Elsevier

Aggression and Violent Behavior

Volume 19, Issue 4, July–August 2014, Pages 322-339
Aggression and Violent Behavior

Attachment & violent offending: A meta-analysis

https://doi.org/10.1016/j.avb.2014.04.007Get rights and content

Highlights

  • We conducted a systematic review and meta-analysis on attachment and violent offending.

  • Offenders were less secure in their attachments than controls.

  • Insecure attachment was strongly associated with all types of criminality.

  • This association was still present in the absence of mental disorder.

Abstract

Attachment theory provides a useful framework for understanding violence as it acknowledges the importance of both interpersonal and developmental factors. The literature suggests that attachment is associated with violence, but the research evidence is equivocal as to whether insecure attachment was a risk factor for criminality, psychopathology more generally, or both. The current study therefore conducted a systematic review of the literature using meta-analytic methods. Results indicated that insecure attachment was strongly associated with all types of criminality (i.e. sexual offending, violent offending, non-violent offending, and domestic violence) even in the absence of psychopathology. Further sub-group analyses indicated differences in attachment patterns between sexual offenders and violent offenders, for example. The implications of the findings are discussed and suggestions for further research are made.

Introduction

Interpersonal violence often occurs between individuals who are bonded or attached to each other; regarded by Meloy (2003) as “one of the great paradoxes of human existence” (p. 509). It is therefore not surprising that there has been a growing interest in the relationship between attachment and violence. Originally conceptualised by Bowlby (1969) and studied in infants, attachment is considered to be a bio-psychosocial system that promotes survival by ensuring that children maintain proximity to their caregivers, particularly when in the presence of threat. Thus, attachment is inevitable, but the formation and course of the bonds forged (i.e. whether adaptive or maladaptive, secure or insecure) will be determined largely by environmental factors, the quality of caregiver interactions, and how the relationship is perceived by the infant (Rich, 2006).

As part of the attachment behavioural system, infants have an innate repertoire of behaviours (e.g. smiling, crying, etc.) that are designed to elicit caregiver proximity (Wallin, 2007). Such proximity and attachment bonds also promote the establishment of a “secure base” (Bowlby, 1979). That is, when attachment figures are available as a source of protection and support, their children then feel safe to explore their environment, thus aiding emotional, cognitive, and physical development (Wallin, 2007). The secure base also provides a safe haven and optimally a source of comfort should retreat become necessary (Bowlby, 1979). As children develop representational skills, they begin to construct mental representations of their own secure base experiences which serve to retain the lessons of prior experience whilst remaining malleable to a degree in light of significant new experience (Waters, Crowell, Elliott, Corcoran, & Treboux, 2002).

However, physical proximity of a caregiver in and of itself is not sufficient for the formation of a secure attachment bond as parents can be emotionally unavailable to their children for a variety reasons (e.g. due to mental health difficulties or substance misuse) even if they are physically present (Wallin, 2007). Furthermore, the child's appraisals of the attachment relationships with caregivers (i.e. “felt security”) are crucial, as attachments are not only relevant during the early years, but rather serve as the template for how we behave in and perceive our relationships throughout the lifespan (Rich, 2006). Consequently, for the developing child, the function of the attachment relationship goes beyond providing security; it is the mechanism through which the infant learns about themselves and the social world (Shaw & Dallos, 2005).

Gergely and Watson (1999) suggest that children have an innate “contingency detection module” through which the earliest forms of self-awareness develop. During the first two to three months of life infants have been observed to seek out and explore stimuli that are perfectly response contingent (e.g. repetitive leg kicking) (Schmuckler, 1996). It is thought that this allows the infant to determine what they have complete control over, thus enabling the development of a primary representation of the bodily self, which in turn helps equip the baby to cope with its environment (Gergely & Watson, 1999). From the age of approximately three months and beyond, infants switch from self exploration to representation and exploration of their social environment, commencing with caregiver interactions (Gergely & Watson, 1999). For example, the child learns a series of “if-then” contingencies that will predict future behaviour. Any learning that takes place in the context of the attachment relationship will differ depending on the nature of the relationship itself (i.e. whether secure or insecure, with a secure attachment being the optimal bond). For example, the securely attached child might learn “if I cry, then I will be comforted”, whereas the insecurely attached child might learn “if I cry, then I will be ignored” (Shaw & Dallos, 2005). Bowlby (1969) suggested that infants internalise these early social experiences by the end of their first year into “internal working models” which then facilitate the child's understanding and predictions of how their world might unfold. “Thus an unwanted child is likely not only to feel unwanted by his parents, but to believe that he is essentially unwantable, namely unwanted by anyone” (Bowlby, 1973, p. 204).

