High-grade prostatic intraepithelial neoplasia (PIN) is a strong predictor of carcinoma when identified in small-needle biopsy specimens. However, the diagnostic variability of PIN is unknown. Eight pathologists reviewed 321 prostatic biopsy specimens to assess the variability of the diagnosis of high-grade PIN and carcinoma. All of the specimens were classified as negative, high-grade PIN, suspicious for high-grade PIN, carcinoma, or suspicious for carcinoma, more than one diagnosis was permitted, except for negative. We compared diagnoses made by two observers by pairing them for negative versus high-grade PIN, negative versus carcinoma, high-grade PIN versus carcinoma, and all diagnostic categories together. Mean kappa coefficient values for 28 interobserver combinations were 0.451, 0.845, 0.669, and 0.482, respectively, for each of the four comparison combinations considered. Our results indicate a high level of agreement, "almost perfect" (kappa = 0.81-1.0) for carcinoma, "moderate" (kappa = 0.41-0.60) for high-grade PIN, and "substantial" (kappa = 0.61-0.81) for high-grade PIN versus carcinoma. We found that variability was related to the level of interest in prostatic pathology, the conditions of the study, the subjective application of diagnostic criteria, and the influence of peers and clinical colleagues.