Reinforcement Based Learning and Use-Dependent Processes Function Independently

• Nicholas Muscara, Biomedical Engineering, University of Delaware

• Joshua G.A. Cashaback, Department of Biomedical Engineering, University of Delaware

Throwing a curveball requires learning from repetitions and past success. Human motor learning consists of multiple mechanisms, including use-dependent learning and reinforcement learning. Use-dependent learning is driven by Hebbian plasticity, which leads to the replication of past movements, while reinforcement-based learning is driven by past rewards. Diedrichsen et al. 2010, suggest that use-dependent learning might be reward-weighted. However, it remains unclear if use-dependent learning and reinforcement-based learning are independent processes or if they interact.

Subjects (N=60) held onto a robotic arm and moved an arc cursor into an arc target on a screen. Subjects underwent baseline, probes, and passive training blocks. During baseline, participants were instructed to reach toward the target’s center to capture any movement biases. Within the passive block, early and late probes measure changes in straight-ahead reach angle. The early probe measures changes after a few repetitions, representing mild use-dependent learning. In contrast, the late probe measures changes after many repetitions, representing a large amount of use-dependent learning.

Additionally, to measure the effects of rewards on straight-ahead reach angle, subjects were randomly assigned to a reward and no reward group. The reward group was told passive movements would be on the target, and they would receive reward feedback, including a pleasant noise and a brief glow. The no-reward group was told passive movements would be to an unsuccessful location and that they would receive no reward feedback. Importantly, all subjects were moved to the same location to control for the effects of use-dependent learning.

The results showed significant main effects of reinforcement (p <0.001) and repetition (p<0.001) but no significant interaction effect (p <0.407). These results suggest that use-dependent learning and reinforcement learning are independent processes that, in parallel, affect sensory-motor adaptation.