Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2010 Oct 27.
Published in final edited form as: J Fam Psychol. 2008 Feb;22(1):144–152. doi: 10.1037/0893-3200.22.1.144

Deeper Into Divorce: Using Actor–Partner Analyses to Explore Systemic Differences in Coparenting Conflict Following Custody Dispute Resolution

David A Sbarra 1, Robert E Emery 2
PMCID: PMC2964495  NIHMSID: NIHMS243259  PMID: 18266541

Abstract

Divorce is an inherently interpersonal experience, yet too often adults’ reactions to marital dissolution are investigated as intrapersonal experiences that unfold outside of the relational context in which they exist. This article examines systemic patterns of interpersonal influence between divorced parents who were randomly assigned to either mediate or litigate a child custody dispute in the mid-1980s. Reports of coparenting conflict and nonacceptance of the divorce were assessed 5 weeks after the dispute settlement, 13 months after the settlement, and then again 12 years later. One hundred nine (N = 109) parents provided data over this 12-year period. Fathers reported the highest initial levels of conflict when their ex-partners were more accepting of the divorce. Mediation parents reported decreases in coparenting conflict in the year after dispute settlement, whereas litigation parents reported increases in conflict. Litigation parents evidenced the greatest long-term increases and decreases in coparenting conflict. Mediation is a potent force for reducing postdivorce conflict, and this article highlights the usefulness of adopting a systemic lens for understanding the long-term correlates of marital dissolution.

Keywords: divorce, child custody mediation, emotional adjustment, actor-partner interdependence model, coparenting conflict


Divorced families are still families, and, consistent with systems perspective, former partners have considerable influence on each other as families negotiate a marital transition. These issues are particularly apparent when parents are embroiled in disputes over child custody. Coparenting conflict can escalate quickly in response to practical child-rearing concerns or the relational dynamics between parents who are redefining the boundaries of their relationship. In many instances, child-rearing conflicts result from relational dynamics that, at most, are only tangentially related to the dispute (Emery, 1994). For instance, parents may fight for child custody as a means of emotionally (and literally) contesting the end of marriage.

Although a great deal is written on the interpersonal dynamics of divorce (Amato, 2000; Emery, 2004; Hetherington, Bridges, & Insabella, 1998; Sbarra & Emery, 2005; Wallerstein, Lewis, & Blakeslee, 2000), few empirical studies have examined the ways in which one parent’s behaviors and emotions influence or are influenced by their former spouse. In an effort to move beyond static, intraindividual perspectives on divorce adjustment, we examined patterns of interpersonal influence in the context of mediated or litigated child custody dispute resolution in the present article. Our main goals were to investigate how custody dispute resolution alters relational dynamics between divorced parents and to better understand how these patterns of interpersonal influence impact short- and long-term changes in coparenting conflict.

The study of divorce mediation provides an ideal context for exploring patterns of interpersonal influence in cases of disputed child custody. As an alternative to the adversary legal system, custody mediation emerged in the late 1970s, premised on the ability to diffuse coparenting conflict, enhance party satisfaction with the court system, and promote the equitable administration of justice (Beck & Sales, 2001; Emery, 1994; Emery & Wyer, 1987). In contrast to adversarial win–lose attorney negotiations and formal litigation, mediation is based on a model of cooperative (“win–win”) dispute settlement. In taking a more cooperative approach to custody disputes, mediation also may begin to alter patterns of interaction between hurt and angry parents, or, at least, prevent the potential escalation of hostilities that can result from adversarial legal negotiations and court proceedings (Beck & Sales, 2001; Myers, Gallas, Hanson, & Keilitz, 1988).

The Interpersonal Dynamics of Divorce

Relative to litigation, mediation has demonstrated the ability to help parents reach equitable settlements, enhance consumer satisfaction, and positively impact some aspects of postdivorce parenting and coparenting (Emery, Laumann-Billings, Waldron, Sbarra, & Dillon, 2001; Emery, Sbarra, & Grover, 2005; Kelly, 1996). Of course, a number of questions still exist about exactly how and for whom mediation works best (Beck & Sales, 2001). For example, Bickerdike and Littlefield (2001) recently demonstrated that a disparity in degree of attachment between ex-partners is corrosive to the mediation process. The findings from this study suggest that in an effort to forestall the dissolution of marriage, nonaccepting spouses obstruct mediation settlements by engaging in prolonged discussions about how the process of mediation works, for instance, rather than actually negotiating substantive issues (Bickerdike & Littlefield, 2001). These findings are consistent with the idea that coparenting conflict can be viewed in terms of attachment-related reunion behavior, in which one parent desperately seeks to maintain the relationship through prolonged, angry protest (Emery, 1994).

Other than Bickerdike and Littlefield’s (2001) research, no other investigations have examined interpersonal dynamics within mediation. In the present article, we built on this work in three specific ways. First, we investigated whether a parent’s own reports of coparenting conflict immediately following the custody settlement are associated with their former partner’s level of nonacceptance of the divorce. This serves as a conceptual replication of the Bickerdike and Littlefield (2001) research, extending their findings in time and outside of the mediation setting itself. Second, we investigated whether the association between nonacceptance of divorce and coparenting conflict is moderated by intervention group membership and gender. This approach involves determining whether patterns of interpersonal influence operate differently between mothers and fathers as well as between the two custody resolution groups. Finally, we examined how these processes unfold over time; in particular, we investigated whether initial patterns of non-acceptance in one parent predict elevated levels of coparenting conflict over a decade after the initial dispute resolution.

