THE EFFECTS OF DIVORCE AND MARITAL

DISCORD ON ADULT CHILDREN'S

PSYCHOLOGICAL WELL-BEING

PAUL R. AMATO

The Pennsylvania State UniversityJULIANA M. SOBOLEWSKIThe Pennsylvania State University

Previous research has demonstrated associations between exposure to parentaldivorce and marital discord while growing tip and children's psychological distressin adulthood. Few studies, however, have attempted to explain these associations.Three pathways are evaluated through which family disruption and discord mayaffect offspring's well-being: children's socioeconomic attainment, children's maritaland relationship stability, and the quality of children's relations with parents. Using17-year longitudinal data from two generations, results show that divorce and mari-tal discord predict lower levels of psychological well-being in adulthood. Parent-child relationships mediate most of the associations between parents' marital dis-cord and divorce and children's subsequent psychological outcomes. Marital discordappears to erode children's emotional bonds with mothers, whereas both divorce andmarital discord appear to erode children's emotional bonds with fathers. The resultshighlight the continuing importance of parent-child ties for children's well-being inearly adulthood.

I NDIVIDUALS with divorced parents arel a t increased risk of experiencing psycho-logical problems in adulthood. Although thistendency has been documented in manystudies, the explanation for this phenom-enon remains elusive. A few studies haveconsidered the possibility that low educa-tional attainment or poor interpersonal skillsmay mediate the association between paren-tal divorce and adult psychological well-be-ing. Because of the lack of appropriate lon-gitudinal data, however, few studies have at-tempted to explain the link between child-hood family structure and adult psychologi-cal functioning.

We use data from a 17-year longitudinalstudy of two generations to explain the esti-mated effect of parental divorce on adultoffspring's psychological well-being. In ad-

Direct all correspondence to Paul R. Amalo.Department of Sociology. The PennsylvaniaStale University, University Park, PA 16802-6207 (pxa6@psu.edu). This research was sup-porled by the National Institute on Aging (grantROI AG04I46). We thank the anonymous ASRreviewers for their helpful comments.

dition to examining the mediating role of so-cioeconomic attainment and relationship sta-bility, we consider a third mechanism: thequality of relations between offspring andparents in adulthood. Although good reasonsexist for assuming that the quality of parent-child ties mediates some of the long-termnegative effects of parental divorce, priorstudies have not considered this possibility.Another unresolved issue concerns the long-term impact of discord between parents whoremain married. In the present study, we ex-amine whether exposure to cbronic discordbetween parents while growing up predictselevated levels of psychological distressamong adults, and whether offspring's socio-economic attainment, relationship instabil-ity, and quality of ties with parents accountfor this association.

BACKGROUND

PARENTAL DtvoRCE

Previous studies have demonstrated thatadults with divorced parents, when comparedwith adults with continuously married par-

9 0 0 AMERtCAN SOCIOLOGtCAL REVIEW, ZOO 1 , VOL. 6 6 ( D E C E M B E R : 9 O O — 9 2 1)

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 90 1

ents. report greater unhappiness, less satis-faction with life, a weaker sense of personalcontrol, more symptoms of anxiety and de-pression, and a greater use of mental healthservices. These findings have emerged fromanalyses based on the General Social Survey(Biblarz and Gotlainer 2000; Glenn andKramer I98.i). the National Survey of Fami-lies and Households (Amato 1991), the Na-tional Survey of Children (Furstenberg andTeitler 1994; Zill. Morrison, and Coiro1993). the Marital Instability over the LifeCourse study (Amato and Booth 1991, 1997),the Detroit Area Study (McLeod 1991). andthe Survey of Aging. Status, and Social Con-trol (Ross and Mirowsky 1999). Similar as-sociations appear in national samples fromother countries, including the National Brit-ish Child Development Study (Cherlin.Chase-Lansdale, and McRae 1998). the Brit-ish National Survey of Health and Develop-ment (Rodgers 1994). the Canadian GeneralSocial Survey (Le Bourdais and Marcil-Gratton 1998), and the Australian FamilyFormation Study (Amato 1988). The greatmajority of studies, with few exceptions,support the existence of a link between child-hood family structure and adult psychologi-cal well-being.

Although it is difficult to establish, the pre-ponderance of evidence suggests that the linkbetween divorce and children's psychologi-cal well-being is causal rather than spurious.Cherlin etal. (1998) found that children withdivorced parents exhibited poorer adjustmentthan did children with continuously marriedparents prior to marital dissolution, suggest-ing that some of the apparent "effect" of di-vorce was a result of predivorce factors (alsosee Sun 2001). However, the gap in adjust-ment between these two groups of childrencontinued lo grow throughout adolescenceand early adulthood, suggesting that parentaldivorce had a cumulative impact not ac-counted for by predivorce factors. The use ofa fixed-effects model, which controlled forall time-invariant differences between chil-dren with divorced parents and children withcontinuously married parents, provided fur-ther support for a causal interpretation of thislink (also see Hetherington 1998). Otherstudies have shown that the association be-tween parental divorce and long-term out-comes for children persists after controlling

for a large number of predivorce family andchild characteristics (Furstenberg andKiernan 2001), parents' personality (Simonsand associates 1996), unobserved heteroge-neity (MeLanahan and Sandefur 1994). ge-netic similarity between siblings (Kendler etal. 1992). and the adoption status of children(Brodzinsky, Hitt, and Smith 1993). Overall,most scholars working in this area have con-cluded that parental divorce has real conse-quences for children's long-term well-being(Amato 2000; Cherlin 1999; Emery 1999;Furstenberg and Kiernan 2001; Hetherington1998; MeLanahan and Sandefur 1994).

PARENTS' MARITAL DISCORD

The long-term consequences of parents'marital discord have been studied less exten-sively than the effects of divorce. Neverthe-less, exposure to chronic interparental dis-cord appears to have long-term conse-quences similar to those of divorce. Adultswho recall a high level of conflict betweenparents while growing up tend to report adisproportionately large number of psycho-logical and marital problems in Iheir ownlives (Amato and Booth 1991; Booth andEdwards 1990; Kessler and Magee 1993;Overall, Henry, and Woodward 1974).Amato and Booth (1997) used prospectivedata on parents' reports of marital discordand adult offspring's reports of well-being,thus avoiding problems with commonmethod variance. Their analysis (based on anearlier wave of data from the present study)indicated that parental discord was posi-tively associated with adult offspring's psy-chological distress 12 years later. Althoughfew studies focus on parents' marital qual-ity, both parental divorce and growing up ina high-conflict two-parent family appear tobe linked with long-term decrements inchildren's psychological adjustment.

EXPLANATIONS

Marital dissolution is a process that beginsbefore physical separation and continues af-ter the marriage is legally ended. Separationis typically preceded by a period of conflictor mutual disengagement between spouses.Moreover, divorce is usually followed by aseries of stressful circumstances for chil-

9 0 2 AMERICAN SOCIOLOGICAL REVIEW

Offspringsocioeconomic

attainment

Parent maritaldiscord and

divorce* •

OffspringrelationshipInstability

Offspringpsychological

well-being

Offspring-parent

relationships

Figure 1. Mediation Model Relating Parents' Marital Discord and Divorce to Offspring'sPsychological Well-Being

dren, including reduced contact with non-custodial parents {usually fathers), increasedtension between children and custodial par-ents (usually mothers), continuing rancorbetween parents, a decline in standard of liv-ing, and an increase in residential mobility—often involving moving to neighborhoodswitb fewer community resources (for a re-view, see Amato 2000). In addition, divorceis usually followed by the remarriage of oneor both parents, and dealing with stepparentsrepresents a new source of stress for manychildren (Hetberington 1998). Parental re-marriage also makes it possible for childrento experience multiple parental divorces.

All of these stressful circumstances arelikely to impact negatively on children's psy-chological adjustment. Consequently, therelatively high level of psychological distressamong adult children who grew up in di-vorced families may represent a simple con-tinuation of emotional problems that beganin cbildbood. Most researchers, however, as-sume that cbildhood adversities are linkedwith adult mental health through a variety ofintervening structural and psychological pro-cesses (e.g.. Brown and Harris 1989). More-over, a life-course perspective (Caspi andElder 1988; Elder 1994) suggests that expe-riences in adulthood have the potential toameliorate, maintain, or exacerbate problemsthat originate in the family of origin.

