D. D. Paschall
The Univ. of Texas at Dallas, P.O. Box 830688, GR4.1, Richardson, TX 75083-0688
Experiments were conducted to determine whether listeners with normal hearing can exploit information in a prior sample of speech to segregate a target voice from a second, competing voice. There are at least two ways that exposure to a prior sample of speech might aid the perceptual segregation of voices. First, peripheral adaptation processes can reduce the auditory response to preexisting sounds in order to enhance newly arriving energy associated with the onset of a second voice. Second, perceptual grouping based on harmonicity or spectrum envelope cues can reduce the interfering effects of preexisting sounds. To separate the contribution of these factors, concurrent vowel pairs were preceded by a precursor which had either the same F[sub 0] and spectrum envelope as one of the vowels (``voiced vowel'' precursor), the same F[sub 0] but a flat spectrum envelope (``buzz'' precursor), or the same spectrum envelope as one of the vowels, but with noise excitation (``whispered vowel'' precursor). All three types of precursors resulted in increased accuracy when the precursor was brief (200 ms). When the precursor was 1 s, only the voiced vowel precursor led to an increase in identification accuracy. These results suggest that both adaptation and perceptual grouping may play a role.