To address these topics, we reported new analyses from the Charlottesville Mediation Project (CMP), a field experiment comparing the effectiveness of problem-focused but emotionally informed mediation with the adversarial settlement of child custody disputes (see Emery, 1994; Emery et al., 2005). Seventy-one families who petitioned a Virginia court in the mid-1980s for a contested custody hearing were approached at random and asked to consider settling their custody disputes through a newly developed mediation program or continuing with a litigation-as-usual research program. Self-reported outcomes of the CMP were assessed on three occasions: a few weeks after the initial settlement, roughly 13 months after the settlement, and, finally, in a long-term follow-up 12 years after the settlement. Details on the methods and findings from the initial and follow-up studies can be found elsewhere (Emery, 1994; Emery, Matthews, & Kitzmann, 1994; Emery, Matthews, & Wyer, 1991; Emery & Wyer, 1987).

The Actor–Partner Interdependence Model (APIM)

To address the topics discussed above, we used data from the three waves of the CMP but analyzed it in a new way. One of the chief limitations for examining patterns of systemic interaction is statistical: Few studies of divorce have incorporated advances in dyadic data analysis to deal with statistical nonindependence between two individuals in the same family (cf. Kenny, Kashy, & Cook, 2006). Consequently, most research in this area has considered fathers and mothers separately, and only a few studies have addressed how ex-partners’ emotional states may influence each other (Bickerdike & Littlefield, 2001). One specific approach for dyadic data analysis, the actor-partner interdependence model (APIM; Kashy & Kenny, 2000; Kenny, 1996; Kenny & Cook, 1999), deals with problems of non-independence by treating individuals as nested under couples.

When estimated, the two main features of APIM are the actor and partner effects (see Campbell & Kashy, 2002). The actor effect reflects the association between one’s current state and his or her own scores on the same variable at an earlier occasion or on another variable. For example, an actor effect is reflected in the association between one’s report of coparenting conflict and his or her own report of nonacceptance of the divorce experience. The partner effect reflects the association between one’s current state and his or her ex-partner’s scores on the same or another variable. For example, this association is observed between a person’s own report of conflict and his or her ex-partner’s report of nonacceptance (controlling for his or her own nonacceptance scores). Actor and partner effects are of considerable interest for investigating the interpersonal dynamics of divorce, and analyses of this kind permit a rich specification of the ways in which the relational dynamics of divorce unfold over time. From this systemic perspective, one’s adjustment to divorce is a function not only of one’s own psychological experiences but also of their ex-partner as well.

The Present Study

The present study was guided by three goals. First, we sought to examine the immediate effects of the CMP and to extend the understanding of mediation and litigation by using the APIM (Kenny, 1996). The initial reports from the CMP were limited to basic group and gender comparisons (see Emery et al., 2005), and little attention was paid to predicting coparenting conflict using multiple variable models, although the available data provide a rich source for doing so. Thus, we sought to use APIMs in service of building a more detailed picture of parents’ conflict immediately following custody dispute resolution (the Time 1 assessment). Coparenting conflict and changes in conflict are the primary outcome variables across all analyses; the set of predictor variables includes custody group membership (mediation or litigation), gender, and nonacceptance of divorce. We expected parents to be sensitive to their ex-partners’ degree of nonacceptance of marital termination and predicted that a partner effect would be observed such that one’s own reports of coparenting conflict immediately following the custody settlement would be associated with their ex-partner’s level of nonacceptance. The second goal of the study was to determine whether the effect of nonacceptance on coparenting conflict depends on gender and group membership. We expected that the highest levels of coparenting conflict would be observed among litigation parents who were the most nonaccepting of the divorce. Support for this hypothesis would suggest that, relative to mediation, litigation fails to deal with important interpersonal dynamics that have consequences for the degree of coparenting conflict. The third major goal for the study was to evaluate whether and how these processes persist over time. Short- and long-term changes in conflict were examined as a function of nonacceptance, gender, and custody group membership. Because we had no a priori hypotheses about how early interpersonal dynamics would shape short-and long-term changes in conflict, analyses of the follow-up data proceeded in a largely exploratory fashion, with the main goal of assessing the persistence of the initial (Time 1) effects.

Method

Participants

Participants were 71 families who had requested a child custody hearing from a Juvenile and Domestic Relations District Court in Central Virginia between 1983 and 1986. Of these 71 families, 35 were randomly assigned to divorce mediation and 36 to traditional litigation to resolve their custody disputes in the initial study (Emery & Wyer, 1987). By virtue of seeking custody settlements, the sample represents a relatively conflicted group of divorced parents; approximately 10% of divorced parents seek litigation settlements to solve child custody disputes (Emery, 1994). The initial (Time 1) assessments took place an average of 5 weeks after dispute resolution, and a follow-up assessment (Time 2) was conducted approximately 13 months later. The long-term follow-up assessment took place an average of 12 years and 2 months after the initial custody decision (range = 10.6–13.3). The original sample consisted of 63 mothers and 59 fathers; the 12-year follow-up sample included 50 mothers and 43 fathers. Table 1 displays the individual- and dyad-level sample sizes at each occasion of measurement.

