To determine the accuracy of thermal-dye indicator-dilution measurements of lung water during perfusion abnormalities, we embolized air into the lungs of 10 dogs anesthetized with pentobarbital and mechanically ventilated. A control period was followed by a period of air injection (10 ml) and subsequent air infusion (E1), a second period of air injection and infusion (E2), and a recovery period. Thermal and dye-dilution curves were obtained during each period, and cardiac output (CO), total thermal volume, intravascular volume, and extravascular thermal volume (ETV) were calculated. Pulmonary arterial pressure (Ppa) increased from 16 +/- 2 cmH2O during control to 29 +/- 3 cmH2O during E1 and to 40 +/- 3 cmH2O during E2 and decreased to 21 +/- 2 cmH2O during recovery. CO did not change, and pulmonary vascular resistance changed in a pattern similar to Ppa. ETV fell from 113 +/- 8 ml during control to 59 +/- 11 ml during E1 and to 29 +/- 9 ml during E2 and rose to 136 +/- 23 ml during recovery. There was no increase in extravascular lung water or mass determined gravimetrically at the end of the experiments. We conclude that the thermal-dye technique will underestimate lung water in nonedematous lungs if emboli prevent complete diffusion of the thermal indicator.