Spectral integration in vowel perception: Matching and discrimination studies

Chistovich [J. Acoust. Soc. Am. 77, 789–805 (1985)] reviewed evidence suggesting that two vowel formants within 3.5 Bark of each other are integrated into a single spectral center of gravity in vowel perception. This center of gravity hypothesis, which has been an important concept in theories of vowel perception, was tested in a series of experiments utilizing two-formant stimuli (a continuum from [a] to [æ]) and one-formant stimuli. Listener’s one-formant matches to the two-formant standards had formant frequencies between the frequencies of the two-formant standards (as predicted by the center of gravity hypothesis), however there were no abrupt changes in the matching function over a range of two-formant separations from 0.5 to 6 Bark. The best-matching one-formant frequencies were generally higher when the bandwidth of the single formant was narrower, but the results were not affected by manipulations of overall amplitude. A discrimination experiment showed that listeners could discriminate between a two-formant stimulus and its best-matching one-formant counterpart. A second discrimination experiment showed that when the stimulus durations were very small (50 ms) and the interstimulus interval was large (3 s) discrimination between one- and two-formant stimuli was poorer when the two formants were close to each other in frequency. However, contrary to the center of gravity hypothesis, there was no sudden change of performance around the hypothesized 3.5 Bark boundary. [Work supported by UCLA, Linguistics Dept.]