Table 1.

Individual and Dyadic Sample Sizes Over Time

Variable Time 1 Time 2 Time 3 Number of dyadsa Number of parentsb
Gender makeup
 Father 59 41 43
 Mother 63 49 50
Dyad-level makeup
 Complete dyads (data from both dyad members) 108 56 74
 One member (data from one dyad member) 28 68 38
 Neither member 6 18 30
Data over time
 Entirely complete (Time 1 → Time 2) 25 85
 Missing one parent at a single point in time (Time 1 → Time 2) 24
 Missing one parent at each point in time (Time 1 → Time 2) 11
Entirely complete (Time 1 → Time 3) 31 85
 Missing one parent at a single point in Time (Time 1 → Time 3) 17
 Missing one parent at each point in time (Time 1 → Time 3) 6
a

Multilevel analysis is capable of handling missing information by estimating all available data points; thus, from Time 1 to Time 2, data from 62 of 71 dyads were included in the analysis, although only 25 of these dyads provided fully complete data.

b

The number of parents who had complete data across each time interval. The same numbers do not indicate that the same parents had complete data at each of the intervals, only that the same number of parents had complete data from Time 1 to Time 2 and from Time 1 to Time 3, respectively.

At entry into the study, the average age for mothers in the sample was 28 years (range = 18–45); fathers’ average age was 31 (range = 20–47). Seventy-nine percent of the sample was Caucasian, and the remainder was African American. Reflecting the court’s population of clients, the sample was largely of low socioeconomic status. Eighty-seven percent of the men and 88% of the women were working or had last worked in clerical or blue-collar occupations; 10% of the men and 20% of the women were unemployed at the time of the study. At the 12-year follow-up, about half the men and women were either remarried or living with a partner at the time of the long-term follow-up interview. Both the men and the women had an average of 2.5 biological children (range = 1–7); children were an average of 4 years 10 months old (SD = 21 months) when custody was initially disputed. Two thirds of the target children were boys. At the Time 1 assessment, no significant differences were found between the mediation and litigation groups on any of these background characteristics.

Although a majority of the participants in the original sample participated in the 12-year follow-up, a substantial number did not. Selective attrition effects were examined by comparing 13-month follow-up participants with nonparticipants on the initial (Time 1) coparenting conflict and non-acceptance measures; analyses were conducted collapsing across custody groups and then within the mediation and litigation groups specifically. No selective attrition differences were observed across the short-term follow-up period. Similar analyses were conducted over the long-term by comparing 12-year follow-up participants with nonparticipants on the initial measures. Although no attrition effects were observed to be group specific, three findings were observed when collapsing across groups. For mothers, significant attrition group differences were found for fathers’ reports of early coparenting conflict and nonacceptance. Mothers who did not complete the long-term follow-up assessment were married to fathers who reported more conflict at the initial assessment than the ex-partners of mothers who remained in the sample, F(1, 57) = 4.04, p = .04. The opposite was true for fathers’ early nonacceptance. Mothers who remained in the study were married to fathers who reported more nonacceptance at the initial assessment than the ex-partners of mothers who left the study, F(1, 58) = 4.37, p = .04. Finally, at the initial assessment, fathers who remained in the study reported less coparenting conflict than fathers who did not participate in the follow-up, F(1, 58) = 3.75, p = .05.

Procedure

The details of the original field study involving random assignment to mediation/litigation conditions are described at length elsewhere (Emery, 1994; Emery et al., 2001, 1991). Parents were approached at random about either attempting mediation or participating in an evaluation of the court’s services (litigation control group). Following random assignment, families either proceeded through the usual court settlement processes or entered the mediation service. Mediation took place inside a courthouse, was conducted by one of four pairs of male and female comediators, and was limited to no more than six 2-hr sessions (average = 2.4 sessions). For the 12-year assessment, attempts were made to locate all participants from the original sample by using original telephone numbers; contacting neighbors, friends, and family members; and using Internet search directories. If participants were located and gave consent for their participation, then they received $50 compensation for a 2-hr interview session. Interviews were conducted in participants’ homes unless they requested to come into the laboratory or complete the questionnaire through the mail. Each phase of the research was approved by the University of Virginia Institutional Review Board for the Protection of Human Subjects; participants provided full informed consent prior to all study assessments.

