Paper presented at the International Symposium on
Degrees of Restructuring in Creole languages,
Regensburg, 24-27 June 1998

Salikoko S. Mufwene
University of Chicago

1. Introduction

Like Mufwene (1986), Thomason (1997:73) observes that "identifying creoles by means of a laundry list of grammatical features has proved to be an elusive goal: there are no features that are exclusive to, nor universal in, languages generally thought to be creoles." Indeed those putative "creole features," which these new vernaculars share with isolating languages such as Chinese, Thai, and the Kwa languages of West Africa (in which they are actually more consistent), are helpful only when they are related to the particular social history of their developments. She is also in agreement with Mufwene (1991a) in stating that creoles vary among themselves in the way they share these features.

In addition, Thomason states that as a group creoles form a prototype category rather than a classical one (in which all the members of a category putatively share all the features associated with it). In the latter respect, I take issue with her. I argue that the kind of variation that obtains among creoles does not support her claim that there are some which are more prototypical than others; rather, they differ among themselves on the Wittgensteinian family resemblance model (Mufwene 1986). Accidents of history have led us to know more about some creoles than about some others. However, we should be cautious about conflating this first-specimen view (so far justified in our case only, and disputably, by which creoles linguists investigated first and know the most about) with the best-exemplar conception of prototypes.¹ My argument is primarily against the best-exemplar approach which Thomason's discussion (1997:78-80) suggests, although I also question the first-specimen interpretation, in the sense of which creoles developed first. I advocate this contra-prototype stand despite our practice in the literature to identify Saramaccan, for instance, as a prototypic or radical creole, i.e., one in which we are likely to find the greatest number of putative deviations from its lexifiers, and, perhaps in their best forms and combinations, structural features that we associate with creoles.

I argue in Mufwene (1998a) that there are no particular linguistic evolutionary processes likely to yield (prototypic) creoles; these are produced by the same restructuring processes that bring about change in any language, as stated by Hock & Joseph (1996:15). Although we have identified particular ecological settings of language contact that have yielded very divergent outputs from the restructuring of some colonial languages (most typically, English, French, or Portuguese), we are still far from knowing what a prototypic creole is, either in the sense of 'the first specimen to have evolved' or in the sense of 'the best exemplar'.² Note that McWhorter's (in press) attempt to articulate three structural features which taken together may identify creoles as a structural type of languages is defeated in part by exceptions that he points out for each of the features within the putatively prototypic creoles themselves. Aside from the fact that most creoles lack this combination of features, there is typically a "prototypic creole" which lacks one of them. His claim is also weakened by the presence of this combination of features in some non-creole languages which he cites.³

I also contend that linguists' self-licence to go around the world baptizing some vernaculars "creoles," when in some cases their speakers do not even know the word creole, let alone how it is used in linguistics, is questionable. This behavior is part of what I have described in Mufwene (1996a, 1997a, 1998a, in press-a) as the disfranchising act by which some vernaculars are marginalized from other normal, natural developments of their lexifiers. I argue that the main reason for this seemingly a-prioristic attitude to creoles is our reluctance to re-examine whether those other new vernaculars that have been identified as dialects of their lexifiers and have been kept within the "franchise" have not developed by the same processes that produced those identified as creoles.⁴

A close examination of Thomason's (1997) position, which is consistent with Thomason & Kaufman (1988), reveals some disputable assumptions about what distinguish creoles from other languages from the point of view of language evolution:

1) "Prototypical creoles develop in a contact situation involving more than two groups of speakers" (78). History suggests that this criterion applies equally well to the development of, for instance, North American varieties of English, in settings where the European founding populations, a large proportion of whom were indentured servants (Tate 1965, Kulikoff 1986), had diverse nationalities and spoke different languages. The main differences between the development of creoles and that of North American English vernaculars and the like lie, on the one hand, in the multitude of languages that came in contact with the lexifier and, on the other hand, in the structural differences among the languages in contact. I am purposely leaving alone here several factors, such as social integration in the community of native speakers, which had a greater impact on the restructuring process and on the divergence of white and black vernaculars in North America than the often-invoked disproportion of native and non-native speakers in the contact setting (Mufwene in press-a).

2) "The grammars of creole languages may be accounted for in large part as cross-language compromises among the grammars of their creators' native languages" (1997:78). Such an same explanation has also been provided for the development of Romance languages, more recently by Posner (1996). If we dismiss the assumption of contact of multitude of languages (which we have typically floated when dealing, for instance, with Berbice Dutch), the same kind of explanation also applies to, for instance, the development of Irish English, which bears strong Celtic substrate influence. Such counter-evidence suggests that in the development of new language varieties the restructuring formula involves the same general competition-and-selection process advocated in Mufwene (1996a, 1996b, and later publications), subject to relevant ethnolinguistic-ecological factors. What is especially significant in this observation which applies to both creole and non-creole vernaculars is the allowance it makes for substrate languages to influence the restructuring of a language.⁵

3) Creoles have developed "by some historical process other than normal transmission" (75). Thomason characterizes "normal transmission" as "complete and successful transmissions, by native speakers to child or adult learners, of an entire language, i.e., a complex interlocking set of phonological, morphological, syntactic, semantic, and lexical systems" (74). Unfortunately, she does not take into account the fact that language is a communal construct which no native or fluent speaker possesses entirely. Therefore no native speaker acquires his/her language completely. Nor does she take into account the fact that ecology determines what aspects of a language are transmitted to meet the relevant communicative needs of the next group or generation of speakers.

Recall that unlike in the scholastic setting, nobody teaches others the system of their language; every learner acquires (some of) the vocabulary to which they have been exposed and infers a system that enables them to communicate like those speakers with whom they have interacted. Hagège (1993) reminds us that restructuring, however minimal, takes place in all cases of spontaneous language transmission, from one generation of speakers to another, even in communities to which a language variety is native. Lass (1997:112) leads more or less to the same interpretation of language transmission, with his notion of "imperfect replication." According to him, "systems with true histories develop them because errors creep into the replication process (...) In language transmission the errors are linguistic innovations [in the traditional sense of the term in historical linguistics], which may be selected (...) by the speech-community."

Imperfect replication is indeed one of the factors that bring about "internally-motivated language change," if there is any reason at all for sustaining the distinction between "internally" and "externally-motivated change." I argue that the history of the development of creoles provides no evidence of broken language transmission, only evidence of cases where feature replication was more imperfect and restructuring was more rapid and extensive than in other communities involving primarily inter-idiolectal contacts and little or no population contacts. Even in contact communities such as 17th- and 18th-century Surinam, language seems to have been transmitted normally from one group of speakers to another (Mufwene 1998a), in conditions where the local vernacular was of course diffuse (in the sense of LePage and Tabouret-Keller 1985) and changing gradually (Arends 1989).

Since the restructuring formula is apparently the same-and speed and extent of restructuring vary according to diverse ecological factors-it seems arbitrary to assume that creole vernaculars have developed by diachronic processes different from normal language change. What we have been reluctant to deal with is the continua of outcomes of restructuring from those identified as creoles to those identified as new dialects of the same lexifiers. One may of course be tempted to think of such continua on the model of the color spectrum, in which subareas of hue saturation are normally broken into different categories articulated around some perceptual foci. That is, although such categories have fuzzy boundaries, they are phenomena justified by their respective perceptual prototypes.

The case of creoles is actually quite different, independently of the fact that we are dealing here not with a multitude of categories that fade into each other but with two basic categories we have wanted to have for reasons we have not explained satisfactorily, viz., creole vs. non-creole vernaculars, at most three if we allow the disputable category of semi-creoles (Mufwene 1997b). Aside from the fact that creoles have no structural prototypes, an important problem with distinguishing creoles from non-creoles on structural grounds lies in the fact that creoles do not share their features universally nor exclusively (Mufwene 1986, Thomason 1997). Research of the past decade or so has also made it more and more obvious that a lot of the features associated with them have been selected, though not preserved intact, from the nonstandard varieties of their lexifiers from which they developed or from the substrate languages that the latter came in contact with. Thus, creoles are structurally related both to their lexifiers and to their substrate languages, an observation that has led us to characterize them quite correctly as "mixed languages" but incorrectly as lacking genetic parentage (Thomason & Kaufman 1988, disputed in Mufwene 1998a).

Unfortunately, this characterization has also disfranchised them, because we have typically also ignored Hjelmslev's (1938) observation that all languages are mixed to some extent, that the Romance languages started as mixed languages too, that English itself has mixed language characteristics.⁶ We have ignored that there is a continuum of structural mixedness and non-monolithicity which may be identified in any language, regardless of its developmental history (Mufwene 1992). In the absence of creole prototypes (pace Thomason 1997 and McWhorter, in press), we have not made it clear where the boundary, however fuzzy, lies in degree of mixedness, nor whether the nature of mixedness changes necessarily depending on whether the varieties in contact are considered dialects of the same language (which may be typologically dissimilar) or separate languages (which may be typologically similar). Even if we ignore the ideological, rather than structural, foundation of the language/dialect distinction, the following question remains: do the competition and selection which lead to restructuring operate differently in dialect contact than in language contact? Trudgill (1986) suggests just the opposite, invoking for dialect contact the same kinds of markedness considerations as Mufwene (1996b) does for the development of creoles.

Contrary to Thomason's assumption that some typological options are universally marked relative to others, note that no speaker has access to all options available on any parameter around the world; one's choices are usually constrained by what one has access to. Hence, we can also argue non-trivially that all speakers select options unmarked to them, and what may be marked in one setting of language transmission may be unmarked in another.⁷ The trick, as expressed in Mufwene (1991a, 1996b), is to identify which factors determine markedness values and what role they apply in the evolution of any language under specific ecological conditions.

Our adherence to the distinction between creole and non-creole vernaculars as if it were structurally based has been fostered also by the above-disputed assumption that creoles are byproducts of language contact but other language varieties are not. A correlate of this in historical linguistics is the distinction between internally and externally-motivated changes to which I also alluded above. Thomason (1997:86) states that "a considerable amount of contact-induced language change can occur without disrupting normal transmission of a language." However, she does not show-not any more than do Thomason & Kaufman (1988) do-what distinguishes normal, internally-motivated language change structurally from externally-motivated language change, nor what light this long-standing distinction in linguistics sheds on language evolution, i.e., the long-term change that the structural and/or pragmatic systems of a language may undergo (Mufwene 1998b), which I hope we seek to understand even in studying creoles.

As expressed in Mufwene (1996b, 1998a, 1998b), internal variation within a language qua species has a lot to do with what has been identified as internally-motivated language change. If one believes in accommodation theory and in the role of "linguistic missionaries" in spreading features from one network of communication to another and thereby causing change, then the basic explanation is that contact among idiolects is the fundamental cause of language change. It is at this level of interaction that competition and selection of features take place in the system of a language qua species. Where contact among languages is also involved, the pool of features in competition becomes larger, but the mechanisms of competition and selection remain the same, operating still at the level of interacting individuals before impacting the speech or language community. Once the distinction between "internally-motivated" and "externally-motivated" language change is dropped, grounds for distinguishing between creole and non-creole vernaculars on criteria other than sociohistorical should also vanish.

