Attachment behavior and hostility as explanatory factors linking parent-adolescent conflict and adolescent adjustment

Abstract

This study examined whether adolescents’ behavior in a support-seeking context helped to explain associations between increases in mother-adolescent conflict during early adolescence and changes in adolescents’ internalizing and externalizing symptoms. A sample of 194 adolescents aged 12–14 (51% female) and their mothers were followed over one year. Adolescent-mother pairs participated in a Speech Task introducing an external social stressor into the parent-child relationship. Using a latent difference score model, adolescents’ observed attachment behavior and hostility were compared as potential explanatory processes. Analyses suggest specificity in the spillover process from conflict to adolescent behavior in a non-conflictual parent-child interaction context, with hostility uniquely linking increasing mother-adolescent conflict and externalizing problems, and disruptions in adolescent attachment behavior uniquely explaining the link with internalizing problems.

Parent-adolescent conflict is an ubiquitous feature of early adolescence that can play an important developmental role in renegotiating family relationships to support increasing autonomy needs (Branje, 2018). However, frequent and intense conflicts are consistently shown to predict poorer adolescent adjustment across both internalizing and externalizing dimensions (Huey, Hiatt, Laursen, Burk, & Rubin, 2017; Weymouth, Buehler, Zhou, & Henson, 2016). In order to better understand this multiplicity of outcomes, the current study compares two complementary explanations for why increases in parent-adolescent conflict in early adolescence may be associated with adolescent adjustment problems: through the spillover of adolescent hostility and by undermining adolescents’ secure attachment behavior.

The spillover of adolescent hostility

Multiple conceptual approaches propose the proliferation of adolescent hostility across relationship settings as an explanatory process linking parent-adolescent conflict and adolescent adjustment, particularly for externalizing behaviors. For example, social learning theory suggests that parent-child interactions serve as models for behavior in other social contexts (Bandura, 1977). Broader externalizing problems may develop as youth recapitulate hostile, aggressive, and defiant behavior learned in the context of family conflicts across settings (e.g., in non-conflictual situations) and relationships (e.g., with peers) (Patterson, 2002; van Doorn, Branje, VanderValk, De Goede, & Meeus, 2011). Accordingly, exposure to parental hostility and rejection has been linked to both adolescent aggression and rule-breaking behavior (Hoeve et al., 2009; Weymouth et al., 2016). Family systems models pose a similar hypothesis with the concept of “spillover,” in which emotions and behaviors produced during interactions in one context transfer to another (Erel & Burman, 1995). Spillover theory has been applied to understanding how parent-adolescent conflicts influence adolescents’ subsequent mood, interactions with peers, and problems at school (Timmons & Margolin, 2015; van Doorn et al., 2011).

Moving beyond traditional conceptualizations of spillover in explaining continuity of behavior across relationships, the current study examined whether adolescents’ experiences in conflicts with their mother would increase their likelihood of expressing hostility towards their mother in a non-conflictual setting. We hypothesized that increasingly intense mother-adolescent conflicts would “spillover” by heightening adolescents’ feelings of anger and frustration when attempting to engage mothers in coping with an interpersonal stressor outside of the parent-child relationship. The proliferation of hostility across mother-child interaction settings may help to explain associations between parent-adolescent conflicts and adolescents’ broader externalizing problems, as an early marker for the generalization of adolescents’ anger, frustration, and defiance across novel settings and relationship contexts.

Adolescents’ secure attachment behavior

Attachment theory offers a complementary hypothesis. Even as adolescents’ increasing desire for autonomy manifests as reluctance to turn to parents for support, confidence in the availability and responsiveness of parents when needed remains as a consistent predictor of adolescent well-being (Allen & Tan, 2016). Escalation of parent-adolescent conflict may undermine this trust, impeding adolescents’ ability to access parental support in coping with stressful challenges and undermining secure attachment. Attachment security is proposed to be evident in youth’s ability to effectively use attachment figures as both a safe haven, a source of support, comfort, and protection, and as a secure base from which to autonomously explore (Jones & Cassidy, 2014). Bowlby referred to each of these components as part of a larger “secure base” construct at the core of the attachment behavioral system (Bowlby, 1988). To avoid confusion in terms, we refer to adolescents’ observed ability to flexibly use their parent as both a safe haven and a secure base as “secure attachment behavior.”

Prior research suggests that a history of security in the parent-child attachment relationship predicts more salubrious, less hostile conflicts in early adolescence (Allen et al., 2003; Kobak, Cole, Ferez-Gillies, Fleming, & Gamble, 1993). What is less clear is whether changes in parent-child conflicts influence adolescent attachment. Increasing frequency and intensity of conflicts with parents may represent discontinuities with the potential to alter adolescent’s working models of attachment and, in turn, their secure attachment behavior. For example, although most teens move through adolescence maintaining relatively stable, secure attachment to parents, stability is reduced for those who experience more frequent parent-adolescent conflicts, parental pressure and criticism, and lack of parental support (Allen, McElhaney, Kupermind, & Jodl, 2004; Jones et al., 2017; Ruhl, Dolan, & Buhrmester, 2015).

