A computational neuroimaging account of impulsive premature decision-making

Impulsivity is a multidimensional construct with distinct implications for the pathophysiology and treatment of various neuropsychiatric conditions, including substance use disorders, behavioural addictions, attention deficit hyperactivity disorder, psychosis, and personality disorders. One proposed behavioural subtype, premature responding impulsivity (PRI), appears to be influenced by neuromodulatory pathways frequently implicated in these disorders, in particular dopamine neurotransmission, known for its role in contingency learning. Still, the neurobiological basis of PRI in humans remains insufficiently understood. Here, we theorize that PRI reflects the brain’s capacity to adapt to environmental uncertainty. To test this hypothesis, twenty-four healthy adults (mean age 22.6 years; 12 females) completed a novel decision-making task featuring alternating stable and volatile probabilistic cue contingencies while undergoing functional magnetic resonance imaging (fMRI). A hierarchical Bayesian model estimated PRI as an urgency-to-respond process, whose parameters were dynamically modulated by volatility. These model-derived indices correlated with established trait impulsivity measures, supporting their construct validity. Model-based fMRI analyses identified a distributed cortico-subcortical network including anterior insula, dorsal anterior cingulate cortex, striatum, and monoaminergic midbrain regions, whose activity tracked within-trial PRI estimates as they evolved over time. Connectivity analyses further showed that high volatility enhanced interactions between subnetworks typically associated with promoting or inhibiting impulsive action. Together, these results outline a neurocomputational account in which environmental uncertainty modulates PRI through interacting brain circuits, offering a principled framework for further probing the transdiagnostic role of impulsivity across neuropsychiatric and neuropsychopharmacological contexts.