We measured cardiac outputs at forty points from five patients by pulse dye densitometry and compared these with those measured by thermodilution method. We obtained a good correlation (y = 1.090 x = . 0.030, n = 40, 5 cases) and small mean bias (0.348 +/- 0.830 l.min-1, n = 40, 5 cases) between the two methods. We suggest that this method of cardiac output measurement by pulse dye densitometry is not invasive and may be more effective than that by thermodilution method.