The ability to detect infiltration in bone marrow biopsy specimens from patients with disseminated neuroblastoma was assessed by immunohistological and routine histological methods. Frozen cores from 33 staging procedures were tested with UJ13A and UJ127.11. Immunopositive tumour cells were found in 10 of 17 staging procedures in which tumour was detectable by routine histological methods. Positive cells or stromal material were also found in eight of 12 staging procedures in which distorted architecture and fibrosis, but no obvious tumour, had been noted. Paraffin wax embedded cores from 29 of the same staging procedures were tested with antibodies against neurone specific enolase and neurofilament. Only a single core reacted with anti-neurofilament antibody. Neurone specific enolase positive cells or stromal material were found in nine of 15 staging procedures in which obvious tumour was detectable. Although these immunohistochemical techniques proved inferior to routine histology in their ability to detect obvious tumour, the demonstration of immunopositive stromal tissue which was not frankly malignant supports the view that distorted, fibrotic marrow may reflect persistence of neuroectodermal tissue and justifies its distinction from normal marrow when reporting the response to treatment.