Vocoder Output and Whispered Speech in a Tone Language: Thai

Abstract

Phoneticians have tried in various ways to investigate the maintenance of tonal distinctions in whispering, where there is no variation of a fundamental frequency. In the present experiment it was found with whispered speech that native speakers of Thai, a language with five phonemic tones, could identify words minimally distinguished by tone, though with somewhat less accuracy than in normal speech. To determine whether such concomitant features as helped the subjects to make systematic distinctions in the whispered speech were also present in normal speech, the spoken words were passed through the Vocoder with the fundamental frequency of the buzz kept constant. No discriminations were made! With hiss alone, however, the results were better than in the natural whisper. We conclude that the features were present in the normal speech but that in the presence of the buzz, listeners were set to hear pitch variations. Inspection of spectrograms suggests that tonal oppositions in Thai whispering lean on such concomitant features as changes in intensity, relative durations of vowels, and small variations in formant frequencies.