There are several results that support this finding. When subjects make reaching movements with their two arms and have the endpoint of one arm perturbed to either side of the movement, the reflex response in the perturbed arm only will act to return the hand back to the trajectory. However, when the two arms are acting together in a reaching movement, controlling a single cursor that is displayed at the spatial average of the two hands, a physical

perturbation of a single limb elicits feedback responses in both limbs to adjust the cursor’s position (Diedrichsen, 2007). This demonstrates the flexibility of OFC. Because noise is signal dependent, the optimal response is to divide the required change in the control signal BMN 673 cell line between the actuators. Another example involved manipulating the visual environment in which subjects reached. During reaching movements a sensory discrepancy produced by a difference between the visual location and the proprioceptive location of the hand could be either task relevant or irrelevant. By probing the visuomotor reflex gain using perturbations, it was shown that the reflex gain was increased in task-relevant but not for task-irrelevant

environments (Franklin and Wolpert, 2008). Similarly it has been shown that target shape modulates the size of the visuomotor reflex response (Knill et al., 2011). Liu and MAPK inhibitor Todorov (2007) investigated another predicted feature of optimal control. The theory itself predicts that feedback should be modulated differently during a movement depending on the distance to the target. At the beginning of the movement, the feedback is less important because there is sufficient time to correct very for errors that might arise in the movement. However, near the end of the movement, errors are likely to cause the target to be

missed. This was investigated by having subjects make reaching movements to a target, and jumping the target lateral to the direction of movement at different times (Figure 1A). As predicted, the subjects responded more strongly when the target jump occurred close to the end of the movement (e.g., blue paths), producing both a change in the movement speed and lateral movement to the target (Figures 1B and 1C). Interestingly, in this case, subjects also failed to completely compensate for the target displacement. For target jumps occurring near the start of movement, no change occurred in the movement speed, and the movement trajectories slowly converged to the shifted target location over the rest of the movement. These results were explained by an OFC model of the task that was able to reproduce the characteristics of the human movements (Figures 1D–1G). The optimal control model has three time-varying feedback gains that act throughout the movement (Figure 1E).