The main goal of this essay is to constructively deconstruct the age-old myth that "the world's simplest grammars are Creole grammars" and to demystify the methodological (mis)practices that underlie this myth and its corollaries throughout Creole studies and beyond.
I start with some notes on historiography and methodology, connecting certain trends in 20th and 21st century creolistics to outdated (quasi) Darwinian concepts in early 19th-century comparative-historical linguistics. Then I move to linguistics per se, inspecting the empirical and theoretical bases of creolists' foundational assumptions about Creole diachrony and synchrony. This will (re-)establish the epistemological limits of certain key terms in Creole studies, including "pidgin(ization)", "creole/creolization", "young" vs. "old", "simple(st)" vs. "(most) complex", etc. I will argue that these terms, although perhaps useful as a-theoretical heuristics and as sociohistorical approximations, cannot serve as theoretically-grounded linguistic-structural taxa: Given Universal Grammar and its Cartesian-Uniformitarian foundations (Chomsky 1966, 1981, 1986, 1995, etc.), there cannot be any invariant and sui generis set of structures and processes that fall under the labels "Creole" and "creolization". Assuming Universal Grammar (UG), creolization reduces at the individual level to the same sort of cognitive processes that underlie idiolect formation through language change, and so are Creole languages aprioristically undistinguishable from non-Creole languages--that is, there is no synchronic Creole typology that excludes non-Creole languages. As I show below, such claims go against the grain of the most ancient and the (still) most prevalent dogma in Creole studies.
My essay is best introduced by the following quotes from Schleicher (1863), Saint-Quentin (1872) and Adam (1883) on simple(st) grammars and from Osthoff & Brugmann (1878), Foucault (1972) and Popper (1965) on methodology. The first two sets of quotes ((1)-(2)) are words from the past and the last three ((3)-(5)) are words of caution for the future, and all five are relevant to linguists' time-honored search for simplest grammars. These quotes speak for themselves, and eloquently so.
(Schleicher 1863 [1869: 30], emphasis in original; also see 1869: 16-17)
(Schleicher 1863 [1869: 79])
(Saint-Quentin 1872 [1989, 40-41]; my translation)
"[Cayenne Creole] grammar ... is nothing but the grammar that is common to the languages of Guinée. The latter we can call langues naturelles as opposed to langues cultivées. For the botanist, plants that are naturelles are superior to plants that are cultivées to the extent that the former are pristine products that are free of intentional adulteration. Likewise, for the linguist, the speech of peoples considered primitive has primacy over the speech of civilized peoples: the former is closer to the sort of grammatical instincts of which children's utterances reveal processes that are simple, logical and fast. ... [Cayenne Creole] grammar is more naturelle than that of Sanskrit, Latin and French. But this grammar did not spontaneously emerge in Guyane; it was imported from Africa. ... "
(Adam 1883: 4-5; my translation)
"In reality, the Malagasy slaves [who created Mauritian Creole] have brought along to Mauritius their native grammar, but not the forms that I have previously mentioned [inflectional gender marking, inflectional plural marking on verbs, the avoir (`to have') auxiliary , the verbal copula]. These forms are the product of an evolution that has not happened in the Polynesian and Melanasian Languages. In dealing with such forms, Malagasy speakers kept their native grammar ... In Mauritius, this native grammar reasserted its influence. ... "
(Adam 1883: 7; my translation)
(Osthoff & Brugmann 1878 [1967: 204])
(Osthoff & Brugmann 1878 [1967: 205-206])
(Foucault 1972: 223)
(Popper 1965: 50)
Taken together, the quotes in (1)-(5) suggest that the search for "the world's simplest grammars" has been going on for quite a while, and so has the dogma that "the world's simplest grammars are Creole grammars". This search for "simplest grammars" is part of a larger search for, and larger myths about, the origins and evolution of our species. This search seems driven by an apparently innate drive to (in Schleicher's phrase) "classify humanity".
Schleicher's Glottik is linguistics as natural history. By definition, its morphological taxonomy of languages qua organisms is regulated by universal natural laws and reducible to the genealogical mapping of their teleological development toward "idioms of higher organization" (Schleicher 1850, 1863, 1865; etc).
In the intellectual climate of early 19th-century, Schleicher's pre-Darwinian language-as-organism evolutionary approach belonged to the "normal science" of his period ("normal" in the Kuhnian sense). In retrospect, Schleicher's Glottik can be viewed as one of the central myths that gave vigor, status and popularity to comparative-historical linguistics (see Hoenigswald & Wiener 1987 and Alter 1999 for comprehensive overviews and, specially, for an array of positions similar to Schleicher's through much of the 19th century and beyond).
At the core of Schleicher's language-as-organism evolutionary hypothesis was a then-attractive congruence between, on the one hand, the simple-to-complex Darwinian evolution of biological organisms from "one-celled organisms" to "higher living being" and, on the other hand, the postulated historical progression of languages, through variation and subspeciation, from isolating to agglutinative to inflectional/fusional (Schleicher 1863 [1869: 50-60]; also see Schleicher 1850 [1852: 6-13]). In the Schleicherian organic school, linguistic evolution, on a par with biological evolution, was to be modeled by a "tree of life", a Stammbaumtheorie--a genealogical ("family" tree) diagram depicting phylogenetic relationships, with (at most) one parent for each daughter node. 3
Schleicher tries to establish his morphology-as-biology congruence by, inter alia, metaphorically taking the "simple cell" and the "simple root" as "the common primitive forms" of biological and linguistic evolution, respectively (Schleicher 1863 [1869: 55]). More explicitly, Schleicher takes "the radical elements [i.e., isolating root morphemes] as the cells of [prototypical primitive] speech" (1863 [1869: 53], emphasis as in original) and posits an isolating proto-language--one made up exclusively of affixless roots--at the evolutionary beginning of each language phylum. The proto-language's monomorphemic words are the linguistic analogues of the "one-celled organisms" at the roots of biological evolution. Thus, Chinese with its tendency toward isolating morphology is most primitive while Sanskrit with its inflectional/fusional morphology is most advanced since, according to Schleicher et al, inflectional/fusional morphology marks the highest degree of complexity and perfection (see (1b) and (2b); also see Schleicher 1850). It is thus that morphology--inflectional morphology, in particular--has long served as the chief measure of evolutionary progress and/or structural complexity (but see note 1 ). 4 Related measures are found in 20th- and 21st-century linguistics, as in the works of (e.g.) Jespersen 1922: 233-234; Whinnom 1971: 109-110; Samarin 1980: 221; Seuren & Wekker 1986; Bickerton 1988: 274-276; Comrie 1992: 208-209; Seuren 1998: 292-293; McWhorter 1998, 2000a,b, 2001; etc (cf. note 6 ; see Chaudenson 1994 and DeGraff 2001 for overviews and critiques).
Time (i.e., language age) is a critical factor in Schleicherian models of linguistic complexity. In such genealogical-cum-teleological approaches to the linguistic systema naturae, time is built-in as a pre-requisite for the development of complexity in both biological and linguistic evolution: complexity qua "higher organization" takes "very long time spans" to evolve (see (1c)). Thus, the following three related propositions:
"Systematization of humanity" via linguistic structures and (alleged) lack thereof is also found at the very inception of creolistics and throughout its existence. Creolists' own `classic' systematization is based on the age-old orthodoxy that Creole morphology is EXTRAORDINARILY simple or simplified from a diachronic and/or synchronic perspective. 6
In the 17th-19th centuries, such orthodoxy had explicitly "race"-based underpinnings. One taken-for-granted piece of "normal science" (in Kuhn's terminology) revolved around the notion that non-whites were inferior human beings, and so was non-whites' speech considered inferior to whites' speech. For candidate (perhaps ambiguous) illustrations of such beliefs, see the works of early creolists such as Pelleprat 1655, Saint-Quentin 1872 (see (2a)), Baissac 1880, Adam 1883 (see (2b)), along with 18th/19th-century dictionary and encyclopedia entries for the word "Creole" (see, e.g., Pierre Larousse's 1869 Grand Dictionnaire Universel du XIX Siècle and Vinson's entry in the 1889 Dictionnaire des sciences anthropologiques). In such works, the distinctive features of Creoles were not due to socio-historical factors only, but they were also taken to reflect the inferiority of their (non-European) speakers. The latter were deemed to be cognitively unable to master the "complexities" of European languages. Per this orthodoxy, "primitive" people spoke "primitive" languages (see 3.1; cf. (2) and note 35 ).
At the turn of the 19th-century the Neogrammarians, while adopting Schleicher's seminal insights in comparative-historical methodology (e.g., with respect to the reconstruction of unattested proto-forms), fought hard to get rid of his excessively organic and teleological metaphors. The Neogrammarians' stated goal was to understand Language and language change through the study of variation among (related) idiolects and through the study of universal laws (e.g., sound-change laws) that are ultimately rooted in the psychology and physiology of individual speakers. This was a shift of interest from the reconstruction of Ursprachen from archives to the analysis of contemporary idiolects in vivo (i.e., as manifested in individual speech). For Neogrammarians (see (3)), grammars live in speaker's minds, not in society; thus, the study of individual grammars as manifestations of "psychical [i.e., psychological] organism" should take epistemological priority over "Historical Grammar" (i..e, "descriptive grammars of different periods ... tacked together"): only the former is truly "scientific" (see Paul 1890 [1970: xxxvii,xliii,1-19]). As Osthoff and Brugmann 1878 [1967: 198] put it, what needed correction is the methodology whereby "[l]anguages were indeed investigated most eagerly, but the man who speaks, much too little".
Continuing this "Cartesian" (i.e., universalist and mentalist) trend into the 20th-century, Boas and his students, including Sapir, assumed both the existence of language universals, the "psychological reality" of idiolects and the independence of phenotypes, language and culture across humanity (see, e.g., Boas 1911: 11; Sapir 1921: ix,207ff). Both Boas and Sapir reacted strongly against the sort of Romantic ethnocentrism inherent in Schleicherian correlations between morphology, age and complexity. About the (non-)rapport between linguistic morphological types and cultural evolution, Sapir wrote:
The Schleicherian genealogical classification of language by age cum complexity loses further grounding with the advent of Chomsky's "Cartesian Linguistics" in the second half of the 20th-century (see, e.g., Chomsky 1966, 1981, 1986, 1995). As in the Neogrammarian dogma in (3), grammars are inherently parts of human biology, not autonomous living organisms that undergo birth, age, senility and death independently of their speakers. Generative linguistics' objects of study are, in Cartesian mode, internal properties of individual minds (i.e., mental grammars qua I(NTERNAL)-LANGUAGES). Thus, generative linguistics is "internalist biolinguistic inquiry" (Chomsky 1995: 1-11, 2001: 41-42) and is intrinsically Uniformitarian (cf. Descartes's assumption that "reason is by nature equal in all men" and Descartes's notion of knowledge as mental representations; see Chomsky 1966 for relevant discussion).
Per current assumptions in biology, the basic morphology of the human brain is uniform across the species. It has thus become more difficult, if not impossible, to theoretically correlate biological evolution with cross-linguistic variation, unless one adopts a quasi-Lamarckian view of language change whereby cross-linguistic typology can be reduced to genetic variation across human groupings. In this quasi-Lamarckian scenario, language-specific structures (e.g., isolating vs. agglutinative vs. inflectional/fusional morphology) would be correlated with variations in the human genetic blueprint. This is not a likely scenario given current results in language acquisition and biology: linguistic structures do not evolve and are not transmitted like DNA. 7
Notwithstanding current advances in Cartesian and Uniformitarian (bio)linguistics, there is a sense in which Schleicher's approach has survived the 19th century into the 20th and now 21st century, albeit under new theoretical guise. While modern linguistics is making steady advances in its exploration of our intrinsically human and species-uniform Universal Grammar (UG), "creolistics" has kept up, and even revived, early 19th-century notions of language evolution. Indeed, creolistics is perhaps the only field where the search for a genealogical and typological class of "simplest grammars" is still at the center of contemporary research. It is thus that certain trends in creolistics are reviving Schleicher's Glottik with Creole languages as the new class of youngest, thus structurally simplest, linguistic species. In this modern Glottik, Creole languages are living specimens of Ursprachen (i.e., contemporary proto-languages--"the world's only instantiation of spoken language having been "born again"" in McWhorter's (§2.3) evangelical phrase).
However, an empirically and theoretically grounded implementation of the claim that the world's simplest and/or most optimal grammars are Creole grammars still constitutes the holy grail of creolists through an un-broken lineage of research programs (see note 6 ). In the 20th and 21st century, starting with (e.g.) Jespersen 1922 and up to McWhorter 2001, linguists have used morphosyntax to try and identify sufficient and necessary conditions that would make Creoles deeply special in a structural and synchronic sense. Perhaps unsurprisingly, many of these contemporary observations on Creole morphosyntax are quite reminiscent of those encountered in 17-19th century texts. (See DeGraff 2001b: 88-98 for further details). For example, consider the statement that "creoles are natural languages reborn from a radical reduction of their source languages into makeshift jargon" (McWhorter 2001: §4). This statement finds direct antecedents throughout Creole studies, from its very inception (see note 6 ).
An explicit quantification for the alleged maximal simplicity of Creole grammars is offered in McWhorter 2001. Even though it makes no reference to Schleicher, McWhorter's recent work (1998, 2000a,b, 2001) is, of late, the most sustained and perhaps most widely read effort to articulate a structural basis for Schleicherian linguistics, with the aim of categorizing languages according to some explicit complexity hierarchy. McWhorter tacitly assumes the Schleicherian dogma whereby linguistic typology must, at all costs, include a genealogically and structurally well-defined class of "simplest grammars". In his revamped complexity scheme, Prototypical Creoles are the new Ursprachen, languages created ab ovo from virtually "ground zero" complexity (McWhorter 2001: §4.4). This is somewhat reminiscent of Saint-Quentin's structural claims in (2a). Thus, McWhorter's hierarchy continues the Schleicherian tradition of putting certain languages (here Prototypical Creoles) in a deeply-special class of linguistic neonates--contemporary fossils of Language at its evolutionary incipience. In a nutshell, the argument is that: "because so much of a grammar's complexity results from the operation of random accretion over time, creoles display less complexity than the rest of the world's natural languages" (McWhorter 2001: §2.3).
The central assumptions here are (i) that Creole languages are markedly younger languages than non-Creoles; and (ii) that this age difference is linguistically measurable and significant, contra Osthoff & Brugmann's (1878) admonition in (3). These assumptions are related to THE foundational claims in Creole studies, namely the oft-repeated dualist statement that Creoles are "non-genetic" languages that emerge via an abnormal "break in transmission" whereas non-Creole languages gradually evolve "genetically" via "normal transmission" (the modern locus classicus for this claim is Thomason & Kaufman 1988: 8-12, 206 and passim); see 3.3 for discussion). In the classic Creole-genesis scenarios, Creole youth stems from the "pidgin-to-creole life cycle" as signaled by the concomitant morphological bottleneck (see references in 1.2 and in note 6 ). In the most recent exponent of this dogma, Creoles are "born as pidgins, and thus stripped of almost all features unnecessary to communication" (McWhorter 2001: Abstract).
As will become obvious through the development of the present critique, current research on Creoles as contemporary Ursprachen (e.g., McWhorter 2001--"The world's simplest grammars are Creole grammars" (this volume)--hereafter abbreviated WSG 8 ) presents us with a "modern" collage of pre- and neo-Darwinian claims about language evolution. The antecedents of such claims go back to the structural-cum-genealogical speculations of (e.g.) Schleicher 1863 (see (1)) and Saint-Quentin 1872 (see (2a)). Thus, a full-fledged Popperian (see (5)) critique of (neo-)Schleicherian-cum-Quentinian creolistics can proceed via a close examination of the proposal in WSG about age-complexity correlations. This proposal conveniently provides us with an updated adaptation of linguistic-genealogical arguments that have run virtually uninterrupted through the past two centuries. A critique of this proposal will, I hope, clear up the scene for empirically-responsible and theoretically-grounded Cartesian-cum-Uniformitarian creolistics (i.e., the sort of creolistics that does not assume any a priori fundamental structural distinction between Creoles and non-Creoles). 9
As we'll see below, modern creolists' measures for age-complexity correlations--like their early 19th-century intellectual antecedents--amount to an empirically, theoretically and logically flawed view of Creole formation and language change. Bearing in mind that "science must begin with the criticism of myths", these re-formulations of Schleicherian linguistics provide a valuable point of departure for learning from our mistakes and for advancing our knowledge (in Popperian mode; see (5)). My critique will thus exemplify "learning by debunking" (in Gould's (1996: 351-353) terminology). Indeed I share in Umberto Eco's optimism in his book Serendipities where he so describes various brands of "lunatic" linguistics (cf. Foucault's allusion to "chimera and reverie" in (4)): "[E]ven the most lunatic experiments can produce strange side effects, stimulating research that proves perhaps less amusing but scientifically more serious" (Eco 1998: iix). 10 Here are the two "serendipities" to be derived from the discussion of neo-Darwinian (or neo-Schleicherian) creolistics below:
Fundamental to neo-Schleicherian arguments about Creoles' lack of complexity are the two notions "pidgins" and "features unnecessary to [basic] communication". These notions still remain ill-defined. I will argue in Section 4 that there can hardly be a core set of basic-communication features that defines the structural essence of ALL pidgins. As for the exclusively-Creole combination of structural features claimed to be "predictable from the history of creoles in pidginization", this typology will be argued to be deeply problematic, on both empirical and theoretical grounds (see Section 2 and references in note 12 ). Problematic as well, at least from a linguistic-theoretical standpoint, is the notion "age of languages": I will argue that thus far no well-defined and independent (i.e., testable and non-circular) LINGUISTIC metric can objectively measure the age of languages.
Yet it is the age of languages that in Creole studies is often taken (in a manner reminiscent of early nineteenth-century--Schleicherian--linguistics) as one crucial factor leading to complexity differentials between Creoles and non-Creoles. For neo-Schleicherians, Creoles' lesser complexity is yet another "predictable result of their youth" (see, e.g., WSG: §1), notwithstanding the fact that there is still no reliable litmus test for young languages as a linguistic class (see (7)-(8); also see Sections 3 and 6).
What about "complexity" per se? Here too creolists enlist terminology and assumptions that will be shown to be ill-defined and empirically- and theoretically-controversial in fundamental ways. For examples, the "complexity metric" in WSG: §2.4 is theoretically peculiar: it is based on an arbitrary list of superficial linguistic features with no psychologically-relevant, theoretically-grounded or independently-motivated unifying bases. In fact, the conceptual foundations of this metric are either left undefined (along with, e.g., "communicative necessity", "basic [human] communication" and language "age") and/or are made largely incompatible or irrelevant to what we (seem to) know about linguistic typology, historical linguistics, language acquisition, language processing and theoretical linguistics (cf. WSG's interpretation of "complexity", "Universal Grammar", etc.). Furthermore, the choice and testing of creolists' complexity metrics is often circular, tendentious and empirically flawed (see Sections 5 and 6).
Chomsky once wrote that "linguistic theory must be constructed with explicit and precise definitions and operational tests" (1957: 233). In the absence of empirically- and theoretically-grounded definitions of "creoles", language "youth", "pidgins", "basic [human] communication", "complexity" and so on, what are we to make of the proposition that "the world's simplest grammars are Creole grammars" because Creoles, and only Creoles, are "born again" languages? I take it that, in the absence of "explicit and precise definitions and operational tests" this most simplistic proposition can provide us with only a mismeasure of Creole languages and, indeed, a mismeasure of Creole SPEAKERS if one assumes a Cartesian (or Humboldtian) approach whereby languages are properties of minds. 11
In what follows, I will document the theoretical, empirical, and bibliographical lapses that undermine the recent (and not-so-recent) claims that "the world's simplest grammars are Creole grammars". In particular I will review the foundational assumptions upon which traditional and recent claims about Creole complexity (or lack thereof) rest. These assumptions touch on the following questions, which I address in turn:
What's in a name? Long ago, Francis Bacon (1620 [1994: 64-65]) warned us that that there "are names for things that do not exist (for just as there are things without names because they have never been seen" and that certain "obscure and deep seated [terminology] is derived from an incorrect and unskilled abstraction". Furthermore, he wrote,
What exactly are creolists from Saint-Quentin to McWhorter comparing when they make claims to the effect that "the world's simplest grammars are Creole grammars"? What is the exact scientific purpose of this comparison? The postulation of some structural essence common to all "Creole" languages is central to the present discussion as it has been from the beginning of Creole studies. That this structural essence is the outcome of "simplification" is another age-old tenet in Creole studies. As Chaudenson (1994: 41) notes, "the idea that Creole are simplified versions of their European ancestors is as old as Creole studies".
