The publication of the DSM-III in 1980 was intended to be a reaction to the evidence of the embarrassingly low reliability of psychiatric diagnoses, which was perceived as a major threat to the credibility of the psychiatric profession. The aims of the DSM-III project were actually two. First, the reliable definition of the diagnostic categories was expected to lead to the collection of research data that would validate those diagnostic entities and in particular elucidate their etiopathogenetic underpinnings. Second, there was an expectation that, by increasing reliability, communication among clinicians would be improved and clinical decisions made more rational. Today, one could say that the first aim of the project has not been achieved, while the fulfilment of the second aim has never been tested appropriately. The crisis of confidence in the DSM paradigm, clearly emerging from the debate following the publication of the DSM-5, has led on the one hand to a renewed emphasis on clinical utility, which is featuring prominently in the ongoing process of development of the ICD-11. On the other hand, it has led to a radical attempt to reform psychiatric nosology starting from neurobiological and behavioural phenotypes. This attempt does have its weaknesses, but may also represent a stimulus to reconceptualize some psychopathological constructs, especially in the area of psychoses, in order to reduce the gap between the level of neuroscience and that of clinical phenomenology.