Deficits in emotion perception in patients with first episode of psychosis have been reported by many researchers. Till now, training programs have focused mainly in patients with schizophrenia and not in first psychotic episode (FEP) patients. We used a new intervention for facial affect recognition in a group of 35 FEP patients (26 male). The emotion recognition intervention included coloured pictures of individuals expressing six basic emotions (happiness, sadness, anger, disgust, surprise, fear) and a neutral emotion. The patients were trained to detect changes in facial features, according to the emotion displayed. A comprehensive battery of neuropsychological tests was also administered, measuring attention, memory, working memory, visuospatial ability and executive function by using specific tests of the Cambridge Neuropsychological Test Automated Battery (CANTAB). We tried to explore whether cognitive performance can explain the difference noted between the original assessment of emotion recognition and the post-intervention assessment. According to our data, overall cognitive performance did not correlate with post-intervention change in emotion recognition. Specific cognitive domains did not correlate with this change, either. According the above mentioned results, no significant correlation between neuropsychological performance and post-intervention improvement in emotion recognition was noted. This finding may suggest that interventions for emotion recognition may target specific processes that underlie emotion perception and their effect can be independent of general cognitive function.