A working model of the self, or the foundation for the self-concept, begins to develop in this way, as does the infant's internal working models of other people (Shaw & Dallos, 2005). Research has demonstrated that the extent to which these models contain positive or negative information is a function of attachment security (Mikulincer & Orbach, 1995), and that affect regulation strategies are organised around these beliefs, thereby affecting the development of the child's emotional functioning (Mikulincer, Shaver, & Pereg, 2003). Furthermore, any insecure patterns established in childhood are likely to transfer to adolescence and beyond, thus potentially having consequences for all future social interactions (Shaw & Dallos, 2005). Attachment theory has therefore provided a model of psychopathology as well as a model of normal development since its inception. The formation and normal course of attachment relationships, as well as the implications of atypical attachment patterns (and their relationship to psychopathology), are central to the theory (Sroufe, Carlson, Levy, & Egeland, 1999).

Indeed, Bowlby (1973) proposed that the nature of the internal representations of attachment experiences (i.e. whether based on secure or insecure bonds) would lead to individual differences in personality development, as well as varying behavioural and emotional responses in the context of interactions with attachment figures. These ideas were further developed and operationalized by Ainsworth, Blehar, Waters, and Wall (1978) who examined individual differences in infant attachment security under experimental conditions using the “strange situation” procedure, observing children's reactions to brief periods of separation from their mother, as well as contact with a stranger, in a controlled environment. The infant's response to the departure and return of their caregiver, and their levels of exploration and play throughout were noted to fall broadly into one of three patterns.

Ainsworth et al. (1978) categorised these behavioural responses as being either secure, anxious/ambivalent, or avoidant; whereby the securely attached child would show distress initially during separation but appeared happy when the mother returned. The anxious-ambivalent child, on the other hand, would show an extreme distress response in the face of separation but demonstrates ambivalence upon reunion (e.g. seeking proximity whilst rejecting comfort). Finally, the avoidant child would demonstrate little in the way of emotional responsivity during the initial separation or reunion, almost as if the presence of the mother was of no significance. A fourth category termed “disorganised/disorientated” was added later when Main and Solomon (1990) observed a range of behaviours that could not be coded under the three way classification system. These behaviours included, for example, stereotypies, freezing, an apparent fear of the parent, confusion, and disorientation. Such actions were considered to be characteristic of a lack of coherent attachment strategy, and were most commonly observed in abused/traumatised infants (Solomon & George, 1999).

The strange situation procedure was the starting point for an abundance of research which indicated that different attachment strategies are adopted depending upon the individual's history of interactions with attachment figures (Holtzworth-Munroe, Stuart, & Hutchison, 1997). A number of methods for assessing attachment in adulthood have also been developed, with one of the most well-known being the Adult Attachment Interview (AAI; George et al., 1985) which is used to categorise an individual's attachment representations using a semi-structured interview methodology. Different terminology and categories are used throughout the attachment literature to describe variance in attachment strategy, but the three major categories originally identified by Ainsworth et al. (1978) (i.e. secure, avoidant, and anxious-ambivalent) are usually present in most attachment measures as either categorical or dimensional constructs (Fraley & Waller, 1998). When assessing adult attachment using the AAI, the “disorganised/disorientated” category from the strange situation procedure corresponds with the “unresolved/cannot classify” AAI category (Main & Solomon, 1990). It is also worth noting that in the classification of adult attachment patterns, the term “dismissive” is often used interchangeably with “avoidant”, and that the term “preoccupied” is often used to represent attachment anxiety (Ross, 2004).

