1. Short periods of coronary artery occlusion protect the heart against the effects of a subsequent prolonged period of ischaemia. This phenomenon is known as preconditioning of the ischaemic myocardium. 2. In mongrel, chloralose-urethane anaesthetized open-chest dogs, within a restricted body weight range, two 5 min periods of occlusion of the anterior descending branch of the left coronary artery markedly reduced the severity of the early ischaemic arrhythmias resulting from a prolonged (25 min) occlusion of the same coronary artery starting 20 min later. Thus, the number of ventricular premature beats (VPBs) was reduced from 528 +/- 140 in controls to 78 +/- 27 in preconditioned dogs, the incidence of ventricular fibrillation (VF) was reduced from 47% to 0% and the incidence of ventricular tachycardia (VT) from 100% to 20%. ST-segment elevation recorded from electrodes within the ischaemic area, and the degree of inhomogeneity of conduction within the ischaemic area were markedly reduced in these preconditioned dogs. 3. The incidence of VF following reperfusion of the ischaemic myocardium at the end of the 25 min occlusion period was reduced in the preconditioned dogs from 100% to 60%; there was thus a 40% survival from the combined ischaemia-reperfusion insult compared with 0% in the controls. 4. NG-nitro-L-arginine methyl ester (L-NAME) an inhibitor of the L-arginine nitric oxide pathway, given in a dose of 10 mg kg-1 intravenously on two occasions, both before the initial preconditioning occlusion and then again before the prolonged occlusion, partially attenuated the protective effects of preconditioning.There were more VPBs (220 +/- 75), a higher incidence of VT (60%) and more episodes of VT (11.5 +/- 6.0 compared to 0.7 +/- 0.3 episodes in the preconditioned dogs not given L-NAME); none of the animals survived reperfusion (incidence of VF 100%). The improvement in the severity of the degree of in homogeneity which resulted from preconditioning was abolished by L-NAME administration.5. L-NAME itself elevated blood pressure (from 96 +/- 5 mmHg diastolic to 119 +/- 7 mmHg), reduced heart rate (from 155 +/- 7 to 144 +/- 4 beats min-') but did not change LVEDP, LVdP/dt,,,,, coronary blood flow, ST-segment elevation or the degree of inhomogeneity of conduction. When given 10 min before the prolonged coronary artery occlusion in dogs not subjected to preconditioning, L-NAME had no significant effect on the severity of arrhythmias except for more periods of VT (a mean of 11.7 +/- 4.7episodes per dog).6. It is concluded from these studies that the generation of nitric oxide contributes to the marked antiarrhythmic effects of preconditioning in the canine myocardium, probably through elevation of cyclic GMP.