The Developmental Costs and Benefits of Children’s Involvement in Interparental Conflict

Patrick T Davies; Jesse L Coe; Meredith J Martin; Melissa L Sturge-Apple; E Mark Cummings

doi:10.1037/dev0000024

. Author manuscript; available in PMC: 2016 Aug 1.

Published in final edited form as: Dev Psychol. 2015 Jun 8;51(8):1026–1047. doi: 10.1037/dev0000024

The Developmental Costs and Benefits of Children’s Involvement in Interparental Conflict

Patrick T Davies ¹, Jesse L Coe ², Meredith J Martin ³, Melissa L Sturge-Apple ⁴, E Mark Cummings ⁵

PMCID: PMC4516573 NIHMSID: NIHMS690884 PMID: 26053147

Abstract

Building on empirical documentation of children’s involvement in interparental conflicts as a weak predictor of psychopathology, we tested the hypothesis that involvement in conflict more consistently serves as a moderator of associations between children’s emotional reactivity to interparental conflict and their psychological problems. In Study 1, 263 early adolescents (M age = 12.62 years), mothers, and fathers completed surveys of family and child functioning at two measurement occasions spaced two years apart. In Study 2, 243 preschool children (M age = 4.60 years) participated in a multi-method (i.e., observations, structured interview, surveys) measurement battery to assess family functioning, children’s reactivity to interparental conflict, and their psychological adjustment. Across both studies, latent difference score (LDS) analyses revealed that involvement moderated associations between emotional reactivity and children’s increases in psychological (i.e., internalizing and externalizing) problems. Children’s emotional reactivity to interparental conflict was a significantly stronger predictor of their psychological maladjustment when they were highly involved in the conflicts. In addition, the developmental benefits and costs of involvement varied as a function of emotional reactivity. Involvement in interparental conflict predicted increases in psychological problems for children experiencing high emotional reactivity and decreases in psychological problems when they exhibited low emotional reactivity. We interpret the results in the context of the new formulation of emotional security theory (e.g. Davies & Martin, 2013) and family systems models of children’s parentification (e.g., Byng-Hall, 2002).

Keywords: interparental conflict, child coping, child emotional reactivity, child psychopathology, emotional security

Children’s involvement in conflicts between their parents has long been regarded as a risk factor for psychological maladjustment (Davies & Cummings, 1994; Emery, 1989; Johnston, Roseby, & Kuehnle, 2009). Consistent with this thesis, a meta-analysis of over a dozen studies indicated that children’s involvement in interparental conflict was a significant correlate of their internalizing and externalizing symptoms (Rhoades, 2008). By the same token, the modest to moderate strength of these associations highlights the need to better understand the role children’s involvement plays in their coping and adjustment. According to family process models (e.g., Byng-Hall, 2002; Cummings & Davies, 1996), children’s involvement is an inextricable part of understanding the nature and developmental sequelae of their broader constellations of responding to parental difficulties. Therefore, our primary objective in this paper is to examine whether involvement may function more consistently as a moderator of associations between children’s distress responses to interparental conflict and their psychological difficulties. More specifically, we test the hypothesis that the multiplicative interaction between involvement and distress is a consistent predictor of children’s internalizing and externalizing symptoms even after the inclusion of each of these forms of responding to conflict as additive predictors.

Although several conceptualizations share the assumption that involvement in interparental conflict is an important component in understanding how children cope in high conflict homes (e.g., Crockenberg & Langrock, 2001; Davies & Cummings, 1994; Grych & Fincham, 1990), involvement is most prominently featured in emotional security theory (EST; Davies & Cummings, 1994). According to EST, children’s involvement in interparental conflict is a central barometer for difficulties in preserving a sense of emotional security in the interparental relationship. Interpreted within this model, children’s efforts to become involved in parental conflicts reflect their attempts to exert control over parental emotionality and behavior before disputes can escalate and result in more serious, negative consequences. However, children’s efforts to serve as protectors, confidantes, or emotional caretakers of parents in the midst or aftermath of conflict require considerable forethought, emotional investment, and personal risk. Thus, EST proposes that protracted involvement in parental disputes over time increases children’s risk for developing internalizing and externalizing symptoms through several mechanisms. For example, high levels of involvement are proposed to reflect the disproportionate allocation of psychobiological resources toward personal safety at the expense of investing in other significant goals necessary for successful adaptation (e.g., mastery of physical and social environment). The progressive accumulation of difficulties in resolving multiple developmental challenges, in turn, may broadly increase children’s risk for experiencing psychopathology. Likewise, high levels of involvement are commonly accompanied by preoccupation, helplessness, distress, and loyalty pulls with parents and, as a result, may set the stage for more pervasive adjustment problems (Davies, 2002).

Family systems theory converges with EST in proposing that forms of involvement in the interparental relationship that reflect emotional preservation or enmeshment incur significant psychological costs (e.g., Byng-Hall, 2002; Jurkovic, Morrel, & Thirkfield, 1999; Minuchin, 1974). These models maintain that distinguishing between harmful and benign forms of involvement in adverse family circumstances requires analysis of the substance and quality of children’s involvement strategies. Benign forms of involvement are defined by mild forms of child caretaking behavior that maintain the child’s distinct identity and autonomy. Conversely, developmentally destructive types of “parentification” involve excessive emotional and psychological entanglement. The resulting dependency and enmeshment in interparental difficulties are further proposed to increase children’s vulnerability for subsequent psychological problems.

