In the literature, it is established that people are better at recognizing faces from their own race relative to faces from other races. This phenomenon is called other-race effect. There have been studies to try to explain this phenomenon. The results indicated that same race faces involved more configural/holistic processing than other race faces. But the theoretical explanation of this phenomenon is still controversial. Especially there were few studies in this field in our country. This study aimed to investigate the neural mechanism underlying the other-race effect, in particular at what stage of face processing differences between same-race and other-race stimuli occur. Event-related potentials (ERP) technique and a race categorization task were used to explore the neural mechanism of face processing. The participants in the research were undergraduate students. Each participant took part in test individually. The materials consisted photographs of Caucasian, African, Asian faces (10 per race and half women). The faces were presented in a frontal view with a neutral expression without distinctive features. The participants were asked to identify as fast and correctly as possible whether face stimuli belonged to their own race or not by pressing with one of two computer keys, during which ERPs were recorded. At the start of each trial, a fixation point appeared at the center of the screen for 300 ms, followed by a face stimulus lasting 900 ms. The offset of each stimulus was followed by a blank-screen during a random inter-stimulus interval (ISI) ranging from 1000 to 1600 ms (mean ISI: 1300 ms). Each of the three ethnic groups was presented in upright and inverted orientations. It was repeated 4 times in one orientation. Thus, a total of 480 trials were presented in 6 experimental blocks with self-timed breaks in-between. Each block consisted of 80 trials. They were presented in a pseudo-random order to avoid immediate repetition of the same stimulus. Before a formal experiment, there was a practice. The participants can enter a formal experiment until accuracy rate was reached 80%. The analyses were based on measurements of behavioral data (reaction, accuracy rate) and electrophysiological data (amplitudes of peak activity). The results were analyzed by repeated measures ANOVA. Greenhouse – Geisser corrections were applied with adjusted degrees of freedom. The results show that (1) for the upright orientation, the peak N170 amplitude was lower for Asian than Caucasian (p<.05) faces, and the former lower than African (p<.05) faces. Significant difference was not found between Caucasian and African faces (p>.05). And inversion effects were significant for all racial groups. It suggested that the peak N170 amplitude was lower for the upright orientation than the inverted orientation. Thus, the inverted faces involved more featural processing than the upright faces. What’s more, we indicated that same-race faces involved more configural/holistic processing, and other-race faces contained more featural processing. More importantly, we found that there was no significant difference between Caucasian and African faces, which can prove our hypothesis to some extent. Previous literatures always presented the differences between the amplitude of same-race and other-race faces. However, this study not only presented the differences between the amplitude of same-race and other-race faces, but also presented the differences between the amplitude of other-race faces. (2) There was no significant interaction between race and orientation. This result was different from some literatures. We considered that Chinese subjects were better at configural/holistic processing. Regardless of races and orientations, configural/holistic processing was involved more or less. Thus there was no significant interaction. This study demonstrates that same-race faces involve more configural/holistic processing than other-race faces, and there is a right hemisphere superiority of N170. Also, we find that the results are differences between Chinese subjects and Caucasian subjects.