+ All Categories
Home > Documents > Nominal constituents in French lexifier creoles: Probing the structuring role of grammaticalization

Nominal constituents in French lexifier creoles: Probing the structuring role of grammaticalization

Date post: 19-Nov-2023
Category:
Upload: rutgers
View: 0 times
Download: 0 times
Share this document with a friend
47
Nominal constituents in French lexifier creoles Probing the structuring role of grammaticalization* Viviane Déprez Rutgers University/ISC,CNRS e paper offers a comparative study of the syntactic structure of nominal con- stituents in French Lexifier Creoles (FLC). In spite of the variability observed in the distribution of FLC determiners, the paper argues that FLC have a common functional architecture and presents both conceptual and empirical arguments in support of this view. Order variability, the paper proposes, is the result of extensive phrasal movement inside this common architecture that is triggered by the functional heads of FLC. Whether a given FLC determiner is a functional head or not is taken to reflect grammaticalization processes modeled in the Minimalist framework. A novel approach to grammaticalization and particularly to ‘semantic bleaching’ is shown to have a structuring effect that accounts for the ordering differences manifest across FLC determiners. Keywords: French based Creoles, determiners, DP, plural, demonstratives, grammaticalization, definite, indefinite, phrasal movement 1. Introduction As is obvious to even the most casual comparative observer, French Lexifier Cre- oles (FLC) are quite remarkable in featuring both a notable uniformity in the inventory of their overt nominal determiners and a rather striking diversity in the syntactic distribution of these elements. For instance, the indefinite singular * My warmest gratitude to the informants who have generously contributed their native speak- ers knowledge and sometimes their own precious insights about their language. anks in par- ticular to E.Vedrine, W. Paul, F.Canal and W.Douce, D.Fattier for Haitian Creole, D. Adone, F.Henri, D.Guillemin, P.Baker, D.Veronique for Mauritian Creole, O.Cyrille, C. Broussous, for Guadeloupean Creole, T.Montoute for Martinique Creole. For fruitful comments at various stages of development of this work, I would also like to thanks R. Kayne, R.Larson, V. Dayal, M. Baker, and E.Aboh. All remaining errors are my own. Journal of Pidgin and Creole Languages 22:2 (2007), 263–308. issn 020–0 / e-issn –70 © John Benjamins Publishing Company
Transcript

Nominal constituents in French lexifier creolesProbing the structuring role of grammaticalization*

Viviane DéprezRutgers University/ISC,CNRS

The paper offers a comparative study of the syntactic structure of nominal con-stituents in French Lexifier Creoles (FLC). In spite of the variability observed in the distribution of FLC determiners, the paper argues that FLC have a common functional architecture and presents both conceptual and empirical arguments in support of this view. Order variability, the paper proposes, is the result of extensive phrasal movement inside this common architecture that is triggered by the functional heads of FLC. Whether a given FLC determiner is a functional head or not is taken to reflect grammaticalization processes modeled in the Minimalist framework. A novel approach to grammaticalization and particularly to ‘semantic bleaching’ is shown to have a structuring effect that accounts for the ordering differences manifest across FLC determiners.

Keywords: French based Creoles, determiners, DP, plural, demonstratives, grammaticalization, definite, indefinite, phrasal movement

1. Introduction

As is obvious to even the most casual comparative observer, French Lexifier Cre-oles (FLC) are quite remarkable in featuring both a notable uniformity in the inventory of their overt nominal determiners and a rather striking diversity in the syntactic distribution of these elements. For instance, the indefinite singular

* My warmest gratitude to the informants who have generously contributed their native speak-ers knowledge and sometimes their own precious insights about their language. Thanks in par-ticular to E.Vedrine, W. Paul, F.Canal and W.Douce, D.Fattier for Haitian Creole, D. Adone, F.Henri, D.Guillemin, P.Baker, D.Veronique for Mauritian Creole, O.Cyrille, C. Broussous, for Guadeloupean Creole, T.Montoute for Martinique Creole. For fruitful comments at various stages of development of this work, I would also like to thanks R. Kayne, R.Larson, V. Dayal, M. Baker, and E.Aboh. All remaining errors are my own.

Journal of Pidgin and Creole Languages 22:2 (2007), 263–308.issn 020–0 / e-issn –70 © John Benjamins Publishing Company

264 Viviane Déprez

determiner, a phonetic variant of the French indefinite/numeral un, is found in all FLC, the definite determiner (if there is one) is always la and the most common demonstrative determiner is regularly sa (la). Contrasting with this inventorial uniformity the distributional variation of some of these elements is rather star-tling. The demonstrative sa, for instance, can either precede or follow the nominal projection, or it can precede or follow other determiners, which themselves either precede or follow the nominal projection. Yet alongside this diversity, some distri-butional uniformity is also found across FLC. A number of determiners — such as the indefinite singular determiner, the numeral determiners and various quanti-fiers — invariably occur in a fixed pre-nominal position. Finally, both uniformity and variation are observed in the distribution of plural markers. They occur either pre-nominally in Seychelles Creole (SC), Martinique Creole (MauC), Antillean-Creole (AntC) and Guadeloupe Creole (GuaC), or post-nominally in Guyanese Creole (GuyC) in Haitian Creole (HC) and Louisiana Creole (LC), in a position that seemingly follows most other determiners.

The syntax of FLC nominal constituents thus clearly presents a particularly fer-tile and challenging ground for any comparative approach that aims at providing a cross-linguistic characterization of what nominal constituents have in common and of how they can differ. Faced with such striking similarities and differences, interesting and challenging empirical and theoretical questions arise. Do the ob-served variations entail profound differences across FLC in the basic hierarchy of the functional projections that make up the structural backbone of their nominal constituents and if so, how and why do the FLC differ in this manner? Or could it be that, despite superficial appearances, the determiner systems of FLC have in fact a unique underlying functional architecture that is common to all varieties?

Embarking on a foray into the syntax of FLC nominal constituents, this paper has two central goals: First, it seeks to provide arguments for a common basic architecture for all FLC nominal functional projections. Second, it explores an analysis of the observed ordering variations that features no head movement but extensive and highly constrained phrasal movement within the proposed DP ar-chitecture. The first step in developing this analysis is to search for evidence of a common underlying architecture. The structure proposed here parallels in most respects the functional architecture developed for nominal constituents in recent generative research. The comparative Creole data are argued to provide new em-pirical support for the existence of a number of hierarchically organized func-tional projections above the lexical noun phrase (DP/DP > DemP/AgrP > NumP > NP). Once the basic architecture is established, the distinct determiner orders are shown to be systematically derivable through a cascade of phrasal movements governed by a single general principle:

Nominal constituents in French lexifier creoles 265

(1) Specifier of FLC nominal functional heads must be filled

A theoretical motivation for this principle is explored based on a novel Minimal-ist approach to grammaticalization. In short, the claim defended here is that FLC nominal functional heads include uninterpretable formal features that must be valued under movement to their specifiers (Chomsky 2000). Assuming that the characteristic ‘semantic bleaching’ that accompanies grammaticalization involves a change from interpretable to uninterpretable features, it follows that determiners that are heads are expected to be more ‘grammaticalized’ than those that are not, and henceforth attract interpretable constituents to value their ‘bleached’ features. This, I suggest, is what motivates the movements that are at the basis of the ob-served distributional diversity in the determiner structure of FLC.

The paper is organized as follows. Section 2 reviews the basic inventory and possible sequences of determiners in a variety of FLC and briefly discusses some of their general properties and uses. Section 3 provides arguments for the proposed basic common architecture. Section 4 details the movements involved in deriving the distributional diversity of the distinct FLC here considered. Section 5 summa-rizes the motivation behind these movements and concludes the paper.

2. The empirical landscape of FLC nominal constituents

The purpose of this section is to introduce the inventory and distribution of the various FLC determiners. The description proceeds by classes of determiners and emphasizes the similarities and differences encountered in the distinct Creoles.

2.1 Parallel distribution: indefinite determiners

The indefinite determiners of the FLC here surveyed can be descriptively charac-terized as follows:

– In all the FLC, the singular indefinite determiner is derived from the French numeral/indefinite marker un (with a variety of allomorphs)

– This singular indefinite determiner is consistently placed in a pre-nominal po-sition (and systematically precedes pre-nominal adjectives) as are numeral, universal and many-type quantifiers

– The singular indefinite determiner admits both specific and non-specific readings

– There are no overt plural counterpart to this indefinite determiner in the FLC under consideration, yet the singular indefinite determiner is quite generally in complementary distribution with the plural marker

266 Viviane Déprez

Characteristic examples for the indefinite determiner are provided below. The spelling of these examples follows that of their sources (see references):

(2) a. Seychelles Creole (SC): ~e zom a man b. Mauritian Creole (MauC): en En lakaz a house c. St Lucie (StLC): õ jõ õ koSõ a pig d. Martinique (MarC): an an tab a table e. Guadaloupean Creole (GuaC): õ õ timoun a child f. Louisiana Creole (LC): ein ein fomme a woman g. Haitian Creole (HC): yõ yõ fij a girl

The pair of examples in (3) shows respectively the specific and non-specific uses of the indefinite determiner in Seychelles Creole. Comparable pairs can be found for all the other FLC considered here:

(3) Specific use a. ~E fwa ti truv ana ~e pti garsõ ki ti apel Tizã Bollée 77:160 a/one day find there is a small boy REL Past call Tizã Once upon a time, there was a little boy named Tizã Non specific use b. u ti asir mwa ki zame ~e zur u a tap mwa ~e kalot Bollée 77: 184 you Past assure me that never a day you Fut hit me a slap you assured me that you would never hit me

2.2 Diverse distribution: definite, demonstrative and plural markers

Most FLC, (except for Seychelles Creole (SC) and Basilectal/Old Louisiana Creole (OLC)), feature both a definite marker: la and a distinct demonstrative marker: sa, which quite commonly co-occurs with the definite marker la. The relative order of these definite and demonstrative markers is quite variable in the distinct FLC as is the order of the plural marker that also commonly co-occurs with the definite marker or has itself a definite reading (see below). The range of possibilities is illus-trated here in Table 1. where the combined distribution of definite, demonstrative and plural markers across FLC is shown.

Following are some concrete data, illustrating the structures given in the Table 1. and a brief discussion of the central properties and uses of these determiners:

– Reunion Creole (RC):

(4) a. la kaz the (fem) house (Baker 2002:15) b. (sa) kaz (la) this house c. le(o) kaz the houses d. se kaz la these houses

Nominal constituents in French lexifier creoles 267

– Seychelles Creole (SC): Dem N Dem PL N

(5) a. sa kuto this knife b. ban lerat the rats c. sa ban zako these/ the monkeys

– Maurician Creole(MauC): Dem N (Def) Dem PL N (Def)

(6) a. liv la the book b. sa loto-la this car c. ban politisyen politicians d. ban liv la the books e. Sa ban lakaz dibwa These wooden houses

– Antillean Creole:

– St Lucie (StLC): N Dem Def PL N Dem Def (7) a. Tab la The table b. Kat mile sa la Those four mules c. se koSõ an the pigs

– Martinique(MartC): N Dem Def PL N Dem Def (8) a. Timanmay la the child b. Bel kay tala this beautiful house c. se tab la the tables d. se boug tala these guys e. *se tab Pl guys

Table 1. Definite, Demonstrative and Plural marker distribution in FLC

Singular PluralPre-Nominal Pl

RC Def/DmNP (def) La/saNP(la) Dfpl/Dmpl(PL) NP (def) L/s/e (ban)NPlaSC Dem NP sa NP Dem PL NP Sa ban NPMauC Dem NP (Def) sa NP-la Dem PL NP (Def) Sa ban NP-laStLCMarC

NP Dem Def NP-sa-(l)a,N-ta-(l)a

PL NP Dem Def Se NP-sa-(l)a,Se NP-ta-(l)a

GuaC N Def Dem N-la-sa PL N Def Dem Se N-la-saMLC (Def) N Dem-(Def)(la) N sa-la (Def /PL) N Dem-(Def) Le N sa-la

Post-Nominal PlBLC N Def/m N-la N (Dem= sila) PL N-(sila)yeGuyC Dem N Def Sa N la Dem N PL Def Sa N-y(e-l)aHC N Dem Def N-sa-a N Dem PL N-sa-yo

268 Viviane Déprez

– Guadaloupean Creole (GuaC) N Def Dem Pl N Def Dem

(9) a. tab la la table b. boug lasa this man c. sé boug la the men d. sé tab wouj lasa these red tables e. *sé timoun the children f. Nonm pasé an chimen-sala Men were passing on this road

– Basilectal /Old Louisiana Creole (OLC)

(10) a. Nom la et fom la rivé hiè The man and the woman arrived yesterday b. Nom yé et fommes yé rivé hiè The men and the women arrived yesterday

– Mesolectal Louisiana Creole (MLC)

(11) a. la fij vini reste avec mwa isi The girl came stay with me here Def girl came stay with me here b. je te pele vjφ mile-la Fiva They called the old mule Fiva they Past call old mule-Def Five c. Eu twa to tõde parle pu dõs-sa-la where you Past hear speak of dance Dem-Def Where did you hear about this dance d. Le vje mun parle e ta kreol The old people talk Creole a lot Def+Pl people talk a lot Creole e. Le kokodri-sa-la These crocodiles d. Le kokodri-la-la These crocodiles there f. (Le) kokodri sa ye These crocodiles

– Guyanese Creole (GuyC) Dem N Def Dem N PL Def (Baker 2002)1

(12) a. mé pagay-la, uben si u pimigò, pran guvernay-la (St Quentin 1872:116) b. agouti mo manjé kan mo té là bitachon yéla (St Quentin 1872:116) c. sa wom-la ‘cet homme’; sa fam-yéla ces femmes (St Quentin 1872:122)

(13) a. V[… sa PSEUDO-entélèktuèl-ya-ya (Ludwig & Telchid 2001) b. V [ sé PSEUDO-entélèktuèl-ya] These pseudo intellectuals c. Se moun-yan ces gens d. Sé difikulte ya ces difficultes e. Se problem yan les problemes

1. These examples are taken from Baker’s paper. As no glosses or translations are given, this author would rather not make them up, for fear of erring in the wrong direction.

