A key question in epilepsy is the organization and size of the neuronal networks necessary for generating seizures. Hypotheses include: a single focal neuronal network drives seizure discharges across the brain, which may or may not be identical with the circuits that generate interictal spikes; or multiple neuronal networks link together in re-entrant loops or other long-range networks. It remains unclear whether any of these hypotheses apply to spontaneous seizures in freely moving animals. We used the tetanus toxin chronic model of epilepsy to test the different predictions made by each hypothesis about the propagation and interaction of epileptic discharges during seizures. Seizures could start in either the injected or noninjected dorsal hippocampus, suggesting that seizures have multifocal onsets in the tetanus toxin model. During seizures, individual bursts propagated in either direction, both between the right and left dorsal hippocampi, and between CA3 and the dentate gyrus in the same hippocampus. These findings argue against one site "driving" seizures or seizures propagating around a limbic loop. Specifically, the side leading each burst switched a median of three times during the first 20 s of a seizure. Analysis of bursts during seizures suggested that the network at each recording site acted like a neuronal oscillator. Coupling of population spikes in right and left CA3 increased during the early part of seizures, but the cross-correlation of their whole-discharge waveforms changed little over the same period. Furthermore, the polarity of the phase difference between population spikes did not follow the phase difference for complete discharges. We concluded that the neuronal aggregate necessary for seizures in our animals comprises multiple spatially distributed neuronal networks and that the increased synchrony of the output (population spike firing) of these networks during the early part of seizures may contribute to seizure generation.