Sociolinguists and generativists seem to have incompatible views of the psycholinguistic or cognitive status of variation. The sociolinguistic notion of inherent variability points to a single mental construct (i.e. a grammar) in which alternative realizations are possible. Generativists suggest that variation results only from the choice of different structures, an understanding that would imply different grammars, leaving only grammar switching as an alternative to inherent variability. Only some very recent contributions (e.g. Cornips and Corrigan 2005) attempt to deal with these apparently opposing views, and, in fact, historical attention to the problem has been minimal (but see Butters , Fasold , Preston [1991a, b, 1996a, b, 2000b, 2001b, 2002, 2004], andWolfram  for earlier attempts to deal with the psycholinguistic validity of the notion variable rule). To approach this problem, we propose at least three kinds of variationist sociolinguistics, although we might more properly speak here of levels rather than kinds.
Some few variationist studies have concerned themselves only with the correlation of linguistic and social facts, and the outcomes from such Level I studies do not seem to lead to ready psycholinguistic interpretations. This does not mean that such studies have no theoretical interest; such interest, however, seems to lie principally in the area of social theory or in the interaction of social forces and linguistic forms (see note 2). For example, in a study of doctor–nt interaction (Marsh 1981), the occurrence of definite article versus pronominal in such sentences as How's the pain in the/your hand? is investigated. Table 3.1 shows how this choice is distributed for patients and physicians, patient social status, and long-term versus short-term physician–patient relationships.