Therefore, disruptions in adolescent secure attachment behavior may serve as a potential explanatory process linking mother-adolescent conflict and adolescent adjustment. Repeated exposure to maternal hostility is proposed to erode adolescents’ trust and confidence in their mother as an attachment figure, as evidenced by greater difficulty and reluctance to rely on their mother for secure base and safe haven support (Booth-LaForce et al., 2014). Moreover, both attachment theory and empirical evidence suggests that insecurity, as evidenced by adolescents’ difficulties accessing parental support, may mediate the relationship between parent-adolescent conflict and both internalizing and externalizing symptoms (Madigan, Brumariu, Willani, Atkinson, & Lyons-Ruth, 2016). On the one hand, insecure adolescents’ failure to effectively use attachment figures to regulate distress may contribute to heightened anxiety and emotion regulation difficulties that, over time, canalize into stable internalizing problems (e.g., Brumariu & Kerns, 2010). On the other hand, difficulty relying on parents for attachment support may lead to a lack of trust and dismissing views of interpersonal relationships shown to contribute to broader externalizing behavior (e.g., de Vries, Hoeve, Geert, & Asscher, 2016).

The current study

The current study seeks to compare these two complementary explanations for how or why mother-adolescent conflicts might contribute to adolescents’ psychological adjustment problems. Guided by both spillover and attachment theory, we compare adolescent hostility (i.e., expressions of anger and aggression towards their mother) and secure attachment behavior (i.e., use of mother as a safe haven and secure base) as potential explanatory processes. Both theories propose that frequent and intense parent-child conflicts influence adolescents’ broader behavioral problems through a process of generalization, whereby youth begin to apply their experiences within conflicts across contexts. Therefore, the current study expands beyond previous observational research (e.g., Jones & Cassidy, 2014) by explicitly testing whether mother-adolescent conflicts are associated with adolescents’ ability to effectively draw on parental support outside the context of parent-child conflict.

Conflicts represent a salient stressor in adolescence with relevance to the attachment relationship (Allen et al., 2003). However, conflicts with parents also represent a unique situation in which the parent is both the source of comfort, and the source of adolescent distress. Particularly if parents exhibit hostile and controlling behavior, adolescents’ primary motivation may be to reduce their exposure to the interpersonal threat cues rather than increasing proximity to and support from their parent (Davies & Martin, 2014). One reason for the limited findings regarding the influence of parent-adolescent conflict on attachment may be that the context in which we commonly assess adolescents’ secure attachment behavior can’t precisely distinguish between those for whom disruptions in secure attachment behavior is limited to conflicts and those for whom conflicts have begun to erode attachment behavior even in settings in which the likelihood of conflict is low and the need for assistance is high.

With this in mind, we investigate the following research questions. First, we seek to replicate previous research (e.g., Huey et al., 2017; Weymouth et al., 2016) by testing whether increases in parent-adolescent conflict with mothers in early adolescence is associated with increases in adolescents’ internalizing and externalizing symptoms. Second, we examine whether adolescent hostility and secure attachment behavior partially explain this link. By observing adolescent’s behavior during a support-seeking task, we test for potential specificity in the spillover of mother-adolescent conflict experiences to the way in which adolescents interact with their mother in coping with stress outside of parent-child conflict. Moreover, to help ensure that adolescents’ behavior in the support-seeking context is predicted by changes in mother-adolescent conflict and not due simply to preexisting individual differences in attachment security, we control for earlier parent reports of adolescent attachment behavior.

Methods

Participants

Participants included 194 families recruited from a city in the Northeastern U.S. as part of a broader study examining relationship functioning across multiple family subsystems. Interested families were included in the project if: (1) they had an adolescent between the ages of 12 and 14; (2) the target adolescent and two parental figures had been living together for at least the previous three years; (3) at least one parental figure was the biological parent of the target teen; (4) all participants were fluent in English; and (5) the target adolescent had no significant cognitive impairments. Families participated in two measurement occasions spaced one year apart. A majority of parents were married or engaged (88%), with another 11% reporting being in a committed, long-term relationship. Adolescents lived with their biological mother in a majority of cases (94%). Adolescents averaged 12.4 years of age at the first wave of data collection (51% female; n = 96). Median household income ranged from $55,000 to $74,999 and 14% of families reported a household income under $23,000. Median parental education was an Associate’s degree and most (85%) attended at least some college. A smaller subset (14%) earned a high school diploma or GED as their highest degree. The sample largely identified as White (74%), followed by Black (13.5%) and mixed race (10%), and 12% identified as Hispanic or Latino ethnic identity. The retention rate from Wave 1 to 2 was 91% (177 families).

Procedures

At each wave of data collection, two parental figures and their adolescent visited the laboratory for a single, three-hour visit. The lab included a room designed to resemble a living room and equipped with audiovisual equipment to record family interactions, as well as other comfortable rooms for participants to complete interviews and surveys. The study was approved by the Institutional Review Board of the University of Rochester (Protocol#: RSRB00030791; Title: Family Relationships in Early Adolescence) and parents provided consent. Families received monetary payment for their participation.