Nominal constituents in French lexifier creoles 269

– Haitian Creole (HC)

(14) a. Bèl pye-bwa (l)a the beautiful tree b. manje sa a this food c. pitit sa a this child d. moun yo the people e. mouvman sa yo these movements

Before exploring empirical motivations for a common architecture, some general remarks about the uses of determiners in Table 1. are in order. Below, I consider each type in turn.

2.3 The definite determiner la

It is important to note that although called a definite determiner here, la is not the exact counterpart of either the French or the English definite determiners in any of the surveyed FLC. For instance, in contrast to the French definite determiner, la does generally not occur in generically interpreted noun phrases, although this use appears possible in some FLC (HC) (Joseph (1989:64)). Moreover, la is often not used in contexts in which definite determiners occur in languages like French or English. For instance, in Seychelles or Mauritian Creole, la does not occur with ordinal numbers (the first, the second etc) or in superlative constructions (the best), although it may occur in these contexts in other FLC (MarC, HC) (Déprez 2005). La is often said to have deictic force in the Creole literature, although, usu-ally, no precise empirical test is provided to support this claim.2 The claim re-lates in part to conjectures on the likely sources for this marker.3 For instance, the French post-nominal demonstrative reinforcer là found in expressions such as ce livre-là (this book there) is often considered a likely source for the FCL la. Should this conjecture be correct, then FLC la would indeed derive from a former de-monstrative form (a fairly common cross-linguistic source for definite determin-ers) and plausibly have preserved some of its deictic force. Interestingly, however, candidate substrate languages also feature post-nominal definite determiners that are likely sources for the FLC la. Ewe for instance has a lexical definite article lá that occurs in final position in the noun phrase: ati lá (the tree) (Lyons 99:77). The same is true of Wolof: Nadp-la (the vase) (Cérol 91:85). In further similarity with these languages, the marker la can apparently serve as a clausal determiner

2. An exception is Bollée (2004) in an interesting comparative study of definiteness in Sey-chelles vs Mauritian Creole.

3. For an in depth discussion on the origin of the post-nominal marker LA in Haitian Creole, see Fattier (2000).

270 Viviane Déprez

in many FLC (see for instance, Joseph (1989) for HC), a feature that is also shared with some informal French dialects. Most FLC also feature a locative adverbial marker la clearly derived from the French locative adverb là (here/there), also often considered a likely source for the post-nominal determiner la. There are, however, in some FCL rather clear arguments that the adverbial marker and the definite marker are currently distinct. In HC, for instance, only the determiner la presents an allomorphic paradigm /la, a, an, lan, nan/ governed by the phonologi-cal form of the immediately preceding lexical item (see for instance Joseph 1989 for the conditions of this allomorphy in HC) not the adverbial la. The same is true in MarC and more generally throughout the Antillean Creoles (Baker, pc.).4

The marker la has characteristic features of definiteness in at least some FLC. For instance, in HC, it exhibits the characteristic uniqueness and familiarity fea-ture of definite determiners in the sense that it occurs in so called anaphoric con-texts i.e. in reference to a unique entity previously introduced in the discourse. As noted in Fattier (2000:13) moreover, HC la can designate a unique object identifi-able by situational reference or even common knowledge reference.5 Addition-ally, la is usually unstressed and shows no proximate/distal contrast. Moreover, to our knowledge, in the all FLC considered here, la differs from demonstrative determiners in being incompatible with contrastive contexts. That is, just like the English or the French definite determiner, FLC la is infelicitous in contexts like (15) where a contrast is required:

(15) a. * Give me the book and the book b. * Donnes-moi le livre et le livre c. * Ban mwen liv la epi liv la (HC Vedrine pc) d. * Donn mwa liv la ek liv la (MC Guillemin pc)

La alone cannot be used to mark such a deictic opposition. In this respect, la clear-ly contrasts with demonstrative determiners, even of the so-called weakest types such as French ‘ce’, and manifests a feature characteristic of definite determin-ers. Finally la often can co-occur with other demonstrative markers such as sa as in (5b) or (14b) respectively, a feature that is cross-linguistically uncharacteristic of demonstrative determiners according to Lyons (1999) but rather common for definite ones. As noted by Lyons (1999:230) indeed, while the co-occurrence of a

4. Allomorphy on the definite determiner is also manifest in Guyanese Creole

Kochon-an the pig (Ludwig&Telchid 2001: 146)

5. Empirical work with a native speaker of Martinique Creole have led me to comparable con-clusions for this creole. In contrast to HC la, MC la, however, does not seem to be striclty inter-preted as singular. I am are currently conducting further field work to verify this distinction.

Nominal constituents in French lexifier creoles 271

demonstrative marker with a definite one is rather common, the co-occurrence of two distinct demonstrative markers in the same noun phrase is unusual.6 In sum, the failure to be used in contrastive contexts, the familiarity and uniqueness properties and the co-occurrence factor concur to suggest that la is currently in at least some FLC a definiteness marker. However, uses of la vary across FLC in quite subtle ways and although it may be that ‘deictic force’ plays a role in these differences, such a notion remains too vague to allow useful distinctive character-izations. Clearly, a detailed empirical comparison of the exact conditions of use of la in each of these FLC would be needed to further specify the precise semantic properties of this marker. This, however, is beyond the scope of the present paper whose focus is more on the syntactic distribution than on the semantic value of these determiners.

Concerning the syntactic position of la, Table 1. shows that it occupies a post-nominal position in most FLC considered here. Yet, although this order clearly reflects a strong tendency across FLC, it is not absolute. Pre-nominal la is found in RC (3a) and also commonly occurs in M Louisiana Creole (10a) as well as in St Thomas Creole (Valdman 1973: 512). While in RC, pre-nominal determiners are believed to be a direct inheritance from French (Baker 2002), in MLC, they clearly are a recent development and are sometimes taken as a sign of de-creolization (Newman 94:57), as earlier forms of the Creole did not feature them.7 RC and cur-rent MLC both seem to also feature a post-nominal marker la that can sometimes concurrently occur with a pre-nominal one (Newman 1984: 1348). The detailed distribution and respective interpretations of pre-nominal and post-nominal la in these FLC remains to be further explored.9

6. Demonstratives are known to be able to co-occur with reinforcers. La, however, being un-stressed, can hardly qualify as a demonstrative reinforcer, since these are usually stressed (cf. for instance the French reinforcer in expressions like ce livre-là.)

7. ’Malgré un certain degré de variation dans l’emploi des formes basilectales et non-basilectales, il est clair qu’un système pré-déterminé est en voie de se substituer au système post-déterminé’.(Newman 94:57) Despite a certain degree of variation in the use of the basilectal and non-basilectal forms, it is clear that a pre-determination system is currently gradually replacing the post-determination system.”Les déterminants définis l(e), la le antéposés au SN ne se trouvent pas dans les textes anciens’ N94:57). The definite determiners l(e) la le preposed to the SN are not found in the older texts.

8. La gros mezõ zon la se pu mwa the big house yellow the be for me The big yellow house is mine (Newman 94: 134)

9. According to Newman, the distribution appears sensitive to gender or phonological factors in current LC.(Newman.94:58). Baker also seems to note a gender difference in the pre-nominal

272 Viviane Déprez

Given the possible presence a pre-nominal la in RC and MLC, it would ap-pear that both pre- and post-nominal positions are in principle possible for all the markers featured in Table 1. Demonstative sa and plural markers also occur both pre-nominally and post-nominally. That is, although the position of these elements is usually fixed within a given Creole, the spectrum of variation across FLC does not strictly preclude either a pre- or a post-nominal position for any of them, even though some elements like la seem to favor a post-nominal one. This state of variability contrasts, for instance, with the strictly pre-nominal position of the indefinite singular determiner, which has never varied either across FLC or across time.

2.4 The demonstrative marker sa

Although sa is often not the only demonstrative determiner in a given FLC, it is nevertheless the most common one. Only two FLC seems to have lacked this de-monstrative at some point in their history. Broussard (1942) for instance, makes no mention of this determiner in his description of Louisiana Creole. ‘The demon-strative adjective to distinguish position is lacking. Cette femme-ci, cette femme-lá would both be rendered fomme-la. (Broussard:1942:5)’. Sa is apparently possible only as a pronoun in Broussard’s description. Determiner sa, however, is mentioned in Newman’s (1992) description of LC. Similarly, Fattier (1994) does not note the presence of determiner sa in the earlier HC texts of Ducoeur-Joly (1802) stressing on the contrary the strong deictic value of la at the time. Thus, it is plausible that at some point in earlier time, la may have supplanted sa as a deictic/demonstrative determiner, the distinction definite/demonstrative being in such cases the result of later a development along the characteristic path of grammaticalization that links definite determiners to earlier forms of demonstrative markers. In current FLC, it is rather uncommon to encounter sa alone. In many of its occurrences, this determiner co-occurs with la. La on the other hand can and does occur alone and is, in this respect, more autonomous than sa. Sa can nonetheless occur alone in a least one FLC, i.e. Seychelles Creoles (4a). It is then quite remarkable that in this FLC, determiner la is uncommon, and perhaps even altogether missing. For Bollée (1977), la in SC is only adverbial. The SC situation, in other words, seems to be the mirror image of Old Louisiana Creole or Old Haitian Creole with sa sup-

determiner of RC. La kaz : the house vs lε syen : the dog (the ε in the definite marker here stands for a shwa). The relation between pre-nominal determiners and the expression of gender needs to be further investigated. In variants that feature a pre-nominal definite marker, post-nominal la seems to have a stronger deictic value than the pre-nominal marker.

Nominal constituents in French lexifier creoles 273

planting la i.e. occupying the whole territory of the demonstrative-definite space instead of the reverse.

2.5 Plural markers

Although FLC have both pre-and post nominal number marking, it is notable that the markers themselves are lexically distinct according to their position. Post-nominal number in HC and Guy is marked by the morphemes yo and ye, respec-tively, which independently function as 3rd person plural pronouns in the relevant FLC. Both ye and yo have been argued to derive from the French strong pronoun ‘eux’ (them) (Broussard 1942), but as Sylvain (1979) noted, HC yo may also be related to the post-nominal Ewe plural marker wo. Whatever their actual origins, both markers are associated with definiteness in their pronominal use and in their determiner uses. In Guyanese Creole, moreover, there is evidence that ye is start-ing to fuse with the definite marker la, deriving a composite form ya. Frequent co-occurrence is quite plausibly at the source of the fusion of these two elements.

