Attributive possession in 19 th century Mauritian Creole

The paper examines the attributive possessive structures attested in 19 th century Mauritian Creole. It is first shown that these include a construction with the preposition pou ‘for’, which has not been mentioned in the literature. The situation in Mauritian Creole is then compared to that in Seychelles Creole, a closely related variety. Also discussed are some of the more general implications of the findings.


Introduction
As is well known, possession is formally encoded in two main types of constructions: predicative possession 1 and attributive possession 2 (cf.Heine 1997a: 86;Herslund/Baron 2001: 3;Stassen 2009: 26).These types differ in several respects.In predicative possession structures the relation of possession is expressed by a verb, being therefore explicitly asserted; their typical interpretation is that of ownership or temporary possession, although they may also be polysemous; they have clausal syntax (cf. Heine 1997a: 86;Herslund/Baron 2001: 3;Stassen 2009: 26-27).As for attributive possession structures, the relation of possession is expressed by a nominal and it is presupposed; they are polysemous to a far greater extent; 3 they have phrasal syntax (Seiler 1983: 199, Heine 1997a: 86, Herslund/Baron 2001: 1 and 12, Seiler 2001: 33, Stassen 200926-27, Nichols/Bickel 2011).Given that both their formal encoding and their semantics differ considerably, Stassen (2009: 28) rightly concludes that "predicative and attributive possession are probably to be considered as belonging to two different (or at least partially different) typologies".
Following Heine (cf. 1997aHeine (cf. , 1997bHeine (cf. , 2001) ) it is useful to distinguish two subtypes of predicative possession structures: the have-construction and the belong-construction.The haveconstruction is defined as "that construction which is used canonically to express 'I have a car' or 'we have no money'"; the possessee is typically indefinite; there is "emphasis on the possessor" (Heine 1997a: 86-87).The belong-construction is typically exemplified by a sentence such as 'The car belongs to me'; the possessee is the clausal subject and/or topic and it is typically definite; the possessor is the complement or an oblique constituent (cf. Heine 1997a: 86-87).
have-construction Ron has a dog.

b. belong-construction
The dog is Ron's.
(2) Attributive possession Ron's dog Following Heine (1997aHeine ( , 1997bHeine ( , 2001)), the theoretical framework adopted in the present paper, the various means of expressing the rather abstract relation of possession are derived from concrete domains.Their "stereotypical descriptions" are called "event schemas" (Heine 1997a(Heine : 91, 1997b: 142): 142).The eight basic event schemas identified (see Heine 1997aHeine : 91-97, 1997b: 142-148: 142-148) are believed to account for the various possessive structures attested crosslinguistically, i. e. they develop into various possessive constructions.Each of the event schemas is associated with a particular set of morphosyntactic correlates.In what follows the presentation of the event schemas is limited to those relevant for the aims of the present paper:4 the Goal Schema and the Topic Schema.
In the Goal Schema the possessor is conceptualized as goal, encoded as an adjunct introduced by means of the allative/directional or of the benefactive/dative case marker, while the possessed NP is the subject.This schema develops into both belong-constructions and attributive possession structures.The morphosyntactic correlates of the Topic Schema are as follows: the possessor NP is the topic and the possessed NP the subject.According to Heine (1997a: 96) the Topic Schema "as a source for attributive possession provides one of the most common templates".
Several studies have examined the syntactic structures employed for expressing attributive possession in pidgins and creoles.Heine and Kuteva (2001), for instance, analyze the attributive possession structures found in over 40 pidgin and creole languages, with various lexifiers.Vintilă-Rădulescu (2003) discusses attributive possession constructions as well as the lexicalization of the meaning of 'to have' in French-lexifier creoles.Attributive possession patterns are also illustrated in the volume edited by Holm and Patrick (2007) on the comparative syntax of 18 creoles.
