|Home | About | Journals | Submit | Contact Us | Français|
Towards advancing conceptualizations of the spillover hypothesis, this study examined the conditions and mechanisms underlying the transmission of distress from the interparental relationship to parenting difficulties over a two year period in a sample of 233 mothers (M = 35.0 years) and fathers (M = 36.8 years) of kindergarten children. Findings from autoregressive structural equation models indicated that parent gender moderated associations between interparental conflict and parental psychological control and insensitivity to child negative affect. Pathways between interparental conflict and parenting difficulties over the two year period were significant for fathers but not mothers. Analysis of insecurity and depressive symptoms as affective mechanisms of spillover revealed that adult relationship insecurity was a significant mediator in the pathways between interparental conflict and parenting difficulties experienced by fathers.
In serving as the relationship hub of the family system, the quality of the interparental relationship has significant implications for how adults approach the developmental challenge of raising children. For example, novel use of experimental designs have shown that mothers randomly assigned to participate in a conflictual interaction with their partners were less attentive to their sons in an ensuing parent-child interaction task than mothers who participated in a non-conflictual interaction (Jouriles & Farris, 1992). Longitudinal research has shown that observations of destructive interparental conflict predicted subsequent decreases in parent emotional availability one year later (Sturge-Apple, Davies, & Cummings, 2006). Lending further support to the impact of marital conflict on parenting, parent participation in interventions that are specifically designed to improve the marriage has been shown to promote more effective parenting practices. In contrast, participation in another program designed to improve parenting did not yield any enhancement in interparental relationship quality (Cowan & Cowan, 2002). Taken together, the various findings provide strong support for the thesis that interparental conflict progressively undermines child-rearing practices (Cowan, Cowan, Pruett, & Pruett, 2007; Mikulincer & Goodman, 2006). However, the question remains as to why interparental conflict compromises parenting. Toward the goal of addressing this gap, the objective of this study is to examine interparental relationship security as an explanatory mechanism in the link between interparental conflict and child-rearing practices.
As the prevailing model guiding research on the interplay between interparental conflict and parenting processes, the spillover hypothesis proposes that “the emotions, affect, and mood generated in the marital realm transfers to the parent-child relationship (Krishnakumar & Buehler, 2000, p. 26).” Research documenting concurrent and longitudinal associations between marital and parent-child difficulties is commonly cited as supporting the tenets of the spillover hypothesis (e.g., Engfer, 1988; Erel & Burman, 1995; Gerard, Krishnakumar, & Buehler, 2006; Krishnakumar & Buehler, 2000; Lindsey, MacKinnon-Lewis, Campbell, Frabutt, & Lamb, 2002; Sturge-Apple, Davies, & Cummings, 2006). However, these studies have not tested the core tenet that affective mechanisms resulting from marital conflict actually predict subsequent impairments in parenting. Moreover, this general assumption provides little guidance in identifying exactly how and why anger and distress in the interparental relationship is transmitted to the parent-child subsystem. As a first step toward advancing a process-oriented model of spillover, we examine the relative viability of two potential affective processes as explanatory mechanisms in pathways between interparental conflict and parenting processes over the span of two years: adult negative affect and adult insecure representations of their intimate relationship.
The central role of mood difficulties as a mediator of associations between interparental conflict and parenting difficulties is evident in many conventional formulations of the spillover hypothesis. For example, the conceptualization of spillover by Krishnakumar and Buehler (2000) underscores the role of “emotions, affect, and mood (p. 26)” as operative mechanisms. Similarly, Erel and Burman (1995) define spillover as the process in which a “conflictual marital relationship may cause parents to be irritable and emotionally drained and therefore less attentive and sensitive to their children (p. 109)” (also see Easterbrooks & Emde, 1988). Drawing from these conceptualizations, a plausible thesis is that protracted experiences with conflict in the interparental relationship may increase the likelihood of subsequent parenting problems by progressively amplifying parental depressive symptoms characterized by negative mood, anhedonia, lethargy, and helplessness. Negative affect may further undermine parenting through the amplification of negative appraisals and attributions of child behavior, constriction of attention resources devoted to child-rearing activities, disruption of problem-solving abilities and facilitation of reflexive, dysregulated response patterns (Dix, 1991)).
By the same token, it is questionable whether relatively short-term feelings or displays of negative affect are primary processes underlying the occurrence of spillover over relatively long developmental lags of months or years. Although many seminal accounts of the spillover hypothesis conceptualized negative affect as a temporary, dynamic state stemming from conflict between parents (e.g., Almeida, Wethington, & Chandler, 1999; Christensen & Margolin, 1988; Jouriles & Farris, 1992; Mahoney, Boggio, & Jouriles, 1996), spillover from the interparental to parent-child relationship in this sense may only be particularly pronounced within narrow developmental windows of minutes, hours, or days. Alternatively, the notion of “spillover” may be viewed as a short-hand or even valuable metaphor for the active processes, but it remains that the active processes must be identified. That is, it is unlikely that relatively brief, transient, changes in mood or emotion can account for lasting changes in parenting. Thus, to cogently affect parenting over time, the dynamic emotional underpinnings of spillover must transform into stable individual differences, such as negative affect (e.g., depressive symptoms), resulting in cumulative decrements in parenting over periods of months or years. These questions have yet to be examined at an empirical level. For example, consistent with the proposed link in the chain between marital conflict and lasting changes in parental negative affect, prior research has shown that marital distress increases depressive symptoms (Beach, Katz, Kim, & Brody, 2003; Fincham, Beach, Harold, & Osborne, 1997; Whisman, 2001). Supporting another link in this pathway, depressive symptoms are associated with decrements in domains of parenting ranging from diminished parental warmth and sensitivity to difficulties in disciplining and managing children (Cummings & Davies, 1994; Lovejoy, Graczyk, O’Hare, & Neuman, 2000).
