The perception of stop consonants: Spectral tilt revisited

Abstract

The claim that invariant features involving spectral tilt, as proposed by Lahiri et al. (1984), are used by human listeners when classifying stop consonants was explored. Subjects were presented with stop‐vowel syllables representing [b], [d], and [g] with a range of vowels in an identification task. Stimuli were either natural, synthetic (based on the natural set), or modified tokens (synthetic tokens containing a burst which, according to the Lahiri et al. metric, should be misidentified). Results indicate that the change in spectral tilt plays little role in stop perception. Only [g] showed any evidence of a systematic effect. To investigate further the difference found in the identification task, tone analogs mimicking the [b], [d], and [g] CV sets were produced. Both speech and nonspeech groups were run to examine the perceptual locus of any effects of change in spectral tilt. The results of these two groups will be compared to those of the speech groups in the first experiment. The implications for theories of auditory to phonetic coding of stops and possible invariant attributes in stop perception will be discussed.