The filter in situ hybridisation (FISH) method for detection of HPV in cervical swabs was evaluated against the Southern blot technique on concomitant cervical biopsies. Of 73 biopsies, HPV 16 DNA sequences were found in 26 biopsies and HPV 18 sequences in 2 biopsies. Analysis by FISH of the corresponding smears detected 58 and 100% of these, respectively. Of the smears corresponding to the HPV-negative biopsies, 17% were HPV 16-positive and 3% were HPV-18 positive by FISH. Re-hybridisation with cold plasmid added for competition did not change these results. To estimate the risk of spurious hybridisation between vector remnants in the probe and bacterial DNA sequences present in smears, we have hybridised by FISH to preparations of the 19 most common vaginal microorganisms. Of these, E. coli, which is present in about 10% of cervical smears, hybridised strongly with a probe of the plasmid vector pBR322 and may be a significant cause of false positive FISH results. None of the bacteria hybridised with probes of purified HPV when cold, denatured plasmid was added for competition. Analysis by FISH with probes of purified pBR322 to 167 smears of a patient control group resulted in 6% positive reactions. In hybridisations with probes of HPV 16 and 18 to 2 or 3 different filter preparations of the same smear, identical results were obtained in 18 of 19 smears, indicating a good reproducibility by the FISH method. The high percentage of HPV negative smears is equivalent to the rates known from cytology and may reflect sampling errors.