Contrasting with post-nominal plural markers, the pre-nominal markers se and ban do not function as pronouns in their respective FLC. A likely source of the Guadeloupean and Martinique Creole se is the French demonstrative plural deter-miner ces. If so, se does not have a pronominal source, although alternatively, se could perhaps just as well derive from the French possessive pronoun ses. Whatev-er its ultimate source, it is clear that se like the post-nominal plural markers of HC and Guy is necessarily associated with definiteness. The form of this association, however, manifests itself differently. As illustrated in (7e) and (8e), pre-nominal se must co-occur with the post-nominal definite marker la. Thus, it would appear, although the marker does not by itself encode definiteness, it must obligatorily be associated with a definiteness marker. Such a co-occurrence, although possible, does not seem to be necessary for ban in SC or MauC. Ban in contrast to se can occur alone with a noun phrase (5c), and although according to Bollée (1977), ban is often associated with a definite interpretation, there are SC examples where ban is compatible with an indefinite interpretation.

(16) I ana ban move voler (Bollée 77 VI:158)) there are bad thieves’

Such examples are also possible in MC as shown in (8c) above.In sum, plural markers do seem to differ lexically according to their position.

Only post-nominal plural markers also function as pronouns in their respective FLC. However, the importance of the pronominal factor in restricting the position of such markers or their definiteness value remains at present unclear. What is clear is that being functionally pronominal does not govern an observed necessary

274 Viviane Déprez

association with definiteness. The common and narrow association between plu-rality and definiteness in FLC, whose foundation need to be further understood (see Déprez 2006 for a suggestion) also arise for pre-nominal plural markers that do not have a pronominal use in the language.

3. A common architecture for FLC

The goal of this section is to discuss both conceptual and empirical arguments favoring a common basic structure for the FLC nominal constituents. The section begins with general conceptual arguments that support a common structure and then goes on to discuss empirical evidence for a particular organization of the functional projections within this common structure.

3.1 Ordering generalizations

Creoles, as is well known, are quite generally SVO languages, or to put it more formally languages that manifest a fairly consistent head-complement order (Bick-erton 1981). This rather solid generalization, clearly verified across FLC in verbal, prepositional, and adjectival constituents, appears at first view to be contradicted by the determiner final organization of FLC nominal constituents. Assuming in line with much generative work that functional elements project a syntactic struc-ture formally comparable to that of lexical constituents, the expectation is that, in similarity with the lexical structure, FLC functional structure should manifest a rigid head-complement order. This expectation is verified in sentential projec-tions where the TMA order narrowly reflects the head-complement functional structure proposed in the works of Cinque (1998) and others. But it appears con-tradicted in nominal functional structure, if, as is likely on the DP hypothesis, at least some of the determiners that occur in post-nominal positions are heads. Clearly, however, the consistent head-complement-order of FLC could be main-tained, if the post-nominal distribution of determiners could be argued to derive from an underlying head initial structure concealed after syntactic derivation. As it turns out, the comparative FLC data provide suggestive empirical evidence that favor such a view. This empirical evidence concerns three interesting asymmetries observed in the respective pre-nominal and post-nominal distributions of the FCL determiners in Table 1.

First, a cursory look at Table 1. suffices to reveal that variability in the respec-tive order of the determiners occurs essentially in post-nominal positions, not pre-nominally. Determiners in pre-nominal positions manifest a fixed order with the definite or demonstrative always occurring before the number marker, and never

Nominal constituents in French lexifier creoles 275

the opposite. In post-nominal position, in contrast, the plural marker may either precede or follow the definite determiner, la yo in HC vs. ye la in Guy, and the de-monstrative may either precede or follow the definite marker, la sa in Gua vs sa la in HC and ta la in MarC. This variability in relative ordering is never exemplified in pre-nominal position where the Def/Dem> Num> NP sequence is rigid.

Second, while there are some FLC determiners like the indefinite singular de-terminer, the numeral and universal quantifiers that consistently occur in rigid pre-nominal positions in all FLC and the known historical stages of their develop-ments, the reverse is never true. That is, there are no FLC determiner that consis-tently occurs in post-nominal position across all current FLC or their historical stages. Even the definite determiner la, which strongly favors a post-nominal posi-tion, clearly occurs pre-nominally in at least some FLC.

Finally, although there are FLC in which all determiners occur pre-nominally, see in particular SC, the converse is again never found. That is, there are no FLC that currently has, or ever had, all and only post-nominal determiners.10 In sum, as these observations reveal, there are important asymmetries in the pre-nominal and post-nominal distribution of FLC determiners: while the pre-nominal posi-tion is always minimally instantiated and shows a rigid and consistent relative or-der, the post-nominal position can fail to be instantiated and manifests a high degree of ordering variability. As these asymmetries demonstrate, the pre-nominal distribution of determiners is far more stable across FLC than the post-nominal one. This stability, in turn, suggests that it is the pre-nominal distribution of de-terminers that plausibly reflects the common base order of FLC nominal projec-tions, not the post-nominal one. If the pre-nominal order is the common base order for FLC nominal constituents, the existence of FLC with only pre-nominal determiners is expected and the absence of FLC with a complete converse order ceases to be surprising. For the post-nominal distribution, in contrast, a greater flexibility is expected if the different orders result from syntactic transformations. This analysis further makes possible the straightforward extension of the general head-complement order to FLC nominal projections, thus bringing nominal pro-jections in line with the wide-ranging right headedness tendency of FLC and more generally of Creoles.

Taking the pre-nominal Def/Dem Num NP order as the common base order of nominal functional projections in FLC, the remainder of this section examines

10. That a complete post-nominal order is permitted by UG, however, is in fact confirmed by the existence of languages such as Gungbe. In Gungbe, all determiners are found in post-nominal position in a rigid order. As argued by Aboh (1997), however, even in such a rigidly post-nominal determiner language, there are evidence that the post-nominal order is derived and not basic.

276 Viviane Déprez

empirical arguments that support the existence of distinct functional projections and of their respective order. Because FLC are strongly analytical languages and have a TMA architecture that so narrowly reflects the sentential structure proposed in a large number of generative work, they have been argued to wear, so to speak, their sentential functional structure on their sleeves (Rottet 1992). Although the observed order variability may at first view cast some doubt that the same is true for nominal projections, a more thorough inspection giving primacy to the pre-nominal order encourages this approach. If so, the observed data indicate that there are at least three potentially distinct functional projections above NP in FLC, a definite projection, a demonstrative projection and a number projection, each representing the distinct overt markers commonly found in FLC. (17) presents the common base structure proposed for FLC nominal constituents:

(17) DP De�nite Phrase

(la) D

la DemP

(sa) Dem

(sa) NumP Number Phrase (= NumP)

(Pl) NumYon

(Pl) NP

Demonstrative Phrase

Notably, (17) which is based on the observed order of FLC pre-nominal deter-miners, parallels the structure proposed in much recent generative works — such as Abney (1987), Ritter (1992), see in particular Bernstein (2000) for a summary of the theoretical and empirical arguments supporting this architecture based on much cross-linguistic research — confirming the view that Creoles also wear their nominal functional structure on their sleeves. In (17), the different markers are presented as either potential heads or potential specifiers of their containing pro-jection. I suggest below that both possibilities are instantiated in the distinct FLC and that the difference is an important source of ordering diversity. As for the singular indefinite determiner, I propose that it is merged invariably in the Spec of NumP across FLC, just like other number terms in a variety of languages (Ritter 1992). This will ultimately account for its constant pre-nominal order, once the derivations for other possible orders have been considered.

Nominal constituents in French lexifier creoles 277

In the rest of this section, evidence supporting the proposed independent pro-jections of the definite and demonstrative markers and their relative order is ex-amined. Next we consider evidence for the projection of number.

3.2 D(ef)P and DemP

In strictly pre-nominal position, definite and demonstrative markers do not gen-erally co-occur in the FLC surveyed.11 As a consequence, no obvious pre-nominal order for these determiners emerges from the pre-nominal available comparative FLC data. However, the fact that both determiners can, on the one hand, co-occur in a single nominal expression as in (4b) (5b)(7b) or (14c) and, on the other hand, occur alone separately on either side of a nominal constituent as in (3a) (4e) for sa and (2a) (4a) (7a) for la nevertheless suggests that each could be associated with a separate functional projection in most of the FLC considered.12 Comparative typological studies of languages in which definite and demonstrative regularly co-occur (see for instance Schloorlemmer 1998) indicate that the relative Def > Dem order is universally the default one. This is not to say that the reverse Dem > Def order is never instantiated, but rather, that whenever it is, this reversed order is the result of syntactic derivation. This is convincingly demonstrated by Panagiotidis (1999) for Greek, a language that clearly manifests the unexpected Dem> Def or-der as one of its possible demonstrative structure. Note, incidentally, that if correct, this generalization provides another argument that the post-nominal position of FLC determiners is derived. At it turns out indeed, the marked order Dem > Def is instantiated in FLC, but only when Dem is pre-nominal and Def post-nominal as in MauC (5b) or when both are post-nominal as in HC (14b) among others. If we take the sequence Dem> …Def to always be derived, no matter what element intervenes between them, then examples like (5b) and (14b) provide an additional argument that the post-nominal position of la and sa is not a base one. Even in an FLC like RC, where sa and la are attested alone in a pre-nominal position (3a, b), when both co-occur they have the marked order Dem > Def with la in a post-

11. That is, when demonstrative and definite co-occur, they are usually either distributed around the NP: Dem > NP> Def or both post-nominal. Thus, it may be that the definite and the demon-strative determiners are in complementary distribution when they both occur pre-nominally, like for instance in English. However, since in any event the pre-nominal position of the definite marker is quite rare across FLC, the lack of pre-nominal co-occurrence could as well be an ac-cidental gap in the distributional paradigm.

12. It may well be that Seychelles Creole and at least one dialect of Louisiana Creole (the one reported on in Broussard (1942); do not distinguish the DP and the DemP projection, since they apparently feature only one of these heads at a time, respectively sa for Sey, and la for Old Louisiana Creole. The same may be true of Old Haitian Creole. (see Fattier 1998).

278 Viviane Déprez

nominal position. The order sa la, which is in fact the most common one in FLC, is never attested pre-nominally, but only post-nominally or with sa pre-nominal and la post-nominal.

In his monography on definiteness, C. Lyons notes that ‘…there is a strong ten-dency for the definite article, if a free form, to occur initially in the noun phrase, independently of a language’s general constituent order pattern’ (Lyons 1999:64). Giusti (1993) Bernstein (1997) and Panagiotidis (1999) on the other hand, have independently argued that demonstrative markers are base generated in a con-stituent separate from DP, and generally occur in a base position that is lower than that of definite determiners. If all this is correct, then ordering a separate left headed initial definite projection on top of a demonstrative one is both compatible with the available FLC data and motivated by general cross-linguistic consider-ations. As will be shown in Section 4, moreover, this configuration has the further advantage of allowing an elegant derivation of the possible post-nominal orders of these FLC markers. Giusti (1993) has argued that demonstrative markers are uni-versally specifiers generated in a projection that is below DP. Panagiotidis (1999), however, has argued that demonstratives can head their own functional projec-tions and attract a deictic constituent or operator in their specifier in a way quite comparable to wh-complementizers in sentential projections.13 The proposition that demonstratives can head their own projection seems particularly attractive for those FLC that manifest ‘independent’ definite and demonstrative markers, the criteria for ‘independency’ being in the present case 1) the presence in an FLC of distinct definite and demonstrative markers (both la & sa) and 2) the possibility in the same FLC for the definite marker to occur alone. The second of this criteria attests that la is an independent head in the language and not a morphological suffix of the demonstrative. Evidence for a distinction is also provided by MauC (5a) data where sa precedes and la follows the nominal constituent, a configura-tion quite unexpected if the two belonged to a single projection. There are at least two FLC in which these two criteria are not met, SC and Old (Basilectal) Louisi-ana Creole. These FLC are interesting because they suggest that D(ef) and Dem projections may in some cases not be separate projections. As noted above, these Creoles exhibit either the determiner la (BLC) or the determiner sa but not both,14

13. That there are striking parallels between question and deictic structures is in fact recently argued in Diesel (2005), a paper that review the properties of a very large set of cross-linguistic demonstrative constructions.

14. ‘Les formes sila/sila-ye post-posees au nom sont les demonstratifs par excellence dans les textes anciens’ …la forme sa-la ne se trouve pas dans ces documents (Newman 1985: 137). The forms sila/silaye post-posed to the noun are the demonstative forms is the old texts… the form sa-la does not occur in these documents.

Nominal constituents in French lexifier creoles 279

suggesting that these markers have competed for the same position in these FLC, and in some sense, eliminated their competitor. Notably, in both SC and BLC, the unique available marker has strong deictic/demonstrative properties (Bollée 2004, Broussard 1942), being more like a demonstrative than a definite determiner. As is well known, cross-linguistically, languages may lack independent definite de-terminers, they only very rarely (if ever) lack a demonstrative one. What seems to be happening in these FLC then is that in contrast to others, they have not yet acquired an independent grammaticalized definite determiner. Consequently, we suggest that in these FLC there is only one functional projection fusing Dem and Def, and no separate DP. In other words, DP and DemP are separate projections in only a subset of the FLC — i.e. the FLC in which la has arguably acquired the status of an independent definiteness marker through subsequent grammaticalization.

