DONNA RUANE MORRISOX Georgetown University

MARY Jo COIRO Johns Hopkins University*

Parental Conflict and Marital Disruption:

Do Children Benefit When

High-Conflict Marriages Are Dissolved?

A million children experience divorce each .vear,and some policymakers argue for policies thatwould muke it more difficult for parents to divorce.Ho\+!ever: being e.rposed to a high degree qf nturi-tal conjlict ittrs been siu,nn to place children at riskfor a varien of problems. Using mother-child datufrom the Nationul Longitudinul Survey oj’ Youth(NLSY) and a prospective design, this researchesplores two questions: Do the e8ect.s of maritaldisruption on child wel[-being varq’ for childrenwhose parents leave high-conflict marriages ver-sus low-conjlict marriages? HOW do children farewhen their high-conjlict parents remain together?WeJnd that sepamtion and divorce are associ-ated brith increases in hehaCor problems in chil-dren, regardless of the level of conflict betweenparents. However. in marriages that do not breakup, high levels of muritul conjlict are associatedwith even greater increases in children ‘s behaviorproblems.

Georgetown Public Policy Institute and Department of Demog-

raphy. Georgetown University, 3600 N Street, NW. Washing-

ton, DC 2OOO7 Imorrisod@gunet.georgetown.edu).

*Department of Health Policy and Management. Johns Hop-

kins University School of Public Health, 624 North Broadway.

Baltimore. MD 21205.

High rates of divorce have prompted many ob-servers, largely out of concern for the children in-volved, to advocate measures to keep marriagestogether. The accumulated evidence suggests thatchildren, particularly boys, not only have problemsin the immediate aftermath of marital disruption,but have difficulties that persist into adulthood aswell (e.g., Amato, 1994: Amato & Keith, 1991;Cherlin. Chase-Lansdale, & McRae, 1998). Alongwith these findings, however, there is evidencethat being exposed to a high degree of conflict be-tween married parents also places children at riskfor a variety of problems. Consequently, the diffl-culty for parents, legal professionals, and policy-makers weighing what is best for the child lies indetermining whether the effects of divorce will beless deleterious than the effects of remaining withtwo parents in a disharmonious relationship.

Another question is whether the effects ofmarital disruption on child well-being vary ac-cording to the level of marital conflict that childrenexperience before the separation. If the break-uprepresents an exit from severe marital disharmony,children may make an easier adjustment than ifthe separation was unexpected. Moreover. childrenremoved from intense parental conflict may farebetter than those whose high-conflict parents re-main together. In a recent study Amato. Loomis,and Booth (1995) used longitudinal data from a

626 Journal of Marriage and the Family 6 1 (August 1999): 626-637

R e p r o d u c e d w i t h pe?mission o f C o p y r i g h t o w n e r . F u r t h e r r e p r o d u c t i o n p r o h i b i t e d w i t h o u t pemirsion.~

Coqflicr und Mm-ital Disruption

study of marital instability over the life courseand documented that the well-being of young adultsafter a parental divorce was highest among thosewho experienced high levels of contlict before thedisruption and lowest among those who experi-enced less conflict before the divorce. Amato andhis colleagues argued that when young people arenot aware of the level of their parents’ unhappiness,divorce is likely to be unanticipated and unwel-comed by the youth.

We attempt to build on the lindings of Amatoet al. (1995) by examining children who wereyounger when their parents divorced and whoseparents divorced more recently. Children in oursample ranged in age from 4 to 9 years in 1988(the average child is 6 years old) when all of thechildren were in married. two-parent families.Controlling for children’s preexisting levels of be-havior problems. we examine their mottler-ratedbehavior problems scores 6 years later. By then.some of the children’s parents had separated ordivorced. (The average time since disruption is3.4 years.) Our analysis addresses two main ques-tions: Do the consequences of marital disruptionfor children’s behavior problems vary, dependingon the level of marital conflict that preceded thedisruption? That is, do children benefit whenhigh-conflict marriages are dissolved, but do theyshow elevated problems when the couple was lessconflictual before separation? And how do chil-dren fare when their high-conflict parents remaintogether?

BACKGROUND

Accumulated evidence suggests that children whoexperience divorce and the associated disruptionsin parent-child relationships, living arrangements,and economic circumstances fare less well thanchildren in two-parent families who do not divorce(See Amato & Keith, 1991.) Although there isconsiderable heterogeneity in outcomes, childrenwhose parents separate or divorce are, on average,more likely to exhibit problematic behavior, havepoorer mental health and academic performance,and have more social difficulties and lower self-concepts than children whose parents remain mar-ried (Amato. 1994). These negative effects havebeen documented in the immediate aftermath ofdivorce (e.g., Hetherington, Cox. & Cox. 1982)and many years later (e.g.. Cherlin et al., 1998;Cherhn, Kiernan, & Chase-Lansdale, 1995: Kier-nan, 1992: Zill, Morrison, & Coiro. 1993).

627

Alternatively. several prospective studies havesuggested that many of the negative effects attrib-uted to divorce are apparent before the actual phys-ical separation (e.g.. Baydar. 1988; Block, Block.& Gjerde, 1986: Cherlin et al., 1991: Doherty &Needle, 199 I ; Morrison & Cherlin. 1995). Promi-nent among the factors that researchers assumeaffect children’s well-being prior to marital disrup-tion is interparental conflict. Parental conflict ispositively associated with a variety of indicatorsof children’s maladjustment. including conductdisorders, anxiety, and aggression. (See Emery.1982, 1988; Grych & Finchman. 1990. for re-views.) Exposure to conflict has been shown to in-fluence children directly. Witnessing adult angeris physiologically and affectively stressful forchildren, and exposure to conflict has been shownto influence children indirectly through its effecton parenting and parents’ psychological welt-being (e.g., El-Sheikh, Cummings, & Goetsch.1989: Emery, Fincham. & Cummings. 1992).Some researchers have shown that the effects ofparental conflict can be more harmful to childrenthan parental absence through death or divorce(Emery, 1982: Jekielek. 1998: Mechanic & Hansell,1989; Peterson & Zill, 1986). For example, in alongitudinal study of adolescents. Mechanic andHanscll found that those in high-conflict, mat-Tiedfamilies had significantly poorer adjustment thanthose in low-conflict, divorced families. Similarly,Slater and Haber (1984) found that adolescentsfrom high-conflict families reported lower self-esteem and greater anxiety, regardless of whetheror not their parents had divorced or remainedmarried.

However, the question of which is more detri-mental for children-divorce or exposure to mari-tal conflict-has not been satisfactorily resolved.For example, focusing on children in married, two-parent families and examining the consequences ofdivorces in a subsequent 2-year interval. Morrisonand Cherlin (199.5) found that boys in maritallydisrupted families experienced increases in behav-ior problems and decreases in reading achieve-ment. For girls. there were no statistically signift-cant effects of disruption. Because the adverseeffects of marital dissolution held for boys, evenwhen a predisruption counterpart to the dependentvariable was held constant, the results suggestedthat most of the explanation for the postdisruptiondifficulties could be attributed to the experience ofthe disruption itself. However, it may be that ex-plicit measures of predisruption marital conflictwould have altered the apparent effect of divorce.

Reproduced with permission of Copyright wrier. Further reproduction prohibited without permlrsion.-

628

An alternative to the hypotheses that eithermarital disruption or marital conflict is the keyexplanatory variable is that there is an interactionbetween the two. The interaction hypothesis arguesthat the effects of divorce vary according to thelevel of predisruption conflict that the child expe-rienced. Evidence for an interaction between theeffects of marital disruption and conflict can befound in the recent work of Amato et al. (1995).They showed that when levels of marital conflicthad been high, young adults whose parents haddivorced up to 12 years previously actually showedmore favorable outcomes than their counterpartswith continuously married parents: suggestingthat the youth were better off when removed froma highly negative situation- However, youth inlow-conflict homes prior to disruption showednegative responses to divorce, presumably becausethey were surprised by and less well prepared fortheir parents’ break-up.

Although provocative, Amato et al.‘s (1995)work leaves several questions unanswered. First,with its focus on outcomes among youth 19 yearsold or older. Amato and his team did not addresswhether the effects of divorce would also vary ac-cording to the level of predisruption contlict foryounger children, a group for whom the negativeeffects of marital disruption appear to be particu-larly salient (e.g.? Block et al., 1986: Morrison dtCherlin. 1995; Zill et al., 1993). Moreover. it maybe that children in newly disrupted families facechallenges (e.g., new living arrangements, changesin neighborhoods or schools. parental quarrelsover the division of income and property, and cus-tody battles) that place them at risk for difficultiesin adjustment, regardless of how well their par-ents got along before they separated. Finally,Amato et al. examined four measures of psycho-logical well-being (psychological distress. overallhappiness, marital happiness, and social resources)as a function of marital disruption and conflict butdid not include youths’ problematic behaviors, anoutcome that has been consistently and negativelyassociated with divorce (e.g., Amato & Keith,1991) and intcrparental conflict (e.g., Emery.1982).

Jekielek (199X), in an extension of Amato etal.‘s (1995) study and of the original analyses onwhich our article is based (Morrison, Coiro, & Blu-menthal. 1994) also examined interactions betweenmarital disruption and marital conflict using datafrom the NLSY. Jekielek focused on internalizeddimensions of children’s behavior problems (anx-iety and depression-withdrawal), as well as the role

of time since disruption in children’s adjustment.She found main effects of disruption and conflict onchildren’s adjustment, as well as interactions indi-cating that children who remain in high-conflictenvironments have more internalizing problemsthan children whose high-conflict parents divorce.However, Jekielek examined a shorter postdisrup-tion interval than we do and only a single measureof conflict.

We test two hypotheses related to the associa-tions of marital disruption, predisruption conflict(both frequency and breadth), and children’s behav-ior problems.

Hypothesis I: When predisruption maritalconflict is high. children whose parents di-vorce will exhibit a decrease in behaviorproblems, whereas children whose parentshave low Icvels of marital conflict will ex-perience an increase in behavior problemsafter divorce.

Hypothesis 3: Children whose high-contlictparents remain together will show greaterincreases in behavior problems than thosewhose high-conflict parents divorce.

METHODS

Data

WC use the 19881994 waves of the NLSY-ChildSupplement, a large longitudinal data set, to explorethe relation between predisruption marital conflictand child well-being after marital disruption. TheNLSY began as a nationally representative longi-tudinal survey of American youth who were14-21 years old in 1979. These youth have beenfollowed annually since 1979. From the 1979 sur-vey to the 1994 survey, excluding individuals whowere explicitiy dropped from the study, the aver-age respondent completed 14.8 out of 16 inter- /views. In 1986. when the participants were 21-29years old, the Center for Human Resource Re-search expanded its data.collection to include thechildren of female respondents. Biannually sincethat date, information has been collected frommothers about their children, and a substantialbattery of assessments has been administered di-rectly to the children. Roughly 60% of the womenin the NLSY youth cohort had become mothersby 1988 (Mott & Quinlan. 1991) the start of theinterval examined in the study presented here.

reduced with pern~Ls.sion of Copyright owner. Further rep

A noteworthy advantage of our study is thatthe disruptions we examine are contemporary. Wehave a unique opportunity to examine the conse-quences for children of divorces that occurred re-cently-between 1988 and 1994--when rates ofdivorce in the nation were high.

Despite the advantages of the NLSY child datafor a longitudinal analysis of the consequences ofmarital conflict and divorce. there are also somelimitations. First, the children in the NLSY arenot fully representative of all children in their agegroup because the NLSY is a random sample ofyoung women. not of children. The children in-cluded in our study represent a cross-section ofchildren born to a sample of women who werebetween the ages of 2 I and 28 years on January 1,1986. Therefore, the children tend to be born tosomewhat younger and disadvantaged mothers(especially the oldest children), The average ageat first birth for mothers in our sample is 20.3years. The national average age was 23.7 years inthe late 1980s. The nature of marital conflict andthe process and effects of disruption may differfor children born to comparatively older mothers.A second disadvantage of the NLSY is that mea-sures of marital quality and contlict are onlyavailable in 1988, 1992. and 1994, precludinganalysis of divorces before 1988.

Purticipan ts

Participants in our study are limited to 727 chil-dren in undisrupted families with two biologicalparents in 1988 who had no missing values on theBehavior Problems Index in 1994. We excludechildren born out of wedlock, children whose par-ents died or divorced before 1988, and thosewhose primary residence was not with their mother.To maximize our sample size? we include siblingsin our analysis. Nineteen Percent (135 cases) ofthe children in our sample have siblings. and wediscuss the statistical implications of jncludingthem in the analysis section. The average child is 6years old (SL) = l..i), and 50% are male. We limitour analysis to children aged 4-9 years in I988 forseveral reasons. First, the assessments availablefor younger children have less satisfactory psycho-metric properties. Second. older children in theNLSY-Child Supplement are disproportionatelyborn to teenage mothers. Finally, we are particu-larly interested in how marital disruption and pre-disruption conflict affects children who are rela-tively young at the time of the break-up. Wecontrol for child’s age and mother’s age at firstbirth in all multivariate models.

