Patients with schizophrenia consistently fail to perform facial affect recognition tasks as accurately as healthy controls. So far, the effects of different antipsychotic drugs on this deficit, as well as the mechanisms through which these drugs may exert their influence on face processing models, remain poorly understood. Twenty-five hospitalised acute schizophrenic patients were randomised to either haloperidol or risperidone following their admission. Patients performed a facial affect discrimination task. In order to differentiate drug effects on perceptive versus semantic processing of faces, the evaluation also included two facial identity discrimination tasks (inverted and upright faces). All evaluations were performed before treatment initiation, after 2 weeks and after 4 weeks of treatment. Compared with patients receiving haloperidol, patients receiving risperidone showed better discrimination of facial affect and inverted faces, but not of upright faces. Differential drug effects were most prominent 2 weeks after treatment initiation. Performance on inverted faces accounted for only 1% of the variance in affect discrimination. Additionally, there was an inverse relationship between negative symptoms and performance on upright-face identification and affect discrimination, with the strongest and most significant correlation being for affect discrimination. Our findings suggest that the preferential effect of risperidone on affect discrimination is unlikely to be due to global effects on general face processing. Moreover, although risperidone appears to enhance the processing of individual salient facial features, this effect cannot explain the improvement in affect recognition. We conclude that risperidone may specifically act on the processing of emotion-laden information.