Date of Defense

6-2002

Department

Speech Pathology and Audiology

First Advisor

James Hillenbrand, Speech Pathology & Audiology

Second Advisor

Michael Clark, Speech Pathology & Audiology

Abstract

The purpose of this study was to shed light on the longstanding debate between the two major speech perception theories offered by previous research, Formant Theory and Spectral Shape Theory. While proponents of Spectral Shape Theory believe human listeners make judgements of vowel identity based upon a sound signal's overall spectral envelope, Formant Theory claims that speech perception is accomplished simply through the analysis of information provided by the frequency locations of a signal's spectral peaks. Two experiments were conducted. Experiment I presented listeners with 900/hVd/ utterances to identify. These were either unmodified, original, naturally spoken syllables, a +9dB/octave uptilt modification of the signals, or a -9dB/octave downtilt modification. Listener response produced no significant differences in intelligibility across the three conditions, suggesting that vowels remain intelligible as long as spectral peaks are preserved. Experiment II consisted of three separate tests, all of which utilized two synthesizers. The envelope vocoder synthesizer retains a speech signal's spectral envelope and the damped sinewave synthesizer retains a speech signal's spectral peak locations. For test A listeners identified CV utterances and produced a notable difference between intelligibility rates for the three conditions, suggesting that only spectral envelope, let alone simply spectral peak locations, are sufficient for transmission of consonant identity. Both tests B and C presented sentences for listeners to identify in two conditions. Results revealed no notable differences in intelligibility between the two conditions for either test. These seemingly contradictory results are investigated at length in the discussion section.

Access Setting

Honors Thesis-Campus Only

Download

Share

COinS