At each measurement occasion, mothers, fathers, and adolescents completed questionnaires. In addition, mothers and adolescents completed the Speech Task (Martin, Sturge-Apple, Davies, Romero, & Buckholz, 2017), a structured activity designed to elicit adolescent attachment behavior. A similar procedure has been used to evaluate adolescents’ stress in the context of the mother-adolescent attachment relationship (i.e., Spangler & Zimmermann, 2014; Zimmermann, Mohr & Spangler, 2009). The task took place in a room set up with a podium in front of a videocamera prominently displayed in order to highlight the evaluative nature of the task. Capitalizing on social evaluation as a developmentally salient stressor, we asked adolescents to give a speech describing their personal strengths and weaknesses. The task was introduced as an attempt to understand how adolescents communicate about themselves, much as they would for a job or college interview. Adolescents were told that their videotaped speech would be evaluated by a “communications expert” to help us understand how teens can improve their oral communication skills. They were given five minutes to prepare and two minutes to complete the speech. Mothers were present in the room and given the single task of pushing the “on/off” button for the camera at the start and end of the speech. A total of eight minutes was recorded. This included three segments: (1) the five-minute preparation period, (2) the two minute speech, and (3) a one-minute post relaxation period. Video records were later used to assess adolescents’ behavior.

Similar to the Strange Situation paradigm for young children, our goal was to introduce an interpersonal stressor with the goal of maximizing the saliency of the attachment system. To complete the task effectively, adolescents had to engage in exploratory behavior (e.g., thinking about their strengths and weaknesses, preparing an engaging speech), while managing distress and maintaining a balance of autonomy and relatedness with their mother. Therefore, the task provided an opportunity to observe how adolescents interact with their mother in a stressful context in which the source of distress comes from outside of the parent-adolescent relationship.

Measures

Mother-adolescent conflict (Wave 1 and 2).

At each measurement occasion, mothers and adolescents described the frequency and intensity of conflict in the mother-child relationship using parent- and adolescent-report versions of The Conflict Behavior Questionnaire – 20 (CBQ-20; Robin & Foster, 1989). Each includes 20 true/false items reflecting different aspect of conflict in the parent-child relationship from the adolescent’s (e.g., “My mother and I sometimes end our arguments calmly (reverse-scored),” “At least once a day my mother and I get angry at each other.”) or the mother’s (e.g., “In general, I don’t think we get along very well,” “My child and I have big arguments about little things”) points of view. Previous research supports the validity and reliability of the CBQ-20 (Prinz, Foster, Kent, & O’Leary, 1979; Robin & Foster, 1989). Scores were summed, with positive items reverse-scored, to yield a single score for each reporter. Higher values reflect more frequent and intense conflict. Internal consistencies were α = .85 and .88 in Wave 1 and α = .87 and .86 in Wave 2 for mothers and adolescents, respectively. To take advantage of multiple informants and supported by their moderate correlation (r = .31, p < .001 at Wave 1 and r = .36, p < .001 at Wave 2), mothers’ and adolescents’ scores were averaged to yield a single score for mother-adolescent conflict at each wave.

Adolescent behavior during The Speech Task (Wave 2).

Adolescents’ behavior was assessed using the Teen Attachment Behavior System (TABS; Martin, Sturge-Apple, & Davies, 2015), an observational coding system designed for this study. The TABS was adapted from the Attachment Behavior Classification Procedure (ABCP; Hilburn-Cobb, 1998; 2004) with the goal of assessing adolescents’ attachment-relevant behaviors in the context of the Speech Task. Two independent coders were initially trained on the TABS to reliability, after which a single primary coder completed the TABS for every mother-adolescent Speech Task video. The second, reliability coder then independently coded 19% of the videos.

Secure attachment behavior.

Adolescents’ secure attachment behavior reflected confidence in their mother’s comfort and protection (e.g., help, unconditional support) in the face of the stressful task. Secure behaviors were reflected in the effective use of their mother to both regulate negative affect and as a secure base from which to pursue task completion. An adolescent scoring high on security evidenced flexibility in balancing proximity-seeking and exploration based on a realistic appraisal of the situation and their abilities. In this context, exploration involved autonomously preparing for the speech (e.g., seeking maternal input only after attempting the task themselves), while maintaining an open, connected, and non-defensive orientation to their parent. The emotions of secure adolescents appeared to be easily expressed and easily regulated within the relationship. Thus, although negative emotions could be evident (e.g., modest distress), secure adolescents were able to draw satisfactory support and comfort from interaction with their mother. Secure adolescents appeared to genuinely hear and take in their parent’s reassurances or gave evidence of relief without embarrassment, shame, or overly effusive appreciation. Overall, interactions with their mother were relaxed (i.e., relaxed body posture, confident tone, no contorted postures or excessive hiding/fidgeting) and avoided dismissing or rejecting either their own distress or their mother’s attempts at support. Given this prototypic profile of secure attachment behavior, coders provided a single, molar rating of the degree to which adolescents’ overall pattern of behavior across the eight-minute task evidenced security from 1 (no evidence of a secure pattern) to 9 (a whole lot of evidence of a secure pattern). A “9” was assigned only when the adolescent consistently exhibited telltale signs of secure attachment behavior across the interaction. The final reliability was excellent, ICC = .92.

Hostile behavior.

