Abstract:

Johnson et al. [J. Acoust. Soc. Am. 94, 701--714 (1993)] suggest, on the basis of observed inter-speaker variability in discriminant analyses of articulatory measures, that speakers utilize acoustically defined targets in speech production. The present paper compares inter-speaker variability in simultaneously recorded acoustic and articulatory data (st words) from five New England speakers. The articulatory data were x and y coordinates of coils on the tongue, lips, and jaw, transduced by the Haskins Laboratories EMMA system and recorded at three locations in the target vowel; acoustic data were F1, F2, and F3 measures at the same temporal locations. Discriminant analyses of the articulatory and acoustic data sets reveal congruent patterns of inter-speaker variability in the two domains. The inter-speaker differences do not reflect superficial dialect or idiolect differences (e.g., extent to which /(open aye)/ and /(open oh)/ contrast, or tendency to glottalize syllable-final /t/). Rather, they reflect differences in the way subjects vary jaw position, especially height, suggesting an anatomical basis for the observed differences. Thus, within the limits of their oral morphology, these speakers are using comparable articulatory targets for speech sounds.