The present paper is concerned with the attributive possession structures in 19 th century Mauritian Creole, although reference will occasionally be made to other types of possessive constructions as well.Following Holm (cf. 1988Holm (cf. : 199, 2000: 220): 220), a distinction is made throughout the paper between possessive structures consisting of two nominals (full NPs) or of a nominal (full NP) and a possessive determiner (possessive adjective), on the one hand, and pronominal possession, i. e. involving possessive pronouns, on the other.The textual evidence examined is illustrative of an already stable variety.The pidginized form of French which had emerged after the French took possession of Mauritius in 1721 had already been creolized: on the basis of demographic and linguistic evidence, Baker and Corne (1986: 169) conclude that "it seems entirely clear that a stable Creole language […] existed in Mauritius by not later than the 1770s".5 The paper is organized as follows.In 2 I present the types of attributive possessive structure in 19 th century Mauritian Creole identified by Syea (1994Syea ( , 1995Syea ( , 2007Syea ( , 2013) ) and Guillemin (2007Guillemin ( , 2009Guillemin ( , 2011a)).Also discussed are a number of issues related to previous work on attributive possession in Mauritian Creole.In section 3 I show that there is a fourth type of attributive possessive structure in Mauritian Creole, attested in 19 th century texts.Section 4 is concerned with a comparison of the attributive possessive constructions in Mauritian Creole and those recorded in the closely related Seychelles Creole.The findings and some of their implications are discussed in section 6.
The examples appear in the orthography or system of transcription used in the sources mentioned, and include glosses6 and translations; early Mauritian Creole examples also include the year of the attestation if different from the date of publication.The length of quotations has been kept to a reasonable minimum.
The last type of attributive possessive structure is called "mixed" by Syea (1994: 86), "possinitial" in Syea (1995: 181) or "possessive pronoun-N-Possessor type" in Syea (2013: 2).In work on the comparative syntax of creole languages this type is known as "[his + possessed + possessor]" (see Holm/Patrick 2009: 349) Listed below are Syea's (1994Syea's ( , 1995Syea's ( , 2007) )  This is a possessor climbing construction, in which the possessive determiner so/son is not adjacent to the possessor NP, but to the possessed NP.Syea (1995: 182) writes that texts pre-dating 1855 (Lolliot 1855) contain no examples of the [possessor + his + possessed] and [his + possessed + possessor] structures.According to Syea (1995: 182), the [possessed + possessor] pattern is "the first and only means of expressing [attributive] possession for well over a century".The [possessor + his + possessed] construction is first recorded in 1855.As for the [his + possessed + possessor] structure, it is attested for the first time in a text dating from 1867.However, according to Syea (1995Syea ( : 187, 2007)), this type might have emerged before the [possessor + his + possessed] structure.As acknowledged by Syea (1995: 187), this "sequence of developments […] appears at first sight to run counter to the textual evidence".Nonetheless, according to Syea (1995: 187) several arguments can be adduced in defence of this scenario, as follows.The time gap between the first attestations of the two constructions is of only 12 years.The [possessor + his + possessed] structure is found in 1855 in a single example (Lolliot 1855: 14) (Descroizilles 1867: 2, 4, 7, 11).Finally, the examples of [his + possessed + possessor] outnumber the instances of [possessor + his + possessed] in the texts collected by Baissac (1888).
The descriptions of the attributive possession patterns in 19 th century Mauritian by Syea (1994Syea ( , 1995Syea ( , 2007Syea ( , 2013) ) and Guillemin (2007Guillemin ( , 2011a) ) call for a number of remarks.Consider first the [possessed + possessor] pattern.The meaning of (3a) -from Syea (1995Syea ( : 181, 2013: 25) : 25) -is not that of literal possession.For ease of exposition, example (3a) is repeated below as ( 10 In all these examples -the earliest on currently available evidence -the second NP is not a possessor, but rather it qualifies the first one, by specifying its kind.In other words, the second NP is a modifier.Similar cases occur abundantly in modern Mauritian Creole as well.As shown by Corne (1999: 171), in addition to "that of possession (the genitive)", structures consisting of a head noun and an NP complement "cover a wide variety of relations between the head noun and the NP complement", such as: ( 12 As can be seen, the constructions ( 10) and ( 11) parallel the structure in (13a), and semantically they express quality or provenience, to use Corne's (1999: 171) terms.They are therefore examples of possession understood lato sensu, i. e. as typically extending to various kinds of attributive modification of one nominal by another.This is in line with the use of the term "possession" by e. g.Nichols and Bickel's (2011) Syea (1994Syea ( , 1995Syea ( , 2007) ) quotes from the second edition in 1831.
