11.1 Introduction: On the Universality of Admixture

This excerpt from Derek Bickerton’s book Bastard Tongues: A Trailblazing Linguist Finds Clues to Our Common Humanity in the World’s Lowliest Languages contrasts the purity and clarity of Creole languages to older languages, these elders which, during their evolution, developed rules full of exceptions and vocabularies whose etymologies are at times justified, at times dubious. On the one hand, this quote pays tribute to Creole languages as symbolizing a victory of the human spirit, resilient and unbroken in the face of slavery. On the other hand, the very title of the book from which this excerpt is drawn shifts the spotlight to the low status of Creole languages, victims of preconceived notions inherited from the colonial times in which they emerged and affected to this day by neocolonial distorted narratives as being the world’s simplest languages.Footnote ¹ Such is the fate of Creole languages, beleaguered languages that are at times reviled by both ignorant outsiders and misguided native speakers and at times venerated by their users and by the linguists who can see the challenging puzzles they offer to their analyses and the deep complexity of their grammatical systems.

They are a shining example of what humanity can offer when faced with strife, grief, and unrecoverable loss. Humans may be abducted away from their mothers, fathers, brothers, sisters, and other kin, they may be stripped of all their social ties, displaced and thrown into new environments where they encounter strangers from different lands and speaking different languages, subjugated to the same fate. And yet, these individuals still manage to preserve their native languages (at first), become familiar with the new languages in their environment and, ironically, contribute to the creation of a new full-fledged language (the Creole) for the purpose of broader communication. If historians (Andrade Reference Andrade1996) are correct in assuming that the original slaves were purposefully taken from different ethnic groups speaking different languages to prevent communication, plotting and mass insurrections on the plantations, it is indeed ironic that Creoles became the enslaved individuals’ best weapon to thwart and counter such dark calculations. It is also a testimony to the resiliency of the human spirit that, out of such chaos, can emerge such orderly and efficient linguistic systems, full-fledged languages that thrive for generations and become sources of pride and markers of identity for their speakers.

That Bickerton calls them Bastard Tongues in the title of his book is not surprising, as such labels referring to admixture have been used from times immemorial to refer to Creole languages. The Greeks had no specific term for the term “Creole language,” but Herodotus is said to have used a variety of expressions to describe them. For instance, he might say that a Creole was “between” Greek and Scythian, or a “mixture” of Greek and Scythian (Sara Forsdyke, personal communication). Herodotus also uses the term soloikizõ (whence solecism) which literally means ‘to speak like the inhabitants of Soloi,’ Soloi being a Greek city in Pamphylia in Asia Minor that was apparently influenced by its Carian linguistic surroundings (Edward Nolan, personal communication). The concept of Creole conjures up the notions of mixture and hybridity due to the mixing of the languages in contact that contributed to their genesis. Traditionally, languages that are perceived as having mixed properties from two or more languages can be the target of demeaning qualifiers. Spanglish (a mix of Spanish and English), for instance, has been reported as “the bastard jargon,” “una lengua bastarda: illegitimate, even wrongful,” “español Bastardo,” “inglañol,” “el habla de los bárbaros” (Stavans Reference Stavans2003). And yet, all languages necessarily come in contact and influence one another at some level, given that no language ever evolves in a vacuum (Thomason Reference Thomason2001). The range of changes they may undergo spans from mere lexical borrowing (Yiddish shlep was borrowed into the English lexicon) to dramatic grammatical restructuring. For instance, the Viking invasion of England in the late eighth century led Old Norse to leave its imprint on the English language. Similarly, one of the outcomes of the Battle of Hastings in 1066 leading to the Norman conquest of England is that the English aristocracy became fluent in French and due to long-term contact, 45 percent to 70 percent of the vocabulary of the English language is said to originate from French.Footnote ² Many other languages (dead and alive) have left their lexical imprint on English whether directly (due to direct contact) or indirectly (due to inheritance): Latin, Greek, Dutch, Italian, Spanish, Hebrew, Yiddish, Persian, Hindi to name just a few. And yet, the layperson would not think of the English language as a bastardized jargon or as being a hybrid of sorts. Powerful linguistic and cultural ideologies meshed with robust hierarchical structures can lead to the dissolution of such perceptions before they can even take shape. There are no “pure” languages, only speakers who come in contact with others and influence each other’s language.

The objective of this chapter is twofold: First, we examine the notion of converging voices, converging identities and converging languages (Baptista Reference Baptista, Lang, Holm, Rougé and Soares2006, Reference Baptista2009, Reference Baptista2020; Corne Reference Corne1999; Faraclas et al. Reference Faraclas2014) to convey the dynamic interactions underlying language contact. Second, we connect the notion of “multiple voicing” (Faraclas et al. Reference Faraclas2014) to Weinreich’s (Reference Weinreich1953) notion of interlingual identification. We argue that interlingual identification is ground zero for language mixing and language change. We propose that the same mechanism that contributes to Creole formation is also present in other contact settings such as bilingual and other language learning situations. If all languages are in contact, this means that the mechanism of interlingual identification is pervasive in all contact settings,Footnote ³ and this, in turn, allows us to expand and reframe Bickerton’s original notion of bastard tongues, while arguing that admixture is one of the major clues to our common humanity (altering Bickerton’s [Reference Bickerton2008] original book title: Bastard Tongues: A Trailblazing Linguist Finds Clues to Our Common Humanity in the World’s Lowliest Languages). We propose that all languages are mixed (see Aboh Reference Aboh2020; Mufwene Reference Mufwene2023), that such admixture is inherent to all human linguistic interactions, and hence mixing is only a matter of degree.

This chapter is organized as follows: In section 11.2 below, we briefly illustrate how “multiple voicing” (Faraclas Reference Faraclas2012; Faraclas et al. Reference Faraclas2014) operates in the linguistic ecology in which Creoles emerge. In section 11.3, we connect the notion of multiple voicing to the notion of interlingual identification, expanding it to all cases of language contact. Section 11.4 summarizes our key points and concludes the chapter.

11.2 Converging Voices, Converging Identities, Converging Languages

Faraclas et al. (Reference Faraclas2014: 177) argue that speakers of Atlantic Creoles make full use of structural and functional convergence between the languages in contact to maximize their use of Creole languages as instruments of multiple voicing. Those speakers also capitalize on the converging words and structures to express multiple identities: “pluri-lingual” and “pluri-cultural” Afro-Atlantic identities, Euro-Atlantic identities, and Atlantic Creole identities. Faraclas et al. (Reference Faraclas2014) admit that it is often challenging to identify a single source for many Atlantic Creole words and constructions, because such forms are identical or at least similar to those found in numerous African languages and in the nonstandard European varieties spoken in the Afro-Atlantic during the colonial period and beyond (Faraclas et al. Reference Faraclas2014: 178). Faraclas and his co-authors argue that due to convergence, the same lexical item can receive different interpretations from listeners of different linguistic backgrounds who may assign different grammatical roles to the same item based on their native language. In other words, their native language will serve as the mold through which they will shape differently their interpretation of the same linguistic element. For instance, the same word may be interpreted by those familiar with European languages to be an adjective and those familiar with West African languages to be a stative verb (Faraclas et al. Reference Faraclas2014: 178), which allows a word to become multifunctional. As Faraclas et al. (Reference Faraclas2014: 181) put it:

This lens on the complex interactions between speakers and listeners of different linguistic backgrounds repositions the original enslaved people and colonists as equal contributors to the genesis of Creole languages and certainly restitutes to the plurilingual African slaves the agency they deserve (see Faraclas Reference Faraclas2012). By means of illustration, we will briefly refer here to the spatial gram/locative marker na that can be found in Guinea-Bissau Creole as being derived from the contracted Portuguese preposition na (derived from the Portuguese preposition em contracted with the feminine article a, em+a > na) meaning ‘in.’ The same spatial gram can be found as na in Igbo, Serer, and Susu, three African languages believed to have contributed to the emergence of Guinea-Bissau Creole (Faraclas et al. Reference Faraclas2014: 188–189)Footnote .⁴ Faraclas et al. use such examples to convey that a speaker of an Iberian lexifier Atlantic Creole can maximize their use of locative marker na by exploiting the similar form and function across multiple source languages in order to express their multiple Afro-Atlantic and Euro-Atlantic identities, correlated with multiple voicing.

On this topic, expanding on previous work (Baptista Reference Baptista, Lang, Holm, Rougé and Soares2006, Reference Baptista2009), Baptista (Reference Baptista2020) collected data from multiple Creoles and other contact languages to illustrate how multiple voicing can showcase convergence as a key cognitive process in the emergence of Creole languages in particular and in language contact in general (see sample list of “converged features” in Table 11.1, adapted from Baptista Reference Baptista2020, which documents congruent forms in twenty languages and across nineteen grammatical domains). The view of convergence we defend here is based on the idea that language users converge on linguistic features that they perceive as being similar in the languages in contact. In other words, the trigger to the convergence are features that happen to be congruent in the languages in contact, may it be at the phonological, morphosyntactic, or lexical levels. The congruent dimension that Table 11.1 emphasizes is that of form/function mapping, focusing (though not exclusively) on grammatical morphemes that happen to have a similar form and similar meaning in the languages in contact. We argue in this chapter that language users’ perception of and use of the congruent forms are the very first step to language mixing.

Grammatical domains include negation, tense, mood, and aspect markers, auxiliaries, passives, conjunctions, complementizers, determiners, plural marking, possessives, causatives, copulas. Lexical categories include nouns, pronouns, adjectives, adverbs, prepositions. Phonemic inventories have also been argued to be subject to convergence.

Some of the congruent forms emerged from both the Creole lexifier and substrates, while others attest to the power of substratal reinforcement. In this respect, Chaudenson’s (Reference Chaudenson2003: 199) generalization that “les substrats ne peuvent pas passer en force” (substrate languages cannot forcefully infiltrate) the emerging superstrate koine language unless the substrates and superstrate show converging structures does not fully capture the complexity of the linguistic ecology present in the formation of many Creoles. We have ample evidence that African languages were preserved and used in the early part of the colonial period (Long 1774; DeGraff Reference DeGraff2002, footnote 45) and that such features could be selected even if they did not converge with the superstrate. As Aboh reports (Reference Aboh, Aboh and Smith2009: 321, footnote 4), Chaudenson’s position can hold only if we have a clear definition of convergence such that we are able to reconstruct the distinctive features that may converge in the competing languages. Baptista (Reference Baptista2020) is an attempt at responding to this challenge by providing a clear definition of convergence and reports on how distinctive congruent features have been reconstructed by various scholars. The set of congruent forms Baptista (Reference Baptista2020) reports on shows that the converged features may rise from contact between the superstrate and substrates or among substrates (substratal reinforcement), as in the case of Guinea-Bissau bin (see Table 11.1). The latter scenario clearly is in line with Mufwene’s (Reference Mufwene2005) idea of Creoles resulting from a Pyrrhic victory where the prevailing lexifier koine (that evolves into a Creole) is clearly affected by the displaced languages (the substrates) (p. 12).

11.3 Interlingual Identification (Weinreich Reference Weinreich1953)

In his Reference Weinreich1953 book, Languages in Contact, Weinreich first proposes the concept of interlingual identification. In discussing the effects of language contact, he divides crosslinguistic influence, or “interference,” as he calls it, into two types. In addition to borrowing, “there is also a type of interference, extremely common in language contact, that does not involve an outright transfer of elements at all,” he writes (Weinreich Reference Weinreich1953: 7). Instead, the use of an existing element in a language is modified to mirror the use of a similar element in another language in contact. Why does this happen? According to Weinreich, bilinguals identify a feature in one language with a feature in their other language, because of a perceived equivalence or overlap in form, function, or syntax (p. 39). The two features are rarely identical or perfectly equivalent, which means bilinguals who have made this interlingual identification – who perceive the features to be equivalent – will likely use at least one of the features differently from monolinguals.

Weinreich claims that this process is at the root of contact-induced language change and can lead to extensive parallelisms between languages with “long and intensive contact” (p. 8). This should be an expected outcome, he tells us, because overlaps in features across languages are extremely common, and therefore interlingual identifications are common. Bilinguals make extensive use of interlingual identification because it reduces the cognitive load, or “linguistic burden,” of speaking two languages (p. 8).