Figure 1 outlines our explanatory frame-work. Our model assumes that three pro-cesses in offspring's lives mediate the long-term effects of parental divorce and maritaldiscord on adult psychological well-being:socioeconomic attainment, relationship in-stability, and the quality of relationships be-tween offspring and parents. To tbe extentthat parental discord and divorce interferewitb children's educational attainment, leavetbem with inadequate interpersonal skills anda history of unstable intimate relationships,or undermine close ties witb tbeir parents andkin, children's distress is likely to be rein-forced or even amplified after reaching adult-hood. Because causal relations among themediators are unclear, the model assumesthat these processes are correlated but doesnot specify the direction of influence.

SOCIOECONOMIC ATTAINMENT

Parental divorce is associated witb lower so-cioeconomic status in adulthood. Comparedwitb cbildren from two-parent families, chil-dren with divorced parents are more likelyto drop out of high scbool, less likely to at-tend college, and complete fewer years ofeducation overall (Biblarz and Gottainer2000; Furstenberg and Teitler 1994; McLeod1991; Ross and Mirowsky 1999; Zill et al.1993). Tbis educational disadvantage ap-

DIVORCE. MARITAL DISCORD, AND CHILDREN'S WELL-BEING 903

pears to be a result of several postdivorcefactors; a decrease in children's standard ofliving, moving to neighborhoods with poorerschools, and declines in parental monitoringand school involvement (McLanahan andSandefur 1994). Given the importance ofeducation in determining later socioeco-nomic attainment, it is not surprising thatoccupational status, earned income, and thevalue of accumulated assets also are rela-tively low among adults with divorced par-ents (Amato and Keith 1991; Biblarz andGottainer 2000; McLeod 1991; Ross andMirowsky 1999).

Parents' marital discord, even in the ab-sence of divorce, may impact negatively onchildren's school achievement. Observingovert conflict between parents is a directstressor for children. In addition, parentswho fight frequently, compared with lesscombative parents, tend to display lesswarmth toward their children and disciplinethem more harshly. Presumably for thesereasons, children in high-conflict householdsare at increased risk for antisocial behavior,anxiety, depression, and difficulty in concen-trating—factors known to influence schoolperformance (Davies and Cummings 1994;Emery 1999). Few studies have related pa-rental discord to children's educational out-comes in adulthood, although Snarey (1993),using longitudinal data, found that parents'marital quality when children were growingup was positively associated with daughters'(but not sons') later educational and occupa-tional attainment.

Previous studies indicate that socioeco-nomic status is positively associated withpsychological well-being. For example, edu-cational achievement is positively related topeople's reports of being happy and havingstimulating, pleasant, and rewarding experi-ences at work and at home (Campbell 1981).Education also decreases the risk of depres-sion (Kessler 1982) and increases people'ssense of personal control (Ross and Wu1995). Moreover, income and economichardship are associated with psychologicalwell-being independently of education (Rossand Mirowsky 1999). Therefore, if educa-tion and income both promote psychologicalwell-being, and if offspring from divorced ordiscordant families have lower levels of botheducation and income, then socioeconomic

attainment may account for the link betweendivorce, marital discord, and offspring's psy-chological well-being in adulthood.

RELATIONSHIP INSTABILITY

Evidence suggests that parental divorcenegatively affects the quality and stability ofchildren's intimate relationships in adult-hood. Individuals with divorced parents,compared with individuals with continu-ously married parents, report more dissatis-faction., problems, and conflict in their ownmarriages (Amato and Booth 1991, 1997).Similarly, the risk of marital disruption ishigher for those who experienced parentaldivorce as children (Amato and Booth 1997;Bumpass, Martin, and Sweet 1991; Glennand Kramer 1987; McLanahan and Bumpass1988; McLeod 1991). Most observers as-sume that children from divorced familiesreach adulthood with traits that predisposethem to relationship problems, such as adeficit in interpersonal skills, a weak com-mitment to the norm of lifelong marriage, orpersonality characteristics that interfere withrelationship harmony and stability. Premari-tal cohabitation also is more common amongadult children of divorce (Cherlin. Kiernan,and Chase-Lansdale 1995; Furstenberg andTeitler 1994; Le Bourdais and Gratton1998). Cohabiting unions, however, are rela-tively tenuous and short-lived, with a largeproportion dissolving before marriage(Bumpass and Lu 2000). This trend, coupledwith their higher risk of marital dissolution,means that adults with divorced parents ex-perience more union disruptions than doadults with continuously married parents.

Growing up with discordant but continu-ously married parents also appears to be arisk factor for offspring's later marital dis-cord and instability. Several cross-sectionalstudies have shown that people who recallunhappiness in their parents' marriages tendto report less happiness, more conflict, andmore problems in their own marriages(Belsky and lsabela 1985; Booth andEdwards 1990; Overall et al. 1974). In alongitudinal study, Caspi and Elder (1988)found that parents' ratings of marital con-flict (using data collected when childrenwere growing up) were positively associ-ated with children's later reports of conflict

9 O 4 AMERICAN SOCIOLOGICAL REVIEW

in their own marriages. Existing evidence,therefore, suggests that marital discord, aswell as divorce, is transmitted across gen-erations.

Research generally shows that marriedpeople, compared with single people, reportgreater happiness, fewer symptoms of psy-chological distress, more positive self-con-cepts, and better physical health (Aseltineand Kessler 1993; Ross 1995; Waite 1995).Most observers assume that the protectiveeffects of marriage are due to the social sup-port provided by spouses, the capacity ofspouses to monitor one another's health, andthe general benefits associated with partici-pating in an institutionalized relationship(Nock 1998). Regardless of current maritalstatus, the number of divorces people accu-mulate appears to be a factor in poor mentalhealth (Kurdek 1990). Therefore, if indi-viduals with histories of unstable relation-ships are prone to psychological distress,and if growing up in a discordant and un-stable home environment makes it more dif-ficult for children to maintain stable rela-tionships in adulthood, then relationship in-stability may mediate the effects of parentaldiscord and instability on adult offspring'spsychological well-being.

RELATtONS WtTH PARENTS

Parental divorce is associated with relativelyweak parent-child ties in adulthood. Com-pared with adults with continuously marriedparents, adults with divorced parents haveless frequent contact with parents, exchangeless assistance with parents, and describetheir relationships with parents less posi-tively (Amato and Booth 1991, 1997;Aquilino 1994; Cooney 1994; Silversteinand Bengtson 1997; Umberson 1992; Zill etal. 1993). Parental divorce and remarriagealso are associated with early home leavingamong youth—another indicator of tensionbetween parents and children (Amato andBooth 1997; Cooney 1994). Presumably,weak bonds with parents emerge from theturmoil that precedes and follows maritaldisruption. Research consistently shows thatdivorce is associated with fewer expressionsof parental affection, greater parental harsh-ness in dealing with children's misbehavior,and more inconsistency in dispensing disci-

pline (Davies and Cummings 1994;Hetherington and Clingempeel 1992). Al-though divorce appears to weaken children'sties with both parents, most studies indicatethat the consequences are stronger for fa-thers than for mothers, presumably becausemost children reside witb mothers followingmarital dissolution. Studies also show thatinterparental discord—even in the absenceof divorce—is linked with poor relation-ships with parents among young children(Davies and Cummings 1994; Emery 1999;Hetherington and Clingempeel 1992) aswell as adult offspring (Amato and Booth1997; Rossi and Rossi 1990).

Even though problematic parent-child re-lationships may originate in childhood, theytake on new significance as children makethe transition to adulthood—a time whenyouth leave home, complete their educations,form career plans, become economically in-dependent, marry, and begin their own fami-lies. During these critical years, children re-ceive many potential benefits from parents,including emotional support; companion-ship; advice with educational plans, jobs,homes, and family life; practical assistancewith everyday tasks such as child care; andmoney for special purchases, such as a downpayment on a car or home. Parents also con-nect children with kin (such as grandparents)and other adults (such as family friends) whocan serve as sources of support or assistance.These transitional years have become moredifficult in recent decades because of a de-cline in wages for young men, the rising costof housing, and an increase in the cost of acollege education (Amato and Booth 1997).Correspondingly, the number of years thatyouth are economically and emotionally de-pendent on parents appears to he increasing(Furstenberg 2000). For example, recent co-horts of youth have been relatively slow toleave the parental home, and among thosewho do, "returning to the nest" is common(Goldscheider and Goldscheider 1994). In-deed, only when parents reach the last de-cade of the life course does the flow of assis-tance between generations shift primarilyfrom children to parents (Rossi and Rossi1990). Because the early adult years presentmany challenges to youth, and because par-ents represent a key resource for making thetransition to adulthood, weak ties with par-

DIVORCE, MARITAL DISCORD. AND CHILDREN'S WELL-BEING 905

ents may exacerbate existing psychologicalproblems or precipitate new ones.

