Coid et al (Reference Coid, Yang and Roberts2006) reported an important cross-sectional survey of 8397 persons in UK households and found that psychosis was independently associated with a sixfold increase in the reporting of five or more violent incidents. Given the controversy and sensitivity over the stigma associated with psychiatric illness, particularly concerning public perceptions of links between psychosis and violence, this kind of result is prone to generate misleading impressions.

In a recent comprehensive review Hiday (Reference Hiday2006) points out that surveys of this type are prone to exaggerate the contribution of mental illness and other diagnostic labels to violence as a result of several methodological weaknesses. The first is associated with the issue of comorbidity. It was not clear from the presentation of their data whether Coid et al were able to investigate the comorbidity of psychosis and other diagnostic categories and violence. It is possible that once comorbid substance misuse, personality disorder or other issues were taken into account, the unique contribution of psychosis to violence might have diminished dramatically (Reference HidayHiday, 2006).

There is an even more fundamental problem that underpins violence surveys of this type: a neglect of the confounding factor that those with mental illness are more likely to reside in violent neighbourhoods and this could be the key predictive variable, not the illness itself. The term now used to describe the places where most people with severe mental illness live is ‘socially disorganised communities’, and these combine a multiplicity of factors that promote violence completely independently of psychiatric dysfunction (Reference Silver, Mulvey and MonahanSilver et al, 2001). Features of these environments include chronic disabling poverty, few employment prospects or educational opportunities, decaying buildings and few amenities. In these neighbourhoods families and similar social institutions have broken down, leaving most individuals devoid of traditional social guidance and control (Reference Swanson, Swartz and EssockSwanson et al, 2002).

Living and growing up in such environments is possibly the key variable that predicts violence, not the mental illness of the individual (Reference HidayHiday, 2006). Community household surveys such as that reported by Coid et al (Reference Coid, Yang and Roberts2006) represent a unique opportunity to explicate the contribution of ecological factors when violence appears to be linked to mental illness. It would therefore be useful in terms of advancing the debate over the link between violence and mental illness if a wider theoretical background to such analyses could be encouraged in the future.