Scepticism about whether relatively transient changes, such as increases in negative affect, can serve as a primary underpinning of spillover encourages identifying how other dynamic properties of the emotional aftermath of interparental conflict might become organized into stable psychological structures that may progressively erode effective parenting patterns over time. From an emotional security perspective, adult representations of the security of the interparental relationship may serve as another mechanism linking interparental conflict and parenting difficulties. Life span formulations of attachment theory posit that conflict between adult partners is intimately linked with the ability of the partners to use each other as sources of support in the family, particularly when they occur during stressful or conflictual times (Cowan, Cohn, Cowan, & Pearson, 1996; Crowell et al., 2002; Mikulincer, Florian, Cowan, & Cowan, 2002; Treboux, Crowell, & Waters, 2004). Expressions of hostility and disengagement during conflicts may specifically serve as signals that the partner in the relationship cannot reliably serve as a secure base and, in some relationships, may even pose a threat to the welfare of the adult. Moreover, as relatively stable relational processes, underlying adult worries about the security of their relationship with their spouse may progressively undermine abilities to serve as an effective and responsive caregiver. In support of this hypothesis, adult endorsement of insecurity in romantic relationships during pregnancy was associated with less desire to have children and greater parenting stress when the children were 6 months of age. Moreover, underscoring the potential robustness of the adult attachment as a predictor, these associations remained even after taking into account potential third variables, including the participants’ and their partners’ depressive symptoms and marital dissatisfaction (Rholes, Simpson, & Friedman, 2006). In light of this support for attachment as a carrier of spillover, the goal of this study was to examine the relative roles of adult attachment insecurity and depressive symptoms as intermediary mechanisms in pathways between interparental conflict and parenting difficulties.
To maximize the power to identify spillover mechanisms, we focused on examining two affective dimensions of parenting during the early elementary school years. Focus on the developmental period of middle childhood is supported by the results of a quantitative review of the literature indicating that strength of associations between interparental conflict and parenting difficulties are accentuated during middle childhood (Krishnakumar & Buehler, 2000). In addition, if adult attachment insecurity and depressive symptoms serve as the operative underpinnings of the spillover process, then it follows that mediational pathways between interparental conflict and parenting should be more consistent when the child-rearing domain is shaped explicitly by emotional relationship processes in the family. More specifically, marital conflict would be most likely to impact emotion-laden domains of parenting by virtue of the difficulties regulating emotion and utilizing the partner as a base of support. Lending evidence to this contention, a recent review indicated associations between interparental conflict and parenting, across the body of literature, were significantly stronger for child-rearing domains reflecting low levels of acceptance and harsh punishment than lax control (Krishnakumar & Buehler, 2000). Accordingly, this study focused on parental insensitivity to child expressions of negative emotion and attempts to control and discipline the child in psychologically controlling ways characterized by guilt induction, instilling anxiety, and intrusiveness.
However, although a fundamental assumption of the spillover hypothesis is that interparental discord plays an etiological role in organizing the affective mechanisms that undermine parenting, an alternative hypothesis is that the proposed mediational paths involving interparental conflict are spurious artifacts resulting from their non-causal association with the more active, transactional roles of adult relationship insecurity and depressive symptoms. Adult depressive symptoms may play a primary etiological role in promoting attachment insecurity in intimate relationships by fueling negative interpersonal appraisals and evoking negative responses from partners (Coyne, 1976; Sacco & Vaughan, 2006). Likewise, attachment theory posits that insecurity in adult relationships may engender elevated depressive symptoms by undermining emotion regulation abilities and reinforcing negative working models of the self and close relationships (Bowlby, 1988; Carnelley, Pietromonaco, & Jaffe, 1994; Mikulincer & Florian, 1998; Roberts, Gotlib, & Kassel, 1996). Supporting these paths, several studies suggest relations between insecure attachment in adult relations and elevated depressive symptoms (Feeney, Alexander, Noller, & Hohaus, 2003; Whiffen, Kallos-Lilly, & MacDonald, 2001).
Thus, as a significant challenge to the spillover hypothesis, this body of work raises the possibility that adult depressive symptoms and attachment insecurity are such strong mediators of one another in models of parenting difficulties that they supersede any role of interparental conflict as a factor in their prediction. That is, adult depressive symptoms and relationship insecurity may serve as mediators of both interparental conflict and one another in predicting interparental conflict. Therefore, this study tests the tenets of the spillover hypothesis and possibly viable alternative models.
Conceptual models and literature reviews have also proposed that differences in the magnitude of the spillover process across families differ for mothers and fathers. The father vulnerability hypothesis specifically postulates that paternal parenting practices are more susceptible to deteriorating in the face of interparental conflict than maternal parenting (Belsky, Gilstrap, & Rovine, 1984; Cummings, Goeke-Morey, & Raymond, 2004; Cummings & O’Reilly, 1997). In support of this prediction, some studies have reported that associations between interparental conflict and parenting difficulties are stronger or more consistent for fathers than they are for mothers (see Davies & Lindsay, 2001; Krishnakumar, & Buehler, 2000). However, the role of parent gender in the spillover process has yet to be definitively delineated. For example, the meta-analysis conducted by Erel and Burman (1995) revealed that these gender effects were inconsistent across groups of studies that differed in methodological characteristics. Therefore, as a step toward resolving this inconsistency, we examine whether the mediational pathways between interparental conflict, adult relationship insecurity and depressive symptoms, and parenting difficulties are moderated by gender. Further complicating the picture, it is also plausible that spillover may vary as a function of child gender or the gender composition of the parent-child dyad (e.g., Davies & Lindsay, 2001; Erel & Burman, 1995; Krishnakumar & Buehler, 2000). For example, there is some, albeit inconsistent, evidence to suggest that spillover processes may be particularly pronounced for daughters (e.g., Erel & Burman, 1995; Krishnakumar & Buehler, 2000) and for father-daughter dyads (e.g., Davies & Lindsay, 2001; Erel & Burman, 1995). To address these possibilities, we also examine the alternative predictions that child gender and the gender composition of the parent-child dyad moderate spillover processes.
