Abstract
The aim of the current study was to investigate whether the association between parental attachment and externalizing behavior is parallelly mediated by dysfunctional emotion regulation and callous-unemotional traits. The community sample included 296 adolescents (Mage = 14.90, SDage = 1.31), who completed the Inventory of Parent and Peer Attachment, the Inventory of Callous-Unemotional Traits, the Regulation of Emotions Questionnaire, and the Youth Self-Report Child Behavior Checklist. A mediation analysis revealed both direct and indirect effects. Secure attachment representations were negatively associated with internal- and external dysfunctional emotion regulation strategies and callous-unemotional traits, which in turn were positively associated with externalizing behavior problems. The current study expands previous research by simultaneously investigating familial and cognitive factors that foster externalizing behavior problems. An attachment-based perspective offers new possibilities for theory expansion, research directions, and the development of interventions.
Externalizing behavior problems are one of the most prominent psychopathologies that place adoles-cents at risk of concurrent and lifetime maladjustment, including school dropout (Hetlevik, Bøe, & Hysing, 2018), suicidal behavior (Soto-Sanz et al., 2019), and substance use (Pedersen et al., 2018). Given these severe consequences, an extensive exploration of the predictive factors is essential to guide theory and future interventions. Based on the developmental psychopathology perspective of externalizing behavior, existing (compartmentalized) conceptual models and empirical evidence should be integrated to simultaneously investigate factors at the individual and contextual level (Hinshaw & Beauchaine, 2015). In the current study we investigate externalizing behavior within an attachment theory perspective, delineating a cascading model that incorporates individual differences in emotion regulation (ER) and callous-unemotional (CU) traits in adolescents.
Theoretical Background
Attachment refers to a strong emotional bond that is established during infancy between a child and their parents (Bowlby, 1982). A secure attachment is fostered when parents are consistent and sensitive towards their infant’s needs and stress signals. In contrast, insecure attachments are fostered by parents who are dismissive of their child’s needs or are inconsistent in their responsiveness. Based on these early attachment interactions, children develop internal working models that guide expectations, emotions, and behaviors within relationships over the lifespan. During and beyond adolescence, the internalized working model can be described as a mental representation of the self (e.g., self-worth, self-efficacy) and others (e.g., support and responsiveness expectations; Mikulincer & Shaver, 2007). Hence, research on attachment in adolescence places a greater focus on the cognitive-based representations of the internal working models (Gullone & Robinson, 2005). Representations are reflected in the appraisal of the relationships with parents, i.e., the individual’s perceived quality of parental attachment (Raja, McGee, & Stanton, 1992). As such attachment representations can be conceptualized dimensionally, with the ends of the continuum signifying high secure attachment versus low secure attachment, i.e., insecure attachment (Armsden & Greenberg, 1987; Madigan, Brumariu, Villani, Atkinson, & Lyons-Ruth, 2016). Within this conceptualization secure parent-child attachments are characterized by trust, good communication, and a lack of interpersonal alienation (Gullone & Robinson, 2005; Raja et al., 1992). Attachment is a complex developmental construct, in which early parent-child interactions, established internal working models, and attachment representations in adolescence cannot be viewed in isolation, but rather as an intertwined holistic construct.
Attachment representations play an essential role in children’s and adolescents’ social-emotional development. Secure attachment representations foster self-reliance, emotion regulation capabilities, and social competence (Sroufe, 2005). Conversely, insecure attachments lay the foundation for maladaptive developmental processes that often lead to psychopathology (Brown & Wright, 2001). A meta-analysis focusing on childhood and adolescence by Madigan et al. (2016) confirmed this; specifically, they found support for the link between attachment and internalizing and externalizing behavior problems. These analyses reveal that insecure attachment representations plays a significant role in the presentation of externalizing behavior in adolescence 1 , yet a closer examination of the mediational mechanisms is needed to fully understand this association.
