Computational modelling of reward learning and social information processing in major depression : a functional MRI and behavioural investigation

It was hypothesised that major depressive disorder (MDD) is associated with a dysfunctional reinforcement processing system, affecting both simple (“primary”) and social reinforcers.  The main objective of this thesis was to test this hypothesis.  MDD patients and matched healthy controls participated in a Pavlovian reward learning and a social inclusion task, and their neural responses were measured using functional MRI.  In addition, subjects participated in a behavioural facial recognition task.  It was additionally hypothesised that antidepressants may have an effect on reward-learning signals.  Hence, the acute effect of antidepressant (SSRI) administration on healthy controls was investigated in all three tasks.  During the Pavlovian task, neural reward-learning signals that conformed to a computational (temporal-difference) model were found to be abnormal in various brain regions in MDD patients.  In addition, acute SSRI administration reduced the reward-learning signals in controls.  These findings are consistent with existing evidence for opposing interactions between serotonin and dopamine.  Dopamine reward-learning signals are believed to be a fundamental neural substrate for reinforcement learning, hence abnormal signals implies abnormal reinforcement learning.  Consistent with the hypothesis, patients were also found to be impaired in recognising facial expressions of emotions, and acute SSRI administration was found to alter emotional information processing in controls.  Patients were found to have abnormal brain responses during the social inclusion task.  Abnormal emotional and social information processing, both related to abnormal reinforcement learning, may be a cause or effect of impaired social interactions in MDD.