Until death do us part

Louise Dixon and Nicola Graham-Kevan look at the role psychology can take in preventing intimate partner homicide

Just under one quarter of all homicide victims in England and Wales were killed by an intimate partner in the year 2008/9, according to Home Office statistics. In the aftermath of such fatalities, where the offender was clearly well known to the victim, questions are often raised about whether the attack could have been foreseen and whether services had failed the victim in not preventing the sometimes seemingly inevitable event. This article considers how psychological theory and research can lend itself to the prevention of serious and fatal intimate partner violence and looks at the current state of practice in this domain.

‘Stalker guilty of stabbing ex-girlfriend to death in brutal attack’ (London Evening Standard, 2010)

‘Bridge game fights “led man to murder wife”’ (Reuters, 2010)

‘Horror as man kills himself after stabbing ex-partner outside Tesco store’ (Daily Record, 2010)

Attention-grabbing headlines like these that depict murderous scenes of male-perpetrated violence against a female intimate partner, are commonplace in the British media.

Lethal acts that hit the headlines are often preceded by risk factors that, with hindsight, show warning signs of the event that was about to unfold (as in the 2010 case of Raoul Moat, see Case 1 opposite). As psychology can be used to evidence patterns in behaviour, it is plausible that this branch of the social sciences will be able to contribute to the prediction and prevention of intimate partner homicide

Ideology or psychology?

It is important to understand the underlying theory of any social problem, as this will inevitably guide professionals’ understanding about the appropriate action that should be taken to prevent, reduce or eliminate it (Loseke et al., 2005). During the last four decades, feminist theory has exerted a large influence on how intimate partner violence is considered and responded to (White & Gondolf, 2000). This perspective views partner violence as a common event, acted out by men toward their female partners, caused by wider societal rules and patriarchal beliefs that encourage male dominance and female subordination (Dasgupta, 1999). From this vantage point, prevention and/or treatment in the short term concerns educating men so they understand their aggressive behaviour is caused by their gendered beliefs towards women. Ultimately addressing or eliminating the root causes of men’s aggression will be achieved by overturning patriarchal social structures, to prevent, reduce and/or eliminate violence to women (Dutton, 2006). However, for as long as this theory has been proposed, there has been considerable debate between academics and practitioners working in this area about its accuracy (Dutton, 2006).

Social science would dictate that in order to predict behaviour, it is necessary to comprehend its specific causes and correlates through an empirically proven understanding of the problem. From a psychological perspective, understanding the aetiology of an offender’s behaviour over time can be achieved from exploration of multiple factors, which range from those pertinent to the individual (e.g. personality disorder, childhood abuse history, substance abuse), to those resulting from the person’s interaction with their immediate (e.g. attachment style of the couple and psychopathology of the partner) and wider environment (e.g. influence of deviant peers, job stress and societal beliefs). Patriarchal beliefs may (or may not) be one factor that needs to be considered within an offender’s aetiological risk profile, such as Raoul Moat, but it is imperative that this is not the only factor that is considered. This nested ecological approach (see Dutton, 2006) stresses the importance of individual differences in a complex set of interacting factors. Thus from a psychological stance, prediction of behaviour is possible from a wide range of risk and/or protective factors that have been associated with an increased likelihood of severe and lethal intimate partner violence in past cases.

However, a recent parliamentary briefing paper that presents figures for ‘domestic violence and homicide’ suggests this may be problematic (Thompson, 2010). Since 1995, 2344 people have been killed by their intimate partner in England and Wales, 65 per cent of these were women. This low base rate makes it difficult to predict lethal violence with a high degree of accuracy. Furthermore, although many intimate partner homicides take place in the context of a violent relationship (approximate range 65–85 per cent: Dixon et al., 2008), a history of partner violence will overpredict lethality and, importantly, will not predict all lethal incidents. Such factors make it difficult for professionals (and victims) to predict which cases are likely to result in severe or lethal violence.

Is prediction of intimate partner homicide possible?