Yet it has often been argued in Uniformitarian (non-dualist) fashion that "creolization is a sociohistorical, not a structural, process" (Mufwene 2000a). In this view, Creoles cannot be distinguished a priori from non-Creoles on strictly synchronic structural grounds. I myself have now adopted a strictly language-external definition whereby "Creole" is a sociohistorical attribute that connotes the results of particular types of (abrupt) language contact marked by exacerbated social distance cum power imbalance (see DeGraff 1999a,b and references therein for overviews). I also take it, as a working hypothesis within the generative framework I work in, that: 12
The Cartesian-Uniformitarian position in Creole studies has long been opposed by the dualist orthodoxy that assumes a strong form of `Creole exceptionalism'. Per this dualism/exceptionalism, Creole languages--thus Creole speakers--are deeply special, with genealogical and structural properties that are fundamentally distinct from their non-Creole counterparts. This is grosso modo the position of neo-Schleicherian creolists who posit both a diachronic exceptionalism ("pidgin ancestry" and Creole "youth") and a synchronic exceptionalism (a "Creole Prototype"). These distinctions are straightforwardly Schleicherian in the rapport they try to establish between complexity and time: "creoles are the world's only instantiation of spoken language having been "born again", when speakers expanded pidgins" (WSG: §2.3), and Creoles' exceptional youth (i.e., their exclusive "pidgin ancestry") entails that "the world's simplest grammars are Creole grammars". As in Schleicher's evolutionist scenario, it is assumed that "much of a grammar's complexity results from the operations of random accretion over time" (WSG: §2.3). However, these genealogical and structural distinctions lack theoretical and empirical substance.
On the synchronic structural front, the Creole Prototype is equated to the combination of the following traits (WSG: n1):
There is already a plethora of data and observations that, taken together, invalidate the empirical and theoretical claims in (11)-(12) about Creole morphogenesis. 13 14 Yet, in creolistics and in other human sciences, complexity rankings and genealogical scenarios are no simple matters, scientifically and sociologically (see, e.g., Gould 1996 for relevant caveats). Thus, the need to address the claims in (11)-(12) from ground-zero up, starting with the theoretically and empirically central aspects of the Creole Prototype.
To start with, let's take Haitian Creole (hereafter HC). This sociohistorically prototypical plantation Creole has been argued in DeGraff 2001b: 69-88 to manifest both "inflectional affixation" and "opaque lexicalization of derivation-root combinations", contra the predictions in (11). Moreover virtually all HC affixes are etymologically related to French affixes (i.e., virtually none of these affixes result from grammaticalization), thus shedding doubt on the claims in (12).
Well-documented facts of Haiti's sociohistory and demographics teach us that HC should count as a bona fide Creole, even a "most Creole of Creoles", one whose historical conditions "have been perfect for the preservation of a basilectal Creole" (McWhorter 1998: 809, 812, 2000b: 206). But, in Creole studies, even established historical facts can be tinkered with to fit the scenario du jour: After robust HC data were advanced as counter-examples to pro-prototype claims (see DeGraff 2001b), HC's privileged status as "basilectal Creole" and "most Creole of Creoles" got revoked. HC is now taken to "not exemplify the Creole Prototype in the purest possible form" because of alleged "contact over the centuries with French" (WSG: 143).
Historically, linguistic interaction in colonial Haiti between Europeans and Africans was by far the most intense at the onset of contact, with French structures having had the most influence in the formation of (proto-)HC quite early on, in late 17th through early 18th-century. Thereafter, contact with French speakers was greatly reduced after the sugar boom in the middle of the 18th-century: the labor needs of expanding sugar plantations led to a drastic increase in the arrival rates of Africans. Before the sugar boom, the colony was still made up of mostly small homesteads--the société d'habitation--many of which subsequently and gradually gave way to the brutal segregation of the plantation economy--the société de plantation. (See Baker & Corne 1982, Chaudenson 1992, Chaudenson & Mufwene 2001 and Singler 1996 for an overview of the historical and demographic details and for pointers to the relevant literature.)
After the independence battles of 1791-1803, the French presence was virtually eliminated. There have always been, and still exist, various degrees of contact between HC and French, specially at the higher echelons of society, with concomitant contact phenomena in both languages as naturally expected. However Haiti's history has allowed post-genesis HC to evolve in relative isolation from its socially-remote, albeit prestigious, lexifier. Haiti is the only New World plantation society that eliminated most of its "lexifier" population through war. From the société de plantation onward, French has been especially remote from (most of) the monolingual peasantry, Haiti's numerical majority. More generally, the vast majority of contemporary Haitians in Haiti are monolingual Creolophones with relatively little contact with other languages (one notable exception is the region alongside the Haitiano-Dominican border where HC speakers are regularly exposed to Spanish varieties; see DeGraff 2001b: 68 for one linguistic-structural consequence of this contact in the domain of morphology). Such a degree of linguistic isolation is quite unlike the situation in other Creole-speaking communities in the Caribbean. These well-known sociohistorical and demographic facts explain why Haiti does not offer the same sort of "Creole continuum" that is found in the former English colonies of the Caribbean. Unlike (say) the English-lexicon Caribbean Creoles, HC by and large (i.e., the majority of HC speakers) did not "remain in contact with the lexifier" (contra WSG: n17). It thus seems unlikely that the bulk of HC morphology would have been created post-genesis via late borrowings from French (e.g., after the elimination in 1803 of the majority of potential "lenders"). 15
Furthermore there is no evidence that there ever was an earlier stage of (proto-)HC where all Creole varieties were uniformly devoid of all affixes. Such an affixless stage is even less likely considering that most of the HC lexicon is etymologically French and that HC speakers--like any other human speakers in our documentable past--are, in principle, able to extract morphological (e.g., affixal) information from stored patterns in their lexicon. In fact, whatever mental capacities would enable late-borrowing(-cum-restructuration) of French affixes in the course of HC's post-genesis diachrony would have also enabled the adoption/adaptation of similar French affixes at the earliest stages of (proto-)HC. This is specially so given the numerical preponderance of French over Africans and the lesser racial segregation at the early stages of contact. In other words, the sociolinguistic context was more learner-friendly at the onset of contact--during the société d'habitation--than later on--during the société de plantation (see note 15 ). No known (psycho- or socio-) linguistic principle could have forced all the speakers in the (early) contact situation to systematically ignore all morphological patterns in the available lexifier varieties (also see Appendix I).
Similar remarks apply to the diachrony of "opaque lexicalizations of derivation-root combinations" in HC--throughout I am assuming Uniformitarianism (e.g., that innate mental capacities for Language have remained uniform across the species in the past few millennia and across socioeconomic contexts; see (3)). Indeed, if HC speakers could create such morphology via "borrowing" through POST-genesis (reduced) contact with French speakers, then no known (psycho)linguistic constraint would prevent the creators of HC from creating this morphology at the ONSET of contact (e.g., during the société d'habitation, when the demographic and sociolinguistic factors were MORE favorable for the acquisition-cum-reanalysis of such forms via relatively HIGHER exposure to French speakers and, thus, more intimate familiarity with the lexifier). I must note again that, as far as we can tell, the bulk of the HC lexicon (both roots and affixes) has always been etymologically related to French (see 3.3). This means that Creole creators from the get-go massively adopted (at the very least) French lexical forms, including opaque lexicalizations.
Thus, we can conclude that HC--a bona fide prototypical Creole in the sociohistorical sense--with its inflectional and derivational morphology and its opaque lexicalizations represents a robust counter-example to classic neo-Schleicherian claims about Creole morphogenesis via an affixless pidgin (also see Appendix I).
On the theoretical front, given the very nature of the lexicon (in the Bloomfieldian sense) as the repository of Saussurean arbitrariness and given now-standard results in (psycho)linguistics, there seems to be no reason to expect "opaque lexicalization of derivation-root combinations" to take millennia to develop. 16 Indeed, it can be reasonably argued that "lexicalization entails that the form is no longer generatable via metaphorical inference and now requires storage as an independent form" (WSG: n10). In this statement, "lexicalization" is clearly defined, in Cartesian fashion, in terms of mental processes in the heads of individual speakers, namely "metaphorical inference" and lexical "storage". The latter is what's normally (and crucially) involved in the creation of "opaque lexicalizations" as Saussurean (thus, holistic) signs in the speaker's mental lexicon. To also claim that such a mental process necessarily requires "millennia" to unfurl (see McWhorter 1998: 792f, 798, 812, etc.) leads to a scenario in which individual speakers can live for millennia--not a likely scenario.
A more realistic and theoretically-grounded alternative is to assume that opaque lexicalizations as properties of I-languages (thus, of individual minds, in concert with communal conventionalizations and cultural transmissions) do not require millennia to develop: like other Creole lexical items, opaque lexicalizations can be innovated in Creole grammars or they can be "inherited" from (and via exposure to speakers of) the source languages (see Fattier 1998; DeGraff 2001b: 76-82). Regarding the "inheritance" of opaque lexicalizations, a pro-prototype creolist could try and argue that such "inherited" lexicalizations should not count for testing the prediction in (11c) because they were not developed over millennia by the Creole speakers themselves. But this argument is circular since (Prototypical) Creoles are a priori defined as young languages by pro-prototype creolists. The argument is fallacious in yet another way: many bona fide opaque lexicalizations in (e.g.) contemporary English were not innovated by the current generation of English speakers--these lexicalizations were "inherited" by modern English speakers'. 17
On the strictly-empirical front, tone, inflection and opaque lexicalizations have been documented in a variety of Creole languages; see, e.g., the observations and references in Muysken & Law 2001 and Muysken (to appear). Muysken & Law make the important observation that:
The widespread consensus across time, across space and across theories in creolistics and beyond is that Creoles are young languages--linguistic neonates that embody an evolutionary prior stage in relation to non-Creole languages (see DeGraff 2001b for an overview). Girod-Chantrans (1785 ) found HC to be "nothing but French back in infancy". Adam (1883: 3) reports that "[i]n Europe, Creole speech is universally considered an infantile jargon" (also see (2b)). Jespersen (1922: 228) wrote that Creole creators spoke "as if their minds were as innocent of grammar as those of very small babies ... [thus, Creoles'] inevitable naïveté and ... childlike simplicity ... ". In Hall's (1962) pidgin-to-creole life cycle, Creole languages are the one exception to the principle of ""normal" language [being] handed down from generation to generation". Bickerton's Language Biogrogram Hypothesis turns Creole speakers into linguistic "Adams and Eves" (in Richard Price's terminology, as cited in Corcoran 2001). Seuren and Wekker (1986: 66, 68) and Seuren (1998: 292) contrast languages that are "older or more advanced" and "sophisticated" to Creoles, which are "younger or less advanced"--"beginning"--languages (see note 16 ). Thomason & Kaufman (1988: 8-12, 206, 211, etc.) consider Creoles as non-"genetic" (i.e., "parentless") languages that, unlike "genetic" languages like English, have evolved via some kind of "abnormal [break in] transmission" (see note 22 ). WSG: 131 follows suit: "[A]ll [of the world's natural languages'] grammars ... trace back tens of millennia. ... with one exception ... : ... Creole languages have, by definition, existed only for several centuries at the most. ... The oldest known creoles today ... trace back to the late fifteenth century ...".
In quasi-Schleicherian mode (see (1)) and against Neogrammarian warnings (see (3)), the notions "younger" vs. "older" language often form the cornerstones of dualist claims about complexity differentials between Creoles and non-Creoles (see note 17 ). For example, the proposition that "the world's simplest grammars are Creole grammars" is argued to be "a predictable and, in the end, rather unremarkable result of creole languages' recent origins" (WSG: 162). More explicitly (WSG: 132):
There is one fundamental question that is evaded throughout: How does one scientifically measure the "age" of languages for the (presumably) scientific purpose of correlating language age with language complexity?
Individual speakers (be they Creole-speaking or not) do not live through millennia. Neither do their respective "I(nternal)-languages" qua Cartesian mental grammars. So the statement that non-Creole languages--and only non-Creole languages--have existed for "tens of millennia" surely cannot refer to I-languages. Such "tens of millennia" estimates must apply to some extensional notion of language--perhaps some (reified succession of) socially-determined "E(xternal)-languages" or communal languages (see (3) and note 17 above on relevant methodological caveats, going back to the Neogrammarians).
And still we are faced with the fundamental question: When do E-languages (or some communal/sociohistorical reification thereof) commence? Under certain definitions and for certain well-defined purposes, language age can certainly serve as a useful heuristic in various disciplines, including archeology, ethnography, anthropology, historical linguistics, etc. For example, one can say that, in a sociohistorical sense, HC is a "young" language as it marks the identity of a newly-created community--a community that did not exist before the 17th/18th-century (I thank Heliana Mello for an enlightening discussion of this point). In a somewhat related sense, varieties of (say) Indian English and the corresponding (speech) communities may also count as "young". But notice that this sort of youth is not necessarily correlated with increased structural simplicity: in some ways, "young" Indian English may even be more complex than the Queen's "old" English (consider, say, the phonology of retroflex stops in Indian English). Similar remarks could, in principle, apply to any "old"-"new" language pair, as with (e.g.) European and Brazilian Portuguese.
Notions such as language birth, age, and death are also assumed implicitly and a-theoretically when we use terms such as "Proto-Indo-European", "Latin", "Old French", "Middle French", "Modern French", etc., as classificatory devices. But, notwithstanding the popularity and sophistication of Stammbaumtheorie qua "Tree of Language" (cf. Darwin's Tree of Life), old vs. new linguistic species cannot be discriminated by any measure that looks like biological genetic criteria (e.g., DNA, interfertility). There is no clear notion whereby E-languages can be taken to reproduce like living organisms. Neither do we have clear linguistic-structural analogues for the DNA sequences that have now become so handy in tracing biological phylogenesis (see 1.4 above about the (im)possibility of quasi-Lamarckian linguistics). 18
There are a number of fascinating sociological factors vis-à-vis why, when and how certain languages start being (perceived as) "new" languages with birth certificates that distinguish them from their relatives. As a facile illustration, one can compare the status of Danish, Norwegian and Swedish (or Spanish, Portuguese and Italian; or Serbian and Croatian) as "distinct" languages vs. that of Chinese as "one" language. Suffice it to say that these distinctions are more relevant for issues of identity (and) politics than for strictly-linguistics (typological) matters.
Another example will drive the point home (in a metaphorical and litteral, if not statistically-representative, sense): In New York, I once saw a sign that advertised (in Haitian Creole!): Isit nou pale Franse `Here we speak French' (I have updated the sign's orthography to fit the official HC phonemic norms; see Dejean 1980). The HC-as-French sign had been proudly displayed by a Haitian employee next to a Spanish sign advertising Aquí se habla Español.
One may well chuckle at this story and argue that it is surely not representative of Haitians' meta-linguistic attitudes, but we still need to ask: What are the precise linguistic-structural criteria--the operational typological threshold--that would classify (e.g.) Missouri French and Cajun French as BONA FIDE varieties of French while HC is usually not so classified? Perceived notions of (non-)distinctness (in this case, between French and historically-related varieties) may be quite useful (or harmful) for creating and promoting stereotypes, political identities and community boundaries, but they seem to have little to do with linguistic typology per se. It is thus not so surprising that the perception of separateness between a Creole and its source languages is not uniform across all Creole speakers (see, Winford 1994: 45-48; Mühleisen 2000: 84-92 for recent discussions).
For typological and sociological reasons (e.g., regarding Creole-based education), I myself consider my native HC to be "distinct" from its Romance and Niger-Congo ancestors. For linguistic-typological reasons, one may well consider some variety of modern continental French (as spoken by Jean-Yves Pollock, say) to be distinct from its Latin and Old and Middle French ancestors as well as to its American relatives (Québec French, Missouri French, Cajun French, etc.); and so do Modern English fall into distinct varieties that in turn are distinct from Proto-Germanic and Old and Middle English varieties. In a somewhat related vein, I can also envisage, for methodological reasons, that children's early grammars are "distinct" from the grammars of their models--the older peers and caretakers that provide children with Primary Linguistic Data (see, e.g., Rizzi 1999 and references therein for discussions of children's "grammatical invention").
If our discussion of age-related complexity differentials (and the concomitant claim about "the world's simplest grammars") is to advance in a scientifically viable manner, we need an independent, theoretically-grounded measure for measuring the birth and age of (new vs. old) languages in some linguistically-relevant fashion that is impervious to our (often tacit) preconceptions, otherwise creolistics may well become the world's most simplistic science. Recall Chomsky's (1957: 233) statement that "linguistic theory must be constructed with explicit and precise definitions and operational tests". Such explicitness and precision is even more urgent when dealing with languages that have generally been stigmatized from the very moment that they were identified and "baptized". 19
In the absence of such measure (and given the discussion in Section 2; see note 19 ), can we rely on our intuition to discriminate newborn from multi-millenarian languages? The question is not so simple.
Let's get back to the case of Jean-Yves Pollock as a speaker of Modern French. What exactly does it mean to say that (Modern) French is an old language? Here we must be terminologically picky in order to try and make sense of this question whose presuppositions are infected with metaphors that contradict one another. Do the idiolects of Pollock and his peers underlie an E-language that has existed for tens of millennia? If so, his (E-)French would actually include, and be the continuation of, a very long ancestry, including at least Middle French, Old French, Latin, etc. and all the varieties in between. And so would Spanish, Portuguese, Italian, Roumanian, etc., be continuations (i.e., Schleicherian growths) of Latin and its descendants along the Stammbaum lines that connect Latin to the corresponding Romance variety. In other words, if each Romance language qua old language has, by definition, existed for "tens of millennia", then Romance speakers all speak a selfsame continuation of their common ancestor (say, Latin), which, in turn, implies that all Romance languages constitute the selfsame multi-millenarian language, by transitivity. Not a satisfying result, at least not for Romance linguists who have done much work to isolate robust parametric differences within Romance.
Let us ask again: Can linguistic typology help us decide whether a given language has existed for "tens of millennia" or is "born again" and "begins anew"? Can neo-Schleicherian (i.e., genealogical) creolistics constructively engage linguistic theory in any scientific way (e.g., in any way that resembles how evolutionary biology engage (phylo)genetic analysis)?
Take morphosyntax. Can relatively well-understood typological properties (e.g., lexical semantics, derivational morphology, underlying word-order, scrambling, nominal case morphology, presence of definite articles) be used as genetic tracers for writing up the "birth certificate" of languages? Let's assume so (only for the sake of argument), and let's take these typological properties as the analogues of, say, DNA sequences in the dating of biological species (but see notes 18 and 20 and Appendix II). Then, along the afore-mentioned typological dimensions, HC can be argued to be more similar to Modern French than Modern French is to Old French or to Latin. Thus, on the counts of both perceived similarity (in the eyes of certain Creole speakers like that HC speaker in the Brooklyn store mentioned in 3.2) and typological closeness (along certain typological variables), HC seems no younger (or no older) than Modern French. 20
What about morphology? It can be straightforwardly argued, as in 2.1 (also see Fattier 1998 and DeGraff 2001b, contra Lefebvre 1998), that most HC affixes historically derive from French affixes. Here the lexifier-creole similarities greatly exceed, both in cardinality and in systematicity, the sparse correspondences that have been used to argue for the "Afrogenesis" of French-lexicon creoles such as HC and Mauritian Creole from a single 17th-century West-African pidgin ancestor (see Appendix II). In the case of HC, the lexifier-creole morphological continuity is not at all surprising: most of the HC lexicon is etymologically related to French. In turn, the French lexicon itself is mostly derived from Latin--with French emerging through language contact as occasioned by Roman imperialist conquests (compare with Creole genesis in the context of Europe's imperialist conquests in Africa and the Americas).
