The aim of this paper is to reconsider aspects of the relationship between sociolinguistic variation and linguistic theory. According to Casillas-Martínez, there have been two main approaches to sociolinguistic variation in linguistic theory. The first of these, the ‘exclusion approach’, has been a dominant trend in generative linguistics throughout the twentieth century (2003:33). Section 1 begins by tracing the history of this approach and questions the evidence that has been provided in support of it. Section 2 reviews a sample of research that is characteristic of the second approach—‘variation as a side effect’ (2003:34). This approach often results from the recognition that linguistic theory should be capable of explaining sociolinguistic variation and so attempts are often made to modify existing approaches that are, in all other respects, asocial. I will argue that, while these accounts appropriately question the legitimacy of the distinction between linguistic theory and sociolinguistic variation, their proposals to incorporate variation are problematic because they do not fully incorporate the social meaning of linguistic variants. In order to improve the synthesis between sociolinguistic variation and theories of language structure, I will argue that it is necessary to adopt a linguistic theory in which the social meaning of linguistic variation is a pre-existing aspect of the framework. Section 3 explains that this is the case with the theoretical approaches of the Cognitive Linguistics movement. I will therefore argue that these frameworks are fundamentally compatible with accounts of sociolinguistic variation and provide a more viable approach to socially motivated variation in linguistic theory.