Consistent with this reasoning, cross-sec-tional research shows that young adults' re-ports of emotional closeness to parents arepositively associated with psychological ad-justment (Amato 1994; Barnett et al. 1991;Barnett, Marshall, and Pleck 1992; Rossiand Rossi 1990; Umberson 1992). Not allstudies are in agreement about the impor-tance of fathers: Some studies find indepen-dent effects of both parents, while otherstudies show no father influences once themother-child relationship is taken into ac-count. Nevertheless, the preponderance ofevidence indicates that the quality of parent-child relation.ships continues to be a salientcorrelate of well-being over the life course.Therefore, if strong parent-child ties pro-mote psychological well-being, and if paren-tal discord and divorce weaken these ties,then the quality of parent-child relationsmay explain the association between paren-tal discord, divorce, and offspring's psycho-logical well-being.

TESTING THE MODEL

Five prior studies have tested aspects of themodel in Figure 1. These studies started withan observed association between parental di-vorce and a measure of offspring well-being,then added mediating variables to the regres-sion equation, with the decline in the divorcecoefficient between models reflecting thedegree of mediation. Glenn and Kramer(1985). using the Genera! Social Survey, andAmato (1991), using 1987-1988 NationalSurvey of Families and Households, foundthat controlling for offspring's education andmarital status resulted in modest reductionsin the magnitude of associations betweenparental divorce and several indicators ofpsychological well-being. Amato (1988), us-ing a national probability sample from Aus-tralia, found thai controlling for educationalattainment reduced the estimated effect ofparental divorce on people's sense of controlby about one half, suggesting a mediatingrole for this variable. McLeod (1991), usingIhe Detroit Area Study, found that parentaldivorce was negatively associated withwomen's reports of marital happiness, andcontrolling for marital happiness eliminated

the association between parental divorce anddepression. Similarly, Ross and Mirowsky(1999) found that parental divorce was asso-ciated with higher depression scores amongmen and women, and five variables—n^duca-tion. economic hardship, early marriage, be-ing in an unhappy relationship, and being inno relationship—mediated tbis association.In summary, these five studies provide somesupport for the mediation model illu.stratedin Figure I. No prior study, however, hasconsidered the possible mediating role ofparent-child relations, despite the fact thatthis variable is related to parental divorce aswell as to adult well-being. Moreover, noprior study has attempted to determine whichprocesses mediate the long-term effects ofparents' marital discord on offspring.

Our study tests the model in Figure 1 usingdata from the 1997 wave of the Marital In-stability over the Life Course Study. Analy-ses of the 1992 wave of data established thatparents' marital discord and divorce are as-sociated with children's psychological dis-tress in adulthood (Amato and Booth 1997).Previous work, however, was not able to de-termine the factors that mediate these asso-ciations. The 1997 wave of data has severaladvantages over the 1992 data for this pur-pose: (I) The sample size is larger, increas-ing from 471 to 691, with a correspondingincrease in the number of parental divorces;(2) the sample is older, and hence a largernumber of offspring had completed their edu-cations, become economically independent,cohabited, married, or divorced; and (3) theaddition ol" new variables to the survey (suchas parents" descriptions of their current rela-tionships with offspring) made it possible tolest the mediating role of parent-child rela-tionships.

The present study also has three supple-mentary goals. The first is to explore genderdifferences in the estimated effects of divorceand discord. Some observers have arguedthat boys suffer more than girls, partly be-cause boys (unlike girls) lose access to thesame-gender parent (Hetherington andClingempeel 1992). In contrast, several stud-ies have reported no gender differences(Amalo 1991; Zill et al. 1993) or found thatthe estimated long-term effects of parentaldivorce are stronger for daughters than forsons (Cooney and Kurz 1996; Glenn and

906 AMERICAN SOCIOLOGICAL REVIEW

Kramer 1985; McLeod 1991; Rodgers 1994).Given the uncertainty In the literature aboutgender differences in the effects of parents"discord and divorce, we test the model sepa-rately for sons and daughters. The secondgoal is to see if weak parent-child bonds inadulthood can be traced to a time when chil-dren were still living with parents, and todetermine, in cases of marital disruption, ifproblematic parent-child relationships werepresent prior to (rather than following) pa-rental divorce. The final goal is to considernot only whether parents divorced, but alsowhether these divorces were followed by ad-ditional family transitions (remarriages anddivorces). Several observers have suggestedthat the number of family transitions towhich children must adjust is a better predic-tor of children's problems than is a singleinstance of marital disruption (Aquilino1996; Pryor and Rodgers 2001; Wu andMartinson 1993). In the present study, weconsider alternative models in which familystructure is treated as an ordered variable (thenumber of transitions) rather than as a di-chotomy (divorced versus not divorced).

METHOD

SAMPLE

Our analysis is ba.sed on a 17-year longitu-dinal study titled Marital Instability over theLife Course. The target population consistedof all married individuals in households inthe contiguous United States with a tele-phone, both spouses present, and bothspouses 55 years of age or less. In 1980,telephone interviewers used random-digit di-aling to select a sample of households and asecond random procedure to determinewhether to interview the husband or wife.Seventeen percent of targeted telephonenumbers could not be reached after 20 call-backs. Of those individuals contacted, 78percent gave complete interviews. The finalsample consisted of 2,033 married persons(not couples). When compared with U.S.census data, the sample was representativeof married individuals with respect to age,race, household size, housing tenure, pres-ence of children, and region of the country.In 1983, 1988, 1992, and 1997, we re-inter-viewed 78 percent, 66 percent, 58 percent.

and 53 percent, respectively, of the originalrespondents. Because of deaths during thisperiod, the 1997 wave included interviewswith 56 percent of living respondents.

A sample of offspring (children of themain respondents) was included as part ofthe 1992 and 1997 waves of data collection.To be eligible, offspring had to (I) be 19years of age or older at the time of the inter-view, and (2) have resided in the parentalhousehold in 1980. Eighty-seven percent ofthe parents in 1992 and 1997 with eligibleoffspring provided the names and telephonenumbers of their children. We obtained in-terviews with 88 percent of those offspringfor whom we had names, for an overallcompletion rate of 77 percent. When parentshad more than one eligible child, a randomprocedure was used to select one child forinclusion in the study. Our analyses pooleddata from three groups: 431 offspring inter-viewed in 1992 and 1997, 40 offspring in-terviewed in 1992 but not in 1997 (mainlybecause we were unable to locate them), and220 offspring who turned age 19 after 1992and were interviewed for the first time in1997. Because the quality of parent-child re-lationships was a central variable in theanalysis, we omitted 36 offspring with a de-ceased biological parent, resulting in a finalsample of 655.

Because the analysis pooled the threegroups of offspring, we checked to see howthese groups differed. The 40 offspring whodropped out between 1992 and 1997, com-pared with the 431 offspring who remainedin the study, scored significantly lower onyears of education, self-esteem, happiness,and life satisfaction (p < .05). The 220 off-spring interviewed for the first time in 1997were younger than the 431 interviewed for asecond time in 1997 (mean age = 23 yearsversus 30 years, respectively). Because ofthis age difference, those interviewed for thefirst time in 1997 were less likely them-selves to have married or to have experi-enced a divorce or a disrupted cohabitation.The three subsamples did not differ on par-ents" marital discord and divorce—the mainindependent variables in our analysis. In pre-liminary analyses, we controlled for mem-bership in the three groups using dummyvariables, but this procedure made little dif-ference to parameter estimates. Conse-

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 907

quently. for the sake of parsimony, we do notcontrol for sub-sample membership in theanalyses reported here.

With respect to attrition, 1,371 respon-dents (parents) had children in the targetedage range living at home in 1980. We ob-tained interviews with offspring (one perfamily) from 691 of these parents in 1992 or1997 (50.4 percent of the families). We re-lied on Heckman's (1979) method to correctfor attrition bias. We began by constructinga probit regression equation to model the at-trition of parents from the panel, based on avariety of potential predictors. Attrition wassignificantly greater among African Ameri-cans, relatively young or old parents, fathers,renters, parents with less education, parentsmarried for fewer years, and parents livingin the South. Marital discord and divorce,along with several additional demographicvariables, were not significant predictors ofattrition. Based on the significant predictorvariables, we calculated lambda to reflectthe probability of dropping out of the panel.We carried out a second probit analysis pre-dicting failure to complete an offspring in-terview, given that parents remained in thepanel, and calculated a second lambda vari-able on this basis. Both variables serve ascontrols.

