Only inmates convicted of certain offenses were affected and the change was not retroactive, creating control groups not affected by the policy change. Data sources included misconduct reports (for relatively minor incidents), critical incident reports, and lists of inmates who participated in riots. Analyses of variance indicated that inmates not covered by the good time policy committed more misconducts and were more likely to participate in riots, but that main effects for time and type of offense accounted for these results. Critical incident data indicated a similar effect for type (but not time) of offense. These findings are discussed in terms of factors affecting inmate behavior and its measurement, prison management, and behavior theory. (Publisher abstract)