A model for reaching control

In this paper we propose that reaches are made to target postures which are selected by evaluating stored postures. Target postures are chosen by taking a weighted average of the stored postures, where the weights assigned to the stored postures depend on their effectiveness for the task. Movements from starting postures to target postures are achieved by reducing the distance, in joint space, between the two. The form of the movement depends on drive (assumed to decrease as less distance remains) and inertia. The model predicts the Power Law of learning, compensation for immobility of joints, changes in limb contributions depending on movement speed, asymmetric bell-shaped velocity profiles, velocity-amplitude relations, Fitts' Law, and position-dependent variations in hand-path curvature. Planned extensions of the model may broaden its application - for example, to handwriting.