VARIABLES

We rely on the 1997 wave of data for themajority of offspring variables. As notedabove, a few offspring were interviewed in1992 but not in 1997. Because omittingthese cases could bias our results, we pooledtheir 1992 data (using reports of psychologi-cal well-being and parent-child relationshipsin that year) with the 1997 data from the restof the sample. Most of the parent variableswere obtained from the 1980, 1983, and1988 interviews. Means and standard devia-tions for al! variables appear in Table 1.

INDEPENDENT VARIABLES. Of the 655offspring in the sample, 137 (21 percent) ex-perienced a parental divorce or permanentseparation while growing up. Of these indi-viduals. 50 experienced divorce prior to 1980and were living in stepfamilies during thefirst wave of data collection. Another 87 off-spring experienced parental divorce between1980 and 1997. (Only two cases involved

permanent separation, and we counted theseas "divorces'" in all analyses.) Offspring var-ied widely in age at the time of parental di-vorce: 27 percent were 0 to 4 years old, 23percent were age 5 to 11, 35 percent were age12 to 19. and 15 percent were age 20 to 25.We include the older offspring for concep-tual and empirical reasons. Because educa-tional levels are rising and because manyyouth are postponing full-lime employmentand marriage, many researchers now viewthe period of dependency and "adolescence"as extending well into the third decade of life(Furslenberg 2000). Moreover, Cooney andKurz (1996) and Furstenberg and Kiernan(2001) found that parental divorce when chil-dren were in their twenties was associatedwith elevated levels of depression, less inti-macy with fathers, more relationship insta-bility, and other personal problems. In thepresent study, we carried out analyses bothincluding and excluding offspring in theirearly twenties at the time of divorce, and theresults were nearly identical. Furthermore,age at parental divorce was not related tooffspring's psychological well-being (r= .10.p > .10). Time since divorce also was unre-lated to psychological well-being.

As an alternative to a dichotomous vari-able measuring divorce, we create a variablereflecting the number of family transitionsto which children were exposed prior toreaching adulthood. Among the 137 childrenwith divorced parents, 62 percent saw theirmothers remarry. 62 percent saw their fa-thers remarry, 23 percent saw their mothersexperience a second divorce, and 12 percentsaw their fathers experience a second di-vorce. Counting all parental divorces and re-marriages, the number of family transitionsexperienced by offspring in this sampleranged from 0 to 5.

Three scales (completed by parents) mea-sure parents" marital discord in 1980, 1983,and 1988. The first is a 5-item measure ofmarital conflict ( a = .54). Items asked aboutfights over the household division of labor,fights over child care, the frequency of dis-agreement in general, the number of seriousquarrels in the past two months, and whetherviolence (initiated by either partner) had oc-curred in the marriage. The second scale is a13-item measure of marital problems (a =.78). Questions addressed the existence of

9 0 8 AMERICAN SOCIOLOGICAL REVIEW

Table L Means and Standard Deviations of Variables Used in the Analysis: Marital Instability overIbe Life Course, 1980 to 1997

Variable

Independent VariablesParents' divorce .^

Parents' marital discord '.

Number of family transitions . '

0

1

2

3

4 or 5

Dependent VariablesOffspring's self-esteem

Offspring's psychological distress

Offspring's life satisfaction

Offspring's happiness

Control VariablesOffspring is female

Offspring's age

Offspring is nonwhite

Parenls' education

Mother was interviewed

Mediating VariablesOffspring's education

Offspring's income (in $l,OOOs) ' •

Offspring is married

Number of offspring's disrupted unions

Quality of relations:

Mother-chi I d (parent) •

Father-child (parent)

Mother-child (offspring)

Father-child (offspring)

Early parent-child (parent)

Mean

.21

r24

.79

.05

.05

.06

.05

3.38

1.68

3,72

2.40

.50

27.39

.08

13.87

.62

14.73

18.39

.45

.42

2.71

2.50

2.68

2.47

3.34

StandardDeviation

.41

.43

.41

.22

.22

.24

.24

.41

.35

.60

.54

.50

5.71

.28

2.17

.49

2.40

13.39

.50

.84

.42

.56

.40

.53

.44

Range

0-1

0-1

0^1

Q^I

0-1

2.17-4.00

1.00-3.00

1.86-5.00

1 00-3.00

0-1

19^0

«MH '1 jfc * •

9-220-65

0-1

0-7

1.00-3.00

1.00-3.00

1.00-3.00

1.00-3.00

1.25^.00

Note: Sample includes 655 adult children.

problems in the relationship, such aswhether one (or both) partners is jealous, iscritical, is not home enough, is unfaithful, orhas difficulty controlling anger. The numberof reported problems serves as the scalescore. The third scale is a 13-item measureof divorce proneness (a = .91). Items dealwith cognition (thinking about divorce) aswell as behavior (discussing the possibility

of a divorce with spouse, friends, or familymembers), and the sum of the items servesas the scale score. Correlations between thethree scales (within the same interview year)range from .49 to .57 (all p < .001) (seeAmato and Booth !997 for additional detailson these scales).

To form an overall measure of discord fora particular year, we compute 2-score trans-

DIVORCE, MARITAL DISCORD. AND CHILDREN'S WELL-BEING 909

formations of each scale (to weight themequally) and took the sum. Correlations ofthis composite measure across time arehigh: r = .71 between 1980 and 1983. r =.66 between 1983 and 1988, and r = .59 be-tween 1980 and 1988 (all p < .001). Forparents who remained continuously mar-ried, we calculate the overall discord scoreas the mean of the scores for 1980. 1983,and 1988. For parenis who divorced, werely on the mean of all scores prior to mari-tal dissolution. We split the scale at ap-proximately the 75"^ percentile to distin-guish high-discord couples from moderate-and low-discord couples. This step allowsthe creation of three groups: parents whodivorced (N = 137), high-discord parentswho did not divorce (N = 158), and low- ormoderate-discord parents who did not di-vorce (N = 360). The first two groups arerepresented as dummy variables in theanalysis, and the third group serves as theomitted reference category.

DEPENDENT VARIABLES. We use fourscales to measure offspring's psychologicalwell-being: the Rosenberg (1965) measureof self-esteem (a= .11), the Langner (1962)measure of distress symptoms (a = .73), a7-item scale of satisfaction with various do-mains of life, including job, home, friends,and neighborhood (a = .65), and a single-item rating of overall happiness with life (I= "not very happy," 2 = "pretty happy," 3 ="very happy").

CONTROL VARIABLES. We control foroffspring's gender, age, and race. Half ofoffspring were female, and 8 percent werenonwhite. The ages of offspring rangedfrom 19 to 40, and two-thirds were in theirtwenties. Because parents' education (repre-sented as the mean of the mother's andfather's years of education) could be acause of parental discord and divorce, aswell as of children's well-being, we use itas a control variable in all analyses. Moremothers than fathers served as the inter-viewed parent.

MEDIATING VARIABLES. We use off-spring's years of education and annualearned income as indicators of socioeco-nomic attainment. We rely on two variablesto measure offspring's relationship instabil-ity. One is the offspring's current maritalstatus. Nearly half of offspring (45 percent)

were married at the time of the interview.The second variable is the total number ofdisrupted unions offspring themselves hadexperienced, including prior divorces andcohabitations that ended without marriage.Twenty-nine percent of offspring had expe-rienced one or more union disruptions, re-sulting in a mean for this variable of .42.Because this variable is positively skewed,we also tried a dichotomous version reflect-ing any disrupted relationships, but the di-chotomous variable yielded similar results.

With respect to the quality of parent-childrelations, offspring rated their mothers andfathers on five items, including "How welldo you feel that your mother understandsyou?" and "How much respect does yourfather show you?" (1 = "not very much," 2= "somewhat," 3 = "a great deal"). Otheritems dealt with trust, fairness, and theoverall closeness of the relationship (a =.87 for mothers and .90 for fathers). In1997, the interviewed parent rated the focalchild on three comparable items referring tounderstanding, respect, and overall close-ness ( a = .87). Parents also responded toitems reflecting the relationship betweenthe noninterviewed parent and the focalchild ( a = .87). Because these data werecollected in 1997, in all cases of parents'marital dissolution they reflect the parent-child relationship after divorce. Correla-tions between parents' and children's rat-ings are moderately high: /• = .45 for themother-child relationship and r = .54 for thefather-child relationship (both/7 < .001).'

We also include a variable reflecting thegeneral quality of parent-child relationswhen offspring were living at home. In

' Using one parent to provide proxy data onihe other parent .s relationship with the focalchild was a less than optimal procedure bin bet-ter than having no data al all. When fathers wereinterviewed, children's reports of the mother-child relalionship correlated more strongly withfathers' reports of the mother-child relationshipthan wiih fathers' reports of ihe faiher-child rela-tionship. Similarly, when mothers were inter-viewed, children's reports of the father child re-lationship correlated more strongly wiih mothers'reports of the father-child relationship than withmothers' reports of the mother-child relalionship.These trends support the discriminant validity ofparents' reports.

