Based on a 6 1/2-year study of bone marrows (BM) from 74 consecutive adult patients with newly diagnosed primary myelodysplastic syndromes (MDS), we have evaluated the role of immunophenotyping (IP) for subclassification purposes with a selected panel of monoclonal antibodies (Mab). Compared to normal BM IP revealed increased numbers of Mab-defined immature myeloid cells (defined by Mab MY7 [CD13] and MY9 [CD33] (P less than 0.05). Furthermore, decreased numbers of mature myeloid cells (defined by Mab NAT-9 II:3F-6F [NAT-9]) (P less than 0.001) were demonstrated in all French-American-British (FAB) subtypes except refractory anaemia with sideroblasts (RA-S). Since expression of CD13 and CD33 antigens on immature myeloid cells is variable, IP based on a single Mab was often found to be non-informative. However, the construction of myeloid antibody ratios (MAR), designed to give higher figures with increasing numbers of immature myeloid cells and/or decreasing numbers of mature myeloid cells in the BM, disclosed increasing mean ratio values with progressing FAB subtypes. More significant however, was that different prognostic subgroups based on the MAR could be identified independently of the FAB classification. We conclude that the use of IP--especially when MAR is included--is useful in prognostic scoring systems in MDS.