This paper explores the use of multi-channel event-related potentials (ERP) to identify mentality facticity or detect lie. Some identifiably meaning information, such as subjects' name and birthday, were selected as concealed information to be identified, 15 subjects were tested by concealed information test (CIT) paradigm and their electroencephalographs (EEG) were recorded from 30 electrodes. In virtue of analysis on the statistically significant difference between multi-channel ERPs evoked by probe information and that evoked by irrelevant information, the P300 amplitudes of 15 electrodes were selected as F-test samples. The significant difference of feature sample values between probe and irrelevant information was applied to identify mentality facticity. The results indicate that P300 amplitudes evoked in many electrodes are significantly different between probe and irrelevant information (P < 0.01). According to the significant difference of space sample values between probe and irrelevant information, the detection correctness to probe information reaches to 93.3% and is better than that of methods based single-channel ERP. The method proposed has the advantages of non-invasion and better accuracy, which could be used to identify mentality facticity effectively.