Overgeneralization and falsifiability in phonological theory
Sylvia Blaho, Curt Rice
December 2014

A prevalent trend in current phonological practice, usually justified by invoking a simplistic version of Popperian falsifiability, is data-fitting: fine-tuning phonological models, often with the help of ad hoc restrictions, in order to exclude unattested languages from the set of possible grammars generated by the model. We argue that this practice should be abandoned, for three main reasons. First, it is based on a dubious interpretation of Popperian falsifiability. Second, ad hoc restrictions do not further our understanding of language, but they decrease the coherence of a model. Third, the practice is based on data from a small subset of existing languages. We examine each of these arguments in detail.
Format: [ pdf ]
Reference: lingbuzz/002323
(please use that when you cite this article)
Published in: La phonologie de français: normes, périphéries, modélisation, ed. by Jacques Durand, Gjert Kristoffersen & Bernard Laks, 2014, Presses universitaires de Paris Ouest
keywords: phonology, falsifiabilitiy, methodology, overgeneralization, theory construction, hungarian, phonology
previous versions: v1 [December 2014]
Downloaded:1790 times


[ edit this article | back to article list ]