The Creolist Archives
PROTOTYPE AS A TYPOLOGICAL YARDSTICK TO CREOLENESSHenri Wittmann
[Since I suggested that true believers can't be outargued in any useful scientific fashion, I have been subject to a lot of negative feedback on this, last not least from McWhorter himself. Against all evidence experience has led me to fancy on probabilites, I have nevertheless decided to give it a try and argue my way to maybe nowhere. Footnotes (numbered in parentheses) and references appear at the end of the text. The present version is identical to the one posted in two parts on the list except for appendices on Magoua which have been added here. A longer version of this will follow most certainly and be announced as it becomes available in one form or another.]
SCOPE OF DISCUSSIONWhat I'd like to do in this small contribution on the uses and misuses of typological yardsticks is to show that non-creole languages such as Manding (as exemplified by Bambara), Sooninke and Magoua French, regular languages representing the supposed substrate and superstrate of at least some creole languages, conform to McWhorter's (1998) "creole prototype" just as good or better than languages stereotyped as "creole" such as Haitian Creole, Saramaccan and Berbice Dutch when weighed against McWhorter's additional yardstick of GRADIENCE. The point I'd like to make is that the "structural yardstick" McWhorter offers to measure "creoleness" rather impressionistically couldn't possibly succeed where quantitative approaches to the morphological typology of language failed to expose prototypical dispositions in language beyond evolutionary tendencies and patterns of drift accessible through time to all natural languages (Greenberg 1960, Wittmann 1969, 1973:89-90, 1983).
INFLECTIONThe most salient feature rendering creoles synchronically distinguishable from other languages seems to be identifiable with a minimal degree of inflection in their structure. Ideally, creoles have no inflection at all. What McWhorter here talks about is of course "fusional" inflection since there are no languages with no inflection. Anyway, when weighed quantitatively under the criterion of gradience, this amounts to saying that creole languages under inflection should not have more than two affixes. This is easily contradicted with the test case of Berbice Dutch, a creole that has admittedly four inflectional affixes, three aspectual ones (perfective, imperfective, iterative) and a plural one. That score could most probably be improved with Algonquian-lexifier pidgins, for example, if we give any credence to Bakker's (1995:31-32) hunch on this which wouldn't be at all incompatible with the idea that synthesis can rise under pidginization-creolization if supported by the right substrate conditions. Even if we could muster pidgins or creoles with seven or eight affixes, that still wouldn't be very much compared to Turkish where we have 42 inflectional affixes. Coming around to Manding, we have a mere one to count: the affirmative intransitive form of the perfective -ra (variants in Bambara -na, -la ; in some other dialects -ta). The pluralizer u, orthographically rendered as -w, doesn't qualifiy as an inflectional affix since, just like Haitian Creole yo , it's a 3PL pronoun adjuncted externally onto DP. The same sort of judgment must apply to the postverbal progressive la, spelled -la, also a postposition, if we compare it to Haitian Creole pou, a preverbal TMA marker and a preposition. The so-called "participles" in -len (variant -nen ) and -tò are in fact deverbal adjectives with none of the features implying movement of the verb to a functional head ASPECT participles would have in the verbal inflection of languages such as English or French. In Sooninke, the count goes up to three: the pluralizer -u (variants -o, -ni in western dialects, -nu in eastern ones, historically related to the Manding pluralizer), the imperfective -nH (historically related to the Manding progressive la ; H for vowel harmony with the preceding syllable), and the transitive-to-instransitive -e/-i, the latter however being treated traditionally in this language as a derivational suffix. With Magoua French, the count goes down to two: the non-progressive imperfective past -e and, as McWhorter would put it, the "weakly" or "marginally" productive conditional -re . McWhorter doesn't seem to consider verb syncopation as found in one of his test languages, Mauritian Creole (Wittmann 1972, Seuren 1990), to be inflectionally of consequence. If we were to disagree with him on this, we would have to admit that verb syncopation [as already obvious from an earlier posting of mine with examples from Magoua, see Appendix II] is shared by a number of varieties of languages derived from French, notably Mauritian Creole, Seychelles Creole, Reunion Creole, Louisiana Creole AND Magoua, which puts the score for the latter up to three. However, no matter how fine-grained our methods of counting and weighing, the three non-creole languages under review here satisfy perfectly McWhorter's requirement for creolehood under the first parameter of his typological yardstick.
TONEThe second parameter in McWhorter's yardstick is tone. Creole languages should make little or no use of tone to lexically contrast otherwise identical monosyllables or to syntactically encode grammatical distinctions, but if they do, grammatical tone should be historically derivable from non-tonal segments. Weighed in such a fashion, creole languages I thought to be tone languages such as Saramaccan and Papiamentu turn happily into very untoney almost no tone ones. Compared in this way, however, the "functional load" of tone in Manding and Sooninke is even lower than in Saramaccan and Papiamentu. They have less than a "handful" of tonally distinguished monosyllables and only some dialects of Manding (such as Bambara) have an example of syntactically encoded "grammatical tone", the marker of definiteness, what in Sooninke and other dialects of Manding is still a segmentally encoded post-NP enclitic, just like in Haitian Creole. The use of tone in most Mande languages in general is so "distinctly marginal" that most official orthographies don't have any use for it which isn't true, as far as I can see, for the offical orthographies of Saramaccan and Papiamentu. There is of course no tone in Magoua or any of the French Creoles.
