The authors hypothesized that the persistence of abnormal atrial conduction detected by serial P wave signal-averaged electrocardiograms (P-SAECGs) can identify patients who are at high risk of recurrent atrial fibrillation (R-AF) following electrical cardioversion (ECV). P-SAECGs were recorded in 60 consecutive patients after ECV (53 men, age 66 +/- 10 years) and repeated in those who had remained in sinus rhythm (SR) 1 week, and 1, 3, and 6 months later. Filtered P wave duration (PD), root mean square (RMS) voltage of the terminal 40, 30, 20 ms (RMS-40, RMS-30, RMS-20) of the filtered P wave, RMS voltage of the entire filtered P wave (RMS-p), and the integral of the voltages in the entire P wave (integral-p) were analyzed. Thirty-one (52%) patients returned to AF within 1 week, an additional 11 (18%) by 1 month, and a further 2 (3.3%) at each subsequent assessment (3 and 6 months). The patients with R-AF had longer PD (157 +/- 24 vs 143 +/- 17 ms, P < 0.0001) and lower RMS-40, RMS-30, RMS-20 (5.3 +/- 2.0 vs 6.1 +/- 3.4 microV, P = 0.007; 4.3 +/- 1.5 vs 5.7 +/- 3.2 microV, P < 0.0001; 3.6 +/- 1.4 vs 5.2 +/- 3.0 microV, P < 0.0001, respectively) than those who remained in SR. These measurements did not change significantly in either group. RMS-p increased in SR patients (P = 0.009) but decreased in those who subsequently reverted to AF (P = 0.032) with a significant difference between the slopes of the RMS-p change (P = 0.006). Integral-p decreased from the time of ECV in the R-AF group only (P = 0.0028) and created a significant difference between the two slopes (P = 0.0004). The evolution of P-SAECG parameters after ECV differs in patients whose AF recurs versus patients who remain in SR.