629

Behavior problem. We examine the effects ofmarital dissolution on the 1994 Behavior ProblemsIndex, which measures mothers’ reports of thefrequency and types of behavior problems mani-fested in the last 3 months by their children aged3 years or older. The Behavior Problems Indexcontained in data from the NLSY-Child Supple-ment comprises 32 iterns developed by Petersonand Zill (.Zill. 1990). primarily from Achenbach’s(,l978) Child Behavior Checklist. The 32 itemswere selected for inclusion in the NLSY test bat-tery because of their demonstrated ability to dis-tinguish children who were referred for psycho-logical treatment from typical children. TheBehavior Problems Index measures both extemal-izing and internalizing behaviors (for example,“he/she bullies or is cruel or mean to other,” and“he/she is stubborn. sullen or irritable”). Zill stan-dardized the index separately for boys and girlsusing the 1981 National Health Interview Survey-Child Supplement of 15,000 children. These normsthen were used by the NLSY staff to convert rawBehavior Problems Index scores to standardscores with a mean of roughly 100 and a standarddeviation of about IS. We use these same-sexstandard scores in our analyses. Higher scores on-the index indicate more behavioral problems.

In order to consider the extent to which sepa-ration or divorce changes child well-being. we in-clude a lagged version of our dependent variable.children’s behavior problems, drawn from a sur-vey taken when all families were still together.Using pretest and posttest measures in regressionanalysis to examine change is considered superiorto calculating raw gain scores due to better relia-bility of measures. (See Hummel-Rossi & Wein-berg, 1975.) To lessen the chance that it would beconfounded with current levels of conflict, wedrew the 1986 Behavior Problems Index whenpossible. For children who were 2 years old in I986(too young for the assessment), however. it wasnecessary to take their 1988 measure.

Means and standard deviations of all variablesare included in Table I. Measurement of key in-dependent variables is described below.

Muritcrl disruption. Our measure of family disrup-tion is a dummy variable that indicates the experi-ence of either separation (without reunion) or di-vorce between the 1988 and 1994 interviews. Weused detailed marital histories provided by themothers as part of the NLSY 1979-1994 surveys

e-c 3?mF-02015-00008

I Reproduced with permission of Copyright owner. Further reproduction prohLb,.tod without permission.

630 Journal of Marriage and the Family

TABLE I. MEANS. PERCEXTAGES, AND ST.&WARD DEVIATIOXS FOR INDEPENDENT VARIABLES IN OLS MODELS

Independent Variable M S D

Mother and family characteristicsMother’s age at first birth

Mother’s AFQT score

Years of mother’s education in 1994

Family income in prior year ( 1993)

Number of mother’s children in household (1994)

Below poverty threshold in prior year (1993)Total HOME score prior to disruption

Child characteristics

Male

Black

H i s p a n i cBirth order

Birth weight (ounces)

Age in 1988 (months)

Behavior Problems Index score before disruption

Marital disruption

Disrupted between I988 and I 994

Predisruptioo marital contlict

High frequency

High breadth

2 0 . 3 2 . 73 9 . 0 2 6 . 312.1 2 . 2

$xi,l4Y 2 3 , 7 0 02 . 8 1 . 2

10%1 0 1 . 8 1 3 . 9

50%

10%

20%1 . 7 0 . 9

1 1 9 . 7 1 9 . 072.5 1 5 . 8

106.0 14.1

19%

10%

IO%.

Kate: Sample is restricted to children aged 49 years in 19X8 who were living with two married parents with no previousdisruptions. who had no missing data on the Behavior Problems Index in 1994. and who usually rcsidcd with their mothers;

n = 7 2 7 .

to determine the timing of marriages, separations,divorces, reunions? and spousal deaths. For childrenborn into married two-parent families that werestill together at the 1988 child interview, we con-structed a dummy variable to indicate whether aseparation or divorce occurred after that date butbefore the 1994 child interview. Our sample con-tains 137 children in disrupted families-76 girlsand 61 boys.

Marital c.o~$‘ict. In the 1988 and 1992 youth inter-views. respondents were asked to report the fre-quency with which they and their spouse argueabout each of nine topics: children, money, choresand responsibilities, showing affection. religion,leisure time, drinking, other women. and the re-spondent’s relatives. We used responses to theseitems to tap two dimensions of interparcntal con-flict. frequency or how often arguments occur andbreadth or the range and content of arguments.Evidence indicates that the more frequent the con-flict, the more it potentially exposes children tonegative parental interactions and the more prob-lematic it is for the child (e.g., Porter & O’Leary,1980). It is less well known whether children aremore severely affected when parental argumentscover a comparatively wide range of topics. To cre-ate a measure of frequency. we summed responses,ranging from 3 (@en) to 0 (never), across the ninecontent areas (observed range: O-25; M = 9.6; SD =

4.2). To create an overall measure of breadth, werecoded each response into yes (1 = often or some-times) or no (0 = hardly ever or never) and summedthe number or content areas that the mother re-ported arguing about (observed range: O-9; M =3.2; SD = 1.9). Although there is precedence in theliterature for using either a continuous measure(e.g., Amato & Booth, 1991; Amato, Loomis, &Booth, 1995; Jekielek, 1998; Kline, Johnston, &Tschann, 1991) or a dichotomous measure (Bu-chanan, Maccoby, & Dornbusch, 199 1: Mechanic& Hansel], 1989; Webster-Stratton, 1989) of mar-ital conflict, we have chosen the latter strategy inorder to differentiate clearly between routine andsevere levels of conftict. We created dichotomousvariables from each of our continuous scales usingthe 90th percentile as a cut-off.

To determine the timing of the measure ofconflict for different subgroups in the sample. weused an approach analogous to Amato et al. (1995).In 1988. all marriages were intact. Among childrenwhose parents subsequently separated or divorced.our marital conflict variables are drawn from thesurvey most proximate, but prior, to the disruption.For children whose parents remained together, webased our measure of marital conflict on the aver-age across the two time periods (1988 and 1992).Measuring parental conflict in this way. we argue,provides the appropriate measure of the level ofconflict from which separation or divorce repre-

Co~fTict and Marital Disruption

sents a potential exit for the child. We test the ro-bustness of our results using alternative strategiesfor measuring conflict.

Control wriablrs. Because divorce does not occurrandomly (see White. 1990, for a review) and be-cause many of the characteristics that may predis-pose families to divorce may also be related to be-havior problems in children, we include a varietyof measures intended to account for family circum-stances before disruption. Our maternal controlvariables include the mother’s years of completededucation as of 1988, her age at first birth, her scoreon the Armed Forces Qualification Test (AFQT)in 1979, and the number of children she had as of1994. Given the clear inverse relationship betweensocioeconomic status and divorce (e.g.. Martin &Bumpass. 1989), we also control for total familyincome and poverty status for the year before thetiming of our dependent variable (i.e., 1993). Weargue that the appropriate counterfactual for as-sessing whether children who exit high-conflictsituations are better off than those who remain insuch situations is to examine differences in levelson the Behavior Problems Index, controlling foraccompanying deficits in the family’s economiccircumstances.

