Perception of vowels with missing formant information

Abstract

The dominant formant-based model of vowel perception has been challenged by several whole-spectrum approaches. Recent arguments in favor of the importance of cues other than formant frequencies come from a study by Ito et al. [J. Acoust. Soc. Am., 110, 1141–1149] in which suppressing either of the first two formants did not radically change the identification of Japanese vowels. The present study replicates the experiment using the larger vowel system of the English language. Visual inspection shows that even when a formant is suppressed, listener responses do not deviate as much as would be expected if formants were the sole cue for vowel identification. However, quantitative analyses indicate that participant agreement in which vowel they heard is significantly lower when they respond to stimuli with suppressed formants. Additionally, the suppressed formant value becomes a less important predictor of vowel identity. These changes in responses become even larger when stimuli (original, F1-suppressed, and F2-suppressed vowels) are presented together rather than in separate blocks. Taken together, these results show that, although formants are not the only correlate to vowel identity, they seem to be the most important.