Acoustic and perceptual correlates of nasal vowels

Abstract

Various versions of different nasal vowels have been synthesized by systematically modifying the all-pole transfer function corresponding to nonnasal vowels. These modifications were achieved by manipulating the frequencies and bandwidths of the first ferment and of an added pole-zero pair over ranges that are consistent with acoustic theory and analysis of nasal vowels, including the effect of the sinuses. The stimuli were evaluated or identified by listeners (principally Gujarati speakers) whose language included a nasal-nonnasal opposition. Nasal judgments were obtained when the pole-zero pair was in the vicinity of 400 Hz, but for some vowels (especially high vowels) nasal responses were also obtained when the additional resonance was at a higher frequency. It is concluded tentatively that the nasal-nonnasal distinction in language is based on the fact that the auditory system responds distinctively when the spectrum in the vicinity of the first formant is flattened to yield a less prominent low-frequency spectral peak. This modification is achieved by the introduction of additional low-frequency resonances and by increased bandwidth for some of the low-frequency peaks. [Supported in part by a grant from NINCDS.]