To investigate the mechanisms by which oxytocin improves socioaffective processing, we measured behavioral and pupillometric data during a dynamic facial emotion recognition task. In a double-blind between-subjects design, 47 men received either 24 IU intranasal oxytocin (OXT) or a placebo (PLC). Participants in the OXT group recognized all facial expressions at lower intensity levels than did participants in the PLC group. Improved performance was accompanied by increased task-related pupil dilation, indicating an increased recruitment of attentional resources. We also found increased pupil dilation during the processing of female compared with male faces. This gender-specific stimulus effect diminished in the OXT group, in which pupil size specifically increased for male faces. Results suggest that improved emotion recognition after OXT treatment might be due to an intensified processing of stimuli that usually do not recruit much attention.