Taken together, family systems theory and EST share the assumption that more precisely understanding the developmental implications of children’s involvement in conflict requires distinguishing between different forms of involvement based on whether or not they carry significant personal risk and emotional investment. Guided by these models, research has differentiated between two dimensions of involvement: mediation and coercion. Mediation, which is also known in the literature as “involvement” (Buehler, Lange, & Franck, 2007), agentic involvement (Schermerhorn, Chow, & Cummings, 2010; Schermerhorn, Cummings, DeCarlo, & Davies, 2007), facilitative behavior (Davis, Hops, Alpert, & Sheeber, 1998), and proactive mediation (Shelton & Harold, 2008), is characterized by relatively benign involvement behaviors manifested in helping, intervening, refereeing, and comforting parents. Conversely, coercive involvement, which is synonymously referred to as aggressive involvement (Davis et al., 1998), dysregulated involvement (Schermerhorn et al., 2010), and overinvolvement (Shelton & Harold, 2008) in the literature, is specifically comprised of intervention behaviors carrying high psychological burden and risk in the form of protecting a parent, siding with one parent against the other, and confronting or arguing with parents.

Although research identifying more precise forms of involvement is still in its early stages, empirical evidence is mixed on whether the concept of coercive involvement yields incremental leverage in identifying children most at risk for developing psychological problems. For example, Shelton and Harold (2008) reported that coercive involvement (i.e., overinvolvement), but not mediation, was a predictor of increases in children’s internalizing and externalizing problems. However, other studies have failed to document any relationship between coercive forms of involvement and children’s psychological problems in cross-sectional and prospective designs (e.g., Buehler et al., 2007; Davies & Cummings, 1998). Still other research has shown that the role of coercive involvement as a predictor of changes in children’s adjustment was modest in magnitude and limited to a specific informant (i.e., child) or outcome (i.e., externalizing problems) (Davis et al., 1998; Jouriles, Rosenfield, McDonald, & Mueller, 2014). Thus, although disaggregating involvement into multiple dimensions may provide some traction in better characterizing children’s risk for psychological problems, the inconsistent findings from this research highlight the importance of adopting complementary approaches.

Our primary thesis is that the discrepancies in findings across studies reflect that children’s involvement in interparental difficulties is an inextricable part of a broader pattern of responding to conflict. Previous approaches to documenting associations between children’s emotional insecurity and psychological adjustment have largely examined involvement as an additive predictor or as part of a linear composite with other forms of reactivity to conflict. At a theoretical level, this approach is at odds with the organizational perspective of EST and its assumption that the developmental meaning and implications of any single form of reactivity varies as a function of children’s other dimensions of responding to interparental conflict (Cummings & Davies, 1996). Thus, against the backdrop of linear, additive approaches to assessing involvement, the present paper is designed to break new ground by examining the synergistic interplay between the two prevailing forms of involvement (i.e., coercion, mediation) and emotional reactivity in predicting children’s psychological problems.

EST provides firm bases for examining emotional reactivity to interparental conflict as a critical context for understanding the multivariate role of involvement. As with involvement, children’s concerns about emotional security are proposed to be expressed through high levels of emotional reactivity that are comprised of intense, prolonged fearful and distressed reactions to interparental conflict. With its strong conceptual emphasis on children’s emotionality in the context of interparental conflict, EST frames children’s emotional reactivity as the cornerstone of the goal-corrected system of emotional security (Davies, Winter, & Cicchetti, 2006). In accord with this conceptualization, the EST component of emotional reactivity is a consistent predictor of children’s psychological problems across a wide array of methodological approaches (e.g., naturalistic, experimental), designs (e.g., cross-sectional, longitudinal), and analytic models (e.g., multivariate) (e.g., Buehler et al., 2007; Davies, Cicchetti, & Martin, 2012; Davies & Cummings, 1998). Based on the central role emotional reactivity plays in defining security, we specifically investigate whether coercive involvement and mediation may function as vulnerability or susceptibility factors that moderate or alter the well-established association between children’s emotional reactivity and children’s internalizing and externalizing symptoms.

Although questions remain about precisely how children’s forms of involvement may moderate the risk posed by emotional reactivity, the translation of developmental psychopathology models to EST yields three prevailing classes of models: vulnerable-stable, vulnerable-reactive, and vulnerable-adaptive. Each of these models is explicitly designed to offer a comprehensive characterization of involvement as a moderator by examining both dimensions of the interaction (see Luthar, Cicchetti, & Becker, 2000; Roisman et al., 2012). In fully delineating involvement as a moderator of emotional reactivity, the models vary along the dimension of how involvement alters associations between emotional reactivity and psychopathology. By the same token, guidelines for distinguishing between the three models in the literature also call for the analysis of whether associations between children’s involvement and their psychological problems vary at high and low levels of emotional reactivity.

In drawing on the developmental psychopathology taxonomy of vulnerability effects (Luthar et al., 2000), the vulnerable-stable model shown in Figure 1a proposes that responding to conflict characterized by high involvement, emotional reactivity, or both are comparable in the degree to which they increase children’s vulnerability to psychological problems. Thus, the combination of emotional reactivity and involvement does not incrementally increase children’s disadvantage above and beyond either form of reactivity considered singly. As denoted by the solid line in the figure, high levels of involvement specifically take a significant and stable psychological toll on children regardless of the level of emotional reactivity they experience during interparental conflicts. Thus, children’s emotional reactivity in this model only predicts increases in their internalizing and externalizing symptoms when their involvement in conflict is low (see the dotted line in Figure 1a). Likewise, due to the mutual potency of involvement and emotional reactivity risk factors, involvement is proposed to predict higher levels of children’s psychological problems only when emotional reactivity is low. Although no studies have examined the interplay between emotional reactivity and forms of involvement, the moderating role of coercive involvement may be more likely than mediation to assume a vulnerable-stable form. More specifically, if coercive involvement is a more potent risk factor than mediation, as some empirical findings indicate (e.g., Davis et al., 1998; Shelton & Harold, 2008), it is plausible that both emotional reactivity and coercion would predict children’s problems when the other form of responding to conflict is low.

a. Conceptual illustration of the vulnerable-stable form of moderation.

b. Conceptual illustration of the vulnerable-reactive form of moderation.

c. Conceptual illustration of a vulnerable-adaptive form of moderation.

