Background: It has been shown that most paroxysmal atrial fibrillation (AF) can be terminated by pulmonary vein (PV) isolation alone, suggesting that rapid discharges from PV drive AF. To define the driving mechanism of AF, we compared the activation sequence in the body of left atrium (LA) to that within PV. Methods: Endocardial noncontact mapping of LA body (LA group; n = 16) and selective endocardial mapping of left superior PV (LSPV) (PV group; n = 13) were performed in 29 paroxysmal AF patients. The frequency of pivoting activation, wave breakup, and wave fusion observed in LA were compared to those in LSPV to define the driving mechanism of AF. Circumferential ablation lesion around left PV was performed after right PV isolation to examine the effect of linear lesion around PV on AF termination both in LA and PV groups. Results: The frequency of pivoting activation, wave breakup, and wave fusion in PV group were significantly higher than those in LA group (36.5 AE 17.7 vs 5.0 AE 2.2 times/seconds, p < 0.001, 10.1 AE 4.3 vs 5.0 AE 2.2 times/seconds, p = 0.004, 18.1 AE 5.7 vs 11.0 AE 5.2, p = 0.002). Especially in the PV group, the frequency of pivoting activation was significantly higher than that of wave breakup and wave fusion (36.5 AE 17.7 vs 10.1 AE 4.3 times/seconds, p < 0.001, 36.5 AE 17.7 vs 18.1 AE 5.7 times/seconds, p < 0.001). These disorganized activations in LSPV were eliminated by the circumferential ablation lesion around left PV (pivoting activation; 36.5 AE 17.7 vs 9.3 AE 2.3 times/seconds, p < 0.001, wave breakup; 10.1AE1.3 times/ seconds, p = 0.003, wave fusion; 18.1 AE 5.7 vs 5.7 AE 1.8, p < 0.001), resulted in AF termination in all patients in both LA and PV groups. Conclusions: Activation sequence within PV was more disorganized than that in LA body. Frequent episodes of pivoting activation rather than wave breakup and fusion observed within PV acted as the driving sources of paroxysmal AF.