Difference in Second-Formant Transitions between Aspirated and Unaspirated Stop Consonants Preceding [α]

Perceptual experiments with synthetic speech have shown that the category boundary on an acoustic [pa]— [ta] (/ba/-/da/) continuum (obtained by varying the onset frequencies of the second and third formants) is closer to the labial endpoint than the boundary on a [p h a]-[t h a] (/pa/-/ta/) continuum. of several possible explanations, the most plausible seems to be that natural unaspirated and aspirated stops have different formant transitions. To supplement limited data on this point in the literature, we conducted an acoustic analysis of CV syllables produced by 10 male speakers of American English. The results show very clearly that the second formants of [p h a] and [t h a] start 100–200 Hz higher than those of [pa] and [to] and reach comparable frequency values only at voicing onset. This difference, which is probably an acoustic consequence of subglottal coupling during aspiration, seems to be part of a listener's tacit knowledge of phonetic regularities and thus explains the perceptual boundary shift. It also needs to be taken into account in realistic speech synthesis.