Respiratory constraints on speech production: starting an utterance.

The aim of this research is to model the sequences of respiratory and articulatory events that occur when a speaker starts talking. In this paper, the focus is placed on a set of timing landmarks--utterance onset, phonation onset, and the initial peak in subglottal pressure--in relation to the first sound segment in the utterance and the location of the first syllable in the utterance that is judged as prominent by a listening panel. Simultaneous recordings of the acoustic signal, airflow, lung volume, and subglottal pressure as estimated from esophageal pressure were collected from four native speakers of American English. The timing of the initial alveolar pressure peak generally coincides with a prominent syllable if that syllable is the first or second syllable in the utterance, and precedes later prominences. There is considerable scatter in phonation pressure and phonation onset time across speakers and utterances. Normalizing these parameters to the initial pressure peak provides a more coherent view of the pressure trajectory and the influence of the initial sound segment on phonation parameters. Use of the data in the development of a model of the respiratory system during speech is discussed.