valliamee Eine synchron-dynamische Phonologic des Reunion Creole als Ausgangspunkt zur Annäherung an Kreolisierung und Sprachwandel

139

Preface The present book contains the proceedings of the workshop on "Creolization and language change", organized by the editors and held as part of the 15th annual conference of the Deutsche Gesellschaft für Sprachwissenschaft at Jena, March 3-5, 1993. Due to various reasons, some of the participants were prevented from submitting a written version of their talk to appear in this book: Jacques Arends & Liliane Adamson, Hans den Besten, Norbert Boretzky, Adrienne Bruyn, and Petra Thiele. Due to other reasons, Philip Baker and Anand Syea were prevented from attending the workshop, but have agreed to contribute to this volume. We are grateful to Richard Wiese, one of the editors of this series, and to Rüdiger Zimmermann for their support. Special thanks goes to Hauke Börngen for his help with the manuscripts. The editors

Contributors Dany Adone University of Sydney Dept. of Linguistics Sydney, 2006 New South Wales Australia tel: 2-6924348 fax: 2-5521683 e-mail: [email protected]

Philip Baker University of London S.O.A.S. (African Dept.) Thomaugh St. GB-LondonWClHOXG fax: 71-4363844

Pieter Muysken Universiteit van Amsterdam Instituut voor Algemene Taalwetenschap Spuistraat210 NL-1012 VT Amsterdam tel: 020-525 e-mail: [email protected]

Ingo Plag Philipps-Universität Marburg Institut fur Anglistik und Amerikanistik Wilhelm Röpke-Str. 6 D D-35032 Marburg tel: 06421-285560 fax: 06421-287020 e-mail: [email protected]

Norval Smith Universiteit van Amsterdam Instituut voor Algemene Taalwetenschap Spuistraat210 NL-1012VT Amsterdam tel: 020-525 e-mail: [email protected] Gillette Staudacher-Valliamee L.A.C.I.T.O./C.N.R.S. 44, Rue de L'Amiral Mouchez F-75014 Paris Anand Syea University of Westminster Faculty of Law, Language, and Education Euston Centre GB-LondonNWl 3ET e-mail: [email protected] Tonjes Veenstra Universiteit van Amsterdam Instituut voor Algemene Taalwetenschap Spuistraat 210 NL-1012 VT Amsterdam tel: 020-525 e-mail: [email protected] Daniel Veronique Universite de Provence 29 Avenue Robert Schumann F- 3621 Aix-en-Provence Cedex tel: 42-592271 fax: 42-206487

Ingo Plag and Dany Adone

Introduction The relationship between creolization and language change lies at the heart of creole studies from its beginnings in the 19th century up to the present day. This volume contains articles that tackle old and new problems like the substrate/ universale issue, the problem of continuity of transmission, the gradual creolization hypothesis, and the role of first and second language acquistion in creolization. In his article, Plag discusses the problem of continuity of transmission and addresses the question whether we can, on strictly linguistic grounds, differentiate between linguistic innovations as they typically occur in creolization, and developments as they can be observed in 'natural' language change. On the basis of cross-linguistic data it is argued that, in terms of linguistic mechanisms at work, creolization is not fundamentally different from canonical cases of language change in non-creole languages. In her diachronic study of Mauritian Creole, Adone comes to the opposite conclusion. She investigates the development of null subjects, reflexives, and relativization and points out that her data support the idea that creolization is instantaneous, while language change is gradual. The contributions by Muysken and Smith, and Baker raise new questions about the traditional substratist and universalist approaches to creolization. In their crossCreole study of reflexives and their emergence, Muysken and Smith ascertain that neither of the current theories of creolization (superstate, substrate or universals) can straightforwardly account for the cross-creole similarities and differences, nor can they explain some peculiar developments in individual Creoles. Baker's study of the Mauritian creole lexicon (including phonological, etymological and morphological aspects) is an elaboration of his 'creativist' theory of creolization. He states that earlier theories are fundamentally flawed in that they regard creolization as the unsuccessful attempt to approach some target language and argues that pidgins and Creoles are "successful solutions to problems of human intercommunication rather than the unhappy consequences of botched language learning or failed language maintenance" (p. 66). A nativist perspective is taken by Syea and by Veenstra. Syea's article is a study of the development of two different genitive constructions in Mauritian. He presents evidence against earlier analyses that explained the development of one of the constructions as the result of substrate or superstate influence. In his view, the development is best accounted for by assuming a language-internal development, guided by independent principles of Universal Grammar. Such principles are also evoked by Veenstra to account for the development of functional categories in

2

Plag and Adone

Saramaccan. He argues that Saramaccan is a null subject language, and that this feature is an outcome of the creolization process itself, guided by a principle of first language acquistion. Veronique addresses the role of mechanisms of second language (L2) acquisition in creolization by comparing natural L2 acquisition data of L2 learners of French with French-based Creoles. The results show both similarities and differences between the two. However, as the author himself points out, these results should be interpreted very carefully, since the social context of natural L2 acquistion not comparable to a typical creolization setting. Staudacher-Valliamee's article is a minute study of Reunion Creole phonology. Her analysis of some 30 idiolects shows a complex interaction of internal developments and contact-induced changes in the creolization and later development of Reunion Creole. The diversity of opinions expressed by the contributors to this book shows that the issues discussed in their articles are still far from being settled. The editors of the present volume thus continue a good tradition among creolists, namely to agree to disagree.

Ingo Plag

Creolization and language change: a comparison 1. Introduction The last two decades have seen an increasing interest by linguists in the languages commonly called Creoles, and in the processes through which these languages emerge. One of the reasons for this trend is the assumed significance of the study of Creole languages for various sub-fields of linguistics and for general linguistic theory.1 The present article is concerned with what historical linguistics may learn from the study of creolization and vice versa. The formation of a Creole language has conveniently been labelled 'creolization1, but the mechanisms of this process are still a matter of controversy. One widely held view of creolization is that it denotes the lexical and grammatical expansion of a rudimentary pidgin into a full-fledged language. There is, however, no consensus about the mechanisms involved in this process. Relexification, the transmission of substrate features, the transmission of superstratum features, processes of first language acquisition, and processes of second language acquisition have all been advocated as being central to creolization. The only point on which all researchers agree is that drastic changes occur when pidgins become Creoles. The nature of these changes is, however, not very clear. The relationship between creolization and language change has been in the focus of creole studies since Schuchardt cited Creole languages as evidence against neogrammarian doctrine. In his debate with Meillet (cf. e.g. Meillet 1914, Schuchardt 1917), Schuchardt argued that the mixed nature of creole languages invalidates the genetic model, whereas his opponent regarded Creoles as direct descendants of their lexifier language. According to Schuchardt, Creoles cannot be classified in language family trees, because they have no direct genetic link to their lexifier and substratum languages. Today, the issue of creolization and language change is still far from being settled, as can be seen from the fact that it plays a major role in the current debate on gradual vs. sudden creolization. Recent diachronic studies of Haitian Creole (Garden & Stewart 1988), and Sranan (Arends 1989, Plag 1993, Plag 1994) have tried to show that some of the syntactic developments that are generally believed to be part of the creolization process take much more time than previously conceived. Contrary to earlier claims by scholars like Bickerton, the aforementioned authors conclude on the basis of their findings that creolization is not a sudden, single1

This interdisciplinary interest is, for example, reflected in the title of Thomason & Kaufman's widely-praised book Language contact, creolization and genetic linguistics (1988).

4

Plag

generation process, but a gradual one, lasting for several generations.2 In his critique of Arenas (1989) and Garden & Stewart (1988), Bickerton defends his position by stating that the changes observed by these authors may not be regarded "as part of the creolization process" (1991:54), but rather should count as subsequent later "natural changes" of the languages under discussion (see also Bickerton 1981:53).3 Bickerton thus assumes that there must be a fundamental difference between what he calls natural changes and creolization. His main argument for rejecting the changes observed by Arends and Garden & Stewart as not being part of creolization is that "these types of change are well attested in languages that are not and probably never have been Creoles." (Bickerton 1991:54).4 This argumentation logically entails that changes that are part of the creolization may only be found in Creoles. In other words, Bickerton tries to establish a definition of creolization which is based on the assumption that the types of change occurring in creolization cannot be found in the development of other languages. Thus, Bickerton maintains that the linguistic mechanisms of creolization are unique and do not parallel developments in non-creole languages. We will refer to this position as the 'uniqueness hypothesis'. An argument against the uniqueness hypothesis has been put forward by Thomason and Kaufman (1988), who characterize 'abrupt creolization' as a type of contact-induced language change, namely as the emergence of a new language through language shift and imperfect learning of the target language, involving borrowing and interference to varying degrees. Bickerton & Givon (1976) have shown parallels between pidginization and natural change. They argue (1976:33) that the changes occurring "in the 'forced1 context of pidginization, and [...] the kind of'spontaneous' change that takes place diachronically in non-pidginized languages [...] are essentially similar and both are governed by possibly universal linguistic principles." In his later work Bickerton rejects a similar parallelism between creolization and natural change. The present article is an attempt to show that Bickerton & Givon's conclusion with regard to pidginization must be extended to the process of creolization. The uniqueness hypothesis will prove to be untenable. We will compare developments that are assumed to be significant of creolization with documented cases of similar The gradualist position has been argued before by Goodman (1985), Mühlhäusler (1986) and Lefebvre (1986), among others. The studies by Garden & Stewart and Arends are, however, the first to use diachronic pidgin/creole data to prove their point. In this article I will adopt the convention to speak of 'natural change' when referring to language change as it occurs in non-creoles. The terminology does not entail that creolization is something unnatural. Note that Bickerton's arguments refer to the strictly linguistic aspects of creolization, and not to socio-historical aspects. In the following we will also concentrate on strictly linguistic, i.e. structural phenomena.

Creolization and language change

5

changes in non-creole languages. In particular we will look at cross-linguistic studies of grammaticalization which propose universal continue of development. From this comparison it will become clear that there is no fundamental difference between natural change and creolization. Instead, in both processes the linguistic mechanisms of change are by and large the same and seem to be a reflex of universal linguistic principles. The refutation of the hypothesis that there is a difference in principle between creolization and natural language change does, however, not mean that no differences can be observed. They seem, however, to be a matter of quantity rather than of quality. The article is structured as follows: in the next section I will clarify some of the theoretical problems involved with our topic, section 3 reviews the developments of several grammatical properties in pidgin/creole and non-creole languages, and in section 4 the results are summarized and discussed.

2. Theoretical preliminaries: some thoughts on continuity, transmission and the mechanisms of change One general theoretical problem in historical linguistics is the question of identity and continuity. When is it justified to say about two language systems that one is the descendant of the other? Consider, for example, New High German and Old High German. What makes us conceive that both languages are 'German'? Obviously, we feel that they somehow they share an identity, and one system is believed to be the continuation of the other (through intermediate stages like Middle High German). The reasons for treating both varieties as 'German' are twofold, namely social and linguistic. The varieties were spoken in roughly the same area, by the 'same people', i.e., there is continuity in terms of the speech community concerned. In strictly linguistic terms, systematic correspondences in various linguistic subsystems substantiate the relationship between the two varieties. Evidence of continuity can be found in written sources, which document the gradual development from one stage of the language to the next. If we think of longer periods of language development in Europe, as exemplified by the Romance language family, the problem is more complicated, since one language, Latin, may have several daughter languages in different places, which makes the question of identity less easy to solve. However, with the help of written sources and our knowledge of social history we can reconstruct the continuity of development. With pidgins and Creoles matters become utterly confusing. In respect to their linguistic properties pidgins seem to be very different from Creoles, and both Creoles and pidgins are very different from the superstratum and substratum languages involved in their formation. To complicate the situation further, pidgins are often

6

Plag

abandoned before they develop into Creoles, or Creoles become repidginized. These linguistic facts seem to be the direct consequences of the social instability which is characterisic of- or even necessary for - the genesis of pidgins and Creoles. So, how can we relate pidgins to Creoles? How can known principles of language change account for their relationship? The debate between Schuchardt and Meillet mentioned above was resumed by Taylor (1956), Hall (1958), and Weinreich (1958). The crucial question in this debate is what the criteria for continuity should be. Thomason (1980) shows that the restriction to exclusively linguistic facts, i.e. synchronic feature clusters, is essentially ahistorical, and argues that "continuity of transmission in a speech community should be regarded as criteria! for the establishment of genetic relationships between languages, because the notion of genetic relationship is historically vacuous unless it is based on some notion of continuity." (1980:33). Thomason defines normal transmission, which is the prerequisite for unimpaired continuity, as follows (1980:28): When we say that language B is a changed later form of language A , we certainly do not mean that all of B's lexicon and grammar were inherited from A. We do, however mean that A speakers passed their language on to their linguistic descendants, and those descendants to their descendants, and so on, until enough changes accrued in A to necessitate calling it a new language, B. Let us reconsider pidgins and Creoles in the light of this view. Two questions have to be answered. Is there a continuity in the linguistic systems? Is there a continuity of transmission with regard to the social processes involved? In answering these questions, earlier treatments have focused on the relationship of the lexifier language to the resulting Creole. In the following, we will restrict ourselves to the relationship between the pidgin and the Creole. Let us look at the linguistic facts first. For the vast majority of the now existent Creoles written sources of early stages of these languages are scarce or non-existent. The only available and reliable information is often on fairly recent stages of the Creole, with the life-cycle of the pidgin/creole being already far beyond the creolization stage. Thus, we are faced with the serious problem that, for very many of the languages under discussion, we do not have direct evidence of the changes that took place from pidgin to Creole. Thus, for lack of linguistic data, in many cases we do not know whether it is justified to speak of a continuity in the development from pidgin to Creole. Our knowledge of the nature of creolization is primarily not derived from the observation of diachronic linguistic facts but from linguistic reasoning based on comparative evidence. Such reasoning goes roughly as follows: Pidgins generally share the properties A, B, and C. In Creoles we also find A, B, and C, sometimes with a few minor changes. Some of the properties that we find in

Creolization and language change

7

the Creole, let us call them X, Y, and Z, do not occur, however, in the antecedent pidgin. From these facts we can conclude that it is the process of creolization that is responsible for the evolution of properties X, Y, and Z. Although the pidgin may differ greatly in structural terms from the Creole, there must be some bridge of continuity between the individual pidgin and its follow-up Creole. Thus, in those pidgins/creoles that are documented in written sources we find at least a shared stock of lexical items in the two systems, and strong phonetic/phonological correspondences. Somehow, these elements have survived through all turmoils of contact and disruption. Let us consider the socio-linguistic side in more detail. Mühlhäusler (1984) cites several examples of discontinuity in the development of pidgins and Creoles. However, even in his examples, one system (or many systems) somehow evolved from other systems in the same place in the same, though highly unstable, community. His examination of the development of Tok Pisin, for example, shows that there have been disruptions and discontinuities concerning the constituency of the speech communities, the languages in contact, and the social status of the vernacular. Nevertheless, even Mühlhäusler still speaks of "the development of Tok Pisin" (1984:127, my emphasis), which entails the contradiction that, while explicitly denying continuity in the development of this language, he still refers to the different vernaculars as varieties of one language, namely Tok Pis in: All these external factors have left traces in the development of Tok Pisin. It is possible to identify at least three and possibly five qualitatively different and mutually only partially or hardly intelligible varieties. (Mühlhäusler 1984:130, my emphasis) In the plantation society of the West Indies large numbers of speakers were frequently displaced, new slaves arrived continually, mortality was high, and the birth rate low, to mention some of the factors involved in discontinuities. Nevertheless, the vernaculars somehow were passed on to new speakers. People had to communicate at all times, which means that newcomers - be they slaves imported from Africa, indentured servants from Europe, or slave children born in the New World - somehow had to learn the existing vernaculars. In the typical pidgin/creole society of the Caribbean there was, of course, no transmission in the sense that a homogeneous community of native speakers passed on their language to their children, sometimes in combination with successful second-language acquisition, as was the case in many of the Indo-European languages. The language of the slave holder was not passed on to the slave, since, due to the non-availability of the target language, successful second language learning was impossible for the vast majority of the slave workers. Instead, we find what I call impaired transmission, which means that a rather rudimentary pidgin was passed on to a small number of children, and high numbers of newly arrived slaves. Hence, we cannot deny that

8

Plag

some sort of language was somehow transmitted to new speakers, though of course with a great deal of restructuring due to the disruptions and discontinutities that indeed occurred. It is the instability of the community that is responsible for the differences between standard cases of normal transmission, and impaired transmission as can be found, for example, in a plantation setting. From the point of transmission the changes and innovations to be expected in creolization should therefore be much more drastic than in normal transmission. The question of whether the developments under impaired transmission are different in kind from the ones occurring under normal transmission will be dealt with in section 3. Before we start with the actual comparison of instances of creolization and natural change, let us consider very briefly the major mechanisms of syntactic change. The two probably most important mechanisms have been labelled analogy (involving reanalysis, or reinterpretation) on the one hand and grammaticalization on the other. Without going into details we can say that analogy and reanalysis typically result in the redistribution of items or patterns to different, though not entirely new categories.5 Grammaticalization, on the other hand, may either result in redistribution or in true innovation, i.e. the creation of categories that have not existed before.6 Given that pidgin languages lack many categories characteristic of human language we may predict the frequent occurrence of grammaticalization in creolization. If so, we will see whether these processes differ from the ones encountered in non-creole languages.