910 AMERICAN SOCIOLOGICAL REVIEW

1980, 1983, and 1988. parents responded tosix questions about their children, includ-ing, "How often do you wish that youdidn't have to live with your children?" (I= "very often," 4 = "never"), "Would yousay that your children have given you morethan the usual number of problems?" (1 =yes, 0 = no), and "Overall, how close doyou feel to your children? (I = "not close atall," 4 = 'Very close"). To form a generalmeasure of early parent-child relations, weequally weight and average tbe items acrossthe three time periods (a = .65). Eor chil-dren who left home prior to 1988, we basethe score on 1980 and 1983 data only. Incases of divorce after 1980, we base tbescore on the mean of all items answered byparents prior to marital dissolution. We didnot calculate scores for the 50 childrenwhose parents had divorced prior to 1980and hence were living in stcpfamilies inthat year. Note that this variable refers torelations between tbe interviewed parentand all children in Ihe household. Informa-tion about parents* relations with the focalchild in particular would have been prefer-able, but data were not available that farback in time. Eocal children, on average,were 10 years old in 1980 and 18 years oldin 1988.

MISSING DATA. Data were missing forseveral variables, with the greatest percent-age (15 percent) occurring for the parent'srating of the relationship between the childand Ihe other parent. We use three strategiesto deal with missing data. Eirst, we carryout all analyses using listwise deletion ofmissing data. This reduces the sample sizeby 20 percent, however. Second, we use fullmaximum-likelihood estimation with miss-ing data (Arbuckle 1997). Although thisprocedure allows us to include all cases inthe analysis, it is difficult to calculate fitindice.s for models. Third, we rely on miss-ing data imputation using the expectation-maximization algorithm (Allison 2001).This method provides accurate estimateswith up to 50 percent missing data. Dataimputation also allows the use of all casesand the computation of fit indices. Becausethe results of the three methods for han-dling missing data are essentially identical,we present the results based on dataimputation.

RESULTS

Eor tbe main analysis, we rely on structuralequation modeling using the Analysis ofMoment Structures (AMOS) software(Arbuckle 1997). We treat offspring's psy-chological well-being as a latent variablewith four observed indicators: self-esteem,distress, life satisfaction, and happiness. Wealso treat mother-child and father-child rela-tions as latent variables with two observedindicators reflecting parents' and children'sreports. We weight parents' and children'sreports equally by constraining the unstan-dardized loadings to be the same.

Eigure 2 shows the measurement modelfor these three latent variables. All of thecoefficients in the model are statisticallysignificant ai p < .05. Eit indices reveal thaithe model fits the data reasonably well. Thelateni mother-child and father-child rela-tionship variables are positively correlated,suggesting that children who had a positive(or negative) relationship with one parentalso tend to have a positive (or negative) re-lationship with tbe other parent. Moreover,each latent relationship variable is posi-tively correlated with the latent psychologi-cal well-being variable, indicating tbat tbequality of ties witb parents is bound up witbthe offspring's well-being in adulthood.With respect to the relationship variables,note that we allow the error terms for par-ents to be correlated, along with tbe errorterms for offspring. Although the model inEigure 2 does not include correlations be-tween the errors for the indicators of well-being and children's ratings of parent-childrelationships, correlations are included inthe structural model (described below) tocontrol for common method variance.

Table 2 presents the results of six struc-tural equation models derived from sixseparate computer runs. Model 1 shows thatdivorce and marital discord predict lowerpsychological well-being among offspring.Although the Iwo variables account for only4 percent of the variance in offspring well-being, recasting the h coefficients as effectsizes tells a different story. Because thestandard deviation of the latent well-beingvariable (not shown) is .612, the differencebetween offspring with divorced parentsand those with continuously married par-

DIVORCE. MARITAL DISCORD, AND CHILDREN'S WELL-BEING 91 1

Self-esteem Distress Satisfaction Happiness

Relationswith mother

(Parent)

Relationswith mother

(Child)

Relationswith father(Parent)

Relationswith father

(Child)

Figure 2. Measurement Model For Mother-Child Relations, Father- Child Relations, and Offspring'sPsychological Weil-Being

Note: All coefficients are standardized and are signiflcanl at /; < .05 (two-tailed tests); x^ = 44.6, d.f. =17, GFI = .98, CFI - .98, RMSEA = .05.

ents (with low-discord marriages) repre-sents .42 of a standard deviation. The corre-sponding difference for offspring with con-tinuously married, high-discord parents rep-resents .38 of a standard deviation. Theseeffect sizes are moderate in magnitude andlarge enough to be nontrivial (Cohen 1988).

Model 2 includes the five control vari-ables. The h coefficients for parental divorceand marital discord become slightly larger,rather than smaller, indicating a small sup-pression effect. No control variables are sig-nificant. Models 3. 4, and 5 add the proposedmediating variables in separate blocks. (Tosave space, the table does not show ihe equa-tions in wbich mediators served as depen-dent variables.) Model 3 indicates that edu-cation and income are positively associated

with offspring's psychological well-being.Adding these variables to the equation, how-ever, reduces the coefficients for parental di-vorce and marital discord slightly (Model 3versus Model 2), indicating little mediation.Model 4 removes the socioeconomic vari-ables and enters offspring's marital statusand disrupted union.s to the equation. Al-though married offspring report higher psy-chological well-being than do single off-spring, a history of unstable relationships isnot related to well-being. Moreover, includ-ing these variables in the model reduces theb coefficients modestly, indicating little me-diation.

Model 5 reveals that the quality of par-ent-child relations is positively associatedwith offspring's psychological well-being.

912 AMERICAN SOCIOLOGICAL REVIEW

Table 2. Unstandardized Coefficients from the Regression of Offspring's Psychological Well-Bcingon Parents' Divorce and Marital Discord: Marital Instability over the Life Course.1980 to 1997

IndependentVariable

Parents' divorce

Parents' marital discord

Offspring is female

Offspring's age

Offspring is nonwhite

Parents' education

Mother was interviewed

Offspring's education

Offspring's income

Offspring is married

Number of offspring'sdisrupted unions

Mother-child relations

Father-child relations

R- . :

X'd.f.

GFI

CFI

RMSEA

Model 1

-.259*"(.075)

-.234"(.071)

—

—

—

—

—

—

—

—

—

—

—

.039

21.3

8

.989

.973

.050

Model 2

-.261*"(.076)

-.243*"(.072)

.064(.058)

-.003(.005)

-.030(.104)

.010(.013)

-.070(.059)

—

—

—

—

—

—

.046

62.6

23

.983

.925

.051

Model 3

-.246*"(.073)

-.240"*(.070)

.103(.059)

-.015*(.006)

-.028(.101)

-.004(.014)

-.067(.058)

.033*'(.013)

.008"(.003)

—

—1

—

— .

.086

81.8

29

.981

.941

.053

Model 4

- .231"(.074)

-.230"*(.070)

.061(.056)

-.012*(.006)

-.029(-101)

.015(,013)

-.055(.058)

—

—

.274"*(.068)

-.037(.036)

—

.092

84.5

29

.981

.933

.054

Model 5

.088(.107)

-,103(.073)

.047(.058)

-.003(.005)

-.103(.103)

-.004(.013)

-.046(.060)

—

—

—

.731*"(-171)

.345*(.14!)

.259

143.9

48

.972

.941

.055

Model 6

.079(.102)

-.111(.069)

.083(.057)

-.020"(.006)

-.101(.098)

-.003(.013)

-.029(.057)

,017(.013)

.006*(.002)

277".(.067)

-.029(036)

.597"*(.147)

.320*(.136)

.301

186,1

67

.971

.949

.052

Note\ The sample size for all equations is 655. Models are based on maximum-likelihood estimation.Standard errors are in parentheses.