Weinreich states that interlingual identification can take place in any part of the language: phonology, morphosyntax, and semantics. Nonetheless, the primary example Weinreich gives of interlingual identification is in the phonological domain, as discussed below.

11.4 Implications and Conclusion

Taking as a point of departure Bickerton’s view that admixture is one of the chief characteristics of Creole languages (Reference Bickerton2008), the objective of this chapter has been to bring to light the tight connections between the congruent forms observed across Creole languages (Baptista Reference Baptista, Lang, Holm, Rougé and Soares2006, Reference Baptista2009, Reference Baptista2020; Faraclas Reference Faraclas2012; Faraclas et al. Reference Faraclas2014), which have been argued to result from speakers’ perception of similarities between the languages in contact, and Weinreich’s notion of interlingual identification. A close review of interlingual identification (as it was laid out in Weinreich Reference Weinreich1953) and of how the concept has been applied and experimentally tested in both situations of bilingualism (Flege Reference Flege1991) and multilingualism (Kresić & Gulan Reference Kresić and Gulan2012) attest to how speakers use their native language as the mold through which they shape differently their interpretation of the same linguistic element in another language. As a result, we argue that interlingual identification is ground zero for language mixing and language change. We propose that the same mechanism that contributes to Creole formation is also present in other contact settings such as bilingual and other language learning situations. If all languages are in contact, this means that the mechanism of interlingual identification is pervasive in all contact settings, and this, in turn, allows us to expand and reframe Bickerton’s notion of bastard tongues, while arguing that admixture is one of the major clues to our common humanity. We propose that all languages are mixed (see Aboh Reference Aboh2020; Mufwene Reference Mufwene2023) and that such admixture is inherent to all human linguistic interactions, and hence mixing is only a matter of degree. Examining this issue from a different angle, studies have shown that language learners are constantly looking for similarities and equivalences in the languages they acquire, perceiving and interpreting such similarities in different ways. Some scholars, such as Richie and Bhatia (Reference Richie and Bhatia2021: 1), view the quest for similarities as the bedrock of cognition: “Similarity is one of the most important relations humans perceive, arguably subserving category learning and categorization, generalization and discrimination, judgment and decision making, and other cognitive functions.” Others (Ringbom & Jarvis Reference Ringbom, Jarvis, Long and Doughty2009: 1) view the quest of interlingual similarities as an inherent part of language learning:

To sum up, interlingual identification, as developed by Weinreich, is a concept that aims to explain the cognitive process behind crosslinguistic influence in situations of language contact. It places bilinguals’ perception, as well as their desire to reduce the cognitive load of speaking multiple languages, at the center of contact-induced language change.

We argue that language learning involves perceiving or misperceiving similarities in the context of interlingual identification, making it the site of temporary admixture in the case of L2 or L3 learning and permanent admixture in the case of contact languages like Creoles, thus making admixture the true clue to our humanity, as it affects all languages when they come in contact with one another.

12.1 Bickerton’s Legacy on Negation and Negative Dependencies

In his Roots of Language chapter on creole languages, Bickerton discussed a list of features argued to be common to creoles languages in regards to the following constructions:

1. 1. Movement rules

2. 2. Articles

3. 3. Tense-modality-aspect (TMA) systems

4. 4. Realized and unrealized complements

5. 5. Relativization and subject-copying

6. 6. Negation

7. 7. Existentials and possessives

8. 8. Copula

9. 9. Adjectives as verbs

10. 10. Questions

11. 11. Question words

12. 12. Passive equivalents

The one focused on in this chapter is negation. Although earlier lists of pan-creole features had been offered, for instance in Taylor (Reference Taylor and Hymes1971), the twelve features discussed there did not include negation. A later list from Markey (Reference Markey1982) does mention negation, but only with respect to its positioning in relation to verbs. So by and large, it would seem that Bickerton was possibly the first to note the commonality of negative doubling constructions in the creole languages. Bickerton’s remarks on this topic in the original volume are quite brief, and yet also quite remarkably informative. They could be summarized as follows. First, Bickerton perceptibly noted that negated indefinite nominal expressions, subjects and objects often tended to be redoubled by the presence of negation on the verb in creole languages, supporting his remark with examples like (1):

1. (1)

   Non	dag	na	bait	non	kyat
   no	dog	not	bite	no	cat
   ‘No dog bit any cat.’

Then he further remarked that sentences of this type, while occasionally found in Hawaiian English Creole (HEC), tended to manifest negation doubling more frequently with ‘negated VP constituents’ than with subjects Bickerton (Reference Bickerton1981: 61).Footnote ¹ That is, in his very short paragraph on negation, we can see with hindsight that Bickerton put his finger first on the common occurrence in creole languages of the construction now known as ‘negative concord’ (NC) and second, on one of its major possible variants, namely, the distinction between so-called strict NC, in which the relevant indefinite expressions are uniformly redoubled by negation in all of their possible syntactic positions as in (1), and non-strict NC, which characteristically manifests a preverbal, postverbal asymmetry with respect to negation doubling by a sentential negative morpheme, as in (8) below. These are indeed the two types of NC constructions that have further been noted to occur in creole languages, as well as in other non-creole languages and whose distribution is discussed below. In his Roots of Language volume, Bickerton further takes up the topic of negation in his chapter on acquisition (1981: 123). There, he ponders the why and how at the root of the creole negative doubling dependencies. Reflecting first on the doubling of subject nominals, he asks whether sentences like (2) are found in child language, guessing at the time, in the absence of relevant quantitative empirical data, that they may be rather rare in child productions.Footnote ²

1. (2) Nobody don’t like me

His initial suggestion for English was to offer the hypothetical speculation that the root of this negative subject doubling construction could perhaps be found in the order of acquisition of the quantifiers somebody, nobody and anybody. If somebody were to be acquired first with negation, with nobody acquired second coming to replace it, this could lead to the child production of a doubling construction that the acquisition of anybody, presumably occurring last, would eventually come to replace. But Bickerton quickly abandoned this English-centered speculation, noting that even if it could turn out to explain data like (2) in child language, it would be unlikely to generalize and account for the frequency of negative doubling constructions across creoles. At this point, while questioning his own acquisition speculation, Bickerton takes a quick stab at the then common belief that faulty ‘child constructions’ could be a solid source of pan-creole features. As he characteristically puts it in his frank language, rather than limiting our search to what children may erroneously produce “There must be some way in which multiple negation is more natural than single negation, despite the pedagogues and logicians” (p. 171). This remark prefigures what is at present a current claim, namely that NC is amongst the most frequent types of negative dependencies not just in creole languages, but more generally across the world’s languages. Penka (Reference Penka2011) and De Swart (Reference de Swart2010), following Haspelmath (Reference Haspelmath1997), claim that “It is actually the non-cooccurrence of sentential negation with negative indefinites that is remarkable” (cited in van der Auwera & Van Alsenoy Reference van der Auwera and Van Alsenoy2016: 80Footnote ³). Bickerton ends his brief remarks speculating that if sentences like (2) turn out to be rare in the child language, then “in light of the creole evidence, the workings of the bioprogram must again be suspected” (1981: 171). More recent works on acquisition have in fact concluded that even in English, a language whose standard dialect is known for not allowing NC, this construction turns out to be a default interpretation for young children when tested on doubling negative constructions they may or may not actually produce (Thornton et al. (Reference Thornton, Notley, Moscati and Crain2016). As Thornton et al. (Reference Thornton, Notley, Moscati and Crain2016) argue, their study provides evidence for the existence of an NC grammar in English-speaking children, for which they should not receive much input from adult standard productions. This grammar is later on switched to the standard more ‘logical’ dialect where there is a one-to-one correspondence between negative marking and negative interpretation. In short, current works in child language acquisition have updated our knowledge in this respect and turn out to corroborate the view that NC is in a sense a default construction and interpretation for children, even when they are exposed to dialects that do not produce it. Hence, as long as it is understood along with Lightfoot’s remarks on Bickerton’s (2014) bioprogram hypothesis that beyond Bickerton’s original views, the bioprogram should extend to non-creole languages as well, the current crosslinguistic and acquisition evidence provides little grounds to disagree with Bickerton’s insights. Given its commonality within and beyond creoles, NC must indeed result from very general/universal principles of the language faculty. The questions that are now left to be pondered concern a deeper understanding of what language faculty principles it may build upon, a question for which a variety of hypotheses have been proposed in the literature but for which there remains much current debate and little consensus. But of course, as Bickerton makes clear, first and foremost is the need for a refined attention to the actual data, and in this sense it is useful to revisit what the landscape of negation and negative dependencies looks like in creole languages at the present time to summarize what turns out to have been discovered since Bickerton’s first observations. This is what the following sections endeavor to accomplish looking first at negation itself and second at negative dependencies primarily in French-based creoles, with an eye, however, to what happens elsewhere, when it is known.

12.2 Standard Negation in Creole Languages: Does the Position of Negation Matter for Negative Concord?

One set of hypotheses entertained by a variety of authors (Jespersen Reference Jespersen1917; Zanuttini Reference Zanuttini1991; Zeijlstra Reference Zeijlstra2004) under a diversity of theoretical approaches is that, in given languages, the possibility of NC dependencies could at least partially depend on the nature of the sentential negative marker. The goal of this brief overview is to assess what such views may entail for creole languages given the type of negation they harbor. I will start by inspecting the position of negation in the creole languages of the Atlas of Pidgin and Creole Language Structures (APiCS) database and then discuss what expectations this may have for NC dependencies. The position of negation markers in the surveyed creoles is given in Table 12.1.

The majority of creole and pidgin languages in the APiCS database manifest a preverbal particle negation, a feature once thought to be common to all creole languages, but since shown to include a few interesting exceptions. This finding simply confirms that in negation ordering, creole languages essentially align with the most frequent negation type and position in the world’s languages (525 languages in the WALSFootnote ⁴ 1,325 language sample). So creoles quite clearly are not unique in any way with respect to negative constructions – quite the opposite. They manifest rather what appears to be default options, preverbal particle negation and NC dependencies, under at least some linguists’ understanding of this type of dependency. As is generally the case, however, such broad crosslinguistic surveys focus largely on the most common forms of negation in the sample languages, leaving aside less prominent alternative constructions that can reveal unexpected but fascinating variation within a broader quite uniform picture.

If we narrow our perspective to look only at the French lexifier creoles in the APiCS, only Réunion Creole and Louisiana Creole are noted as allowing postverbal negation as a variant to the preverbal one. Although this is surely correct in simple regular declarative sentences, Déprez (Reference Déprez and Govain2021) shows that a postverbal negation variant can be found in essentially all the French lexifier creoles, with a surprising and interesting variety of uses. Table 12.2 offers a summary of the diverse extent of these postverbal uses and a list of examples is provided in (3) below:

1. (3)

   Fixed form

   Guadeloupean Creole

   an

   vlé

   on

   negress

   bombé,

   an

   vlé

   pa

   vouè

   pon

   fèy

   tol’

   I

   want

   a

   woman

   shapely

   I

   want

   not

   see

   bride

   of

   metal

   ‘I want a shapely woman not one that looks like a metal stick’

   (lesnyck’gwada, Fresh Badam [radio show])

   Modal

   Martinique Creole

   pé

   pa

   rivé

   bien

   lwen

   can

   neg

   arrive

   very

   far

   ‘could not have gone far’ (Ti Prens Lan: 16)

   Neg Raising

   Mauritian Creole

   Mo

   krwar

   pa

   [ki

   Li

   pou

   vini]

   1sg

   believe

   neg

   that

   3sg

   would

   come

   ‘I think he would not come’ (Native informant)

   Short form + aspect

   Lousiana Creole

   Mo

   lav

   pa

   mo

   figi

   1sg

   wash

   neg

   1sg.poss

   face

   ‘I don’t wash my face’ (Klingler Reference Klingler2003: 134)

   Finite/non-finite

   Réunion Creole

   M’i

   touch

   pa

   aou

   1s-

   touch

   neg

   obl-2

   ‘I do not touch you’ (Cellier Reference Cellier1982: 42)

I will make two observations with respect to this kind of data which is discussed in far more details in Déprez (Reference Déprez and Govain2021). First, they clearly complexify the overall picture inviting caution on the validity of too strong typological generalizations or sweeping statements about creole simplicity. That a language is classified as displaying a specific ordering for negation does not entail that it never permits other ones nor that these possible exceptions are of little interest. Second, as Déprez (Reference Déprez and Govain2021) notes, the ordering exceptions observed in the French-based creoles often finds echoes in other languages, indicating that as far as exceptions are concerned, these may not in fact be uncommon. For instance, the fixed use of a postverbal negation is also found in English expressions such as he loves me, he loves me not, a potential frozen remnant of a previous stage when the language displayed verb movement. Marking the scope of negation with respect to modals with a position change is a possibility also exemplified in French, English or German. Thus, in French the differing positions of a negative marker ne can be used to mark the negation scope (Tu ne dois pas voter en faveur de cette loi, tu dois ne pas voter en faveur de cette loi.). Yet, while in languages like English scope marking with the positioning of negation is not obligatory, since negation can clearly take scope higher or lower than where it is syntactically expressed, it seems more rigid in the French-based creoles, possibly owing to the rather rigid and transparent ordering of negation and TMA in these languages. The position of negation is also sometimes used in English or French to differentiate Neg raising interpretations from non-Neg raising ones, as we can see in the following examples. Do you think he will win? I don’t think so (ambiguous negation scope). I think not. (low negation scope only). Finally, the distinction of negation with respect to finite vs nonfinite V is found in a number of languages and obviously in French, the clear source of this alternation in these creoles. As Déprez (Reference Déprez and Govain2021) argues, it is quite likely that these observed negation ordering exceptions exemplify remnants of the verb movement option in French, generally no longer available in these creoles apart from in Louisiana and Réunion Creole, and with differing oppositions. The same may be true for the parallel constructions found in English, inviting the speculation that exceptions to negation ordering could be more common in the creoles that have verb movement languages as substrate or superstrate, a speculation left here for further research.

