OBJECTIVE To explore the diagnostic application of receiver operating characteristic (ROC) curve of P300 amplitude and latency in schizophrenia. METHODS ROC curve of P300 amplitude and latency was plotted from 91 first episode schizophrenia (FES) and 141 normal controls (NC). Youden's index and distance were calculated in ROC curve to determine the optimal cutoff point of P300 amplitude and latency for schizophrenic diagnosis. Then the subjects were layered by gender and age for improved diagnostic accuracy. RESULTS The area under ROC curve for P300 amplitude for predicting schizophrenia was 0.746 (P < 0.001), while P300 latency showed poor power in predicting schizophrenia (P = 0.373). The optimal cutoff point of P300 amplitude was 5.097 µV with a sensitivity of 80.14% and a specificity of 63.74%. After gender and age stratification, Youden' s index was enhanced in all 5 subgroups (from 1.8% to 14.88%) and the sensitivities of corresponding cutoff were enhanced obviously in subgroups of males (25.01%), ≤ 25 years (22.22%) and >40 years (28.80%) patients. CONCLUSIONS P300 amplitude has certain values in schizophrenic diagnosis. Gender and age stratification may enhance the diagnostic efficiency of ROC curve in the diagnosis of frequent schizophrenia.