This paper reviews and critiques major studies that have examined a possible relationship between mental disorder and violent offending. Such studies have either examined a) the prevalence of violence in the mentally disordered, or b) the prevalence of mental disorder amongst the violent. Indices of violence used include a) arrest rate, b) conviction rate and c) prevalence of violent behaviour in psychiatric inpatients, representative community samples, and birth cohorts. It is concluded that although the mentally disordered do face an elevated risk of violent behaviour/offending, this risk is largely confined to those suffering from severe or long term psychotic symptoms, or psychopathic disorders.

Even the most enlightened reader might express unease at the prospect of a psychiatric-care unit being built in his or her neighbourhood. A major concern might be whether psychiatric patients pose a danger to members of the local community. Notorious crimes committed by the mentally ill are often wheeled out by the press to support such fears and undermine policies such as de-institutionalisation of psychiatric patients. Civil and criminal law (by providing for the detention of the mentally ill likely to pose a danger to others) lend credence to the notion of the mentally ill as dangerous, as did early forensic psychiatry and criminology by dwelling on dramatic acts of violence committed by "the mad".

Such stereotyping of the mentally ill as a threat to personal and community safety is challenged by mental health organisations. A US National Mental Health Association pamphlet (1987) claims that "people with mental illness pose no more of a crime threat than do other members of the general population". Ironically, a central feature of mental health policy, risk assessment, casts immediate suspicion on such a statement.

In order to examine a link between mental disorder and violent offending, the terms involved must be defined. Mental disorder refers to a psychological or behavioural condition that deviates significantly from those typical of persons with "normal" mental health, causing a functional impairment (e.g. emotion, perception or memory). The current edition of the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association - DSM-IV) distinguishes between organic (somatic) psychoses (e.g. Alzheimer’s dementia); functional or other psychoses (e.g. schizophrenia, affective psychosis); and neurotic disorders (those without a demonstrable organic basis e.g. depression, anxiety, obsessive-compulsive disorder). Personality disorders (maladaptive behavioural patterns distinguished by no pre-morbid/morbid course) compose a fourth group of psychiatric disorders.

In particular, investigation has focused on schizophrenia - a group of disorders characterised by disturbances of thought (e.g. delusions), perception (e.g. hallucinations and other ‘positive’ symptoms), emotion (e.g. blunted affect and other ‘negative symptoms’), and social functioning. Described by Kandel (1991) as "perhaps the most devastating disorder of mankind" schizophrenia affects both the sexes equally, occurring in approximately 1% of the population.

Violent crimes are intentional acts of aggression in humans that violate criminal law. Principal violent crimes are murder, non-negligent manslaughter, robbery, aggravated assault (with/without a weapon or by means likely to produce death or bodily harm), and forcible rape. Conclusions as to the relationship between mental disorder and violent crime should take into account various studies’ differing criteria for violent offending.

Ostensibly, it might seem easy to establish if there is a relationship between mental disorder and violent offending. However, experimental investigation of such a relationship is hampered by major methodological difficulties. We are unlikely to find a symptomatic relationship between mental disorder and criminality as we would sneezing and a common cold. Monahan (1992) has classified experimental studies on the relationship between mental disorder and violence (Table I):

Table I: Methods of estimating the relationship between mental disorder and violence

Arrest rate may be used as an index of violent behaviour. Shuerman & Soloman (1984) compared the arrest rate for Los Angeles County Department of Mental Health patients (n=65,599) against that of the remaining county adult population (n=4,919,000). Patients were 1.7 times more likely to be arrested for a violent offence. In their review, Link et al. (1992) find an elevated patient versus public arrest rate of 3:1.

There are problems with the interpretation of evidence from arrest-rate studies. The general public are not an adequate control group for patients discharged from public psychiatric hospitals. Such patients may have been hospitalised because of their violence (Craig, 1982) - the best predictor of future violence is past violence. Other risk factors for violent behaviour may be over-represented in discharged patients e.g. being male, poor, and from an ethnic minority. Arrest procedure and police bias may contaminate arrest rate studies - police may arrest a mentally ill person who has committed an offence rather than organise hospitalisation. This "criminalisation of mental illness" (Abramson, 1972) diverts patients away from the health care system to the criminal justice system. Teplin’s (1984) field study found those showing signs of overt mental disorder were more likely to be arrested than non-mentally disordered persons, even when type of offence was held constant. Klassen & O’Connor (1988), however, found that released mental patients who committed violence in the community were twice as likely to be re-hospitalised than arrested.

