Changing a health behavior and maintaining a positive change can be very difficult. This study examined attributions for health behavior change by using retrospective reports to elicit college students' (N=466) current views of successes and failures at adopting health promoting behaviors. In completing the Health Behavior Questionnaire, 229 subjects provided details of a successful attempt to change a health behavior and 237 subjects provided details on an unsuccessful attempt. Subjects freely generated their own causal ascriptions and used the Causal Dimension Scale to locate their attributions on the dimensions of locus, stability, and controllability. Causal dimension scores indicated that the average attribution was internal, unstable, and controllable and that success attributions were more stable and controllable than were failure attributions. This suggests a tendency to make attributions that enhance perceived control over health behavior outcomes. A self-serving attributional bias was observed for reports involving exercise and substance use, but not for eating. Successful road safety habit changes were more external, more stable, and less controllable than were unsuccessful changes. Stable attributions were associated with expectations that previous outcomes would continue into the future. (Author/NB)