The development of an accurate method for the detection and typing of genital human papillomavirus is of substantial clinical importance. This virus has been implicated as an etiologic agent in the development of cervical neoplasia. To detect human papillomavirus infection with maximum sensitivity, cells must be collected and assayed for human papillomavirus deoxyribonucleic acid. We compared two noninvasive methods of sampling exfoliated cervical cells--cervicovaginal lavage and scrape-Cytobrush. Seventy-four patients newly referred to the colposcopy clinic were divided randomly for cell sampling by either cervicovaginal lavage followed by scrape-Cytobrush or, conversely, scrape-Cytobrush followed by cervicovaginal lavage. Restriction analysis and Southern blot hybridization were used to test all the samples thus obtained for human papillomavirus. Overall, test results from 42 patients (56.8%) were positive for human papillomavirus deoxyribonucleic acid. Twenty-six (31.1%) tested positive for human papillomavirus by both sampling methods, and 32 (43.2%) tested negative for human papillomavirus by both methods. One (1.4%) tested positive with scrape-Cytobrush sampling but negative with cervicovaginal lavage, while 15 (20.3%) tested negative with scrape-Cytobrush but positive with cervicovaginal lavage (p less than 0.001, McNemar's test). These data, combined with previous work from our group, suggest that, of the available methods, cervicovaginal lavage, coupled with human papillomavirus deoxyribonucleic acid hybridization, is the most sensitive noninvasive method for harvesting cells for molecular identification of human papillomavirus in the female lower genital tract.