As it is, one can feel but uncomfortable with McWhorter's appreciation of the perceptual saliency of tone if compared to the so far unchallenged opinion that the grammatical use of interrogative tone is a quasi-universal throughout natural languages and that the tonal parameter developed earlier in language than vocally encoded segmental phonemes (Hermann 1942, Bolinger 1964, Wittmann 1980, 1991).
DERIVATIONTransparent derivation as a means to enlarging vocabulary constitutes McWhorter's third and last parameter for measuring creolehood. DeGraff (1999) has shown that derivation in Haitian Creole is much more opaque than McWhorter would have it, Lefebvre & Lumsden (1994) compared successfully the derivational means of Haitian with that of a regular language, Fongbe, whereas Wittmann & Fournier (1996), comparing the derivational transparency of Haitian Creole, Fongbe and Magoua French, found Haitian and Magoua to be identical in ways which can only be explained as a case of transmission from an older Koine French1 . If derivational opacity is quite common in Creole French and Magoua 2 , the same cannot be said for derivation in Manding and Sooninke or for Mande languages in general. The degree of semantic transparency is such that Wittmann & Soumaré (1990), in a study of language planning for Sooninke, upheld that "les méthodes d'élaboration lexicale proposées en milieu de langue allemande ont l'avantage d'être fondées sur l'exploitation systématique des moyens de la motivation morphologique naturelle, moyens qui ne sont que faiblement productifs dans des langues comme le français ou l'anglais." Weighing this against the combined powers of the instrinsic and not so instrinsic features of McWhorter's yardstick, it would not be possible to say that Mande languages, though no longer considered Niger-Congo affiliated (cf. Dwyer 1989), are phylogenetically to be considered more pidgin in any useful way than Hittite or Germanic.
CONCLUSIONMcWhorter's quantitative approach to morphological typology fails to define the creole prototype in any meaningful way except as a fuzzy set including languages such as Mande that evolved the prototype over thousands of years. The typological distance from Magoua French to Haitian Creole cannot be interpreted diachronically as a "radical break in transmission" stripping away inherited affixal morphology in favor of a new affixal morphology evolving via internal gradual grammaticalization of erstwhile lexical items. As a matter of fact, McWhorter's typological yardstick is useless as tool if we wanted to determine in which ways Magoua differs from both Haitian Creole and Standard French (cf. Wittmann 1998b).
Magoua inflection is characterized by a Bantu-like class system of agreement which unlike Bantu is clitic in nature instead of being affixal as in Bantu. It also has serial verbs unlike both Haitian Creole and Standard French: It distinguishes four-way directionals of movement toward and away of the type GO-TAKE (àlé-mné), GO-GET (àlé-kri), COME-BRING (vyenn-mné), COME-GET (vyenn-kri) and makes use of a narrative instrumental TAKE 3 . However, the loss of the copula with the ensuing rise of proclitic subject-object agreement for person-number-gender-definiteness distinctions and preverbal tense-aspect marking combined with the semantic bleaching of the etymological determiner system are clear signs that the origin of Magoua and related dialects indeed involved a break in the transmission of French, though not in the manner fancied by McWhorter.
Appendix I: On MagouaSome of my work on Magoua and related subjects (as well as contributions by Robert Fournier) will soon be available at the following site:
Here are some of the references which might be of use.
Wittmann & Fournier (1996): A refutation of Lefebvre's relexification hypothesis according to which Haitian Creole would be (to paraphrase DeGraff) a Fongbe clone in French clothing. For a bibliography on relexification à la Lefebvre, see Lefebvre (1994: 83-93). Lefebvre's more recent publications [in English] don't add anything new to what is alraedy known from her on the subject to date..
Wittmann (1995): A comparision of Magoua with related dialects lined up against 26 diagnostic features. Also a hypthesis on the rise of a koineisized (pidgin-like) variety of French in Paris and a refutation of Lefebvre (1992).
Wittmann (1998a): Specifically on the rise of agreement features in Koine French, the "sontaient' syndrome and what happened to all of that in Creole French.
Wittmann (1998b): A specific refutation of Lefebvre & Lumsden (1994) where they attempted to demonstrate that Haitian Creole differs morphosyntactically form Standard French on 14 counts. Had they bothered to base their comparison on any basilectal variety of Québec French (such as Magoua), the outcome would have been very different. As it is, Claire Lefebvre is a native speaker of basilectal Québec French which differs from Magoua in trivial ways.
