Barrett's mucosa is a precancerous condition. The goal of endoscopic surveillance for patients with Barrett's esophagus is to grade dysplasia. The diagnosis of dysplasia is difficult. The aim of this study was to determine the observer variation in diagnosis of dysplasia. Thirty-seven biopsies selected from Barrett's mucosa reported as having dysplasia between 1987 and 1988 were reviewed by 4 pathologists. The biopsies were classified according to the method of Riddell, as no dysplasia, indefinite for dysplasia, low-grade dysplasia and high-grade dysplasia. Each pathologist reviewed the biopsies alone and then together. There was a total agreement between the 4 pathologists in 20 out of 37 biopsies (54%). The pair-wise agreement varied between 65 and 84%. It was 84% when the pathologists belonged to the same institution. Of 222 double reviews made by 6 pairs of pathologists (37 biopsies x 6 pairs of pathologists), there were 58 disagreements: 19 low grade/high grade dysplasia, 2 high grade/indefinite for dysplasia, 1 high grade/no dysplasia, 18 low grade/indefinite for dysplasia, 2 low grade/no dysplasia, 16 indefinite for dysplasia/no dysplasia. At the second review when the pathologists evaluated the biopsies together, the agreement was 92%; there were 3 biopsies on which there was total disagreement: twice between high grade and low grade dysplasia, once between low grade dysplasia, indefinite for dysplasia and no dysplasia. This retrospective study has shown that disagreement between pathologists was rarely found on the biopsies diagnosed as having high grade dysplasia, more often on those diagnosed as having low grade dysplasia or indefinite for dysplasia.(ABSTRACT TRUNCATED AT 250 WORDS)