Although Doppler echocardiographic measurements of aortic flow have been found to correlate with stroke volume, the reliability of this technique is unknown. The purpose of this study was to measure the reliability of Doppler estimates of cardiac output by identifying and estimating the magnitude of different sources of error. We measured the reliability of Doppler estimates of cardiac output by identifying the magnitude of sources of error in 11 subjects with studies performed by two technicians and read by two readers. Analysis with generalizability theory demonstrated that the largest portion of the total variance was from differences among patients, with a smaller contribution due to day-to-day variability. Variability due to technician was low for continuous wave Doppler (2.0%), but high for pulsed wave (23.2%). Thus continuous wave, but not pulsed wave Doppler measurements, can be used to detect serial changes in cardiac output due to an intervention.