In the second conceptualization of moderation shown in Figure 1b, the vulnerable-reactive model proposes that high involvement amplifies predictive pathways between children’s emotional reactivity and their psychological difficulties (Luthar et al., 2000). Whereas emotional reactivity is a relatively weak predictor of changes in children’s psychological problems at low levels of involvement (see dotted line in Figure 1b), it may take an exponentially greater toll on children when they are heavily involved in interparental problems (see solid line in Figure 1b). A primary premise is that the greater psychological immersion underlying children’s involvement in interparental conflict magnifies the impact of children’s emotional reactivity to parental disagreements on their psychological adjustment problems. In a corresponding fashion, the vulnerable-reactive model also hypothesizes that children’s extensive involvement in interparental conflict only carries significant psychological costs when it is organized by high levels of emotional reactivity. Without an existing empirical base to guide predictions, it is premature to formulate any firm hypotheses on the viability of identifying the vulnerable-reactive form of moderation. However, because mediation and coercion are each characterized by some level of emotional entanglement in conflicts (Buehler et al., 2007; Davies & Cummings, 1998; Harold & Shelton, 2008), it is possible that associations between children’s emotional reactivity and their psychopathology could be magnified in a vulnerable-reactive way in the presence of either form of involvement.

As the third conceptualization of involvement as a moderator, the vulnerable-adaptive model proposes that the meaning of children’s involvement shifts drastically from carrying psychological disadvantages at high levels of emotional reactivity to conferring significant adaptive benefits at low levels of emotional reactivity. Thus, as shown by the steep slope of the solid line in Figure 1c, the vulnerable-adaptive and vulnerable-reactive models share the hypothesis that associations between emotional reactivity and psychopathology will be particularly evident when children are highly involved in their parents’ conflicts. However, as the central source of distinction between the two models, the vulnerable-adaptive model proposes that children exhibiting higher levels of involvement in parental conflicts may actually fare substantially better in their psychological functioning when they experience low levels of emotional reactivity. Although the form of the moderation is comparable to the interaction proposed by differential susceptibility theory (Belsky & Pleuss, 2009), the mechanisms postulated to underlie the effect are different. Thus, when children experience heightened emotional reactivity to interparental conflict, high levels of involvement reflect their prolonged difficulties preserving emotional security and, as a consequence, disproportionately increase their vulnerability to psychological problems (Davies, 2002). Conversely, under conditions of low emotional reactivity (e.g., low fear, distress, and vigilance), involvement is theorized to be motivated by well-regulated, empathetic concern for the welfare of their parents. Accordingly, involvement coupled with low emotional reactivity may have benign ramifications for children’s mental health (Davies & Martin, 2013). Vulnerable-adaptive models presuppose that there are null associations between the proposed moderator (i.e., mediation, coercive involvement) and outcomes (Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007). Therefore, by virtue of its proposed innocuous implications for children’s development, mediation may be particularly likely to assume the moderating role proposed in the vulnerable-adaptive model. However, due to inconsistencies in the identification of coercion and mediation as risk factors (Buehler et al., 2007; Davies & Cummings, 1998), it is also plausible that the both forms of involvement may evidence vulnerable-adaptive moderating effects.

In summary, the present article is the first, to our knowledge, to test the theoretically guided hypothesis that the multiplicative interplay between children’s emotional reactivity and involvement in the face of interparental conflict uniquely predicts their psychological problems. Children’s internalizing and externalizing symptoms not only have substantial societal implications but are the most commonly investigated correlates and sequelae of their reactivity to interparental conflict (e.g., Rhoades, 2008). Therefore, to afford a base of comparison with the existing knowledge, we examine these two forms of psychological problems as outcomes of the interaction between emotional reactivity and involvement. To provide an authoritative test of the replicability of our findings, we capitalize on data drawn from two large studies (n > 240) of children and their families that vary in their measurement techniques, methodological designs, and developmental characteristics of the children.

Study 1

As a first step to addressing our aims, Study 1 examined the synergistic interplay between children’s emotional reactivity and involvement as predictors of subsequent change in psychological maladjustment for young adolescents. Early adolescence is a highly significant developmental period for understanding children’s coping with interparental conflict. Research supports the premise that adolescence is a period of heightened sensitivity to interparental problems (e.g., Buchanan, Maccoby, & Dornbusch, 1991; Cummings, Schermerhorn, Davies, Goeke-Morey, & Cummings, 2006). Although no studies, to our knowledge, have examined age differences in children’s forms of behavioral involvement over developmental periods that include adolescence, teens have been shown to experience stronger impulses to mediate conflicts in comparison to younger children (Cummings, Ballard, El-Sheikh, & Lake, 1991; Davies, Myers, & Cummings, 1996). Given that both parents and children report that the effectiveness of interventions diminishes substantially from childhood into adolescence (Covell & Miles, 1992), heightened involvement in conflicts may do little more than embroil adolescents in protracted discord between parents and, in the process, increase their risk for psychological problems.