Conversely, if violence is seen as an indicator of psychiatric morbidity, then offenders will be diverted into the patient population (the "psychiatrisation of criminal behaviour" - Monahan, 1973). Thus arrest and hospitalisation become correlated - the more a person is defined as a criminal (arrested), the more frequently he will be defined as mentally ill (admitted), resulting in "changing clientele" for the mental health care system (Steadman et al., 1978a).

Steadman et al. (1978b) thus supposed that differential arrest rates among patients compared to the general public could be accounted for by differential demographic factors - patients have a lower average age, lower social class, more previous arrests etc. By statistically adjusting for such factors, the differential arrest rate was nullified. However, controlling for demographic variables may be statistically unsound (Monahan, 1992) - if a symptom of schizophrenia is social drift, controlling for social class will obscure the relationship between schizophrenia and arrest rate. As patient samples will rarely contain many "first-episode" schizophrenic patients, controlling for prior arrest rate may have a similar effect.

Furthermore, prior arrest rates in patients have increased since the 1960’s - possibly due to de-institutionalisation (so violence results in arrest instead of treatment) and lack of provision for discharged patients (associated with medication non-compliance and alcohol/substance abuse, increasing the risk of violent offending). In summary, arrest rate studies do not constitute a sensitive tool in investigating the relationship between mental disorder and violent crime.

Conviction rate studies may suffer from similar difficulties, due to differential treatment of the mentally disordered in the criminal justice system. Wessely et al.’s (1994) study of all incident cases of schizophrenia in a London borough found schizophrenic females were 3.3 times more likely to have received a criminal record than those with other mental disorder (the control group). This effect was only present in males for violent offences - schizophrenic males were 3.8 times as likely to receive a conviction for a violent offence as controls, but no more likely to receive a criminal record in general. An interaction effect between gender, schizophrenia, and ethnicity showed black schizophrenic males to be most at risk of any sort of conviction, although criminal careers of schizophrenics began later and were shorter than those of controls. Subjects were not restricted to hospital discharges - cases were obtained from a psychiatric case register, and time spent in hospital was controlled for.

However, in a similar study of 644 schizophrenics born between 1920 and 1959 in Stockholm and discharged in 1971, Lindqvist & Allebeck (1990) showed that violence recorded by schizophrenics is almost exclusively of minor severity. Although methodologically sound (patients were re-diagnosed by an independent clinician along DSM-III criteria), considerable subject-attrition through death may have attenuated any relationship - studies have shown that those who kill others are at an increased risk of killing themselves. Even so, schizophrenics committed 4 times as many violent offences as the general population.

Is studying violent behaviour established independently of arrest or conviction a better method? Although violent behaviour is not equivalent to violent offending, there is likely to be equivalence in aetiology. Psychiatric hospital reports typically contain retrospective information about behaviour in the days or weeks preceding admission. Estimates of pre-admission violence vary from 9.9% (Tardiff & Sweillam, 1980) to 22% (McNiel et al., 1988), although studies differ in definitions of violence, patient’s type of mental disorder (including/excluding substance/alcohol abuse), and retrospective time period. These results cannot readily be generalised to the wider population of community-dwelling psychiatric patients, as violence probably peaks just before admission and may be over-reported by parties interested in the patient’s admission (e.g. family members). Such qualification is consistent with Tardiff & Koenigsberg’s (1985) finding that only 3% of private outpatients recently assaulted other community members. Thus, by relying on psychiatric hospital admission reports, studies of violent behaviour established independently of arrest may be as methodologically unsound as arrest-rate studies.

What about violent behaviour of patients during their hospital stay? Binder & McNiel (1990) showed that within 3 days of hospitalisation, 17.4% of patients physically attacked another person at a short-term psychiatric facility. However, such studies are of limited use in answering the present question as there is no data on offences (attacks in hospital, unless very serious, are unlikely to elicit police intervention), and no control studies exist of the violence of normal people made to live on psychiatric wards. Self-reports of previous arrest and conviction are the best discriminator of those likely to commit hospital violence. The hospital may, however, be an appropriate arena to examine how symptomatology/medication etc. affects violent behaviour.

