This study of cervical squamous precursors addressed the consistency with which pathologists could agree on diagnosis using a Bethesda system and the degree to which the classification system discriminated "high-risk" human papillomavirus (HPV) types. Four pathologists independently assessed biopsies of 75 squamous lesions; all contained HPV DNA amplified from archival fixed tissue with polymerase chain reaction (PCR) and typed by restriction digestion of the PCR product. Lesions were categorized as low or high grade using published criteria. In independently performed histologic evaluations a majority (three or more) of observers agreed on the classification of 63 of the 75 cases (84%) with good to very good interobserver (kappa values, 0.43 to 0.63), and fair to excellent intraobserver (kappa values, 0.32 to 0.83) agreement. A majority of the observers classified as high grade 15 of 17 (88%) HPV 16-positive lesions (P < .002), but only 15 of 21 (71%) lesions associated with other high-risk HPV types 18, 31, 35, and 39 (P = .089). Concurrence among observers also varied with HPV type; majority agreement between three or more observers was present for 100% and 94%, respectively, for lesions associated with HPV 6/11 and HPV 16 versus 82% and 76% for lesions associated with HPV 18/31/35/39 and other HPV types. A binary system for grading cervical precursor lesions was applied with good reproducibility among pathologists, and segregated as high-grade virtually all lesions associated with the prototype high-risk HPV (HPV 16). Conversely, other presumed high-risk HPV types, particularly type HPV 18, were not distinguished by this grading scheme and were segregated frequently with low-grade lesions. This finding suggests that variables other than HPV type alone will influence lesion grade. Resolution of these variables will determine whether lesion grade is a more potent biologic parameter than HPV type.