Previous studies have established an association between cytomegalovirus (CMV) and classic and endemic Kaposi's sarcoma (KS) which can be extended to include the epidemic form of KS. The identification of nucleic acid sequences homologous to CMV but not to Epstein-Barr virus (EBV) and Herpes simplex virus type 1 (HSV-1) and type 2 (HSV-2) as well as the detection of CMV gene products, particularly early antigens, in tumour biopsies and/or early cultures derived from them is an important criterion in the establishment of the type of association. Detection of CMV early antigens but not late antigens as well as the failure to demonstrate virus particles in primary tumour biopsies rule out a simple passenger role of this virus or a preferential site for virus replication in neoplastic tissue. In the setting of a profound immune dysfunction, as in the case of acquired immune deficiency syndrome (AIDS), renal transplant recipients under immunosuppressive therapy, and other iatrogenically immunosuppressed patients (the three groups in which KS has been most frequently observed), the specific involvement of CMV oncogenes, of which at least two have been so far identified in experimental systems, is strongly indicated. Moreover, in vitro transformation experiments indicate that the virus is clearly involved in the initiation of transformation, while other factors (co-carcinogens, activation of cellular oncogenes?) might be required for the maintenance of the transformed phenotype. Endothelial cells, permissive for CMV infection, might be the prime target for these events, since there is evidence that they are a main constituent of KS tissue. Data obtained during 14 years of research on KS are reviewed and possible etiopathogenetic mechanisms giving rise to this type of tumour are discussed.