The defibrillation threshold is markedly reduced very early following the initiation of ventricular fibrillation. The purpose of this study was to determine if the same finding holds true for atrial defibrillation. Sustained, reproducible AF was induced with programmed atrial pacing using acetyl-beta-methylcholine chloride (40-640 microL/min) in six adult sheep (heart weight 245-300 g). Seven timing intervals (125 ms, 200 ms, 1 s, 3 s, 10 s, 30 s, and 5 min after AF induction) and two lead configurations: (1) RA as cathode and CS as anode; and (2) RA as cathode and RV apex as anode were tested. Single capacitor biphasic waveforms (3/1 ms) were delivered and atrial defibrillation thresholds (ADFTs) were determined in random order. No significant differences in leading edge voltage and total energy were detected for the RA-CS configuration for the seven timing intervals. For the RA-RV configuration, a significant difference was detected comparing the voltage for 125 ms to the 5-minute timing interval. For all times except 125 ms, the RA-RV threshold was significantly higher than the RA-CS level. In contrast to ventricular defibrillation, the ADFT does not change significantly within the first 5 minutes after the initiation of AF for the RA-CS configuration. However, if the shock is given very early (125 ms after AF induction) with the RA-RV configuration, the ADFT is lowered almost to the RA-CS level.