Les classifications internationales en psychiatrie, la CIM et le DSM, ont permis un accroissement rapide des connaissances factuelles sur les troubles qu’elles définissent, sans que cela se soit accompagné d’un accroissement de leur validité. La notion de validité n’est applicable qu’aux classifications scientifiques. Celles-ci cherchent à modéliser la réalité de phénomènes naturels. Ce qui est validable dans une classification scientifique, c’est ce modèle. En médecine, une maladie correspond à un modèle étio-physiopathologique de la déviation morbide d’un processus naturel. Or, la CIM et le DSM étant athéoriques, ils ne reposent sur aucun modèle et sont donc invalidables. Ainsi, l’utilité de ces classifications doit être limitée à la pratique technique. À vouloir les appliquer en science, ces instruments ne peuvent que ralentir toute avancée basée sur la méthode classique des corrélations cliniques.

The international classifications in psychiatry, i.e. ICD-10 and DSM4R, enabled a fast increase of factual information on the disorders they have defined. The improvement of reliability accounts for most of their success. But this did not come with an increase in validity. We will argue that these classifications are not “validatable” by questioning some of the postulates that guided their elaboration.

We will state the epistemological principles underlying scientific and medical classifications and compare them with those at work in psychiatric classifications. Examples, mainly from the field of psychoses, will be given to illustrate the core differences.

Validity is a notion that can only apply to scientific classifications. Scientific classifications aim at providing a model for an external reality, i.e. a natural phenomenon. The validatable part of a classification is the model on which it relies, not the definition that ensues from the model. Medicine has adopted the scientific way of classifying. Accordingly, in nosology, the term “disease” applies to natural morbid entities defined by a model, relevant to their aetiology or pathophysiology. This aetio-physiopathological model alone can be validated. ICD and DSM utility are, by construction, oriented towards practical and/or technical objectives. Hence, their framework is not adapted to the pursuit of scientific goals. The most challenging is the atheoretical orientation, which deprives the classifications from their only validatable part, i.e. the model underlying symptoms grouping. It is therefore worthless to attempt to accredit the invalidable, i.e. the symptomatic definitions of the disorders. Nevertheless this has been attempted by the “external validators” approach. The above-mentioned epistemological error is embedded in the external validator approach, which has inevitably failed to hold its promises.

It is urgent to clarify the utility of ICD and DSM in psychiatry, i.e. epidemiology and clinical practice. To rely exclusively on them for aetio-physiopathological scientific investigations can only hinder any progress towards the discovery of the cores of psychiatric diseases.