This study was undertaken to determine the optimum number of Doppler velocity waveforms required to calculate cardiac output in atrial fibrillation with the same degree of accuracy as that for sinus rhythm. Twenty-one patients in atrial fibrillation underwent calculations of cardiac output derived from aortic Doppler waveform velocity time integrals and RR intervals. The variability in estimates of the cardiac output was calculated with the successive addition of sequential beats and compared with that determined in a control group of 12 subjects in sinus rhythm. For the group in atrial fibrillation, a mean of 13 beats (range 4 to 17 beats) was required to achieve an estimation of cardiac output with a variability of less than 2%, compared with a mean of four beats in sinus rhythm. In atrial fibrillation, the mean number of beats required to determine cardiac output was approximately three times that necessary in sinus rhythm.