The "kindling" phenomenon is associated with long-lasting facilitation of synaptic transmission. A possible mechanism of such facilitation could involve changes in the number of synaptic contacts. However, previous attempts to demonstrate a synaptic morphological alteration that could account for the long-term effects ofkindling had failed, possibly due to the unavailability, at the time, of unbiased methods for synapse quantitation. Using the unbiased stereological disector technique, we estimated the number of synapses per neuron in the middle molecular layer of the hippocampal dentate gyrus in rats kindled by electrical stimulation of the medial perforant path with implanted electrodes. Unkindled but stimulated (coulodibit control) and unstimulated but implanted rats served as controls. Animals were coded and killed 4 weeks after reaching the kindling criterion of five generalized seizures. The most important results were obtained when axospinous synapses with continuous or discontinuous postsynaptic densities ("nonperforated" or "perforated" synapses) were differentially analyzed. Kindling resulted in a selective loss of nonperforated synaptic contacts in contrast to preservation of perforated ones. Furthermore, the ratio of perforated to nonperforated synapses was increased by 45% or 40% in kindled rats relative to unstimulated or coulombic controls, respectively. These findings suggest that synaptic efficacy may depend on a balance of the two synaptic types; selective elimination of nonperforated synapses may augment the potency ofremaining synaptic contacts, a process reminiscent of synaptic remodeling during development."Kindling" may be viewed as an experimental model of the formation of an engram in the brain (1, 2). It differs from all other models of synaptic plasticity (or neuronal models of memory) by reason of its extraordinary duration. After only a few exposures to a low-level, localized electrical stimulus, the synaptic responsiveness of the stimulated circuit undergoes an augmentation that persists, without further reinforcement, for many months.In the kindling paradigm (3, 4), a subthreshold electrical stimulus is repeatedly delivered to a local brain area once or twice a day for a duration of 1-2 sec. When first applied, the electrical stimulation results in only brief afterdischarge (AD) and no behavioral alteration. Without any change in stimulation parameters, the AD gradually increases in duration and spreads from the stimulated area to increasingly distant, though synaptically connected, brain regions. A progressive alteration of behavior is also seen, beginning with momentary arrest of ongoing locomotor activity and proceeding through localized twitching to a generalized seizure. Once generalized seizures have occurred, cessation of stimulation for weeks, months, or even years does not result in loss of the newly acquired, electrophysiologically and behaviorally defined change (2, 5, 6). Reintroduction of the original stimulus, which was behaviorally ineffective to begin with, ...