Interparental Conflict and Adolescent Emotional Security Across Family Structures

Abstract

This study investigated whether interparental conflict was differentially related to forms of emotional security (i.e., family, interparental, parent-child) and whether forms of emotional security were differentially associated with mental health problems for adolescents in married vs. divorced/separated families. Participants were 1,032 adolescents (ages 10–15; 51% male, 49% female; 82% non-Hispanic White, 9% Black/African American, 5% Hispanic, 2% Asian or Pacific Islander, 2% Native American) recruited from a public school in a middle-class suburb of a United States metropolitan area. We used multiple group multivariate path analysis to assess (1) associations between interparental conflict and multiple measures of emotional insecurity (i.e., family, interparental, and parent-child), (2) associations between measures of emotional insecurity and internalizing and externalizing problems, and (3) moderation effects of parent-child relationships. The patterns of association were similar across family structures. A high-quality parent-child relationship did not mitigate the harmful effects of interparental conflict on emotional insecurity or mental health problems. Findings suggest that regardless of family structure, emotional security across multiple family systems may be a critical target for intervention to prevent mental health problems, in addition to interventions that reduce conflict and improve parent-child relationships.

Exposure to interparental conflict (IPC) is a significant stressor across all family structures (Cummings & Davies, 2010; Harold & Sellers, 2018). IPC is a broad concept that ranges from minor disagreements to physical violence. In the current study, the operational definition of IPC includes the frequency, intensity, content, and resolution of disagreements between parents. Although well-established as a robust risk factor for maladjustment across domains of adolescent health and well-being (Harold & Sellers, 2018; van Eldik et al., 2020), little is known about whether processes underlying the link between IPC and adolescent adjustment vary by family structure. This is a critical gap in the literature because process-oriented research provides the foundation for designing theory-based, effective interventions for youth exposed to high levels of IPC. If, for example, processes differ as a function of family structure, it would indicate the need for targeted intervention strategies depending on whether IPC occurs between married or divorced parents. We used emotional security theory (EST; Cummings & Davies, 2010) to test relations among interparental conflict, adolescent emotional insecurity, and psychological adjustment across different family structures. We specifically examined whether pathways among IPC, various forms of emotional insecurity in family relationships, and internalizing and externalizing symptoms were comparable across married/cohabitating and divorced/separated families.¹ We also investigated whether the interactive effects of parent-child relationship processes varied across family structures.

Adolescent Emotional Insecurity in the Family

EST posits that IPC increases risk for mental health problems by threatening the psychological need to maintain a sense of security and safety in family relationships, as evident in emotional reactivity, regulation behaviors, and internal representations related to IPC (Davies, Harold, et al., 2002). Although youth’s responses to IPC may have short-term adaptive value, the substantial expenditure of emotional resources undermines the achievement of developmental milestones and can become generalized to create rigid, pathogenic patterns of functioning that are stable across time and situations (Davies, Harold, et al., 2002). EST identifies three primary forms of emotional insecurity – parent-child, interparental, and family – as critical explanatory mechanisms in risk models of IPC. A fundamental assumption of EST is that sources of emotional security (i.e., parent-child, interparental, family) are intercorrelated yet make unique contributions to children’s adjustment (Davies, Harold, et al., 2002). For example, how children perceive their family functioning as a whole predicts adjustment above and beyond the effects of individual dyadic subsystems (see Forman & Davies, 2005, for an extensive literature review and discussion). Thus, evaluating the unique and useful information gained from assessing family processes through both subsystem and whole system perspectives is critical.

IPC is proposed to undermine insecurity in multiple family contexts, including the parent-child relationship, interparental subsystem, and overall family unit. First, consistent with attachment theory, youth derive a general sense of security, safety, and protection from their dyadic relationships with parents or caregivers (Ainsworth et al., 1978; Davies, Harold, et al., 2002). Witnessing frightening or frightened parental behaviors during IPC events may undermine attachment security, as reflected in diminished abilities to utilize parents as sources of support and protection. Second, youth are also influenced by their sense of security in the interparental relationship (Davies, Forman, et al., 2002). When the interparental relationship is turbulent, a sense of security is threatened because exposure to IPC raises fears about potential family separation, whether that be the dissolution of the interparental relationship or losing contact with one or both parents. Finally, a holistic sense of safety and security in the family extends beyond the aggregate effects of individual family subsystems. It can be conceptualized as a collective set of experiences within the overall family system (Forman & Davies, 2005). Ongoing exposure to IPC may reduce confidence in the predictability and stability of family bonds.

Findings from several empirical studies support emotional security in multiple family subsystems as key mediators linking exposure to IPC and subsequent maladjustment across developmental phases (Cummings et al., 2006, 2012, 2015; Davies et al., 2008; de Silva et al., 2021; El-Sheikh et al., 2008). For example, a longitudinal study of 243 preschoolers found that internal representations reflecting emotional insecurity in the family system mediated the association between IPC and externalizing problems two years later (Parry et al., 2020). Another longitudinal study (N = 235) tested interparental security as an intervening process explaining exposure to IPC and mental health outcomes across development from childhood to adolescence (Cummings et al., 2012). Exposure to IPC in kindergarten was indirectly related to internalizing and externalizing problems in adolescence through emotional security in middle childhood. In a study of 141 families, insecurity about the parent-child relationship was a unique mediator in the association between IPC and adolescent adjustment problems (de Silva et al., 2021).

