The interplay of relexification and levelling in creole genesis and development(*).
This paper bears on the relationship between relexification and levelling, two processes that play a role in creole genesis and development. It is a well-documented fact that situations where creoles develop involve several substratum languages and one (or at least one major) superstratum language. Relexification of several substratum languages on the basis of a single/major superstratum language provides the creole community with a common lexicon, hence a common language. As a mental/cognitive process, relexification is an individual process. Consequently through relexification, the specificities of the various substratum languages are reproduced in the incipient creole, thus creating what might be referred to as different "dialects" of the new language. Relexification of several substratum languages, on the basis of a single superstratum language, can be viewed as the major source of variation in an incipient creole. This provides a sound explanation for the fact that different substratum languages may contribute different features to a particular creole. Dialect levelling, as discussed in the literature on dialects in contact, refers to the reduction of variation between dialects of the same language, in situations where these dialects are brought together. On the basis of three sets of data from Haitian creole, it is shown that, in creole development, dialect levelling operates on the output of relexification. The role of levelling in creole genesis and development accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages.
Dialect levelling, as discussed in the literature on dialects in contact (see e.g. Domingue 1980, 1981; Siegel 1995, 1997; Trudgill 1986; etc.), refers to the reduction of variation between dialects of the same language in situations where these dialects are brought together. As Siegel (1997: 128) puts it, "dialect differences are reduced as speakers acquire features from other varieties as well as avoid features from their own variety that are somehow different. This may occur over several generations until a stable compromise dialect develops."
Dialect levelling is thus a social process, which can be described as a negotiation between speakers of different dialects of a given language. The process results in agreement on a form or on the properties of a given form in the context where either competing forms or competing properties for a given form coexist. Examples of well-documented cases of dialect levelling in noncreole communities include Bhojpuri as spoken in Mauritius (Domingue 1980, 1981) and English as spoken in America, Australia, New Zealand, and South Africa (Trudgill 1986).
In the literature on pidgins and creoles, it has been proposed that the process of levelling plays a role in the development of these languages (see e.g. Harris 1991; Mufwene 1990, 1994, 1996; Muhlhausler 1980; Siegel 1997). For example, Siegel (1997: 131-132) asserts that
Mixing and levelling may [...] be important in the development of stable pidgin and creole languages [...]. When the various versions of the superstrate are [...] used as the main means of communication among speakers of different substrate languages (in other words, when vernacularisation occurs), and when the speakers begin to view themselves somehow as a "community," then levelling begins.
The above references all acknowledge the fact that speakers of incipient creoles manifest variation among themselves. This raises two questions. The first one can be formulated as follows: what is the source of the variation found in incipient creoles? There is a consensus in the literature to the effect that communities where pidgin and creole languages emerge generally involve several substratum languages, spoken by the majority of the population, and one superstratum or lexifier language spoken by a relatively small social group (e.g. Whinnom 1971). One option is to say that the variation observed in an incipient creole comes from dialectal variation found in the superstratum language. Another option is to say that it comes from the variation manifested between the various contributing substratum languages. Building on Lefebvre (1998), I argue below that the bulk of the variation found in an incipient creole is attributable to the variation manifested between the substratum languages (sections 3 and 4), and even to the variation found within the individual substratum languages of this incipient creole (section 5). The second question can be formulated as follows: how does the variation observed between the various substratum languages of an incipient creole get transferred into the creole? I demonstrate below that this is achieved through the process of relexification.
This paper addresses the topic of levelling in creole communities as it relates to the relexification account of creole genesis developed in the course of the UQAM Haitian projects (see Lefebvre 1998 and the references cited therein). I would like to acknowledge the important contribution of John Lumsden in the articulation of these two processes, that is relexification and levelling (see mainly Lefebvre and Lumsden 1989a, 1994a, 1994b; Lumsden and Lefebvre 1994). In this paper, I am building on theoretical work that we did together. I take full responsibility for the implementation of the data within this framework. I will assume the general theory of creole genesis extensively outlined and documented in Lefebvre (1998) on the basis of Haitian creole and its contributing languages. In order to preserve the autonomy of this paper, however, I will summarize the pertinent facts and assumptions in each section. The question of how relexification and levelling are implemented in everyday life is not discussed here (for a discussion of this topic, see Durie forthcoming). The processes identified above as playing a role in the genesis and development of creole languages take place in communities and, hence, they interact with the social components that define the features of these communities. Although social factors are not discussed here, I assume that they interact with linguistic processes (particularly levelling) in a way similar to that described in Labov's (and his associates') meticulous work on the interplay of linguistic and social factors in communities. Principles that guide levelling are not discussed either (for a preliminary discussion of this point, see Jourdan 1985 and Siegel 1997).
The paper is organized in the following way. Section 1 summarizes the theoretical framework referred to above. The basic idea is that, while relexification of various substratum lexicons on the basis of a single superstratum language provides the speakers of an early creole community with a common vocabulary, it is the major cause of variation in an incipient creole. In this view, levelling is hypothesized to operate on the output of relexification, thus reducing the variation produced by the relexification of several lexicons. Section 2 summarizes the situation that prevailed in Haiti when Haitian creole was formed. Sections 3, 4, and 5 implement the theory outlined in section 1 on the basis of a subset of data drawn from Haitian creole and its contributing languages. A full discussion of the general research methodology may be found in Lefebvre (1998: 52-78). In this paper, I consider cases involving distinctive properties of specific lexical entries in the substratum languages and compare them with those of the corresponding lexical entries in the creole. Third person plural pronouns and plural markers, ways of encoding reflexivity, and deictic terms will be discussed in turn. The discussion of these data goes beyond that in Lefebvre (1998), and it considers questions and comments raised since 1998 by various scholars. Section 6 concludes the paper.
1. Relexification and levelling
This section shows how the process of relexification is predicted to account for the variation found in the early creole communities and how levelling is hypothesized to operate on the variation produced by relexification of various substratum languages on the basis of a single lexifier language.
The first formal definition of relexification was provided by Muysken (1981: 61): "Given the concept of lexical entry, relexification can be defined as the process of vocabulary substitution in which the only information adopted from the target language in the lexical entry is the phonological representation."(1) For the purpose of this paper, I will adopt the representation of the process proposed in Lefebvre and Lumsden (1994a, 1994b). It is illustrated in (1).
A relexified lexical entry thus has the semantic and syntactic properties of that of the lexicon it has been copied from and a phonological representation derived from a phonetic string in the lexifier language.(2) In line with the terminology adopted in Lefebvre and Lumsden (1994a, 1994b), I will refer to the process of assigning a new phonological representation to a copied lexical entry as relabelling. In the representation in (1), relabelling proceeds on the basis of phonetic strings found in the superstratum language. Furthermore, it is hypothesized that, if no form is available in the lexifier language to relabel a copied lexical entry, this lexical entry may be assigned a null form at relabelling, represented as [[Phi]] in the schema in (1).(3)
According to Muysken's (1981: 62) proposal, relexification is semantically driven: "For relexification to occur, the semantic representations of source and target language entries must partially overlap; otherwise, the two entries would never be associated with each other. Other features of the two entries may, but need not, be associated with each other." Muysken's idea of partial semantic overlap between the source and target lexical entries is preserved in the representation in (1) by specifying that the meaning of the phonetic string selected to relabel a copied lexical entry is deduced from its use in specific semantic and pragmatic contexts. In Lefebvre (1998), it is assumed that copying applies to all lexical entries and that it is relabelling that is semantically driven. Thus, only functional categories with some semantic content (e.g. determiners, demonstrative terms, etc.) may be assigned a new label when relexified. Functional categories that have no semantic content (e.g. case markers, operators, etc.) can be copied but not relabelled. They are assigned a null form at relabelling. As is noted in Lefebvre and Lumsden (1994b), practically speaking, this means that these lexical entries are not pronounced.(4)
In Lefebvre and Lumsden (1994a), it is proposed that the functional-category lexical entries of the copied lexicon are relabelled on the basis of major-category lexical items of the superstratum language and that the distributional properties of the superstratum form must be similar to those of the copied lexical entry. Relabelling of functional-category lexical entries is thus constrained by what the superstratum language has to offer in terms of major-category lexical items. The semantics and distribution of a lexical item of the superstratum language have to be appropriate in order to provide a phonetic matrix to a copied functional-category lexical entry. The constraint imposed by the superstratum language on the relabelling of functional-category lexical entries is discussed in Lefebvre (1998). Constituent order in the superstratum language also acts as a constraint in establishing constituent order in the incipient creole. Such a case is documented in detail in Lumsden (1991) (see also section 5 of this paper).
In Lefebvre and Lumsden (1989a, 1994a, 1994b; see also Lefebvre 1998: 10, 30-41), it is claimed that, in creole genesis, the process of relexification is used by speakers of the substratum languages as a tool for acquiring a second language, the superstratum language. As is pointed out in Lefebvre and Lumsden (1989a), this account is a further development of the second-language acquisition theory of creole genesis. For example, Alleyne (1971, 1981), Andersen (1980), Chaudenson (1993), Mufwene (1990), Schumann (1978), Thomason and Kaufman (1991), and Valdman (1980) have proposed that pidgin and creole languages constitute a crystallized and incomplete stage of second-language acquisition. As is discussed in Lefebvre (1998), without relexification, this approach to creole genesis does not explain why creole languages have crystallized the way they have (see also Lefebvre 1984; Lefebvre and Lumsden 1989a). The relexification hypothesis does explain why creole lexicons reflect the properties of both their superstratum and substratum source languages the way they do (see ). This raises the question of whether relexification plays a role in second-language acquisition. As is discussed in Lefebvre (1998:33-35 and the literature cited therein), substratal features in radical creoles have long been considered as cases of transfer, shown to play a major role in second-language acquisition (see e.g. Weinreich 1953). On the one hand, in the theory of creole genesis advocated here, the type of data claimed to be associated with the notion of transfer in creole genesis corresponds to the result of the process of relexification (see Lefebvre 1998). On the other hand, Lumsden (1999), Mous (1995), and Siegel (1997) argue that, in some cases of second-language acquisition, relexification (or paralexification, in Mous's term) plays a role. Thus, in both creole formation and more simple cases of second-language acquisition, relexification may be argued to be used as a means of acquiring another language. As is discussed in Lefebvre (1998 and the references cited therein), the amount of relexification in both cases appears to be a function of the amount of exposure to the lexifier language. Little exposure to L2 triggers recourse to relexification, whereas more exposure to L2 increases the degree of successful acquisition.(5)
As has been pointed out many times (e.g. Lefebvre and Lumsden 1989a, 1994a, 1994b), relexification is a mental process and thus an individual process. Typically, situations where pidgins and creoles emerge involve several substratum languages, as we saw above. Each individual speaker of the substratum languages relexifies his or her own lexicon in the early creole community. Relexification of several lexicons on the basis of a single superstratum language provides the early creole community with a common vocabulary. But, as is also pointed out in Lumsden and Lefebvre (1994c), in relexification, speakers of various substratum languages reproduce the idiosyncratic semantic and syntactic properties of their own lexicons; thus the product of relexification is not uniform across the early creole community. In this view, the variation found in an early creole community is the product of relexification of various substratum lexicons. For example, relexification of the lexicons of languages X, Y, and Z on the basis of a single superstratum language will yield three slightly different lexicons in an incipient creole. This is schematically represented in (2).
