Transdiagnostic symptom clusters show separable dimensional effects on reinforcement learning in children and adolescents

Past reinforcement learning (RL) studies showed effects of psychopathology on learning computational parameters in relation to valence and uncertainty. However, gaps remain in understanding the developmental trajectories of RL, their generalizability across different contexts, and the nuanced impact of individual symptom severity that is often overlooked in case-control designs. In this study, we adopted a dimensional symptom approach to investigate RL in a sample of 122 clinical and typically developing individuals aged 8 to 18 years. By deconstructing symptoms of depression, anxiety, and ADHD into five transdiagnostic symptom clusters – anhedonia, negative affect, fear, inattention, and hyperactivity/impulsivity – we identified distinct effects on RL. Our findings revealed inverted-u-shaped relationships for anhedonia and negative affect, with anhedonia linked to positive prediction error learning rate, and negative affect to differences in learning rate between punishment and reward. Additionally, u-shaped associations were observed in fear and inattention, with fear associated with uncertainty-directed exploration and inattention with random exploration. In contrast, hyperactivity/impulsivity showed a linear relationship with learning rate in uncertain, volatile environments. These results suggest that symptom-related alterations in learning and decision-making manifest in children as young as 8 years old, with larger effects in children compared to adolescents. Furthermore, nonlinear relationships only became apparent when examining the full spectrum of symptom severity. Through this comprehensive approach, we aim to enhance the understanding of the interplay between psychopathology, development, and RL processes, ultimately informing targeted interventions and treatment strategies.