Exploring the role of spectral integration in the perception of vowel sounds
Randy L. Diehl · The Journal of the Acoustical Society of America · 2005
Building on the important work of Chistovich and Lublinskaya [Hearing Res. 1, 185–195 (1979)], Syrdal [Speech Commun. 4, 121–135 (1985)] proposed that vowel category boundaries tend to occur in quantal regions corresponding to a 3.5 Bark bandwidth of spectral integration. This proposal was shown to be consistent with acoustic distributions of certain American English vowels. Some initial perceptual results [K. A. Hoemeke and R. L. Diehl, J. Acoust. Soc. Am. 96, 661–674 (1994)] also tended to support the role of a 3.5 Bark integrator in determining major perceptual boundaries among vowel categories. However, other evidence, reviewed here, raises doubts about the generality of Syrdal’s proposal. The negative evidence includes both within- and cross-language variation in the location of vowel category boundaries assumed to be governed by the 3.5 Bark integrator and the absence of elevated discriminability in putative quantal regions corresponding to the spectral limit of integration. [Work supported by NIDCD.]