Laboratory of Phonetics & Psycholinguistics

P. C. Shinn, S. E. Blumstein, and A. Jongman

Two series of experiments were performed to test the effect of varying vowel length (reflecting different speaking rates) on the crossover boundary of a [b-w] continuum. Earlier research (Miller & Liberman, 1979) indicated context- dependent perception of [b] and [w] as a function of syllable duration. The object of our study was to determine whether the syllable-duration effects shown for [b-w] would be lost if the parameter values of the test stimuli were more similar to natural speech. Different combinations of acoustic parameters were manipulated, as was the method of stimulus presentation. Results showed that the magnitude of the syllable-duration effect was greatly reduced and that the effect was eventually lost as the parameters increasingly approximated natural speech. In addition, the syllable-duration effect was differentially affected depending on the experimental design being used: in most experimental conditions the syllable-duration effect was more affected when the stimuli were presented in a blocked, as compared to a mixed design. Nevertheless, when parameters similar to natural speech were used, the syllable- duration effect was completely lost, regardless of which experimental design (mixed vs. blocked) was used. These results suggest that the syllable-length effect for the perception of the [b-w] contrast may not be as robust as originally believed, and, furthermore, that the role of context-dependent cues in the perception of speech may have been overestimated.

Limitations of context conditioned effects in the perception of [b] and [w]