3. The development of some grammatical properties in Creoles and non-creoles For the present investigation, the choice of the grammatical properties we want to investigate is crucial. Bickerton, for example, rejects Garden & Stewart's argument for gradual creolization on the grounds that a priori the development of a system of

5

For definitions, examples, and discussion of the notions of analogy and reanalysis cf, for example, Hock 1991:329-370, Anttilla 1972: chapter 5, Lehmann 1973:195ff, and the references cited in these works.

6

In simplified terms, grammaticalization is a process by which a lexical morpheme becomes a grammatical morpheme, or a partially grammatical morpheme becomes more grammatical. For an up-to-date discussion see, for example, the contributions in Traugott & Heine 1991, and the references cited therein. Very often, grammaticalization involves the reanalysis of an item as belonging to a different category. On the other hand, analogy does not entail grammaticalization. Grammaticalization thus cannot be neatly separated from analogy and reanalysis.

Creolization and language change

9

reflexive pronouns is not part of the creolization process, because the differentiation between anaphors and pronouns is not criterial for a full-fledged language.7 For the sake of the argument, we will choose only those grammatical properties that Bickerton himself (along with numerous other researchers)8 has acknowledged as being part of (and indeed crucial for) creolization, namely the tense-mood-aspect (TMA) system, complementizers, and articles. Limitations of space do not permit a broader coverage of more properties. Future research will have to tell whether the arguments presented below can be extended to other properties. We can now turn to an examination of the developments in non-creole languages in the grammatical subsystems outlined above and compare them to the pidgin/creole case.

3.1 The TMA system Pidgins lack a grammaticalized TMA system, i.e., TMA categories are not expressed by obligatory grammatical morphemes. The temporal relations between events, the internal temporal constituency of situations, and the modality of propositions are either not expressed at all and must be inferred from the context, or are optionally expressed by adverbials.9 According to Bickerton (e.g., 1981:58f), the prototypical Creole TMA system, on the other hand, has three pre-verbal markers for encoding anterior tense, nonpunctual aspect, and irrealis mood. This analysis has not gone unchallenged, and recent studies (see, for example, Bruyn & Veenstra 1993, or the articles in Singler 1990) have raised serious doubts concerning the homogeneity of the TMA system across various Creoles. As we go along, some of the objections against Bickerton's basic premise will have to be taken into account.

7

For a critical discussion of Bickerton's notion of'full-fledged language1 see Plag 1993:144-146.

8

The emergence of these properties as being indicative of creolization has been mentioned by, among many others, Romaine 1988, Mühlhäusler 1986, Washabaugh 1975)

9

Cf. e.g., Labov (1990:16) on tense: One of the most striking characteristics of most pidgins is that they have [...] no grammatical category of tense. By that I mean [...] the absence of any obligatory element in the verb phrases to identify the time reference of each predication, (original emphasis)

10

Plag

3.1.1 Aspect In the creation of a grammatical subsystem through creolization certain items are picked from the input and assigned to a newly created category.10 The Creole TMA categories, for example, are typically expressed by items that are regularly taken from the superstrate.11 Two main questions arise with regard to the emergence of the TMA system. Which items become grammaticalized, and how do these items acquire their function? According to Bickerton, it is the innate language capacity, the notorious bioprogram, that has the relevant categories in store for any item that may come along to pick it up. What is worrying about this hypothesis is that crosslinguistically, i.e. across Creoles and non-creoles, very often the same lexical items 'attract' certain functions, i.e. are grammaticalized, so that it is hard to believe that in all of these cases creolization is involved. Note that this is not necessarily an argument against the innateness of certain principles that may be responsible for these innovations. It is, however, an argument against the restriction of these phenomena to creolization. Take, for example, the Creole non-punctual marker, which is often identical with a locative/spatial element (copula, verb or adverb) borrowed from the superstrate: (s)1a < Spanish/Portuguese es tar in many Ibero-based Creoles, stei < English stay in Hawaiian Creole, da/a/(d)e < English there in English-based Creole. Cross-linguistically, the grammaticalization of locative expressions as markers of imperfective, especially progressive, aspect has taken place in numerous non-creole languages (cf. Comrie 1976:98f, Jespersen 1931, Mufwene 1984:28f) and can thus not be regarded as unique to creolization. Recently, Ebert (1989) has shown that across the Germanic languages grammaticalized spatial expressions, like prepositions (plus copula) or verbs expressing a postural position, are used to mark progressive aspect. Consider the examples in (1): (1) a. Ich bin am Nachdenken (German) I am at thinking Ί am pondering' b. Ik ben aan "t schrijven (Dutch) I am at the writing Ί am writing*

Bickerton has labelled this process 'reconstitution', which "consists of making a new grammatical morpheme originally retained for its lexical meaning" (1989:23) Note that this is not without exception. Berbice Dutch, for example, employs postverbal clitics, borrowed from its substratum language Eastern Ijo.

Creolization and language change

11

c. Hat sat tu leezen (Frisian) She sits to read 'She is writing (while sitting)' d. Er setst in de stuv lesen (Westerwald dialect, Germany) He sits in the room read 'He is reading (while sitting in the room)' (Examples (c) and (d) are taken from Ebert 1989) In Spanish and Portuguese the use of the locative copula estar with gerund or infinitive is common to mark progressive: (2) a. Esta lluviendo (Spanish) It is raining b. Esta a espargir (European Portuguese) (He) is at scattering The comparative evidence strongly suggests that in creolization speakers make use of the same strategies of grammaticalization that we know from other languages. Not only the same fundamental categories are grammaticalized; moreover, these categories seem to be expressed by similar items taken from similar domains. One objection to this conclusion could be that in non-creole languages the punctual/non-punctual distinction is not encoded as systematically as it is in Creole. This argument is flawed in several respects. First of all, it is far from clear whether in creole the aspectual marking is really as nice and consistent as Bickerton would have it. Recent studies have emphasized the differences between the actually occurring systems and the 'ideal' creole type (cf., e.g., Muysken 1981, the articles in Singler 1990, Bruyn & Veenstra 1993, or Dahl 1985:167). Secondly, crosslinguistic evidence shows that the basic aspectual distinction encoded by languages is the one between perfective and imperfective (where imperfective can be split up in various subcategories like habitual, progressive, etc.12). The notion of (im-) perfectivity is essentially the same as Bickerton's (non-)punctuality. Singler (1990:xii), for example, notes with regard to the term 'nonpunctual' that "in terms of the study of aspect more generally, the term imperfective is more widely used. That is, Bickerton's punctual:non-punctual opposition is usually (if not invariably) a perfective:imperfective opposition." From all this we can conclude that the punctual:nonpunctual distinction is not unique to Creoles, but rather the unmarked option for the encoding of aspect (cf. also Muysken 1981: 191f, for a similar argument). Another, and perhaps more serious, objection against the view advocated here may be that Creoles are unique in developing an aspectual category which has not 12

See Comrie (1976) for details.

12

Plag

existed before. Although it is true that the creation of entirely new categories is rather exceptional in natural change, we still have to state that sometimes, though admittedly not too often, even non-creole languages create new categories. An example of the evolution of an aspectual category that has not existed before is Standard German, which currently seems to be developing the category of progressive aspect (cf. Ebert 1989, Lehmann 1991). We may conclude for the grammatical subsystem of aspect marking that Creoles and non-creoles tend to grammaticalize the same lexical items for the encoding of at least similar, if not identical, aspectual distinctions. Of course, there is great diversity among different languages concerning the mapping of morphemes onto the different aspectual subcategories, but this is not the point. From all we know about Creole TMA systems the mapping of categories is probably never identical in two languages. Given the prototype effects in the structure of categories this does not come as a surprise. The absence of an identical mapping of morphemes onto categories may not serve as an argument for a fundamental difference between natural change and Creole formation. What matters is that neither the Creole categories, nor the type of grammaticalization of the morphemes expressing those categories are restricted to Creole languages. There is, nevertheless, one important difference that can be observed, namely the rate of change. The evolution of the marking of progressive in English, for example, started most probably in the 14th century and reached its present form roughly in the 17th century (cf., e.g., Jespersen 1931). For many Creoles we can assume a development of much less than a century. In general this means that, in order to incorporate a new a grammatical category into its system, natural change proceeds very slowly, whereas creolization seems to be able to do the same job in a much shorter period of time.

3.1.2 Mood It has been argued that the creation of the modal categories irrealis and realis is significant of creolization. However, the realis :irrealis distinction is a very basic distinction that can be found in numerous languages, creole and non-creole. In many Austronesian languages, for example, it is the major morphological distinction in the TMA system. Givon (1984) illustrates this with Bikol and Chuave. Bikol employs the realis marker nag- for all past, present and habitual, while irrealis, expressed by the prefix mag-, covers future and all other non-realized modalities. In the New Guinea language Chuave realis is zero-marked, irrealis morphologically marked (see Givon 1984:309-314). The creole mood category is therefore in no sense a unique development of creole languages. The parallelisms between creole and non-creole languages again extend to the kinds of morphemes that are used for the encoding of modality. Many creole

Creolization and language change

13

languages employ some fonn of the motion verb 'go' to encode irrealis, a strategy also commonly found in numerous non-creoles to encode future/conditional/ hypothetical meaning.

3.1.3 Tense According to Bickerton, the temporal reference of Creole verbs is established in the following manner. Unmarked stative verbs are interpreted as having present time reference, unmarked non-stative verbs are interpreted as having past time reference. If non-stative verbs are preceded by the imperfective marker, they receive a present time reference. If verbs are preceded by the creole tense particle, which expresses anterior tense, the interpretation is, roughly, past-before-past for action (i.e., nonstative) verbs, and past for stative verbs (Bickerton 1975: chapter 2, 1981:58). An alternative description of these facts would be that the interpretation of the time reference of the unmarked verb is simply a secondary consequence of an essentially aspectual distinction. According to this analysis unmarked non-stative verbs are perfective and are interpreted as having past time reference. Unmarked stative verbs are treated as inherently imperfective, and receive a present time interpretation. This would be in line with many tenseless languages, which generally associate imperfective aspect with present time reference and past time reference with perfective aspect (Comrie 1976:83, Dahl 1985:79). A system strikingly parallel to the creole one can be found in Yoruba and in Igbo. Comrie (1976:82) gives the following account: in Yoruba, nonstative verbs have no marker if they have perfective meaning, and the marker n before the verb, if they have imperfective meaning. Stative verbs have only imperfective meaning, and take no marker. In Igbo the Imperfective of nonstative verbs is marked by na before the verb. [...] the Imperfective forms [...] are interpreted as referring to the present, while the Perfective forms [...] are interpreted as referring to the past. Another argument against the uniqueness of the creole system is that Proto-IndoEuropean probably had a similar system, "with aspect being marked overtly and time reference at best as a secondary consequence of aspectual distinctions." (Comrie 1976:83). Using markedness theory, Muysken (1981:190) shows that the creole tense categories constitute the unmarked case.

14

Plag

3.2 Complementizers The emergence of complementizers in creolization has been frequently discussed in the literature (see, among many others, Washabaugh 1975, Bickerton 1981, Woolford 1979, Plag 1993). It is generally assumed that pidgins lack any kind of sentence embedding. In Creoles, however, embedded clauses are not uncommon. On the basis of this observation it has been argued (e.g. by Washabaugh 1975) that the development of complementizers is part of the creolization process. Bickerton even goes a step further by saying that alongside the syntactic development we can observe the evolution of the semantic distinction between realized and unrealized complements, which is encoded by the Creole complementizers) (cf. Bickerton 1981:30-33). This complementizer is typically derived from a preposition 'for' or a movement verb 'go1. Two different developments have to be considered separately, the emergence of complementizers on the one hand, and the emergence of the realized:unrealized distinction on the other. Most of the Creole complementizers under discussion fall into two categories. One category consists of items that are homophonous with a general speech act verb 'say', the other of items that are homophonous with a preposition 'for' in the relevant language. The use of a general speech act verb 'say' as complementizer is a well-known phenomenon in non-creole languages around the globe (see, for example, Ebert 1991, Plag 1992, and the references cited in these works), and therefore not a phenomenon unique to Creoles. In a diachronic study of the development of the Sranan complementizer taki (< English talk) Plag (1993: chapter 5) has shown that this language follows the universal path of grammaticalization proposed by Ebert (1991) on the basis of more than twenty languages representing different parts of the world and several language families. A similar situation can be observed with the evolution of a preposition into a complementizer, of which Washabaugh says that it "is not a unique process for each distinct lamguage but that this development in creolization, and perhaps in all cases of grammar elaboration, proceeds alomg the same route. That is, the infinitivizing complementizer evolves from a preposition, ultimately a locative preposition, which is made to perform multiple syntactic functions in the expanding language" (Washabaugh 1975:109). And what about the realized :unrealized distinction? Again, a terminological problem seems to be involved. Bickerton's terminology strongly reminds one of the factive:nonfactive distinction, which is a familiar one, especially since Kiparsky & Kiparsky's seminal paper. Given that Bickerton's realized:unrealized is more or less identical with factive-non-factive, Bickerton is certainly right in pointing out that, for example, in English the infinitival element to does not encode this distinction in a straightforward manner. Consider Bickerton's examples (1981:60):

Creolization and language change 7317 7327 7337 7347

IS

I managed to stop (entails "I stopped") I failed to stop (entails "I did not stop") I went to see Mary and we talked about old times. I went to see Mary but she wasn't at home.

However, Bickerton (ibid.) is certainly wrong when he assumes that the factivernonfactive distinction, does not exist in English grammar. The systematic contrast in sentences like She remembered closing the door vs. She remembered to close the door shows that the factive:nonfactive distinction is indeed grammaticalized in English. The above English examples only show that to does not encode (non-) factivity iconically, and that to obviously cannot be equated in distribution and function with the Creole complementizer derived from 'for'. An even stronger point against the uniqueness of the development of the factive:nonfactive distinction in creolization can be made with regard to Haspelmath's 1989 cross-linguistic diachronic study of the grammaticalization of infinitives. Haspelmath shows that the development of an allative preposition into an infinitival marker follows a universal path of grammaticalization:13

Haspelmath's diachronic study lends independent support to Washabaugh's conclusions cited above, which were based on a cross-linguistic survey of synchronic data.

