Reconciling static and dynamic aspects of the speech process
Björn E. F. Lindblom, Randy L. Diehl · The Journal of the Acoustical Society of America · 2001
In speech, movement and spectral change are pervasive whereas steady-states are rare. Yet descriptive frameworks (e.g., IPA) focus mainly on static properties. Similarly, in quantitative models, dynamics is generally attributed to response characteristics while the underlying control consists of static units, e.g., spatial ‘‘targets,’’ ‘‘equilibrium points,’’ or ‘‘stable auditory goals.’’ Accordingly, the input to the production of a diphthong would not be some dynamic aspect of the event, but an ordered string of two articulatory states. This approach creates a paradox when placed in the context of findings on visual and auditory perception that show that perceptual systems are more sensitive to changing stimulus arrays than to purely static ones. Clearly nervous systems have evolved to detect change. We are led to ask: If perception likes change, why assume production control in terms of static targets? We maintain that the two perspectives can be reconciled by according a significant role to positional targets in movement control and, at the same time, fully acknowledging the importance of dynamics in perception. In support of this position we present some recent computational work on simulating vowel systems by selecting optimal sets of vowels seen as objects of spectral change. [Work supported by NIH.]