To identify predictors of physical activity levels in patients with chronic heart failure, 43 patients, aged 33 to 91 years, who had well-compensated heart failure were asked to perform a symptom-limited exercise treadmill test and to complete activity logs for 2 consecutive days while wearing an ambulatory heart rate activity monitor. Activity logs included information on the type of activity, duration, rating of perceived exertion, symptoms experienced, and the intensity of symptoms. Subjects also completed the Duke Activity Status Index, a brief self-administered questionnaire that assesses physical functioning, and a self-efficacy for general activity questionnaire. Simultaneous multiple regression analysis was used to predict physical activity levels from a model that included: personal variables of physical fitness (peak oxygen consumption); knowledge, attitudes, and beliefs including self-efficacy for general activity, and rating of perceived exertion during daily activity; and environmental factors such as social support (marital status). The overall model explained 38% of the variance (p < 0.001). Self-efficacy (p = 0.015) was the strongest predictor of physical activity in this group. From this initial descriptive study, we conclude that self-efficacy is a better predictor of performance of physical activity than measures of physical fitness or rating of perceived exertion during activity. Additional studies are needed to examine other behavioral and physiologic mediators as well as behavioral strategies that may be used to increase participation in physical activity programs. Particularly promising are strategies to enhance self-efficacy for exercise.