The time and frequency domain characteristics of the human electroencephalogram continue to be parameters of significant interest to both research workers and clinicians alike. In nearly every analysis of the e.e.g., the question of stationarity is raised. This paper provides a brief review of the literature and presents working definitions governing the theory for evaluation of the stationarity of the e.e.g. An easily implemented stationarity-analysis method is detailed and the technique is applied to 104 clinically normal e.e.g.s. The results verify that epoch lengths used intuitively by many electroencephalographers are also appropriate from a theoretical consideration of the signal's characteristics. The method presented provides a simple and useful assessment of whether a given sample is sufficiently stationary for meaningful analysis.