Bowlby's (1969) theory of attachment was originally conceptualised to explain the evolutionary function of infant–caregiver bonds. However, the theory has since been expanded, and it is now widely accepted that the attachment system is active throughout the human life span (Cassidy & Shaver, 1999).

In adulthood, close relationships between individuals (e.g. intimate partners) can take on the quality of attachment relationships in that they can provide a sense of security and belonging (Ainsworth, 1991). However, as the individual goes through different life stages, the strategies that maintain attachments will change. For example, in infancy, physical proximity to caregivers is important whereas this is much less the case in adulthood, particularly for those who are secure in their attachment style (Cassidy & Shaver, 1999). That said, the interpersonal style of the adult and their corresponding attachment strategies will continue to be influenced by the internal working models developed in childhood, meaning that the securely attached adult will regard themselves as deserving of attachment and others as being able to meet their needs, whereas the opposite is likely to be true for those considered insecurely attached (Ma, 2006).

There is a sizeable literature concerned solely with the measurement of attachment in adulthood, partly due to the emergence of two distinct traditions (Bartholomew & Shaver, 1998). These two largely independent schools of thought are based on different assessments and conceptualisations of adult attachment (Cassidy & Shaver, 1999). As described above, Ainsworth et al. (1978) began the first line of research by introducing the strange-situation procedure to assess attachment in childhood. This was later expanded upon with the introduction of the AAI (George, Kaplan, & Main, 1985), an interview method designed to assess an adult's state of mind with regard to their childhood experiences of attachment relationships (Shaver, Belsky, & Brennan, 2000). The AAI assessment process involves discussing emotionally laden attachment related experiences whilst a trained rater codes the individual's discourse in terms of coherence (Hesse, 1999). The individual being interviewed is assigned one of four adult attachment categories (i.e. secure/autonomous, dismissing, preoccupied, unresolved/disorganised) that are conceptually comparable to the infant categories in the infant strange situation procedure.

The AAI is often referred to as the “gold standard” in the measurement of adult attachment (Waters et al., 2002), partly due to the high convergent validity between parental AAI classifications and infant strange situation classifications (van Ijzendoorn, 1995), and the clinical utility of the information gained during the interview. However, in order to be able to use the AAI, extensive training is required, and the procedure itself takes several hours of transcription and coding (Ma, 2006). Unfortunately this means that it is not particularly accessible to the average clinician. It can also be difficult to utilise for research purposes as a great deal of resources are required due to the fact that it is essentially a qualitative process with a quantitative (i.e. categorical) outcome.

The second tradition began in the mid 1980s when Hazan and Shaver (1987) applied attachment theory to romantic relationships on the premise that the strategies employed in such partnerships would be derived from childhood attachment experiences. The dominant form of measurement employed by this tradition is that of self-report questionnaires, a methodology that has been criticised at times for its “…(theoretically) limited ability to tap into unconscious attachment strategies and…vulnerability to defensive reporting” (Ma, 2006, pp. 442–443). However, it can be counter-argued that conscious and unconscious processes often function simultaneously in order to obtain the same goal, and that most adults have the ability to describe their behaviour in close relationships on the basis of reflections on their behaviour within them (Bartholomew & Shaver, 1998). Due to ease of administration and scoring, self-report measures of attachment are accessible to interested clinicians and widely used in research.

The four category model of adult attachment devised by Bartholomew and Horrowitz (1991) on the basis of self-report outcomes is well validated and frequently used in research concerning adult psychopathology and offending. The model conceptualises attachment across two dimensions of anxiety and avoidance. Four prototypic attachment styles can then be defined in relation to the intersection of the two underlying dimensions: secure; preoccupied; dismissing; and fearful.