(2) Substratum lexicons X Y Z ... Early creole lexicons L1 L2 L3 ...
The features that are common to all the relexified lexicons (that is, to all the substratum languages) will most probably be maintained in the creole (see also e.g. Gambhir 1988: 77; Harris 1991: 199; Sankoff 1984; Siegel 1997; Singler 1988; Thomason and Kaufman 1991; Trudgill 1986: 98). The idiosyncratic features, however, are those that are subject to levelling. The criteria governing the selection of one or the other of the competing forms require further research (see Mufwene 1986, 1990). A preliminary discussion of this point may be found in Siegel (1997).
In the scenario of creole genesis presented in Lumsden and Lefebvre (1994c), relexification applies when the speakers of the substratum languages are targeting the superstratum language. It is hypothesized that, when they stop targeting the lexifier language and start speaking to each other using the relexified lexicons, that is, the early creole, they begin learning aspects of the incipient creole lexicons from each other, and they begin levelling out the differences between the relexified lexicons. This is conceptually in line with the authors cited in the Introduction. The originality of the above proposal lies in the claim that, in this case, levelling reduces the variation produced by the relexification of the various substratum lexicons. It is thus hypothesized that, in the context of creole genesis, levelling operates on the output of relexification.
As is mentioned above, the process of levelling starts when the speakers of the incipient creole (thus the creators of the creole) begin using the creole to communicate among themselves. By hypothesis, the creators of the incipient creole themselves may get involved in levelling. This does not mean, however, that the process of levelling takes place only in the initial formative stage of the creole, nor that it levels out within a few years all the variation in the incipient creole. As will be seen in the data sections (sections 3 to 5) of this paper, even if there are early cases of levelling, there are also cases that take much longer to be resolved. Furthermore, as is illustrated in Lefebvre (1998), there are cases of variation in modern Haitian that parallel variation found in one of its substratum languages, which shows that, even after approximately three hundred years, levelling has not affected all parts of the lexicon of this language. The reason some differences are levelled out early, others later, and still others not at all in the development of a creole remains a topic for future research. However, it should be pointed out that there is no reason to believe that the situation observed in creole communities with respect to variation is any different from cases of linguistic variation observed in noncreole communities (see e.g. Labov 1972). Likewise, there is no reason to believe that the process of levelling involved in cases of linguistic change should apply differently in creole communities than in noncreole communities. We know from the literature on linguistic change that some changes may occur rather rapidly, while others take more time to be completed.
In the view advocated above, while the purpose of relexification in creole genesis is to provide speakers of different languages with a common vocabulary, the purpose of levelling is to reduce variation between the various dialects (or even idiolects) produced by relexification of different substratum lexicons.
Finally, as has been pointed out in Lumsden and Lefebvre (1994c: 8), "Since the process of reconciling the various dialects of the creole created by relexification of the various substratum languages must depend on the interactions of these dialects in the community, it follows that different communities will arrive at a somewhat different compromise." Evidence that this is the case may be found in regional dialect differences observed in a given creole (for Haitian, see Orjala 1970; Romain 1959; Tinelli 1970) or in differences among creoles formed on the basis of similar languages (e.g. both Haitian and Martinican creoles arose in historical situations involving basically the same languages [see Singler 1993, 1996]; nevertheless, these two languages present lexical differences [see e.g. Goodman 1964], which shows that levelling did not produce exactly the same results in these two communities).
The above proposal provides a sound explanation for what has been referred to in the literature on creole studies as the "cafeteria principle" -- a term first used by Dillard (1970), which was borrowed by Bickerton (1981: 49). As Bickerton (1981: 49) puts it, "As things stand, we are asked to believe that different African languages contributed different rules and features to particular creoles [...]. It is [...] absurd to suppose that a creole could mix fragments of Yoruba, Akan, Igbo, Mandinka, and Wolof [...]." On the one hand, the proposal that relexification plays a major role in the formation of creole languages accounts for the fact that specific features of the various substratum languages will be manifested in the early creole. On the other hand, the proposal that levelling operates on the output of relexification provides a principled explanation for the observation that different substratum languages may contribute features to a given creole. This is because, as is pointed out in Lefebvre (1998: 394), in cases of competing forms, it is not always the same linguistic group that comes out the winner. For example, Lefebvre (1998: 219-246) shows that, in the Haitian case, Fongbe speakers were quite successful with regard to the clausal determiner, whereas the Ewe speakers won on other cases (see also section 3 of this paper). Finally, as is pointed out in Lumsden and Lefebvre (1994c), the role of levelling in creole genesis accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages. These points will be illustrated in the data sections (sections 3 to 5).
2. The linguistic situation in Haiti at the time Haitian creole was formed
This section summarizes the facts pertaining to the linguistic situation in Haiti at the time Haitian creole was formed.
On the basis of the historical study that he carried out on the economy, demography, and linguistic diversity of early Haiti, Singler (1996) provides evidence that Haitian creole was probably created between 1680 and 1740. He demonstrates that the bulk of the Haitian population, at the time Haitian creole was formed, was adult. On the basis of this fact, Singler (1996: 199) concludes that the principal agents of creole genesis must have been adults. As is documented in Singler (1996: Table 3), the majority of the Haitian population during that period was of African origin and people of mixed race; thus day-to-day contacts between French speakers and the Haitian-African population were very limited. Furthermore, Singler (1993: Table 3 and 4) shows that the African people were from the following language families: Atlantic, Mande, Kwa (including Akan and Gbe), Gur, Nigerian Benue-Congo, Ijoid and Bantu. These African languages, although not entirely disparate from a typological point of view, are mutually unintelligible (see Lefebvre 1998:58-61 and the references therein).
Thus, as is pointed out in Lefebvre and Lumsden (1994b), the situation that prevailed in Haiti between 1680 and 1740 presented the two basic prerequisites for the emergence of a creole language. First, the community was a multilingual one, speaking languages that were not mutually intelligible. Given the number of languages present, it was obviously in need of a lingua franca. This lingua franca was needed to permit communication not only between the speakers of the substratum languages and those of the superstratum language, but also between the speakers of the substratum languages themselves (see e.g. Foley 1988; Lefebvre 1998: 1; Singler 1988: 47; Thomason and Kaufman 1991). Second, the African people could not have had much direct access to native speakers of French during that period, given the figures provided by Singler. Thus, between 1680 and 1740, the social conditions in Haiti were such as to permit the development of a creole language. In the view advocated in Lefebvre and Lumsden (1994a, 1994b), the first step toward the creation of the lingua franca was achieved by relexifying several African lexicons(6) on the basis of a single language, in this case, French.
This brings me to the question of what type of French data the creators of Haitian were exposed to. This topic is documented at length in Lefebvre (1998: 62-65), based on the literature. Here, I only summarize the most salient points. What variety of French did the French colonists speak in Haiti? There is a consensus in the literature (see e.g. Faine 1937; Hull 1979; Valdman 1978, 1979; etc.) that the bulk of the French colonists who went to the Caribbean in the seventeenth and eighteenth centuries were from Western and Central France (i.e. Normandy, Ile-de-France [Paris], Poitou, Saintonge, Brittany). The dialects spoken in this area of France constitute the langue d'oil (as opposed to the langue d'oc). These dialects were not necessarily mutually intelligible among themselves or with the variety of French spoken in Ile-de-France (Paris). (For various views on this matter, see the papers in Mougeon and Beniak 1994.) But, as noted by Meyer-Lubke (1909: 123), these regional dialects were influenced by Parisian French: "On sait tres bien que, dans les provinces de l'Ouest, les dialectes ont subi, plus profondement que partout ailleurs en France, et meme avant l'epoque de la colonisation de l'Amerique, l'influence de la langue officielle" [It is very well known that, in the Western provinces, more than anywhere else in France, the dialects were influenced by the official language, even before the colonisation of America]. Furthermore, there appears to be a consensus in the literature on the characteristics of the French colonists of the seventeenth and eighteenth centuries (see e.g. Chaudenson 1992; Mougeon and Beniak 1994; Poirier 1994; Rivard 1914; Wolf 1994; etc.). First, the colonists were mainly from the domaine d'oil. Second, even though they were born in the provinces, they had lived in cities before leaving France. Third, most of them spoke French. Finally, about half of them were literate (that is, they knew how to read and write). According to Juneau and Poirier (1973: 191-193), the variety of French spoken by the colonists was close to the variety of French spoken in Ile-de-France in the seventeenth and eighteenth centuries: "Les traits syntaxiques sont facilement rattachables a des tendances de la syntaxe de l'epoque. Il en est de meme pour les traits morphologiques" [The syntactic features can very easily be linked to syntactic tendencies of that time. The same is true of the morphological features]. Wolf (1991) draws a similar conclusion.(7) There is also a consensus in the literature that the French colonists who went to North America in the same period spoke the same languages as those who went to the Caribbean (see Juneau 1972; Valdman 1978, 1979; etc.). Various authors hypothesize that the seventeenth- or eighteenth-century French spoken by the French colonists, regardless of whether they went to North America, the Caribbean, or Africa, was similar. For these reasons, in Lefebvre (1998:62), it is assumed that the French colonists in Haiti spoke a variety of French similar to that referred to as seventeenth- or eighteenth-century French in dictionaries and grammars.
It is perhaps worth noting that, as is pointed out in Lefebvre (1998: 63), some scholars who have worked on French-based creoles have adopted a different methodology, which consists in looking for traces of French regional dialects in the creoles. For example, Picard, the French dialect spoken in Picardy, is often cited as the source of a creole form or feature. According to Juneau (1972), however, only 2% of the French colonists sent to the Americas were from Picardy. Consequently, it is unlikely that the bulk of the African population in the French Caribbean was exposed to this dialect of French in a way that would have had a significant impact on the French-based creoles. Furthermore, it is likely that, in the colonies, the French colonists who spoke regional dialects natively used the variety of seventeenth- or eighteenth-century French that they had in common, as attested in the historical work of the authors cited above. Given these considerations, I assume that the creators of Haitian were mainly exposed to the variety of French spoken in Ile-de-France in the seventeenth and eighteenth centuries.
The creators of Haitian, who spoke the various mutually unintelligible African languages enumerated above, were thus exposed to a variety of seventeenth- or eighteenth-century French, which probably contained regionalisms from the dialects of the langue d'oil. In line with the general theory of relexification outlined in section 1, I take this variety of French to have provided the phonetic matrices for the relabelling of the lexical entries of the African lexicons and to have imposed the surface constraints upon this process (as discussed in section 1). The phonological representations of some Haitian lexical entries directly reflect this situation. For example, the Haitian verb bay `to give' is derived from the French form bailler `to give', no longer in use in modern standard French. Another example in point is provided by Lumsden (1991), who shows that the creators of Haitian creole abandoned the prenominal complement of their original grammar in order to conform to the consitituent order of French, where the nominal complement always follows the nominal head (e.g. la maison de Pierre `the house of Peter'). So, the variety of French the creators of Haitian creole were exposed to provided the superstratum material on which Haitian creole was formed.
