Abstract:
Chistovitch and Lublinskaya [Hear. Res. 1, 185--195 (1979)] reported that listeners' judgments of vowel quality are influenced by the relative level of spectral peaks that fall within a critical distance (3.5 Bark). They suggest that the listeners respond to a ``perceptual formant'' at the spectral center-of-gravity (COG) for the peaks falling within the critical distance. This study was designed to determine whether the COG of the presumed neural excitation pattern might serve as a better predictor of listener performance than the COG of the signal spectrum. Listeners adjusted the center frequency of a simple spectral peak to match the location of the ``perceptual formant'' in several two- or three-formant signals. The relative level of the surrounding peaks was varied (-20 to +20 dB in 10-dB steps) to shift the location of the perceptual formant. In general, listeners were able to match the perceptual formant at each relative level. However, the matches were tightly grouped about the lower frequency peak when its level was 20 dB greater but more scattered when the higher frequency peak had the higher level. Neither COG of spectra, nor COG of excitation patterns are adequate to explain these differences. [Work supported by AFOSR.]