The Dynamic Model and Inherent Variability: The Case of Northern France

This paper explores the claims of the ‘dynamic’model of variation by testing it against data recorded in Avion, northern France. Parallels are drawn between langue d'oïl areas of France and decreolization situations in which proponents of the dynamic model have generally worked. Evidence to support the rejection of inherent variability within the dynamic model is found to rest on invariant performance for very small numbers of tokens, and on the dubious assumption that conversational code-switching does not occur. A lack of homogeneous speaker data in this case makes determination of lect structure all but impossible. In an attempt to isolate uni-code sequences, co-occurrence between variables is studied at the level of the clause, as in Burträsk by Thelander (1976, 1982). Significant differences from the Burträsk data emerge, and the claims of the dynamic model prove ultimately untest-able in the absence of reliably homogeneous data, even in relatively short talk sequences. The researcher is ultimately forced to abandon the model or dismiss the misfit between theory and data as “performance”, which, it is argued, may well be no more than inherent variability under a different name.