Objective: 1.Characterize deficits in reinforcement learning processes in PD during a strategic game analogous to Rock-Paper-Scissors (RPS). 2.Examine the effects of dopaminergic (DA) treatment on learning rates during RPS. 3.Investigate whether dysfunction develops at the same rate across multiple motor systems.

Background: During RPS, each player’s actions and associated outcomes change dynamically based on their opponent’s actions. Optimizing strategies requires choosing among several actions, the likelihood of which is adjusted dynamically based on reinforcement information, a process involving corticostriatal networks. The dorsal striatum is thought to compute a reward prediction error¹, and patients with PD show decreases in prediction error signaling¹. DA medication alters learning rates in reinforcement learning tasks², potentially contributing to impulse control disorders.

Methods: PD patients (stage 1-3) and age-matched controls competed in a game of RPS against a computer opponent that exploited biases in choice patterns. Participants maximized reward by minimizing predictabilities in choice sequences (i.e., choosing stochastically). Choices were indicated with either a saccade or a button press. Both groups completed 2 sessions; patients both on- and off- medication. Reinforcement learning processes and predictabilities in choice sequences were examined using logistic regression modelling, and behavior was correlated with neurocognitive scores including the Baratt Impulsiveness Scale and MoCA.

Results: Patients were impaired in choosing optimally during the RPS, particularly in the saccade condition. Patients were more variable, exhibited more predictabilities in choice patterns, and had lower reward rates compared with controls in the saccade condition. These results were exacerbated by DA medication. Patients performed better during the button-press than saccade trials in most aspects of RPS, suggesting greater dysfunction within the oculomotor loop through the basal ganglia.

Conclusions: We propose a new tool to investigate reinforcement learning in PD, which could lead to novel insights into optimizing treatment protocols and maximizing cognitive function. Further investigation into individual differences in the PD group could provide insight into the traits that predispose certain patients to impulse control disorders.

Preliminary data presented at the 2016 MDS congress, June, 2016

References:

1. Schonberg, T., O’Doherty, J.P., Joel, D., Inzelberg, R., Segev, Y., & Daw, N.D. (2010). Selective impairment of prediction error signaling in the human dorsolateral but not ventral striatum in Parkinson’s disease patients: evidence from a model-based fMRI study. NeuroImage, 49, 772-781.
2. Rutledge, R.B., Lazzaro, S.C., Lau, B., Myers, C.E., Gluck, M.A., & Glimcher, P.W. (2009). Dopaminergic drugs modulate learning rates and perseveration in parkinson’s patients in a dynamic foraging task. The Journal of Neuroscience, 29, 15104-15114.

« Back to 2017 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/quantifying-reinforcement-learning-deficits-in-early-stage-parkinsons-patients-using-a-strategic-decision-making-task/