Thus, the very morphology and history of HC (as compared to the morphology and history of, say, French) challenges the exclusive "born again" or "recent origins" status bestowed on radical Creoles as a class by neo-Schleicherian creolists. As far as I can understand neo-Schleicherian techniques for linguistic phylogenetic analysis and for identifying language birth (but see notes 18 and 20 and Appendices I and II), HC's etymological longevity and uniformity and its morphosyntactic inheritances-cum-restructurations should be the exclusive province of multi-millenarian languages. 21
Recall the classic "pidgin-to-creole life-cycle" scenarios whereby pidginization creates a radical bottleneck for lexical and morphological development, thus forcing Creoles to emerge from affix-less ancestors (see note 6 ; also see (12) and discussion in 4.2 below). A related and more general assumption is that "affixation ... emerges from the grammaticalization, reanalysis, or reinterpretation of material which was not originally inflectional" (WSG: §5.2). Taken together, these assumptions entail that affixes in allegedly young languages such as HC must, in general, emerge via the grammaticalization of erstwhile free morphemes (see (12)). These scenarios are robustly contradicted by HC where, from genesis onward, almost all affixes have had, and still have, cognates in French affixes--which in turn often have cognates in Latin morphology. Etymologically these cognates diagnose millennia of seemingly un-broken transmission, quite an ancient pedigree for the morphology of a "most creole of creoles". Indeed, there is no documented stage in HC diachrony where the language was affix-less or with most affixes derived from "erstwhile free morphemes" or with most affixes derived from outside of French (also see Section 2 and, specially, Appendix I).
In this respect, Goodman (1964: 26-28, 122-124) gives a variety of HC examples in the nominal and verbal morphosyntactic domains that suggest that "the initial speakers were exposed to a French which was virtually as complex inflectionally ... as is standard French". Putting aside the fact that "standard French" (specially back then) was an artificially constructed language with few, if any, native speakers, Goodman's contention seems somewhat extreme, for at least two reasons: (i) most whites in colonial Haiti, and even in France, were more likely to be illiterate speakers of rural "patois"--Langue d'Oc, Langue d'Oïl, Norman French, etc.--than fluent speakers of standard/literate French (see, e.g., Chaudenson 1995: 18; Chaudenson & Mufwene 2001: 151-153); and (ii) not all Creole creators were exposed to the same (non-native approximations of) native French varieties: right from the onset of contact, there must have existed a continuum of contact varieties, which were subsequently modulated through sociolinguistic factors into later varieties, including those known to us today (see, e.g., Alleyne 1971, Mufwene 2001; also see note 15 and Appendices I and II).
Nonetheless Goodman's point about HC's morphological complexity vis-à-vis its lexifier is valid to the extent that the available archival and comparative evidence suggests a robust degree of etymological continuity, which in turn disconfirms the radical morphological bottleneck posited by the pidgin-to-creole scenarios. Alleyne (1971: 172-174) makes a similar point, subject to similar caveats, when he gives linguistic evidence from French- and English-lexicon Creoles that their lexifiers "in their full morphological systems where used in the contact situation". In the same vein, Mufwene (2000b: 9) writes that "to the extent that English pidgins and creoles, as well as indigenized Englishes, can ultimately be traced back to Old English, they all have a long history". Thus, as Goodman, Alleyne, Mufwene and many others have argued before, there seems to be little, if any, evidence that Creole genesis must prototypically proceed via "a radical reduction of [the] source languages into makeshift jargon" (cf. WSG: 144).
To recapitulate: Language age has long been taken as the crucial factor that determines level of complexity--this is in keeping with Schleicher's intuition about the genealogy of morphology (see 1.2). Schleicherian linguistics takes for granted the existence of some independent, precise and operational "language dating" algorithm for genealogical/phylogenetic analysis. Yet the language-dating heuristics that have thus far been used to diagnose language youth (e.g., pidgin-to-creole symptoms such as development of new affixes via grammaticalization) simply fail to account for robust data in HC. The latter's morphology is incompatible with its postulated ancestry in some hypothetical affixless pidgin. HC, as a sociohistorically prototypical Creole language, manifests multi-millenarian morphological wrinkles.
At this stage, this reader is left begging what sorts of criteria are tacitly applied in creolists' genealogical heuristics. Thus far, it looks like we're dealing with either some arbitrary (perhaps sociologically motivated but unstated) presuppositions and/or some circular argument. The circularity would go something like this: Creole languages are "new" because they are Creole languages whereas non-Creole languages are "old" because they are not Creoles. 22
Can creolists' theoretical elaborations on the concept "pidgin(ization)" and its import in the "pidgin-to-creole life-cycle" get us out of this conundrum?
One time-honored tradition in Creole studies views pidgins and the proto-creoles they gave birth to as paragons of "basic [human] communication" with near-zero complexity. The precursors of this view go back to (e.g.) Saint-Quentin (1872) and Schuchardt (1914); see (2a) and (19). We find similar views in 21st-century neo-Schleicherian creolistics (WSG: 126): 23
I myself don't trust my intuitions on the elusive notion "basic communication", and specially not so in the teleological-functional context of (15) and its congeners in Creole genesis scenarios. Above all, I don't know what the evolutionary and structural correlates of basic communication are. Has UG specifically evolved to perfectly implement basic communication? Is UG a, or THE, perfect (post-)pidgin grammar? What are the morphosyntactic requirements of "basic communication"? Are syntactic categories like N(P)s, V(P)s, C(P)s, etc., "functionally central"? What about XP-movement and other structural transformations? Couldn't "basic communication" do without them on a par with (e.g.) formal languages? Ditto with respect to (abstract) Case marking and the X-vs-XP distinction. Is there a maximum number of thematic roles per verb in "basic communication"? What is the shape of the lexicon in "basic communication"? What is the minimal phonemic inventory required by "basic communication"? Etc., etc. (Cf. note 35 .)
From what I gather in the UG literature (e.g., from Linguistic Inquiry articles), there isn't much there that can be straightforwardly related to "basic communication" needs. And it even seems that there are sociolinguists interested in language change who, like generativists, are quite skeptical about teleological functionalism. For example, Labov 1994: Ch. 19 bears the title "The overestimation of functionalism". In the next chapter on "The maintenance of meaning", Labov concludes:
The reasoning in (17) is the theoretical essence of neo-Schleicherian creolistics while the list in (18) is explicitly offered as a negative litmus test for basic communication. Yet, no independent justification and theoretical argument is advanced to explain why the features in (18) are "incidental to basic communication". Given the vast array of superficial cross-linguistic distinctions, why the ad hoc list of 15 features in (18) and not others? For example, why should ergativity, but not accusativity, be dispensable in basic communication? Pending answers to these and other questions above, it seems to me that appealing to some arbitrary list of scattered features to derive creoles' simplicity via a "basic communication" pidgin that lacks such features runs immediately into theoretical trouble (e.g., circularity, theoretical vacuity and unfalsifiability).
What are "pidgins" and how do they emerge? One thing that we seem to know, based on a variety of comparative evidence, is that pidgins (be they "early"/"reduced" or "extended"/"expanded") cannot be uniformly reduced to some sort of lowest-common-denominator "basic communication" natural language: In standard descriptions, "early" and "extended" pidgins fall at opposite ends of the structural and functional continuua.
About "EARLY pidgins", one common observation is that they often emerge as reduced communication systems used in restricted and specialized contexts (e.g., for sporadic limited exchanges outside of one's speech community). The prototypical, if controversial, definition for (early) pidgins is that they arise as "makeshift adaptations, reduced in structure and use, no one's first anguage" (Hymes 1971: 3; also see Schuchardt 1909; Jespersen 1922: Ch. 12; Bloomfield 1933: 472-473; Hall 1962: 151-153; etc.; see references in 1.2 and note 6 ). Bickerton (1999: 49), for one, considers early pidgins to be "reduced well below the minimum required by natural languages". It may even be argued that early pidgins may be unlike native languages to the extent that early pidgins' (lack of) structure seems to fall outside the formal boundaries for natural languages as set by UG (see DeGraff 1996b, 1999b: 499-500 for some discussion). It seems then that the "drastically reduced linguistic structure and lexicon" of early pidgins with restricted functions is the result of "the very first stage of rudimentary language learning" (Hall 1962: 151-153). These structural and functional restrictions immediately dis-qualify early pidgins as candidates for viable full-fledged systems for human communication, IF--and this is a big "IF"--"viable full-fledged ... human communication" entails the ability to encode for transmission the expressive needs of normal human beings across a functional range of topics. Something along the lines of this assumption is adopted in WSG: §2.3 where it is claimed that pidgins in the pre-expansion stage are "universally agreed to be rudimentary codes not fulfilling the needs of full language".
As for "EXTENDED pidgins" (e.g., in Melanesia), they seem to function beyond and above "basic" communicative needs: these pidgins incorporate various structural properties that are as formal (i.e., not "functionally central") as that of any full-fledged natural language (see, e.g., Hall 1962: 154-155; Keesing 1998, 1991, Siegel 1999; below I illustrate non-"basic" features in pidgins).
It thus appears that, as in other cases of (I-)language creation by adults, the making of pidgins leads to distinct-looking results depending on contingent sociohistorical specifics. In Hymes's (1971a: 69) words, "the characteristics found in development to, and of, a pidgin admit of degrees. ... pidgins and pidginization are instances par excellence of variable adaptation of means to an audience and situation".
Cartesian-Uniformitarian methodology (see, e.g., (3)) invites us to sort out historical processes and the external entities they create ("E-pidgins", say--social entities) from psychological processes and their concomitant individual-level creations ("I(nternal)-Pidgins", say--mental entities); see Paul 1890, Andersen 1983 and Siegel 1999 for useful overviews and methodological caveats (also see note 17 ). My (null) working hypothesis is that the making of I-Pidgins (I-Pidginization, if you will) enlists cognitive processes that commonly unfurl, not just in situations of abrupt and/or limited language contact, but also in the various instances of second language acquisition in "ordinary" contexts of language contact.
In this Cartesian (i.e., mentalist and internal) perspective, I(NTERNAL)-pidgins are not sui-generis: they are the internal linguistic states--(transient or crystallized) interlanguages, if you will--in which adult language learners (qua second-language creators in need of a lingua franca) routinely find themselves. When viewed EXTERNALLY a stabilized (E-)pidgin, on a par with a communal (E-)language with native speakers, is a reification and conventionalization of the creations of individual speakers interacting in specific sociohistorical contexts, with their particular linguistic ecologies and their particular communicative requirements. From that perspective, the fact noted above that pidgins in distinct sociohistorical matrices may widely differ from one another (i.e., that pidginization and pidgins "admit of degrees", in Hymes's words) can be naturally and constructively related to, inter alia, the observed variability in the outputs of second-language acquisition (e.g., with respect to the structural profiles and functional characteristics of adult learners' interlanguages and fossilizations thereof); see DeGraff 1999b: 479-508 for one overview and some references.
This Cartesian "I-pidgin(ization)" working hypothesis seems heretical to the field at large (see Siegel 1999 for one recent overview of diverging perspectives). Indeed, one (now familiar) truism in creolistics is that pidginization is a sui generis process that eschews (virtually) all morphology (see Jespersen 1922; Hjelmslev 1938; Bickerton 1988; Seuren & Wekker 1986; Seuren 1998; McWhorter 1998, 2000a,b, 2001; etc.). Per this truism, pidgins are uniformly (and, for some, teleologically) designed ab ovo as simplest languages. The corollaries of this truism--the pidgin-to-creole life-cycle and its concomitant morphological bottleneck--constitute received wisdom in language-contact and historical-linguistics textbooks. By definition, the pidgin-to-creole cycle is exclusive to Creole formation and is radically different from processes underlying the diachrony of non-Creole languages. It is noteworthy that this scenario is still part of the communis opinio in Creole studies notwithstanding the fact that the "classic" pidgin-to-creole litmus test fails on representative Creoles, including HC (see Section 2 above; also see Alleyne 1971). At this point, a bit of critical historiography is in order, before exploring modern exponents of the pidgin-to-creole cycle.
The view that pidginization entails a morphological bottleneck--a "stripping" of language-particular morphology--is already found in Schuchardt's (1914) description of "the creole before [it] become[s] the native language of the majority":
Schuchardt's hunch echoes through much of contemporary creolistics. We still find allusion to "stripp[ing]" as in Bickerton's (1988: 272-278) claims that "a sharp, and in some cases quite radical, reduction in the structural properties of the original target language was an essential prerequisite for new language formation" and that such reduction entails that, in the formation of radical Creoles, "the target's bound morphology [is] stripped ... thoroughly".
More recently, Schuchardt's "simpl[e] matter of mutual comprehension" has been linked (as in (20a)) to "basic communication" and its "functionally central" features (see, e.g., (15) and (20b)). Here, one ill-defined term--Schuchardt's "middle ground"--is replaced in WSG by another ill-defined term--"basic communication"--which, in turn, is based on some ill-defined mechanics for "stripping away virtually all of a language's complexity (as defined in [WSG]), such that the complexity emerging in a Creole is arising essentially from ground zero". And both the early-20th and the early-21st century scenarios involve "deliberate design" (somewhat reminiscent of "naive or teleological design" as critiqued by Labov 1994: Ch. 19-20; see (16); also see Paul 1890 [1970: xliv-xlvi]):
In (19), Schuchardt takes some intuitive impression of "basic variety" cum "baby/foreigner talk" (cf. Bloomfield 1933: 472, Ferguson 1971, 1975, 1981) to a structural extreme. He assumes that speakers in contact situations can systematically suppress structures that are "peculiar"/"distinctive" to their respective native languages in order to create an "emergency language" for "mutual comprehension". But this entails that, whenever speakers of (say) languages and need an "emergency language" for "mutual communication", speakers can correctly decide which of their native structures will "meet with [ speakers'] total incomprehension", and vice versa. And this would be why, in the European-African "emergency language", the Europeans eliminated European affixes (e.g., plural-marking suffixes such as English -s) while the Africans, in analogous fashion, suppressed the expression of African affixes (e.g., plural-marking prefix such as Duala ma-); see Schuchardt 1914 [1980: 91-92].
I find Schuchardt's claims in (19) and its modern implementations (see paragraphs below) theoretically and empirically challenging, even if seemingly common-sensical. In particular, I don't understand the psycholinguistics of finding the "middle ground". Ferguson (1971, 1975, 1981) seems right that every native speaker can resort to some recognizable and negotiable simplified register for speaking to linguistically-handicapped foreigners. Yet, notwithstanding the broad tendencies identified by Ferguson and others toward universals of simplification (see, e.g., contributions to Clyne 1981), foreigner talk doesn't seem to constitute a cross-linguistically well-behaved "simplest" structural type. Ferguson himself (1971: 146, 148; 1981) shies away from positing (simplification in) "foreigner talk" as a general, absolute and sui generis process in pidginization. Instead, regarding simplification, Ferguson cautiously notes:
It may well be the case that all contact situations entail simplification (however defined--e.g., as reduction in structural irregularities) to some noticeable extent. It may also be the case that simplification can happen without (large-scale) language contact (I return to this below). 24 However, speakers engaged in language contact are neither telepaths, nor (psycho)linguists, nor fluent in each other's languages. Therefore they cannot SYSTEMATICALLY decide what in their native speech should unambiguously count as "peculiar"/"distinctive" to the foreigner's ear. Furthermore, in deciding what's "peculiar"/"distinctive", the "middle ground" creators must, strangely enough, abstract away from phonetics--the language-particular component that is most accessible to the foreigner's ear. This is much easier said than done. For now, the (psycho-)linguistics of negotiating this "middle ground" strikes me as quite mysterious. One (perhaps less mysterious) alternative then is to posit that speakers of any pair of languages and know in advance what the "middle ground" ought to be, independently of any contrastive analysis of vs. . A cognitive pre-requisite for successfully establishing this "middle ground" is that speakers of and (and of all other languages) share a universal set of hardwired instructions for finding this "middle ground" as "simply a matter of mutual comprehension" (see Clyne 1981 for bibliographies on foreigner talk; also see 4.4 on simplification in pidginization).
One possible non-innatist, TELEOLOGICAL answer to the puzzle of Schuchardt's "middle ground" equates the latter to a pidgin qua (near-)perfect "basic communication" system (see (20)). This hypothetical pidgin is built almost exclusively on "functional central" features (i.e., it is "stripped off of almost all features unnecessary to communication"). However, it is still not clear how speakers (or linguists for that matter) can "deliberately" sort out between "functionally central" vs. communicatively "unnecessary" linguistic properties. Deciding what is "functionally central", and why, remains an ever-elusive task that has long frustrated expert linguists who are deliberately tackling this problem in the leisure of their research offices, thus the lively debates in functional linguistics toward the discovery of deep-seated, (i.e., non ad hoc) correlations between function and structure. (Also see Mufwene 2000a: 72-76 for socioshistorical arguments agains the teleological view of pidginization and creolization.)
There's one camp though where something like "basic communication"--qua universal set of INNATE instructions for finding the "middle ground"--has been proposed. Recently, Klein & Perdue (1997) have proposed that all second language learners go through a stable and universal "basic variety" (BV) stage. This BV, although not mentioned in WSG, somewhat looks like "basic communication" in WSG, at least in spirit: BV is a well-defined I-language, a pre-determined state of the language faculty. The BV's prototypical features are: no inflectional morphology, no grammatical morphemes, (NP)-V-NP order, tense-marking via adverbials, no movement, no complex hierarchical structure, etc. (Klein & Perdue 1997: 311-326, 332, 336) These "organizational principles" are genetically wired via some sort of "core UG", thus relatively independent of the native and target languages. Here's how K&P locates BV vis-à-vis UG:
The exclusively teleological-functional definition of pidgins--as speech that "eschew[s] all but the functionally central" (see (20))--is incompatible with a variety of well-documented facts from psycholinguistics and contact linguistics.
On the empirical front, alongside robust evidence for various kinds of simplification in pidginization, there are well-documented pidgin structures that are `inherited' from (some of) the source languages (see below). In other words, these structures were not "eschewed" from the Creole creators' native languages, whether or not these structures were "functionally central". The retention and reanalysis of source-language structures in pidginization (and creolization) is not surprising given what psycholinguists and sociolinguistics have taught us about language transfer in second-language acquisition (SLA) and about the dynamics of contact linguistics (see Weinreich 1953; Labov 1994, 2001; Mufwene 1990, 2000a,b 2001; and DeGraff 1996b, 1999b,d for some overview and further relevant comments). Besides, the fact that certain pidgin structures are absent in many "old" languages across phyla and across time suggests that not all pidgin structures are required for basic communication. This too is unsurprising. The sociolinguistic specifics of each instance of language contact are contingent on history. It is thus tautological that pidgins' source languages (and other relevant sociolinguistic factors in language contact) vary across time and space. Pidgins will thus `inherit' (and re-analyze) selected patterns from the languages in contact. Since it cannot be the case that every such "inheritance" exists in every ("old") language, these "inheritances" cannot all be taken as "functionally central" to "basic communication", lest many "old" languages are communicatively dysfunctional.
Counterexamples to (20) go as far back as Schuchardt 1914, if not earlier. Schuchardt documents widespread substrate influence on the developing "emergency language" at all levels of structure: syntax, lexicon, lexical semantics, proverbs, etc. For Schuchardt (1914 [1980: 93]), such transfers are quite natural: 25 "[T]he slaves spoke the creole not only with the Whites but also among themselves while their mother tongue was still in existence, the latter being moreover constantly revived to some extent by the continual immigration from Africa".
