A comprehensive training program for reliable use of the ICD/10 in Consultation-Liaison (C-L) psychiatry was conducted with 220 psychiatrists and psychologists from 14 European countries. The training included rating of written test cases and development of a coding manual to avoid diagnostic pitfalls not addressed in the ICD-10 manual. Following this training, all consultants rated 13 written case histories. One hundred sixty-seven consultants (76%) had a kappa (kappa) of at least 0.70. Only 13 (6%) had a kappa 0.40. The percentage of high reliability raters was evenly distributed among the different countries. Consultants had some problems in the differentiation between adjustment disorders and depressive disorders, and in the classification of disorders where ICD-10 differs from the DSM-III-R system. National biases in diagnostic practice were found with regard to the "case" concept and the role of alcohol in confusional states. Finnish consultants coded "no psychiatric disorder" significantly more often, whereas German and Italian consultants attributed delirious state more often to alcohol than consultants from other European countries. The study demonstrates that it is possible to achieve acceptable interrater reliability in applying the ICD-10 guidelines, through training programs designed for C-L psychiatrists and psychologists. Nevertheless, this first cross-national study shows the importance of addressing differences in national diagnostic practice.