Wittmann (1996): An introduction to Magoua phonology and
transcription. Heavy high vowels: /é eú ó/;
heavy middle vowels: /e (eu) o/; heavy low vowel: /a/; light high
vowels: /i u ou/; light middle vowels: /è eù ò/;
light low vowel: /à/. The aigu /´/ is used exclusivement
to mark high heavy vowels, the grave /`/ solely for the middle and low
light vowels. Heavy vowels are diphthongized under stress except in open
syllables and monosyllabic affixes and clitics. Light vowels are never
diphthongized even when they are lengthened in the context of the "lengthening"
consonants /v z j/: /chèv/ = [chè:v] "goat" with a long always
monophthongal light middle vowel, /fev/ "bean" always with a long diphthongized
heavy middle vowel. The lengthening of light vowels has acquired minmal
functional load with /à/ as in /bàté/ "threshing" (from
/bàt/ "to beat") : /bà:té/ "to bat (with a basball
"dough" turn out as /pàt/ with a short vowel and /pat/ with a long diphthongized vowel). To imitate the so-called accent "en bouche de cul de poule" of the French "imported" from France, the "extra-high" /i: u: ou:/ are used, in fact the tense high vowels of Standard French. A nasal after heavy vowels is phonetically resolved as nasalization on the preceding vowel. All other nasals are consonants. The system allows for high nasal vowels such as in /sénñ/ "sign" et /mónn/ "people" comparable to high nasal vowels in Haitian Creole and other related dialects. As might be expected, nasal vowels diphthongize under stress.
References to Appendix I
Appendix II: Stress-shift as an indicator of morphological productivity[The appendix reproduces here the relevant passages of an exchange on the subject with Philip Baker and Albert Valdman. Reproduced by permission of the authors.] Recent postings re prototypical Creoles etc. from DeGraff, McWhorter and Parkvall lead me to think that some data from Mauritian Creole may be of interest to some of you.
Mauritian Creole also has an inversive prefix, de- and a repetitive prefix re-. What is of special interest is that these productive affixes are normally stressed. In consequence, there are some minimal pairs. In the examples below, stressed vowels are written in upper case while upper case N indicates that the preceding vowel is nasalized. (Mauritian Creole has five not seven peripheral oral vowels so there are no mid-high vs mid-low contrasts)
fEr 'do' (Fr faire), refEr 'get well', rEfer 'do again' (Fr refaire has both the latter senses but there is no distinction in pronunciation) (also: rErefer 'get well again', rErEfer 'do yet again', dEfer 'undo', rEdEfer 'undo again', etc.)
pAre 'get ready' (Fr parer), repAre 'to repair' (Fr reparer), rEpare 'to get ready again' (Fr *reparer does not appear to have this latter sense)
remoNte 'comeback' [n] (Fr remontee), rEmoNte 'climb up again' (Fr remonter), mONte 'climb up' [Philip Baker]
Same thing for certain varieties of Québec French such as (Three-Rivers) Magoua:
However, this pattern (of stress shift) gets blurred, both in these varieties of Québec French and in Mauritian Creole French, whenever syncopation rules apply in the context of Verb Movement (also refererred to as Verb Raising).
So the second r in EÙRrparé is a reduction/contextually conditioned variant form of EÙR, I assume. [Philip Baker]
Double consonants and consonant clusters not inherited from Old French (inherited clusters are obligatorily simplified like in Creole) are rather common because of the highly cliticizing behaviour of Magoua (and American varieties of French in general). Problems with this are resolved by a universal rule of schwa insertion (which incidently is the EÙ with a grave on the U).
How about DE- in Magoua? [Philip Baker]
With stress shift:
Without stress shift:
débité "cut up,
Minimal pair: dépann "unhook" vs dépann "depend"
Comparing a contrastive pair: dézàkroché and dépàrlé, does the stress in the first go all on the prefix and in the second on the final syllable? [Albert Valdman]
As it is, for the first one, you can have déz àkròché, dézàkròch and dézàkròch, the latter phrase finally only; for the second one, you'll get only dépàr lé and dépàrl where other dialects might have for the latter dépàrl or dépàr lé (as noted already by La Follette 1969:67).
At first look, it looks like most of the items in the first set are fairly semantically transparent inversives (except for dézosé which carries inversive meaning but for which there is no base *osé "mettre les os") whereas those of the second set are simplexes (except for dépàrlé, which is a not too transparent inversive on pàrlé). One could claim that in this second set dé- does carry a sort of negative meaning but there is no obvious base (bité). Would you agree with his generalization? [Albert Valdman]
Except that I wouldn't say that "unstressed de- and re- ... are found only in words adopted direcly from French". I'd rather say that these are examples of (relatively more) "opaque" derivational affixation, but affixation nontheless.
This paper appeared originally as Wittmann I and Wittmann II on LINGUISTLIST. Its publication here caused the following exchange between McWhorter and Wittmann: