The purpose of the study was to examine psychological factors associated with maintenance of exercise behavior in a population of middle-aged individuals with elevated risk factors for cardiovascular disease. 191 males and 17 females took pan in a one-year diet and/or exercise intervention during 1990-1991. Four years later questionnaires were sent out to the 200 former participants who were still available for contact. 67.9% of those who answered ( n= 140) were categorized as exercisers, and 30.7% were categorized as nonexercisers. The majority of the exercisers had exercised at least one and a half years. A chi-squared analysis showed that whether the individuals were exercising or not at present was independent of whether they had exercised or not during the intervention study. Discriminant analyses were used to determine how well physical self-perceptions at different times would categorize exercisers and nonexercisers. Current physical self-perceptions categorized the Active Exercisers (86.9%) and the Nonexercisers (63.3%) the best (in total 79.1% correct classifications). Neither change in physical self-perceptions during the intervention nor change in physical self-perceptions from the end of the intervention until four years later, classified the exercise behavior as well. Three social cognitive models, The Self-perception model, The Health Belief model, and The Self-efficacy model, were investigated as discriminators between Active Exercisers and Nonexercisers. Active Exercisers were classified better than Nonexercisers, and current physical self-perceptions showed the highest percentage of total correct classifications. The proposed models were also analyzed as predictors of the variance in self-raced Motivation for Exercise. Outcome Expectations, Compliance Self-efficacy, Perceived Fitness, and Exercise Mastery explained 45% of the variance in self-rated Motivation for Exercise.