Hostility, as assessed in the TABS, included affective, verbal, and physical expressions of anger, aggression, and active rejection of maternal attempts to engage. Affective expressions of hostility included anger, frustration, or irritation towards the parent. The degree of hostile behavior was determined based on its potential to disrupt the quality of the parent-child relationship. Thus, high scores on hostility included expressions of dysregulated anger towards their mother combined with attempts to hurt, demean, or attack her either verbally (e.g., “You’re so stupid,” “God, just shut up about it!”) or physically (e.g., slapping mother’s hand away, grabbing the stopwatch from her). As with secure behavior, coders observed the entire 8-minute video and provided a single, continuous score reflecting the degree to which the adolescents exhibited hostility from 1 (very little or no hostility) to 9 (a whole lot of hostility). To score a “9,” adolescents had to exhibit marked signs of hostility for the majority of the task. Final interrater reliability was excellent: ICC = .83.

Adolescent psychological adjustment (Waves 1 and 2).

Adolescents reported on their psychological adjustment using the Youth Self-Report (YSR11–18: Achenbach & Rescorla, 2001) at each wave. Adolescents are considered to be ideal reporters of their own psychological adjustment, particularly for internalizing symptoms (Smith, 2007). Consistent with recommended procedure (Achenbach & Rescorla, 2001), internalizing symptoms were assessed by calculating the sum of the anxious/depressed (13 items; e.g., “I am unhappy, sad, or depressed”) and withdrawn (8 items; e.g., “I keep from getting involved with others”) subscales. Items from the aggression (16 items; e.g., “I get in many fights”) and delinquency (13 items; e.g., “I break rules at home, school, or elsewhere”) subscales were similarly summed to yield a single, continuous measure of externalizing problems. Response options ranged from 0 (not true) to 2 (very or often true). Internal consistencies were α = .80 and α = .88 for the internalizing subscale and α = .86 and .90 for the externalizing subscale at Waves 1 and 2, respectively.

Covariate: Parent-reported attachment to mother (Wave 1).

Mothers and fathers completed the parent-report version of the Parental Attachment Scale (PAS-P; Davies et al., 2002). The PAS-P includes nine items reflecting how the adolescent relates to the parent as an attachment figure (e.g., “When my child is upset, s/he comes to me/my partner,” “My child appears to like being around me/my partner”). Each parent reported on the adolescents’ relationship with their mother. Response options range from 1 (Not at all like my child) to 5 (A whole lot like my child). The internal consistency was .89 for maternal self-report and .86 for father reports. Mother and father reports of adolescent attachment to mother were correlated (r = .43, p < .001) and averaged to yield a single score for adolescent attachment to mom at Wave 1.

Covariate: Socioeconomic status (Wave 1).

Mothers and fathers completed a short demographic survey in Wave 1. Both reported on their family’s average yearly income. Their reports were averaged and ranged from less than $6000 to over $125,000 (Median = $55,000 - $74,999). Each parent also reported their highest degree of education. Mothers ranged from 10^th or 11^th grade to doctoral degree (Median = Bachelor’s degree) and fathers from 8^th or 9^th grade to doctoral degree (Median = Associate’s degree). Each parents’ education and the average family income were standardized and averaged to yield a single score, with higher values reflecting higher socioeconomic status. Internal consistency for the aggregated variable was α = .81.

Results

Families who dropped out of the study and those who returned for the second wave evidenced no differences on any of the demographic characteristics or variables collected at Wave 1. Little’s MCAR test suggested that missing data (9.38% of all values) were missing completely at random: χ² (50) = 53.23, p = .35. To maximize our sample size, missing data were estimated using full information maximum likelihood (FIML; Enders, 2001) estimation in Amos 25.0 (Arbuckle, 2017). Table 1 includes descriptive statistics and correlations.

Table 1.

Means, standard deviations, and bivariate correlations among the primary variables in this study

	Reporter	Mean	SD	1	2	3	4	5	6	7	8	9	10
Wave 1
1. Adolescent Gender	A	---	---	---
2. Socioeconomic Status	M, D	0.00	0.87	−.11	---
3. Parent-child Attachment	M, D	4.09	0.55	−.06	−.06	---
4. Parent-adolescent Conflict	M, A	3.15	2.90	.12	−.05	−.42*	---
5. Internalizing Symptoms	A	6.57	4.73	.11	.02	−.11	.24*	---
6. Externalizing Problems	A	7.11	5.92	.12	−.16*	−.13	.34*	.62*	---
Wave 2
7. Secure Attachment Behavior	O	4.49	2.50	.06	.30*	.21*	−.31*	.04	−.02	---
8. Adolescent Hostility	O	4.26	1.74	−.08	−.20*	−.14	.12	−.09	.10	−.42*	---
9. Parent-adolescent Conflict	M, A	3.14	3.20	.06	−.18*	−.23*	.58*	.18*	.34*	−.35*	.33*	---
10. Internalizing Symptoms	A	8.40	6.62	.12	−.09	−.09	.04	.45*	.31*	−.18*	.08	.30*	---
11. Externalizing Problems	A	8.40	7.05	.01	−.22*	−.13	.23*	.35*	.63*	−.15†	.31*	.40*	.51*

^*Note. p < .05,

^†p < .07;

Reporters: M = Mom, D = Dad, A = Adolescent, O = Observer

Parent-adolescent conflict and adolescent psychological adjustment

The first step was to test whether increases in mother-adolescent conflict from Wave 1 to 2 were associated with increases in adolescent psychological adjustment over the same period. Three latent difference scores (LDS) were specified, one assessing change in mother-adolescent conflict across the two annual measurement occasions and two assessing change in adolescents’ psychological adjustment (i.e., internalizing, externalizing). LDS models include a latent intercept and a latent change score, modeled as a function of both the linear slope (i.e., constant change) and the previous score (i.e., proportional change). This provides an assessment of interindividual differences in intraindividual change (McArdle, 2009).

