In considering the relationship between criminal behaviour and severe mental illness two schools of thought can be identified. One, arbitrarily labelled the 'criminological', holds that criminality and mental illness are only weakly associated, if at all. The second, labelIed the 'psychiatrie' view, holds that there is areal and consistent relationship between mental illness, in particular psychosis, and violen ce. There are at least two explanations for these apparently conflicting views. Methodological difficulties, such as case definition, information bias, selection bias and confounding, have led to flawed results. The opposing conclusions are, howeve,r. not just the result of methodological shortcomings but reflect more fundamental differences in approach. The criminological emphasis on populations and offending, largely in quantitative terms, is contrasted with the psychiatrie emphasis on individuals and violen ce. Longitudinal cohort studies, using the methodology of criminal careers, may allow the two approaches to be combined, and lead to conclusions of more practical value.