Kendler argues for the reality of psychiatric diagnostic classes in terms of two realist theories of truth, coherence and correspondence. I would advocate an alternative interpretation of the truth status of diagnostic classifications that leads to different conclusions. This is based firstly on Karl Popper's ideas on the growth of knowledge, whereby hypotheses developed from theoretical conjectures are deliberately subjected to attempts at refutation (we refine our always provisional views of what is true by increasing our knowledge of what is false). My second source of argument is John Wing's view that diseases are theoretical constructs on which disease theories may be based and tested. Such theories relate variously to aetiology, pathology, treatment, course and outcome. Rejecting a disease theory does not force rejection of the disease construct it seeks to qualify. We adhere to disease constructs more strongly than to the disease theories based on them. However, if it becomes apparent that the information obtained by testing disease theories is incoherent, we may eventually jettison particular disease constructs, as has happened regularly in the history of medicine. The disease constructs used in psychiatry may be approaching this point.