A conclusion in the abstract of Large et al ^{Reference Large, Smith, Swinson, Shaw and Nielson1} is illogical. If the same sociological factors causing increase in ‘other’ homicides up until the 1970s had caused the increase among the mentally ill, then they should have continued to have exerted this effect, with a continuing rise corresponding to ‘other’ homicides instead of a fall. Similarly, if the subsequent decline in homicides among the mentally ill were due to improvements in psychiatric treatment and service organisation as the authors suggest, then the rise in their rates prior to that period must have been due to the converse: a deterioration in quality of treatment and service organisation. The obvious explanation (which is now politically incorrect) is the closure of mental hospitals and rehabilitation at that time because of almost non-existent community care.

In reality, it is highly unlikely that there has been a true rise and fall in homicide among mentally ill people in England and Wales over the past 50 years. These figures are entirely based on statistics which reflect the workings of the Criminal Justice system (a charge to which I plead guilty). ^{Reference Coid2} They merely reflect changes in processing defendants by the courts. The probable culprit for declining diminished responsibility was declining enthusiasm for treating personality disordered and sexually deviant killers under the Mental Health Act legal category ‘Psychopathic Disorder’. The authors did not provide statistics on other forms of man-slaughter. These have increased in recent years, suggesting that defence lawyers have become more successful in putting forward alternative defences to murder than diminished responsibility.

I agree with the authors that sociological and legal factors (mainly the latter) have effects on rates of homicide due to mental disorder. But it is the overall base rate of homicide in the population that matters and with which these figures must be compared. This differs markedly between different countries. In those where it is very high, such as South America and Sub-Saharan Africa, mental disorder is almost irrelevant as an epidemiological risk factor. The authors refer to a small number of studies suggesting a correlation between rates of homicide among the mentally ill and rates among the rest of the population. It may well be that the ‘laws’ ^{Reference Coid2} they refer to are too rigid. For example, it makes sense that a country that allows handgun ownership is more likely to have killers with schizophrenia who use a handgun, and at a rate higher than in countries where handguns are banned, although the evidence for this remains thin on the ground. But from the public health perspective does it matter? Handguns are the key risk factor, not schizophrenia.

England and Wales have a low but steadily rising rate of homicide. It is unrealistic to propose mental health services as a public health intervention, but will be popular with politicians. Social geographers have demonstrated that social exclusion and growing social inequalities are the strongest correlates with this phenomenon affecting young men in England and Wales. ^{Reference Shaw, Tunstall and Dorling3}