Promoting Emotional Security through Parent-child Relationships

In addition to the extensive evidence that IPC undermines children’s emotional security, other family processes have been identified to positively affect children’s sense of safety and security in their families. Generally, high-quality parent-child relationships are robustly protective in the face of adversity (Mortensen & Mastergeorge, 2014; Yap et al., 2016), particularly following separation/divorce (Sandler et al., 2012). However, prior research is mixed about whether parent-child relationships are protective against IPC specifically (Davies, Harold, et al., 2002; DeBoard-Lucas et al., 2010; El-Sheikh & Elmore–Staton, 2004; Grych et al., 2004; O’Hara et al., 2021). In samples of children from married families, two studies found that supportive parent-child relationships reduced the extent to which children blamed themselves for IPC (DeBoard-Lucas et al., 2010; Grych et al., 2004). However, a study of children from divorced/separated families did not observe the hypothesized protective effect of a high-quality parent-child relationship to buffer children’s fear of abandonment in the context of post-separation/divorce IPC (O’Hara et al., 2021). Other studies have shown that there may be complex relations among IPC, parent-child relationship quality, and children’s outcomes in divorced/separated families (Sandler et al., 2008, 2013).

The Role of Adolescent Emotional Insecurity in Different Family Structures

Although research supports parent-child, interparental, and family insecurity as explanatory mechanisms in the association between IPC and mental health problems, little is known about whether the three forms of insecurity operate similarly across family structures. Most empirical studies on EST have been conducted using samples primarily or exclusively comprised of youth from married or cohabitating families (exception: Cummings et al., 2011). Findings have been theoretically extended to youth from divorced or separated families. Still, there is a lack of empirical studies investigating emotional security in youth from diverse family structures and living arrangements (exception: López-Larrosa et al., 2019).

The literature examining IPC and emotional insecurity in youth from divorced families is limited. First, a few studies have examined measures conceptually related to emotional security in separated/divorced families (Lee, 1997; van Eldik et al., 2020). For example, a meta-analysis supported general indices of parent-child relationship quality as an explanatory variable in the link between post-divorce IPC and externalizing problems. In a sample of 559 youth from divorced families, fear of abandonment mediated the effects of IPC on increased youth- and teacher-reported internalizing problems and teacher-reported externalizing problems (O’Hara et al., 2021). Still, little is known about whether IPC is related explicitly to insecurity in the parent-child relationship (van Dijk et al., 2020).

Second, some studies assessed emotional insecurity in youth from diverse family structures but did not explicitly examine whether family structure moderated associations between IPC and emotional security. For example, one longitudinal study examined the role of perceived family functioning to explain the link between IPC and depression in a sample of 107 adolescents from married and divorced families (Unger et al., 2000), but they did not test differential effects as a function of family structures. Another study (cross-sectional; N = 142) examined the relation between IPC and adjustment problems separately for youth from married and divorced families (Forehand et al., 1989). They showed distinct pathways from IPC to adjustment problems based on family structure (i.e., a direct effect for youth from married families and an indirect effect for youth from divorced families). Still, they did not test for group differences directly.

Testing Emotional Security Theory Across Family Structures

The current study builds on prior research to test generalizability of EST’s key hypotheses by assessing whether associations among adolescents’ exposure to IPC, three types of emotional insecurity, and psychological symptoms varied as a function of family structure (i.e., married versus divorced). Given the limitations of empirical work on differential associations between IPC, emotional security, and adjustment across family structures, our hypotheses were based on the notion that as the interparental relationship dissolves, the holistic family system becomes more salient than the interparental subsystem as a source of security and safety in the family. Thus, given the absence of one parent, family insecurity may be a more robust correlate of youth mental health problems than interparental insecurity in divorced families. Conversely, given the presence of both parents, interparental insecurity may be a more robust correlate of youth mental health problems than family insecurity in married families. Across family structures, IPC may be less detrimental if youth have a strong, supportive relationship with their parents to buffer or offset emotional security threats from exposure to IPC (Davies, Harold, et al., 2002; DeBoard-Lucas et al., 2010; El-Sheikh & Elmore-Staton, 2004; Grych et al., 2004; O’Hara et al., 2021; Yap et al., 2016).

Current Study

This study used a large sample of adolescents ages 10–15 to examine whether associations among IPC, emotional insecurity, and mental health problems differed for youth from married versus divorced families. We predicted that IPC would be related to emotional insecurity and emotional insecurity related to mental health problems in both groups. However, more specific questions about the relative significance of different forms of insecurity across family subsystems and relations with child adjustment have not been systematically addressed in past research. Based on the relative presence of the parents, we expected family insecurity would be more strongly associated with IPC and mental health problems for adolescents from divorced families. In contrast, interparental insecurity would be more strongly associated with IPC and mental health problems for adolescents from married families. We also expected high-quality parent-child relationships to attenuate relations between IPC, emotional insecurity, and mental health problems in married and divorced families.

Method

Participants

Participants were 1,032 adolescents ages 10–15 (51% male) recruited from a pool of 1,290 sixth-, seventh-, and eighth-grade students at a public school in a middle-class suburb of a United States metropolitan area. Teachers reported on mental health symptoms for 798 (77.3%) adolescents. The two primary reasons teachers declined to participate included (1) substitute teachers who did not know enough about their students to provide reports; and (2) teachers who were uninterested or too busy. Most adolescents (69.7%; n = 719) lived with married parents, and 28.4% (n = 293) lived with divorced parents. Information on parents’ marital status was missing for 20 cases (1.9%). The racial and ethnic distribution matched that of the school: 82% non-Hispanic White, 9% Black or African American, 5% Hispanic, 2% Asian or Pacific Islander, and 2% Native American. In the school district, 23.1% received free (14.4%) or reduced (8.7%) lunch.

