To evaluate albumin transport across the pulmonary capillary endothelial and interstitial barriers, we simultaneously measured blood-to-tissue (QA,t) and blood-to-lymph (QA,l) clearances of 125I-radiolabeled albumin as well as endogenous albumin clearance (Qa,l) in the canine lung in vivo (n = 10). Steady-state prenodal lung lymph flows (Qw,l) and protein clearances were measured over a 2-h period at a constant capillary pressure (Pc, 13-33 cmH2O). Comparison between QA,t and QA,l as a function of Pc suggests that little of the albumin that crossed the capillary wall remained in the lung tissue, with most leaving in the lymph. Qw,l increased significantly as Pc increased, but lung tissue water was minimally affected. From the ratio of the clearance-Pc slopes for albumin and water, the albumin reflection coefficient was estimated to be 0.81 using QA,l and Qw,l and 0.56 using Qa,l and Qw,l. The permeability surface area product for the sum of blood-to-tissue and blood-to-lymph fluxes of labeled albumin (QA,t + QA,l) was 31 +/- 9 microliters/min, whereas that calculated from the blood-to-lymph flux of endogenous albumin (Qa,l) was 97 +/- 22 microliters/min. These data suggest that 1) both tissue and lymph accumulations of albumin must be considered when microvascular permeability is evaluated using protein tracers; 2) lymph clearance, but not tissue accumulation of albumin, was filtration dependent; and 3) lymph flow was an important contributor to the safety factor against edema formation over a moderate range of capillary pressures.