Beyond possible variations in the ordering of negation, of particular interest here is the question of whether the ordering of negation could have an impact on the type of negative dependencies that a creole manifests. Although not a question that Bickerton raised, it is one that has arisen in the study of NC in the Indo-European languages and that is worth glancing at here, given Bickerton’s more general interrogation for possible motivations behind the common occurrence of NC in the creole languages. While Zanuttini (Reference Zanuttini1991) suggested that NC could correlate with preverbal negation in a number of Romance languages, the generality of this correlation was dispelled in further research, as there are clear evidence of languages with postverbal negation that manifest NC. Van der Auwera and Van Alsenoy (Reference van der Auwera and Van Alsenoy2016), while revisiting this question within a larger and more diverse sample of 31 NC languages, concluded again that there is no discernible correlation. As they note “negative concord does not seem to be related very strongly to any particular word order regularity” (van der Auwera and Van Alsenoy Reference van der Auwera and Van Alsenoy2016: 19). Included in their survey were both the main constituent word order of languages (SVO) and the relative order of negation with respect to the V. In their sample, 10 out their 31 NC languages have preverbal negation, 13 have postverbal negation and finally 6 combine both possibilities. To reconsider this question with a focus on creole languages, we can start by looking at the few creoles that do not manifest a Neg V order and ask whether they could fail to instantiate NC with any systematicity. Skipping for now the creoles that manifest an alternating Neg V order, or a double negation on both sides of the verb, we are left with 5 creoles in the APiCS sample with V O Neg order and 4 where Neg immediately follows the V. As has previously been noted, neither the WALS nor the APiCS furnish easily quantifiable information on NC, largely because of how negative indefinites are defined in the feature inventory used. As Haspelmath specifies, “in this feature (feature 102 in APiCS) we are not interested in whether the indefinite pronoun itself carries negative meaning, i.e. the issue of ‘double negation’ is left aside. Sentences like I didn’t see anybody and I didn’t see nobody have the same status” (Haspelmath & APiCS Consortium 2013). What this means for our purpose is that the feature does not distinguish negative polarity dependencies from NC ones. Consequently, although all 5 of the creoles with a V O Neg order manifest a positive result for the co-presence of negation with indefinites, a closer inspection is required. For Berbice Dutch, the APiCS indefinite expression in the examples provided involve expressions with the indefinite ‘one’ respectively en gutu for one thing and en kene for one person as in (4). From the examples provided, we see that the interpretation is one of single negation, but this is expected if the relevant expressions simply never allow for a negative interpretation. From the examples provided we can only deduce that the sentence-final negation can have scope over an indefinite in subject position, but not that the language manifests NC.

1. (4)

   Example 28-178 (Kouwenberg Reference Kouwenberg, Kahrel and van den Berg1994: 249)
   tut	ju	drai	wɛrɛ	ju	drai	mɛt	en	gutu	ka
   until	2sg	turn	again	2sg	turn	with	one	thing	neg
   ‘When you return, you return with nothing.’
   Example 28-179 (Kouwenberg Reference Kouwenberg, Kahrel and van den Berg1994: 249)
   en	kɛnɛ	kan	kapo	ka
   one	person	can	cut 3sg	neg
   ‘Nobody can cut it down.’ (referring to a certain tree)

Turning to the other languages with the same Neg order, we find that three present the same facts as Berbice Dutch, that is, use indefinite expressions with the numeral determiner one. For these languages there is, in short, no evidence that the relevant indefinites carry negative morphology or can ever be interpreted negatively. This is not conclusive of course, as more semantic information would be required to strongly dismiss this possibility, but at first view it does not seem that these creoles permit NC. Yet at least one of these creoles appears to provide strong evidence for NC, namely Principense. Here the indefinite expressions used clearly carry negative morphology and/or are inherited from Portuguese expressions that do. This language, then, provides solid evidence that NC can occur in creoles that manifest a negation in final position after the V and its object.

1. 1. a.

      N	vê	ko	nhon	fa.
      neg	1sg	see	thing	no
      ‘I didn’t see anything.’

   2. b.

      Ami	n	têndê	ningê	nhon	na	nixi	ki	gita	fa	ô.
      1sg	1sg	hear	person	no	loc	here	rel	shout	neg	val
      ‘I didn’t hear anybody shouting myself.’ (Maurer Reference Maurer2009: 139)

If we consider the creoles with negation immediately after the verb, only two have this order as their exclusive option, Eskimo Pidgin and Yimas-Arifundi Pidgin. Unfortunately, the APiCS provides no information about their indefinite negation relations, most likely for lack of empirical data. For languages that present a bipartite negation, one preverbal and one postverbal after the object, they all have a Portuguese lexifier and appear to manifest NC. And finally, concerning languages that feature a post-V negation as one of their options, the best examples are the Louisiana and Réunion French lexifier creoles, which both solidly feature NC, albeit slightly differently, since the former features strict NC and the latter non-strict NC with a preverbal negative indefinite that suffices to render the sentence negative without the co-presence of negation. (See example 8 below.)

To summarize, in this section, we have first looked at the position of sentential negation in creole languages and evoked some interesting variations to the generalizations provided in the APiCS for a subset of the creoles, namely the French-based creoles. We have then examined the question of whether the syntactic position of the negative marker could have an impact on the occurrence of NC and concluded that wherever matters were decidable, this factor did not seem to matter more in creole languages than crosslinguistically, beyond creoles. The discussion, however, revealed difficulties in deciding whether or not a given creole does manifest NC. In the following section, we further discuss this important matter before launching into a more narrowly empirical investigation of NC and its possible variations in the French-based creole.

12.3 Questions of Frequency and Typological Representativity of Creole Negative Concord

After exploring the negation order in relation to the occurrence of NC in creole languages, one obvious question that arises is how frequent this phenomenon actually is. Clearly the fact that Bickerton noted NC as a possible pan-creole feature suggests high frequency. But the actual reality is a bit harder to establish. Indeed, as noted in many works on NC, the distinction between what can be considered a negative polarity dependency and what can be considered a NC one is often both empirically rather tenuous and theoretically contentious. The basic concept of NC can seem deceivingly straightforward; one can speak of NC if there is evidence that the terms involved in the negative dependency can sometimes manifest an independent negative meaning, which at other times appears to dissolve. Gianakidou (Reference Giannakidou, Everaert and van Riemsdijk2006) proposes the following descriptive definition for the terms that participate in an NC relation, which I refer to here as negative concord items or NCI, following Watanabe (Reference Watanabe2004).

1. (6)

   ‘NCI’: An expression α is an ‘NCI’ iff:

   (i) α can be used in structures that contain sentential negation or another α-expression, yielding a reading equivalent to one logical negation; and

   (ii) α can provide a negative fragment answer (i.e., without the overt presence of negation).

This definition clearly takes interpretation as the core criterion for the determination of what an NCI is but curiously only allows for a rather narrow condition under which an NCI’s negative value can be tested, namely fragment answers. In their paper on the typology of negative concord, van der Auwera and Van Alsenoy (Reference van der Auwera and Van Alsenoy2016) criticize the validity of this test on the basis of Yiddish and Brabantic Dutch data. In Yiddish, they note, indefinite negative expressions like keyn, which are etymologically directly related to the German negative quantifier kein, participate in NC dependencies. Yet in fragment answers, they require the co-presence of negation. Likewise, in Brabantic Dutch, similar indefinite expressions that can optionally co-occur with negation in regular sentences, also allow for the optional co-presence of negation in fragment answers. In their view, since both of these languages manifest NC, Giannakidou’s criterion (ii) cannot be taken as definitional. In the end, they leave quite open the definition of what NCIs ultimately are, but it can be surmised that in practice, in their own survey of NC in the world’s languages, they rely more on morphological evidence for negation than on semantic evidence. As has been discussed since at least the 1990s, however (Laka Reference Laka1994), the use of negative morphology to determine whether given expressions are NCIs rather than negative polarity items also encounters problems. For instance, in a language like French where expressions like personne (nobody) or rien (nothing) can clearly have a negative interpretation, there is no evidence of negative morphology. The picture gets even more blurry in languages like Catalan, where some negative indefinite expressions, such as nadie (nobody), appear to bear negative morphology, while others like res (from the Latin noun for ‘thing’) clearly do not. Yet, in spite of their morphological differences, both manifest the same behavior in negative dependencies. At the other end of the spectrum, Basque indefinite expressions like i-nor (no-body), e-zer (no-thing), i-noi (no-time), i-non (no-where) are formed with a negative prefix that combines with an indefinite also used for questions and so, they overtly feature negative morphology. Yet, as Etxeberria et al. (Reference Etxeberria, Tubau, Borras-Comes, Déprez and Espinal2018) demonstrate, these expressions are better analyzed as NPIs given their distributional and semantic properties as well as the fact that they can never occur in fragment answers, be it with or without a sentential negation, nor ever lead to double negation readings. Observations of this sort underscore the difficulty of determining whether or not given languages actually manifest NC and highlight the need for detailed analyses of each case, of the kind carried out in Déprez and Henry (Reference Déprez and Henri2018), for a better understanding of this type of negative dependencies, their common characteristic across languages and their potential variations. Nonetheless, since the goal here is to assess the current state of knowledge concerning Bickerton’s contention that NC is a good candidate for a pan-creole feature, and since Bickerton himself appeared to rely on the morphological criterion to characterize the construction, it is useful to attempt some inquiry about the frequency of this construction in creole languages on the basis of the morphological criterion, acknowledging all the while the fragility of such an enterprise as long as more detailed knowledge about the interpretation possibilities of these dependencies in the different creoles is still in large part missing. This is why, although here we offer a quick overview of whether different creoles have been observed to manifest NC, we endeavor in the next section to focus on a more detailed study of the similarities and interesting variations that a closer study of NC in a subset of the French- based creoles can reveal.

