The purpose of this study was to improve the immunogold labeling of epoxy sections and to increase our knowledge of the mechanism for how antigens become immunolabeled on resin sections. Tissues from pancreas, thyroid and fibrin clots were embedded in an epoxy resin and LR-White. The epoxy mixture was composed and treated in different ways, especially with respect to altered amounts of accelerator (DMP-30). Immunogold labeling was performed with anti-glucagon, anti-thyroglobulin and anti-fibrinogen respectively. By increasing the amount of DMP-30 in the infiltration steps and/or embedding step, we observed a significant rise in the immunogold labeling. For the largest proteins the labeling was up to 8 times more intense than the labeling achieve with epoxy sections produced by 'normal' amount of accelerator in the embedding mixture and without accelerator in the infiltration mixture. For the smallest protein, glucagon, the differences were almost absent. The labeling of thyroglobulin and fibrinogen on the high accelerator epoxy sections was up to 70% of the labeling of LR-White sections, while conventional epoxy sections showed a labeling of 5-10% of that obtained with acrylic labeling. The cutting qualities of the high-accelerator blocks were similar to that of conventional epoxy embedding. The ultrastructure of the sections from the high-accelerator epoxy blocks were good, and the contrast was improved when tannic acid was used as enhancer. Our theory to explain the improved labeling is that the antigens are less tightly incorporated in the polymer network when the concentration of the accelerator is increased. The method outlined significantly improves the detectability of antigens on epoxy sections, which is the embedding resin routinely used in many laboratories.