1Sobell Department of Motor Neuroscience & Movement Disorders, UCL Institute of Neurology, London,
OBJECTIVES: A growing body of evidence suggests that dopamine plays a key role in reinforcement learning. Animal studies have shown that phasic bursts of dopamine neurons are observed after reward, whereas the firing of dopamine neurons drops below baseline after choices that do not result in reward. Therefore, disruption of the midbrain dopamine system as in Parkinson’s disease (PD) may lead to deficits on tasks that require learning from positive and negative feedback. Previous studies have shown that in a probabilistic selection task and a transitive inference task, learning from positive or negative feedback is differentially affected by PD and the dopaminergic medication used to treat it. However, previous studies have not directly examined these effects during learning, but rather have shown deficits after learning during transfer. The aim of the present study was to examine stochastic sequence learning in PD patients.
METHODS: A group of medicated PD patients and age-matched healthy controls were assessed. Participants were required to learn sequences of four button presses, where each button press was performed either with the left or right index fingers. We used six sequences in the task, each of which had two left and two right button presses. After each button press, participants received feedback whether or not it had been correct. Importantly, the feedback was noisy (85% correct feedback). Thus, on 15% of the trials, even if the participants had pressed the correct buttons, the feedback informed them that they were incorrect. Their task was to use this ’noisy’ feedback to learn the correct sequence of button presses and then repeat it eight times. After participants had executed the sequence correctly eight times, a new sequence was introduced and the participants once again had to learn the new sequence by trial and error. We used a randomized block design with eight blocks in total .
RESULTS: Both medicated PD patients and age-matched controls were biased to preferentially learn from positive feedback relative to negative feedback. Although the medicated PD group required more trials to learn the sequences they showed an enhanced learning effect from positive feedback relative to the control group.
CONCLUSION: Our finding that medicated patients with PD showed increased sensitivity to positive feedback in stochastic sequence learning, may reflect a dopamine ’overdosing’ effect in the fronto-striatal circuits which are less affected in PD. Future assessment of stochastic learning in PD patients in the ”off” medication state would be of interest.