The statistical identification of temporal region-of-interests (ROIs) of the significant difference in event-related potentials (ERPs) was popularly achieved using the cluster-based approach, in which the clustering was achieved based on the temporal adjacency of statistical significance if data from single-electrode were tested, or based on the spatial and temporal adjacency of statistical significance if data from multi-electrodes were tested. However, this cluster-based approach would be problematic if the significant differences were strong and sustained in time, but varied greatly in space. In other words, neural generators, which contributed to the detected significant differences, changed markedly within the explored temporal-cluster. To solve this problem, we implemented a statistical approach based on topographical segmentation analysis, which did not only make use of the temporal adjacency of significance, but also utilized the scalp distribution of statistical difference. We applied this technique to assess the significant difference of SEPs between deviant and standard conditions, and we observed that temporal ROIs, captured distinct spatial distributions of statistical difference, could be correctly identified using the topographical segmentation analysis be means of quasi-stable scalp distribution.