Dr Bennett's, Dr Ryle's and Professor Pilgrim's letters raise several very different issues, which makes it impossible for me to respond to, or even comment upon, more than a few of them.

Dr Bennett's argument that the concept of mental illness assumes an ‘abnormality of higher mental function’ and that personality disorders ‘lack good-quality evidence of altered higher mental function’ is essentially the same as Aubrey Lewis's contention that mental illness involves an ‘evident disturbance of part-function as well as of general efficiency’, and that ‘until the category (of psychopathic personality) is… shown to be characterised by specified abnormality of psychological functions, it will not be possible to consider those who fall within it to be unhealthy’ (Lewis, 1953). Lewis's views had a considerable influence on my generation of psychiatrists but now, 50 years on, the limitations of this criterion for distinguishing between personality disorder and mental illness are increasingly apparent, mainly because of the evidence that some personality disorders and some mental disorders share the same genetic diathesis, and are sometimes amenable to the same treatments. As a result, confusion reigns. The affective personality disorder of ICD-9 has been replaced by two new mental disorders, cyclothymia and dysthymia, in ICD-10; schizotypal disorder is classed as a personality disorder in DSM-IV but with schizophrenia and delusional disordres in ICD-10; and the authors of DSM-IV wonder whether avoidant personality disorder may simply be an ‘alternative conceptualisation’ of generalised social phobia.

Dr Ryle argues that the behaviour of people identified as having ‘borderline personality disorders’ is understandable in the light of their childhood experience and is amenable to cognitive—analytic therapy. This may indeed be so. Unfortunately, from my perspective, it does not solve the problem, if only because the psychoanalytic concept of borderline personality only embraces a small part of personality disorder as a whole. Dr Ryle also asserts that ‘the need is… for an understanding of persons’ and has no sympathy with my (and, I presume, Dr Bennett's) interest in ‘underlying cerebral mechanisms’. Many psychotherapists would agree with him. But I still have to insist that we must agree what is implied by the term mental illness before we can decide whether personality disorders are mental illnesses or not, and that the forensic issues involved mean that this is not a trivial issue.

Professor Pilgrim asks, perhaps with tongue in cheek, whether Scadding's definition of disease (not mental disorder) implies that being male or poor are diseases. The answer in both cases is, of course, no. Scadding's definition refers explicitly to variation ‘from the norm for the species’, so the reference group for a putative male disease would be the life expectancy of other males. Likewise, poverty is a handicap imposed by the environment which may increase the risk of several diseases, and thereby reduce life expectancy, but is not itself a disease. For similar reasons, living in a zoo rather than a natural habitat is a disadvantageous environment for many wild animals, not a disease of wild animals, despite the implications for longevity. More importantly, Professor Pilgrim refers to the ‘logical superiority of a dimensional over a categorical approach’ to the classification of personality disorders and chides psychiatrists for what he regards as their inappropriate attachment to categories. I would argue that the relative merits of categorical and dimensional classifications is an empirical issue rather than a matter of logic, and that their relative advantages and disadvantages may vary with the purpose for which the classification is to be used. In fact, it is explicitly recognised in ICD-10 that personality disorders ‘represent either extreme or significant deviations from the way the average individual… perceives, thinks, feels, and particularly relates to others’. It is also on the cards that in DSM-V the American Psychiatric Association will replace its present categorical classification of personality disorders with a set of dimensions.