Syntactic Variation and Linguistic Competence:
The Case of AAVE Copula Absence
Emily M. Bender, 2001
Ph.D. thesis, Stanford University
Chapter 1: Introduction
The original impetus for this project was the observation that generative grammar (especially generative syntax) seemed incompatible with the data on variation found by sociolinguists working in the Labovian tradition. It seemed to me that the intricate patterns of non-categorical grammatical constraints on variation might reflect some underlying linguistic knowledge. If they do, they pose several problems for generative theory, which is generally ill-equipped to deal with probabilistic or quantitative knowledge.
But how was I to show that the non-categorical grammatical constraints are indeed a matter of linguistic knowledge? They answer lay in the growing body of literature on the social value of variation (e.g., Labov 1963, Ochs 1992, Eckert 2000). I hypothesized that knowledge of non-categorical constraints would be apparent in their effect on the social value of sociolinguistic variables, and further that this effect would be detectable in a matched-guise experiment (Lambert et al. 1975).
The case study taken up here is variable copula absence in African American Vernacular English (AAVE). This variable was chosen because it is the single most well-studied syntactic variable and the non-categorical grammatical constraints are therefore well-established. This variable also turns out to have interesting syntactic properties.
The structure of this dissertation is as follows. Chapter 2 presents the theory of syntax assumed here (a version of HPSG, see Pollard and Sag 1994 and Ginzburg and Sag 2000). Three aspects are highlighted. The first is that HPSG is a sign-based, surface oriented grammar. These design features are what makes it plausible to extend HPSG to the kinds of knowledge required to model non-categorical constraints on variation. The second aspect emphasized is the general lack of empty categories in HPSG (and related theories). It is against this background that the syntax of AAVE copula absence becomes particularly interesting. The third aspect highlighted is an account of the detailed properties of English auxiliaries, drawing on the extensive literature this topic has generated in HPSG and its precursors. This detailed background is required in order to develop and evaluate the possible analyses of copula absence discussed in Chapter 3.
Chapter 3 begins by establishing that copula absence is indeed a syntactic, rather than phonological, variable, contrary to the assertions of Labov (1969, 1995). Labov's claim is that AAVE simply carries the general English process of auxiliary contraction one step further, resulting in copula absence. For this to be true, contraction must feed deletion, and copula absence must therefore be possible only where contraction is. The data in Chapter 3 show that this is not the case; copula absence is possible in a couple of environments that exclude contraction. I then go on to consider four different syntactic analyses: two constructional analyses designed to avoid positing an empty category for AAVE copula absence, an analysis in terms of a phonologically empty verb, and an analysis in terms of a potentially phonologically empty construction. Somewhat surprisingly, given the general success of lexicalist frameworks at avoiding empty categories, the latter two analyses are shown to be the only adequate ones.
With this background on the syntax of AAVE copula absence, the next two chapters concentrate on the problem of non-categorical constraints on variation. Chapter 4 presents the matched-guise experiment. The results of this experiment provide preliminary evidence that AAVE speakers (and other African Americans who are familiar with, but do not speak, AAVE) have knowledge of the non-categorical constraint tested: the effect of the following grammatical environment.
Chapter 5 reviews previous approaches, formal and functional, to non-categorical constraints against the background of three aspects of the social value of variation: socially meaningful patterns of variation, socially meaningful individual tokens of variants, and the effect of the grammatical environment on the social value of variants observed in Chapter 4. The approaches considered are the Variable Rule approach of Labov (1969) and others, the Optimality Theoretic approach of Reynolds and Nagy (1994), Anttila (1997), and Boersma and Hayes (1999), and the functionalist approach of Kiparsky (1972, 1988). Of these, only the functionalist account is capable of modeling all three aspects of the social value of variation. However, in order to capture the results of the experiment, it must posit that speakers have knowledge of the (putatively functional) constraints behind the patterns of variation across different grammatical environments.
Chapter 6 discusses three types of linguistic knowledge that seem to be involved in the judgments the participants gave in the experimental task: knowledge of social meaning attached to linguistic forms, direct knowledge of grammatical structure that is computable from more basic signs already in the grammar, and knowledge of frequencies or probabilities. While all three are usually excluded from models of linguistic competence, there are in the literature proposals for including each in competence grammar, and sign-based grammars appear to be uniquely qualified to do so. The conclusion of this discussion is that the location of the boundaries of competence grammar should be considered an open issue.
Back to dissertation page