Using arabinogalactan (stractan II) densitygradient centrifugation. normal washed rabbit platelets and platelets that had been treated with thrombin were separated into three density populations. Under the conditions chosen, about 6 % of the normal platelets appeared in the least dense fraction. 32 % in the fraction of intermediate density. and 62 % in the most dense fraction. In contrast, when thrombin-treated platelets were separated on the gradient. 46 % were found in the least dense fraction, 40 % in the intermediate fraction, and 14 % in the most dense fraction. Although the serotonin and adenine nucleotide contents and the heparin-neutralizing activity of thrombin-treated platelets were less than those of control platelets, there was little difference among the platelets of different densities with respect to these platelet constituents. These observations indicate that the density differences are not solely attributable to loss of granule contents. In addition, platelets treated with thrombin and labeled with 51Cr in vitro were injected into rabbits, reharvested, and separated on the gradient. The distribution on the gradient of this mixture of thrombin-treated and normal platelets obtained from the circulation was 26 % in the least dense fraction, 46 % in the fraction of intermediate density. and 28 % in the most dense fraction; 68 % of the 5tCr-Iabeied thrombin-treated platelets were in the least dense fraction. Thus, it may be possible to use density gradients to detect platelets that have been exposed to thrombin in vivo during thrombus formation and have subsequently returned to the circulation when the thrombi break up.