Reduction of a rapid ventricular rate in atrial fibrillation results in a longer diastolic filling period and a higher left ventricular stroke volume but this is offset by reduced contractility and fewer beats per minute; the net effect on cardiac output is uncertain. Sequences of stroke distances were measured by Doppler ultrasound in 60 resting patients with atrial fibrillation to determine the relation between ventricular rate and linear cardiac output. The slope of the cardiac output/ventricular rate relation was positive in all 20 patients with a ventricular rate less than 90 beats per minute and negative in 16 (40%) of 40 patients with a ventricular rate greater than 90 beats per minute. In atrial fibrillation the ventricular rate can be regarded as "controlled" when the cardiac output/ventricular rate slope is positive and "uncontrolled" when the slope is negative--that is when reduction of ventricular rate would lead to increased cardiac output. As so defined, ventricular rate at rest was controlled in every patient when the ventricular rate was less than 90, controlled in 44 (73%) patients when the ventricular rate was 90-140 beats per minute, and uncontrolled in every case when it was greater than 140 beats per minute. Achieving a target ventricular rate of 90 beats per minute in patients with atrial fibrillation at rest would result in control with the least compromise of cardiac output.