Interrelatedness of Children’s Psychological and Physiological Responses to Interparental Conflict: A Moderating Role of Harsh Parenting

Abstract

Children’s psychological and physiological responses to interparental conflict have received considerable attention due to their implications for later adjustment, yet limited research has investigated the interplay between these two response systems. This study investigates patterns of association between children’s psychological responses (e.g., emotional distress) and cortisol reactivity to interparental conflict, including possible moderations by negative caregiving environment. Participants included 193 families (mother, father, and child). Parents completed questionnaires relating to their caregiving behaviors toward the child (107 girls and 86 boys, M age = 7.99 years, SD = 0.53 years) and children’s psychological responses to interparental conflict. Children provided three saliva samples over the course of watching videos depicting conflicts between two adults, whom children were asked to pretend were their parents. Based on a series of Latent Growth Curve Models, only children’s emotional responses to interparental conflict (indicated by increased distress) were associated with greater cortisol reactivity. Additionally, fathers’ harsh parenting behavior moderated the relation between children’s emotional reactivity and cortisol reactivity, yet the moderation effect was not found for mothers’ parenting. Findings are discussed in terms of the importance of exploring both psychological and physiological reactivity to conflict and the possible moderating role of harsh parenting.

Children from high-conflict homes are at risk for developing behavioral, emotional, and social problems (Harold & Leve, 2012; Jouriles, Rosenfield, McDonald, & Mueller, 2014; Morris, Silk, Steinberg, Myers, & Robinson, 2007). To understand the processes through which interparental conflict affects children’s socioemotional development, research has been devoted to the examination of children’s responses to conflict across multiple levels of analysis (Davies & Cummings, 2006), including behavioral and emotional responses (Davies & Cummings, 1994), social-cognitive reactions (Grych & Fincham, 1990), and physiological reactivity (Katz, 2001). Although these responses are all in action when children are exposed to interparental discord, little is known about the interplay among different response systems due to the lack of research integrating these levels of analysis (Cummings, Davies, & Campbell, 2000; Katz, 2001). Are children’s different response systems co-activated or coordinated in the face of interparental conflict? How does harsh parenting, another source of family risk, affect the interrelatedness among different response systems? Guided by the neurobiological framework (Repetti, Taylor, & Saxbe, 2007; Repetti, Taylor, & Seeman, 2002) and Emotional Security Theory (EST; Cummings & Davies, 1996; Davies & Cummings, 1994), the present study aimed to address the above questions by investigating the interplay between children’s psychological and physiological reactivity to interparental conflict, including consideration of children’s caregiving environment as a possible moderator.

Witnessing interparental conflict is particularly stressful for children because it threatens children’s personal safety (Cicchetti & Rogosch, 2001) and disrupts the secure internal representations of their attachment figures (Cassidy, Ehrlich, & Sherman, 2013). Children’s physiological functioning is undermined by repeated exposure to high levels of interparental conflict (Katz, 2001; Repetti et al., 2002), disrupting children’s ability to adapt to new, stressful situations. Conceptual frameworks in the broader literature on family risk highlight the role of the hypothalamic-pituitary-adrenal (HPA) axis in children’s physiological responses to stress (Repetti et al., 2007; Susman, 2006). Stressful events activate the HPA axis, which subsequently leads to children’s release of the glutocorticoid hormone, cortisol. The elevated cortisol level in children’s bloodstream marshals resources, such as mobilizing energy (e.g., glucose, oxygen), increasing cardiovascular activity, and amplifying cognitive processing of emotionally significant events (Cahill & McGaugh, 1998; Gunnar & Vazquez, 2006). In light of the role that the HPA axis plays in allocating resources, cortisol reactivity has been used as a measure of children’s physiological response to environmental stressors, such as interparental conflict (Davies, Sturge-Apple, Cicchetti, & Cummings, 2008; Pendry & Adam, 2007; Sturge-Apple, Davies, Cicchetti, & Manning, 2012).

In addition to their automatic physiological reactions, children also respond psychologically to interparental conflict in a conscious and reflective way. Children’s behavioral, emotional, and cognitive reactivity to interparental conflict has been systematically examined by work rooted in EST (e.g., Cummings, Schermerhorn, Davies, Goeke-Morey, & Cummings, 2006; Davies & Cummings, 1998; McCoy, Cummings, & Davies, 2009). According to EST, an important goal for children in contexts of interparental conflict is to preserve a sense of security. Witnessing conflicts between parents can be particularly disruptive to children’s security system, demonstrated by various psychological responses including children intervening in their parents’ conflict, acting out with aggressive behaviors, and showing distressed reactions. In particular, elevated distressed reactions, such as increased anxiety, vigilance, and emotional upset are theorized (Cummings & Davies, 1996) and empirically supported (Davies et al., 2008) to be reliable indicators of children’s insecurity in the interparental relationship.

A complete understanding of children’s responses to conflict requires a simultaneous examination of children’s psychological responses with their adrenocortical reactivity. Empirical work stemmed from psychological traditions has primarily focused on analyzing children’s psychological responses (e.g., increased distress and hostility; Davies, Sturge-Apple, Cicchetti, & Cummings, 2007; McCoy et al., 2009). Studies rooted in neurobiological models have mostly examined the undermined functioning of children’s physiological systems in exposure to family risk factors (Heim, Meinischmidt, & Nemeroff, 2003; Katz, 2001; Repetti et al., 2002) with the relative exclusion of their psychological functioning. Driven by the repeated calls to integrate multiple levels of analysis for children exposed to family stress, research as a field is moving toward a direction in which the concurrent correspondence or interrelatedness among different types of children’s responses (i.e., psychological and physiological), elicited by witnessing interparental conflict, becomes the focus (Cummings et al., 2000; Davies et al., 2008; Fox, Hane, & Perez-Edgar, 2006; Katz, 2001).

Studies are limited in providing guidance on the nature of interrelatedness between children’s psychological and physiological reactivity. As one of the first forays into identifying this interplay, the present study relies on EST and a dual-process perspective of emotional response coherence (Evers et al., 2014), with the former suggesting a synchronized association and the latter indicating the existence of individual differences in the degree of correspondence across systems.

