We examined epidermal growth factor receptor (EGFR) overexpression and EGFR gene amplification using immunohistochemistry (IHC) and chromogenic in situ hybridization (CISH) in 109 glioblastomas, including 98 primary glioblastomas and 11 secondary glioblastomas. EGFR overexpression and EGFR gene amplification were found in 33% and 24% of glioblastoma, respectively, and all of those cases were primary glioblastoma. Large ischemic necrosis was significantly more frequent in primary glioblastomas than in secondary glioblastomas (54% vs. 18%), but pseudopalisading necrosis was not (65% vs. 54%). EGFR gene amplification was detected significantly more frequently in cases with both types of necrosis. Although glioblastomas with EGFR gene amplification invariably exhibited EGFR overexpression at the level of the whole tumor, tumor cells with EGFR gene amplification did not always show EGFR overexpression at the level of individual tumor cells. Cases of "strong" EGFR overexpression on IHC could be regarded as having EGFR gene amplification, and cases without EGFR overexpression could not. Cases of "weak" EGFR overexpression should be tested with CISH to confirm the presence of EGFR gene amplification. We found that 54% of glioblastomas with EGFR gene amplification were composed of areas with and without EGFR gene amplification; however, there were no obvious differences in morphology between tumor cells with and without EGFR gene amplification. Although small cell architecture might be associated with EGFR gene amplification at the level of the whole tumor, it did not always suggest amplification of the EGFR gene at the level of individual tumor cells. In one case, it seemed to suggest that a clone with EGFR gene amplification may arise in pre-existing tumor tissue and extend into the surrounding area. In cases of overall EGFR amplification, CISH would be a useful tool to decide the tumor border in areas infiltrated by tumor cells.