Proceeding with our rapid frequency review of the set of creole languages in the APiCS database, we can first rely on the results of van der Auwera (Reference van der Auwera, Ziegeler and Bao2017) who discusses NC frequency considerations for the English lexifier creoles. There he finds that all the Caribbean English creoles manifest NC, that African Creoles also do with one exception, namely Nigerian Pidgin English, which seems instead to feature either negative quantifiers or negative dependencies that involve indefinite expression inherited from the English anything. Concerning Pacific creoles, Bislama and Tok Pisin do not appear to manifest NC, and NC appears to be very rare in Norf’k but is otherwise attested in the other creoles of this region. Van der Auwera concludes that in the English lexifier creoles, NC is frequent, basing his observations on the co-occurrence of sentential negation, generally no and negative indefinite expressions inherited from the English negative quantifiers nobody or nothing. At present, however, it is not always known whether these expressions are in fact able to take on any negative meaning in these creoles. The possibility of fragment answers and a more detailed distribution of the co-occurrence of negation and these negative indefinites are studied for a number of English creoles, more specifically in Jamaican Creole, in Belize Creole (van der Auwera Reference van der Auwera2022) and in Pichi (Yakpo Reference Yakpo, Déprez and Henri2018) as well as in Singlish by (Zhiming & Luwen Reference Zhiming, Luwen, Déprez and Henri2018). Turning to Portuguese lexifier creoles, the conclusions are essentially similar. All appear to manifest NC, since in most cases their negative indefinite expressions are directly inherited from their lexifiers which already manifest NC. Detailed studies of the distribution and meaning of NCI can be found for Cape Verdean Creole (Pratas Reference Pratas, Déprez and Henri2018), San Tome (Hagemeijer Reference Hagemeijer, Michaelis, Maurer, Haspelmath and Huber2013), Korlai (Clements Reference Clements, Déprez and Henri2018) and Guinea Bissau Krio (Khim Reference Kihm, Déprez and Henri2018), which all confirm the use of NC and the possible negative meaning of these creole NCIs in fragment answers. For the other Portuguese lexifier creoles, the APiCS database furnishes relevant examples for all except Angolar, whose NC status remains unclear as the APiCS example uses indefinite expressions equivalent to ‘one thing’ for which the availability of a negative semantics would need confirmation. Concerning the Dutch-based creole, Berbice Dutch was already discussed above, and the remaining ones are Afrikaans, for which evidence and details of NC are much discussed in the works of Biberauer (Reference Biberauer2007) among others, and Negerholland for which the APiCS example provides positive evidence of at least morphological negation in its negative indefinite expressions. French lexifier creoles are discussed in more detail below, but for now it can be noted that they all manifest NC under the Giannakidou definition. For creoles with Spanish as a lexifier, matters appear to be more complex, which in itself is interesting given that the lexifier language clearly manifests NC. Only Papiamentu is noted as manifesting NC, as its negative indefinites are inherited from the Spanish NCI. For all varieties of Chabacano, APiCS only provides examples in existential constructions. Some of these use Spanish inherited negative indefinites doubled with negation in these constructions, but this remains insufficient evidence. Finally, for Palenquero, Schwegler (Reference Schwegler, Déprez and Henri2018) provides evidence of NC with postverbal indefinites but leaves undiscussed the question of preverbal ones (as do the APiCS examples) and that of the occurrence of negative indefinite in fragment answers, so the information remains at present incomplete. For the Bantu-based creoles, as well as a few others grouped with them, the APiCS examples are inconclusive as they are based on negation co-occurring with what appears to be positive indefinites. Finally, for Malay lexifier creoles, here again APiCS data are inconclusive since the relevant indefinites noted to co-occur with negation are either of the positive type (one NP) or questions-based items (what NP) for which there is no evidence of negative morphology or interpretation.

Overall, with this rapid tour of the APiCS creoles, we find that NC is indeed quite frequent, but also that this seems in large part an inheritance from the lexifiers they are built upon. While standard English or Dutch are usually assumed not to be NC languages, most of their dialects are, and so it still comes as no surprise that lexifier creoles should also manifest NC. In reviewing the frequency of NC in a larger language set, van der Auwera and Van Alsenoy (Reference van der Auwera and Van Alsenoy2016) concluded that NC appears to be most frequent in Indo-European languages. In this respect, then, we again find a great parallel between what happens in creole languages to what happens in the world languages, namely the influence of the Indo-European subset. Yet one cannot but wonder whether NC is really an areal feature or whether this could rather be a consequence of the fact that these languages are the better-studied ones so that the characteristics of negative dependencies are better researched there. Both for the non-creole sphere and for the creole one, the question remains whether or not, upon further detailed inspections, non-Indo-European lexifier creoles could display just as high a frequency for NC once their empirical landscape is better mapped out.

In favoring NC dependencies, it is of interest to note that creole languages clearly depart from fiercely analytic languages such as Mandarin Chinese or Cantonese to which they are not uncommonly typologically compared. As is well known, these languages severely preclude concordant negative constructions. If creoles are indeed analytic languages, why is there such a strong difference? In considering the question of creoles as analytic languages, Szeto et al. (Reference Szeto, Lai and Ansaldo2019) endeavor to compare common features of two groups of such languages, Sinitic and Kwa, to creole languages. They conclude that, on the one hand, creole languages do indeed arguably belong to an analytic typological type, in as much as such a type can be defined, and on the other hand, not unlike other analytic languages, their features also reveal their structural debt to their relevant substrates. Indeed, not surprisingly, Szeto et al. (Reference Szeto, Lai and Ansaldo2019) find that creole languages may be divided into those that most resemble the analytic type of the Kwa languages and those that don’t. Even though these authors do not include negation among the features they discuss, it seems certain that creole negative dependencies would end up owing more to the Kwa analytic type than to the Sinitic one. When peeking at the other end of the spectrum, we see Singlish stand out among the creoles as one that clearly lacks the concord type of dependencies, no doubt due to the influence of Chinese, as is explicitly argued in Zhiming et al. (Reference Zhiming, Luwen, Déprez and Henri2018). In sum, the common occurrence of NC in creole languages seems typologically surprising only if the Sinitic languages are taken as a model for analyticity. So here again, in the frequency of NC, creole languages affirm strong roots both in their Kwa substrate analytic type and in their lexifiers, leaving little space for any considerations of creole uniqueness.

12.4 Variations in Negative Concord: A Look at French Lexifier Creoles

One question that a discussion of the frequency of NC leaves untouched is that of its diversity. For the creoles that display NC, do these essentially exhibit a uniform landscape or can we detect substantial variations, and if so are any of these variations specific to the creole languages in any way? These questions are addressed here by discussing the constants and variations in the NC dependencies of a subset of creole languages, namely the French lexifier creoles.

To begin our empirical survey of NC dependencies in the French lexifier, let’s start by observing that in most French lexifier creoles, NC dependencies appear at first view to be of the strict NC type, requiring negative doubling with both preverbal and postverbal NCIs, as exemplified in (7). Note that, since there is empirical confirmation that all of these NCIs can be used without negation in fragment answers, we are here dealing with NC dependencies as defined by Giannakidou (Reference Giannakidou, Everaert and van Riemsdijk2006) and not on the basis of negative morphology.

1. (7)

   a.

   HC:

   Pèsonn

   *(pa)

   wè

   anyen

   b.

   Mau C:

   Personn

   *(pa)

   vwar

   nanye

   c.

   Mart C:

   Pèsonn

   *(pa)

   wè

   ayen

   d.

   Guad C:

   Ponmoun

   *(pa)

   vwè

   ayen

   e.

   Guy C:

   Pésonn

   *(pa)

   wè

   ayen

   f.	Sey C:	Personn	*(pa)	vwar	naryen
   Nobody	not	see	nothing
   ‘Nobody saw anything’

Despite this first appearance of strong homogeneity, note that there is at least one among the French lexifier creoles, that is, Réunion Creole, that displays non-strict NC. As the example in (8) clearly illustrates, the same NCI expression requires a doubling negation when occurring in a postverbal position, but not in a preverbal one.

1. (8)

   Persone	la	aprivoiz	a	zot	e	zot	la	pa	aprivoiz	persone.
   Nobody	past	tame	prep	3pl	and	3pl	past	neg	v	nobody
   ‘Nobody tames anyone and they do not tame anyone.’ (Lo ti Prins 1999: 72)

On a first pass, then, negative concord in the French-based creoles appears to uphold the distinctions noted in Bickerton as well as in cross-linguistics studies for two main types of negative concord relations, strict and non-strict ones, with the former clearly more common than the latter, as has also been observed in the Romance languages (Benini & Ramat Reference Bernini and Ramat1996) and beyond (van der Auwera & Van Alsenoy Reference van der Auwera and Van Alsenoy2016).

Upon closer inspection, however, this neat picture dissolves rather quickly. Starting with Réunion Creole, observe that the co-presence of negation with preverbal arguments if dis-preferred is not fully precluded as the example in (9) taken from the same corpus indicates. Although it is clear that doubled subject cases are rarer than non-doubled ones in this text, the exact conditions under which doubling is possible for preverbal argument more generally remain at present to be further investigated.

1. (9)

   Me	persone	pa	vouli	kroir	a	li	akoz	la	manière	li
   But	nobody	not	want	believe	to	him	because	the	way	he
   te	kostime.
   past	dressed
   ‘But nobody wanted to believe him because of the way he was dressed.’ (Lo ti Prins 1999: 18)

In this respect, the type of NC manifested in Réunion Creole resembles the one observed in Catalan in the Romance area. There are, however, other interesting cases in the other French lexifier creoles in which negative doubling with preverbal arguments also appears to be optional. Looking at Haitian Creole, for instance, Déprez (Reference Déprez2018) reports that for one of her native informants of a Gonaive dialect and his family, while the negation pa is obligatory with the NCI pèsonn, it is optional with expressions such as Pyes moun, but again, only in preverbal position. A similar observation is recorded in Albert Valdman’s (2017) English Haitian-Creole bilingual dictionary (p. 629) for a complex nominal expression using the NCI oken in preverbal position. Valdman provides the example (10b) in which the doubling negation is missing.

1. 1. a.

      Pyes	moun	(pa)	pati
      Piece	person	not	left
      ‘Nobody left.’  (Native informant)

   2. b.

      Okenn	lòt	moun	vini
      No	other	person	came
      ‘Nobody else came.’  (Valdman 2017: 629)

Finally, still in Haitian Creole, we observe that with more complex subject nominal expressions, the negation pa is also optional. Hence, examples like (11) found on the internet are accepted by our informants.

1. 1. a.

      Sa-k	kapab	geri	tout	kè-m?
      What-that	can	heal	whole	heart-1s
      Anyen	pase	san	Jezu-Kri
      Nothing	over	blood	Jesus-Christ
      ‘What can heal my heart ? Nothing but the blood of Jesus Christ’
      (Religious song: http://reyolaudiomp3.blogspot.com/2011/12/55-chandesperans-kreyol-sa-k-kapab-lave.html)

   2. b.

      Pèsonn	ki	te	wè	sa a	ta	rapòte	li
      Nobody	who	pst	see	that	cond	report	it
      ‘Nobody who would see this would report it’ (Native informant)

Similar examples can be also found for Mauritian Creole, where the speakers consulted reported a distinction. When comparable complex DPs were used in a preverbal subject position as opposed to a postverbal one, only in the former did negation doubling become optional.

1. 1. a.

      Nanye	ki	to	fer	kapav	aret	li
      Nothing	that	you	do	able	stop	him
      ‘Nothing you do can stop him.’ (Native informant)

   2. b.

      Nanye	apart	to	lakor	neseser	pou	gagn
      Nothing	except	your	agreement	necessary	to	win
      sa	draw	la.
      dem	right	def
      ‘Nothing but your agreement is necessary to win this right.’
      (Native informant)

   3. c.

      Mo	pa’nn	invit	personn	apart	twa	pou	sa	dine		la.
      1st	neg-past	invite	nobody	except	you	For	dem	diner		def
      ‘I did not invite anyone but you for this dinner.’ (Native informant)

The generalization across this novel data seems to be that there is a difference between simple pronominal NCIs and complex DP NCIs (Déprez Reference Déprez2018). While the former generally require doubling in preverbal subject position, doubling becomes optional for more complex DP NCIs. Yet this complexity does not have the same effect for postverbal NCIs where doubling remains obligatory. Moreover, although syntactic complexity seems to be a factor, there can also be lexical differences. Hence, while our Haitian native speaker accepted (10), he did not like the same sentence with another expression Pèsonn moun where he required negative doubling. Comparable distinctions in the optionality of negative doubling in preverbal position between bare pronominal expressions and complex DPs are also reported for Jamaican Creole (van der Auwera and De Lisser Reference van der Auwera and De Lisser2019) and Belize Creole (Van der Auwera Reference van der Auwera2022). Beyond creole, Déprez and Poletto (Reference Déprez and Poletto2019) also discuss comparable variation for the Northern Italian dialects. This type of variation is thus not specific to creole NC, and it calls in question the distinction between strict and non-strict NC, since here doubling optionality concerns not language types but specific expression types within the same languages.

