This prospective study examined effects of expectancy of exercise benefits, value of benefits, expectancy violation, and self-efficacy (SE) on exercise behavior and study dropout in sedentary women (n=86). SE predicted exercise for study completers but did not predict study dropout. After 6 and 12 weeks, participants evidenced expectancy violations, particularly for fitness and weight. Initial value, expectancy, and Expectancy X Value did not predict exercise for study completers, but dropouts had higher initial positive expectancies than did completers. Expectancy violations produced in Weeks 1-6 did not predict exercise in Weeks 7-12 in study completers, but dropouts in Weeks 7-12 had higher expectancy violations for weight in Weeks 1-6 than did completers. Follow-up revealed that study dropouts stopped exercising. Identifying individuals with high initial expectancies and expectancy violations may improve health behavior intervention and research.