Purpose. We hypothesized that patients randomized to an exercise program would demonstrate a measurable improvement in behavioral outcomes with no adverse clinical outcomes, as compared with control patients. Methods. This randomized, prospective, parallel, and controlled study spanned 12 weeks. Twenty-eight patients were randomized either to participate in a supervised exercise program (Exercise) or to continue their current level of activity with no planned intervention (Control). The Exercise group worked with an exercise physiologist three times per week. At specific intervals, behavioral (QOLIE-89, POMS, PSDQ, Self-Esteem) and clinical (seizure activity, antiepileptic drug (AED) concentrations) outcomes were measured. Results. Twenty-three patients completed the study (Exercise n = 14, Control n = 9). Of the four patients in the Exercise group with active seizures, two had no change, one had an increase, and one had a decrease in seizure activity. Of the three patients in the Control group with active seizures, one had no change, one had an increase, and one had a decrease in seizure activity. In all patients, there was < 26% coefficient of variation in AED concentrations over the 12-week study, suggesting little or no impact of the exercise intervention. The overall quality of life and two domain scores improved from baseline to Week 12 in the Exercise group (P = 0.031), while the Control group score did not change (P = 0.943). In the Exercise group, there were several measures of physical self-concept and vigor that improved and total mood disturbance decreased from the beginning to the end of the program. Conclusion. This is the first randomized, controlled study of exercise in patients with epilepsy. Behavioral outcomes are positively influenced by moderate exercise and there is no impact on seizure frequency. This suggests that exercise should not be discouraged in the care of epilepsy patients. The ability to offer an exercise program adds a health promotion component to the current plan of care provided by our comprehensive epilepsy program. (C) 2001 Elsevier Science