The investigation of reward-based learning in obsessive-compulsive sub-clinical checkers.

People with Obsessive Compulsive Disorder (OCD) have deficits in decision making under ambiguity, also known as probabilistic reward-based learning (RBL). In these tasks participants are not aware about the target probabilities of each option and need to use trial-by-trial feedback to learn these probabilities during the task in order to improve their performance. An open question from the previous literature is why OCD participants present this specific deficit. The review of the limitations of previous RBL studies with OCD patients and the use of a new task paradigm allowed us to explore new directions in this research field. Firstly, previous studies had the following limitations: sometimes they did not control for the use of medication, for the presence of common comorbidities associated with OCD, or for differences between distinct OCD subtypes. In addition, previous tasks usually manipulated several factors at the same time, such as feedback direction (positive vs. negative feedback) and feedback magnitude (e.g. less vs. more positive feedback), for instance. Therefore, it is not possible to know which factor affected the RBL performance in OCD patients. Furthermore, previous studies did not consider the effect of symptom-related feedback on RBL. Secondly, task performance was only measured with accuracy, even though RBL involves the use of previous trials to make predictions for the current decision. Finally, while it is reported that participants do use distinct previous trials to perform these tasks, sequential effects were not previously investigated in OCD patients, neither for random-sequences nor for patterns sequences. Based on this, the presented PhD thesis addressed some of these gaps in the literature. The investigation was restricted to subclinical checkers which is the most common OCD subtype. A binary decision making task was designed that allowed us to separately manipulate feedback direction and feedback magnitude. In addition, the study compared probabilistic RBL performances between checkers and non-checkers for random and pattern sequences. New analyses techniques were employed to investigate the use of previous trial information for the current decision making, e.g. win-shift, lose-shift, and cross-correlations. Studies 2 and 3 examined the effect of feedback direction (positive vs. negative feedback) on decision making in subclinical checkers. Results showed that subclinical checkers were more biased towards exploitation when using negative feedback. Studies 4 and 5 examined the effect of the feedback magnitude for both the positive and the negative feedback direction. Results showed that subclinical checkers were able to change their bias towards exploitation within the positive direction experiments depending on the feedback magnitude of the experiment. They were more biased towards exploitation in the positive task presenting a higher error magnitude. In addition, checkers were always biased towards exploitation within the negative direction experiments, irrespectively of the feedback magnitude. Study 6 examined the effect of the manipulation of feedback magnitude when a symptom related feedback was added as an increment of the negative feedback. Results showed that subclinical checkers were not affected by the presence of the symptom-related feedback and, in terms of feedback magnitude, they continued to present a bias towards exploitation. The findings show that checkers were able to adapt towards exploitation in the positive experiments when higher feedback magnitudes are given. This strategic shift might be related to the higher error magnitude associated with the absence of reward, when the incorrect option was chosen. In contrast, checkers were not able to adapt their exploitation behaviour within the negative experiments when enhancing the feedback magnitude. In this way, this indicates that subclinical checkers present a deficit that bias their responses towards exploitation, when the magnitude associated with the error in the task surpasses a certain value of negative magnitude. This bias could reflect deficits regarding negative prediction errors in OCD or a hyperactivation of brain areas related with exploitation. Both explanations could be linked with a hyperactive dopaminergic system in OCD, so these results could encourage new research about the role of dopamine and prediction errors deficits associated with OCD. Additionally, heightened emotions and reward magnitudes might reduce treatment success because of the enhanced exploitation behaviour, so one crucial aspect of future therapies might be to carefully study the employment of stimuli with such higher error magnitudes