In summary, our goal is to identify the affective mechanisms underlying emotional spillover by examining adult relationship insecurity and depressive symptoms as mediators of associations between interparental conflict and subsequent changes in parenting over a two year period. Inclusion of three measurement occasions each spaced one year apart generated rigorous tests of our mediator predictions by temporally disaggregating the putative risk factor, mediator, and outcome (Cole & Maxwell, 2003). Therefore, temporally ordered prospective analyses were possible using Wave 1 measures of interparental conflict, Wave 2 assessments of adult relationship insecurity and depressive symptoms, and Wave 3 assessments of parenting difficulties. Autoregressive paths controlling for prior values of the affective mediators and parenting difficulties were also specified to more definitively identify directionality of relations among the constructs (Gollob & Reichardt, 1987).
Participants were drawn from a larger project that originally included 235 parents and children recruited through local school districts and community centers in a moderate-sized metropolitan area in the Northeast and a small city in the Midwest. Interested families were included in the project if they met the following eligibility criteria: (a) the primary caregivers had a child in kindergarten; (b) the kindergarten child and two primary caregivers lived together for at least the preceding 3 years; and (c) the primary caregivers and child were fluent in English. Across the three annual measurement occasions, twenty families who initially participated in Wave 1 did not participate by Wave 3 (8.5%). Comparisons of families who did and did not participate in all three waves (i.e., n = 215 and 20, respectively) along the 22 primary variables at Wave 1 (i.e., interparental conflict, adult insecurity and depressive symptoms, parenting) yielded one significant difference: mothers who did not participate in all three waves reported higher levels of parental intrusiveness (M = 21.55, SD = 2.61) than mothers who participated in all three waves (M = 20.03, SD = 3.07), t (231) = 2.41, p < .05. Because differences between the two groups of families were lower than would be expected by chance (4.5%), we estimated missing data for all 235 families in the larger project using an EM algorithm. Two families were excluded from analyses because they were identified as multivariate outliers. Accordingly, the final sample for this paper consisted of 233 mothers (M age in years = 35.0, SD = 5.61; Range: 22 – 52) and fathers (M age in years = 36.8, SD = 6.17; Range: 22 – 52) of kindergarten children (127 girls and 102 boys; M age in years at Wave 1 = 6.0 years, SD = .48; Range 5 – 8).
Median family income for the 233 families was in the range of $40,000 to $54,000, with 13% of the sample reporting household income below $23,000. The average number of years of education completed by mothers and fathers were 14.48 (SD = 2.34) and 14.61 (SD = 2.70), respectively. A large proportion of the parents reported being White (77%), followed by smaller percentages of Black (17%), Hispanic (4%), Asian American (1%), mixed (1%), and other races (1%). Parents were generally married (89.3%) and lived together an average of 11 years prior to Wave 1 (SD = 4.88). Most caregivers reported being the biological parents of their children (91.0%), followed by smaller numbers of stepparents (3%), adoptive parents (3%), and other types of guardianship (3%). In accordance with the goal of obtaining diversity in the experiences with interparental discord, 33% of mothers and 43% of fathers in our sample could be classified as martially dissatisfied based on scores below 100 on the Short Marital Adjustment Test at Wave 1, with 53% of the couples containing at least one maritally dissatisfied partner (e.g., SMAT; Locke & Wallace, 1959).
Families visited the laboratories at the research sites two times spaced one week apart at each of three waves spaced one year apart. Visits lasted about three hours. At Wave 1, mothers and fathers participated in an interparental interaction task. Following similar procedures used in previous research (DuRocher-Schudlich, Papp, & Cummings, 2004), spouses independently selected three topics that they perceived as problematic in their marriage. After selecting one topic they both felt comfortable discussing from each of their lists, the couples discussed each topic for ten minutes while they were alone in the laboratory room. Videotaped records of the session were subsequently coded for interparental conflict behavior. To examine the validity of the assumption that the interparental interaction task assessed characteristic ways of managing conflict in the interparental relationship, mothers and fathers individually responded to the question, “Overall, how much did the discussion resemble disagreements that usually occur between you and your partner at home” immediately following the interaction task. Response alternatives ranged from: (1) a lot more negative to (7) a lot more positive. Supporting the comparability of the interactions during the laboratory visit to conflicts that occur in the home, the means of mother and father responses fell between “about the same” and “a little more positive” on the seven-point scale (M = 4.71, SD = .91 and M = 4.76, SD = 1.11, respectively). Across all three waves, mothers and fathers also independently completed questionnaires to assess interparental conflict, adult relationship security and depressive symptoms, and parenting difficulties. The study was approved by the Institutional Review Board at each research site.