The problems of the above methodologies are avoided in studies of persons with mental disorder who are detected in representative community samples. In a rigorously designed study, Swanson et al. (1990) administered the Diagnostic Interview Schedule (DIS - Robins et al., 1981) to household residents participating in the Epidemiologic Catchment Area (ECA) program (Reiger et al., 1984). The DIS measured the occurrence (but not the frequency) of violent behaviour (hitting someone, fighting after drinking alcohol, using a weapon etc.) and also assessed the community prevalence of DSM-III Axis I (major) mental disorders. Percentage of population violent during the previous year by diagnosis was: no disorder: 2.1%, schizophrenia: 12.7%, mania/bi-polar disorder: 11%, major depression: 11.7%, alcohol abuse/dependence: 24.6%; drug abuse/dependence: 34.7%. Subjects who were co-morbid (suffering from more than one diagnosis) were included in all relevant categories.

The results of the ECA study suggest that meeting criteria for a DSM-III Axis I diagnosis increases prevalence of violence by at least 5 times compared to those with no diagnosis, with substance abuse increasing prevalence by up to 16 times. Surprisingly, there is little difference between the prevalence of violence for those meeting the criteria for a neurotic disorder like major depression and those suffering from a psychotic disorder such as schizophrenia. The association of "major" mental illness (schizophrenia and affective disorders) with increased violent behaviour remained significant when substance abuse, demographic variables, and institutional contacts (prior arrests and hospitalisations) were controlled for. Volavka (1995) thus seems misguided in suggesting that differential aetiology, symptomatology and outcome of distinct mental disorders causes distinct and differential associations with violent crime.

Confidence in the ECA conclusions is bolstered by the results of a similar study conducted by Link et al. (1992). Mentally disordered groups (as measured by the Psychiatric Epidemiology Research Interview - Dohrenwend et al., 1980) were 2 or 3 times as violent (measured by self-report and police data) compared to never-treated community residents. This association remained intact even when a huge range of demographic variables were controlled for. Regression analysis showed that recent violence among patients was related to recent level of psychotic symptoms (especially delusions of thought control, thought insertion and persecution). Furthermore, current psychotic symptoms also predicted recent violent behaviour in the never-treated community residents. The results of these community studies cannot be explained away by sample bias and suggest that active psychotic symptoms might explain any relationship between violent behaviour and mental disorder.

Birth cohort studies also minimise selection bias. Hodgins (1992) examined an unselected Swedish birth cohort of persons born in Stockholm in 1953 (n=15,117). At age 30, 1.1% of the subjects suffered from a major mental disorder. Mentally disordered males were 4 times more likely than normal males to be convicted of a violent offence (which in this study assumed a broader definition including ‘molestation’ and ‘unlawful threat’). Mentally disordered females were 27 times more likely than normal females to have been similarly convicted (female statistics may be more subject to the effect of mental disorder due to a much lower base rate of violent offending). These estimates are likely to be conservative - persons sent to hospitals by the courts were not included, and violence committed during hospital stays (average for males = 522 days) would be unlikely to result in recorded offences. This study confirms the results obtained by Ortmann (1981), using a Danish birth cohort, who found an elevated conviction rate for all types of offence amongst those who at some time had been admitted to a psychiatric hospital.

As Table (1) shows, there is a second major empirical direction available - the examination of mental disorder among the violent. Teplin (1990) reviewed 18 studies of mental disorder among U.S. jail inmates (of whom an average of 66% have a current or past conviction for a violent offence). These studies established that between 5% and 16% of inmates were psychotic. Roth (1980) finds a rate of psychosis in prisoners of below 5%, with a further 10-15% suffering from other mental disorder. Using hospitalisation as an index of mental disorder, Steadman et al. (1987) reported rates of disorder among inmates across six U.S. states. Considerable geographical variation among hospitalisation records of prisoners emerged - Arizona, with 2.2% of prisoners having been hospitalised, is physically adjacent to California, where 15.2% of inmates have been hospitalised. Such discrepancies are likely to be due to local mental health and criminal justice policy.

Teplin (1990) avoided such problems by administering the DIS to male inmates in a Chicago jail (n=728). Combined with the results of a similar study of male inmates (n=362) by the California Department of Corrections (1989), a prevalence of schizophrenia 3 times higher than that of the general population was shown. Prevalence of major depression was 3 to 4 times higher, and bipolar disorder 7 to 14 times higher. These results are similar to those of Taylor & Gunn (1984) - 9% of their sample of inmates showed major symptoms of psychiatric illness, and a further 8.6% evidence of withdrawal from drugs/alcohol. 45% of those with schizophrenia had been charged with a violent offence. Problems with such studies include false reporting of disorder by inmates, possibly to influence their legal/prison status.

