Objectives. The purposes of this study were to examine the first activations following atrial defibrillation shocks to help understand how and where atrial fibrillation (AF) relapsed following failed shocks and to assess the difference in postshock activation between failed and successful shocks. Background. While many studies have investigated the mechanism of ventricular defibrillation, much less is known about the mechanisms of AF. Methods. Sustained AF was induced electrically after pericardial inf usion of methylcholine in 10 sheep. Biphasic subthreshold shocks were delivered to three configurations: right atrium to distal coronary sinus (RA-CS), sequential shocks with RA-CS as the first pathway followed by proximal CS to superior vena cava as the second pathway (Sequential), and right ventricle to superior vena cava plus can (V-triad). In eight sheep, global atrial mapping was performed with 504 electrodes spaced 3 to 4 mm apart. Results. Earliest postshock activations mostly arose from the left atrium for V-triad but arose from either atrium for RA-CS and Sequential. Preshock AF cycle lengths were significantly shorter at the earliest activation sites than at seven of eight other sites globally distributed over both atria. In all type B successful episodes in which one or more rapid activations occurred after the shock and in 50 of the 72 failed episodes analyzed, activation fronts spread away from the earliest site in a focal pattern, and discrete nonfragmented activation complexes were present in the first derivatives of the electrograms. In the other 22 failed episodes, earliest activation fronts spread in a nonfocal pattern, and earliest postshock electrogram derivatives were fractionated. To better interpret the activation pattern in the fragmented regions, a 504 electrode plaque with 1.5-mm electrode spacing was placed on the right atrial appendage in two additional sheep. In 11 of 108 failed episodes, earliest postshock activation appeared inside the plaque and spread in a focal pattern with nonfragmented electrogram derivatives in 10 episodes and in a reentrant pattern with fragmented electrogram derivatives in the other. Conclusions. (1) The electrode con figuration influenced the location of earliest postshock activation. (2) Earliest postshock activation occurred where the preshock AF cycle length was short. (3) Earliest activations following all type B successful and most failed episodes were not fragmented and spread in a focal pattern. (4) The region of earliest postshock activation in the failed episodes without a focal postshock activation pattern exhibited regions of fragmented electrogram derivatives that may represent conduction block and possibly reentry. © 2004 Heart Rhythm Society. All rights reserved.