This rudimentary sketch of the socio- and psycho-linguistics of language contact in Creole genesis has since been confirmed and refined across a wide variety of cases. It has now been painstakingly documented that pidgins are pregnant with (reanalyzed) structures from the languages originally in contact, alongside structural innovations. Such massive transfers, restructurations, innovations give rise to an array of syntactic options for any given semantic function within and across pidgins. Unsurprisingly distinct pidgins select distinct functions for morphosyntactic marking and their morphosyntactic options often enter into competition for the expression of similar functions. 26
Now, one could well try and argue that whatever source-language properties survive pidginization and make it into the Creole must be "functionally central" properties that are "necessary to human [basic] communication". In other words, given evidence for admixture in Creoles (see, e.g., references in notes 26 and 31 ), if Creoles begin anew at virtually ground zero, then it must be the case that language transfers do not introduce (substantial) complexity in Creole genesis. Witness the following quote from WSG: 27
This postulated "stripping away" entails that pidginization systematically filters out from the languages in contact "the results of tens of thousands of other accretions", allowing the retention of only features that are "functionally central" (cf. WSG: n11, 13). However, given the right ecology, admixtures do carry along "incidental" features from the languages in contact. This carry-over of source-language features is favored by (e.g.) relative homogeneity of (some of) the languages in contact, relative exposure to these source languages, the socio-psychological profiles of the speakers in contact and other sociolinguistic incidental factors (see the references in 26, specially works by Alleyne, Mufwene, Siegel, Singler, and Thomason & Kaufman). Such factors may even include the whims of "one [pidgin] speaker"; see Nichols's (1986: 240) speculations about the grammaticalization of evidential marking in Chinese Pidgin Russian (see (24b)). No matter how they get carried over into the emergent contact language, these admixtures cannot all be subsumed under the (still elusive) category of "functionally central" properties. Indeed these admixtures and concomitant restructurations-cum-innovations do not seem required by basic communication; they even include some of the "incidental" features in (18). To wit, the very preliminary sample in (24) (also see the discussion in 6.4).
In a related vein, the sheer structural diversity of pidgins refutes the claim that pidgins are "communication vehicles deliberately designed to eschew all but the functionally central" (see (20)). 30 Pidgins' structural diversity includes patterns that are unfamiliar to speakers of "old" languages like English and French. Commenting on such diversity while sketching the `exotic' character of Chinese Pidgin Russian, Thomason & Kaufmann sum up this picture much better than I can, so I quote at length:
"Like the other pidgins described in [the section on diversity
in pidgin structures], Chinese Pidgin Russian has features that are
unusual among the better-known pidgins and creoles with European lexical
sources: SOV word order, postpositions as well as preposition(s), V NEG
word order, and a few inflectional and derivational affixes. ... None of these features could be predicted as the result of the operation
of universal structural tendencies alone, because the suffixes represent
marked constructions, and the word order features are different from the
ones found in other contact languages. The presence of both
preposition(s) and postpositions is itself rather highly marked in
universal terms. ...
We should emphasize, finally, that the examples given in this section do not by any means exhaust the instances of pidgin structures that are not promising candidates for simplified lexical source language features or features of universal grammar. ... Our goal here has been to demonstrate that origin theories based solely on evidence from well-known, well-documented mainstream pidgins and creoles are inadequate to the extent that they fail to predict the kinds of features we have illustrated". (Thomason & Kaufman 1988: 191)
WSG systematically skirts all empirically "strong antidote" against the view that Creole grammars are simplest. WSG offers no reference whatsoever to the "exotic" pidgins and the cross-linguistically rare distinctions that have been documented in Chinook Jargon, Chinese Pidgin English, Chinese Pidgin Russian, Fanagalo Pidgin, Hiri Motu, Kitúba, Mobilian Jargon, Ndyuka-Trio Pidgin, Nubi Arabic, Pidgin Delaware, Russenork, Sango, Taimyr Peninsula Russian-based Pidgin, etc. Yet there exist valuable treatises on many of these "exotic" pidgins. Mobilian Jargon is one case in point: it is closely examined by Drechsel (1993, 1997) who observes that "pidgins need not reflect universal patterns as thought earlier ... but may actually exhibit highly marked features of syntax" (1993: 344).
Pidgin structures should then count as extremely diverse, with structures that often have no counterpart in many "old" languages. There's no space here to illustrate the complete range of such diversity: the possibilities may not be endless, but they do seem to go beyond what can reasonably be imagined as "basic communication"/"functionally central" requirements. This is not surprising given the history of many pidgins in conditions of intensive language contact. I encourage the reader to consult the relevant references in order to appreciate the extent to which the definition in (20) is empirically untenable (also see DeGraff 1999b: 479-508 for additional observations and references on the ontology of pidgins). 31
With Schuchardt's and others' detailed evidence of admixtures, restructurations and innovations in Creole genesis, any sui-generis definition of "creole" as "language [that] `begins anew' amidst pidginization" is a fallacy. The available evidence thus far doesn't fare well for the teleological proposal in (20) that pidgins are "communication vehicles deliberately designed to eschew all but the functionally central". The semantics of "functionally central" should not be up for grabs: the "functionally central" in basic human communication--if definable--should be universal and should not be left to be determined on a case-by-case basis by some ecological roll of dice, lest we run into empirical and theoretical incoherence.
We've now tasted our "strong antidote to the still common view that all pidgins and creoles have similar and simple structures". Pidgin structures are typologically diverse and they are definitely not a subset of "old" language structures--many pidgin structures are nowhere to be found in functional "old" languages such as English or French.
This said, one must reckon that the products of (large-scale) language contact do give the impression that they are, to a certain degree and in certain domains, simpler than their corresponding source languages. For example, overt morphological paradigms (e.g., phonetically-realized inflectional affixes on nouns and verbs) tend to decrease in size, morphological irregularities tend to be filtered out, various sorts of semantic transparency tend to increase, etc. But it must be stressed that such simplification is not absolute: as documented through much of the creolistics literature, what we're dealing with is gradient simplification with respect to the languages in contact and their respective complexity in PARTICULAR domains of grammar. For each specific terminus a quo whose composition is determined by contingent sociohistorical factors, simplification leads to a necessarily-distinct terminus ad quem. Given incidental variations in the particular linguistic ecology and in the relevant socio-psychological and demographic factors, the terminus ad quem in certain cases will be more complex in certain grammatical domains than the terminus a quo in other cases (for case studies, see references in note 26 ; also see Romaine 1992: 217 for a similar observation and Muysken & Law's (2001) important caveat in (13)).
What are the sources of simplification in language contact as in the Creole-genesis cases? This is a complex question. DeGraff 1999b: 491-499;517-518 gives one, admittedly incomplete, answer, which is rooted in a Cartesian perspective that views pidginization and creolization as reducible to individual-level mental processes that are shared across the species (cf. (3)). There I argue that simplification stems from the cognitive limitations of adult language learners and the concomitant mechanics of second-language acquisition "under duress, in the initial stages of language acquisition in the context of language contact--contact that may be massive and abrupt, and that may involve considerable social and psychological distance between speakers in different language groups" (p. 491). This Cartesian (i.e., mentalist) view of simplification is Uniformitarian: the underlying (psycholinguistic) causes of simplification in Creole genesis are not, and could not be, exclusively "Creole" (see 4.2). Simplification patterns are due to the idiolect-formation mechanisms that are necessarily employed in the creation of both Creole and non-Creole languages. In fact, we find similar simplification patterns in well-studied cases of language change via language contact and, at the individual level, in the creation of interlanguages and in child-language acquisition (see, e.g., contributions to DeGraff 1999e and references therein). 32
With this in mind, let us re-examine the conceptual basis of the frequent claim that simplification in pidginization, unlike simplification elsewhere, creates linguistic neonates that, as a class, start life from virtually "ground zero" of complexity (see (23)). Let's put aside the grotesque claim that the terminus a quo of pidginization (i.e., the pidginizers' native languages) is invariably at "ground zero" complexity (cf. note 27 ). How can it, then, be guaranteed that simplification of, and transfer-cum-restructuration from, the source languages, alongside structural innovations, uniformly creates pidgins at ground zero of complexity?
Let's consider the four relatively uncontroversial propositions in (27):
All the afore-mentioned facts and observations about transfer, innovation, simplification and typology in Creole genesis and beyond should wake us up from any "chimera and reverie" whereby pidgins become systems for "basic communication" that eschew "all but the functionally central" (cf. (20)). 33
Popular scenarios for the emergence of "the world's simplest grammars" are fraught with epistemological problems: their theoretical foundations have long been undermined by notions left critically ill-defined--including "creoles", "pidgins", "young" vs. "old" languages, "basic communication", "functionally central" features. Regarding the latter two notions, in the absence of any independently-justified theory for "functionally central" properties and their cross-linguistic realization, we still lack a coherent, non ad hoc notion of "basic communication" as a linguistic-theoretical concept. For now, "basic communication" remains vague and elusive.
What would be needed to adequately define "basic communication" is a theoretical framework (some universals of "basic communication", say) that would independently motivate the "functionally central" ingredients of "basic communication" and spell out how they are minimally realized cross-linguistically at all levels of grammar. Such a framework would, for example, predict the exclusion of the features in (18) from basic communication and explain why these features alongside other "incidental" features go beyond the requirements of "basic communication" (but see Labov's caveat about "naive or teleological design" in (16)). In contradistinction, the list of features that are "incidental to basic communication" (see (18)) is scattered through the space of typological variation and is constructed outside any independent theory of "basic communication".
At this point, some pro-prototype creolist may optimistically respond that it's Universal Grammar (UG) itself, or some version of the Language Bioprogram à la Bickerton, that tells us about (the structure of) "basic communication". Or perhaps "basic communication" in creole formation is a community-wide fossilized instantiation of the "basic variety" seen in 4.2 (cf. (22)). And recall that the "basic variety" itself is postulated as the product of some `minimal/core UG'. So we should ask: Does UG define the "functionally central" requirements of "basic communication"? Before answering this question, I first need to spell out some working assumptions about UG.
It is usually assumed that UG, by its very nature, does underspecify all idiolects, whether Creole or non-Creole. As far as I can tell, there is no sense (yet?) in which UG defines a scale whereby languages can be ranked as being more or less overspecified across all domains of grammar simultaneously--or more or less removed from some innate system for basic communication. Underspecification is the very essence of UG qua biological template for Human Language. UG only defines the set of PERMISSIBLE languages; no ACTUAL language is defined by UG alone; see (e.g.) Chomsky 1986: 145-152, 1995: 6. This is akin to the way in which genotypes underspecify phenotypes. Language (with capital "L", in the singular) is innate, but languages (with small "l", in the plural) are not. That is, humans are ultimately hardwired for Language whereas the individual expression of this capacity as idiolects--(I-)languages with their particular phonetics, lexicon, morphology, syntax, semantics, etc.--is not biologically programmed, even though it is biologically-constrained. 34
Generativists aim at one abstract implementation this underspecified template for all human languages (see, e.g., Chomsky 1986: 145-152, 1995: 6-7, etc.). In this research program, this abstract Human Language template (aka UG) consists of "principles" and "parameters". Principles are presumably universal, hardwired in human biology. They exist alongside an array of under-specified parameter settings and/or an array of open slots for a language-particular lexicon with its concomitant phonemic inventory, morphology, lexical semantics, etc. The parameter-settings and lexical slots become (over)specified only after exposure to Primary Linguistic Data (PLD) on an idiolect-specific basis (compare with the emergence of phenotypes via the interaction of innate genotypes with incidental environmental variables). In other words, UG specifies no actual parameter settings and no actual lexicon: it's the inevitably contingent linguistic experience that fills in--that "specifies"--the idiolect-specific information. This framework makes it axiomatic that every actually-occurring idiolect (including Creole idiolects) will be "overspecified" with respect to UG. The latter only defines the space of--the boundary conditions on--possible human languages; it does not specify any one particular language or any one particular class of languages (pace Bickerton 1988; see (28)). Neither does UG specify a global hierarchy for classifying languages in terms of overspecification at all levels of grammar taken simultaneously.
Let's contrast this view with the proposition that Creole languages, because of their alleged youth, represent the minimal--"simplest"--instantiation of some universal set of structural requirements as dictated by UG. This is the essence of Bickerton's Language Bioprogram hypothesis. 35 Here UG is taken as a sort of lowest-common-denominator grammar with respect to which specific languages are more or less overspecified. Such overspecification is claimed to go hand-in-hand with complexity: "old" languages are the most overspecified and they are the most complex:
As I have already pointed out, UG is not a "basic communication", or lowest-common-denominator, grammar in the sense of a minimal, set of "functionally central" requirements. This is perhaps made clearer by taking the lexicon as example. The lexicon is yet another area of grammar where one language can "give overt and grammaticalized expression to more fine-grained semantic and/or pragmatic distinctions than another" (cf. WSG: 136). Individual lexica and the distinctions therein (e.g. vis-à-vis semantic distinctions as for, say, kinship terms--along with distinctions in phonemic inventory, contents of functional heads, affixal inventory, etc.) become fully specified only upon exposure to contingent PLD in sociohistorically-determined environments. Independently of (say) phonemic inventory and complexity thereof, the PLD give rise to arbitrary signifiant-signifié semantic oppositions of arbitrary complexity, including word-level ("opaque lexicalizations") and phrase-level non-compositional semantics (e.g., idiom chunks)--much of this goes back to Saussure. The point here is that the lexica of natural languages are in no way fully specified sensu stricto by UG, no matter the eventual number of distinctions therein. Furthermore, lexical(alized) distinctions and the arbitrary semantic partitions they establish are, a priori, orthogonal to (say) phonemic inventories and/or their complexity: there is no reason to expect complexity qua number of distinctions (as in WSG) to increase in lockstep across all levels of grammar.
In any case, the actual (phonetic and semantic) make-up of lexical items IS constrained by universal laws of phonetics, by universal constraints on semantic interpretation, constraints on argument structure and its linking to surface representations, etc. Take HC as an example. At the phonemic level, HC may look simpler than French or English. But, at the lexical and morphosyntactic levels, HC uses operations like reduplication and predicate-clefting for semantic stress. The PRECISE morphological, syntactic and semantic details of these operations are not necessarily "simpler" (i.e., with fewer overt distinctions) than the reduplication and cleft patterns that exist in "old" families like Romance and Germanic. Yet reduplication and predicate-clefting in HC must also obey UG strictures on a par with, say, affixation and verb-movement in Romance and Germanic.
As I have already mentioned, it is axiomatic that any given idiolect, though biologically bound by UG, will be overspecified with respect to UG's initial parametric and lexical slots and so on. These slots are necessarily left underspecified (i.e., open to parametric choices) in the initial state defined by UG. The crux of the matter is that, notwithstanding apparent cross-linguistic quirks, all such quirks and myriad others (including those found in Creole languages) will, by assumption, fall within the boundaries defined by UG and indeed will help us discover the make-up of UG. No amount of complexity-building via diachronic drift can take languages "beyond the bounds of the genetic specification for language" (contra WSG: §6.3; see (14) above). Indeed, it is tautological that our "genetic specification for language" (i.e., the genetic encoding of UG) enables us to learn ANY "overspecification"--ANY ""ornmamental" elaboration"--in ANY human language; that is, the mind/brain is genetically pre-wired to acquire, store, produce and process any and all the "overspecifications" that exist across the world's languages (see note 17 ). Any linguistic feature that could not be so acquired, stored, processed, etc., would just not exist in any natural language, assuming with Descartes, Humboldt, Osthoff & Brugmann, Paul, Chomsky and others that natural languages are mental properties of Homo Sapiens. Thus, the necessity for cross-linguistic research: it is linguistic diversity that will help us elucidate the boundary conditions imposed by UG. It is by apprehending the diversity of specific languages that we will elucidate the unity of Human Language.
This Cartesian methodology puts an ironic epistemological twist on the (neo-)Schleicherian claims in (29). Is it really "such a challenge to glean [UG] in older languages" while "Creoles represent a fundamental layer of natural language [that is] unobscured"? If Creoles were really "the world's simplest grammars" with the fewest distinctions possible, then this would actually make it HARDER for pro-prototype creolists to "glean UG" and the diversity it affords. Prototypical Creoles make the "prototypical" creolist's job most trivial: "the world's simplest grammars" require no more than the world's simplest analyses. In this perspective, Prototypical Creoles, as defined in WSG, would have little, if anything, to contribute to theoretical progress in linguistics (be it in phonology, morphology, syntax, semantics, etc.). As Chomsky (1986: 149) writes, "Quite often, the study of exotic phenomena that are difficult to discover and identify is much more revealing, as is true in the sciences generally".
Let's imagine a would-be analogue of (29e) in the hard sciences. Imagine, say, some hypothetical claim to the effect that chemists have their best shot at discovering the molecular make-up of nature by examining the world's simplest molecules. If the latter were the only data that chemists had to experiment with, then they certainly would have no clue about the diverse complexity of nature's "Universal Chemistry". In particular, chemists working with `prototypically simplest' molecules (chemistry's equivalent of "Prototypical Creoles") would have no opportunity to glean the complex structures of proteins and DNA--the very molecules that make our existence possible. Similarly, if our field notes and intuitions were exclusively about Prototypical Creoles as defined in WSG, then we would have no clue about inflection, tone, Saussurean lexicalizations of root-affix combinations, ergativity, grammaticalized evidential marking, inalienable possessive marking, etc. The list in (18) hints at other features that UG makes available, but that linguists could never glean from the hypothetical "simplest grammars" defined by that list (cf. note 36 ). In DeGraff 2001b: 76-78, 86-88, I argue that the Creole Prototype defined by (11)--by (11c) in particular--even lacks some of the basic Saussurean properties that are usually associated with natural languages and their lexicon (see Section 2).
If UG both defines the language learner's innate initial state and imposes boundaries on the outcome of acquisition and, thus, on each and every I-language, then it cannot be the case that UG is "a challenge to glean in older languages" because of their "incidental" features. To the contrary, UG is best studied through our exploration of the diversity of languages, and this exploration is best carried out when guided by our theoretical results about the unity of Human Language. In this view, Prototypical Creoles as theoretical constructs with simplest, unexpected and ad hoc properties do constitute a typological-ontological (and epistemological) challenge to (the study of) our faculté de langage and the cross-linguistic structural possibilities it affords. From the perspective of UG as sketched in this Section, Prototypical Creoles become "linguistic monstrosit[ies]" (cf. (4)).
Be that as it may, we still lack any operational criteria for "basic communication" and its "functionally central" properties. Yet at this point we must raise the central question: How is "complexity" defined?
First, some words of caution: Any descriptively and scientifically adequate complexity metric requires an independent theory of complexity (that explains what is to be counted, why and how) along with exhaustive descriptions for the languages to be compared (so we can list all that is to be counted), lest our complexity metric have exclusive scope on arbitrary bits of grammar with no consequence whatsoever for linguistic theory and global complexity across languages. No general claim about cross-linguistic levels of complexity is reliable if it focuses solely on a small set of disparate superficial patterns that are not unified in any kind of linguistic theory or psycholinguistics. More generally, simplistic and highly selective measures of complexity whose benchmarks focus on arbitrary and isolated aspects of surface strings in some handpicked sample of languages seem largely orthogonal to the theoretical and/or psychological foundations, and to the descriptive goals, of linguistic typology and theoretical linguistics (see notes 37 and 40 ).
I thus agree with Muysken (1988: 288) that "the idea that creole languages are not grammatically complex in general only makes sense if one has a theory of grammatical complexity to fall back on ...". Chaudenson (1994) makes similar points, as he notes the absence of any coherent evaluation metric in past and current allegations of extraordinary Creole simplicity. Without any independent theory and formal criterion for complexity, we cannot even begin to determine how particular properties (or absence thereof) contribute to global complexity. With this in mind, let's proceed to evaluate the most recent complexity metric in neo-Schleicherian creolistics.
The complexity metric in (30) is simply a count of "overt distinctions and/or rules" in (30a), which in turn is related to "length [of] descriptions" in (30b).
Bit-complexity immediately faces a number of unresolved methodological and theoretical problems that render it scientifically un-usable at best and tendentious at worst.