With reference to the [possessor + his + possessed] pattern, Guillemin (2007: 290, f. n. 1) writes that "I have come across only one example of [such] genitives in the early texts (19th century)", and quotes the example in ( 6).This claim is disconfirmed by the earlier attestation in 1855, see example (4a).Some of Syea's (1995Syea's ( , 2007Syea's ( and 2013: pages) : pages) claims regarding the [possessor + his + possessor] and the [his + possessed + possessor] structures also need to be amended.The [possessor + his + possessed] pattern does not occur only once in Descroizilles (1867), as stated by Syea (1995: 187), but twice.Here is the second example: More importantly, the [possessor + his + possessed] construction does not occur at all in any of Baissac's (1880Baissac's ( , 1888) ) texts.Reproduced below is the only such alleged example, as glossed and translated by Syea (1995Syea ( : 181, 2013 As can be seen, Zean so manze is not an instance of [possessor + his + possessed], but of a double object construction.Baissac's translation constitutes further support for this analysis: example (17a) is translated 'Jeanne qui porte à Jean son déjeuner' (Baissac 1888: 78), while the French version of (17b) is 'Jeanne arrive pour apporter à Jean son déjeuner' (Baissac 1888: 80).The [his + possessed + possessor] pattern is found, even if only once, in an 1850 text: In spite of the comma separating son malice 'the wickedness' from sa satte la 'that cat'.12This is the first occurrence in Mauritian Creole of the [his + possessed + possessor] construction.This example precedes by some 17 years the one in (3a), dating from 1867, considered by Syea (1995Syea ( : 187, 2007: 17) : 17) and Guillemin (2009Guillemin ( : 133, 316, 2011a: 273) : 273) to be the earliest attestation in Mauritian Creole of this attributive possessive structure.Note, however, that the occurrence of this pattern in 1850 confirms Syea's (1995: 187) hypothesis that it chronologically preceded the [possessor + his + possessed] construction, first attested in an 1855 text (see example (4a) in section 2). 13Further support is provided by the fact that this pattern is better attested in Descroizilles (1867).According to Syea (1995: 187), "the 1867 text -has four examples" illustrative of the [his + possessor + structure], but actually there are 11 such cases in Descrozilles (1867).There are several examples attesting to the use of this structure.Seven such instances are found in three of the texts also published in Chaudenson (1981): Annexe et note de renvoi (Chaudenson 1981: 80-82), Le catéchisme de 1828 (Chaudenson 1981: 107-114), and Proclamation pour noirs esclaves dans Maurice (Chaudenson 1981: 118-120).One more example is recorded in Chrestien (1831: 46) (1997a: 95, 2001: 316) calls "Goal Schema".As shown by Heine andKuteva (2001, 2002), the languages in which this schema develops into attributive possession structures include creoles and pidgins.For instance, a number of English-lexifier creoles 18 use reflexes of the English benefactive preposition for.In French-lexifier creoles constructions with a reflex of pour 'for' are far less frequent.However, there are French-lexifier creoles which use it in structures expressing attributive possession.One such variety is Tayo, spoken in New Caledonia.Consider the following examples from the only book-length description of this creole (Ehrhart 1993): (25) a. meso pou per (Ehrhart 1993: 149) house According to Chaudenson (1981: 228), the uses illustrated in ( 28) and ( 29) are "the origin of the formation of possessive pronouns in Mauritian".Indeed, the "long" variants, e. g. ça qui pour vous 'yours' in (30b), are actually belong-constructions.
