Attachment and Delinquency
Two main theories addressing the relation between attachment and delinquency are the social control theory (Hirschi
1969; Sampson and Laub
1993) and attachment theory (Ainsworth
1979; Bowlby
1973). The first theory is a criminological theory that was developed by Hirschi (
1969), who conceptualized attachment as an affective bond through which children internalize conventional norms of society. According to Hirschi, delinquency will be low in families with strong affective ties, because juveniles who are strongly attached to their parents are more likely to care about the normative expectations of their parents, which protects against delinquent impulses. The quality of attachment functions as an
indirect parental control: conventional behavior of the child is achieved as a by-product of strong child-parent attachments (Hirschi
1969). Delinquent behavior, however, will increase if the bond to the parent is weak.
The second main theory that addresses the link between attachment and delinquency is attachment theory. Observing mothers and their infants, Bowlby (
1973) identified an intense distress experienced by infants who had been separated from their parents. These children were crying and clinging to their mother when they were to be separated. Bowlby called these behaviors attachment behaviors, which are natural adaptive responses to separation that serve an evolutionary function. If the parent-child attachment relationship is disrupted during infancy, long-term negative consequences are the inability to show affection or concern for others and aggressive and delinquent behavior (Bowlby
1944,
1973).
Over the years, scholars began to investigate attachment relationships between adolescents and their parents and attachment relationships in young adulthood (e.g., Ainsworth
1989), focusing on
attachment representations rather than observing attachment behaviors in infants. Bowlby (
1988) believed that from infancy, children internalize and organize patterns of relating to people. They can mentally represent their experiences with attachment figures and construct ideas and expectations about relationships with primary as well as secondary attachment figures. As such, an internal working model of attachment is gradually shaped during development (Bretherton
1990). Next to observational methods that measure attachment in early childhood in terms of behavioral responses to caregivers, questionnaires and (in-depth) interview methods were developed to measure attachment representations in middle childhood (e.g., Dwyer
2005; Kerns et al.
2000), adolescence (Allen and Land
1999) and (young) adulthood (Bretherton
1990).
In sum, attachment theory is a theory of both normal and abnormal development that focuses on the impact of parent-child attachment relationships on healthy development and psychopathology, including juvenile delinquency (Sroufe et al.
1999). Control theory is a theory of abnormal development that mainly focuses on explaining criminal behavior on the basis of social control. Essential for control theory is the affectional bond that the child forms to his or her parents, which is thought to influence child behavior through the psychological presence of the parent. Whereas an affectional bond is a necessary but not a sufficient condition for attachment in attachment theory (see Schuengel and Van IJzendoorn
2001), it is both a necessary and sufficient condition for attachment in control theory. Despite these differences between attachment theory and social control theory, it is safe to hypothesize that poor or disturbed attachment is associated with risk for delinquency.
Several previous meta-analyses have been conducted that focused on the link between insecure attachment relationships and problem behavior in children. For example, Van IJzendoorn et al. (
1999), analyzing 12 studies, found a significant link between disorganized attachment and externalizing problem behavior. Also, in a more recent meta-analysis it was found that attachment insecurity was significantly linked with externalizing behaviors in 69 studies (Fearon et al.
2010). Unfortunately, these meta-analyses did not differentiate between the two main types of externalizing problems, that is, aggression and delinquency (e.g., Achenbach
1991). Given the significance of delinquency among adolescents and young adults as a problem in society, we exclusively focus on delinquency in this study. Previous research has shown that delinquency and aggression are conceptually distinct types of behaviors. For example, the trajectories of aggression and delinquency have been found to show a different developmental pattern (Bongers et al.
2004). Furthermore, different risk factors have been found for aggressive and delinquent behavior (Dishion and Patterson
2006). Therefore, it is seems important to conduct a separate meta-analytic study of the association between attachment and delinquency, which enables studying a number of specific moderators of the association between attachment and delinquency, such as the source of information on delinquency, and perhaps most importantly, the concept of attachment.
Sex-differences
It was not until feminist criminologists criticized mainstream theories for exclusively focusing on criminal behavior of males (Daly and Chesney-Lind
1999) that scholars began to investigate potential explanations of sex differences in delinquency, referred to as the gender gap. Feminist scholars argued for gender sensitive theories of crime (Miller and Mullins
2009). According to control theorists, processes that enhance or prevent delinquency are gender neutral, and the gap in delinquency between males and females can be explained by differences in the quality of bonds to parents (Agnew
2009). Social ties to parents are stronger in females than in males, which explains why females’ levels of delinquency are lower than those of males.