Characteristics of the children, themselves,also may have destabilizing effects on marriage(Corman & Kaestner, 1992; Koo, Suchindran, &Griffith. 1084; Morgan, Lye, & Condran, 1988)and are predictive of levels of behavior problems.Hence, we control for the child’s sex, birth order,birth weight, race-ethnicity. and age at the 1988assessment. To account for differences in the qual-ity of the home environment provided to the childin 1988, we include scores on the Home Observa-tion Measurement of the Environment-ShortForm. This is a mother-rated and interviewer-ratedmeasure of cognitive and emotional aspects ofparenting, as well as aspects of the physical envi-ronment (Mott & Quinlan, 1991).

631

In our multivariate models, we use mean sub-stitution to impute the mean values for 11 casesmissing values for child’s weight at birth. 141 casesmissing values for family income and poverty sta-tus in 1993. four cases missing values for mother’seducation, 34 cases missing AFQT scores, 28 casesmissing values for the Home Observation Measureof the Environment-Short Form, and 66 cases miss-ing values for measures of marital conflict. Weconstructed a dummy variable to identify cases withimputed values and include it in our regressionmodels.

We use linear regression models to assess the re-lationship between parents’ maritai disruption andmarital conflict and children’s level of behaviorproblems in 1994. Because the NLSY child sam-ple contains siblings (and we want to maintain alleligible cases to maximize our sample size). weneed to address the potential problem of noninde-pendence in the random error terms in our modelsfor behavioral problems. Because there may beunobserved family factors that influence the out-comes of children in the family, the error terms forsiblings will not be independent, and, consequently.ordinary least squares anaIysis will underestimatethe standard errors for the point estimates of inter-est. Our solution is to estimate Huber-White stan-dard errors, which correct for the nonindependenceof observations across siblings. (See Huber, 1964:White, 1980.)

RESULTS

To set the stage for our analyses of an interactioneffect between marital disruption and predisruptionconflict, we begin by establishing whether mar-riagcs that end in divorce or separation are charac-terized by higher levels of predisruption conflictthan those that remain continuously intact from

TABLE 2. MEANS ON MEASURES 01; PREIXSKLWIOK MARITAI. COKFLI~T ANI?V~WFHER-RATED BEHAVIOR PROBLEMS. t)~ MARITAL DISRUPTION STATUS IK 1994

Parents Separated or Divorced Parents Remained Married

Predisruption marital conflictHigh frequency of arguments 2WC 7 ‘yrLarge breadth of topics argued about 22% 8%

1994 score on Behavior Problems Index 110.2 106.1

Note: Sample is restricted to children aged 49 years in 1988 who were living with two married parents with no previousdisruptions, who had no missing values on the Behavior Problems Index in 1994 or for measures of parental conflict. andwho usually resided with their mothers; II = 123 for separated or divorced parents, )I = 538 for children whose parents rc-mained married.

IbNF-02015-00008~eprocluccd with permission of Copyright owner. Further reproduction prohibited without pennission.~

632 Journal of Marricrge and ihe Farnil

1988 to 1994. Table 2 reveals that, compared withchildren in continuousiy married families, morethan double the percentage of children whosemothers were separated or divorced in 1994 werecoded high on each dimension of conflict betweentheir parents. Specifically, 20% of the childrenwho experienced separation or divorce were clas-sified as high on frequency of parental quarrels.compared with 7% of those in the never-disruptedsample @ < .OOl). Twenty-two percent of childrenin disrupted families were rated high on breadthof parental marital conflict. The comparable figurefor children whose parents remained together was8% (p < .OOl ).

disruption, we see significantly higher levels of be-havior problems among those exiting high-conflictsituations (t = 1.57, p < .I0 for frequency; t =1.78, p < .05 for breadth). For the two groups inhigh-conflict marriages (frequency as well asbreadth), there is no statistical difference in thelevel of behavior problems observed for childrenwhose parents separated or divorced and for chil-dren whose parents remained together.

Table 2 also allows us to examine whethermarital disruption is associated with higher levelsof behavior problems. Children whose parentsseparated or divorced scored. on average, roughly4 points higher than children whose parents re-mained together (1 10 and 106, respectively; t =3.34, p < .OOl.) Having established an associationbetween conflict and marital disruption and be-tween marital disruption and subsequent behaviorproblems in children, we tackled the next ques-tion: whether the relationship between maritaldisruption and child well-being varies accordingto levels of predisruption conflict.

Table 3 is arranged to allow us to make twocomparisons in 1994 scores on the Behavior Prob-lems Index: among children who experiencedseparation or divorce, those with low versus high-predisruption conflict; and among children inhigh-conflict families, those whose parents didand did not separate or divorce. A difference ineither of these contrasts would suggest an interac-tion between the effects of predisruption conllictand marital disruption. The scores presented for thefour groups of interest are unadjusted for mater-nal, child. and family control variables. Beginningwith the two groups of children who experienced

The results presented thus far have not takenaccount of behavior problems in children that al-ready may have been apparent before separationtook place. Our next step is to examine whetherour bivariate evidence of an interaction betweenthe effects of marital disruption and prior conflictholds when we control for children’s initial scoreson the Behavior Problems Index as well as laterfamily circumstances. Controlling for prior scoresallows us to determine whether divorce and highlevels of conflict predict increases in children’sbehavior problems over the &year interval. Table4 presents the coefficients for marital disruption,prior marital conflict, and the interaction of thetwo effects in models predicting children’s behav-ior problems scores in 1994. Separate modelswere computed for frequency and breadth of con-flict. All of the models control for child’s age, sex,birth weight, birth order, predisruption score onthe Behavior Problems Index, Black. Hispanic.mother’s age at first birth, mother’s years of com-pleted education, mother’s AFQT-R score, numberof mother’s children in the household, log of fam-ily income in 1993, poverty status in 1993, pre-disruption quality of the home environment, and aflag for imputed data, although the coefficientsfor control variables are not reported in Table 4.

In Model 1, where we enter only the main ef-fect of separation or divorce, we find a statisticallysignificant deleterious effect of disruption on chil-dren’s scores on the Behavior Problems Index.

TARL~ 3. MEWS ON MOTHER-RATED BEHAVICIK PKO~ILEMS SCOKES FOR CHILDREV 1~ 1994.BY PRCDISRWTIO?I MARI’IAL CONFIKT AND MARITAL DISRWTION S14~u.s IK 1994

High Conflict

Parents ParentsSeparated or Remained

Divorced Married

Low Conflict

Parents ParentsSeparated or Remained

Divorced Married

.ooion Of Copyright owner. Further reproduction prohibited without permission.