According to Chaudenson (1981: 228), the use of pou in such structures is due to the fact that the French preposition à is not preserved in Mauritian Creole. 21As is well known, the uses of the preposition à in French include various possessive constructions.These include attributive possession structures with nominals (full NPs) (31a) and with pronominals (31b), possessor doubling constructions with nominals (full NPs) (31c) and with pronominals (31d), belongconstructions (31e), and wh-possessors (31f): (  Baissac (1880: 75) writes that à only exists in a few what he calls "compositions adverbiales", such as "àcote, where; àçthére, now".Similarly, Guillemin (2009: 129-130) states that in early Mauritian Creole "the preposition à is found only in fixed expressions, e. g. jusqu' à 'until', à bord 'on board'" and that such expressions "become one morpheme" -ziska and abor respectively -in the modern variety.In fact, there is at least one occurrence of the preposition à in an early 19 th century source.Significantly, à is recorded in an attributive possession construction: (32) moi venir voir femme à moi [1802] (Milbert 1812: 271) 1.SG come see woman to 1.SG 'I have come to see my wife' Since the French-derived preposition à is not preserved in Mauritian Creole, pou takes over the uses in French of the preposition à in possessive constructions.Circumstantial evidence is provided by the fact that in some French-lexifier creoles a reflex of this preposition functions as a marker of attributive possession, both with nominals (full NPs) and with pronominals.One such variety is Guadeloupe Creole: (33) a. timun -a-Pyè (Corne 1999: 131) child to Pyè 'Pyè's child' b. timun -a-mweṅ (ibd.)child to 1.SG 'my child' Consider next the situation in modern Mauritian Creole.According to Guillemin (2009: 130), a is found in "relatively modern MC [= Mauritian Creole] texts", particularly in written Creole, which is presumably an effect of decreolization. 22Crucially, even in such texts the preposition à "is never used in possessive constructions" (Guillemin 2009: 130) for 2.SG 'yours' Note, here again' that the "long" variant sa ki pu twa, literally "that which is yours", is actually a belong-construction, which, as mentioned above, is believed to be the source from which possessive pronouns were formed in Mauritian Creole.The analytic type is first attested relatively early.This is to be expected since, as put by Guillemin (2009: 131), these "genitive constructions without a preposition have always been grammatical in MC [= Mauritian Creole]". 25Rather surprisingly, perhaps, the hitherto unnoticed attributive possession construction with for is first attested equally early.Until about the half of the 19 th century these two structures appear to have been the only means available in Mauritian Creole for expressing attributive possession.Two patterns, the poss-initial and the one with for, are no longer in use: while the former survived into the early decades of the 20 th century the latter fell out of use towards the end of the 19 th century.Finally, the synthetic type -although the last one to emerge -not only still survives, but, on some accounts, appears to 24 Keeping in mind Baker's (1997: 93) caveat that "old text are not an accurate guide to when features first appeared in the spoken language" because, obviously, "features must generally have become established in the spoken language before their first appearance in a text". 25The earliest record dates from 1734, an "example of what appears to be pidginized French" (Baker/Corne 1986: 69).No attributive possession constructions are attested in any of the eight 18 th century texts (see Chaudenson 1981: 77-78, Textes créoles anciens n. d.).
be the most widespread attributive possession structure in the modern variety of Mauritian Creole.

Mauritian Creole and Seychelles Creole
Consider next the situation in Seychelles Creole (cf.Bollée 1977, Corne 1977, Chapuis 2007: 349, Michaelis/Rosalie 2013).This is a variety closely related to Mauritian Creole, of which it is an offshoot, having evolved independently starting with the year 1770.
