The international classifications in psychiatry, i.e. ICD-10 and DSM4R, allowed a fast increase of factual information. The improvement of reliability account for most of it. But it did not came with an increase in validity. We will argue that these classification are not “validable” unless questioning some of the postulates that guided their elaboration.

We will state the epistemological principle underlying scientific and medical classifications and compare the ones at work in psychiatric classifications. Examples from the field of psychoses will be given.

Validity is a notion that can only apply to scientific classification, i.e. that aim at providing a model for an external reality. The validable part of a classification is the model on which it relies, not the definition that ensue the model. In medicine, the term “disease” apply to natural morbid entities defined by an hypothesis, i.e. a model, on their aetiology or pathophysiology. Only this hypothesis can be validated.

ICD and DSM utility is oriented toward a practical objective. This constrains choices in their elaboration not adapted to pursue a scientific goal. The most challenging is the atheoretical orientation which take away from the classification their only validable part, i.e. a model.

There is urgent need to clarify the utility of ICD and DSM in psychiatry, i.e. epidemiology and clinical practice. Requiring their use in scientific inquiry could only slow down any progress toward the discovery of real disease in psychiatry.