Yet further distinctions in doubling can be observed for other NCIs in the French lexifier creoles as well as in the Northern Italian dialects. As Déprez (Reference 230Déprez, Ziegler and Zhiming2017) observes, adverbial expressions equivalent to never display interesting variation in their doubling requirement in Seychelles Creole as well as in Mauritian Creole. As the examples in (13) show, preverbal adverbs that occur most likely in a focused position at the beginning of the clause do not require doubling. Moreover, they display sufficient negative force to be able to license other NCIs under negative spread, that is, without the co-presence of the sentential negation marker.

1. 1. a.

      e	zanmen	son	ansennyan	in	bezwen	pran	roten	avek	li
      and	never	his	teacher	cl	need	take	hard	with	3sg
      ‘And never has his teacher needed to be hard with him.’

   2. b.

      Remon	pa	frekant	dimoun	isi	e	Zanmen	i
      Raymon	neg	mingle	people	here	and	Never	cl
      Pou	anmenn	okenn	zanmi	dan	son	lakaz.
      would	bring	no	friends	in	his	House
      ‘Raymond did not mingle here and never would he bring any friends to his house.’

   3. c.

      Mis	mwan	mon	pa’n	zanmen	fer	mon	zanfan	soufer
      but	1sg	poss	neg	never	make	poss	child	suffer
      ‘But I would never make my child suffer.’

Note that in contrast, argumental NCIs within the very same creole clearly continue to require negative doubling. We see again that within the same language, some NCIs can require strict NC, while others can display non-strict NC. This shows that languages are not homogeneous with respect to NC and exhibit differences that could either concern a particular class of NCI, namely adverbial ones as opposed to nominal ones, the syntactic complexity of the NCI or a lexical difference. What these data clearly demonstrate is that strict vs non-strict NC is not a distinction between languages here but can concern a subset of the terms involved in NC dependencies within the same language. These data provide clear counter examples to the idea that the strict and non-strict distinction in NC could be a matter of crosslinguistic parametric difference.

Distinctions in doubling requirements are not the only type of variations that the creole NC dependencies display. As discussed in Déprez (Reference 230Déprez, Ziegler and Zhiming2017), different NCIs or the same NCIs in different creoles also display variations in their possibility to occur in characteristic negative polarity contexts without a doubling negation. Table 12.3 sums up the different type of licensing contexts in which argument NCIs in Haitian Creole vs. Mauritian Creole can occur.

As Déprez (Reference 230Déprez, Ziegler and Zhiming2017) argues, these variations very much parallel distinctions that are also found crosslinguistically with NPIs and that have come to be characterized in terms of the strength of the negative relation involved in the dependency. In particular, distinctions in licensing context types differentiate so-called strong NPIs licensed in anti-additive contexts from super-strong NPIs licensed only in antimorphic contexts. Here, the very same distinctions are seen to characterize the argument NCIs of Haitian vs. those of Mauritian creoles. That NCIs are sensitive to the same type of distinctions as NPIs provides strong evidence that the former could well turn out to be a mere sub-kind of the latter type of expressions, as Déprez (Reference 230Déprez, Ziegler and Zhiming2017) concludes, following Laka’s (Reference Laka1994) initial proposal among many others. It is to be noted additionally that distinctions in the behavior of NCIs in non-negative contexts is not limited to distinctions across creoles. In Guadeloupean Creole, for instance, Petijean and Schang (Reference Petijean, Schang, Déprez and Henri2018) have found that two distinct forms for the adverb never, jam vs. jamen actually behaved differently in non-negative contexts with the former being excluded while the latter is possible. Given such fine-grained lexical distinctions, we are again strongly reminded of the diversity that studied NPIs have been shown to display language internally.

Yet another type of variation that distinguishes NC relations in the French lexifier creoles concerns their locality conditions or more specifically the distance at which a licensing negation can be separated from the licensed NCI. For Mauritian Creole and Guadeloupean Creole argument NCIs, for instance, the negation must essentially occur within the same clause as the NCI. But this is not the case in Haitian Creole or Martinique Creole where the distance can be longer as the examples in (15), taken from the online Haitian Bible and verified with our informants, attest. Note again that this distinction in locality is one that is also known to differentiate among types of NPIs crosslinguistically, which provides yet another argument to conceive of NC dependencies as a subtype of NPI.

1. 1. a.

      	Mauritian Creole
      	*Mo	pa	ti	dir	personn	ti	vini
      	1sg	neg	past	say	nobody	past	come
      	‘I did not say anyone would come.’ (Syea Reference Syea2013: 155)

   2. b.

      Mauritian creole
      *Mo	pa	kone	si	li	finn	apel	person
      1sg	neg	know	if	3sg	asp	call	nobody
      ‘I do not know if he called anyone.’ (Syea Reference Syea2013: 162)

   3. c.

      Guadeloupe Creole
      *Marie	pa	di	ké	ou	vlé	ban	mwen	anyen.
      Marie	neg	say	that	2sg	want	give	me	nothing
      ‘Marie didn’t say that you don’t want to give me anything.’
      (Petijean & Schang 2018)

1. 1. a.

      Haitian Creole
      M’pa	kwe	Mari	di	li	we	pèsònn
      1sg neg	believe	Mary	say	3sg	see	nobody
      ‘I don’t believe Mary said she saw anyone’ (Déprez Reference Déprez and DeGraff1999)

   2. b.

      Haitian Creole
      M	pa	kwe	m	pale	ak	pyes	moun	non,	pa	fache	ou
      1sg	neg	believe	1sg	speak	with	bit	person	no,	neg	anger	2sg
      ‘I don’t believe I spoke with anyone no, do not get mad.’ (https://www.facebook.com/video)

   3. c.

      Seychelles Creole
      Mon	pa	ekspekte	li	rann	mwan	naryen	an	retour
      I	neg	expect	he	returns	me	nothing	in	return
      ‘I didn’t expect him to give me anything back in return.’

   4. d.

      Martinique Creole
      Man	pa	di	ou	pesonn	telephone
      1sg	neg	say	2sg	anybody	call
      ‘I did not tell you that anybody called.’ (Native informant)

As these more fine-grained comparisons reveal, the NC landscape of the French lexifier creole is far from homogeneous, displaying variations in negative doubling, contexts of licensing and locality conditions that in the end come to question whether NC could usefully be taken as a criterion to classify ‘languages’ rather than merely ‘types of dependencies.’ There has been a strong tendency in recent years to try to separate languages that manifest NC from languages that don’t in terms of a parametric divide (Zeijlstra Reference Zeijlstra2004). At the same time nothing comparable has ever been hinted at for other negative dependencies such as polarity item dependencies. To my knowledge, indeed, no one has ever argued that languages that manifest negative polarity item dependencies should be parametrically distinguished from those that don’t. If NC is but a sub-kind of NPI dependencies, that is, a kind of dependency that is already known to come in different types and forces, why should one type of dependency be understood as manifesting a parametric divide but not the other? As is well known, both types of dependencies can coexist in given languages, as they do, for instance, in French, and they are also historically related, since polarity dependencies can evolve into concord ones and the other way around, though perhaps less frequently. In recent work, Zeijlstra (Reference Zeijlstra2021) has argued that polarity dependencies display far more diversity than NC dependencies and that the diversity that NC manifests is more systematic and not unlike what is observed in other types of dependencies like agreement. I contend, however, that this kind of assertion mostly stems from insufficient knowledge and focus on the extent of empirical NC diversity and that increased attention to the details of NC diversity has and will unearth far more diversity than is presently discussed in the literature. In my work, I hence proposed that there should be no conceptual divide between negative polarity dependencies and NC but rather that these exemplify the two ends of a spectrum of the manifestation of negation in the sentential vs. the constituent domain and that the focus of research should be on understanding the nature of the terms that participate in these two types of dependencies and their properties and not on the formulation of a putative parametric divide. Furthermore, when focusing on creole languages, one cannot help but notice that the pervasiveness of negative concord sharply contrasts with the rarity of agreement relations, raising immediate empirical questions for any views that claim that both exemplify the same type of syntactic relations. Why should NC be so common a phenomenon in creoles, while number or gender agreement is so rare? Contrasting with Zeijlstra’s parametric agreement approach to negative concord, Preminger Reference Preminger2013 and Preminger and Polinsky (Reference Preminger and Polinsky2015) have forcefully argued that agreement and NC do not, in fact, involve the same kind of syntactic dependencies, and that the similarities that proponents of such a view invoke represent a spurious unification. It is clear that at first sight the empirical landscape of creole NC as opposed to that of number or gender agreement in creoles tends to support this position. It is of course always possible to claim that behind empirical evidence, there could be abstract principles at play. One should then wonder, however, if this does not amount to a claim subsuming all syntactic dependencies into a single one, a possible view of course but one that could risk missing crucial distinctions of important interest.

For creole languages, the facts remain that if NC were subsumed under agreement, the meat of the question would then have to be why is the former so prominently overt in the creole languages while the latter is so visibly absent. And until an interesting answer can be provided to this puzzle, these two syntactic relations may perhaps best be kept apart.

12.5 Conclusions and Questions

While our survey of negative concord in creole languages from the perspective of Bickerton’s legacy has supported most of his conjectures, it has also generated more questions than actual answers. We started out by confirming the absence of relation between the position of negation and the occurrence of NC seen also crosslinguistically, as well as the frequency of NC in the creoles of the APiCS Atlas. On both of these counts, creoles were found to align with a majority of other languages in the WALS. In this sense, it is evident that in the make-up of their negative syntax and negative concord, creole languages are not unique. It turns out, however, that in the course of our assessment, criteria for the distinctiveness of NC as a potential classificatory language feature have become eroded and blurred. As noted above, to be able to raise questions of frequency for this type of dependency, we have needed to rely on the rather superficial criterion of negative morphology without being able to always answer the important question of its corresponding interpretive correlate. Reliance on this criterion, however, may force arbitrary classificatory choices of what is or is not an NC language, since NCI terms observed to behave in the same way can in fact differ on this criterion, language internally and crosslinguistically. And clearly, without evidence of actual negative interpretation for the terms involved in a negative relation, the very notion of negative ‘concord’ could turn out to be essentially moot. Furthermore, when investigating the diversity in the NC relations of the French-based creoles, the often assumed neat divide between strict and non-strict NC languages could in the end not be upheld, as this pattern of diversity was also found to characterize distinct NCI terms language internally, and not just as crosslinguistic distinctions. This calls into question the idea that strict vs. non-strict NC could distinguish language types rather than construction types within and across languages. Other patterns of diversity concerning the licensing of NCIs in non-negative contexts as well as their locality conditions have revealed strong similarities between NCI dependencies and the observed crosslinguistic behavior of NPIs. Such similarities lead us to question the wisdom of considering NC as a potential classificatory feature for languages, given in contrast that NPI dependencies have never been assumed to form classificatory distinctions among language types. Our survey also turns out to question whether it could actually be possible that focused attention to NC rather than NPI relations in creole languages are more a product of an Indo-European lens than an actual characteristic of creole languages. The majority of creoles assumed to display NC appear to do so because they make use of indefinite terms inherited from their lexifiers that in turn tend to display NC relations. So could focus on NC relations in creole be essentially an artifact of focusing on the characteristics of a restricted set of indefinite expressions language internally, mostly assumed to have inherited their negative import from their lexifiers? An answer to this question surely calls for further investigation of this phenomenon, deepened by a more detailed, conjoined look at both of its facets, the morphosyntax and the semantic aspects of its realizations. If it turned out that the diversity found is in many respects close to the one otherwise observed for NPI dependencies, then the idea that NC could be a crosslinguistic or a pan-creole classificatory feature could make no more sense than the constatation that NPI dependencies arise in most languages in a great diversity of types. But even if NC could no longer be considered a language classificatory feature, but rather a mere construction type that characterizes the behavior of a certain class of indefinite expressions, this would of course not mean that its study is fruitless. As one type of negative dependencies, its detailed study undoubtedly reveals the subtle properties of the working of negation and the multiple interactions it displays between its sentential manifestations and its incarnation in the argument or dependent domain. So while classificatory considerations may turn out to be artificial, the properties of NC relations and their potential diversity certainly are not.