According to EST, children’s patterns of psychological responding to interparental conflict, developed over repeated exposure to interparental hostility, reflect their tendencies to allocate resources to cope behaviorally, cognitively, and physiologically with family adversity (Cummings & Davies, 1996; Davies & Sturge-Apple, 2007). For example, when children witness interparental conflict, their elevated distressed reactions (e.g., heightened vigilance and anxiety) may indicate the underlying adrenocortical changes that are necessary to organize resources to cope with the stress. In other words, EST postulates an association between children’s elevated psychological responses and their heightened cortisol reactivity, and it is the synchronized co-activation of these responses that characterizes children’s responses to interparental conflict.

Providing empirical evidence to this speculation, significant links between children’s psychological response and cortisol reactivity have been found at different developmental stages, from toddlerhood to adolescence (Blair et al., 2008; Davies et al., 2008; Davies, Sturge-Apple, Cicchetti, Manning, & Zale, 2009; Spies, Margolin, Susman, & Gordis, 2011). For example, Davies and colleagues (2008) found that increased cortisol reactivity was uniquely associated with children’s distress, above and beyond other forms of psychological responding (i.e., involvement and hostility). This result supported EST’s contention that children’s distressed reactions are hallmark symptoms of their concerns about safety (Cummings & Davies, 1996; Davies & Cummings, 1994). These submissive responses indicated by emotional distress and anxiety are more robust markers of children’s emotional security, compared with more active responding (i.e., involvement and hostility), and thus are more closely related to children’s cortisol activity.

The dual-process perspective of emotional response coherence (Evers et al., 2014), on the other hand, suggests that the dissociation between children’s psychological and physiological responses may be meaningful, and thus should be expected when children are exposed to interparental conflict. Research in the field of functioning and processing of basic emotions describes the emotion response system as a system that involves coordinated responses across multiple components, including experiential, behavioral, and physiological components (Levenson, 1994). The dual-process framework allocates these components into two largely independent systems, one automatic and the other reflective. Physiological responses are considered in the automatic system because they are unconscious, fast, and efficient, while psychological responses (e.g., experiential or behavioral) are categorized in the reflective system because individuals go through conscious, deliberate, and effortful consideration. The dual-process model postulates that response coherence is minimal across the two systems, and uncoupling of the two systems may be advantageous because the conscious, reflective system (e.g., behavioral) can override the automatic system (e.g., physiological) to guide optimal responses (Evers et al., 2014).

To further advance the study of children’s responses to interparental conflict, research that explores individual differences in interrelatedness between child emotional and adrenocortical reactivity is needed. Davies and colleagues (2008) found that the confluence of behavioral distress and elevated cortisol reactivity was amplified at higher levels of child involvement in conflict. In other words, children’s hypersensitivity to threat caused by witnessing conflict between their parents was particularly evident when their distress responses were coupled with high levels of involvement in the conflict. As a next step in exploring the variability in children’s response patterns to interparental conflict, the current study aimed to investigate the psychological-physiological link in the context of negative parenting practices.

Difficulties in providing consistent, responsive, and engaged parenting are documented threats to children’s sense of security; these parenting difficulties are also correlated with variations in children’s adrenocortical functioning (Cicchetti & Rogosch, 2001; Davies et al., 2007; Sturge-Apple et al., 2012), above and beyond the effects of interparental conflict. Harsh caregiving behaviors undermine children’s emotional security, indicated by distressed, overinvolved, and dysregulated behavioral reactions to the psychological stress posed by interparental conflict (Cummings & Davies, 2002). In addition, frequent engagement in negative parenting practices has been found to alter patterns of children’s physiological stress reactivity. For example, punitive parenting practices, such as spanking or slapping, have been related to young children’s elevated cortisol levels after stressful situations (Blair et al., 2008; Hastings et al., 2011).

Little is known, however, about whether negative parenting practices play a role in the association between children’s psychological and physiological reactivity, despite that the two response systems are likely to be activated simultaneously during stressful events. Providing support for the role of the caregiving environment in the associations between psychological and physiological responses, Blair and colleagues (2015) found that the interrelatedness between toddlers’ emotional reactivity and cortisol reactivity to a fear-inducting task was only evident for those who experienced sensitive maternal parenting. An implication is that mothers’ sensitivity may facilitate the coordination of these regulating systems in young children. Questions remain, however, about whether such results will hold for older children in the context of a negative parenting environment.

To better explicate the role of caregiving negativity, it is important to differentiate fathering from mothering as contexts for investigating the interrelatedness of children’s psychological and physiological responses to interparental conflict. Despite the similarities between paternal and maternal parenting (Fagan, Day, Lamb, & Cabrera, 2014), fathers and mothers are involved in various parenting activities to different extents, including engagement, control, warmth, and responsiveness to the child (Barnett, Deng, Mills-Koonce, Willoughby, & Cox, 2008; Braver & Griffin, 2000). Accumulating evidence has also suggested that maternal and paternal parenting contribute uniquely to children’s social emotional functioning (Lamb & Lewis, 2010) and stress physiology (Mills-Koonce et al., 2011; Owen et al., 2013). Therefore, maternal parenting and paternal parenting were both examined in the present study, to account for the variability in children’s stress response patterns to interparental conflict.

In sum, the present study aimed to examine 1) the associations between school-aged children’s psychological responses and cortisol reactivity to interparental conflict and 2) whether these associations varied as a function of children’s negative caregiving environment. Consistent with EST’s emphasis on distress, we hypothesized that children’s distressed reactions, rather than other reactions (i.e., involvement and hostility), were uniquely related to their adrenocortical reactivity. Moreover, we aimed to examine the role of negative parenting practices in the association between children’s psychological responses and cortisol reactivity to interparental conflict. Due to the limited literature documenting related findings that might support specific hypotheses, we did not advance specific hypotheses on these relations, but posited that the study of these relations was exploratory.