Relative to earlier developmental periods, dimensions of sensitivity to conflict are also theorized to increase in adolescence by virtue of teens’ greater awareness of subtle interparental difficulties and longer histories of exposure to interparental conflict (Cummings et al., 2006; Davies et al., 1999). Increases in the temporal stability in patterns of reactivity to family adversity in conjunction with concomitant rises in developmental (e.g., autonomy from parents, identity development) and contextual (e.g., peak occurrence of stressful life events) challenges during adolescence may further amplify their vulnerability to psychological problems (e.g., Davies & Cummings, 2006; Rhoades, 2008). Consistent with this conceptualization, metaanalytic findings revealed that associations between children’s involvement and emotional reactivity and their internalizing and externalizing symptoms were significantly stronger during adolescence in comparison to childhood (Rhoades, 2008).

To rigorously address our objective, we utilized a multi-informant (i.e., adolescents, mothers, and fathers) measurement battery nested within a longitudinal design. We selected a two-year time window between measurement occasions based on empirical and conceptual considerations. At an empirical level, multiple laboratories examining the sequelae of adolescent responses to interparental conflict have examined changes in psychological problems over two year periods (e.g., Buehler et al., 2007; Cummings et al., 2006; Shelton & Harold, 2008). At a conceptual level, EST proposes that children’s responses to interparental conflict set in motion the emergence of risk mechanisms that, over months or years, gradually intensify, broaden, and crystallize into the development of children’s psychological problems (e.g., Davies et al., 2006). This unfolding sequence of processes is theorized to take place over a longer period (i.e. years rather than months) during adolescence than earlier periods by virtue of progressive constraints in the plasticity of functioning during the teen years (Frankenhuis & Del Giudice, 2012).

Methods

Participants

Data for this study were drawn from a longitudinal project on family relationship processes and adolescent development. Participants in the larger study consisted of 280 families with adolescents who were recruited through local school districts and community centers in a moderate-sized metropolitan area in the Northeast and a small city in the Midwest. However, due to our aim of examining children’s responses to interparental conflict, families were only included in this paper if the mothers, fathers, and adolescents had regular contact with each other. Regular contact was specifically defined as maintaining contact as a triad for an average of 2 to 3 days per week during the year. Instituting this criterion resulted in the exclusion of 17 families from this paper, yielding a sample of 263 mothers, fathers, and adolescents. Participants returned two years later to participate in the second wave of data collection. The retention rate across the two-year period was 85%. The average age of adolescents at Wave 1 was 12.62 years (SD = .57), with 50% of the sample consisting of girls. Median household income of the families was between $55,000 and $74,999 per year. Median education level of mothers and fathers was between some college education and an Associate’s degree. Most parents (i.e., 89%) were married at the outset of the study. The majority of the adolescents were White (74%), followed by smaller percentages of African American (17%), multi-racial (8%), and other races (1%). Approximately 4% of the adolescents were Latino. Adolescents lived with their biological mother in most cases (93%), with the remainder living with an adoptive or stepmother (4%) or a female guardian (3%). In addition, children lived with their biological father in the majority of cases (79%), with the remainder of the sample living with either an adoptive or stepfather (16%) or a male guardian (5%).

Procedures and Measures

Families visited the laboratory at one of the two research sites. During the first wave, adolescents completed survey measures to assess their responses to interparental conflict, and their mothers filled out demographic and family background information. At each wave, mothers, fathers, and children completed questionnaires to assess children’s internalizing and externalizing symptoms. All research procedures were approved by the Institutional Review Board at both research sites prior to conducting the study. Families were compensated monetarily for their participation.

Adolescent responses to interparental conflict

Adolescents completed three scales derived from the Security in the Interparental Subsystem (SIS) scales to assess their emotional reactivity and involvement (Davies, Forman, Rasi, & Stevens, 2002). Children’s emotional reactivity was specifically indexed by children’s reports on the Emotional Reactivity scale. The scale consists of nine items designed to capture children’s multiple, prolonged fear and distress in response to conflict (e.g., “When my parents argue, I feel scared”). The newer, longer version of the SIS contains two involvement scales that distinguish between mediation and coercion (see Shelton & Harold, 2008). The “Mediation” scale consists of six items assessing children’s efforts to intervene in the conflict through comforting (e.g., “I try to comfort one or both of them”), distracting (e.g., “I try to distract them by bringing up other things”), solving the problem (e.g., “I try to solve the problem for them”), or directly interceding (e.g., “I tell them to stop fighting”). In contrast, the SIS Coerciveness scale, which is also referred to as the Overinvolvement scale (Shelton and Harold, 2008), contains five items that capture bossy, aversive, and domineering approaches to interrupting parental conflicts (e.g., “I argue with one or both of them,” “I end up taking sides with one of them,” “I tell one of my parents that he or she is wrong”). Response alternatives for the SIS items were as follows: 1 = Not at all true of me; 2 = A little true of me; 3 = Somewhat true of me; and 4 = Very true of me. To confirm the factor structure of the items in the two involvement scales, we conducted a principal components analysis with varimax rotation. The results supported a two-factor solution (e.g., eigenvalues were greater than 1.0 for the two factors) that discriminated between the items on the Coerciveness and Mediation subscales. Alpha coefficients were .89, .75, and. 74 for the Emotional Reactivity, Coerciveness, and Mediation scales, respectively. The validity of the scales is supported by their concurrent and prospective associations with children’s exposure to family conflict and their psychological adjustment (e.g., Davies, Sturge-Apple, Bascoe, & Cummings, 2014; Shelton & Harold, 2008).

