Auditory Contrast versus Compensation for Coarticulation: Data from Japanese and English Listeners

English listeners categorize more of a [k-t] continuum as "t" after [integral] than [s] (Mann & Repp, 1981). This bias could be due to compensation for coarticulation (Mann & Repp, 1981) or auditory contrast between the fricatives and the stops (Lotto & Kluender, 1998). In Japanese, surface [integral k, integral t, sk, st] clusters arise via palatalization and vowel devoicing from /sik, sit, suk, sut/, and acoustic vestiges of the devoiced vowels remain in the fricative. On the one hand, compensation for coarticulation with the devoiced vowel would cancel out compensation for coarticulation with the fricative, and listeners would not show any response bias. On the other hand, if the stop contrasts spectrally with the fricative, listeners should respond "t" more often after [integral(i) under circle] than [s (u) under circle]. Experiment 1 establishes that [k] and [t] coarticulate with preceding voiced [i, u], voiceless [(i) under circle, (u) under circle], and [integral, s]. Experiment 2 shows that both Japanese and English listeners respond "t" more often after [integral(i) under circle] than [s (u) under circle], as predicted by auditory contrast. English listeners' "t" responses also varied after voiced vowels, but those of Japanese listeners did not. Experiment 3 shows that this difference reflects differences in their phonetic experience.