Objective: To evaluate the diagnostic utility of bone marrow (BM) sampling in HIV positive patients. Design: Retrospective cohort analysis. Setting: Specialist HTV/AIDS service in London. Subjects: 215 consecutive HIV infected patients undergoing 246 BM samplings for investigation of pyrexia without localising signs, haematological abnormalities, or staging/investigation of lymphoma. Main outcome measure: Diagnostic yield from (and impact on management of) BM sampling. Results: Of 122 BM samples taken to investigate pyrexia, 33 (27%) revealed the cause on microscopy: unexpected lymphoma in seven (6%), mycobacteriosis in 25 (20%), and toxoplasmosis in one (1%). Marrow infiltration was confirmed in 11 of 38 BM samples taken for staging/investigation of lymphoma/leukaemia. In afebrile patients, of 22 with pancytopenia, BM samples showed HIV associated changes in 17 and specific diagnoses in five (mycobacterial infection in three, haemophagocytic syndrome in one, and megaloblastic change due to vitamin B-12 deficiency in one); of 21 with isolated thrombocytopenia, 20 (95%) BM samples showed immune thrombocytopenic purpura to be the cause and the remaining patient had BM changes of aplasia; of 29 with isolated anaemia, 28 had BM changes of HIV associated dysplasia/ erythroid dysplasia and one had unsuspected iron deficiency; all 10 with isolated leucopenia/neutropenia had BM changes ascribed to HIV infection exacerbated by concurrent sepsis or medication; of four BM samples taken for other reasons, one showed mycobacterial infection. Conclusions: BM sampling has diagnostic utility in HIV infected patients with pyrexia without localising signs, pancytopenia, and staging/investigation of lymphoma; this test has little value in the investigation of afebrile patients with isolated thrombocytopenia, anaemia, or leucopenia as HIV is usually the underlying cause.