1 This paper has benefited from the comments of several colleagues since it was first written in 1994. I am particularly indebted to Guy Bailey, Michael Montgomery, Salikoko Mufwene, and Don Winford for feedback on earlier verions. Since I have not always heeded their wise counsel, however, I alone bear responsbility for the final version of this paper. I am also indebted to Angela E. Rickford for facilitating the writing of this paper in innumerable ways.
2 See Reinecke et al (1975:482) and Holm (1988:32-33, 55) for discussion of the early contributions of these pioneers on the creole origins issue. And for discussion of a previously unpublished ms by Schuchardt which bears on this topic, see Gilbert (1985).
3 While Ewers' study is substantive and very valuable, his assumption that morphological and syntactic textual analysis is not as seriously affected by poor recording quality as phonetic and phonological analysis represents a common error (cf. Schneider 1989:49, whom he quotes). Labov's (1972b:190, fn. 9) view is the exact opposite, and corresponds more closely to my own experience : "In phonology, we can wait for the clear, stressed forms to emerge from the background noise. But many grammatical particles are reduced to minimal consonants or even features of tenseness or voicing which are difficult to hear in less than the best conditions, and many are so rare that we cannot afford to let one escape us.".
4This is not a great example, because as Beryl Bailey (1966) has noted, niem ("name") is a special naming verb in Jamaican (and other Creoles) which requires no predicating copula.
5 The sources include John Leacock (1776) The fall of British tyranny, John Murdock (1795) The triumphs of love, James Fenimore Cooper (1821) The spy, and Edgar Allan Poe (1843) "The gold bug." These are admittedly late in the evolution of African American dialects, but they still take us further back than the ex-slave narrative recordings and Hoodoo texts, as well as diaspora data from Samaná.
6 The early twentieth century evidence is in Repka and Evans (1986). In their analysis of four works written by African American authors (Chestnutt, Toomer, Hughes, and Hurston) between 1899 and 1937, copula absence is 20% (61/310) with plural and second person subjects, 3% with first singular subjects, and 3% (20/720) with third singular subjects. For evidence on modern AAVE, see Rickford et al (1991:117), who report a .67 variable rule feature weight for copula absence with second person and plural subjects in the East Palo Alto data (as computed by the "Labov deletion" formula in which deletions are computed as a proportion of deleted and contracted forms, with full forms excluded from consideration), but a much lower feature weight (.33) for third person singular subjects.
7 The evidence of Barbadian and Jamaican is somewhat more ambiguous. Rickford and Blake (1990:267, table 2) found that plural and second person subjects did favor copula absence in Barbadian speech more than first singular or third singular subjects did (variable rule feature weights of .58, .47 and .45 respectively), but the difference was not statistically significant. Rickford (1991c) found that plural and second person subjects were slightly less favorable to copula absence in Jamaican creole than third singular subjects, and slightly more favorable than third singular subjects (variable rule feature weights of .50, .54 and .46 respectively), but the differences were again, not statistically significant.
8 Since AAVE is primarily a lower and working class phenomenon, fictional characters from other socioeconomic strata should not be expected to show the same frequencies of copula absence, although we might perhaps expect them to show the same kinds of conditioning if they show sufficient copula absence for conditioning to be evident.
9 In Rickford et al's (1991:117) study of copula absence in East Palo Alto, California, the relative positions of __Locative and __Adjective in the feature weight hierarchy were reversed when the same data were analyzed as Labov Deletion (__Loc: .42, __Adj: .47) and Straight deletion (__Adj: .45, __Loc: .47).
10 If Samaná subjects with there, where, here, these, those, this and them are also excluded--on the grounds that some or all of these were excluded in the AAVE studies as well (see Winford 1992a:55, fn 6)--the overall deletion rate would remain at 40% (57/144).
11 The Observer's Paradox (Labov 1972b:209): "the aim of linguistic research in the community must be to find out how people talk when they are not being systematically observed; yet we can only obtain these data by systematic observation."
12 One anomaly is the fact that a following adjective remains the least favorable following environment for copula absence in Samaná, but this is not the case in Hannah's more recent Samaná data (row two, table 4).
13 The statistics from Hannah (1996) cited here are for "Labov Deletion" (cf. Rickford et al 1991), the same method used by Poplack and Sankoff (1987). Hannah's "Straight Deletion" varbrul weights for copula absence are: .14 __NP, .43 __Adj, .47 __Loc, .88 V+ing, .97 __Gon.
14 Omitted from this table, as it is also from tables 2 (second row) and 3, is the feature weight reported by the authors for a following Wh-clause; in all three cases it is higher than the feature weight for __Gonna.
