1. This study was undertaken to determine the accuracy of a modification of a single breath method for estimation of cardiac output. The technique incorporated a single rebreathing stage followed by a prolonged expiration. Cardiac output was determined from the O2 uptake and the instantaneous changes in O2 and CO2 in the expired gas during the prolonged expiration.
2. The mean values and the random errors (determined from the differences between pairs of estimates) of cardiac outputs in normal subjects at rest and exercise were 5.42 and ± 0.60 litres/min (2 sd, 60 pairs) and 14.1 and ±1.8 litres/min (40 pairs).
3. Larger random errors were obtained in a group of cardiac patients but, except in hypoxic patients, the mean values obtained by the single breath and the direct (Fick) methods were almost identical.
4. We conclude that our modification of the single breath method is simple to use and sufficiently reliable for use in humans both at rest and during steady states of light exercise.