Kenneth N. Stevens
Res. Lab. of Elec. and Dept. of Elec. Eng. and Comput. Sci., MIT, Cambridge, MA 02139
The prosody of an utterance is often described in terms of the locations of attributes such as pitch accents, boundary tones, and vowel reduction. These prosodic descriptions are usually represented in the sound by changes in fundamental frequency, duration, and other temporal and spectral attributes of the segments. This paper presents data showing that the amplitude and spectrum of the glottal pulses, as well as their frequency, undergo modification depending on locations of pitch accents and boundary tones and depending on vowel reduction. The following results were obtained: (1) Reduced vowels tend to have significantly lower amplitude (7 to 13 dB) than nonreduced vowels, together with an increased first-formant bandwidth and an increased tilt in the source spectrum; (2) a nonreduced vowel has a somewhat reduced amplitude (by about 6 dB) when it follows a nuclear pitch accent; and (3) a nonreduced vowel in utterance-final position in a declarative sentence shows a decreased amplitude of the first-formant peak, an increased open quotient, and an increased spectrum tilt. Production mechanisms that could account for these changes include modifications in subglottal pressure, adjustments of the glottal configuration, and influences of constrictions in the supraglottal airway. [Work supported in part by a grant from NIDCD.]