16

Plag

Figure 1: Semantic grammaticalization of the infinitive (from Haspelmathl989:298)

benefactive

N.

allative —>

/*

purposive -»irrealis-14 —» irrealis—» realisdirective

potential

non-factive

> (realisfactive)

causal At the first three stages the former allative, benefactive, or causal preposition expresses purposive, then irrealis-directive, and then irrealis-potential modalities. Note that at these stages our former preposition is consistently used to express nonfactive meanings, i.e., we have a situation identical to the Creole system. Only at later stages realis-factive meanings emerge and the unequivocality in the encoding of the distinction between factivity and non-factivity is lost. In the framework of the above figure, the Bickertonian Creoles have not gone beyond the irrealis-potential stage. Coming back to our comparison of developments in Creole and non-creole languages, let us look at Old High German, a language that is currently assumed to be free of any suspicion of being a Creole language. Haspelmath (1989) presents diachronic evidence which shows that in Old High German the allative preposition zi, which roughly corresponds to English ίο, is exclusively used with purposive and irrealis-directive modality. Later, in Middle High German, it is used also in complement clauses with irrealis-potential meaning. Yet, it is only in late Middle High German and Early New High German that the realis-nonfactive use emerges. 14

Haspelmath (1989:298f) defines the modality of complement clauses as follows: (a)

(b)

c)

(d)

Irrealis-directive is the modality of complements to manipulative verbs like Order', 'ask', 'cause', and to desiderative verbs like 'desire', 'want', 'prefer1. The complement situation is presented as not realized, and its possible realization is expected for the future [...] Irrealis-potential is the modality of complements to modal predicates like *be possible', "be able', *be necessary', "have to', and to evaluative predicates like 'interesting', 'funny', 'regret', etc. Here the situation is not realized either, but it is not expected to be realized sometime in the future; rather it is presented as potentially occurring any time. Realis-non-factive is the modality of complements to verbs of thinking (or 'prepositional attitude') like think', "believe1, 'seem', and verbs of utterance, e.g. 'say', 'claim', 'report'. Here the situation is presented as real, although the speaker is not committed to its truth. Realis-factive is the modality of complements to verbs of cognition, like "know1, 'realize', 'find out', and of evaluative predicates, (original emphasis)

In a footnote Haspelmath correctly remarks that many evaluative predicates can be used both in irrealis-potential and realis-factive modalities.

Creolization and language change

17

This means that until late Middle High German, the formerly prepositional element zi consistently marked unrealized German complements. A comparison with the Creole pattern shows that the way of marking unrealized complements is very similar, if not identical in the two systems. Hence, the emergence of complementizers in creolization seems to have close parallels in quite ordinary instances of natural language change. It should, however not go unnoticed that, as in the case of the TMA system, the observed developments happen in a comparatively very short period of time. In Plag (1993) it is argued on the basis of diachronic evidence that, for example, the Sranan complementizer fit took less than a century to reach - roughly - its present status of encoding irrealis-directive and irrealis-potential. 3.3 Articles Bickerton maintains that the emergence of a Creole-specific article system is significant of creole formation. Contrary to the pidgin, which lacks articles, in the Creole we find definite and indefinite articles, which encode specific/non-specific reference: "Virtually all Creoles [...] have a definite article for presupposed-specific NP; an indefinite article for asserted-specific NP, and zero for nonspecific NP" (Bickerton 1981:56). The indefinite creole article is usually derived from or completely identical with the numeral One1. In a paper that appeared in the same year as Bickerton's Roots of Language Givon shows the development of the numeral One1 as an indefinite-referential marker to be a universal process of grammaticalization. Givon uses the term 'referential' synonymously with 'specific'. His 'non-referential1 is identical to Bickerton's 'non-specific1 or 'generic'. Givon cites Creoles, Israeli Hebrew, Mandarin, Sherpa, Turkish, Neo-Aramaic, and Persian as languages that illustrate the first step in the grammaticalization of One1, with One' being exclusively used to mark referential-indefinite, i.e. asserted-specific, NPs. In those languages non-referential (i.e., non-specific) NPs are marked differently. At the other end of the developmental continuum One' marks indefinite referential and non-referential NPs in all environments (as is the case in English): "Romance languages such as Spanish and Italian exhibit an intermediate stage, whereas French, English, and German represent the latest perhaps the terminal stage along this diachronic continuum" (Givon 1981:35). With regard to the encoding of the specific/nonspecific distinction we can therefore conclude that the developments in creolization have corresponces in non-creole languages. These correspondences concern the nature of the underlying categories and the type of forms that are employed for marking these categories.

18

Plag

4. Conclusion In this paper we have compared the emergence of some grammatical properties in Creole and non-creole languages. The comparison revealed that, at least with respect to the categories under discussion, there are striking parallels between grammatical developments that can be observed in creolization on the one hand and natural change on the other. Which conclusions should be drawn from this fact? First of all, there seems to be no fundamental difference between the kind of change we encounter in creolization and natural change. The evolution of new categories follows universal principles of grammaticalization that can be observed at work in Creoles and non-creoles alike. Bickerton's uniqueness hypothesis thus seems not to be borne out by the (cross-linguistic) facts. On the contrary, the developments in creolization are the ones that are typically wide-spread, and represent the unmarked options for language development. The changes observed thus seem to be outstanding not because of their being resticted to Creoles, but just because of the opposite, namely their universality. According to current approaches to grammaticalization, the expansion or reorganization of grammar can often be represented by universal developmental continua, some of which have been mentioned above (cf. Ebert 1991, Haspelmath 1989, or Givon 1981). The idea of such continua is that they show a universal path of grammaticalization or grammar expansion.15 Different languages exhibit different possible stages on the continuum. With regard to the development of complementizers and articles (and the semantic categories expressed by these forms) we observed that Creole languages exhibit the early stages of these developmental continua, whereas non-creole languages may be found at all stages. The Creole TMA system shows the development of categories that seem to be fundamental, or unmarked, in human language. In general, speakers (be they first or second language learners, children or adult language users) seem to draw on universally available strategies of grammatical expansion for innovations. It is in Creoles (but not only in Creoles) that we find the development of the basic, unmarked options that are at the disposal of the speakers. Whether one wants to label the responsible mechanisms 'universal grammar', 'language bioprogam', or 'markedness' seems to be rather a matter of taste than of scientific judgment. Creolization is unique not because of the restriction of certain developments to this type of language evolution, but because of the fact that the changes proceed very fast and affect many subsystems at the same time. In natural change innovations are generally not as drastic, and tend to take much more time. This fact 15

'Universal' in this context does not mean that all languages must follow this path, but that the relevant development can be found frequently in the languages of the world (independent of genetic or areal relationship).

Creolization and language change

19

is a direct reflection of the type of transmission the speakers are confronted with. In impaired transmission, due to the mixed, variable, rudimentary or limited input, speakers have to draw heavily on their creative capacities, whereas in cases of normal transmission the choice of possible innovations is much more restricted because of the availability of a model system. The more impaired the transmission, the more we can expect developments that correspond to the early stages of diachronic continua and that reflect unmarked categories. The study of Creole genesis thus does not receive its importance for general theory from the uniqueness of the mechanisms involved, but from the fact that the historical linguist will hardly find a comparable abundance of interesting phenomena in non-creole languages. In creolization we are confronted with a sort of greenhouse effect: just as plants grow faster and in higher numbers in a greenhouse than in their natural environment, in creolization language change proceeds faster and more drastically than in natural change. The differences between creolization and natural change are, however, not differences in kind, but in quantity. After all, even in greenhouses one cannot grow tomatoes from lettuce seeds.

References Anttila, Raimo. 1972. An introduction to historical and comparative linguistics. New York: Macmillan. Arenas, Jacques. 1989. Syntactic developments in Sranan. Nijmegen: Catholic University dissertation. Bickerton, Derek. 1975. Dynamics of a Creole system. Cambridge: Cambridge University Press. . 1981. Roots of language. Ann Arbor: Karorna. . 1984a. "The language bioprogram hypothesis", The Behavioral and Brain Sciences 7, 172-221. . 1989. "The lexical learning hypothesis and the pidgin-creole cycle", in Martin Pütz and Rene Dirven (eds.) Wheels within wheels: papers of the Duisburg symposium on pidgin and Creole languages, 11-31. Frankfurt: Peter Lang. . 1991. "On the supposed 'gradualness' of Creole development", Journal of Pidgin and Creole Languages 6.1,25-58. Bickerton, Derek, and Talmy Givon. 1976. "Pidginization and syntactic change: from SXV and VSX to SVX", in Sanford B. Steever et al. (eds.) Papers from the parasession on diachronic syntax, 9-39. Chicago: Chicago Linguistic Society. Bruyn, Adrienne, and Tonjes Veenstra. 1993. "The creolization of Dutch", Journal of Pidgin and Creole Languages 8.1, 29-80. Carden, Guy, and William A. Stewart. 1988. "Binding theory, bioprogram, and creolization: evidence from Haitian Creole", Journal of Pidgin and Creole Languages 3.1, 1-67.

20

Plag

Comrie, Bernard. 1976. Tense. Cambridge: Cambridge University Press. Dahl, Osten. 1985. Tense and aspect systems. Oxford: Blackwell. Ebert, Karen H. 1989. "Aspektmarkierung im Fering (Nordfriesisch) und verwandten Sprachen", in Werner Abraham and Theo Janssen (eds.) Tempus - Aspekt - Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen, 293-322. Tübingen: Niemeyer. Ebert, Karen H. 1991. "Vom verbum dicendi zur Konjunktion - Ein Kapitel universaler Grammatikentwicklung", in Walter Bisang and Paul Rinderknecht (eds.) Von Europa bis Ozeanien Von der Antonymie zum Relativsatz: Gedenkschriftßir Meinhard Scheller, 77-94. Zürich: Seminar für Allgemeine Sprachwissenschaft. . 1981. "On the development of the numeral One' as an indefinite marker", Folia Linguistica Historica 2, 35-53. .1984. Syntax: a functional-typological introduction. Vol.1. Amsterdam: Benjamins. Goodmann, Morris. 1985. "Review of Roots of Language by Derek Bickerton", International Journal of American Linguistics 51.1, 109-137. Hall, Robert. 1958. "Creolized languages and 'genetic relationships'", Word 14, 367-373. Haspelmath, Martin. 1989. "From purposive to infinitive - a universal path of grammaticization", Folia Linguistica Historia 10, 287-310. Hock, Hans Henrich. 1991 (2nd. ed.). Principles of historical linguistics. Berlin: Mouton. Jespersen, Otto. 1931. A modern English grammar based on historical principles. Part IV: Syntax. 3rd vol.: Time and tense. London: Allen & Unwin/Copenhagen: Munksgaard Labov, William. 1990. "On the adequacy of natural language", in Singler (ed.), 1-58. Lehmann, Christian. 1991. "Grammaticalization and related changes in contemporary German", in Traugott and Heine (eds.), 493-535. Lehmann, Winfried. 1973 (2nd ed.). Historical linguistics. New York: Holt, Rinehart and Winston. Meillet, Andre. 1914. "Le probleme de la parents des langues", Scientia (Rivista di scienza) XV, 403425. Mühlhäusler, Peter. 1984. "Continuity and discontinuity in the development of pidgins and Creoles", in Werner Enninger and Lilith M. Haynes (eds.) Studies in language ecology, 118-134. Wiesbaden: Steiner. ———. 1986. Pidgin and Creole linguistics. Oxford: Blackwell. Mufwene, Salikoko. 1984. Stativity and the progressive. Bloomington: Indiana University Linguistics Club. Muysken, Pieter. 1981. "Creole tense/mood/aspect systems: the unmarked case?", in Pieter Muysken (ed.) Generative studies in Creole languages, 181-199. Dordrecht: Foris. Plag, Ingo. 1992. "From speech act verb to conjunction: the grammaticalization of taki in Sranan", Journal of Pidgin and Creole Languages 7.1, 55-73.

Creolization and language change

21

.1993. Sentential complementation in Sranan: on the formation of an English-based Creole language. Tübingen: Niemeyer. . 1994. "The emergence oftaki as a complementizer in Sranan: on substrate influence, universals, and gradual creolization", in Jacques Arenas (ed.) Creoles: the early stages. Amsterdam: Benjamins (to appear). Romaine, Suzanne. 1988. Pidgin and Creole languages. London: Longman. Schuchardt, Hugo. 1917. "Sprachverwandtschaft", Sitzungsberichte der Berliner Akademie der Wissenschaften, XXXVII, 518-529. Singler, John Victor (ed.). 1990. Pidgin and Creole tense-mood-aspect-systems. Benjamins.

Amsterdam:

Taylor, Douglas. 1956. "Language contacts in the British West Indies", Word 12, 399-414. Thomason, Sarah Grey. 1980. "Continuity of transmission and genetic relationship", in Elizabeth C. Traugott et al. (eds.) Papers from the 4th International Conference on Historical Linguistics, 2735. Amsterdam: Benjamins. Thomason, Sarah Grey, and Terrence Kaufman 1988. Language contact, creolization, and genetic linguistics. Berkeley/Los Angeles/London: University of California Press. Traugott, Elizabeth, and Bemd Heine (eds.). 1991. Approaches to grammaticalization. 2 Volumes. Amsterdam: Benjamins. Washabaugh, William. 1975. "On the develpoment of complementizers in creolization", Working Papers on Language Universals 17, 109-140. Weinreich, Uriel. 1958. "On the compatibility of genetic relationship and convergent development", Word 14, 374-379. Woolford, Ellen. 1979. "The developing complementizer system of Tok Pisin: syntactic change in progress", in Kenneth Hill (ed.) The genesis of language, 108-124. Ann Arbor: Karoma.

Dany Adone

Creolization and language change in Mauritian Creole1 1. Introduction The present paper is concerned with three processes which are central to the field of Creole studies: creolization, language change and decreolization. The discussion about creolization and language change is as old as the field of Creole studies itself (cf. Meillet 1914, Schuchardt 1917). Two prevailing views on creolization and language change have been articulated up to now. One group of scholars (e.g. Goodman 1985, Lefebvre 1986, and Mühlhäusler 1986) argue that creolization is a gradual process lasting for several generations. Creolization then is a form of language change. Scholars like Bickerton (1984, 1986, 1991) claim that creolization is sudden and abrupt and it takes place within one generation. Thus creolization is not language change. Closely related to creolization is the process of decreolization. It has been assumed that if creolization takes place, then decreolization can also happen in the life-cycle of a Creole language. It is an interesting question whether decreolization can be viewed as a form of language change. Three aspects of modem Mauritian Creole (henceforth MC), namely null subjects, the pronouns/reflexives and relativization, will be investigated here to illustrate the phenomenon of language change in a Creole language. It is argued that two of these structures, namely null subjects and reflexives with mem 'self were not present in the grammar of MC during the creolization process. Their presence in modern MC is due to changes in the language which took place gradually as MC expanded grammatically. Furthermore it is claimed that these changes are due to contact between French and MC. But inspite of the similarity to French, there is no reason to regard these changes as signs of decreolization. The paper is organized as follows. In the next section the issue of creolization and language change will be discussed. In section 3 some sociolinguistic facts of modern MC are presented. Section 4 discusses the cases of language change in MC and addresses the question of decreolization in MC and section 5 contains some conclusions.

1

I thank Ray Fabri, Stefan Freinatis, Ingo Plag and Tonjes Veenstra for very helpful discussions and constructive comments. All shortcomings are my own.