Measures

The Acrimony Scale (AS)

The AS is a 25-item measure of coparenting conflict between separated or divorced parents that yields a single acrimony score, the mean of all items, with higher scores indicating greater conflict and coparenting difficulties (Shaw & Emery, 1987). Many of the AS items refer directly to problems in coparenting. Example items include “My child’s mother/father and I agree on discipline for him/her” and “Visitation is a problem between myself and my children’s mother/father.” Several items also refer to interparental conflict more explicitly, such as, “I have angry disagreements with my former spouse.” Throughout the present article, we refer to the AS as a measure of coparenting conflict, which best characterizes its emphasis on practical difficulties ex-partners have about child rearing in the wake of divorce. Responses are made on a 4-point Likert scale, rating the degree to which each statement characterizes the relationship ranging from 1 (almost never) to 3 (almost always). Items are worded in a counterbalanced format to control for response bias. This scale has high internal consistency (α = .86) and test–retest reliability (r = .88) (Shaw & Emery, 1987). AS reliabilities across the three measurement periods were adequate, ranging from .80 for fathers at the Time 1 assessment to .90 for mothers at the Time 1 assessment. Partial intraclass correlations (pICCs) were computed for each of the three conflict outcomes models (i.e., Time 1 conflict, changes from Time 1 to Time 2, and changes from Time 2 to Time 3). The pICC represents the proportion of the variance due to dyads from which the effects of the predictor variables are partialed (see Kenny et al., 2006); pICCs are a measure of nonindependence and quantify the extent to which variation occurs at the level of the dyad (vs. individual) within a given APIM. The pICCs were .30, .59., and .38, respectively, for each of the conflict outcome models, which indicates that individual-level observations are not independent and that the APIM would be an appropriate tool for addressing the within-dyad correlations observed in these data.

The Nonacceptance of Marital Termination (AMT) Scale

The AMT is an 11-item scale that taps a range of feelings about accepting the end of the marriage (Kitson, 1982). It yields a single scale, which is the mean of the items. Higher scores indicate greater nonacceptance. Example items include, “I find myself thinking a lot about my former spouse” and “I feel that I will never get over this separation/breakup.” Responses are made on a 4-point Likert scale, rating the degree to which each statement characterizes current feelings, ranging from 1 (not at all my feelings) to 4 (very much my feelings). AMT reliabilities across the three measurement periods were adequate, ranging from .78 for mothers at the Time 2 assessment to .94 for fathers at the Time 1 assessment.

Data Analysis

The main outcome variables in the present study were parent reports of coparenting conflict immediately following the initial dispute resolution, changes in conflict between the initial settlement and the 13-month follow-up (Time 2), and changes in conflict between the initial settlement and the 12-year follow-up (Time 3). For all analyses, a series of APIMs were specified using SAS Proc Mixed (Kenny et al., 2006), which is a widely used and flexible program for fitting multilevel or hierarchical linear models within the SAS programming package (Little, Lilliken, Stroup, & Wolfinger, 1996; Singer, 1998). APIMs can be specified within a two-level model (consisting of individuals nested under couples), a key feature of which is analyzing data from both members of the dyad (in this case, fathers and mothers) simultaneously. For each outcome, a series of increasingly complex regression models were compared in a forward stepwise fashion. The intervention group membership and actor gender variables were entered in the first step, followed by their interaction. The actor and partner nonacceptance variables were then entered into the model. The parameter estimate for the actor variable reflects the effect of one’s own nonacceptance on one’s own conflict scores, whereas the parameter estimate for the partner variable reflects the effect of a given parent’s ex-partner’s nonacceptance on one’s own conflict scores (controlling for one’s own nonacceptance scores). All significant parameters were retained, and interactions were then computed between the actor and partner nonacceptance variables and the two grouping variables (actor gender and intervention group membership). Gender and group membership were effect coded (−1 for mothers and mediation parents; 1 for fathers and litigation parents). Finally, in cases in which an interaction effect was of interest but one or both of the simple effects were not significant, the interactions were tested, and, if significant, the simple effects were controlled (Aiken & West, 1991). For the cross-time analyses, gain scores were created by subtracting participants’ follow-up conflict score from their initial Time 1 score on those variables. All variables were centered (based on their grand mean) prior to the creation of the gain score; thus, positive gain scores were associated with increases over time, whereas negative gain scores were associated with decreases over time. In the short-term change model, actors’ initial conflict score was controlled; in the long-term change model, actors’ short-term gain score was controlled.

Results

The first series of APIMs investigated coparenting conflict immediately after the initial dispute settlement as a function of nonacceptance, custody group membership (mediation or litigation), and actor gender. The unstandardized parameter estimates from this model are displayed in the top portion of Table 2. As shown, actor nonacceptance was negatively associated with coparenting conflict. Participants who reported greater nonacceptance of the end of the marriage reported lower coparenting conflict. Coparenting conflict also was negatively associated with ex-partner nonacceptance, and this trend approached significance. After controlling for one’s own level of nonacceptance, actors tended to report higher conflict when their ex-partners reported lower levels of nonacceptance (i.e., more acceptance). The effect of ex-partner nonacceptance on conflict also depended on actor gender, and the simple slopes for the interaction are displayed in Figure 1. Given the multilevel framework of the APIM, we probed this interaction using the procedures and tools recently described by Preacher, Curran, and Bauer (2006), which are conceptually similar to the well-known simple slopes approach discussed by Aiken and West (1991). Post hoc probing revealed that the differences between mothers at low and high levels of nonacceptance was not different from zero, t(103) = 0.66, p = .50, whereas the simple slope for fathers was significantly different at low and high levels of nonacceptance, t(102) = −2.19, p = .02. Fathers with ex-partners reporting low levels of nonacceptance (i.e., more acceptance) reported higher levels of conflict than fathers whose ex-partner reported greater nonacceptance.

