The single-breath method of Kim, Rahn and Farhi (1966) enables the estimation of the pulmonary blood-flow from the PCO2 vs. PO2 curve recorded during a prolonged expiration. Studies on the repeatability of the method have given very different results. At one extreme the estimated pulmonary blood flows ranged from -47 to 173 litre/min and at the other extreme the repeatability was found to be better than that usually reported for invasive methods. A possible explanation for these important differences is the use of different data-reduction procedures to calculate the pulmonary blood-flow from the measured PCO2 and PO2 values. In the present study we have investigated the repeatability of the single-breath method in 20 test subjects using four previously published data-reduction procedures. Furthermore, we have studied the sensitivity of the estimated pulmonary blood-flow to measurement noise for each of the four data-reduction procedures, by using a previously described technique based on computer simulations. Three of the data-reduction procedures gave standard deviations on repeated single determinations of the pulmonary blood-flow of about 1 litre/min whereas the last method gave results which fluctuated wildly from one determination to the next. The simulation showed that about 1/3 of the random variations can be ascribed to the measurement noise in PCO2 and PO2.