Study Procedures

Researchers used a two-step consent/assent process. First, parents were allowed to refuse permission for their adolescent to participate by returning a form via prepaid envelope or contacting the school or research team. Second, adolescents were allowed to participate or not participate after learning about the study. Reasons for non-participation included absence from school (n = 71), parental refusal (n = 115), and adolescent refusal (n = 72). Participants completed paper and pencil measures in their classroom with guidance from a trained research assistant. All participants completed the same study measures regardless of family structure. They earned a chance to win several gift certificates as an incentive to participate.

Measures

Interparental Conflict

Participants completed items from selected subscales of the Children’s Perception of Interparental Conflict Scale (23 items, CPIC; Grych et al., 1992). They rated perceived truth of each item on a 3-point scale, true (1) to false (3). Selected items were mean-scored on four subscales: frequency (6 items; α = .78; e.g., “I often see my parents arguing”), intensity (7 items; α = .82; e.g., “My parents get really mad when they argue”), resolution (6 items; α = .85; e.g., “Even after my parents stop arguing, they stay mad at each other), and content (4 items; α = .80; e.g., “My parents’ arguments are usually about something I did”). All items were scored such that higher scores indicated higher levels of IPC. CPIC has well-established psychometric properties in samples of children and adolescents (Grych et al., 1992; Holt et al., 2020; Moura et al., 2010).

Emotional Insecurity

Family insecurity.

Participants completed the Security in the Family System Scale (24 items; SIFS; Forman & Davies, 2005). SIFS assesses perceived security in the family as a whole during the past year. They rated their agreement with each statement. SIFS items are rated on a 4-point scale ranging from completely disagree (1) to completely agree (4). Items were mean-scored on three subscales: preoccupied (8 items; α = .86; “I feel like something could go very wrong in my family at any time”), secure (7 items; α = .85; “I feel I can count on my family to give me help and advice when I need it”), dismissive (7 items; α = .83; “I don’t care what goes on in my family”). Items on the secure subscale were reverse-coded so that higher scores indicated higher levels of family insecurity. We fitted a second-order factor model where items load on subscales and subscales load on the factor, as presented in Forman & Davies, 2005 to compute a factor score to represent family insecurity in structural models. SIFS has evidence of good internal consistency, test-retest reliability, and discriminant validity and predicts mental health problems across samples (e.g., Cummings et al., 2015; Forman & Davies, 2005).

Interparental insecurity.

Participants completed the Security in the Interparental System Scale (33 items; SIS; Davies et al., 2002). SIS assesses perceived security in the interparental subsystem during the past year. They rated their agreement with each statement during the past year. SIS items are rated on a 4-point scale ranging from not at all true of me (1) to very true of me (4). Items comprise six subscales: emotional reactivity (9 items; α = .87; e.g., “I can’t calm myself down”), behavioral dysregulation (3 items; α = .56; e.g., “I yell or say unkind things”), avoidance (7 items; α = .79; e.g., “I feel like staying far away from them”), involvement (6 items; α = .73; e.g., “I try to solve the problem for them”), destructive family representations (4 items; α = .78; I worry about my family’s future”), and conflict spillover representations (4 items; α = .82; e.g., “I feel like it’s my fault”). Higher scores indicated higher interparental insecurity. We fitted a second-order factor model where items load on subscales and subscales load on the factor, as presented in Davies et al., 2002 to represent interparental insecurity in structural models. SIS has evidence of good internal consistency, test-retest reliability, and convergent validity, and predicts youth mental health problems across samples (e.g., Davies et al., 2002; Gao et al., 2019; Holt et al., 2020).

Parent-child insecurity.

Participants completed the 15-item (α = .89 - .90) Child-Parent Attachment Security Scale (CPAS-C; Davies, Harold, et al., 2002) for each parent separately. Although they reported on each parent separately, a mean score was used to equate all three measures of emotional security in this study. CPAS-C was designed to assess perceived security in the parent-child subsystem during the past year. They rated how well each item described their relationship with their parents on a 4-point scale, not at all true of me (1) to very true (4). An example item is, “When I am upset, I go to my [mother/father] for comfort.” Higher scores indicated higher levels of parent-child insecurity. CPAS has evidence of good internal consistency and test-retest reliability and is associated with measures of attachment security, parenting quality, and mental health problems (Davies, Harold, et al., 2002).

Parent-child Relationship Quality

Acceptance/rejection.

Participants completed the 10-item (α = .92 - .93) acceptance scale of the Parental Acceptance and Rejection Questionnaire (PARQ; Rohner, 1980). They reported on each parent separately. They rated how much they agreed with each statement on a 4-point scale, almost always true (1) to almost always false (4). An example item is, “My [mother/father] talks to me in a warm and loving way.” High scores indicated higher levels of parental rejection. Meta-analytic data supports PARQ as a reliable measure of perceived parental rejection/acceptance that predicts psychological adjustment (Khaleque & Rohner, 2002).

Psychological control.

