Extravascular lung water (idQw1) is measured in vivo from the difference in mean transit times, computed by extrapolating the dilution curves, of two indicators, one freely diffusible, the other confined to the intravascular space. Using 3H2O it has been shown that idQw1 is smaller than the amount of extravascular water obtained from the difference between wet and dry lung weight (Qw1). Extrapolation allows one to use dilution curves for a short time, i.e., up to onset of obvious recirculation. Clearing the dilution curves or recirculation by deconvolution extends the observation time, which then becomes limited by sampling duration rather than onset of recirculation. This procedure entails recording recirculating tracers in the pulmonary artery (PA). Dilutions of tracers at input in PA and output in a systemic artery must be related to each other as continuous time functions. This is accomplished by means of a convolution integral. Deconvolution yields the frequency function of water molecule transit time in the extravascular lung space, l(t). In dogs and men, in both normal and edematous lungs, l(t) exhibits a knee and a fairly long tail. Extravascular lung water computed from l(t), idcQw1, agrees with Qw1 and correlates with data on the extravascular thermal volume of the lung and with radiographic findings of lung edema. A radiographic score of pulmonary edema may be used clinically to assess extravascular lung water in cardiac patients and in patients with adult respiratory distress syndrome.