Both theory and research suggest that the association between insecure attachment and externalizing behavior is in part attributable to the link between attachment and ER (Brumariu, 2015). The functionalist account of emotions proposes that emotions provide valuable information about situations and enhance an individual’s capacity to cope with them (Phillips & Power, 2007). Regulatory strategies that use this information (i.e., retaining and processing the emotions) facilitate behavior that is direct towards well-being, thereby proving to be functional (Phillips & Power, 2007). Regulatory strategies that do not use this information (i.e., suppression) are considered dysfunctional (Phillips & Power, 2007). According to attachment theory, it is through attachment interactions that individuals learn how to regulate their distressing emotions, which integrated into their working models. Specifically, children learn about emotion regulation from their parents through socialization methods, which include direct communication about strategies and observation (Brumariu, 2015). Mothers who formed secure attachments with their children, tend to utilize emotion coaching and better regulate their own emotions when interacting with their child (Brumariu, 2015). Furthermore, secure relationships offer a consistent source of support, and parent-child conversations help shape emotion appraisal and understanding (Thompson, 2014). In contrast, insecure relationships result in the child having a limited understanding of emotions and regulation thereof (Thompson, 2014). Insecurely attached children are believed to either minimize or maximize negative affect (i.e., suppress or heighten expression; Cassidy, 1994). This paves the way for emotion dysregulation, which in turn increases vulnerability to stress and the development of psychopathology (Cole & Hall, 2008). Specifically, difficulties with ER may impede empathic or prosocial responding in social situations (Lemerise & Arsenio, 2000), with studies demonstrating that this results in externalizing behavior in adolescents (Compas et al., 2017).
Early in its conceptualization, attachment theory was used to explain the “affectionless” personality of juvenile offenders, whose lack of an established bond with caregivers resulted in the absence of concern for others (Saltaris, 2002). Individuals with insecure attachments develop inner working models in which others are deemed unworthy of trust and empathy; they fail to internalize and display moral and emotional commitment to others, which paves the way for the development of CU traits (Saltaris, 2002). CU traits refer to an enduring affective and interpersonal style, commonly characterized by a lack of guilt and empathy, limited display of emotions, and using others for personal gain (Frick et al., 2003). Extending on an earlier framework by Frick and Viding (2009) with a developmental psychopathological perspective, Frick, Ray, Thornton, and Kahn (2014) state that the presence of high CU traits designates a distinct group of (children and) adolescents with conduct problems with deficits in their conscience development. The framework is based largely on research which has demonstrated that CU traits are positively associated with various externalizing behaviors in adolescence, including aggression and delinquency (Frick & White, 2008). Concerning their association with attachment, a meta-analysis revealed a small-to-medium positive association between psychopathic traits and insecure attachment, with the largest effect size reported for CU traits (van der Zouwen, Hoeve, Hendriks, Asscher, & Stams, 2018). Hence, CU traits could present another mediational mechanism to explain the association between attachment and externalizing behavior.
The Current Study
It is important to investigate the factors that lead to externalizing behavior, including their interplay and pathways, to provide vital insights that can advance theory and interventions. Although a review of the literature revealed connections between the above described factors, only a few studies integrated these in mediational models and even less focused on the adolescent age-range. Furthermore, Rowe et al. (2010) note that relatively little is known about children with high CU traits but without conduct disorder; thus, highlighting the need for studies to investigate these associations with community (non-clinical) samples. The aim of the current study was to investigate whether the association between parental attachment and externalizing behavior is mediated by CU traits and dysfunctional ER in adolescence. We postulated that a more secure parental attachment representation will be negatively associated with externalizing behavior. Specifically, we hypothesized that (1) attachment will be negatively associated with dysfunctional ER, which in turn will be positively associated with externalizing behavior, and that (2) attachment will be negatively associated with CU traits, which in turn will be positively associated with externalizing behavior. Thus, we expect dysfunctional ER and CU traits to mediate the association between attachment and externalizing behavior in parallel.