Let's paraphrase (30) in a transparent information-theoretic way. As sine qua non for a rigorous and objective application of the sort of complexity metric sketched in (30), we must at the very least get straightforward answers to the following questions:
In order to adequately answer the "" question in (31a) for any given language, no less is needed than (an approximation of) the description of this language at all linguistic levels (i.e., for all "area[s] of grammar [with] overt distinctions and rules", including phonology, lexicon, morphology, syntax, semantics, discourse, etc.; cf. (30a)). Much progress has been made in typological linguistics, yet such exhaustive descriptions are not generally available for all of the world's languages. In the absence of such descriptions, no can be reliably estimated toward an unbiased global comparison of all natural languages (e.g., in order to discover what "the world's simplest grammars" are). 36 37
Pending such exhaustive descriptions for all areas of Creole and non-Creole grammars, what one should expect from an analysis that partitions natural language into simplest and most complex classes is a general theory whereby one can safely extrapolate from the (apparent) complexity of isolated and arbitrary linguistic properties (see, e.g., (18)) to global complexity. No such theory is hinted at: the very features in (18) from a very small and selective set of (apparently) "far out" languages seem to have been picked exactly so that the few Creoles chosen as "test cases" show less bit-complexity in the corresponding domains than the few non-Creoles chosen as "control cases" (see note 37 ). What we've had thus far in many searches for the world's simplest languages are formulas for "rigged" experiments--experiments that are designed to guarantee the desired "results". 38
For the sake of the argument, I will now abstract away from the methodological issues in 5.2. Instead let's ask this: Is complexity only, or primarily, a matter of counting information bits--"overt distinctions and/or rules" as in (30a)--no matter the source of these bits (i.e., no matter the theory underlying these distinctions and rules)? If a complexity metric involves counting, then we better make sure we know what we're counting. Linguistic distinctions and rules are not pre-theoretical objects that we can gather and count without prior analysis. 39
In fact, bit-complexity bears no relation to any theory where grammatical phenomena are independently identified and analyzed. WSG (§§2.4.1-2.4.3 and n6) explicitly cuts off its complexity metric from the better understood areas of (psycho)linguistics, including grammatical theory, acquisition, production and processing. In other words, bit-complexity may well have no basis in (what we know about) Language in the mind/brain--our faculté de langage. Bit-complexity, as defined in (30)-(31) is strictly a-theoretical: this is literally bit counting with no concern for psychological-plausibility and theoretical insights. 40
At the onset, let's note one of the (bizarre) logical consequences of bit-complexity: The languages with the biggest lexica would be the most complex--independently of, say, their phonology and their syntax. Indeed, each new lexical item further partitions the speaker's semantic space (recall Saussure's view of the lexicon as a system of oppositions). For any given degree of complexity, a large enough lexicon automatically carries enough "overt distinctions" to make the corresponding (I-)languages complex to the degree. In this view, the proposed complexity metric (in particular, (30a)) applies even within the "same" (E-)language of a given speech community: within that community, the (I-)languages with the biggest lexica will unavoidably require the "length[iest] descriptions" (cf. (30b)) and will thus be the most complex languages ceteris paribus. This strikes me as a rather naive view of language complexity. Number of (superficial) "overt distinctions and/or rules" without regard for linguistic theory (assuming for the sake of argumentation that such sets can be made available) seems, to me at least, a rather crude and uninteresting way to approach linguistic typology.
A simplistic bit-complexity creolist could well try to save the "overt distinction" metric and argue that the lexicon altogether lies outside the scope of his metric. So let's now move from bit-complexity in the lexicon to bit-complexity in the syntax. For the syntax too, bit-complexity simplistically implies counting--here, counting of "rules":
One example will straightforwardly illustrate the extent of this fallacy. Let's consider the statement that "asymmetries between matrix and subordinate clauses (e.g. Germanic V2 rules)" entail an increase in complexity via the "processing of more rules" (see (32)). This is presumably because such asymmetries involve distinct "rules" for root vs. embedded clauses, thus an increase in the number of "overt distinctions and/or rules" (cf. (31a)). No analysis is presented for the (added) rules that underlie this added complexity--the latter is taken for granted, and mistakenly so.
Since at least den Besten 1981, it has been argued, and it now seems quite likely, that there isn't any "asymmetric ... Germanic V2 rule" per se. Typically, V2 in German(ic) results from the application of X- and XP-movement rules, both of which are made universally available by UG. The finite verb moves quite high outside of VP (e.g., to the C(omplementizer) position) while a maximal projection (e.g., some topic or operator) moves to the left of the verb (e.g., in Spec(CP)). The root-embedded "asymmetry" itself is understood by many germanicists to be epiphenomenal, emerging as a surface side-effect of the interaction between abstract syntactic ingredients (e.g., head and phrasal movement plus the features and contents of V, C, Spec(CP), etc., and their associated functional projections in root vs. embedded contexts).
Let's consider the German(ic) V2 asymmetry as an example--grossly oversimplifying the available analyses. German matrix clauses have a C(omplementizer) head that is usually empty, thus available as a possible landing site for V(erb)-movement, with the finite verb surfacing right-adjacent to a moved XP in Spec(CP), thus matrix V2. In German, the embedded C head is usually filled by an overt complementizer, which blocks V-to-C movement; the embedded finite V is thus stuck in the IP, in clause-final position; thus, the root-embedded asymmetry with respect to V2. Yet, in embedded clauses that allow an EMPTY C head, the "V2 asymmetry" disappears: V-to-C and XP-to-Spec(CP) take place, giving rise to an embedded V2 pattern that is "symmetric" with the root V2 pattern. Similar symmetry is robustly displayed in Germanic languages such as Icelandic and Yiddish. V2 in these languages can be argued to result from movement of the finite verb, not to C, but to a head lower than C, thus the lack of asymmetry since V2 does not depend on the (lack of) contents of C. Cross-linguistically, observed V2 (a)symmetries (as in German, Dutch, Yiddish, etc.) reduce to the complex interaction between X and XP movement (and the triggers and/or semantics thereof) and language-specific properties of clause structure and functional heads, etc. (See Vikner 1995 for an overview.)
Notwithstanding current debates about the exact mechanics of V2 in Germanic and beyond, the lesson from syntax about the alleged complexity of "asymmetric" rules is clear. The above (simplified) analyses for V2 teach us that "V2 asymmetry" does not necessarily entail "the processing of more rules" than (say) "V2 symmetry". Even "symmetric" languages such as Yiddish and Icelandic instantiates similar sorts of movements, albeit within a different clausal topology. At the right level of analysis, the so called Germanic "movement rule asymmetries between matrix and subordinate clauses" become a rather superficial side-effect of a single uniform operation--head-movement of V into C and XP-movement to Spec(CP)--which applies whenever possible. The blocking of V-to-C due to overt C arises via independent morphosyntactic requirements (e.g., selectional requirements and the morphology and semantics of the CP layer). In this view, root-vs-embedded (non-)V2 patterns in Germanic (and elsewhere) are not the result of distinct (root vs. embedded) transformations; instead such patterns result from deeper universal principles of the computational system of our faculté de language (e.g., structure building, selectional requirements, movement transformations, etc.) interacting with language-specific morphology and/of functional heads.
From this theoretical perspective, root-embedded asymmetries do not necessarily increase complexity in ways that are alien to Creole languages. In fact, Creole languages too manifest (superficial) root-embedded asymmetries, contra the erroneous generalization in WSG: §6. HC, for one, has such asymmetries with respect to wh-phrases in direct questions: the moved wh-phrase, no matter its underlying position, overtly moves to the root Spec(CP). So in HC direct wh-questions, the wh-phrase is pronounced in the root clause, never in an embedded clause; thus the appearance of a root-embedded asymmetry. This is unsurprising in any theory where (direct) questions need to be typed as such at the root level and/or where semantic operators need to take scope over their quantification domains. (Also see Syea (1997) for another possible instance of Creole root-embedded asymmetry, with respect to copula distributions and V-to-C movement; cf. the Mauritian Creole data in Appendix II.) In Creoles (as in non-Creoles), apparent asymmetries result from deeper universals of syntax. 41
More generally, syntactic theory in the generative framework has witnessed a fundamental move away from lists of language-specific and/or construction-specific "rules". Current generative syntacticians have adopted the Principles-&-Parameters/Minimalist hunch that "constructions" arise via the complex interaction between, on the one hand, operations and constraints that are universal and, on the other hand, language-specific properties that reside mostly in the lexicon--in particular, in (the morphology of) functional heads. The number of operations made available by UG may well be few (e.g. Merge, Move, Agree, etc.). Yet these few universal operations interact in complex ways with numerous language-particular properties, thus the vast and intricate array of superficially-distinct cross-linguistic phenomena (e.g. the "V2 asymmetry" in German, the wh-movement asymmetries in Haitian Creole, the copula asymmetries in Mauritian Creole, the Chinese in-situ wh-phases, etc). In such a framework, complexity does not reside in the number of different "overt distinctions and/or rules" (e.g. distinct/asymmetric rules for root vs. embedded German clauses). Such construction-particular and language-particular "rules" may not even exist, although their labels are often retained for taxonomic descriptive purposes (Chomsky 1995: 170).
The morale of the story is clear: There can't be any "counting" in syntax without an explicit theory of syntax that independently tells us what needs to be counted, and how. Any comparative approach that gives the "looks" of languages priority over the "essence" of grammar runs the risk of becoming a most simplistic and misleading linguistic measure. (McWhorter's (2001b) Afrogenesis Hypothesis runs a similar risk; see Appendix II.)
Bit-complexity in (30) is quite ambitious: it is meant to rank the entire grammars of the entire set of Creole languages against the entire grammars of the entire set of non-Creole languages, with perhaps the few exceptions noted in WSG: §6. Recall the central claim that Creole languages and non-Creole languages tend to fall at opposite ends of the bit-complexity cline. This claim is not only about Lahu vs. Saramaccan; or Tsez vs. Saramaccan; or Maori vs. Saramaccan. Nor is this claim to be evaluated with respect to only a handful of linguistic features such as those in (18). Bit-complexity has universal scope: it is a claim about all languages across all areas of grammar (phonology, lexicon, syntax, semantic, pragmatic, etc.)
With this in mind and given the arguments above, it is worth stressing again that bit-complexity as "tested" in WSG enlists only an arbitrary set of linguistic properties (such as those in (18)). These properties are picked from a handful of exotic non-Creole languages (e.g., those mentioned in paragraph above) without recourse to any independently-motivated theory of grammar, processing and/or acquisition. Thus, this metric has no principled implications for Language in the mind/brain (but see note 40 ). Given our current state of knowledge and the complex nature of Language, we can't yet afford a global complexity metric with global cross-linguistic scope. In the meantime, the handpicking of languages and linguistic features in implementing and testing the metric in WSG belies the purpose stated therein to elaborate "a direct comparison of certain Creole grammars with older language grammars, with a view towards making more precise my grounds for the claim that Creole grammars constitute a synchronically identifiable class" (WSG: §1). If the few "test" languages and the few "test" properties are both prejudicially chosen without regard to any independently-established criteria, then whatever we may learn from this comparison is not enough to equate Creoles to the (natural?) class with the label "world's simplest grammars".
Empirically it has already been argued above that a slightly larger sample of cross-linguistic data and typological/diachronic observations undermine neo-Schleicherian creolistics. The rest of this critique brings additional methodological and empirical observations that further undermine the bases of the new Schleicherian linguistics.
Here I focus on some of the specific empirical problems that undermine the "testing" of age-complexity correlations. In a nutshell, what we're dealing with is a set of "rigged" experiments where the "test" cases and "control" cases seem carefully handpicked to provide support for neo-Schleicherian creolistics hypothesis. But this empirical support will be shown to be illusory: the empirical claims in neo-Schleicherian creolistics ultimately lead to theoretical incoherence, specially vis-à-vis purported complexity-age correlations.
Pre- and Neo-Darwinian linguistics from, say, Schleicher 1863 to McWhorter 2001 rests on the following premise: Complexity increases with age--as a language gets older, it gets more complex (but see notes 5 and 24 ). It is further postulated that, after tens of millennia, old languages "all come to rest at a certain "surplus complexity quotient""--an evolutionary plateau of maximum complexity that excludes Creole languages (WSG: §2.3), even though a few old languages like the Riau dialect of Indonesian and Southeast Asian languages may have slipped from the maximal-complexity plateau and acquired "pidgin-level syntax" due to "extensive adult acquisition" (WSG: §4.4). Temporarily putting aside the theoretical elusiveness of the notions "language birth" and "language age" qua linguistic constructs sensu stricto (see Sections 2 and 3 above), I will argue that (neo-)Schleicherian complexity-age correlations are robustly disconfirmed by the available diachronic and typological data.
Consider inflection, for example, which is taken as a marker of complexity-cum-age:
Be that as it may, does inflectional morphology always increase with age? It has long been observed in the grammaticalization and historical-linguistics literature (see, e.g., Meillet 1912) that, from a diachronic perspective, grammatical systems (e.g., Case and Tense-Mood-Aspect--TMA--marking) often evolve along analysis-synthesis cycles whereby overt markers go through the ebbs and flows of syntax (analysis / periphrasis) and morphology (synthesis / word-level processes). Free-standing auxiliaries can become verbal inflectional affixes; pre-/post-positions can become nominal case affixes; and both verbal and nominal affixes can fuse to their stems and erode over time. Given such morphology-syntax cycles, hope springs eternal for any affix-less language with affix envy: "Weep not, my children, for today's syntax is tomorrow's morphology" (Givón 1971: 413n1; also see Hodge 1970, Giacalone Ramat & Hopper 1998, Heath 1998, Haspelmath 2000 and Janda 2001 for some discussion of (de)grammaticalization phenomena and their theoretical bases; also see note 46 ).
Schleicher himself was well aware of the erosion of morphology in "old" languages and clever enough to try and incorporate inflectional decay in his evolutionist scenario, distinguishing between "evolution" and "history" while strenuously holding on to his complexity-age correlations (see note 5 ). Classic instantiations of the rise-then-decline of morphological marking are common in the history of Romance and Germanic and, more generally, throughout Indo-European and elsewhere. Compare, say, verbal inflection and nominal case in Old English vs. Modern English; ditto regarding, say, the evolution of nominal case and verbal inflection from Latin to Romance.
Here, we get the exact opposite of the Schleicherian complexity-age correlation: Vis-à-vis case and verbal morphology in the relevant stages of Germanic and Romance diachrony, "older" implies "simpler", assuming bit-complexity as in (30). Modern English, for example, has substantially fewer overt distinctions in, and fewer combinations of, verbal inflection than Old English. Ditto in the domain of overt case morphology on non-pronominal noun phrases: Modern English has none of the overt case affixes that were once productive in Old English. WSG: 138 does note that "English expression of case is simpler overall than Latin's", but it is not noted that the Modern English expression of case is also much simpler than that of Old English, and so is the Modern French expression of case simpler than that of its ancestors--in terms of bit-complexity.
The English and French cases are counter-examples to the claim that "diachronic drift ... encrusts older grammars [with complexities]": diachronic drift has reduced the overt inflectional paradigms of both English and French, thus decreasing their inflectional bit-complexity. Therefore, it is not at all clear that inflectional paradigms can be used as reliable indicators of language "age" (also see the case of Riau Indonesian above). In fact, within single language families we do find languages with drastically different degrees of overt inflectional morphology. Compare (e.g.) Malayalam to other Dravidian languages, English to Icelandic, French to Italian, or Western European languages to Balto-Slavic languages as in (8) (also see notes 5 and 24 and the comparative data in Hodge 1970). It has long been established that much morphological variation is expected within any single linguistic phylum. Assuming that all languages within a particular family "go back" to a single ancestor proto-language (abstracting away from the theoretical difficulties in language "dating"), then such variation within single families goes against the notion that language age can be correlated with bit-complexity. Of course, this is reminiscent of Edward Sapir's memorable quote in (7); also see (8). 43
It must also be noted that acquisition processes (either in ordinary situations or in situations of abrupt language contact) exert an inevitable pressure toward regularization and/or morphological "simplification" in certain domains, even if this pressure is counteracted elsewhere in the grammar by other factors such as certain types of language transfer, grammaticalization, innovation and the like. It's not only in the diachrony of Riau Indonesian that we find "a [certain] degree of pidginization ... due to extensive acquisition by adults, having `shaved away' a large degree of accreted complexity" (cf. WSG: §4.4). Acquisition by adults with its potential for morphological erosion is a widespread phenomenon in language contact and language change (see notes 5 and 24 ). Furthermore, the standard fare in historical and contact linguistics teaches us, for example, that: (i) sound change often leads to loss/assimilation of phonemic distinctions; (ii) morphological change often leads to regularization and analogical leveling (i.e. to loss of morphological distinctions); (iii) syntactic reanalysis often leads to structural simplification (or structural transparency); etc. (see Campbell 1999 for an overview). 44
In fact, functionalists have often argued that ease of articulation, regularization and rule extension are at the roots of language change, which is driven by functional factors--principles of least effort, economy, optimization (see Labov 1994: 547-568, 2001: 16-28 for a critical overview in the context of "maladaptive" language change). Some of these principles of economy entail a reduction of bit-complexity: as Labov notices, there is a long series of arguments that language change--sound change, in particular--may be `dysfunctional', and in the 19th-century some of these arguments were advanced within language-as-evolving-organism approaches (but see note 5 for a sample of concurring and diverging opinions). Labov (1994: 586-599, 2001: 10-14) surveys a number of areas where sound change does reduce overt distinctions, thus bit-complexity. Labov remarks: "The almost universal view of linguists is ... that the major agent of linguistic change--sound change--is actually maladaptive, in that it leads to the loss of the information that the original forms were designed to carry. Though there is a wide range of divergent opinions on the nature of sound change ... there is general agreement on the negative character of this fundamental process" (Labov 2001: 10)".
In effect, this means that the "older" a language, the more opportunity these complexity-reducing "negative" processes would have had to reduce its bit-complexity, thus producing "loss of information" (in Labov's terminology). Here again, we get the equation "older = simpler" at least in certain domains, with "old age" entailing less, not more, overt distinctions (as with English and French case and verbal morphology). It is also expected that processes like grammaticalization (e.g., of free morphemes into bound morphemes) may ultimately offset some of the results of regularization, leveling, morphological erosion, etc. (cf. Givón's afore-mentioned quip "yesterday's syntax is today's morphology", which goes back to e.g. Bopp's, Humboldt's and Meillet's insights--Humboldt (1836 [1988: 205]; see note 5 ) talks about "the wearing-down of inflection [as] undeniable fact"; also see note 24 ).
What these cycles suggest is a picture in which absence of (overt) distinctions in one domain of grammar (e.g., morphology with its affixes and other word-internal processes) can well be compensated by distinctions in another domain (e.g., syntax with its periphrases, word-order distinctions, selectional requirements, etc.).
