Event-related potential (ERP) technique was used to study the brain mechanism of cartoon face classification processing，and investigate the differences between cartoon face classification and cartoon face recognition processing. Cartoon face classification experiment: 26 volunteers were chosen from a domestic university, aged from 18 to 23. All the volunteers were right-handed and had normal vision or corrected vision. None of them were color blind, nor had they any history of mental illness or taking psychotropic drugs. The materials used in Experiment I were 450 cartoon human faces and cartoon animal faces. Neuroscan EEG acquisition system was used in the experiment. 32-channel EEG was recorded at a sampling rate of 1000Hz using averaged mastoids reference. With invalid data rejected, 25 volunteers were accepted. The amplitude and latency of N170 (T5,T6), VPP (Cz) and LPP(F3，Fz，F4，FC3，FCz，FC4，C3，Cz，C4，P3，Pz，P4，O1，Oz and O2) components were obtained by off-line analysis. Behavioral data and ERP data were analyzed by SPSS 13.0 through multi-factors repeated measure ANOVA method. The results: The RT(reaction time) of human faces were significantly shorter than that of animal faces; RT of square faces were significantly shorter than those of round faces and triangle faces; RT of females were shorter than that of males; The amplitudes of VPP of square faces were significantly larger than those of round faces and triangle faces; VPP latency of animals were significantly longer than that of humans, and VPP latency of girls were significantly shorter than that of boys; The amplitude of N170 showed right hemisphere dominance in face classification; The average amplitude of LPP in parietal lobe and central area were larger than those in frontal lobe, post frontal area and occipital lobe. Cartoon face recognition experiment: The subjects were the same as tested in the cartoon face classification experiment. The materials used in Experiment II were 255 cartoon human face pictures and 255 cartoon cars pictures, in which the human faces pictures were the same as used in the first experiment. The subjects’ task was to tell if the picture was about a car or a face. Data recording, processing and analysis were the same as used in the previous experiment. The results: The RT of recognition task were significantly shorter than that of classification task; N170 showed right hemisphere dominance in face classification; The amplitudes of VPP and LPP of recognition task were significantly lower than those of classification task; The VPP latency of females were shorter than that of males; The average amplitude of LPP in parietal lobe and central area were larger than those in the frontal lobe, post frontal area and occipital lobe. Conclusion: Face classification processing in brain is slower and stronger than the recognition processing, which may be because the former processing needs more psychological resources; People are better at processing human faces than animal faces; Parietal lobe and central area are probably the principal brain regions for face classification processing; Females have an advantage over males in intensity and speed in face processing; Structural coding and feature coding may be parallel during face processing. In the averaged mastoids reference condition, VPP is a good indicator to reflect face structural encoding, and is better than N170; LPP is an important indicator to reflect face classification processing.