Lefebvre's (1998) extensive comparison of the lexical properties of Haitian creole with those of its contributing languages argues, however, that the bulk of the variation found in modern Haitian creole reflects not the variation found in the variety of French that the creators of Haitian were exposed to, but rather the variation found between the substratum languages. I refer the reader to Lefebvre (1998 and the references cited therein) for a thorough demonstration of this point across a significant portion of the lexicon. The examples discussed in sections 3 to 5 of this paper illustrate this point. This situation is congruent with the fact that, as was mentioned earlier, the creators of Haitian did not have enough exposure to French to acquire the properties of its lexical entries or its syntax.
In light of this historical background, I now turn to the discussion of the data.
3. Third person plural pronouns and plural markers
This section discusses the variation between the substratum languages of Haitian with respect to the ways third person plural pronouns and plural markers are encoded. I look at how the lexicon of Haitian creole compares with the lexicons of its substratum languages. The data show that, in this case, levelling must have taken place early in the development of Haitian. The content of this section builds on Lefebvre (1998:84-87), based on preliminary work in Lumsden and Lefebvre (1994c).
Fongbe has two distinct morphemes to encode the third person plural pronoun and the plural in noun phrases. This is shown in (3).
(3) FONGBE a. ascn le b. Ye yi. crab PL 3pl leave `the crabs' `They left.'
In contrast, Ewe has the morpheme wo, indicating both notions (see e.g. Wallace 1995:10; Westerman 1930:47, 58). This situation is also observed in other West African languages: Yoruba (see Goodman 1964:45-46; Migeod 1911:170), Mandingo (see Delafosse 1929:175), Igara and Igbira, and Vai, a Mande language (see Mufwene 1986:138).
Haitian differs from Fongbe in this case and follows the pattern manifested by Ewe and the other West African languages mentioned above. As is shown in (4), the Haitian morpheme yo serves as both a plural marker and a third person plural pronoun.
(4) HAITIAN a. krab yo b. Yo pati. crab PL 3pl leave `the crabs' `They left.'
As Mufwene (1986) asks, how can the difference between Haitian and Fongbe be accounted for within the framework of the relexification hypothesis? As is suggested in Lefebvre (1998) and Lumsden and Lefebvre (1994c), the answer to this question lies in the interaction between relexification and levelling as it is hypothesized to apply in creole communities (see section 1).
With Goodman (1964) and Sylvain (1936), let's hypothesize that the phonological representation of Haitian yo is derived from the strong form of the third person plural pronoun eux in French. As is argued in Lefebvre (1998:84-87), this is a likely hypothesis. First, Fongbe ye, Ewe wo, etc., share some element of meaning with French eux: all serve as a third person plural pronoun. Second, according to Brousseau (n.d.), the phonological derivation of French eux as Haitian yo is a predictable one. Note that, as is pointed out in Lefebvre (1998:411, note 4), although the French form eux is pronounced [yo] in Gascon and Auvergne French (see Goodman 1964), these French dialects were not represented in Haiti at the time Haitian creole was formed (see section 2). Thus, it is unlikely that Haitian yo was phonologically derived from [yo] in Gascon or Auvergne French. Third, as is shown in Lefebvre (1994, 1998:86), the substratum pronominal forms and French eux share distributional properties. The pronoun eux in French occurs as an emphatic form before a clitic, as in (5).
(5) FRENCH (=  in Lefebvre 1994) Eux, ils mangent du riz chaque jour. them, they eat PART rice every day `Them, they eat rice every day.'
Furthermore, in French, the third person plural pronoun eux also occurs as an emphatic pronoun after a nominal constituent, as is shown in (6a). Note the neutralization in gender that extends to ils as well as to eux in (6b), also discussed in Gougenheim (1973).
(6) a. FRENCH Les gars, eux, ils ... DET-PL guy them they `The guys, them, they ...' b. POPULAR FRENCH (=  in Lefebvre 1994) Les filles, eux, ils ... DET-PL girl them they `The girls, them, they ...'
In short, the French form eux was arguably a suitable candidate for relabelling the third person plural pronoun of Fongbe, Ewe, and other related languages, but as is also noted in Lefebvre (1998:86), the Haitian form yo does not derive all of its properties from French eux. For example, eux cannot serve as a plural marker in French nominal structures.
With these assumptions in mind, Lumsden and Lefebvre (1994c) proposed the following general account of the historical derivation of Haitian yo. On the one hand, the Ewe speakers and speakers of similar languages had one lexical entry to relexify. Relexification of this lexical entry yielded yo, serving as both a third person plural pronoun and a plural marker in the early Haitian lexicon. On the other hand, the Fongbe speakers had two forms to relexify. Presumably, Fongbe speakers also relexified the third person plural pronoun of their lexicon as yo. As a result of interacting with other speakers of early Haitian, who had relexified their lexicons on the basis of Ewe-type lexicons, Fongbe speakers noticed that yo was also used to encode plural in nominal structures. As a result of levelling, they abandoned the form le of their own lexicon and extended the usage of yo so it also served as a plural marker in noun phrases. Presumably, this is an early case of levelling for, to the best of my knowledge, there is no attested dialect of modern Haitian that has two distinct forms for the third person plural pronoun and the plural marker.
An alternative proposal was submitted to me for discussion. According to this proposal, the Fongbe speakers could have relexified both the third person plural pronoun and the plural marker occurring in noun phrases as yo on the basis of the two surface positions of French eux (see  and ), thus yielding two homophonous lexical entries in the early Haitian lexicon. This proposal is a likely one for, as is shown in Lefebvre (1998), there are cases where different substratum lexical entries can be argued to have been relexified on the basis of the same French form. This proposal is also an attractive one for, if confirmed, it would illustrate a case where a difference between the substratum lexicons would have been obliterated in the creole as a result of the restricted number of available forms in the lexifier language. The plausibility of this scenario is weakened, however, in view of the following points. First, as we saw in section 1, relabelling is semantically driven; thus functional categories that have some semantics may be relabelled (e.g. the definite determiner), while those that do not have some semantics (e.g. case markers) may not be relabelled. Do plural markers in noun phrases have enough semantics to undergo relabelling? Having thought about this question for a long time, I still have no definite answer to give, although I tend to think that they do not. But, for argument's sake, let's suppose that, since plural markers have some semantics by virtue of being plural, they can be relabelled. Then, the question we should ask is whether or not the third person plural pronoun eux occurring in (6) is an appropriate form to relabel the Fongbe plural marker le. French eux in (6) and Fongbe le share the property of being plural and of occurring postnominally. However, while eux is pronominal, hence referential, le is not: it is only plural. While I can easily conceive of a third person plural pronoun being relabelled on the basis of a third person plural pronominal form (as in the first scenario), I can hardly conceive of a plural marker being relabelled on the basis of a third person plural pronominal form (as in the second scenario). Pending an independent theory on lexifier-language processing in second-language acquisition, it is not useful to pursue the discussion of this point any further here. Given the problems raised by the second scenario, however, I conclude for the time being that the first scenario is the most plausible one.
Yet a third scenario can be considered, according to which levelling would have taken place between speakers of different African languages prior to relexification. On this view, at the time of relabelling, only one lexical entry having both functions for, say, Ewe and Fongbe speakers had to be relabelled. Although this proposal is attractive, I think it is unlikely for the following reasons. Recall from the Introduction that levelling is a process that applies when varieties of the "same language" are brought together. These varieties have to be mutually intelligible and they have to be the means of communication between the speakers for levelling to take place. As we saw in section 2, the various West African languages that were brought together in early Haiti, although not completely disparate from a typological point of view, were not mutually intelligible. Moreover, as is argued in Capo (1984), the Gbe languages constitute a dialect cluster or continuum, the two extremes of which (Ewe and Fongbe) are not mutually intelligible. Although one cannot exclude a priori the possibility that speakers of more Gbe dialects (e.g. Fongbe and Gungbe or Fongbe and Ajagbe) tried to communicate among themselves in their native tongues in spite of their differences, the hypothesis that the West African speakers, in early Haiti, learned each other's languages in order to communicate among themselves is not supported by the historical facts. As we saw in section 2, the languages involved were numerous and their speakers needed a lingua franca (that would provide them with a common vocabulary) to help them communicate among themselves. This is why Haitian creole was created. On this view, and as is discussed at length in Lefebvre (1998), the African population in early Haiti was busy creating this lingua franca. Their target was French rather than the other West African languages present. Thus, on the basis of known historical facts, if any levelling took place prior to relexification, its impact on the creole was minor, if noticeable at all.
Pending further theoretical work on language processing in second-language acquisition and given what we know of the historical situation in Haiti, I conclude that the first scenario is the most plausible one; that is, the interaction between relexification and levelling, as hypothesized in the first scenario above, accounts for the fact that, in this particular case, Haitian differs from Fongbe and pairs with languages of the Ewe type. The specific example under discussion further illustrates an early case of levelling where no competing forms are visible in the creole.(8)
This section shows that the variation between the substratum languages of Haitian with respect to how they encode reflexivity was carried over into the creole through relexification, creating variation in creole. This variation offers a situation for levelling to take place. In contrast to the case discussed in section 3, the variation among competing Haitian forms to encode reflexivity is still visible in the creole, thus making the process of levelling observable as an ongoing one in the Haitian community. The discussion of this case (based on Lefebvre 1998:159-183) illustrates the fact that not all cases of levelling take place in the early stage of creole formation.
Fongbe has a reflexive anaphor but no BODY-part reflexives. According to Hazoume (1990:62), all Gbe languages encode reflexivity by means of a personal pronoun + SELF. In Fongbe, this lexical item is -dee. Kinyalolo (1994) argues that the semantic and distributional properties of this Fongbe lexical item are similar to those of the English -self For example, in English, pronouns combine with -self (e.g. He washes him-self; We wash our-selves; etc.). Similarly, in Fongbe, the strong pronominal forms (but not the syntactic clitics, as is emphasized by both Brousseau 1995a and Kinyalolo 1994) combine with -dee, as is illustrated in (7).
(7) FONGBE (=  in Brousseau 1995a) a. [N.sub.i] na xu nye-[dee.sub.i]. 1sg DEF-FUT kill me-SELF `I will kill myself.' b. [Bayi.sub.i] mc e-[dee.sub.i]. Bayi see she-SELF `Bayi saw herself.'
Fongbe has no BODY-part reflexives. Both Brousseau (1995a) and Kinyalolo (1994) report that, in Fongbe, the word meaning `head' does not participate in the expression of reflexivity. Furthermore, they also show that wu, the Fongbe word meaning `body', does not participate in such constructions either (for an extensive discussion of this point, see Lefebvre 1998:167-171). As is documented in Awoyale (1986), however, BODY-part reflexives are widely used in other Kwa languages. For example, in Yoruba, the word meaning `body' is ara; this word may appear with personal pronouns, yielding a reflexive interpretation (e.g. ara mi `body' + 1sg pronoun = `myself'). Awoyale (1986) provides similar examples from Igbo, Urhobo, Ebira, and Bassa-Nge. Similar data in Efik and Akan are reported by Faltz (1985). Sylvain (1936) discusses a Wolof reflexive formed on bob `head', and Faltz (1985) mentions a Fula reflexive formed on hooremum (lit. `his head') `himself'.
Thus, the substratum languages of Haitian appear to encode reflexivity either by means of a reflexive anaphor or by means of a BODY-part reflexive. This is summarized in (8).