We utilized observations derived from the interparental interaction task and parent reports of interparental conflict in the home to obtain a multi-method latent construct of interparental conflict at Wave 1. For the observational component of the measurement battery, trained coders provided behavioral ratings using three codes from the System for Coding Interactions and Dyads (SCID; Malik & Lindahl, 2004). The SCID yields molar ratings of discord for each of the two marital interactions, ranging from 1 (very low) to 5 (high). Coders separately rated maternal and paternal Negativity and Conflict, which reflects the extent to which each individual in the dyad displays anger, frustration, and tension. At a dyadic level of analysis, coders provided ratings of couple Cohesiveness. Low Cohesiveness scores indicate overall interpersonal disengagement, aloofness, or indifference in the dyad as a unit. Therefore, for the subsequent analyses, Cohesiveness was reverse scored and labeled as Disengagement so that higher scores indicate greater withdrawal. For each conflict code, ratings were averaged across the two interparental interactions to obtain more comprehensive, parsimonious measures. Ratings from the primary coder, who rated all the marital interactions, were used in the analyses. To assess interrater reliability, another coder independently rated 25% of the tapes. The resulting intraclass correlation coefficients for the three codes of interparental conflict ranged from .95 to .96.
For the parent report component of the measurement battery, mothers and fathers each completed the 8-item Verbal Aggression scale from the Conflict and Problem Solving Scales (CPS; Kerig, 1996). Parents rate each of the items (e.g., “name-calling, cursing, insulting”) along three point scales ranging from (0) “never” to (3) “often” in relation to both their own and their partner’s use of verbally aggressive behaviors during interparental conflicts. Internal consistency, test-retest reliability, and various forms of validity for this measure are well-established (Kerig, 1996; 1998). Mothers and fathers composite ratings of self and partner’s verbally aggressive conflict behaviors were significantly correlated with one another (r = .49). Accordingly, the scales were summed to form a single composite of parent report on overall use of Verbal Aggression in the interparental dyad. Internal consistency of parent report of verbal aggression in this sample, as indexed by Cronbach’s α was .92.
Mothers and fathers reported on their depressive symptoms at Waves 1 and 2 using the Center for Epidemiological Studies – Depression Scale (CES-D; Radloff, 1977). The CES-D is a 20-item self-report measure designed to assess depressive symptoms, particularily negative affect (e.g., “bothered by things,” “could not shake off the blues,” “sad,” “depressed,” “fearful”), in community samples. For each item, mothers and fathers used a 4-point response format ranging from 0 (rarely or none of the time) to 3 (most of the time) to indicate the way they felt or behaved during the past week. The CES-D has well-established psychometric properties including internal consistency, test-retest reliability, and convergent validity with clinical and self-reports of depression (Radloff, 1977; Weissman, Sholomskas, Pottenger, Prussoff, & Lock, 1977). Internal consistency coefficients for the CES-D in this sample were satisfactory for mothers (αs = .87 at Wave 1 and .90 at Wave 2) and fathers (α=.85 at Wave 1 and .90 at Wave 2).
To assess adult insecurity in the interparental relationship, mothers and fathers completed the Attachment Style questionnaire at Waves 1 and 2 (AS; Becker, Billings, Eveleth, & Gilbert, 1997). The development of the AS consisted of adapting selective items from established adult attachment questionnaires (i.e., Bartholomew & Horowitz, 1991; Collins & Read, 1990; Hazan & Shaver, 1987; Mikulincer, Florian, & Tolmacz, 1990). The AS contains 19 items that assess three scales, including: (a) Preoccupied Attachment (i.e., “I often want to get closer to my partner than s/he wants to get to me.”), (b) Fearful Attachment (i.e., “I am not sure that I can always depend on my partner to be there when I need him/her.”), and (c) Secure Attachment (i.e., “I find it relatively easy to get close to my partner.”). Items were rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Internal consistency coefficients among the three AS scales in this sample were satisfactory for mothers (αs = .74 – 85 for Wave 1 and .78 – .87 for Wave 2) and fathers (αs = .74 – .84 at Wave 1 and .84 to .85 at Wave 2). The three scales were highly correlated for mothers (Mean r = .58; range .50 to .71) and fathers (Mean r = .54; range .43 to .70). Therefore, to retain methodological comparability with the manifest measure of depressive symptoms, the three scales were averaged together after reverse scoring the Secure Attachment Scale to correspond with the directionality of the other two scales.
Measures of parental insensitivity to child distress were obtained through maternal and paternal reports of their non-supportive reactions on two subscales (i.e., Minimization Reactions, Punitive Reactions) of the Coping with Children's Negative Emotions Scale (CCNES; Fabes, Poulin, Eisenberg, & Madden-Derdich, 2003). The CCNES consists of twelve hypothetical vignettes depicting child distress or anger (e.g.,” If my child is playing with other children and one of them calls him/her names, and my child then begins to tremble and become tearful, I would:”). After reading each vignette, parents independently rated the likelihood of responding to their children’s distress in each of six ways using a seven-point scale ranging from 1 (very unlikely) to 7 (very likely). Whereas the Minimization Reactions subscale reflects parental dispositions to downplay or devalue their children’s distress (e.g., “tell my child not to make a big deal out of it”), the Punitive Reactions subscale denotes parent tendencies to punish or reprimand their children for expressing negative affect (e.g., “tell my child to behave or we’ll have to go home right away”). The validity of these non-supportive scales is supported by its associations with indices of children’s social and emotional adjustment (e.g., Fabes et al., 2003; Jones, Eisenberg, Fabes, & MacKinnon, 2002) and negligible links with indices of social desirability (Fabes et al., 2003). Reliability for the Minimization and Punitive Reactions subscales was good for mothers (αs = .80 and .74 for Wave 1 and .86 and .77 for Wave 3) and fathers (αs = .82 and .81 at Wave 1 and .82 to .78 at Wave 3). Each of the two subscales were used as indicators of latent constructs of maternal and paternal insensitivity to child distress.
