AIMS: To determine the relative diagnostic sensitivity of non-isotopic in situ hybridisation (NISH) for the diagnosis of human papillomavirus (HPV) on matched smears and biopsy specimens; to compare the NISH signal type in the two samples; and to correlate the NISH data with the morphological diagnosis. METHODS: HPV samples were assayed individually by NISH with digoxigenin labelled probes (HPV6, 11, 16, 18, and 33) on routinely collected paraffin wax embedded cervical biopsy specimens and for high risk HPVs with a cocktail of similarly labelled probes (HPV16, 18, 33) on matched smears. These were taken at the same colposcopic examination from 32 patients investigated for an abnormal cervical Papanicolaou (PAP) stained smear. RESULTS: An HPV signal was present in 18 (56%) biopsy specimens and in 14 (44%) smears. There was higher concordance of sets of data in the presence of cytopathic wart virus changes. The superiority of biopsy over smear in detecting HPV was mainly the result of examining the entire cervical biopsy specimen rather than cells scraped from the cervical surface. The NISH signal type in both biopsy specimen and smear was similar; it has been shown that NISH type 1 signal correlates with episomal viral replication and type 2 and 3 signals with viral integration. CONCLUSIONS: These data show that NISH on cervical smears is a worthwhile primary screen for HPV infection. The NISH signal types in cervical smears are similar to those previously described in cervical biopsy specimens.