Techniques were developed for measurement of signal to noise ratio and response variability in single trial evoked potentials. These techniques were extended and verified using a digital computer simulation of signal plus noise trials. When applied to real data the measures appear to have high reliability and to demonstrate that human evoked potentials are more variable than would be expected from background noise variation alone. Empirical equations are presented which can be applied to existing single trial EP data to estimate both signal-to-noise ratio and its expected variance.