A Computational Mechanism for Forming Arm Motor Memories Differentiated by Dynamic Gaze States

Various kinds of natural reaching behaviors, such as reaching for a cup or learning a new tennis shot, require adaptive changes in response to error signals. Since reaches are usually coordinated with preceding eye movements to a reach target, learning to reach often occurs under coordinated eye-hand movements. To understand the computational mechanism of eye-hand coordination, we examined whether/how preceding eye movements contribute to the formation of motor memory for reaching. In the experiment, participants were asked to perform sequential eye and hand movements toward a visual target (i.e., eye-then-reach) and to learn two opposing visuomotor perturbations (i.e., clockwise and counterclockwise visuomotor rotations), which normally prevent learning due to memory interference. We found that two visuomotor maps for reaching could be learned simultaneously when each was uniquely associated with preceding static and dynamic gaze states. This gaze-dependent learning effect also attenuated as the pause between eye and hand movements increased. These results demonstrate that the brain encodes reaching memories in tight association with the recent history of preceding dynamic gaze states.