Formantless vowels and theories of vowel perception

Abstract

Most theories of vowel perception rely on a stage of formant peak extraction. However, combining the indications from earlier studies with single formant vowels [Chistovich and Lublinskaya, Hear. Res. 1, 185–193 (1979)] and formantless vowels with “flat” spectra [Carpenter and Morton, Lang. Speech 5, 203–214 (1962)], it would appear that perceived vowel quality does not depend on extraction of two or more formant peaks. In this study, Carpenter and Morton's results are extended by constructing a series of tones with “step” spectra with F0 (125 Hz) represented by the first harmonic. It was discovered that a full range of clear, satisfactory front vowel qualities ([i] to [a]) resulted with a band of high‐frequency harmonics, where the upper frequency cutoff was 3.5 kHz, and the low‐frequency edge varied from 3‐1 kHz. Similarly, a back vowel series could be created with a low‐frequency band consisting of the first two to the first nine harmonics. The results are interpreted in terms of gross spectral dominance as a prime determinant of phonetic vowel quality. Further experiments with formant and formantless vowels are planned to test this hypothesis.