The mechanism(s) underlying ventricular fibrillation (VF) remain unclear. We hypothesized that at least some forms of VF are not random and that high-frequency periodic sources of activity manifest themselves as spatiotemporal periodicities, which drive VF. Twenty-four VF episodes from 8 Langendorff-perfused rabbit hearts were studied using high-resolution video imaging in conjunction with ECG recordings and spectral analysis. Sequential wavefronts that activated the ventricles in a spatially and temporally periodic fashion were identified. In addition, we analyzed the lifespan and dynamics of wavelets in VF, using a new method of phase mapping that enables identification of phase singularity points (PSs), which flank individual wavelets. Spatiotemporal periodicity was found in 21 of 24 episodes. Complete reentry on the epicardial surface was observed in 3 of 24 episodes. The cycle length of discrete regions of spatiotemporal periodicity correlated highly with the dominant frequency of the optical pseudo-ECG (R2=0.75) and with the global bipolar electrogram (R2=0.79). The lifespan of PSs was short (14.7±14.4 ms); 98% of PSs existed for <1 rotation. The mean number of waves entering (6.50±0.69) exceeded the mean number of waves that exited our mapping field (4.25±0.56;P<0.05). These results strongly suggest that ongoing stable sources are responsible for the majority of the frequency content of VF and therefore play a role in its maintenance. In this model, multiple wavelets resulting from wavebreaks do not appear to be responsible for the sustenance of this arrhythmia, but are rather the consequence of breakup of high-frequency activation from a dominant reentrant source.