2026 ECNP Congress

PS01-1007

Computational profiling of reinforcement learning under dopaminergic modulation in stimulant use disorder and obsessive-compulsive disorder

Walker Kennedy Cobb — T.V. Lim, N.A. Fineberg, T.W. Robbins, K.D. Ersche · University of Cambridge, Department of Psychiatry, Cambridge, United Kingdom

2026 ECNP Congress

Introduction

Stimulant use disorder (SUD) and obsessive-compulsive disorder (OCD) are characterised by maladaptive behaviours that continue despite negative consequences. These behaviours are thought to reflect impaired learning from outcomes, a process mediated by dysfunctional dopaminergic signalling [1]. Yet whether this dopaminergic dysfunction represents a shared computational impairment across disorders, or instead manifests as disorder-specific impairments in learning, remains unclear. Evidence points towards the latter: patients with SUD tend to show impaired learning from negative outcomes [2], whereas OCD pathology is associated with disrupted translation of learned values into action [3]. Characterising these profiles is essential for identifying disorder-specific targets. To this end, we applied reinforcement learning modelling to decompose behaviour into learning and choice components and examined how dopaminergic receptor modulation affected each. We hypothesised that patients with SUD show a reduction in sensitivity to negative outcomes, whereas patients with OCD demonstrate an impaired translation of learned values into actions. Furthermore, we predicted that an acute D2/3 receptor agonism and antagonism remediates these deficits.

Methods

We investigated how dopaminergic modulation affects probabilistic learning in SUD and OCD. In a randomised, double-blind, placebo-controlled crossover design, 18 SUD and 18 OCD patients (DSM-IV-TR), and 18 matched healthy controls, completed the Probabilistic Selection Task [4] across three separate sessions. Participants received placebo, an agonist (0.5 mg pramipexole), or an antagonist (400 mg amisulpride). Computational modelling separated behaviour into four components: positive and negative learning rates (α+, α−), choice consistency (β), and tendency to repeat (κ).

Results

SUD patients showed lower negative learning rates (d = -1.14, 95% HDI = [-2.30, 0.15], PNZ = 0.935), whereas OCD patients showed no credible differences (0 ∈ 95% HDI). Both clinical groups showed reduced choice consistency relative to controls (SUD: d = -0.32, PNZ = 0.931; OCD: d = -0.44, 95% HDI = [-0.79, -0.06], PNZ = 0.981), suggesting learned values exerted less influence on choice. Pharmacological effects were group-selective: amisulpride increased reward learning rates only in SUD patients (d = 0.55, 95% HDI = [0.002, 1.179], PNZ = 0.950), with no other parameter changes. Pramipexole increased the tendency to repeat only in controls (d = 0.384, 95% HDI = [0.075, 0.684], PNZ = 0.985).

Conclusions

Broadly consistent with our hypotheses, patients with SUD showed a selective impairment in learning from negative outcomes, whereas both clinical groups showed reduced choice consistency compared with controls. Acute dopamine D2/3 receptor modulation further dissociated the two disorders: antagonism selectively enhanced reward-based learning in SUD patients, agonism increased response repetition in healthy controls, and neither manipulation influenced OCD patients. Although reduced choice consistency represents a transdiagnostic feature shared across both disorders, the pharmacological findings indicate that the underlying neurobiological drivers are distinct. In SUD, the dopaminergic system remains a potential target for modulating the cognitive processes supporting adaptive behaviour; in OCD, the absence of any modulatory effects points towards dopaminergic-independent mechanisms of the pathology. These results demonstrate that convergent behavioural phenotypes can mask fundamentally different neural substrates, with direct implications for how pharmacological interventions should be stratified across diagnostic boundaries.

References

  1. Schultz, W., Dayan, P., Montague, P.R., 1997. A neural substrate of prediction and reward. Science 275(5306), 1593-1599.
  2. Lim, T.V., Cardinal, R.N., Bullmore, E.T., Robbins, T.W., Ersche, K.D., 2021. Impaired learning from negative feedback in stimulant use disorder: dopaminergic modulation. Int. J. Neuropsychopharmacol. 24(11), 867-878.
  3. Gillan, C.M., Robbins, T.W., 2014. Goal-directed learning and obsessive-compulsive disorder. Philos. Trans. R. Soc. Lond. B Biol. Sci. 369(1655), 20130475.
  4. Frank, M.J., Seeberger, L.C., O’Reilly, R.C., 2004. By carrot or by stick: cognitive reinforcement learning in parkinsonism. Science 306(5703), 1940-1943.