Participants completed the 10-item (α = .82 - .83) Psychological Control Scale-Youth Self-Report Scale (PCS; Barber, 1996) to assess parents’ use of psychological control tactics (those that manipulate, distort, or limit the youth’s emotional experience). They reported on each parent separately. Response options were on a 4-point scale, not like them (1) to a lot like them (4). An example item is, “My [mother/father] would like to be able to tell me how to feel or think about things.” High scores indicated higher psychological control. PCS predicted youth mental health problems in cross-sectional and longitudinal studies (Kaniušonytė & Žukauskienė, 2016; Romm et al., 2020).

Mental Health Problems

Participants completed anxious/depressed, withdrawn, delinquent behavior, and aggressive behavior subscales of the Youth Self-Report (YSR; Achenbach, 1991) to assess internalizing (20 items; α = .89; e.g., “unhappy, sad and depressed”) and externalizing (30 items; α = .89; “gets in many fights”) problems. YSR items are rated on a 3-point scale, not true (1) to very true/often true (3). Teachers completed the Teacher Report Form (TRF; Achenbach, 1991) on adolescents’ internalizing (5 items; α = .84) and externalizing (5 items; α = .80) problems. YSR and TRF have well-documented reliability and validity evidence (Achenbach, 1991) to identify mental health disorders in youth in clinical and community settings (Warnick et al., 2008).

Data Analysis Approach

We used Mplus Version 8.5 (Muthén & Muthén, 1998–2017) and a structural equation modeling framework. Multiple group models were estimated using maximum likelihood robust standard error correction estimator for unbiased estimates with missing data (i.e., MLR; Yuan & Bentler, 2000). We conducted Satorra-Bentler Scaled Chi-Square difference tests to assess for a significant detriment to model fit when we constrained paths to be equal across groups compared to an unconstrained model in which paths were freely estimated for both groups. Missing data were handled with full information maximum likelihood (FIML). Data were missing across key study variables for 4.8–22.7% (mean = 11%, median = 8%) of the sample. We used SPSS to identify potentially influential cases by conducting regression diagnostics to assess observed change in significant coefficients of theoretical predictors when cases were excluded (using cutoffs proposed by Cohen et al., 2003) and Cook’s distance (using a cutoff of >.20; Bollen & Jackman, 1985). We identified no potentially influential cases; thus, we reported results with all cases included.

For significant interactions in our moderation analyses, we conducted post-hoc analyses based on Roisman et al.’s (2012) recommendations to assess the nature of the interaction effect. First, we simplified the model to focus on the interaction of interest; we dropped outcomes for which there was no significant interaction effect, retaining only the interaction effect that was significant and covariates. Second, we used the R-based interActive application (McCabe et al., 2018) to (1) plot interactions for illustration in a way that includes information about parameter uncertainty, (2) estimate simple slopes at various levels of the moderator (e.g., +/− 1 SD, mean), and (3) conduct a region of significance on Z test to identify values of relationship quality for which IPC and the focal outcome were significantly correlated in our sample. Finally, we quantified the portion of interaction index (PoI index; Roisman et al., 2012) to assess whether the interaction was theoretically consistent with diathesis-stress (e.g., risk factors are most likely to impact particularly vulnerable adolescents) or differential susceptibility (e.g., vulnerability factors that amplify maladaptation under certain conditions and promote positive adaptation under others). PoI values of .20 – .80 support differential susceptibility (Del Giudice, 2017).

Statistical Power & Inferences about Negligible Associations

A sample size calculator (Soper, 2022) indicated that our sample size of N = 1032 exceeded the minimum sample size of N = 200 needed to detect a small effect (z = 0.1; Cohen, 1992) in our structural equation models (i.e., comprised of one latent variable and six manifest variables with a statistical power level of .8 and a probability level of .05). There are well-documented limitations of drawing a conclusion of “no group differences” based on nonsignificant results from null hypothesis significant tests (see Alter & Counsell, 2021; Altman & Bland, 1995; Counsell & Cribbie, 2015; Jabbari & Cribbie, 2022). Thus, we also conducted equivalence, or “negligible effect” tests, using the R-based negligible application (Cribbie et al., 2022) for associations that were significant for at least one group. The application uses both a symmetric confidence interval approach (100•[1–2α]%) and the Anderson and Hauck (1983) procedure. Our smallest effect size of interest (SESOI) was a small-medium effect (b = .03; Cohen, 1992). The null hypothesis is the group difference regression coefficient is non-negligible (i.e., larger than the SESOI); the alternative hypothesis is the parameter falls within the equivalence bounds and, thus, a negligible effect (i.e., smaller than the SESOI) can be concluded.

Measurement Invariance

We tested measurement invariance across family structure groups for our key measures of emotional security. We fitted and evaluated the evidence for configural, weak, and strong invariance models through a sequential process of imposing model constraints and evaluating corresponding decreases in model fit. The SIS, SIFS, and CPAS evidenced no meaningful reduction in model fit according to absolute fit statistics (i.e., CFI/TLI change of less than .01, RMSEA remained within the confidence interval of the less restrictive model) as a function of measurement invariance between intact and divorced families. The CPAS also evidenced no meaningful reduction in model fit according to global fit statistics (i.e., no significant change in χ2) as a function of measurement invariance between married and divorced families.