US-based researcher Jacquelyn Campbell has pioneered work that aims to identify which women are at risk of severe/lethal partner violence. Campbell’s early work identified correlates of male-perpetrated lethal and severe violence. She found an increase in the frequency or severity of physical violence over the preceding year, forced sex, threats to kill, controlling behaviours, separation, drug abuse, jealousy and violence outside the relationship and/or to children or during pregnancy, to be associated with lethality (see Campbell et al., 2003). Her more recent research identified pre-incident risk factors that differentiated lethal and non-lethal cases in 11 US cities with a high level of predictive accuracy. These were a perpetrator’s access to a gun, previous threats with a weapon, a perpetrator’s step-child in the home and estrangement, especially if the victim’s partner was controlling. Never living together and prior arrests for domestic violence lowered risk (Campbell et al., 2003).

This work has led to the development of the Danger Assessment-Revised (DA-R: Campbell et al., 2009), a 20-item risk-assessment tool that can be used as a predictive assessment of severe and lethal violence with female victims of partner violence. Therefore, Campbell has shown it is possible to identify high risk of partner homicide drawing upon multiple risk factors and empirical measures. However, are men who murder their female partner a homogeneous group – or are their differences between perpetrators that professionals should be aware of to aid prediction and intervention?

Profiles of killers

In attempting to understand the aetiology of male perpetration, empirical research has repeatedly demonstrated the presence of different types of offenders, each with a different aetiological risk profile. Holtzworth-Munroe and Stuart (1994) proposed three types of non-lethally partner-violent men from an examination of multiple factors: the Family Only perpetrator (FO), the Generally Violent Antisocial perpetrator (GVA), and the Dysphoric/Borderline perpetrator (DB) (see box on p.823). The typology has gathered support from several empirical studies that find evidence for some or all of the proposed subtypes (see Dixon & Browne, 2003). It is evident that the GVA and DB offenders are most likely to carry out severe and lethal harm. Indeed, the applicability of this typology to men convicted for intimate partner femicide has been demonstrated in an English sample, with the majority of 90 lethal offences committed by men who shared profiles consistent with the GVA (49 per cent) and DB (36 per cent) subtype (Dixon et al., 2008). This information supports the need for a multi-factor framework to guide a coherent assessment of a perpetrator’s risk of harm. Individual and comprehensive assessment is the key to understanding and preventing violent behaviour.

Female perpetrators?

Although the aforementioned has been crucial in developments of prediction of male-perpetrated homicide, such research fails to explain cases where women murder their husbands and boyfriends for example: ‘Wife kills womanising husband with a hammer’ (Daily Mail, 2010), ‘Wife blasted sick husband with shotgun at £75,000 home before drowning herself in the bath’ (Daily Mail, 2010) and ‘Wife kills husband in booze row’ (see Case 2 above). Thompson, (2010) has shown that 35 per cent of people killed by their intimate partner in England and Wales since 1995 are men. Furthermore, whilst more women who kill their male partners claim self-defence in comparison to men in the same situation, only a minority of these have their claim accepted (Nutall, 1993). Research demonstrates that men and women who use partner violence share similar risk factors, such as childhood maltreatment, substance use and personality disorders (e.g O’Leary et al., 2007), and although limited studies have examined typologies of female perpetrators, those that have find similarities with non-lethal female offenders (e.g. Babcock et al., 2003). Therefore, whilst this area of aggression remains underresearched in comparison to male-perpetrated partner homicide, the findings warrant a more in-depth understanding of female lethal and non-lethal partner violence than is currently available to date.

Can psychology prevent fatalities?

Psychological understanding of serious and lethal intimate partner violence can aid the prevention of the social problem through assessment and prediction. Several risk-assessment tools have been developed in North America. The majority of these have been designed to assess the likelihood of male non-lethal recidivism and include both actuarial tools with cut-off scores to indicate risk level and structured frameworks that act as aide-memoires to guide clinical judgement without precise cut-off scores. The Spousal Assault Risk Assessment (SARA: Kropp et al., 1999), a structured framework, is most commonly used in British practice, although it has been most commonly validated with North American samples. In terms of predicting severe or lethal violence, the 20-item DA-R has been developed to assess this risk for female victims with an acceptable degree of predictive validity in the US, and as such can act as an actuarial measure (Campbell et al., 2009).

