The perceptual classification of speech

Abstract

By employing new methods of analysis to the physical signal, a number of researchers have provided evidence which suggests that there may be invariant acoustic cues which serve to identify the presence of particular phonetic segments (e.g., Kewley-Port, 1980; Searle, Jacobson, & Rayment, 1979; Stevens & Blumstein, 1978. Whereas previous studies have focused upon the existence of invariant properties present in the physical stimulus, the present study examines the existence of any invariant information available in the psychological stimulus. For this purpose, subjects were asked to classify either a series of full-CV syllables ([bi], [be], [bo], [∂⁁], [di], [d∈], [do], [∂⁁]) or one of two series of chirp stimuli consisting of information available in the first 30 meec of each syllable. The full-formant chirp stimuli consisted of the first 30 msec of each syllable, whereas the two-formant chirps were composed of the first 30 msec of only the second and third formants. The object of the present study was to determine whether or not there was sufficient information available in either the full- or two formant chirp series to allow subjects to group the stimuli into two classes corresponding to the identity of the initial consonant of the syllables (i.e., [b], or [d]). A series of classification tasks were used, ranging from a completely free sorting task to a perceptual learning task with experimenter-imposed classifications. The results suggest that there is information available in the full-formant chirps, but not in the two-formant chirps, which allows subjects to group the sounds into classes corresponding to the identity of the initial consonant sounds.