Method
Procedure and participants
The recruitment of participants was undertaken with a voluntary response sampling technique. Following a convenience sampling method, secondary schools in northern Germany were asked in 2018 if they would like to participate. In the schools which agreed to participate (seven in total), informed consent forms were distributed to the parents of students in Grade 7 to 12. Adolescents who obtained parental consent, also received consent forms before partaking in the study. The data collection was anonymous. Each adolescent received a questionnaire booklet, which they completed independently in supervised group settings at school.
A total of 296 adolescents participated in the study. The sample included 165 girls and 131 boys, and their ages ranged from 13 to 18 (M = 14.90, SD = 1.31). Table 1 depicts the gender distributions across the grades. Although being homogenous, the sample was not equally distributed across school type (1.4% general secondary school [Hauptschule], 20.6% intermediate secondary school [Realschule], 0.3% grammar school [Gymnasium], 15.2% unified comprehensive school [Oberschule], 7.8% special education school [Förderschule], and 54.7% integrated comprehensive school [IGS]). In total 71.3% indicated that they themselves and both their parents were born in Germany, and 83.8% stated that German is their mother tongue.
Gender Distributions Per Grade (n)
Measures
The Inventory of Callous-Unemotional Traits (Frick, 2003) was used to assess CU traits. The questionnaire consists of 24-items, with participants responding on a scale ranging from (0) not at all true to (3) definitely true. The questionnaire assessed their callousness (e.g., feelings of others are unimportant), unemotionality (e.g., hiding feelings from others), and tendency to be uncaring (e.g., easily admit being wrong – reverse scored). Combining these three aspects as a total score provides a general indicator of CU traits (Ray & Frick, 2020). A meta-analysis has shown acceptable reliability and validity for the total CU score (Cardinale & Marsh, 2020), as have studies with a German sample (Essau, Sasagawa, & Frick, 2006).
To assess externalizing behavior, participants completed the Youth Self-Report checklist (Döpfner, Plück, & Kinnen, 2014), which is a direct adaptation of the Child Behavior Checklist (CBCL; Achenbach, 1991). On a scale ranging from (0) not true to (2) very or often true, participants indicated how often they displayed certain behaviors in the past six months. A total of 32 items assessed both rule-breaking behaviors (e.g., lying and/or fibbing) and aggressive behavior (e.g., being mean to others), which were combined to externalizing behavior. Walter and Remschmidt (1999) report that the German version of the CBCL has adequate reliability and validity.
The short version of the Inventory of Parent and Peer Attachment (IPPA; Raja et al., 1992; Armsden & Greenberg, 1987), was used to assess parental attachment. The inventory dimensionally assesses attachment security, i.e., adolescents perceived atta-chment quality with their parents. In the current study we combined the first 10-items, which focused on communication (e.g., telling parents about problems), trust (e.g., parents respect feelings), and alienation (e.g., not getting attention at home – reverse scored). Regarding the interaction with their parents, participants responded on a scale ranging from (1) almost never or never to (4) almost always or always. Variations of the IPPA have shown adequate reliability and validity (Gullone & Robinson, 2005), and it has been successfully applied with German samples (e.g., Essau, 2004).
To assesses the usage of dysfunctional ER strategies, participants were asked to complete the Regulation of Emotions Questionnaire (REQ; Phillips & Power, 2007). Dysfunctional ER was assessed with 10 items, and included both internal-dysfunctional (e.g., keep feelings locked inside) and external-dysfunctional strategies (e.g., take feelings out on other people). Participants responded on a scale ranging from (0) not at all to (4) always. Phillips and Power (2007) report adequate validity, and the REQ has been utilized with German samples (Kullik & Petermann, 2013).