The option of investigating the development of creole vernaculars with the goal of learning more about language evolution seems evaded by statements such as the following by Baker (1997:96):

in most contact situations which resulted in the emergence of pidgins or creoles, the real if unconscious aim of most of the participants was the creation of a medium for interethnic communication (MIC). By that I mean, with specific reference to slave plantation societies, that slaves did not aspire to acquire the language of the plantation owner as such. Their aim was to communicate, particularly with their fellow workers. The most readily available lexical source for the MIC among a multilingual workforce was that of the language to which they were all exposed in the workplace, that of the plantation owner. (It is no accident that the lexical items from languages of slaves which became established in creoles relate overwhelmingly to activities associated with slaves' free time.) If one abandons the concept of T[arget]L[anguage], one can begin to see that they might have created the MIC they needed, not only by drawing on the range of available resources but also by innovating (...)

With words such as aim, create, and aspire, the above quotation suggests surprisingly that, unlike in other cases of language evolution, the slaves who developed creoles could have deliberately been engaged in this process. While Thomason & Kaufman (1988), whose position Baker adopts, claim that there simply was no target for the slaves to aim at-which is incorrect in the first place (see below)-this passage claims that the slaves were more interested in developing a MIC than in learning anything that was already in place or promised to be useful under the circumstances of multi-ethnolinguistic contact. This characterization of the situation is more mistaken than that claimed by Thomason & Kaufman.

Note that the notion of target in settings where creoles developed need not have a different interpretation from its meaning in other language acquisition settings, where a language is assumed to have been transmitted and acquired "normally" and "perfectly" from one group to another. In all such settings, one must deal with interindividual variation and some heterogeneity of input, a state of affairs from which generative grammar derives evidence for innateness and creativity with respect to child language development.

Regarding the suggestion that creoles were deliberately created by the slaves, let us note non-trivially that in natural, non-scholastic cases of language transmission, the development of any communicative competence is not a planned activity; it is a byproduct of efforts to communicate. Things must not have been different in settings were creoles developed, as the populations were under pragmatic pressures to establish communication and had a vague idea who spoke the lexifier, the target language that was selected by the contact ecology as an MIC, fluently and who did not, especially among the Europeans themselves. One must also remember that as rapidly as the plantation colonies were peopled, the whole process was still gradual, with the early homestead communities, characterized by intimate interactions and regular attempts to communicate between speakers and learners of the lexifier.

Even by the time the plantation communities developed and non-Europeans became the majorities, language was still transmitted normally from one group to another, from the creole or seasoned slaves to the bozal slaves, regardless of the structural variation in the target. The difference between those settings identified misguidedly in the literature as involving no contact versus those identified as involving contact really lies in how many systems competed with the lexifier and how they differed structurally from each other (Mufwene 1998a). One may also observe that the contact setting made language boundaries less rigid and the lexifier more osmotic, allowing more influence from the substrate systems than in other settings of language transmission where there is more interaction with a majority of fluent speakers of a less diffuse target language.⁸ One needs only check the literature on code-mixing to notice why this is a plausible explanation. (I return to these observations below.)

There is no doubt that the lingua francas, if not the vernaculars spoken by the founder populations, i.e., those who arrived earlier, were less focused (in the sense of LePage & Tabouret-Keller 1985) than may be expected in those settings involving no contact of populations and minimal restructuring of the kind described by Hagège (1993). However, should we equate lack of focusing with absence of a target? Is this not just a matter of degree, like almost anything else. Is the Chomskyan creationist view on language acquisition not also predicated partly on chaotic input? Isn't this a correlate of the notion of idiolect in linguistics, viz., that different speakers have internalized (slightly) different grammars and therefore present different kinds of inputs to whoever is acquiring language from interacting with them? Isn't this what led Hagège (1993) to observe correctly that the transmission of a language from one generation to the next is in part inheritance and in part creation, which explains "internally-motivated" change in the first place?⁹

Arguing, as Baker does, that those who developed creoles did "did not aspire to acquire the language of the plantation owner as such" because they wanted to communicate among themselves raises interesting questions about similar situations, such as in Europe, where foreign workers speaking the same vernaculars and living together wound up developing language varieties which have been claimed to share some structural features with creoles. Is the formation of foreign workers' varieties deliberate too? Does it reflect some sort of defiance or lack of interest in speaking like native speakers. Effects of the post-"critical period' in language development set aside, don't such varieties reflect the complicity of ecological factors in not providing learners of the target with optimal interactive settings for language acquisition? Is there any particular reason why we should think that the economically or politically dominant language on a plantation could not be a target for the slaves and thereby function as an MIC? Is the notion of a target in plantation contact settings really in conflict with a scenario where the African languages would have disappeared not quickly but gradually? Did the extended coexistence of some substrate languages with the lexifier really preclude targeting the lexifier as an MIC? Why have similar situations not prevented the entrenchment of European languages as lingua francas in former European colonies? As a matter of fact, these varieties have generally indigenized, reflecting both substrate influence during usage of these lingua francas (Mufwene, in press-b) and the fact that languages are normally adapted to the communicative needs of those who adopt them. Interestingly in this case, the association of these colonial languages with domination and/or oppression did not prevent its adoption by the dominated populations.

