The steady-state distribution of endogenous albumin in mouse diaphragm was determined by quantitative postembedding protein A-gold immunocytochemistry using a specific anti-mouse albumin antibody. Labeling density was recorded over vascular lumen, endothelium, junctions, and subendothelial space. At equilibrium, the volume density of interstitial albumin was 18% of that in circulation. Despite this large difference in albumin concentration between capillary lumen and interstitium, plasmalemmal vesicles labeling was uniformly distributed across the endothelial profile. 68% of the junctions displayed labeling for albumin, which was however low and confined to the luminal and abluminal sides. The scarce labeling of the endothelial cell surface did not confirm the fiber matrix theory. The kinetics of albumin transcytosis was evaluated by injecting radioiodinated and DNP-tagged BSA. At 3, 10, 30, and 60 min, and 3, 5, and 24 h circulation time, blood radioactivity was measured and diaphragms were fixed and embedded. Anti-DNP antibodies were used to map the tracer in aforementioned compartments. A linear relationship between blood radioactivity and vascular labeling density was found, with a detection sensitivity approaching 1 gold particle per DNP-BSA molecule. Tracer presence over endothelial vesicles reached rapidly (10 min) a saturation value; initially localized near the luminal front, it evolved towards a uniform distribution across endothelium during the first hour. An hour was also needed to reach the saturation limit within the subendothelial space. Labeling of the junctions increased slowly, out of phase with the inferred transendothelial albumin fluxes. This suggests that they play little, if any, role in albumin transcytosis, which rather seems to proceed through the vesicular way.