1. The mechanism of generation of epileptiform excitatory postsynaptic potentials (e-EPSPs) induced by bursting activity in vitro was examined in slices of piriform cortex. 2. Previous study revealed that e-EPSPs in piriform cortex are generated in the subjacent endopiriform nucleus, perhaps with a contribution from the claustrum and deep part of layer III of piriform cortex. A puzzling feature of these e-EPSPs was their abrupt origin at long latency with little sign of preceding abnormal activity. 3. Systematic mapping revealed that within spatially restricted regions of the endopiriform nucleus there is an irregular buildup in extracellularly recorded multiunit activity and intracellularly recorded depolarization that precedes the onset of e-EPSPs. Analysis of latency revealed that these "slow-onset" e-EPSPs precede the more widely distributed "abrupt-onset" e-EPSPs, suggesting that they occur at sites of initiation. 4. The hypothesis was tested that the buildup associated with slow-onset e-EPSPs is dependent on synaptically mediated excitation. According to this hypothesis, all-or-none e-EPSPs originate when mutually excitatory (positive feedback) interactions within a population of cells in the endopiriform nucleus become self-regenerative. 5. Predictions from the regenerative positive feedback hypothesis that were successfully verified include the presence of excitatory synaptic connections between cells in the endopiriform nucleus; the consistent prediction of a subsequent e-EPSP from the occurrence of the accelerating buildup in population activity; the occurrence of inhibitory postsynaptic potentials (IPSPs) together with EPSPs during the buildup period; and the blockage of the buildup and e-EPSP by a low concentration of a specific excitatory amino acid antagonist 6,7-dinitroquinoxaline-2,3-dione (DNQX). 6. Blockage of e-EPSPs by a concentration of DNQX that was much less than that required to block monosynaptic EPSPs in the endopiriform nucleus indicates that synaptic reverberation is mediated by alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) type excitatory amino acid receptors. 7. D-2-amino-5-phosphonovaleric acid (D-APV) reduced the duration and amplitude of e-EPSPs but did not block their occurrence, indicating that N-methyl-D-aspartate (NMDA) receptors have a boosting effect on e-EPSPs but are not required for their generation. This is in contrast to the induction of e-EPSPs by bursting activity for which NMDA receptor activation is required. 8. Outside the region of initiation e-EPSPs propagated through the endopiriform nucleus at a velocity of 0.1 m/s.(ABSTRACT TRUNCATED AT 400 WORDS)