Preliminary, Primary, and Exploratory Analyses

In preliminary analyses, we tested for group differences (i.e., married vs. divorced) on all study variables, including IPC, emotional security, parent-child relationship quality, and mental health problems. In primary analyses, we followed our preregistered analytic plan (available at OSF link). We tested a series of multiple group multivariate path models to assess (1) associations between IPC and measures of emotional insecurity and (2) associations between measures of emotional insecurity and internalizing and externalizing problems (separate models for self-report and teacher-report). See Figure 1 for an illustration of the models tested. Note that although prior research supports emotional security as a mediator of the relation between IPC and children’s mental health problems, the cross-sectional data used in this study precluded an unbiased test of mediation (see Maxwell & Cole, 2007 for an extensive discussion of the reasons against empirical tests of mediation with cross-sectional data, including biased estimates and misleading conclusions). Thus, the two models were tested separately with the intention to examine basic relations among key constructs and conduct well-powered tests of group differences to establish justification for future longitudinal studies. Adolescent gender and grade were included as covariates in all models. In exploratory analyses, we tested whether measures of parent-child relationship quality (parent-child insecurity, rejection, psychological control; three separate models) attenuated associations between IPC and four outcomes: interparental insecurity, family insecurity, internalizing problems, and externalizing problems.

Figure 1.

Conceptual models

Note. Conceptual models were tested separately for youth from divorced/separated and married/cohabitating families. IPC = interparental conflict.

Based on preliminary evidence that there may be complex relations among IPC, parent-child relationship quality, and outcomes in divorced families (Sandler et al., 2008, 2013), we planned to test three-way interactions among father-child relationships, mother-child relationships, and IPC. However, we discovered that scores on relationship quality were highly correlated in married families (i.e., PARQ: r = .84, PCS: r = .83; CPAS: r = .77). Thus, we created composite scores by taking the mean of mother-child and father-child relationship quality in married/cohabiting families. We conducted an exploratory analysis to test higher-order interactions only in the divorced subsample, where scores on father-child and mother-child relationship quality were less strongly correlated (PARQ: r = .48, PCS: r = .46; CPAS: r = .38).

Results

Complete code for statistical modeling is available at Open Science Framework.

Preliminary Analyses

Supplemental Table 1 displays descriptive statistics and correlations. Supplemental Table 2 displays correlations separately for married and divorced families. Adolescents from divorced families reported significantly higher scores than adolescents from married families on measures of IPC; interparental, family, and parent-child insecurity; and parental control and rejection (Supplemental Table 3). According to self- and teacher-report, adolescents from divorced families had significantly higher externalizing problems.

Primary Analyses: Associations Between IPC and Emotional Insecurity

Exposure to IPC significantly predicted all forms of emotional insecurity, including family (β_M = .598, 95% CI = .531, .666, p < .001; β_D = .470, 95% CI = .341, .599, p < .001), interparental (β_M = .504, 95% CI = .428, .580, p < .001; β_D = .414, 95% CI = .287, .541, p < .001), and parent-child (β_M = .497, 95% CI = .419, .574, p < .001; β_D = .226, 95% CI = .121, .331, p < .001), for adolescents from married and divorced families (Table 1). Associations between IPC and family insecurity and interparental insecurity did not differ between groups; however, the deleterious association between IPC and parent-child insecurity was stronger for adolescents from married families (β_diff = .271, p < .001). The chi-square difference test yielded a significant detriment to model fit when we constrained paths from IPC to each form of emotional insecurity to be equal across groups (χ2diff [3] = 16.23, p = .001). When we constrained one path at a time to understand whether group differences were driven by the association between IPC and a specific form of insecurity, we found that constraining the association between IPC and parent-child insecurity to be equal across groups resulted in a significant detriment to model fit (χ2diff [1] = 25.45, p < .001). Conversely, the chi square difference tests did not yield a significant detriment to model fit with constraints to associations between IPC and family (χ2diff [1] = 1.29, p = .256) or interparental security (χ2diff [1] = 3.75, p = .053). Equivalence tests indicated that a negligible effect could be concluded for group differences in the association between IPC and family and interparental insecurity, but not IPC and parent-child insecurity (AH T = 4.04, p =.033).

Table 1.

Associations, between-group difference tests, and equivalence / negligible effect tests

	Married/Cohabitating (n = 717)	Divorced/Separated (n = 293)	Between-Group Difference Test	Equivalence / Negligible Effect Test *
Model 1: IPC and Forms of Emotional Insecurity
CPIC → SIFS	β = .598 p < .001	β = .470 p < .001	βdiff = .129, SE = .074 p = .082	95% CI = [−0.016, 0.274] AH T = 1.74, p =.01
CPIC → SIS	β = .504 p < .001	β = .414 p < .001	βdiff = .090, SE = .075 p = .233	95% CI = [−0.057, 0.237] AH T = 0.98, p =.01
CPIC → CPAS	β = .497 p < .001	β = .226 p < .001	βdiff = .271, SE = .067 p < 0.001	95% CI = [0.14, 0.402] AH T = 4.04, p =.33
Model 2: Forms of Emotional Insecurity and Mental Health Problems (child report)
SIFS → INT	β = .379 p < .001	β = .290 p < .001	βdiff = .089, SE = .091 p = .329	95% CI = [−0.09, 0.268] AH T = 0.98, p =.01
SIS → INT	β = .328 p < .001	β = .439 p < .001	βdiff = −.111, SE = .069 p = .108	95% CI = [−0.246, 0.024] AH T = −1.61, p <.001
CPAS → INT	β = .041 p = .342	β = .083 p = .207	βdiff = −.042, SE = .079 p = .597	-
SIFS → EXT	β = .446 p < .001	β = .210 p = .013	βdiff = .235, SE = .102 p = .021	95% CI = [0.035, 0.435] AH T = 2.30, p =.26
SIS → EXT	β = .079 p = .051	β = .185 p = .017	βdiff = −.106, SE = .088 p = .226	95% CI = [−0.279, 0.067] AH T = 1.20, p =.01
CPAS → EXT	β = .105 p = .027	β = .146 p = .083	βdiff = −.041, SE = .097 p = .674	95% CI = [−0.231, 0.149] AH T = −0.42, p <.001
Model 3: Forms of Emotional Insecurity and Mental Health Problems (teacher report)
SIFS → INT	β = .126 p = .075	β = .025 p = .787	βdiff = .101, SE = .117 p = .385	-
SIS → INT	β = −.020 p = .701	β = .048 p = .544	βdiff = −.068, SE = .095 p = .473	-
CPAS → INT	β = −.056 p = .375	β = −.061 p = .507	βdiff = .006, SE = .112 p = .959	-
SIFS → EXT	β = .110 p = .094	β = .278 p = .068	βdiff = −.168, SE = .167 p = .313	-
SIS → EXT	β = .107 p = .057	β = −.088 p = .298	βdiff = .195, SE = .101 p = .055	-
CPAS → EXT	β = .013 p = .817	β = .021 p = .883	βdiff = −.008, SE = .152 p = .956	-
covariates = gender, grade