Mothers and fathers reported on their own use of psychologically controlling parenting behaviors using the Control through Guilt (e.g., “You tell your child that s/he would do what you want if s/he loved you”), Instilling Persistent Anxiety (e.g., “If your child breaks a promise, you don’t trust him/her again for a long time”), and Intrusive (e.g., “You ask your child to tell you everything that happens when s/he is away from home”) scales from the parent form of Child Report of Parenting Behaviors Inventory at Waves 1 and 3 (CRPBI; Margolies & Weintraub, 1977; Schluderman & Schluderman, 1970). Items on each of the five-item subscales were rated on 5-point Likert scales ranging from 1 (never) to 5 (always). The subscales have satisfactory internal consistency and previous research has supported the convergent and discriminant validity of the measures (Fauber, Forehand, Thomas, &Wierson, 1990). Internal consistency coefficients among the three CRPBI scales in this sample were adequate for mothers (αs = .66 – 71 at Wave 1 and .63 –68 at Wave 2) and fathers (α= .70 –.75 at Wave 1 and .73 –.80 at Wave 3). The three indicators for mothers and fathers were retained as manifest indicators of latent constructs of maternal and paternal psychological control.
Table 1 and Table 2 show the means, standard deviations, and correlations among the main variables for mothers and fathers, respectively. The moderate to high magnitude of correlations among the proposed manifest indicators of the main constructs (i.e., Mean r = .48 and .49 for mothers and fathers, respectively) provided initial support for the operationalization of the measurement model. Consistent with the proposed mediational pathways between the main constructs, significant interrelationships were evident among dimensions of interparental conflict and Wave 2 adult relationship insecurity and depressive symptoms for both mothers and fathers. Moreover, Wave 1 interparental conflict and Wave 2 adult relationship insecurity and depressive symptoms were more consistently associated with greater parenting difficulties for fathers than for mothers.
We conducted structural equation modeling using maximum likelihood estimation with AMOS software (Arbuckle, 2006) to examine adult relationship insecurity and depressive symptoms as explanatory mechanisms in links between interparental conflict and parenting difficulties. As a first step toward parsimoniously testing the mediational hypotheses, we conducted three sets of multi-group comparisons to evaluate whether the proposed pathways differed as a function of child gender, parent gender, and the combination of parent and child gender. Due to the complexity of the multi-mediator models in relation to our sample size, separate multi-group SEM analyses were performed for each type of parenting difficulty. For two of the three analyses involving child gender, the small sample sizes resulting from splitting the data by child gender necessitated the specification of path analyses using manifest, rather than latent, composites for our measures of interparental conflict, adult insecurity, and parenting difficulties. Manifest variables of these primary constructs were constructed by first standardizing their respective manifest indicators and subsequently summing them to create composites. Each multiple group comparison for the structural paths in Figure 1 and Figure 2 consisted of comparing a model in which all parameters were allowed to vary freely with a model in which comparable paths across the relevant groups of child gender, parent gender, and parent gender by child gender were constrained to be equal.
The results indicated that the free-to-vary models for child gender and the parent gender by child gender models did not generate significantly better representations of the data than the models constraining paths across gender to equality. Therefore, the results indicated that child gender or the configuration of parent and child gender did not moderate the proposed mediational pathways. However, the multi-group comparisons for parent gender indicated that the free-to-vary model provided a significantly better fit than the constrained model in the prediction of parental insensitivity to child distress, χ2 diff (7, N = 466) = 15.25, p < .05, and psychological control, χ2 diff (7, N = 466) = 14.70, p < .05. Because the mediational pathways varied as a function of parent gender, the primary analyses consisted of multi-group comparisons designed to simultaneously estimate the mediational paths for mothers and fathers in the prediction of each type of parenting difficulty.
Mediational tests often require that a significant association exist between the predictor and outcome (Baron & Kenny, 1986; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). Therefore, our first pair of SEMs constrained paths between each of the mediators (i.e., adult relationship insecurity and depressive symptoms) and parenting difficulties to zero. In the same model, we estimated the direct, predictive paths between (1) Wave 1 interparental conflict and the proposed mediators at Wave 2 and (2) the Wave 2 proposed mediators and Wave 3 parenting difficulties. To provide a conservative test of prospective associations, autoregressive paths were specified for the proposed mediators from Wave 1 to Wave 2 and each parenting difficulty from Wave 1 to Wave 3. In support of the precondition for testing mediation, interparental conflict at Wave 1 was a significant predictor of increases in paternal insensitivity, β = .18, p < .05, and psychological control, β = .16, p < .05, from Wave 1 to Wave 3. In contrast, comparable associations between interparental conflict and maternal insensitivity, β = −.03, and psychological control, β = −.03, were negligible and non-significant. Tests of parent gender as a moderator through estimation of pairwise parameter comparisons further revealed that interparental conflict was a significantly stronger predictor of parental insensitivity, z = 2.43, p < .05, and psychological control, z = 2.26, p < .01, for fathers in comparison to mothers.
Given support for the direct paths between interparental conflict and paternal parenting, we proceeded to test the full model in which the paths between each of the proposed mediators and parenting difficulties were freely estimated. Although results did not yield initial support for mediation in the maternal model, we estimated the full model for mothers and fathers simultaneously through multi-group comparisons based on two considerations. First, even if interparental conflict is unrelated to parenting difficulties for mothers, it is still possible that interparental conflict may be indirectly associated with greater parenting difficulties by undermining maternal depressive symptoms and adult relationship insecurity. Second, the multi-group analysis provided an examination of the mediational pathways involving depressive symptoms and adult relationship insecurity as a function of parent gender. In addition, the full model also estimated cross-lagged, longitudinal paths between the proposed mediators to test our model in relation to an alternative model which proposes that adult depressive symptoms and attachment insecurity are such strong predictors of one another in models of parenting difficulties that they supersede any role of interparental conflict as a factor in their prediction.
