For most individuals, information about the self is associated with a positive valence. Previous research demonstrates such self-positive bias can be modulated by the third-person perspective-taking, which means when people evaluate oneself through the third-person perspective, they evaluate themselves less positive as through the first-person perspective: as the cognitive neuroscience studies showed the event-related potentials (ERPs) component such as P300, N400, and LPC attributed to the modulation phenomenon, while it’s still not quite clear what happened at the very beginning of the self-reflect task or whether the modulation start at the early stage of the self-evaluation processing. Since the emotional studies found the processing of negative words attract much more attention than the positive ones at the early stage of the emotional processing. We hypothesis the modulation of the self-positive bias by the third-person perspective started at the early stage when attention was paid for the emotional words which were used to describe one’s personality. In the present research, 15 participants were enrolled in a 2 (perspective: first-person perspective, third-person perspective) ×2 (emotion: positive, negative) within-group experimental design. All participants had to make a judgment whether the emotional words appeared on the computer screen is appropriate to describe themselves through their own perspective or other perspective respectively. During the processing, the electroencephalogram (EEG) was recorded using the 256-channel HydroCel Geodesic Sensor Net (Electrical Geodesics, Inc., Eugene, OR). Components, source and spectral analysis were used to explore the processing of the self-reflective task in 4 different conditions. The mean amplitude and power data were then analyzed with repeated measures ANOVAs. The components results showed that under the third-person perspective, the processing of negative words elicited larger P2 (120-200ms) amplitude than the positive ones in the left frontal electrodes, while, there’s no significant difference between the emotional words under the first-person perspective. Then we analysis the source of the P2 component and found the same results: at the left prefrontal area, the activity is significant different between processing the negative words through different perspective while no difference was found when processing the positive words through the 2 perspectives. Furthermore, in order to confirm the hypothesis, we analysis the event related spectral power (ERSP). The result is consistent with the previous: when processing negative words, significant increasing ERSP in the Theta (3-6Hz) and Gamma (30-45Hz) rhythm have been found than the positive words under third-person perspective, while the Alpha (8-13Hz) band showed significant decreasing, there’s no significant difference in 3 rhythms when processing emotional words through the first-person perspective. All the results showed that when we make the self-reflective processing or self-evaluation, we cared more or paid more attention when we take the third-person perspective than the first-person perspective, especially for the negative information. In consistent with the previous research, these findings suggested during the self-reflective processing, taking the third-person perspective can decrease the self-positive bias, or we can say, the modulation of the self-positive bias by the third-person perspective start at the early stage of the processing when attention began.