Perceptual classification of information in vowel-consonant syllables

Abstract

Two experiments are reported which examined whether information specifying consonantidentity was available in brief segments at the offsets of vowel-consonant syllables. The first experiment employed a classification task in which the subjects were required to sort eight synthetic stimuli composed of two stop consonants, /b/ and /d/, in four vowel environments. It was found that the subjects' responses were best described by a classification strategy based on overall acoustic similarities between the stimuli. It was hypothesized that these acoustic similarities could be predicted by averaging the frequencies of the second and third formants at offset. A perceptual learning task was used in Experiment 2. Although the salience of the acoustic similarities was again evident, the results also indicated that the subjects were able to learn classification schemes based on acoustic-phonetic similarities. Subjects made fewer errors in learning to sort the stimuli by both consonant-identity and vowel-similarity rules in comparison to an arbitrary division when all of the formants were left intact. These data are interpreted as an indication that brief segments of speech contain invariant cues to phonetic identity and that the salience of phonetic classifications increases as the sounds retain more of the information found in speech. © 1985 Psychonomic Society, Inc.