Baker would have probably been better off arguing that in the settings in which creoles developed, language boundaries were less rigid than they may be in other "non-contact" communities and there was thus more of what Chaudenson (1992) describes as "osmosis," which, as noted above, allows mixing of systems and therefore substrate influence in the lexifier. Research on code-mixing is powerful proof of osmosis, although, pace Myers-Scotton (1997), I doubt that code-mixing or code-switching itself, as a phenomenon associated with bilingualism, sheds significant light on the development of creoles other than in demonstrating that structures of languages can indeed be mixed in all sorts of ways.Rather, imperfect replication as a principle applicable to all cases of language acquisition and substrate influence account well for their development.¹⁰

As explained in Mufwene (1991b), the main difference between child language acquisition and second language acquisition lies in the addition of features from languages previously spoken by the learners to the pool of competing features. One need not invoke code-mixing or switching (in the sense of typically associated with these terms in the bilingualism literature) in order to develop a new language variety in the communication process. Otherwise, the same principles invoked in the competition-and-selection approach (Mufwene 1996b), drawing from the best of universalist, substratist, and superstratist hypotheses, complement Thomason & Kaufman's basic "imperfect learning" explanation, extended here to all cases of language acquisition

The literature on the development of creoles has focused on language contact more from the perspective of communities being in contact with each other, than from the point of view of interacting individuals, at whose level actual contact takes place. It is really the tacit negotiations and innovations at the inter-idiolectal level of interaction which ultimately produce the phenomena that have concerned us at the language level, more or less like biological selection which, operating at the level of individual organisms first, ultimately affects a species. In settings where large proportions of speakers do not have normal monolingual fluency in the language they intend to speak, language boundaries become less and less rigid and the degree of osmosis is accordingly increased, an ecological factor that favors interference, which is so central in Thomason & Kaufman's (1988) account of the development of creoles. This leads to innovations in its traditional sense in historical linguistics, viz., new patterns or structures not attested previously in the relevant language (variety).

Chaudenson (in press) captures the above idea by speaking of relaxation of metropolitan norms in colonial communities both among Europeans and among non-Europeans. Like in other settings, children born to a contact community select and reproduce with minor modifications systems used by others, per Hagège's account of language acquisition and change, and thereby contribute to the development of the new vernaculars.

Thus, the following other questions arise about Baker's statement: Where does the difference lie between what Thomason characterizes as "ordinary" language change and his observation that the slaves created the MICs "not only by drawing on the range of available resources but also by innovating?" Isn't this all part of the traditional language change as discussed in historical linguistics? Also, is there any fundamental difference between targeting a European colonial lexifier and taking most of a creole's lexicon from the same language? Can such a massive selection of the lexicon from a particular language be distinguished from identifying the same as a target system? Would the vocabulary have been learned alone without the concurrent structural and pragmatic constraints on their usage, regardless of the ensuing restructuring? Were Thomason & Kaufman (1988) so wrong in invoking imperfect language learning, except in making it peculiar to the development of "contact varieties?"

To be sure, most of us have disputed earlier Bickertonian UG-based innovations in the sense of new creations ex-nihilo (Mufwene 1996b). Much of the valid efforts of superstratists and substratists have indeed consisted in showing that many of the so-called innovations stemmed from models available in either the lexifier (typically a set-theory union of nonstandard varieties) or from some of the substrate languages. Consistent with Lass's (1997) "imperfect replication," Boretzky (1993) shows clearly why substrate influence need not have been preserved unmodified in creoles any more than in other languages, just as Chaudenson (especially 1989, 1992) shows that creoles often pushed to their logical conclusions evolutionary tendencies already observable in the lexifier itself. This is one of the thrusts of his notion of "français zéro" as a set-theory union of varieties of French, from which structural variants were selected into creoles and often adapted to new communicative functions, under specific substrate influences. I cannot help thinking that the development of creoles was a byproduct of acquiring a diffuse but nonetheless targeted system with relatively greater ecology-prompted restructuring than in less heterogeneous and more focused settings of language transmission. Thus, as also observed by Hagège (1993), the difference between the development of creole and non-creole languages lies not in the structural processes that produced them but in the outcomes of the same processes. Mufwene (1996c) tries to account for this variation in outcomes in terms of ecological factors affecting restructuring.¹¹

Several things in the history of the development of creoles, including much of the above discussion, lead to the conclusion that these vernaculars are socially disfranchised dialects of their lexifiers, especially since dialects of the same language need not be mutually intelligible. As also recognized by Thomason (1997:86), "the dividing lines between ordinary languages and contact languages [if they are non-ordinary] (...) are as hard to find as the dividing lines between dialects and languages." We know well that the latter distinction serves no structural purpose, because we do not need different ways, nor techniques, to discuss characteristics of dialects than to describe features of languages, nor to explain the development of dialects than to account for the formation of languages. As some recent studies have shown (Mufwene 1996c included), grammaticization processes that have taken place in the development of creoles are not different in kind from those that have taken place in non-creole languages.¹² The morphemes or constructions which have been idiomatized for specific grammatical functions start from models available in some source, typically in the lexifier. Influences that bear on the grammaticization process may come from within the lexifier or from outside the lexifier, while the whole process is constrained by other grammatical conventions in place in the system.