Following conventional procedures (McArdle, 2009), we regressed the Wave 2 assessment onto both the Wave 1 assessment and the latent difference score, constraining each pathway to 1, for each of the three LDS models. We then examined the unconditional LDS by specifying a correlational pathway between each initial value and the latent change score. This allows for the examination of mean change. Each initial value (i.e., conflict, internalizing, externalizing) was also allowed to covary with the other two. Three covariates (i.e., adolescent gender, SES, and parent-reported attachment to mother) were included as predictors of each LDS. This resulted in a perfectly identified model. Results indicated that the frequency and intensity of mother-adolescent conflicts, on average, did not increase across the one-year period (−.04; z = −.17, p = .86). By contrast, both internalizing (1.88; z = 4.06, p < .001) and externalizing (1.25; z = 2.90, p = .004) symptoms demonstrated an average increase over the same period. All three constructs evidenced significant interindividual variability in intraindividual change: conflict (var = 7.81; z = 9.45, p < .001), internalizing (var = 37.99; z = 9.73, p < .001), and externalizing (var = 32.34; z = 9.26, p < .001).

Therefore, we proceeded to examine our first hypothesis, that increases in mother-adolescent conflict over the one year period would be associated with increases in psychological adjustment problems over the same period. A conditional LDS was specified for mother-adolescent conflict and the respective adjustment variables by replacing the correlation between the initial score at Wave 1 and the latent change score with a regression path. The three study covariates were specified as predictors of the latent change in mother-adolescent conflict and each form of psychological adjustment. A pathway was also specified between initial levels of conflict and adjustment and the others’ latent change score. This provided a test of whether initial values of each variable impacted change in the others. Lastly, a direct path was specified from the latent change score for mother-adolescent conflict and for each adjustment variable.

The results are illustrated in Figure 1. The model was perfectly identified. Paths from the latent change score for conflict to latent change in internalizing and externalizing symptoms were significant at p < .001; β = .34 and β = .24, respectively, suggesting that increases in conflict across the one-year period were uniquely associated with increases in adolescent internalizing and externalizing symptoms over the same period. In addition, adolescents’ initial reports of externalizing problems at Wave 1 uniquely predicted increases in conflict; β = .23, p = .02. The same was not true for initial reports of internalizing; β = −.06, p = .51. None of the three covariates evidenced a unique relationship to any of the latent change scores.

Figure 1.

LDS path model testing latent change in parent-adolescent conflict from Wave 1 to 2 as a predictor of latent change in adolescent-reported internalizing and externalizing symptoms over the same period, controlling for covariates. Covariates are not included in the model for clarity. All other specified parameters are illustrated in the figure. Light grey lines represent paths that were not significantly different from zero at p < .05. * p < .05, ** p < .01.

Indirect effect analyses

We next proceeded to test whether adolescents’ behavior (i.e. hostility, secure attachment behavior) during the Speech Task could partially account for the association between increases in mother-adolescent conflict and increases in adolescent adjustment problems. Building on the previous models, adolescent hostility and secure attachment behavior, assessed at Wave 2, were included as intervening variables. Pathways were specified from the latent change score for conflict to hostility and secure attachment behavior. Additional pathways were specified from each form of adolescent behavior to the latent change score for adolescent internalizing and externalizing symptoms. Paths were also specified from each Wave 1 variable, including the initial scores for mother-adolescent conflict and adolescent adjustment, to hostility and secure attachment behavior. Given their moderate correlation (r = −.41, p < .001), the error terms for adolescent secure base behavior and hostility at Wave 2 were allowed to covary. This model was perfectly identified.

The primary results of interest for this model are illustrated in Figure 2. With the proposed explanatory factors included in the model, the latent change score for mother-adolescent conflict continued to evidence a unique association with concurrent change in adolescent internalizing, β = .31, p < .001, and externalizing symptoms, β = .17, p = .03. Initial Wave 1 levels of externalizing problems continued to predict increases in mother-adolescent conflict, β = .23, p = .02. The same was not true for initial internalizing symptoms, β = −.06, p = .50. As with the previous model, none of the covariates evidenced a relationship with any of the three latent change scores. However, in this model, the latent change score for mother-adolescent conflict was uniquely associated with both adolescent secure attachment behavior, β = −.22, p < .001, and adolescent hostility, β = .30, p < .001 at Wave 2. Neither initial internalizing nor externalizing symptoms uniquely predicted adolescent behavior during the Speech Task.

Figure 2.

