Learning from the path not taken: Sensory prediction errors are sufficient for implicit adaptation of withheld movements

ABSTRACT

Current theories of motor control emphasize forward models as a critical component of the brain’s motor execution and learning networks. These internal models are thought to predict the consequences of movement before sensory feedback from these movements can reach the brain, allowing for smooth, continuous online motor performance and for the computation of prediction errors that drive implicit motor learning. Taking this framework to its logical extreme, we tested the hypothesis that movements are not necessary for the generation of predictions, the computation of prediction errors, and implicit motor adaptation. Human participants were cued to move a computer mouse to a visually displayed target and were subsequently cued to withhold those movements on a subset of trials. Visual errors displayed on both trials with and without movements to the target induced single-trial learning. Furthermore, learning on trials without movements persisted when accompanying movement trials were never paired with errors and when movement and sensory feedback trajectories were decoupled. These data provide compelling evidence supporting an internal model framework in which forward models generate sensory predictions independent of the generation of movements.