Variability and modularity: A response to Hudson

Abstract

Hudson begins his response to Adger (2006) (Combinatorial variability) with the following question: 'are the findings of variationist sociolinguistics compatible with standard versions of Minimalism' (p. 683) and ends with the statement that a rapprochement between variationist sociolinguistics and 'any theory which assumes that language is a purely linguistic module free of all social and usage information' is 'inherently impossible' (p. 692). In Adger (2006), I provided an analysis of the variable patterns of use of was and were in Buckie English which did indeed assume a 'purely linguistic' module, free of social and usage information. I assumed that speakers could have such information about lexical items, but that it was not accessible to the processes that put words together into phrases, or that establish agreement relations or dependency relations between words and between phrases. In that sense, I defended a model of Socio(linguistics)-free Syntax and Usefree Syntax, with the same kind of general architectural claims as models of Phonology-free Syntax (see, e.g. Zwicky & Pullum 1983). The first point to address, then, is whether the findings of variationist sociolinguistics are indeed compatible with such a model. I think the answer is that they straightforwardly are, and I'll outline the general model I've been proposing immediately below, showing how this is trivially the case. The second question is then whether the analysis I put forward for the Buckie case supports the general claim. Note that even if it didn't, the architectural points I make still stand : the model is in principle compatible with variability. However, I'll argue that Hudson's criticisms of Adger (2006) are not convincing.