Note. CPIC = Children’s Perception of Interparental Conflict; SIS = Security in Interparental System; SIFS = Security in Family System; CPAS = Child-parent Attachment Security; INT = internalizing problems; EXT = externalizing problems. AH T = Anderson-Hauck T statistic; Bolded = statistically significant.

^*The smallest effect size of interest (SESOI) was a “small-medium” effect (i.e., standardized regression coefficient of .30; Cohen, 1992); thus, negligible effect tests are based on an equivalence interval of ± .30. Note that the null hypothesis associated with the AH T statistic is that the regression coefficient is non-negligible (i.e., larger than the SESOI).

To test the hypothesis the family system may be more salient than the interparental subsystem as a source of security and safety in separated/divorced families, we examined whether the relative association between IPC and family insecurity versus IPC and interparental insecurity differed across groups. The association between IPC and family insecurity was larger than between IPC and interparental insecurity for adolescents from married and divorced families. The relative difference was not significant for either the married group (β_diff = .094, p = .059) or the divorced group (β_diff = .056, p = .506).

Primary Analyses: Associations Between Emotional Security and Mental Health Problems

Internalizing Problems

Family insecurity and interparental insecurity each significantly predicted self-reported internalizing problems (SIFS: β_M = .379, 95% CI = .277, .481, p < .001; β_D = .290, 95% CI = .143, .436, p < .001 and SIS: β_M = .328, 95% CI = .245, .411, p < .001; β_D = .439, 95% CI = .332, .545, p < .001) for adolescents in married and divorced families. Parent-child insecurity did not significantly predict self-reported internalizing problems for either group (PAS: β_M = .041, 95% CI = −.044, .041, p = .342; β_D = .083, 95% CI = −.046, .212, p = .207).

Externalizing Problems

Family insecurity predicted self-reported externalizing problems in married and divorced families (SIFS: β_M = .446, 95% CI = .334, .557, p < .001; β_D = .210, 95% CI = .044, .377, p = .013). Although all associations were in the same predicted direction, interparental insecurity significantly predicted self-reported externalizing problems for adolescents from divorced families only (SIS: β_M = .079, 95% CI = .000, .158, p = .051; β_D = .185, 95% CI = .033, .337, p = .017), whereas parent-child insecurity significantly predicted self-reported externalizing problems for adolescents from married families only (CPAS: β_M = .105, 95% CI = .012, .198, p = .027; β_D = .146, 95% CI = −.019, .311, p = .083) (See Table 1). Despite a different pattern of significance within groups, comparisons of group differences in path coefficients were not significant (SIS: β_diff = −.106, p = .226; CPAS: β_diff = −.041, p = .674).

Full Model: Mental Health Problems

The strength of associations between forms of emotional insecurity and self-reported mental health problems did not differ for adolescents from married versus divorced families, except that the positive association between family insecurity and externalizing problems was stronger for adolescents in married families (β_diff = .235, p = .021). However, the chi-square difference test did not yield a significant detriment to model fit when we constrained paths from each form of emotional insecurity to internalizing and externalizing problems to be equal across groups (χ2diff [6] = 9.48, p = .148). Equivalence tests indicated that a negligible effect could be concluded for group differences in all associations between forms of emotional insecurity and self-reported mental health problems, except family insecurity and externalizing problems (AH T = 2.30, p =.26). No form of emotional insecurity predicted teacher-reported problems. We also examined whether the relative association between family versus interparental insecurity and mental health problems differed across groups. The relation between interparental insecurity and self-reported internalizing problems was stronger for adolescents from divorced families. The association between family insecurity and self-reported externalizing problems was stronger for adolescents from married families; however, group differences were not statistically significant. There were no statistically significant differences between family versus interpersonal insecurity on teacher-reported problems.

Exploratory Moderation Analyses

We found that the association between IPC and interparental insecurity was moderated by parent rejection and parent psychological control in the divorced group but not the married group and moderated by parent-child insecurity in the married group but not the divorced group. However, multiple-group model comparisons indicated that the between-group differences in interaction parameters were not statistically significant (PARQ: b = .001; p = .065; PCS: b = .001, p = .171; CPAS: b = .000; p = .801); thus, we did not probe these interaction effects further. Also, the chi-square difference tests did not yield a significant detriment to model fit with constraints for any of the interaction models (PARQ: χ2diff [4] = 3.262, p = .515; PCS: χ2diff [4] = 3.231, p = .520; CPAS: χ2diff [4] = 0.676, p = .954).

