An analysis of the observer variation between 4 endoscopists was made assessing the frequency of visual interpretative discrepancies in the areas of the esophagus, stomach and duodenum. The observer variation in relation to the experience of the endoscopist was compared. The trial was carried out in 104 patients. Two observers were staff members with a minimum of 5 years experience in endoscopy. The other two were fellows with one year experience in endoscopy. The procedure was carried out by one endoscopist while the other observed the procedure through the teaching attachment. The findings were immediately recorded on a protocol. Discrepancies observed were classified in two categories: "Major" and "Minor" discrepancies. Results were evaluated in subgroups according to the different endoscopist's training (staff vs fellow, staff vs staff, fellow vs fellow). An analysis of the lesions where discrepancies occurred was made. The minor discrepancies per case was 2.18 +/- 1.32. Major discrepancies were reduced to 1.19 +/- 1.08. There was a significantly decreased incidence of minor and major discrepancies in the comparison of the staff vs staff as opposed to cases performed by fellow vs fellow (p less than 0.01 and p less than 0.005). The diagnoses where discrepancies occurred most often in order of decreasing frequency included. Esophagitis, hiatal hernia, gastritis, gastric ulcer, erosions and duodenal mucosal lesions.