Research has shown that listeners accommodate talker differences in speech production by adjusting phonetic boundaries in light of a talker's unique productions. However, phonetic categories are not marked solely by boundaries, they also have a graded internal structure in that not all members of a category are considered equally good members. The current work examines whether listeners adjust the internal structure of a phonetic category for individual talkers and whether such adjustments would transfer to a novel word. Listeners participated in training and test phases. During training, two groups of listeners heard a talker produce "cane". Voice-onset-time (VOT) was manipulated such that one group heard the talker produce /k/ with relatively short VOTs and the other group heard relatively longer VOTs. During test, listeners were presented with a VOT continuum from "gain" to "cane" or "goal" to "coal" and asked to rate each member for goodness as /k/. The results showed that the best exemplar region at test differed between the two groups in line with exposure during training, though this effect was greater for the training word compared to the novel word. These results suggest that internal category structure is dynamically adjusted in order to accommodate talker-specific phonetic variation.
Bibliographic reference. Theodore, Rachel M. (2013): "Talker-specific perceptual processing: influences on internal category structure", In INTERSPEECH-2013, 2802-2806.