The defining feature of a usage-based model of language is that there is assumed to be an unquestionable relationship between language structure and language use and that language use plays a defining role in shaping the grammar of individual speakers. In Bybee’s terms, "experience affects representation" (2001:67). If this is the case then it seems reasonable to suggest that the frequency with which different parts of the language system are used will also affect the way in which the linguistic system is organised and stored in cognition. This talk explores the role of lexical frequency in usage-based models of linguistics and the possibility of including this as an explanatory factor in accounts of variation and change.
The data are from a corpus of 38 hours of conversation (roughly 370,000 words), compiled over a two year period from a community in west Fife using the ethnographic technique of participant observation.
The first part of the paper deals with the role of lexical frequency in a phonological change in progress in this community. The change in question is th-fronting i.e. the realisation of (th) as [f]. Results from the analysis of this variable suggest that lexical frequency is only one of a number of factors influencing the direction of change in th-fronting. These results suggest that it is vital to consider the role of lexical frequency alongside other social, structural and cognitive motivations for variation and change.
The second part of the paper explores the role of lexical frequency in an apparently stable variable, the BIT vowel. There is very little discussion of the role of lexical frequency in cases of stable variation in the literature. Preliminary results from the analysis of variation of BIT suggest the existence of a paradox. On the one hand, certain lexical items show evidence of word specific phonetics (Pierrehumbert 2002), a phenomenon which can only be accounted for in a usage-based model of language structure. On the other hand, however, it would appear that the very existence of stable variation may pose a problem for the usage-based approach. How is it possible that a situation of variability can exist in such a model without the more frequently used variants becoming more entrenched and, over time, winning out?