In the light of research by Epstein (1979) (which reported that error of measurement in the analysis of behavior stability may be reduced by examining the behavior of aggregate stability coefficients computed for measurements with known stability characteristics), this study examines stability coefficients for computer-generated data sets systematically varying in their stability characteristics. Score for 200 cases across 40 trials were generated to examine the behavior of aggregate stability coefficients for data with known stability characteristics. Forty replications were performed of seven studies, each study varying in the degree of stability characteristics of the scores. Aggregate stability coefficients were computed by correlating the average of even-trial scores with the average of odd-trial scores according to the recommendations of Epstein. Since high stability coefficients were found when only a small percentage of the cases exhibited score stability, the use of aggregate stability coefficients as evidence of stable behavioral dispositions was questioned. (Author/AEF)