(8) a. reflexive anaphor: Gbe languages b. BODY-part reflexives: Yoruba, Igbo, Urhobo, Ebira, Bassa-Nge, Efik, Akan, Wolof, Fula
The hypothesis outlined in section 1 predicts that the early Haitian lexicons manifested both a reflexive anaphor and BODY-part reflexives and that this variation was afterward subjected to levelling. Is there evidence supporting this hypothesis?
As has been extensively documented in the literature, reflexivity can be encoded in Haitian in two ways. These are shown in (9). In (9a), the object pronoun is interpreted as a reflexive even though there is no overt reflexive morphology (see Brousseau 1995b; Carden and Stewart 1988; Dechaine and Manfredi 1994; Kinyalolo 1994; Muysken and Smith 1995). In (9b) and (9c), reflexivity is encoded by means of tet `head' + pronoun and by means of ko `body' + pronoun, respectively. The last two forms are also attested in Goodman (1964:57) and in Sylvain (1936:63-64).
(9) a. HAITIAN (=[1a] in Brousseau 1995b) [Mwen.sub.i] we [mwen.sub.i] nan glas la. 1sg see me in mirror DET `I saw myself in the mirror.' b. HAITIAN (= [2a] in Brousseau 1995b) [M.sub.i] ap touye tet [mwen.sub.i]. 1sg DEF-FUT kill head me `I will kill myself.' c. HAITIAN (= [6d] in Brousseau 1995b) [Li.sub.i] blese ko [li.sub.i]. 3sg hurt body him `He hurt himself.'
Let's assume that all these three possibilities have been produced by relexification. Hence, on this view, the phonological representations of the BODY-part reflexives of Haitian would be derived from French tete `head' and corps `body', respectively. In French, these two words cannot be assigned a reflexive interpretation. As has been pointed out by Carden and Stewart (1988:11), Goodman (1964:57), Muysken and Smith (1995: 276), and Sylvain (1936:65), among others, there are instances in Old French of the word corps `body' being used as a reflexive (e.g. Por lor cors deporter `to amuse themselves'; see Muysken and Smith 1995:276), but there is no evidence showing that this use of the French word corps was available to the creators of Haitian. In fact, grammarians report that the reflexive use of corps had disappeared by the middle of the sixteenth century (see e.g. Brunot 1905). Therefore, like Goodman (1964:57), I conclude that it is unlikely that the creators of Haitian were exposed to a variety of French that used BODY-part reflexives. This conclusion challenges that in Chaudenson (1973). As a result of having been created through the process of relexification, these two Haitian lexical entries would have the properties of the corresponding substratum lexical entries. Consequently, in addition to being used to refer to the actual body part or body, they are also used to encode reflexivity, just as in the substratum lexicons that they were relexified from, that is, those listed in (8b).(9)
Now, what happened in the case of the Gbe speakers who had a reflexive anaphor? In Lefebvre (1998:163-167), it is hypothesized that these speakers looked for a form in the superstratum language to relabel the -dee anaphor of their lexicon. It is a well-documented fact in the literature that French does not have a reflexive anaphor of the type of -self in English or of the type of -dee in Fongbe. French expresses reflexivity in other ways (see Lefebvre 1998:163-167, and the references therein). Furthermore, the French word meme (sometimes unofficially suggested as a potential form for the relexification of the Fongbe reflexive anaphor) did not have the suitable properties, as is extensively argued in Lefebvre (1998:164-166). Building on an idea suggested by John Lumsden during a research seminar in 1993, Lefebvre (1998:164) proposes that, since the Gbe speakers did not find an appropriate form in the superstratum language to relabel the -dee anaphor, the Haitian lexical entry relexified from -dee was assigned a null form at relabelling (see ), yielding a covert reflexive anaphor in the Haitian lexicon (see [9a]).(10)
As is argued in Lefebvre (1998:166), independent evidence for this claim comes from Berbice Dutch and Gullah, whose lexifier languages, Dutch and English, respectively, have a lexical entry meaning `-self'. According to Robertson (1993:307), Berbice Dutch may express reflexivity by means of a pronoun +SELF. The Berbice Dutch form corresponding to English -self is -selfu, whose phonological form probably derives from Dutch -zelv. Similarly, according to Mufwene (1992:169), Gullah may express reflexivity by means of a pronoun +SELF. The Gullah form corresponding to English -self is -self, whose phonological form obviously derives from English. Another similar example is provided by data from Saramaccan. The substratum languages of Saramaccan are very much the same as those of Haitian (see Smith 1987). Like Berbice Dutch and Gullah, Saramaccan has an overt -self anaphor: seei (Veenstra 1996:43) in addition to having BODY-part reflexives (Muysken and Smith 1995). Thus, while the creators of Berbice Dutch, Gullah, and Saramaccan found appropriate forms in Dutch or English to relabel the lexical entry meaning `-self', the creators of Haitian did not find an appropriate form in French because French does not have such a form. Consequently, they had to assign a phonologically null form to this lexical entry at relabelling. Assuming the representation of a pronoun +SELF in English to be as in (10a), following Chomsky (1981), the representation of a pronoun +SELF in the languages discussed above would be as in (10b).(11)
(10) [ILLUSTRATION OMITTED]
In summary, the Gbe speakers in (8a) produced a Haitian lexicon that had a phonologically null reflexive anaphor, accounting for the data in (9a), while the speakers of the other substratum languages in (8b) produced a Haitian lexicon with BODY-part reflexives, accounting for the data in (9b) and (9c). This is summarized in (11).
(11) GBE (8a) OTHER KWA (8b) HAITIAN a. reflexive anaphor overt ?(12) covert b. BODY-part reflexives -- + +
The reflexive forms of Haitian constitute a clear example of how relexification proceeds when various substratum lexicons offer different forms to encode a particular notion. The relexification hypothesis thus provides a straightforward account of the fact that we find several reflexive forms in Haitian creole (and in other creoles as well; see Muysken and Smith 1995). These forms reflect the differences found in the substratum lexicons. This suggests that, in the early creole, there were different Haitian dialects reflecting the differences between the substratum languages. Hence, speakers who had relexified Gbe lexicons would use pronominal forms and a phonologially null anaphor. Speakers who had relexified their lexicons based on other languages such as those mentioned above would use BODY-part reflexives. While communicating with each other, speakers of the early Haitian dialects presumably learned the forms that were originally foreign to them. As a result, it is not rare to find speakers of modern Haitian who use all three reflexive forms.
The availability of several forms encoding the same notion constitutes an ideal situation for levelling to take place. Recall from the Introduction that the process of levelling refers to the reduction of variation between dialects or even idiolects in situations where they are brought together. Hence, it is likely that levelling would operate to reduce the number of forms encoding reflexivity in the various Haitian communities. Evidence that this is the case is provided by Carden and Stewart (1988), who show, on the basis of varius types of data (e.g. geographical, historical, etc.), that levelling is now taking place with respect to the ways to encode reflexivity.(13) According to them, the beginning of this process goes back to 1790 and it is not completed as of today. This shows that, while some cases of levelling are completed in the early stages of the creole (e.g. the case of the plural marker in section 3), other cases take much longer to be resolved. This should not come as a surprise if cases of levelling are considered as cases of regular linguistic change that gradually (over generations) take place in communities.(14)
5. Demonstrative terms
Demonstrative terms constitute yet another case where levelling appears to operate on the output of relexification from various substratum languages. I limit my discussion to the two Haitian deictic lexical entries: sa and sila. Speakers of Haitian appear to divide into two groups with respect to these lexical items. Some have both terms in their lexicon; others only have sa. Most Haitian speakers I worked with on this topic have both terms. The following authors attest the existence of these two forms in the Haitian lexicon that they discuss: d'Ans (1968), DeGraff (forthcoming), Etienne (1974), Faine (1974), Ferere (1974), Goodman (1964), Hall (1953), Joseph (1989), Orjala (1970), Peleman (1978), Pompilus (1976), Sylvain (1936), Tinelli (1970), Valdman (1978), and Valdman et al. (1981). The following authors mention speakers who only have the form sa: Valdman (1996) and Vilsaint (1992). I will discuss these two lexicons in turn.
5.1. The Haitian lexicon that has both sa and sila
Consider the nominal structures of Haitian and Fongbe in (12). The demonstrative terms are glossed as DEM and are translated as `this/that/these/those' for now. I return below to the interpretive patterns associated with these terms.
(12) HAITIAN (=  in Lefebvre 1998: 78) bag [mwen [Phi]] sa/sila a yo FONGBE alcke [nye tcn] (e)lc/(e)ne c le ring me GEN DEM DEF-DET PL `these/those rings of mine (in question/that we know of)'.
The striking fact about the nominal structures in (12) is that, in both Haitian and Fongbe, a possessor phrase (in the genitive case), a demonstrative term, the definite determiner, and the plural marker may all be present. In both languages, the order of these lexical items is the same. Furthermore, all the "determiners" occur postnominally. This is in direct contrast with French, where the determiners are prenominal and where the three types of determiners are mutually exclusive, as is shown in (13).(15)
(13) [ILLUSTRATION OMITTED]
In Lefebvre (1997: 184-193), it is extensively argued that the other syntactic properties of Haitian sa and sila contrast with the properties of the French demonstrative terms (e.g. gender, animacy, syntactic category, distribution, etc.) and that there is no French form that has the same properties as sa and sila in Haitian (a conclusion that challenges that in Chaudenson 1993).(16) Furthermore, in Lefebvre (1997: 193-196), it is extensively argued that the details of the syntactic properties of Fongbe (e)lc and (e)ne parallel those of Haitian sa and sila, respectively (e.g. gender, animacy, syntactic category, distribution, etc.).
On the basis of these facts (and semantic facts discussed below), Lefebvre (1997: 196-198) proposes a relexification account of the two Haitian lexical entries: while the syntactic properties of Haitian sa are derived from Fongbe (e)lc, the phonological representation of Haitian sa is derived from French ca [sa]. Haitian sa and French ca are argued not to have the same syntactic properties. Likewise, while the syntactic properties of Haitian sila are derived from Fongbe (e)ne, the phonological representation of sila is derived from either one of the following French forms: cela pronounced [sela] or [sla], both yielding sila in Haitian, or celui-la pronounced [shila], also yielding sila in Haitian. This is illustrated in (14).
(14) [ILLUSTRATION OMITTED]
I now turn to the discussion of the semantics of the lexical items involved. Three semantic patterns have been attested so far for sa and sila. These are given in (15). The first pattern is identified as G1 and it manifests a distal distinction between the two terms. In this grammar, sa is [+ proximate] and sila is [- proximate]. This grammar is attested in the following sources: Goodman (1964: 51)(17) and Tinelli (1970: 28).(18) None of the Haitian speakers I worked with so far manifested this grammar. The second pattern is identified as G2. In this grammar, while sa is [[Alpha] proximate], thus a general deictic term, sila is [- proximate].(19) This pattern is attested in Lefebvre (1997).(20) The third pattern is identified as G3. In this grammar, there is no distal distinction between sa and sila, which can both be used to point at objects that are either close to or far from the speaker. (I will come back to the semantics of this grammar below.) There are speakers with whom I worked who have this grammar. It is also attested in Ferere (1974: 103), Joseph (1989: 112), Valdman (1978: 194) and Valdman et al. (1981).