13.1 Introduction

Since D. Bickerton’s Reference Bickerton1981 seminal work in Roots of Language on the properties of Creole NPs, a vast literature has been published on the syntax, semantics and pragmatics of Creole NPs (see Baptista & Guéron Reference Baptista, Guéron, Baptista and Guéron2007; Bobyleva Reference Bobyleva2013 inter alia).Footnote * Bickerton’s discussion of specificity in Creole NP has been extremely influential (Bobyleva Reference Bobyleva2013: 73). Despite critical remarks passed on the specific ~ non-specific dichotomy (Mufwene Reference Mufwene, Muysken and Smith1986; Wespel Reference Wespel2008; Wiesinger 2017 inter alia), it is still thriving (Baptista & Guéron Reference Baptista, Guéron, Baptista and Guéron2007: 19). In light of Bickerton’s initial analysis of Creole NP and of subsequent literature, the present chapter compares and contrasts the NPs of the three main French-related Creoles of the Indian Ocean – Mauritian Creole (MC), Seychelles Creole (SC) and Réunion Creole (RC).Footnote ¹ It is principally based on a discussion of recent monographs devoted to the Creoles of the area.

The term NP or Nominal Expression (NE) (Bobyleva Reference Bobyleva2013) will be used alternately in the chapter; they encompass both the lexical NP and the functional DP defined in formal linguistics (Chierchia Reference Chierchia1998). The syntax of Indian Ocean (IO) Creole NPs has been widely discussed, especially within generative linguistics (see Déprez Reference Déprez2007a, Reference 254Déprez, Baptista and Guéron2007b; Guillemin Reference Guillemin2009; Syea Reference Syea2013 and Alleesaib Reference Alleesaib2012 inter alia); it will not be addressed per se in this chapter. Here the focus is mainly on the semantics and pragmatics of IO Creole NPs in a broadly functionalist perspective (Givón, Reference Gívón1979).

The chapter is organized as follows. Section 13.2 explains the theoretical framework of the study. Section 13.3 is devoted to a presentation of IO Creole NPsFootnote ² and of its determiners.Footnote ³ Section 13.4 focuses on semantic and pragmatic definiteness in IO Creoles. Section 13.5 discusses other features of determiner and bare NPs in nominal reference. Section 13.6 examines the findings of the chapter in light of the grammaticalization of determiners in IO Creoles.

13.2 Frame of Analysis

In order to analyse the semantics and pragmatics of IO Creole NPs, different dimensions of nominal reference – identifiability, individuation, definiteness, specificity and genericity – have to be defined (Bobyleva Reference Bobyleva2013: 60ff.). In the vast literature available, I shall restrict myself to some references that provide an answer to the following question: How do overt and non-overt nominal markers and lexical noun types contribute to nominal reference?

As a rule, NPs denote KindFootnote ⁴ (i.e. classes of objects or individuals sharing a certain property; e.g. MC LATAB/ Eng. TABLE is Kind) and properties (see Lyons Reference Lyons1999; Bobyleva Reference Bobyleva2013: 60 inter alia). The semantic and pragmatic values of NPs are related to the lexical noun types involved and to their context of use. As signalled by Löbner (Reference Löbner2013), certain nominal semantic features contribute to the definiteness of NE whilst others such as number or count and mass are central for the marking of indefiniteness on NP. In this chapter, the lexical noun typology proposed by Löbner (Reference Löbner1985, Reference Bobyleva2013) will be resorted to but the count noun vs mass noun dichotomy, a component of individuation (Hopper & Thompson Reference Hopper and Thompson1980), will also be taken into account.

Löbner (Reference Löbner1985) distinguishes non-inherently unique nouns from inherently unique nouns. Amidst the first group of nouns, Löbner teases apart sortal nouns that classify objects (e.g. MC fler, Eng. flower), from relational nouns that describe “nouns as standing in a certain relation to others”; MC latab and Eng. table are sortal nouns, while MC ser and Eng. sister are relational nouns. Individual nouns such as MC soley (Eng. sun), or functional nouns such as MC levek (Eng. bishop) or MC sofer (Eng. driver), form a subcategory of relational nouns, are “inherently unique”.

The analysis of the identifiability and inclusiveness of NPs has given rise to a vast literature on the unicity and the familiarity of referents (for a summary of the discussion, see Lyons Reference Lyons1999). In this chapter, for all intents and purposes the identifiability of NPs will be defined through the semantic and pragmatic notions of individuation, definiteness, specificity and genericity, which are explained below (see also Bobyleva Reference Bobyleva2013: 9).

According to Hopper and Thompson (Reference Hopper and Thompson1980: 253), the individuation of a nominal entity is defined by a set of features including the count-mass distinction, number and animacy. Mufwene (Reference Mufwene, Muysken and Smith1986 inter alia) has shown that the features “individuated – non-individuated”, “definite – non-definite” and “specific – non-specific” overlap because specificity and definiteness presuppose individuation, but individuation entails neither. Heed will be paid to Mufwene’s word of caution, but the notion of individuation as defined by Hopper and Thompson will be retained to parse head nouns referring to count and animate entities.

Hawkins (Reference Hawkins1978) provides an extensive definition of definites in discourse. Definite NPs are identified through the immediate situation where they are used, through broader explicit situations, through immediate anaphoric use or through associative anaphora. Wiesinger (2017: 164–203) discusses at length the contexts of use of definites as defined by Hawkins (Reference Hawkins1978) and Himmelman (Reference Himmelmann, Haspelmath, König, Oesterreicher and Raible2001) inter alia. Wiesinger (2017) proposes a tripartite classification of definiteness: definiteness via deixis, via anaphora and marked through inferable contexts, which include restrictive contexts (i.e. through relative clauses for instance), and associative anaphora. Following Löbner (Reference Löbner2013: 75), I shall consider that the definiteness of NEs expresses either semantic uniqueness, independently of the condition of use of NE, or pragmatic uniqueness, conveyed by the conditions of use of NE, hence endophoric, deictic and associative definites.

According to Löbner (Reference Löbner1985: 279), semantic definiteness is fostered by nouns and concepts classified as either individuals (e.g. MC lalin, ‘moon’) or functional (e.g. MC levek ‘bishop’), while pragmatic definiteness is established through the use of sortal (MC fler ‘flower’, zanimo ‘animal’) or relational head nouns (MC matant ‘aunt’). Löbner (Reference Löbner1985) insists on the fact that relational nouns are not always count nouns.

Löbner (Reference Löbner2013: 79ff.) shows that in many languages definite determiners in NEs mark both pragmatic and semantic uniqueness. These definite determiners share deictic and anaphoric uses with demonstratives. However, only definite determiners may convey semantic uniqueness. While a definite determiner indicates that the NP it determines refers to an entity uniquely determined in a given context, an indefinite determiner signals that the referent of its NP is not given beforehand and is not uniquely determined. Indefinite determiners associate with singular count nouns, mass nouns and plural nouns to form indefinite NPs.

F. Schwarz (Reference Schwarz2013) distinguishes two types of definites: strong definites related to anaphoric uses and weak definites which map with “uniqueness-based situational uses”. Schwarz’s distinction offers an alternative to Löbner’s distinction between semantic and pragmatic definiteness. On the basis of Wespel’s analysis of definites in MC (Wespel Reference Wespel2008), Schwarz (Reference Schwarz2013) equates MC la with a strong definite article and considers bare NP in MC to be an instance of weak definite. The application of Schwarz’s distinction to IO Creoles needs to be further discussed.

In his seminal description of the Creole prototype, Bickerton poses that Creoles have two overt markers: a definite article to mark “presupposed-specific” NPs and an indefinite article for “asserted-specific” NPs, together with a zero marker for non-specific and generic NPs. Bobyleva (Reference Bobyleva2013: 73) and Wiesinger (2017: 145) consider that Bickerton’s definition of specificity refers to pragmatic specificity or to Givón’s pragmatic referentiality (Givón Reference Gívón1984). For Guillemin (Reference Guillemin2009: 27),

Von Heusinger (Reference von Heusinger2002: 248) notes that specificity was introduced in relation to indefinite NPs as an analogy to referentiality in definite NPs. Specificity is a feature of NP that cuts across both definites and indefinites, just like genericity (von Heusinger Reference von Heusinger2002: 252). Von Heusinger (Reference von Heusinger2002) supports the view that specificity is sentence-bound, whereas definiteness is discourse-bound. However, he favours the idea that specificity is referentially anchored in discourse and identifies at least four types of specificity, characterized in terms of variation in the extension of the scope of the NP. According to Wiesinger (2017: 142ff.), two perspectives on the notion of specificity may be identified: a semantic perspective which investigates the possible ambiguity of the scope of NP in discourse and a pragmatic perspective which focuses on the enhancement of the particular entity the hearer has in mind. Wiesinger (2017: 144) shows how semantic and pragmatic specificity cut across the definite/non-definite distinction.

13.3 Lexical Entities and Determiners in Nominal Expressions in IO Creoles

Since the 1970s, IO Creole grammars have been widely described (see Baker Reference Baker1972; Chaudenson Reference Chaudenson1974; Bollée Reference Bollée1977; Cellier Reference Cellier1985; Corne Reference Corne1977 and Staudacher-Valliamee Reference Staudacher-Valliamee2004 inter alia). More recently, various extensive analyses have been devoted to the syntax and semantics of IO Creole NPs. The IO Creoles exhibit similar lexicons where approximately 90% of the lexical items have been provided by the French lexifier, the rest coming from other languages such as Malagasy, West African Bantu languages and Tamil, considered to be substrate languages, and English (Chaudenson Reference Chaudenson1974). The lexicons of SC and MC are quite close; RC lexicon differs to a certain degree from those of SC and MC. RC shares an ancient stock of words with MC and SC but differs from these two languages in its modern vocabulary (Chaudenson Reference Chaudenson1974). According to Corne (Reference Corne1989: 271), although MC and SC have a cognancy rate of 95.7% on the Swadesh 100-word list, these languages seem to share only c. 70% of their total vocabulary. Corne (Reference Corne1989: 273) indicates that on the Swadesh 100-word list, the rate of cognancy between RC and MC is 66.4% and rises to 68.7% between RC and SC.

Déprez (Reference Déprez and Quer2001, Reference Déprez2007a: 263) notes that “French Lexifier Creoles (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.” This statement applies particularly well to IO Creole determiners, as seen in Table 13.1. A subset of these determiners excluding possessives will be studied in this chapter.

IO Creoles share a common pre-posed indefinite marker (inn/in/enn/en), a pre-posed demonstrative marker sa and a marker of plurality bann, which may associate with the demonstrative and the definite marker. RC is the only IO Creole to exhibit a pre-nominal definite singular marker lo as well as the post-nominal marker la shared with MC, and pre-nominal plural markers lé for the definite and dé for the indefinite together with bann. Besides, determinerless NPs are used in all IO Creoles.

Determiner NPs in the three languages compared are used in nearly the same syntactic and discoursal contexts as bare NPs. However, semantic features attached to Determiner NPs are partly different from those conveyed by bare NPs. In IO Creoles, NEs may be found in both argument and non-argument positions (Allesaib Reference Alleesaib2012; Syea Reference Syea2013: 58).

Most, if not all, of the studies of IO Creole NPs have been conducted within the realm of generative linguistics. In these studies, IO Creole nouns have been classified either under the twofold distinction of count and mass nouns or within the nominal classification developed by Löbner.

13.4 Definiteness in IO Creoles

The review of previous research shows that IO Creoles differ mainly in the marking of definiteness. Besides similar uses of overt determiners and bare NPs, it is expected that some differences will be identified, especially in the use of sa and -la in MC, RC and SC. Following the theoretical framework sketched in section 13.1 above, definiteness will be understood as the marking of pragmatic and semantic uniqueness of NEs., that is, uniqueness of N as bestowed by discoursal and situational circumstances of use, and kind reference principally (Löbner Reference Löbner1985, Reference Löbner2013).

13.5 Determiners and Bare NPs in IO Creoles

This section briefly recaps the marking of indefiniteness, specificity, genericity and number in the three Creoles compared (13.5.1) and discusses some differences in their use of determiners (13.5.2).