For instance, the chances of going to becoming a full syntactic auxiliary verb for FUTURE depends in part on whether the variety in which it is used allows copula-less predicates and/or on whether or not going is still identified as a progressive form in the relevant variety. This accounts for why going to in Standard English will have different syntactic distributional properties from gon in AAVE and a go in Jamaican Creole. Likewise, the use of self in compound pronominal forms depends largely on how possessive constructions behave in the relevant varieties of English. Thus, there is no requirement to combine it with a possessive pronoun in Gullah and most English creoles because they have no possessive pronouns and the same invariant personal pronouns are typically preposed to the possessive noun. Thus one finds mi/yu buk 'my/your book', mi/yu own 'my/your own, mine/yours', and misef/yusef 'myself/yourself'.

Understanding much of the position I present here depends on rejecting the common assumption that a language is an organism, and starting thinking of it as a species, which is internally variable, and a complex adaptive system (Mufwene 1996b, 1998b). Such a change of perspective gives more sense to the competition-and-selection approach which I advocate and to such ethnographic notions as mutual accommodation, uneven language change in a population, or decreolization for that matter (wrongly or justifiably). Space limitations prevent getting into all that here.

After all the above arguments against treating the development of creole vernaculars as non-ordinary linguistic evolutions, I should perhaps explain why I consider myself a creolist and continue to distinguish between creole and non-creole vernaculars. My position does not contradict this professional specialization of mine. Chaudenson (1989, 1992) justifies the identification of creoles quite adequately as a group of vernaculars whose developments are similar especially in their temporal and geographical positions, viz., in tropical colonies settled by Europeans practicing slave-based economy from the 17th to the 19th centuries.¹³ The lexifiers are typically nonstandard varieties of European languages.

As explained in Mufwene (1997b), in several of these colonies there are parts of the populations which have been identified as creoles and creole vernaculars have typically been associated with parts of these creole populations, viz., those who do not descend entirely from European ancestors. Thomason (1997) joins me in acknowledging that these languages are not typologically unique, nor do they share universally even some of the features associated with them. If we had to interpret history faithfully, some of the languages that have retained our attention would not probably fit in this category, because their speakers, or those who disfranchised them, have not called them creoles, for instance Gullah. This could have been the case for Jamaican Creole, typically identified as patois, although I am sure it falls in the subset that Thomason identifies questionably as "prototypical."¹⁴

However, there is nothing wrong with delineating a group of languages on sociohistorical grounds and making them the subject matter of one's research, just as one may be interested in postcolonial discourse or in precolonial communication. This position does not presuppose that creoles, as identified in the social histories of some polities, share any more structural features than, say, the Germanic languages do among themselves. One thing worth remembering for creoles is that as a group they are not genetically related among themselves (although those with the same lexifier certainly are). If they happen to share some features, grouping them together arouses more academic interest, as it has justifiably for the past century and a half, especially over the past three decades. We would like to learn what the shared features reflect from the different sociohistorical ecologies of their developments. There is plenty to learn from them about language universals, Universal Grammar qua biological endowment for language, and about language evolution (characterized in Mufwene 1998b as "long-term change that a language qua species undergoes" in the direction of more, or less, complexity, diversification into other varieties, etc.).

On the other hand, we need not assume more than is justified by the facts. Creoles vary among themselves; and the nature of both their cross-systemic variation and their structural similarities does not warrant lumping them in a prototype kind of category. So far we have identified neither the best exemplar of the sociohistorical conditions nor of the few creoles that possess exclusively the combination of structural features associated with the category, pace McWhorter (in press). All there is so far is family resemblance, typical of any grouping of languages, even those that are genetically related and/or typologically related, for which we have sought to identify no prototypes.¹⁵

The problem arises when one starts claiming some structural features to be peculiarly creole, pretending that there is a yet-to-be-articulated optimal combination of such features that produces a creole, otherwise one has a semi-creole or something else. Until good evidence is adduced, we must simply recognize that some of our working assumptions are social but have not yet been validated structurally. While there are sociohistorical reasons for isolating creoles in a separate category-undoubtedly because we hope to learn something new or special about language (as when we study any other language, for that matter)- we still have not produced that particular battery of structural features that justify treating them as a particular structural type of languages. We do not even know what features are the most critical for a vernacular to be a creole, which makes it more difficult to determine, for whatever purpose, when a variety qualifiers as a semi-creole or as a dialect of its lexifier, a position for which I can draw support from Thomason (1997) too.

One thing I cannot help noticing is that we have often evaded the question of what the study of creoles contributes to understanding language. For instance, what does the study the development of creoles contribute to research on language evolution? What does the study of creoles' structural features, especially how they tend to cluster, tell us about language typology and universals? What does the study of the particular features which they seldom do without reveal about Universal Grammar, for instance, in regard to that core of features that a language cannot do without, regardless of how extensively it is restructured? What are the parameters or paths of restructuring that they validate or question in respect to grammaticization? Many of the tough questions that we face in the practice of our profession should prompt us not only to ask ourselves what we creolists have done wrong but sometimes also what the linguistic establishment may not have understood, simply because they were not confronted with the questions we face (Mufwene 1996c).

The above discussion suggests that the verb creolize and its nominalization creolization mean no particular kind of structural diachronic process, no special kind of restructuring, only the normal kinds of linguistic evolutionary processes observable in various combinations and in various degrees elsewhere; they have simply been branded with special social values in this case (Mufwene 1997a). In the spirit of Mufwene (1997b) these terms could also be dispensed with, unless we assign them interpretations that are consistent with the sociohistorical characterization presented above (Mufwene 1986, Chaudenson 1989, 1992, in press). Perhaps only ideologically, we could thus focus more of the kinds of questions formulated above and contribute more adequately to the endeavor to understand language, how it works and how it evolves over time. I maintain that creoles could also be treated as dialects of their lexifiers, at least where their native speakers think so. However, whether they are treated as dialects of their lexifiers or as separate languages should not really concern us that much, because this distinction changes nothing to the nature of structural questions we address, certainly not about the kind of restructuring that produced them nor about their undeniable genetic ties with their lexifiers.

