BACKGROUND Tragic and high profile killings by people with mental illness have been used to suggest that the community care model for mental health services has failed. It is also generally thought that schizophrenia predisposes subjects to homicidal behaviour. OBJECTIVE The aim of the present paper was to estimate the rate of mental disorder in people convicted of homicide and to examine the relationship between definitions. We investigated the links between homicide and major mental disorders. METHODS This paper reviews studies on the epidemiology of homicide committed by mentally disordered people, taken from recent international academic literature. The studies included were identified as part of a wider systematic review of the epidemiology of offending combined with mental disorder. The main databases searched were Medline. A comprehensive search was made for studies published since 1990. RESULTS There is an association of homicide with mental disorder, most particularly with certain manifestations of schizophrenia, antisocial personality disorder and drug or alcohol abuse. However, it is not clear why some patients behave violently and others do not. Studies of people convicted of homicide have used different definitions of mental disorder. According to the definition of Hodgins, only 15% of murderers have a major mental disorder (schizophrenia, paranoia, melancholia). Mental disorder increases the risk of homicidal violence by two-fold in men and six-fold in women. Schizophrenia increases the risk of violence by six to 10-fold in men and eight to 10-fold in women. Schizophrenia without alcoholism increased the odds ratio more than seven-fold; schizophrenia with coexisting alcoholism more than 17-fold in men. We wish to emphasize that all patients with schizophrenia should not be considered to be violent, although there are minor subgroups of schizophrenic patients in whom the risk of violence may be remarkably high. According to studies, we estimated that this increase in risk could be associated with a paranoid form of schizophrenia and coexisting substance abuse. The prevalence of schizophrenia in the homicide offenders is around 6%. Despite this, the prevalence of personality disorder or of alcohol abuse/dependence is higher: 10% to 38% respectively. The disorders with the most substantially higher odds ratios were alcohol abuse/dependence and antisocial personality disorder. Antisocial personality disorder increases the risk over 10-fold in men and over 50-fold in women. Affective disorders, anxiety disorders, dysthymia and mental retardation do not elevate the risk. Hence, according to the DMS-IV, 30 to 70% of murderers have a mental disorder of grade I or a personality disorder of grade II. However, many studies have suffered from methodological weaknesses notably since obtaining comprehensive study groups of homicide offenders has been difficult. CONCLUSIONS There is an association of homicide with mental disorder, particularly with certain manifestations of schizophrenia, antisocial personality disorder and drug or alcohol abuse. Most perpetrators with a history of mental disorder were not acutely ill or under mental healthcare at the time of the offence. Homicidal behaviour in a country with a relatively low crime rate appears to be statistically associated with some specific mental disorders, classified according to the DSM-IV-TR classifications.