For example, loss of overt inflectional morphology has often been correlated with increased rigidity in word-order (see, e.g., van Kemenade & Vincent 1997 and van Kemenade 1999). Vis-à-vis case morphology and word-order, it has long been noted, starting perhaps with Meillet 1912 [1958: 147-148], that languages seem to balance out overt case marking with grammatical marking in the syntax (via, e.g., pre-/post-position and word order). The exact correlation is hard to pin down, but the general, if overly simplistic, impression is that richness of Case morphology tends to be proportional with the scrambling of noun phrases away from their base positions (to wit: the history of English and of Romance). If so, measuring only overt morphological distinctions at the expense of abstract syntactic information leads to an incomplete and misleading metric. Similar caveats apply to the interface between any two (abstract) linguistic modules and the trade-offs therein in terms of grammatical information encoding. It seems to me that computing global complexity requires a theory of how grammatical information is encoded within and across the various linguistic modules (cf. note 39 ). 45
Hawkins 1986 is one such attempt toward formalizing the trade-off between different sources of "complexity". Hawkins argues that, along the dimension of form-meaning opacity, English is more "complex" than German because of increased case syncretism in English. Independently of the merits of this proposal, the point here is that it can be logically argued that reduction of overt morphological distinctions (i.e., reduced bit-complexity in morphology) increases some other kind of complexity, namely form-meaning opacity which in turn increases semantic ambiguity. Hawkins places English and German morphology at opposite ends of the semantic-opacity continuum: German morphology allows for "a `tigher fit' between surface form and semantic representation" (p. 122). Then again, Enlish's increased "form-meaning opacity" due to case syncretism is somewhat compensated by its word-order which is more rigid than in German. Hawkins speculates that:
No matter the fate of Hawkins' arguments, his speculations shed further doubts on the theoretical validity of the simplistic sort of (bit-)complexity metric advocated in (30): it can indeed be argued on empirical and theoretical grounds that, at least in some cases, ABSENCE of "overt distinctions" (compare, e.g., case affixes in English vs. German) can itself be a distinction that should enter into some global complexity metric. This is reminiscent of Meillet's (1912) argument on how rigid (grammaticalized) word-order emerges to replace case morphology. Here again, we see why bit-complexity is of little use in the absence of a theory of grammar that motivates the items to be counted and relates them to larger linguistic concerns--with respect to mental representations, language acquisition, language processing, language change, etc. 46
Keeping these opposing trends in the picture, it must then be concluded that, in Creole genesis as in other cases of language contact, distinct grammatical domains in the languages in contact will belong to distinct points on (e.g.) their respective morphology-syntax cycles (à la Meillet-Hodge-Givón-etc.) and semantic-opacity cycles (à la Hawkins). With this in mind, one is forced to conclude that creolization (on a par with language change via language contact) will start, not "essentially from ground zero" (contra WSG), but from a contingent (i.e., sociohistorically determined) array of non-"ground zero" structural termini ad quo. The latter fit distinct points on diachronic cycles (e.g., with respect to their respective analysis-synthesis and semantic-opacity clines). These non-"ground zero" structural termini ad quo determine the ecology of the PLD available in the formation of Creole idiolects. In other words, various components of the (pre-)Creole grammar will be extrapolated (via language transfer cum restructuration, innovation, grammaticalization, simplification, etc.) from patterns that are located across the analysis-synthesis and morphological-syncretism continua, as a synchronic reflection of prior cycles in the respective diachrony of the "old" languages in contact. In a similar manner, the diachronic termini ad quem will unavoidably be scattered across the relevant cycles in various grammatical domains. Ultimately, the trace of this scattering (i.e., the resulting I-language grammar(s) whose structures are necessarily bounded by UG) will depend both on the PLD and the sociodemographic conditions--all of which result from a very complex set of historical contingent factors.
So, here too, the creolist-cum-complexity-creolist must employ much caution in dealing with the competing pressures on language complexity that arise from the sociohistorical cum typological specifics of the contact situation. As Hymes (1971a: 70) reminds us, we need "to recognize pidginization as a complex process, comprising the concurrence of several component process". At this stage, we're far away from the Schleicherian "older = more complex" perspective on language evolution. We're also quite far away from the ab ovo genesis of "born again" languages from "ground zero" complexity via "a radical reduction of [the] source languages into makeshift jargon" (cf. WSG: 149).
The empirically and theoretically responsible scenario is much more complex, thus much more fascinating, even if it removes Creole languages from the category of contemporary fossils of Language evolution, and even if it deprives creolists of a most simplistic account for some of the most complex cases of diachronic development.
How many non-Creole languages could pass the WSG-type (neo-)Schleicherian structural litmus test for old-cum-complex languages?
WSG `carefully' picks its benchmarks for "old language" complexity, not from (e.g.) Romance and Germanic (the well-documented source languages--the terminus a quo--of its Prototypical Creoles), but from Caucasian, Tibeto-Burman and Polynesian languages (which have nothing to do with the Prototypical Creoles being surveyed). The complexity "control" cases--Tsez, Lahu and Maori--would strike many linguists, not just creolists and Creole speakers, as "fearsomely elaborated" or "extraordinarily complex by any linguistic standard" (cf. WSG: §3.2; also see Trudgill 1989: 237 who can imagine "mainstream Europeans and North Americans find[ing Caucasian languages] "exorbitant" or "incredible""). Plus these three "control" language are all "esoteric" languages, languages spoken in small communities in relatively low-contact situations where, for social-network reasons, speakers are most likely to maintain (and perhaps even promote) features that may APPEAR complex and/or linguistically unusual (cf. Hymes 1971a: 73; Trudgill 1989: 236-237). Trudgill specifically cites the languages of the Caucasus, to which Tsez belongs, as prime examples of "low contact" languages. According to Grimes 1996, Tsez has approximately 7,000 speakers; Lahu (Shi), 600,000; and Maori, 110,000 (these are approximate totals across the relevant dialects). This sociolinguistic factor alone is a delibitating confound in these complexity experiments comparing Saramaccan with Tsez, Lahu and Maori. What should be asked is whether an hypothetical Creole derived from contact among, say, Caucasian languages of the Tsez-type (or Tibeto-Burman languages of the Lahu-type or Polynesian languages of the Maori-type) would end up looking like Saramaccan. (See notes 30 and 37 above for related methodological remarks.)
As it turns out, the neo-Schleicherian "deck-stacking" methodology, to the extent that I undertand how it can be applied without bias, unsurprisingly ranks the complexity of English alongside that of Saramaccan. Yet Saramaccan, but not English, is usually regarded as a most "radical" Creole. In fact, WSG considers English to be an "old" language--much "older" than "born again" Saramaccan.
This is how we can try and check whether English is "born again" or "old" by applying the criteria and methodology exemplified in WSG:
In a coda to the Lahu-vs-Saramaccan discussion, it is noted that "English is also less complex than Lahu in all but one of the features cited (derivation)" (WSG: 149). Yet, in the same paragraph, an ad hoc list of English properties is pulled out in order to argue that "English is more complex, according to our metric, than Lahu in a great many aspects". This is yet another vacuous argument since similar ad hoc lists of non-Lahu "incidental" properties could be produced for any language. Indeed it remains possible in principle to produce a similar ad hoc list of Saramaccan features whereby Saramaccan too would look "more complex, according to [the WSG] metric, than Lahu in a great many aspects". In fact, given the structure of UG (see 4.5 above) and given the contingent aspect of parameter-setting and of the lexicon, it is always possible (in principle) for any pair of languages and to produce an ad hoc list of grammatical distinctions that exist in but not in . Such lists exist for any pair of languages no matter whether and/or are Creole or non-Creole. The handpicking of scattered "incidental" features does not constitute explanation or result. The contents of such lists are truly "incidental", that is, language- and construction-particular in a relatively superficial way (i.e., with no explanatory power).
As a (facile) exercise, I will now overtly rig a neo-Schleicherian experiment and produce a list of "incidental" Haitian Creole (HC) features of the Lahu/Maori/Tsez type above whereby HC "is more complex, according to [the WSG] metric, than [certain varieties of French and English] in a great many aspects": 47
The above ad-hoc "incidental" list of HC non-English properties would thus lead to the conclusion that HC is more complex than English, if one assumes the style of argumentation in WSG. In addition, HC also manifests some (quasi) English-like features that constitute "overspecification [that] goes beyond the needs of a human grammar" (cf. WSG: §4.2). Such (quasi) English-like "ornamental"/"incidental" properties include:
One could argue that all the complex features of HC above--which, as in English, seem "incidental to basic communication"--have arisen "due to contact over the centuries with French" (cf. WSG: 143). But this would not work for a simple reason. Many of the features noted above have no direct counterpart in French. Furthermore, the French-like HC features (e.g. the auxiliaries ap and pral) seem to have been part of the language from very early on, resulting from the grammaticalization of French periphrastic constructions with après and après aller, respectively. Furthermore cognates of these preverbal TMA markers are regularly found in French varieties that are not labelled "Creole" (cf. Appendix II). 48
Such arbitrary and superficial comparison whereby "born again" languages may APPEAR more complex than "old" languages given some ad hoc list of language-specific features can be extended ad libitum. Indeed, for any choice of (an integer), (a so called "born again" language) and (a so called "old" language), can be the source of an arbitrary list of features that are absent in . Take, say, dual marking, whose presence in any language is considered (in WSG: §5.3) to increase complexity/markedness in the Greenbergian sense. Dual marking, which is absent in many "old" languages, does exist in a number of "born again" languages. In Cape York Creole, as described in Crowley & Rigsby 1979, the pronominal system offers both Dual and Plural in the non-singular forms (for 2nd and 3rd person), plus there is a grammaticalized distinction between inclusive vs. exclusive pronouns in the 1st person. If dual is dispensable from a communicative viewpoint (which it must be given its absence in many "old" languages), then Cape York Creole is surely more complex than (say) English in that respect. A similar example is provided by Chaudenson (1994: 50) who argues that dual marking in French-lexifier Creoles in the Indian Ocean (e.g., in Réunion) makes these Creoles more complex then their lexifier vis-à-vis the grammaticalization of number marking. Dual marking is also found in Taimyr Peninsula Russian-based Pidgin (Wurm 1996: 83). Such "incidental"/"ornamental" system of pronominal reference and number marking are not found in "old" languages such as many English and French varieties. Here too, we have "born again" languages with arbitrary distinctions that are not found in "old" languages (also see Section 4 for further examples of pidgin/creole features that are "incidental to basic communications").
It is worth repeating that, given the partially contingent nature of parameter-setting and lexicon formation, any number of arbitrary (and so called "incidental"/"ornamental") distinctions can, in principle, be found in ANY language, whether "old" or "born again". Indeed, the list of "incidental" features provided by creolists-cum-complexity-theorists in their Creole-vs-non-Creole comparisons (see, e.g., (18)) is truly incidental.
Lastly, one could argue that the reason why English's bit-complexity looks low when compared to Tsez, Maori and Lahu is that English is, after all, a "Creole" as can be perhaps adduced from its history. Such argument would go on to claim that English's history of contact is the reason why so many Creoles (such as HC and Cape York Creole, say) have features that are at least as "complex" as their English counterparts. Well, I suspect that many other languages besides English would also rate low given the complexity metric in WSG and its exotic benchmarks and "incidental features" list. In any case, the English-as-Creole argument would drastically weaken the theoretical bite that many would like to assign to the term "Creole", specially in the absence of an unambiguous STRUCTURAL definition for "Creole". 49 In any case, English IS claimed as one of these multi-millenarian languages that are generally more complex than young languages (see, e.g., WSG: §4.2).
In fact, it can be argued that language contact in the history of English did contribute to reducing bit-complexity in restricted domains, as in (e.g.) inflectional morphology. Meillet, Weinreich, Trudgill and Chambers, among many others (see notes 5 and 24 ) have argued that language learning in contact situations induces "erosion" of morphology and/or regularization of overt morphological distinctions. And it's been argued many times before that language acquisition itself can induce various degrees of `simplification' in the concomitant new idiolects. 50 Thus, any language (whether "old" or "born again") can undergo certain sorts of "simplification", independently of its Creole status or origins.
What the above remarks suggest is that the complexity-reducing effects of "pidginization" in certain domains (e.g., in overt inflectional morphology) are quite widespread--even when there are no recognizable pidgins(-to-be) in sight (recall that (I-)pidginization as a individual-level process is in principle distinct from the creation of stable (E-)pidgins--the latter crystallize through the focusing of norms and other group-level sociolinguistic processes). This, of course, is not a novel observation. Schleicher himself considered that language contact (e.g., in the history of English) was a degenerative factor while Humboldt took the "undeniable fact" of inflectional erosion as a sign of intellectual maturity (see note 5 )! It's been commented over and over again in the sociolinguistics and historical linguistics literature that language contact, across space and time, often entails structural simplification in various domains (see note 24 ). As Hymes (1971: 73) wrote, "simplification may prove to be, not an isolated phenomenon, but one pole of a continuum applicable to ... all languages" (see note 50 ). Others have proposed that one key factor driving structural simplification in various domains is vehicularization (i.e., the sociohistorical process by which certain languages become lingua francas, as in the case of English; see Mufwene 2000b, 2001). At the individual level, the simplification processes often noted in language contact and vehicularization must be ultimately rooted in the cognitive capacities and constraints that underlie second language acquisition (see, e.g., DeGraff 1996, 1999b,c,d for discussion and references).
The various lists elaborated by (neo-)Schleicherian linguists in the past three centuries in order to isolate "the world's simplest grammars" still seem irreducible to any fundamental principle(s) of linguistic theory. Thus far, these lists cannot count as scientific explanation for any robust set of linguistic phenomena. At best, they identify scattered "historical accidents" in scattered domains of grammar in scattered samples of languages.
In any UG-based framework (along the lines sketched in 4.5), language-specific "incidental overspecifications" (such as opaque lexicalizations, pro-predicate morphemes, dummy verbs, dual marking, ergativity, grammaticalized evidential marking, pharygealized uvulas, clicks, etc.) ARE "historical accidents"--the sociohistorically contingent choices made by particular idiolects cum societal conventions within the boundaries set by the biological necessities of our faculté de langage. Given the rich linguistic ecology and the (socio)linguistics of language contact, it is no surprise that "historical accidents" of various sorts also happen in Creole genesis, as documented above.
One key question facing modern linguists is: What is the structure of UG such that, across the species, language learners faced with incidental and relatively shallow PLD unfailingly (re-)create idiolects with all sorts of abstract complex properties that are not evident from the PLD (cf. Chomsky's (1986) "Plato's Problem").
The central assumption here is that every idiolect is somewhat created anew at every instance of acquisition. A related, but distinct, assumption is also found in the grammaticalization camp. There too, notions like "old" and "new" languages seem to make little theoretical sense: "Students of grammatization realize that worrying about where one gramamr ends and the next grammar begins is a totally meaningly and futile pursuit. For the `new gramar is constantly being created on top of the wiling and yielding ruins of the old' ..." (Matisoff 1991: 447).
From the Cartesian-Uniformitarian perspective espoused here, it can be reasonably argued that ALL (I-)languages evolve via an initial "break in transmission": grammars are not inherited, but (re-)created (Paul 1890; Meillet 1929; Halle 1962; Chomsky 1981, 1986, 1995; Lightfoot 1999; etc.). In each of its individual instantiations, language acquisition sensu stricto is not language transmission, but UG-guided language (re-)creation with contingent, sparse and heterogeneous PLD drawn from idiolects (i.e., speakers) in contact (Chomsky 1981). The exact and particular nature of this "contact" is sociohistorically determined and so are the tempo, amplitude and group-level effects of the individual "breaks". Yet, whether in "Creole genesis" or through "language change", the PLD always underdetermine the attained grammar: there always exist structural "breaks" (however subtle) between "old" and "new" idiolects. Thus, the ineluctability of language change/creation. Whatever general tendencies may exist across instances of Creole genesis partake of the same mental processes that underlie all other cases of (I-)language creation.
As of "(I-)pidginization", to the extent that it reduces to (aspects of) second-language learning (L2A) in adulthood, the "pidgin-to-creole life-cycle" will have congeners far and wide, in all cases of language contact where non-native utterances contribute to the PLD that are used in the formation of native-idiolects (L1A). In a Cartesian-Uniformitarian perspective, the "pidgin-to-creole cycle" is naturally mirrored by an L2A-L1A cyle--or L2A-L1A "cascade relationship" in the terminology of DeGraff 1999b: 497, 504. In this vein, the discrepancies between "old" and "new" idiolects may seem more dramatic in the Creole cases than in the non-Creole cases, but it can reasonably be argued that the difference (if any) is a matter of degree, rate of spread, and/or subjective perception, not of quality.
If so, the sui generis structural category "world's simplest grammars" with "born again" genealogical status is only "chimera and reverie" and, worse yet, "linguistic monstrosity", as hinted at by Foucault in (4). Thus, the fallacy of dualist Neo-Darwinian scenarios for the origin of Creoles. Given UG and given the sociolinguistics and ubiquity of language contact, there is not, and there could not be, a constant and exclusive set of Creole structures that are FUNDAMENTALLY special, across time and across space, independently of the specific linguistic ecology (see references in note 12 ). I thus agree with, inter alios, Mufwene and Muysken, as per the following quotations (compare with (10)):
"What we have everywhere seems to be simple evolution of languages from one state to another in different ecological conditions" (Mufwene 1998: 324)
In my native Haiti and elsewhere in the Caribbean and wherever else we find Creole speakers, arbitrary (pseudo-)linguistic measures are still employed to "classify humanity" and de-humanize (monolingual) Creole speakers (see refererences in note 9). This "linguistic apartheid" is undermined (theoretically, at least) by observations like those in (36); also see the quote in (10) and the references in note 12 . As a creolophone creolist, I find that there is grandeur in this (Cartesian-Uniformitarian) view of Language: Creoles, on a par with ALL other languages (irrespective of genealogy), are reflections of our (species-uniform and species-specific) human biology, which is among the "most beautiful and most wonderful [forms that] have been, and are being, evolved" (cf. Darwin 1859 [1979: 459-460], Chomsky 1995: 1-4, 2001: 2). 51
As we saw throughout the main text (see, e.g., 2.1, 3.3, 6.1, 6.4), there is robust, inescapable evidence that HC--the Creole formerly known as "most Creole of Creoles", " `pure' [Creole Prototype] case" and "basilectal Creole" (McWhorter 1998: 809, 812; McWhorter 2000b: 206)-- is far removed from a structurally "simplest" Creole prototype with ancestry in a structurally "simplest" affixless pidgin. For instance, HC affixes (with cognates in French affixes) straightforwardly disconfirm the catastrophic pidgin-to-creole scenario whereby HC affixes would have emerged via grammaticalization of erstwhile free morphemes. How would a neo-Schleicherian creolist reconcile the postulation of a pidgin-to-creole catastrophic cycle (and its radical morphological bottleneck) with the well-documented French-based affixes of HC qua "most Creole of Creoles", "basilectal Creole", etc.?
McWhorter (2000a: 107) claims that "[HC] fits quite neatly into the Creole Prototype model as a case ... in which contact with the lexifier over the centuries pulled the creole away from the Prototype to which it honed at its genesis". This claim is somewhat echoed in WSG: §3.2 where HC, "due to contact over the centuries with French", is considered to "have borrowed many French lexicalized derivation-root combinations and thus do not exemplify the Creole Prototype in the purest possible form". Such a scenario is ahistorical given established socio-demographic facts of Haitian history as sketched in 2.1 above (also see notes 15 and Appendix II).
On the empirical front, McWhorter (2000a: 106) mistakenly relies on what he calls "[Goyette's (2000) demonstration] through painstaking historical analysis that the derivation markers in modern [HC] ... cannot have been incorporated into the Creole at its birth, and in fact were borrowed from French in later periods."
Not only is Goyette's scenario as a-historical as McWhorter's (see the sociohistorical sketch in 2.1) , Goyette's empirical generalizations about (the diachrony of) HC and French are flawed, and so are his deductions:
Take, for example, Goyette's discussion of the timecourse of the HC prefix re- /re/ (cf. French re-). From Fattier 1998, he quotes the lone HC form ekile /ekile/ `to move back' (a variant of HC rekile /rekile/). Goyette claims ekile as a cognate of some (dialectal) metathesized French variant erculer (cf. Standard French reculer /røkyle/). In Goyette's scenario, erculer's counterpart in HC is ekile and not erkile since most HC dialects forbid syllable-final /r/. Why is the single HC form ekile, taken from Fattier's (1998) 6-volume dissertation, of such significance? Goyette's argument is based on the premise that "[i]n 17th-century French, [the prefix re-] was CONSISTENTLY metathesized" (emphasis added) while "where /re/ is a prefix [in HC] such metathesis is wholly unknown". Thus, Goyette argues, the first syllable /e/ in HC ekile is the un-productive remnant of the metathesized French prefix er-; crucially the productive prefix re- in contemporary HC was not part of proto-HC morphology. Goyette's conclusion: re- in contemporary HC was borrowed late, not inherited early.
To the extent that I can understand Goyette's argument, it seems to contain at least one empirical flaw, even if one abstracts away from Goyette's overly simplistic claims about the phonology, distribution and diachrony of the er-/re- alternation across French dialects (the complex diachrony of French dialects sheds doubt on the categorical claim that "[i]n 17th-century French, [the prefix re-] was consistently metathesized"). Given (inter alia) that ekile and rekile have the same initial vowel /e/, Fattier (1998) reasonably takes ekile as a case of apheresis, not metathesis. Fattier also documents a third variant, which she writes ekile, where the first segment (the superscript /r/) is a phonetically-weakened variant of the /r/ in rekile (see Fattier 1998: v1: 230-232, v2: 449 for details). Furthermore, throughout her 6-volume thesis, Fattier documents robust cases of apheresis in a variety of environments. Similar apheresis is documented in Ducurjoly's (1802) Creole-teaching manual. The latter also documents robust affixation in early HC, virtually all of it derived from French affixes (including apparently UN-metathesized re-). Thus vanishes Goyette's single data point in arguing for the late borrowing of HC re-.
