Qualitative separateness in children's speech

Abstract

Acoustic analyses of fricative‐vowel (FV) syllables spoken by adults and young children (ages 3 to 7 years) have demonstrated several differences. In children's speech: (1) the gross spectra of /∫/ and /s/ segments were more similar than those of adults; (2) fricative F2 frequencies varied more as a function of vowel context; and (3) the relative F2 amplitude was greater. These acoustic differences between children's and adults' speech provided an opportunity to test two competing hypotheses of perceptual segmentation. According to one, listeners divide the signal into temporally discrete, context‐sensitive allophones that can be used to make inferences about later‐arriving segments. The second hypothesis suggests that listeners can separate simultaneously arriving acoustic information into successive phonetic units. In the present set of experiments, brief portions of fricative noise extracted from FV syllables spoken by adults and children to normal adult listeners were presented. Results showed that listeners were less accurate in fricative identification for child speakers, but more accurate in identifying the vowel context from which the fricative noise came. Thus the greater coproduction found in children's speech was more easily recovered by listeners, even though the most prominent phone was more poorly perceived. This finding was taken as support for the second hypothesis. [Work supported by NIH Grants NS‐07237 and HD‐01994 to Haskins Laboratories.]