Using the acoustic signal to make inferences about place and duration of tongue‐palate contact

Abstract

Productions by four native speakers of American English of /t/ and /d/ in various intervocalic contexts were examined. Closure duration was measured from the acoustic signal (using both waveform and spectral analysis) and from dynamic palatography data. Place of contact (mean of front‐most and rear‐most contact) was observed from the palatographic signal; formant‐frequency measurements (F2, F3, F4) in the closure interval were made. Measurements were averaged over six productions of each stimulus by each talker. Duration measurements by the two techniques correlated well, except (as expected) in instances in which the second vowel had a glottal onset, in which case the durations of low amplitude in the acoustic waveform were longer than the durations of the tongue‐palate closure measured palatographically. Place of contact correlated best with F3 measurements. Normalization across talkers was achieved by dividing each F3 measurement by the talker's mean F2. [Work supported by NICHD.