AIMS--To assess the relative diagnostic performance of the polymerase chain reaction (PCR) and non-isotopic in situ hybridisation (NISH) and to correlate these data with cytopathological assessment. METHODS--Paired analysis of human papillomavirus (HPV) detection was performed by PCR and NISH on exfoliated cervical cells from 122 women attending a routine gynaecological examination. PCR amplification followed by generic and HPV type specific hybridisation was compared with NISH on a parallel cervical smear. RESULTS--Overall, 32 cases were positive by NISH and 61 positive by PCR. Of the 105 cases in which both PCR and NISH were interpretable, 76 (26%) were normal smears, 20 of which were HPV positive by NISH and 37 (49%) by PCR. Of 17 borderline smears, two were NISH positive and 12 PCR positive. Eight of nine smears containing koilocytes were positive by NISH and seven by PCR. Of three dyskaryotic smears, none were NISH and two were PCR positive. The concordance of NISH and PCR in these samples was 57%. To assess sampling error, NISH and PCR were performed on an additional 50 cases using aliquots from the same sample. This increased the concordance between assays to 74%. Filter hybridisation of PCR products with the cocktail of probes used in NISH (under low and high stringency conditions) demonstrated that several cases of NISH positivity could be accounted for by cross-hybridisation to HPV types identified by PCR but not present in the NISH probe cocktail. CONCLUSIONS--Sampling error and potential cross-hybridisation of probe and target should be considered in interpretation of these techniques. PCR is more sensitive because it provides for the amplification of target DNA sequences. In addition, the PCR assay utilised in this study detects a wider range of HPV types than are contained in the cocktails used for NISH. However, PCR assays detect viral DNA present both within cells and in cervical fluid whereas NISH permits morphological localisation.