Table 2.

Unstandardized Parameter Estimates for Actor–Partner Interdependence Models Predicting Coparenting Conflict at Each Wave of Assessment

Model outcome Estimate (B) SE 95% confidence interval of B t
Coparenting conflict (Time 1 [T1])
 A-AMT-T1 −.16 .07 −.30, .02 −2.36*
 P-AMT-T1 −.14 .07 −.28, .00 −1.86
 A-Gender −.10 .06 −.22, .02 −1.52
 A-Gender × P-AMT-T1 −.14 .06 −.28, .01 2.10*
Short-term changes in coparenting conflict (Time 1 → Time 2 [T2])
 A-Con-T1 −.44 .07 −.58, −.30 −6.74***
 Group .21 .07 .07, .35 2.78**
 ΔA-AMT (T1/T2) −.13 .06 −.25, −.01 −2.03*
 Group × ΔA-AMT (T1/T2) .14 .06 .26, .02 2.10*
Long-term changes in coparenting conflict (Time 1 → Time 3)
 ΔA-Con (T1/T2) .30 .10 .10, .50 2.98**
 Group −.02 .10 −.08, .22 0.88
 Group × ΔA-Con (T1/T2) .23 .10 .03, .43 2.41*
 A-Gender .09 .10 −.11, .29 0.41
 P-AMT-T1 −.01 .11 −.23, .21 0.98
 A-Gender × P-AMT-T1 .18 .10 −.02, .38 1.72

Note. Degrees of freedom range from 49 to 115 for these analyses. Gender and group membership were effect coded (−1 for mothers and mediation parents; 1 for fathers and litigation parents). A = Actor variable; AMT = acceptance of marital termination; P = Partner variable; Gender = actor gender; Con = coparenting conflict (Acrimony Scale); Group = dispute resolution group.

p < .10.

*

p < .05.

**

p < .01.

***

p < .001.

Figure 1.

Figure 1

Simple slopes illustrating the interaction of Time 1 ex-partner nonacceptance by actor gender on initial coparenting conflict. Low and high levels of ex-partner nonacceptance are displayed for individuals scoring −1 and + 1 standard deviation above and below the mean of ex-partner nonacceptance, respectively.

To examine change from Time 1 to Time 2 (the one follow-up), a conflict gain score was used as the primary outcome variable, with negative scores indicating a decrease in conflict over time and positive scores indicating an increase; all analyses controlled for conflict at Time 1. Significant effects from the short-term conflict change model are presented in the middle portion of Table 2. (Although not reported in the table, it is notable that the intercept for the overall model was not different from zero, t[41] = 0.42, p = .75, indicating that the average dyad evidenced no change in conflict over the 13 months between the first two measurements once actors’ initial conflict was controlled.) As shown, actors reporting high initial levels of conflict reported greater decreases in the year after dispute settlement. Custody group membership was significantly associated with changes in conflict. In particular, participants in the litigation group evidenced short-term increases in conflict, whereas the mediation group evidenced short-term decreases in conflict. A significant effect also was observed for short-term changes in actor nonacceptance. Actors who reported increases in nonacceptance reported decreases in conflict. Finally, this simple main effect for actor nonacceptance on conflict was qualified by an interaction with custody group membership, which is displayed in Figure 2. Post hoc probing of the interaction revealed that the effect of litigation group membership on changes in conflict was similar across all levels of changes in nonacceptance, t(76) = −1.23, p = .22. In contrast, the effect of mediation membership on changes in conflict varied as a function of changes in nonacceptance, t(55) = −3.33, p = .04. Mediation parents reporting decreases in nonacceptance (i.e., increases in acceptance of the end of marriage) evidenced no changes in conflict, whereas mediation parents reporting increases in nonacceptance evidenced the largest decreases in conflict.

Figure 2.

Figure 2

Simple slopes illustrating the main effect of custody group membership and the interaction of short-term changes in actor nonacceptance by custody group membership on short-term changes in coparenting conflict. Low and high levels of changes in actor nonacceptance are displayed for individuals scoring −1 and + 1 standard deviation above and below the mean of changes in actor nonacceptance, respectively.

The analysis of long-term change in conflict proceeded in a manner similar to the analysis of short-term change. The significant parameter estimates for the long-term change model are displayed in the bottom portion of Table 2. As with the short-term conflict change model, the intercept for the overall model was not different from zero, t(37) = 0.43, p = .66, indicating that the average dyad evidenced no long-term change in conflict once short-term change was controlled. Participants who reported short-term decreases in conflict reported greater decreases over the decadelong follow-up period. Although custody group membership did not predict long-term changes in conflict, mediation and litigation membership moderated the effects of early conflict on later conflict, and the simple slopes for this interaction are displayed in Figure 3. Post hoc probing of this interaction revealed that the effect of mediation on long-term changes in conflict was not different from zero across different levels of short-term change in conflict, t(54) = 0.79, p = .43. In contrast, the effect of litigation on long-term changes in conflict varied across levels of short-term changes in conflict, t(50) = 5.90, p < .0001. Parents in the litigation group who reported short-term increases in conflict reported the greatest long-term increases in conflict, whereas litigation parents who reported short-term decreases in conflict reported the greatest long-term decreases in conflict.

