This paper reviews translational research in psychiatry, focusing on those programs addressing the problem of the validity of psychiatric diagnoses. In medicine in general, and in psychiatry in particular, the term “translational” is used with different meanings. A conceptual analysis suggests that there are at least seven different types of translational research in psychiatry: T1 (“bench-to-bedside” development of tools and treatments), T2 (application of animal models to human psychiatry), T3 (papers focusing on the mind-brain gap, studying biological, neurobiological and cognitive dysfunctions), T4 (personalised therapies and prediction of treatment responses), T5 (“bedside-to-bench” translation of population data for laboratories), T6 (implementation of treatments at the population level, including accessibility and quality of services), and T7 (improving translational knowledge in residents’ trainings and researchers’ careers).
Concerning the problem of validity of psychiatric diagnoses, new neurocognitive models like the Research Domain Criteria project are considered, in particular the translational program of cross-validation aimed at reducing the gap between neuroimaging data and psychopathological scores derived from rating-scales. It is shown that these programs are useful, filling some of the current research gaps, but it is also stressed that they carry implicit realist and reductionist assumptions. It is finally suggested that the formation of mental symptoms is a complex process involving both neurocognitive and semantic factors, which raises doubts about the possibility of complete translations, without residuals.

Keywords : Translational research, psychiatric classification, psychopathology, validity, Research Domain Criteria, Cambridge model