Background: A new endoscopic classification of gastroesophageal reflux disease (GERD) has been proposed, and the term mucosal break has been introduced to describe mucosal damage. This new classification was evaluated by endoscopists with different levels of experience. Methods: Fifty endoscopic photographs for each of 20 randomly selected patients with GERD were assessed retrospectively by three groups of seven endoscopists classified by experience: group 1 (100 to 500 procedures), group 2 (500 to 3000), and group 3 (more than 3000). The new classification was modified by adding grade 0 to describe healed mucosal breaks, so that there were five grades. All photographs were assessed twice at an interval of more than 1 week, and κ statistics were used to determine observer variation. Results: Interobserver variation within group 3 (κ = 0.39, n = 21) and between groups 3 and 2 (κ = 0.36, n = 49) was significantly different (p < 0.01) from that between groups 3 and 1 (κ = 0.26, n = 49). Intraobserver variation in group 1 (κ = 0.39, n = 7) was significantly different (p < 0.01) from that in group 2 (κ = 0.51, n = 7) and group 3 (κ = 0.54, n = 7). Conclusions: Observer variation depends on level of endoscopic experience. Only experienced endoscopists should use the new classification for grading of GERD. (Gastrointest Endosc 1999;49:700-4.)