Data analyses
Data analyses were completed with SPSS and AMOS. As each variable consisted of multiple items, we opted to compute means if up to 90% were completed. If more than this amount was incomplete, the variables were treated as missing values. An analysis of missing data patterns revealed that 1.26% of values were missing, afflicting 13 cases. Although Little’s MCAR test was significant (χ 2 =45.62, p = .03), we proceeded with listwise deletion as the amount of missing data is considered negligible (see Jakobsen et al., 2017). The following data analysis is hence based on 157 girls and 126 boys (M age = 14.94, SD age = 1.31). Preliminary analyses were conducted to examine the associations between the variables, and to identify age and gender differences. A linear regression revealed that multicollinearity between the variables was unlikely (VIF <1.50). An examination of the Q-Q plots showed that most variables were approximately normally distributed, except for external-dysfunctional ER. In our mediation model, attachment was added as the independent variable, externalizing behavior as the dependent variable, and CU traits as well as internal- and external-dysfunctional ER as parallel mediators. Gender was added as an independent variable, including a covariance with attachment. The model included covariances of error terms between CU traits and the ER strategies. Total, direct, and indirect effects were calculated with 5000 bootstrap samples and 95% confidence intervals. As the development of externalizing behavior encompasses numerous factors (Hinshaw & Beauchaine, 2015), we expected a complementary mediation, in which both the direct and indirect effect remain significant (Zhao, Lynch, & Chen, 2010).
Results
Table 2 depicts the correlations between the in-cluded variables, as well as the recorded mean scores. Almost all correlations amongst the variables were significant, and in the expected direction. The only exception was external-dysfunctional ER, which did not correlate with attachment nor with internal-dysfunctional ER. Age only correlated with attachment (r = – 0.13, p = 0.03); this led to the decision not to include age in the mediation model. Testing for gender differences, we found that boys reported higher attachment, t(281) = 2.58, p = 0.01, external-dysfunctional ER, t(281) = 3.13, p < 0.01, and CU traits, t(281) = 2.34, p = 0.02, whilst girls reported higher usage of internal-dysfunctional ER, t(281) = – 5.59, p < 0.01. No gender differences were observed for externalizing behavior, t(281) = 1.19, p = 0.24.
Means, Standard Deviations, Correlations, and Internal Consistency of the Variables
Note. *p < 0.05, **p < 0.01.
The mediation model demonstrated good fit values (χ2/df = 1.188, p = 0.276, CFI = 0.999, NFI = 0.997, TLI = 0.991, RMSEA = 0.026 [CI 90% (0.163, 0.428)]). Attachment demonstrated a direct effect on externalizing behavior (β= – 0.44, p < 0.01, 95% CI [– 0.53, – 0.35]). Table 3 depicts the direct effects from the mediation model. Attachment was negatively associated with internal- and external-dysfunctional ER, CU traits, and externalizing behavior. Internal- and external-dysfunctional ER, as well as CU traits were positively associated with externalizing behavior. Furthermore, attachment had an indirect effect on externalizing behavior (β= – 0.17, p < 0.01, 95% CI [– 0.25, – 0.09]), whilst gender did not (β= – 0.08, p = 0.06, 95% CI [– 0.15, 0.003]). The amount of variance explained by the model was 15% for CU traits, 23% for internal-dysfunctional ER, 5% for external-dysfunctional ER, and 48% for externalizing behavior.
Direct Effects from the Mediation Model
Notes. 1 = boys, 2 = girls. *p < 0.05, **p < 0.01.
Discussion
The aim of the current study was to demonstrate that the association between attachment and externalizing behavior is mediated by both dysfunctional ER and CU traits in adolescence. As expected (see Madigan et al., 2016), we found that parental attachment negatively predicted externalizing behavior, i.e., the more secure the attachment representation, the lower the reported externalizing behavior problems. Thus, even in adolescence, when it is often assumed that peers become the central social entities, a secure attachment to parents remains important. In line with our first hypothesis, we found that the more secure the attachment representation was, the lower the reported dysfunctional ER strategies, and the higher the dysfunctional ER strategies the higher the externalizing behavior. This aligns well with previous research that has supported a direct effect of attachment on ER (Brumariu, 2015), and ER on externalizing behavior (Aldao, Gee, De Los Reyes, & Seager, 2016). Brenning and Braet (2013) found that in adolescence anger dysregulation acted as a mediator between insecure attachment (dimensions) and aggressive behavior, whereas anger/sadness suppression and sadness dysregulation did not. Although not quite comparable to our study, due to the different conceptualizations of attachment and ER, their findings also highlight that ER plays an important mediational role.