An LDS model testing the indirect effect of adolescent hostility and secure attachment behavior (SAB) during the Speech Task in linking parent-adolescent conflict and latent change in adolescent internalizing and externalizing symptoms over one year. Covariates were included in the model but are not illustrated in the figure. Only pathways significant at p < .05 are included in the figure. Bolded pathways represent the paths of primary interest. * p < .05, ** p < .01

In turn, adolescent secure attachment behavior at Wave 2 was associated with latent change in adolescent internalizing symptoms; β = −.22, p = .01. By contrast, the relationship between adolescent hostility and change in adolescent internalizing symptoms was negligible, β = −.05, p = .50. We examined the indirect effect from change in conflict to change in adolescent internalizing symptoms through secure attachment behavior using Tofighi and MacKinnon’s (2011) RMediation program. Results indicated that the indirect effect estimate (μ = 0.10) was significantly different from zero, 95% CI [.02, .23]. This indirect effect held over and above adolescent hostility and each of the Wave 1 predictors, including initial mother-adolescent conflict and parent-reports of adolescent attachment to mom. In addition, adolescent hostility evidenced a unique association with increasing adolescent externalizing symptoms from Wave 1 to 2, β = .22, p = .01. Adolescent secure attachment behavior was negligibly associated with changes in externalizing symptoms, β = −.02, p = .78. We again tested the indirect effect from change in conflict to change in externalizing problems, this time through adolescent hostility. Results supported the indirect effect, μ = 0.13, 95% CI [.03, .27].

A number of additional pathways were estimated to be different from zero at p < .05. First, the initial level of mother-adolescent conflict evidenced a unique association with Wave 2 secure attachment behavior, β = −.37, p < .001. By contrast, the pathway between Wave 1 conflict and Wave 2 adolescent hostility was β = .17, p = .07. This raised the question of whether there was a difference in the degree to which secure attachment behavior and hostility were influenced by initial conflict versus increases in the frequency and intensity of conflict over time. To examine this further, a set of exploratory pairwise parameter comparisons were estimated comparing the magnitude of the association between the pathways to secure attachment behavior and hostility from initial parent-adolescent conflict and change in conflict and secure attachment behavior and hostility. Results of this analysis suggest that initial mother-adolescent conflict has a stronger effect on secure attachment behavior compared to hostility, z = −2.74, p = .01, but that change in mother-adolescent conflict was associated with each form of behavior equally, z = −.01, p = .99. We also examined whether initial level of mother-adolescent conflict was a stronger predictor of secure attachment behavior than changes in conflict. This exploratory pairwise comparison suggests that both were equally associated with adolescent attachment behavior: z = 1.55, p = .12. Second, parent reports of SES evidenced a unique predictive relationship with both adolescent secure attachment behavior, β = −.30, p < .001, and hostility, β = −.17, p = .02, over and above the influence of mother-adolescent conflict. Third, parent reports of adolescents’ attachment relationship with their mother evidenced some unique associations with both secure attachment behavior and hostility in the expected direction, although each only approached significance: β = .14 and β = −.14, p = .07, respectively.

Discussion

This study examined whether adolescents’ ability to effectively draw on maternal support in managing an interpersonal stressor would help to explain associations between mother-adolescent conflict and adolescent adjustment. Drawing on a novel observational assessment, the Speech Task, two complementary explanations were explored. From a spillover perspective, we proposed that increases in the frequency and intensity of mother-adolescent conflicts would be associated with adolescents’ anger and hostility towards their mothers in non-conflictual situations. Adolescent hostility, in turn, was predicted to be associated specifically with increases in externalizing problems. Drawing on attachment theory, we posited that increasingly frequent and intense mother-adolescent conflicts would undermine adolescents’ confidence in their parent as a source of safety and security, as evidenced by reductions in secure attachment behavior. These difficulties, in turn, were predicted to uniquely explain the association between mother-adolescent conflict and both internalizing and externalizing problems. Together, the results provide some support for each framework, albeit through distinct pathways.

First, replicating previous work (e.g., Hofer et al., 2013), increases in mother-adolescent conflict over the one-year period were associated with increases in both adolescent internalizing and externalizing problems over the same period. Notably, at the group level, there was no mean change in mother-adolescent conflict. Rather, significant variability was evident. Consistent with recent research (e.g., Huey et al., 2017), this points to important individual differences in the stability of parent-adolescent conflict in early adolescence. In addition, the use of latent difference scores revealed unique patterns in the relationship between initial values and change for each form of adjustment. Only initial levels of externalizing symptoms predicted increases in mother-adolescent conflict. By contrast, initial mother-adolescent conflict was not predictive of changes in externalizing problems. This pattern suggests an evocative effect whereby adolescents’ aggressive and rule-breaking behavior elicits increasingly frequent and intense parent-adolescent disagreements.

We next sought to compare adolescent hostility and secure attachment behavior during the Speech Task as alternative explanatory processes linking increases in mother-adolescent conflict and increases in adolescent psychological adjustment. Researchers note how difficult it has been to find a developmentally appropriate, valid, and ethically viable method of eliciting distress in adolescence (McElhaney et al., 2009; Warmuth & Cummings, 2015). Drawing on the success of similar procedures (i.e., Zimmerman et al., 2009), our goal was to observe adolescent behavior in a context that would mirror the central components of the Strange Situation. Accordingly, the Speech Task provided a situation that: (a) was high in experimental realism, (b) was distressing, but in which the mother was not inherently the source of the distress, (c) required the adolescent to balance coping with autonomous exploration, and (d) allowed the mother to be present (and thus utilized as a secure base/safe haven) without explicitly instructing her to provide support. Importantly, the Speech Task provided an opportunity to observe whether adolescents generalized hostile and insecure attachment behaviors in a setting in which the need for caregiving support is likely to be the prevailing motivation (Davies & Martin, 2014).

