The existence of endothelium-derived vascular relaxant factor (EDRF) was postulated by Furchgott and colleagues when they observed that acetylcholine paradoxically relaxed preconstricted aortic strip preparations by an endothelium-dependent mechanism. This phenomenon has since been demonstrated in different blood vessels and mammalian species and it can be elicited by several other agents. EDRF has been thought to be a humoral agent, a lipoxygenase derivative and possibly a free radical. In the study reported here, by using aortic preparations from the rabbit, alone and in cascade experiments with isolated perfused coronary preparations, we demonstrate definitively that EDRF is a humoral agent. It is released from unstimulated aortic preparations containing endothelium, its release can be stimulated for prolonged periods by acetylcholine, and it is not a lipoxygenase derivative or free radical but an unstable compound with a carbonyl group at or near its active site.