1. In one of her seminal papers, Rosch (1977) reminds us of this ambiguity in the usage of prototype. We should remember not to confuse them, especially since neither implies the other.

2. Given the intimate nature of social interactions in the early stages (15 to 50 years) of the contact communities where creoles have also evolved in the New World and Indian Ocean, it has been rather illusive to determine which creole is the oldest and whether it, rather than any other, deviates the most from its lexifier. Even creoles such as Sranan, on which there seems to be the oldest textual evidence, are hard to document before the early 18th century. We cannot tell whether Sranan or Saramaccan, which is apparently among the most different structurally from English, started before Haitian Creole, which is likewise among the most structurally different from French. (Surinam and Haiti, where these vernaculars are spoken, are not among the oldest English and French colonies, respectively; and both are marked by departures of metropolitan speakers of the lexifier, quite early in the same of Surinam.) As often as Guinea Coast Creole English has been invoked as maybe the ultimate source of New World English creoles, the still scant, indirect, and disputable documentation of its widespread existence (Huber 1996) dates only from the early 18th century. Hancock (1986) presents no documentation of the varieties spoken in the mixed marriages of European lanÁados and African women, especially by their children. It is possible that next to regular colonial varieties spoken in these households, pidgin varieties were spoken by the grumettoes, who need not have communicated with the other Africans in the European language. Besides, the nature of contacts between the lançados and the grumettoes does not make such an early development of pidgins the only possibility. It must have taken several decades, if not more than a century since the establishment of trade forts on the Western coast of Africa, before the European languages were widely targeted as trade lingua francas and developed into pidgins. The evidence of Dillard's (1972:142) West African Pidgin English, which he then claimed to be the ancestor of African-American English, is also from the early 18th century.

3. The irony of the exercise itself lies in the fact that no non-creole language, or group of languages, has ever been defined structurally. Even genetic classifications such as Germanic or Romance have not been defined on structural grounds; their classification was lexically motivated initially. Attempts to characterize Bantu languages typologically have not excluded the inclusion of other languages in such structural classifications. Besides, within the Bantu group, such typological generalizations are often disjunctive, and they display a significant amount of family resemblance patterns. Since genetic connections between creoles lexified by different languages are likely to be very remote, if valid at all, one wonders what rewards are to be reaped from McWhorter's endeavor. Where creoles are related genetically, because they descend from the same lexifier, efforts to identify prototypes among them are similar to trying to determine which offspring in a family are more prototypical of it while there are apparently no combinations of genes or of morphological features that make an individual the best exemplar of that family, even if there are features shared by most members of the family. Note that typicality is different from, and more inclusive than, prototypicality. McWhorter's exercise is even more futile if his goal is to "vindicate a typological class" and all he can show is a disputable set of a handful of "prototypical creoles." I will leave alone the general problem with global, as opposed to parametric, typology.

4. This is simply an invitation to turn things around and ask ourselves whether contact has played no role at all in putatively "normal" cases of language change. If it has-at the level of varying and competing idiolects, as I argue below-how qualitatively different is the kind of contact overlooked in "normal" language change from that which has produced vernaculars called creoles?

5. In clarifying his position on the development of French creoles as similar to that of new French dialects in the New World, Chaudenson (in press) proposes a similar explanation, consistent with his earlier work for that matter, viz., the substrate languages are the additional factor that made creoles structurally different from their colonial non-creole congeners such as Québécois French or St. Barths's French patois. (I discuss the varying folk usage of the term patois below.)

6. As argued in Mufwene (1998a), this is no reason for jumping to the mistaken conclusion that the Romance languages and English developed by a putative structural process of creolization. The point here is simply that there are similar evolutionary processes that have resulted in the varieties identified as creoles in some cases and as non-creoles in others, on the basis of sociohistorical factors. We should not overlook those similarities if we stick to the position that creoles have developed by processes different from normal language change.

7. The assumption of a universal, ecology-independent scale of markedness leads to the following question: why would some speakers prefer marked options instead of following the natural principle of least effort? The model of markedness I have advocated accounts for reversals of markedness values, as suggested by choices made by speakers, by highlighting changes in the ecology.

8. I speak of "majority" tongue in cheek here. Cases involving the development of African-American vernacular English and foreign workers' language varieties in Europe show clearly that integration within the native-speaking community is a more critical factor than the demographic disproportion, on which the literature on the development of creoles has capitalized (Mufwene, in press-a).

9. My comments are not meant to deny that a language variety may be formed deliberately. One such language variety is described by Childs (1997) about Tsotsitaal and Isicamtho in South Africa. Note, however, that in this case the speakers had another vernacular in common from which they wanted to distinguish their speech patterns.

