In the current study, we examined the role of task-related top-down mechanisms in the recognition of facial expressions. An expression of increasing intensity was displayed at a frequency of 1.5 Hz among the neutral faces of the same model that was displayed at a frequency of 12 Hz (i.e., 12 frames per second, with the expression occurring every eight frames). Twenty-two participants were asked either to recognize the emotion at the expression-specific frequency (1.5 Hz) or to perform an orthogonal task in separate blocks, while a scalp electroencephalogram (EEG) was recorded. A significant 1.5 Hz response emerged with the increase in expressive intensity over the medial occipital, right and left occipitotemporal, and centro-frontal regions. In these three regions, the magnitude of this response was greater when participants were involved in expression recognition, especially when the intensity of expression was low and ambiguous. Time-domain analysis revealed that engagement in the explicit recognition of facial expression caused a modulation of the response even before the onset of the expression over centro-frontal regions. The response was then amplified over the medial occipital and right and left occipitotemporal regions. Overall, the procedure developed in the present study allowed us to document different stages of the voluntary recognition of facial expressions, from detection to recognition, through the implementation of task-related top-down mechanisms that modulated the incoming information flow.