Frequency of arguments I 13.6 117.2 109.7 105.1II 25 40 9x 498

Breadth of topics argued about 114.3 I IO.0 109.5 105.5n 27 41 96 497

Note: Sample is restricted to children aged 4-9 years in 1988 who were living with two married parents with no previousdisruptions. who had no missing values on the Behavior Problems Index in 1994 or for measures of parental conflict, andwho usually resided with their mothers.

e-0 BWF-02015-00008Reproduced with pennioo+on of Copyright owner. Further reproduction prohibited without pe-

Model I Model 2 Model 3

High frequency Separated or divorced x22** 2.1) ] *:.: bJ.x-l;::“:,:

of arguments High frequency of conflict d,,(J*“:* 7.02”C”kSrparated or divorced) x (high cuntlict) - 6 . 7 5 ”

R’ .27Large breadth Separated or divorced 3,2**. .- 7 ‘)x”:”-. 2.yp:::

of topics Large breadth of conflict ‘,jtp’ 2.70argued about (Separated or divorced) x (high conflict) - 0 . 0 ’R? .28

Note: Sample is restricted to children aged 4-9 year> in 1988 who were living with t\vo married parents with no previousdisruptions. who had no missin, m values on rhe Behavior Problems In&x in 19% (M = 106.9. SD = 14.3). and who usuallyresided with their mathcrs; IT = 727. Tables of coefficients for all variables are available from authors.

*p 5.05. **ps,o1. “**/7~.001.

The experience of marital disruption significantlyraises children’s scores over predisruption levelsby roughly one fifth of a standard deviation (3.2points). By adding a main effect for each high-conflict variable in Model 2. we explore whetherpredisruption marital conflict affects child well-being over and above the effect of divorce orwhether the apparent effect of marital disruption isattributable to the disharmony that existed beforethe break-up. As shown in the second column ofTable 4, both of our measures of prior marital con-flict have a statistically signCant main effect-adding 3--5 points to the Behavior Problems Index.controlling for other child and family factors.Note that the coefficient for high frequency ofconflict exceeds the effect of divorce, but the effect

of high breadth of conflict is fairly comparable withthe effect of marital disruption. It is noteworthythat the magnitude of the disruption coefficient isbarely reduced and remains statistically significantwhen we add measures of conflict. In other words.children who experience their parents’ separation ordivorce show a notable increase in behavior prob-lems scores, and this increase is not attributable tothe effects of predisruption conflict. Furthermore.because we are controlling for current income andpoverty status. this effect is arguably net of the ef-fects of the downward economic mobility thatoften accompanies divorce.

Our final question in this multivariate contextis whether the effect of marital dissolution differsaccording to the level of conflict that preceded thedisruption. As shown in Model 3, we observe astatistically significant interaction between the ef-fects of divorce and conflict but only for conllictthat is frequent. Fully understanding this statisti-cally significant interaction requires that we maketwo separate comparisons of outcomes: across thetwo disrupted groups and across the two high-

conflict groups. Moreover, we need to take accountof children’s predisruption scores on the BehaviorProblems Index to appreciate the association be-tween disruption and children’s behavior problemsat the end of the period.

Figure I illustrates both average predisruptionscores on the Behavior Problems Index and in-creases in scores on the Behavior Problems Indexthat are predicled by the interaction model fromTable 4 for children in the four disruption-conilictcategories. Among children whose parents sepa-rated or divorced (first and third bars). there wasan equivalent increase in behavior problems overprior levels (4 points), regardless of whether thechild exited a high-conflict or low-conflict situa-tion. To test this, we entered dummy variablesrepresenting the different combinations of maritalstatus and conflict in a model (not shown) andfound that the coefficient for low-conflict divorcewas not statistically different from the effect ofbeing in the high-conflict divorce group. Thus,contrary to our expectations. exiting high-conflicttnarriages did not attenuate behavior problems inchildren. That is. net of prior levels of behaviorproblems, disruption was no less deleterious forchildren leaving high-conflict situations than it wasfor children whose parents had had relatively har-monious marriages. The bivariate results presentedin Table 3 and thebe multivariate results attest tothe importance of accounting for children’s priorlevels of behavior problems. Specifically, childrenliom high-conflict marriages already showed ele-vated behavior problems scores before their par-ents’ separations.

To answer more completely the question ofwhether children bcnetit when high-conflict mar-riages dissolve. we need to know how these childrenmight have fared if their parents had stayed to-gether. A comparison across the two high-conflict

634 Journuf of Murriage and the Family

F IGURE 1. CHILCNEN’S I994 B~H.WIOK PKOBLEM IWES Scorn AS FUIKTI~K OF IKTIXAC~-IOK BETWEEN !MAKITAL CONFIKTAND MARITAL Drs~urr~ou. AUJUSIU FOR PRIOR BF.H+WOR PRIJHLEM LEWLS ANU FPIMILY BACKGRCXJYD

Disrupted R e m a i n e d

MarriedHigh-Confl ict

M a r r i a g e s

Disrupted R e m a i n e d

MarriedLow-Confl ict

M a r r i a g e s

categories (first and second bars of Figure I)gives some insight into this. Here we see that chil-dren whose high-conflict parents remained togetherhad a larger increase in behavior problems thantheir counterparts whose parents broke up (7 pointsvs. 4 pointA, re5pectively, p < .05 j.

To test the robustness of our results to the waywe measured parental contlict and marital disrup-tion, we examined several alternatives. We wereconcerned with three issues related to the interac-tion that we observed between marital disruptionand conflict: (a) We may have confounded con-flict and parental divorce by using a conflict mea-sure that was proximate to the separation; (b) wemay have missed information by using a dichoto-mous measure of high conflict: (c) we may haveunderestimated the effect of marital disruption bycontrolling for current economic circumstances. Toaddress the first concern, we estimated the samemodels shown in Table 4 but substituted 1988predisruption measures of marital conflict for allrespondents, rather than the most proximate mea-sure oi average of 1988 and 1992. With this strat-egy we observe roughly equivalent, statisticallysignificant main effects of disruption (p = 3.7,1.1 <.05) and frequency of parental conflict (p = 3.2, p <-10) but not a statistically significant interactionbetween the two. Thus, if we control for parentalconflict at the beginning of the period (whenquarrels may not yet have reached severe levels).disruption does not provide the same level of re-

lief to the high-conflict and disrupted group as itdoes when we account for the level of arguingthat occurs more immediately before the separa-tion or divorce. Our second step was to explorealternatives to our dichotomous measure of severeconflict. We created a trichotomous measure ofconflict. with the middle group comprising childrenwith values on parental conflict ranging from themean plus or minus one half a standard deviation.We found, as we did above, that the effect of divorceis no better or worse, depending on predisruptionlevels. But unlike our results using dichotomousmeasures. we also found no statistical differencein the change in scores on the Behavior ProblemsIndex fur children in the disrupted group versusthe intact high-conflict group. Instead, the signifi-cant interaction effect was driven by the fact thatchildren in low-conflict and medium-conflictfamilies who remained together had fewer behav-ior problems than those in the high-conflict, dis-rupted families. Finally, we removed contempora-neous measures of family income and poverty andobtained results that were virtually identical to theresults reported above.