Some scholars disagree on whether sex-differences exist in attachment relationships. According to Del Giudice (
2009), sex-differences exist in the quality of the attachment relationship, yet secure attachments are similarly often found in males and females. In a nutshell Del Giudice states that females are more vulnerable to develop anxious-ambivalent patterns of attachment, while males are expected to adopt the avoidant attachment type. As a consequence, males are more likely to adopt competitive and aggressive traits and externalizing problems, such as delinquency, whereas females are more likely to develop internalized problems, such as anxiety and depression (Del Giudice
2009). However, Bakermans-Kranenburg et al. (
2005) criticized Del Giudice’s model for lack of empirical evidence. Their meta-analyses did not support the hypothesis of sex-differences in attachment.
Studies of sex-differences in the link between attachment and delinquency are scarce and their findings are contradictory. Some studies reported stronger effects of attachment in girls (e.g. Nye
1958), whereas others concluded that the family bond is more important to boys. For example, it was found that family strain (Hay
2003) and the quality of parental caregiving (Rothbaum and Weisz
1994) had stronger effects on males than on females. Insecure attachment was more strongly linked to externalizing behavior in samples with boys than in samples with girls (Fearon et al.
2010). A third group of researchers found very few sex-differences in family related factors of delinquency (Hubbard and Pratt
2002; Loeber and Stouthamer-Loeber
1986). Moffitt et al. (
2001) extensively investigated potential sex-differences in the prevalence of risk factors and the impact of family risk factors on delinquency and concluded that, in general, boys seem to be more exposed to risk factors of delinquency, rather than that they are more vulnerable for risk factors of delinquency compared to girls. Thus, findings as to whether there are sex differences with regard to the link between attachment and delinquency have been inconsistent and a meta-analysis could shed light on this issue.
Apart from potential differences between males and females in the link between attachment and delinquency, differences may appear depending on the sex of the parent for several reasons. First, the
quantity of the time fathers and mothers spend with their children is different. In general mothers spend more time in taking care of their children (Dubas and Gerris
2002), which could result in stronger ties to mothers and a stronger effect of the quality of the mother-child bond on the child’s behavior. Second, there are indications that parental involvement is also
qualitatively different (Videon
2005). Offending behavior by the father predicts delinquent behavior of their sons (Farrington et al.
2001). Moreover, the longer antisocial fathers live with their families the higher the risk for their children’s antisocial behavior (Jaffee et al.
2003). A possible explanation for this finding is that attachment relationships to these fathers are more problematic. Attachment relationships are influenced by the extent to which the parent is sensitive to the needs of the child (De Wolff and van IJzendoorn
1997). Although fathers and mothers have been found to be equally capable of sensitive responding, fathers tend to exercise their sensitivity less often (Lamb
1982). This is essential since delinquency is more common among boys than among girls. Despite this possibility, both control theory and attachment theory are not very clear about the effects of attachments to fathers. Moreover, relatively little research has examined the quality of attachment relationships to fathers compared to attachment to mothers in relation to the child’s delinquent behavior (Williams and Kelly
2005). Finally, it is unknown whether the consequences of insecure attachment relationships or poor bonds to parents with regard to delinquency gradually change over time and whether this potential change is different for boys and girls.
The Present Study
Many studies found evidence to suggest that poor attachment relationships to parents increase the risk of delinquent behavior. The inconsistencies in the literature, however, make it difficult to summarize the results in a narrative review. Previous studies did not summarize potential sex-differences in the link between attachment and delinquency, and it is therefore unknown whether the attachment-delinquency link is gender-specific or neutral. Given that studies examining attachment to mothers and fathers separately are scarce, it remains unclear whether attachment to father has a different effect on delinquency in sons and daughters than attachment to mother. Finally, previous studies have not systematically investigated whether the attachment-delinquency link varies by age, that is, whether this link is stronger in younger children than in adolescents and early adults.
The aim of the present meta-analytic study is therefore to examine the extent to which attachment is related to delinquency and to gain insight into sex and age differences. This study could advance theories on attachment and delinquency, particularly with regard to potential conceptual differences of attachment constructs in social control and attachment theories, potential gender differences (debate on whether gender specific theories are necessary) and age differences (static versus dynamic theories). It may also provide starting points for developing or improving interventions for delinquent youth or children at risk for delinquency.
The present meta-analytic study addresses the following research questions: Is poor attachment significantly associated with high levels of delinquency? How strong is the association between attachment and delinquency? Is age a moderator of the attachment-delinquency link? Is the association between attachment and delinquency moderated by sex of the child and the parent? We also test whether the attachment-delinquency link depends on the theoretical background of the study (control theory or attachment theory) and measure of attachment (observation, self-report, parent report or multi-method), and whether conceptual differences in attachment constructs (presence of parental control items) influence the relation between attachment and delinquency. Further, we examine moderator effects of methodological characteristics (sample size, publication status, study design and delinquency source) in order to investigate potential influences of study quality on effect size. We also address the relative importance of the moderators and interaction effects of sex and age in a multivariate model.
