Different reinforcement learning biases in Parkinson's disease, Narcolepsy, and healthy subjects

ContributorsHinnen, Gabriel
Number of pages65
Master program titleMaîtrise universitaire interdisciplinaire en neurosciences
Defense date2023

Learning through rewards and punishments is an essential part of our everyday life, as it helps us attribute values to things, and shapes our behaviour (approach rewards or avoid punishments). However, reinforcement learning processes can be altered in certain neurological conditions, such as Narcolepsy and Parkinson’s disease.

In this project, 12 narcoleptic patients (NC), 15 Parkinson’s disease patients (PD) and 14 healthy controls (HC) were compared for a reinforcement learning task in an fMRI protocol. We also fitted a computational model to each participant’s learning to extract prediction error (PE) values at each trial as well as learning rates.

Behaviourally, no differences were observed between the three populations. Similarly, a whole-brain analysis revealed no between-groups differences, highlighting a higher activity in the bilateral Nucleus Accumbens (NAcc) for positive feedbacks and in the bilateral insular cortex for negative ones, for all subjects. However, learning rates showed that HC subjects were more sensitive to positive PEs, whereas PD patients were more sensitive to negative PEs, during their learning process. At the cerebral level, activity in the bilateral VTA, NAcc and dACC, as well as the bilateral amygdala and anterior insula, tends to underly these biases, which we propose to interpret in terms of different learning profiles. Regarding NC patients, while no specific bias was shown behaviourally, activity in the caudate nucleus head suggested a tendency towards a negative-PE bias.

These results indicate that NC and PD patients display behavioural and/or cerebral reinforcement learning alterations, materialized by different PE sensitivities. Different learning profiles might therefore be engaged, influencing these patients’ everyday functioning and well-being.

Citation (ISO format)
HINNEN, Gabriel. Different reinforcement learning biases in Parkinson’s disease, Narcolepsy, and healthy subjects. 2023.
Main files (1)
Master thesis
  • PID : unige:172129

Technical informations

Creation10/05/2023 11:31:03 AM
First validation10/10/2023 9:46:30 AM
Update time10/10/2023 9:46:30 AM
Status update10/10/2023 9:46:30 AM
Last indexation02/01/2024 10:49:06 AM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack