Interparental Discord and Child Adjustment

Interparental Discord and Child Adjustment: Prospective Investigations

of Emotional Security as an Explanatory Mechanism

E. Mark Cummings and Alice C. Schermerhorn University of Notre Dame

Patrick T. Davies University of Rochester

Marcie C. Goeke-Morey and Jennifer S. Cummings University of Notre Dame

Advancing the process-oriented study of links between interparental discord and child adjustment, 2 multi- method prospective tests of emotional security as an explanatory mechanism are reported. On the basis of community samples, with waves spaced 2 years apart, Study 1 (113 boys and 113 girls, ages 9 – 18) identified emotional security as a mediator in a 2-wave test, whereas Study 2 (105 boys and 127 girls, ages 5 – 7) indicated emotional security as an intervening mechanism in a 3-wave test. Relations between discord and emotional security increased as children moved into adolescence in Study 1. Emotional security was identified as an explanatory mechanism for both internalizing and externalizing problems in children.

A first generation of research has established rela- tions between interparental discord and children’s adjustment problems, including links between mar- ital discord and child maladjustment measured several years apart (e.g., Davies & Cummings, in press; Grych & Fincham, 1990; Ingoldsby, Shaw, Owens, & Winslow, 1999; Neighbors, Forehand, & Bau, 1997). In recent years a second generation of research has focused on identifying the mechanisms or processes that explain these effects on children because of exposure to interparental discord (Cum- mings & Davies, 2002). Several conceptual models have been proposed to explain maladjustment in children, with common agreement that emotional, cognitive, and behavioral processes are potential mediators, and that negative forms of marital conflict (i.e., interparental hostility, see Goeke-Morey, Cum- mings, Harold, & Shelton, 2003) are predictive of maladjustment (Crockenberg & Langrock, 2001; Da- vies & Cummings, 1994; Emery, 1989; Grych & Fincham, 1990).

A next step is longitudinally testing theoretical models for relations between interparental discord, explanatory mechanisms, and child maladjustment. Most of the process-oriented studies characteriz- ing second-generation research have used cross- sectional designs exploring concurrent relations among variables (for exceptions, see Grych, Harold, & Miles, 2003; Harold & Conger, 1997; Harold, Fincham, Osborne, & Conger, 1997). Initial cross- sectional tests of explanatory models have been promising with regard to identifying key processes accounting for relations between interparental dis- cord and child adjustment (e.g., Grych, Fincham, Jouriles, & McDonald, 2000). Longitudinal research is essential to further advance the cogency of theo- retical models about mediational pathways, includ- ing testing relations between variables pertinent to inferences about cause and effect relations (Fincham & Grych, 2001). For example, in a two-wave longi- tudinal design, Grych et al. (2003) reported relations between interparental conflict, children’s appraisals of threat and self-blame, and child adjustment.

A mediational pathway receiving considerable support is drawn from emotional security theory (EST; Davies & Cummings, 1994). For years theorists speculated that children’s emotional security in the context of the family derived from the quality of marital as well as parent – child relationships (Blatz, 1966; Byng-Hall, 1995; Cummings, Zahn-Waxler, & Radke-Yarrow, 1981; Marvin & Stewart, 1990). For example, Bowlby (1949) pointed out the need to

r 2006 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2006/7701-0010

This research was supported in part by a grant from the Na- tional Institute of Child Health and Human Development (HD 036261) to the first author and by a grant from the National Inst- itute of Mental Health (MH 57318) to the first and third authors. We gratefully acknowledge the assistance of Scott Maxwell, PhD, for consultation on advanced statistical matters.

Correspondence concerning this article should be addressed to E. Mark Cummings, The Notre Dame Endowed Chair in Psy- chology, Department of Psychology, 118 Haggar Hall, University of Notre Dame, Notre Dame, IN 46556. Electronic mail may be sent to Cummings.10@nd.edu.

Marcie C. Goeke-Morey is now at The Catholic University of America.

Child Development, January/February 2006, Volume 77, Number 1, Pages 132 – 152

consider family in understanding children’s security and distress. On the basis of this foundation, EST was advanced as a family-wide model of emotional se- curity as a regulatory process, including emotional security about marital relations (Cummings & Da- vies, 1996). Emotional security is an appraisal that family bonds are positive and stable, even in the face of everyday stressors (e.g., marital conflict), and that family members can be expected to remain respon- sive and emotionally available for the child, even in times of stress. Consistent with a functionalist per- spective on emotions, appraisal processes may in- clude emotional as well as cognitive processes.

As in attachment theory, protection, safety, and security are held to be among the most salient goals in the hierarchy of human goals (Bowlby, 1973; Wa- ters & Cummings, 2000). Another similarity is that EST is a developmental theory that assumes the child’s emotional security can be enhanced or un- dermined by the quality of family relations (Bowlby, 1973; Cummings, Davies, & Campbell, 2000). A di- vergence (e.g., Ainsworth, Blehar, Waters, & Wall, 1978) is that multiple family influences on emotional security in addition to parent – child attachment are posited (Cummings & Davies, 1996; for a discussion of additional influences, see Waters & Cummings, 2000). For example, Davies, Harold, Goeke-Morey, and Cummings (2002) showed that security in the context of marital conflict and attachment, respec- tively, independently influenced child adjustment.

EST posits that preserving a sense of security in the face of marital conflict is a critical goal that or- ganizes a child’s responding. Drawing from core notions of attachment theory (Bowlby, 1969), EST posits a control system model in which preserving emotional security is a set-goal that influences chil- dren’s reactions to marital conflict. That is, children evaluate marital conflict in terms of the set-goal of emotional security, with the emotional security behavioral system activated if that set-goal is threatened. Sroufe and Waters (1977) introduced felt- security as a set-goal for the attachment behavioral system. Although an important and influential con- tribution, this approach has been criticized because insecurity may also occur in nonconscious processes that may not be reported as ‘‘felt’’ (Ainsworth, 1990; Kobak & Sceery, 1988). EST importantly extends Sroufe and Waters’ (1977) concept to include more than self-reported feelings as regulatory processes (e.g., behavioral or physiological reactions; Cum- mings & Davies, 1996; Davies, Cicchetti, Sturge- Apple, & Cummings, 2005). In summary, EST posits that children appraise marital conflict in terms of its implications for children’s sense of emotional

security, and are motivated to respond accordingly (e.g., negative emotional reactivity; mediation in marital disputes) if the desired level of security about marital conflict (i.e., the set-goal) is violated.

To give a concrete example, if children observed aggression by one parent toward the other, responses of negative emotional reactivity and overinvolve- ment in the marital dispute would be among the responses expected by the children. These responses reflect the children’s emotional insecurity regarding the interparental relationship (see Cummings, Goeke- Morey, & Papp, 2003). The responses also function as a regulatory response system toward regaining the set-goal of emotional security about the interparental relationship. Applying secure base and control sys- tems concepts to the family context, it follows that a child’s emotional security about marital conflict can be assessed by the organization of regulatory processes occurring in the context of interparental conflict that serve the goal of preserving the child’s emotional security (Cummings & Davies, 1996; Davies, Harold et al., 2002).

The emotional security response system thus reg- ulates, organizes, and motivates a child’s responses to interparental discord, including emotional experi- ences (e.g., negative emotional reactivity) and behav- ioral action tendencies (e.g., involvement; behavioral dysregulation) (e.g., Cummings, Goeke-Morey, & Papp, 2004). These responses have been identified as especially pertinent to the control system concerned with regaining or preserving emotional security (e.g., Davies & Cummings, 1994; Davies, Harold et al., 2002). Notably, some degree of interdependency is expected among these response systems, as they serve the same goal of preserving security, but each is also assumed to represent distinct aspects of the emotional security system (Davies, Harold et al., 2002). Children’s heightened emotional and behavi- oral dysregulation, and elevated involvement, in response to marital conflict therefore provide theory- based assessments of children’s underlying emo- tional insecurity about marital conflict (see Davies, Forman, Rasi, & Stevens, 2002). Moving beyond simply an emphasis on single response domains, emotional security is conceptualized as a higher or- der system that can be understood from an organi- zational perspective on the functioning of the regulatory systems serving the goal of maintaining or regaining emotional security (Davies, Cummings, & Winter, 2004; Davies & Forman, 2002). Thus, sim- ilar to attachment security (Ainsworth et al., 1978; Cicchetti & Cohen, 1995; Sroufe & Waters, 1977), emotional security in the context of marital conflict is an organizational construct, with children’s underly-

Interparental Discord and Child Adjustment 133

ing sense of emotional security about marital conflict served, and indicated, by a relatively broad class of interrelated responses.

These responses may be adaptive in the short term. For example, they elevate vigilance in the threatening context of destructive conflict. At the same time, emotional insecurity may increase chil- dren’s risk for adjustment problems in the long term (Cicchetti, Cummings, Greenberg, & Marvin, 1990; Thompson, 2000; Waters, Weinfeld, & Hamilton, 2000), fostering the development of psychopatholo- gy (see Cassidy & Shaver, 1999; Cummings & Cic- chetti, 1990). With regard to emotional insecurity about marital conflict, vigilance, fear, and preoccupa- tion with adult problems may increase risk for more pervasive internalizing problems. Parental conflict may promote negative expectations or representations of the aftermath of conflict, serving as a (maladap- tive) guide for interpreting other challenging con- texts (e.g., peer conflicts). Insecurity may heighten tendencies for maladaptive responding in new set- tings, including being less flexible, open, and skilled in forming and maintaining social relationships (Da- vies, Harold et al., 2002). The energy required to re- gain emotional security in the face of interparental conflicts may limit the psychological and physical re- sources needed to pursue significant developmental goals (Cummings & Cummings, 1988; Davies, Har- old et al., 2002). Children may be sensitized to inter- personal stress because of continual challenges to security, resulting in persistent emotional or behav- ioral dysregulation, or negative views of the self and others (Cummings & Davies, 1994).

A useful analogy is to think about emotional se- curity as a bridge between the child and the world. When the marital relationship is functioning well, it serves as a secure base, a structurally sound bridge, supporting the child’s exploration and relationships with others (Waters & Cummings, 2000). When de- structive marital conflict erodes the bridge, children may become hesitant to move forward and lack confidence, or may move forward in a dysregulated way, unable to find appropriate footing within themselves or in interaction with others.

Several recent studies provide initial support for the explanatory mechanisms outlined by the emotional security hypothesis. On the basis of cross- sectional data, Davies and Cummings (1998) indi- cated initial empirical support for emotional security as mediating relations between interparental discord and child maladjustment (see also Davies, Forman et al., 2002). In a first longitudinal test of the theory, Davies, Harold et al. (2002, Study 2) demonstrated that child emotional security was an intervening

process linking interparental discord (Time 1) with subsequent child externalizing and internalizing symptoms (Time 2), even when other mechanisms (e.g., perceived threat, self-blame; Grych et al., 2003) were considered in the same analytic models. In another two-wave longitudinal study, Harold, Shel- ton, Goeke-Morey, and Cummings (2004) provided support for the notion that components of emotional security were a key mechanism linking interparental discord with children’s risk for maladjustment, even when emotional security about parenting was also entered in the same statistical model (see also Da- vies, Harold et al., 2002, Study 3).

On the basis of two large-scale (2001) U.S. samples of children and families, this report further explores emotional security as an explanatory mechanism pro- spectively accounting for relations between interpa- rental discord and children’s maladjustment. Recent prospective investigations of longitudinal models are based on British samples (Davies, Harold et al., 2002; Grych et al., 2003; Harold et al., 2004); therefore, this research provides opportunities to examine the generalizability of this theoretical model to a U.S. sample. In addition, the present two studies examine emotional security as a latent construct, rather than examining separate components of emotional secu- rity as predictors (e.g., Davies & Cummings, 1998; Harold et al., 2004), thereby providing a conceptually integrative test of emotional security as an explana- tory, organizational mechanism (see also Davies, Harold et al., 2002). Moreover, the earlier longitudi- nal tests (e.g., Harold et al., 2004) were based on analog designs that raise questions about the appli- cability to children’s actual reactions, that is, concerns about ecological validity. In this study, assessments of emotional security are based on reports of children’s actual reactions in the home, thereby further advanc- ing the ecological validity of tests of the propositions of the emotional security hypothesis in the context of longitudinal research designs. Thus, the present two studies, involving different samples and different age groups, provide further programmatic exploration of emotional security as an explanatory mechanism for the effects of interparental discord on children’s ad- justment. The central hypothesis is that emotional security will contribute to the explanation of the im- pact of marital discord on children throughout a sig- nificant span of childhood, that is, for children varying in age from kindergarten through adolescence.

Study 1

Study 1 presents an initial exploratory, longitudinal analysis of multiple pathways between interparental

134 Cummings, Schermerhorn, Davies, Goeke-Morey, and Cummings

discord, emotional security about marital conflict, and child adjustment. A secondary goal for Study 1 was to investigate age differences. Samples in the few longitudinal studies conducted have had re- stricted variability in age at Time 1. Although this fact has advantages for addressing some questions, it has restricted possibilities for testing the viability of emotional security as an explanatory mechanism as a function of age. Although age is acknowledged as a potential moderator in models of interparental con- flict, there is little clear support for the notion that children of any specific age group or developmental level are, in any broad sense, more vulnerable to interparental conflict (Buehler, Anthony, Krishnaku- mar, & Stone, 1997). Even less is known about the viability of process models about explanatory mechanisms for the impact of marital conflict on children in different age groups, or about the effects of marital conflict on different forms of child mal- adjustment as a function of age. Children in Study 1 were selected to vary across a relatively wide age span (i.e., about 8 years) from middle childhood to middle adolescence at the onset of the study, facili- tating the examination of age-related differences in the viability of the emotional security hypothesis as a process model. However, given the exploratory na- ture of the tests for emotional security as a process mechanism, no specific hypotheses are advanced as a function of age.

Method

Participants

Participants were 226 primarily middle-class chil- dren (113 boys, 113 girls) and their cohabiting parents. In order to be eligible to participate, couples had to have cohabited for at least 2 years before the begin- ning of the study. Two hundred and twenty-two (98%) of the couples were married, and the average length of cohabitation was 13.44 years (SD 5 6.26). The mean age for mothers was 37.61 years (SD 5 6.02), and the mean age for fathers was 40.03 years (SD 5 6.90). The target child’s mean age was 11.12 years (SD 5 2.28, range 5 8 – 17) at Time 1. The measure of emotional security was not available for wave 2. Thus, the data used for examining the viability of emotional security as an explanatory mechanism are based on waves 1 and 3 from the larger study, described as Times 1 and 2, respectively, in this report. This data set has previously been de- scribed by DuRocher Schudlich and Cummings (2003).

In order to obtain a sociodemographically repre- sentative sample of the geographic area, efforts were

made to actively recruit families of low socioeco- nomic status (SES) and of racial and ethnic diversity, including school districts, events and sign-ups at community agencies and events for diverse com- munities, and postcard mailings for lower SES areas. Representative of the Midwest community from which this sample was drawn, 87% of the children were European American, 8.5% were African American, 3% were biracial, 0.5% were Asian, and 1% were Hispanic. Families participated as part of a larger longitudinal project. The attrition rate was low at Time 2, with 221 of the families retained.

We compared the socioeconomic characteristics of our sample with those of the county from which this sample was drawn. In our sample, 98% of parents had completed at least high school, and 38% had completed at least a bachelor’s degree. Based on U.S. Census Bureau statistics (2000), 82% of the popula- tion in this county had completed at least a high school education, with 24% completing at least a bachelor’s degree. The median household income in our sample was in the range of $40,000 – 65,000, and the median household income in 2000 for the county was $49,653.

Providing a basis for comparing the marital functioning of our sample with those of other com- munity samples, mothers and fathers reported their global marital satisfaction on the Marital Adjustment Test (MAT; Locke & Wallace, 1959). The MAT is a widely used measure, with good content and con- current validity. Scores can range from 2 to 158, with scores below 100 suggesting marital distress (Crane, Allgood, Larson, & Griffin, 1990). The mean marital satisfaction score for mothers was 112.99 (SD 5 25.25, range 5 34 – 154) and the mean for fathers was 111.67 (SD 5 21.84, range 5 28 – 156). Sixty-one mothers (27%) and 59 fathers (26%) had MAT scores below 100, suggesting marital distress. Eight-nine of the 226 couples (39%) included at least one partner with a score below 100. Although the percentages of participants scoring in the dis- tressed range are somewhat higher than those reported in other studies based on community samples, the average level of distress is comparable to that of other community samples (e.g., McHale, Kuersten-Hogan, Lauretti, & Rasmussen, 2000). Of the 226 families participating in the study, nine couples separated or divorced. Given that the goal of this study is to test emotional security theory with a community sample, these families were retained as participants at all time points in which they were willing to participate, on the basis that they contributed to the representativeness of our com- munity sample.

Interparental Discord and Child Adjustment 135

Measures

Marital functioning. The O’Leary – Porter Scale (OPS; Porter & O’Leary, 1980) is a 10-item measure of child exposure to hostile interparental conflict. The OPS consists of 9 scored items completed on a 5- point scale ranging from 0 (never) to 4 (very often). A sample item from the OPS is How often do you com- plain to your partner about his/her personal habits in front of your child? Porter and O’Leary reported a test – retest reliability coefficient of .96, good internal consistency (Cronbach’s a5 .86), and good conver- gent validity. Cronbach’s as for mothers and fathers in this sample were .85 and .80, respectively.

The Conflict Tactics Scale (CTS, Straus, 1979) was completed by both mothers and fathers, and lists 19 conflict tactics for which parents report the fre- quency of occurrence by either spouse over the past year. Responses are indicated using an 8-point or- dinal scale from 0 (never) to 6 (more than 20 times). For the purpose of the current study, the Verbal Ag- gression subscale was used; sample items include did or said something to spite the other one and insulted or swore at the other one. Previous research has sup- ported the construct and concurrent validity of this measure. Cronbach’s as for mothers and fathers in this sample were .79 and .79, respectively.

The Negative Marital Quality dimension of the Positive and Negative Quality in Marriage Scale (Fincham & Linfield, 1997) was completed by both mothers and fathers. Parents were instructed to rate negative qualities of their spouse, negative feelings toward their spouse, and negative feelings about their marriage, while ignoring the positive aspects. Spouses rated three items on a scale ranging from 1 (low) to 10 (high). Fincham and Linfield (1997) reported high in- ternal consistency for the Negative Marital Quality dimension for both mothers and fathers (Cronbach’s a5 .89 and .91, respectively) and demonstrated good validity for this dimension. In this sample, Cronbach’s as for mothers and fathers were .92 and .89, respec- tively. Mothers’ and fathers’ reports on the CTS, NMQ, and OPS were standardized and summed for each reporter to yield two indicators of interparental discord (one for mothers, one for fathers). Cronbach’s as computed on these three scales were .72 for mothers and .70 for fathers, providing support for the compositing of these variables.

The Children’s Perceptions of Interparental Con- flict Scale (CPIC; Grych, Seid, & Fincham, 1992) is a child-report measure of marital conflict, consisting of 51 items answered on a 3-point scale consisting of T (true), ST (sort of true), and F (false). The Conflict Properties subscale is a measure of children’s per-

ceptions of the characteristics of their parents’ con- flict, and includes such items as When my parents have disagreements, they yell at each other. The Conflict Properties subscale has demonstrated internal con- sistency greater than .70 and good test – retest relia- bility. The Cronbach’s a computed for this sample was .83.

Emotional security about marital relations. The Se- curity in the Marital Subsystem-Parent Report Inventory (SIMS-PR, Davies, Forman et al., 2002) is a parent- report measure of child emotional security, and in- cludes the 9-item Direct Involvement, the 5-item Behavioral Dysregulation, and the 10-item Negative Emotional Reactivity subscales. Mothers reported on their child’s reactions to witnessing arguments be- tween parents in the past year. Each item consists of a different emotional or behavioral response, and is completed on a 5-point ordinal scale from 1 (not at all like him/her) to 5 (a whole lot like him/her). Sample items include Tries to comfort one or both of us, Yells at family members, and Appears upset. Davies, Forman et al. (2002) reported Cronbach’s as for mothers and fathers of .78 and .64 for Emotional Reactivity, .69 and .70 for Involvement, and .64 and .65 for Behav- ioral Dysregulation. Cronbach’s as in our sample were .71 for Behavioral Dysregulation, .77 for In- volvement, and .84 for Emotional Reactivity.

Child maladjustment. Mothers and fathers com- pleted the Child Behavior Checklist (CBCL; Achen- bach, 1991a), a parent-report measure of both internalizing and externalizing child adjustment problems. The Internalizing and Externalizing sub- scales consist of 58 items completed on a 3-point ordinal scale. Parents rate whether or not each statement is true of their child from 0 (not true as far as you know) to 2 (very true or often true). The CBCL is composed of three Internalizing subscales: Somatic Complaints (e.g., headaches), Anxious/Depressed (e.g., nervous, high strung or tense), and Withdrawn (e.g., withdrawn, doesn’t get involved with others). The Externalizing scale consists of two subscales: Ag- gressive (e.g., gets into many fights) and Delinquent (e.g., vandalism). The test – retest reliability of CBCL scale scores is good, with a mean r of .89 for the problem scales over a 7-day period. The CBCL has also demonstrated both content and construct va- lidity (Achenbach, 1991a). For mothers and fathers in this sample at Time 1, Cronbach’s as were .84 and .85 for Internalizing and .89 and .87 for Externalizing, respectively. At Time 2, Cronbach’s as for mothers and fathers were .87 and .89 for Internalizing and .89 and .87 for Externalizing, respectively.

Children completed the Child Depression Inven- tory (CDI; Kovacs, 1981), a 27-item self-report

136 Cummings, Schermerhorn, Davies, Goeke-Morey, and Cummings

measure of children’s depressive symptoms. Each set of items consists of three sentences representing varying levels of depression, and children are in- structed to choose the sentence that best describes them for the past 2 weeks from each set of items. Sample sentences include I feel like crying everyday and I feel alone all the time. Each sentence is assigned a numerical value from 0 to 2, with higher scores corresponding to higher degrees of depression. The values for the endorsed sentences are summed to compute a CDI score for each child. The CDI has demonstrated good reliability and validity (Kovacs, 1981). Cronbach’s as for this sample were .78 at Time 1 and .85 at Time 2.

Procedure

Participating parents and their children complet- ed questionnaires and tasks in the laboratory every year, with laboratory sessions lasting approximately 3 hr. Parents completed questionnaires about demo- graphic information, marital functioning, and other measures in separate rooms. Children also complet- ed questionnaires in a separate room with the as- sistance of a trained research assistant.

Results

Descriptive statistics and intercorrelations be- tween all of the variables are shown in Table 1. There was a 2-year period between Times 1 and 2. Struc- tural equation modeling was used to test our central hypotheses regarding the effects of interparental discord on child adjustment. The structural equation modeling analyses were conducted using Analysis of Moment Structures (AMOS, v. 4.01; Arbuckle & Wothke, 1999), a computer software package that utilizes the full information maximum likelihood approach to handling missing data. We tested a se- ries of models examining the direct effect of inter- parental discord and the mediational effect of emotional security. We also tested child age as a moderator of the mediational models. An a level of .05 was used for all statistical tests.

We report multiple fit indices to facilitate evalua- tion of the degree to which our models fit the sample data. The traditional w2 discrepancy test is presented, although this statistic is artificially inflated by sam- ples as large as ours (Bentler & Bonett, 1980). Several additional fit measures are also presented. For the relative w2 index (w2/df) values below 3 are con- sidered indicative of an acceptable fit between the hypothetical model and the sample data (Arbuckle & Wothke, 1999). The root mean square error of

approximation (RMSEA; Browne & Cudeck, 1993) penalizes models that include many parameters, and values of the RMSEA less than, or equal to, .08 in- dicate a reasonable fit. Finally, values of the com- parative fit index (CFI; Bentler, 1990) should be at least .90 for an acceptable fit. In order to minimize artificial inflation of scores reported by the same observer, as well as scores from different observers for the same measures, we allowed the error terms for these measures to be correlated (e.g., maternal reports of maladjustment at Time 1 were allowed to correlate with maternal reports of maladjustment at Time 2).

Consistent with our expectations and previous work, we found a direct effect of Time 1 interparental discord on Time 2 child adjustment, controlling for Time 1 adjustment. Internalizing symptoms were highly stable over this 2-year period (b5 .75, po.001). Nonetheless, interparental discord was a significant predictor of increases in internalizing problems (b5 .19, po.05). Similarly, even controlling for the autoregressive path from Time 1 externalizing symptoms to Time 2 externalizing symptoms (b5 .78, po.001), interparental discord predicted increases in externalizing problems (b5 .19, po.05). Demonstrating these significant direct effects repre- sents the first step in Baron and Kenny’s (1986) guidelines for testing mediation.

Mediational Model

Next, rigorous testing of mediational models was conducted, to test hypotheses about emotional se- curity as an explanatory mechanism. We tested for mediation of the effect of Time 1 interparental dis- cord on Time 2 child adjustment through children’s emotional insecurity at Time 2. Providing a stringent test of EST, we controlled for the effect of Time 1 child adjustment on both Time 2 emotional security and Time 2 child adjustment. As described by Cole and Maxwell (2003), when some of the relevant paths are not included in the model, the paths that are modeled are biased, with the possibility that vari- ance that is actually due to other variables is attrib- uted to the paths that are modeled. In addition, we were concerned that by not testing age as a moder- ator, we might be omitting an important variable of interest for this sample. Toward providing a com- prehensive and precise test of emotional security theory, we therefore tested models that included both tests of age as a moderator and controls for initial child adjustment.