The effectiveness of an exercise prescription and unsupervised home exercise programme was tested on 37 subjects with rheumatoid arthritis and 34 with systemic lupus erythematosus. Subjects were randomly assigned to control or stationary bicycling at home, using loaned bicycles. Exercise subjects (with bicycles) did better than controls, but not significantly, on all outcomed measures (exercise tolerance test, fatigue, depression and helplessness) at 3 months. Bicycles were reclaimed at 3 months and all subjects in both groups given instructions for home exercise. Exercise in the second 3 months was predicted primarily by baseline exercise habits and fatigue. It is concluded that although safe, unsupervised home exercise programmes may benefit few patients. Future research should address methods of stimulating and maintaining unsupervised exercise programmes in patients with systemic rheumatic disease.