Facial expression processing specificity was investigated by means of event-related potentials (ERPs). Stimuli consisted on happiness facial expressions (H) along with three other visual stimuli with different physical and affective characteristics: neutral facial expressions, landscapes and grey slides. The 32 subjects had to rate the stimuli with respect to two affective dimensions: valence and arousal . EEG was recorded at T7, T8, Fz and Cz, and the epoch was of 1000 ms (100 ms prior to stimulus onset). Analyses revealed that the variability of N200 and P200 components, which were not able to detect a differential activity in response to H stimuli, could be explained in terms of physical complexity and relevance of the stimuli. N300 did detect a differentiation between H and the rest of stimuli, sharing some characteristics of previous and later components. With respect to late components, P 3 0 0 seemed to reflect the trends of arousal ratings and SW results suggest that its amplitude was influenced by the stimulus recognition effort. Not a clear right or left hemisphere advantage in facial expression processing was found. These ERP results suggest that the specificity of facial expression processing is not strictly circumscribed to a particular latency or to a specific neural area, but it resides in a particular combination of discrete processes.