W e have demonstrated that human plasma contains a heparin-dependent inhibitor of thrombin that is distinguishable from antithrombin III (AT III). When a 1:50 dilution of plasma was incubated with greater than or equal to 0.01 U/ml heparin and 1 U/ml 125I-thrombin, the labeled thrombin B-chains became incorporated into two complexes of Mr-96,000 and Mr-85,000 that were separated by polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate and beta-mercaptoethanol. Neither complex was detectable at heparin concentrations less than 0.01 U/ml. When a limiting amount of 125I-thrombin was present, the proportion of radioactivity incorporated into each of the two complexes varied with the heparin concentration. Thus, the Mr-85,000 complex predominated at 0.01-5 U/ml heparin, whereas the Mr-96,000 complex predominated at 5-100 U/ml heparin. The Mr-85,000 complex reacted with antibodies to human AT III and comigrated with the purified thrombin-AT III complex. The Mr-96,000 complex did not react with antibodies to AT III or to alpha 1-antitrypsin, and it was detected in normal quantities after incubating 125I-thrombin with plasma immunodepleted of AT III, alpha 2-antiplasmin, alpha 2-macroglobulin, C1 inactivator, alpha 1-antichymotrypsin, or inter-alpha-trypsin inhibitor. The protein that combines with thrombin to form the Mr-96,000 complex was estimated to be present at a minimum concentration of 90 +/- 26 micrograms/ml (mean +/- SD) in identical to any of the known plasma protease inhibitors and that at relatively high heparin concentrations in vitro it reacts with thrombin more rapidly than does AT III.