AIMS: To compare the use of biotinylated and digoxigenin labelled probes for diagnosis of human fetal parvovirus B19 infection in formalin fixed, paraffin wax embedded tissues; and to assess the cellular distribution of the virus in positive cases. METHODS: Sections of lung tissue from 23 cases of anatomically normal non-immune fetal hydrops presenting between 1984 and 1989, and from 13 control cases of hydrops due to chromosomal abnormality were probed for B19 DNA by in situ hybridisation using both biotinylated and digoxigenin labelled probes. The distribution of the virus was then investigated in all cases of fetal B19 infection confirmed in this laboratory to date (n = 11) by combining in situ hybridisation for viral DNA (using the digoxigenin system) with immunohistological labelling for a range of cellular antigens. RESULTS: Five unequivocal cases of B19 infection were identified among the 23 fetuses with unexplained hydrops using both probe labels. When combined with data from previous studies of the period 1974-1983, the results indicate that B19 infection was responsible for 27% of cases of anatomically normal non-immune hydrops and 8% of all cases, of non-immune hydrops presenting to this hospital over 15 years. False positive signal was seen in an additional three cases, using biotinylated probes. Digoxigenin labelled probes gave greater specificity and permitted detailed investigation of tissues high in endogenous biotin. Though most cells containing B19 DNA colabelled as erythroid precursors, viral DNA was frequently detected within mononuclear-phagocytic cells. In three cases viral signal was also found within occasional myocardial cells labelled by antibody to desmin. CONCLUSIONS: A relatively high proportion of cases of anatomically normal, non-immune hydrops are caused by B19 infection. Digoxigenin is a more reliable probe label than biotin for in situ hybridisation in archival fetal tissues. Double labelling for cellular antigens and viral nucleic acid is a powerful technique for investigating virus-host cell interactions, and provides evidence that cell types other than those of erythroid lineage may have a role in human fetal parvovirus infection.