The diagnosis of Barrett's esophagus (BE) requires an accurate recognition of the columnar-lined esophagus at endoscopy. However, a universally accepted standardized endoscopic grading system of BE was lacking prior to the development of the Prague 'circumferential and maximal' criteria. In this system, the landmark for the esophagogastric junction (EGJ) is the proximal end of the gastric folds, not the distal end of the palisade vessels, which are used to endoscopically identify the EGJ in Japan. Although the circumferential and maximal criteria are clinically relevant, an important shortcoming of this system may be failure to identify short-segment BE, a lesion that is found frequently in the Japanese. To compare the diagnostic yield for BE when using the palisade vessels versus gastric folds as a landmark for the EGJ, we evaluated interobserver diagnostic concordance. The endoscopic identification of the EGJ using both landmarks resulted in unacceptably low kappa coefficients of reliability. However, there was a statistically significant improvement after the participants were thoroughly trained in identification of the EGJ during the endoscopic study. Although it remains controversial which landmark is better for the endoscopic diagnosis of BE, it is important to systematically educate and train endoscopists in order to improve diagnostic consistency in patients with BE.