In our study, we found that both external- and internal-dysfunctional ER mediated the association between attachment and externalizing behavior. Effect sizes indicate that attachment is more strongly associated with internal-dysfunctional ER, whilst external-dysfunctional ER is more strongly associated with externalizing behavior. However, there are a variety of methodological and statistical problems with the external-dysfunctional ER strategies, and findings should be interpreted with caution. Specifically, (1) there may be an overlap of the assessed strategies and externalizing behavior (i.e., the content of the items), (2) external-dysfunctional ER was right skewed (i.e., potential floor effect), and (3) external-dysfunctional ER did not correlate with attachment, yet was significant in the model. However, it is still noteworthy that similar disparities between external- and internal-dysfunctional ER have been found in mediation models of attachment and internalizing behavior (Kullik & Petermann, 2013), as well as with other maladaptive ER strategies (Brenning & Braet, 2013).
In line with our second hypothesis, the more secure the attachment representation was, the lower the reported CU traits, and the higher the CU traits, the higher the externalizing behavior. This conforms with previous research, which found that CU traits predict externalizing behavior (Dadds, Fraser, Frost, & Hawes, 2005; Fanti, Frick, & Georgiou, 2009), and attachment predicts of CU traits (van der Zouwen et al., 2018). A recent review illustrated that attachment plays a role in fostering empathy (Stern & Cassidy, 2018), indicating that empathy development might be a mechanism with which attachment counteracts the development of CU traits. Furthermore, moral disengagement has been found to mediate the association between attachment and delinquency in adolescence (Bao, Zhang, Lai, Sun, & Wang, 2015). This may indicate another mechanism by which attachment influences CU traits and externalizing behavior, as bi-directional associations between moral disengagement and callous-unemotional traits have also been found to predict disruptive behavior in adolescents (Paciello, Ballarotto, Cerniglia, & Muratori, 2020).
Theoretical and practical implications
Most studies have focused on ER as a mediator between attachment and internalizing behavior, and not many have investigated the mediational association with externalizing behavior. This is surprising, given the theoretical support for this association, as well as the observed comorbidity between internalizing and externalizing behaviors. Furthermore, very few of the studies which have investigated this mediation, have focused on the adolescent age-range. Thus, our findings provide a much-needed contribution by supporting the mediational significance of ER for attachment and externalizing behavior in the phase between (early) childhood and adulthood, and indicating that previously established ER models of attachment should be expanded into adolescence.
Regarding familial factors, parenting practices have been implicated int the development of CU traits (Waller, Gardner, & Hyde, 2013). Although parenting and attachment are linked, they are separate constructs with the ability to distinctly contribute to the theoretical understanding of CU traits. Hence, we hope that the current study can draw focus to the predictive role of parental attachment in the development of CU traits. Previously identified constructs that are intertwined in the development of CU traits and externalizing behavior, such as temperamental traits and conscience development (Frick & White, 2008), should be further explored with an attachment-based perspective (see Kochanska & Thompson, 1997).
Frick and Viding (2009) proposed a developmental psychopathology framework for the etiology of antisocial behavior in childhood. The framework involves two pathways, marked by deviations in developmental mechanisms, which are responsible for the manifestation of antisocial behaviors; one involving deficits in emotional and behavioral regulation, and another involving deficits in conscience development, specifically marked by CU traits. The results of the current study demonstrate that attachment is a precursor to both pathways leading to externalizing behavior in adolescence. Hence, the current study merges previous frameworks which have looked at externalizing behavior from distinct attachment, emotion regulation, and psychopathic perspectives, thereby integrating fragmented associations into a more complex model that allows for a better (developmental) understanding of the involved constructs.
