Abstract. Since the early trials using human interferon (hIFN) derived from blood leukocytes or cell lines, fever has been a prominent component of IFN therapy. Human protein impurities might account for the fever to cell-derived hIFN, but recombinant hIFN, free of extraneous human proteins, has produced fever in nearly all recipients during clinical trials. Our present studies were carried out to determine the mechanisms of fever due to recombinant hIFN currently being used in humans. Because recombinant hIFN is produced in Escherichia coli, in these experiments we considered contaminating endotoxin as the cause of fever. Polymyxin B, which blocks endotoxin, had no effect on the pyrogenicity of hIFN in rabbits. In addition, hIFN injected into an endotoxin-resistant strain of mice produced fever. The pyrogenicity of hIFN does not appear to involve production of leukocytic pyrogen (LP), since no circulating LP was detected in rabbits during IFN fever. Furthermore, human mononuclear cells incubated with hIFN in vitro at 104_106 U/ml did not release LP. However, hIFN stimulated prostaglandin E2 (PGE2) release from rabbit hypothalamic tissue in vitro. Intracerebroventricular injection of hIFN into the awake cat also produced fever and a rise in PGE2 levels in the cerebrospinal fluid; both effects were reversed by treatment with indomethacin. We conclude that the fever of recombinant hIFN is not due to endotoxin but that