Decision Making and Reinforcement Learning in Children with ADHD on and Off Methylphenidate

Children with attention-deficit/hyperactivity disorder (ADHD) show impairments in decision making in tasks where choices are associated with positive and negative consequences. Reinforcement learning is crucial for decision making processes, and abnormalities therein play a key role in neurobiological models of ADHD. In a reinforcement learning paradigm children were asked to select stimuli with the largest reward probability rate (stimulus-pairs AB, CD, EF with probability rates 100/0%, 85/15%, 70/30%). We expected children with ADHD (7-13 years) to show difficulties with the acquisition of stimulus-reward mapping, with generalization of knowledge to new contexts (new stimulus-pair combinations, AC, BE etc.) and with reversal learning (e.g., 100/0%-0/100%). We expected methylphenidate to ameliorate these impairments. Overall, the ADHD group (study 1; n=116) less often than controls selected stimuli with the largest reward probability, but the learning curves did not differ between the groups. Further, the ADHD group showed lower generalization of knowledge. Study 2 (n=120) confirmed lower generalization of knowledge and showed MPH-related improvements in ADHD for all aspects of learning, except reversal learning for which performance rates did not differ from controls. Given the relative difficulty of the stimuli (6 Greek characters) and relatively low overall reward probability compared to studies that show intact stimulus-reward mapping in ADHD, impairments may be stressed when learning is more implicit / memory load is high. Findings are discussed in relation to a proposal that ADHD is characterized not by impairments in learning, but by an increase in ‘exploration rate’ (expressed in a greater variability in choice behavior).