Evidence for spillover

Increases in mother-adolescent conflict over the one-year period was uniquely associated with adolescent hostility in the Speech Task. Adolescent hostility, in turn, was solely and uniquely associated with increases in externalizing problems from Wave 1 to 2, over and above secure attachment behavior. Moreover, the indirect pathway through adolescent hostility in the Speech Task was specific to externalizing symptoms. These findings are consistent with spillover theory in suggesting that anger and hostility honed in the context of parent-adolescent conflicts may spillover into non-conflictual interactions with parents. The spread of hostile feelings across parent-adolescent interaction settings may set the stage for recapitulating aggressive and defiant behaviors in extrafamilial settings that, ultimately, become instantiated in persistent externalizing symptomology. At a practical level, adolescents are regularly faced with interpersonal stressors. The added challenge of having to regulate pervasive feelings of anger and frustration towards parents may be especially likely to undermine their ability to manage social situations, such as conflicts with peers, in adaptive ways (Ehrlich, Dykas, & Cassidy, 2012). Additionally, although the current study design limited our ability to consider bidirectional links, these are highly likely. Adolescents’ expressions of hostility when faced with stressors from outside the parent-adolescent relationship may represent another way in which adolescents evoke future conflicts with parents.

It is also highly likely that adolescents’ previous levels of anger, aggression, and rule-breaking behavior play a substantial role in these family processes. However, a broader examination of the pathways through hostility show that, although adolescents’ initial levels of externalizing symptoms predicted increases in mother-adolescent conflict, neither initial conflict nor externalizing problems predicted hostility at Wave 2. These findings argue against the possibility that our assessment of hostility in the Speech Task was merely capturing the expression of adolescents’ persistent, stable externalizing problems across contexts. Moreover, the specificity of the link from increases in mother-adolescent conflict to adolescents’ hostility in the Speech Task lends further support to both the social learning and spillover hypotheses by suggesting that increases in mother-adolescent conflicts contribute to adolescents’ hostility in non-conflictual situations. Although clearly a first step, these findings suggest a possibility worthy of further study: that the bleeding of adolescent hostility from within parent-child conflicts across interaction settings, and particularly into settings with relatively little cause for disagreement, may be an especially strong marker of risk for increasing externalizing symptoms. An important next step will be to examine whether these processes set the stage for the development of externalizing psychopathology.

Evidence for attachment

We further sought to test an attachment theory explanation by examining adolescents’ secure behavior as a second potential explanatory process linking mother-adolescent conflict and adjustment. In support of this hypothesis, increases in mother-adolescent conflict over the one-year period were associated with adolescent secure attachment behavior. Individual differences in adolescents’ secure attachment behavior, in turn, were negatively associated with increases in adolescents’ internalizing symptoms, but not externalizing symptoms. The findings support the notion that increases in parent-adolescent conflict may undermine adolescents’ ability to use parents as a safe haven to regulate distress and a secure base from which to explore, contributing to a small but growing literature suggesting that adolescent attachment security is susceptible to change as a function of changes in the parent-adolescent relationship (e,g., Pinquart, Feubner, & Ahnert, 2013). This was also, to our knowledge, the first attempt to consider associations between mother-adolescent conflict and adolescents’ observed attachment behavior in a non-conflict interaction. Our findings extend those of Jones and Cassidy (2014), who found that parental hostility undermined adolescents’ attachment behavior in the context of conflict, by suggesting that hostile parent-adolescent conflict may influence adolescents’ attachment behavior even in settings in which attachment needs should be salient. Moreover, these findings are consistent with other empirical work suggesting that meaningful individual differences in attachment may be observable in adolescents’ interactions with parents (e.g., Hershenberg et al., 2011; Hilburn-Cobb, 2004).

The hypothesis that adolescent secure attachment behavior would help to explain the association between mother-adolescent conflict and both forms of adjustment was not supported. Compared with adolescent hostility, an indirect effect of secure attachment behavior was evident only in the association between mother-adolescent conflicts and adolescent internalizing problems. This contributes to a broader literature demonstrating that secure attachment allows children and adolescents to better regulate negative emotions (e.g., Brumariu, Kerns, & Seibert, 2012; Zimmer-Gembeck et al., 2017) by demonstrating how increases in the frequency and intensity of conflicts with parents may disrupt adolescents’ ability to draw on parental support in coping with interpersonal stress. Difficulty relying on parents for support and encouragement may deprive adolescents of an important resource at a time when they are at increased risk for internalizing problems (Kerns & Brumariu, 2014).

Moreover, even the specificity of hostility in predicting externalizing symptoms may not fully rule out an attachment explanation. We chose, based on the centrality of the secure base construct as a behavioral expression of security, to limit our assessment of attachment to a continuous measure of attachment behavior. However, attachment theory also argues for the existence of distinct constellations or profiles of attachment insecurity (Fearon et al., 2014). Since the Speech Task was designed to elicit attachment behavior, it may be that our observational assessment of adolescent hostility was confounded with a subtype of disorganized insecure attachment, namely a “controlling” pattern. A controlling pattern of disorganized attachment is characterized by the subjugation of attachment goals in favor of dominant strategies of aggression, defiance, and minimization of the attachment figures’ role in one’s security (Hilburn-Cobb, 2004; Lecompte & Moss, 2014). Although research on this pattern in adolescence is relatively sparse, previous findings do suggest that disorganized attachment is linked primarily to externalizing problems (Lecompte & Moss, 2014). Future research on adolescent behavior during the Speech Task would benefit from a closer examination of distinct patterns of attachment and attempts to tease apart these possible patterns.

