Lexically biased perceptual adaptation

Abstract

Listeners are able to rapidly shift their acoustic-phonetic categories in the face of speaker-specific variation. However, an open question is how top-down expectations can influence this perceptual adaptation. Under Adaptive Resonance Theory [Grossberg (2003)], if an acoustic token provides initial activation of a linguistic category, top-down expectations about that category can boost the resonance and lead to activation of the category even with an ambiguous token. Ambiguous tokens in positions with stronger expectations are hypothesized to lead to less perceptual adaptation than in positions with lower expectations. To test this hypothesis, two groups of participants will be exposed to 10 ambiguous s-∫ sounds during a lexical decision exposure task. The groups differ in where in the word these sounds will occur, either in the onset or coda of CVC words. Lexical bias, as a form of top-down expectations, is stronger for ambiguous coda sounds [Pitt and Szostak (2012)]. I predict that in a s-∫ categorization task following exposure, the coda group will differ less than the onset group from a control group that only completes the categorization task, indicative of less perceptual adaptation.