The averaging of individual late component event related potential (ERP) responses, particularly P300, has revealed significant differences between schizophrenic patients and normal subjects. However, the averaging process removes the variability of the individual epochs that constitute that average. The response-variance-curve (RVC) method quantifies the variability of the individual epochs and allows examinations of windows of maximum variance. In this study, we examine the complementary nature of the RVC method to the traditional averaging approach. The averaged N200 and P300 ERP components differed significantly between the schizophrenic and normal groups, but not between the unmedicated and medicated schizophrenic patients. The RVC measure, on the other hand, revealed systematic differences in variability, maximal between 190 and 240 ms, between the unmedicated and medicated schizophrenic patients. The RVC measure therefore provides a focused time frame in which to examine dysfunctions in information processing and macroscopic scale changes in brain function due to medication. © 1995.