10. The kind of bilingual ecology presupposed by Myers-Scotton's ahistorical approach did not obtain as a general rule in settings where creoles have developed, although there were unavoidably cases in the New World and the Indian Ocean in which Africans speaking a common language were on the same plantation. However, during the development of creoles, the earlier slaves who "seasoned" the newcomers and therefore served as their linguistic models did not necessarily share a language with the latter. Therefore their verbal interactions did not encourage codeswitching. Given strong pressures on the newcomers to communicate with the locals before they had full command of the local vernaculars, interlanguage stages, rather than full-scale bilingualism and code-switching in individual speakers, are responsible for the restructuring that produced creoles. The more frequently speakers of interlanguages communicated with each other, the more likely some features of such interactions were to normalize in the community, albeit as variants of other features closer to the lexifier's, hence the development of basilects. Myers-Scotton's approach suggests also that a substrate language functions as the "matrix language." A charitable interpretation of this is that, by the least effort principle, learners first try structures familiar to them, until they are proved wrong. This is may be true. On the other hand, one can be sure that when there is a constant native model (which was not the case during the basilectalization stage of creoles), even the most inept learner can recognize when constituents are structured differently in the target language. Being (fully) successful in attempts to master the target language is another story, which makes imperfect-learning or approximation accounts plausible, especially because they make allowance for interference from previously known systems. Like the relexification hypothesis, Myers-Scotton's approach makes no, or little, allowance for structural features of the lexifier to prevail even in modified ways. After all, creole DO share several features with their lexifiers, which, we must recall, were nonstandard vernaculars. There are indeed cases involving no decreolization, such as in Réunion, where a putative creole has remained close enough to structures of its lexifier.

11. This explanation seems also to be the gist of Chaudenson (in press) as he speaks of relaxation of metropolitan norms and makes allowance for influence of substrate languages on the restructuring of the lexifier.

12. It has been argued recently that the development of creoles involves no processes of grammaticization. This is a mistaken conclusion based on the gratuitous assumption that creoles did not develop by ordinary or normal language change, not on the unproven absence of developmental paths similar to those of grammaticization in the development of non-creole vernaculars. See my discussion of this topic in Mufwene (1996c).

13. There are perhaps others like McWhorter (in press) who assume that denying that creoles can be defined structurally amounts to claiming that there is no empirical justification for singling them out as an interesting group of vernaculars. I never argued for such an extreme position. I have maintained since Mufwene (1986) that as a group of language varieties, creoles are sociohistorically defined by the similarities of the socio-economic settings that produced them. Sociohistorical justification is as empirical as alternative justifications for singling creoles out as a group worth investigating, especially regarding the relation between ecological factors and language evolution. However, one justification need not entail another, and this has been the point of my arguments all along against defining creoles structurally.

14. It is sociologically interesting that in the Anglophone Caribbean the prevalent folk term for creole vernaculars is the French word patois (often represented as patwa in the relevant creolistic literature), a term which is not used for the same kind of vernacular in the Francophone Antilles. Here the term créole is used, whereas on St. Barths, for instance, patois applies to the nonstandard vernaculars spoken by the white descendants of French settlers. Calvet (1999) reports that white creoles in Louisiana will not refer to their French varieties as créole, although they are structurally similar to those spoken black creoles. To be sure, they sometimes characterize those spoken by black creoles as créole, but they would rather use the terms cajun or patois in reference to their own varieties.

15. Perhaps it will help to note that not all natural categories may be organized on the prototype model. There are good reasons why such an organization applies to color and kinship categories: their members (though not so discrete in the case of color categories) have perceptual and behavioral correlates that justify it. In the few cases where the model has been applied to membership in grammatical categories (e.g., McCawley 1988, ch. 7), it seems more realistic to invoke family resemblance instead. It is practically impossible to identify the best exemplars for categories such as Noun and Verb, though it is possible to structure them into several subcategories, which still have fuzzy boundaries. Not even child language acquisition data support such a prototype model to grammatical categories. Although one can argue that peripheral category members are acquired later, there is no convincing evidence that the best exemplars of particular categories are acquired first. However, see GivÛn (1986) for a somewhat different perspective, which assumes that prototype categories may be the compromise between Platonic categories and the family resemblance model.

Arends, Jacques. 1989. Syntactic developments in Sranan. Doctoral thesis, University of Nijmegen.

Baker, Philip. 1997. Directionality in pidginization and creolization. In The structure and status of pidgins and creoles, ed. by Arthur K. Spears & Donald Winford, 91-109. Amsterdam: John Benjamins.

Boretzky, Norbert. 1993. The concept of rule, rule borrowing, and substrate influence in creole languages. In Africanisms in Afro-American language varieties, ed. by Salikoko S. Mufwene, 74-92. Athens: University of Georgia Press.

Calvet, Louis-Jean. 1999. Mei you tiao cha jiu: pour une écologie des langues du monde. Paris: Plon.

Chaudenson, Robert. 1989. Créoles et enseignement du français. Paris: L'Harmattan.

Chaudenson, Robert. 1992. Des îles, des hommes, des langues: essais sur la créolisation linguistique et culturelle. Paris: L'Harmattan.

Chaudenson, Robert. in press. Variation, koïnèisation, créolisation: français d'Amérique et créoles. pp. 163-179.

Childs, Tucker. 1997. The status of Isicamtho, a Nguni-based urban variety of Soweto. In The structure and status of pidgins and creoles, ed. by Arthur K. Spears & Donald Winford, 341-367. Amsterdam: John Benjamins.

Dillard, J. L. 1972. Black English: Its history and usage in the United States. New York: Vintage Books.

Givón, T. 1986. Prototypes: Between Plato and Wittgenstein. In Noun classes and categorization, ed. by Collette Craig, 77-102. Amsterdam: John Benjamins.

Hagège, Claude. 1993. The language builder: An essay on the human signature in linguistic morphogenesis. Amsterdam: John Benjamins.

