Factors thought to affect the success of and energy requirements for cardioversion of atrial fibrillation were studied in 80 (49 male, 31 female) patients aged 21-88 (mean 61.5 years). Transthoracic impedance was measured in advance of the countershock using a 30 kHz low amplitude AC current passed through self-adhesive ECG/defibrillator pads (diameters 8-12 cm) applied to the chest in the antero-posterior (AP) position in 57 patients and the anteroapical (AA) position in 23 patients. Mean transthoracic impedance for all patients was 69.3 +/- 16 (SD) ohms (range 39-131 ohms), but transthoracic impedance was significantly greater in the AA than the AP position (75.4 +/- 13 vs. 66.7 +/- 16 ohms, p = 0.02). Initial energy was 50 J (delivered) and was gradually increased to a maximum of 360 J if required. Cardioversion was successful in 73 of 80 (91.2 per cent), and low energy shocks (less than or equal to 200 J) were successful in 45 of 80 (56.2 per cent) patients. Using single factor analysis, sex, left atrial enlargement, electrode pad positions, aetiology of atrial fibrillation, presence of left ventricular failure, and prior treatment with verapamil or beta-adrenergic blockers were not significant determinants of cardioversion success or success of low energy shocks but prior treatment with digoxin was, both for cardioversion success and success at low energies. In patients with transthoracic impedance less than or equal to 70 ohms, low energy shocks were more often successful (33 or 50, 66 per cent) than in patients with transthoracic impedance greater than 70 ohms (12 of 30, 40 per cent), p = 0.04. Using univariate analysis, cardioversion success with low energy shocks was not only significantly associated with prior treatment with digoxin but also with the duration of atrial fibrillation (24 hours to one month and one month to three years) and for shocks of less than or equal to 100 J, with prior treatment with amiodarone. Multifactorial linear regression analysis selected, in rank order, only duration of atrial fibrillation of 24 hours to less than one month and one month to three years as significant predictors of both cardioversion success irrespective of shock strength, and success of low energy shocks.