PURPOSE: The exercise status in patients with rheumatoid arthritis, associations between exercise behavior and personal factors, and associations between exercise behavior and exercise-specific cognitions and their effects were assessed. METHOD: Four hundred thirty nine outpatients with rheumatoid arthritis were studied. The exercise status was measured by a single item. The intensity was multiplied by the frequency and duration of each exercise. The product of these intensity values for all exercises was defined as exercise behavior. Based on the Pender's revised health promotion model, exercise benefit, barrier, self-efficacy, enjoyment and social support were chosen as exercise specific cognitions and affect variables. Path analysis was used to identify the predictors of exercise behavior. Results: Compared to the duration before being diagnosed, the number of subjects who exercised regularly increased after being diagnosed. However over half of the subjects refrain from any sort of exercise and the type of exercise is very limited. Among the variables, exercise barrier, self-efficacy, and social support were found to be significant predictors of exercise behavior, and only previous exercise experience was found to be significant predictors of all behavior specific cognitions and affect variables. CONCLUSION: These findings suggest that studies should explore exercise behaviors and strategies to emphasize the cognitive-motivational messages to promote exercise behaviors.