Facial expressions are fundamental emotional stimuli as they convey important information in social interaction. Most empirical research on facial expression processing has focused on isolated faces. But in everyday life, faces are embedded in surrounding context. For example, fearful faces always accompany with tight bodies, and happy faces appear in birthday parties more often than in sickrooms. Scenes which faces are embedded in provide typical visual context. Recently some studies attempted to investigate the influence of emotional scenes on facial expression processing. Although a few previous studies in this field demonstrated the scene effects of facial expression processing existed, the studies did not further explore the specific processing mechanism of the scene effects. Because of its excellent temporal resolution, the present study used event-related potentials (ERPs) to investigate the effects of scenes that contain different spatial frequencies on facial expression processing. Our hypothesis was that the different spatial frequencies of scenes affected facial expression processing in different ways.

Eighteen right-handed college students (11 females; age range 17~24 years; mean age_20.67±1.91 years) were paid to participate in the experiment. Thirty-two face pictures (16 females and 16 males) with fearful and neutral expressions and thirty-two scene pictures (16 negative scenes and 16 neutral scenes) were presented. Spatial frequency content in the original scene stimuli (broad-band, BSF) was filtered using a high-pass cut-off that was > 16cpi for the higher spatial frequencycpi for the lower spatial frequency (LSF) scene stimuli. In the present study, we used a paradigm in which the scene and the facial expression were presented together, i.e., simultaneous processing of the scene and the facial expression. In simultaneous processing of the scene and the facial expression, the early mental representation of the scene has to be constructed in parallel to the construction of the facial expression. In the paradigm, after 500 ms into the presentation of the fixation, the face-scene compound stimuli appeared centrally on the scene for 800 ms. The scene and face emotions were either congruent or incongruent. Participants were instructed to perform a gender categorization task (task-irrelevant). (HSF) scene stimuli, and a low-pass cut-off of < 8

Because the task-irrelevant task was used, our behavioral data was only used to evaluate the degree of attention. Accuracy of target stimuli did not show significant differences between conditions (above 90% for all conditions, ps > 0.1). Our ERP results showed that for scenes with broad-band spatial frequency, fearful faces which appeared in neutral scenes elicited larger N170 amplitudes than these faces which appeared in negative scenes in both right and left hemispheres. But the effects were not found for scenes with high and low spatial frequencies. In addition, neutral faces which appeared in neutral scenes elicited larger N170 amplitudes than the faces which appeared in negative scenes. The effects that were found for scenes with different spatial frequencies were distributed in different hemispheres. For scenes with broad-band and low spatial frequencies, the effects were distributed in the right hemisphere. But for scenes with high spatial frequency, the effects were distributed in the left hemisphere.