Doctor of Philosophy
Paul, Gribble L.
Human motor control is highly adaptive to new tasks and changing environments. Motor adaptation relies on multiple dissociable processes that function to increase attainment of reward and to reduce sensory error and physical effort as costs. This thesis tests the hypothesis that fronto-striatal and dopaminergic neural systems contribute to specific aspects of motor adaptation that occur through reinforcement of rewarding actions.
Behavioral tasks were designed to isolate learning in response to feedback conveying information about reward, error, and physical effort. We also measured behavioral effects of savings and anterograde interference, by which memories from previous motor learning can facilitate or impair subsequent learning. Electroencephalography (EEG) was used to record neural event-related potentials (ERPs) elicited by task-related feedback. We measured the feedback-related negativity/ reward positivity (FRN/RP), a midfrontal component of ERP responses to feedback stimuli that correlates with neural activity throughout fronto-striatal circuits. Levodopa, a dopamine precursor, was used to manipulate dopamine release in healthy volunteers, as it has been shown to impair reward-based learning in various cognitive tasks.
We first determined that medial frontal feedback processing indexed by the FRN/RP is a specific neural correlate of reward prediction error during motor adaptation, and that the FRN/RP is not elicited by sensory error. Next, we found that levodopa did not affect either the FRN/RP, reward-based motor adaptation, savings, or anterograde interference. Finally, we determined that medial frontal activity indexed by the FRN/RP does not respond to physical effort as a cost that discounts the value of reward. However, effort increased neural sensitivity to reinforcement outcomes in activity measured by a midfrontal ERP component that was spatially and temporally distinct from the FRN/RP.
These findings suggest that mid-frontal feedback processing measured by the FRN/RP may play a specific role in reward-based motor learning that is distinct from error- and effort-based learning processes. Our findings also indicate that reinforcement learning mechanisms that contribute to motor adaptation do not depend on the same dopaminergic processes that are impaired by levodopa in cognitive learning tasks.
Summary for Lay Audience
People have a remarkable ability to adapt their movements to changing conditions. Most research about motor adaptation has studied how people adjust movements according to spatial errors. For example, when the wind begins to push a tennis player’s serve in an unexpected direction, the player sees the placement of each serve and adapts to compensate for the errors they experience visually. Movements can also be adapted and refined through reinforcement learning. In reinforcement-based motor learning, variable movements are produced during repeated practice, and we learn to repeat the movements that result in successful or rewarding outcomes. For example, a tennis player might learn to adjust their stance in a way that provides more power and wins more points. People also adapt their movements to be more efficient and require less effort. One hypothesis is that the brain treats effort as a cost that discounts the value of rewards.
When decisions result in rewarding outcomes, a chemical called dopamine is released in the brain. Dopamine is thought to cause changes in reward-processing brain areas that reinforce successful decisions. We tested whether similar mechanisms also contribute to reinforcement-based motor learning. We performed experiments in which people learned to adapt reaching arm movements based on rewards and errors. Some people took a drug called levodopa that impairs reinforcement-based learning in cognitive decision making tasks by overstimulating dopamine release. We placed electrodes on participants’ heads to measure electrical activity from the brain in response to reward and error during learning. We also performed an experiment to test whether physical effort affects brain responses to reward as a cost.
We identified electrical signals from frontal areas in the brain that may reflect a specific mechanism for reinforcement-based motor learning. This brain mechanism was sensitive only to reward or success, while signals from different brain areas reflected effort and error. Levodopa did not affect these specific brain responses to reward, nor did it affect motor learning across a variety of tasks. These results indicate that motor learning does not depend on the same dopamine learning mechanism that is impaired by levodopa in cognitive decision-making tasks.
Palidis, Dimitrios J., "Neural substrates of reward, error, and effort processing underlying adaptive motor behaviour" (2021). Electronic Thesis and Dissertation Repository. 7887.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License