The purpose of this study was to examine predictors of exercise adherence (i.e. exercise in the intervention group) and exercise contamination (i.e. exercise in the control group) in a randomized controlled trial of home-based exercise in colorectal cancer survivors. At baseline, 102 participants completed measures of the theory of planned behavior, personality, past exercise, exercise stage of change, physical fitness, and medical/demographics and then were randomly assigned in a 2:1 ratio to an exercise (n=69) or control (n=33) group. Exercise was monitored weekly for 16 weeks using self-reports by telephone. Ninety-three (91%) participants completed the trial. Adherence was 76% in the exercise group and contamination was 52% in the control group. Hierarchical stepwise regression analyses indicated that baseline exercise stage of change (β=0.35; p=0.001), employment status (β=−0.28; p=0.010), treatment protocol (β=−0.26; p=0.018), and perceived behavioral control (β=0.20; p=0.055) explained 39.6% of the variance in exercise adherence. Intentions (β=0.36; p=0.049) and baseline exercise stage of change (β=0.30; p=0.095) explained 29.9% of the variance in exercise contamination. These findings may have implications for conducting clinical trials of exercise in colorectal cancer survivors and for promoting exercise to colorectal cancer survivors outside of clinical trials. Copyright © 2004 John Wiley & Sons, Ltd.