J. E. Flege
A. C. Walley
L. Randazza
Depts. of Biocommun. and Psychol., Univ. of Alabama at Birmingham, Box 503 UAB Station, Birmingham, AL 35294-0019
Two vowel continua were created using the Klatt synthesizer. F1 frequency varied in the same eight equal mel steps in both. F2 frequency also varied in eight equal mel steps but, whereas it increased relative to /i/ in one continuum, it decreased relative to /i/ in the other. As a result, vowels in the ``native--native'' continuum ranged from /i/ to /i/, those in the native-foreign continuum from /i/ to a vowel symbolized as /Y/. Adults and children ranging in age from 4 to 12 years (ten per group) were trained to identify the endpoints of both continua as either /i/ or ``the other vowel.'' There was no effect of subjects' age on the number of stimuli identified as /i/. This failed to support the hypothesis that vowels not bounded by another native language category---such as /i/ in the /i/--/Y/ but not the /i/--/i/ continuum---will expand outward with age. Slope analyses revealed that the 4- to 6-year-old children had shallower identification functions than the adults. Much the same effect of age was evident for the /i/--/i/ and /i/--/Y/ continua. This argues against attributing the observed age effect on slopes to a poorer definition of vowel categories by the 4- to 6-year-olds than by the adults. In metalinguistic tests, the 4-year-olds performed a rhyming task more poorly than 5- to 8-year-olds. The 4- and 5-year-olds were less able than 6- to 8-year-olds to segment initial and final consonants. The slope differences may therefore have arisen from age-related differences in ability to perform perceptual tasks involving localized sound segments. [Work supported by NIH.]