(15) HAITIAN G1 sa [+ proximate] sila [- proximate] G2 sa [[Alpha] proximate] sila [- proximate] G3 sa [[Alpha] proximate] sila [[Alpha] proximate]
What is the source of these three interpretive patterns associated with sa and sila? Are the French forms from which sa and sila are phonologically derived the source of these three patterns? The forms ca, cela and celui-la are all [[Alpha] proximate] (see Lefebvre 1997:192-193 for an extensive discussion of this point, based on literature on seventeenth-century French; see note 14 of this paper for a list of authors consulted on the semantics of French deictic terms). So the semantics of these demonstrative terms could at best account for Haitian G3, if Haitian sa and sila in G3 were indeed equivalent in meaning to the French forms ca and cela/celui-la, respectively. I doubt it, however, for reasons I come back to below. Also, as is shown in Lefebvre (1997), there are no demonstrative terms in French that lexically encode the value [- proximate]; hence, the [- proximate] feature of sila in G1 an G2 in (15) has no equivalent in French. Furthermore, French lexically encodes the feature [+ proximate] with the form ceci. In this case, the distal distinctions that are lexically encoded in French are as in (16).
(16) FRENCH ceci ca/cela/celui-la `this' `that/this' [+ proximate] [[Alpha] proximate]
The interpretive pattern in (16) does not correspond to any of the patterns found in Haitian (see ). I thus conclude that the interpretive patterns associated with Haitian sa and sila in (15) are not derived from French.
By hypothesis, the semantics of sa and sila would come from the substratum languages. (The phonological forms are obviously derived from the superstratum language.) So let's look at what we find in Fongbe. The interpretive patterns that are attested for (e)lc and (e)ne are shown in (17). The first grammar (G1) shows a distal contrast between the two terms. This grammar is attested in Anonymous (1983) and Segurola (1963). I also worked with a speaker who has this distinction. In the second grammar (G2), (e)lc is a general deictic term ([[Alpha] proximate]) and (e)ne is [- proximate] (see Lefebvre 1997, forthcoming). In the third grammar (G3), there is no distal distinction between the two terms.(21)
(17) FONGBE G1 (e)lc [+ proximate] (e)ne [- proximate] G2 (e)lc [[Alpha] proximate] (e)ne [- proximate] G3 (e)lc [[Alpha] proximate] (e)ne [[Alpha] proximate]
The Fongbe interpretive patterns in (17) parallel the Haitian ones in (15). Is this a coincidence? No, it follows directly from the relexification hypothesis, which predicts that both the syntactic and semantic properties of the Haitian lexical entries sa and sila should be provided by the corresponding ones in the substratum languages.
Although demonstrative terms in Kwa languages have not been given much attention in the literature, the following interpretive patterns are attested for languages other than Fongbe. The G1 pattern of Fongbe in (17) is also attested for the two Ewe corresponding lexical entries (see Rongier (1979: 521, 1995; Westerman 1930). The G2 pattern of Fongbe in (17) is also attested for Ewe in Lefebvre (1997) and Wallace (1995: 6) and acknowledged by Chris Collins (personal communication). The G1 pattern in (17) is also attested in other Kwa languages such as Ega (see Bole-Richard 1983: 379), Eotile (see Herault 1983b: 415), and Krobou (see Mensah 1983: 454).
So, presumably, and following the methodology adopted above, the interpretive patterns of the substratum lexicons were carried over into the creole in the way schematized in (18).
(18) substratum lexicons G1 G2 G3 (see ) Haitian lexicons G1 G2 G3 (see )
The first two grammars identified for Haitian in (15) and Fongbe in (17) involve a distal distinction between the two demonstrative terms. The third one, however, does not. For Haitian, DeGraff (forthcoming and the references cited therein: e.g. Ferere 1974; Joseph 1989; Valdman 1978) claims that sa and sila are interchangeable. As DeGraff (forthcoming) puts it, "spontaneous examples abound where sa and sila are used in nominal phrases referring to objects that are both near and equally distant." This remark, and similar ones cited in DeGraff, suggests that sa and sila are synonyms; they apparently show no difference in meaning. Could it be, however, that in G3 another dimension of deixis is involved? We know from the literature that there are various dimensions to deixis: person, time, place, speaker's point of view, etc. (see e.g. Fillmore 1975; Grenoble and Riley 1996; Gundel et al. 1993; Levinson 1983; Maclaren 1980; Piwek and Cremers 1996; etc.). Although I did not investigate the matter, I would not be surprised if, in this case, there were another dimension of deixis involved, and this for two reasons. The first is that there are West African languages whose deictic terms have been reported to be interpreted along other dimensions of deixis. For example, Jaggar and Buba (1994) have argued that, in Hausa, deictic terms do not always manifest a distal distinction (in much the same way as G3 in  and ) and that they can be interrupted along nonspatial deixis dimensions (anaphoric and symbolic). The second reason is that, as is discussed in Lefebvre (1998: 213-217), both Fongbe and Haitian creole (but not French) have a series of markers that give the point of view of the speaker on the proposition. Could it be that deictic terms are to the nominal structure what these markers are to the clause structure? I submit this as a promising topic for future research.
5.2. The Haitian lexicon that has sa but not sila
Speakers who have sa but not sila in their lexicon are attested in the literature (e.g. Valdman 1996; Vilsaint 1992). There is one speaker with whom I worked who has only sa. Are there substratum languages of Haitian whose lexicons have only one demonstrative term? Strikingly enough, in the Atlas des langues kwa de Cote d'Ivoire (Herault 1983c), several such languages are attested. For example, for Abidji (Tresbarats 1983: 62-63), Adioukrou (Herault 1983a: 145), and Alladian (Duponchel and Mel 1983: 204), only one demonstrative term is reported.
Presumably, the speakers of the substratum languages of Haitian who had a lexicon containing only one demonstrative term relexified this term on the basis of the French form ca, yielding sa in Haitian creole. Consequently, in the early creole, in addition to the lexicon with the two demonstrative terms showing the three interpretive patterns in (15), there was also a lexicon with only one form used as a general deictic term. This is schematically represented in (19).
(19) Haitian lexicon 1 sa sila (three interpretive patterns) Haitian lexicon 2 sa (general deictic term)
In this kind of situation, we expect dialect levelling to take place. Is there evidence from the literature that levelling is taking place in this area of the Haitian lexicon?
Recall from section 1 that, in the process of dialect levelling, different communities may arrive at a somewhat different compromise. There is evidence from the literature that different Haitian communities have arrived at different compromises with respect to the hypothesized early Haitian lexicons schematically represented in (19). For example, in his dialect survey of Haitian creole, Orjala (1970: 117) points out that all dialects have sa and that the northern and central dialects also have sila. This statement is corroborated by Joseph (1989: 112): the form sila "[...] se rencontre surtout dans le nord d'Haiti et dans le dialecte de certaines regions de l'Artibonite [...]" [is mostly found in the north of Haiti and in the dialect of some areas of the Artibonite region]. Valdman (1978: 194) also points out that sa and sila constitute dialectal variants. These observations suggest that the two hypothesized early Haitian lexicons in (19) have been subjected to dialect levelling and that different Haitian communities have arrived at different compromises in the process.
Another indication that there might be some levelling taking place with respect to the two lexical items sa and sila is frequency of use. Several authors point out that sa occurs more frequently than sila (e.g. Etienne 1974: 154; Ferere 1974: 105; Joseph 1989: 112; Pompilus 1976: 38). Furthermore, while Valdman et al. (1981) list both sa and sila. Valdman (1996) only lists sa. These observations suggest that the second Haitian lexicon in (19) might be replacing the first one, at least in some regions of Haiti. Only history will tell if regional dialects will stabilize as they are now or if sila will disappear from all the lexicons of Haitian.
The data discussed in this section illustrate the fact that the variation between substratum lexicons may be transferred into the early creole by means of the process of relexification. They further show that this variation may be subjected to levelling, which operates on the output of relexification.
In this paper, I have shown that relexification of various substratum lexicons on the basis of a single superstratum language provides the speakers of the early creole community with a common vocabulary. However, through this process, individuals bring into the incipient creole idiosyncrasies of their native lexicons; thus, the properties of an early Haitian lexical entry may vary slightly among the speakers of the incipient creole. The three sets of data discussed in this paper show that the bulk of the variation found in the incipient creole is attributable to the variation found in the substratum languages rather than that found in the superstratum language. Other similar cases discussed in Lefebvre (1998) lead to the same conclusion. Furthermore, this approach is congruent with the claim that the creators of Haitian did not have enough exposure to French, at the time the creole was formed, to acquire the details of the properties of French words, as we saw in section 2. Relexification of various substratum lexicons also accounts in a principled way for the observation that different substratum languages may contribute to an incipient creole. On the basis of three sets of Haitian data, I have illustrated how levelling operates on the output of the relexified lexicons, reducing the variation. Other cases of dialect levelling are discussed in Lefebvre (1998) (see, for example, the discussion involving case markers [1998: 110] and tense, mood, and aspect markers [1998: 137-139]).(22) The role of levelling in creole genesis and development accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages.
The data discussed in this paper illustrate two different types of levelling with respect to the time span required to complete the change. The first case (section 3) (and other similar cases discussed in Lefebvre 1998) is hypothesized to be an early case of levelling. The two others (sections 4 and 5) have not been completed as yet. This situation raises a number of questions for future research. For example, why are some forms levelled out early, while others take more time, if they are levelled out at all? Are there principles underlying levelling that would make the data predictable in this respect? How does early levelling operate?
Finally, the data discussed in this paper are interesting from another point of view. Singler (1988: 29) advocates the view that "when universals, substrate, and lexifier converge with regard to a given phenomenon, such a phenomenon is more likely to enter the creole grammar than when the sources compete." The data discussed in this paper constitute clear cases of the sources competing in different ways. First, as was extensively documented above, the superstratum and substratum languages do not have much in common in terms of how they lexically encode the various notions discussed in this paper. Second, the substratum languages are not homogeneous with respect to the lexical items discussed. Nevertheless, the idiosyncratic properties of the substratum languages have made their way into the creole, as was argued above. This shows that creole lexicons and grammars cannot just be a product of universals.
University of Quebec at Montreal
Received 27 June 2000 Revised version received 26 January 2001
(*) The content of this paper was presented at the Linguistic Society of America (Society for Pidgin and Creole Linguistics) meeting in January 1999, in Los Angeles, under the title "Dialect levelling in creole communities." I would like to thank Renee Lambert, Catherine Leger, and Olivier Tardif for assisting me in doing some of the research that led to this paper. Thanks to Andree Belanger for formatting the manuscript and to Catherine Leger for proofreading it. The research in this paper was financed by SSHRCC and FCAR. I would like to thank Bernard Comrie and Jeff Siegel for their very useful comments on an earlier draft of the paper, as well as a number of anonymous reviewers for their questions and comments. I am also grateful to Christine Jourdan and Jeff Siegel for fruitful discussions on several of the issues raised here. Correspondence address: Departement de linguistique, Case postale 8888, succursale A, Universite du Quebec a Montreal, Montreal (Quebec) H3C 3P8, Canada. E-mail: email@example.com.
