The prolongation of intraatrial and interatrial conduction time and the inhomogeneous propagation of sinus impulses have been shown in patients with atrial fibrillation. Recently P wave dispersion (PWD), which is believed to reflect inhomogeneous atrial conduction, has been proposed as being useful for the prediction of paroxysmal atrial fibrillation (PAF). Ninety consecutive patients (46 men, 44 women; aged 55 +/- 13 years) with a history of idiopathic PAF and 70 healthy subjects (42 men, 28 women; mean age 53 +/- 14 years) were studied. The P wave duration was calculated in all 12 leads of the surface EGG. The difference between the maximum and minimum P wave duration was calculated and this difference nias defined as P wave dispersion (PWD = Pmax - Pmin). All patients and controls were also evaluated by echocardiography to measure the left atrial diameter and left ventricular ejection fraction (LVEF). There was no difference between patients and controls in gender (P = 0.26), age (P = 0.12), LVEF (66 +/- 4% vs 67 +/- 5%, P = 0.8) and left atrial diameter (36 +/- 4 mm vs 34 +/- 6 mm, P = 0.13). P maximum duration was found to be significantly higher in patients with a history of PAF (116 +/- 17 ms) than controls (101 +/- 11 ms, P < 0.001). P wave dispersion was also significantly higher in patients than in controls (44 +/- 25 ms vs 27 +/- 10 ms, P < 0.001). There was a weak correlation between age and P wave dispersion (r = 0.27, P < 0.001). A P maximum value of 106 ms separated patients with PAF from control subjects r with a sensitivity of 83%, a specificity of 72%, and a positive predictive accuracy of 79%. A P wave dispersion value of 36 ms separated patients from control subjects with a sensitivity of 77%, a specificity of 82%, and a positive predictive accuracy of 85%. In conclusion, P maxim um duration and P wave dispersion calculated on a standard surface ECG are simple ECG markers that could be used to identify the patients with idiopathic paroxysmal atrial fibrillation.