Atrial fibrillation is a daily cardiological problem which poses three types of questions, which, though old, are only partially mastered: anticoagulation, reduction and prevention of recurrence. It is a potent source of embolism. The risk is the greatest in patients with rheumatic valvular disease when the fibrillation is recent and when underlying cardiac disease is uncompensated. Long term anticoagulation is mandatory when the cause is rheumatic heart disease. In other pathologies, though anticoagulation has not been shown to reduce mortality, it significantly reduces the number of cerebrovascular accidents, including in the elderly and with low-dose vitamin K antagonist drugs. The efficacy of anticoagulation in preventing arterial embolism has not been established. Reduction of atrial fibrillation is not essential if the arrhythmia is well tolerated, chronic, especially in elderly patients and when several recurrences have occurred despite preventive therapy. In other cases, medical reduction is to be preferred to cardioversion if the fibrillation is recent and well tolerated. Of the oral and injectable preparations, amiodarone seems to be the drug with best benefit/risk ratio. Prevention of recurrence of fibrillation is unnecessary for many after a first episode, especially when idiopathic. In other cases, there are many available drugs but results are uncertain except in those observed in atrial fibrillation related to the autonomic nervous system. Strictly controlled and statistically exploitable studies show comparable efficacy of quinine and other Class I drugs. Beta-blockers are not very useful and the excellent long term results with amiodarone require confirmation.(ABSTRACT TRUNCATED AT 250 WORDS)