However, the need for separate tools to predict intimate partner violence above and beyond well-established tools that predict general violence recidivism remains to be proven. A 2007 meta-analysis from Hanson and colleagues demonstrated that general violence risk assessment tools showed similar levels of accuracy in predicting male partner violence recidivism as partner violence specific tools. Whilst the jury is still out on this issue, it is apparent at least that empirically informed tools fare best. Their completion facilitates a thorough assessment, as they involve interviews with relevant parties, appropriate psychometric assessment and review of collateral information where possible.

Currently, the MARAC model of intervention (Multi-Agency Risk Assessment Conference) is implemented in England and Wales (see Robinson, 2004). In this model cases of domestic assault that come to the attention of professionals in relevant agencies (e.g. police, health, social care, specialist domestic violence agencies) are assessed to identify those at highest risk of severe harm. For very high-risk cases a specialist multi-agency response is conducted whereby information is shared between local agencies in order that a more effective coordinated safety plan can be assembled. Professionals use recommended risk-assessment forms to identify very high-risk cases. For example, police officers will complete initial risk-assessment forms at the scene, such as the South Wales Police Victim Initial Risk Indicator Form, which was developed from a review of 47 domestic homicides investigated by their force (Robinson, 2004). Similar factors to those identified in the US homicide literature are used, such as past physical abuse and its escalation, weapon use, financial problems, substance abuse, pregnancy, jealous or controlling behaviour, relationship separation, threats, sexual abuse and suicidal thoughts.

Small-scale research on the use of the MARAC process found the majority of victims (66 per cent) were not re-victimised up to six months post-MARAC (Robinson, 2004). It is difficult to make sense of these figures without corresponding re-victimisation data for similar cases where MARACs were not used. Additionally, the predictive validity of the Risk Indicator Form in differentiating high-risk women from other risk categories remains to be established. Many risk assessments for partner violence are based upon historical factors, so are not appropriate for treatment planning or predicting situations that may enhance or reduce risk. Considering this, and the findings from Hanson’s 2007 meta-analysis, it may be considered that general violence risk assessments such as the Historical Clinical Risk-20 (HCR-20: Webster et al., 1997), or empirically informed partner violence specific tools like the SARA, are better placed to predict future risk and plan interventions. However, it is clear that in order to predict lethal or non-lethal partner violence with any degree of accuracy professionals must carry out comprehensive, non-biased assessments that examine the presence of multiple risk factors for both partners. Evidence-based risk-assessment tools help this process.

In addition to the use of empirically developed measures, thorough psychological assessment should try to understand the perspective of the perpetrator of partner homicide, if we are to be able to recognise high-risk situations in the future. With the exception of assaults by psychopathic individuals, most violence is preceded by a perceived provocation by the victim (Reidy et al., 2010). Such provocation may appear to outsiders as very minor (e.g. an argument about who is going to the shops), but this needs to be understood not by the appearance of the actions of those involved, but by the feelings invoked and hence the function of those actions to the individuals. Other ‘provocations’ are objectively extremely difficult for most people (e.g. losing your home and access to your children) and should be automatically flagged as ‘high-risk’ situations by professionals working with perpetrators or couples and considered in the context of the presence of other risk factors. Many people who have personality disorder traits find conflict extremely threatening and, coupled with an inability to self-soothe, this can lead to even minor disagreements resulting in escalating distress. Such individuals are effectively in crisis, and their aggression is an expressive manifestation of this. Professionals involved in cases of known partner violence should be particularly vigilant to signs of crisis and adopt crisis-intervention techniques (Ireland et al., 2011).

Therefore, thorough assessment may be able to reduce the likelihood of severe or lethal partner violence in cases that come to the attention of relevant professionals, although further development of tools applicable to British cases is needed. However, it remains difficult to predict the small number of over-controlled offenders that do not show some of the commonly recognised antecedents. In addition, the majority of work to date examines male-to-female violence; applicability of such tools to female heterosexual and same-sex perpetrators remain to be established.

