Three pathologists evaluated a number of designated architectural features to assign grades to 41 cases of well- to moderately differentiated adenocarcinoma, and their opinions were compared. The consensus opinion was obtained and evaluated against objective measurements of glandular architecture that were obtained by morphometric techniques. The observers agreed on gland size, gland uniformity, and the number of glands per field in only 49%, 31%, and 39% of cases, respectively. There were significant differences in the Gleason grades assigned by observers. Paired matching of individual Gleason grades showed agreement among observers in 44% (18 of 41), 56% (23 of 41), and 75% (31 of 41) of cases, respectively. This level of interobserver disagreement occurred even though cases with predominant patterns were selected carefully and those with variable patterns were excluded. A direct relationship appears to exist between increasing Gleason grade and increasing glandular variability, and there is an inverse relationship between Gleason grade, gland lumen area, and the number of glandular nuclei, as assessed by a group of pathologists.