Viscosity of blood was measured in the absence of anticoagulants by means of an improved cone-in-cone viscometer. Data obtained in a study of a series of patients suffering from venous thrombosis and coronary occlusion were compared with data from a control group of healthy subjects. The probability that the large difference in viscosity observed between normal and abnormal blood is due to chance is approximately 1 in 1,000,000. The viscosity of blood remains constant, at any rate of shear, during the first two to four minutes after removal of blood. Clotting starts after various time periods, depending not only on the intrinsic properties of the particular blood sample, but also on the velocity gradient at which such clotting takes place.The following general trends were established: at 36°C the onset of clotting is much more rapid at high rates of shear than at low rates of shear. It appears, however, that at lowered temperatures this trend is reversed.The causes of aggregation of red cells, and the role of this aggregation in clotting, are discussed. It is emphasized that blood is thixotropic, i.e., that its viscosity is both time and rate-of-shear dependent. Thixotropy of blood is due, mainly, to a reversible aggregation of red cells. This suggests the existence of intermittent sedimentation of red cell aggregates during flow in blood vessels. Owing to the existence of a velocity profile, in which the axial region is under zero or near-zero rate of shear, such aggregates should be present especially in large veins and arteries. It appears that an increased thixotropy and viscosity are symptomatic of some thrombotic states.