What can we conclude?

Can psychology lend itself to the prevention of fatal attacks against intimate partners, such as that by Raoul Moat, and many others like it? It is difficult to predict individual behaviour that occurs at a low base rate, and therefore it is not possible to predict risk correctly 100 per cent of the time. However, it is apparent that a psychological understanding can promote thorough assessment, which incorporates the use of valid empirically derived risk-assessment tools, an awareness of the multi-factor nature of intimate partner violence and the context in which it occurs – all of which can certainly increase the possibility that professionals understand and estimate risk of lethal violence in relationships with a good degree of accuracy.

Box: Case 1

The case of Raoul Moat, 37 from Newcastle upon Tyne, put partner homicide and its preceding risk factors in the media spotlight. Moat, with a history of assault and aggression and recently released from prison, believed his ex-partner, Samantha Stobbart (who was the mother of their child), had started a new relationship with a police officer. While in prison Moat made threats to seriously harm Stobbart, and within two days of release posted threats to the police and Stobbart on Facebook. He then shot her, her new partner Chris Brown and police officer David Rathband. Brown died, the others were seriously injured. After a six-day police hunt he eventually shot himself in a dramatic standoff after hours of negotiation. The questions remain: Could this serious and fatal violence have been predicted from the risk factors that were present (e.g. history of serious violence outside the relationship, jealousy and threats to seriously harm victim/s). Could the victims have been warned and protected from Moat?

Box: Case 2

‘A Leeds woman [Buck] has been jailed for life after stabbing her partner to death in a row over whose turn it was to buy booze from the off-licence… Leeds Crown Court heard how a series of drunken 999 calls had been made from Buck’s home…in the months leading up to the death… Three months earlier Buck had received a formal police caution for assaulting Mr Rider with a glass at the property. Neighbours had also described seeing Mr Rider slapping Buck across the face on another occasion. The court heard how the couple both struggled with chronic alcohol problems.’ (Yorkshire Evening Post, 2009). This case highlights how risk factors (such as previous use of/threats with weapons, high relationship conflict and substance abuse) can also be present in the lives of females who perpetrate severe and lethal partner violence. As such both sexes warrant investigation as potential perpetrators, victims, or both.

Box: Types of perpetrator

The Family Only (FO) perpetrator limits violence to family members; this is of low severity and frequency and demonstrates the least criminal behaviour and psychopathology. Their profiles of distal and proximal risk most closely resemble non-violent controls (distal factors occur in childhood or before – e.g. genetic factors; proximal factors are adult characteristics).
Intimate partner violence is likely to result from an amalgamation of low-level risk factors, such as: some exposure to family violence in childhood; poor communication skills with their partner; mild impulsivity; dependency on their partner; exposure to violence in the family of origin; alcohol and drug abuse. They are likely to use reactive aggression when involved in conflict and emotions are high.

The Generally Violent Antisocial (GVA) perpetrator possesses multiple distal and proximal risk factors that increase the likelihood of moderate to severe levels of violence both within and outside the family. They have the highest levels of exposure to violence in their childhood; extensive involvement with delinquent and deviant peers; high levels of impulsivity, substance abuse, criminality, antisocial personality, and narcissism; negative attitudes toward women; attitudes supportive of violence in general. They lack conflict-resolution skills in a wide variety of situations and have a dismissive attachment style. They display low levels of empathy, psychological distress and depression, and moderate levels of anger. They are likely to engage in instrumental aggression to a partner to get their own way or in situations where they feel the need to keep or regain control and feel powerful, such as if they perceive the other had disrespected or rejected them in some way.

The Dysphoric/Borderline (DB) perpetrator is also likely to carry out moderate to severe violence, primarily to family members. They have some experience of family of origin violence and involvement with deviant peers; demonstrate the highest levels of psychological distress, emotional volatility, depression and anger; hold moderate attitudes supportive of violence and hostility to women; display low to moderate levels of empathy, criminality and substance abuse; show moderate impulsivity; and have low marital communication skills. They display characteristics of borderline personality and preoccupied or fearful attachment, reacting with anger when they feel rejected, abandoned or slighted. They are most likely to use reactive aggression, particularly around issues related to estrangement, where they may engage in stalking behaviours in attempts to maintain or re-establish the intimate relationship.


