Attachment and Violence
Measures of attachment have been negatively associated with violence in many studies (e.g., Alarid, Burton, & Cullen, 2000; Bellair et al., 2003; Benda & Corwyn, 2002; Ellickson & McGuigan, 2000; Franke, 2000; Gottfredson et al., 1991; Kesner, Julian, & McKenry, 1997; Kesner & McKenry, 1998; McNulty & Bellair, 2003; Rebellon & Van Gundy, 2005; Salzinger, Rosario, & Feldman, 2007). There has been a paucity of reviews on this subject, and it is yet unclear whether attachment has a special relationship with violence or not. Dozier, Stovall-McClough, and Albus (2008) conclude that the findings on separation and antisocial behavior are “relatively consistent” (p. 736) when researchers look at violence rather than antisocial personality. In a companion paper, we have published a longer treatment of our own review of the association between attachment, physical aggression, and violence (Savage, 2014a).
Because the research on attachment and violence spans all ages, beginning in very early childhood, this topic is particularly interdisciplinary. As such, measures of attachment vary a great deal. We note that no seminal review has produced a consensus about the attachment-violence association. Therefore, in this chapter, we elected to include all studies of associations between attachment indicators and measures of physical aggression and violent behavior, not just those studies where the authors also reported findings for nonviolent antisocial behavior as we will do in some of the coming chapters. We did attempt to locate all studies of attachment/parental bonding and nonviolent-only offending as well (mainly theft and other property crime; not including drug-only offending). We note that, because so much of the research centers on young children, among whom measures of purely nonviolent antisociality are practically nonexistent, what we learn about attachment and nonviolent offending is based on data from adolescents.
We have elected to be inclusive in our operationalization of attachment. Separation and loss are often seen as proximate for severe attachment problems (e.g., Harlow, 1958; Rutter et al., 2001). Theoretically, the loss of a relationship with an important caregiver should have significant effects at least partly because of the loss of the attachment relationship. Therefore, studies of parent loss and changes in caregivers early in life are included. We will also consider all indicators of parental bonding and caregiver sensitivity. Parental attachment and bonding among older children and adolescents is likely to reflect several important pieces of the puzzle-the adolescent’s capacity to form such relationships (which may be indicative of early attachment relationships), the ongoing relationship between parent and child (which is likely to be autocorrelated over time), and the parent’s sensitive and responsive behavior. Caregiver sensitivity is the crux of the attachment construct. It is less likely to be confounded by child effects, and it is most the amenable among the various operationalizations of attachment to intervention.
A search of the Proquest Research Library, Criminal Justice Abstracts, and PsycINFO databases was conducted. The search terms combined the word “attachment” or related terms (maternal sensitivity, paternal sensitivity, parental sensitivity, caregiver sensitivity, maternal bonds, paternal bonds, parental bonds, caregiver bonds, bonds with mother, bonds with father, bonds with caregiver, social bonds, social control theory) and terms for antisocial behavior
(delinquency, violence, violent, property crime, property offenses, theft, status offenses, nonviolent or non-violent, aggression, antisocial, arrest, conviction, conduct disorder, criminal, deviance, externalizing, externalization, offending, and oppositional defiant). Here we include attachment to parents and caregivers, not attachment to romantic partners or peers. Thus the many studies about attachment relations and domestic violence are mostly absent (unless the authors report associations between domestic violence and the perpetrator’s attachment to his or her parents). In addition, all relevant items we came across through other means-previous readings, looking through reference lists-were also added to our list. This resulted in a master list of more than 500 sources.
That list was culled for studies likely to provide estimates of associations between attachment and violence, physical aggression, or nonviolent antisocial behavior. Browsing the master list resulted in 83 tier one items, those with the highest likelihood of having both violent/physically aggressive and nonviolent/ general aggressive outcomes reported, and nearly 400 tier two items, with titles that were more ambiguous—suggesting that the studies tested associations between attachment and any measure of antisocial conduct. We looked at all of these studies.
We used the following criteria for determining whether or not each study’s findings would be included in our own summary tables. Regarding outcomes, we included only studies with measures of physical aggression or violence. Measures of “externalizing” were only included if the description suggested that the measure was limited to physical aggression. Therefore, most studies using the Child Behavior Check List are excluded, unless the authors limited their analysis to the aggression subscale. Regarding measures of attachment, we included any study using measures that appeared to be indicative of parental attachment, but not attachment to other parties, such as domestic partners. This included measures of parental bonding, attachment categories determined using the Strange Situation procedure, indicators of parental loss or separation, and parental sensitivity. Because we separately examine the association between parental warmth and rejection and violence in Chapter 7, we included studies of those constructs elsewhere. Regarding laboratory-manipulated attachment, such as attachment interventions, we did not include findings merely related to the intervention itself, but only those where attachment was actually measured and not assumed from intervention participation.