In their comprehensive review Gill et al. (Reference Gill, Donohoe and Corvin2010) convincingly argue for the contribution of genomics to the understanding of psychoses. It is of note that the authors have centred their discussion around psychosis and not used official nomenclature in the title. Indeed, deconstructing complex phenomena like schizophrenia (Allardyce et al. Reference Allardyce, Gaebel, Zielasek and van Os2007) increases the rigour of genomic research, e.g. clearly defined symptom dimensions or non-clinical correlates yield higher effect sizes for susceptibility loci or allelic variants.

However, official communication and funding of science dictate uniform language, and findings from novel integrative research (van Os & Rutten, Reference van Os and Rutten2009) are attuned to the definitions of the Diagnostic and Statistical Manual of Mental Disorders (DSM). Much of the discussion on the methodology of genome-wide association studies (GWAS) focuses on power calculations (Psychiatric GWAS Consortium Coordinating Committee, Reference Cichon, Craddock, Daly, Faraone, Gejman, Kelsoe, Lehner, Levinson, Moran, Sklar and Sullivan2009), with the implication that reliability and validity, prerequisites for sound methodology, are taken for granted.

A question that precedes the one addressed in the article is, thus: ‘Do current official psychiatric classifications facilitate research as much as they did at the time they were rebuilt for this purpose?’ At a time when rapid progress in genomic research coincides with the development process of DSM-V (APA, 2009), the answer warrants a reconsideration of the epistemological assumptions in DSM-III, apparently maintained in the upcoming edition.

In order to accelerate replicable research, DSM-III prioritized reliability over validity and emphasized standard clinical assessment (Kendler, Reference Kendler2009). Inherent to the major paradigm change reflected in the atheoretical perspective of DSM-III, however, was the assumption that many of its definitions would eventually fit the disease model, as missing biological information was acquired. Widespread clinical use of DSM definitions and adoption of a strictly medical model in psychiatry solidified this view, so that the polythetic constructs of DSM have been regarded as diseases, and the Manual as a natural classification. Today, given the small effects of individual genetic variants on the distal phenotype of schizophrenia and evidence of ‘shared’ susceptibility loci or genes with other disorders (Carroll & Owen, Reference Carroll and Owen2009), it is time to re-emphasize that this construct may not be a natural entity. Part of the challenge in the design, funding and interpretation of research comes from using one nosological system for both research and clinical purposes, and DSM is too influential to overlook this problem.

Level of agreement on the categorical diagnosis is another consideration, since genomic research necessitates multisite collaboration. Structured clinical interviews largely rely on patients' accounts and barely emphasize the experienced clinical reasoning critical for proper assessment of mild intellectual deficits, psychosis or impaired functioning in a developmental perspective. These stigmata have been overemphasized as characteristic for ‘chronic psychosis’, and their overlap with antipsychotic side-effects further decreases the likelihood of questioning an established diagnosis in future assessments, even when other neurodevelopmental or early-onset conditions (e.g. velocardiofacial syndrome or severe anxiety disorders) are properly recognized as ‘comorbid’ disorders. Overdiagnosis could be reduced, if DSM-V criteria relevant to differential diagnosis address individual symptoms rather than the global picture.