Method

Participants

The data for this study were drawn from a larger longitudinal study investigating the effects of marital and family functioning on child well-being. The original sample consisted of 235 families (father, mother, and child) who participated when children were 6-years-old on average. Families were recruited through local school districts and community centers in two moderately sized cities in the Midwest and the Northeast. Families were eligible to participate if family members lived together for at least three years, were proficient in English, and if they had a child in kindergarten at the start of the study.

Of the 214 families who participated in the wave of data collection used for this study (the third wave), 19 children provided none or inadequate amounts of saliva samples and the cortisol data from 2 children were identified as outliers (i.e., ±2 SD from the mean), and thus their data were excluded from the analyses. This resulted in a sample that included 193 families, comprising both parents and child (107 girls and 86 boys, M age = 7.99 years, SD = 0.53 years), for the analyses. The families whose data were not included in the analyses (n = 21) did not differ from the families in the final sample (n = 193) on family income, child age, or child gender. The families who were not included in the analyses did, however, report significantly higher levels of negative parenting behaviors than parents in the families whose data were included in the analyses, p < .05.

On average, both mothers and fathers included in our sample had completed some college. The majority of the parents included were White (78.2% mothers; 73.0% fathers), while smaller percentages were Black (15.5% mothers; 19.2% fathers), Hispanic (3.6% mothers; 3.6% fathers), or Other races (1.6% mothers; 2.6% fathers). Most female caregivers reported to be the biological parents of their children (94.3%), followed by smaller proportions of adoptive or step parents (3.6%) and other types of guardianship (2.0%). The sample had a median annual family income ranging between $40,000 and $54,999, and their annual income ranging from less than $6,000 to more than $100,000.

Procedure

Data in this study were collected from families during one laboratory visit. Visits were scheduled in the afternoon or early evening to control for the diurnal pattern of cortisol levels. Cortisol levels have been shown to decline over the course of a day, but the decline tends to be more gradual in the afternoon and evening hours than in the morning hours (Knutson et al., 1997; Stansbury & Gunnar, 1994). The research was approved by the Institutional Review Boards at both sites prior to conducting the study.

Marital Conflict Vignettes.

Seven videotaped clips depicting conflict between two adults were presented to each child (see Koss et al., 2011 for a more detailed description of the videotaped clips). Children were instructed to pretend that these disagreements were taking place between their parents. Each vignette was approximately 1-minute long and depicted an adult couple engaging in a conflict with varying degrees of intensity (Shamir, Cummings, Davies, & Goeke-Morey, 2005). Children were randomly assigned to one of two sets of videos: although the topic of each conflict varied between the two sets of stimuli, ordering of the conflicts was the same. After being shown a warm-up story vignette, children subsequently watched an unresolved conflict, a resolved conflict, an unresolved child-rearing conflict, another resolved conflict, an unresolved escalating conflict, and a resolution to all previous scenes.

Saliva Collection.

Three saliva samples (one pre-conflict and two post-conflict task samples) were collected through the passive drool technique with the aid of a straw (see Granger et al., 2007 for a more detailed descripton of the technique). Prior to the baseline assessment, children rinsed their mouths with water to reduce the number of contaminants in their saliva samples. Following a 7-minute period in which they were invited to play with toys, children chewed Trident sugarless gum for approximately three minutes to induce saliva production immediately prior to saliva collection. The pre-conflict sample was then collected prior to the viewing of marital conflict vignettes. The average sampling time for the pre-conflict task cortisol sample provided by children was 3:51p.m. (SD = 2 hr 7 min). Two post-conflict saliva samples were also obtained. Previous research has suggested that cortisol levels peak 20–40 min after the occurrence of the stressor (Dickerson & Kemeny, 2004). Therefore, the first post-conflict sample (post I) was collected approximately 25 minutes after the end of the child-related conflict episode (fourth in the series of seven vignettes), to assess children’s cortisol reactivity to the destructive conflict shown in the vignette. The second post-conflict sample (post II) was collected approximately 25 minutes after the final resolution (last in the series of seven vignettes) to assess children’s cortisol reactivity to the resolution of the distressing event, as shown in the final vignette.

The three samples were immediately stored at −36°C until they were shipped on dry ice to Salimetrics (State College, PA). Cortisol concentrations were analyzed in duplicate form using 25μl of saliva, and the test sensitivity ranged from .007 μg/dl to 3.0 μg/dl. The average intra-assay coefficient was 4.2% for the current sample.

Measures

Emotional security.

Children’s psychological reactions to witnessing interparental conflict were assessed by the Security in the Marital Subsystem Scale (SIMS; Davies, Forman, Rasi, & Stevens, 2002). Both mothers and fathers completed the emotional reactivity (7 items), involvement (9 items), and behavioral dysregulation (5 items) subscales. The emotional reactivity subscale assessed the degree to which children displayed distressed emotions (e.g., “still seems upset after we argue”, “appears angry”) to marital disputes. The involvement subscale measured children’s behavioral intervention in their parents’ conflicts and their aftermath (e.g., “tries to comfort one or both of us”, “watches and listens very closely”). The behavioral dysregulation (hostility) scale measured children’s use of aggressive or angry reactions (e.g., “starts hitting, kicking, slapping, or throwing things at family members,” “yells at family members”) to witnessing marital discord. Parents rated each item on a 5-point Likert scale ranging from 1 (not at all like him/her) to 5 (a whole lot like him/her). Scores for each subscale were averaged, with higher scores indicating greater levels of psychological reactions to witnessing interparental conflict. Father-reported and mother-reported scores were further averaged for each subscale to create a composite score for children’s psychological reactions to marital discord. The internal consistencies, as indexed by alpha coefficients, for the emotional reactivity, involvement, and behavioral dysregulation subscales were .76, .86, and .74, respectively.

Parenting.