24

Adone

2. Creolization, decreolization and language change 2.1 Why creolization cannot be gradual In Creole studies creolization is generally defined as the process by which Creole languages are formed. In the current debate on creolization there are two issues which are still unclear: the mechanisms involved in creolization and the pace of creolization. With respect to the first issue, was it relexification or transmission of substrate languages? Or perhaps processes of first or second language acquisition? On the basis of observations in previous studies we are led to argue that it is first language acquisition which plays a primary role in creolization. But the fact that the transmission factor (of superstrate and substrate features) in creolization cannot be denied, does not automatically mean that its presence alone can explain the process of creolization. A sociolinguistic model of creolization in which either the transmission of superstrate or substrate features is regarded as the primary explanation for the emergence of Creole languages is inadequate because it cannot properly account for the well known similarities among the English-, French-, Dutch-, Portuguese- and Spanish-based Creoles2. Thus, such a model fails to be psychologically oriented, that is it does not take into account that part of the human knowledge of language is innate, i.e. due to the nature of human organism. With respect to the second issue, there are two positions. The first position is that of Bickerton (1984, 1988, 1991), the second has been argued by Goodman (1985) among others. Recently, proponents of the gradualist theory have provided empirical evidence for the long and gradual development of Haitian Creole and Sranan respectively (cf. Garden and Stewart 1988, Arends 1989, Plag 1993)3. Their findings lead them to argue that creolization is gradual and that creolization and language change are not fundamentally different (cf. Plag this volume). Although it is possible that some of the structures have taken some generations to develop to the present state, this does not necessarily imply that creolization as a whole was slow and gradual. To argue that creolization was slow and gradual obviously contradicts two well-known facts in creolization, namely the pressure for communication and the fact that children born in these societies needed a language*. As far as the parents of these children Although these similarities might be limited. For more detail see Bniyn and Veenstra (1993). Garden and Stewart worked on Haitian reflexivization. Arends on clefting, comparatives and copulative verbs in Sranan and Plag on complementation in Sranan. Bickerton (1990a) showed on the basis of French census data, that in 1681 15% of the slave population of Haiti were children. These data are fully consistent with the thesis that early and rapid creolization took place.

Creolization and language change in Mauritian Creole

25

are concerned, it is clear that they spoke a pidgin to communicate with the rest of the population and had their own mother tongue to communicate with other speakers of their language. But what did children of these speakers do? Did they stay without any native language? If creolization is gradual as the gradualist theory claims, then it means that generations of speakers (starting with children) did not have any native language. This is not at all likely. To answer the above questions we have to look at the input. The first question that comes to mind here is what was the nature of the input? Since the input children were confronted with, comes from their parents, who were pidgin speakers, the input must have been degenerate. Degenerate here means that the input was not syntactically consistent. In contrast to normal language acquisition in which children receive variable but 'consistent' syntactic input, Creole children did not. So they had no other choice than to rely on their genetic equipment for language to create a novel system, as Bickerton argues. This argument is supported by studies in language acquisition research. Especially studies on abnormal language acquisition have shown that deaf children go much further than any models available to them (Goldin-Meadow 1990, Singleton and Newport 1987). Whether it is the Language Acquisition Device (LAD) as formulated by Chomsky, or Bickerton's Language Bioprogram (LB), is not relevant to the discussion here. If creolization is sudden or 'catastrophic', it cannot be language change, which takes place over a long period of time. Contrary to creolization, language change affects all systems, stable as well as instable, including Creole languages. The question here is why creolization cannot be language change. Given the fact that it is children who 'created' Creole languages, it cannot be change because change logically implies that there is an alteration in a system which at a certain period of time is affected by something else, and this is obviously not the case in creolization.5 What children did, was to create something new. Thus, creolization is nativization. One question related to the definition of creolization given so far is, whether the process of creolization is unique to Creole languages. The field of sign language, particularly the case of deaf children with hearing parents seems to offer a good comparison. Gee and Goodhart (1985) argue that a deaf child with hearing parents is in a similar situation as a Creole child. But this similarity is only superficial. There are two reasons for this. Firstly, because Creoles represent a case of normal language development, in contrast to the deaf situation which is a case of abnormal language development. The terms normal and abnormal are used here to refer to situations in which language development takes place without any specific health deficiency or illness such as deafness. Secondly, because there is a difference in terms of input between these two cases. In the pidgin-creole cycle the first generation speakers have an incoherent input, which the second generation turns 5

Similarities between creolization and language change are observed with respect to the mechanisms. Grammaticalization is one of them.

26

Adone

into a coherent and fluent system. In the deaf situation if parents do not know how to sign, then it is clear that the children do not receive any consistent input and thus they have to use natural signs to communicate. Thus it is the second generation which moves from an incoherent to a coherent system and it is finally the third generation which has a coherent and fluent system. In the Creole case we can speak of restricted input, while in the deaf case one can speak of disruption of input. This shows that deprivation in the deaf cycle is more radical than in Creole languages and it obviously takes longer than in the Creole cycle. Summarizing, the view taken here is nativist. It is assumed that part of our knowledge of language is innate and that there is a biological predisposition in human beings to acquire languages. Following Bickerton, creolization is thus regarded as first language acquisition determined by the human biological language capacity in cases of restricted input.

3. Sociolinguistic facts about modern MC Before we start with our analysis of MC, some theoretical issues have to be discussed. Decamp (1971:349) argues that "a Creole can continue indefinitely without any substantial change, as Haitian Creole seems to be doing." This picture of a Creole coexisting with another language without any change is not realistic because any language (creole or non-creole) in contact with other systems undergoes changes in the course of time. And this is what I will show for MC. But before, one might ask whether there is a continuum between French and MC. Though no studies have been conducted regarding the dialectal variations in MC, I do not assume, contrary to Bickerton (1980), the presence of a continuum between French as the standard language and several lects of MC, moving from basilect to acrolect. Instead, I argue that there are two more or less distinct varieties of MC, which I call urban and rural. One reason for the absence of a continuum is the structure of Mauritian society. French is still the prestige language in the country, as witnessed in other Francophone countries. It is also used as first language by a small part of the population (that is Franco-Mauritians and 'gens de couleur'6). It is also the language used in daily life for interethnic communication but it is not the only language used for this purpose, since the population is not homogenously francophone. A look at the ethnic groups and languages distribution in Mauritius shows the following picture. There are 51.8% Indo-Mauritians (Hindu), 16.6% Indo-Mauritians (Muslim), 2.9% Chinese and 28.7% 'population generate1 (creoles, FrancoMauritians, gens de couleurs). This term is used for the mixed blood population.

Creolization and language change in Mauritian Creole

27

Apart from French and MC, there are other languages such as Bhojpuri, Urdu, Tamil, Telegu, Marathi and Hakka among others which are spoken by elderly people (not by the younger generation, though) and which are given much religious importance by the government7. English, for instance, is hardly ever spoken except for official purposes. MC is the only language which is spoken by 90% of the population. For more than 70% of Mauritians it is the first language. While French is preferred on certain occasions in urban areas, MC is the language most frequently used in rural areas. Differences between urban and rural MC can be witnessed on two levels; on the lexical as well as on the phonological level as illustrated in Tables 1 and 28.

Table 1: Some examples of lexical differences between urban and rural MC. French la maison la rue epoux le journal arbre

Urban lameson lari mart zurnal zarb

Rural lakaz semen bonom/misie lagazet pye

Table 2: Some of the phonological differences between urban and rural MC French du riz le cheval le chemin avec

Urban djuri seval semen avek

Rural diri/duri suval/seval simen ek

One reason for these differences is certainly the fact that in urban areas MC speakers have more contact with Mauritian French and that they are more conscious of the value of French in the society. Thus an urban MC speaker would prefer lameson "house1 which is nearer to French 'la maison' than lakaz Tiouse' on table 1. With this choice the speaker makes also clear that s/he speaks French also. Massive lexical and phonological borrowings take place in different fields such as 7

The government is in favor of a policy of multiculturalism and multilingualism. Oriental languages are still taught at primary and secondary schools even though with little success.

8

These tables were designed according to the model presented by Valdmann (1991).

28

Adone

essential for the development of MC as a standard language. However, crucial is the fact that there are no syntactic differences between these two variants of MC. Summarizing, we can say that modern MC has two varieties, urban and rural. On the lexical and phonological levels there are differences between these two varieties, while on the syntactic level there is no difference.

4. From creolization to language change in MC 4.1 Properties of creolization Bickerton (1981) argues that there are features in Creole grammars which are proper to creolization. He identifies twelve of them: TMA system articles movement rules realized and unrealized complementizer negation relativization existential and possessive copula adjectives as verbs questions question words passive equivalents In this study three features, null subjects, pronouns/reflexives and relative clauses will be discussed. As far as null subject is concerned, Bickerton did not make any particular claims in his Bioprogram. We note however that it has been implicitly assumed that Creole languages, in contrast to pidgins, developed the minimal sentence structure: S -> NP AUX VP (Cf. Koopman and Lefebvre 1981) In other words, lexical subjects can be taken to be present in creolization. With pronouns and reflexives it will be shown that during the period of creolization MC used so lekor for reflexivity more frequently than limem. In modern MC both expressions are under strong competition. Relativization, which was present in the creolization of MC, is still present and productive in modern MC. With these three features, this study aims at showing some developments in MC

Creolization and language change in Mauritian Creole

29

With these three features, this study aims at showing some developments in MC with focus on the issue of what was part of creolization process and what can be regarded as subsequent language change.

4.2 Null subjects 4.2.1 Previous studies Since Perlmutter (1971) observed that there are languages which allow phonologically null subjects in tensed clauses, several proposals have been made to account for this phenomenon. A standard assumption was that null subject languages such as Italian and Spanish have a rich agreement system. But a look at various other languages reveals a different picture. There are languages like Italian with agreement and thus null subjects, languages like German and Dutch with agreement but no null subjects, languages like English with no agreement and no null subjects and languages like Chinese, MC and Haitian Creole with no agreement but null subjects.

4.2.2 Early MC Before we start discussing the MC texts, it is important to recall that all the available texts of MC mentioned in Chaudenson (1981), the only source of early MC, were written by French-speaking people native speakers of French. Documents from pidgin or Creole speakers are not available. This means that one should be careful in evaluating the importance of early documents for the development of MC9. With the term early MC I refer to the Creole variety which appeared immediately after creolization. According to Baker and Come (1986), creolization of MC took place around 1730-1770. My corpus consists of four texts from four authors Pitot (1805), Freycinet (1827), Descroizilles (1867) and Baissac (1880). These authors were chosen because they postdate the period of creolization. The texts were written between 1805 and 1880. Pilot's text is a small text of about 77 sentences. It is about an imaginary interview of a Mozambique slave by a white journalist. This text was written by Pitot to refute all the criticisms made by Bemadin de St Pierre concerning the conditions of slaves on the island. Freycinet's text is part of Freycinet's 'Voyage 9

Roberts (pc.) has also noted the same in Hawaiian texts. The HCE texts written by schoolchildren which Reinecke (1969) collected, showed a more sophisticated TMA system than revealed in contemporary texts written by Anglophones.

30

Adone

autour du monde'. It consists of 91 sentences and tells the adventure of a hunter. Descroizilles's text is written in the form of a poem in which he recalls the good life in Mauritius and the battle of Vieux Grand Port10. It consists of 259 sentences. Baissac's text is the creole version of the 'conte du chat botte'. It has 280 sentences. A look at the texts shows the following picture. Of 624 sentences in the corpus with a null subject possible in modern MC, there were four ambiguous constructions which could be interpreted as lacking a phonologically realized subjects. The first one is from Pitot11 and the three others come from Descroizilles: (1)

Qui manze? what eat? 'what do (you) eat?1 (taken from Pitot (1805) in Chaudenson (1981:81))

(2)

Cotte ti fine diboute, laissel li coma li tende where TNS ASP stand let him as him hear 'let him stand where (he) stood' (taken from Descroizilles (1867) in Chaudenson (1981:129))

(3)

dans so chapeau guette in his hat look '(he) looks in his hat' (taken from Descroizilles (1867) in Chaudenson (1981:128))

(4)

Sipri dimound aster, fine vini bien malin; surprise everybodynow ASP become very smart 'to the surprise of everybody (he) became very smart.' (taken from Descroizilles (1867) in Chaudenson (1981:128))

In all the sentences there could be a // 's/he' missing, in (1) u 'you' could be missing. However, for the above examples there is no translation accompanying the text for clarification. Sentences (5)-(9) show the use of lexical subjects, as it occurs in most of the sentences:

10

This battle is historically significant because the French lost the battle and Mauritius became a British colony.

11

This example is from the white journalist and not as might be expected from the black slave.

Creolization and language change in Mauritian Creole

31

(5)

Catte mete so botes, li prend so sac, li amare li dans so lereins cat puts on his boots, he take his bag, he tie it in his waist. 'the cat puts his boots on, he takes his bag, he ties it around his waist' (taken from Baissac (1880) in Chaudenson (1981:134))

(6)

Liiere mo fine tire so botes, pour dire moi grand merci li envoye moi ene coudepied ki fere moi tonbe ici pour raconte vous ca zhistoire la. as I ASP take off his boots, to tell me big thank he give me a kick which make me fall here to tell you this story 'as I have taken his boots off, to thank me. he gave me a kick which makes me fall here to tell you this story' (taken from Baissac (1880) in Chaudenson (1981:140))

(7)

mo n'apa volor, moi mo n'apas maron,... I NEG thief, me I NEG maroon am not a thief, I am not a maroon' (taken from Pitot 1805 in Chaudenson 1981:80)

(8)

Ene zur nu te la sasse cerf Grand-Riviere one day we TNS hunting deer Grand-Riviere One day we went hunting in Grand-Riviere' (taken from Freycinet (1827) in Chaudenson (1981:100))

(9)

Et langage Blanc meme, aster li 1'autrement: dans salon grand Misie, quant zot vie cause, mo tende dire lemots, qui mo näpas comprend. and language white even now it different: in dining-room big master when they want talk I hear say words that I NEG understand. 'and even the language of the white is different today: in the dinning room of my boss when they want to speak, 1 hear them say words which I do not understand.' (taken from Descroizilles (1867) in Chaudenson (1981:127))

At this point it is interesting that inspite of the majority of the sentences with lexical subjects, there are empty expletive subjects found in the texts:

32

Adone

(10)

non! n'apas comm' ca asper va! no NEG like that wait go 'no, it won't be like that, wait!' (taken from Freycinet (1827) in Chaudenson (1981:100))

(11)

Faut vous dire moi, mais ... must you tell me but... Ί must tell you' (taken from Descroizilles (1867) in Chaudenson (1981:128))

(12)

te iena ene ois ene vie blanc qui ti ena trois pitits TNS EXI a time a old white who TNS have three children Once upon a time there was an old white (man) who had three children' (taken from Baissac (1880) in Chaudenson (1981:134))

The examples in (10) to (12) indicate quite clearly that there were no expletive subjects in early MC contrary to French12. In Freycinet's texts there were 2 cases, in Descroizilles's text there were 4 cases and in Baissac's text there were 7 cases. A look at other Creoles shows a similar development. Byrne (1987) and Veenstra (this volume) for Saramaccan, Kouwenberg (1990) for Papiamentu, Degraff (1992) for Haitian Creole have also shown that empty expletive constructions occur in these languages. This, together with the fact that early MC had null expletive constructions, leads one to conclude tentatively that null expletive is a feature found in creolization. The fact that early MC had empty expletive constructions with existentials shows that this language had only one property of pro-drop languages. The other property, i.e. null subject pronouns in subject position, is probably a more recent development13.

4.2.3 Modern MC Below are some examples from different MC speakers belonging to different social classes and ethnic groups, thus showing that phonologically null subject constructions is a syntactic development present in both rural and urban MC. e stands for an empty category:

I mean written French here. In colloquial modem French the expletive il 'it' is normally dropped. Recent development here means any development which did not take place in the creolization period.

