In advanced gastric and gastroesophageal junction (GEJ) adenocarcinomas that overexpress human epidermal growth factor receptor 2 (HER2), treatment with trastuzumab confers a survival benefit. To select patients for treatment, HER2 status is evaluated by immunohistochemistry (IHC) and in situ hybridization. Gastric and GEJ adenocarcinomas demonstrate heterogeneity in HER2 expression. Nonetheless, testing is often performed on biopsies alone, which raises the issue of nonrepresentative sampling. We investigated the correlation of HER2 status between matched biopsy and resection specimens and the role of tumor heterogeneity in contributing to discrepancy. A total of 128 patients with gastric or GEJ adenocarcinoma had tissue available from a biopsy and subsequent resection. HER2 IHC was performed and evaluated by the criteria used in the Trastuzumab for Gastric Cancer clinical trial. In situ hybridization was performed if IHC was equivocal (2+) in either the biopsy or resection and in discrepant cases. Tumor heterogeneity was defined as 3+ or 2+ staining in 10% to 60% of tumor cells. Overall, HER2 was overexpressed in 18 tumors (14%), with a biopsy-resection concordance of 96.1%. Five cases were discrepant; 2 were positive on biopsy only, and 3 were positive on resection only. Tumor heterogeneity was seen in 80% of discrepant biopsies and resections, compared with 24% of concordant cases (P = .016). Our study demonstrates strong concordance between biopsy and resection specimens for HER2 overexpression in gastric cancer. Discordance was correlated with tumor heterogeneity. Overall, both biopsy and resection specimens are appropriate for HER2 testing, but generous sampling for biopsy specimens is necessary to ensure accurate assessment.