Increasing Realism of Auditory Representations Yields Further Insights into Vowel Phonetics

The ear’s remarkable ability to cope with noisy signals is linked to its use of a spatio-temporal mechanism that distributes information about strong spectral components across neural units whose characteristic frequencies (CFs) often span broad frequency ranges. Speech formant information is thus carried not only by channels with CFs near spectral peaks but also by adjacent channels. This paper examines several implications of this mechanism for vowel phonetics: (1) Although there is no evidence of explicit ‘formant tracking’, spectral peaks are nonetheless granted a special status. (2) There is an auditory warping of the vowel space that enhances contrasts along the open-close dimension relative to the front-back dimension. (3) Incorporating the above mechanism into auditory models helps to unify accounts that appear to take somewhat different approaches to explaining the structure of vowel systems (viz., Stevens’s Quantal Theory, the Grenoble group’s Dispersion-Focalization Theory and Lindblom’s Adaptive Dispersion Theory).