An Investigation into the Neurobehavioral Interactions Between Sensory- and Reward-Prediction Errors During Motor Skill Learning

Abstract

In motor skill learning, two different types of prediction errors have been identified: Sensory prediction errors (SPEs) and reward prediction errors (RPEs). SPEs occur when the predicted sensory consequence of a movement does not align with the observed consequence, while RPEs occur when the result of said movement deviates from the desired goal. These errors were previously believed to be processed independently, with SPEs driving implicit sensorimotor adaptation in the cerebellum and RPEs driving explicit decision-making in the striatum. Recent research, however, has suggested otherwise, finding that SPEs influence decision-making by elevating risk tolerance, while RPEs influence implicit adaptation by increasing recalibration. What remains unclear is whether RPEs influencing SPE-driven adaptation extends to the neuroanatomical level. To explore this, we utilized the visuomotor reach adaptation task, which enables isolation of SPEs and RPEs and their subsequent effects on implicit adaptation behavior. Using fMRI, we asked if cerebellar activity associated with SPEs is modulated by RPEs when both error signals are experienced in tandem compared to just SPEs alone. From this, we replicated previous behavioral findings that RPEs are insufficient to drive implicit adaptation while SPEs can regardless of RPE presence. We also observed a numerical trend consistent with literature that the combination of SPEs and RPEs drive greater adaptation than SPEs alone. While strictly exploratory and below threshold due to a limited sample size, we observed preliminary evidence of cerebellar activity associated with RPEs when both error signals were experienced. This pattern may suggest that RPEs modulate SPE processing, which could explain the enhanced adaptation when both error signals are experienced together. These results overall support the viability of our MRI-adapted task for future prediction error imaging studies, and that the current trends may manifest under more conventional thresholds with a complete sample size.