The four attachment categories are indicative of prototypic strategies for regulating felt security in close relationships (Griffin & Bartholomew, 1994). A distinct pattern of interpersonal behaviour and style of emotion regulation characterises each category (Ma, 2006). For example, secure individuals (positive self and other models, low anxiety and low avoidance) have an adequate sense of self worth and can find the balance between intimacy and independence (Bartholomew & Horrowitz, 1991). In contrast, fearful individuals (negative self and other models, high anxiety and high avoidance) are highly dependent on others in an attempt to validate their self-worth, but tend to shy away from intimacy in an attempt to avoid the rejection expected on the basis of their negative model of others (Bartholomew & Horrowitz, 1991). Dismissing individuals (positive self and negative models, low anxiety and high avoidance) also avoid intimacy because of their negative other model, but retain their high self-worth by over-valuing independence and denying the value of relationships (Bartholomew & Horrowitz, 1991). Finally, preoccupied individuals (negative self and positive other models, high anxiety and low avoidance) seek excessive closeness and idealise others as a result of their positive other model, but are often rejected and experience extreme distress when the attachment figure is unable to fulfil their intimacy needs (Griffin & Bartholomew, 1994). These four attachment styles can be assessed using a range of self-report measures (Ma, 2006).

There is much debate in the literature as to whether attachment should be measured in categorical or dimensional terms. Crittenden (2000) asserts that it is inaccurate to consider attachment as a categorical phenomenon as attachment behaviours are complex and variable by degree, and thus can be better understood dimensionally. Brennan, Clark, and Shaver (1998) share the same view, suggesting that categorical models have become commonplace merely as a result of convenience. Categorical models have also been criticised as they apply a label to a person (e.g. Mr X is dismissing in his attachment style) that is seen as all encompassing in that this classification is thought to apply in all circumstances and to all attachment figures (Crittenden, 2000). Dimensional models, on the other hand, allow for variance in attachment across time and context, thus an individual would be described as more or less secure at the time of assessment as opposed to “secure” or “insecure” (Rich, 2006). However, despite support for dimensional models, categorical measures are still used routinely, and are often considered best practice as is the case with the AAI, for example. Waters and Beauchaine (2003) argue that it makes no difference whether a categorical or dimensional classification system is used as the underlying principles of attachment theory are sound, and as such, it is at the discretion of the clinician or researcher as to how they wish to assess attachment.

From an evolutionary perspective, violence can be seen to be adaptive for survival, and thus would be considered part of normal development (Buss & Shackelford, 1997). From a developmental perspective, violence is also seen as part of normal development, but is thought to be unlearned rather than learned through “…social processes that…regulate and tame it” (Fonagy, 2003, p. 190). This is contradictory to what has long been one of the primary explanatory models of violence, that is, that aggression is learned in response to frustration and through observing others model such behaviour, and reinforced through the achievement of goals (Tremblay, 2008). However, a number of longitudinal studies would support the notion that violence is unlearned, as the data suggest that physical aggression peaks between the ages of two and three and steadily decreases thereafter for most individuals (Coté et al., 2006, Liben and Bigler, 2002, Tremblay et al., 2004). Fonagy (2003) proposes that violence is unlearned as part of normal development in that secure attachment experiences facilitate the development of control over innate aggressiveness. However, where the home environment is dysfunctional and the child is unable to form secure attachment bonds, the process of unlearning violence becomes disrupted.

Support for this hypothesis comes from a study by Gilliom, Shaw, Beck, Shonberg, and Lukon (2002) in which children's ability to manage anger in a frustration task was examined. Attachment was assessed at the age of 18 months in a sample of 310 boys, who were followed up until the age of 6 when performance on the frustration task was assessed. Children who were assessed as securely attached at 18 months were found to be more likely than those who were assessed as insecurely attached to seek support and to disengage from the task as opposed to expressing inappropriate levels of anger. Furthermore, studies of children experiencing attachment difficulties have found them to be significantly more disruptive, aggressive and antisocial (Levy and Orlans, 2000, Lyons-Ruth et al., 1993). In adults, Fossati et al. (2009) found a relationship between insecure attachment and impulsive aggression and Critchfield, Levy, Clarkin, and Kernberg (2008) found an association between high levels of attachment anxiety and avoidance and multiple forms of aggressive behaviour (e.g. verbal assault, physical assault, and self harm) in a sample of participants with borderline personality disorder.

