Abstract
Exploration is important when attempting to relearn motor skills following neurological disorders. Converging neuroanatomical evidence shows bidirectional connections between reinforcement-based (basal ganglia) and error-based (cerebellum) neural circuitries. However, it is unknown whether these bidirectional neural connections would influence exploratory behaviour. Here we designed two experiments and a computational model to investigate the unique and interacting roles of reinforcement and error feedback on motor exploration. Participants grasped the handle of a robotic manipulandum. They made reaching movements to a large target that promoted exploratory behaviour, without vision of their hand. Participants received either reinforcement feedback (pleasant sound, monetary gain for a success) and/or error feedback (small cursor showing hand position) at the end of their reach. We computed trial-by-trial statistical random walks (lag-1 autocorrelations) to quantify exploration. Aligned with model predictions, in Experiment 1 we found that neurologically intact individuals displayed significantly greater exploration with reinforcement feedback compared to error feedback (p < 0.001). Participants displayed moderate levels of exploration when receiving both forms of feedback, which was greater than isolated error feedback (p = 0.035) and less than isolated reinforcement feedback (p < 0.001). In Experiment 2, we considered those with Parkinson’s disease, who have compromised reinforcement-based neural circuits. Individuals with Parkinson’s had less exploration with reinforcement feedback compared to healthy age-matched controls (p < 0.001). Taken together, our results and model suggest that reinforcement-based and error-based processes respectively boost and suppress exploration, while in concert these processes oppose one another to result in moderate exploratory behaviour.