4th International Conference on Spoken Language Processing
Philadelphia, PA, USA
Stevens suggested that certain features of consonants are auditorily robust. Such features are abrupt; manifest in 10-30 msec, e.g., [voice], [nasal], [continuant]. Other features such as [palatalized], [pharyngealized] are less robust and are carried on top of (and presumably require more time to be manifested than) the robust features. Robust features are used first-and sometimes exclusively-by languages in constructing a consonant inventory. The less robust features may not be used at all but if they are, the language has already used features from the robust set. We sought to test a similar hypothesis regarding vowels: within the first few tens of msec. only as many vowel contrasts can be differentiated as are attested most commonly in languages of the world - something like (IPA) /i e a o u/. Analysis of the confusion matrices made by listeners identifying end-gated versions of 11 N. Am. English vowels lends support to the hypothesis.
Bibliographic reference. Lang, Carrie E. / Ohala, John J. (1996): "Temporal cues for vowels and universals of vowel inventories", In ICSLP-1996, 434-437.