I I

p < .05 p < .01 ' "p < .001 (two-tailed tests)

The effect for mothers is especially strong,with each step on the relationship scale(range = 1 to 3) increasing the latent well-being variable by over one standard devia-tion. Furthermore, adding parent-child rela-tions to the model reduces the coefficientsfor parental divorce and marital discord tononsignificance. The coefficient for divorcechanges direction (from negative to posi-

tive), although it is no longer significantlydifferent from zero, whereas the coefficientfor marital discord is reduced to less thanhalf its value In Model 2. These results sug-gest that a decline in the closeness of par-ent-child relations mediates much of the as-sociation between parental divorce, maritaldiscord, and offspring's psychological well-being in adulthood. Model 6 includes all of

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 913

Table 3. UnNtandardized CoefTicients from the Full Structural Model; Marital Instability over theLife Course, 1980 to 1997

IndependentVariable

Parents' divorce

Parents' marital discord

Offspring is female

Offspring's age

Offspring is nonwhite

Parents' education

Mother was interviewed

Offspring's education

Offspring'.s income

Offspring is married

Number of offspring'sdisrupted unions

Mother-child relations

Father-child relations

R=

Off-spring's

Education

-.320(.223)

.149(.214)

.439"(.173)

.089'"(.019)

-.357(.312)

.340'"(.040)

-.021(.178)

—

—

—

—

—

—

.154

Off-spring'sIncome

.387(1.130)

-.125(1.085)

-6.808""(.879)

1.160"'(.077)

1.123(1.585)

.243(.204)

-.135(.905)

—

—

—

—

—

—

.299

Dependent Variiible

OffspringIs Married

-.031(.045)

-.024(.043)

.003(.035)

.038"*(.003)

.003(.062)

-.026"(.008)

-.058(.036)

—

—

—

—

—

—

.214

DisruptedUnions

.408""(.081)

.099(.078)

-.039(.063)

.032"*(.006)

.115(.113)

-.026(.015)

-.054(.065)

—

—

—

—

—

.086

Relationswith

Father

-.642"''(.045)

-.151"*(.039)

-.037(.031)

.002(.003)

.091(.057)

.012(.007)

- .113"(.032)

—.

—

—

—

—

—

.395

Relationswith

Mother

-.182"*(.038)

- . 1 3 1 " '(.037)

.041(.030)

-.002(.003)

.101(.053)

.015(.007)

.001(.030)

—

—

—

—

—

—

.069

Psycho-logical

Well-Being

.079(.102)

-.111(,069)

.083(.057)

-.020"(.006)

-.101(.098)

-.003(.013)

-.029(.057)

.017(.013)

.006"(.002)

.211'"(.067)

-.029(.036)

.597"*(.147)

.320*(.136)

.301

Note: Sample size is 655. Models are based on maximum-likelihood estimation. Standard errors are inparentheses. ;f= = 186.1. d.f. = 67. GFI = .971. CFI = .949. RMSEA - .052. This model also contains correla-tions between error terms for all mediating variables.

><.O5 "/7<.0I "V<.OOI (two-tailed tests)

the mediating variables. The coefficients fordivorce and marital discord change little be.,tween Models 5 and 6. indicating thatoffspring's education, income, marital sta-tus, and relationship disruptions contributelittle to the mediation model.

To confirm the mediating role of parent-child relations, we use the method developedhy Clogg, Petkova, and Haritou (1995) tocompare regression coefficients. The coeffi-cients for divorce and marital discord inModels 5 and 6 are significantly different ip< .05) from each of the corresponding coef..

ficients in Models 1 through 4. No othercomparisons are significant.

The full set of structural equatiotis forModel 6 is shown in Table 3. The first rowindicates that offspring from divorced fami-lies (compared with offspring with continu-ously married, low-discord parents)achieved about one-third year less educa-tion, although the coefficient is not signifi-cant. Parental divorce is not related tooffspring's earned income or offspring'smarital status, although it is positively andsignificantly associated with offspring's

914 AMERICAN SOCrOLOGICAL REVIEW

number of relationship disruptions. Parentaldivorce is negatively associated with both ofthe latent parent-child variahles. and theseassociations are substantial. Compared withoffspring with continuously and happilymarried parents, offspring with divorced par-ents score one and one-half standard devia-tions lower on father-child relations andabout two-thirds of a standard deviationlower on mother-child relations. (The stan-dard deviations for the latent variables are.424 and .277, respectively.) Parents' mari-tal discord is not related to offspring's edu-cation, income, marital status, or relation-ship disruptions, although it is associatedwith poorer relations with mothers and fa-thers.

Table 3 also reveals a number of signifi-cant associations involving the control vari-ables. Daughters are better educated but earnless money than sons. Older offspring havemore education, have more income, aremore likely to be married, and have experi-enced a greater number of relationship dis-ruptions. Parents" education i.s positively re-lated to offspring's education and negativelyrelated to offspring being married. Eacb ofthese correlations is consistent with prior re-search, as well as with common sense. Fi-nally, the quality of father-child relations islower when mothers, rather than fathers,were interviewed. This association is a re-sult of the fact that mothers rated the father-child relationship less positively than did fa-thers. (The gender of the interviewed parent,however, is not related to the offspring's rat-ings of fathers.) This finding demonstratesthe necessity of controlling for the gender ofthe interviewed parents.

As noted earlier, the model in Figure 1does not specify causal paths between themediating variables. Although not shown inTable 3. however, the model includes corre-lations between the residuals of all media-tors. Significant correlations includeoffspring's education and the number of dis-rupted relationships (r = -.17. p < .001).offspring's education and the quality of fa-ther-child relations (r = .10, p < .05),offspring's education and the quality ofmother-child relations (r = .12, /J < .01),offspring's income and being married {r =.11, /J < .01), being married and the numberof disrupted relationships (r = -.30, /) <

.001). and the quality of relations with moth-ers and the quality of relations with fathers{r - .51, p < .001). All of these correlationsare consistent with prior research.

ASSESSING THE EFFECT OF ATTRITION

To assess whether sample attrition affectedour results, we reran the models in Table 2controlling for lambda 1 (reflecting attritionof parents from the pane!) and lambda 2 (re-flecting failure to obtain an offspring inter-view, given that parents remained in thepanel). Neither lambda variable is associatedwitb offspring's psychological well-being.Adding the lambda control variables toModel 2 in Table 2 reduces the magnitude ofthe h coefficient for divorce by 6 percent andthe h coefficient for discord by 2 percent,and both coefficients remain significant {p <.001). Results are comparable In other mod-els. These findings indicate that adding con-trols for sample attrition results in only mi-nor changes in the estimated effects of di-vorce and marital discord.

GENDER DIFFERENCES

Because some prior studies suggest differ-ences between sons and daughters in thelong-term effects of parental divorce, wetested for gender differences using multi-group models. Structural coefficients forsons and daughters (Model 2 in Table 2) areconstrained to be the same in one model andare allowed to vary in another model, withthe difference in chi-square values serving astbe significance test. This procedure revealsthat the estimated effects of parental divorceand marital discord on well-being are notsignificantly different for sons and daugh-ters. Additional tests reveal that the associa-tions between the parental divorce, maritaldiscord, and the mediating variables do notdiffer significantly between sons and daugh-ters, with one exception. The estimated im-pact of parental divorce on father-child rela-tions is stronger for daughters {h = -.11)than for sons (/7 =- .47, ^-= 13.8, d.f. = \,p< .001). This difference, however, has no im-plications for the mediation process. Forsons as well as daughters, parent-child rela-tions appear to mediate the estimated effectsof divorce and discord.

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 915

Self-esteem Distress Satisfaction Happiness

Parent-childrelations(Time 1)

.15—

.12'")

Maritaldiscord

Parent-childrelations(Time 2)

-.08 (-.08"

Parentreport

Childreport

Figure 3. Unstandardized CoerficienU rrom the Structural Equation Model Showing Links hctweenMarital Discord, Divorce, and Parent-Child Relations at Time 1 and Time 2, andOrfspring's Psychological Well-Being

Note: Numbers in parentheses are b coefficienls for mothers. For fathers, x^ = 39.8, d.f. = 34, GFI = .97,CFI = .98. and RMSEA = .03. For mothers, x^ = 62.2, d.f, = 34, GFI = .98. CFI = .93. RMSEA = .05.

'p < .05 "p < .01 "'p < .001 (two-tailed tests)

THE: R O L E OF EARLY PARENT-CHtLDRELATtONS

To see if problems in parent-child relationscould be traced to a time when children werestill living at home with parents, we con-struct a structural equation model that in-cludes the 1980-1988 measure of parents*relations with children. In this model (shownin Figure 3), we assume that marital discordaffects parent-child relations at time I (^|),subsequent divorce, and parent-child rela-tions in adulthood ((T)- The model also in-cludes the full marital discord scale ratherthan the dichotomized version.

To test this model, it is necessary to omitthe 50 cases of divorce that occurred priorto 1980 (and for whom we had no data onparent-child problems or marital discordprior to divorce), leaving 87 cases of di-vorce. Recall that this parent-child relationsscale is based only on items completed byparents prior to marital dissolution. Modelswere run separately for fathers and mothersbecause the measure of parent-child rela-tions (f|) refers only to the interviewed par-ent and not the interviewed parent's spouse.