Mothers and fathers completed portions of the Parent Behavior Inventory, a parent-report questionnaire adapted from the Children’s Report of Parent Behavior Inventory (Schaefer, 1965). Three subscales were used, including intrusiveness (5 items; e.g., “You ask your child to tell you everything that happens when s/he is away from home”), control through guilt (5 items; e.g., “You tell your child that s/he would do what you want if s/he loved you”), and instilling anxiety (5 items; e.g., “You think that any misbehavior is serious and will have future consequences”) subscales. Both parents rated their own and their partner’s behaviors on 5-point Likert scales ranging from 1 (never) to 5 (always). Fathers’ reports of themselves and mothers’ reports of their partners’ scores were summed to reflect fathers’ parenting, while mothers’ reports of themselves and fathers’ reports of their partners’ scores were summed to reflect mothers’ parenting. The three subscales (i.e., intrusiveness, control through guilt, and instilling anxiety) were positively correlated with each other (r ranges from .515 to .747 for mothers and from .478 to .730 for fathers), and they were loaded on a higher-level factor psychological control (Schludermann & Schludermann, 1983), we averaged them to create a composite score of harsh parenting for fathers and mothers. The Cronbach alphas for the composite harsh parenting score were high (α =.87 for mothers and .88 for fathers).

Covariates.

Five covariates were included in our model: child gender, family income, child medication use, blood contamination in saliva, and child’s previous exposure to interparental conflict. Child gender was dichotomously coded (boys contrasted with girls) and family income was entered as a categorical variable (from 1 to 10, with higher numbers representing higher family income). Additionally, Salimetrics conducted immunoassays of transferrin on saliva samples to assess blood contamination. The minimum detection limit of this test, which used 20 µl of saliva per assay, was 0.12 mg/dl. The mean level of blood contamination in this study was .29 mg/dl (SD = 0.30 mg/dl; range = 0 to 2.08 mg/dl). Mothers also reported children’s over-the-counter and prescription medication use. Most children in the study (86%, 166 children) were not on any medication at the time of the study. The remaining 14 percent of children had taken at least one medication. The most frequently used medications were ibuprofen (4.7% of children in total sample), cough/cold medicine (3.7%), and vitamins (3.1%). Mothers also reported lower levels of antibiotics (1%), allergy treatment (1%), and other medicines (0.5%).

Child exposure to various forms of interparental hostility (e.g., physical abuse, quarrels, sarcasm) was assessed by the O’Leary–Porter Scale (OPS; Porter & O’Leary, 1980). On a 5-point Likert scale ranging from 0 (never) to 4 (very often), both fathers and mothers reported the frequency of child exposure to interparental hostility. The scale has been demonstrated to have good test–retest reliability, internal consistency, and concurrent validity (Porter & O’Leary, 1980). The internal consistency of the OPS in the present sample was high (α = .80 for mothers and .79 for fathers). Due to the high correspondence between mother and father reports on the OPS (γ = .61, p < .001), maternal and paternal reports were averaged to yield a composite indicator of children’s previous exposure to interparental conflict.

Data Analysis

After examining the bivariate correlations among all studied variables, we fit Latent Growth Curve (LGC) models to children’s three salivary samples across the pre-conflict, post-I, and post-II cortisol assessments. Children’s cortisol reactivity was indicated by slopes of the LGC models, rather than the predominantly-used difference scores between pre-stressor and post-stressor cortisol levels in analyses of variance or regression models (e.g., Powers, Pietromonaco, Gunlicks, & Sayer, 2006). Depicting children’s cortisol reactivity with LGC slopes allows researchers to separate measurement error from measures of reactivity and to model cortisol trajectories across three measurement occasions used in the study (see Hruschka, Kohrt, & Worthman, 2005). All models were conducted using Mplus Version 7 (Muthén & Muthén, 1998–2012) with robust maximum likelihood estimation. Model fit was evaluated with Satorra-Bentler chi-square (χ²), Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), and Standardized Root Mean Square Residual (SRMR). According to Hu and Bentler (1999)’s recommendation, nonsignificant χ², RMSEA < .06, CFI > .95, and SRMR < .08 indicate good model fit.

A series of LGC models, with covariates included, were conducted in sequence. First, a baseline model was estimated prior to the inclusion of children’s psychological responses, parents’ harsh parenting, and their interactions as predictors. Time of measurement during collection of the baseline sample was specified as a covariate of the three cortisol measures to account for the diurnal rhythm of cortisol. Second, we examined children’s different psychological responses (i.e., emotional reactivity, involvement, and behavioral dysregulation) as predictors of children’s cortisol functioning (both levels and slopes). Subsequently, we tested whether the significant association between children’s psychological–physiological responses was moderated by paternal or maternal negative parenting. Satorra-Bentler scaled χ² difference test was used to evaluate whether the proposed moderation models fit the data significantly better than the models with only the psychological responses and negative parenting (i.e., model without the interaction term). Lastly, given the potential moderating role of child gender in models of emotion regulation and cortisol reactivity, we used multiple-group analysis to examine whether the effect of parenting on the interrelatedness of children’s responses differed as a function of child gender.

Results

Table 1 shows the means, standard deviations, and correlations of the raw scores among the studied variables. The three cortisol measures were log transformed in order to reduce skewness and normalize the distributions prior to the primary analyses, yielding the following means and standard deviations: preconflict cortisol (M = − 2.33, SD = 0.51), post I conflict cortisol (M = − 2.69, SD = 0.47), and post II conflict cortisol (M = − 2.73, SD = 0.50) assessments.

Table 1.

Means, standard deviations, and correlations of the raw scores among the study variables

Variables	M	SD	1	2	3	4	5	6	7	8
Child Cortisol reactivity
1. Preconflict	.112	.066	–
2. Post I	.077	.044	.69***	–
3. Post II	.075	.047	.58***	.89**	–
Harsh Parenting
4. Mothers’ parenting	28.03	5.20	− .07	− .02	− .03	–
5. Fathers’ parenting	25.61	5.19	− .11	− .02	− .04	.79***	–
Child Psychological Reactivity
6. Emotional reactivity	10.14	2.58	− .13	− .08	− .03	.22**	.20**	–
7. Involvement	19.49	5.81	− .02	− .06	− .01	.23**	.22**	.43***	–
8. Behavioral dysregulation	6.85	2.05	− .08	− .03	− .01	.20**	.20**	.37***	.37***	–
Exposure to Interparental Conflict
9. OPS	10.29	4.53	− .01	− .02	.03	.25***	.15**	.23**	.21**	.34***

Note.

