Au tournant du siècle, la valeur DSM s’est effondrée outre-Atlantique, entraînant dans sa chute la notion médicale classique de phénotype même pour la recherche fondamentale sur les psychoses. Sur leurs ruines ont émergé de nouveaux concepts : les dimensions prônées par l’association des psychiatres américains dans la partie recherche du DSM-5, les constructions mises en avant par le NIMH avec ses Research Domain Criteria (RDoCs), les biotypes du NIH, les endophénotypes des généticiens ou les propositions mixtes mêlant phénotype et biotype. Si les concepts traditionnels s’inscrivent dans le cadre paradigmatique naturaliste des sciences de la vie, la plupart des nouveaux sont issus du cadre normativiste des sciences humaines. Cela n’est pas sans conséquences en termes applicatifs : l’hypothèse mono-factorielle du premier lui confère une efficience supérieure. Plutôt que de faire un nouveau choix d’autorité, comme l’avait été le DSM, comment laisser la nature nous guider?

At the turn of the century, the decline of the DSM-value in psychosis research not only casted doubt about the disorder concept, but also to the one of phenotype, paving the way to new proposals. What are the paradigmatic frameworks of its proposed alternatives such as the dimensions of the DSM-5 research section, the constructs of the NIMH's Domain Research Criteria (RDoCs), the biotypes of the NIH, the endophenotypes of the geneticists or even the mix of phenotypes and biotypes?

Strength and weakness of each alternative are discussed from the perspective of philosophy of knowledge and illustrated.

ICD and DSM still valuable for evidence-based medical practice but do not define valid scientific entities, i.e. diseases funded on biology. There is room for further research in the classical naturalist framework. It assumes the existence of naturally founded phenotypes, validated by the finding of their single causes, translating into a biomarker and efficient treatment. At the opposite, dimensions and constructs are defined at the phenotypic level. Questioning their biological foundations rely on the normativist framework of social sciences, i.e. dimensions and constructs are assumed to be supported by multiple non-interacting causes of small effects which merely add to one another. Although they are a good way to summarize the variance in a population their translation into efficient treatment remains questionable. Biotypes are clusters of biological measures which admit single causes defined at a more fundamental level and which could ultimately be converted in a biomarker. Endophenotypes are instantiations of biotypes based on genetic biomarkers. Last mix of clinical and biological definitions could be an option as illustrated in other field of medicine. An adversarial collaborative study could confront these different models on the same biological measure and help to disentangle this issue.