Hippocampal sclerosis involves the selective loss of some hippocampal cell populations, in a process that may disturb the excitatory/inhibitory balance of the remaining cells and produce the epileptic focus. Endfolium sclerosis is the minimal common pathological change found in epileptic patients with hippocampal damage. This subtle lesion is characterized by extensive dentate hilar cell loss without a similarly severe loss of dentate granule cells or hippocampal pyramidal neurons. We attempted to reproduce endfolium sclerosis experimentally by producing dentate granule cell seizure discharges with focal electrical stimulation in anesthetized rats, thus avoiding generalized seizure activity and motor convulsions. With this model, dentate hilar neurons and CA3 pyramidal cells were selectively and irreversibly injured, replicating the pattern of human endfolium sclerosis, with hilar cell damage and survival of dentate granule cell layer GABA-containing basket cells. This results in permanent granule cell disinhibition and hyperexcitability. Excitatory deafferentation of GABAergic basket cells was probably secondary to the loss of hilar mossy cells that normally excite the GABA neurons, rendering these neurons dormant. I propose that endfolium sclerosis in humans represents a selective loss of intrinsically vulnerable dentate hilar cells that normally govern dentate granule cell excitability and that this process leads to epileptiform discharges.