Although therapeutic and toxic serum concentrations of digoxin have been established, there is sparse information permitting correlation of drug level with clinical effect. This study was undertaken to assess the radioimmunoassay serum digoxin levels in 30 patients with acute atrial fibrillation (38 determinations) and 30 patients with chronic atrial fibrillation (54 determinations). Those with chronic fibrillation were subdivided into those in clinically stable condition (14 patients), and those seriously ill and in clinically unstable condition (16 patients). Slowing of ventricular rate in patients with stable, chronic atrial fibrillation was accomplished in 10 of 16 instances by "therapeutic" and "subtherapeutic" levels of digoxin (less than 2 ng/ml). Ventricular rate was "controlled" (65 to 95 beats/min) with therapeutic levels of serum digoxin in only five instances of acute atrial fibrillation and seven of unstable chronic atrial fibrillation. In 43 studies (23 of acute atrial fibrillation, 20 of chronic atrial fibrillation), a rapid ventricular rate (95 to 140 beats/min) persisted in the presence of "therapeutic" or high levels of digoxin. Thirty-nine of these were in patients who were seriously ill with conditions such as infection, hypoxia or recent thoracotomy. Slowing of the ventricular rate required "toxic" concentrations of digoxin (2.5 to 6 ng/ml) in 15 instances. We conclude that sufficient amounts of digoxin to achieve "therapeutic" serum concentrations may fall to lower the ventricular rate in atrial fibrillation to less than 100 beats/min, especially when a serious, complicating illness coexists.