Does a listener need to recover the dynamic vocal tract gestures of a talker to recognize his vowels?

Abstract

Although it has been reported that vowels are identified better in the context of a CVC syllable than in isolation, the basis for this advantage is still the subject of some controversy. At the Providence meeting of the Society, Fowler and Shankweiler presented the results of a study comparing vowel identification in speech and nonspeech contexts and argued that the advantage observed for vowel identification in CVC contexts was a consequence of the listener's sensitivity to “acoustic patterning in the speech signal that specifies the vocal tract gestures of the talker.” In the present study, we compared the identification of eight vowels presented in isolation to those presented in the following three different CVC contexts: (1) English fricative-vowel-fricative [sVs], (2) Non-English fricative-vowel-fricative [XVX], and (3) Nonspeech-vowel-non-speech [nsVns] that was generated with a broad-band flat spectrum noise source. In contrast with the results of earlier studies, we found that vowel identification was substantially better in isolation than in any of the CVC contexts. This result suggests that dynamic spectral change is probably not a necessary requirement for accurate vowel recognition to be observed. Furthermore, these findings indicate that a listener does not have to “recover” the dynamic vocal tract gestures of a talker to recognize his vowels. [Supported by research grants from NIMH and NINCDS.]