The other HC suffix discussed by Goyette (2000) is agentive -è as in mantè. For Goyette:
"[HC -è] is the normal reflex of final French -eur [pronounced /r/] ... IN THE SEVENTEENTH-CENTURY, [THIS FRENCH] SUFFIX LACKED A FINAL /R/ ... (menteux instead of menteur `liar'). ... We should expect the modern [HC] form of the suffix to be [/e/ since French /ø/ is mapped into HC /e/]". (emphasis added)Here too, Goyette's scenario is overly simplistic and empirically inaccurate. Although exact dates are uncertain, it seems that the reduction and dropping of final /r/ in French -eur (as part of a larger pattern of final-consonant reduction) was most robust in the Middle French period up until the 16th century. The opposing trend (pronunciation of final /r/) gained strength in the second half of the 17th-century, becoming re-established in many 18th-century dialects and their descendants up to the present.
So could it be true that "[i]n the seventeenth-century, [the -eur] suffix lacked a final /r/"? The crucial observation here is that, even at its apex, final-/r/ reduction in French was not categorical across all dialects (see, e.g., Fouché 1966: 669 and Morin 1986: 173-175). Instead many dialects manifested (phonologically-, morphologically-, semantically- and sociolinguistically-conditioned) variation between dropping and retention of final /r/. In some cases, final /r/ would enter into external sandhi phenomena, being pronounced before a vowel or a pause. In other cases, final /r/ was pronounced for emphasis or semantic nuance or for ill-understood sociolinguistic reasons (e.g., hypercorrection). A very telling case is the 1547 example Ajoustes si tu veux les Perfumeux, les Balleurs `Add if you want the Perfumeux and the Balleurs' where -eux and -eur alternate in the very same sentence (Brunot 1906: 290; also see Brunot 1913: 211-212, 1924: 671, for other instances of variation with semantic and/or sociolinguistic nuances and for further remarks on the diachronic course of final-/r/ reduction).
In addition to the variable rule of final-/r/ reduction, there are other factors affecting -eur in French diachrony. Discussing the passage from /o/ to // or /ø/ through Old and Middle French (cf. flor fleur, dolor douleur, etc.), Nyrop (1899: 163) mentions 15th- and 16th-century dialects where and ø do not exist and where sur and sur enter into "near rhymes that are said to be either `Provencal' or `Gascon', either `Normand' or "from Chartres"". Nyrop adds that such near rhymes are widespread in the 15th and 16th-centuries. Also relevant here is the following synchronic alternation: chur/choral, docteur/doctoral, fleur/floral, murs/moral, pasteur/pastoral, etc. (I thank Dominique Fattier for pointing this alternation to my attention.)
Pending further research, one can reasonably speculate that the above alternations in French diachrony and synchrony are related to another case of variation in HC's synchronic morphophonology. As Fattier (1998) repeatedly notes (also see Freeman & Laguerre's (1998) dictionary), è/ò alternation in HC word-final syllables is quite widespread, although not generalized: flatè/flatò `flatterer' (cf. flate `to flatter') gadè/gadò `watchman' (cf. gade `to watch'), vòlè/vòlò `thief' (cf. vòlè `to steal') etc. (also see fiyèl/fiyòl `godchild', lè/lò `time', sè/sò `sister', etc.); see Fattier 1998: v1: 132, 269, v2: 501, 740, 755. This è/ò alternation does not seem to have been borrowed late from French as spoken in Haiti: as far as I can tell, contemporary Haitian French manifests no analogous alternation or remnants thereof (see Pompilus 1961 for a sketch of Haitian French). (Although Goyette mentions HC mantè from Fattier's thesis, he fails to mention the variant mantò and the corresponding alternation vis-à-vis the -è suffix.)
On the conceptual font, it must noted that, unlike (say) Québec French (per Goyette), many contemporary (non-Creole) French dialects (e.g., Standard French and Haitian French), on a par with HC, do not productively use er- for re- and neither do they productively drop their final /r/ in -eur /r/ (contrast with Québec French erculer and menteux as cited by Goyette). Yet the absence of such alternations surely does not constitute evidence that the contemporary affixes re- and -eur in these French varieties sans er- and sans final-/r/ dropping "were borrowed ... in later periods". Where would these "later period" French speakers "borrow" the (UN-metathesized) re- prefix and the /-r/ suffix (WITH phonetic realization of final /r/) if ALL prior dialects consistently lacked such affixes?
Now, let's assume (purely for the sake of argument) that (some? most?) French dialects in 17th-century Haiti did metathesize re- into er- and did drop final /r/ in (e.g.) -eur. Then, whatever (socio)linguistic processes led to contemporary absence of er- and the contemporary retention of final-/r/ in (the "later periods" of) Standard French, Haitian French, etc. could, in principle, also account for the corresponding facts in HC without evoking an (unattested) affixless-pidgin stage. In any case, we must reckon with the complex French and HC variations noted in this appendix. Pending detailed work on the (socio)linguistics of Caribbean French colonists and their entourage, it is quite unlikely that all the relevant French varieties would "consistently metathesize [-er]" and that all "lacked a final /r/ [in their pronunciation of -eur]". (See, e.g., Chaudenson & Mufwene 2001: 145-153 for preliminary remarks and caveats on the intricate mix of Romance lects--standard, "patois", non-native varieties, koines, etc.-- in the colonial French Caribbean and elsewhere in the New World.)
To sum up, the data and observations that disconfirm Goyette's (and McWhorter's) late-borrowing scenario are taken from Goyette's own bibliographical source (namely, Fattier 1998) and from well-known facts about French diachrony (see, e.g., Nyrop 1899, 1903; Brunot 1906, 1924; Zink 1986; Morin 1986; Pierret 1994). Lastly, Goyette's (2000) categorical claim about the postulated (but undocumented) existence of a totally affixless proto-HC is based both on faulty logic and on erroneous empirical generalizations about (the diachrony of) only two affixes--not a representative database, by any measure.
All available evidence, coupled with theoretical considerations, suggests that the ancestors of HC as spoken in colonial Haiti (then known as Saint-Domingue) were never affixless, contra the claims of the "classic" pidgin-to-creole scenario (see the data, argumentation and references in 2.1, 3.3 and Appendix I above). But what if the "real" ancestor of HC goes back even further, to some UNDOCUMENTED French-based proto-pidgin spoken somewhere in West Africa, say in Senegal, around a slave fort? Is it, then, this hypothetical pidgin qua HC's proto-ancestor that would have most closely honed to the ("simplest") Creole Prototype?
McWhorter's (2000b) "Afrogenesis Hypothesis" (hereafter AH) pushes the origin of Caribbean and Indian Ocean Creoles back to a small number of hypothetical pidgins that would have been created around West African slave castles (cf. Goodman 1964: 129-132). The AH considers Caribbean and Indian Ocean French-lexicon Creoles as expansions of a SINGLE French-based pidgin ancestor that was created around a slave-trading fort on the Senegalese coast in mid-17th century. The AH is empirically, methodologically and theoretically flawed. Only a small sample of these flaws can be discussed in this appendix, abstracting away from the lack of historical evidence (see, e.g., Bickerton 1998).
Empirically, the AH's linguistic evidence is based on a misleading and skewed survey based on superficial comparisons. The latter are string-based with no attempt whatsoever at structural, distributional and semantic analyses. Independently of robust and documented morphosyntactic and interpretive differences, any superficial similarity between, say, TMA markers across Caribbean and Indian Ocean French-lexicon Creoles is judged "too close to be attributed to chance" and taken to "suggest common ancestry" in the hypothesized SINGLE Senegalese French-based pidgin (McWhorter 2000b: 149). Given the granularity of such comparison, a host of fundamental differences among French-lexicon creoles (vis-à-vis their syntax and semantics) are either ignored altogether or dismissed as "minor paradigmatic variation [that] is not counterevidence to a common [Senegalese pidgin] parent".
Take Tense-Mood-Aspect markers. On one hand, robust TMA-related differences across French-lexicon Creoles are amply documented. On the other hand, non-Creole French varieties (e.g., Québec French, Missouri French, Cajun French and 17th-/18th-century French) exhibit cognates of the same preverbal TMA FORMS that are enlisted from HC and Mauritian Creole as support for AH. (The references in note 48 document a variety of (dis)similarities in TMA and clause structure across Creole and non-Creole French-related varieties.) Do the "Creole-like" preverbal TMA markers in regional and diachronic French varieties trace back to a single pidgin spoken somewhere in Africa? As many have noted, the origin of these markers is, in all likelihood, not from a Senegalese French-based pidgin, but from (the grammaticalization and, in some cases L1-influenced, restructuration of) verbal periphrases in earlier French varieties.
More generally, do loose and superficial similarities in the phonetics, distribution and interpretation of preverbal TMA markers, IN ADDITION TO SYSTEMATIC MORPHOLOGICAL AND LEXICAL CORRESPONDENCES, constitute "conclusive evidence of a common origin" (cf. McWhorter 2000b: 148-151, 178-179, etc.)? If so, then HC and Mauritian Creole readily join Québec French, Cajun French, Missouri French, etc., as bona fide co-descendants with common origins in FULL-FLEDGED varieties of French. At this point, the discussion of (non)common PIDGIN origins for French-lexicon Creoles becomes moot. (Also note that Baker's (1995: 14) survey of pidgin and creole characteristics tentatively takes the combination of preverbal markers as one of the few "potential candidates for linguistic features which might distinguish creoles from pidgins". If Baker is right, then the "rudimentary" would-be pidgin spoken around 17th-century West African slave castles could not have provided any stable structural model for the complex structures of HC and Mauritian Creole's TMA systems, contra the premises of the AH.)
Related methodological remarks apply to McWhorter's (2000b: 151-155) use of HC's ye and Mauritian Creole's ete as evidence for the AH. (See, e.g, HC ye in Se yon lengwis Bouki ye `It's a linguist that Bouki is' and Mauritian ete in En voler Malis ete `A thief Malis is'.) HC ye and Mauritian ete are classified, without structural analysis, as (so called) "exposed copulas" and taken as tell-tales of HC and Mauritian Creole's common ancestry in a single Senegalese pidgin.
One problem here is that McWhorter skirts around central empirical and theoretical details of HC and/vs Mauritian predication patterns and their (language-particular and universal/theoretical) implications. Here are some tidbits of ongoing research and debate on the nature of HC ye vs. Mauritian Creole ete, just enough to illustrate the fragility of the facile comparisons that underlie the AH.
On the Haitian side, I myself have analyzed ye as a morpheme that spells out certain traces of non-verbal (i.e., [V]) predicates when the latter move outside the clause (i.e., outside IP into some operator position) for contrastive stress or wh-formation; see DeGraff 1992a,b, 1995, 1998, 1999c (cf. (34h)-(34j) above). In Se yon lengwis Bouki ye, ye would spell out the trace of the displaced nominal predicate yon lengwis `a linguist', even when the (phonetically realized) trace is not in sentence-final position as in Se yon lengwis Bouki ye vre `It's a linguist that he really is' (like HC ye, Mauritian ete does surface even when not "exposed" in sentence-final position; Syea 1997: 34). This disconfirms McWhorter's (2000b: 152) generalization that "copular overtness is sharply restricted to sentence-final position"; "exposed copula" is somewhat a misnomer. In reality, spell-out of [V] predicate traces by ye is subject to subtle syntactic and semantic constraints, having to do with the licensing of movement and traces and with the syntax-semantics of quantification. These constraints are rooted in UG, even if they result in distributional and interpretative details that appear specific to HC.
On the Mauritian side, Baker & Corne (1982: 46, 103) argue explicitly against the copula status of ete while Baker & Syea (1991: 172) take ete to result from Case-assignment requirements. As of Syea's (1997) analysis, it takes ete to result from the need to strongly head-govern a predicate trace; in root questions such as Kot Malis (ete)?, strong head-government of kote's trace is ensured either via ete or via (the co-indexed trace of) a null copula that has moved from V to C and agrees abstractly with the moved predicate in Spec(CP). Though elegant, this analysis does not account for the (apparent?) restriction of ete to wh-, nominal and prepositional predicates (do we get ete with movement of verbal and adjectival ([+V]) projections?). Neither does it account for the facts noted by Baker & Syea (1991: 167-170) whereby ete in contemporary Mauritian Creole is OBLIGATORY in matrix non-negated present-tense questions with ki, ki kote and other wh-phrases distinct from kot (but see Syea (1997: 28-29) for crucially different (dialectal?) judgements). Lastly, if the null-vs-overt alternation is driven by economy considerations vis-à-vis ECP satisfaction ("null except when it can't be"; Syea 1997: 52), then the null-vs-overt alternation in Kot Malis (ete)? is incorrectly ruled out. Of course, these are all delicate theoretical problems that McWhorter's superficial "description" has nothing to say about.
Most relevant to the discussion here, four observations are in order on the cross-creole/cross-linguistic syntax of (so called) "exposed copulas":
Observations 1 and 2 disconfirm McWhorter's (2000b: 152) claim that HC and Mauritian show "the same occurrence pattern", which would originate from a single common pidgin. Here, again, detailed distributional and structural analysis is of the essence: linguistics, after all, is about structure, not strings, since what you see is often NOT what the structure gets (see 5.3 above for related remarks). Observations 3 and 4 illustrate the methodological perils of scant and skewed comparisons as "linguistic evidence" in phylogenetic speculations: the structural resemblances in 3 and 4 surely do not suggest that HC, Mauritian Creole, English, Irish, Hebrew, Arabic and Russian all descend from a Senegalese French-based pidgin.
In any event, as analyzed thus far, the HC predicate-movement strategies (see (34h)-(34j)) and their Mauritian analogues are built on the intricate interaction of delicate morphosyntactic and/or semantic constraints, and so are their TMA systems. As such, they could hardly qualify as pidgin features, specially given the definition of pidgins in WSG: §2.3 as youngest/simplest "rudimentary codes" that eschew all but the "functionally central" (also see Baker's (1995: 14) comment, cited above, on pidgins' apparent lack of TMA combinations). The predicate-movement and TMA strategies in HC and Mauritian do not exist in many (functional) "old" languages. Therefore, the syntax and semantics of TMA and predication in HC and Mauritian--alleged telltales of a common pidgin ancestor--could not have been part of any pidgin that was created on the Guinea Coast as "rudimentary code [that is unlike] full language".
Another conceptual flaw in the AH argumentation concerns its idiosyncratic use of the comparative method. Take, say, Romance languages and the uncontroversial fact that they, like French-lexicon Creoles, exhibit structural correspondences aplenty at various levels of grammar (including across their Latin-derived lexica). Pan-Romance correspondences are more reliably documented and more numerous than McWhorter's few HC-Mauritian correspondences (the latter number a dozen or so). The logic of the AH comparison and argumentation, when applied to Romance, would have us erroneously conclude that all Romance languages originated in a single locale via a single contact language created in a single encounter. However pan-Romance similarities are not due to a single encounter with Latin; they are due to historically and geographically separate encounters between related varieties of Latin and diverse "substrate" languages. The Romance case teaches us that it is an error to claim monogenesis from a single pidgin in order to explain grammatical correspondences among certain French-lexicon Creoles (e.g., HC and Mauritian Creole), lest we throw away our usual comparative-historical heuristics when dealing with Creole genesis.
A logical flaw in AH concerns the (non-)evidence for a common French-based PIDGIN ancestor vs. related evidence for FULL-FLEDGED French ancestors. If pidgins are structurally-reduced lowest-common-denominator compromises among the source languages--"rudimentary codes not fulfilling the needs of full language ...[that] eschew all but the functionally central" (WSG: §2.3; also see discussion in 4.2 above)--then the few bits of superficial comparative data in the AH also count, in principle, as evidence for common ancestry in (native and non-native) FULL-FLEDGED French varieties, as spoken in the relevant contact situations. Given assumptions in the AH (and in WSG), full-fledged French varieties (or, rather, simplification thereof) contributed to the reduced patterns in the hypothesized Senegalese (proto-)pidgin--recall that pidginization ""shaves away" a large degree of accreted complexity" from the source languages (WSG: §4.4; also see McWhorter 2000b: 4). If so, there logically is no way to demonstrate single ancestry for (e.g.) HC and Mauritian Creole in a common French-based PIDGIN while excluding separate ancestry in independent situations of contact between the corresponding FULL-FLEDGED varieties of French and (some of) the corresponding FULL-FLEDGED substrate(s), unless it can be documented that speakers of the relevant varieties entered into contact once and only once, in one and only one locale.
As den Besten, Muysken & Smith (1994: 88-89) write, any sort of theory of monogenesis from a single pidgin is "fundamentally flawed" and "completely irrational" because:
"A UNIQUE example of any TYPE of phenomenon connected with human conceptual and cultural activity is just inconceivable--anything that can happen once can also happen more frequently".If (some) adult plantation slaves in Haiti, Jamaica, Barbados, Cuba, Columbia, Mauritius, Seychelles, Mauritius, Réunion, New Caledonia, etc., did approximate some (ANY) variety of some European language (cf. (ii) and (iv) in note 15 ), then they, like language learners everywhere, could not have acquired that variety overnight. This is specially so in the psycho-social context of colonial plantations from the perspective of the African-born who were taken to the colony as adults (see the contemporary reports in Pelleprat 1655, Girod-Chantrans 1785, Moreau de Saint-Méry 1797, Descourtilz 1809, etc.). These African-born slaves (the bossales) must have passed through a "pidgin(ized)" (QUA EARLY INTERLANGUAGE) stage with some structural features similar to those of the corresponding "pidgins" created by their in-situ compatriots who dwelt around African slave forts. It is not accidental that the speech of the Bossales--the numerical majority on 18th-century Haitian plantations--was often ranked as markedly "inferior" and "unintelligible" as compared to the speech of the locally-born ("Creole") slaves. Sociolinguistic factors, some of which remain to be elucidated, would determine the eventual fate of these early pidgins/interlanguages in the Old and New Worlds. However Cartesian-Uniformitarian assumptions about language acquisition/creation (see 4.2 and 4.4) guarantee the existence of these "pidgin(ized)" varieties, at least as transitory individual-level lects, across all instances of language contact in the Old and New Worlds and beyond.
As for full-fledged and stable French-lexicon Creoles, systematic correspondences between them and the corresponding FULL-FLEDGED varieties of French as spoken in, say, the then-colonized Caribbean and Indian Ocean islands (e.g., systematic correspondences at the level of morphology and lexicon; see 3.3) are many times more robust and numerous than the few (about a dozen) superficial correspondences claimed by the AH as genetic tracers linking Caribbean and Indian Ocean French-lexicon Creoles. Yet these systematic lexical-morphological correspondences across Creoles and their respective lexifiers, while they belie exclusive ancestry in a structurally-reduced pidgin, are not taken into account by the AH. In effect, the AH enlists scant and skewed correspondences among Caribbean and Indian-Ocean French-lexicon Creoles in order to argue for the un-broken transmission of some hypothetical French-based Senegalese pidgin while it discards massive and robust correspondances between these Creoles and their French lexifier in order to argue for a radical break in the transmission of French. This methodological paradox is unlike standard practices in comparative-historical linguistics.
Keeping the latter in mind, we can conclude that CERTAIN commonalities among, say, Mauritian and HC need not be due to "the SAME encounter with French" and need not "trace back to the same pidgin" (contra McWhorter 2000b: 147, 150). Like in the better understood Romance case, commonalities can independently arise from SEPARATE encounters (in the plural) among overlapping sets of languages and from universal strategies of language acquisition/creation. Common ancestry and common patterns do not necessarily entail common birthplace. In this perspective, differences among French-lexicon creoles (SOMEWHAT on a par with differences across Romance) are due to, inter alia, ecological variations across contact situations--variations that are now being documented (which varieties were spoken and learnt where, how, by whom, to what ends, by how many, for how long, etc.?).
