Human papillomavirus (HPV) infection has been strongly implicated as a cause of genital neoplasia. Although Southern blot DNA hybridization techniques are regarded as the most accurate means of identifying HPV infection, studies using this technique to measure infection have provided varying estimates of the prevalence of HPV among healthy and diseased groups. This investigation provides a possible explanation for study-to-study differences, by demonstrating inter-laboratory variability in the detection and typing of HPV by Southern blot. Four experienced laboratories tested 40 identical, masked samples of DNA extracted from presumed infected and non-infected tissues. The pairwise percentage agreement between the laboratories in judging samples as HPV-positive or negative ranged from 66 to 97%. Among samples judged to contain HPV, agreement as to type ranged from 77 to 96%. We conclude that inter-laboratory differences are an important consideration in any discussion of HPV prevalence estimates.