The computer pattern recognition system for the study of spontaneous rat behavior has allowed new analytical techniques which expand the definition of experimentally induced changes in behavior. As with any technique, the stability of the measures must be considered when evaluating overall sensitivity. This study evaluates the stability and reproducibility of three behavioral measures: a measure of the number of initiations of specific behavioral acts, a measure of the total time of each act, and a measure of behavioral time structure. Normal statistical parameters are used to evaluate the significance of changes detected using the first two measures, but the third measure utilizes K-functions, the bootstrap and ad hoc criteria to evaluate significance of observed changes. This study compares the stability of results from these three measures as applied to fourteen different groups of control Sprague-Dawley male rats. All three measures provided stable and reproducible results, but the measure of time structure, the K-function analysis, provided the greatest consistency. Behavior, particularly spontaneous behavior, has traditionally been perceived as being intrinsically variable. However, this study shows that the computer pattern recognition system and its analytical techniques provide stable and reproducible values that vary only a few percent.