Toward a perception based model of the production of prosody

Abstract

The speech production model proposed by Guenther, Perkell, and co-worker’s [Guenther et al., Psych. Rev. 15 (1998); Perkell et al., J. Phon. 28 (2000)] is generalized from the predominantly segmental perspective to a new theory of the production of prosody. It is posited that the only invariant targets of the production of prosodic events are perceptual targets, characterized as multidimensional regions in the perceptual space. Prosodic gestures are planned as trajectories that reach and traverse the perceptual target regions. Successfully executed gestures produce acoustic realizations of perceptually relevant prosodic events. The prosodic interpretation of the speech production model is structured around a hierarchy of prosodic domains: discourse structure, information structure, and accentual structure. It is assumed that the linguistically relevant and phonologically distinctive functions of prosodic features are represented by internal models. Once acquired, the phonemic settings pertinent to these models are stable and resistant to change, and they do not rely on continuous auditory feedback. We further suggest that the mechanisms involved in prosody control, may not differ categorically from those that control segmental speech production.