Discussion
The present meta-analytic study summarizes and integrates previous findings on the link between attachment and delinquency, and examines factors that could have a moderating effect on this association, including age and sex. Results of this meta-analysis confirm previous findings in that poor attachment to parents is associated with more delinquent behavior. According to the criteria of Cohen (
1988) for small, moderate and large effect sizes, the strength of the association between attachment and delinquency should be considered small to moderate (
r = 0.18). This finding is relatively similar for studies examining the association between attachment and delinquency from the perspective of control theory and attachment theory.
Testing the potential effects of conceptual differences in attachment among studies revealed that not the theoretical background but the type of attachment measure moderated study outcomes. Studies that used parent reports and combined methods to measure attachment resulted in a relatively strong link between attachment and delinquency. This is consistent with findings of the meta-analysis by Fearon et al. (
2010) showing that the relation between attachment and externalizing behavior problems was moderated by the type of attachment measure. A first explanation for this effect could be that type of attachment measurement is confounded with age. Particular attachment measures are typically used with subjects of specific ages, for example, behavioral observation for infants and attachment interviews (e.g., AAI) and self-report questionnaires for late adolescents and (young) adults. We found larger effect sizes with parent-report and mixed methods, which are commonly used with younger children, than with the AAI and self-report measures, which are used with late adolescents and (young) adults. However, given that both age and type of attachment measurement remained significant in the multivariate models, age cannot (entirely) explain the effect of measurement type.
Second, some scholars have argued that the type of attachment measure can have an effect on the study outcomes, because different attachment instrument may measure different attachment concepts (Van IJzendoorn and Bakermans-Kranenburg
2004). Parent reports of attachment and mixed methods had a stronger focus on the quality of the parent-child relationship and the bond between parent and child, but less on attachment security or internal working models of attachment. It should be noted that in particular security and internal working models of attachment are key concepts in the theory of attachment that was originally developed by John Bowlby. The most valid way to assess attachment in adolescents and (young) adults is considered to be the Adult Attachment Interview (AAI; Schuengel et al.
2011). It is in particular interesting to observe that (juvenile) delinquents in (forensic) institutional facilities show high rates of insecure and disorganized attachment (Zegers et al.
2008; Van IJzendoorn et al.
1997), which may have important consequences for treatment. Smith et al. (
2010), for instance, showed that therapeutic alliance is affected by attachment (in) security. Schuengel and Van IJzendoorn (
2001) considered attachment to be important in institutionalized treatment because the internal working model of attachment may affect quality of therapist-client relationships, treatment outcomes, and because institutionalization may have an effect in itself on attachment representations and future attachment behavior of the client.
Further, the combination of attachment and parental control, including supervision, rules setting and strictness had more impact on the child’s behavior than child-parent attachment relationships alone. Studies that used instruments to measure attachment, including parental control items, revealed larger effect sizes. This is not in line with the assumption of Hirschi (
1969) and Nye (
1958), who state that supervision and parental control is probably limited for adolescent children, because youths in this age period spend less time with their parents and are relatively autonomous. Given that most of the time parents are not able to supervise their children, direct parental control, such as supervision and discipline, are considered by control theorists as etiologically less relevant as a predictor of delinquency compared to indirect control, such as attachment (e.g., Hirschi
1969; Nye
1958). Our findings with regard to parental control do, however, not concur with our meta-analytic results on the link between parenting and delinquency, as we found moderate mean effect sizes for several types of parental control, including poor parental supervision (
r = 0.23), and psychological control (
r = 0.23; Hoeve et al.
2009). Possibly, attachment is more important at younger ages, while parental control remains to be an important risk factor of delinquency, given that the attachment-delinquency link was found to become weaker at older ages (current meta-analysis) while the parental control-delinquency link was not moderated by age (Hoeve et al.
2009). In conclusion, the supposition of control theory and attachment theory that disturbed attachment is linked to delinquency is supported by the current findings. However, in contrast to these theories, it seems that parental control and discipline are at least as or even more important for predicting delinquent behavior.
Effect sizes of more or less similar magnitude were found for males and females, indicating that poor bonds to parents similarly explain delinquency in both boys and girls. This finding may not refute the assertion of feminist criminologists that gender-sensitive theoretical models are needed to explain sex-differences in delinquency. Notably, the attachment-delinquency link may partly explain female delinquency, but does not provide an explanation as to why boys are more often involved in delinquent behavior than girls. Hagan (
1999) explains this sex difference by pointing to the impact of gender-specific socialization processes within families that are related to differences in the distribution of power between the father and the mother. For example, in patriarchal families power relations between fathers and mothers are unequal. Such inequality may result in more parental control toward daughters than sons, which could offer an explanation for the gender gap in delinquency. It is also possible that different types of insecure attachment are related to different problem behaviors, as has been suggested by Del Giudice (
2009). This is a promising future line of research.