The existence of a functional projection between DP and NumP has previ-ously been proposed in Rottet (1997) for Mauritian Creole. The projection is la-belled AgrP in his work, and is assumed to host both possessive pronouns and demonstrative markers. Since on the one hand, several FLC seem to allow the co-occurrence of a demonstrative with a possessive marker (see in particular Fattier 2004 for HC) and on the other hand, there are reasons to doubt the well founded-ness of agreement nodes in languages that manifests no overt agreement, I main-tain the DemP label for this projection. The label, however, is ultimately of little importance, the relevant claim being rather that demonstrative markers can, in some but not all FLC, occupy a functional head position that is both distinct from D0 and lower than DP in the nominal functional architecture.

3.2 The Number Phrase

In contrast to English or French, FLC manifest no number agreement between a noun and a verb or between a noun and an adjective. Furthermore, nominal expressions are not always marked for number in FLC. Bare nouns, for instance, manifest what Corbett (2000) has termed ‘general number’, permitting both a singular and a plural interpretation with the same unmarked nominal form. The hypothesis that there is a functional projection for number in languages that fail to obligatorily express grammatical number thus clearly merits careful consider-ation. This section discusses empirical arguments in support of what may perhaps seem the most controversial of the nominal functional projections in FLC, namely the projection of number.

The first consideration that supports positing a number projection is that number is expressed by distinct number words in all FLC. The origin and the form of this number word varies: recall that as noted above, SC and MauC pre-nominal ban clearly derives from the noun bande, Gua and Mart se from the demonstrative

280 Viviane Déprez

(or possessive) ces, Ha yo and Guy ye plausibly from the pronoun ‘eux’. Assuming that words always have an associated projection, the question is what the nature of these words are, i.e. is there evidence that they have a functional or grammati-cal nature and not a lexical one? We begin by considering the grammatical status of the number word ban arguing that its syntactic properties clearly distinguish it from nouns and measure expressions and characterize it as a grammaticalized marker of plurality. Next we discuss the syntactic position of pre-nominal marker se and finally that of the post-nominal markers yo/ye in the structure (17).

Since SC ban derives from a noun, it could be tempting to analyze ban + NP combinations as a special kind of bi-nominal construction akin to measure phras-es like ‘a group of people’. However, convincing empirical arguments can be given against such an analysis based on a point-by-point comparison between the be-havior of common nouns in such bi-nominal constructions and that of the plural marker ban (Rottet 1997). The arguments presented here are based on Mauritian Creole data but comparable data are also found is Seychelles Creole. They all point to the conclusion that ban is no longer a noun but a grammaticalized expression of number.15

Measure expressions, as is well known, entertain a very narrow semantic rela-tion with the nouns they serve to measure. Not just any measure phrase can be used with just any measured noun. There are restrictions that make expressions like ‘a school of children’ vs ‘a school of fish’ inappropriate. Such restrictions, how-ever, are not observed in ban constructions. Ban is compatible with nouns refer-ring to + human (5c) as well as to inanimate object noun (5d) indiscriminately. There are, in other words, no apparent semantic selection between ban and its complement: the only requirement is that the complement be a noun. In regu-lar bi-nominal constructions of MC, as (31) shows, both the head noun and the complement noun can host independent modifiers:

(18) a. En gro sak puasõ A big bag (of) fish b. En sak gro puasõ A bag (of) big fish

(19) shows, in contrast, that ban can never be modified independently of its com-plement noun:

(19) a. * gro ban zanfan big PL children b. ban gro zanfan la

15. Data for this section either come from Rottet (1993) or were directly elicited from informants.

Nominal constituents in French lexifier creoles 281

PL big children the the big children

Similarly in MC, both the head and the complement noun of a bi-nominal con-struction can have independent determiners (20). As (20b) shows, this is never possible for ban:

(20) a. Sa karotsu-la sa loto-a That tire of that car b. * sa ban-la sa zanfan la That PL the that children the

The comparison between [ban + NP] with a superficially synonymous measure noun [grup + NP] serves to further emphasize their difference, revealing besides syntactic differences important interpretative distinctions. As (21a) shows, the most natural interpretation of [ban+NP] construction is that of a (definite) plural. When ‘grup’ combines with a noun, however, it only has a partitive interpretation, never a plural one. As further shown in (21b), grup but not ban can follow a nu-meral quantifier. In such constructions again, ‘grup’ has a partitive interpretation that ban simply cannot have. Somewhat surprisingly, as (21c) reveals, ban appears to be able to combine with the indefinite determiner. However, as it turns out, ban in such constructions has a non-compositional interpretation meaning some that contrasts with ‘grup’, which must be interpreted partitively.

(21) a. Mo fin truv ban/*grup zanfan pe zue dan lakur I have found the kids playing in the yard b. Mo ti pe get trua grup/*ban zanfan pe zue dan lakur I was watching three big groups of kids playing in the yard c. Mo fin truv en grup/en ban zanfan pe zue dan lakur I found a group of/some kids playing in the yard

Since this combination en+ban is unusual in other respects, numerals being in general in complementary distribution with ban, it may be best regarded as an unanalyzed lexicalized expression [en ban] with the non-compositional mean-ing some directly merged in Spec NumP.16 To sum up, the preceding arguments show that the construction [ban +NP] is neither syntactically nor semantically equivalent to a [measure noun + NP] construction. In such constructions, the measure nouns usually maintain some nominal properties. Ban clearly has a more functional behavior since it imposes no semantic restriction on its complement, it

16. In this respect, the combination [en ban] recalls the French quantifier quelqu’un/quelques uns (someone) historically derived from the combination of the adjectival quelque and the in-definite numeral un, which also means ‘some’.

282 Viviane Déprez

precludes both modification and determination and it is associated with a particu-lar non-nominal interpretation.

Further supporting evidence for the functional status of ban comes from his-torical considerations. As noted in Baker (2002), the earlier uses of ban in MauC occurred in combination with an indefinite determiner and had the compositional meaning ‘a group of ’. (Ein band’p’tits miletons (Lolliot 1855:67) a group or small mules). It is only latter (end of 19th) that ban started appearing alone, i.e, without an indefinite determiner and began to take on the function of an optional plural marker. What the historical data indicate is that ban began to function as a plural marker when it began to enter in complementary distribution with the indefinite en. This is turn, suggests that when ban took on a more functional meaning (i.e. and thereby lost its capacity to be determined) it came to occupy the same syntac-tic position of en, i.e. Spec NumP in our view.

The exact position of ban within the NumP projection, however, is a matter of some debate. Rottet (1997) argues that ban is the head of NumP. I locate ban in the Spec of NumP. As Bollé (1977:38) notes: “The plural marker ban can be preceded by the demonstrative, the possessive and tu but not by a numeral or by a quantifier, apart from a few very rare exceptions”. Relevant MC or SC examples, including one with the rare combination of ban with a numeral are given below:

(22) a. ban liv la (MauC) the books b. Sa ban lakaz-la (MauC) these/those houses c. tu sa ban gardjè (SC) tous les gardiens d. ?? Sa ban trua gro zom (MauC) those three big men

Particularly relevant from the point of view of determining the position of ban within NumP are considerations concerning the respective ordering of ban and numerals. If as many have proposed (see Zamparelli 1995 among others), numer-als are projected within the NumP projection and merged in its specifier position, it follows that ban cannot both occupy the head of NumP and precede a numeral also contained in the NumP projection. As clearly shown by examples like (22d), however, ban must precede a numeral whenever they co-occur — however rare this co-occurence may be. It can never follow it. Examples like (22d) thus suggests that at least in these context, ban cannot be in the head of NumP. It is important to note, however, that the co-occurence of a numeral with a plural marker, although clearly ordered Pl> Numeral, is not a common way of expressing cardinality in FLC. As it turns out, several of our informants reject (22d), or only accept it with

Nominal constituents in French lexifier creoles 283

a lexical reading for ban comparable to that of grup. In regular cases, numeral expressions are used without any plural marker, the number term alone being suf-ficient to indicate plurality. That is, in contrast to languages like English where in cardinal expressions like three books, number is redundantly marked by the cardinal and the morphological plural on the noun (-s), FLC in general are more ‘economical’ in that number is expressed only once by the presence of the cardinal numeral and there is no additional marking of plurality for the noun. Across FLC, numerals are, in fact, quite generally in complementary distribution with plural markers when part of an indefinite cardinal expression. If as suggested above, the plural marker occupies the Spec of NumP and not its head, this complementary distribution arguably results from a competition between the number term and the plurality marker for the same syntactic position. This in a sense, assimilates number and numerals in languages like FLC, a potentially attractive perspective that put singular and plural (cf. the number one as a marker of singularity in Spec NumP) on the same level and is in line with the idea that plurals can be cardinal-like expressions in some languages (Heycock and Zamparelli 2000).

To sum up, the data reviewed above suggest that ban and numerals compete for the same syntactic position. In regular indefinite cardinal expressions they are in complementary distribution with each other. These data are accounted for if as proposed here, ban like numerals is merged in Spec NumP.

3.2.2 Position of the plural marker se in Guadeloupe and Martinique CreoleTurning now to the pre-nominal plural marker se of Antillean and Guadeloupe Creole, we first note that its distribution with respect to other determiners is rather similar to the one discussed above for ban. Like ban, the plural marker se follows universal quantifiers like tout (23a). Like ban again, se does not co-occur with nu-merals in cardinal nominals with an indefinite interpretation like (23b).

(23) a. Tout se timoun la (sa) all PL three children the these All these children b. twa timoun Three children c. * Sé twa timoun

(24) a. Tout se moun ta la anfrans (MarC) all PL people these the France All these people are in France b. Man ni senk yich I have five children

284 Viviane Déprez

Note that in MarC and GuaC, as in SC and MauC, semantically plural nominal expressions with cardinal numeral do not need to contain a plural marker. This again suggests that numerals and se could be in complementary distribution and compete for the same syntactic position Spec NumP. Cyrille (1997) who also ar-gued for the existence of a NumP projection in Guadeloupe Creole proposes in-stead that se projects as the head of NumP. But this view turns out to be unable to account for the complementary distribution of numerals and plural markers in indefinite cardinal nominals and for their relative order when they do co-occur. The complementary distribution of se and numeral, indeed, is not as systematic in GuaC and MarC as in MauC and SC. There are cases where the plural marker se can co-occur with a cardinal term completely naturally. In such cases, se must precede the cardinal term and never follow it.

(25) I vini épi sé dé misye a he come with PL two men the He came with the two men

These examples, however, are always semantically definite i.e. equivalent to the combination of a definite + a cardinal expression as in the English: the three books. They can never simply mean three books. This necessary interpretation is related to an important distinction between the plural markers se and ban. In contrast to ban, se in Martinique and Guadeloupe Creole never occurs alone with nomi-nal expressions but must always obligatorily co-occur with the definite marker la. That is, expressions such as *se liv (books) are always ungrammatical. (Cyril 1997, Damoiseau 1999) and la must always be present whenever se is: Se liv la. Conse-quently, while the pre-nominal plural marker ban seems compatible with an indef-inite reading as in (16) above, this is impossible in Martinique and Guadeloupean Creole where plural invariably associates with the overt definite marker and with a definite interpretation. Given this property, it turns out that the co-presence of a plural marker with a numeral in MarC and GuaC causes an obligatory change in the meaning of cardinal expressions from indefinite to definite. The association of a numeral with a plural marker in these FLC thus serves to introduce definiteness rather than plurality. I would like to suggest that the pre-nominal plural marker ban and se differ in that se is associated with a [+ definite] feature that must be overtly checked, while ban in contrast, is unmarked for definiteness, being in this respect compatible either with a [+definite] marker or with indefinites.17 The con-sequences of this proposal will be seen in the derivations of the respective orders, discussed in Section 4.

17. The indefinitness marker on the other hand, can be assumed to be [-def]. As a result, it will be generally incompatible with definite markers.

Nominal constituents in French lexifier creoles 285

As an alternative account of the difference between SC, MauC and Gua,Mart plural markers, structure (17) could be complexified. Based on a study of nominal structure in Gungbe, Aboh (1999) indeed proposes to subdivide a projection that contains a cardinal expression from a projection that contain grammatical num-ber proper which he terms PlP.