^***p < .001

^**p < .01

OPS, the O’Leary–Porter Scale

Preliminary Analyses

We fit an unconditional LGC model to examine child cortisol reactivity in response to the simulated interparental conflict. Given the diurnal rhythm of cortisol, time of measurement was included as a covariate in the model, in addition to other covariates (i.e., child gender and race, blood contamination in saliva, medication use, as well as the child’s previous exposure to interparental conflict). Time of measurement was significantly associated with preconflict (b = − .085, SE = .016, β = − .359, p < .001), post I conflict (b = −.120, SE = .014, β = − .536, p < .001), and post II conflict (b = − .125, SE = .014, β = − .540, p < .001) cortisol assessments in this unconditional LGC model. Weights for the manifest assessments of cortisol in the model were specified to correspond with the time elapsed since the pre-conflict cortisol measure. Specifically, the weight of 0.45 was assigned to the post I conflict cortisol assessment because it occurred approximately 45 minutes after the pre-conflict saliva sample, and a weight of 0.56 was assigned to the post II conflict cortisol assessment because it took place approximately 56 minutes after the pre-conflict sample.

The model fit the data well, Satorra-Bentler χ² (4, N = 193) = 6.596, p = .16, RMSEA = .060 (90% CI = [.000, .138]), CFI = .995, SRMR = .010. The mean level of the intercept characterized the estimate of the pre-conflict cortisol level, which was significantly different from zero (μ_i = − 1.30, z = − 4.21). The mean slope parameter, which estimated the average, constant cortisol change across the three assessments, was not significant (μ_s = .72, z = 1.41). Notably, both the intercept and slope factors had statistically significant variances of .18 (z = 4.65) and .26 (z = 1.98), respectively, indicating significant individual differences in children’s pre-conflict cortisol levels and in the rate of change in cortisol across three sampling occasions.

Primary Analyses

We proceeded to examine how different patterns of children’s psychological reactivity were associated with their cortisol reactivity in response to the interparental conflict. We included children’s distress, involvement, and hostility simultaneously as predictors of cortisol activity. Pathways were specified from each type of behavioral reactivity to the initial cortisol levels (intercept) and cortisol change (slope), see Figure 1. The model fit the data well, Satorra-Bentler χ²(27, N = 193) = 30.849, p = .79, RMSEA = .028 (90% CI = [.000, .067]), CFI = .993, SRMR = .042. As for the paths between the psychological responses and cortisol reactivity, children’s higher distressed emotional reactivity was associated with greater cortisol reactivity to interparental conflict (slope), b = .043, SE = .022, β = .219, p < .05. Neither involvement nor hostility, however, were associated with initial level or change in cortisol.

Figure 1.

Associations between children’s psychological reactivity and cortisol activity to interparental conflict.

Note. Covariates (i.e., child gender, family income, child medication use, blood contamination in saliva, and child’s previous exposure to interparental conflict) are included in the model but are not presented in this path diagram for the sake of simplicity. The model fits the data well, Satorra-Bentler χ²(27, N = 193) = 30.849, p = .79, RMSEA = .028 (90% CI = [.000, .067]), CFI = .993, SRMR = .042 ^* p < .05; ^*** p < .001

Parenting Behaviors as Moderators

Results in the primary analyses indicated that distress was significantly and uniquely related to higher levels of cortisol reactivity in response to interparental conflict. Guided by our conceptual interest in the role of maternal and paternal parenting behavior, we proceeded to examine whether negative parenting behaviors moderated the above association. Because a child’s response pattern toward interparental conflict may be shaped by both the father’s and mother’s parenting practices, paternal harsh parenting and maternal harsh parenting were included simultaneously in the model and the covariance between these constructs were estimated to account for their interdependency. As such, the full LGC model (Model 0) included distressed emotional reactivity, the proposed moderator (i.e., harsh parenting for both fathers and mothers), as well as the interaction terms (i.e., emotional reactivity × proposed moderator) as predictors of cortisol intercept and slope. In the next two models (Model 1mom and Model 1dad), we constrained the interaction term for fathers (i.e., emotional reactivity × paternal negative parenting) or mothers (i.e., emotional reactivity × maternal negative parenting) to be 0, to examine if children’s psychological–physiological reactivity link was moderated by mothers’ parenting or fathers’ parenting, respectively. Lastly, we fit a restricted model (Model 2) where we constrained both fathers’ and mothers’ interaction terms to be 0. Satorra-Bentler scaled χ² difference test (Satorra, 2000) was then conducted to compare nested models (i.e., Model 1mom/1dad against Model 2) in order to evaluate whether adding the interaction term would significantly improve model fit. For these models involving interaction terms, all variables were standardized across the sample prior to analyses.

Table 2 presents unstandardized coefficients of the interaction terms (emotional reactivity × parenting) predicting initial cortisol level (intercept) and cortisol change (slope), along with the fit indices of each model. All of the examined models had good fit, in accordance with the recommendation proposed by Hu and Bentler (1999), see Table 2. In Model 0, with all four interaction terms included, no interaction term was a significant predictor of either the initial level or slope of cortisol reactivity for fathers or mothers. In Model 1mom, the interaction term for mothers was significant in predicting children’s initial level of cortisol (b = − .084, SE = .034, β = − .202, p < .05), but was not significant predicting the slope of cortisol reactivity (b = .090, SE = .042, β = .173, p ＝ .07). Although the maternal parenting did not moderate children’s psychological–physiological reactivity link, Satorra-Bentler scaled χ² difference test between Model 1mom and Model 2 showed that including the interaction term of maternal harsh parenting and child distressed reactions did explain variance in the data significantly better, Satorra-Bentler scaled ∆χ²(∆df = 2) = 7.099, p < .05.

Table 2.

