Experiments in guinea pig heart Langendorff preparations assessed the effect of KATP channel blockade on the coronary vasoactivity of adenosine and 17 analogs chosen to represent a variety of purine and ribose modifications. Although glibenclamide is a functional antagonist that acts at the level of an effector rather than at a receptor, it caused parallel rightward shifts of agonist dose-response curves. The size of the shift of EC50 differed according to the kind of analog: the ranking was, generally, N6-phenethyladenosines > 2-aryl-aminoadenosines = 2-(1-alkyn-1-yl)adenosines > N6-cycloalkyladenosines = adenosine-5' -uronamides. The coronary vasoactivity ranking of agonists in the presence of supramaximal concentrations of glibenclamide was 2-(1-alkyn-1-yl)adenosines = 2-aralkoxyadenosines > 2-aralkylaminoadenosines > 2-arylaminoadenosines > N6-substituted adenosines. Glibenclamide did not affect the vasoactivity of adenosine itself, perhaps because avid uptake by endothelial cells prevented penetration of the agonist to receptors deeper in the vascular wall. The results exclude a model consisting of one kind of receptor acting exclusively through a KATP channel, argue against one kind of receptor coupled to a KATP channel as well as to an additional effector but is consistent with two kinds of vasodilatory adenosine receptors, one of which activates a KATP channel. The identity of the adenosine receptor coupled to the KATP channel is uncertain; the other receptor has the pharmacological profile of an A2a-adenosine receptor.