The subjects of this study are 445 patients with advanced cervical cancer treated by standardized radiation therapy. Upon entry into one of two Gynecologic Oncology Group (GOG) protocols, original pathologic diagnoses and histologic tumor descriptions for each patient were compared with separate evaluations made by a consensus opinion of two GOG pathologists. A review diagnosis using grade, cell type, and the Stendahl scoring system was then made by the first author (R.J.S.) without knowledge of the prior diagnoses. Of the original pathologists' diagnoses, 21% did not include grade or cell type. There was little agreement among the different pathologists as to the use of either specific grade or cell type. Histologic grade, irrespective of the pathologists making the diagnosis, had no correlation to prognosis. The Reagan and Wentz large-cell keratinizing (LCK) cell type, when applied by the author to tumors with any form of squamous keratinization present, identified a group of patients with a poorer prognosis, although not independently of other prognostic factors. The Stendahl scoring system identified a number of patients with both a poorer and better prognosis. This was statistically significant and independent of other risk factors. A major limitation, however, was the number of patients evaluable because of inadequate biopsy material in 23.6% of the study group.