Until recently, serological typing has been the primary technique used for HLA class I analysis. But because of limitations, molecular-typing techniques have replaced or supplemented the microlymphocytotoxicity test. It has been assumed that HLA class I serological typing was more accurate than serological HLA-DR typing; the latter has been shown to have 10-25% errors. But several studies have shown that HLA-AB typing was poorer than expected, and error frequencies between 5-25% were reported. This study systematically investigated the accuracy of HLA class I serological AB typing in healthy, bone-marrow registry donors, necrokidney donors, kidney-transplantation patients (on waiting lists), and haematological disorder patients. Genomic HLA class I typing, which uses polymerase-chain reaction with sequence-specific primers (PCR-SSP), gave discrepant results in 3-24% of the patients, compared to serological typings. The highest error rate (24%) was found among haematological disorder patients. Among the kidney waiting-list patients and necrokidney donors, 11% discrepancies were found. In the consecutively typed bone-marrow donors group, 3% errors were found. But among those with only one detected HLA-A specificity, 12% discrepancies were found, and among donors with only one detected HLA-B specificity, 19% errors were found. Based on these results, we recommend that patients with haematological disorders should be typed using genomic techniques. In investigations of bone-marrow registry donors and kidney patients, in which only one serological specificity is found, additional typing by genomic methods should be done.