Hemostatic reaction in the mural thrombus formation (up to 2 hrs) was examined by scanning electron microscopy after removal of the endothelium in the rabbit carotid artery by the use of a rough-surfaced needle. The endothelium was almost completely removed leaving a network of subendothelial microfilaments which sometimes appeared half embedded in the basement membrane. Platelet adhesion occurred on this subendothelial tissue in the following steps: attachment of discoid platelets, pseudopod formation and spreading. The subendothelium was thus covered by a layer of platelets in about 10 min. During adhesion they caught the microfilaments by their pseudopods and never produced hemispherical protrusions. Loose aggregates of the rounded platelets were then formed on them but they were frequently reversible, resulting in 1-2 hrs, in coverage by only one or two layers of the adhered platelets. This platelet reaction was weaker than that to collagen in case of bleeding in the previous report. Leukocyte participation in thrombus formation began at around 30 min. Fibrin strands appeared as tiny filaments which were attached exclusively to the activated platelets and later grew into thick and long fibers forming a network. Thus activated platelets seemed to be very important as the base of development of fibrin thrombus. Many erythrocytes were demonstrated to be destroyed at the mural thrombus after attaching to either the subendothelial components or activated platelets or fibrin strands and being deformed by the blood stream. This finding supports the hypothesis of microhemolysis in the hemostatic process by HELLEM (1961). This type of hemostatic reaction was proved to be caused by a slight manipulatory pressure on the arterial wall, suggesting the occurrence of thrombus formation in our daily life.