Date of Completion

January 1986


Language, Linguistics




Due to the coarticulation of vowels with consonants, the formant frequencies of U and I are lower than produced between labial consonants than when produced in isolation. When asked to categorize matched U - I and wUw - wIw stimuli, listeners reveal their sensitivity to these acoustic differences by placing vowel boundaries closer to U for the /wVw/ continuum than for the /#V#/ continuum. Since a vowel's phonetic quality is constant across the two contexts, the "shift" in boundary locations must index some perceptual operation. Most likely, a speech-specific process keyed to the underlying articulatory equivalence of vowels accounts for the perceived phonetic equivalence. However, speech-neutral auditory mechanisms which isolate and interactively process rapidly time-varying ("consonantal") and more stable ("vocalic") regions of the signal might be posited instead.^ Three experiments examined the plausibility of this auditory hypothesis. In Experiments 1 and 2, listeners categorized sine wave U - I ("steady-state") and wUw - wIw ("time-varying") stimuli using either vowel ("U" vs. "I") or pitch ("LOW" vs. "HIGH") labels. A basic difference in response patterns was observed: Category boundaries along the /wVw/ continuum were displaced relative to those along the /#V#/ continuum for the vowel judgments, but not for the pitch judgments.^ The articulatory and the auditory theories can both account for the vowel results in the first two experiments. But, to account for the lack of a shift in the pitch results, the auditory theory must claim that listeners based their judgments solely on the apex frequencies of the time-varying stimuli. In order to test whether listeners are, in fact, able to isolate the "target" regions of these signals, subjects in Experiment 3 were asked to judge the vowel and pitch categories of the /wVw/ stimuli against those of the /#V#/ stimuli (and vice versa). As predicted by both theories, vowel category boundary locations were not altered by the "crossing" manipulation. Contrary to the auditory theory, however, it was found that pitch category boundary locations were strongly affected: For both stimulus sets, boundaries were shifted in a direction opposite to that characterizing the influence of consonantal context in vowel judgments. Thus, the results clearly support the inference that processes keyed to inherently dynamic, articulatory, information play an important role in perceiving coarticulated vowels. ^