Intracerebral or intraperitoneal injections of kainic acid, an agonist at a class of glutamate receptors, have been extensively used to model temporal lobe epilepsy. In the present study we compared the types and distributions of selectively vulnerable neurons in the ipsi- and contralateral hippocampi following unilateral kainate injections into the CA3 subfield in order to examine whether "proximal" or "distant" neuronal damage resembled the pathology, and possibly also the mechanism, of human temporal lobe epilepsy. The degeneration of principal cells in the different hippocampal subfields was visualized by silver impregnation, and the loss of various types of non-principal cells was studied by immunostaining for the calcium binding proteins parvalbumin, calbindin-D28k and calretinin, as well as for somatostatin. In the first series of experiments various concentrations (ranging from 0.1 to 1 mg/ml) and volumes (0.5-2 microliters) of kainate were tested to induce reproducible damage in the contralateral hippocampus. The optimal dose, employed in the subsequent vulnerability studies, was found to be 3 x 0.5-microliter injections (over a period of 10 min) of a concentration of 0.33 mg/ml under ether anaesthesia, which was discontinued immediately after injection. Anaesthesia with equithesin was found to prevent contralateral cell death. Most if not all pyramidal cells in the CA3 region degenerated on the ipsilateral side, whereas the dentate granule cells, and the majority of CA1 pyramidal cells were resistant. A strikingly different pattern was found on the contralateral side, where CA1 pyramidal cells were almost completely lost, but the CA3 region (with the exception of CA3c) and the dentate gyrus remained intact. Three subpopulations of non-principal cells were found to be vulnerable in both hemispheres, the hilar somatostatin cells, spiny calretinin cells and mossy cells, as well as the spiny calretinin cells in stratum lucidum of CA3. The other subpopulations were resistant, except for those within the effective injection site. We propose that the "distant" (contralateral) damage resembles the pattern, and probably also the mechanism, of cell death in human temporal lobe epilepsy, whereas the ipsilateral damage does not.