The two analyses involve 376 mothers (in-cluding 59 cases of divorce) and 229 fathers(including 28 cases of divorce). Althoughnot shown in the figure, the model includescorrelations between the error term for allvariables based on parents" reports (parent-cbild relations at t\, parent-child relations att2, and marital discord), as well as the corre-lation between the error terms for children'sreports of parent-child relations at tj andchildren's reports of psychological well-be-ing. In addition, although parent-child rela-tions at t^ and marital discord are treated asobserved variables, tbe model incorporatesestimates of measurement error for thesevariables to provide more accurate param-eter estimates.

Figure 3 shows tbe results for fathers andmothers; the unstandardized h coefficientsfor mothers are in parentheses. Note that forfathers as well as mothers, marital discord isnegatively related to parent-child relations inchildhood (/|) as well as adulthood (fo). Inaddition, for fathers and mothers, parent-child relations (t\) are positively related toparent-child relations in adulthood (̂ 2), indi-cating a degree of continuity over time.

916 AMERICAN SOCIOLOGICAL REVIEW

These results suggest that part—but notall—of the negative effect of early maritaldiscord on parent-child relations in adult-hood can be traced to a time when childrenwere still living at home. Not surprisingly,marital discord increases the likelihood ofdivorce, regardless of whether mothers orfather reported on discord. The results formothers and fathers differ in one respect,however. For fathers (but not for mothers)divorce is associated with more negative fa-ther-child relations in adulthood, net ofpredivorce marital discord and predivorcefather-child relations. These findings suggestthat for mothers, the apparent effects of di-vorce are largely accounted for by discordthat preceded marital dissolution. For fa-thers, however, di.scord and divorce appearto have independent effects on later relationswith offspring.

The quality of parent-child reiations inadulthood (t2) is the only significant predic-tor of offspring's psychological well-beingin the model. For fathers, marital discord,early parent-child relations (/|). and divorceappear to have indirect effects on offspring'spsychological well-being through their im-pact on parent-child relations in adulthood.For mothers, marital discord and early par-ent-child relations (but not divorce) appearto have similar indirect effects. These resultssuggest that parent-child relations in adult-hood (net of parent-child relations in child-hood) mediate much of the effect of parentaldiscord and divorce on adult offspring's psy-chological well-being.

1Nu\fBER OF FAMILY TRANSITIONS

Although most studies consider parental di-vorce as a discrete event, one also can con-ceptualize divorce as the first step in a seriesof family transitions involving additionalparental remarriages and divorces. In thisanalysis, we replaced the dichotomous pa-rental divorce variable with the family tran-sition variable, coded as a series of dummyvariables to capture 1. 2. 3. and 4 or moreparental divorces and remarriages. No tran-sitions (that is, parents continuously andhappily married) served as the omitted com-parison group. (See Table 1 for group sizes.)A structural equation model, which includesthe full set of control variables, the dichoto-

mous discord variable, and all 655 cases, re-veals a monotonic association between tbenumber of transitions and offspring's psy-chological well-being. For I. 2, 3. and 4 ormore transitions, the respective b coeffi-cients are -.17 (n.s.), -.19 (n.s.), -,27 (p <.05). and-.42 (p < .01). Interestingly, it wasonly when three or more transitions occurredthat children's well-being was significantlylower than that of children from low-discordtwo-parent families. Comparable results areobtained when the father-child relationsvariable is treated as the dependent variable.For 1. 2, 3, and 4 or more transitions, the re-spective b coefficients are -.53, -.62, -.62.and -.84 (all p < .001). (The coefficients formothers did not vary witb the number oftransitions.) Further analysis reveals that thecoefficients for multiple and single parentaltransitions do not differ significantly, al-though this may be a result of the smallnumber of cases in each group. We cannotsay, therefore, that the estimated effect ofmultiple transitions is significantly greaterthan the estimated effect of a single divorce.Moreover, the small number of cases limitsthe extent to which we can decompose thetransition variable into different types oftransitions involving fathers versus mothers.Nevertheless, these results suggest the util-ity for future studies of treating changes infamily structure as an ordered variable ratberthan as a dichotomy.

DISCUSSION

Our study makes four contributions. First,we have replicated previous research show-ing that children who experience parentaldivorce while growing up tend to report acomparatively low level of psychologicalwell-being in adulthood. The gap in well-be-ing between offspring wiih divorced parenisand offspring witb continuously marriedparents appears for sons as well as fordaughters and is large enough to be non-trivial.

Second, we have shown that parents' mari-tal discord has long-term implications forboth sons and daughters that are comparableto marital dissolution. Although other stud-ies have offered similar findings, our re-search design is particularly strong, withparents reporting on marital discord in the

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 917

1980s and offspring reporting on psycho-logical well-being in the 1990s.

Third, rather than focusing on divorce asa single event, our findings suggest the util-ity of viewing divorce as the first step in aseries of family transitions to which childrenmust adjust. In the present study, over three-fourths of children wbo experienced divorcealso had to contend with at least one paren-tal remarriage, and many of these childrenexperienced a second parental divorce aswell. Our data suggest that children's psy-chological well-being generally declineswith the number of family transitions^afinding consistent with recent work indicat-ing that multiple transitions may be moreimportant than a single divorce in under-standing children's long-term adjustment(Aquilino 1996; Pryor and Rodgers 2001;Wu and Martinson 1993).

Fourth, we have gone beyond prior stud-ies by testing a causal model to explain ihelingering effects of parental discord and di-vorce on offsprings' psychological well-be-ing. One might expect thai children wouldgradually recover from stress in the familyof origin as they leave home and become in-dependent adults. Tbe persistence of a gapin well-being between offspring with di-vorced parents and offspring with continu-ously married parents suggests that somefactor in adulthood continues to disadvan-tage these individuals. We assumed tbat lowsocioeconomic attainment and a hi.story ofdisrupted relationships would be determinedto be the mediating factors, but the analysisdid not support our assumptions. Instead, ouranalysis suggests that parents" marital prob-lems weaken the emotional bonds betweenparents and children in adulthood, and weakparent-child bonds, in turn, place adult off-spring at risk for distress, low self-esteem,and general unhappiness. The present studyis the first to consider and demonstrate themediating role of parent-child relationshipsin linking marital discord and divorce withchildren's long-term well-being.

Our analysis also revealed that problemsin parent-child relationships associated withmarital discord were apparent when childrenwere still living at home. Interestingly, thelink between divorce and weak mother-childrelations was not significant when pre-divorce discord was controlled. But prob-

lems in the father-child relationship ap-peared to result from divorce as well as frommarital di.scord. Although we do not havedata on children's adjustment prior to adult-hood, previous research suggests tbat chil-dren were exhibiting an elevated number ofpsychological and behavioral problems dur-ing this time. A continuation of weak bondsbetween parents and children into adulthoodhelps to explain why psychological difficul-ties associated with early marital discord anddivorce do not go away. Meeting tbe chal-lenges of early adulthood with relativelyliltle support from parents is likely to be akey factor in maintaining, and perhaps am-plifying, earlier levels of distress.

This assumption is consistent witb a largebody of research on children and adoles-cents. Many developmental psychologistsargue that family problems, including dis-cord and divorce, affect children primarilyby disrupting parent-child relationships(Davies and Cummings 1994; Fauber et al.1990; Hetherington and Clingempeel 1992).Sociologists, too, have relied on the qualityof parenting and children's emotional bondswith parents to explain the impact of mari-tal discord and divorce on children and ado-lescents (Conger and Elder 1994; Simonsand associates 1996). Researchers appear tohave assumed, however, that parent-childbonds become less important after offspringreach adulthood. Yet, many studies showthat children's emotional ties to their par-ents continue to be associated with psycho-logical adjustment and subjective well-be-ing throughout the adult years (Amato1994; Barnett et al. 1991; Barnett et ai.1992; Rossi and Rossi 1990; Umberson1992). Moreover, a focus on parent-childrelationships in adulthood is consistent withone of the fundamental insights of lifecourse theory, tbat is, that parents andchildren's lives are inextricably linkedthroughout the life course {Elder 1994). Ofcourse, poorly adjusted children also mayalienate their parents' affection, although itwas beyond the limits of our data to incor-porate this assumption into our statisticalmodel. Nevertheless, the existence of recip-rocal causation does not undermine the no-tion that supportive parents are importantresources for young adults, and that parentshave the potential to contribute (positively

9 1 8 AMERICAN SOCrOLOGICAL REVIEW

or negatively) to their adult children's senseof well-being.