Important parameter estimates and fit indices all the LGC models examined

	Model 0			Model 1mom			Model 1dad			Model 2
Effects	b	SE	β	b	SE	β	b	SE	β
Mothers
Distress × Parenting → Level	− .003	.068	− .008	− .084*	.034	− .202	–	–	–	–
Distress × Parenting → Slope	− .076	.102	− .147	.090	.042	.173	–	–	–	–
Fathers
Distress × Parenting → Level	− .082	.058	− .225	–	–	–	− .084**	.029	− .232**	–
Distress × Parenting → Slope	.167	.103	.371	–	–	–	.109*	.047	.242*	–
Fit Indices
S-B χ2		48.028			50.787			48.457		58.113
df		39			41			41		43
c		.806			.835			.816		.837
RMSEA		.036			.036			.032		.044
CFI		.983			.982			.986		.972
SRMR		.048			.048			.048		.055

Note.

^**p < .01

^*p < .05;

“Distress × Parenting →Level” indicates the interaction term of children’s distressed reaction and harsh parenting predicting children’s initial cortisol level; “Distress × Parenting →Slope” indicates the interaction term of children’s distressed reaction and harsh parenting predicting children’s cortisol reactivity in response to simulated interparental conflict. S-B χ² = Satorra-Bentler χ²; c = weighting constant for computing the χ² statistic using robust estimation method; RMSEA = root mean square error of approximation; CFI= comparative fit index; SRMR = standardized root mean square residual.

As for the moderating role of fathers’ harsh parenting, in Model 1dad, the interaction term for fathers was significant in predicting both children’s initial level (b = − .084, SE = .029, β = − .232, p < .01) and the slope of cortisol reactivity (b = .109, SE = .047, β = .242, p < .05). Comparing Model 1dad and Model 2, the Satorra-Bentler scaled χ² difference test showed that the moderation model with paternal harsh parenting as the moderator (Model 1dad) fit the data significantly better than the model without interaction terms (Model 2), Satorra-Bentler scaled ∆χ²(∆df = 2) = 7.240, p < .05. Hence the interaction added significant variance above and beyond the reactivity and parenting measure.

To clarify the significant interaction of child distress and fathers’ harsh parenting in LGC Model 1dad, we plotted regression slopes of child distress on cortisol reactivity (slope) at high (1 SD above the mean) and low (1 SD below the mean) levels of fathers’ harsh parenting (see Figure 2), following Preacher, Curran, and Bauer’s (2006) procedure. Subsequent simple slope analysis showed that for children whose fathers had high levels of negative parenting practices, greater emotional reactivity predicted increased cortisol reactivity (b = .183, t = 2.588, p < .05). When paternal harsh parenting was low, however, the association between emotional reactivity and cortisol reactivity was not significant (b = − .027, t = − 0.3382, p = .703).

Figure 2.

Plots of the significant interactions between children’s emotional reactivity to interparental conflict and fathers’ harsh parenting predicting children’s cortisol reactivity to interparental conflict.

Alternative Models.

Children’s psychological reactivity and parenting were composite scores reported by both parents, and thus it is necessary to rule out the possibility of the reporter confound by independently examining mothers’ reports of distressed reactivity and fathers’ reports of parenting (and vice versa). More analyses were conducted in which mother-report of child distress and father’s self-report of parenting were used (as well as father-report of child distress and mother’s self-report of parenting). Results from these alternative models indicated that using different reporters did not alter the direction or significance of paths from those in previous models, and thus only the results from the models in which reports from both parents were used have been presented for this study.

Additionally, given the potential moderating role of child gender in the cortisol reactivity and emotion regulation literature, the proposed influence of harsh parenting on children’s emotional–cortisol reactivity interrelatedness examined in the present study may differ as a function of child gender. Although child gender has been included in previous LGC models to control for its influence on cortisol reactivity, it is also necessary to examine whether child gender is a potential moderator of the model. To accomplish this, we excluded child gender as a covariate and instead used it as a grouping variable in multiple-group analysis to simultaneously examine the models for boys and girls. Specifically, we constrained the influences (including the interaction effect) of child emotional distress and parenting on child cortisol reactivity to be the same for boys and girls, and compared these models against their corresponding unconstrained model in which the predictor pathways were allowed to vary across groups. Model comparisons revealed no differences in fit for each pair of the models. Thus, in our sample, child gender did not moderate the model influences.

Discussion

Witnessing interparental conflict has been shown to elicit children’s psychological responses, such as exhibiting distress (Davies & Cummings, 1994), as well as physiological responses, such as changed cortisol levels (Davies et al., 2007; Sturge-Apple et al., 2012). Different levels of responses to interparental conflict, however, have typically been examined independently. Little is known about the interconnection between children’s psychological and physiological responses to interparental conflict. Accordingly, the current study aimed to advance our understanding of children’s responses to interparental conflict by first simultaneously examining the association between psychological responses and children’s cortisol reactivity and then by further studying how such an association manifests under negative parenting.

Consistent with our hypothesis, children’s psychological distress was uniquely and significantly associated with elevated cortisol reactivity to interparental conflict, whereas children’s involvement in conflict, as well as their hostile responses to conflict were not significant predictors. This result supported a distress specificity model over a general arousal model. In contrast to the general arousal model, which would indicate that all three types of psychological activation assessed in the study would be associated with elevated cortisol reactivity (Gunnar & Vazquez, 2006), only children’s distress reactivity, relative to the other types of psychological responding (i.e., involvement and hostility), was associated with cortisol reactivity. At least in the context of interparental discord, children’s distress seemed to be a particularly salient indicator of psychological responses that were coupled with greater HPA activity.

Similar results supporting distress specificity models were also found in past studies (Davies et al., 2008; Susman, Dorn, Inoff-Germain, Nottelmann, & Chrousos, 1997). One possibility is that heightened cortisol reactivity is most likely to occur when children exhibit submissive behavioral reactions, such as emotional distress and anxiety, under stressful events (Adam, 2006; Kagan, Reznick, & Snidman, 1988). This interpretation is also consistent with EST’s proposition that elevated emotional distress is a hallmark symptom of children’s concerns about security (Cummings & Davies, 1996; Cummings et al., 2006), such that children’s distress (characterized by anxiety, vigilance, and emotional upset) is distinctively underpinned by their adrenocortical reactivity.

