It is suggested that disturbances of excitation velocity in optic nerve disease go along with a temporal heterogenity of impulse conduction which subsequently may be responsible for an increased variability of latencies of cortical visual evoked potentials. Since the latencies of each single evoked response, being superimposed by EEG-background activity, cannot be measured, we developed a computer-assisted statistical technique in an attempt to quantify this variability. Random samples (n = 30) were taken from the whole of stimulus-correlated EEG-segments recorded during a clinical routine examination (n = 200) and averaged to give a number (n = 30) of randomly sampled VEP's. The results of the statistical analysis, by calculating the variability of the latency values of these samples, can be taken as an approximation of the real variability of latency of single evoked potentials. It could be demonstrated that the variability of latency of the first clearly defined positive component P2 of the pattern-induced VEP is increased even in cases of clinically suspected retrobulbar neuritis or disseminated encephalomyelitis, where the summed VER-latency turned out to be within normal limits. Based upon these results, the statistical analysis proposed here has proven to be of differential diagnostic significance as an additional parameter - besides the ordinary measurement of latency - especially in borderline cases.