Feedback Processing and Observation of Errors It wasn ’ t me ... or was it ? How false feedback affects performance

The reinforcement learning model (Holroyd and Coles, 2002) argues that a feedback ERN is generated when the feedback is worse than expected. In the current experiment, we investigated whether the reinforcement learning system was also activated by unexpected feedback in a task in which feedback was not necessary to perform properly. A flankers task, in which false negative feedback was presented on four percent of the trials, was used. As expected, response-locked ERP data only showed an Ne/ERN after incorrect responses. Feedback-locked ERP data showed no feedback ERNs for the different feedback conditions. However, P300 amplitude increased significantly following false feedback. Performance adjustments, indexed by post-error slowing, were only present after incorrect responses. Interestingly, five out of nine participants also showed post-error slowing following false feedback. Participants were divided into two subgroups, on the basis of whether they showed post-error slowing following false feedback (PES) or not (No-PES). These subgroups revealed a difference between P300 amplitude and the presence of post-error slowing: the PES group had a larger P300 compared to the No-PES group, suggesting that the PES group attributed more meaning to false feedback. Overall, the results show that a dissociation between performance adjustments and the reinforcement learning system exists.