Consonant/vowel ratio: an improbable cue in speech.

Abstract

The idea of the consonant duration relative to the vowel duration as a cue to voicing of postvocalic consonants is almost three decades old. Port and Dalby (1982) offer what they believeto be convincing support for such a view. We argue that the acceptance of this idea is premature and probably incorrect. In every instance that the concept of consonant! vowel (C/V) ratio is contrasted with an alternative view, the latter provides a significantly better description of the results. This alternative view considers consonant and vowel duration as independent cues to voicing, following the evaluation and integration of cues within the fuzzy-logical model of speech perception (Massaro & Cohen, 1976, 1977;' Massaro & Oden, 1980a, 1980b;Oden & Massaro, 1978). The first relevant study was carried out by Denes (1955), who varied vowel duration and the final consonant duration in the perception of the voicing difference between the pronunciations of the word use. Without any formal analysis, Denes interpreted the results in terms of the ratio of consonant duration to vowel duration, serving as a cue to voicing of the final consonant. The perceived voicing should decrease systematically with increases in the C/V ratio. However, in an earlier paper (Massaro & Cohen, 1977), we provided a quantitative test of the C/V ratio hypothesis against the observed results of Denes (1955). We also tested the idea that vowel duration and consonant duration provide independent cues to voicing of the consonant. The fit of the latter model was twice as good as that of the former, even though the better model required only three-fifths as many free parameters. In the present note, we evaluate the C/V ratio model against the data of Derr and Massaro (1980) and Port and Dalby (1982). The present tests offer as much flexibility as possible for the quantification of the C/V ratio idea. In addition, we contrast the C/V ratio model against the predictions of the fuzzylogical model, assuming independent vowel and consonant duration cues. The fuzzy-logicalmodel does a