Plasticity of categories in speech perception and production
A longstanding issue in speech processing involves the relationship between the categories used in perception and production. Evidence shows that perceptual categories adapt to recently heard speech, but evidence that this adaptation affects production is mixed. In this talk I will describe three experiments where we measured baseline production and perception of bilabial voiced and voiceless English stop consonants, and then exposed participants to an unlabelled bimodal distribution of VOT, in which we artificially shifted the implicit category boundary between voiced and voiceless stops. Our results show that participants are able to adjust their perceptual category boundaries on the basis of new distributional information, and this leads to changes in production categories influenced by the new category boundary. I will argue that these data are consistent with models in which speech perception and production categories are closely linked, but constraints exist on the transfer between perception and production. I will also discuss how models of phonemic accomodation, exemplar models and Bayesian models may have difficulty accounting for our findings.