Value representation in youth psychopathology: evidence of a transdiagnostic risk mechanism for psychosis.

Negative symptoms of schizophrenia are associated with deficits in representing the value of actions, observed in reinforcement learning (RL) tasks as impaired learning from gains but intact learning from losses. This RL profile contrasts with depression, where enhanced loss sensitivity is common. Whether a schizophrenia-like RL pattern characterizes youth at clinical high-risk (CHR) for psychosis - who show modest psychosis transition rates but high rates of co-occurring depression - remains unclear. We tested whether CHR youth show a schizophrenia-like RL profile and whether RL parameters relate differentially to negative vs. depressive symptoms by estimating RL in CHR (n = 292), clinical controls (CC; n = 241) with other psychopathologies, and healthy controls (n = 175). Participants completed symptom interviews, neuropsychological assessments, a dimensional psychosis risk calculator, and an RL task in which participants could seek gains or avoid losses. A computational gain-loss Q-learning model decomposed task behavior into components indexing value updating and value-guided choice. Although overall RL performance was similar across groups, in both CHR and CC youth, higher negative, but not depressive symptoms, were selectively associated with reduced win-stay behavior following gains. Computational parameters suggested this behavioral pattern reflected disrupted updating and expression of positive value representations. Lower win-stay rates were also linked to lower premorbid intelligence and higher psychosis risk calculator scores. Here, a schizophrenia-like RL profile was unrelated to depression but associated with multiple indicators of psychosis risk across clinical groups, suggesting a transdiagnostic psychosis risk mechanism distinct from affective disturbance.