Figure 3.

Figure 3

Simple slopes illustrating the interaction of short-term changes in actor conflict by custody group membership on long-term changes in conflict. Low and high levels of short-term changes in actor conflict are displayed for individuals scoring −1 and + 1 standard deviation above and below the mean of short-term changes in conflict, respectively.

We also evaluated whether the initial Partner Nonacceptance × Actor Gender interaction effect predicted long-term changes in conflict. Controlling for the simple effects of each of these variables on long-term changes in conflict, the interaction term approached significance. Consistent with the post hoc probing of this interaction at the Time 1 assessment, an inspection of the simple slope revealed that the effect operated differently across levels of partner non-acceptance for fathers, t(69) = 1.68, p = .09, but not for mothers, t(69) = −0.83, p = .41. However, in contrast to what was observed at the initial Time 1 assessment, the direction of the simple slope effect for fathers was positive, indicating that the greatest increases in fathers’ conflict over the decadelong follow-up period were observed when mothers reported high levels of nonacceptance at Time 1.

Discussion

The APIM analyses conducted in the present article revealed a number of important findings not previously documented in our longitudinal study of couples randomized to different forms of custody dispute resolution. In particular, three of the findings allow for a more systemic view of the relationship between former partners and are worthy of further consideration. First, partner acceptance of the end of the marriage predicted increased actor reports of conflict, especially among fathers. This finding may seem puzzling, but, in fact, it is quite consistent with our conceptualization of a key aspect of parents’ relationship dynamics in custody disputes. Second, coparenting conflict decreased for mediation parents and increased for adversary settlement (litigation) parents, a finding that underscores critical, alternative interpretations of the major results of this study reported here and elsewhere. Third, coparenting conflict decreased notably at 12-year follow-up for a small subgroup of families for whom adversary settlement also lowered conflict in the short term. This last finding highlights the need to consider a taxonomy of custody conflicts.

We predicted that, controlling for one’s own level of divorce nonacceptance, partner nonacceptance would predict one’s own initial level of conflict. Findings supported this hypothesis, and the effect was particularly strong for fathers. For fathers, there was a strong association between their own conflict and their ex-partners acceptance (controlling for their own levels of acceptance). Although this may seem paradoxical, Emery and colleagues (Emery, 1994; Emery et al., 2005; Sbarra & Emery, 2005) have speculated that divorce-related custody conflict can reflect, in part, attachment-related protests over the end of marriage. Some custody disputes can be viewed as desperate reunion behavior, an attempt to hold on to the former spouse (and the marriage) by making the consequences of leaving especially dire (losing your children). Given the traditional gender roles in the families in this sample, the threat of losing one’s children may be especially great for women. This may explain why fathers in this study were both less accepting of the end of the marriage and why their partner’s acceptance was linked with fathers’ increased coparenting conflict. The present results provide the first empirical support for our previous speculations, and they are consistent with the findings of Bickerdike and Littlefield (2001), who observed that discrepancies in parents’ acceptance of divorce stalled mediation negotiations.

In the year following dispute resolution, mediation parents reported decreased conflict, whereas litigation parents reported increased conflict. Because the CMP was a true field experiment, we can conclude that mediation caused decreases in conflict and that litigation caused increases in conflict. Given the increases in power gained by considering mothers and fathers together within a single APIM, these findings extend those reported in the first CMP follow-up report, which failed to demonstrate a significant Time × Group effect on coparenting conflict (Emery et al., 1994). More important, initial changes in conflict have long-term implications: Parents who reported the greatest short-term decreases in conflict reported the greatest long-term decreases in conflict (and vice versa). These findings also illustrate alternative interpretations of the major results of the CMP reported here and elsewhere: Is mediation a good choice for families because it changes relationships for the better? Or, does the true benefit of mediation come from protecting parents from the potentially harmful effects of litigation (cf. Beck & Sales, 2001; Sbarra & Emery, 2005)? The present results suggest that both explanations have merit. Mediation helps to reduce coparenting conflict, and it also reduces the likelihood that parents will enter adversary legal processes that, on average, increase coparenting conflict.

Other analyses revealed that short-term changes in conflict interacted with custody group membership to predict changes in conflict up to 12 years later. This effect was observed for parents in litigation but not in mediation. Litigation parents who reported short-term decreases in conflict (from the time of the dispute to 13 months later) evidenced the greatest decreases in long-term conflict. In contrast, litigation parents who reported short-term increases in conflict evidenced the greatest increases in long-term conflict. We interpret this finding as suggesting the need, in future research, to consider a taxonomy of conflicts in custody disputes. Many conflicts (we would hypothesize most) can be viewed as a form of attempted reunion behavior. However, other disputes are “real” in the sense that they involve substantive issues that the legal system can resolve, for example, whether a parent and a child are allowed to relocate a significant distance from the other parent. However one conceptualizes the nature and kind of custody disputes (and we believe there are several further categories of dispute), it is essential to interpret the interaction between group membership and short-term changes in conflict in light of the short-term main effect for group membership. The average litigation parent evidenced a short-term increase in coparenting conflict, which was caused by their experience with the adversary process. From these results, we would expect nonadversarial approaches to work best with the majority of parents, but we also see a role for adversarial approaches to dispute settlement and expect positive effects over time, albeit in a minority of cases.