The multi-group SEM results for the model predicting parental insensitivity are shown in Figure 1. The model provided a good representation of the data, χ2 (80, N = 466) = 149.85, p < .001, RMSEA = .04, and χ2/df ratio = 1.87, CFI = .97, and TLI = .94. Results for mothers and fathers are depicted in the same model for the sake of comparison, with maternal findings bolded to differentiate them from the paternal findings. In support of the measurement models for both mothers and fathers, all loadings of the manifest indicators onto their respective latent constructs were significant (i.e., p < .001) and moderate to high in magnitude (i.e., .53 to .92).
Consistent with mediational hypotheses, analysis of the structural paths in the father model indicated that Wave 1 interparental conflict predicted Wave 2 paternal relationship insecurity, β = .19, p < .05 even after specifying autogressive paths and the cross-lagged paths between depressive symptoms and relationship insecurity (see Figure 1). Adult relationship insecurity at Wave 2, in turn, was a significant predictor of Wave 3 parental insensitivity, β = .15, p < .05. Inclusion of the structural paths between the proposed mediators and parenting difficulties resulted in notable reductions in the magnitude of the path coefficients between interparental conflict and paternal insensitivity (β reduced by 50% from .18, p < .05 to .09, p = n.s.). In further support of the mediational role of adult relationship insecurity, follow up tests using MacKinnon et al. (2002) procedures for indirect effects indicated that the indirect pathways involving interparental conflict and adult relationship insecurity were significant in the prediction of paternal insensitivity at Wave 3, z’ = 1.82, p < .05.
In accordance with the father model, the results for mothers indicated that Wave 1 interparental conflict was a predictor of increased insecurity in the interparental relationship from Wave 1 to Wave 2, β = .17, p < .05. In addition, Wave 1 maternal relationship insecurity was a predictor of increased depressive symptoms from Wave 1 to Wave 2, β = .16, p < .05. However, maternal depressive symptoms and adult relationship insecurity at Wave 2 were not significant predictors of maternal insensitivity at Wave 3.
Figure 2 depicts the results of the SEM multi-group comparison analyses in the prediction of parental psychological control. Results for mothers are represented by the bolded values to distinguish them from the findings for the fathers. The overall fit of the model was good, χ2 (232, N = 466) = 257.41, p < .001, RMSEA = .05, and χ2/df ratio = 2.11, CFI = .95, and TLI = .92. As with the parental sensitivity model, support for the measurement models for paternal and maternal psychological control is evidenced by the moderate to high loadings of the manifest indicators onto their proposed latent constructs (range: .50 – .90; p < .001).
The structural paths for the psychological control model in Figure 2 evidence a similar pattern of findings to the parental sensitivity analyses. For fathers, Wave 1 interparental conflict was a significant predictor of increases in paternal relationship insecurity, β = .19, p < .05 from Wave 1 to Wave 2. In turn, Wave 2 adult relationship insecurity was associated with Wave 3 paternal psychological control, β = .14, p < .05. In contrast, Wave 2 depressive symptoms were not significantly related to Wave 3 father psychological control, β = .03, n.s. Inclusion of the mediational paths resulted in a considerable reduction in the magnitude of the standardized path between Wave 1 interparental conflict and Wave 3 psychological control (β dropped by 53% from .17, p < .05 to .08, p = n.s.). Significance tests for mediation outlined by MacKinnon and colleagues (2002) further revealed that the indirect pathway among interparental conflict, adult relationship insecurity, and paternal psychological control was significant, z’ = 1.84, p < .05.
Similar to the findings from maternal sensitivity, Wave 1 interparental conflict predicted subsequent increases in maternal relationship insecurity, β = .17, p < .05, but not maternal depressive symptoms, β =−.08, n.s. Moreover, no support was found for maternal depressive symptoms or adult relationship insecurity as predictors of subsequent parenting difficulties.
The spillover hypothesis has garnered considerable empirical and theoretical attention, but the conditions and mechanisms underlying the transmission of distress from interparental relations to parenting difficulties remain largely unidentified. Moreover, because the spillover hypothesis was originally developed to understand why interparental discord was related to parenting difficulties over relatively small developmental periods (e.g., minutes, hours), even less is known about when and why interparental conflict may insidiously undermine parenting practices over relatively lengthy developmental spans of years. The goal of our longitudinal study was to address this gap by identifying affective mechanisms that mediate prospective associations between interparental conflict and parenting difficulties within a broader gender-differentiated model. Our findings indicated the interparental conflict predicted increases in parental psychological control and insensitivity to child distress for fathers but not mothers. Moreover, in supporting the moderating role of parent gender, interparental conflict was a significantly stronger predictor of increases in parental psychological control and insensitivity to child distress for fathers than for mothers. Mediational analyses further revealed that fathers’ susceptibility to experiencing increases in adult relationship insecurity explained why interparental conflict was associated with their greater parenting problems over time.