The poss-initial type, i. e. [his + possessed + possessor] also occurs, in which "the possessive determiner is not adjacent to the relevant noun, but "climbs" up the first noun in the nominal phrase" ( Unlike Mauritian Creole, the type with for, i. e. [possessed + for + possessor], is not attested in Seychelles Creole.It cannot occur with two nominals (full NPs).As for structures consisting of a nominal (full NP) and a possessive determiner, the preposition pur can be used "for emphasis" (Corne 1997: 49), i. e. "to emphasize possession" (Chapuis 2007: 350).In such cases, the possessed NP is preceded by a possessive determiner and followed by pur and a pronominal.Consider the example below: (39) eski u n uar u ser pur u? (Corne 1977 This usage is a reflex of the French possessor doubling construction, illustrated in (31d).Note that unless the possessive determiner and the possessive pronoun co-occur, the sentence below, as an alternative to (39), is ungrammatical: *eski u n uar ser pur u? QP 2.SG COMPL see sister for POSS 2.SG 'did you see your sister?'As in Mauritian Creole, the preposition pur 'for' is found in several other possessive constructions.For instance, "pronominal possession is marked by pur + NP" (Corne 1977: 49).Bollée (1977: 53) writes that "(sa ki) pu mwa, pu u, etc." correspond to French 'le mien, le tien, etc.' In other words, pur occurs in the forms of possessive pronouns (see also Chapuis 2007: 350) The existence both in Mauritian Creole and in Seychelles Creole of the analytic type is not surprising.It represents a case of what Heine and Kuteva (2001: 15) call "retention".As put by Guillemin (2009: 310) about Mauritian Creole, this type "is derived from French, minus the preposition de".The French preposition de has not survived in the two creoles at issue,27 which do, however, preserve the order in which the possessed NP and the possessor NP occur in the lexifier language.
The fact that the synthetic type is also found both in Mauritian Creole and in Seychelles Creole is not surprising either.This attributive possession structure instantiates what Heine (1997a: 96, Heine 2001: 317) and Heine and Kuteva (2001: 16-17) call "Topic Schema".This accords with similar views expressed in the literature on Mauritian Creole and respectively Seychelles Creole.Guillemin (2009: 134), for instance, writes that "the modern MC [= Mauritian Creole] genitive with a possessive adjective is a Topic-Possessive structure".As for Seychelles Creole, Corne (1986: 167) considers this structure a form of topicalization.Heine and Kuteva (2001) show that this schema is implemented in a number of creoles, with various lexifier languages and spoken in various geographical areas. 28It is therefore conceivable that the synthetic type of attributive possession structure may have emerged independently in Mauritian Creole and in Seychelles Creole.Also, both creoles have French as their lexifier, which makes frequent use of topic structures in colloquial speech (see e. g.Chaudenson 1990Chaudenson , 2003)) Chaudenson (1990Chaudenson ( , 2003)) On the other hand, as already mentioned, the synthetic type is quite common, possibly the most frequent, in Mauritian Creole, whereas it is rare in Seychelles Creole.This significant difference in the frequency of occurrence of the synthetic type can only be explained with reference to an important difference between Mauritius and the Seychelles in terms of the socio-linguistic situation.Beginning with 1835, large numbers of indentured labourers and their families arrived in Mauritius (Holm 1989: 398, Syea 1995: 183), whereas in Seychelles there was no importation of indentured workers from India (Holm 1989: 402, Syea 1995: 1985).As a consequence, while at present Mauritians of Indian descent make up "two thirds of the island's population" (Holm 1989: 398), in the Seychelles "the population includes few Indians" (Holm 1989: 402).The languages spoken by the Indian immigrants in Mauritius included both Indo-European (Indic) languages -Bhojpuri, Gujarati, Hindi and Marathi -and Dravidian languages -Tamil and Telugu. 