Studies on HLA antigens in schizophrenia have produced conflicting results, but there has been greater agreement when clinical subtypes of the disorder have been separated. In view of this, we reassessed 68 previously studied patients with hospital diagnoses of schizophrenia and, while blind to their HLA types, used operational criteria to define clinical subtypes. We compared and combined the results with those from all available similar studies. Those of our patients who fulfilled operational criteria for paranoid schizophrenia showed a nonsignificant increase in HLA A9 as compared with controls. The magnitude of the increase was similar to that from all previous reports, and when data from all sources were combined, the evidence for an association between HLA A9 and paranoid schizophrenia was consistent and highly significant. Patients who were diagnosed as suffering from hebephrenic schizophrenia showed significant increases in HLA A1 and B8 compared with controls. An association between hebephrenia and A1, but not B8, remained on combining the results with those of other studies.