A Case of Control Induced Relevance

Previous work has shown that immediate and predictable sensory events after a movement rapidly increase the execution speed of these movements. Recently, we suggested that this speedup is indicative of a mechanism that reinforces the motor programs credited with the successful sensorimotor predictions. Anecdotal evidence suggests that reinforcement from sensorimotor predictability (RSP) is modular, and is insulated, for example, from the agent’s movement goals and from other, more abstract action goals. Here, we tested the dependence of RSP on the agent's own and instructed goals. In a series of experiments, we increased or decreased the goal-relevance of the sensory information that followed directly on the participants' motor responses. In general, RSP occurred independently of the goal- or task-relevance of these sensory events. These results are consistent with the supposition that this type of reinforcement is driven by a modular, dedicated mechanism that depends primarily on the confirmation of sensorimotor predictability.