Formants as a Voicing Cue in the Perception of /z/

Abstract

In earlier synthetic speech studies by researchers at Haskins Laboratories of the cues for the voiced-voiceless distinction of fricatives in initial position, hiss duration, intensity of the hiss relative to the vowel, separation of hiss from the vowel by a short period of silence, and the consonant-vowel transitions all had been shown to have an effect. In the present study, the addition of voiced formants to the hiss portion has been investigated. These formants are a rather obvious device for signaling voicing since a voice bar and often higher resonances as well can be seen clearly on spectrograms. In this experiment versions of the syllables /si/ and /zi/ were synthesized on PAT using the following variables: hiss amplitude: −5, −10, −15, and −20 db from the peak intensity of the vowel: voiced formant amplitude during fricative portion: −10, −20, and −30 db from the peak intensity of the vowel, as well as hiss without these formants; duration of the fricative portion: 85, 124, 170, and 213 msec; a 20-msec separation of the fricative portion from the onset of vowel transitions vs no separation; and vowel portions copied from spectrograms of spoken /si/ and /zi/. Low hiss intensity, short duration, no separation, and vowel formants from spoken /zi/ all tended to produce z judgments, but, as anticipated, the addition of voiced formants was the overriding cue.