The value of signal-averaged P-wave electrocardiogram and echocardiography for predicting recurrent atrial tachyarrhythmias was prospectively investigated in 60 patients presenting with paroxysmal atrial fibrillation or flutter. All patients were followed up for 1 year after restoration of sinus rhythm. A stepwise discriminant function analysis was used to identify variables predicting recurrent atrial tachyarrhythmias. Analyzed variables included signal-averaged P-wave duration in 3 bipolar orthogonal leads (X,Y,Z) and their vector magnitude, as well as left and right atrial dimensions and volumes. During follow-up, 25 patients had recurrent atrial tachyarrhythmias, while 35 did not. Using discriminant function analysis, the left atrial antero-superior dimension was found to be the only variable predicting the recurrence of atrial tachyarrhythmias (p < 0.0038) and was able to correctly classify 65% of the study patients. It was concluded that, in patients with paroxysmal atrial fibrillation or flutter, the traditionally used determination of left atrial dimension was the variable most closely associated with a high risk for recurrent tachyarrhythmias. The signal-averaged P-wave duration did not improve tachyarrhythmia prediction.