Change of desired trajectory caused by training in a novel motor task

When the human motor system adapts to novel dynamics of the arm during reaching, hand trajectories tends to converge toward a roughly straight line. This straight line is thought to be the desired trajectory of the system. Trial-to-trial changes in performance are well described by a first order state-space model: errors in a given trial affect performance on the subsequent trial as a function of the distance in state space between the two trials. This function describes the generalization patterns that govern adaptation. Whereas the desired trajectory and the generalization function have been quantified for short-term adaptation, little is known about their behavior with long-term training. We report that when subjects are trained to reach in novel force fields over multiple days, the state-space model suggests that the desired trajectory undergoes systematic changes. In a constant field, the desired trajectory becomes curved. The direction of change of the desired trajectory is affected by the training protocol, such that occasional unperturbed trials (catch trials) caused subject to increasingly under-compensate for the perturbations whereas a lack of such trials caused subjects to increasingly overcompensate for the perturbations. We suggest that the desired trajectory is not constant, but is a result of an optimization that considers the success rate. In this optimization process, subjects weigh more heavily the importance of certain trials: either those that are infrequent or those that cause large errors.