Gradience and Variation: How to Integrate Social Variation into the Grammar Model?

Aria Adli

Zentrum für Allgemeine Sprachwissenschaft (ZAS), Berlin, Germany


Variation at the level of syntax is still an understudied field in variationist sociolinguistics. This is due to important methodological and conceptual challenges: (i) While syntactic theory works with judgments, sociolinguists often consider this data source as unreliable and focus on speech production. However, a solid definition of a linguistic variable in the sense of Labov (1966) must rely on formal-syntactic arguments. (ii) Syntactic theory assumes all speakers of a dialect to have the same underlying rule system.

This paper shows that the key to this problem lies in a gradient treatment of judgments and a corresponding extension of the formal theory. We will show that a careful experimental approach can provide us with gradient judgments that fulfill both criteria of statistical reliability and construct validity: They show high Cronbach’s ?-values, and they do map systematic social variation. Furthermore, gradient judgments enable direct comparisons with frequency data from speech production on a metrical scale.

We studied the issue of variation at the level of syntax with data from French wh-questions. They show a high number of possible structural variants and are therefore a suitable linguistic variable. 102 French native speakers from Paris participated in a specially designed gradient grammaticality judgment test. The sentences distinguish between 2 types of wh-words (wh-adjunct vs. wh-NP) and 6 different structural wh-variants (complex inversion, subject-clitic inversion, est-ce que, Q S V order, wh-in-situ, and clefts). We calculated a series of 3-way variance analysis with the syntax-related variables A ‘wh-word’, and B ‘structural wh-variant’, as well as with the social variable C. C covers socio-demographic (e.g. gender, age), socio-economic (e.g. income, housing, etc.), and socio-cultural (lifestyle) factors that were collected with a questionnaire. The results show significant patterns of social variation with the variables housing density and lifestyle, in particular for those wh-variants which are at opposite regions of the register continuum, namely wh-in-situ vs. subject-clitic inversion.

This paper then turns to the initial question to what extent the systematic variation in the judgments amount to inter-individual differences in the grammar itself. As a preliminary step, we need to “enrich” the formal-syntactic framework with the concept of gradience. We build on previous work in which we proposed a distinction between well-formedness constraints – core principles of syntax, whose violation results in ungrammaticality, and preference constraints – principles that realize the interaction with other components of language, and whose violation lowers the degree of grammaticality, but without necessarily causing ungrammaticality. Given that preference constraints correlate with non-syntactic components of language, the integration of social variation into grammar becomes now straightforward. The basic idea is that inter-group social differences correspond to systematic patterns of variation in the violation costs of preference constraints. The well-formedness constraints are the same for all speakers. We therefore do not have different grammars but different uses of the (same) grammar. This model also accounts for language change which is explained by the transformation of a preference constraint into a well-formedness constraint.

Session: Paper session
Variation 8 / Change
Friday, April 4, 2008, 13:45-15:15
room: 15