In the present study, we found that attachment to mothers was more important for girls, while attachment to fathers was more important for boys with regard to the development of delinquency. Given that delinquency is more prevalent in boys, this finding has important implications for theory and practice. First, disturbed attachment to fathers should be recognized as a risk factor for delinquency in boys. Second, interventions that target delinquency in boys could profit from the involvement of fathers. This is in accordance with a recent study of Lundahl et al. (
2008) showing that children benefited more if fathers attended a parent training compared to programs that focused on mothers only.
We found that the strength of the association between attachment and delinquency was negatively related to age, suggesting that the influence of attachment to parents on delinquency weakens as youngsters become older. These findings contradict the static perspective of Gottfredson and Hirschi (
1990), who state that criminal potential results from several factors early in life, including poor attachment, and that criminal propensity remains relatively stable over time. Our finding is more consistent with the theory of Sampson and Laub (
2005), which assumes that correlates of delinquency may change during the life-course. According to Sampson and Laub, changes in life circumstances can generate turning points in an individual’s criminal career. Delinquent behavior is inhibited during childhood and adolescence by bonds to the family and school. During (young) adulthood, social ties to labor or marriage and other turning points in life can modify trajectories of criminal offending.
Several limitations of the present meta-analysis should be noted. First, for reasons of comparability, we only included studies reporting bivariate associations. Multivariate analyses give insight into the unique contribution of attachment to delinquency by simultaneously controlling for other factors. In general, attachment to parents was significantly linked to delinquency in these studies (e.g. Bernburg and Thorlindsson
1999; Ford
2005; Laundra et al.
2002; LeBlanc
1994) or attachment was even the most important predictor of delinquency (e.g., compared to other family factors and economic factors, Mack et al.
2007), while some studies found only weak links between social bonding and delinquency (e.g., compared to prior delinquency and delinquent peers, Agnew
1991). We found 23 manuscripts reporting only multivariate results. Meta-analyzing multivariate associations, however, is problematic, because the effect size statistics of interest depend on other variables in the multivariate model (Lipsey and Wilson
2001, p. 69). The effect sizes derived from multivariate models with different control variables are therefore not comparable with each other and also not comparable with effect sizes derived from bivariate analyses. Moreover, conducting moderator analyses could result in artifactual findings if study characteristics that are tested as moderators would have been used as control variables in the multivariate models of primary studies from which effect sizes are included in the meta-analysis.
Second, this meta-analysis is about
associations between parental attachment and delinquency, which limits the causal interpretation of our study findings. Although it can be derived from attachment theory that poor attachment can lead to delinquency (Bowlby
1944; Hirschi
1969), research has shown that parents do not only influence their children and the attachment relationship, but that parent-child relationships develop as a result of complex interactions between parents and children (e.g., Crouter and Booth
2003; Granic
2000; Holden
1997). Thus, relationships with parents may also have been affected by the individual characteristics of youngsters, which does not allow drawing conclusions about the direction of the effect. Also, the main aim of this study was to examine the association between attachment and delinquency, however, it would be interesting to know which insecure attachment patterns are most strongly linked to delinquency. Surprisingly, only five studies distinguished between insecure patterns of attachment. More work needs to be done on this issue.
Despite these limitations, this meta-analysis has several strengths. This study connected two areas of research by focusing on empirical studies that have tested hypotheses drawn from social control theory and assumptions from attachment theory. The results were consistent across disciplines. Second, applying a multi-level model made it possible to include all effect sizes with their unique potentially moderating characteristics in the analysis, at the same time correcting for statistical dependency. Third, this meta-analysis tested specific hypotheses drawn from the literature with regard to moderating effects of gender and age rather than testing a series of moderators that are not theory driven as is common in meta-analytic investigations, but increases the risk of chance capitalization. Finally, focusing on a set of limited theoretically based moderators made it possible to build a multivariate moderation model in which the main and interaction effects of age and gender were analyzed, controlling for methodological moderators.
In conclusion, this meta-analysis confirmed that attachment to parents is linked to delinquency in boys and girls. We found several sex differences and age effects: stronger attachment-delinquency links were found in same-sex parent-child pairs compared to cross-sex pairs, and the attachment-delinquency link was stronger for younger participants compared to older participants. These findings suggest that attachment to parents is a viable target for interventions aimed at reducing or preventing delinquency. To be more specific, boys could especially benefit from interventions that focus on relationships with their father, whereas delinquent girls could especially benefit from interventions that focus on the attachment relationship with their mother. When carrying out (preventive) interventions for young delinquents or children at risk for delinquency, one should consider the attachment relationship with parents as a target of intervention. Next to insecure attachment or week bonds to parents these interventions should also aim to improve discipline techniques of parents.