The experience gained using intracardiac cardioverter defibrillators for the treatment of ventricular arrhythmias has prompted the development of an automatic atrial defibrillator capable of detecting and automatically terminating atrial fibrillation (AF). Experimental studies in sheep have shown that it is possible to terminate AF with energies ranging from < 1 to 7 joules [J], using biphasic shocks. The best electrode configuration using intracardiac catheters and/or a subcutaneous patch was two catheters, one in the right atrium and the other in the coronary sinus. Current studies in man focus on the answers to three questions. First, can the experimental results of atrial defibrillation derived from healthy anesthetized sheep without spontaneous AF be extrapolated to AF in man with areas of fibrosis within the atria and/or underlying heart disease in 80% of cases? Preliminary studies in man suggest that cardioversion of AF of short duration is feasible using a mean energy of 2 J. Second, are these energies well tolerated in an awake nonsedated patient? Energies < 1 J were well tolerated, but pain resulting from higher energies needs further investigation. Third, is low-energy atrial defibrillation safe, i.e., is there a risk of ventricular arrhythmias induced by an atrial shock? Experimental results in sheep have shown that the risk of R wave synchronized shock to induce ventricular arrhythmias was only present when the preceding RR interval was shorter than 300 msec. The risk of proarrhythmia in man is undergoing evaluation and must be sufficiently low (< 0.1) before sanctioning implantation of a stand-alone (without associated ventricular defibrillator) automatic atrial defibrillator. Preliminary data on 1212 shocks showed no proarrhythmia.(ABSTRACT TRUNCATED AT 250 WORDS)