(1.) The term "target language" might not be appropriate given the various circumstances in which relexification applies. Baker (1990), Muhlhausler (1986), and Thomason and Kaufman (1991) have discussed this issue. In this text, I will use the term "lexifier language" instead.
(2.) The question of how phonetic strings of the superstratum language are interpreted by the substratum speakers is discussed at length in Brousseau (n.d.) on the basis of Haitian creole.
(3.) For other representations of the process, see Mous (1995, forthcoming) and Muysken (1981). For a general discussion on the formal representation of relexification, see Lefebvre (1998: 15-19, 22, 27, 384-386).
(4.) For an extensive discussion on null forms in creoles, see Lefebvre (1998:17-18, 37-38, 44-45, 78, 108-110, 128, 164, 375-386).
(5.) It has been argued many times that the process of relexification plays a major role in the formation of mixed languages (see e.g. Bakker 1992; Mous 1994, forthcoming; Muysken 1981). It has also been argued by the same authors that second-language acquisition is not involved in the formation of mixed languages for they are created by bilingual speakers. As is pointed out in Lefebvre (1998: 29), relexification may sometimes be used as a tool in second-language acquisition (but not in every case). Being a cognitive process, relexification is autonomous and it may be used in different situations of language creation, whether acquisition of L2 is involved (as in creole formation) or not (as in mixed-language formation).
(6.) Fongbe, a Gbe language of the Kwa family, was chosen as a case study for reasons extensively discussed in Lefebvre (1998: 52-77).
(7.) For extensive discussions of these issues, see Mougeon and Beniak (1994) and the papers and references therein.
(8.) The following question was addressed to me: "[...] why levelling would not have countered the role of relexification in the initial stages of what would become Haitian creole?" There are two ways in which this question can be interpreted. The first one is whether levelling could take place prior to relexification, thus countering relexification. In this case, levelling would have to operate on the native languages of the early creole population, in this case, the West African languages present at the time the creole was formed. Such a possibility was argued against in the discussion of the third scenario. The second interpretation of the above question is whether levelling could start taking place before the relexification of whole lexicons was completed, thus countering the relexification of some lexical entries. In my view, this possibility is a likely one. It is compatible with the fact that, for levelling to take place, the members of a community must communicate in the same language, in this case the incipient creole, consisting of West African lexicons (partially) relexified on the basis of French phonetics matrices. It is also compatible with the first scenario proposed to account for the properties of Haitian yo.
(9.) For a different view, see Carden and Stewart (1988: 22), who advocate that BODY-part reflexives might constitute an independent development in creoles.
(10.) I refer the reader to Lefebvre (1998:160-167) for further discussion of this point.
(11.) Dechaine and Manfredi (1994: 228) claim that there is no morphological form in Haitian creole. In their analysis, the lack of such form "allows referential indices to be freely assigned to li." On the analysis proposed in this paper, according to which Haitian has a reflexive form that is phonologically null, a pronoun is assigned a reflexive interpretation if it is part of a structure of the type in (10b), and it is assigned a free interpretation if it is the head of DP, like any other pronominal form. While in Dechaine and Manfredi's analysis, "Haitian lacks binding condition A because it lacks morphological reflexives" (1994: 266), on the analysis proposed here there is no need to say that Haitian lacks binding condition A. The reflexive anaphor is simply nonovert.
(12.) The literature on reflexives cited earlier only provides examples of BODY-part reflexives. Whether there are Kwa languages other than Gbe that have reflexive anaphors is a question for future research.
(13.) According to these authors, during the levelling period, different geographical areas did not choose the same variant, in such a way that, in this case, levelling appears to also result in dialect differentiation where Port au Prince, the north, and the south of Haiti manifest geographical dialectal differences with respect to how reflexivity is now encoded.
(14.) It is a well-documented fact that regular linguistic change may take place gradually across the lexicon. This dimension also appears to be involved in the change concerning reflexive forms in Haitian. As has been mentioned by several authors (e.g. Carden and Stewart 1988; Dechaine and Manfredi 1994; Brousseau 1995b; Lefebvre 1998) the reflexive forms are not all acceptable with all verbs. For example, while the verb we to see' may take a phonologically null anaphor or tet as a reflexive, the verb mode `to bite' can only take tet (see e.g. Dechaine and Manfredi 1994: 203, 209). Similar facts are also reported by Brousseau (1995b), who divides up all the reflexive verbs in terms of the typical reflexive form that they take. Some verbs, such as blese `to hurt', are shown to take any of the reflexive forms. Other verbs are shown to take two of the forms, and so on and so forth. A detailed documentation of how the ongoing change spreads through the verbal lexicon as well as through time is a topic for future research.
(15.) An anonymous reader pointed out the similarity between the position of the possessor phrase in Haitian and the position of the emphatic pronoun in popular French in examples such as la bague a moi la `the ring of mine there' and raises the question of "[...] whether non-standard French structures just had no role to play whatsoever in the development of Haitian creole." A full discussion of this subject is beyond the scope of this paper. However, since the question was raised, I will answer it by giving the main points. Although French structures of the type cited above served as a constraint on constituent order in the formation of Haitian creole nominal structures, as is mentioned in section 2 and as is argued in Lumsden (1991) -- precisely on the basis of the position of the possessor phrase in Haitian and its contributing languages--the French and the Haitian possessor phrase in (12) differ in case. As is extensively argued in Lumsden (1991), whereas the French possessor phrase is marked for objective case it, as in la bague a moi la, the Haitian one in (12) has all the characteristics of a genitive phrase (the genitive case marker is phonologically null in Haitian). According to Lumsden's (1991) analysis, the Haitian possessor phrase in (12) has the same properties as the Fongbe possessor phrase marked for genitive case tcn. The fact that Fongbe tcn marks genitive case rather than objective case is extensively argued in Brousseau and Lumsden (1992). For further discussion of this point, see also Lefebvre (1998:101-110).
(16.) The authors who worked on French deictic terms cited in Lefebvre (1997) are Brunot (1926), Dees (1971), Grevisse (1975), Guiraud (1966), Lommatzsch (1925), Rosset (1911), and Yvon (1951).
(17.) In his comparative grammar of French-based creoles, Goodman (1964:51) writes, "In Haiti [...], these two meanings [`this' and `that'] are distinguished formally, sa `this' and sila `that'."
(18.) Tinelli (1970: 194) attests two grammars: "sa is sometimes replaced by sila, with no syntactic change [...], and no clear modification of meaning, except for some speakers who distinguish the remote deictic sila from the proximate sa [...]."
(19.) This pattern is similar to the first one in the sense that sila is [- proximate]. It differs from the first one in the sense that sa is [[Alpha] proximate] rather than [+ proximate]. There is independent evidence that sa is a general deictic term for some speakers. Although sa is glossed as `this' at the beginning of Hall's (1953: 29) book, the 195 cases of this lexical item found in his grammar and Haitian texts are translated either as `this' or `these' (57 occurrences) or as `that' or `those' (138 occurrences), which shows that sa is a general deictic term in this source. For similar facts, see d'Ans (1968), Peleman (1978), and Sylvain (1936). This argues that the analysis of sa as a general deictic term is common in Haitian.
(20.) Although no other author has explicitly reported a variety of Haitian where sa is [[Alpha] proximate] and sila is [- proximate], some Haitian sources provide evidence that suggests such a distinction. For example, in Sylvain (1936), all occurrences of sila are translated by a French deictic term + la: Loze sila-yo k'tro lwe ba mwe `allongez-moi ceux-la qui sont trop loin'/`stretch (give me) those ones which are too far for me' (p. 58); Sila-a se fre-m `celui-la est mon frere'/`that one is my brother' (1936: 108). In contrast, sa is translated by any of the French deictic terms: Burik-sa-a move `cet ane est mechant, cet ane-ci est mechant'/`this/that//this donkey is nasty' (1936: 56); Sa-a gu `celui-ci est agreable au gout'/`this one is pleasant to taste' (1936: 57); Tut sa-a pa bo `celui-la n'est pas bon en entier'/`that one is not completely good' (1936: 76). The following example shows that sa may also be used to point at something that is far: M'pito sa k'pi lwe-a `je prefere celui qui est plus loin'/`I prefer the one that is farther' (1936: 149). I take Sylvain's (1936) translations to reflect the pattern I describe: while sila is [- proximate], sa is a general deictic term [[Alpha] proximate]. Etienne's (1974: 154) remark seems to make the same statement: "Le morpheme/sila/comporte une indication d'eloignement [...]. Il s'oppose a/sa/[...] comme demonstratif locatif. [...] Les Haitiens du Nord semblent faire de/sa/un double emploi: celui de demonstratif general (proche ou lointain) ou de locatif." [The/sila/morpheme comprises information of distance [...]. It contrasts with/sa/[...], a locative demonstrative. [...] Haitians of the North seem to have two uses for/sa/: one is a general demonstrative term (close or far), the other one is a locative.]
(21.) Lefebvre (1997) is based on fieldwork done prior to 1992 (G2). G1 and G3 were found after this date during fieldwork done with additional speakers.
(22.) An anonymous reviewer points out that the data presented in the paper are "overfamiliar" to researchers in this field. I would like to emphasize the fact that these particular sets of data have been chosen because they represent different types of cases of dialect levelling. These cases are complementary in several ways that are identified at the beginning and at the end of the data sections. Also, choosing familiar data for a paper of this kind allows the researcher to concentrate on the analytical and theoretical points that have to be made; having to present, analyze and justify new sets of data might detract from the main line of argumentation.
Alleyne, Mervyn C. (1971). Acculturation and the cultural matrix of creolisation. In Pidginization and Creolization of Languages, Dell Hymes (ed.), 169-187. Cambridge: Cambridge University Press.
--(1981). Comparative Afro-American: An Historical-Comparative Study of English-Based Afro-American Dialects of the New World. Ann Arbor, MI: Karoma.
Andersen, Roger W. (1980). Creolisation as the acquisition of a second language as a first language. In Theoretical Orientation in Creole Studies, A. Valdman and A. R. Highfield (eds.), 273-295. New York: Academic Press.
Anonymous (1983). Elements de recherche sur la langue fon. Cotonou, Benin.
Arends, Jacques; Muysken, Pieter; and Smith, Norval (eds.) (1995). Pidgins and Creoles: An Introduction. Creole Language Library 15. Amsterdam and Philadelphia: Benjamins.
Awoyale, Yiwola (1986). Reflexivization in Kwa languages. In Current Approaches to African Linguistics, G. J. Dimmendaal (ed.), 1-14. Dordrecht: Foris.
Baker, Philip (1990). Off target? Journal of Pidgin and Creole Languages 5, 107-119.
Bakker, Peter (1992). A language of our own: the genesis of Mitchif, the mixed Cree-French language of the Canadian Metis. Unpublished doctoral dissertation, University of Amsterdam.
--; and Mous, Maarten (eds.) (1994). Mixed Languages. Studies in Language and Language Use. Dordrecht: ICG.
Barz, Richard K.; and Siegel, Jeff (eds.) (1988). Language Transplanted: The Development of Overseas Hindi. Wiesbaden: Harrassowitz.
Bickerton, Derek (1981). Roots of Language. Ann Arbor, MI: Karoma.
