Endoscopic mucosal resection (EMR), which is advocated for the treatment of early (superficial) gastroesophageal neoplasms, has also been alluded to represent a superior diagnostic and staging modality. We compared the diagnostic concordance of preceding biopsies with EMR specimens in 31 gastric and 10 esophageal EMRs consisting of 6 low-grade and 12 high-grade dysplasias, 21 intramucosal adenocarcinomas, and 2 submucosal invasive adenocarcinomas. Discrepancies were considered as either major or minor if the histologic grades differed by 2 or more, or by only 1, respectively. Discrepant and concordant cases were compared with regard to the size of lesion (maximum dimension and surface area), number of biopsy fragments, and extent of biopsy sampling (ratio between lesion size and number of biopsy fragments). These same variables were used to evaluate the differences seen between gastric and esophageal cases. Of the 41 cases, 16 (39%) had discrepant diagnoses, including 14 gastric and 2 esophageal neoplasms. A major discrepancy was seen in 2% of the cases (n = 1, gastric) and a minor discrepancy, in 15 cases. All but 2 of the discrepant cases were found to have a higher grade on EMR. The average number of biopsy fragments was 4.4 in both concordant and discrepant groups. The maximal dimension, surface area, and biopsy sampling ratios of the lesion were significantly greater in the discrepant cases than in the concordant cases. The esophageal cases trended toward having smaller size and a significantly extensive biopsy sampling. We conclude that EMR is superior to biopsy for diagnosing superficial gastroesophageal tumors. Discrepancies between the specimens occur in larger lesions (>10 mm) with less extensive biopsy sampling. EMR can substantially modify the diagnostic grade of a lesion and therefore facilitate optimal therapeutic decisions by avoiding undertreatment and overtreatment based on inaccurate grading and staging.