E xperimental thrombocytopenia results in endothelial alterations associated with bleeding. In this study prednisone was shown to prevent or reverse these changes, which supports the clinical inference that adrenocorticosteroids decrease capillary fragility in thrombocytopenia. Rabbits (3-4 kg), intraperitoneally injected with busulfan, developed 98-99% reductions in platelet count and hemorrhaged profusely. Orally administered prednisone (0.2 mg/kg or 1.0 mg/kg daily) reduced bleeding despite persistent thrombocytopenia. Tongue biopsies obtained after 3 days of prednisone treatment were examined by electron microscopy. Normal rabbits served as controls. 25 consecutive capillaries or venules from each of four animals in the control group and each of five experimental groups were examined for fenestrations, “thin spots” (<800 A thick), and mean wall thickness as determined by planimetry. Vessels from control animals had no thin spots or fenestrations, and the mean vessel wall thickness was 4,254±105 A SEM. The 100 vessels from the thrombocytopenic animals had a mean vessel wall thickness of 2,081±218 A (P < 0.001), and 42 had thin spots of fenestrations. After administration of the smaller dosage of prednisone, the mean vessel wall thickness increased to 3,556±40 A (P < 0.001), and only nine vessels had thin spots or fenestrations. With the larger dosage, only six vessels had thin spots or fenestrations and the mean vessel wall thickness of this group increased to 3,704±206 A (P < 0.005). All preparations demonstrated normal endothelial junctions. The data are consistent with the hypothesis that the bleeding of thrombocytopenia is caused by altered capillary and venule endothelium and that diminished bleeding observed with prednisone administration results from amelioration of these endothelial changes.