Abstract:The Draft of “ICD‐10, Chapter V, Clinical Descriptions and Diagnostic Guidelines” was tested in a multicenter field trial in Japan. We have previously reported good results in suitability, confidence and ease of diagnosis, and adequacy of descriptions of the Draft. In this paper, the interdiagnostician reliability of the Draft is reported. Among the two‐character categories, “Schizophrenia, Schizotypal States and Delusional Disorders (F2)” (ICC = 30) and “Mood Disorders (F3)” (ICC =.80) proved reliable. “Neurotic, Stress‐Related, and Somatoform Disorders (F4)” was less reliable (ICC =.65). The ICCs of the 17 major categories (three‐character code) and the 21 subcategories (four‐character code) were also calculated. The finding that in Japan subtyping schizophrenia with ICD‐10 was more reliable than that made using DSM‐III Diagnostic Criteria supports the need to use a descriptive version of ICD‐10 as the basis for several versions serving different purposes. The nature of disagreements with unreliable categories was also investigated. The results are discussed with special reference to the changes in the final Draft of Chapter V, which contained a feedback of the results from field