In sum. the results reported in Table 4 suggestthat the experience of parental separation and di-vorce is uniformly harmful to children (at least inthe first years after disruption). regardless of howoften their parents quarreled beforehand. However.parents remaining married is not a better alternativefor children when conflict between the parents is

Reproducrd with permission of Copyright owner. Further reproduction prohibited without pcrmirrion.

Cmflict und Marital Disruption

high. Indeed, the largest increases in scores on theBehavior Problems Index were observed for thechildren whose parents remained in such marriages.

We sought to examine whether the effect of par-ents’ separation and divorce on children’s behaviorproblems varies according to the level of conflictbetween parents before the disruption. We usedthe NLSY mother-child data to examine this issuein a sample of 4 to 9-year-olds in 1988 who beganthe study in two-parent. married families. An av-erage of 3 years later, 19% of the children experi-enced the break-up of their parents’ marriage. Wewere able to compare the 1994 levels of behaviorproblems of children whose parents separated ordivorced with the levels of children whose parentsremained married, net of their levels of behaviorproblems at the outset. This prospective strategyallowed us to examine changes in behavior prob-lems scores after marital dissolution and enabledus to make inferences, although not firm conclu-sions, about causality.

We observed a sizable negative effect of mari-tal disruption on children’s behavior problems.controlling for predisruption scores on the Behav-ior Problems Index and other child and familycharacteristics. We explored the main effect ofboth frequency and breadth of marital conflict onchild well-being and also observed the extent towhich these predisruption measures accounted forthe apparent effects of divorce. We found that thenegative effect of divorce remained relativelylarge and statistically significant when conflictwas added to our models. This suggests that factorsassociated with divorce, itself, such as parentalabsence, changes in custody and relationships.and declines in parents’ psychological well-being:explain increases in children’s behavior problemsover and above the effect of predisruption parentalquarrels. Nonetheless, we found that prior reportsof high levels of marital conflict have a large andstatistically significant adverse effect on children’sbehavior problems. This is particularly true forour measure of the frequency of parental conflict.Indeed. the adverse effect of frequent maritalquarrels is larger than the deleterious effect ofseparation and divorce. Finally, consistent withAmato and colleagues’ (1995) study. we found alarge. statistically significant interaction betweenthe effects of frequent marital conflict and disrup-tion. Unlike our predecessors, however, our multi-variate results did not indicate a benefit for chil-

--02015-00008Reproduced with peIlnLssion of Copyright owner. Further reproduction prohibited without permission.-

635

dren exiting high-conflict marriages. Children ex-perienced a 4-point average increase in behaviorproblems after parental separation and divorce, re-gardless of the level of conflict that predated thedisruption. Furthermore, in our sample, the greatestincrease in behavior problems was observed amongchildren whose parents remained married, despitevery frequent quarrels.

Our failure to uncover the expected differentialsin the effect of divorce may be due to the youngerages of the children in our sample at the time ofdisruption or may be due to the comparativelyshorter duration since disruption in our analysis,compared with the study by Amato et al. (1995).Any relief or benefit gained by being removedfrom a high-conflict situation may not be apparentfor several more years or until adulthood. It is pos-sible that if we were to follow these children, wewould find a pattern similar to the Amato team’sfindings. In her examination of the relative impor-tance of parental conflict and marital disruption onchildren’s internalizing behavioral problems.Jekielek (1998) reports that the stress-relieving ef-fects of leaving a high-conflict marriage becomemore apparent over time. However, she only ob-served children up to 4 years after disruption.

In addition, the set of conflict measures avail-able in the NLSY is limited in scope. comparedwith the types of information often explored inanalyses of parental conflict and children. For ex-ample. although presumably there is variation inthe extent to which children are exposed to orshielded from parental conflict, we are not able toaccount for these differences. Nor does our mea-sure provide an indication of the process by whichdisagreements in each area are resolved. For ex-ample, arguments may be calm and amicably re-solved, or they may be heated or openly hostile.Moreover, our predisruption measure of conflictis a snapshot measure. It would be preferable tomeasure whether the child has had relatively long-term or short-term exposure to intensive parentalquarrels. Furthermore, these measures of conflictare from a single rater, the mother. Ideally, onewould also have reports about parental conflictfrom her spouse or the children. Moreover, the Be-havior Problems Index score, our dependent vari-able, is also mother rated. In sum. respondents whorate relatively low on the measures used here maybe quite a heterogeneous group.

The fact that the association between maritaldissolution and the Behavior Problems Index doesnot differ according to prior levels of conflict inthe expected way also may relate to the dimen-

6 . 3 6

sions of conflict that we were able to measure.The conflict measures employed in the study byAmato et al. (1995) reflect not only the frequencyof arguments, but also their severity-specificallywhether marital disputes ever involved physicalaggression. Research on the effects of witnessingand experiencing physical violence (Sternberg elal.. 1993) points to the serious and long-lastingconsequences of such exposure for children. Itseems likely that the beneficial effects of removalfrom a violent household through divorce couldoutweigh the costs of marital disruption for thechild. Because the NLSY does not include a mea-sure of physical marital aggression. our analysiscould not detect such a pattern. Moreover. it isperhaps not surprising that our measure of the

breadth of marital conflict showed fewer andweaker associations with children’s beha\/iorproblems than did our measure of frequency 01conflict. Our results suggest that children are sen-sitive to the amount of arguing. rather than therange of topics about which parents argue.

Finally, after divorce. parental conflict may di-minish when former spouses establish a morefunctional family life. or conflict may intensifywhen couples work through custody arrangements(Emery & Coiro, 1997). Levels of postdivorce con-flict have been shown to predict chiid and adoles-cent adjustment (Buchanan et al.. I991 ; Forehand,Thomas. Wierson, Brody. & Fauber, 1990; Luep-nitz. 1982). Because the NLSY does not include ameasure of postdivorce conflict. we were unable toexamine whether the behavior problems of childrenin disrupted families reflect exposure to on-goingconflict, rather than prior conflict.