High levels of attachment insecurity have been observed in violent offenders (e.g. Ross & Pfafflin, 2007), sexual offenders (e.g. Wood & Riggs, 2008), and perpetrators of domestic violence (e.g. Chiffriller & Hennessy, 2010). However, insecure attachment is surprisingly common in the general population (approximately 40% would be considered insecure in their attachment orientation) in comparison with offending, which is relatively rare (van Ijzendoorn, 1995). Consequently, insecure attachment in isolation does not provide an adequate model of offending. Levinson and Fonagy (2004) hypothesized that deficits in the ability to mentalise (i.e. the ability to understand mental states in self and others) may mediate the relationship between insecure attachment and offending and found some support for this. However, the relationship between insecure attachment and offending in and of itself must be understood before the influence of mediating variables is explored.

The primary aim of this review and meta-analysis is to synthesize the studies available on attachment in offending populations. A clear understanding of how attachment relates to violence is crucial to inform the development of therapeutic interventions for violent offenders, violence risk formulation and management of offenders, with the aim of reducing violent recidivism (Webster & Hucker, 2007). In addition, the meta-analysis aims to examine whether insecure attachment is related to mental health problems, criminality, or both, as there is conflicting evidence across studies. van Ijzendoorn et al. (1997) in their study of attachment in 40 personality disordered offenders concluded that “…insecure attachment may be a general mental health risk factor, rather than a specific determinant of severe criminal behaviour” (pp. 456). However, this conclusion was later brought into question by Baker and Beech (2004), who found high rates of attachment insecurity in non-offenders with mental disorders. The aim of this review is to investigate these issues in a systematic manner. Thus, in addition to considering whether offenders were more insecure in their attachments than normal controls, differences between offending populations were explored in more detail. Recommendations for further research are made.

  • Main Research Question

    • a)

      Are offenders more insecure in their attachments than non-offending controls?

  • Subsidiary Research Questions

    • b)

      Do offenders with mental disorders differ from non-offending mental health populations in terms of attachment security?

    • c)

      Do offenders with mental disorders differ from non-offenders with mental disorders in terms of attachment security?

    • d)

      Are there differences in attachment security between violent and non-violent offenders?

    • e)

      Does attachment vary according to offence type (e.g. sexual offences, violent offences)?

Section snippets

Search strategy

A comprehensive search was conducted using the following procedures.

Results

Thirty papers were included in the review (27 published and 3 unpublished); the papers included a total of 2798 offenders. Table 1 outlines each study, including the sample characteristics, the measures used, and the key findings. The following abbreviations are used in Table 1 and throughout the Results section: SOs = sexual offenders, VOs = violent offenders, DVM = domestically violent men, NVOs = non-violent offenders, PD = personality disordered offenders and MDOs = offenders with mental disorders. The

Discussion

Through a meta-analysis of 30 studies that included an overall total of 2798 offenders, the relationship between attachment security and offending was examined. The results indicated that offenders were less secure in their attachment than non-offending controls; medium to large effects were observed. Furthermore, small to medium effects were noted with respect to attachment anxiety and avoidance when comparing offenders with non-offending controls, with offenders significantly more anxious and

References (90)

  • K. Bartholomew et al.

    Attachment styles among young adults: A test of a four-category model

    Journal of Personality and Social Psychology

    (1991)
  • K. Bartholomew et al.

    Methods of assessing adult attachment: Do they converge?

  • S. Bogaerts et al.

    Recalled parental bonding, adult attachment style, and personality disorders in child molesters: A comparative study

    The Journal of Forensic Psychiatry & Psychology

    (2005)
  • J. Bowlby

    Attachment and loss: Vol 1

    Attachment

    (1969)
  • J. Bowlby

    Attachment and loss: Vol 2

    Separation: anxiety and anger

    (1973)
  • J. Bowlby

    The making and breaking of affectional bonds

    (1979)
  • K.A. Brennan et al.

    Self-report measurement of adult attachment. An integrative overview

  • J. Cassidy et al.

    Handbook of attachment: Theory, research, & clinical applications

    (1999)
  • S.H. Chiffriller et al.

    Empirically generated typology of men who batter

    Victims & Offenders

    (2010)
  • C.R. Cloninger et al.