Bole-Richard, Remi (1983). Systematique phonologique et grammaticale d'un parle ewe: le gen-mina du Sud-Togo et Sud-Benin. Paris: L'Harmattan.
Brousseau, Anne-Marie (1995a). Les pronoms en creole haitien, en francais et en fongbe. In Research Report, Claire Lefebvre (ed.), vol. IV, 33 pages. Montreal: Universite du Quebec a Montreal.
--(1995b). Reflexivite et classes de verbes en creole haitien et en fongbe. In Research Report, Claire Lefebvre (ed.), vol. IV, 15 pages. Montreal: Universite du Quebec a Montreal.
--(n.d.). The genesis of the phonological system of Haitian Creole (provisional title).
--; and Lumsden, John S. (1992). Nominal structure in Fongbe. Journal of West African Languages 22: Topics on the Syntax and Semantics of Fongbe (special issue), Claire Lefebvre (ed.), 5-26.
Brunot, Ferdinand (1905). Histoire de la langue francaise des origines a 1900. Paris: A. Colin.
--(1926). La Pensee et la Langue. Paris: Masson.
Capo, Hounkpati B. C. (1984). Elements of Ewe-Gen-Aja-Fon dialectology. In Peuples du golfe de Benin: Aja-Ewe, F. de Medeiros (ed.), 167-178. Paris: Karthala.
Carden, Guy; and Stewart, William A. (1988). Binding theory, bioprogram, and creolization: evidence from Haitian creole. Journal of Pidgin and Creole Languages 3, 1-68.
Chaudenson, Robert (1973). Pour une etude comparee des creoles et parlers francais d'outre-mer: survivance et innovation. Revue de linguistique romane 147/148, 342-371.
--(1992). Des iles, des hommes, des langues. Langues creoles/cultures creoles. Paris: L'Harmattan.
--(1993). De l'hypothese aux exemples. Un cas de creolisation: selon la formation des systemes de demonstratifs creoles. Etudes creoles 16, 17-38.
Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris.
Christie, Pauline (ed.) (1996). Caribbean Language Issues: Old and New. Kingston, Jamaica: University of the West Indies Press.
d'Ans, Andre-Marcel (1968). Le Creole francais d'Haiti: Etude des unites d'articulation, d'expansion et de communication. The Hague: Mouton.
Dechaine, Rose-Marie; and Manfredi, Victor (1994). Binding domains in Haitian. Natural Language and Linguistic Theory 12, 203-257.
Dees, Anthonij (1971). Etude sur l'evolution des demonstratifs en ancien et moyen francais. Groningen: Wolters-Noordhoff.
DeGraff, Michel (ed.) (1999). Language Creation and Language Change, Creolization, Diachrony, and Development. Cambridge, MA, and London: MIT Press.
--(forthcoming). Empirical quicksand: probing two recent articles on Haitian creole. Journal of Pidgin and Creole Languages.
Delafosse, Maurice (1929). La Langue mandingue et ses dialectes (malinke, bambara, dioula). Paris: Geuthner.
de Medeiros, Francois (ed.) (1984). Peuples du golfe du Benin: Aja-Ewe. Colloque de Cotonou. Paris: Karthala.
Dillard, Joey Lee (1970). Principles in the history of American English: paradox, virginity, and cafeteria. Florida FL Reporter 8, 32-33.
Dimmendaal, Gerrit J. (ed.) (1986). Current Approaches to African Linguistics. Publications in African Languages and Linguistics. Dordrecht: Foris.
Domingue, Nicole (1980). Syntactic innovations in Mauritian Bhojpuri. Unpublished manuscript, McGill University, Montreal.
--(1981). Internal change in a transplanted language. Studies in the Linguistic Sciences 4, 151-159.
Duponchel, L.; and Mel, G. (1983). L'alladian. In Atlas des langues kwa de Cote d'Ivoire, vol. 1, G. Herault (ed.), 181-225. Paris: Agence de cooperation culturelle et techniques; Abidjan, Ivory Coast: Universite d'Abidjan, Institut de linguistique appliquee.
Durie, Mark (forthcoming). Language, Function and Time. London: Oxford University Press.
Etienne, Gerard (1974). Le creole du nord d'Haiti: etude des niveaux de structure. Unpublished doctoral dissertation, Universite des sciences humaines de Strasbourg.
Faine, Jules (1937). Philologie creole: etudes historiques et etymologiques sur la langue creole d'Haiti. Port-au-Prince: Imprimerie de l'Etat.
--(1974). Dictionnaire francais-creole. Ottawa: Lemeac.
Faltz, Leonard M. (1985). Reflexivization: A Study in Universal Syntax. New York and London: Garland.
Ferere, Gerard A. (1974). Haitian creole sound-system, form-classes, texts. Unpublished doctoral dissertation, University of Pennsylvania.
Fillmore, Charles J. (1975). Santa Cruz Lectures on Deixis 1971. Bloomington: Indiana University Linguistics Club.
Foley, William A. (1988). Language birth: the processes of pidginisation and creolisation. In Language: The Socio-Cultural Context, F. J. Newmeyer (ed.), 162-183. Cambridge: Cambridge University Press.
Gambhir, Surendra K. (1988). Structural development in Guyanese Bhojpuri. In Language Transplanted: The Development of Overseas Hindi, R. K. Barz and J. Siegel (eds.), 69-94. Wiesbaden: Harrassowitz.
Goodman, Morris F. (1964). A Comparative Study of Creole French Dialects. The Hague: Mouton.
Gougenheim, Georges (1973). Grammaire de la langue francaise au seizieme siecle. Paris: A. et J. Picard.
Grenoble, Lenore; and Riley, Matthew (1996). The role of deictics in discourse coherence: French voici/voila and Russian vot/von. Journal of Pragmatics 25, 819-838.
Grevisse, Maurice (1975). Le Bon Usage, 10th ed. Gembloux, Belgium: Duculot.
Guiraud, Pierre (1966). Le Moyen Francais, 2nd ed. Que sais-je? Paris: Presses universitaires de France.
Gundel, Jeanette K.; Hedberg, Nancy; and Zacharski, Ron (1993). Cognitive status and the form of referring expressions in discourse. Language 69, 274-307.
Hall, Joan H.; Doane, Nick; and Ringler, Dick (eds.) (1992). Old English and New: Studies in Language and Linguistics in Honor of Frederic G. Cassidy. New York and London: Garland.
Hall, Robert A. (1953). Haitian Creole: Grammar, Texts and Vocabulary. Menasha, WI: George Banta.
Harris, John W. (1991). Kriol/the creation of a new language. Language in Australia, S. Romaine (ed.), 195-203. Cambridge: Cambridge University Press.
Hazoume M. L. (1990). Essai de classification synchronique. Etude comparative des parlers Gbe du Sud-Benin. Cotonou, Benin: Centre national de linguistique appliquee.
Herault, Georges (1983a). L'adioukrou. In Atlas des langues kwa de Cote d'Ivoire, vol. 1, G. Herault (ed.), 129-153. Paris: Agence de cooperation culturelle et technique; Abidjan, Ivory Coast: Universite d'Abidjan, Institut de linguistique appliquee.
--(1983b). L'eotile. In Atlas des langues kwa de Cote d'Ivoire, vol. 1, G. Herault (ed.), 403-424. Paris: Agence de cooperation culturelle et technique appliquee.
--(ed.). (1983c). Atlas des langues kwa de Cote d'Ivoire, vol. 1, 2nd ed. Paris: Agence de cooperation culturelle et technique; Abidjan, Ivory Coast: Universite d'Abidjan, Institut de linguistique appliquee.
Highfield, Arthur R.; and Valdman, Albert (eds.) (1981). Historicity and Variation in Creole Studies. Ann Arbor, MI: Karoma.
Horiot, Brigitte (ed.) (1991). Francais du Canada -- Francais de France. Proceedings of the [2.sup.e] Colloque international de Cognac, France, September 27-30, 1988. Tubingen: Niemeyer.
Hull, Alexander (1979). Affinites entre les varietes du francais. Le Francais hors de France, A. Valdman (ed.), 165-182. Paris: Honore Champion.
Hymes, Dell (ed.) (1971). Pidginization and Creolization of Languages. (Proceedings of a conference held at the University of West Indies, Mona, Jamaica, April 1968.) Cambridge: Cambridge University Press.
Jagger, Philip J.; and Buba, Malami (1994). The space and time adverbials nan/can in Hausa: cracking the deictic code. Language Sciences 16, 387-421.
Joseph, Frantz (1989). La determination nominale en creole haitien. Unpublished doctoral dissertation, Universite de Paris VII.
Jourdan, Christine (1985). Sapos yumi mitim yumi: Urbanisation and creolisation of Solomon Islands Pidgin. Unpublished doctoral dissertation, Australian National University.
Juneau, Marcel (1972). Contribution a l'histoire de la prononciation francaise au Quebec. Paris and Sainte-Foy: Klincksieck and Presses de l'Universite Laval.
--; and Poirier, Claude (1973). Le Livre de comptes d'un meunier quebecois (fin [XVII.sup.e]-debut [XVIII.sup.e] siecle). Sainte-Foy: Presses de l'Universite Laval.
Kinyalolo, Kasangati K. W. (1994). Pronouns and relexification. In Etudes syntaxiques, C. Lefebvre and J. S. Lumsden (eds.), volume III, 259-293. Montreal: Universite du Quebec a Montreal.
Klemperer, Victor; and Lerch, Eugen (eds.) (1925). Jahrbuch fur Philologie. Munich: Max Hueber.
Labov, William (1972). Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press.
Lefebvre, Claire (1984). Grammaires en contact: definition et perspectives de recherche. Revue quebecoise de linguistique 14(1), 11-49.
--(ed.) (1992). Topics on the Syntax and Semantics of Fongbe (special issue). Journal of West African Languages 22.
--(1994). Functional categories and relexification. In The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole, C. Lefebvre and J. S. Lumsden (eds.), 74 pages. Montreal: Universite du Quebec a Montreal.
--(ed.) (1995). Research Report, 4 vols. Prepared for FCAR on the project L'Organisation des lexiques et des entrees lexicales. Montreal: Universite du Quebec a Montreal.
--(1997). Relexification in creole genesis: the case of demonstrative terms in Haitian creole. Journal of Pidgin and Creole Languages 12, 181-201.
--(1998). Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole. Cambridge Studies in Linguistics 88. Cambridge: Cambridge University Press.
--(forthcoming). On the empirical reliability of some Haitian data. Journal of Pidgin and Creole Languages.
--; and Lumsden, John S. (1989a). Les langues creoles et la theorie linguistique. Revue canadienne de linguistique 34(3): La Creolisation (special issue), C. Lefebvre and J. S. Lumsden (eds.), 249-272.
--; and Lumsden, John S. (eds.). (1989b). La Creolisation (special issue). Revue canadienne de linguistique 34 (3).
--; and Lumsden, John S. (1994a). Le role central de la relexification dans la genese das langues creoles. Plurilinguismes 8, 47-93.
--; and Lumsden, John S. (1994b). Relexification in creole genesis. In The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole, C. Lefebvre and J. S. Lumsden (eds.), 28 pages. Montreal: Universite du Quebec a Montreal.