Adolescent internalizing and externalizing symptoms

To obtain multi-informant assessments of internalizing and externalizing symptoms, mothers, fathers, and adolescents completed child psychopathology questionnaires. Mother and father reports of adolescent psychological problems were derived from the Internalizing (e.g., “Nervous, high-strung, or tense,” “Unhappy, sad, or depressed”) and Externalizing (e.g., “Temper tantrums or hot temper,” “Gets in many fights”) scales from the Child Behavior Checklist (CBCL; Achenbach, Dumenci, & Rescorla, 2003). Previous studies have provided consistent evidence for the reliability and convergent and discriminant validity of the CBCL scales (e.g., Achenbach et al., 2003). Consistent with recommendations, we used the sum of raw scores on the CBCL scales in the analyses. Alpha coefficients indexing internal consistency ranged from .82 to .91 for parents across the two waves. In addition, children completed the Conduct Problems (5 items; e.g., “I fight a lot”) and Emotional Symptoms (5 items; e.g., “I worry a lot”; “I am often unhappy, downhearted, and tearful”) scales from the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1999) to obtain assessments of externalizing and internalizing symptoms, respectively. The psychometric properties of the two SDQ scales have been supported in previous studies (e.g., Goodman, 1999). Internal consistencies for the two scales across the two waves ranged from .64 to .74. The three-informant approach specifically permitted use of mother, father, and child reports as manifest indicators of latent constructs of internalizing and externalizing symptoms at each wave.

Covariates

Two covariates were derived from parent reports of demographic background characteristics: (1) children’s gender (1 = boys; 2 = girls) and (2) total annual family income based on a 13-point ordinal scale ranging from 1 (less than $6,000) to 13 ($125,000 or more). As the third and final covariate, we assessed destructive interparental conflict based on mother and father reports on three scales. Mother and fathers completed the 10-item O’Leary Porter Scale to assess children’s exposure to marital hostility (OPS; Porter, & O’Leary, 1980; e.g., “How often do you and/or your partner display verbal hostility [raised voices, etc.] in front of your child?”). The remaining scales, which were derived from the Conflict and Problem-Solving Scales (CPS; Kerig, 1996), reflected the frequency with which parents and their partners engaged in (a) physically violent acts (e.g., Physical Aggression; e.g., “Push, pull, shove, grab partner”) and (b) impasses in conflict characterized by unresolved hostility, distress, and disengagement (e.g., Stonewalling; e.g., “Storm out of the house”). Internal consistencies for the maternal and paternal reports on the four scales ranged from .78 to .91. The six scales were standardized and aggregated to form a single, parsimonious composite of interparental conflict (α = .81).

Results

Descriptive and Preliminary Analyses

For descriptive purposes, Table 1 provides the means, standard deviations, and correlations for primary variables in the study. Inspection of the means for individual scales of the interparental conflict composite revealed means that were similar to community samples of families in previous studies. For example, means on the CPS Physical Aggression (M = 2.11, SD = 3.60) and Verbal Aggression (M = 22.92, SD = 9.21) scales were comparable to levels identified in prior working and middle class samples of families (CPS Physical Aggression Means ranging from 2.06 to 2.14; CPS Verbal Aggression Means ranging from 22.15 to 23.87; Crockenberg, Leerkes, & Lekka, 2006 ; El-Sheikh, Harger, & Whitson, 2001; Fosco & Grych, 2007; McConnell & Kerig, 2002).

Table 1.

Means, Standard deviations, and Correlations for the Primary Variables in the Study 1 Analyses.

	Mean	SD	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
Wave 1 Covariates
1. Child Gender	0.50	0.50	–
2. Household Income	8.47	3.14	−.06	–
3. Interparental Conflict	0.00	0.71	.03	−.22^*	–
Wave 1 Child Reactivity to Conflict
4. Emotional Reactivity	14.63	5.61	.09	−.15^*	.20^*	–
5. Mediation	12.48	3.97	.06	−.10	.14^*	.34^*	–
6. Coerciveness	6.36	2.23	−.04	−.17^*	.17^*	.35^*	.30^*	–
Wave 1 Internalizing Symptoms
7. Paternal CBCL	5.58	5.10	.05	−.13^*	.22^*	.15^*	.02	.12	–
8. Maternal CBCL	5.87	5.43	.11	−.10	.24^*	.02	−.05	−.01	.42^*	–
9. Child SDQ	2.12	2.23	.24^*	−.19^*	.13^*	.41^*	.05	.26^*	.26^*	.25^*	–
Wave 1 Externalizing Symptoms
10. Paternal CBCL	5.65	6.53	−.08	−.26^*	.13^*	.07	.02	.30^*	.52^*	.26^*	.22^*	–
11. Maternal CBCL	5.62	6.14	−.13^*	−.16^*	.19^*	.03	−.11	.23^*	.31^*	.48^*	.23^*	.66^*	–
12. Child SDQ	1.79	1.82	−.10	−.22^*	.08	.22^*	.00	.35^*	.17^*	.09	.46^*	.53^*	.45^*	–
Wave 2 Internalizing Symptoms
13. Paternal CBCL	4.51	5.38	.09	−.22^*	.21^*	.09	−.07	.11	.63^*	.25^*	.26^*	.42^*	.32^*	.17^*	–
14. Maternal CBCL	4.78	4.82	.16^*	−.18^*	.18^*	.04	−.04	.12	.37^*	.61^*	.19^*	.29^*	.44^*	.09	.42^*	–
15. Child SDQ	2.33	2.32	.34^*	−.16^*	.11	.15^*	.06	.00	.24^*	.25^*	.46^*	.11	.04	.11	.26^*	.29^*	–
Wave 2 Externalizing Symptoms
16. Paternal CBCL	4.93	6.28	−.07	−.23^*	.15^*	.17^*	−.01	.39^*	.51^*	.13	.22^*	.74^*	.60^*	.39^*	.62^*	.30^*	.07	–
17. Maternal CBCL	4.76	5.55	−.08	−.25^*	.12	.14^*	−.04	.39^*	.31^*	.27^*	.26^*	.59^*	.76^*	.45^*	.32^*	.47^*	.08	.70^*	–
18. Child SDQ	1.84	1.80	−.12	−.32^*	.13	.21^*	−.01	.26^*	.14^*	.07	.18^*	.35^*	.34^*	.35^*	.20^*	.17^*	.20^*	.37^*	.48^*

Open in a new tab

Note. Child gender: 1= boys; 2 = girls. p ≤ .05.

To index change in adolescent internalizing and externalizing problems, we used latent difference score (LDS) modeling across the two measurement occasions. LDS offers a powerful way of testing whether the interaction predicts interindividual differences in intraindividual changes in adjustment over time through the specification of latent change scores (McArdle, 2009). Because our goal of identifying predictors of adolescent psychological problems first requires demonstrating that there are significant individual differences in intraindividual change in their symptoms, we tested this assumption by specifying successive unconditional LDS models for internalizing and externalizing symptoms following procedures for testing strong factorial variance (Widaman, Ferrer, & Conger, 2010). Thus, as Figure 2 depicts, measurement equivalence across time was maximized by placing the following constraints in the analyses: (1) fixed and free factor loadings are identical over time; (2) factor loadings of each of the indicators of internalizing and externalizing symptoms were constrained to be equal across time; and (3) intercepts of the same indicators were fixed to be invariant across time. Both models fit the data well: χ² (9, N = 263) = 15.72, p = .07, RMSEA = .05, CFI = 0.98, and χ²/df ratio = 1.75, for internalizing symptoms; and χ² (9, N = 263) = 14.59, p = .10, RMSEA = .05, CFI = .99, and χ²/df ratio = 1.62, for externalizing symptoms. Moreover, statistically significant variances of 7.31 (z = 5.07) for externalizing symptoms and 5.99 (z = 3.18) for internalizing symptoms indicated that there were significant individual differences in mean changes in adolescent problems from Time 1 to Time 2. Therefore, given our aim of examining why teens differ in their psychological problems over time, we proceeded to our primary analytic goal of examining the interplay between adolescent emotional reactivity and involvement as predictors of changes in their internalizing and externalizing symptoms.

A conceptual illustration of the specification of unconditional latent difference score models. τ and λ denote factor loadings and intercepts, respectively. λ₁₁ and λ₂₁ are fixed to 1. Loading constraints include λ₂₁ = λ₂₂ and λ₃₁ = λ₃₂. Intercept invariance is as follows: τ₁₁ = τ₁₂; τ₂₁ = τ₂₂; and τ₃₁ = τ₃₂.

Primary Analyses

To test coerciveness and mediation as moderators of associations between adolescents’ emotional reactivity and change in internalizing and externalizing symptoms, four analytic models were conducted based on consecutively pairing emotional reactivity with the two forms of involvement in predicting LDS change for internalizing and externalizing symptoms separately (i.e., 1 emotional reactivity predictor × 2 forms of involvement × 2 outcomes). For each model, we examined emotional reactivity, the specific form of involvement, and their multiplicative product as predictors of LDS change in adolescent psychopathology over the two-year lag in measurement occasions using structural equation modeling (SEM). Emotional reactivity, mediation, and coercion were centered to reduce multicollinearity in the moderator analyses. In addition, we included exposure to interparental conflict, child gender, and family income as covariates in the prediction of change in adolescent psychological symptoms. The same strong factorial invariance constraints in the unconditional LDS were specified in the primary analyses (Widaman et al., 2010). In addition, correlations were also specified between: (a) the primary predictors (i.e., emotional reactivity and involvement), (b) the covariates, (c) the covariates and primary predictors, (d) the Wave 1 latent construct of adolescent psychological functioning and each of the covariates and predictors, and (e) the same informant’s measure of psychological difficulties across the two time points. Finally, consistent with analytic recommendations in LDS approaches, an autoregressive path was specified between the Wave 1 latent construct of adolescent psychological functioning in each model and the latent change in functioning from Wave 1 to Wave 2 (McArdle, 2009). SEM analyses were conducted using full-information maximum likelihood (FIML) in Amos 22.0 to estimate missing data (i.e., data were missing for 10.0% of the values) and retain the full sample for primary analyses (Enders, 2001).

Internalizing symptoms

The two models examining the moderating roles of adolescents’ forms of involvement in analyses of change in their internalizing symptoms provided satisfactory to marginal representations of the data: χ² (31, N = 263) = 79.85, p < .01, RMSEA = .08, χ²/df ratio = 2.58, and CFI = .90 for the mediating type of involvement; and χ² (31, N = 263) = 108.86, p < .01, RMSEA = .10,χ²/df ratio = 3.48, and CFI = .87 for the coercive form of involvement. In support of the measurement models, the loadings of the manifest indicators onto the Wave 1 and 2 latent constructs of internalizing symptoms were all significant (p < .001) and moderate to strong in magnitude (ranging from .47 to .65). Further analysis of the structural model revealed that none of the structural paths running from the primary predictors to latent change in internalizing symptoms were significant. Therefore, for succinctness, the findings from the model are not presented.

Externalizing symptoms

SEM analyses testing adolescent mediation in parental conflicts as a moderator of the association between their emotional reactivity and change in their externalizing problems provided a good fit with the data: χ² (33, N = 263) = 53.20, p < .05, RMSEA = .048, χ²/df ratio = 1.61, and CFI = .97. As shown in Figure 3, the model for adolescent externalizing symptoms yielded significant findings for the predictor and interaction. Consistent with prior research, adolescent emotional reactivity predicted increases in externalizing symptoms over time, β= .19, p < .05. Pertinent to the study hypotheses, the interaction between adolescent emotional reactivity and mediation was also significantly associated with LDS increases in externalizing symptoms over the two-year period, β = .22, p < .01.

