In order to provide a reliable diagnosis for the presence and type of human papillomavirus (HPV) DNA in a case-control study of cervical cancer in Colombia and Spain, 926 cervical scrapes from female subjects were examined by ViraPap (VP) and Southern hybridization (SH), and 510 of these (263 cases and 247 controls) were also tested by polymerase chain reaction (PCR) using the HPV L1 consensus primers. HPV DNA prevalence was much higher in cases than in controls by each of the three tests. There was complete agreement between the results of the three tests for 64.9% of the 510 specimens; 53.5% were negative and 11.4% were positive (regardless of type) by all tests. An additional 29.0% of the specimens were positive by PCR: 19.4% by PCR alone, 6.7% by PCR and VP, and 2.9% by PCR and SH. SH and/or VP gave positive results for 6.0% of the specimens for which the PCR finding was negative: 2.7% by SH alone, 2.5% by VP alone, and 0.8% by both VP and SH. When specimens which were positive by VP alone or only by SH at low-stringency conditions were excluded, PCR confirmed all but four specimens which were positive by other tests. The concordance between type-specific diagnosis by SH and PCR was 86% when HPVs were typed in both tests. HPV-16 accounted for over 80% of the typed HPVs in each test. The presence of blood in case specimens did not appear to inhibit HPV positivity by VP or by PCR at the dilution tested. Low amounts of cellular DNA of specimens resulted in some underestimation of HPV positivity by VP and SH but not by PCR. Compared with that of PCR, the sensitivities for case specimens were 38% by SH and 50% by VP; the sensitivity for control specimens, although it could not be measured precisely because there were few positive specimens, appeared to be lower than for case specimens. It was concluded that PCR-based tests are best suited for epidemiological investigation of HPVs.