Examination of the associations between changes in mother-adolescent conflict and adolescents’ attachment behavior also revealed specificity. Initial levels of conflict evidenced a unique predictive effect over and above changes in conflict for attachment behavior, but not hostility. This finding is consistent with attachment theory’s view that security will both evidence moderate stability, due in part to consistency in the caregiving environment, and be open to change following changes in caregiving (Waters, Hamilton, & Weinfeld, 2003). Thus, although we were particularly interested in whether changes in parent-adolescent conflict would serve as a source of discontinuity influencing adolescent attachment behavior, as early adolescents begin to increasingly self-regulate distress (e.g., Allen & Tan, 2016), youths’ history of conflict within the parent-child relationship is likely to continue to play an important role.

Socioeconomic status

A somewhat surprising finding was the unique association between family socioeconomic status and adolescents’ behavior during the Speech Task, over and above changes in parent-adolescent conflict. Several explanations might account for this effect. For one, the Speech Task may have been experienced as more stressful for adolescents from lower SES families. Prior research with adolescents has demonstrated a correlation between low SES status and greater perceived stress (Goodman et al., 2005). Adolescents from low SES families are also less likely to cope by engaging interpersonal support (Finkelstein, Kubzansky, Capitman, & Goodman, 2007). A related possibility is that family SES plays a unique role in shaping adolescent attachment security. Particularly as youth age and become more aware of the broader socioeconomic context of the family, factors associated with low SES (e.g., financial hardship, job loss) may both increase adolescents’ perceived stress and tax parental resources, limiting their ability to provide support (Allen et al., 2004). This also raises the possibility that the influence of SES on adolescent behavior during the Speech Task is mediated through other family processes, either within the parent-adolescent relationship (e.g., parent withdrawal, parent stress) or in the broader family unit (e.g., interparental conflict; father-adolescent conflict), all of which interact in predicting adolescent adjustment (Conger, Conger, & Martin, 2010).

Limitations and future directions

Any conclusions must be considered in light of several study limitations. First and foremost, data for this study were collected across only two time points. This prevents the full disaggregation of predictor, mediator, and outcome that allows for testing a putative mediational change model (e.g., Preacher, 2015). Although the use of latent difference scores provided some interesting tests of the relative influence of initial levels versus change in mother-adolescent conflict on adolescent behavior and psychological adjustment, caution is warranted in drawing strong conclusions about the direction of effects. Moreover, although findings were consistent with the hypothesized spillover process, conclusions are limited by the inclusion of the observational assessment of adolescent behavior at only a single time point. Demonstrating that changes in parent-adolescent conflict lead to subsequent changes in adolescents’ behavior outside of conflict would provide a more thorough test of this model.

Adolescent adjustment was assessed using a single measure (the YSR) and single reporter. Adolescents are considered accurate reporters of internalizing and externalizing symptoms (Smith, 2007), but more comprehensive, multi-informant assessments of adolescent psychological adjustment would strengthen future studies. Similarly, although both mothers and adolescents reported on conflict, more in-depth approaches that account for multiple aspects of conflict may be warranted. For example, the resolution of conflicts likely play a unique role in shaping adolescents’ confidence in parental attachment support (Branje, van Doorn, van der Valk, & Meeus, 2009).

Although the focus on behavioral observations of attachment in adolescence is a strength of this study, adolescents’ internal representations play an important role in attachment theory’s explanation for how and why changes to the parent-adolescent relationship lead to changes in attachment security (Jones et al., 2017; Madigan et al., 2016). Understanding the role of internal representations in the proposed process model will be an important next step. In addition, our study focused on the maternal relationship to the exclusion of fathers. Although, as a first step, this places our exploratory study within the broader literature on adolescent attachment, we do not intend to suggest that the father-adolescent relationship is irrelevant. Future work should absolutely broaden the scope of research to examine similar questions with fathers. For example, conceptual and empirical work suggests that fathers may play a unique role as secure-base providers compared to mothers (Bretherton, 2010).

Lastly, consistent with previous approaches (e.g., Jones & Cassidy, 2014), our observational measure of adolescent attachment behavior integrated adolescent’s use of mothers as both a safe haven (i.e., to reduce distress) and a secure base (i.e., to increase confidence and autonomous exploration). Although this was an important first step, future research may benefit from a more fine-tuned analysis that disaggregates these two aspects of attachment behavior (e.g., Kerns, Mathew, Koehn, Williams, & Siener-Ciesla, 2015).

Despite these limitations, this study provides an important, incremental step towards better understanding how parent-adolescent conflicts may contribute to adjustment problems in early adolescence. It explicitly examined whether and how changes in mother-adolescent conflicts are associated with adolescents’ behavior in the context of an interpersonal stressor outside of the parent-child relationship and employed a novel observational approach for assessing adolescent attachment behavior. Although strong conclusions certainly require replication and further, longitudinal analysis, these findings do provide initial evidence for the value of examining the processes by which disruptions in the context of parent-adolescent conflicts unfold across situations.