The influence of attachment has often gone unexplored in the treatment of externalizing behavior in adolescence (Brenning & Braet, 2013), although they have been proven effective. For instance, Moretti, Obsuth, Craig, and Bartolo (2015) report that their attachment-based intervention (targeted at parents and adolescents with serious behavior problems) resulted in a decrease in externalizing symptoms, as well as improving ER. Our findings demonstrate that similar effects could be found for CU traits, which although being relatively stable (Frick & White, 2008), can be modified by psychosocial interventions (Hawes & Dadds, 2007; McDonald, Dodson, Rosenfield, & Jouriles, 2011). Although most have focused on the buffering effects of parental warmth (Henry et al., 2018; Pasalich, Dadds, Hawes, & Brennan, 2012), the notion that attachment can be another vital intervention aspect has also been supported by researchers (e.g., Dadds et al., 2016). First steps could involve combining existing parenting- and attachment-based interventions for adolescents (see Gregory, Kannis-Dymand, & Sharman, 2020; Moretti & Obsuth, 2009), yet further research is needed to identify for which age range interventions would be most effective.
Limitations and future research
One methodological limitation is that we only relied on self-report measures. Furthermore, utilizing different theoretical conceptualizations of ER and attachment (e.g. avoidant versus ambivalent) in future studies, could expand our understanding of the interplay between these factors. Our study relied on a community sample, which was later reflected in the low to medium reports of CU traits, dysfunctional ER, and externalizing behavior (see Table 2). Thus, our results cannot be generalized to clinical samples without an appropriate replication. Although main effects exist, previous research indicates gender and age do not moderate the association between attachment and externalizing behaviors (Madigan et al., 2016), nor between attachment and CU traits (van der Zouwen et al., 2018). However, future developmental research with a bigger sample size should give more thought to these. Furthermore, the internal consistency of the ER scales is questionable, which corresponds with reports in other studies (e.g., Kullik & Petermann, 2013); future work should be dedicated to refining the REQ to increase reliability. The effect sizes call for us to be modest in the interpretation of the results and highlight the need for future studies to incorporate additional predictive factors.
Also, it is important to consider the cross-sectional design of the current study. Although we identified attachment as a predictor for CU traits (based on theory and past studies), a few authors have proposed that CU traits are predictors for attachment (Stylianou, Charalampous, & Stavrinides, 2019; Walsh, Roy, Lasslett, & Neumann, 2018). Considering that CU traits are in part biologically-driven and are already present in early childhood (Frick et al., 2014), it is plausible that CU traits could also disrupt attachment. A study by Dadds, Jambrak, Pasalich, Hawes, and Brennan (2011) found that boys with high CU traits showed consistently less eye contact with attachment figures that those with low CU traits. This in turn may interfere with parent-child bonding, ultimately reducing the influence of parenting and attachment processes (Hawes, Price, & Dadds, 2014). It is important for future research to examine causal bi-directional associations, as well as distinguishing between the biologically-heritable and the environmentally-driven variant of CU traits.
Lastly, we found that CU traits and dysfunctional ER significantly correlated in our study. Thus far, very few theoretical models have integrated emotion (dys)regulation and psychopathic traits, and empirical investigations have only recently begun. However, the hypothesized interplay between these constructs has varied greatly, with researchers noting that these run as parallel developmental processes (Frick & Viding, 2009), whilst others note that ER predicts psychopathic traits (Donahue, McClure, & Moon, 2014), or that psychopathic traits predict emotion dysregulation (Garofalo, Neumann, & Velotti, 2020). Further investigations into early familial/environmental factors, such as parental attachment, as well as the use of person-centered analysis will provide vital insights. Only through such investigations will we be able to fully understand the complex interplay of predictors in the developmental process leading to externalizing behavior.
Footnotes
Bio Sketches
Dr. Naska Goagoses is a postdoctoral researcher at the Carl von Ossietzky University of Oldenburg. Her research is embedded in the field of developmental and educational psychology. Her work focuses on the motivation and the social-emotional development of adolescents whilst considering individual, social, and contextual factors.
Dr. Ute Koglin is a professor at the Carl von Ossietzky University of Oldenburg in the Department of Special Needs Education and Rehabilitation. Her research is grounded in the field of developmental, clinical, and educational psychology. Among many others, her research interests include the social-emotional development of children and adolescents, as well as the prevention of behavioral problems/psychopathology.