13.6 Discussion

Wiesinger (2017: 422–36) and Albers (Reference Albers2019: 108ff.) have aptly summarized and discussed the paths of grammaticalization of determiners propounded by Greenberg (Reference Greenberg, Ferguson, Greenberg and Moravcsik1978), Lehmann (Reference Lehmann2002) and Himmelmann (Reference Himmelmann, Haspelmath, König, Oesterreicher and Raible2001). In a nutshell, according to these authors, the first stage of development of the definite article/determiner is represented by the existence of a deictic marker in deictic and anaphoric contexts, where the referent may be identified through the immediate situation or in discourse. At the next stage, the determiner applies to inferable contexts where the referent may be identified. A further stage of evolution happens when the determiner marks “both definite determination and non-definite specific uses” (Greenberg Reference Greenberg, Ferguson, Greenberg and Moravcsik1978: 62). In the last stage, the determiner turns into a mere nominal marker.

MC, RC and SC seem to be at different stages along the same path of grammaticalization. The determiner system of SC and its marker sa illustrates a first stage of grammaticalization where a deictic marker is used in anaphoric contexts (see 2c, 4c or 6c for instance). Although RC presents social and regional dialectal variation in the use of pre-nominal markers, its determiner system lies in an intermediate position where pre-nominal definite lo- and demonstrative post-nominal -la are used in distinct specific contexts. The RC system seems to be in a state of flux. MC has developed a definite article -la which differs markedly from demonstrative sa … la. According to Wespel (Reference Wespel2008) strong definite marker la is specifically used with sortal lexical head nouns in deictic, endophoric and anaphoric discoursal contexts.

Whatever their genetic links (see Chaudenson Reference Chaudenson1974 and Baker & Corne Reference Baker and Corne1982), the determiner systems of IO Creoles exhibit important formal and functional similarities. The inventories of their overt markers are close and they resort to the same plural marker bann. The use of bare NPs and the placement of their markers are identical except for RC lo-. However, it should be noted on the one hand that -la does not have the same status in RC and MC and on the other hand that sa- does not have the same status and frequency of use in MC, SC and RC. The differential marking of nominal reference in MC, SC and RC seems to be due to differences in the degree of grammaticalization of determiners in these Creoles.

14.1 Introduction

Serial Verb Constructions (SVCs) in Kreol Seselwa (KS) have a long and disputed history in creolistics.Footnote ¹ Bickerton (Reference Bickerton1989) was the first to suggest that SVCs exist in KS and he takes their occurrence as an argument for his Language Bioprogram Hypothesis (Reference Bickerton1984). After the publication of his seminal paper, much of the following discussion revolved around the existence of SVCs in KS with a focus on their origin. While Seuren (Reference Seuren1990) and Corne, Coleman and Curnow (Reference Corne, Coleman, Curnow, Baker and Syea1996) argue that SVCs cannot be found in KS, Bickerton (Reference Bickerton1990, Reference 276Bickerton, Baker and Syea1996) maintains the position that SVCs are part of KS grammar. Even though the origin of KS serials is still being debated, most scholars nowadays assume that some types of SVCs exist in KS (e.g. Adone Reference Adone2012; Michaelis & Rosalie Reference Michaelis, Rosalie. and Michaelis2013; Adone et al. Reference Adone, Brück and Gabel2018).Footnote ²

Based on a detailed study of SVCs in KS (Gramatke Reference Gramatke2019), this chapter ascertains that SVCs are indeed part of KS grammar and presents a brief typology of serials in KS. Furthermore, this chapter analyses the syntactic nature of SVCs in KS by investigating claims first raised in Bickerton (Reference Bickerton1989, Reference Bickerton1990, Reference 276Bickerton, Baker and Syea1996). It shows that KS serials exhibit prototypical forms and functions while also exhibiting non-prototypical types, thereby taking a continuum approach to SVCs as suggested in Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018). Finally, it discusses the distribution and variation of SVCs in KS and traces their development in recent years.

14.2 Serial Verb Constructions

Following Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018), SVCs can be defined as structures that contain two or more verbsFootnote ³ that form one predicate and appear in a single clause without linking elements such as coordinators or subordinators. Verbs used in an SVC are independent verbs, that is, they can appear as the only predicate in another sentence and the two verbs do not have a complement relationship with each other. Serial verbs have the same configuration for tense, mood, aspect (TMA) and negation, and they may share arguments. They are conceptualized as one event, they exhibit one intonation contour and are typically uttered without a pause between the verbs (Aikhenvald Reference Aikhenvald2018: 1ff.). SVCs are present in numerous languages throughout the world with different typological profiles, yet they predominantly cannot be found in European languages (Aikhenvald Reference Aikhenvald2018: 1). Two examples of SVCs in Saramaccan, a Creole language spoken in Suriname, are given below, both taken from Veenstra and Muysken (Reference Veenstra, Muysken, Everaert and van Riemsdijk2017: 3916, 3906):

1. (1)

   A	kúle	gó.
   3sg	run	go
   ‘He ran away.’

1. (2)

   A	téi	dí	fáka	kóti	dí	beée
   3sg	take	det	knife	cut	det	bread
   ‘He cut the bread with the knife.’ [approx.]

There is a wide range of forms and functions of SVCs, which is why a prototype and continuum approach is adopted in Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018). For instance, subjects and objects are prototypically shared. However, examples of SVCs exist in which the object of the first verb simultaneously functions as the subject of second verb and are therefore classified as switch-subject or switch-function SVCs (Aikhenvald Reference Aikhenvald, Aikhenvald and Dixon2006: 14). Hence, Veenstra (Reference Veenstra1996: 85) concludes that “subject sharing is not an essential property of serial verb constructions” even though it is a prototypical feature.

According to Aikhenvald (Reference Aikhenvald2018: 55ff.), SVCs can be asymmetrical or symmetrical in their composition. In asymmetrical SVCs one verb comes from an open and unrestricted class, whereas the other verb is semantically or grammatically restricted. Therefore, Aikhenvald (Reference Aikhenvald2018: 81) speaks of a major verb which “can be considered the ‘head’ of the construction inasmuch as it determines its meaning” and a minor verb which adds “further specification”. Apart from their composition, SVCs can also be classified based on other formal properties.

Aikhenvald (Reference Aikhenvald2018: 92ff.) distinguishes between contiguous or non-contiguous SVCs depending on whether elements, such as adverbs or objects, can appear between the two verbs or not. Additionally, SVCs can be multi-word serials if the two verbs constitute separate grammatical/phonological words or one-word serials if the two verbs are combined into one word. Finally, SVCs can be concordant or non-concordant depending on whether grammatical categories such as TMA and pronominal subjects are marked on each verb or not.Footnote ⁴ Aikhenvald (Reference Aikhenvald2018: 100ff.) stresses that even though a defining criterion of SVCs are shared values, these values can indeed be overtly expressed more than once. Hence, TMA markers as well as pronominal subjects can be repeated on each verb (Aikhenvald Reference Aikhenvald2018: 102), a notion that was also proposed by Seuren (Reference Seuren1994) in his ‘Tacit Subject Condition’ which states that either no subject is present before the second verb, or at most, a subject pronoun may be repeated. In general, non-concordant SVCs are considered to be more prototypical than concordant SVCs (Aikhenvald Reference Aikhenvald, Aikhenvald and Dixon2006: 44; Veenstra & Muysken Reference Veenstra, Muysken, Everaert and van Riemsdijk2017: 3909). Finally, since argument sharing is more prototypical and single-word, non-contiguous SVCs are only attested very rarely (Aikhenvald Reference Aikhenvald2018: 97), it is assumed in this chapter that contiguous SVCs are more prototypical than non-contiguous SVCs.

In contrast to formal approaches, functional approaches to a typology of SVCs classify them according to the semantic content that they express. This classification, even if different terminology is used, can be found in many publications on SVCs, such as in Jansen, Koopman and Muysken (Reference Jansen, Koopman and Muysken.1978), Sebba (Reference Sebba1987), Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018) and Ansaldo (Reference Ansaldo and Brown2006), to name only a few. As Veenstra and Muysken (Reference Veenstra, Muysken, Everaert and van Riemsdijk2017) address all of the terminology used in the literature, their classification is taken as a basis in this chapter. An overview of the functional types investigated here can be found in Table 14.1.

14.3 SVCs in Kreol Seselwa

14.3.3 The ‘Quacking Duck’ Revisited

Based on the spoken as well as the written corpus, a total of 1,069 tokens identified and analysed as SVCs were found.Footnote ⁶ All of them contained at least two independent verbs without a complement relationship and linking elements. This can be shown exemplarily by the following four sentences, one of each data type:

1. (3)

   […]	mon	ti	war	latet	papa	pe	monte	desann […]
   […]	1sg	tns	see	head	papa	asp	ascend	descend […]
   ‘[…] I saw my dad’s head going up and down […]’ (Choppy Reference Choppy2012: 40) (written)

1. (4)

   epi	mami	anmennen	sorti	travay
   after	mother	bring	exit	work
   ‘After my mother brought this from work.’ (semi-spontaneous)

1. (5)

   Son	tas	i	tonbe	kraze
   prn	cup	pmFootnote 7	fall	break
   ‘His cup fell and broke.’ (elicitation)

1. (6)

   I’n	anmenn	liv	i’n	donn	nou
   prn.asp	bring	book	prn.asp	give	us
   ‘He has brought the book and given it to us.’ (judgement)

All of these examples contain multiple independent verbs which also appear on their own in other contexts. No overt subordinators or coordinators are present. In addition, they exhibit the same TMA as well as NEG setting. The subjects are shared, which is a prototypical but not a necessary feature of SVCs, and in example (6) the direct object is shared as well. Furthermore, no examples exhibit a pause or a rise or fall in intonation contour. This is verified by the Praat pictures for the semi-spontaneous data in (4) as well as the elicitation data in (5), as can be seen in Figures 14.1 and 14.2.

Figure 14.1 Intonation contour of example (4)

Figure 14.2 Intonation contour of example (5)

The monoclausality for the written data in example (3) is determined by the orthography without a comma and for the judgement data in (6) by a comment by one participant that this SVC should be written without a comma and is understood as one event. The verbs in these examples appear in their long form and, thus, no complement relation holds between the two verbs.Footnote ⁸ All of these aforementioned characteristics of the sentences conform to the defining criteria of SVCs. Therefore, it has to be concluded that SVCs are part of KS grammar.

Three strategies in KS can be observed that border on SVCs. The first of these is overt coordination or subordination, second, covert coordination, also called asyndetic coordination, and the third refers to verb complementation. The differentiation between the first of these and SVCs is in fact straightforward, as overt linking elements are present and, hence, these can be classified immediately as coordination/subordination structures. However, the second and third neighbouring phenomena are not as readily identified based on the lexical items present in the sentence.

Since no overt coordinator is present in asyndetic coordinations, other identificational criteria have to be chosen in order to delimit these structures from SVCs. These are evidenced in some of the generalizations that follow from the definition of SVCs above. First and foremost, SVCs necessarily have the same TMA and NEG configuration on all verbs since they are contained within the same clause and describe one event. As Bickerton (Reference 276Bickerton, Baker and Syea1996) shows, this is not necessarily the case for asyndetic coordinations. In KS, the constraint that in SVCs the same TMA configuration has to be present on both verbs is easily discernible, since TMA markers are usually overtly repeated on V₂, that is, all verbs are concordantly marked.Footnote ⁹ Furthermore, during the judgement task of this study, sentences were included that exhibited different TMA settings. Some of the participants explicitly stated that the marker has to be the same. Some accepted the different TMA setting; however, in these cases they repeated the sentence with comma intonation. Hence, a covert coordination structure was uttered instead of an SVC. In the sentences produced during the other tasks, only structures with the same TMA value surfaced that also fulfil all of the other criteria of SVCs. Hence, the same-TMA constraint for SVCs in KS can be confirmed.

A second distinguishing criterion between SVCs and covert coordinate constructions that follows from the definition of SVCs is the assumption that the actions represented by the two or more verbs in an SVC constitute one event. As Givón (Reference Givón, Heine and Traugott1991a) has shown, the idea and reality of what constitutes a single event is problematic. However, two observations surfaced during the judgement as well as elicitation data within this study. First, one participant remarked during the judgement task that a certain sentence has to be written with commas since it contains different actions rather than one event. The same can be stated based on the results of the sentence I pran laliny met dan delo tire ‘He takes the fishing rod, puts it in the water and pulls it out’, administered during the judgement task. The first and second verb pran and met were uttered without an overt coordinator or a comma intonation by the participants, whereas the third verb, tire, was explicitly coordinated, either by an overt coordinator or a clear intonation break. A participant remarked that the last action (tire) constitutes a separate event as one has to wait some time for the fish to bite.

