This study tested the capability of low-power radiofrequency current delivered through a bipolar "epicardial-endocardial" catheter electrode configuration to produce discrete epicardial left atrial (LA) and left ventricular (LV) necrosis adjacent to the mitral anulus for potential application in ablating left free-wall accessory atrioventricular pathways. In 15 anesthetized, closed-chest dogs, a 6F electrode catheter was inserted via the jugular vein into the coronary sinus (CS). A second catheter was inserted via the femoral artery into the left ventricle and positioned beneath the mitral valve, high against the anulus, and directly opposite the CS electrode. The LV tip electrode was positioned to record the largest LA potential to ensure proximity to the anulus. Thirty-four sites were tested (five anterior, 14 lateral, and 15 posterior). Radiofrequency current (continuous wave, 625 kHz) was delivered between the CS and LV electrodes at 37-55 V (median, 41 V) for 4-60 seconds (median, 20 seconds). Current ranged from 0.10 to 0.35 A (median, 0.18 A), resulting in power ranging from 4.3 to 19.2 W (median, 7.3 W) and total energy of 51-446 J (median, 152 J). Dogs were sacrificed 2-9 days later. The CS was grossly intact in all dogs and thrombosed in one dog. The circumflex artery was grossly normal in all dogs. Necrosis of a small segment of the arterial wall was found microscopically in one dog. Lesions were identified at 30 of the 34 sites. Twenty-two (73%) of the 30 lesions consisted of a cylindrical-shaped area of necrosis extending between the anulus and CS with diameter ranging from 2.1 to 15.0 mm (median, 4.0 mm). Atrial and ventricular epicardial necrosis extended 0-7.0 mm (median, 2.5 mm) and 0-6.8 mm (median, 2.6 mm) beyond the anulus, respectively. At the remaining eight (27%) sites, little or no epicardial injury occurred, possibly because of downward displacement of LV electrode (four sites) or positioning of LV electrode within a trabecular recess (four sites). We conclude that 1) radiofrequency current delivered between CS and LV produced, at 22 (65%) of 34 sites, LA and LV necrosis adjacent to the anulus without rupture of the CS and that 2) large, sharp LA potentials help identify an optimal anular location of LV electrode. This technique may have clinical usefulness for ablating left free-wall accessory atrioventricular connections.