Being descendants of the Kraepelinian nosology, DSMN and ICD-10 rely largely on the internal cohesion of the clinical picture and the pattern of course and outcome as validating criteria of the definitions of mental disorders. The majority of the research diagnostic criteria are provisional and should be extensively tested against empirical evidence. The crucial issue is whether psychiatric disorders, as currently defined in DSM-IV and ICD-10, are clearly separated from one another and from normality. Options for future revisions of the classifications include categorical typologies, dimensional models and empirically derived prototypes. The advantages and disadvantages of each option are outlined, highlighting the need for new research focusing on these critical issues. Copyright (c) 2005 S. Karger AG, Basel.