Creolization and language change in Mauritian Creole

33

(13)

e pe rod sanz konstitisyon brit e ASP try change constitution brutally 'they are trying to change the constitution quickly'

(14)

e ti boykot en paket e TNSboycot QUA many "he boycotted many Creoles in his work'

kreol Creole

(15)

e pu return dan peis e MOD return in country Ί will go back to the country soon'

biento soon

(16)

si ena syklon zot lakaz pu if there is cyclone their house MOD 'if there is a cyclone their house will be damaged'

dan travay in work

kase break

A related issue concerns the status of null subjects in French which is the lexifier language in the genesis of MC. French does not allow null subject constructions, thus it cannot be regarded as the source for null subjects. However, even if we assume that French is pro-drop, we still cannot argue that French has influenced MC for the simple reason that in Mauritius we have a Mauritian variety of French, which I will call Mauritian French. It is, in contrast to metropolitan French (the French spoken in France), very conservative and has no null subjects14. If there is no external reason for the development of null subjects, one might probably have to look for some language-internal reason. One could assume that MC is changing from a non-null subject language to a null subject language because MC syntax is changing. If this is so, we might ask for the factors motivating this development. Without going into a detailed analysis which would be beyond the scope of this paper, I tentatively suggest that there are at least two changes in other domains of syntax which are responsible for this development. Within GB framework it is assumed that there are two necessary conditions to account for the presence of null subjects in languages: licensing and identification (cf. Jaeggli and Safir 1989). The fact that null subjects in MC are possible in constructions where TMA markers appear, shows that TMA markers license null subjects, and the fact

I have even collected examples of corrections of null subjects in MC in primary classes at school. The use of null subjects is very common in the speech of MC children, but teachers force children to use overt subjects in MC by correcting them permanently.

34

Adone

that these null subjects are discourse-oriented15, leads us to assume that the presence of null topics in MC can account for the identification of these null subjects.

4.3 Pronouns and reflexives 4.3.1 Previous work Recently, there has been a substantial amount of work on pronouns and reflexives in creole languages (cf. Adamson 1993, Garden and Stewart (1989), Corne (1988), Muysken and Smith this volume). As shown by Muysken and Smith (this volume), the systems described in these studies vary considerably. This variability among creole languages has led many creolists to different hypotheses. For instance Garden and Stewart (1989) argue that the historical development of Haitian Greole reflexivization supports the gradual creolization theory. Bickerton argues that reflexives were not part of creolization. In Bickerton (1986:228-30) he states that "Creoles [...] could be understood in terms of three key concepts: retention, loss and reconstitution... . What was lost and what was retained played an important role in determining syntactic differences. For instance, reflexive forms were, in general, retained in English Creoles, but lost in French Creoles." The reason for the slow replacement of reflexives in contrast to tense, which was replaced quickly, could be that reflexives are not essential to a language. As far as MC is concerned, we will first have a look at Corne's study of early MC16. Corne (1988) argues, on the basis of early MC texts, that the 19th century MC expressed reflexivity by the use of so lekor "his body' as in example (17). But this construction was restricted to physical action only. Moreover, in imperatives as exemplified in (18) to (20) the postposed reflexive pronoun (here bare pronoun) "is in fact a feature of French-influenced 19th century Mau[ritian Creole, D.A.]." (Come 1988:83). Finally, the attestation of a postposed pronoun used reflexively, with or without mem 'self is also taken by Come to be an example of French influence. In other words, so lekor "his/her body', bare pronouns and sporadic pronouns with mem 'self were means of marking reflexivity in early MC.

15

SeeSyea(1985).

16

Corne's (1988) remark on modern MC pronoun and reflexive system can be disregarded because the discussion here is focussed on the early form of MC.

Creolization and language change in Mauritian Creole

35

(17)

sipa docteir va soulaze mo lecorps perhaps doctor MOD relieve my body 'perhaps the doctor will relieve me/my pain' (Baissac (1888:353) in Corne (1988))

(18)

aranze zautes arrange you 'get yourselves organized!' (Baissac (1880:133) in Corne (1988))

(19)

napa bizoein tracasse vou NEG must worry you 'take no thought!' (Anderson (1885:64) in Corne (1988))

(20)

faudrait to corize toi must you correct you "you must correct yourself (Segrais (1939) in Come (1988))

Since creolization took place around 1730-1770, all these texts above by Corne are not representative of the early period of creolization because they are dated 1885, 1888 and 1939. A look at Pitot's and Freycinet's texts which were written between 1805 and 1827 would be more appropriate. Unfortunately these texts do not have any reflexives. This of course, does not mean automatically that there were no reflexives at this time. In modem MC we also have several ways of marking reflexivity which are illustrated in the following examples. In 21 we have zero-marking with the verb benye. In 22 reflexivity is marked by // which is a bare pronoun. In 23 we have nu + lekor 'we + body' and in 24 we have limem 's/he + self: (21)

mo ti beny dan en lak kontamine I TNS bath in a lake contaminated Ί bathed in a contaminated lake*

(22)

li ti amiz li s/he TNS amuse her/himself 's/he amused her/himself well1

(23)

nu ti amiz nu lekor we TNS amuse our body 'we amused ourselves'

byefi well

36

(24)

Adone

tifi- la pe benylimem dan larivyer girl DET ASP bath herself in river 'the girl is bathing in the river'

4.3.2 Experiments In 1992/93 I conducted a series of experiments to test different groups of MC speakers on their use of reflexivization. The following observations were made: Bare pronouns in MC are used to mark reflexivity with a particular group of inherently reflexive and transitive verbs such as benye "bath1, amize 'amuse' among others, thus confirming Come in his analysis of modern MC. Other verbs such as save 'save', sape 'save' are preferred with zero marking, that is without reflexive pronouns. The [pronoun + mem} form is preferred to the reflexive expression so lekor "his body' in some constructions and the competition between both forms is obvious. Compared with early MC, it seems that the use of the [pronoun + mem} to mark reflexivity is more common. Note that this development has not yet been witnessed in Haitian Creole, where limem 'him/herself has more of an emphatic function (Michel Degraff, personal communication). So lekor 'his/her body', which was witnessed in creolization, is used rarely nowadays. In urban MC, for instance, so lekor is rarely used with verbs such as amize 'amuse1 or deside 'decide'. In these contexts the [pronoun + mem} is preferred, thus suggesting that the latter is overtaking partly the functions of so lekor. These developments are relatively new, i.e. they seem to have occurred in the last ten or fifteen years. These observations together with the diachronic data on early MC lead me to propose the following scenario to account for the development of the pronoun/reflexive system in modern MC. In creolization or even pidginization, as Bickerton has already suggested, reflexive forms are lost, because they do not belong to the core features of a new language. In addition, in French-based Creoles both reflexive forms (me, te, se) and regular nonemphatic pronouns (je, tu, if) were lost. This loss was predictable, according to Bickerton, because these elements are clitics which obviously carry no stress and cannot occur sentence-finally. Bare pronouns, which are essential in the grammar of a language, were first reconstituted. This would explain their use as reflexive pronouns sometimes with mem or sometimes alone, together with so lekor to express reflexivity sporadically. This explanation seems more natural than Corne's view, which considers the reflexive pronouns in MC, early and modern to be the result of French influence. In fact, the well-established use of bare pronouns in the speech of monolinguals of modern MC, rural and urban, is evidence against

Creolization and language change in Mauritian Creole

37

Gome's scenario. If the use of bare pronouns were really due to French influence, one would not expect to find them in environments in which French has little or no influence. The use of limem (more frequently than so lekor) as a reflexive pronoun in modern MC can be regarded as an ongoing change. Although such changes can be motivated by the influence of French, there is no reason to assume that MC is decreolizing. We will come to the issue of decreolization below.

4.4 Relativization 4.4.1 Previous Work In creole studies, relativization is of particular importance because it represents complexity in grammar. Pidgins, it has been argued, lack these rules and thus have no formal marking to indicate embedded clauses, whereas Creoles, being fullflegded languages, feature such structures. Hence, relativization can be taken to emerge in the creolization period.

4.4.2 Early MC The corpus consists of four texts from four authors, Pitot (1805), Freycinet (1827), Descroizilles (1867) and Baissac (1880), which were already analyzed with respect to the null subject issue in section 4.2. A look at the texts shows frequent use of relative clauses. There are 7 relative clauses in Pitot's text, 3 in Freycinet's text, 10 in Descroizilles' text and 16 in Baissac's text, all of them with the relativizer ki 'who/which'. Some examples are given in sentences (25) to (28): (25)

... ca qui mauvais n'apas capave fer ...that RELbad NEG can make 'the bad god cannot harm me' (taken from Pitot (1805) in Chaudenson (1981:81))

(26)

Ah ah! qui dou-monde qui te pass'la? oh! who people REL TNS pass by Oh! who are the people who passed by?1 (taken from Freycinet (1827) in Chaudenson (1981:100))

moi dimal me harm

38

Adone

(27)

Yena qui dimande, pourquoi criole nous cause? there are who ask why Creole we speak 'there are (people) who ask why we speak Creole?' (taken from Descroizilles (1867) in Chaudenson (1981:127))

(28)

te iena ene fois ene vie blanc qui te ena trois pitits TNS EXI a time a old white REL TNS have 3 children Once upon a time there was an old white man who had three children' (taken from Baissac (1880) in Chaudenson (1981:134))

A look at other French-based Creoles shows a similar development. Degraff (1992) and others have also shown that ki in Haitian Creole is also a relativizer. It is interesting that with ki, all French-based Creoles have retained a form which is nearer to the French qui. In MC ki is used both for subjects and objects of relative clauses. Que was obviously not a good candidate for retention because it was not salient enough in terms of stress. 4.4.3 Modern MC At this point we should look at modern data. These examples come from spontaneous speech and from two texts by Virahsawmy: (29)

sa bug ki ti fer holdup dan labank DET man REL TNS make holdup in bank 'the man who did the holdup in the bank was a white17'

(30)

ti ena en blan ki ti marye ek en sinois TNS EXI a white REL TNS marry with a Chinese 'last year there was a white man who married a Chinese girl'

(31)

sa lavyon ki final krase DET plane REL ASP crash 'the plane which crashed, was full'

(32)

so lizie ki nwaye... his eye REL drowned 'his eye which was drowned...' (taken from Virahsawmy 1991:6)

ΒΙαΛ in MC is the name for a Franco-Mauritian.

la, ti ranpli DEF TNS full

ti en blan. TNS a white Ian pase year last

Creolization and language change in Mauritian Creole

(33)

Enn zwazo plim mouye ki pe mor lamor a bird feather wet REL ASP die dead 'a bird with wet feather which is dying a stupid death' (taken from Virahsawmy 1991:6)

39

bet. stupid

The use of ki as relativizer is well attested in both rural and urban MC. It cannot be denied that ki is derived from French qui. But in contrast to French, ki in modern MC can be used together with some wh-words like kisana 'who', a development which is not found in early texts: (34)

kisana ki ti vini yer? who REL TNS come yesterday? 'who was the one who came yesterday?1

Summarizing, one can say that ki was witnessed in the creolization process of MC, a fact that leads us to argue that relativization belongs to the features of creolization. The fact that it is still present in modem MC reinforces the view that relativization is a necessary feature of grammar. Furthermore, it shows that features of creolization are also subject to change. Although ththe MC relativizer is derived from French, it differs from French qui in some respects.

4.5 Is MC decreolizing? The notion of decreolization plays an important role in Bickerton's theory on Creole languages. Bickerton (1980) characterizes decreolization as a process which occurs whenever a Creole is in permanent contact with a superstrate language. An essential condition for decreolization to take place is the existence of a continuum between the Creole language and the lexifier language. In this case the Creole and its lexifier language are not two strictly separate systems. Bickerton (1980)) proposes that at one end of the continuum there is a basilectal variety which represents the most conservative Creole, and at the other end there is an acrolectal variety which represents the decreolized form of the language concerned. Following Bickerton, and Valdmann (1991), I take decreolization to be any form of linguistic restructuring taking place in a Creole 'in the direction of the lexifier language*. Decreolization is reserved only for cases in which Creole languages lose their typical Creole structures in favor of the standard language structures. Marshall (1991), for instance, pointed at two concrete cases of decreolization in Mon Louis Island taking place in the noun and verb phrases, in which there is an increasing use of preposed definite articles, as in French, and a change in the position of the negator pa 'not': contrary to early texts, modern Louis Island Creole has pa postverbally, as in French.

40

Adone

Decamp (1971:349) distinguishes four possibilities of change in the life-cycle of a Creole language: 1) a Creole continues without change, 2) it may become extinct like e.g. Negerhollands, 3) it may further develop into a normal language, 4) it may gradually merge with the standard language. If we assume that all languages are dynamic systems, then possibility 1 is not realistic, because languages do not seem to survive without changes. Possibility 2 is a case of language death. Both possibilities 3 and 4 are instances of language change. MC illustrates possibility 3, since it is developing grammatically as already shown in previous sections and it is not moving closer to French.18 Possibility 4 is decreolization as evidenced, for example, in Reunion Creole, Louisiana Creole and Mon Louis Island Creole. All these Creole languages have one factor in common, namely that they all exist in a linguistic continuum with French, an undeniable fact documented by several specialists (cf. Carayol and Chaudenson 1977 for Reunion Creole, Neumann 1985 for Louisiana Creole, Marshall 1985 for Mon Louis Creole). As seen in section 3 the relation between French and MC is however different from other French-based Creoles and we cannot speak of a continuum between French and MC. Summarizing, it can be said that a necessary condition for decreolization is that there is a continuum between the Creole concerned and French, since decreolization usually takes place in a society in which the Creole language merges with the lexifier language. This is not the case for Mauritius. Inspite of the pressure of French on MC, it seems that MC is acquiring more complexity. The increasing use oflimem and ki for instance, shows clearly that MC is expanding grammatically.

5. Conclusion In Creole studies there has been a tendency recently to place the burden of proof on those who argue for nativization. In the theoretical part of the foregoing paper, I have given a brief account of what constitutes the processes of creolization and language change. On the whole it has been argued, following Bickerton, that creolization is 'catastrophic', while language change is considered to be gradual. In the empirical part, the case of MC has been studied. On the basis of data from both early and modern MC it has been shown that ki was present in creolization, but null subjects and the use of limem as a reflexive are relatively new developments in MC. 18

MC is today the national language of Mauritius, though not yet recognized officially.

Creolization and language change in Mauritian Creole

41

The fact that null subjects were not witnessed in the early creolization of MC suggests that recent changes in MC syntax must be responsible for the use of null subjects. In the pronoun and reflexive domain there are also changes which have taken place in modern MC, such as the increasing use of the pronoun + mem to express reflexivity. These changes, together with others19, can be viewed as the results of MC becoming more complex, a natural development in the evolution of a language. Given that creolization produced "a language that was structurally equivalent to any other natural language, on the basis of input that quite clearly was not structurally equivalent to a natural language" (Bickerton 1991: 28), it is obvious that, like other languages, this language continues to develop.

References Adamson, Liliane. 1993. The distribution of personal and reflexive pronouns in Sranan. Paper delivered at the conference of the society for pidgin and Creole linguistics in Amsterdam. June 1993. Arends, Jaques. 1989. Syntactic developments in Sranan. Nijmegen: Catholic University dissertation. Baker, Philip, and Chris Corne. 1986. "Universals, substrata and the Indian Ocean Creoles", in Muysken and Smith (eds.), 163-183. Bickerton, Derek. 1980. "Creolization, linguistic universal, natural semantax and the brain", in Richard Day (ed.) Issues in English Creoles, 1-18. Heidelberg: Groos. . 1981. Roots of Language. Ann Arbor: Karoma. 1984. "The language bioprogram hypothesis", The Behavioural and Brain Sciences 7, 173-221. 1986. "Beyond Roots: the five-year test", Journal of Pidgin and Creole Languages 1, 225-32. 1988 "Creole languages and the bioprogram", in Frederick Newmeyer (ed.) Linguistics: The Cambridge survey. Vol. 2, 268-284. Cambridge: CUP. .1991. "On the supposed 'gradualness' of Creole development", Journal of Pidgin and Creole Languages 6, 25-58. Bruyn, Adrienne, and Tonjes Veenstra .1993. "The creolization of Dutch", Journal of Pidgin and Creole Languages 8, 29-80. Carayol, Michel, and Robert Chaudenson. 1977. "A study in the implication analysis of a linguistic continuum: French-Creole", Journal of Creole Studies 1, 179-218. Carden, Guy, and William A. Stewart. 1988. "Binding theory, bioprogram, and creolization: Evidence from Haitian Creole", Journal of Pidgin and Creole Languages 3, 1-67.

