Kalveram, K.Th. (1995) A neural-network modeling sensorimotor learning: Implications for speech motor control and stuttering. In C.W. Starkweather& H.F.M. Peters (Hrsg.), Proceedings of the First World Congress on Fluency Disorders, 140-146.
Abstract
In speech motor control, the sensorimotor tool transformation is defined transforming the force pattern of the articulatory muscles into speech sounds. This transformation, also called vocal tract transformation, is - similar to the model ot the two jointed arm - partitioned into two parts, namely the transformation relating the muscle forces to the mechano-spatial states of the vocal tract (which is analogous to the arm's forward dynamics and includes also 'natural' interarticulatory couplings), and the transformation relating the mechano-spatial states to the speech sounds (which is analogous to the arm's forward kinematics). Low level speech motor control then requires to invert both transformations. Assuming reflex-like processing as the principle of control, the inversion of the force to mechano-spatial state transformation can be performed using the self-imitation algorithm. Due to erraneous learning of this inversion, the controller can fail to decouple the natural inter-articulatory coupling. This causes abnormal feedback loops through the reflex-like operating neural network, which in turn can cause stuttering if audio-phonatoric coupling is involved in learning.
Back