Chemoradiation therapy is used widely for locoregional esophageal cancer. Patients with persistent disease may benefit from surgery. Preoperative esophagoscopy can identify persistent tumor but its accuracy is uncertain. The primary objective of this study is to assess the extent of agreement between esophagoscopy and surgical pathology in patients treated with neoadjuvant chemoradiation. A retrospective chart review of patients who underwent chemoradiation, preoperative endoscopy and surgery from January 1996 to December 2002 was performed. Cohen's kappa statistic was used to measure the degree of agreement between findings at endoscopic biopsy and surgical pathology. Thirty cases were identified. All patients received chemoradiation followed by surgical resection. There was insufficient agreement between tumor size (kappa 0.25, standard error 0.17, P = 0.07) and appearance (kappa 0.19, standard error 0.18, P = 0.14). Preoperative endoscopy revealed atypia/inflammation in 15 cases and dysplasia in eight. Of these 23 cases, 11 were adenocarcinomas at surgery. Only nine patients had concurrence between surgical pathology and endoscopy. The positive and negative predictive values of esophagoscopy for identifying residual tumor were 100% and 11%, respectively. Our data suggests that after chemoradiation, esophagoscopy is unreliable for excluding residual disease. The roles of other modalities need to be explored.