Two other changes in the TMA system have been noted by the author. Pu is taking over the function of να, which was witnessed in creolization, ana fin is taking over some functions of ft'.

42

Adone

Chaudenson, Robert. 1981. Textes orioles anciens (La Reunion et He Maurice): Comparaison et essai d'analyse. Hamburg: Buske. Come, Chris. 1988. "Mauritian Creole reflexives", Journal of Pidgin and Creole Languages 3, 69-94. Decamp, David. 1971. "The study of pidgin and Creole languages", in Dell Hymes (ed.) Piginization and creolization of languages, 13-43. Cambridge: CUP. Degraff, Michel. 1992. "Is Haitian Creole a pro-drop language?" in Claire Lefebvre and Paul Law (eds.) Haitian and Null Subject Languages. Travaux de Recherche sur le Creole Haitien 11,121. Montreal: Universiti du Quobec ä Montreal. Deuchar, M. 1986. "Sign languages as Creoles and Chomsky's notion of universal grammar", in Modgil, S. and C. Modgil, (eds.) Noam Chomsky: consensus and controversy. Brighton. Gee, J.P., and W. Goodhart. 1985. "Nativization, linguistic theory, and deaf language acquisition", in William C. Stokoe (ed.) Sign language studies 49, 293-342. Goldin-Meadow, Susan. 1990. "Beyond the input given: the child's role in the acquisition of language", Language 66, 323-55. Jaeggli, Osvaldo, and Ken Safir (eds.). 1989. The null subject parameter. Dordrecht. Kluwer. Koopman, Hilda, and Claire Lefebvre. 1981. "Haitian Creole pa", in Pieter Muysken (ed.) Generative studies in Creole languages, 201-221. Dordrecht: Foris. Kouwenberg, Silvia. 1990. "Complementizer pa, the finiteness of its complements, and some remarks on empty categories in Papiamento", Journal of Pidgin and Creole Languages 5, 39-51. Lefebvre, Claire. 1986. "Relexification in Creole genesis revisited: the case of Haitian Creole", in Muysken and Smith (eds.), 279-300. Marshall, Margaret M. 1991. "The Creole of Mon Louis Island, Alabama, and the Louisiana connection", Journal of Pidgin and Creole Languages 6, 73-87. Meillet, Antoine. 1914. "Le probleme des langues", Scientia (Rivista di scienza) XV, 403-425. Neumann, Ingrid. 1985. Le Creole de Beaux-Bridge, Louisiane, etude morpho-syntaxique—textes— vocabulaire. Hamburg: Buske. Plag, Ingo. 1993. Sentential complementation in Sranan: on the formation of an English-based Creole language. Tübingen: Niemeyer. Perlmutter, David. M. 1971. Deep and surface structure constraints in syntax. New York: Holt, Rinehart, and Winston. Schuchardt, Hugo. 1917. "Sprachwissenschaft", Sitzungsberichte der Berliner Akademie der Wissenschaft, XXXVII, 518-529. Singleton, J. L., and E. Newport. 1987. "When learners surpass their models", paper presented to the Society for Research in Child Development. Baltimore. Syea, Anand. 1985. Aspects of empty categories in Mauritian Creole. Essex: University of Essex dissertation.

Creolization and language change in Mauritian Creole

43

Thomason, Sarah. G., and Terrence Kaufman. 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. Valdmann, Albert. 1991. "Decreolization or dialect contact in Haiti?" in Francis Byme and Thorn Huebner (eds.) Development and structures of creole languages, 75-88. Amsterdam: Benjamins. Virahsawmy, Dev. 1980. Trip sere logon amare. Mauritius. 1991. Lalangpeyna lezo. Mauritius.

Pieter Muysken and Norval Smith

Reflexives in the creole languages: an interim report 1. Introduction This paper is concerned with a class of function words in pidgin and Creole languages that can contribute to the still on-going debate about the role of universal and substratum features in creole formation, as well as to the debate about gradual versus abrupt creolization: reflexives. As we will see below, these tend to be innovative in Creoles with respect to their lexifier languages. While content words are often reflexes of the lexemes of the colonial languages, for function words, and particularly for reflexives, there is a much more indirect correspondence. Reflexives in creole languages raise all the issues that have been under discussion in the field in recent years: how does the lexical reconstitution of a grammatical morpheme class proceed: through contributing elements from substrate languages, the influence of a linguistic bioprogram, the gradual transformation of superstate patterns, or through processes of grammaticalization of content words? In addition these issues may link creole studies to the mainstream of theoretical linguistics, where the distribution and properties of reflexives have been central issues for many years (Chomsky, 1981; Reinhart and Reuland, 1991). Earlier accounts, typified by such survey studies as Hohn (1989), were mostly focussed on the forms the reflexives took and on their possible resemblance to the superstate languages, with some reference to the substrate issue. Reflexives are often found to consist of two parts, as seen in (1): (1) a. b. c.

koli(body-3) her/himself (cf. Fr. 'se/soi-meme') en srefi (3-self) her/himself (cf. Eng. "himself) my yet (1 -head) myself (cf. Eng. 'myself)

MARTINICAN SRANAN TOK PISIN

The nature of these complex forms will be discussed in much detail below. The orientation of the work in this area has changed due to the publication of Garden and Stewart's seminal article from 1988. They argue on the basis of the distribution of the reflexives in Haitian dialects, coupled with some scant diachronic data, that early Haitian had bare pronoun reflexives. This raises the issue of whether early Creoles are fully natural languages, since this may go against universal grammatical principles (defined in Chomsky's Binding Theory, 1981), or rather resemble the pidgins from which they are derived.

46

Muysken and Smith

Corne's work on Mauritian reflexives (1988; 1989) introduced a new dimension into this research: different sets of verbs often select different reflexive forms. Thus there is an intimate link as well with verb semantics and the way it is reflected in the argument structure and subcategorization frame of verbs. A dimension which needs to be explored further is to what extent principles of discourse organization influence the distribution of reflexive forms in those cases where several different forms are possible with a single verb. The state of the work on reflexives in Creoles is such that only an account in terms of a number of propositions is possible at the present time.

2. Diversity among the Creoles Creole languages exhibit a fair variety of reflexive structures. This section represents a preliminary attempt to classify the forms found. Due to lack of data, we will restrict ourselves to a small number of Creole languages here, so we do not wish to pretend that our conclusions are in any way definitive. In (2) we present an overview of the different types of reflexive forms encountered in the languages of the world. (2)

definition la. Ib. 2. 3a. 3b. 4. 4a. 4b. 4c. 5. 6. 7.

3rd person pronoun lst/2nd person pronoun reflexive pronoun pronoun + identifier possessive + identifier body word ('body1, 'head1, 'skin') pronoun + body word pronoun + identifier + body word possessive + body word a null form verb + reflexive affix verb + body incorporation

example Haitian// French melte French se himself myself Fon wu Saram. en sikin Saram. en seei sikin Papiamentu su kurpa Eng. bathe (comp. Sp. banar-se) Quechua riku-ku-n Bini

In table 1 the distribution of these forms over a number of Creoles is presented:

Reflexives in the Creole languages

TABLE 1:

47

Distribution of types of reflexives HT MA SR SA PA AN NE BE

la. 3pro Ib. l/2pro 2. refl 3 a .p r o + i d 3 b . poss + i d 4. body 4a. pro + body 4b. pro + id + body 4c. poss + body

χ χ

χ χ

χ

χ

χ

χ

5.

null

χ

χ

χ

χ

6. 7.

verb + refl af verb + body inc

x

χ χ x

χ χ x

x

χ χ x x

x χ

xx x χ

χ χ

(HT = Haitian, MA = Mauritian, SR = Sranan, SA = Saramaccan, PA = Papiamentu, AN = Annobon, NE = Negerhollands, BE = Berbice)

The most frequent forms are bare pronouns, pronoun + identifier combinations, and null forms. Only a few of the possibilities attested in the languages of the world are not attested in Creoles.

3. Overlap In several Creoles a number of competing forms exist, partially overlapping in use. We will illustrate this with two examples. The first concerns contemporary Papiamentu (Muysken, 1993). In Papiamentu no less than seven different forms have replaced the Ibero-Romance clitics: (3) a. b. c. d. e. f. g.

pana < Port, pano, Sp. pano 'cloth' kurpa < Port, corpo, Sp. cuerpo 'body' null reflexive possessive + kurpa pronoun pronoun + mes < Port, mesmo 'self, precisely' possessive pronoun + mes

Examples for the principal reflexive forms are given in (4):

48

Muysken and Smith

(4) a.

pena feita sofoka kiupa sofoka yuda su kurpa sisti su kurpa weta su mes yuda su mes sinti e tristo hana e

b. c. d. e.

'comb oneself 'shave (oneself)1 'exert oneself 'stifle' 'help oneself 'serve/stufF oneself 'look at oneself 'help oneself 'feel sad1 'find oneself

In (4a) we find the null reflexive, in (4b) the bare body word. The latter is limited to a specific set of, often idiomatic, expressions. (4c) illustrates the pronoun + body word construction, and (4d) the pronoun + identifier construction. In (4e), finally, there is a bare pronoun. Of course, these forms are not all usable interchangeably. In table 2 a rough outline of their distribution is given, along the dimensions [± physical action] ((n)phys.) and [inherent (inh.) versus transitive (tr.)]:

TABLE 2:

The rough distribution of Papiamentu reflexives 0

identifier phys.inh. phys.tr. n-phys.inh. n-phys.tr.

kurpa

pro + kurpa

pro + mes

pro

always some

some idiom

some

many

many many many

man)

In Papiamentu, mes can be used as an identifier, in addition to being a reflexive, but kurpa cannot: (5) a. b.

mi mes ta hunga Ί myself am playing.' * mi kurpa ta hunga

The main factor in the choice between mes and kurpa as reflexives seems to be whether the verb expresses a physical action or not. With some verbs both forms are possible:

Reflexives in the Creole languages (6)

el a hoga su mes/su kurpa na lama 'He has drowned himself in the sea.1

(7)

bo a yuda bo mes/?bo kurpa 'You have helped yourself.'

49

In other constructions, only mes is possible. These are principally cases where the 'self is purely mental or figurative: (8)

m'a ekiboka mi mes/*mi kurpa Ί made a mistake.'

(9)

el a hasi su mes/*su kurpa sokete 'He made himself out to be stupid.'

(10)

el a lolea/hode su mes/*su kurpa 'He made an asshole of himself

In cases such as (11), which is purely corporeal, kurpa but not mes is possible: (11)

el a dal su kurpa/*su mes na un palo 'He walked into a pole.' (lit. 'he walked himself into a pole')

We will return to the use of bare pronoun forms below. The distribution in Papiamentu is not dissimilar to that in Mauritian, as described by Corne (1988; 1989). The following four categories and distributions are distinguished by Come: (12)

null pronoun pro + mem pro + lekor

inherent reflexives inherent reflexives / transitive verbs / datives preferred transitive verbs; prepositional phrases preferred physical action verbs

Mauritian lekor has a distribution very much like Papiamentu kurpa. We will see below that the same holds for bare pronoun and null forms. A partially different picture is suggested by 18th century Negerhollands (Muysken & van der Voort 1991). Some examples are given in (13): (13) a.

Object reflexive wies ju selv na die Priester (Mat 8,4) show yourself to the priest

50

Muysken and Smith

b.

c.

d. e.

Object reflexive third person singular ... ha openbaar sie sei v ... (Mat 2, 19) TNS reveal himself Adverbial Prepositional Phrase reflexive Partie van die Skriftgeleerden ha seg bie sender selv Part of the Pharisees said among themselves (Mat 9, 3) Small Clause Prepositional Phrase reflexive en Jesus ha ruep sie twaelf Disciplen na sie (Mat 10, 1) and Jesus TNS call his twelve disciples to REFL Inherent reflexive maer die Volk ha verwonder sender (Mat 8, 27) but the people TNS marvel them

In Table 3 it is made clear that there are considerable differences amongst the different contexts where the selv forms occur most frequently:

TABLE 3:

pronoun pronoun + selv

Distribution of selv over different contexts in Negerhollands (van der Voort and Muysken 1994) DO 49 64

IO 12 6

ADVPP 25 97

SCPP 103 23

V 222 18

Forms with selv are very common in direct object position and particularly in adverbial phrases, but much less so elsewhere.

4. Analyticity Reflexives are formed with the analytic word formation procedures characteristic of Creole lexical extension in general. This statement needs no further comment here, given the examples presented.

5. The role of the lexifiers For French lexifier Creoles the colonial lexifier can only have played a limited role. The reflexives in the English-based Creoles are not directly inherited from the lexifier model, either (cf. Smith, 1987). Unlike the question words in the colonial lexifier languages, which tend to be uniformly mono-morphematic in structure (i.e.

Reflexives in the Creole languages

51

consist of one meaning-bearing element), as in (14), reflexive pronouns in these languages are different in their morphological structure (cf. 15): (14)

ENGLISH: DUTCH FRENCH: PORTUGUESE:

who wie qui quern

what wat que que

when wanneer quand quando

(15)

ENGLISH DUTCH FRENCH PORTUGUESE

myself me(zelf) me me

himself herself zich(zelf) se se se se

where waar ou onde

etc. etc. etc. etc.

etc. etc. etc.

Speaking in terms of loss and reconstitution, the problem raised by reflexives is the following. In Portuguese and in Spanish - the languages that have provided most of the lexicon for Papiamentu - we find constructions such as (16): (16) a. b.

Eu me vejo no espelho. Ί see myself in the mirror.1 Maria se corta en la mano. 'Mary cuts herself in the hand.'

(Portuguese) (Spanish)

Ibero-Romance reflexive clitic forms are the following:

(17)

me te se

nos os etc. se

As was the case with the other clitic pronouns, reflexive clitics were lost in the process of genesis of Papiamentu, perhaps in a phase when the language existed only as a rudimentary second language pidgin. The question is of course what replaced them. Superstate explanations are inadequate also. If superstate influence were the proper explanation in most cases, then we would expect the following patterns in French and English lexifier Creoles: (18) French-based English-based

lst/2nd Pronoun Possessive + Identifier

3rd Reflexive Pronoun Pronoun + Identifier

52

Muysken and Smith

Substandard English also has possessive + identifier for the 3rd person: theirselves, hisself. In fact we observe the pattern in table 4:

TABLE 4:

Reflexives in various French and English-based Creole languages

English-based Sranan Saramaccan Jamaican (?)

French-based a. Louisiana Seychelles Cayenne b. Haiti Trinidad St. Lucia

Pron -

χ χ Pron

Pron+Idnt Pron+Body χ χ(?) χ χ χ χ

χ χ χ Idnt+Pron

Pron+Head

χ χ χ χ Body+Pron Head+Pron χ χ χ χ

(Pron = pronoun; Idnt = identifier)

The most striking fact that springs to the eye here is the uniformity among the various systems. There is quite obviously no question of any major superstate influence. The analytic constructions Pron+Idnt (him-self), Pron+Body/Body+Pron (li-kolko-lf) - and to a lesser extent Pron+Head/Head+Pron (li-tet/tet-li) - are shared between English-based and French-based Creoles. There are two possible cases of superstate influence to be discerned. The first concerns the use of the bare Oblique pronoun as a reflexive in Seychellois and some other French-based Creoles. This differs slightly from the French facts in that the third person form is also an Oblique pronoun rather than a true reflexive form as in French, but we could put this down to a regularization of the system, removing what is a minority pattern in French. (26)

Seychellois ...ibey lipartu '...(he) washes himself all over.'

