What are the odds? A Functional Imaging Study on Loss Anticipation, Positive Punishment and Negative Reinforcement in Problem Gamblers and Healthy Controls
Objective: With this study, we aim to arrive at a better understanding of the neural mechanisms underlying the anticipation of losses due to unpleasant electric shock stimuli. According to prospect theory – currently the most important descriptive model of decision making under uncertainty in economics – subjects overweight small probabilities and underweight high probabilities. Here, we analyzed to what extent prospect theory is manifested in biological data. Does the human brain process aversive contingencies of decisions in a nonlinear manner according to the probability weighting function underlying prospect theory? Which brain regions process certainty or uncertainty regarding an outcome regardless of its valence? How do individuals with problematic gambling habits differ in these entities? Methods: The present study included a total of 83 participants (22 problem gamblers, 18 habitual gamblers, 23 age-matched subjects and 20 young control subjects). Each participant was evaluated by psychologists through a semi-structured interview and the self-reported Southern Oaks Gambling Scale. The study participants engaged in a card game in which unambiguous situations with various risk levels for aversive events (unpleasant but not painful electric shocks) were created. Apart from neural responses during anticipation of an aversive event, our paradigm allowed us to measure how neural activity during expectation is modulated by the probability of the outcome. Furthermore, we used the participants’ attitudes towards risk to account for the subjective evaluation of probabilities, as suggested by prospect theory. Results: We confirmed existing knowledge about the active network during anticipation of aversive events in a large sample. Specifically, in response to our experimental stimuli, the BOLD (blood oxygen level dependent) signal increased in the anterior and medial cingulate cortex, the anterior insula, the amygdala, the thalamus and the medial frontal regions. Based on economic studies and on our previous behavioural results, we modelled the expected outcomes according to uncertainty and prospect theory. With increasing chances of avoiding punishment, we found that the BOLD signal increased in the orbitofrontal cortex, the striatum, the hippocampus, the precuneus, and the posterior cingulate cortex. Conversely, with increasing certainty of a positive or negative outcome (U-shaped from most certain loss to no loss), we found activation patterns in the medial superior frontal cortex, the anterior cingulate cortex, the anterior insula, the frontopolar prefrontal cortex and the angular gyrus. Furthermore, we found evidence for group differences in cerebral activation in the bilateral superior temporal pole, the anterior insula and the medial cingulate cortex during the perception of distinct probability levels according to the nonlinear probability weighting function. Interestingly, these regions are relevant both in loss anticipation paradigms and in functional imaging studies describing pathological gambling. Conclusion: Our results support the concept that the neural response to anticipation is determined by its valence. Although it has frequently been reported that gamblers and non-gambling individuals show differences in reward anticipation, during loss anticipation gambling participants do not differ in their neural activation patterns. We also tested whether specific regions of the brain follow a nonlinear brain response according to prospect theory and the certainty of an outcome. When their BOLD responses during probability perception were compared, non-gambling and gambling individuals did show differences in the perception of probabilities. This provides new evidence that simultaneously occurring cognitive processes are represented in different sets of brain regions. Specifically, neurobiological correlates of loss anticipation were found in areas known to be involved in dopaminergic reinforcement learning and in the corticolimbic circuit, while regions known to subserve decision making, conflict monitoring and punishment showed correlation with the certainty of the outcome.