First International Conference on Spoken Language Processing (ICSLP 90)
Two experiments were conducted to examine the distribution of perceptual cues for Japanese intervocalic stop consonants, with a special interest in the perceptual role of the pre-closure portion for Japanese listeners. In Experiment 1, as predicted by Japanese phonological structure, Japanese listeners could not identify stop consonants correctly when only pre-closure portions of vowel-consonant-vowel(VCV) utterances were presented . However, in Experiment 2, despite the apparent ineffectiveness, the pre-closure portion demonstrated strong effects for place perception of intervocalic stop consonants. In natural VCV utterances, even when the release burst and most of the CV formant transitions were replaced by a louder white noise, nearly 90% of original consonants were clearly "heard" under the noise. On the other hand, without VC coarticulation, only about a half of original consonants were restored. Moreover, when a preclosure and a post-closure of different intervocalic stops were spliced together, the pre-closure dominated the consonant restoration as the noise duration increases. These results suggest that Japanese listeners can utilize cues for Japanese intervocalic stop consonants that occur not only after the closure, but which are also distributed in the region before and after it.
Bibliographic reference. Kashino, Makio (1990): "Distribution of perceptual cues for Japanese intervocalic stop consonants", In ICSLP-1990, 557-560.