Some other findings merit our further consideration. The main effect of dispute resolution group membership on short-term changes in coparenting conflict was qualified by a significant Group × (changes in) Nonacceptance interaction. For mediation but not litigation parents, greater short-term increases in nonacceptance were associated with decreases in conflict. Whereas conflict may reflect continued attachment (nonacceptance) in the immediate context of a custody dispute, the causal direction of this effect may be reversed over time; in this sense, parental cooperation that is a goal of mediation may lead to pining for an ex-partner and wondering why the separation occurred in the first place. In short, the success of mediation (decreases in conflict) may come at the cost of some longing (increases in nonacceptance of divorce; see Sbarra & Emery, 2005).

Over the long-term follow-up period, there was a trend for fathers whose ex-partners were the most nonaccepting (immediately after the custody settlement) to evidence greater increases in conflict than fathers whose ex-partners were less nonaccepting of the divorce. All of the long-term effects need to be considered in light of the long-term selective attrition analyses, and this is especially relevant for examining the persistence of the effects observed at the initial assessment. There is clear evidence that the nature of the long-term follow-up sample differed from the initial Time 1 sample. Mothers who were lost to follow-up were married to fathers who reported higher initial rates of conflict and lower nonacceptance than the partners of mothers who remained in the sample; fathers who were lost to follow-up reported more initial conflict than fathers who were retained over time. The observed trend toward an interaction effect may be explained by the loss of the most conflicted fathers over time.

Although this article provides new findings on the short-and long-term interpersonal dynamics of divorce, the effects should be considered in light of several limitations. First, as noted above, the selective attrition effects fundamentally changed the nature of the long-term follow-up sample (compared with the initial sample). This limitation also can be considered a substantive finding on the interpersonal dynamics of divorce. Mothers lost to follow-up were married to the most conflicted fathers. One way of dealing with the acrimony of divorce is to move away, and the selective attrition effects reflect the reality of postdivorce relationships. Second, even though 72% of fathers and 79% of mothers involved in the initial sample were retained over the decadelong follow-up period, complete data on both members of the former couple at multiple time points was somewhat limited. One of the unique features of multilevel modeling is that maximum likelihood estimation procedures use all of the available data to fit a model and can handle missing data at Level 1 (in the APIM, the individual parent). Thus, although missing data from both members of the dyad limit statistical power to detect longitudinal effects, the APIMs make use of all individuals observed at each time point. Finally, it is important to recognize that the time frame and context of this study may limit its external validity. The CMP was conducted in the late 1980s, a time when the tender years presumption (i.e., that young children are best served in the custody of their mothers during the so-called tender years of development) dominated the legal landscape (Emery et al., 1991). Thus, these findings should be viewed within the context of changes in child custody litigation that now provide fathers with a more equitable voice in settlement decisions (Beck & Sales, 2001).

Conclusions

It is common practice in the divorce adjustment literature to study individuals as isolated entities whose trajectories of recovery are influenced almost exclusively by their own psychological adjustment and perceptions of their relationship with their former partner. The results of the present study suggest that researchers look much more carefully at the patterns of interpersonal influence operating after divorce to understand the emergent coparenting conflict and its persistence over time. The present article is one of the first long-term studies of coparenting conflict in the wake of divorce and the first to examine systemic patterns of interpersonal influence between divorcing parents. The primary findings were that (a) fathers whose ex-partners were more accepting of the divorce reported the highest initial rates of coparenting conflict; (b) custody mediation caused decreases in short-term conflict, whereas custody litigation caused increases in short-term conflict; (c) mediation parents who reported the greatest short-term increases in non-acceptance also reported the greatest short-term decreases in conflict; and, (d) litigation parents who reported early declines in conflict reported the greatest long-term decreases in conflict, whereas litigation parents who reported the greatest short-term increases in conflict reported the greatest long-term increases in conflict. Overall, the results suggest that using a systemic framework on adjustment to marital dissolution provides a useful lens for developing a better understanding of how individuals, ex-partners, and parents renegotiate postdivorce relationships.

Acknowledgments

This research was supported by grants from the William T. Grant Foundation awarded to Robert E. Emery.

