The writer introduces the idea of sociolinguistic competence, the ability of a speaker to produce and recognize an infinite number of inter-idiolectal code switches, and discusses two methods of dealing with such language variations: frequency analysis and implicational analysis. In frequency analysis, the method used by sociolinguists such as Labov and Wolfram, the linguistic data are correlated to non-linguistic data, resulting in a statement of the frequency with which a given linguistic form appears in a given non-linguistic context. Frequency analysis, in generalizing the empirical description of a corpus, deals with linguistic performance. Implicational analysis, on the other hand, "attempts not to describe a set of speech acts but to model the idealized competence of the persons involved in those speech acts." Variations are explained by a set of conditional statements of the form: If F1, then F2; if F2, then F3, etc. Implicational analysis is preferred to sociolinguistic work because implicational scales, having been structured without reference to non-linguistic data, permit the correlation of socio-economic and other contextual data to linguistic variation, continuously and without circularity of argument. The significance of the linearity of implicational scales and their place in a generative grammar are also discussed. (FWB)