Formant frequencies of some fixed-mandible vowels and a model of speech motor programming by predictive simulation

Formant frequency data are reported for four Swedish vowels, /i, u, o, α/, produced by six speakers both with the mandible unconstrained and fixed by a bite block. The measurements were made at the moment of the first glottal pulse to confine the interpretation of the results to non-auditory sensory feedback mechanisms. The findings indicate that in spite of bite block production of physiologically unnatural jaw openings, all subjects were able to produce F-patterns the majority of which fell within the ranges of variation observed for a set of normally spoken vowels. It is worthy of note that subjects required no practice although experimental procedures were designed to facilitate the study of possible learning effects. To explain the results it is hypothesized that the “instantaneous” learning of a novel, or highly unfamiliar, task such as the compensatory articulation of fixed-mandible vowels is possible neither because speakers draw extensively upon past similar experience nor because special motor mechanisms distinct from those of natural speech are invoked but primarily because normal speech motor programming is indeed “compensatory”. In other words, it operates in a context-sensitive mode to achieve listener-oriented goals and since “contexts” constitute a practically infinite class of events the programming has to be “creative”, that is it must be capable of handling conditions never experienced before. As a step towards the description of these features a model of speech production is proposed that achieves increased plasticity of motor control owing to the incorporation of a predictive “upstairs” component that is used to simulate “downstairs” processes of execution.