The second observation concerning the generalization that SVCs are thought of as one event stems from the elicitation data. Without any instigation on the part of the interviewer, a participant produced the following utterance:

1. (7)

   Eh. Trwa	keksoz.	I’n	pran	baton,	i’n	zwe	tanbour,
   Eh. Three	things.	prn.asp	take	stick,	prn.asp	play	drums
   apre	i’n	pran	lanmen	met	lo	tanbour	pour
   after	prn.asp	take	hand	put	on	drum	for
   son	arete
   sound	stop
   ‘Eh, there are three things. He has taken the drumstick, he has played the drums, after that, he has taken the hand and put it on the drum for the sound to stop.’

As can be seen from the example, the participant announces three observations and he coordinates three clauses, the last of which includes two verbs (pran and met) which constitute an SVC. Since two verbs are combined together and taken together make up the ‘third thing’, it can be assumed that they are conceptualized as one event. Furthermore, there is a pause of 0.96 and 0.64 seconds respectively that occurs between the three clauses. Within the last clause, two verbs are present, both of which are uttered with a single intonation contour without a pause. Hence, it is evident that the third clause is monoclausal but contains two verbs, therefore constituting an SVC rather than a covert coordination structure.

Apart from overt and covert coordinations, a third phenomenon in KS can be observed that borders on SVCs but has to be seen as distinct from them. Based on the definition of SVCs given above, it becomes clear that no complement relationship should exist between V₁ and V₂ in the sense that the first verb licenses the second. Hence, the TMA configuration has to be the same, and within an SVC no combination of a finite and a non-finite verb can appear. As will be described in more detail below, TMA markers are usually repeated on V₂ in KS SVCs and, hence, their matching configuration and their status as finite verbs is immediately apparent. Further, if a verb is followed by a complement, such as an object, or a VP-complement, the short form of the verb is used, if they belong to the class of verbs with short-form alternations. This is in contrast to adjunction, in which the long form of the verb is used. In SVCs the long form is used (e.g. anmennen, instead of anmenn, or tonbe, instead of tonb; see examples 4 and 5), indicating that no complement relationship holds between the two verbs and their VPs. An example in which a complement relationship between the two verbs exists are structures that contain ale (‘go’) as the first verb, followed by a second one. These are very common in KS (Choppy, personal communication), and numerous examples of these were also found in the data corpus. In these structures, such as al aste, ‘go buy’, or al pran, ‘go take’, the short form of the verb is used and, thus, a complement relationship is assumed. Therefore, they were not considered to be SVCs.

Hence, it can be stated that a distinction can be observed in KS between clear cases of asyndetic covert coordinations with a pause and comma intonation, clear cases of subordination with the short verb form being used and clear cases of SVCs. However, structures were also found that were not as distinct and straightforward as the examples given above, for example in cases in which no pause was present but there was a slight rise or fall in intonation contour. So far, no studies have been carried out on how high or low a rise or fall has to be for it to indicate a coordination structure in contrast to an SVC. Studies such as Givón (Reference Givón, Heine and Traugott1991a, Reference Givón and Lefebvre1991b) investigate pause probability after the verbs. However, comma intonation and a rise or fall of contour has not received wide attention thus far. Gramatke (Reference Gramatke2019) concludes that it is best to treat SVCs and their neighbouring phenomena as a continuum, similar to the analysis of complex sentences in KS suggested by Michaelis (Reference Michaelis1994) with overlapping, non-distinct cases. This also conforms to Aikhenvald’s (Reference Aikhenvald2018: 5) scalar approach of prototypical and less prototypical SVCs. As coordination and complementation may be seen as opposing phenomena, SVCs seem to occupy a position in the middle between those phenomena, as suggested by Figure 14.3, taken from Gramatke (Reference Gramatke2019: 210).

Figure 14.3 Continua of coordination, SVCs and complementation

As was shown above, SVCs are evidently part of KS grammar based on criteria proposed in the literature, for instance by Aikhenvald (Reference Aikhenvald2018). The study found that Directional, Give, Take, Resultative and Open SVCs (see Table 14.1) exist in KS. Again, a prototype approach to the classification of SVCs in KS best captures their semantic properties. This can be illustrated with the help of Directional SVCs in KS. A prototypical Directional SVC is displayed in (8) and a less-prototypical Directional is exemplified in (9).Footnote ¹⁰

1. (8)

   I’n	taye	i’n	ale
   prn.asp	run	prn.asp	go
   ‘He runs away.’

1. (9)

   Myse	ti	monte	desann
   The man	tns	ascend	descend
   ‘The man went up and down/back and forth.’

In both examples, two verbs of motion are combined to express a direction. However, the second example has a figurative meaning that includes an aspectual notion since it describes a repetitive and frantic action, expressed in the English translation of ‘back and forth’.Footnote ¹¹

Degree serials were neither produced in the written or spoken data nor accepted in the acceptability judgements. Hence, it can be stated that Degree serials do not exist in KS. Finally, borderline cases are Say, Aspectual and Causative SVCs in the corpus. Structures that contain two verbs of saying and adhere to the definition of SVCs were only found in the written corpus and almost all of these were contained within the same source, namely Accouche (Reference Accouche and Bollée1976). A similar construction with a verb of saying plus poudir (nowadays a complementizer ‘that’) is very common in KS and appeared quite often in the corpus. However, the status and etymology of poudir is unclear, and therefore these structures were not considered SVCs or remnants of a Say serial in the study.

Likewise, Aspectual SVCs were very rare. Only three instances in which the second verb was fini were present, one in the written data, and two in the semi-spontaneous data. Furthermore, structures with V₁+fini were rejected by all of the participants during the acceptability judgements. Hence, it can be concluded that the presence of prototypical Aspectuals is very marginal in KS. Those that were present, however, can be shown to follow the definition of SVCs proposed here. The low productivity of this type of SVC may be due to the competing structures with fini + V which were not identified as SVCs in this study, since they are most likely instances of verbal complementation, which has been excluded for SVCs.

In terms of Causatives, a total of only two sentences were found in the corpus of produced speech that featured the word fer, ‘make’, one in the written data and one in the elicitation data. Their status as such is questionable due to the fact that Causatives, as defined by Veenstra and Muysken (Reference Veenstra, Muysken, Everaert and van Riemsdijk2017: 3921), include make as one of the verbs present that connect two other actions. Hence, it should be the second verb in a three (or more) verb SVC. Neither sentence fulfils the definition as proposed by Veenstra and Muysken (Reference Veenstra, Muysken, Everaert and van Riemsdijk2017: 3921). It has to be noted that others, such as Durie (Reference Durie, Alsina, Bresnan and Sells1997: 333f.), allow a causative SVC with two verbs, the first of which is make. In these cases the status of the following verb as a finite verb is crucial in order to exclude a complement relationship between V₁ and V₂, akin to English sentences such as, He made her go. To test for this option, sentences with fer were also included in the judgement task in which the second verb was marked for TMA. However, the results clearly indicate that the second verb after fer cannot be marked. Thus, it seems as if fer requires a non-finite verbal complement in these contexts in KS, which disqualifies this structure from SVCs based on the definition employed in this study.

14.4 Conclusion

In conclusion, it can be ascertained that SVCs are part of KS grammar. It was illustrated that structures exhibiting multiple verbs in written as well as oral KS speech adhere to the definition and criteria of SVCs as proposed by Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018). Thus, the structures have to be classified as SVCs: “If it walks like a duck, and swims like a duck, and quacks like a duck, and lays eggs like a duck … it’s a duck!” (Bickerton Reference Bickerton1990: 302). However, while prototypical formal as well as functional types of SVCs can be found, KS does not exhibit the full array of semantic types of SVCs, and some attested types exhibit non-prototypical meanings. In addition, it can be shown that most SVCs in KS are concordantly marked, that is, often the tense, mood or aspect marker as well as the subject pronoun is repeated before the second verb in KS serials. From a cross-linguistic perspective, this is a non-prototypical form of SVCs. Therefore, the continuum approach suggested by Aikhenvald (Reference Aikhenvald, Aikhenvald and Dixon2006, Reference Aikhenvald2018) arguably best accounts for the nature of SVCs in KS and can be extended to capture neighbouring syntactic phenomena as well. KS SVCs are best analysed as adjunction structures, firstly since they exhibit the long form of the verb and secondly because the second verb and its associated phrase are optional and adverbial in nature. Even though KS exhibits SVCs, it can be seen that this is a predominantly oral phenomenon and not particularly productive. For instance, recent written data either do not exhibit any SVCs or are restricted to the occurrence of certain semantic types. In oral speech, a difference in size between produced and accepted SVCs can be observed. Furthermore, the use of SVCs seems to decrease the younger a speaker is. This may be linked to formal education in KS as well as to increasing English influence and may possibly hint at an incipient loss of these structures in the language.

15.1 Introduction

Creoles, languages as any other, are not only young but very dynamic systems because they are in close contact with (several) other languages, some of them equipped with political and educational assets.Footnote * In the preface to the sixteenth edition of Roots of Language, Bickerton (Reference Bickerton2016: vii) explains that the differences that creoles show “have to do with the very different nature and extent of contacts between the participants involved”.

In this chapter, we show that creoles are nowadays driving language changes in a linguistic area. In particular, Cape Verdean Creole, which is in contact with Portuguese in Cape Verde, is contemporarily granting an emergent variety of Portuguese being structured in the country.

The study of this emergent variety of Cape Verdean Portuguese (henceforth, CVP) is a relatively recent research field (see, e.g., Alexandre Reference Alexandre, Oliveira and Antunes2018; Alexandre & Gonçalves Reference Alexandre, Gonçalves, Álvarez, Gonçalves and Avelar2018; Alexandre & Swolkien Reference Alexandre, Swolkien and Reutner2024; Cardoso Reference Cardoso2005; Jon-And Reference Jon-And2011; A. Lopes Reference Lopes2016, Reference Lopes, Álvarez, Gonçalves and Avelar2018a; F. Lopes Reference Lopes2017; Mouta Reference Mouta2019). The language contact scenario in Cape Verde, since its settlement in the late fifteenth century, led to the birth of a new contact language, the Cape Verdean Creole (henceforth, CV; or, as officially designated by the decree-law 8/2009, Kabuverdianu ‘Capeverdean’), and to a variety of Portuguese shaped by a process of second language acquisition (see Alexandre Reference Alexandre, Oliveira and Antunes2018). In fact, the Portuguese colony status of Cape Verde until 1975 grounded the idea that all Cape Verdeans spoke the European variety of Portuguese (henceforth, EP). Nevertheless, in the past decade, teachers, students, intellectuals and especially scholarsFootnote ¹ have noticed that young generations play a fundamental role in the acquisition of Portuguese in Cape Verde and, mainly, in framing the linguistic change and its spread in society. The varied linguistic input that children and youngsters are exposed to, mostly from teachers who also speak Portuguese as a second language, is responsible for this process.

Consequently, researchers have the responsibility to capture and explain the properties of a grammar that is similar, yet different, from the EP model. Therefore, our goal is twofold: to present the properties of reflexive constructions in CVP, and to promote an explanatory approach to the nature of these constructions, centred in language transfer and language universals.

Hence, in section 15.2, we put forward a sustained definition of CVP, rooted in the country’s linguistic ecology, and summarise the features of reflexive constructions both in EP and in CV. This comparative perspective will frame our research questions. Section 15.3 focuses on methodology, that is, characterisation of participants and the description of two experimental tasks that have been carried out in the study. Section 15.4 presents the data analysis and a discussion of the results. The chapter ends with conclusions from our findings, in section 15.5.

15.2 Defining ‘Cape Verdean Portuguese’

This section discusses the contemporary linguistic situation in Cape Verde, followed by a description of reflexive structures in both EP and CV, in an attempt to define CVP and present the research goal.

15.3 Methodology

This section characterises the participants of the study and describes the two experiments which were carried out.