Developmental differences in reward-learning and its connection to resting-state functional connectivity modeled using a hierarchical Bayesian model

Adolescence is a period of heightened sensation-seeking, risk-taking, and reward sensitivity, characterized by structural and functional changes in the brain. Developmental changes in functional connectivity between cortical and subcortical regions may refine communication within reward-related circuitry, influencing learning and decision-making. Here, we compared reinforcement learning behavior and its relationship to resting-state functional connectivity in reward-related circuits in adolescents and adults. Fifty-eight healthy participants (32 adolescents aged 13-16; 26 adults aged 30-40) completed a probabilistic two-armed bandit task and resting-state functional magnetic resonance imaging (fMRI). The learning-related parameters learning rate (α) and inverse temperature (β, an index of the randomness of choices) and their relationship to functional connectivity were modeled from behavioral data using Q learning in a hierarchical Bayesian framework. In the whole sample, learning rate was associated with functional connectivity in several cortico-subcortical pathways, particularly involving the anterior cingulate cortex. Adolescents exhibited lower learning rate and inverse temperature values than adults and had a stronger association between learning rate and fronto-striatal connectivity. Adolescents also showed less tendency to stay with winning options in the task, defined as the proportion of trials where participants repeated the previous choice after a reward. These findings highlight the involvement of the ACC in reward learning and indicate that behavior in a reinforcement learning context is characterized by reduced feedback-driven learning and more variable choice behavior or greater exploration in adolescents compared to adults, and suggest that adolescents rely more on fronto-striatal connectivity during learning.