Cardiac stroke volume was measured simultaneously by Doppler echocardiography and thermodilution in patients with severe coronary artery disease. One group (20 patients) was examined during supine exercise at 25 W and a second group (20 patients) during sitting exercise at 25 W and 50 W. In a third group (10 patients) the reproducibility of the non-invasive stroke volume estimate was determined during upright exercise. There was a considerable variation between results obtained with the invasive and non-invasive technique at rest and during exercise, but no systematic differences were found. Analysis of variance showed that no systematic differences between ultrasonic results were introduced by two observers or by measurements on different days. The coefficient of variation between any pairs of Doppler measurements in each patient was 6%. These findings demonstrate that the ultrasonic technique is suitable for detecting changes in stroke volume induced by low load exercise in patients with severe angina pectoris.