Of 204 patients who underwent a selective amygdalohippocampectomy at our hospital, we studied 30 in whom special pre- and postoperative MR imaging was carried out and who were followed up for at least 1 year postoperatively. We measured the total size of the resection and the extent to which the following specific mediobasal temporal lobe structures had been removed: amygdala, hippocampus, pes hippocampi, dentate gyrus, parahippocampal gyrus, uncus and subiculum. Postoperative seizure control was correlated with the overall size of the resection as well as with the resection scores of the above mentioned limbic subcompartments. The mean size of the removed tissue was 7.2 cm3 (range: 2.1-17.7). The mean resection scores of the limbic subcompartments (in percentages) are: amygdala 92%, hippocampus 46%, pes hippocampi 92%, dentate gyrus 45%, parahippocampal gyrus 32%, uncus 92%, subiculum 40%. Although a small resection did not exclude a good outcome, the general tendency was that a better outcome was obtained from a larger resection. With regard to the resection scores of the limbic subcompartments, a positive correlation emerged between good postoperative outcome and the radicality of the removal of the parahippocampal gyrus (and the subiculum, which has been evaluated separately). These findings support our previously formulated amplifier hypothesis for the parahippocampal gyrus.