Longitudinal studies have shown that physical illness attributions are associated with poor prognosis in chronic fatigue syndrome (CFS). Speculation exists over whether such attributions influence treatment outcome. This study reports the effect of illness beliefs on outcome in a randomized controlled trial of cognitive-behavior therapy versus relaxation. Causal attributions and beliefs about exercise, activity, and rest were recorded before and after treatment in 60 CFS patients recruited to the trial. Physical illness attributions were widespread, did not change with treatment, and were not associated with poor outcome in either the cognitive-behavior therapy group or the control group. Beliefs about avoidance of exercise and activity changed in the cognitive behavior therapy group, but not in the control group. This change was associated with improved outcome. These findings suggest that physical illness attributions are less important in determining outcome (at least in treatment studies) than has been previously thought. In this study, good outcome is associated with change in avoidance behavior, and related beliefs, rather than causal attributions.