--; and Lumsden, John S. (eds.) (1994c). The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole. Research report prepared for SSHRCC on the project La Genese du creole haitien: Un cas particulier d'investigation sur la forme de la grammaire universelle. Montreal: Universite du Quebec a Montreal.
--; and Lumsden, John S. (eds.) (1994d). Etudes syntaxiques, 3 vols. Research report prepared for SSHRCC and PAFAC on the project La Genese du creole haitien: Un cas particulier d'investigation sur la forme de la grammaire universelle. Montreal: Universite du Quebec a Montreal.
Levinson, Stephen C. (1983). Pragmatics. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press.
Lommatzsch, E. (1925). Deiktische Elemente im Altfranzosischen. In Jaarbuch fur Philologie, V. Klemperer and E. Lerch (eds.), 202-244. Munich: Max Hueber.
Lumsden, John S. (1991). On the acquisition of nominal structures in the genesis of Haitian creole. Travaux de recherche sur le creole haitien 4, 37-61 (Universite du Quebec a Montreal).
--(1999). Language acquisition and creolization. In Language Creation and Language Change. Creolization, Diachrony, and Development, M. DeGraff (ed.), 129-157. Cambridge, MA: MIT Press.
--; and Lefebvre, Claire (1994). The genesis of Haitian creole. In The Central Role of Relexification in Creole Genesis: The Case of Haitian Creole, C. Lefebvre and J. S. Lumsden (eds.), 15 pages. Montreal: Universite du Quebec a Montreal.
Maclaran, Rose (1980). On two asymmetrical uses of the demonstrative determiners in English. Linguistics 18, 803-820.
Mensah, E. N. A. (1983). Le krobou. In Atlas des langues kwa de Cote d'Ivoire, vol. 1, G. Herault (ed.), 425-463. Paris: Agence de cooperation culturelle et techniques; Abidjan, Ivory Coast: Universite d'Abidjan, Institut de linguistique appliquee.
Meyer-Lubke, Wilhelm (1909). Le francais au Canada. Bulletin de francais parle au Canada 8, 121-129.
Migeod, F. W. H. (1911). The Languages of West Africa. London: Kegan Paul.
Mougeon, Raymond; and Beniak, Edouard (eds.) (1994). Les Origines du francais quebecois. Sainte-Foy: Presses de l'Universite Laval.
Mous, Maarten (1994). Ma'a or Mbugu. In Mixed Languages, P. Bakker and M. Mous (eds.), 175-200. Dordrecht: ICG.
--(1995). Language intertwining. Paper read at the Amsterdam Creole Workshop: Creole Genesis and Language Contact.
--(forthcoming). Ma'a as an ethno-register of Mbugu. Sprache und Geschichte in Afrika 15.
Mufwene, Salikoko S. (1986). The universalist and substrate hypotheses complement one another. In Substrata Versus Universals in Creole Genesis, P. C. Muysken and N. Smith (eds.), 129-162. Amsterdam: Benjamins.
--(1990). Transfer and the substrate hypothesis in creolistics. Studies in Second Language Acquisition 12, 1-23.
--(1992). Africanisms in Gullah: a re-examination of the issues. In Old English and New, J. H. Hall, N. Doane, and D. Ringler (eds.), 156-182. New York: Garland.
--(ed.) (1993). Africanisms in Afro-American Language Varieties. Athens, GA, and London: University of Georgia Press.
--(1994). Genese de population et genese de langue. Creolistique et grammaire generative 8, 95-113.
--(1996). Creole genesis: a population genetics perspective. In Caribbean Language Issues, P. Christie (ed.), 163-196. Kingston, Jamaica: University of the West Indies Press.
Muhlhausler, Peter (1980). Structural expansion and the process of creolization. In Theoretical Orientation in Creole, Studies, A. Valdman and A. R. Highfield (eds.), 19-56. New York: Academic Press.
--(1986). Pidgin and Creole Linguistics. Language in Society 2. Oxford: Blackwell.
Muysken, Pieter C. (1981). Half-way between Quechua and Spanish: the case for relexification. In Historicity and Variation in Creole Studies, A. R. Highfield and A. Valdman (eds.), pp. 52-79. Ann Arbor, MI: Karoma.
--; and Smith, Norval (eds.) (1986). Substrata Versus Universals in Creole Genesis. Creole Language Library 1. Amsterdam and Philadelphia: Benjamins.
--; and Smith, Norval (1995). Reflexives. In Pidgins and Creoles, J. Arends, P. Muysken, and N. Smith (eds.), 271-287. Amsterdam: Benjamins.
Newmeyer, Frederick J. (ed.) (1988). Language: The Socio-Cultural Context. Linguistics: The Cambridge Survey 4. Cambridge: Cambridge University Press.
Orjala, Paul R. (1970). A dialect survey of Haitian creole. Unpublished doctoral dissertation, Hartford Seminary Foundation, Kansas City, MO.
Peleman, L. (1978). Diksyonne kreol-franse. Port-au-Prince: Bon Nouvel.
Piwek, P. L. A.; and Cremers, A. H. M. (1996). Dutch and English demonstratives: a comparison. Language Sciences 18, 835-851.
Poirier, Claude (1994). La langue parlee en Nouvelle-France: vers une convergence des explications. In Les Origines du francais quebecois, R. Mougeon and E. Beniak (eds.), 237-274. Sainte-Foy: Presses de l'Universite Laval.
Pompilus, Pradel (1976). Contribution a l'etude comparee du creole et du francais a partir du creole haitien. Port-au-Prince: Editions Caraibes.
Rivard, A. (1914). Etudes sur les parlers de France au Canada. Quebec: Garneau.
Robertson, Ian (1993). The Ijo element in Berbice Dutch and the pidginization/creolization process. In Africanisms in Afro-American Language Varieties, S. S. Mufwene (ed.), 296-316. Athens: University of Georgia Press.
Romain, Jean-Baptiste (1959). Quelques moeurs et coutumes des paysans haitiens. Travaux pratiques d'ethnographie sur la region de' Milot a l'usage des etudiants. Port-au-Prince: Imprimerie de l'Etat.
Romaine, Suzanne (ed.) (1991). Language in Australia. Cambridge: Cambridge University Press.
Rongier, Jacques (1979). Description de l'ewe vol. 1. Unpublished doctoral dissertation, Universite de Paris VII.
--(1995). Dictionnaire francais-ewe suivi d'un index francais ewe-francais. Paris: Karthala.
Rosset, Theodore (1911). Les Origines de la prononciation moderne etudiees au 17ieme siecle. Paris: A. Colin.
Sankoff, Gillian (1984). Substrate and universals in the Tok Pisin verb phrase. Meaning, form and use in context: linguistic applications. In Georgetown University Round Table on Language and Linguistics, D. Schiffin (ed.), 104-119. Washington, D.C.: Georgetown University Press.
Schiffin, Deborah (ed.) (1984). Georgetown University Round Table on Language and Linguistics. Washington, D.C.: Georgetown University Press.
Schumann, John H. (1978). The Pidginisation Process: A Model for Second Language Acquisition. Rowley, MA: Newbury House.
Segurola, R. P. B. (1963). Dictionnaire fon-francais. Cotonou, Benin: Procure de l'archidiocese.
Siegel, Jeff (1995). Koine formation and creole genesis. Paper read at the Amsterdam Creole Workshop: Creole Genesis and Language Contact.
--(1997). Mixing, levelling and pidgin/creole development. In The Structure and Status of Pidgins and Creoles, A. K. Spears and D. Winford (eds.), 111-149. Amsterdam: Benjamins.
Singler, John V. (1988). The homogeneity of the substrate as a factor in pidgin/creole genesis. Language 64, 27-51.
--(1993). The African presence in Caribbean French colonies in the seventeenth century: documentary evidence. Travaux de recherche sur le creole haitien 16/17, 1-236 (Universite du Quebec a Montreal).
--(1996). Theories of creole genesis, sociohistorical considerations, and the evaluation of evidence: the case of Haitian creole and the relexification hypothesis. Journal of Pidgin and Creole Languages 11, 185-231.
Smith, Norval (1987). The genesis of the creole languages of Surinam. Unpublished doctoral dissertation, University of Amsterdam.
Spears, Arthur K.; and Winford, Donald (eds.) (1997). The Structure and Status of Pidgins and Creoles. Amsterdam: Benjamins.
Sylvain, Suzanne (1936). Le Creole haitien: Morphologie et syntaxe. Wetteren, Belgium: Imprimerie De Meester, Port-au-Prince: by the author
Thomason, Sarah G.; and Kaufman, Terrence (1991). Language Contact, Creolization, and Genetic Linguistics. Berkeley, Los Angeles, and Oxford: University of California Press.
Tinelli, Henri (1970). Generative phonology of Haitian creole. Unpublished doctoral dissertation, University of Michigan.
Tresbarats, C. (1983). L'abidji. In Atlas des langues kwa de Cote d'Ivoire, vol. 1, G. Herault (ed.), 41-65. Paris: Agence de cooperation culturelle et techniques; Abidjan, Ivory Coast: Universite d'Abidjan, Institut de linguistique appliquee.
Trudgill, Peter (1986). Dialects in Contact. Language in Society 10. Oxford: Blackwell.
Valdman, Albert (1978). Le Creole: Structure, statut et origine. Paris: Klincksieck.
--(ed.) (1979). Le Francais hors de France. Creoles et francais regionaux. Paris: Honore Champion.
--(1980). Creolisation and second language acquisition. In Theoretical Orientation in Creole Studies, A. Valdman and A. R. Highfield (eds.), 297-311. New York: Academic Press.
--(1996). A Learner's Dictionary of Haitian Creole. Bloomington: Indiana University Creole Institute.
--; and Highfield, Arthur R. (eds.) (1980). Theoretical Orientation in Creole Studies. New York: Academic Press.
Valdman, Albert: et al. (1981). Haitian Creole-English-French Dictionary. Bloomington: Indiana University Creole Institute.
Veenstra, Tonjes (1996). Serial Verbs in Saramaccan: Prediction and Creole Genesis. Dordrecht: Holland Institute of Generative Linguistics.
Vilsaint, F. (1992). Diksyone Angle Kreyol English Kreyol Dictionary. Temple Terrace, FL: Educa Vision.
Wallace, Efoe (1995). Le groupe nominal en ewe. In Research Report, Claire Lefebvre (ed.), volume II, 15 pages. Montreal: Universite du Quebec a Montreal.
Weinreich, Uriel (1953). Languages in Contact. The Hague: Mouton.
Westerman, Diedrich (1930). A Study of the Ewe Language. London: Oxford University Press.
Whinnom, Keith (1971). Linguistic hybridization and the "special case" of pidgins and creoles. In Pidginization and Creolization of Languages, Dell Hymes (ed.), 91-115. Cambridge: Cambridge University Press.
Wolf, Lothar (1991). Le Langage de la cour et le francais canadien. In Francais du Canada--Francais de France, B. Horiot (ed.), 115-123. Tubingen: Niemeyer.
--(1994). Postface. In Les origines du francais quebequois, R. Mougeon and E. Beniak (eds.), 325-332. Sainte-Foy: Presses de l'Universite Laval.
Yvon, H. (1951). Cil et cist, articles demonstratifs. Romania 72, 145-181.