Our study also found that the confluence of children’s psychological distress and cortisol reactivity was affected by negative parenting. Specifically, heightened emotional distress was only significantly associated with elevated cortisol reactivity for children whose father tended to exhibit high levels of harsh parenting, whereas the coupling between children’s emotional distress and cortisol reactivity was not significant at low levels of negative parenting. Two potential reasons may account for the moderating role of harsh caregiving environment. First, it is consistent with theoretical contentions that harsh parenting is a salient environmental risk to children’s perceptions of personal safety and wellbeing (Cicchetti & Rogosch, 2001; Margolin, 2005), such that a negative caregiving environment may amplify the interrelatedness of children’s distressed reactions and cortisol reactivity. In the present study, the chronic threat posed by negative parenting, in addition to the acute stressor of experiencing parents’ conflict, might activate a heightened threat detection system. As a result, children’s cortisol reactivity may be particularly related to their distressed emotional responses in order to maximize their ability to cope with a risky family environment.

Additionally, the amplifying role of harsh parenting in the link between children’s emotional and cortisol reactivity may be related to its disruption of effortful control. Both emotional distress and cortisol reactivity have been linked to effortful control (Davies & Martin, 2013; Taylor et al., 2013), which is an important component in children’s self-regulating system. Healthy development of effortful control may be disrupted by high levels of psychological control (e.g., intrusiveness and control through guilt), because the internal, self-driving regulating processes are intruded by external forces that are exerted by parents. Past research has indicated that mothers’ unsupportive (i.e., minimizing or punitive) responses to children’s negative emotions are linked to low levels of effortful control in early childhood (Eisenberg, Fabes, & Murphy, 1996; Spinrad et al., 2007) and that maternal over-control and intrusiveness predict toddlers’ effortful control (Graziano, Keane, & Calkins, 2010). Particularly, Calkins (2007) argued that children’s regulated physiological reactivity might be the foundation for later regulatory competencies, emotional and behavioral regulation abilities included, such that increased coupling of physiological and psychological reactivity might indicate children’s self-regulation difficulties and increased risk for developing adjustment problems. Therefore, the role of negative, controlling parenting in altering the association between children’s emotional and cortisol reactivity could be attributed to its influence on children’s self-regulation.

Interestingly in our study, the confluence of children’s heightened distress and cortisol reactivity was only amplified in the context of fathers’, but not mothers’, harsh and controlling caregiving practices. This indicates that negative fathering may play a unique role in affecting children’s psychobiological functioning. Although this topic has received limited empirical inquiry, one study with samples of younger children is worth noting for its interest in examining the unique role of fathering. Specifically, Mills-Koonce et al (2011) found that high levels of father negativity were related to greater increases in child cortisol levels in response to emotion challenge for infants and with higher general levels of cortisol for toddlers. These associations were present above and beyond any associations with mothers’ parenting.

The more salient role of paternal negativity, compared to maternal negativity, in affecting children’s response patterns toward interparental conflict warrants further discussion. Given that fathers tend to take on less caregiving responsibility and interact less frequently with children than mothers (Lamb & Lewis, 2010), small variations in paternal negative parenting may have more significant implications for children’s psychobiological functioning. In other words, children’s responses to the stress related to interparental conflict may be particularly sensitive to the amount of negativity exhibited in fathers’ parenting. Another possible explanation for this finding is the differences between father-child and mother-child interactions: fathers tend to engage their children in more physical, stimulating, and arousing activities than mothers, forming a so-called “activation relationship” (Paquette, 2004). The nature of this relationship may contribute to children’s development of conflict-coping and stress-regulating abilities. These speculations need to be tested in future research to achieve a better understanding of the unique roles of fathers’ and mothers’ parenting.

As discussed above, the aim of the current study was to understand the unique effects of maternal and paternal parenting in children’s response patterns toward interparental conflict. Accordingly, maternal and paternal parenting were tested separately in Model 1mom and Model 1dad. Child emotional and cortisol reactivity were linked in the context of high levels of paternal, rather than maternal, harsh parenting, indicating that paternal parenting was a significant moderator of the emotional–cortisol reactivity link. However, neither paternal nor maternal parenting was a significant moderator when the other one was controlled (Model 0). We suggest that the moderate shared variance between paternal and maternal parenting (r = .789) in Model 0 resulted in a high standard error, limiting the power to detect effects. Substantively, the results may also indicate that it may be the overlap between maternal and paternal parenting that modulates the risk conferred by emotional distress to child cortisol reactivity.

Although the focus of the present study was children’s cortisol reactivity, indicated by the slopes of LGC models, we also found significant interaction effects between children’s distress and parenting on children’s initial cortisol levels, indicated by the LGC model intercepts. Examination of the interaction effects showed that children’s increased distress to interparental conflict was associated with lower preconflict levels of cortisol, but only for children experiencing greater parental negativity. Thus, among children who experienced high levels of negative parenting, their elevated distress responses were associated with blunted preconflict cortisol but elevated cortisol reactivity. Taken together, our findings indicate the complex interplay between initial cortisol levels and cortisol reactivity in stressful situations (e.g., Sturge-Apple et al., 2012), also highlighting the value of differentiating the two aspects of cortisol activity (Hostinar & Gunnar, 2013).

Questions remain about whether the significant association between distress responses and cortisol reactivity under a harsh caregiving context indicates adaptive or mal-adaptive functioning for children. Little evidence is available in existing studies to support either contention. On the one hand, some theories indicate that there may be potential problems with coupled high reactivity of response systems. For example, distress responding to interparental conflict is a hallmark indicator of child emotional insecurity (Cummings & Davies, 1996), and elevated cortisol reactivity signifies an increased engagement of children’s HPA system to cope with environmental stress, such as interparental conflict (Gunnar & Vazquez, 2006; Saltzman, Holden, & Holahan, 2005). Heightened distress and increased cortisol responses, therefore, may alter the set points of children’s stress-response system by engendering elevated anticipatory vigilance in children, eventuating in later adjustment problems. Particularly, independent lines of research have suggested that children’s distressed reactions (Crockenberg & Langrock, 2001; Cummings, Cheung, & Davies, 2013; Davies & Cummings, 1998) and increased cortisol reactivity (Davies et al., 2007; Dougan, Hastings, Granger, Usher, & Zahn-Waxler, 2001; Gunnar & Vazquez, 2006) toward quarrelling parents were linked with child internalizing problems. Additionally, the dual-process perspective of emotional responses (Evers et al., 2014) argued that the disconnection between the expression of distress (part of the reflective system) and high cortisol reactivity (part of the automatic system) may be functional: overriding the automatic activity with more conscious and reflective reactions contributes to socially appropriate response.

