Variability of single event-related brain potentials

Electrical responses of the brain to stimuli can be conceived as deterministic low-amplitude signals embedded in colored noise. Subjects are repeatedly stimulated to extract the brain's response by averaging. The authors address the issue that single responses may vary from trial-to-trial. Three statistical tests are defined and their theoretical properties are given. Simulations based on real data demonstrate the tests' empirical validity and their differential sensitivity to different types of signal variation. A real data application shows that substantial variations were present even in a simple experimental setting.<>