9308671 ROSENBAUM Understanding how limb movements are controlled can have great theoretical and practical importance. Theoretically, it can integrate findings about motor performance from such fields as experimental psychology, human factors, and neurophysiology. Practically, it can benefit robotics, medicine, and the training of perceptual-motor skills. Despite the need for an understanding of limb-movement control, little is currently known about how such movements are chosen and performed by normal adult humans. This research project will help fill this void by extending and evaluating Rosenbaum's recent model of motor control. The aim of the model is to describe the computations underlying the selection of coordinated motion patterns among the limb segments, especially in reaching tasks. The central idea in the model is that there are stored posture representations in human memory. When a spatial target is selected as an object to be reached, each stored posture is evaluated for the contribution it can make to the task. Weights are assigned to the stored postures based on the evaluations they receive, and a single target posture is found by taking a weighted sum of the stored postures (using their weights and treating the postures as vectors in joint space). Movement occurs from the starting posture to the target posture by reducing the distance between the starting angle and target angle of each joint. The model allows for the selection of movements when infinitely many movements permit an object to be reached, it explains how joints can compensate automatically for reduced mobility of other joints, and it accounts for a large number of facts concerning practice, speed-accuracy tradeoffs, performance errors, and movement kinematics. The research will evaluate and extend the model through further computational work and behavioral experiments. The computational work will examine such issues as the use of more joints and joints other than tho se that have been modelled so far, the hierarchical versus nonhierarchical nature of posture space, the necessity for evaluating all stored postures, and obstacle avoidance. The behavioral experiments will test predictions of the model and provide data about human reaching that can inform the model's future development. ***
