During the years 1982-1989, 2627 women were recruited into eight studies analyzing the relationship between human papillomavirus (HPV) infection and cervical neoplasia. Subsequently, each individual was assigned as either a case or control, and each cervical sample was rescreened for HPV DNA by low-stringency Southern blot hybridization. Positive samples were retested at high stringency with specific probes for HPVs 6/11, 16, 18, 31, 33, 35, 42, 43, 44, 45, 51, 52, 56, and (in most instances) 58. Most cases (153 cancers, 261 high-grade and 377 low-grade squamous intraepithelial lesions) had target or cone biopsies; all 270 borderline atypia subjects and more than 85% of the 1566 normal controls had cytology plus colposcopy/cytology. Scientists performing HPV testing were masked to the clinical diagnoses. Human papillomavirus DNA was detected in 79.3% of specimens from women with definite cervical disease (627 of 791), in 23.7% of borderline atypia subjects (64 of 270), and in 6.4% of normal subjects (101 of 1566). Graphic analysis of odds ratios at each point in the diagnostic spectrum defined four categories: 1) "low risk" (HPVs 6/11, 42, 43, and 44), present in 20.2% (76 of 377) of low-grade lesions but absent in all 153 cancers; 2) "intermediate risk" (HPVs 31, 33, 35, 51, 52, and 58), detected in 23.8% (62 of 261) of high-grade squamous intraepithelial lesions but only 10.5% (16 of 153) of cancers; 3) "high risk/HPV 16," associated with 47.1% of both high-grade intraepithelial lesions (123 of 261) and cancers (72 of 153); and 4) "high risk/HPV 18" (HPVs 18, 45, and 56), found in 26.8% (41 of 153) of invasive carcinomas but only 6.5% (17 of 261) of high-grade intraepithelial lesions. The presence of an oncogenic HPV type conferred relative risks ranging at 65.1-235.7 for the occurrence of a high-grade lesion and 31.1-296.1 for an invasive cancer.