In order to assess the accuracy of schizophrenia diagnoses for genetic studies, we identified all schizophrenia patients (n = 492) in an isolated community with a diagnosis of schizophrenia in the Finnish Hospital Discharge Register (HDR) between 1969-1991. For the accuracy study we identified a sample of 73 patients from registers with Diagnostic and Statistical Manual (DSM)-III-R for schizophrenia (codes 295.10, 295.30, 295.60, 295.90) (n = 62) or “schizophrenia spectrum” diagnoses (295.40, 295.70, 297.10, 301.20, 301.22) (n = 11). When the operational criteria (DSM-III-R) were applied by two senior researchers using information from the original mental hospital records, 93% (68/73) of the cases fulfilled criteria for schizophrenia or schizophrenia spectrum. The results demonstrate that the schizophrenia diagnoses of the registers are accurate when a broad concept of schizophrenia is applied. When using operational DSM-III-R schizophrenia criteria, eight false positive cases were found among the 62 mental hospital schizophrenia diagnoses. Consequently, there may be a need to reassess schizophrenia diagnoses depending on the purpose of the study. We also found good agreement between DSM-III-R (kappa 0.93) and operational criteria (OPCRIT) diagnostic system (kappa 0.89) diagnoses, made by one researcher, compared with operational diagnoses. This indicates the possibility for the reliable use of one of these methods alone for diagnostic reassessment. The information in the HDR on primary diagnoses and on the dates of admission and discharge was accurately transferred from the hospital records.