Louise Dixon is in the Centre for Forensic and Criminological Psychology, University of Birmingham

Nicola Graham-Kevan is in the School of Psychology, University of Central Lancashire


Babcock, J.C., Miller, S.A. & Siard, C. (2003). Toward a typology of abusive women. Psychology of Women Quarterly, 27, 153–161.

Campbell, J.C., Webster, D., Koziol-McLain, J. et al (2003). Risk factors for femicide in abusive relationships: Results from a multi-site case control study. American Journal of Public Health, 93, 1089–1097.

Campbell, J.C., Webster, D.W. & Glass, N.E. (2009). The Danger Assessment: Validation of a lethality risk assessment instrument for intimate partner femicide. Journal of Interpersonal Violence, 24, 653–674.

Dasgupta, S.D. (1999). Just like men? A critical view of violence by women. In M.E. Shephard & E.L. Pence (Eds.) Coordinating community responses to domestic violence (pp.195–222). London: Sage.

Dixon, L. & Browne, K.D. (2003). The heterogeneity of spouse abuse: A review. Aggression and Violent Behavior, 268, 1–24.

Dixon. L., Hamilton-Giachritsis, C. & Browne, K.D. (2008). Classifying partner femicide. Journal of Interpersonal Violence, 23, 74–93.

Dutton, D.G. (2006). Rethinking domestic violence. Vancouver: University of British Columbia Press.

Hanson, R.K., Helmus, L. & Bourgon, G. (2007). The validity of risk assessments for intimate partner violence: A meta-analysis (User Report No. 2007-07). Ottawa: Public Safety Canada.

Holtzworth-Munroe, A. & Stuart, G.L. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116, 476–497.

Ireland, C.A., Fisher, M. & Vecchi, G.M. (2011). Conflict and crisis communication: Principles and practice. Hove: Routledge.

Kropp, P.R., Hart, S.D., Webster, C.D. & Eaves, D. (1999). Manual for the Spousal Assault Risk Assessment guide (3rd edn). Toronto: Multi-Health Systems.

Loseke, D.R., Gelles, R.J. & Cavanaugh, M.M. (2005). Section I: Controversies in conceptualisation. In D.R. Loseke., R.J. Gelles & M.M. Cavanaugh (Eds.) Current controversies on family violence. (pp.1–4). Thousand Oaks, CA. Sage

Nutall, C. (1993). Letter published in The Sunday Times cited in Dewar Research response to Partial defences to murder (Law Commission Consultation Paper No.173, October 2003). Retrieved from www.dewar4research.org/docs/


O’Leary, K.D., Smith Slep, A.M. & O’Leary, S.G. (2007). Multivariate models of men’s and women’s partner aggression. Journal of Consulting and Clinical Psychology, 75, 752–764.

Reidy, D.E., Foster, J.D. & Zeichner, A. (2010). Narcissism and unprovoked aggression. Aggressive Behavior, 36, 414–422.

Robinson, A.L. (2004, June). Domestic violence MARACs for very-high risk victims in Cardiff, Wales: A process and outcome evaluation. Retrieved from www.caada.org.uk/research/


Thompson, G. (2010). Domestic violence statistics (Standard Note SN/SG/950). House of Commons Library. Retrieved from tinyurl.com/5vz6aqu

Webster, C.D., Douglas K.S., Eaves, D. & Hart, S.D. (1997). HCR-20: Assessing risk for violence (version 2). Burnaby, BC: Mental Health Law and Policy Institute, Simon Fraser University.

White, R.J. & Gondolf, E.W. (2000). Implications of personality profiles for batterer treatment. Journal of Interpersonal Violence, 15, 467–488.


BPS Members can discuss this article

Already a member? Or Create an account

Not a member? Find out about becoming a member or subscriber