The recent paper in which Dikeos et al (Reference Dikeos, Wickham and McDonald2006) investigate the distribution of symptom dimensions within a psychosis sample is a valuable contribution to the literature, and we fully support their observation that bipolar disorder is a much more solid construct than schizophrenia.

There are two important issues that were not discussed which we believe deserve consideration. The first is a major limitation of the current conceptual framework of psychopathology where definitions of psychopathology items are not independent of diagnostic concepts. Consider, for example, the relationship between items that measure course of illness and items that represent occurrence of reduced affective response and drive. Episodes of reduced affective response and drive with inter-episode recovery are likely to be interpreted as consistent with the presence of mood disturbance and indicative of a relatively good outcome. In contrast, chronically reduced affective response and drive which may be qualitatively identical to that in the previous example but without inter-episode recovery is likely to be interpreted as consistent with the negative features of a (schizophrenic) defect state and taken as evidence of a relatively poor outcome. In this example, recovery becomes part of the definition of two similar states. It is hardly surprising that one predicts poor outcome. We could give other examples. The only way to overcome difficulties such as these will be to use a set of clinical descriptors that do not have definitions that are enmeshed in our traditional diagnostic concepts. We believe such approaches are needed.

The second issue concerns validity. Dikeos et al addressed validity by considering prediction of clinical characteristics, some of which cannot be considered independent of the other items of psychopathology used to make the predictions. A key goal of diagnosis should be to identify clinical entities that are helpful for making management decisions. Recent developments in neuroscience in general, and molecular genetics in particular, offer the realistic prospects that over the coming years we will be able to identify domains of psychopathology that are associated with abnormal action in specific biological systems (Reference Craddock, O'Donovan and OwenCraddock et al, 2005). This will provide truly independent validators against which to examine the relative merits of diagnostic categories versus psychopathological dimensions (Reference Craddock, O'Donovan and OwenCraddock et al, 2006), will allow us to escape from our historical strait-jacket of traditional psychiatric thinking (Reference Craddock and OwenCraddock & Owen, 2005; Reference MarnerosMarneros, 2006) and has the potential to lead to major benefits for our patients.