OBJECTIVE The aim of the present study was to investigate the reproducibility of the exponential method of CO2 rebreathing with the use of automated curve fitting and to determine whether this method is superior to the equilibrium method in terms of reproducibility and clinical practicability. METHODS Repeated measurements of cardiac output were performed using the automated equilibrium and exponential methods. These measurements were compared in 12 healthy male subjects at rest and during incremental exercise tests. RESULTS Estimated cardiac output was not significantly different between duplicate measurements at rest nor at any level of exercise with either method. At rest the exponential method showed a tendency toward larger variability than the equilibrium method. The exponential method produced significantly higher (P < or = 0.001) estimates at rest (averaging up to 9.8 L x min(-1)) compared with the equilibrium method (averaging up to 6.5 L x min(-1)). Reproducibility improved for both methods with increasing workloads, and a second measurement at rest also seemed more reproducible and valid than the first. During exercise, both methods produced comparable values for cardiac output, and highly significant relations between cardiac output and oxygen uptake were observed for both methods (ranging from r2 = 0.79 to r2 = 0.88, P < or = 0.001). The equilibrium method produced unpleasant side effects more frequently (75% vs 21%, P < or = 0.001) compared with the exponential method and lead more subjects to premature interruption of the exercise test because of the rebreathing maneuver (42% vs 17%, P = 0.058). CONCLUSIONS Automated curve fitting for the exponential method gave reproducible and valid results during submaximal and maximal exercise but not at rest. The equilibrium method on the other hand interfered with exercise. Therefore, the equilibrium method is recommended at rest and at lower levels of exercise and the exponential method at higher intensities.