Jose R. Benki
Christine Bartels
John Kingston
Linguistics Dept., Univ. of Massachusetts, Amherst, MA 01003
Neil A. Macmillan
Univ. of Massachusetts, Amherst, MA 01003
Two of the dimensions along which vowels vary, rate of vocal fold vibration (F0) and tongue height (F1), covary directly in natural speech. These experiments examined whether F0 and F1 integrate perceptually for vowels in contrastive intonation contexts, as they have been shown to do in isolation [Kingston, Phonetica 47, 149--179 (1991)]. The vowels occurred in the second, stressed syllable of a [w(schwa)'w_w(schwa)] frame bearing either a [H+L* H-L%] or [L+H* L-L%] intonation contour. This set of contexts was designed to affect listeners' expectations about the value of F0 in the target vowel. Listeners classified the imperfectly discriminable vowels by F0 and/or F1, according to the different tasks of the Garner paradigm. Both accuracy and response time were measured. The perceptual representation of the vowels can be inferred from a detection-theory analysis of the accuracy data; response times facilitate comparison of the results with the bulk of other work using the Garner paradigm. [Work supported by NIH and NSF.]