On a structural UG-related note, let's ask a question that goes beyond the specific concerns of the AH (this question is related to the general methodology in WSG; see 5.3 above): With respect to modern historical-comparative syntax, what is the status and import of string-based comparisons based on few isolated and superficial patterns? With clever handpicking, any degree of ad hoc superficial (dis)similarity can be established between any pair of languages, whether they are historically related or not (see note 20 ). Yet twelve or so superficial similarities seem enough for the AH to consider HC and Mauritian Creole "too alike not to have had a common [pidgin] ancestor", in spite of the lack of sociohistorical evidence for such a pidgin... One can't help but notice that there do exist, across a wide range of Creoles and non-Creoles across distinct genetic phyla, deep STRUCTURAL similarities in clausal structure, including TMA structure (see, e.g., Cinque 1999). In AH's parlance, these similarities become "too close to be attributed to chance", "suggest[ing] common ancestry". Thus, the widespread structural similarities in Cinque's large-scale and theoretically-grounded cross-linguistic comparisons would erroneously suggest a monogenesis scenario for ALL languages--perhaps from a single (affixless?) Ur-Creole spoken by Eve in pre-historical Africa! (see note 18 ). Humor put aside, Cinque's own, and more reasonable, conclusion is that certain cross-linguistically common patterns, even if they superficially look idiosyncratic, can--actually, MUST--arise independently across languages, simply due to UG. In this view, common patterns do not necessarily entail common ancestry in one single E-language. Common patterns are often due to the common biological ancestry of the species Homo Sapiens and universal constraints on I-languages. (See Marantz 1983: 16 for similar arguments, as part of a critique of Bickerton's Bioprogram Hypothesis.)
For helping me (re-)write this paper with more depth (and more pleasure) than I ever thought I could and for kindness beyond expectations, my heartfelt thanks go to Yves Dejean, Dominique Fattier, Tometro Hopkins, Alec Marantz, Heliana Mello, Salikoko Mufwene, Marilene Phipps and Adrienne Talamas. I am specially grateful to Yves (Papa Iv) for constant guidance, to Dominique for her seminal work on the Haitian Creole lexicon, to Salikoko for his always "heretical" inspiration, and to Marilene for endless brainstorming.
This paper and much else in my thinking about "Creoles" and "creolistics" benefited from many, many discussions with students and colleagues--e.g., in my Topics in Linguistic Theory class at MIT (Spring 2001), at the meeting of the Society for Pidgin & Creole Linguistics in Coimbra, Portugal (June 2001), and on the Internet List CreoLIST. I'll single out Marlyse Baptista, Melissa Cho, Hans den Besten, Yolanda Rivera-Castillo, Maria Carlota Rosa, Michael Tsai and Henri "Nou la!" Wittmann.
I would also like to take this opportunity to express my gratitude to various institutions that, over the past couple of years, have kindly invited me for colloquia or courses where I could share, and improve on, some of the ideas developed in this paper and related work: Bates College, Boston University, Boston's Haitian Scientific Society, Cornell University, Florida International University, University of Maryland (College Park), Ohio State University, UMass Boston, Universidade Federal de Minas Gerais, Universidade Federal do Rio de Janeiro, University of Chicago, University of Rochester, University of Tulsa, and the University of the West Indies (Barbados)-- host of the June 2001 conference (Re)Thinking Caribbean Culture. Much of my "(re)thinking" is due to feedback I received at these venues.
The 19th-century texts and their 20th-century exegeses should be on the reading lists of creolists who (tacitly) promote neo-Schleicherian scenarios for Creole genesis; I myself still have much to learn from the mistakes and insights of our intellectual forebears.
As of Humboldt 1836 [1988: 203-213], he takes "the wearing-down of inflections [which] is an undeniable fact" as a consequence of "the mind's progress", alongside "vernacular[ization]" and language contact (e.g., "foreign immigrations"). Given Schleicher's view above, the interesting, if surprising, observation here is Humboldt's (1836 [1988: 205]) remark that morphological decay results from "the mind's progress":
"The more mature the mind feels itself to be, the more boldly it works in combinations of its own, and the more confidently it casts away the bridges that language constructs for the understanding."For Humboldt, this "more mature" genius, whose maturity is due to inter alia the intellectual development made possible by inflection, can cleverly decide to replace synthetic structures (e.g., affixes for nominal case and verbal tense) with analytic structures (e.g, prepositions and preverbal auxiliaries), thus promoting semantic transparency and ease of articulation (Humboldt 1836 [1988: 206]).
Taking Schleicher's and Humboldt's teleological-genealogical programs to their logical consequences, one must then claim that Prototypical Creoles, with their (alleged) "lack of inflection" (see (11)), are either most "senile"/"retrograde" or most "mature" in revealing "the mind's [utmost] progress" (!).
For further discussion of linguistics-ideology interactions in the terminologies, taxonomies and theorizing of Creole studies as well as in Creole-related language policies and practices, see (e.g.) Prudent 1980; Alleyne 1994; Dejean 1999a,b, to appear; Mühleisen 2000; DeGraff 2001b, in preparation; Corcoran 2001.
In the case of Haiti, propositions (ii)-(iv) straightforwardly entail that many early HC speakers had a "diglossic competence between an L2 variety of the lexifier and a ... pidgin" and that both the founding slaves and plantation-born children could acquire "the local standard" (or, at least, a variety thereof). If so, then one can reasonably argue that "diglossic" early HC speakers were in a position to adopt and adapt affixes and other patterns from their "L2 variety of the lexifier" in order to structurally expand the emerging "Creole". This renders unlikely any scenario whereby incorporation of French-derived affixes into HC is necessarily a late post-genesis phenomenon (cf. Appendix I). The linguistic evidence points to the same conclusion: affixes in all varieties of HC--including any contemporary "preservation of a basilectal creole" (see (v) above)--have cognates in French affixes (see, e.g., Fattier 1998, DeGraff 2001b, contra Lefebvre 1998).
(On the empirical and theoretical status of McWhorter's (2000b) "Afrogenesis Hypothesis" and the "imported pidgins" therein, see Appendix II.)
This also applies to opaque lexicalizations as discussed in the main text and to the other linguistic "ornaments" in (18) that are considered "incidental to basic communication". Thus "`ornamental' elaborations" cannot be excluded a priori from Creole languages. Here too the ill-defined linguistic distinction "old" vs. "young" becomes a methodological trompe-l'uil (see (3b)). The main text elaborates on this E-/I-language distinction in resolving the old-vs-young issue (also see DeGraff 1999a: 8-9).
"[I]f transmission has been interrupted, then there should be ... a lack of correspondence among the various subsystems of the language, most probably between the lexicon as a whole and the grammar as a whole." (Thomason & Kaufman 1988: 11; also see pp. 8-12, 206, 211, etc.)Thomason & Kaufman's litmus test is challenged by the same sort of epistemological-methodological and empirical-structural problems already mentioned in the main text. Their structural criterion is not given any operational and testable measure. In the particular case of HC, lexical and grammatical correspondences between the Creole and its lexifier show no greater amount of discrepancies than their counterparts in French and English diachrony. In fact, in certain domains (e.g., lexicon, morphology, underlying word order and nominal inflection), the discrepancies in French and English diachrony may even be greater than in HC and Jamaican Creole diachrony, respectively (abstracting away from the rate at which discrepancies qua innovations propagate through the speech community). Similar non-Creole, yet (perhaps?) "significant", discrepancies are also found in Indo-European and elsewhere (see, e.g., Thomason 1980 as cited in (8)). Pending an objective and falsifiable measure of "significant discrepancy" (what is Thomason & Kaufman's "significant" threshold?), certain lexical-vs.-grammatical "discrepancies" in non-Creole diachrony seem as "significant" as in Creole genesis. See Mufwene 1998, 2000a,b, 2001, and DeGraff 2001a,b, to appear, for further details on the theoretical abnormality of abnormal transmission in Creole genesis (also see notes 20 and 21 ).
While many of the pidgin structures above can, in principle, be related in a natural fashion to well-known (pyscho)linguistic facts of language acquisition/creation and language contact (e.g., language transfer, restructuration, grammaticalization, regularization, simplification, innovation, etc), the list in (18) hardly forms a theoretically-justifiable natural class for computing complexity (see Section 5 below). Compare, say, the counter-examples and counter-observations referenced in the main text with the arbitrary list of scattered features in (18). The complexity metric defined (negatively) through (18) is based, not on a theoretically-cogent "general tendency", but on features that seem (relatively) rare cross-linguistically--rare, at least, among the "old" languages that Western linguists are most familiar with (including the lexifiers of Caribbean Creoles. WSG even quotes the quite telling remark that some of its non-Creole "test" languages (Kabardian and other languages of the Caucasus) are "extraordinarily complex by any linguistic standard"--if so, the "test" languages in WSG will surely make many other languages (Creole or not) look extraordinarily simple by the very metric in WSG.
In any case, given the very complexity of Language and the vast space for potential distinctions at all levels of grammar (phonology, morphology, syntax, semantics, pragmatics, discourse, etc.), the ad hoc and scattered list of "incidental" features in (18) is truly "scattered": such a list cannot reliably estimate cross-linguistic rankings of overall complexity (see 6.4 below).
"The presence of serial verbs in creoles, then, is not the result of a functional necessity. Their presence is the result of their being A GRAMMATICALLY CENTRAL information-encoding strategy of UNIFORM AND WIDESPREAD AREAL DISTRIBUTION in West Africa, such that there is no reason to suppose that they would not be transferred into an emerging contact language by West Africans, and then retained as the language developed through time and space in the mouths of adults and children." (McWhorter 1997: 155, emphases as in original)Here McWhorter is arguing against Bickerton's view of "Saramaccan as UG in vitro" (McWhorter 1997: 12). Elsewhere we read, still in the pro-substratum anti-Bickerton vein:
"A great many structures that Bickerton designates innate are in fact much more likely to have been transfers from the languages spoken by the slaves first brought to the Caribbean. Bickerton's claim has been that any such similarities between Creole and African structures are mere accidents. While it is hardly impossible that such accidents could have occurred--especially given the "unmarked" nature of many of the structures in question--comparative analysis makes it relatively unequivocal that many of the Caribbean-African correspondences are indeed transfers, not spontaneous creations" (Kegl & McWhorter 1997: 20)While these two quotes capitalize on substratal influence to challenge Bickerton's Bioprogram Hypothesis, McWhorter 2001 (see, e.g., n11) systematically downplays similar substratum-influenced data in favor of a catastrophic Bickertonian scenario. Nonetheless the same sort of arguments levelled against Bickerton's Biogrogram (including those in McWhorter 1997 and McWhorter & Kegl 1997) also applies against McWhorter's (2001) proposal.
(Also see note 27 and DeGraff 2001b for recurring parallels between Saint-Quentin's and later proposals on Creoles as "ab ovo" creations.)
In effect, WSG: §1 ultimately pays only lip service to the stated "intention [for] a sustained investigation of creoles from the perspective of cross-linguistic configurational possibilities, beyond the Western European lexifier languages that have served as the primary focus of creolists' attempts to define the term creole". Such investigation also requires the thorough comparison, at all levels of grammar, of "old" Western European languages (e.g., English and French) with "born again" contact languages whose ancestors exclude European languages and include, say, "fearsomely elaborated" languages such as Tsez, Lahu, Maori--the non-Creole benchmarks for "old language" complexity in WSG.
Such systematic comparison is sorely missing, which makes hypothesis-"testing" in WSG look like a rigged experiment (see 6.4).
"I believe that the difference in degree of complexity between older grammars and a subset of Creole grammars is distinct enough that a complexity metric so fine-grained as to, for example, allow us to rank Romanian, Hausa and Korean in terms of some general complexity quotient would be unnecessary to our project."This a priori belief that "the world's simplest grammars are creole grammars" becomes self-justification for replacing the theoretical necessity of a "general complexity quotient" in favor of a stipulated "metric" based on a scattered list of features based on a skewed sample of languages (see (18)). The metric itself is devised with the expressed goal of separating Creoles from non-Creoles; it ignores a large body of relevant evidence--including data in McWhorter's other work (see note 31 )--and it makes no recourse to, and no prediction vis-à-vis, (psycho)linguistic theory. It thus seems that the main purpose of the proposed metric is to justify the author's latest "belief" about Creole genesis, against all the available counter-evidence and against existing theories (also see notes 37 and 40 ).
"Is the number of genes in an organism's genome an appropriate measure of biological complexity? ... The recent flurry of completed genome sequences, including our own, suggests that this is not necessarily the case ... Rather surprisingly, it turns out that the worm Caenorhabditis Elegans has 18,424 genes in its genome, the fruit fly Drosophila Melanogaster 13,601, the plant Arabidopsis about 25,498, and humans about 35,000. This suggests that there must be other, more sensible genomic measures of complexity than the mere number of genes."Szathmáry et al's proposal is to use "networks of transcription factors and the genes they regulate, rather than ...simply counting the number of genes or the number of interactions among genes". Thus, biologists go beyond simple counting of overt items and they enlist inter alia abstract computational theories about "the connectivity of gene-regulation networks". These theories are at the core of our understanding of how genes work (e.g., the mechanics whereby certain genes are switched on and off in order to represent information and compute over these representations).
No need to say that these theories go beyond my competence. Yet such a development strikes me as normal for the sciences. No matter the field, complexity measures--if they are to be scientifically constructive--must be related to broad empirical and theoretical concerns (cf. Darwin 1871: Ch. 2: 61-62 for relevant remarks on "how easily [our biological and linguistic complexity metrics] may err").
Separating complexity from psycholinguistics leads to incoherence within the very assumptions in WSG. If "creoles represent a fundamental layer of UG" and if "[i]n the realm of syntax, creoles are closer to an ontogenetic foundation than many other languages", than Creoles should be the easiest languages to acquire (à la Bickerton).
Furthermore, the bit-complexity of morphology is explicitly linked to "the development of morphophonological processes, which constitute an added component of a grammar to be LEARNED" (WSG: §2.4.3(d); emphasis added). Acquisition researchers have related complexity issues (say, phonological and semantic transparency in morphology) with ease of acquisition (see, e.g., Clark 1993). Here, complexity is linked directly to acquisition and storage. Thus, WSG does more than "take as a given that all languages are acquired with ease by native learners" (cf. WSG: §2.42). WSG's metric also implies that Creoles should be acquired and/or processed with the most ease (as in Saint-Quentin 1872, Adam 1883, Seuren & Wekker 1986, Bickerton 1988, etc.; see (2) and note 35 ). This contradicts WSG: §2.4.2's attempts at isolating bit-complexity from acquisition and processing.
(DeGraff 1999e, provides an overview of the constructive theoretical linkages among Creole studies, historical linguistics and language acquisition research, in a Cartesian framework that does not postulate fundamental structural differences between Creoles and non-Creoles.)
Here too, as with the V2 case, the view in (30) is theoretically naive; it is too tied to the surperficial "looks" of languages to offer any deep insights into the evolution of grammar and grammatical complexity. Indeed the metric in (30) ignores much of what syntacticians have taught us about the abstract "essence" of grammar. (See Appendix II for similar flaws in McWhorter's (2000b) Afrogenesis Hypothesis.)
"DeGraff ... (1997 1999) argues that the differences between a Creole grammar and its source languages' are due to certain syntactic results following from loss of inflection during second language acquisition (such as lack of verb movement to I), with subsidiary results due to the filtering out of low-frequency features, and the ellipsis of certain functional categories, with the qualification that the effect of the latter two was no more marked than that upon other languages with heavy contact in their histories, such as Yiddish (DeGraff 2000)."Although I myself have not (yet) explored the fascinating and probably enlightening history of Yiddish, what I did argue is that, like in other cases of language creation, CERTAIN (morpho)syntactic properties seem correlated to properties of (overt) inflectional morphology. This is a rather commonplace, if difficult to formalize, guiding intuition in much theoretical and historical work (see, e.g., van Kemenade & Vincent 1997 and van Kemenade 1999 for two recent anthologies on this topic). For example, the abstract linking of (degrees of) verbal inflection and (degrees of) verb raising, although somewhat controversial, is standard fare in current research--see the references in DeGraff 1997, 1999b: 501-502, 518-521, 2001a, to appear; Roberts 1999. Chomsky (1995: 6) attributes to Jespersen the hunch that cross-linguistic syntactic variations can in large part be reduced to variations in morphology. In this particular respect, certain (erstwhile) contact languages such as Capeverdean Creole, Chinook Jargon, HC and Saramaccan (where certain kinds of overt inflectional morphology seem more `economical' than in some of the Creole's source languages) offer valuable databases--natural "test-tubes", if you will--to evaluate and refine current hypotheses vis-à-vis UG's constraints on language creation and language change--in particular, structural constraints on the interaction between inflectional morphology and word-order parameters (see, e.g., DeGraff 1992a, 1997, 1999b,d, 2000, 2001b, to appear; Veenstra 1996; Baptista 1997; Vrzic 1997).
This said, nowhere have I (or any other creolist, as far as I can tell) claimed that ALL "the differences between a Creole and its source languages" are parasitic on (overt) inflectional morphology. In fact, the very work of mine cited in WSG insists on the following caveats:
"In [morphology-driven approaches to syntax], crosslinguistic syntactic differences ... are due to (AMONG OTHER THINGS) distinct inventories of inflectional paradigms". (DeGraff 1999b: 501; emphasis added)These quotes make it clear that, in my view, (loss of) inflectional morphology is not, and could not be, the exclusive factor that determines the overall syntactic shape of Creole languages. DeGraff 1999b goes on to discuss other factors in creolization (e.g., ecological and stochastic factors, substrate and superstrate influences, markedness, processing, learnability). Given the state-of-the-art in linguistic theory, it is not at all clear to me how OVERT morphology could ever be argued to be the SOLE determinant in syntactic change: indeed many aspects of syntax seem orthogonal to properties of overt inflection. Lastly, DeGraff 1999b explicitly argues against "Creole" as a structural type (see, e.g., the quote in (10) in the main text). The arguments in WSG against the view that "the differences between a Creole grammar and its source languages' are due to ... loss of inflection [etc]" address a (theoretically puzzling) straw-man of its own making.
"[O]ne should not expect Creoles to reflect only unmarked parameters ... There are certainly aspects of Creole grammars that were influenced by structures in the source languages ... Furthermore, even granting that Creoles' tendency toward unmarkedness is rooted in their isolating [inflectional] morphology ... , NOT ALL SYNTACTIC PROPERTIES NEED TO BE DIRECTLY TIED TO THE MORPHEMES THAT TEND TO BE LOST IN PIDGINIZATION. It is not clear, for example, whether the properties of (long-distance) wh-movement are parasitic on inflectional morphology" (DeGraff 1999: 519; emphasis added)
Thus, I somewhat agree with Domingue (1977: 90) that:
"[I]f ME ... can be described as another Creole, there must be many more Creoles around than we think. On the contrary, if ME does not fit in the Creole category, it will be because stringent characteristics of that category will have been established, a task yet to be done"Vis-à-vis "stringent [structural] characteristics of [the Creole] category", the "task yet to be done" remains elusive still and seems to me impossible--thus, this paper.
"Scholars are often wary of citing ... commitments [to social justice], for, in the stereotype, an ice-cold impartiality acts as the sine qua non of proper and dispassionate objectivity. I regard this argument as one of the most fallacious, even harmful, claims commonly made in my profession. Impartiality (even if desirable) is unattainable by human beings with inevitable backgrounds, needs, beliefs, and desires. It is dangerous for a scholar even to imagine that he might attain complete neutrality, for then one stops being vigilant about personal preferences and their influences--and then one truly falls victim to the dictates of prejudice. Objectivity must be operationally defined as fair treatment of data, not absence of preference. Moreover, one needs to understand and acknowledge inevitable preferences in order to know their influence--so that fair treatment of data and arguments can be attained! No conceit could be worse than a belief in one's own intrinsic objectivity, no prescription more suited to the exposure of fools. ... The best form of objectivity lies in explicitly identifying preferences so that their influence can be recognized and countermanded." (Gould 1996: 36-37)Gould then proceeds to debunk a number of rankings of human cognition across the "races", in (e.g.) the practice of 19th-century craniometry and 20th-century psychometrics.