Despite these limitations. our results are sober-ing. Regardless of the level of conflict reportedbefore the disruption, separation and divorce areassociated with notable increases in young chil-

dren’s behavior problems, namely a rise of nearlyone third of a standard deviation. This is in kecp-ing with the magnitude of the effect sizes (one fifthto one fourth of a standard deviation) that Amatoand Keith ( 199 1) report in their meta-analysis ofstudies of the long-term consequences of parentaldivorce for adult well-being. Moreover, our studyreinforces evidence from Amato’s team ( 1995)and numerous other researchers that frequentmarital conflict has a deleterious effect on chil-dren, possibly even exceeding the adverse effectof physical separation or divorce. We began withthe question: “Do children benefit when high-contlict marriages are dissolved’?” The answer isa qualified -‘yes.” We found that children whose

high-conflict parents remained together through-out our study demonstrated the largest increase inbehavior problems. However, children whose high-conflict parents subsequently separated or di-

vorced had the highest scores on the BehaviorProblems Index at the outset. In simple terms, this

means that if we took a snapshot view of all thechildren in 1994, we would not be able to distin-guish children whose high-conflict parents sepa-rated from those whose high-conflict parentsstayed together.

Findings such as these do not provide clear an-swers for parents. legal professionals. and policyanalysts trying to put the best interests of childrenabove other concerns when divorce is contem-plated because they underscore that neither alter-native is without cost to children. In the absenceof information about how the children in our sam-ple will fare over the long term, perhaps the clear-est signal comes from the children in two-parentfamilies with low levels of parental conflict-clearly the group who is f‘aring best in this study.

N O T E

An carlier version of this article was presented at the an-nual meetings of the Population Association of Americain Miami. May. 1994. The authors gra tefu l ly acknowl-edge the research ass is tance of’ Connie Blumenthal .Nancy Buchanan. and Amy Ritual0 and the helpfulcomments of Larry Bumpass. Andrew Cher l in , Frank F.Furstenherg. Jr., Kristin Moore. Martha Zaslow, and fouranonymous reviewers . This research was par t ia l ly sup-por ted by Nat iona l Ins t i tu te of Menta l Heal th GrantMH4549 l-01 and National Institute of Child Health andHuman Development Grants 5 R37 HD25936-07 andHD349684 I _

REFERENCES

Achenbach, T. M. (I 978). The Child Behavior Profile: I.Boys aged 6-1 I. JOII~WI/ of Cotzsultiug and ChidP,s~dwlo~,v, 46. 478488.

Amato, P. ( I 994) . Lift-apan adjus tment of chi ldren tothe i r paren ts ’ d ivorce . The Future qfC/liihen: C h i l -dren und fli~wrce, 4, 143- 163.

Amato, P. R., & Booth, A. ( 1 9 9 I ). The consequences ofparenta l d ivorce and mar i ta l unhappiness f o r adu l twell-being. S~ciul F0rcrs. 69, 895-914.

Amato, P R., & Kei th , B. (1991) . Parenta l d ivorce andthe well-being of children: A rneta-analysis. Psycho-logiccrl Bulletin. I IO. 26-46.

Amato. P. R., Loomis. L.. & Booth, A. (1995). Parentaldivorce, mari ta l confl ic t , and offspring well-beingduring early adulthood. Gcial Fnmes, 73, 895-9 IS.

Bay&r. N. ( 1988). Effects of parental separation and re-entry in to un ion on the emot iona l well-being of ch i l -dren. Journd of‘Marriuge and the Fcmily 50, 967-981.

Block, .I. H., Block, J., & Gjerde, P F. (1986). The per-sona l i ty of ch i ldren pr ior tu d ivorce : A prospec t ivestudy. Child Dewlopnent, 57, 827-840.

W - 0 2 0 1 5 - 0 0 0 0 8R e p r o d u c e d w i t h p e r m i s s i o n o f C o p y r i g h t owner. F u r t h e r r e p r o d u c t i o n p r o h i b i t e d w i t h o u t p e r m i s s i o n .

I-

Buchanan. C. M.. Mnccoby. E. E.. & Dombusch, S. M.(1991). Caught between parents: Adolescents’ cxperi-ence in divorced homes. Cl?&/ Dc~+y>///c//r. 62, loOX-1029.

Cherlin, A. 1.. Chase-Lansdale. P. L.. CG McRae. C.(1998). Effects of divorce on mental health throughthe life course. Anwriccrrl S'nc~iolqqicnl Re~ie~r,. 63.

239-249.Cherlin. A. J.. Kiernan. K.. & Chase-Lansdale, P. L.

(1995). Parental divorce in childhood and demographicoutcomes in young adulthood. Drr~~ogeru/$/~ 32. 399-3 1 x .

Cherlin. A. J.. Furstenberg. F. F.. Jr.. Chase-Lansdale,l? L.. Kieman. K. E.. Robins. P. K.. Mol-rison. D. J. R..& Teitler. J. 0. (1991). Longitudinal rtudics of theeffects of divorce on children in Great Britain and theUnited States. Scie/~c,~. 252. 1386- 138’).

Corman. H.. iyr Kaestner. R. (1992). The effects of childhealth on marital status and family structure. D~//rc/g-mph.‘: 29. 3X9-408.

Doherty. W. J.. & Needle, R. H. f 1991). Psychologicaladjustment and substance use anion! adolescents bc-fort and after a parental divorce. Cl/&l nr~,clo//,//rr,/tr.62. 328-337.

El-Shcikh. M.. Cummings. E. .M., & Goetsch, V. L.(1989). Coping with adults’ angry behavior: B&v-ioral, physiological. and verbal responses bypreschoolers. De~~elopnmtd P.yl~olo,rp: 25, 390-4%.

Emery. R. E. (IY82). Interparental conflict and the chil-dren of discord and divorce. Ps~[./lologir,f// Bdletirr,92, 3 I O-330.

Emery. R. E. ( 1 Y88). Mcr~in:p. r/irv//cc>. otrtl c,/ril~l/v// ‘.Ym~~.srnmr. Beverly Hills. CA: Sage.

Emery, R. E., & Coiro. M. J. (1097). Some costs of cop-ing: Stress and distress among children from divorcedfamilies. In D. Cicchetti & S. Toth (Eds.). Dr~,r/q,-tmmd perspcctiws Ott rruum~: Tkw?; rcscwrdz, rrdinrerwruim (pp. 435-462). Rochester. NY: Lniuersityof Rochester Press.

Emery. R. E.. Fincham, F. D.. & Cummings, E. M.(1992). Parenting in context: Systemic thinking aboutparental conflict and its influence on children. Jo//~/c//of Cf~t/.sitltirlg und Clitkul Psyl7olr~,qv, 6, 909-9 12.

