In the present study, 67 patients suspected to be cases of visceral leishmaniasis (VL) were each checked for leishmanial infection by the microscopical evaluation of various biological specimens, in-vitro culture, serology and an assay based on nested PCR. Most (35) of the subjects were immunocompetent (IC) but 32 were immunodeficient (ID) as the result of HIV infection (18 cases), treatment to prevent transplanted organs being rejected (six) or haematological malignancies (eight). Forty-one (61.2%) of the subjects (19 IC subjects, 12 HIV-positive patients, four transplant patients and six patients with malignancies) were considered true cases of VL. For the IC subjects, only the production and microscopical examination of leucocytoconcentrates and cultures of Buffy coats gave sensitivities of <80%, the results of the other methods showing higher sensitivities and almost perfect agreement with the 'gold-standard' diagnoses. For the ID subjects, however, only the serological tests and the PCR gave reasonable sensitivities (of >80%). For the initial diagnosis of leishmaniasis in ID patients, IFAT and western blots may be useful, as, among the present ID patients, they gave sensitivities (of 80.9% and 88.2%, respectively) that were almost as high as that for the PCR, and specificities of 100%. In the diagnosis of VL in either IC or ID patients, the assay based on a nested PCR appeared to be particularly reliable, with sensitivities of 88.9% and 95.2%, respectively, and a specificity of 100% in both groups of patients. The testing of bone-marrow aspirates by PCR revealed very few VL cases who were not found positive when samples of their peripheral blood were checked in the same assay. For both IC and ID subjects therefore, the use of the PCR-based method to test samples of peripheral blood (which can be collected much more easily than bone-marrow aspirates and with much less pain for the subject) is recommended.