Craddock & Owen are certainly right in drawing attention to the ‘complex overlap’ between psychotic and mood disorders, but the same can be said for most other sets of psychiatric symptoms. The neat, mutually exclusive categories described by our present classifications do not exist in nature, and classifications must necessarily draw a line somewhere between the major groups of symptoms. But are these lines at present drawn in the right places, and are there perhaps too many lines already? Their letter is very welcome, and it is to be hoped that many others will join this debate and express their views on what is an important matter.

My main research interest has been in those psychological disorders seen by generalists in primary care and general hospital practice, and here the overlap between symptoms is particularly marked. ^{Reference Löwe, Spitzer, Williams, Mussell, Schellberg and Kroenke1} In this broad group, the reasons for suspecting common ground between the various syndromes are set out at length elsewhere, ^{Reference Goldberg, Krueger, Andrews and Hobbs2} and the arguments considered most certainly included both data and clinical utility. It seems to my colleagues ^{Reference Andrews, Goldberg, Krueger, Carpenter, Hyman and Sachdev3} that if we are to make at least gradual progress towards a more rational system of classification there are other peculiar features that need attention. What sense does it make to classify similar disorders in different chapters of the ICD? Not only is there overlap between adult and child disorders, but the fact that anxiety disorders, mood disorders and somatoform disorders occur in separate chapters make multiple ‘comorbidity’ inevitable for many patients. Craddock & Owen welcome dimensions (without mentioning the problems that are associated with them) but appear to want the chapter structure of the classifications to remain as it is. It is difficult to see the advantage in doing this, and we cannot wait until ‘neuroscientific’ research has allowed us to cross the last frontier before improving it. It is not clear whether epidemiological or psychological research may also be allowed to be considered relevant – they are both respectably scientific, but do not qualify for the prefix ‘neuro-’.

The problem of where to put bipolar disorder is a difficult one to resolve, and for the time being the balance of evidence probably favours a cowardly approach, with bipolar disorders being separate from both schizophrenias on the one hand, and emotional disorders on the other. ^{Reference Goldberg, Andrews and Hobbs4} It is clear that further modifications will inevitably be made in our classifications as knowledge increases, and that changes suggested now can only be provisional. It remains to be seen whether either classification will take any account of the arguments put forward – but at least the issues have been aired. Our suggestions were intended to provoke discussion, in the hope that we might make a little progress towards a still distant goal.