Reflexives in the creole languages

S3

More striking is the use of the Pron.+Ident. pattern in certain English-based Creoles. Once again we have a difference in the overall pattern, however, but this time in the majority of cases. TABLE 5:

Reflexives in English, Saramaccan, and Sranan English

Saramaccan

Sranan

myself yourself himself ourselves yourselves themselves

mi-seei ju/i-seei en-seei wi/u-seei unu-seei den-seei

mi-srefi ju-srefi en-srefi : wi-srefi unu-srefi den-srefi =

In fact only the two patterns indicated with an -' sign are equivalent, and then only if we ignore the fact that plurality is marked in English reflexives. The significant differences in the pattern of Personal Pronouns are as follows:

TABLE 6:

Contrasts between Sranan and English Reflexives Pronoun

English

Sranan

* Sranan

l s Pron.

I me my my-self you your your-self we us our our-selves

mi mi mi mi-srefi ju/i ju/i ju-srefi wi/u wi/u wi/u wi-srefi

*ai mi *mai *mai-srefi ju *juwa *juwa-srefi wi *osi *owa *owa-srefi

Poss. Ident. 2s Pron. Poss. Ident. Ip Pron. Poss. Ident.

If the Sranan reflexives were cognate with their English congeners we would have expected the non-occurring phonetic forms in the *Sranan column (cf. Smith 1987 for details of phonetic developments in the Surinam Creoles). This suggests that

54

Muysken and Smith

neither substrate nor superstate can in themselves offer an acceptable explanation of more than a small part of these phenomena, morphologically speaking. There were, it should be mentioned, cases of 'body' reflexives in Old French (Einhorn, 1974:69), but there is no reason to assume that there is a historical link between these and the 'body' reflexives in the Caribbean Creoles: (19)

por lor cors deporter 'to amuse themselves1

Notice finally that the forms in (20) correspond to each other, but not directly to a European model. (20) a. b.

Papiamentu Negerhollands

su mes, e mes sie-self, am-self

6. Grammaticalization One may hypothesize that self-type forms developed as discourse markers and slowly developed into a grammatical formative. This use of self is illustrated with an example from Quechua: (21) a.

b.

Xwan pay-ta riku-n Juan he-AC see-3 'Juan sees him/*himself Xwan pay-lla-ta-tak riku-n Juan he-DEL-AC-EMP see-3 'Juan sees himself/just him especially.' (AC = accusative; DEL = delimitative; EMP = emphatic)

The evidence for grammaticalization so far is limited, however. We will consider four cases here, namely Negerhollands, Papiamentu, Saramaccan, and Sranan. Did Negerhollands self evolve from an emphatic highlighter to a non-discourseoriented anaphoric marker? Consider first the data in Table 7. Here two periods in the early history of Negerhollands are contrasted, 1780 and 1800 (Van der Voort & Muysken, 1994). While the percentage of self forms (marked with S) increases in this period, it does so more for 1st and 2nd persons, where grammatical disambiguation is not needed, than for 3rd persons, where it is.

Reflexives in the Creole languages

TABLE 7:

1/2 1/2 S 3 3S

55

The relation between the person of the pronoun and the presence of self (in parentheses the non-5/e 3rd person forms) in Negerhollands. I II 99 48 = 33% 250 (229) 107 = 30% (51 = 15%)

22 19 = 46% 59 (59) 34 = 36 % (27 = 29 %)

(I = period around and before 1780; II = period around and shortly after 1800) A similar question can be posed for Papiamentu. Did Papiamentu kurpa evolve from an inalienably possessed body noun to a freely occurring anaphor? The form kurpa is mostly used with physical action verbs, taken in the widest sense of the word: (22)

E ta kana bai bini sin duna su kurpa sosiego. 'He walks back and forth without giving himself rest.'

(23)

E ta kita nan for di su kurpa. 'He takes them off himself/his body.'

(24)

?? M'a sina mi kurpa ingles. Ί taught myself English.'

Notice, however, that it cannot be used together with another inalienably possessed noun: (25) a. b.

M'a korta mi mes/*mi kurpa na mi man. Ί cut myself in the hand.' Mi ta dal mi mes/*mi kurpa na mi kabes. Ί hit myself on the head.'

Here man 'hand' is inalienably possessed by the subject. Even though the action is quite physical, kurpa is impossible. We can interpret this contrast by assuming that kurpa itself is an inalienably possessed element, and hence blocked in (25a). When the anaphor and the antecedent are not co-arguments of the same predicate, kurpa cannot be used either:

56

Muysken and Smith

(26)

Mi a mira un kulebra serka di mi/*mi mes/(*)mi kurpa. Ί saw a snake near me (near my body (as in a dream))'

(27)

Mi a mira mi mes/(*)mi kurpa kai. Ί saw myself fall.'

(Ί saw my body fall (as in a dream)1)

Body part (which take the form en-sinkii) reflexives in Saramaccan (Bolle & de Ruiter, 1993) are quite limited in their occurrence: (28) a. b. (29) a. b.

Janj ta si wan peentju f enj John sees a picture of himself, * Janj ta si wan peentju f en-sinkiij Jan, jei en-seeijta fan John heard himself speak, * Janj jei en-sinkiij ta fan

In Sranan skin and here "belly1 reflexives are also limited, although we do find some 18th century cases: (30) a. b.

anokanshekihemskin(1783) he can't move mi membre datti na mi belle I thought by myself ...

7. Bare pronoun forms Is there evidence for a pidgin or early Creole generalized bare pronoun reflexive (as argued by Garden & Stewart 1988) or are the bare pronouns a late development under the influence of superstate reflexive clitic systems (Come 1988)? Again, several languages provide relevant evidence on this point. The following data show that in present-day Papiamentu bare pronoun reflexives are clitics occurring with lexicallly specified verbs, and even there only with specific meanings: (31) a. b.

Mi ta sinti mi/mi mes/*mi kurpa un tiki tristo. Ί feel a bit sad.1 Mi ta sinti *mi/mi mes/mi kurpa dor di e deklo. I feel myself through the blanket.'

Reflexives in the Creole languages

57

With the two possibilities in (3 la) the two following structures correspond: (32) a. b.

Mi ta sinti+mi [ pro(anaphoric) un tiki tristo]. Mi ta sinti [mi mes un tiki tristo].

Some of the verbs taking bare pronoun reflexives are listed below; the verbs are generally inherently reflexive verbs denoting an abstract action: (33)

sinti e X hana e find gana e okupä e imagina e komporta e duna e kuenta diskulpa e kompromete e establese e dedika e

feel X oneself reach, find oneself occupy oneself imagine oneself behave oneself take into account (lit. give oneself account) excuse oneself commit oneself establish oneself dedicate oneself

Notice also that these verbs are often part of the more 'educated' vocabulary, almost certainly not dating from the early stages of the Creole. Another factor worth taking into consideration is the fact that many of these verbs contain more than two syllables: perhaps their weight favors a light reflexive object pronoun. A similar situation holds in 18th century Negerhollands (cf. Table 8), where the bare pronoun reflexive are all inherent reflexives: TABLE 8:

Verbs taking an inherent reflexive (those marked with an asterisk in Table 5 below, have also been attested as zero-reflexive in Negerhollands) bedink bekeer * beweeg boek * draej/dreij * erger keer * ... etc. etc.

think, (re)consider (lit: think by oneself) convert oneself stir, move (lit: stir oneself) stoop, lean down (lit: to lean oneself down) turn oneself get annoyed at (lit: to irritate oneself) turn oneself

58

Muysken and Smith

The data from 18th century Sranan (Bruyn, in prep.) merit much closer investigation; however, a similar picture emerges: (34)

da zo mi za beri dem zomma di kili den srefi it is thus I will bury the people who kill themselves.

This quotation from van Dyk (+ 1760) indicates that reflexives based on English 'self were present in the oldest known substantial body of textual material in one of the Surinam Creoles - in this case Sranan. There are also bare pronoun reflexives, but often with verbs that take an inherently reflexive direct object in the meaning intended: (35) a. b.

mi gi mi abra na hem (1783) I give myself (over) to him. bunne jorka kibri hem the good ghost hides himself

However, in this context se(freflexives are not excluded: (36)

wan libisomma membre, takki, hem kann helpi hem srefi, a kori hem srefi (1783) someone who thinks he can help himself is deceiving himself

As for 20th century Sranan, Adamson (1993) has argued that with a certain class of verbs bare pronouns can function as reflexive objects. Thus en in (35a) can be interpreted both as a reflexive and as a referential pronoun, non-coreferential with the subject: (37) a. b.

Johnj syi e%j ini a spikri. John saw himself in the mirror, John; syi ensrefij/*j ini a spikri (non-emph. reading).

The reflexive ensrefi in (35b) can only be interpreted as coreferential with the subject. Adamson (1993) argues that reflexive en in (35a) is in fact an object clitic on the verb. The Saramaccan data (Bolle & de Ruiter, 1993) suggest that this same development has not taken place there. In (36a) en can only be interpreted noncoreferentially: (38) a. b.

Janj si en*j/j John saw him. Jan, si en-sinkiij/en-seeii

59

Reflexives in the creole languages

8. Substrate There is also quite a variety of forms to be found in the various (potential) substrate languages:

(39)

GBE (FON): BINI: TWI: YORUBA:

wu gb me ho arami

(•body1) (•body1) etc. ('my body1) etc. Cbody my')

(Segurola, 1963)

If the form of the reflexives in the creole languages were purely a question of substrate or superstrata influence we would expect clear evidence one way or the other, taking the great variety of morphological structures into account. Let us first consider substrate influence. We will only analyse those cases where we appear to have some evidence for particular West African languages having played a major role in the formation of particular Creoles. Can we observe direct substrate influence in the reflexive formation in such languages? The following languages represent such cases:

(40)

Creole Language

Substrate Language

Berbice Dutch Saramaccan/Sranan

E. Ijo (Kalahari) Gbe (Fon)

Haitian Annobonese

Gbe (Fon) Bini

Let us consider these cases one by one. (41)

Berbice Dutch Pron + selfu

Kalahari bu "body1

Here there is no correspondence whatsoever.

(42)

Saramaccan Pron + sei ('self) Pron + sinkii "body" (< skin)

Fon wu "body'

(43)

Haitian kadav + Pron 'corpse1 kor + Pron "body1

Fon wu "body1

(Smith et al. 1987) (Smith 1987; Bakker 1987) (Lefebvre 1986) (Ferraz 1970)

60

Muysken and Smith

Here there is a partial semantic correspondence between Haitian and Saramaccan on the one hand, and Fon on the other. (44)

Annobonese ague 'body'

Bini egbe 'body'

The only case involving a complete equivalence (i.e. morphological, etymologicalphonological and semantic) of these four is the last, that of Annobonese/Bini. The cases of Haitian and the Surinam Creoles, here represented by Saramaccan, are semantically equivalent, but not equivalent either phonologically or morphologically. Overall the claim for substrate influence is not particularly strong for reflexives. The evidence for an African basis for the body reflexives is not very strong at present, but cannot be plausibly denied. What does the use of bare object pronouns as reflexives imply for the bioprogram hypothesis? (a) There is no Ibero-Romance reflex for Papiamentu kurpa, as there is for French Creole kor. (b) No body-part reflexives in Berbice Dutch. (c) Some West-African languages (e.g. Ewe) do not have body-part reflexives; this needs to be studied in much more detail. (d) The absence of grammaticalization of Papiamentu kurpa and Saramaccan sinkii pleads against direct calquing. Note that in cases where we can identify both substrate and superstrate the N is lexically supplied in one of three ways: (45) a. b. c.

the superstrate form the substrate form the substrate form reinterpreted or relexified in the superstrate language

In Saramaccan we have for instance examples of options a. and c. Note that where we have the actual substrate form, as in the case of Annobonese, this is associated with the morphological pattern of the substrate language - in this case the form "body1 alone - as forecast by recent versions of the Language Biogram Hypothesis incorporating the Lexical Learning Hypothesis. We can summarize the alleged substratum cases as in Table 9:

Reflexives in the creole languages TABLE 9:

61

Reconsideration of substratum cases

Saramaccan mi-seei Morphosyntax: Constituency: Universal Order: Sup. (English) Phon.Etym.: Sup. (English) Semantics: Sup. (English) mi-sinkii Morphosyntax: Constituency: Universal Order: Sup. (English) Phon.Etym.: Sup. (English) Semantics: Sub. (Fon)? Haitian kadav-mwe Morphosyntax: Constituency: Universal Order: Sub. (Fon) Phon.Etym.: Sup. (French) Semantics: Sub. (Fon)? Annobonese ogue Morphosyntax: Constituency: Sub. (Bini) Order: irrel. Phon.Etym.: Sub. (Bini) Semantics: Sub. (Bini)

Note that it is conceivably a frequent historical semantic process that reflexives develop from inalienable possessives through the use of words with the meanings "head1 or "body1. This does not necessarily imply that it is the default case that reflexives should be expressed by such words. So, all in all, the explanation of the causation of creole reflexive forms is much more complex than might have been expected. Different factors require to be taken into consideration when these are being analysed. The influence of universals in reflexives seems to be restricted to one aspect of morphosyntax. For this influence to even be present it is necessary for the substrate item not to have been inherited. It also appears that we have to reckon with the effects of relexification. However, extrapolating once again here from the very small amount of clear cases at our disposal we appear to have a situation where

62

Muysken and Smith

relexification lexically does not have maintenance of the morphosyntactic pattern associated with it. This does not augur well for much of the more grammatical interpretation of substratist claims. We suggest, in line with the ideas of Bickerton (1981), and to a lesser extent, those of Seuren and Wekker (1986), that the unexplained morphosyntactic patterns derive from universal aspects of the internalized grammar of the early speakers of the relevant Creole languages. It might be remarked that while the analytic type of reflexive appears to be dominant in Creole languages, the order of the two constituents is not invariable. However, recent versions of Bickerton's Language Bioprogram Hypothesis regard syntactic constituency as universal, but the order of constituents as languagespecific. Note that the universal structure of reflexives would then be: [Pronoun, N]. The problem remains of how the lexical filling of the N is to be defined. The Papiamentu case suggests that there are complex semantic motivations for the choice of either the identifier or inalienable possessive reflexive. If the use of kurpa derives from some African pattern, it was not simply a case of relexifying an African form, but a complex process of reinterpretation of African pattern to fit the [Pronoun, N] mould.

9. Conclusion The above survey of Creole reflexive systems has of necessity been incomplete. It has yielded some preliminary answers, but it has led to further questions as well. Before we can state a more definite set of conclusions, a number of issues need to be looked into. These include: (a) The relation between reflexive formation and the formation of other systems of grammatical morphemes, e.g. quantifiers. These resemble the compoundlike transparent question words of many Creoles. (b) To what extent are the systems found simply the result of the only word formation rules that these languages have available? To answer this question we must study the relation between the morpho-syntactic processes involved in function word formation and those involved in word formation in general. Are we dealing with compound formation, affixation, or phrase formation? Would the difference have syntactic implications? This very important set of questions can only be answered once we know more about the morphology of Creoles. (c) The reflexive systems of the Portuguese and Spanish-based Creoles, about which sufficient information is still lacking. If they are not transparent, and do not particularly resemble the related colonial languages, by what principles are they formed?

Reflexives in the creole languages

63

Most of all a thorough diachronic analysis of binding phenomena in the various stages of a single well-documented creole such as Sranan, is called for.