Contributor Information

David A. Sbarra, Department of Psychology, University of Arizona

Robert E. Emery, Department of Psychology, University of Virginia

References

  1. Aiken LS, West SG. Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage; 1991. [Google Scholar]
  2. Amato PR. The consequences of divorce for adults and children. Journal of Marriage and the Family. 2000;62:1269–1287. [Google Scholar]
  3. Beck CJA, Sales BD. Family mediation: Facts, myths, and future prospects. Washington, DC: American Psychological Association; 2001. [Google Scholar]
  4. Bickerdike AJ, Littlefield L. Divorce adjustment and mediation: Theoretically grounded process research. Mediation Quarterly. 2001;18:1–23. [Google Scholar]
  5. Campbell L, Kashy DA. Estimating actor, partner, and interaction effects for dyadic data using Proc Mixed and HLM: A user-friendly guide. Personal Relationships. 2002;9:327–342. [Google Scholar]
  6. Emery RE. Renegotiating family relationships: Divorce, child custody, and mediation. New York: Guilford Press; 1994. [Google Scholar]
  7. Emery RE. The truth about children and divorce: Dealing with the emotions so you and your children can thrive. New York: Viking/Penguin; 2004. [Google Scholar]
  8. Emery RE, Laumann-Billings L, Waldron M, Sbarra DA, Dillon P. Child custody mediation and litigation: Custody, contact, and coparenting 12 years after initial dispute resolution. Journal of Consulting and Clinical Psychology. 2001;69:323–332. doi: 10.1037//0022-006x.69.2.323. [DOI] [PubMed] [Google Scholar]
  9. Emery RE, Matthews SG, Kitzmann KM. Child custody mediation and litigation: Parents’ satisfaction and functioning one year after settlement. Journal of Consulting and Clinical Psychology. 1994;62:124–129. doi: 10.1037//0022-006x.62.1.124. [DOI] [PubMed] [Google Scholar]
  10. Emery RE, Matthews SG, Wyer MM. Child custody mediation and litigation: Further evidence on the differing views of mothers and fathers. Journal of Consulting and Clinical Psychology. 1991;59:410–418. doi: 10.1037//0022-006x.59.3.410. [DOI] [PubMed] [Google Scholar]
  11. Emery RE, Sbarra DA, Grover T. Divorce mediation: Research and reflections. Family and Conciliation Courts Review. 2005;43:22–37. [Google Scholar]
  12. Emery RE, Wyer MM. Child custody mediation and litigation: An experimental evaluation of the experience of parents. Journal of Consulting and Clinical Psychology. 1987;55:179–186. doi: 10.1037//0022-006x.55.2.179. [DOI] [PubMed] [Google Scholar]
  13. Hetherington EM, Bridges M, Insabella GM. Five perspectives on the association between marital transitions and children’s adjustment. American Psychologist. 1998;53:167–184. doi: 10.1037//0003-066x.53.2.167. [DOI] [PubMed] [Google Scholar]
  14. Kashy DA, Kenny DA. The analysis of data from dyads and groups. In: Reis HT, Judd CM, editors. Handbook of research methods in social and personality psychology. Cambridge, England: Cambridge University Press; 2000. pp. 451–477. [Google Scholar]
  15. Kelly JB. A decade of divorce mediation research: Some answers and questions. Family and Conciliation Courts Review. 1996;34:373–385. [Google Scholar]
  16. Kenny DA. Models of non-independence in dyadic research. Journal of Social Personal Relationships. 1996;13:279–294. [Google Scholar]
  17. Kenny DA, Cook W. Partner effects in relationship research: Conceptual issues, analytic difficulties, and illustrations. Personal Relationships. 1999;6:433–448. [Google Scholar]
  18. Kenny DA, Kashy DA, Cook WL. Dyadic data analysis. New York: Guilford Press; 2006. [Google Scholar]
  19. Kitson GC. Attachment to the spouse in divorce: A scale and its application. Journal of Marriage and the Family. 1982;44:379–391. [Google Scholar]
  20. Little RC, Lilliken GA, Stroup WW, Wolfinger RD. SAS system for linear mixed models. Cary, NC: SAS Institute; 1996. [Google Scholar]
  21. Myers S, Gallas G, Hanson R, Keilitz S. Divorce mediation in the states: Institutionalization, use, and assessment. State Court Journal. 1988;12:17–25. [Google Scholar]
  22. Preacher KJ, Curran PJ, Bauer DJ. Computational tools for probing interaction effects in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of Educational and Behavioral Statistics. 2006;31:437–448. [Google Scholar]
  23. Sbarra DA, Emery RE. In the presence of grief: The role of emotion in contemporary divorce mediation. In: Fine MA, Harvey JH, editors. Handbook of divorce and dissolution of romantic relationships. Mahwah, NJ: Erlbaum; 2005. pp. 553–573. [Google Scholar]
  24. Shaw D, Emery RE. Parental conflict and the adjustment of school-aged children whose parents have separated. Journal of Abnormal Child Psychology. 1987;15:269–281. doi: 10.1007/BF00916354. [DOI] [PubMed] [Google Scholar]
  25. Singer JD. Using SAS Proc Mixed to fit multilevel models, hierarchical models, and individual growth models. Journal of Educational and Behavioral Statistics. 1998;23:323–355. [Google Scholar]
  26. Wallerstein J, Lewis J, Blakeslee S. The unexpected legacy of divorce: A 25 year landmark study. New York: Hyperion; 2000. [Google Scholar]

RESOURCES