1. Anaesthetized, open-chest dogs were subjected to 60 min of left circumflex coronary artery occlusion followed by 90 min of reperfusion. Endothelium-dependent and -independent relaxant responses of the isolated coronary arterial rings were then investigated. 2. The endothelium-dependent, acetylcholine-induced relaxation of ischaemic/reperfused arterial rings was significantly attenuated in comparison to control rings (1.9 fold rightward shift, ischaemic/reperfused maximum relaxation = 57 +/- 13% of control maximum relaxation; P less than 0.05). In contrast, glyceryl trinitrate produced similar relaxant responses in control and ischaemic rings. 3. Pretreatment of dogs with either amlodipine (3 micrograms kg-1 min-1, i.v.) or propranolol (1 mg kg-1, i.v.) completely prevented the postischaemic impairment of endothelium-dependent relaxant responses (100 +/- 3% and 90 +/- 5% of control maximum relaxation, respectively). 4. Allopurinol pretreatment (25 mg kg-1, p.o. 24 h previously, plus 50 mg kg-1 i.v. 5 min before arterial occlusion) partially protected against endothelial dysfunction by preventing the ischaemia-induced rightward shift of the acetylcholine relaxation curve and increasing the maximum relaxation response (83 +/- 7% of control rings). 5. These results confirm that endothelium-dependent coronary vascular relaxation is impaired by ischaemia and reperfusion, and that the ischaemia-induced impairment is reduced by pretreatment with amlodipine, propranolol or allopurinol.