The insertion of pauses in the speech stream contributes to the effectiveness and naturalness of speech. Pauses provide structure to the speech stream (boundaries), signal discourse control (e.g. turn-taking), and serve as indicators of cognitive state (e.g. hesitations). The frequency and duration of pauses are important factors in the perception of fluency for normal and disordered speech (e.g. Goldman-Eisler, 1972;Hammen and Yorkston, 1996;Nishio and Niimi, 2001). In general, a speaker may inhale at a pause or pause in the course of an exhalation (within-breath). Both types are highly frequent in normal and disordered communication. Current models of speech production assign a largely passive role to the respiratory system, yet preliminary data indicate that pauses are actively generated by re-adjustments of the respiratory system (Slifka, 2006). Models of speech production are limited by the lack of a quantitative understanding of the physical requirements for generating within-breath pauses, the impact of those requirements on the resulting speech acoustics, and the relation between within-breath pauses and the speech planning process. We propose three basic aims focusing on aerodynamic, physiological, and muscular factors.
Aim 1 is to develop a baseline quantification of the magnitude, duration, and pattern of respiratory system involvement at within-breath pauses. Net muscular pressure is derived from a graphical representation of pleural pressure and lung volume, and a piecewise-linear fit to the data will be determined for each pause. Degree of involvement of expiratory versus inspiratory muscles is estimated from surface electromyography. Quantifying pause-related activity also requires examining how factors such as articulatory actions and prosodic function influence respiratory system involvement. Suspension of sound sources at within-breath pauses is assumed to result from a coordinated set of actions between the respiratory system and the active articulator(s). Distinctly different behavior is expected for three manner classes: sonorant, obstruent, and fricative.
Aim 2 is to determine the degree of correlation between measures used to quantify pause-actions (Aim #1) and both the duration of the pause and the type of sound segments bracketing the pause.
Aim 3 is to determine the degree to which graded actions are predictably represented along the levels of the prosodic hierarchy. A greater degree of respiratory system involvement is expected at speech boundaries (e.g. utterance and phrase) which are higher in the prosodic hierarchy. Speech tasks will systematically vary the types of sound segments bracketing the pause and sentence structure. Characterization of the planning and production mechanisms involved in pausing will advance theoretical models of the complex workings of motor and cognitive aspects of speech production, extend interpretation of acoustic measures based on pauses, and support advancement of clinical methods that use modification of breathing patterns to remediate speech and voice disorders. Acoustic measures of speech production provide non-invasive methods for assessing the severity and progression of a disorder - only if a clear link has been established between the acoustic measure and the underlying mechanism. Acoustic measures related to pauses in the speech signal are widely used in the clinical and research setting, however, the link to the underlying mechanism for pause generation within a breath is not yet known. In this proposal, we aim to quantify the role of the respiratory system in pause generation with the eventual goal of supporting the use of breathing-related techniques in the clinic and expansion of models of normal and disordered speech production.
