Abstract
Learning from the emotional reaction of others is crucial in our everyday lives. We assumed that additional emotional information could be especially beneficial, when a task is difficult and the limits of working memory capacity are reached. For this reason, we examined whether a potential benefit of emotional feedback during reinforcement learning is dependent on working memory load. In addition to learning performance, we analysed the neural mechanisms of reinforcement learning by examining two event-related potentials (ERPs): feedback-related negativity (FRN) and P3b. Participants were divided into two difficulty groups (with n = 21 in the difficult and n = 22 in the easy group), performing a learning task with emotional or non-emotional feedback. Task difficulty was manipulated by varying the number of stimulus-response associations. Participants' showed learning in all conditions. Emotional feedback led to increased accuracy and decreased reaction times in both groups. However, this benefit occurred earlier in the easy condition. The detection of unexpected events, as reflected in the peak-to-peak FRN, as well as working memory updating, as reflected in the P3b, were enhanced after emotional in contrast to non-emotional feedback for both groups. In contrast, task difficulty had no effect on the detection of unexpected events but led to a P3b that was more evenly distributed over the scalp, which could indicate that additional frontal resources were recruited to perform the difficult task. Our results suggest that working memory load and emotional information independently influence feedback processing without interacting.