In this study, we examined correlates of adherence and contamination in a randomized controlled trial (RCT) of exercise in cancer survivors using the theory of planned behavior and the Five Factor Model of personality (FFM). We randomly assigned cancer survivors in group psychotherapy classes to either a waiting-list control group (n = 45) or a home-based, moderate intensity exercise program (n = 51). At baseline, participants completed measures of the theory of planned behavior, the FFM, past exercise, physical fitness, medical variables, and demographics. We then monitored exercise over a 10-week period by weekly self-reports. Hierarchical multiple regression analyses indicated that the independent predictors of overall RCT exercise across both conditions were past exercise (beta = .36, p < .001), assignment to experimental condition (beta = .34, p < .001), sex (beta = .30, p < .001), and intention (beta = .14, p < .10). For exercise adherence in the exercise condition, the independent predictors were sex (beta = .38, p < .01), extraversion (beta = .30, p < .05), normative beliefs (beta = -.27, p < .05), and perceived behavioral control (beta = .23, p < .10). Finally, the independent predictors of exercise contamination in the control condition were past exercise (beta = .70, p < .001), sex (beta = .20, p < .05), and intention (beta = .17, p < .10). We conclude that the correlates of exercise adherence and contamination differ in kind as well as in degree. Explanations for these findings and practical implications for conducting exercise RCTs in this population are offered.