William D. Clarke III
Dept. of Commun. Sci. and Disord., Univ. of South Florida, Tampa, FL 33620
In a previous study [Clark, J. Acoust. Soc. Am. 93, 2423--2424(A) (1993)], listeners identified ten American English vowels in dVd silent-center (SC) syllables. Syllables were presented in both natural and synthetic form embedded in a naturally spoken sentence. The synthetic syllables were modeled on the natural syllables so that formant trajectories were identical. While performance on both natural and synthetic unmodified syllables was very good (98% correct), errors in the synthetic SC condition exceeded errors on natural SC syllables (23% vs 11%). Most errors were on short lax vowels (44% errors on lax vowels versus 4% on tense vowels) which were misinterpreted as spectrally similar long vowels. The present study tested the hypothesis that ``perceptual lengthening'' may have occurred for SC syllables. Silence duration was reduced 33% and 67% in both natural and synthetic SC syllables. For natural SC syllables the error rate did not change. For synthetic SC syllables errors on lax vowels were reduced to 41% and 20%, respectively. The results suggest that listeners were using syllable duration cues to identify vowels in synthetic SCs more than in natural SCs. Thus perceptual lengthening had a greater effect on perception of synthetic syllables. [Work supported by NIDCD.]