In agreement with the results of previous studies (1), it was shown that intravenous injection of endotoxin into mice bearing 9-d SA1 sarcoma resulted in a tumor hemorrhagic reaction that rapidly caused necrosis of most of the center of the tumor, and then the complete regression of the rim of living tumor tissue that survived the hemorrhagic reaction. The tumor hemorrhagic reaction was confined to the vascular bed of the tumor, and its rate and extent of development were measured in terms of the intratumor extravasation of 51Cr-labeled syngeneic red cells. The development of the hemorrhagic reaction was associated with the presence in the tumor over a 6-h period of endogenous TNF that was measured in terms of its capacity to kill L929B cells in vitro and identified by its susceptibility to neutralization with a monospecific, polyvalent anti-rTNF antibody. The same antibody was capable in vivo of inhibiting the endotoxin-induced tumor hemorrhagic reaction by only approximately 50%, even when present in the tumor in excess. However, it was capable when given in the same quantity of inhibiting the ability of endotoxin to cause complete tumor regression. The fact that TNF was generated in the tumor during the tumor hemorrhagic reaction, and that infusion of a sufficient quantity of anti-rTNF antibody severely interfered with hemorrhagic necrosis and prevented tumor regression represents convincing evidence that TNF is an essential participant in endotoxin-induced regression of an established SA1 sarcoma. Moreover, because tumor regression, as opposed to hemorrhagic necrosis, failed to occur if the tumor was growing in immunoincompetent mice, but did so if the mice were infused with tumor-sensitized T cells, it can be concluded that an adequate level of T cell-mediated immunity is also an essential requirement for endotoxin-induced tumor regression. The participation of other endotoxin-induced mediators in tumor regression cannot be ruled out.