Pulmonary extravascular water volume was obtained as the lung thermal volume (LTV) in mongrel dogs (n = 31) by double indicator dilution method, using negative heat and indocyanine green. We used a technic of one injection site (right atrium) and two sampling sites, i.e., pulmonary artery trunk and aortic root. In 13 dogs, cardiac outputs determined simultaneously from thermodilution curves in pulmonary artery trunk, aortic root and from dye dilution curve in aortic root, were in a good agreement with a coefficient of variation of +/- 12%. Lung thermal volume measured twice within a period of 5 minutes in 17 dogs, gave closely similar values with a coefficient of variation of +/- 9%. Lung thermal volume was 5.7 +/- 1.1 ml/kg (mean +/- SD) and corresponding "weighed lung water" (WLW), which was measured by destructive, direct method, was 5.0 +/- 1.0 ml/kg, the ratio of LTV to WLW (LTV/WLW) being 1.2 +/- 0.2 in control state. LTV/WLW ratio was 1.2 +/- 0.2 in 11 dogs loaded with dextran, and 1.1 +/- 0.1 in 6 dogs loaded with alloxan. Thus the lung thermal volume slightly exceeded WLW. In order to evaluate the extent to which the thermal diffusion into the left ventricular wall would cause LTV to slightly overestimate the pulmonary extravascular water volume, dye dilution curves and thermodilution curves were recorded simultaneously in the left atrium (LA) and aortic root (Ao) in a total of 25 runs in 5 dogs. The ratio of LTV obtained in LA to that in Ao was 0.9 +/- 0.1 (mean +/- SD). It was concluded that the pulmonary extravascular water volume was overestimated, by about 10% due to the thermal diffusion into the left ventricular wall.