Abstract： The processing of facial expression has been well studied. However, how facial expression is processed and memorized when it is presented outside the focus of attention is little known. Previous results of negative distractor processing were contradicted. Some studies found that processing of emotional distractor is resistant to perceptual load, whereas other studies found that emotional distractor processing is modulated by perceptual load. Interestingly, two recent studies have shown that the processing of emotional charged distractor are both automatically and depending on perceptual load. However, the temporal information of automatically and load depending processing of emotional charged distractor, is little known. Whether the individual can detect a distractor during early phase automatically, and suppress the distractor under high perceptual load at late phase of processing. To this end, we investigate how perceptual load affects facial expression distractor processing at different phase of encoding by using event-related potential methods. In ERP studies, the N170 and N250 are two early components which sensitive to the attentional resources and facial expression, while the N700 is a late component which related to saving stimulus into long-term memory when stimulus is disappear. According to previous studies and perceptual load theory, we hypothesized that facial expression distractor would process automatically in the early phase, while it modulates by perceptual load in the late phase. We expect that N170 and N250 would modulate by distractor valence but not perceptual load, while N700 would modulate by the interaction between perceptual load and distractor valence. Participants were asked to perform a letter search task which adopted from (Jenkins et al., 2005), followed by a surprise recognition task for facial expression distractor. In the letter search task, a letter string superposed on a facial distractor, participants were required to detect specific character and make response. Perceptual load (low, high) and distractor valence (neutral, fearful) were manipulated. In the surprise recognition task, participants were asked to judge whether the face was presented in the letter search task, half of the faces were presented in the letter search task, half of the faces were not presented before. In the letter search task, behavioral results showed that fearful distractor, relative to neutral distractor, interfere with visual search performance in the high perceptual load condition, while this interference effect was not significant in low perceptual load condition. ERP results showed that fearful distractor, relative to neutral distractor, elicit larger N170 and N250. The N170 and N250 were not modulated by perceptual load. In contrast, the N700 which related to post-stimuli processing and learning was modulated by the interaction between perceptual load and distractor valence. Fearful distractor elicited larger N700 than neutral distractor under low perceptual load, whereas there was no significant difference between fearful and neutral distractor under high perceptual load. The memory performance was mimic the N700 results, participant recognized fearful distractor better than neutral distractor under low load, whereas there was no significant difference under high load. These results suggest that the processing of facial expression distractor affected by perceptual load at different stage during encoding phase. In the early stage, the facial distractor is process automatically, whereas in the late stage, the post-stimuli processing is modulated by perceptual load and valence. Low perceptual load leads to less filtering of task irrelevant negative information, which induced a memory bias for negative distractor.