Atrial fibrilliation (AF) is often combined with advanced age and structural heart disease, conditions known to invite serious proarrhythmic complications of antiarrhythmic drug therapy. Recent controlled trials comparing two AF treatment strategies-rhythm control requiring atrial defibrilliation and antiarrhythmic drugs to prevent AF and ventiricular rate control obviating sinus rhythm maintenance with such drugs-showed equal or superior results with rate control. AF is associated with derepressions of "fetospecific" expression patterns that may profoundly alter the responsiveness of atrial muscle to antiarrhythmic drugs. Therefore, effects of drugs predicted according to pharmacological classifications evaluating drug actions in intact myocardium only should be interpreted cautiously. The classification proposed by Vaughan Williams fails to distinguish between acute and chronic drug efficacy and toxicity as recommended in classical pharmacology. There is, however, overwhelming evidence that acute and chronic drug effects often differ fundamentally. For instance, amiodarone acts acutely as a sodium channel blocker, whereas chronic effects may be mediated by a downregulation of thyroid hormone receptors. Meaningful direct effects of amiodarone on atrial potassium channels is questionable, since the main candidate target-current (IKr) may not be expressed in human atrial muscle. Multiple biophysical factors contribute to the lack of ion channel-selective actions of antiarrhythmic agents. Nonselectivity becomes particularly important in the context of mechanisms of action of Vaughan Williams Class I and III agents on human atrial muscle. Preclinical studies indicate that Class I agents such as flecainide and propafenone may act in AF predominantly as Class III agents. Meta-analyses of antiarrhythmic agents for the prevention of AF have failed to reveal superior drugs or drug classes. Superiority of amiodarone over other agents may depend on arbitrary amiodarone-favoring loading protocols producing significant differential effects exclusively during the acute phase of treatment. In conclusion, the classification of current antiarrhythmic agents into Class I and III may not be a useful simplification when applied to the pharmacotherapy of AF.