Several rebreathing methods are available for cardiac output (Q (T)) measurement. The aims of this study were threefold: first, to compare values for resting Q (T) produced by the equilibrium-CO(2), exponential-CO(2) and inert gas-N(2)O rebreathing methods and, second, to evaluate the reproducibility of these three methods at rest. The third aim was to assess the agreement between estimates of peak exercise Q (T) derived from the exponential and inert gas rebreathing methods. A total of 18 healthy subjects visited the exercise laboratory on different days. Repeated measures of Q (T), measured in a seated position, were separated by a 5 min rest period. Twelve participants performed an incremental exercise test to determine peak oxygen consumption. Two more exercise tests were used to measure Q (T) at peak exercise using the exponential and inert gas rebreathing methods. The exponential method produced significantly higher estimates at rest (averaging 10.9 l min(-1)) compared with the equilibrium method (averaging 6.6 l min(-1)) and the inert gas rebreathing method (averaging 5.1 l min(-1); P < 0.01). All methods were highly reproducible with the exponential method having the largest coefficient of variation (5.3%). At peak exercise, there were non-significant differences between the exponential and inert gas rebreathing methods (P = 0.14). The limits of agreement were -0.49 to 0.79 l min(-1). Due to the ability to evaluate the degree of gas mixing and to estimate intra-pulmonary shunt, we believe that the inert gas rebreathing method has the potential to measure Q (T) more precisely than either of the CO(2) rebreathing methods used in this study. At peak exercise, the exponential and inert gas rebreathing methods both showed acceptable limits of agreement.