15 The hypothesis is also stated alternatively by the authors as follows: Categorical or high frequency of copula ABSENCE before gonna, Verb+ing, and adjective, and categorical or high frequency of copula PRESENCE before locative and NP.
16 Singler 1993 reports data on fourteen Sinoe Settlers, including Carolina. Of these, six have overall rates of nonpast copula absence between 71% and 83%, indicating that Carolina's high zero copula rate is not atypical; four others have rates between 55% and 67%; and the remaining four have rates between 33% and 40%.
17 The hierarchy of folowing environments holds even when a multivariate, variable rule analysis is performed (on Albert and Slim's speech only, not Carolina's, since hers is too categorical), with feature weights for copula absence (Singler 1991:144, table 15) as follows: __gon 1.00, __loc .86, __V+ing .72, __Adj .35, __NP .29, __Det-NP .13.
18 Singler (1991b:143) argues, in fact, that gon has achieved auxiliary atatus in LSE and no longer co-occurs with nonpast forms of the copula. This same argument could be made for __gwine in Jamaican Creole (it shows 100% "copula absence" in DeCamp's JC texts (see Rickford, 1996:369, table 6), and has already been made for creoles more genreally by Holm (1984:298): "creole go was a calque for a protocreole preverbal marker indicating irrealis which was never preceded by any copula-like particle."
19 One difference is that in the Trinidadian data, copula absence with a following NP is not only low, but almost non-existent; the copula ia almost always retained. See Winford (1992a:35-37) for further discussion.
20 Following C.J. Bailey (1973), we adopt 80% as the cut-off point for (near) categoricality.
21 At the same time, pidgin-creole communities show variation with respect to the relative ordering of these environments, as do AAVE communities and data sets (Poplack and Tagliamonte 1991:323, Rickford et al 1991:121, table 7); for Jamaican and Gullah, the ordering is indeed adjective > locative (Holm 1984, Rickford 1991c, Rickford 1996), perhaps in line with Singler's 1991b:156-7 argument for Non-Settler Liberian English that basilectal varieties tend to show the high adjective copula absence pattern. (Jamaican and Gullah are arguably more basilectal than Barbadian and Trinidadian.)
22 Stewart's (1969) account predates but essentially presumes the "more=earlier, less=later" principle which C.J. Bailey (1973) articulated. Stewart's exact words (p. 244) for the final stage in his hypothetical chain of events are as follows: "Finally, with the introduction of an optional dummy /iz/ in V-ing phrases, and a partial collapse of verbal /is/ = Ø with equative /iz/ = Ø, one can see the historical process--entirely documentable--which could easily have given rise to the statistical difference in copula deletion discovered by Labov ."
23 Day (1973) also provided a qualitative implicational analysis, and interestingly enough, that suggested a different relative ordering of __Adj and __Loc (see his tables 4 and 5, pp. 89-99), one which he was apparently unable to explain to his own satisfaction (see p. 111). As we have noted, the position of these two environments--relative to each other-- is extremely variable in studies of AAVE.
24 One disparity in the case of the (1960) JC data is the fact that __Loc is lower than __NP; one disparity in the case of the (1991) data is that __Adj and __Ving are equivalent. Without additional data, it is difficult to know how substantive these apparent disparities are, or to pursue explanations for their occurrence.
25 The corresponding variable rule feature weights are: __NP .23, __Loc .12, __Adj .75, __Ving .79, __gwain 1.00 (Rickford 1996:369, table 7), and __NP ..00, __Loc .19, __Adj .52, __Ving .45, __gwain .83 (Rickford 1991c). Both the percentages and the feature weights were computed by the "straight deletion" method, in which deletions are computed as a proportion of all possible forms--full forms, contractions, and deletions (Rickford et al 1991:106).
26 This model is represented by the quote from Bickerton (1971) above.
27 Of course, the prediction does not hold for the 1991 Barbadian data in table 10, where __Adj shows more copula absence than __Loc, but it can be argued that the two octogenarians who are the source of these data are more basilectal than the six younger speakers who are the source of the 1980s data.
28 The group data in this table are summed from Winford's (1992a:34) table 5 data on first person singular, third person singular. and plural and second person forms. In table 6 (ibid.) , Winford provides the corresponding VARBRUL feature weights for copula absence: __NP .00, __Adj .64, __Loc .80, __V+ing .85, __goin .88.
29 The interview data in this table are summed from Winford's (1992a:41) table 7 data on first person singular, third person singular. and plural and second person forms. Winford does not provide VARBRUL feature weights for these data.