29Regardless of their genetic affiliation, all these languages are typologically head-final languages and have attributive possession structures in which the possessor precedes the possessed (a typical syntactic correlate of the OV word order).These attributive possession constructions are superficially similar to the synthetic type attested in Mauritian, from which, however, they also differ in several respects (Syea 1995(Syea : 183-184, 2013: 23): 23) Syea (2013: 38) suggests that in spite of differences such as the ones illustrated above, the synthetic type "may have been preferred […] by the Indian immigrants, given its superficial similarities to the genitives in their native languages".These similarities have led Corne (1986: 162) to the conclusion that the synthetic type of Mauritian Creole has its origins in the languages of the Indian immigrants, in particular Bhojpuri and Hindi.However, several facts speak against such a scenario.The Topic Schema, as already mentioned, develops into attributive possession constructions in other creoles, which cannot be traced back to any substrate or adstrate influences.Also, as shown by Syea (1995: 184), given that the synthetic type is first attested in Mauritian Creole in 1855, it is rather implausible to assume that some 20 years only after the beginning of their immigration to Mauritius speakers of Indian languages would have already impacted on this particular aspect of the syntax of Mauritian Creole.Moreover, this would not explain why, on currently available textual evidence, there are only a total of three occurrences of this type of attributive possession construction in 19 th century Mauritian Creole.On the contrary, it seems more reasonable to assume that a possible influence of the adstrate languages spoken by the Indian immigrants may have made itself felt only later, long after the emergence of the synthetic type.This tallies with the significant increase, beginning with the 20 th century, in its frequency of occurrence, to the point of possibly being the most widespread type of attributive possession construction in modern Mauritian Creole.To sum up, the subsequent influence of the adstrate languages spoken by Indian immigrants in Mauritius provides a plausible explanation for the difference in the frequency of occurrence of the synthetic type in modern Mauritian Creole as opposed to modern Seychelles Creole, for which no similar adstrate influence can be invoked. 30onsider next the rather different situation in the case of the poss-initial type.This is an instantiation of what Heine (1997b: 148) and Heine and Kuteva (2001: 18-19) call "anti-Topic Schema".This conceptual schema is only rarely a source for attributive possessive structures.More recently, Syea (2013: 37) derives the poss-initial type in 19 th century Mauritian Creole from "such [French] possessor doubling possessive constructions as son père à lui (lit.his father to him)".Whatever the right account may turn out to be, the fact remains that of the pidgin and creole languages examined by Heine and Kuteva (2001), the poss-initial is attested in only two: Mauritian Creole and Seychelles Creole.Since this type is rare, it is unlikely that its occurrence is the outcome of independent developments in the two creoles.Rather, it is more plausible to assume that it must have existed in Mauritian Creole prior to 1770, the date after which this variety and Seychelles Creole evolve separately.This could account for the attestation of this type of attributive possessive structure in both creoles.While Mauritian Creole has lost the poss-initial attributive possessive construction, this continues to be used, although infrequently, in the modern variety of Seychelles Creole.
Finally, the attributive possessive construction with pou 'for' appears to be an independent evolution in 19 th century Mauritian Creole.

Conclusions
19 th century Mauritian Creole has been shown to have four types of attributive possession structures.As shown by Heine and Kuteva (2001: 24), "that a given language should have more than one type [of] attributive possession does not seem exceptional, however, in creoles this […] is the rule".On their view, to the attributive possession construction preserved from the lexifier language others are added in the process of creolization, by exploiting the various conceptual schemas which may lead to the grammaticalization of attributive possession.19 th century Mauritian Creole is illustrative of this scenario.To the analytic type, retained from French, three other are added: the synthetic, the poss-initial and the one with for respectively.
Variation in different historical periods in the frequency of the patterns expressing attributive possession in Mauritian Creole also reflects the contribution of the languages in the adstratal input.Even if the equivalent attributive possession structures in the adstrate languages of the Indian immigrants are not identical to the synthetic type of Mauritian Creole, the former appear to have contributed to reinforcing its use in the latter. 31  The existence of several attributive possession structures is a further confirmation of the claim that creoles "germinate several options" (Aitchison 1996: 147), even in the same historical period, some of which are preserved, whereas others end up being lost.A similar point is made by Syea (2013: 38), who writes that "old and new structures compete against each other until one gives way".As put, again, by Syea (2013: 38), "this kind of competition between old and new structures is of course not exclusive to the development of creole languages".In the case of Mauritian Creole, the only type of attributive possession structure attested throughout its history is the analytic one, which retains the order of the possessed NP and of the possessor NP in French, its lexifier language.On the other hand, two of the types of attributive possessive construction recorded in the 19 th century are no longer is use today: the poss-initial and the one with pou respectively.One question which might be asked is: why did these two particular types lose out to their competitors, i. e. the analytic and the synthetic type respectively?.As in other cases of language change leading to the loss of previously attested structures, it is difficult to answer such a question.An admittedly speculative account might run as follows.In three of the types of attributive possession construction, i. e. the analytic, with for and the poss-initial one, the order is identical, with the possessed NP preceding the possessor NP: [possessed + possessor], [possessed + for + possessor] and [his + possessed + possessor].In other words, there was a competition between three types with the same ordering of the possessed and of the possessor NPs.Of these types only the analytic one has survived, presumably because, unlike the other two, it is retained from French "minus the preposition de", to use Guillemin's (2009: 132) formulation, being therefore the longest established type, and also the most frequent one, hence unmarked. 32The type with for is not an instance of retention, it replaces with pou the preposition à in the original French construction, which is less common than its counterpart with de.The poss-initial type is possibly modelled on a less frequent French possessor doubling construction with à.According to Syea (2013: 37), possessor doubling structures "would have constituted an ideal way of expressing possession in a contact situation", given the fact that "their use has a functional motivation -they provide emphasis through doubling the possessor".In principle, then, the poss-initial type modelled on such constructions would have also been ideal in a contact situation.However, if or once there is no more need to "provide emphasis", such constructions would have used up their utility.Also, given the fate of the preposition à in Mauritian Creole, the possessor has never been overtly marked twice in the poss-initial type.Consider, finally, the synthetic type, which has survived and which is reported to be quite frequent.Possibly also influenced by similar topic structures in colloquial French, this is the only one with the reverse order, i. e. the possessor NP precedes the possessed NP.In this respect it has not faced competition from any of the other three types of attributive possession construction.It has also had the additional advantage of being reinforced by the similar attributive possession structure attested in the adstrate languages spoken by Mauritians of Indian descent.
The findings also confirm the importance of the historical study of pidgin and creole languages.This "provides […] evidence which attests to the existence, previously not suspected, of certain features in the older stages of pidgins and creoles" (Avram 2005: 245).A case in point is the occurrence of the attributive possessive construction with pou in Mauritian Creole, between 1805 and 1885.
Also, the claim put forth by, among others, Mufwene (1991: 29), that "creole structures have barely changed since their formation" is disconfirmed.On the contrary, as stated by Baker (1995: 6), "wherever early Creole texts exist, they show that the languages in question subsequently changed a great deal".The loss of both the poss-initial type and of the one with pou confirms the fact that creole languages can undergo significant changes in their evolution.Crucially, such changes are not necessarily the result of decreolization.
Last but not least, this analysis of the attributive possession structures in 19 th century Mauritian Creole can be taken as being an argument in support of the so-called "gradualist" views of creolization.Although there are several versions, 33 known under various names, such as the "Gradualist Model" of creolization (Carden/Stewart 1988, Arends 1993) or the "Gradual Creolization Hypothesis" (McWhorter 1992), they all share the basic claim that the process of creolization is gradual and it extends over a number of generations of speakers.Baker (1997) has already shown, on the basis of the first attestations of 42 features of Mauritian Creole, that this language has developed slowly, for about 150 years: some of these features are only recorded after 1888.The development of attributive possession structures in Mauritian Creole appears to have been equally slow.Indeed, currently available data suggest that it is also only after 1888 that the two types found in the modern variety of Mauritian Creole have become definitively entrenched.
