Dialect data, lexical frequency and the usage-based approach

Unlike most work in the generative tradition, researchers employing usage-based models of language change (e.g. see the collection of papers in Barlow and Kemmer 2000) often incorporate variable data in their model building. Most of this variable data comes from large corpora. While this approach appreciates the importance of testing models of language change on observed language use, it is also problematic because it forces the researcher to test theories of language change on abstract language varieties such as ‘American English’. This is particularly problematic because the usage-based approach assumes that the speakers’ linguistic system is abstracted largely from their previous experience and, hence, that no two speakers will have the same grammar. This paper aims to redress this mismatch by considering the role of lexical frequency in a usage-based model of phonological change in light of new data that was collected from a relatively (geographically and socially) homogeneous group of speakers living within a single dialect area in east-central Scotland.