Although gender is increasingly regarded as critical factor for advancing an understanding of the spillover hypothesis (Cummings et al., 2004; Krishnakumar & Buehler, 2000), the role of gender as a moderator has rarely been tested directly in studies. Following calls by Cummings and colleagues (2004), our study separately tested and compared the strength of the spillover pathways across child gender, parent gender, and the gender composition of the parent-child dyad. The results showed that pathways between interparental conflict and parenting varied specifically as a function of parent gender. In accord with prior findings (Jouriles & Farris, 1992; Kitzmann, 2000; Owen & Cox, 1997), interparental conflict was associated with greater parenting difficulties for fathers but not mothers. However, against a backdrop of primarily cross-sectional and short-term experimental designs that do not explicitly examine parent gender as a moderator, a novel finding of this paper is that the magnitude of the paths between interparental conflict and subsequent increases in parental psychological control and insensitivity were significantly stronger for fathers than they were for mothers over a two-year time period. In drawing from the father vulnerability hypothesis (Cummings et al., 2004), one interpretation for the greater susceptibility of fathers is rooted in the notion that mothers may be better able to compartmentalize their spousal and parental roles, resulting in less carry-over from marital relations to parenting (Belsky, Youngblade, Rovine, & Volling, 1991; Lindahl, Clements, & Markman, 1997). Similarly, in keeping with traditional gender roles and the disproportionate prevalence of women as primary caregivers, parenthood might be regarded by women as a more fundamental role than men (Davies & Lindsay, 2001; Thompson & Walker, 1989).
Analysis of the mediational role of adult relationship insecurity and depressive symptoms in the spillover process further elucidated why paternal parenting is more likely to progressively deteriorate following experiences with interparental conflict. In examining the first autoregressive path in the mediational chain, interparental conflict predicted subsequent increases in interparental relationship insecurity, but not depressive symptoms, over a one year period for fathers. However, the father vulnerability hypothesis does not propose that the first link in the mediational chain is the source of their greater vulnerability. In fact, the tacit assumption of this model is that interparental conflict should incur a similar emotional toll on both mothers and fathers (Cummings & O’Reilly, 1997). Thus, the finding that interparental conflict also predicted increases in adult relationship insecurity for mothers is consistent with father vulnerability formulations of spillover. Within the father vulnerability hypothesis, the locus of the gender specificity in the transmission of distress lies in the associations between emotional spillover mechanisms (i.e., relationship insecurity, depressive symptoms) and parenting difficulties. The greater salience of the parenting role and more effective compartmentalization of spousal and parenting roles by mothers relative to fathers should specifically be evidenced in the ability to counteract the unfolding cascade of negativity into child-rearing contexts (Belsky et al., 1991; Cummings et al., 2004; Thompson & Walker, 1989). Consistent with this prediction, the findings in the second link of the proposed mediational chain indicated that adult relationship insecurity predicted subsequent increases in both parental insensitivity to child negative affect and psychological control for fathers but not mothers.
In attesting to its strength as an explanatory mechanism of spillover processes in fathers, the mediational role of adult relationship insecurity in pathways between interparental discord and fathering difficulties remained even when estimating adult depressive symptoms as a possible mediator and rival predictor in the analytic models. Thus, the results lend support to the notion that more global forms of negative affect are not necessarily the operative processes underlying the transmission of disturbances from the interparental to parent-child relationship. Relative to a trait conceptualization of negative affect (i.e., depressive symptoms), our findings not only demonstrate intriguing specificity in adult relationship insecurity as a carrier of spillover but they also substantially reduce the plausibility of competing, methodological explanations. For example, in extending trait negative affectivity models (Harold & Conger, 1997; Watson & Pennebaker, 1989), associations between interparental conflict, adult emotional distress, and parenting difficulties may simply be an artifact of the tendency for individuals with high levels of trait negativity to inflate reports of adverse experiences with others and their own negative attributes. However, the failure to identify depressive symptoms as a mediator within a broader autoregressive design that controls for prior values of the primary constructs militates against the notion that the findings are simply an artifact of a negative affectivity trait.
By the same token, the results beg the question of why adult relationship insecurity serves as a specific affective mechanism of spillover even after taking into account global depressive symptoms. According to ethological formulations of security (Ainsworth, 1985; Crowell et al., 2002; Davies & Sturge-Apple, 2007; Mikulincer, Shaver, Gillath, & Nitzberg, 2005), the attachment system is so closely intertwined with the caregiving system and its adaptive function of protecting dependents through sensitivity and responsiveness to dependent needs that it is commonly viewed as a component of the adult attachment relationship. By extension, attachment insecurity fostered by doubts about the availability of one’s partner as a source of support may also undermine the caregiving system, impeding adults’ ability to be sensitive and responsive to the needs of both their partner and, potentially, their children. The primacy of self-protective systems (e.g., attachment, social defense/wariness) in the context of recurrent threats posed by interparental conflict, may limit opportunities to utilize the adult partnership as a training ground for developing and refining strategies in the caregiving system (Davies & Sturge-Apple, 2007), thereby coalescing into stable patterns in which the caregiving system is consistently superseded by security concerns (Feeney & Collins, 2001; Mikulincer et al., 2005). Therefore, increasing tendencies to utilize punitive, minimizing, and intrusive child-rearing practices may be manifestation of disturbances in the caregiving system.
However, this interpretation, in itself, does not address the question of why relationship insecurity served as a mediator of associations between interparental conflict and parenting only for fathers. The maternal gatekeeping hypothesis may provide one explanation for this pattern of findings. In assuming a greater role in organizing caregiving activities and opportunities within the family (Allen & Hawkins, 1999; Parke, 2002), mothers are theorized to act as “gatekeepers” that increase or decrease (a) paternal access to opportunities to acquire parenting skills from them and (b) paternal autonomy to take on caregiving responsibilities for their children (Grossman, Pollak, & Golding, 1988). High levels of interparental conflict may be associated with diminished paternal involvement as mothers in distressed marriages become increasingly reluctant to collaborate with fathers in caregiving responsibilities (De Luccie, 1995). However, since paternal involvement and maternal gatekeeping were not assessed, it remains for future research to test whether maternal gatekeeping and consequent diminished paternal involvement are factors in relations between adult relationship insecurity and spillover to parenting for fathers.
