We examined the reliability of the OPCRIT system from ratings produced by 30 USA and European clinicians involved in molecular genetic research. The OPCRIT system facilitates a polydiagnostic approach to research on severe psychiatric disorders. OPCRIT comprises a 90-item checklist of signs and symptoms and a suite of computer programs, which together generate diagnoses according to the operational criteria of 12 major classificatory systems (e.g. DSM-III, DSM-III-R, RDC, ICD-10). Thirty summaries of actual cases ranging in signs and symptoms, taken from independent sources, were rated by participants from research centres across Europe and the USA using the OPCRIT system. Each rating was then compared to a standard rating using a kappa statistic. Good levels of reliability were observed within all classifications (e.g. DSM-III-R, kappa = 0.73, RDC, kappa = 0.71; ICD-10, kappa = 0.70) and a similar pattern of ratings was found in both the European and USA samples. We conclude that the OPCRIT system, is both flexible and practicable retaining the 'top-down' advantage of operational definitions as well as the 'bottom-up' potential offered by well defined signs, symptoms and other component items. Within the limitations of an international, multicentre design this study shows that the OPCRIT system affords good reliability with raters from a variety of geographical and theoretical backgrounds.