Disruptive behavior disorders, such as conduct disorder and oppositional defiant disorder, are common reasons for a childhood referral to mental health and educational services and represent a substantial public health cost (1). Children with disruptive behavior disorders are not only at risk of developing antisocial personality disorder in adulthood, but they are also at risk of developing a number of other psychiatric and physical health problems (2, 3). What is striking to people who come into contact with these children, in either a research or a clinical/educational setting, are the repeated poor choices that they make. Children with disruptive behavior disorders often act impulsively, with apparently little ability to compute the consequences of their poor behavioral choices. Many of these children desperately want to make better choices but become upset when they find themselves yet again in trouble that they did not see coming. Others may pursue selfish goals but with overly optimistic predictions of the likely outcome of their actions and experience frustrated aggression when things do not work out. In order to adequately formulate and target interventions that will help children with disruptive behavior disorders, we need to better understand the information processing abnormalities that characterize their decision making.

Behavioral studies have demonstrated that children with disruptive behavior disorders show impairments in standard learning and reversal learning tasks (4). A handful of neuroimaging studies have also documented atypical neural responses to reward-punishment learning in children with such disorders, for example, in the orbitofrontal cortex and caudate (e.g., 5, 6). In this issue of the Journal, White et al. (7) report the first study to examine the neural responses of youths with disruptive behavior disorders (compared with typically developing youths) during decision making and reinforcement, using model-based functional MRI. Successful decision making involves two critical components: 1) the appropriate representation of reinforcement expectancies (i.e., the expected value associated with a stimulus or action) and 2) prediction error signaling (i.e., the signal regarding the difference between the actual and the expected reward or punishment that enables updating of the reinforcement expectancies). Prediction error can be positive, when the actual value is better than expected, or negative, when the actual value is worse than expected. White et al. evaluated expected value processing during the decision-making phase and prediction error processing during the feedback phase of a probabilistic version of the passive-avoidance decision-making paradigm. They show that compared with typically developing youths, those with disruptive behavior disorders exhibited significantly reduced modulation of activity by expected value within the ventromedial prefrontal cortex when choosing objects and within the anterior insula when refusing objects during this task, indicating atypical expected value signaling. The authors also demonstrate that in the caudate, there was reduced modulation of activity by prediction error in response to reward but increased modulation of activity in response to punishment in youths with atypical behavior disorders, indicating disrupted prediction error signaling that appeared qualitatively different from that seen in typically developing youths.

These are important findings that have a bearing on both theoretical accounts of disruptive behavior disorders and translational efforts for these conditions. The data presented by White et al. suggest that a simple deficit model will not capture the decision-making impairment apparent in youths with disruptive behavior disorders. Instead, they argue that we may be seeing a fundamentally different organization of the information processing system responsible for reward-punishment learning and appropriate decision making. Therefore, this study demonstrates that youths with disruptive behavior disorders do not appropriately modulate brain responses in the face of prediction errors. The authors suggest that the atypical neural modulation by expected value and prediction error indexes increased risk for antisocial behavior in youths with these disorders. It is clear that appropriate expected value signaling is critical in making good behavioral choices and that appropriate prediction error signaling is important for an individual’s ability to generate further, more accurate expected values. White et al. argue that disrupted expected value and prediction error signaling will lead to the selection of poor behavioral choices and (given certain environmental contexts) antisocial behavior.

Of course, the study has some limitations that future research could helpfully address. This was the first study of its kind, and it examined expected value and prediction error learning cross-sectionally. The authors appear to suggest that poor prediction error learning will lead to inappropriate expected value signaling over time. However, it is also plausible that inappropriate expected value signaling will contribute to atypical patterns of prediction error over time. It would be interesting to study the development of both expected value and prediction error processing longitudinally using cross-lag models, in both typically developing youths and those with disruptive behavior disorders. Another consideration for the future concerns the type of reward used to study expected value and prediction error processing in youths with disruptive behavior disorders. A specific type of reward, money (although it was not clear whether the money was actually awarded to the participants), was used in this study. In the future, it might be helpful to map rewards that hold high subjective value to each participant and use these to study expected value and prediction error. In other words, are there differences in expected value and prediction error processing across the categories of reward, or does the subjective salience of the reward modulate the extent of expected value and prediction error processing abnormalities that we see in children with disruptive behavior disorders? It is unclear whether the subjective motivational value of the rewards in this study was equal among all participants and whether this contributed to the group differences. Finally, the authors themselves note that “currently, it is unknown what might cause such a fundamental reorganization of prediction error punishment signaling.” We know that disruptive behavior disorders are moderately to strongly heritable (8). We also know that parents of children with these disorders not only pass on a degree of genetic vulnerability to their children but can also provide a poor parenting environment (8). A wealth of research now indicates that the parenting environments of children with disruptive behavior disorders typically involve less positive and more negative reinforcement, as well as less consistent reinforcement contingencies, than the parenting environments of typically developing children. Thus, it may be that there are children who have the unfortunate “double whammy” of being genetically vulnerable to atypical expected value and prediction error processing and who experience childhood learning environments that further derail this processing. Genetically informative, longitudinal study designs would be helpful in investigating the etiology of the expected value and prediction error processing in children with disruptive behavior disorders.

In summary, White et al. have made an important contribution to the literature probing the neural correlates of specific decision-making impairments that are correlated with disruptive behavior disorders. Their study suggests that a simple deficit model of decision making is not appropriate for describing the information processing profile of youths with disruptive behavior disorders. We look forward to future work that both replicates and extends this research and particularly welcome the promise of this research in informing interventions to help children with disruptive behavior disorders. Treatment efforts for disruptive behavior disorders often involve behavior modification programs that rely on rewards and punishments. It will be of considerable interest to build on the findings of this study in order to develop a more precise understanding of the specific reinforcement processing abnormalities in children with such disorders and how these might be best modified with appropriate cognitive-behavioral manipulations.