a slave of war.'This is also the case of all other similar examples in this 1805 text, such as the one below: of that cat'

3
The [possessed + for + possessor] structure 19 th century Mauritian Creole evinces yet another attributive possession pattern, which appears to have gone unnoticed in the literature, 14 which employs pou, the Mauitian Creole reflex of the French preposition pour 'for': [possessed + for + possessor].
Summing up, 19 th century Mauritian Creole employs pou in five possessive constructions.The chronology, based on textual evidence, of the possessive constructions with pou attested in 19 th century Mauritian Creole is set out in Table2

Table 1 : Attributive possessive structures in 19 th century Mauritian Creole identified in previous work
"to refer to all kinds of adnominal constructions regardless of whether the semantics is literal possession".Examples illustrative of literal possession are first attested in 1818:Example (3b) does not date from 1831, it is actually from 1822.The example at issue is already found in Chrestien (1822), 11 whereas Baissac (1888)is taken out of context.The string Zean so manze is found twice inBaissac (1888).Consider the identical sentences in which it occurs: 11 See Textes créoles anciens (n.d.).For example (3b) see the text inChaudenson (1981: 94).
. One last example comes from Anderson (1885: chapter V).The examples are listed below in chronological order.The first attestation of the [possessed + for + possessor] pattern is in 1805: The existence of this type of attributive possessive construction instantiates what Heine In 19 th century Mauritian Creole pou also occurs in other possessive constructions.As mentioned byChaudenson (1981: 228), "pour is especially used in early Mauritian to mark possession", in the belong-construction:In addition, pou is the mark of pronominal possession.As already noted byBaissac (1880:  19), "the Creole develops le mien, le tien, into ça qui pour moi, ça qui pour toi, or, shorter, pour moi, pour toi".Consider the examples below: . Consequently, unlike in French, the following possessive structure, for instance, is ungrammatical:

Table 2 : Possessive constructions with pou in 19 th century Mauritian Creole Note
that pou is first attested in the attributive possession construction with nominals (full NPs).The use of the preposition pou in 19 th century Mauritian Creole in attributive possession structures is not surprising.As already mentioned, such constructions are widespread in many languages.Furthermore, reflexes of the French preposition pour are found in attributive possession structures in other French-lexifier creoles.The various uses of pu in 19 th century Mauritian Creole possessive structures confirmHeine's (1997a: 95)claim that the Goal Schema is a common source for both attributive possession and belong-constructions.

Table 3 : Types of attributive possession constructions
What is striking is the fact that the now extinct type with for is more frequently attested in 19 th century Mauritian Creole than the synthetic one.As shown in this section, there are nine occurrences of the [possessed + for + possessor] pattern with nominals (full NPs), and four more with pronominals; these are recorded in seven sources.On the other hand, as seen in the preceding section, there are only three attestations of the [possessor + his + possessed] construction, in just two sources.The distribution in specific historical periods of the attributive possessive structures of Mauritian Creole, strictly on the basis of their attestations, 24 is set out below:
, optionally preceded, as in Mauritian Creole, by sa ki:The preposition pur also occurs together with the question word ki 'who' to form the whpossessor pur ki 'whose': SG dog for QW 2.SG lit.'you are whose dog' [= 'who the hell do you think you are?'] , spoken French exhibits even recursive topic structures.Consider the example below: . For instance, in Hindi the possessive morpheme shows number and gen-Moreover, the Hindi possessive marker agrees with the possessed NP, i. e. with the NP to its right, whereas in Mauritian so(n) agrees with the possessor NP, i. e. with the NP to its left.On the other hand, unlike Mauritian Creole, the possessive marker in Bhojpuri exhibits no agree-