Ring preparations from the carotid, coronary, renal, mesenteric and femoral arteries of the dog were precontracted with the thromboxane mimetic U46619, after ensuring that the resting conditions were comparable from the Laplace relationship. In the presence of prazosin (1 microM) and propranolol (3 microM), noradrenaline (NA) relaxed the arteries in the order coronary greater than carotid greater than femoral greater than renal = mesenteric. When maximum relaxation to nitroglycerin (10 microM) was taken to be 100% the maximum relaxation to noradrenaline in each artery was: coronary 70%; carotid 34%; femoral 19%; renal 7% and mesenteric 2%. In endothelium-intact arteries UK14304 mimicked the relaxation responses to NA and idazoxan shifted the curves to both agonists to the right, consistent with an alpha 2-adrenoceptor classification. Substance P relaxed the arteries in the same order as for NA but showed higher efficacy i.e.: coronary 100%; carotid 80%; femoral 71% renal 49%; and mesenteric 41%. Removal of the endothelium abolished the relaxation to NA. We conclude that endothelium-dependent relaxation to NA and substance P varies greatly across 5 large arteries of the dog. This may indicate that endothelium-derived relaxing factor (EDRF) release is site-dependent or that the efficacy of EDRF on smooth muscle varies; being greatest in the coronary and weakest in the renal and mesenteric arteries.