In drawing from an evolutionary perspective, another explanation is that perturbations in the balance between security and caregiving concerns in the interparental relationship may be more pronounced for men than for women. Evolutionary models of pair-bonding specifically theorize that maximizing reproductive fitness may demand different strategies for males and females (Geary & Flinn, 2001). For women, the ability to definitively determine their maternity permits the development of the caregiving system within the context of the parent-child relationship, in which effort is devoted toward protection and support of the child (Draper, 1989; Geary, 2000). In contrast, paternal uncertainty for males may pull for the development of the caregiving system within the context of a lengthy, pair-bond relationship with their partner. Caregiving behaviors that reinforce the pair-bond relationship serve the adaptive function of increasing reproductive fitness by insuring continued access to a reproductive partner and by increasing certainty of paternity in potential offspring through reducing her exposure to and desire to mate with other adult males. A corollary of this assumption is that the adaptive value of male parental investment depends on signals of monogamy from the female. Conflict or withdrawal between parents may represent a lack of pair-bond enhancing behaviors, which could signal a potential change in the balance of costs and benefits for reproductive effort. Thus, the significance of the intimate adult relationship as a context for the development of the caregiving system in men may help to explain why mediational paths between interparental conflict, relationship insecurity, and parenting were only applicable to fathers.
Interpretation of the present findings, however, must be qualified by several limitations of the study. First, despite some diversity in sociodemographic (e.g., race, income, education) and family risk (i.e., high levels of interparental distress), care should be taken in generalizing the findings to other samples. Second, although interparental conflict was assessed using multiple methods and informants, maternal and paternal depressive symptoms, relationship insecurity, and parenting practices were assessed using self-report instruments. Although controlling for earlier levels of depressive symptoms, adult relationship insecurity, and parenting practices in our autoregressive models substantially reduces common method and informant variance (e.g., Steinberg, Lamborn, Darling, Mounts, & Dornbusch, 1994), multi-method or multi-informant assessments of each of the key constructs may help to further clarify spillover processes from the interparental to parent-child relationship. For example, our self-report assessment of adult relationship insecurity only captures one facet of a multi-faceted attachment system. Because cognitive, affective, and behavioral components of the attachment system also operate outside of conscious awareness (Mikulincer et al., 2002), obtaining more comprehensive assessments of adult relationship security through semi-structured interview or observation techniques is an important direction for future research (e.g., Crowell et al., 2002; Treboux et al., 2004; Wampler, Riggs, & Kimball, 2004). Third, as is common in lagged panel designs (e.g., Cui, Donnellan, & Conger, 2007), the magnitude of the mediational pathways in our study were relatively modest. However, comparable effect sizes in prior research are still regarded as having considerable practical and theoretical implications (for a discussion, see Cui et al., 2007).
As a further qualification in interpreting the results, the focus on examining mediational pathways among interparental conflict, adult depressive symptoms and relationship insecurity, and parenting difficulties does not reduce the plausibility of other patterns of family process. For example, although the results supported the role of interparental conflict as a predictor of depressive symptoms and adult relationship insecurity, previous research underscores other patterns of directionality among these family characteristics. For example, consistent with prior empirical documentation of concurrent and prospective associations (Carnelley et al., 1994; Feeney et al., 2003; Roberts et al., 1996), maternal relationship insecurity predicted increases in their depressive symptoms over a one year period. Likewise, research has also shown that adults exhibiting elevated depressive symptoms experience subsequent increases in distress in their marital relationships (Davies, Dumenci, & Windle, 1999; Kouros, Papp, & Cummings, 2008). In addition, because adult relationship insecurity is likely to be one component of a multitude of mediational pathways involved in spillover, testing the role of other attributes (e.g., maternal gatekeeping, paternal involvement) as explanatory mechanisms and broadening the conceptual scope of spillover to incorporate more supportive and constructive relationship and emotional processes are important directions for future research (Simons, Lorenz, Wu, & Conger, 1993).
In conclusion, our study constitutes a first step toward beginning to elucidate the affective underpinnings of the spillover hypothesis, that is, the cumulative transmission of difficulties from the interparental to parent-child relationship. Over the relatively lengthy developmental period of two years, interparental conflict was a significant predictor of subsequent parental difficulties only for fathers. Consistent with evolutionary models of parenting, autoregressive SEM models indicated that greater paternal susceptibility to affective forms of parenting problems (i.e., psychological control, insensitivity to child negative affect) following experiences with interparental discord were mediated by increases in their interparental relationship insecurity.
This research was supported by the National Institute of Mental Health (Project R01 MH57318) awarded to Patrick T. Davies and E. Mark Cummings. The authors are grateful to the children, parents, teachers, and school administrators who participated in this project. Their gratitude is also expressed to project staff, including: Courtney Forbes, Courtney Henry, Marcie Goeke-Morey, Amy Keller, Michelle Sutton, Alice Schermerhorn, and the graduate and undergraduate students at the University of Rochester and University of Notre Dame.
Publisher's Disclaimer: The following manuscript is the final accepted manuscript. It has not been subjected to the final copyediting, fact-checking, and proofreading required for formal publication. It is not the definitive, publisher-authenticated version. The American Psychological Association and its Council of Editors disclaim any responsibility or liabilities for errors or omissions of this manuscript version, any version derived from this manuscript by NIH, or other third parties. The published version is available at www.apa.org/journals/dev.
Patrick T. Davies, Department of Clinical and Social Sciences in Psychology, University of Rochester.
Melissa L. Sturge-Apple, Mt. Hope Family Center, University of Rochester.
Meredith J. Woitach, University of Rochester.
E. Mark Cummings, Department of Psychology, University of Notre Dame.