Background— Conduction abnormalities serve as a substrate for ventricular fibrillation (VF) in patients with Brugada syndrome (BS). Signal-averaged electrograms can detect late potentials, but the significance of conduction abnormalities within the QRS complex is still unknown. The latter can present as multiple spikes within the QRS complex (fragmented QRS [f-QRS]). We hypothesized that f-QRS could indicate a substrate for VF and might predict a high risk of VF for patients with BS. Methods and Results— In study 1, we analyzed the incidence of f-QRS in 115 patients with BS (13 resuscitated from VF, 28 with syncope, and 74 asymptomatic). f-QRS was observed in 43% of patients, more often in the VF group (incidence of f-QRS: VF 85%, syncope 50%, and asymptomatic 34%, P <0.01). SCN5A mutations occurred more often in patients with f-QRS (33%) than in patients without f-QRS (5%). In patients with syncope or VF, only 6% without f-QRS experienced VF during follow-up (43±25 months), but 58% of patients with f-QRS had recurrent syncope due to VF ( P <0.01). In study 2, to investigate the mechanism of f-QRS, we studied in vitro models of BS in canine right ventricular tissues (n=4) and optically mapped multisite action potentials. In the experimental model of BS, ST elevation resulted from a large phase 1 notch of the action potential in the epicardium, and local epicardial activation delay reproduced f-QRS in the transmural ECG. Conclusions— f-QRS appears to be a marker for the substrate for spontaneous VF in BS and predicts patients at high risk of syncope.