30 Note that while the nominal copula is replaced by zero in Singler's NSLE model, it is replaced by invariant is in Winford's CEC model, in line with evidence that the CEC mesolect shows very low rates of prenominal copula absence.
31 As Labov (1982:198) notes, this parallelism was said by James Sledd (in a personal communication) to constitute "the first serious evidence for the Creole hypothesis."
32 (This is part of a larger problem in variation studies (both quantitativist and implicational)--the tendency to be satisfied with describing rather than explaining the patterns.
33 It is not true that progressives and futures require no "copula" or auxiliary in the creoles. The basilectal progressive is (d)a or de V, with a clear preverbal marker or auxiliary; the basilectal future is either go V or a go V (see discussion below), the former never requiring a copula at higher levels of the continuum (where it surfaces as gon in some varieties), the other sometimes doing so (where it surfaces as gwain/goin tu).
34 "Copula preceding ADJECTIVES and those preceding LOCATIVES could be expected to delete at substantially different rates in BEV were they to be calculated separately, since in the protocreole there was a copula for location (de) whereas adjectives were a subclass of verbs requiring no copula. The deletion rate for copulas preceding adjectives could be expected to be several times greater than that of copulas preceding locatives."
35 One problem with model B is that it posits simultaneous replacement of the basilect forms in all environments by zero, something which Singler himself admits (1991b:155) may need refining. Differential timing of decreolization processes by environment--as in Winford's model, and in the empirical work of other continuum scholars (Day 1973, Bickerton 1973, Rickford 1979)--seems theoretically more plausible.
36 For further discussion of the relationship between the future tense markers in creoles, see Mufwene (199x--Sali please fill in date here and enter in references--I'm referring to your grammaticaliz. paper).
37 This is not the case in Hannah's (1996) analysis of more recent Samaná data, however. The VARBRUL weight for copula absence (Labov deletion) is .038 for an NP subject, and .239 to .806 for various pronoun subjects.
38 Don Winford (personal communication) suggests that it may involve phonological constraints peculiar to AAVE. However, as the data in Labov (1972a:104) and other studies indicate, the pronoun effect cannot simply be attributed to the fact that most personal pronouns end in vowels, since Pro__ has a consistently favoring effect on copula absence, while a noun ending in a vowel does not (except in the case of the NYC Thunderbirds). Beyond this, no one has given ANY explanation, much less a convincing explanation, as to why pronoun subjects should favor copula absence more than full Noun Phrase subjects.
39 Salikoko Mufwene (personal communication) argues that ri, se and je are the only real copulas in this list, since Yoruba n is really the counterpart of using English progressive -ing as the future marker.
40 From the point of view of the creole hypothesis, of course, universals of simplification and substrate influence are not necessarily in conflict, since both could be elements in the prior pidginization and creolization of AAVE.
41 As Wolfram ((1974:522) goes on to note: "Of course, it must be admitted that the inability to find copula deletion in British varieties does not necessarily mean that it doesn't occur; but since copula deletion is a rather noticeable phenomenon, one would suspect that if it had occurred, there would be some report of its existence in the major sources."
42 Of the eight instances of zero copula which comprise this 1%, seven are with plural or second person subjects--that is, they are instances of are deletion rather than is-deletion, as reported for other Southern dialects (Alabama, Mississippi, Texas).
43 Guy Bailey (personal communication) offered the following helpful remarks after this paper was written, and it seems most relevant to insert them here: "First, the exact order of the constraints of the following predicate on copula deletion is not really crucial to the creole hypothesis. The fact that the following environment matters at all is sufficient to prove that this comes from something other than English. In English the form of the verb always depends on the subject. Even in those dialects that do not have subject-verb concord, the form of the verb is determined by whether the subject is an NP or PRO. It is not surprising that there should be some discrepancies among AAVE and various creoles in regard to the exact effects of the following environment. After all, they've had several centuries of independent development. Second, I think the differing effects of a preceding NP or PRO on zero copula has a simple explanation: it reflects the grafting of an English constraint onto a creole process. This constraint manifests itself in a number of ways in earlier AAVE, and with several centuries of contact, it is only reasonable to assume that other dialects of English affected AAVE just as AAVE affected them. Third, I'm convinced that African and Creole influence not only extended throughout the entire period of slavery but that the period from 1790-1840 saw a real reinfusion of these elements. More than half of the slaves imported to the U.S. were imported after 1790 (most of these after 1793 and the invention of the cotton gin). With the westward expansion of the cotton kingdom, this was the most dynamic period of slavery."