(26) Aboh (1998)DefP

PLP

DemP

NumeralP

NP

As shown in (26) Aboh locates the projection of PPl above DemP, his main moti-vation for this choice being based on the base structure and derivations he propos-es to account for the determiner order of Gungbe nominals. There are, however, both conceptual and empirical arguments against generalizing Aboh’s particular proposal cross-linguistically. Note first that sandwiching the number projection between the definite and demonstrative projections suggests that plurality is as-sociated with definiteness. But although this association is verified in Gungbe as well as in HC, MarC and GuaC in the FLC considered here, it is certainly not cross-linguistically valid. There are many languages such as English or Arabic in which both definite and indefinite noun phrases are specified for number. Even within our FLC sample, there appear to be a division between GuaC MarC HC and GuyC on the one hand, where plural systematically goes with definiteness and SC and MauC where plural is compatible with an indefinite interpretation. There are, moreover, no strong evidence that the association between definiteness and number should be expressed syntactically through the position of number. See for instance Déprez (2006) for an entirely distinct proposal. Other linguists (Heycock & Zamparelli (2003), Borer (2005)) although dissociating one projection for num-ber from one for numerals do not assume the order proposed in Aboh (1999). Fo-cusing on FLC, the exemplified order (se twa liv) would rather locate the number phrase directly on top of the Numeral Phrase as in (27).

(27) PLP

NumeralPse

NPliv

twa

286 Viviane Déprez

Note that under such a proposal, se could be in the Specifier of PlP and cardinal expressions in the Spec of NumeralP, accounting for their relative pre-nominal ordering in examples like (25). Recall, however, that SC/MauC, on the one hand, and GuaC, MarC, HC on the other hand are distinguished by the co-ocurrence they allow between numeral and plural markers. In SC and MauC, co-occurrence is essentially impossible. In GuaC, MarC, and HC, co-occurrence is fully possible but with a change of meaning for the numeral expression from indefinite to defi-nite. Now this distinction could be accounted for if SC/MauC were assumed to only allow the projection of Num(eral)P and not that of an independent PlP. That is, in SC and Mau, only Num(eral)P would project and ban would then be merged in Spec Num(eral)P. This accounts for a complementary distribution between ban and numeral discussed above. In contrast, in Gua and Mar, PlP and Num(eral)P can project independently, therefore allowing co-occurrence.

Note that this proposal parallels in interesting respects the distinction made above for Definite and Demonstrative projections in SC and Louisiana Creole vs other FLC. This suggests that in FLC featuring an independent PlP projection, number is more grammaticalized than in FLC featuring a fused projection for Nu-meral and number. In SC and MauC, NumeralP and PlP are not yet distinguished, a situation that can be taken to correspond to less grammaticalized plurality. Ban on this view has the status of a vague cardinal term that encodes the cardinality > 1. Whether this leads to notable semantic differences for the interpretation of plural noun phrases remains to be further established. This complementarity is not ob-served in GuaC and MarC where se can co-occur with plural, nor it is observed in FLC like GuyC and HC in which plural is pronominal in nature and also systemati-cally associated with definiteness. For these languages, then, grammatical number may be dissociated from numerals in a way encoded in the representation in (27).

Turning finally to the post-nominal pronominal plurals ye and yo, the pro-posal for now is that like se they are merged in the PlP projection. The difference I wish to propose is that these plural markers are merged in the head of the projec-tion rather than in the specifier. Motivation for this view derives essentially from derivational considerations discussed in the next section. ye and and yo will thus be discussed in more details when we turn to the derivations of the determiners order in Guyanese and Haitian Creoles respectively.

In this section, arguments for a unified base structure of the FLC nominal constituents were provided. FLC were shown to present clear asymmetries in the pre-nominal and post-nominal distribution of their determiner, asymmetries that were argued to favor the pre-nominal order as the base structure. (17) was pro-posed as the underlying base structure and argument were provided for each of the functional projections proposed in (17) and their respective hierarchical or-ders. Only a subset of FLC were assumed to distinguish both a demonstrative and

Nominal constituents in French lexifier creoles 287

a definite projection, i.e. those in which evidence could be found for the existence of an independent definite determiner. Similarly, FLC were assumed to differ as to whether or not they subdivide their number projection and an independent PlP is provided in addition to a Numeral projection. In both cases, the FLC that distinguish additional projections were suggested to have a more grammaticalized determiner system than those that do not.

4. Deriving the distinct orders

Given the basic architecture of nominal projection in (17), the only orders that are predicted so far are those of pre-nominal determiners. Consider first attested pre-nominal orders. The la+NP order is found in Reunion Creole and irregularly in current day Louisiana Creole, sa+NP occurs in Seychelles and Mauritian Cre-oles. Pl+ NP is realized as ban+NP in Seychelles and Mauritian Creoles and as se+ NP (+la) in Antillean and Guadeloupe Creole. All these pre-nominal orders are derived straightforwardly given (17). The obvious question is where do the post-nominal orders come from? In this section, we turn to an analysis of the post-nominal ordering and discuss the detail derivations that produce the observed diversity. The analysis provided here is couched within the Minimalist model of syntax proposed in Chomsky (1995–2000). An important feature of this model is that it imposes economy conditions on movement. Syntactic displacement is im-possible unless it is forced. With (17) as the base order, the question then becomes what motivates the syntactic displacement that results in the different FLC order-ing? In the Minimalist model, movement is triggered by operations on features. Certain features present in the structure are uninterpretable at the cognitive inter-faces of language, the sensorial-perceptual interface or phonological component or the conceptual intentional-interface concerned with interpretation. For a deri-vation to be successful, uninterpretable features must be eliminated. As proposed in Chomsly (2000), uninterpretable features can be eliminated either through de-letion or through valuation by a corresponding interpretable feature that enters in agreement relation with them. In short, an uninterpretable feature acts as a probe that looks within a limited search space for an accessible feature that has an adequate value. If such a feature is found in the search space, movement occurs to allow agreement under a Spec-Head configuration and valuation of the uninter-pretable feature by the interpretable one is achieved. If such a feature is not found, an unvalued feature of the same nature can trigger movement and deletion under agreement. If neither valuation nor deletion obtain, the derivation crashes and ungrammaticality results. With this formal apparatus at hand, let us now return

288 Viviane Déprez

to FLC and see how such an approach can explain the diversity observed in the placement of their determiners.

The first question to investigate is what type of movement is involved. Assum-ing the architecture (17) as common to all FLC, displacement may a priori involve the movement of a head or that of a maximal projection, the two entities that on Minimalist assumptions are objects visible to syntactic operations. Let us begin to consider the simple case of a post-nominal definite determiner.

(28) * [DP D0 [ NP N ] ]

In this simple case the order liv la could in principle either derive from the head movement of the noun over the determiner head, in similarity to what happen in Romanian where the definite article cliticizes to the left of the head noun, or from the movement of the entire NP projection. This question, in fact, turns out to have a very simple answer as soon as more complex nominal constituents are consid-ered. As is evidenced by the data in (39) and (40), it is not just the head noun that occurs before the determiner, but rather the noun with some determination mark-ers, all of its possible modifiers, adjective or relative clauses included, and all of its relevant complements. Clearly then, it is the whole nominal constituent, head and dependents included that is displaced, not just the nominal head alone. Final definite determiners, in fact, occur in the absolute final position of the entire noun phrase, a position expected only if a whole nominal projection is moved.

(29) a. [bel kay chè ] la (HC) beautiful house expensive the the beautiful expensive house b. [Sa ban zen zom katolik ki Mari kõtã] la (MauC) Dem PL young men catholic that Mari loves Def Those young Catholic men that Mary loves

As comparable arguments can be given for the post-nominal position of the other determiners, it is clear that phrasal movement is the key to the observed ordering variation here, not head movement.

The question of why this should be naturally arises. As numerous consid-erations in the generative literature have found head movement to be generally linked to rich inflectional morphology, the lack of head movement in FLC should come as no surprise, FLC, like other Creoles, being well known for the paucity of their inflectional morphology.18 A further consideration suggests itself. Given that

18. The case of verb movement in current Louisiana Creole convincingly argued for in Rottet (1992) constitutes an interesting exception for head movement within FLC.

Nominal constituents in French lexifier creoles 289

heads are in effect the minimal part of XP projections, there is a sense in which head movement can be considered as a subcase of XP movement. The reverse, however, is not true. From this point of view then, XP movement appears to be a more basic type of displacement than head movement. If Creole languages mani-fest a tendency to instantiate unmarked choices along available parametric options (see Roberts (1999) for a discussion of this view in regards to verb movement) the choice of XP movement over head movement appears in line with this tendency.

Going a bit further, I would like to propose that within FLC nominal projec-tions, markers that are heads systematically attract constituents into their specifi-ers. To put it differently, all FLC determiner heads systematically contain uninter-pretable features in the sense of Chomsky (2000) and thus, systematically attract a phrasal projection in their specifier to allow the valuation of the uninterpretable feature they contain.

Proposal: XP movement Nominal functional heads are probes in FLC

Again, questions arise as to why this should be, and perhaps more particularly as to why such a systematic attraction pattern is manifest primarily in nominal constitu-ents but not in other functional domains, such as for instance the verbal functional domain. In all FLCs, indeed, verbal functional projections, and more generally Tense Mood and Aspect markers quite generally precede the verbal constituent. They do not follow it. Although an in depth comparison of the verbal functional domain and the nominal functional domain in FLC is beyond the scope of this paper, I’d like to mention a few factors that may help understanding this intriguing difference. First, it may be of relevance that nominal functional projections seem to manifest a certain amount of semantic or pragmatic overlap or ambiguity. That is, markers in the nominal functional system seems to have changed from one meaning to another across time (cf. in particular la and sa), without necessarily changing their morphological shape. This does not seem to have occurred to the same extend in the verbal functional paradigm. Another consideration seems also important namely the importance of substrate and superstrate influence. Both the substrate languages and to some extent the superstrate language do seem to ex-hibit a certain amount of phrasal movement in the nominal system but there is no comparable displacement in the verbal system. More precisely, African languages like Ewe, Fon Gbe and Wolof among others clearly manifest post-nominal deter-miners rather similar to those found in FLC. The positions of these determiners have been argued to be derived from phrasal movement (cf Aboh 1997). This con-stitutes a clear model for the FLC determiner order. Furthermore, as argued by Bernstein (1997), French also seems to manifest phrasal movement in the deriva-tion of its demonstrative reinforcer pattern. Constructions such as ce livre-là, that

290 Viviane Déprez

are a plausible source for part of the definite/demonstrative determiner system of FLC are analyzed by Bernstein as involving phrasal movement of the nominal con-stituent. Phrasal movement has also been proposed by Kayne (1994) as a source of the order of constituents in French possessive constructions. Given on the one hand, that FLC possessive constructions present interesting similarities with those of French and, on the other hand, that the post-nominal position of some deter-miners may have originated in locative/possessive constructions (Fattier 2000), possessive constructions provide another attractive source for the systematization of phrasal movement in the FLC nominal system.

It follows from the proposal made above and from a strict notion of locality discussed below that if a determiner is a head in FLC, it will have a post-nominal position. Since not all determiners are post-nominal, it must also be the case that they are not all heads. We have discussed above potential evidence based on dis-tributional facts that the pre-nominal plural marker ban occupy the Spec of Nu-meralP and that at least a partial similarity in the distribution of se also suggest that it occupies a Spec position in its projection. Although comparable evidence still need to be discovered to characterize the status of demonstrative markers in FLC, it is clear that a survey of the cross-linguistics data provides ample evidence that demonstrative markers can either be heads or specifiers, crosslinguistically. Such a distinction, for instance is explicitly argued for in Shloorlemmer (1996) and Panagiotidis (1999) among others. Pending supporting empirical evidence, I assume that this distinction is also present in FLC and may in fact be an important factor of variation. Within the realm of diachronic data, a change of status from a specifier to a head (or from an XP to an X0) for other elements has been a fre-quently noted evolution. From the perspective of creolization (as well as historical change), this change, I suggest, could also be understood as a further step along the dimension of grammaticalization. That is, on this perspective a head element can be taken to be more grammaticalized than a specifier element. If this is cor-rect, expectations are that more grammaticalized elements should have more ab-stract meaning, be less stress bearing, manifest more clitic like behavior and be morpho-phonologically dependent to a greater extent than less grammaticalized ones. As an exploration of these factors remains to be done in the FLC under study (see Déprez 2005 for empirical evidence relating to the definite/ demonstrative projection), the analysis developed here should be taken as preliminary. Its overall direction, however, seems promising enough to warrant further exploration, as is done in the remaining of this section. Armed with phrasal movement and vari-ability in the specifier/ head distinctions of the elements considered, the rest of this section proceed to offer a derivation for the determiner ordering of each of the FLC in Table 1.

Nominal constituents in French lexifier creoles 291

4.1 Seychelles Creole

The data reviewed for Seychelles Creole seems essentially to manifest the linear order predicted by the structure (17). This is clearly the case for the demonstrative and the plural markers. In this respect, Seychelles seems to exemplify the absence of movement. Things are a bit less clear with respect to the definite marker la. In the data reviewed from (Bollée 77) la never seems to occur as a determiner. There is a locative la in Seychelles Creoles, just as in the other FLC, but this Creole does not seem to make use of an overt definite marker distinct from a demonstrative one (see Déprez 2000). That is sa in Seychelle Creoles seems to serve both as a demonstrative or as a weak deictic determiner, making reference to an entity in the context or to one previously mentioned in the discourse. Possibly, as discussed above, in Seychelles Creole, the DP projection is indistinguishable from the DemP projection or simply not projected. If so there are only two functional projections in Seychelles nominal constituents and their order results from the absence of movement.19 If there is no distinct DefiniteP projection in Seychelles, then there cannot be an uninterpretable feature in D0 that needs to be valued or eliminated at the level of DP. The plural projection moreover, being apparently compatible with either a definite or an indefinite reading may be assumed to be unmarked for definiteness. This would account for why no movement is required as there are no features that need valuation or elimination. In short, in SC, functional projections have a null head that call for valuation. An element is then merged in their speci-fier providing the needed valuation at merge with no need for further derivation.

(30) [DemP SA [NumP BAN [ NP N ]]] Sa ban NP sa: Spec DemP, ban: Spec NumP

19. From a rapid inspection of the Seychelles Creole texts in Corpus Creole (Ludwig &Telchid 2001), there seems to be a few cases of la that could pass for a definite determiner. In all these cases, la is post-nominal, just as it is in Mauritian Creole. It may be then, that in some current dialects of Seychelles Creole there is a DP projection, in which case, SC would be pretty much identical to Maurician Creole from the point of view considered here, with a movement of the DemP projection to the Spec of DP. Further empirical research, however, is needed to better establish this point.

292 Viviane Déprez

Structure of Seychelles Creole:

(31) DefmP Demonstrative Phrase

sa Dem

NumP

ban Num

NP

Number Phrase (= NumP)

4.2 Mauritian Creole

Turning now to MauC, there is evidence in this Creole for the existence of a DP projection separate from DemP. Present day MauC, to a greater extend than Sey-chelles, distinguishes between the definite la and the demonstrative sa. As noted in Baker (2002) for instance, there has been an interesting change in the syntax of la in the history of Mauritian. The first mark of determination in MauC seems to have been a bimorphemic sa N la where neither marker was independent of the other, and both had essentially a indistinct definite/demonstrative value. Later on, a distinction occurred gradually between the la and the sa, in that the former be-gan to appear alone with nominal constituents, i.e. independently of sa and with a more distinct definite value. An important point in the evolution is marked by a change in the ordering of la in relative clauses. Baker (2002) observe that the dis-continous sa…la form used to surround the head of a relative clauses as in (32):

(32) Vous per sa bon Dié là qui mauvais Pitot 1805:373 in Baker (2002) you afraid this god the that bad You are afraid of the bad god

Starting with 19th century texts, however, a new pattern emerged where la came to be placed at the end of the entire relative clause

(33) u per sa bondye ki move la you are afraid of the god that is bad

The latter pattern is the only one possible in current day MauC. Syea (1996) analy-ses this evolution of la as a change from a nominal affixe to a clitic and then to an independent head “which selects as its complement the noun phrase and its modifier” (Syea: 184). From our point of view, MauC exemplifies a case where the definiteness marker has become an independent functional head while both the demonstrative and the plural markers are still specifiers that provide valuation to their own functional projections.

Nominal constituents in French lexifier creoles 293

Within a minimalist framework, the distinction of la as a functional head sug-gests that it now has an uninterpretable feature. Tentatively, I suggest that this feature is a person feature. In other words, la has an uninterpretable person feature uP that must be valued. Any constituent that contains a person feature will then be able to value la under movement to its specifier. Plausibly valuation through movement to Spec will obtain with any phrase containing a noun if a noun has default 3rd person feature. As a result, la systematically finds itself in a final posi-tion in Mauritian Creole.

The uninterpretability of the person feature in la may be understood as result-ing from a process of grammaticalization. Definite determiners in FLC plausibly derive from former demonstrative forms. As is well known from grammaticaliza-tion studies, demonstrative forms are often first pronominal forms and only later determiner ones. The Table 2. below provides a typical cline of grammaticalization for definite determiners showing this tendency.

Table 2. The grammaticalization cline from demonstrative to definite

DEICTIC PARTICLE+CATEGORIAL NOUN

DEMONSTRATIVEPRONOUN

DEMONSTRATIVEDETERMINER

WEAKLY DEMONSTRATIVEDEFINITE DETERMINER

DEFINITEARTICLE

AFFIXALARTICLE

NOUNMARKER

Reinterpreting this cline within our formal system, I suggest that the passage from pronominal to determiner and from demonstrative to definite could plausibly be interpreted as involving the ‘bleaching’ of a person feature. From interpretable, and thus pronominal in nature, the person feature becomes uninterpretable and thus functional in nature. It then acts as a probe for a constituent that may value it, i.e. the role of the moved noun phrase in the present scenario. Assuming that un-der Minimalist assumptions, valuation involves a Spec-Head relation, the move-ment of some nominal constituent to the Spec of D0 la will be necessary, deriving the systematic post-nominal position of this determiner.

Concerning the number marker ban evidence that it is merged in Spec comes from it complementary distribution with numerals as discussed above. Thus to sum up, the structural assumptions made here for Mauritian Creole are that la is a head with an uninterpretable person feature, sa is merged in the specifier of a null Dem node to value it and ban is merged in the specifier of Num for the same reason. The attested order of Mauritian nominal projections is then derived as fol-lows: the DemP projection is attracted to the spec of the DP projection with every-thing that it contains to value uP in la. This derivation is represented in (34):

294 Viviane Déprez

(34) DefP De�nite Phrase

D

la DemP

sa Dem

NumP Number Phrase (= NumP)

Noun Phrase

ban Num

NPzomgro

Demonstrative Phrase

4.3 Antillean Creole

Antillean Creole and Martinique Creole both involve one more step in the deriva-tion of their determiner order than Mauritian. In Antillean, I suggest, both the Def and the Dem markers are heads that contain a uP feature. First, Dem attracts its complement PlP into its specifier to value its uP features resulting in the [ [(Pl) NP] Dem] order and then Def also attracts its complement to its specifier, presum-ably to value its uP feature, resulting in [ [(Pl) NP Dem] Def ] order. This deriva-tion is represented in (35):

(35) DefP De�nite Phrase

D

la DemP

Dem

PLPs/ta Number Phrase (= NumP)

Noun Phrase

se Pl

NumP

Demonstrative Phrase

de NPmoun

Nominal constituents in French lexifier creoles 295

The derivation is a pure case of Comp to Spec movement, with heads always at-tracting their direct complement to their specifiers. As for the plural morpheme se, I have assumed in the structure above that it is merged in the Spec PlP to value a number feature in the null head of PlP. I further suggest that it may also contain an uninterpretable person feature. This feature can be checked by the movement of PlP in the Spec of either Dem0, or D0 which themselves contain an uninterpretable person feature. Note that on this proposal, a nominal constituent that projects only a PlP with se in its Spec will be ruled out, the uninterpretable person feature of the plural morpheme remaining unchecked. If as suggested here, se is merged in the specifier position of PlP, its uninterpretable person feature cannot be valued by the movement of NP to its Spec. Consequently, the projection of a head able to check and delete uP of the plural marker is required, accounting for the obligatory co-presence of the definiteness marker with the plural marker in this FLC.

4.4 Guadeloupean Creole

Recall that what distinguishes Guadeloupean from other Antillean Creoles is the fact that it manifests an unusual post-nominal Def> Dem order. Given the pro-posed structure (17), this order can in fact obtain easily if a Spec to Spec phrasal movement is involved rather than a Comp to Spec one. That is let’s suppose that in Guadeloupe Creole, PlP moves to Spec DemP [se NP] sa and subsequently to Spec DP as in (44). This derives the post-nominal Def> Dem order, appropriately checking the uninterpretable person features of se, sa and la.

(36) se de moun la sa these two persons

DefP De�nite Phrase

D

la DemP Demonstrative Phrase

Dem

sa PlP Plural Phrase (= NumP)

se Pl

NumP Numeral + Noun Phrase

de moun

296 Viviane Déprez

Recall from earlier, that the order Def> Dem is apparently not the only one found in Guadeloupe Creole: variants showing the same order as Antillean (sa la), i.e. with a Comp to Spec rather than a Spec to Spec phrasal movement are also pos-sible. This second order illustrates the same pure Comp to Spec derivation as in the other Antillean Creole. That is, first PlP is attracted to Spec DemP to check its uninterpretable feature against uninterpretable person feature of Dem0, then the DemP is attracted to Spec DP to check D0.20

4.5 Old Louisiana Creole

Old Louisiana Creole is the first Creole here considered to show a post-nominal plural marker. This suggests that the OLC plural marker is a head with an unin-terpretable feature that attracts its NP complement. Again, it is plausible to think that this uninterpretable feature is a person feature distinguishing a pronominal yé from a number marker. That is, while the 3rd person plural pronominal marker yé has an interpretable 3rd person feature, this feature is uninterpretable in the plural marker, turning yé into a functional projection that acts as a probe and triggers the movement of the complement NP. Although there are distinct dialects of Louisi-ana French Creole, the one considered in Broussard 1942 does not seem to clearly distinguish the Def projection from the Dem one. In this dialect, la is clearly used but, there is no mention of sa anywhere and la is said to have a particularly strong demonstrative force. The suggested analysis, here is that la occupies the head of an undistinguished DemP/DP projection immediately above the NumP (as in Sey-chelles Creole), and attracts its complement to check its feature. Given that Num0 attracts its NP complement and that D0 does so as well, we have here again a pure Comp to Spec derivation where heads attract their direct complements. Note how-ever, that in this dialect, the plural marker seems to be in complementary distribu-tion with the Def/Dem one so that an NP yé la order is never attested. A possible explanation for this fact could be that the definite marker la has come to mark singularity in addition to definiteness, hence becoming incompatible with the plu-ral head yé. Alternatively, since plural NPs with an overt plural marker are usually interpreted as definite in this Creole, it may be that there is a definite null head in D0 with the whole NumP moving to its specifier, and that somehow, after phrasal movement to Spec D of the whole NumP projection, yé has come to be absorbed phonologically by the Def head in the Spec of which the NumP projections sits. This alternative is represented in (37):

20. Following Chomsky (2000), it is assumed here that uninterpretable feature do not delete until the end of a phase and that DP is a phase.

Nominal constituents in French lexifier creoles 297

(37) [Def/DemP [NumP [NP N]j YE tj]i 0 ti ]

Louisiana Creole seems to have had an earlier stage or a distinct variant in which la could still co-occur with the plural marker ye. In this diachronic (?) variant the Def /Dem marker preceded the plural marker with the resulting order NP la ye as in (38):

(38) Pikan la yé déchiré ma lapo Valdman from Newman text These spikes have torn my skin in Veronique 1994: 196

As before, this variant can easily be derived with a simple change in the course of the derivation. Instead of having heads systematically attract their immediate complements, a Spec to Spec derivation with NP moving first to Spec NumP and then on to Spec DP gets the desired result, given the assumption made above that only NumP and Dem/DP project in this dialect. That is, in the NP la ye dialect, NP also move to Spec NumP but it then continues to move to Spec Def/DemP leading a post-nominal order of the markers that parallels the one observed in pre-nominal position (Def/Dem > Pl).

(39) [Def/DemP [NP N]j LA [ ti YE ti ] ]

If this is correct, we have here, as above for Antillean C, two (diachronic?) vari-ants that are distinguished by the type of phrasal movement they involve (Spec to Spec vs Comp to Spec) rather than by the status of their markers (Spec/Head). It might be hypothesized then, that the Spec to Spec derivation in time, moved to a ComP to Spec derivation, leading a potential [[ NP ye] la] order, which through phonological absorption itself led to the complementary distribution of la and ye, in similarity with what seems to have happened in Guyanese Creole below.

New Louisiana Creole, as termed here in Table 1. above (cf. the non-basilectal dialect of Newman 1994), presents both a pre-nominal definite determiner and pre-nominal plural marking. Both of these characteristics are said to result from the ‘decreolizing’ influence of Lousiana French, which is closer to French itself. As Louisiana Creole seems at present to be rather irregular, so that a post-nominal determination system co-exists with the newer pre-nominal one, a more detailed study would be required to allow us to integrate this system within our proposal. We thus leave this case for further study.

4.6 Guyanese Creole

Guyanese is also a post-nominal plural Creole, involving a PL head that attracts its NP complement. Dem, however, appears to be a specifier in this Creole, as it always remains pre-nominal. Def, on the other hand, is a head that also attracts

298 Viviane Déprez

its complement as in most other FLC. The resulting derivation is one in which the plural marker attracts its complement (NP > Pl) the Def marker as well ([Dem NP PL] Def) but where movement to Spec Dem does not occur.

(40) safam-y(el) athese women

DefP De�nite Phrase

D

la DemP Demonstrative Phrase

sa Dem/D

PlP Number Phrase

Pl

ye Noun PhraseNPfam

This Creole is the only one where a post-nominal PL > Def order is solidly attested. Interestingly, however, the two markers are distinguished in older texts only. In more recent samples, (Corpus Creole Ludwig &Telchid 2001), a single marker ya occurs, apparently resulting from the fusion of the plural marker ye with the final definite marker (l)a. Rather that being an instance of head movement, the only one that would occur across FLC, I suggest that the fusion ye la > ya results from a phonological absorption due to the linear proximity of the two markers assembled through phrasal movement. If this is correct, this appears to provide some rein-forcement for the scenario suggested above for Old Louisiana Creole in which the post-nominal marker of plurality yé gets reanalyzed as a definite plural marker un-der some sort of absorption process. In OLC, however, fusion would have resulted in the deletion of la rather than in a phonologically fused determiner ya.

4.7 Haitian Creole

Haitian Creole, the last FLC considered here has all of its makers occurring in post-nominal positions (except of course for the indefinite singular placed as al-ways in Spec NumP and thus pre-nominally on our analysis). For the derivation of a plural expression, I will again assume that the plural marker is a head merged in

Nominal constituents in French lexifier creoles 299

PlP that attracts its complement. With HC again, however, distinct dialects need to be considered, although, at this point, too little is known to be able to reach firm conclusions on their distinctions. What follows then is even more speculative than what precedes, clearly awaiting further empirical information.

In one dialect of HC (the most common case?) the plural marker yo is in strict complementary distribution with the definite marker la. Nominal expressions are either singular and definite with la or plural and definite with yo but la and yo do not co-occur. In this same dialect, however, co-occurrence of the demonstrative sa, with the definite la for a singular or with yo in the plural is common and in fact necessary (Fattier 2000). The demonstrative invariably precedes both la and yo. In HC, perhaps more strongly than in other FLC, la seems strictly associated with singularity. It may be then, that in addition to the Def feature, D0 contains an uninterpretable number feature that must always be valued by the movement of a interpretable number marked nominal projection in Spec DP. That is in HC, it may be more an uninterpretable number feature that motivates movement that an uninterpretable person feature. We leave the consequences of this suggestion for further research.

Given (17) and assuming that the Pl marker yo heads the NumP projection, the order NP Dem> Pl can be captured under a Spec to Spec derivation, with NP moving first to Spec NumP and then subsequently to Spec DemP. For the last step of the derivation, however, although in principle for plural nominal expressions with yo, a further Spec to Spec movement bringing NP to the Spec of a null defi-nite determiner would be possible, such a step is incompatible with the Dem > Def ordering attested regularly in the singular (NP sa (l)a). It thus seems that the last step (i.e. the phrasal movement to the D0 projection) must involve the movement of the whole DemP projection to the Spec of D0, thus involving a Comp to Spec step toping an otherwise Spec to Spec derivation.

300 Viviane Déprez

(41) DefP

D

(la) DemP

Dem

sa PlP

Pl

yo NP

De�nite Phrase

Demonstrative Phrase

Number Phrase

Noun Phraseliv

Confirming this derivation for the last step, the only attested (rare) combination of the three markers together manifests the order NP sa yo (l)a (Fattier 1998:84421) that can be derived as described above if NP first moves to the Spec of PlP headed by yo, then move to Spec DemP headed by sa and finally the whole DemP moves to the Spec of DP headed by (l)a. This derivation, with initial Spec to Spec steps followed by a Comp to Spec finish, may seem oddly unbalanced as it combines two distinct types of locality. But it may, in fact, be precisely this lack of balance that is at the source of the largely complementary distribution of the definite and plural markers. The following is what I have in mind. Note first that the unbal-anced derivation suggests that at some point, a plural DemP must have moved in Comp to Spec fashion to Spec DP. In similarity with what was suggested for O Louisiana Creole above, the resulting co-occurrence of a plural marker pres-ent only in definite plural expressions and of a definite marker may have led to the phonological fusion of the two marker (in similarity with what is observed in Guyana Creole ye la → ya) subsequently followed by the reanalysis of the plural marker as a definite plural determiner, possibly merged then directly in the head of D0. Plausibly, this may be how a plural D0 came to inherit a number feature, which was then generalized to la = singular by opposition. Note that given such re-analysis, i.e. with yo eventually becoming the head of D0, a fully homogeneous Comp to Spec derivation becomes possible, with NumP moving to Spec DemP and DemP subsequently moving to the Spec D0, containing yo. The suggestion

21. Fattier 1998: 844 ‘ En ce qui concerne le pluriel, partout est citée la même forme. [sa yó]. En, 1C, on note un tour plus complexe: [tut cat sa yó a, tut move] “tout ces chats, tous sont méchants” (que l’on peut comparer a [tut sa yo 13D) et également à [ay prå cat sa yó a not. 5]

Nominal constituents in French lexifier creoles 301

here then is that the presence of the final definite marker la has been progressively eliminated through the reanalysis of the plural marker as a definite plural marker. The result (and possibly the motivating force for the reanalysis) may have been to allow a simpler homogeneous Comp to Spec derivation, thereby eliminating the more complex mixed Spec to Spec + Comp to Spec derivation represented in (41) above and resulting in the complementary distribution of la and yo.

In addition to the dialect that manifest a complementary distribution between la and yo there is, apparently mostly in the South of Haiti, another dialect where the co-occurrence of the plural marker with la is fully possible and rather com-mon. This dialect appears at first, to raise some difficulty for the scenario devel-oped above. Indeed unexpectedly from the perspective taken above, la turns out to precede the plural marker yo resulting in the order NP la yo, and it cannot follow it. Yet more surprisingly, concurrently but apparently never at the same time, this dialect also allows the co-occurrence of the demonstrative marker sa with the plu-ral maker yo in the already observed order sa yo. It may seems tempting at first to analyze this dialect as one manifesting a uniform Spec to Spec derivation, with NP moving to Spec NumP and then subsequently either to Spec DemP = NP sa yo or to Spec DP, NP la yo, with this derivation reflecting perhaps an older stage of the language where la and sa where not yet distinguished. As noted in Fattier (1998) indeed, older forms of Haitian Creole did allow the co-occurrence of la and yo in the order NP la yo rather frequently but did not have sa as an independent post-nominal demonstrative. With a full Spec to Spec derivation, the expected order of the three markers together would then be NP la sa yo. What is intriguing, however, is that this order turns out in fact to have never been attested (Fattier 2000)22 either in the current dialectal map or in the historical records. What has been observed instead in the dialect that allow la yo combination, is that the sa yo seems to be in complementary distribution with la yo, so that the three markers never in fact co-occur. The only co-occuring order of the three markers that has ever been record-ed is the above noted NP sa yo a and to my knowledge this order has not ever been attested with speakers of the la yo dialect. Unexpectedly from our point of view, it seems that the Southern la yo dialect also admits the co-occurrence of the demon-strative marker with the definite determiner in the regular order NP sa (l)a. This, at first, appears contradictory with a pure Spec to Spec derivation. Indeed, given (17) the only possible way of deriving an NP sa la order with phrasal movement is under a Comp to Spec analysis. In sum, although this dialect seems at first to ex-emplify an earlier stage of the language with a Spec to Spec derivation (NP to Spec NumP + NP to Spec Dem or Def) this derivation is incompatible with the attested NP sa la order that involves a Comp to Spec derivation, if as has been assumed

22. Further empirical research is surely needed to better establish this point.

302 Viviane Déprez

so far, la in this dialect is always merged in the head of D0. This last assumption, however, may in fact be incorrect. Although further empirical research needs to be done on this Southern Haitian dialect, our current knowledge of the data sug-gests that perhaps, the la that precedes the plural marker is not in fact the definite marker discussed so far but rather a distinct marker that has the same phonologi-cal form but not the same interpretation. The existence of two distinct la markers in Haitian Creole — one stress bearing with a deictic locative force and the other non-stress bearing with a definite interpretation — has been noted several times in the literature, in particular by D’Ans (1968), Sylvain (1936) and Fattier (2000). As noted in d’Ans (1968) in fact, combinations of a nominal constituent with two la, NP la a (d’Ans: p 105: mun la a/ mun la yo) are in fact attested in some cases. This suggest that there may well be two different la, one deictic and the other definite, that can in fact combine together in the order Deictic > Definite in some cases. If this is correct, then the complementary distribution of the combination la yo with the combination sa yo noted above suggests in turn that this second la may in fact be a distinct weaker demonstrative marker that also occupies the head of Dem, in alternance with sa. On this view, the southern la yo dialect would turn out in fact to be largely similar to the main dialect — imposing in effect a complementary dis-tribution between definite la and the plural marker yo — the only difference being that this southern dialect has a weak deictic/demonstrative marker la in addition to the demonstrative marker sa and the regular definite la. If as conjectured here (and also Fattier pc) the la yo dialect is in fact an older form of HC (la yo combi-nations are present in the older text of Ducoeur Joly (Fattier 1998).) then it seems plausible that at this stage in the development of Haitian Creole, the DP and the DemP projections had not yet been distinguished. If so, old HC would in fact par-allel Old Louisiana C in having only two functional projections, DP (+deict +def) = sa/la and NumP = yo with a straightforward Spec to Spec derivation leading the observed NP la yo / NP sa yo. In the more modern dialect, the distinction of the D0 projection from the Dem projection, would have lead to a stage with a mixed derivation Spec to Spec up to DemP and Comp to Spec toward DP, subsequently reanalyzed with yo in D0 as a plural definite heading DP and a uniform Comp to Spec derivation. Evidently, it is only through a thorough investigation of the uses of la in the la yo dialect and in the complementary dialect that such speculations could be confirmed or infirmed. What remains clear nonetheless is that given the proposed structure (17) and the type of derivations explored in this section, the la yo dialect can be accommodated on some but not all possible combinations. What the structure excludes is the possibility to have a Spec to Spec derivation for NP la yo and NP sa yo combinations and a Comp to Spec derivation for the NP sa (l)a combination. Since all these combinations do in fact occur in the same dialect, it must be either as suggested above that NP la yo combination are in fact instances

Nominal constituents in French lexifier creoles 303

of Dem Pl order and not Def PL order or else that the structure (17) is incorrect.23 My hope is that further empirical investigation will provide sufficient information to determine the correctness of these predictions.

5. Summary

This completes our discussion on the structures and derivations of the distinct FLC determiner orderings. To recap, the essential features of the proposed analysis are as follows:

– The proposed nominal architecture (17) matches closely the one used in re-cent generative cross-linguistic works on nominal projections. The Creole data turn out in fact to provide interesting additional empirical support for the existence of independent number and demonstrative projections as proposed in (Ritter 94) and (Bernstein 1997 or Panagiotidis 1999) respectively.

– There is no head movement in FLC DP just as in verbal projections (Deprez-Vinet 1992, DeGraff 1998). Nominal projections, however, manifest intense and systematic phrasal movement.