Next, we examined our original research question regarding higher-order interactions among IPC, mother-child relationship quality, and father-child relationship quality on forms of emotional insecurity and mental health problems in the divorced subgroup. We found no evidence of three-way interactions. When we examined two-way interactions for mothers and fathers, we found that the association between IPC and interparental insecurity differed by mothers’ rejection (b = −.112, p = .001) and psychological control (b = −.001, p = .039).

Upon decomposing interactions, we found that the simple slope of interparental insecurity regressed on IPC was positive and significant at low and moderate maternal rejection levels and psychological control levels. For those with high maternal rejection and psychological control (above the 84^th and 80^th percentile, respectively), IPC was not significantly related to interparental insecurity (see Supplemental Figure 1). The PoI value (i.e., the ratio of better outcomes to the sum of better and worse) for maternal rejection was .30, representing the proportion of adolescents that fared better under more benign or supportive conditions (i.e., low IPC and high acceptance/low rejection). These findings support low rejection/high acceptance as a susceptibility factor, consistent with a differential susceptibility model (Roisman et al., 2012). The PoI value for maternal psychological control was .83, representing the proportion of adolescents that fared better under more benign or supportive conditions (i.e., low IPC and low psychological control). These findings support low psychological control as a protective-reactive factor, consistent with a risk saturation model (Luthar et al., 2000).

Discussion

In a sample of over 1,000 adolescents, we tested whether associations among IPC, forms of emotional insecurity, and mental health problems varied as a function of family structure. This dataset offered many benefits, such as a large, community-based sample with ample representation of adolescents from different family structures. The cross-sectional data allowed us to test basic relations among key constructs and conduct adequately powered multiple group tests. Given the dearth of empirical data on emotional security processes in children from separated/divorce families, this is a critical first step to establish a foundation for future longitudinal studies. The most important contribution is to show that IPC had the same implications for emotional insecurity across multiple family structures with few exceptions. All forms of emotional insecurity had the same implications for mental health problems for adolescents residing in married and separated/divorced families. For two reasons, our direct test of EST-based hypotheses across family structures is an important development in process-oriented research on IPC. First, it supports the generalizability of EST as an explanatory framework that applies both theoretically and empirically to adolescents in diverse family contexts. Second, despite decades of research supporting IPC as a robust risk factor that appreciably accounts for the increased risk of maladjustment following parental separation/divorce, there is a lack of empirical evidence illuminating process-oriented pathways through which exposure to IPC confers its harmful effects for youth from separated/divorced families. This study raises our confidence in drawing on the EST literature to hypothesize and test mechanisms accounting for youth adjustment in the context of post-separation/divorce IPC.

The Particularly Deleterious Nature of Post-Separation/Divorce IPC

This study adds to the extensive evidence that IPC confers risk for maladjustment across multiple family structures and highlights that parental separation/divorce is a distinct risk factor for higher exposure to IPC, emotional insecurity, and mental health problems. Adolescents from divorced families showed a higher mean level of risk on these outcomes, and for many variables, the difference was statistically significant. This suggests that the risk for maladjustment associated with separation/divorce may be partially explained by a higher likelihood of exposure to IPC and related risk factors. This emphasizes the need for process-oriented research to inform and evaluate interventions that reduce and prevent maladjustment for youth who experience post-separation/divorce IPC.

General and Specific Associations among IPC, Emotional Security, and Mental Health

We observed that IPC predicted all forms of emotional insecurity, including child-parent, interparental, and family insecurity. Contrary to our predictions, the relative associations between IPC and interparental insecurity and family insecurity did not differ for adolescents from married versus divorced families. This means that even after the dissolution of the interparental romantic relationship, adolescents continue to derive a sense of emotional security, or insecurity, specific to the interparental relationship, which is heavily influenced by parents’ interactions around disagreements. This raises important questions about the quality of co-parenting relationships in divorced families. The current state of the literature on the protective effects of post-divorce co-parenting is confounded by co-parenting being defined as a complex construct that is likely multidimensional and includes both low conflictual behaviors and high cooperative behaviors (e.g., Lamela et al., 2016; Saini et al., 2019). Although there is some evidence that positive co-parenting is protective against post-separation/divorce youth maladjustment (Teubert & Pinquart, 2010), it is not clear whether these effects are driven by an absence of destructive conflict (i.e., aggression, hostility; Cummings & Wilson (1999), presence of constructive conflict (i.e., cooperation, problem-solving), or both. It will be important to assess whether reducing destructive post-separation/divorce IPC is the central goal of protecting adolescents or whether it is also important to increase constructive conflict and cooperation.

We also observed that all forms of emotional insecurity predicted self-reported internalizing and externalizing problems for adolescents from married and divorced families, except that parent-child insecurity only predicted externalizing problems in married families. It is difficult to make sense of this finding in the context of the larger literature that supports the role of high-quality parent-child relationships in preventing externalizing problems, particularly after separation/divorce (Sandler et al., 2012). Parent-child relationships are typically defined as a broad construct that includes attachment, among other qualities, such as warmth and communication. However, in the case of more complex family constellations, parent-child security may be derived from relationships with multiple attachment figures in ways that reduce the impact of any single attachment relationship. For example, adolescents from divorced families may be more likely to have a broader network of caregivers, such as grandparents, aunts/uncles, and stepparents (Hunter et al., 1998). This may be a similar mechanism seen in ethno-racial cultural groups in which adolescents who are embedded in close-knit extended family networks are less influenced by the security of attachment in any single relationship (McLoyd et al., 2000). Another possibility is that because the protective effects of post-separation/divorce parenting often operate through secure attachment in the parent-child relationship and consistent and effective discipline, parent-child insecurity alone may not be a robust predictor of post-separation/divorce externalizing problems.