The Study 1 LDS model of adolescents’ mediation and emotional reactivity to interparental conflict as predictors of change in their externalizing symptoms over a two-year period. Significant structural paths are bolded. For clarity, only significant correlations among predictors and covariates are included in the figure. ^* p < .05.

The moderating effect was first clarified by graphically plotting and calculating simple slopes of emotional reactivity at high (+1 SD) and low (−1 SD) levels of adolescents’ mediation in parental conflicts. To portray the interaction across a relatively comprehensive range of the proposed predictor (i.e., 95%), we conducted simple slope plots and calculations at −2 SD and + 2 SD from the centered mean of emotional reactivity (see Roisman et al., 2012). Findings from the simple slope analyses revealed that emotional reactivity predicted increases in adolescent externalizing symptoms at high, b = 0.21, p < .001, but not low levels, b = −0.01, p = .85, of mediation. As depicted in Figure 5a, the graphical plot of the moderating effects revealed a disordinal (i.e., cross-over) interaction. Although the simple slope analyses and visual inspection of the graphical plot correspond most closely with the vulnerable-adaptive model (see Figure 1c), they do not provide a definitive analysis of its explanatory power relative to the vulnerable-reactive model. Both of these models posit that emotional reactivity is a stronger predictor of psychopathology when mediation is high. Moreover, in spite of the visual similarities between the graphical plot and the interaction proposed in the vulnerable-adaptive model, Roisman and colleagues (2012) have cautioned against using graphical plots to examine whether moderator results are more consistent with models proposing an ordinal interaction (i.e., vulnerable-reactive) or a disordinal interaction (i.e., vulnerable-adaptive).

a. A graphical plot of the interaction between adolescent emotional reactivity and mediation in predicting LDS change in their externalizing symptoms over a two-year period.

b. A graphical plot of the interaction between adolescent emotional reactivity and coerciveness in predicting LDS change in their externalizing symptoms over a two-year period.

To more authoritatively test which of the moderating patterns proposed in the vulnerable-adaptive and vulnerable-reactive models more closely corresponds with the data, statistical guidelines call for calculating regions of significance on X (RoS on X) tests (Dearing & Hamilton, 2006; Kochanska, Sanghag, Barry, & Philibert, 2011; Roisman et al., 2012). RoS on X tests specifically invert the predictor and moderator to yield analyses of the significance of the association between the moderator and outcome within the bounded regions of the proposed predictor (i.e., + or − 2 SDs). Accordingly, the RoS on X test was used to test whether mediation significantly predicted adolescent externalizing symptoms at high and low levels of their emotional reactivity (Roisman et al., 2012). Support for the vulnerable-reactive model would be evidenced by findings indicating that mediation was significantly associated with increases in adolescent psychopathology only at high levels of emotional reactivity. In contrast, the vulnerable-adaptive model would be supported if the results revealed that mediation was a predictor of significant increases in externalizing problems at high levels of emotional reactivity and significant decreases in externalizing problems at low levels of emotional reactivity. Consistent with the vulnerable-adaptive form of moderation, the results indicated that adolescent mediation in interparental conflict predicted increases in externalizing symptoms at high levels (+2 SD) of emotional reactivity, b = 0.30, p < .05, and reductions in externalizing symptoms at low (− 2 SD) levels of emotional reactivity, b = −0.33, p = .01.

By relying on null hypothesis testing, the results of the RoS on X test vary as a function of the sample size. Small sample sizes may specifically produce low power to detect moderating effects proposed in the vulnerable-adaptive mode. Conversely, large samples run the risk of producing findings in favor of the vulnerable-adaptive pattern that have little substantive significance. Thus, as a sample-independent index for testing the relative fit of the interaction with the vulnerable-adaptive and vulnerable-reactive models, we followed statistical recommendations and calculated the proportion of interaction (PoI) index (e.g., Davies et al., 2014; Roisman et al., 2012). The PoI index consists of the ratio of improved functioning (i.e., decreases in externalizing symptoms) for adolescents with high levels of mediation relative to the overall aggregate of improved and impaired outcomes. If mediation confers no significant advantages for adolescents, as the vulnerable-reactive model proposes, then the PoI index should approach .00 (i.e., between .00 and .16). Conversely, the ratio (i.e., PoI index) should be higher than .16 if the findings support the vulnerable-adaptive proposal that beneficial outcomes are common for adolescents exhibiting high levels of mediation. Strong support for the vulnerable-adaptive model would be evidenced by PoI values between .40 and .60. In support of the vulnerable-adaptive hypothesis, the PoI index for mediation was .54.

Figure 4 depicts the results of the SEM tests of adolescent coerciveness as a moderator of associations between their emotional reactivity and their changes in externalizing symptoms. The model provided a good fit with the data: χ² (33, N = 263) = 66.66, p < .01, RMSEA = .062, χ²/df ratio = 2.01, and CFI = .96. In accord with the hypotheses, the interaction between adolescent emotional reactivity and coerciveness significantly predicted LDS increases in their externalizing symptoms over the two-year period, β = .32, p < .001. The graphical plot of the moderating effect in Figure 5b revealed a disordinal interaction similar in shape to the moderating effect of adolescent mediation. Simple slope analyses further indicated that adolescents’ emotional reactivity predicted increases in their externalizing symptoms at high, b = 0.20, p < .001, but not low levels, b = −0.04, p = .50, of coerciveness. To quantitatively compare the utility of the moderating effects proposed in the vulnerable-adaptive and vulnerable-reactive models, we also conducted the RoS on X and PoI tests. Consistent with the vulnerable-adaptive model, adolescent coercive involvement predicted increases in their externalizing symptoms at high levels of emotional reactivity, b = 0.67, p < .001, and decreases in their externalizing symptoms at low levels of emotional reactivity, b = −0.57, p = .05. In addition, the PoI coefficient of .42 revealed strong support for the vulnerable-adaptive model and its hypothesis that there is a relatively equal balance of healthy and unhealthy outcomes for teens exhibiting high coerciveness across the range of the emotional reactivity variable.