– Variations among the distinct Creoles center on two distinct factors: 1. The Spec vs. head status of functional elements: A head forces local movement to its Spec A Spec does not 2. XP movements can involve either a Comp to Spec or a Spec to Spec deriva-

tion with the former seemingly favored over the latter. The two clearly in-volve two distinct types of locality. With the first type of derivation, heads attract their immediate complements and whatever they contain. Attrac-tion is as local as can be and pied-piping as plain as can be, i.e. maximal. With the second type of derivation, locality is a bit relaxed since heads can attract the specifiers of their complements, thus penetrating the comple-ment domain. The tendency to favor Comp to Spec derivations may be governed by Economy considerations favoring the most stringent locality type whenever possible. What remains to be further explored, however, is

23. See in fact Deprez (2005) for yet a third possible scenario that preserves the structure (17). This scenario is built on the observation that a dem >def> pl order is in fact regularly found in Gungbe, one of the possible substrate of HC. For the Gungbe order, Deprez (2005) suggest that the demonstrative marker has cliticized/head moved to D0 deriving the dem>def order. The Def> plural order is then the result of a Spec to Spec derivation. It could be that earlier forms of HC respect a similar derivation.

304 Viviane Déprez

why this type of derivation, although more common is not always real-ized. I will leave this question for further research.

– It was suggested that grammaticalization is at the source of the observed dis-tinctions among FLC. That is the proposed view in grammaticalization, has a structuring role with two different effects. First, grammaticalization appears to foster the subdivision of functional projections into a more structured archi-tecture. This effect was taken to be responsible, on the one hand, for the sub-division of two distinct ordered projections DemP and DefP and, on the other hand, for the differentiation of separate numeral and number projections. In either case, these subdivisions can be modeled as resulting from a fission op-eration on the feature bundle of a particular node producing two distinct but related nodes. It may be that, as a result of this fission, a subset of the original features takes on a more abstract character, i.e. or perhaps, a more linguis-tic one, along the line of the changes observed in the grammaticalization of definite determiners where deixis seems to be abstracted from a perceptually related context (pointing demonstration) to discourse anaphora and unique-ness. Similarly, number grammaticalization could be seen as abstracting away from a numeration function to a more linguistically oriented function such as for instance individuation/identification having to do with discourse intro-duction and discourse tracking of individuals that are realizations of a kind (see Déprez 2006).

The second structuring role of grammaticalization was seen as involving a change in the value of features from interpretable to uninterpretable. Such a change affects the role of lexical items in a structure. From a value provid-ing element with only interpretable features, a lexical item (LI) that contains uninterpretable features become a value-seeking element, or more abstractly, a functor. Formally, the change is akin to that of turning an argument into a function and in certain aspects parallels type changing operations that are the common bread and butter of formal semantics. That semantic ‘bleaching’, i.e. one of the processes characterizing grammaticalization, can be understood as involving a change in the semantic type of an element has been previous-ly suggested in Von Fintel (1995). The proposal made here concerning the change from interpretable to uninterpretable feature ties semantic bleaching to syntactic consequences. Since uninterpretable features must be valued, they become potential probes and as such the motor of syntactic displacement. In the FLC considered here, LI that become functional heads containing unin-terpretable features require that their specifiers be filled with constituents that can provide an adequate valuation for these uninterpretable features. They are thus the motor of syntactic displacement and in the particular case of FLC of distinction in the determiner ordering.

Nominal constituents in French lexifier creoles 305

Conclusion

Focusing on the similarity and differences observed in the determiner inventory and distribution of FLC, this paper has proposed what is, to my knowledge, the first comparative model of the syntax of their nominal constituents. It argued that despite the observed diversity, FLC nominal constituents have a common underly-ing functional architecture that distinguishes up to four distinct functional projec-tions, DefP, DemP, PlP and NumP hierachically organized as in (17). Empirical evidence for this structure have been offered and the paper has proposed that the distinction observed in the surface distribution of these functional projections re-sult from syntactic derivations that are governed by a single generalization: speci-fiers of FLC functional heads must be filled. Variation obtains, the paper argued, when determiners in FLC either have a head or a specifier status and derivations involve either a movement from Comp to Spec or from Spec to Spec. The status of determiners as specifiers or heads has been assumed to be determined by Gram-maticalization. Grammaticalization it was proposed involves either the division of a particular functional node into two distinct nodes under feature fission or a change in the interpretability of features from interpretable to uninterpretable, and possibly both processes concurrently. This changes in turn trigger the syntac-tic movements that are responsible for the apparent distributional diversity.

The proposed approach offers a unique analysis of nominal constituents in FLC, tying syntactic diversity to potential semantic distinctions. As such, it pro-vides an elegant and systematic model of an apparent chaotic diversity and offers an exceptional window of the intricate interaction of syntax and interpretation. The paper further proposes a novel approach to grammaticalization within the Minimalist framework, arguing that interpretability changes in the feature set of a lexical element are sufficient to trigger restructuring operations such as syntactic movement and projection fission that have a dramatic effect on the surface struc-ture of nominal constituents. Although the proposed view of grammaticalization is no doubt in need of further empirical substantiation, the perspective provides a sufficiently coherent picture of FLC nominal projections to prove worthy of fur-ther investigation.

Submitted: 01/21/05Accepted: 02/01/06

306 Viviane Déprez

References

Abney, Steven (1987) The English noun phrase in its sentential aspect. Cambrige, MA: MIT dissertation

Adone, Dany (1992) For a DP analysis in Mauritian Creole’Kognitionswissenschaft, Bericht Nr 17. Oktober 92.

Aboh, Enoch (1997) From the syntax of Gungbe to the grammar of Gbe. Geneva; Switzerland: University of Geneva dissertation

Baker, Philip & Chris Corne (1982) Isle de France Creole. Karoma Publishers.Baker, Philip (2002) Quelques cas de reanalyze et de grammaticalisation dans l’évolution du

Creole Mauritien (ms.)Bernstein, Judy (1997) Demonstratives and reinforcers in Romance and Germanic languages.

Lingua 102: 87–113Bollée, Annegret (1977) Le créole français des Seychelles. Tubingen. Niemeyer.Bollée, Annegret (2004) Le développement du démonstratif dans les créoles de l’Océan Indien,,

Creolica 05–17 16:18:23Broussard. James.F (1942) Louisiana Creole Dialect. Louisiana State University Press. Baton

Rouge. Louisiana.Bruyn, A. 1994. “Noun Phrases.” In Pidgins and Creoles: An Introduction, J. Arends, P. Muysken,

N. Smith (eds), 259–269 Amsterdam / Philadelphia: John Benjamins.Bruyn, A. 1995. Grammaticalization in Creoles: The Development of Determiners and Relative

Clauses in Sranan. Ph.D.dissertation. Amsterdam: IFOTT.Cérol, Marie-Josée (1991) Une introduction au Créole Guadeloupéen. Editions Jasor.Cinque, Guillermo (1998) Adverbs and functional heads: A cross-linguistic perspective. Oxford:

Oxford University Press.Cyrille, Odile, Elzire (1997) Aspect of the Syntax of Guadeloupe Creole Nominal Expressions. Sal-

ford, England. Dept of Modern Languages. University of Salford. Master Thesis.D’Ans André-Marcel (1968) Le Creole Français d’Haiti. Mouton. The Hague-Paris Damoiseau,

Robert (1999) Eléments de grammaire comparée Français-Créole Martiniquais. Mouton. Paris

DeGraff, Michel (1997) Verb syntax in creolization (and beyond). In The New Comparative Syn-tax, L. Haegeman (ed), 64–94. London: Longman.

Déprez, Viviane, Vinet Marie-Thérèse (1992) ‘Une structure predicative sans copule’ La Revue Québécoise de Linguistique 22: 1. 11–44.

Déprez, Viviane, Vinet Marie-Thérèse. (1997) Predicative Constructions and Functional Cat-egories in Haitian Creole. Journal of Pidgin and Creole Languages 12.2. 1–32

Déprez, Viviane (1999) De la nature sémantique des nominaux sans déterminant en créole hai-tien Langues 2.4. 289–300.

Déprez, Viviane (1999a) “The Roots of Negative Concord in French and French based creoles” in language creation and language change: Creole, Diachrony and Development. 375–428. MIT Press. Michel DeGraff ed.

Déprez, Viviane (2001) On the nature of Haitian Bare NPs in Current Issues in Linguistics Theo-ry. D.Cresti, C.Tortora, and T.Satterfield eds. 45–62. John Benjamins

Déprez, Viviane (2003a) Constrants on the meaning ofbare Nouns in Current Issues in Romance Linguistics. R. Nunez-Cedeo, L. Lopez, R. Cameron eds. John Benjamins, Amsterdam.

Nominal constituents in French lexifier creoles 307

Déprez, Viviane (2003b) Haitian Creole se: A copula, a pronoun, both or neither. In recent de-velopments in Crole studies. D. Adoneed. Max Niemeyer Verlag. Tubingen. 135–173.

Déprez, Viviane (2005a) Morphological number, semantic number and bare nouns. Lingua.6. 857–883

Déprez, Viviane (2005b) Nominal Constituents in French lexifier Creoles: probing the structur-ing role of grammaticalization. Paper presented at the Creole Conference on Substrate and superstrate Influence. Institute for Evolutionary Anthropology. Leipzig. June 3–5.

Déprez, Viviane (2006) On the conceptual role of number. In New perspectives on Romance linguistics. Nishida Chiyo and Jean-Pierre Y. Montreuil (eds.), 67–83. John Benjamins.

Déprez, Viviane (to appear) Number Marking and Bare arguments. Proceeding of LSRL 2005.Fattier Dominique (1994) Un fragment de créole colonial: Le Manuel des habitans de Saint-

Domingue de S.J Ducoeurjoly, 1802. Creolisation et acquisition des langues. Ed by D. Vero-nique. Université de Provence. France.

Fattier Dominique (1998) Contribution à l’étude de la genèse d’un créole: l’Atlas Linguistique d’Haïti, cartes et commentaires, Aix-en Provence. Université de Provence. Thèse d’Etat.

Fattier Dominique (2000) Genèse de la determination Post-Nominale en Haitien.’ In D. Vero-nique. Ed.

Grimshaw, Jane (1997) Projection, Heads and Optimality. Linguistic Inquiry. 28. 373–422Heycock, Caroline, Zamparelli, Roberto (2000). “Plurality and NP- Coordination”, in: Proceed-

ings of NELS–30, Rutgers UniversityJoseph, Frantz L (1989) La détermination nominale en créole haïtien. Paris: Université de Paris

VII. Thèse de Doctorat de 3ème cycle.Kayne, Richard (1994) The Antisymmetry of Syntax. MIT Press.Lefebvre, C. 1998. Creole Genesis and the Acquisition of Grammar. Cambridge: Cambridge

University Press.Longobardi, Giuseppe (2000). N-movement in Syntax and LF. University of Venice Working Pa-

pers in Linguistics.91.1.9Ludwig, Ralph, Telchid, Sylviane, Bruneau-Ludwig Florence (2001) Corpus Créole HelmutBuske

Verlag, Hamburg.Lyons, Christopher (1999) Definitness. Cambridge University PressNewman, Ingrid (1985) Le Creole de Breaux Bridge, Louisiane: Etude morphosyntaxque, textes,

vocabulaire. Hamburg. Helmut BuskePanagiotidis, Phoevos (2000) Demonstrative determiners and operators. Lingua 110. 717–742.Ritter, Elizabeth (1992) Cross-linguistic evidence for the number phrase. Canadian Journal of

Linguistics 37. 197–218.Rottet, Kevin (1992) Functional Categories and Verb raising in Louisiana Creole. Probus

4.261–289.Rottet, Kevin. (1993) The internal Structure of DP in Mauritian Creole. Indiana University

(ms)Schloorlemmer, Maaike. 1998. ‘Possessors, Articles and Definiteness. Possessors, Predicates and

Movement in the Determiner Phrase. ed. by Artemis Alexiadou and Chris Wilder, 22. 55–86. Amsterdam: John Benjamins

Syad, Anand 1996. The development of a marker of definiteness in Mauritian Creole. Chang-ing Meaning, Changing Functions. ed. by Philip Baker and Anand Syea, 171–186. London: University of Westminster Press.

Sylvain Suzanne, [1936] 1979: Le créole haïtien. Morphologie et syntaxe, Slatkine Reprints, Ge-nève (réimpression de l’édition de Wetteren-Port- au-Prince 1936)

308 Viviane Déprez

Thràinsson, Hoskuldur, 1996. On the (non)universality of functional categories. Minimal Ideas: Syntactic Studies in the Minimalist framework. ed.by Werner Abraham,, 253–281. Amster-dam. John Benjamins.

Von Fintel (1998) ‘The formal Semantics of grammaticalization’ Proceedings NELS 25: Papers from the Workshop on Language Acquisition & Language Change, Volume 2. 175–189 UMass Amherst: GSLA

Zamparelli, Roberto (1995) ‘Layers in Nominal Structure’. Rochester: Ph.d. dissertation


Recommended