References Adamson, Lilian. 1993. "The binding of anaphors in Sranan". Paper presented at the Amsterdam meeting of the Society for Pidgin and Creole Languages, June 1993. Bakker, Peter. 1987. "Reduplication in Saramaccan", in Mervyn C. Alleyne (ed.) Studies in Saramaccan Structure, 17-40. Amsterdam and Kingston: University of Amsterdam and University of the West Indies. Bickerton, Derek. 1981. Roots of language. Ann Arbor: Karoma. Bolle, Jette, and Ellen de Ruiter. 1993. "Reflexives in Saramaccan". Paper presented at the Amsterdam meeting of the Society for Pidgin and Creole Languages, June 1993. Bruyn, Adrienne. In prep. The development of the nominal constituent in Sranan. Amsterdam: University of Amsterdam dissertation Garden, Guy, and William A. Stewart. 1988. "Binding theory, bioprogram, and creolization: Evidence from Haitian Creole", Journal of Pidgin and Creole Languages 3, 1-68. . 1989. "Mauritian Creole reflexives: a reply to Come", Journal of Pidgin and Creole Languages 4, 65-102. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. 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, 103-114.

Dechaine, Rose-Marie and Victor Manfredi. 1990. "Binding domains in Haitian", to appear in Natural Language and Linguistic Theory. van Dyk, Pieter. +1760. Nieuwe en nooit bevoorens geziene onderwijzinge in het Basiert Engels, of Neeger Engels (...). Amsterdam: Jacobus van Egmont. Einhorn, E. 1974. Old French: A concise handbook. Cambridge University Press. Ferraz, Luis. 1970. "The substratum of Annobonese Creole", Linguistics 173, 37-49 Holm, John. 1989. Pidgins and Creoles 1. Cambridge: Cambridge University Press. Lefebvre, Claire. 1986. "Relexification in creole genesis revisited: the case of Haitian creole", in Muysken and Smith, 279-300. Muysken, Pieter. 1993. "Reflexes of Romance verb + reflexive clitic combinations in Papiamentu", in Francis Byrne and Donald Winford (eds.) Focus and grammatical relations in the creole languages. Amsterdam: Benjamins.

64

Muysken and Smith

Muysken, Pieter, and Norval Smith (eds.) 1986. Substrata versus universal in Creole genesis. Amsterdam: Benjamins. Muysken, Pieter, and Hein van der Voort. 1991. "The binding theory and creolization: Evidence from 18th century Negerhollands reflexives", in Francis Byrne and Thorn Huebner (eds.) Development and structures ofcreole languages, 145-159. Amsterdam: Benjamins. Reinhart, Tanya, and Eric Reuland. 1991. "Anaphors and logophors: an argument structure perspective", in Jan Koster and Eric Reuland (eds.) Long-distance anaphora, 283-321. Cambridge: Cambridge University Press. Segurola, R.P.B. 1963. Dictionnaire Fon-Francais. Cotonou: Procure de l'Archidiocese. Seuren, Pieter, and Herman Wekker. 1986. "Semantic transparency as a factor in Creole genesis", in Muysken and Smith (eds.), 57-70. Smith, Norval. 1987. The genesis of the Creole languages of Surinam. Amsterdam: University of Amsterdam dissertation. Smith, Norval, Ian Robertson, and Kay Williamson. 1987. "The Ijo element in Berbice Dutch." Language in Society 16, 49-90. Voort, Hein van der, and Pieter Muysken. 1994. "Negerhollands reflexives revisited", in Jacques Arends (ed.) Creoles: the early stages. Amsterdam: Benjamins (to appear).

Philip Baker

Creativity in Creole genesis l. Introduction With regard to Creole genesis, most creolists can be broadly categorized as supporting one of the three positions which may be termed "superstratist", "universalist" and "substratist": (1) the superstratist position is that slaves identified the language of their owners as their target and, while they failed to master that, almost all aspects of the resulting Creole can be attributed to features of, or inherent in, the full range of regional, popular and standard varieties of die slave-owners' language; (2) the universalist position similarly assumes that the target language of slaves was the language of their owners and that, in failing to achieve the latter because of restricted access, their descendants repaired (some of) the deficiencies of the presumed resulting pidgin, turning it into a Creole, by drawing on their innate human faculty of language. (3) the substratist position is that slaves sought to maintain as much as possible of their linguistic heritage but the circumstances of slavery forced them to acquire a great deal of the language of the slave-owners, particularly its lexicon. Substratists nevertheless take the view that a significant proportion of the grammatical and other features of individual Creoles derives from one or more of the languages of slave immigrants. What concerns me is that all three positions assume that transplanted nonEuropeans and their descendants had particular aims - to acquire or maintain some pre-existing language - and that their efforts met with little success. Allied to this general impression of failure, there is an apparent assumption that most if not all features of Creoles derive from pre-existing languages or, by default, from the human faculty of language. There is also a widespread further assumption that Creole languages were formed very quickly, within a generation or two, and that, wherever they remain in contact with their lexifier language, little has happened since then that cannot be encompassed under the general heading of "decreolization". The possibility of a very different approach was sketched in an article two years ago in which I cast doubt on the validity of the notion of "target language" in pidgin and Creole development (Baker 1990). I suggested that the real if unconscious aim of people brought from a wide variety of ethnolinguistic backgrounds who were obliged to live and work together - slaves, slave-owners, and others belonging to neither of these categories - may instead have been to create a medium for interethnic communication (MFIC). In other words, pidgins and Creoles are

66

Baker

successful solutions to problems of human intercommunication rather than the unhappy consequences of botched language learning or failed language maintenance. In this paper, I propose to examine data from Mauritian Creole (MC) for evidence which supports any of the four approaches mentioned above, that is the superstratist, universalist and substratist positions and that outlined in Baker (1990) which I shall term, for want of a better word, "creativist".

2. The lexicon as a whole The Dictionary of Mauritian Creole (Baker & Hookoomsing 1987) contains well over 18,000 entries, of which at least 15,000 are individual words, the others being combinations of two or more MC words. Of the individual words, I estimate that around 13,500 derive from French and that 1,600, just over 10%, are of non-French origin. This appears to represent a larger number and proportion of non-French items than has been reported for other French-based Creoles. It is, however, important to understand just what these figures mean. A breakdown of a slightly smaller number of such items, taken from Baker (1982), is given in Table 1.

67

Creativity in Creole genesis

Table 1 Number of MC words of non-metropolitan French origin by source all categories

English Indie unknown Indian Ocean French (IOF) Dravidian Malagasy convergence trade names Bantu names of people Chinese Portuguese Manding Breton Wolof others TOTAL

554

292 257 94 72 69 50 34 32 27 15 12 7 6 5 9 1535

sources oflOF

sources of convergence

total nonEuropean words

16 15

26

333

16 6

14 21

102 96

1

27

60 15

31

9 94

9

16

12 5 1141

17 23 662

The information in Table 1 requires some clarification and comment: (A) Figures in the first column ("all categories") relate to assumed immediate source. (B) There can be little doubt that virtually all the English and Chinese words, trade names, most of the Indie (including all words of Bhojpuri origin) and many of the Dravidian words were adopted in MC in the post-slavery era (from 1835). These account, collectively, for more than half of the total non-metropolitan French vocabulary. (C) One sixth of the total are of unknown origin. (D) Indian Ocean French (IOF) includes words of ultimate non-metropolitan French origin which are known, or assumed for good reason, to have become established in one or more variety of Indian Ocean French earlier than in MC. The 1

The total of 114 includes all the non-European words identified as possible etyma of the fifty items in this category. (In several cases, possible European etyma were also identified but these are excluded from this table.)

68

Baker

two main criteria for inclusion in this category are (i) that they occur in a French text from India, Reunion or Mauritius earlier than in MC, or (ii) that they include, like zantak (from Malagasy antako), an initial element derived from a French article, indicating that they were adopted in MC from a local variety of French. The 94 words in this category can be reclassified according to the likeliest source from which they passed into IOF, as is done in the second column. (E) The category "convergence" includes MC items for which phonetically and semantically identical or very similar words are found in two or more of the languages represented in Mauritius. Fourteen of these words are of ultimate Arabic origin. These include salam which is firmly established in Wolof, Manding, Malagasy, the Bantu languages of coastal East Africa, and both the Indie and Dravidian groups of languages. It would clearly be absurd to attribute this to just one of these languages for it was undoubtedly known to the majority of nonFrancophone immigrants from the 1720s onwards. A somewhat similar situation arises in the case both of Bantu words which are also established in Malagasy, and of Indie words which are also found in Dravidian languages (or vice verso). The remaining examples of convergence consist of chance similarities between etymologically unrelated words in two or more languages. (B) and (C) together suggest that less than a third of the total of 1535 words in the above table were established in MC by the time slavery was abolished and can be reliably attributed to languages other than French. This is 3% or so of the total vocabulary, a figure similar to that found for Antillean Creoles which did not experience a change of ownership or a massive influx of non-Francophone immigrants in the 19th century. Thus, the lexicon of MC does not provide any more support for substratists than other French-based Creoles. Conversely, since up to 97% of the lexemes of MC in 1835 were of French derivation, this seems entirely consistent with the superstratist and universalist assumption that slaves identified French as their target language (rather than aimed to preserve their linguistic heritage or create an MFIC). Nevertheless, other interpretations are possible. The figures in Table 1 show that the Indie and, to a lesser extent, the Dravidian contributions to modern MC are far greater than those of Malgasy and African languages. This may be due to the fact that one Indie language, Bhojpuri, continues to be the first language of 25% of the Mauritian population while two Dravidian languages, Tamil and Telegu, have only comparatively recently ceased to be regularly employed in private homes. For generations, Mauritian speakers of an Indian language have also been fluent speakers of MC, thus facilitating the two-way flow of lexical items. This might suggest that, in an earlier era, slaves may similarly have been fluent speakers of an ancestral language and MC, and that far more words of African or Malagasy origin may once have been current in MC. If the small number of non-French words in MC texts predating 1835 were invoked as evidence against this, substratists might legitimately point out that there are few such texts and all of them were recorded by Whites whose presence alone might

Creativity in creole genesis

69

explain the paucity of non-French words employed in the circumstances. While I suspect that there may well have been far more non-French words in 18th and early 19th century MC than texts ofthat era suggest, there remains one important factor which distinguishes African languages and Malagasy from Indian languages in Mauritius: the evidence of 19th century missionaries indicates that the former did not survive the death of the last foreign-born slaves, i.e. they were not spoken by their locally-born children, whereas Indian languages in general were - and in the case of Bhojpuri still are - passed on from generation to generation. This may be because Indian immigrants arrived at a far more rapid rate, and in far greater numbers, than slaves had done earlier, and that the "camps" where they were typically housed provided an environment in which their languages could continue to flourish. The ovenvhelming French contribution to the lexicon of MC may be explained as follows. In order to communicate verbally with people who did not speak their own or a related language, non-Francophone immigrants had to find words their interlocutors might understand. As indicated in the discussion of convergence above, there were a few non-French words known to members of different ethnolinguistic groups but, in general, French was the only language to which all slaves had at least some direct exposure. This meant that French words would more readily have been understood in interethnic encounters than words of other sources. Thus the bulk of the vocabulary used for interethnic communication was bound to be of French origin, whatever the particular linguistic aims of participants may have been. If this is correct, it is not the proportion of MC words which are of French origin which is significant so much as the ways in which those words differ from French in pronunciation, form and function. The remainder of this paper will thus be devoted to examining each of these differences in turn.

3. Differences of pronunciation - the phonemic inventory of MC Standard French is generally reckoned to have 36 phonemes, of which sixteen are vowels, three are semi-vowels and seventeen are consonants. Eleven of these are not found in the basic set of Mauritian Creole (MC) phonemes: the five front rounded vowels and semi-vowel, the vowels /a/ and Id, and the consonants /// and / 3/; while French Id and /o/ are, in MC, merely allophones of /e/ and /o/, respectively. (This is a more radical change from French than is the case with any of the Caribbean area Creoles, all of which have /J7, /3/, Id and /o/.2 ) Three other sounds which occur in MC have some claim to be considered phonemes: the velar nasal [η] and the affricates [t]] and Some doubt is cast on the status of [ε] and [a] in Antillean Creoles in Baker (1991).

70

Baker

The velar nasal occurs in 200 or more MC words, about half of which are of French derivation and result from a regular phonological process: French nasalized vowel + [g] corresponds to MC oral vowel + [η] word-finally or before a following consonant. However, the fact that high status words tend to resist such processes means that, for a great many Mauritians, lang [lag] '(standard) language' (Fr. langue) and lang [lag] 'angle' (Fr. I'angle) constitute a minimal pair. The other half of the words containing the velar nasal are of non-French origin and include such things as bang 'marijuana' (of Indie origin), murung 'horse radish tree' (Tamil), pingo "bird sp.' (Bantu) and filing 'petrol station' (English) in all of which the sequence of oral vowel + [η] does not alternate with nasalized vowel + [g]. Thus the velar nasal has to be considered a bonafide phoneme in MC. A minority of Mauritians consistently distinguish the affricates [tj] and [ds] from the sequences of corresponding plosive + semi-vowel [tj] and [dj], and thus contrast jet [dset] 'jet plane' and dyet [djet] 'diet', on the one hand, and cule [tjule] 'point at which the chorus takes over from the lead sega singer* and ti ule [tjule] 'wanted', on the other. For such people, the affricates have phonemic status. For the majority of Mauritians, however, both [if] and [tj], and [ds] and [dj], are pairs of allophones which Bhojpuri-speakers tend to pronounce as affricates while non-Bhojpurispeakers generally treat them as sequences of plosive + semi-vowel. There is also some, rather limited, textual evidence to suggest that such variation existed even before the majority of the population of Mauritius was of Indian descent. Throughout its recorded history of two and a half centuries, there is no indication that more than 25 of the 36 phonemes of French were ever fully integrated in MC. If, as both superstratists and universalists suppose, slaves had indeed initially identified French as their target language, it would seem that they must have abandoned their attempt to acquire the full set of French phonemes at a very early stage. While it might be possible to explain the "missing" phonemes by some kind of phonological component in a Bickertonian type bioprogram (cf Wise 1990), the degree of access to French which locally born slaves would have had would not generally have been so limited as to prevent them from subsequently acquiring these phonemes if French had remained their target language. There is also the problem, addressed below, of why Antillean Creoles should have four more French phonemes than MC. The substratists could, if they addressed themselves to the question, provide a straightforward explanation for MC's set of phonemes: the eleven French phonemes not found in MC are also lacking in almost all the main languages introduced in Mauritius by slaves and indentured labourers - Malagasy, Wolof, Tamil, Bambara, Makhuwa and other Bantu languages of Mozambique and Tanzania, and Bhojpuri3 - while virtually all of these same languages do have the three non-French MC Only Wolof of these languages has a basically seven rather than five peripheral vowel system. Wolof, Bhojpuri and Tamil also have a vowel phoneme similar to schwa.

Creativity in creole genesis

71

phonemes, the velar nasal and the palatal affricates. Substratists might similarly account for the fact that Antillean Creoles have four more French phonemes than MC: most languages taken to the Caribbean area by substantial numbers of slaves Kwa languages, and the relevant Bantu languages of Zaire and Angola - have seven vowel systems and [f] is well represented among them (although [3] is not). Impressive though this may be, this does not really provide evidence that nonFrancophone immigrants actively sought to maintain their linguistic heritage. For example, the retroflex consonants of Makhuwa and Tamil are not found in MC, nor are the labiovelar plosives so typical of Kwa languages found in any Antillean creole. What the phonemic inventories of both MC and Antillean Creoles really suggest is that the populations of these territories selected for their medium of interethnic communication precisely those phonemes which were shared by the majority of the languages represented there during the early decades of their settlement. Such a strategy would minimize the number of new phonemes to be acquired by the polyglot population and, to that extent, this is consistent with the creativist approach rather than any of the others.

4. Differences of form Most differences of form between French and MC result from a very strong tendency for nouns and verbs in the latter to have an initial consonant. In some cases, this is achieved by dropping the initial vowel of the French word, as in tanltand& Tiear' (