Hancock, Ian. 1986. The domestic hypothesis, diffusion and componentiality: An account of Atlantic Anglophone creole origins. In Substrate versus universals in creole genesis, ed. by Pieter Muysken & Norval Smith, 71-102. Amsterdam: John Benjamins.

Hjelmslev, Louis. 1938. Etudes sur la notion de parenté linguistique. Revue des Etudes Indo-Européennes 1.271-286.

Hock, Hans Henrich & Brian D. Joseph. 1996. Language history, language change, and language relationship. Berlin: Mouton de Gruyter.

Huber, Magnus. 1996. Atlantic English creoles and the Lower Guinea Coast: Afrogenesis reconsidered. Paper presented at the Westminster Workshop on Creoles.

Kulikoff, Allan. 1986. Tobacco and slaves: The development of southern cultures in the Chesapeake, 1680-1800. Chapel Hill: University of North Carolina Press.

Lass, Roger. 1997. Historical linguistics and language change. Cambridge: Cambridge University Press.

Le Page, R.B. & Andrée Tabouret-Keller. 1985. Acts of Identity: Creole-based approaches to language and identity. Cambridge: Cambridge University Press.

McCawley, James D. 1988. The syntactic phenomena of English. Chicago: University of Chicago Press.

McWhorter, John H. in press. Identifying the creole prototype: Vindicating a typological class. Language.

Mufwene, Salikoko S. 1986. Les langues créoles peuvent-elles être définies sans allusion à leur histoire? Etudes Créoles 9.135-50.

Mufwene, Salikoko S. 1991a. Pidgins, Creoles, Typology, and Markedness. Development and Structures of Creole Languages: Essays in honor of Derek Bickerton, ed. by Francis Byrne and Thom Huebner,123-43. Amsterdam & Philadelphia: John Benjamins.

Mufwene, Salikoko S. 1991b. Review of Pidgins and creoles. vols. 1 & 2. by John Holm. Language 67.380-87.

Mufwene, Salikoko S. 1992. Why grammars are not monolithic. In The joy of grammar: A festschrift in honor of James D. McCawley, ed. by Gary Larson et al., 225-50. John Benjamins.

Mufwene, Salikoko S. 1996a. The development of American Englishes: Some questions from a creole genesis perspective. In Varieties of English Around the World: Focus on the USA. Edited by Edgar Schneider, 231-264. Amsterdam: John Benjamins.

Mufwene, Salikoko S. 1996a. The Founder Principle in creole genesis. Diachronica 13.83-134.

Mufwene, Salikoko S. 1996c. Creolization and grammaticization: What creolistics could contribute to research on grammaticization. In Changing meanings, changing functions, ed. by Philip Baker and Anand Syea, 5-28. University of Westminster Press.

Mufwene, Salikoko S. 1997a. The legitimate and illegitimate offspring of English. In World Englishes 2000, ed. by Larry E. Smith and Michael L. Forman, 182-203. College of Languages, Linguistics, and Literature, University of Hawai'i and the East-West Center.

Mufwene, Salikoko S. 1997b. Jargons, pidgins, creoles, and koinés: What are they? In The structure and status of pidgins and creoles, ed. by Arthur K. Spears and Donald Winford, 35-70. Amsterdam: John Benjamins.

Mufwene, Salikoko S. 1998a. What research on creole genesis can contribute to historical linguistics. In Proceedings of the XIIIth International Conference on Historical Linguistics, ed. by Dieter Stein and Monika Schmidt, 315-338. Amsterdam: John Benjamins.

Mufwene, Salikoko S. 1998b. Language contact, evolution, and death: How ecology rolls the dice. Ms.

Mufwene, Salikoko S. in press-a. Accountability in descriptions of creoles." In Creole genesis, attitudes, and discourse: Studies celebrating Charlene J. Sato, ed. by John Rickford and Suzanne Romaine. Amsterdam: John Benjamins.

Mufwene, Salikoko S. in press-b. Contact and speciation in English: Some dialects of English are creole. In English globalized, Volume 1: Perspectives and prospects, ed. by Anne Pakir, Vincent Ooi, & Ismail Talib. Amsterdam: John Benjamins.

Myers-Scotton, Carol. 1997. 'Matrix language recognition' and 'morpheme sorting' as possible structural strategies in pidgin/creole formation. In Structure and status of pidgins and creoles, ed. by Arthur Spears & Donald Winford, 151-174. Amsterdam: John Benjamins.

Posner, Rebecca. 1996. The Romance languages. Cambridge: Cambridge University Press.

Rosch, Eleanor. 1977. Classification of real-world objects: Origins and representations in cognition. In Thinking: Readings in cognitive science, ed. by P. N. Johnson-Laird & P. C. Wason, 212-222. Cambridge: Cambridge University Press.

Tate, Thad W. 1965. The Negro in eighteenth-century Williamsburg. Williamsburg, VA: The Colonial Williamsburg Foundation.

Thomason, Sarah G. 1997. A typology of contact languages. In The structure and status of pidgins and creoles, ed. by Arthur K. Spears & Donald Winford, 71-88. Amsterdam: John Benjamins.

Thomason, Sarah G. and Terrence Kaufman. 1988. Language Contact, Creolization, and Genetic linguistics. Berkeley: University of California Press.

Trudgill, Peter. 1986. Dialects in contact. Oxford: Blackwell.

Return to Salikoko Mufwene's Home Page

Page created March 12, 1999
Last revised March 12, 1999