    The Temperament and Character Inventory: A Guide to its Development and Use

    (1994)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • N.L. Collins et al.

    Adult attachment, working models, and relationship quality in dating couples

    Journal of Personality and Social Psychology

    (1990)
  • S. Coté et al.

    The development of physical aggression from toddlerhood to pre-adolescence: A nationwide longitudinal study of Canadian children

    Journal of Abnormal Child Psychology

    (2006)
  • K.L. Critchfield et al.

    The relational context of aggression in borderline personality disorder: Using adult attachment style to predict forms of hostility

    Journal of Clinical Psychology

    (2008)
  • P.M. Crittenden

    A dynamic-maturational approach to continuity and change in patterns of attachment

  • V. de Vogel et al.

    Female additional manual: Additional guidelines to the HCR-20 for assessing risk for violence in women

    (2012)
  • D.G. Dutton et al.

    Intimacy–anger and insecure attachment as precursors of abuse in intimate relationships

    Journal of Applied Social Psychology

    (1994)
  • J.A. Feeney et al.

    Attachment style as a predictor of adult romantic relationships

    Journal of Personality and Social Psychology

    (1994)
  • Fischer, A., & Haapasalo, J. (1998). Childhood adversities and adult depression of insane acquittees in light of...
  • P. Fonagy

    Towards a developmental understanding of violence

    British Journal of Psychiatry

    (2003)
  • P. Fonagy

    The developmental roots of violence in the failure of mentalization

  • A. Fossati et al.

    Alexithymia and attachment insecurities in impulsive aggression

    Attachment & Human Development

    (2009)
  • R.C. Fraley et al.

    Adult attachment patterns: A test of the typological model

  • A. Frodi et al.

    Attachment & Human Development

    (2001)
  • George, C., Kaplan, N., & Main, M. (1985). Adult Attachment Interview. Unpublished Manuscript. Berkely: University of...
  • G. Gergely et al.

    Early social–emotional development: Contingency perception and the social biofeedback model

  • M. Gilliom et al.

    Anger regulation in disadvantaged preschool boys: Strategies, antecedents, and the development of self-control

    Developmental Psychology

    (2002)
  • H. Goldstein et al.

    Empathy and attachment in relation to violent vs. non-violent offense history among jail inmates

    Journal of Offender Rehabilitation

    (2001)
  • D. Griffin et al.

    Models of the self and other: Fundamental dimensions underlying measures of adult attachment

    Journal of Personality and Social Psychology

    (1994)
  • S.D. Hart et al.

    The Risk for Sexual Violence Protocol

    (2003)
  • Y. Hawkins-Rodgers et al.

    Nonviolent offenders and college students attachment and social support behaviours: Implications for counselling

    International Journal of Offender Therapy and Comparative Criminology

    (2005)
  • C. Hazan et al.

    Romantic love conceptualised as an attachment process

    Journal of Personality and Social Psychology

    (1987)
  • E. Hesse

    The adult attachment interview: Historical and current perspectives

  • A. Holtzworth-Munroe et al.

    Violent versus nonviolent husbands: Differences in attachment patterns, dependency and jealousy

    Journal of Family Psychology

    (1997)
  • S. Jamieson et al.

    Attachment styles and violence in child molesters

    The Journal of Sexual Aggression

    (2000)
  • Cited by (73)

    • The role of anxious attachment in the continuation of abusive relationships: The potential for strengthening a secure attachment schema as a tool of empowerment

      2022, Acta Psychologica
      Citation Excerpt :

      Dutton and White (2012) explain how attachment theory plays a role in IPV. Studies have shown that many of the individuals who experience IPV demonstrate insecure attachment orientations (Ogilvie et al., 2014; Ponti & Tani, 2019), and these orientations predict both the perpetration of IPV and IPV victimization (Bélanger et al., 2015). For example, compared to their counterparts, individuals high in attachment anxiety are at higher risk of IPV perpetration.

    • Psychological criminology: An integrative approach

      2023, Psychological Criminology: An Integrative Approach
    View all citing articles on Scopus
    View full text