Few techniques have been developed for deriving quantitative measures of activation patterns during ventricular fibrillation (VF). Such measures have many potential applications, for example, assessing the effects of time, drugs, or electrical interventions. We have developed a new framework for quantifying VF patterns as mapped from an array of approximately 500 unipolar electrodes. Individual activation wavefronts are isolated from one another using an algorithm that groups together adjacent active electrogram samples (dV/dt < -0.5 V/sec). Contacts between wavefronts are detected: these include fractionations, in which a single wavefront breaks into multiple wavefronts, and collisions, in which multiple wavefronts coalesce to form a new wavefront. The timing and contact relationships between wavefronts are summarized as a directed graph. From this model of the VF episode, we derive several parameters: number of wavefronts, number of fractionations, number of collisions, mean wavefront size, mean area swept out, and mean duration. As an example of this analysis, we computed these parameters in six open-chest pigs at 5, 10, 15, and 20 sec after electrical induction of VF. The number of wavefronts and the number of collisions decreased, whereas the mean wavefront size and mean area swept out increased during this period. These results are consistent with previous studies showing a recovery of organization during the first minute of VF.