The Study 1 LDS model of adolescents’ coercive involvement and emotional reactivity to interparental conflict as predictors of change in their externalizing symptoms over a two-year period. Significant structural paths are bolded. For clarity, only significant correlations among predictors and covariates are included in the figure. ^* p < .05.

Study 2

Although the Study 1 findings showed that teen involvement served as a moderator of prospective associations between their emotional reactivity and externalizing symptoms, testing the bounds of generalizability and specificity of the results is advanced through replication in a study that differs systematically in its approach. Accordingly, Study 2 was designed to provide a complementary test of the multiplicative interplay between emotional reactivity and involvement by varying methodological and conceptual characteristics. More specifically, we rigorously tested the generalizability of the findings from the predominantly White adolescents in Study 1 by examining reactivity to conflict in a racially and ethnically diverse sample of preschool children.

According to developmental models of interparental conflict (Davies et al., 2006; Kitzmann, Gaylord, Holt, & Kenny, 2003), preschool children adapt to interparental conflict in developmentally distinct ways based on a unique confluence of intraorganismic and family processes. Meta-analytic results have indicated that associations between children’s exposure to destructive interparental conflict and their psychological problems are particularly pronounced during the preschool period (Kitzmann et al., 2003). Moreover, children’s patterns of responding to interparental conflict are theorized to underpin their heightened sensitivity to interparental conflict during the preschool period. As children traverse through the preschool years, advances in social perspective taking and information processing abilities are posited to increase children’s concerns about the welfare of their parents during conflicts (Davies et al., 2006). Consistent with this finding, the limited research findings on involvement during early childhood indicate that intervention in parental and adult disputes increases significantly from toddlerhood to the late preschool years (Cummings, 1994; Cummings, Zahn-Waxler, & Radke-Yarrow, 1984). The limited sophistication and effectiveness of preschooler intervention efforts may also increase their vulnerability to subsequent psychological problems relative to school-aged children (Cummings et al., 1984). Moreover, in comparison to older children, heightened involvement is also commonly accompanied by preschooler experiences of greater fear, threat, and perceived coping inefficacy in response to interparental conflict (Cummings, Vogel, Cummings, & El-Sheikh, 1989; Grych, 1998; Jouriles, Spiller, Stephens, McDonald, & Swank, 2000).

At a conceptual level, our objective was to examine the replicability of the Study 1 findings. Consistent with Study 1, we examined whether children’s internalizing and externalizing symptoms were predicted in similar or distinct ways from the synergistic interaction between emotional reactivity and the two prevailing types of involvement in the literature: (a) mediation, consisting of comforting, distracting, solving the problem, or directly interceding in the conflict; and (b) coercive control, characterized by bossy, angry, and aggressive strategies for altering the course of the conflict. However, as a further test of the specificity and generalizability, we also expanded our multi-dimensional assessment of involvement to examine appeasing forms of intervening in the conflict. Appeasing involvement is specifically characterized by efforts to placate parents through inauthentic, demonstrative displays of charm, positivity, and over-bright behavior (e.g., Emery, 1982; Garcia O’Hearn, Margolin, & John, 1997; Johnston, et al., 2009). Although research has yet to systematically examine the developmental implications of appeasement, different sources of evidence highlight its prevalence in the face of interparental conflict. Practitioners have repeatedly emphasized that responding to interparental conflict with facades of charm and positivity is relatively common in childhood (Emery, 1982; Johnston et al., 2009). Moreover, as an indicator of appeasement, maternal diary reports indicated 15% of children exhibited “unusually well-behaved” responses to interparental conflict during a 5-week period, a percentage that was relatively similar to indicators of coercive (e.g., 14% “misbehaved”; 19% “took sides”) involvement (Garcia O’Hearn et al., 1997). Thus, as a first foray into examining appeasement in models of children’s psychological problems, we tested its role as a predictor and moderator in pathways between children’s emotional reactivity and their psychological difficulties.

Finally, although parental reports of children’s overall levels of involvement in interparental conflict have been shown to be psychometrically sound, parents’ tendencies to direct attention towards their partners during conflicts likely undermines their ability to detect specific displays of emotional reactivity (e.g., veiled expressions of fearful distress) and subtle forms of involvement (e.g., appeasing behaviors). Study 2 was designed to address this barrier through the implementation of an observational coding system that was designed to sensitively identify children’s emotional reactivity and forms of involvement in the context of an actual interparental disagreement. Moreover, we strengthened the multi-informant measurement approach in Study 1 by incorporating multiple informants and methods into the assessment battery of Study 2. In comparison to the prospective design of Study 1, our analysis of the interactions between emotional reactivity and the forms of involvement as predictors of subsequent change in children’s psychological adjustment also took place over a shorter, one-year span. Our selection of narrower developmental interval in Study 2 was guided by conceptualizations of early childhood as a period of considerable malleability in functioning and our goal of minimizing the possibility of missing significant change in children’s adjustment (e.g., Frankenhuis & Del Giuidice, 2012; Shaw, Bell, & Gilliom, 2000).