Cardiac output measurements determined by dye dilution, iced-injectate thermodilution and room temperature thermodilution were compared in man in order to assess the random error of each method and to examine the systematic error of both thermodilution methods in comparison with dye dilution. Results showed that random error was greatest with room temperature thermodilution and least using iced thermodilution. Iced thermodilution correlated well with dye dilution, tending to overestimate cardiac output only at low flows. Room temperature thermodilution, however, overestimated cardiac output by up to 25% in the clinically important range and more so at low cardiac output.