The optimal atrioventricular delay at rest and during exercise was investigated in nine patients with heart block and implanted dual chamber pacemakers. All patients studied had normal left ventricular function and a normal sinus node rate response to exercise. Cardiac output was measured by continuous wave Doppler and was calculated as the product of stroke distance measured by Doppler at the left ventricular outflow, aortic root area measured by M mode echocardiography, and heart rate. Pacemakers were programmed in the DDD mode. Cardiac output was measured with the patient at rest while supine and while erect and at the peak of submaximal exercise (the end of stage 1 of the Bruce protocol) with the pacemakers programmed to the following atrioventricular intervals: 75-80 ms, 100-110 ms, 140-150 ms, and 200 ms. During exercise the basic pacing rate was programmed to 70 beats/min. Cardiac output at rest while supine and erect was greatest with an atrioventricular delay of 140-150 ms and it was significantly higher than that with an atrioventricular delay of 75-80 ms. On average there was a 31% decrease in cardiac output when patients stood up. During treadmill exercise, however, cardiac output was greatest when the atrioventricular delay was 75-80 ms, and this was significantly higher than the cardiac output with atrioventricular delays of 150 and 200 ms. During exercise 1:1 atrioventricular relations were maintained in patients at all atrioventricular intervals. In patients with atrioventricular sequential pacemakers cardiac output at rest is greatest with an atrioventricular delay of 140-150 ms but during exercise the optimal atrioventricular delay is shorter. Rate modulation of the atrioventricular interval may improve the haemodynamic response and possibly exercise tolerance in patients with dual chamber pacemakers.