Forehand, R.. Thomas. A. M.. Wierson, M.. Brody. G..Bi Fauber, R. (1990). Role of maternal functioninsand parenting skills in adolescent functioning follow-ing parental divorce. ~~nurzn1 ofA/mm~~rrl f?syho/r~~~~,99, 278-283.

Crych. J. H.. & Finchman. F. D. ( I990). Marital conflictand children’s adjustment: A cognitive-contextualframework. f.s~cho/o~t$c~// B~/llerr//, 108, 267-2YO.

Hethcrington. E. M.. Cox, M.. & Cox, R. (1982). Effectsof divorce on parents and children. In M. Lamb (Ed. 1.~Vonrrufirionnl ,\~r/rilic.s (pp. 233-288). Hillhdale. NJ:Erlbaum.

Huber, P. H. (I 964). Robust estimation of a local para-meter. Aft//~fI.s of’MarkorlaticaI Sld.5ric.h 35. 7% I 0 I

Hummel-Rossi. B.. & W&&erg, S. L. (1975). Prm.ticdpiilelinct in upplyiqq currei~r r/~fwt?~s to the IN~~I.~I~IY~-rnenr of churlge. Greenwich. CT: Johnson Associates.

Jekielek, S. M. ( 1998). Parental conflict, marital disruptionand children’s emotional well-being. So&l Fovc%v, 76,905-935.

Kieman. K. (1992). The impact of family disruption inchildhocjd on transitions made in young adult life.Populcrtiorl Srudirs. 46. 2 13-234.

WUF-02015-00008

Kline. M.. Johnston. J. R.. & Tschann. J. M. t 1991 1. Thelong shadow of marilal conllict: A model of children‘spostdivorce adjustment. ./r////~r// qf ,21~/+~c, c///tl r//pfiurlilv. 53. 297-309.

Koo. H.-P.. Suchindran. C. M.. & Griffth. J. D. (1085).The cffccts of children on divorce and rcrnarr-iagc: Amultivariatc analysi\ of lift table probabilities. &/IX/-kffhl S/fftks. M. 4s l-17 I.

LuepniU. I). A. ( 19x2). C//i/r/ (.//.t/o~/~: .+1 .rh/rl\. o//~/i////i-/&s ~!ficv &v/r(~,. 1,cxinpton. MA: Lexington Books.

Martin. T. C.. & Bumpass. L. L. (198Y). Recent trendsin marital disruption. f)c,r//r/,s/r//~//~. 26, 37-S I

Mechanic. D.. K: Hansell. S. ( IYXY). Divorce. family con-tlict. and adolc~ccnt~’ w&being. JO//UILI/ (!f’ Hcw///Iirrd SocYril l~rhtnio/: 30. 10% I 16.

Morgan. S. P.. Lye. D.. & Condran. G. A. I 10x8). Sons.dauphtcru. and the risk of mnri~al disruption. Auwr-i-C’U11 .fo/l/7ftrr f!f’.soc~irh,lp. 94. I I o- I ‘Y.

Morrison. D. R.. & Cherlin A. J. (1995). The divorceprocess and young children’s well-being: A prospec-tive analysis. Jor//7/f1l 0f’ililtrrrirc.qc~ turd r/w Fund\; 57.X00-8 1 ’ .

Morrison. D. R.. Coiro. M. J.. & Blumenthal. C. ( 1994.Mty ). Mm-itd ~lic~xp~i~~~~. ~~~~~/li~~f. nml rh rdl-/wir~p<ffc,l/i/tlre//. Paper prcscntcd at the annual meetings ofthe Population Associalion 01’ America. Miami.

Mott. E, & Quinlan. S. t I YYI j . C‘hihhrt of’rk NLSY:/WY ttrlwlr~fiorr.s crrd .~~rfrufw;~~ di.sc~rr.&r~. Cal tim-bus: The Ohio S~atc L’nikcrsity. Center fc>r HumanRrrourcr Rc.\e*lrch.‘

Peterson. .I. I... & Zill. .U. ( 1086). Marital disruption.parent-child relationships. and behavior problems inchildren. .lo/r/~o/ ol’ Mt/r/?tr,q~~ trrrrl tlrc /+/r//i/~, 48.29.i-307.

Portcl-. R.. & O’Leary. K. 0. t IYSO). Marital discord andchild beha\,ior problems. ./~/r//~//o/ ($Ah/rn/.///~1 C//i/r/P.~u~llrllr/,qY, s. 7x7-295.

Slater. E. J.. & Habcr. J. D. ( 198-1). Adolescent adjust-ment followinp divorce 21; a I‘tniction of Ihniilial con-tl ict. ./o///7/(/1 0f Cc///.\rf/ri/r,q co/t/ Clir7iwl P~~~~/~ok/,q~;52. 920-Y? I.

Stcrnbcrg K. J.. Lamb. M. E.. Grccnbaum. C.. Cic-chetti. D.. Dawud. I:.. Cortex. R. M.. Krispin. 0.. &Lorey. F. ( I YYZ). Effects of domsstic violence on chil-drcn’\ behavior prohlcm\ and dcprebsion. I>c*~.e/o/j/-rrw7tcrl P.\,dK/hgY '9. 11-52.

Webster-Stratton. C. t lY8Y J. The relationship of maritalsupport. conflict. and divorce to parent pcrceptinns.behaviors. and child conduct problems. ./o~r/~t/l (t/’!Mo,-ricr,&v t/d /I/e Ffllr/i!\: 3-l. 4 17-130.

White. H. ( l%(l). A Iict~roskedaxticit~-~onsistc/lt co-variance matrix estimator and direct tesl for het-eroskedastici ty, ~~,o//o///rl/‘ic’rr. -IS. 8 I 7-830.

1Vhitc. L. (1YYO). Drtr‘rminant~ of divorce: A rcvicw ofrcscarch in the eighties. ./o///~c/l r!f’Mtrrritr,cy rrd thei+rl/i/\: 52. Y0-LY Il.

Zill. N. f 1990). &~htrv;r//. /‘/r//~/c//r., I/rc/c<r btr.\ct/ O/I IX//.-c//i /q/o/-f. Il’ashington. DC: Child Trends.

Till. N.. Morrison. D. R.. & Cairo. M. J. ( IYY.?). Lonp-term cllizcts of parental divorce on parent-child reln-tionships. adjustnicnt. and achicvcment in )-oungadu l thood. ./o//~/t/l c$/.irr//i/.~ P.\~c~l~olo,~,v. 7 . Y I- 103.

Reproduced with ponnirsion of Copyright owner. Further reproduction prohibited without pormirsion.