In this study, the significant associations between emotional security and youth mental health problems were limited to self-reported mental health problems, raising questions about whether these findings are influenced by shared method variance or simply reflect differences in available knowledge by reporters. One consideration is that participants in this study were in middle school and had multiple teachers, which may have affected the teacher’s awareness of their mental health problems. Another consideration is that low agreement rates among reporters of youth mental health problems are well documented (Achenbach et al., 1987; De Los Reyes & Kazdin, 2005). Scholars theorize that incongruent reports may be due to differences in how youth display symptoms in different contexts and note that reporters have unique positions to observe certain symptom domains (i.e., internalizing versus externalizing problems) (De Los Reyes et al., 2015). In this study, youth and teacher reports of mental health problems were only modestly correlated (r = .13 and r = .27 for internalizing and externalizing problems, respectively).

Protective Under Which Conditions? Low Parental Rejection and Control with High IPC

Contrary to our hypothesis that parenting is protective in the face of adversity, parent-child relationship quality did not moderate the link between IPC and any form of emotional security. We found that in separated/divorced families, IPC was significantly related to interparental insecurity, except under the conditions of high maternal rejection and psychological control. We observed distinct forms of moderation for low maternal rejection versus low maternal psychological control. Low rejection functioned as a differential susceptibility factor in the relation between IPC and interparental insecurity. The pattern suggests that high-quality relationships (i.e., characterized by high acceptance and low rejection) may confer benefits when IPC is low but costs (disproportionate risk for insecurity) when IPC is high. It may be that adolescents who appraised their mothers as warm and supportive have higher emotional stakes in the welfare of the interparental relationship for better or worse. This finding is similar to a prior study demonstrating adolescents from separated/divorced families who had particularly close relationships with their fathers reported higher fears of abandonment in families with high versus low IPC (O’Hara et al., 2021).

In contrast, we found that low psychological control functioned as a protective-reactive factor, conferring protection in the form of greater interparental security only at low levels of IPC. When IPC is high, the risk of IPC may supersede any protection conferred by experiencing low levels of psychological control. However, although the benefits of low psychological control diminish and are no longer evident at high levels of IPC, adolescents who experience low levels of psychological control do not appear to fare noticeably worse in terms of interparental insecurity at high levels of IPC. In other words, unlike high maternal acceptance, low maternal psychological control was not associated with extra vulnerability under adverse IPC conditions. Despite the extensive literature supporting the protective nature of parenting, this adds to the growing number of studies raising questions on the ability of high-quality parenting to counteract the deleterious effects of IPC (Davies, Harold, et al., 2002; Grych et al., 2004; Lutzke et al., 1996; O’Hara et al., 2021).

Limitations and Future Directions

First, as with all cross-sectional studies, although well-suited to test primary associations implicated in an explanatory framework, formal mediation tests are limited, particularly concerning conclusions about temporal precedence. This study explicitly did not include tests of mediation effects. However, given the results of this study, there is a strong argument for conducting future longitudinal studies to assess theoretical mediation effects, in order to better characterize these processes in children from separated/divorced families. Second, the sample is primarily comprised of non-Hispanic White adolescents from one geographic location in the United States. The study of IPC for youth from ethnically and racially marginalized communities is a critical and urgent next step for advancing the generalizability of EST as an explanatory framework and as a guiding model for theory-based interventions. Third, we did not observe cross-reporter effects, raising the possibility of inflated correlations due to shared method variance. Future studies should include multi-informant assessment to provide a more comprehensive picture of these processes. Fourth, we did not have detailed information available about family dynamics, such as the amount of contact with and between parents and time since divorce/separation, or other important family context variables, such as socioeconomic status. These are important variables to test as possible moderators in future studies. Finally, group differences not detected in our multiple group statistical models may exist. However, our use of equivalence or “negligible effect” inference tests bolsters the argument that any group differences likely to be observed are practically and statistically negligible (Alter & Counsell, 2021).

Clinical Implications and Future Directions

This study underscores the importance of multiple intervention strategies for high-IPC families. First and foremost, clinicians working with high-IPC should focus on reducing IPC events. However, we acknowledge that IPC can be an intractable set of behaviors difficult to eliminate, particularly in the context of contentious separation or divorce proceedings. Further, youth exposed to IPC become more sensitized over time (Goeke-Morey et al., 2013). Thus, clinicians and researchers need to work together to evaluate other strategies to promote youth mental health in the context of “breakthrough” IPC events. Possible targets include: (1) increasing the youth’s capacity to cope with IPC, particularly the difficult emotions and maladaptive cognitions it provokes, (2) increasing the parents’ use of constructive conflict strategies (Cummings et al., 2008; Cummings & Schatz, 2012; Cummings & Wilson, 1999; Devonshire et al., 2022), and (3) teaching parents to be sensitive and responsive to their children’s fears, such as reassuring them that the conflict will be resolved and they will be cared for by both parents indefinitely. Empirical studies to test the effects of these strategies will bolster the public health impact of interventions for youth exposed to high levels of IPC.