Lindqvist (1986) studied all persons committing homicide in Northern Sweden between 1970 and 1981 (excluding those not prosecuted - estimation of mental disorder prevalence is thus probably conservative). 53% suffered from major mental disorder, 38% of whom were co-morbid for substance abuse. Of these, 85% were intoxicated at the time of the homicide.

Substance abuse may be a controversial candidate for inclusion as a mental disorder, as may psychopathic personality disorder (PPD). PPD is a persistent disorder or disability of mind resulting in abnormally aggressive or irresponsible behaviour which is not the product of psychosis or other illness. Psychopaths behave in a socially unacceptable manner (though they may show superficial charm and good intelligence) often with adverse affects on themselves and others - they may leave behind them a characteristic "chain of chaos" (Reid, 1978). They may show lack of remorse, shallow affect, impulsivity, a grandiose sense of self worth, and a lack of long-term goals (items from the revised Hare Psychopathy Checklist [PCL-R] - Hare, 1991). They may have under-aroused autonomic nervous systems; indeed, their disorder may be directly opposed to that of the neurotic - development of conditioned fear responses is inhibited.

Psychopaths unsurprisingly show a greater degree of criminality (often characterised by apparent irrationality), and violent crime in particular, than normal persons (Hare & McPherson, 1984). Their criminal career tends to begin earlier than those with major mental disorders. 97% of psychopaths (diagnosed by clinical global ratings) taken from a prison population had at least one conviction for a violent offence, compared to only 74% of non-psychopathic inmates. Psychopaths are more likely than other inmates to show violence in prison (Wong, 1984). Other studies have similarly established a link between violent crime and PCL score (e.g. Kosson et al., 1990; Serin, 1991). Psychopathy also predicts violent recidivism in young male offenders (Forth, Hart & Hare, 1990) - this violent offending persists even when other criminal activity declines with age (Hare, McPherson & Forth, 1988).

Associations of criminality and violence are, however, not surprising as criminal behaviour is often seen as a symptom of psychopathy (DSM-IV describes "repeatedly performing acts that are grounds for arrest" as a diagnostic category for Antisocial Personality Disorder - a condition analogous to PPD). PPD is unassociated with, or even negatively associated with, most Axis I mental disorders, yet PCL scores are associated with violence in both schizophrenics and non-schizophrenics (Heilbrun et al., 1992). Similarities exist between PPD and impulse control disorder, a condition that predicts violent acts on behalf of the patient (especially "motiveless" acts such as stranger homicide for no personal gain). Although personality disorders are often regarded as "less serious" than major mental disorder, their effect on the general public through crime may be more substantial.

A relationship between mental disorder and violent offending may not be inherently caused by mental disorder per se. For example, do schizophrenics mate with criminals? What if genetic pre-dispositions for violence and mental disorder are caused by analogous genetic substrates? Mentally ill offenders may be up to 25 times more likely to be homeless than the mentally ill in general (Martell, 1991). Perhaps the strains of being homeless contribute to violent behaviour, or that the homeless mentally disordered person may be less likely to receive medication and appropriate care. Clearly, whatever the relationship between mental disorder and violence, we must not equate the two.

In summary, the evidence does point to a link between mental disorder and violent offending. However, mental disorder’s contribution to violent crime in the community is generally agreed to be small, if only due to the small numbers suffering from such disorders. Although studies such as that of Link et al. (1992) have shown psychotic symptoms are related to violence "such symptoms are relatively rare and are no way as important as the influx of drugs, the breakdown of communities, and similar factors as causes of violent/illegal behaviour". Elevated rates of violent offending among those with major mental disorder are likely to be due to a small subset of those with severe and long term psychotic symptoms (e.g. type I, or paranoid schizophrenia characterized by a predominance of positive symptoms), co-morbidity in the form of alcohol/substance abuse, or those with serious personality disorder. Although public education campaigns claiming that the mentally ill pose no greater crime threat than the public at large may be ideologically rather than factually inspired, equal caution should be thrown upon the common stereotype of the mentally ill as violent, unpredictable offenders.