A joint-centred model accounts for movement curvature and spatial variability

Hand reaching toward a visual target needs the central nervous system (CNS) to encode target location and initial limb posture. Once these sensory inputs are encoded within a common frame of reference, the motor system builds a motor command to drive the limb towards the target. In order to face the controversy concerning the variables (in task-space versus joint-space) the motor system may manipulate to build a motor command, we propose to compare recorded hand pointing movements to a task-space vector model and to a simplified joint-centred model. Subjects repeated large 3D free movements toward 12 different targets lying on a pointing table. Corresponding endpoint confidence ellipses were then computed. The main directions of these ellipses were finally compared to those predicted by the two models. The present work differs from earlier related studies in the fundamental approach data were recorded. Indeed, we put our emphasis on isolating the motor planning component from other online sensorimotor processes and higher level cognitive processes. In particular, we did not impose cognitively controlled movement features such as movement shape or planar constraint. Furthermore, a precise control of online feedbacks allowed us dissociating motor planning from online feedback processing. The obtained data better fitted the joint-centred model than the task-space vector model. Moreover, the recorded data exhibited curved trajectories very similar to the simulated values obtained from the joint-centred model.