Twenty‐eight clinically compensated cardiac patients with atrial fibrillation have been investigated with the graded exercise test. Twenty‐four patients had one test with a lower dose of digitalis and one with a higher dose, both dose rates being within the therapeutic range. Four patients performed one exercise test with and one without digitalis. The heart (ventricular) rate attained at a given exercise load was significantly lower (with a difference of 5–25 beats/min at the different loads) with the higher dose. This was interpreted as a beneficial effect on working capacity. The individual beat‐to‐beat variability of heart rate was not influenced. The beneficial effect of an increase in digitalis dose thus seems to depend on a decrease in average ventricular rate during exercise and not on a decreased degree of arrhythmia.