The present study was designed to test the hypothesis that chronic gamma-aminobutyric acid (GABA) disinhibition of granule cells could explain permanent kindled epileptogenicity. Quantitative and statistical comparisons of glutamate decarboxylase immunoreactivity (GAD-IR), the synthesizing enzyme for GABA, were made of GAD-IR cells and puncta in stratum granulosum of the fascia dentata. The use of GAD immunocytochemistry in kindled and control tissue was used to allow direct anatomic confirmation that we were measuring changes in GAD-IR which would represent GABA synthesis for release by the recurrent inhibitory system of the fascia dentata. Immediately after the last kindled seizure, optically detected GAD-IR puncta densities were significantly reduced in stratum granulosum. At 3 or 7 days after the last kindled seizure, GAD-IR was normal in puncta, indicating that the transient GAD-IR loss was probably a metabolic response to the recent seizure represented by over-use of GAD needed for synthesis of GABA after a prolonged kindled seizure. When the prolonged kindled seizures were discontinued GAD-IR recovered in the puncta. This transient effect did not occur in other areas such as Ammon's horn (CA3) or substantia nigra. The extent of the GAD-IR loss showed no correlation with the severity of the final behavioral seizure (R = 0.23), or the final afterdischarge (AD) duration in entorhinal cortex (R = 0.17) or motor cortex (R = 0.53). A massed stimulation control group given 19 shorter-duration ADs every 10 min (non-kindling) did not reduce GAD-IR. These findings support the hypothetical model that prolonged kindled seizures release excessive GABA which depletes GAD in axon terminals for 1 day after the seizure. However, such a transient suppression of GAD-IR provides no evidence that disinhibition contributes to the kindling process, because kindling proceeds normally with inter-seizure intervals as long as 1 week. The finding of full recovery of GAD-IR within 1 week does not support the model of loss of GABA inhibition to explain the permanency of kindled epileptogenesis.