Thesis (Ph. D.)--University of Washington, 1997. / Vita. Includes bibliographical references (leaves [84]-91).

The present study examines the effect of context on the use of transitions as cues to vowel perception. Thirty V₁CV₂CV₁ utterances were recorded, with V₁ being one of the three vowels /a,i,u/, and V₂ one of ten English vowels (/ i , I, el, E, ae, a,^,Ou,U,u/). After removal of the outer vowels (V₁), three sets of stimuli were created from the CV₂C parts: (1) unmodified controls (CO); (2) V₂ steady-state only (SS); and (3) transitions only (TR). Twenty subjects were asked to identify V₂. Subjects and speaker were matched for dialect and all subjects had some phonetics training. Results showed significant differences across conditions and contexts. Scores for SS stimuli, for all contexts, were as high as for CO stimuli. Performance on the TR stimuli was as good as on the other two conditions for two of the contexts. However, for the TR condition--/a/ context, performance was considerably worse than for any other combination of conditions and contexts. Possible reasons for this are discussed, and the need for testing of other vowel contexts is emphasised. It is concluded that, in some V₁CV₂CV₁ contexts, transitions can provide information about vowel identity on a level equal to steady-state alone, or to the combined information provided by both transitions and steady-states. This effect, however, is not uniform across contexts. For at least one context, transitions alone are not sufficient to cue vowel identity at a level comparable to steady-state or combined information. This lack of uniformity suggests that the role of transitions varies with the type of vowel context present, and conclusions about general usefulness await systematic testing of a number of vowel contexts. / Medicine, Faculty of / Audiology and Speech Sciences, School of / Graduate

No description available.

No description available.

No description available.

No description available.

No description available.

Digitized by Kansas Correctional Industries

Speech perception is challenging because the acoustic input is extremely variable. This variability partially stems from differences in how talkers pronounce words. For example, Voice Onset Time (VOT) is the primary cue that distinguishes /b/ from /p/. Women tend to use longer Voice Onset Times (VOTs) than men. A VOT of 20 msec could thus be a /b/ spoken by a woman and a /p/ spoken by a man. A critical question is how listeners deal with this variability. Previous research shows that listeners use these regularities (e.g., the systematic relationship between gender and VOT) to compensate for variability. For example, listeners adjust their phoneme category boundary based on talker gender. However, it is unclear the exact mechanisms by which talker gender information influences speech processing. Talker gender could influence only later stages of speech processing, like phoneme categorization. Alternatively, talker gender could modulate the earliest stage: acoustic cue encoding. I use event-related potentials, eye-tracking in the visual world paradigm, and electrocorticography to isolate the specific role of talker gender in speech perception. The results show that the auditory system influences the earliest stage of speech perception by allowing cues to be encoded relative to prior expectations about gender and that gender is integrated with acoustic cues during lexical activation. These experiments give insight into how the brain deals effectively with variability during categorization.

No description available.