Esophageal autopsy specimens from 11 subjects were used to determine observer variation and sampling error in assessing alterations attributed to gastroesophageal reflux. Inter- and intraobserver variation exceeded 20% even when the diagnosis was limited to a normal and abnormal reading. Marked differences in basal cell thickness but not in papillary height occurred when specimens were obtained from different levels of the lower esophagus. The differences were less marked in specimens obtained from the same level. Based on these data the reliability of basal cell thickness and papillary height as an index of gastroesophageal reflux appears limited.