To get an insight into the degree of major histocompatibility mismatches in donor/recipient (D/R) combinations who were 'ABDR-matched' by serology for class I and by oligotyping for DR1-14 (low resolution typing), we performed additional HLA testing using a combination of molecular, biochemical and cellular techniques. For class II we used extended oligotyping, discriminating all the common DRB1/B3/B5-subtypes. For class I (-subtypes) we used oligotyping (HLA-A2,-A3,-B35,-B41,-B44), sequencing (HLA-B35,-B41,-Cw16), isoelectrofocusing (IEF), primary cytotoxic T lymphocyte (CTL) assays and class I-subtype specific T cell clones. In addition, all combinations were serologically typed for HLA-C. This high resolution typing by the combination of techniques revealed numerous histoincompatibilities. Fifty-three per cent of all 'ABDR-matched' combinations tested (n = 198) appeared to be DR incompatible. Moreover, independent of the presence of a class II mismatch, 47% of the donors tested (n = 131) displayed pretransplant cytotoxic activity against the patient. This activity was found to be rigorously correlated with the presence of class I incompatibilities, predominantly HLA-A,-B subtypes and HLA-C. Thus, although the D/R pairs had been originally matched for AB including serological splits and by generic class II typing, only 28% of the pairs were in fact ABCDR identical. As many as 38% of the D/R pairs were mismatched for one, 14% for two, 13% for three and 6% for four A, B, C or DRB1 antigens. We conclude that the presence of such a high number of histoincompatibilities in a group of relatively well matched D/R pairs will severely hinder the analysis of the role of HLA in marrow transplantation and that conclusions from studies in which D/R pairs are matched by conventional typing must be interpreted with extreme caution.