The traditional averaging process used to derive event related potential components (ERPs) is a soundly based method of determining the underlying ERP. Averaging, however, ignores the variability due to the single-trial ERPs that constitute the traditional average ERP. This variability may reflect complementary functional information to the average measure. Our group applied a simple procedure, the response variance curve (RVC), which measures single-trial ERP variability relative to their average. In this study, the average ERP and RVC measures (generated from the same single-trial task-relevant target ERPs) were assessed in an auditory oddball paradigm, in 17 unmedicated male adolescents with attention deficit hyperactivity disorder (ADHD) and in 17 age- and sex-matched normal controls. P300 amplitude, latency and point of maximum variability of the RVC were measured within the P300 processing window (250-500 ms post-stimulus). There were no significant differences in P300 amplitude or latency between the groups. Unmedicated ADHD patients, however, showed significantly increased single-trial variability within the P300 window compared with controls. This variability was significantly reduced with stimulant medication.