Talker-specific influences on phonetic category structure

Abstract

A major goal of research in spoken language processing has been to describe how listeners achieve stable perception given the marked variability in mapping between the acoustic signal and linguistic representation. Toward this end, the research of Joanne L. Miller and colleagues has shown that speech sound categories, like other cognitive/perceptual categories, have rich internal structure, with category membership represented in a graded fashion. Moreover, category structure robustly shifts as a function of variation in the speech signal including variation associated with phonetic context, speaking rate, and dialect. I will discuss evidence indicating that internal structure also reflects talker-specific phonetic variation. These experiments concern talker differences in voice-onset-time (VOT), an acoustic parameter that marks the voicing distinction in stop consonants. I will begin with findings from production experiments that explicate talker differences in VOT with the goal of generating predictions for how listeners might accommodate such differences. I will then present findings from perception experiments indicating that listeners comprehensively adjust the mapping between the acoustic signal and phonetic category to reflect a talker's characteristic VOT distribution. Collectively, these findings demonstrate that listeners accommodate variability in the speech stream by dynamically adjusting internal category structure in light of systematic acoustic variation.