The accuracy and reproducibility of indirect measurement of cardiac output at rest by the carbon dioxide rebreathing (indirect Fick) method with an automated respiratory analysis system (Gould 9000IV) were compared with simultaneous measurements made in duplicate by dye dilution and thermodilution in 25 patients having cardiac catheterisation studies. Measurements of cardiac output by the carbon dioxide rebreathing method were not significantly different from those obtained with dye dilution (mean difference -0.3 l/min, SD 0.76, 95% confidence interval -0.7 to 0.1). Thermodilution significantly over-estimated cardiac output by a mean of 2.2 l/min or 39% (SD 1.5, 95% confidence interval 1.6 to 2.8) compared with the carbon dioxide rebreathing method and significantly overestimated cardiac output by 1.9 l/min or 31% (SD 1.2, 95% confidence interval 1.2 to 2.5) compared with dye dilution. The reproducibility of measurements of cardiac output in individual patients was satisfactory with the dye dilution method but was poor with carbon dioxide rebreathing and thermodilution. Indirect measurement of resting cardiac output by the Gould 9000IV automated carbon dioxide rebreathing method is more accurate but the variability inherent with this method requires that multiple measurements be taken for each determination. Measurement of cardiac output by the thermodilution method by a commercially available cardiac output computer was not satisfactory because not only was there considerable variability between repeat measurements but the method also consistently overestimated cardiac output compared with the dye dilution method.