On the other hand, the confluence of distress and cortisol reactivity can be regarded as synchrony between the psychological and physiological systems, which may be more adaptive than an inconsistent responding pattern between the two systems. Such speculation is consistent with the organizational perspective of development: healthy development involves flexible deployment of behaviors, emotions, and physiology in response to context (Cicchetti & Rogosch, 1996; Cicchetti & Tucker, 1994). Coordinated regulation of different systems (e.g., emotional, behavioral, cognitive, and physiological), rather than isolated responses in a single system, may be indicative of better integrated developmental systems. Providing indirect evidence for the adaptive implications of the significant confluence of psychological and physiological systems, researchers have found that youth’s subjective distress was related to cortisol increases, but only for those in sensitive caregiving environments (Blair et al., 2015) or who reported low levels of internalizing symptoms (Spies et al., 2011).

Limitations of the current study merit consideration when interpreting and generalizing these results. First, children’s psychological reactivity was assessed by a questionnaire, rather than through real-time observations while children were watching conflict videos, and thus may only indicate children’s general psychological response. Future studies should attempt to simultaneously measure children’s in-situ psychological and physiological responses to similar interparental conflict stressors. Behavioral coding used to measure children’s observed psychological responses, in conjunction with cortisol assessments as their responses to stressful events, would also increase the rigor in assessing the link between children’s stress-responding systems.

Second, our study did not include positive parenting behaviors, such as involvement and sensitivity. Parenting that is engaged, nurturing, and responsive may buffer against abnormal HPA axis functioning in children, possibly by providing them easier access to attachment figures and stronger signals of safety (Gunnar & Hostinar, 2015; Hostinar & Gunnar, 2013). Findings from a recent study with a sample of young children suggested a positive role of parental sensitivity in children’s stress response processes (Blair et al., 2015). Given the importance of both positive and negative parenting, future studies should examine both aspects of parenting as developmental contexts for children’s physiological responses to salient familial stresses, such as interparental conflict.

Third, only using three cortisol samples for each child in our study prevented us from examining a more complex change trajectory of cortisol reactivity, other than those involved in linear growth curves. Curvilinear trajectories may present over the course of a stressful event, and the shape of cortisol reactivity may also change from the responding phase to recovery phase. Linear curves used in our study were not able to delineate individual differences in the recovery of cortisol levels to baseline levels; in fact, a significant linear trend of cortisol reactivity over the course of several time points can also be interpreted as a failure to recover, in addition to high reactivity. Future studies with more assessments are needed to better delineate the trajectory of cortisol reactivity and recovery.

Fourth, the cross-sectional nature of our study prevented us from addressing the temporal ordering of relationships in our model. Cortisol reactivity was tested as the outcome, yet it is also reasonable to designate it as a predictor: children who are more physiologically reactive tend to be perceived by their parents as more distressed when exposed to marital conflict. However, theoretical and methodological considerations supported our design and interpretation. As the main theoretical interest in our study, the concurrent correspondence between children’s psychological and physiological reactivity could be readily addressed by the cross-sectional design used without differentiating the predictor and outcome. Besides, children’s heightened psychological reactivity (e.g., increased emotionality, vigilance, and threat) engendered by exposure to family discord has been theorized and empirically supported to alter the HPA axis set point (Davies et al., 2009; Luecken, Appelhans, Kraft, & Brown, 2006). Anticipatory psychological reactivity may sensitize children’s cortisol responses, such that children are better prepared to mobilize and modulate resources to efficiently cope with the stress associated with stressful events (Davies & Cummings, 2006; Lopez, Vazquez, & Olson, 2004). Furthermore, the methodological design of our study also supported our examination of children’s psychological reactivity as a predictor of their cortisol reactivity, instead of the other way around. While children’s psychological reactions to interparental conflict were assessed with parent-report questionnaires based on parents’ perceptions of the child during the past year, their cortisol reactivity was measured over the course of watching simulated interparental conflicts in the laboratory setting. In other words, our assessment battery was designed to measure children’s psychological reactivity within a time window that preceded the assessment of physiological functioning.

Despite its limitations, the present study opens important directions for future research to systematically understand children’s responses to interparental conflict. First, it calls for future studies to identify children’s different responding profiles and investigate the developmental sequelae (e.g., psychopathological symptoms, socioemotional outcomes) of each responding profile. Such an endeavor can help promote researchers’ understanding of the developmental function of children’s multilevel responses. In a recent study, Koss, Cummings, Davies, and Cicchetti (2017) identified four groups of adolescents with distinct responding profiles: adolescents within the physio-reactive group, who displayed increased cortisol responses despite their effective regulation of other types of responses (either affective or behavioral), exhibited pronounced growth in mental health problems over three years. Their findings indicate that activation of the HPA axis in response to conflict, by itself, can be a precursor to later problems. Other types of responding profiles, on the other hand, may not have posed as great a risk as the sole activation of the HPA axis.

Secondly, the present study also encourages future research to investigate the developmental trajectory of the association between children’s emotional and physiological responses. The links between the reactivity of different systems may change as children develop. Questions remain largely unexplored about when children’s stress-response systems become linked and begin to affect children’s adjustment outcomes, yet they may hold significant implications for intervention or prevention. Moreover, longitudinal designs are also beneficial for understanding the transactional influences or temporal orders among psychological and physiological reactivity, which cross-sectional designs are unable to detect.