Our findings may reflect the fact that mostoffspring in the sample were relativelyyoung. Parent-child relationships may be-come less central to psychological well-be-ing, and factors such as occupational successand marital stability may become more cen-tral, as people move through the life course.This assumption may account for the failureof our analysis to replicate previous researchshowing mediating roles for offspring's edu-cational attainment and relationship instabil-ity (Amato 1988; McLeod 1991; Ross andMirowsky 1999). In fact, most prior studieshave used community samples of adultsolder than the adult offspring in our study.We do not argue, therefore, that offspring'ssocioeconomic attainment or relationship in-stability are unimportant— only that parent-child relationships (a heretofore neglectedvariable) are part of the process linking dis-cord and disruption in the family of originto psychological difficulties in later life.

It is possible that the links between par-ents' marital discord, divorce, weak bondswith parents, and psychological distressamong adult offspring are a result of theirmutual association with parental patholo-gies, such as alcohol addiction, mental ill-ness, or a violence-prone personality—problems that often co-occur with divorcein dysfunctional families (Fergusson,Horwood, and Lynskey 1994; Kessler andMagee 1993). Although we cannot rule outthis possibility empirically, we do not thinkthat this explanation can account for ourfindings. Divorce may be common amongseriously disordered individuals, but thatdoes not mean that most individuals whodivorce are seriously disordered. Currently,nearly half of all first marriages are pro-jected to end in divorce, and the projectionsare even higher for second marriages.Rather than interpret divorce as a symptomof individual or family pathology, it makesmore sense to view marital dissolution as anormative family transition. Indeed, re-.search indicates that many divorces occuramong couples who appear to be function-ing reasonably well a year or two prior tomarital dissolution (Amato and Booth1997). Because divorce is common andwidely accepted, however, does not mean

that its consequences for children are be-nign.

Some studies have found that the esti-mated effects of divorce are minimal-—evenpositive^—when marital dissolution is pre-ceded by an especially high level of chronicand overt marital conflict, including vio-lence (Amato and Booth 1997; Hanson1999; Jekieiek 1998). In these cases, divorceappears to benefit children by removingthem from a dysfunctional home environ-ment. We could not model interactions be-tween divorce and predivorce marital dis-cord in the present analysis because we usedall cases of marital disruption, includingthose that occurred prior to the original 1980interview. Pooling all cases of divorce into asingle category, regardless of marital qualityprior to divorce, provides an estimate of the"average" effect of divorce. Estimates in ourstudy would have been stronger if we hadeliminated divorces preceded by especiallyhigh levels of conflict.

In conclusion, our study contributes to agrowing literature demonstrating that dis-cord and disruption in the family of origincan have consequences for offspring thatpersist well into adulthood. This findingshould be of genera! interest, given the highdivorce rate in American society since 1980.Furthermore, an increasing number of chil-dren are born within cohabiting unions, andthese unions are even more likely than mar-riages to end in disruption (Bumpass and Lu2000). The divorce rate leveled off. then de-clined slightly, after 1980. If one counts as"marriages" all unions producing children,however, then the divorce rate, from thechild's perspective, continued to rise duringthe last two decades. The result of this trendmay be a gradual weakening of the bondsbetween the generations and a decline in themean psychological well-being of the popu-lation.

Policymakers should keep in mind that re-stricting access to divorce will not addressthe central problem, as chronic marital dis-cord between continuously married parentsappears to be as detrimental as divorce. In-stead, our findings suggest a need for inter-ventions that promote the positive involve-ment of both parents in children's lives fol-lowing divorce, as well as the developmentof cooperative postdivorce relationships be-

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 9 1 9

tween parents. The availability of mediationservices and parent education classes for di-vorcing parents may help to achieve thesegoals (see Emery, Kitzniann, and Waldron1999 for a review of programs and relevantresearch). More generally, the psychologicalwell-being of the next generation of youthwill be enhanced if emerging social trendsor policies lead to an increase in the numberof children raised by parents with stable andharmonious marriages.

Paul R. Amato is Professor of Sociology, De-mography, and Family Studies at Ihe Pennsylva-tiia Stale University. Hi.s research interests in-clude marital (quality and stability, psychologicalwell-being, and inter gene rational relationships.

Juliana M. Sobolewski is a graduate student insociology at the Pennsylvania State University.Her general interests are in family, work andfamily issues, and close relationships. She is cur-rently conducting research on relationships be-tween parents and young adult children.

REFERENCES

Allison. Paul D. 2001. Missing Data. ThousandOaks, CA: Sage.

Amato. Paul R. 1988. "Long-Term implicationsof Parental Divorce for Adult Seif-Concept."Journal of Family Issues 9:201-13.

. 1991. "Parciilal Absence During Child-hood and Depression in Later Life." The So-ciological Quarterly 32:543-56.

. 1994. "Father-Child Relations, Mother-Child Relations, and Oifspring PsychologicalWell-Being in Early Adulthood." Journal ofMarriage and the Family 56:1031 -42.

-. 2000. "The Consequences of Divorce forAdults and Children." Journal of Marriageand the Family 62:1269-87.

Amato. Paul R. and Alan Booth. 1991. "Conse-quences of Parental Divorce and Marital Un-happiness for Adult Well-Being." SocialForces m.%95-9\A.

. 1997. A Generation at Risk: Growing Upin an Era of Family Upheaval. Cambridge,MA: Harvard University Press.

Amato, Paul R. and Bruce Keith. 1991. "Separa-tion from a Parent During Childhood andAdult Socioeconomic Attainment." SocialForces 70:1 87-206.

Aquilino. William S. 1994. "Impact of ChildhoodFamily Disruption on Young Adults' Relation-ships with Parents." Journal of Marriage andthe Family 56:295-3]3.

. 1996. "The Life Course of Children Bornto Unmarried Mothers: Childhood Livina Ar-

rangements and Young Adult Outcomes."Journal of Marriage and the Family 58:293-310.

Arbuckle. James L. 1997. Amos User's Guide.Version 3.6. Chicago. IL: SmallwalersCorpo-ration.

Aseltine. Robert H.. Jr. and Ronald C. Kessler.1993. "Marital Disruption and Depression in aCommunity Sample." Journal of Health andSocial Behavior 34:237-51.

Barnett. Rosalind C. Nazli Kibria. Grace K.Baruch, and Joseph H. Pleck. 1991. "AdultDaughter-Parent Relationships and Their As-sociations with Daughters" Subjective Well-Being and Psychological Distress." Journal ofMarriage and the Family 53:29-42.

Barnett. Rosalind C . Nancy L. Marshall, and Jo-seph H. Pleck. 1992. "Adult Son-Parent Rela-tionships and Their Associations with Son'sPsychological Distress." Journal ofFamilv Is-.sues 13:505-25.

Belsky, Jay and Russell A. Isabella. 1985. "Mari-tal and Parent-Child Relationship in Family ofOrigin and Marital Change Following the Birthof a Baby: A Retrospective Analysis." ChildDevelopment 56:342-49.

Biblarz. Timothy J. and Greg Gottainer. 2000."Family Structure and Children's Success: AComparison of Widowed and Divorced SingleMother Families." Journal of Marriage and theFflfHi/v 62:533-48.

Booth. Alan and John N. Edwards. 1990. "Trans-mission of Marital and Family Quality over theGenerations: The Effects of Parental Divorceand Unhappiness." Journal of Divorce 13:41-58.

Brodzinsky, David, Jennifer C. Hitt, and DanielSmith. 1993. "Impact of Parental Separationand Divorce on Adopted and Nonadopled Chil-dren." American Journal of Orthop.svchiatrv63:51^61.

Brown, George W. and Tirril O. Harris. 1989.Life Events and Illness. New York: Guildord.

Bumpass, Larry and Hsien-Hen Lu. 2000."Trends in Cohabitation and Implications forChildren's Family Contexts in the UnitedStates." Population Studies 54:29-41.

Bumpass, Larry L., Teresa C. Martin, and JamesA. Sweet. 1991. "The Impact of Family Back-ground and Early Marital Factors on MaritalDisruption." Journal of Famil\ Issues 12:22-42.

Campbell, Angus. 1981. The Sense of Well-Be-ing in America: Recent Patterns and Trends.New York: McGraw Hill.

Caspi, Avshalom and Glen H. Elder Jr. 1988."Emergent Family Patterns: The Intergenera-tional Construction of Problem Behavior andRelationships." Pp. 218-40 in Relationshipswithin Families: Mutual Influences, edited by

9 2 0 AMERICAN SOCIOLOGICAL REVIEW

R. Hinde and J. SievensonHinde. Oxford, En-gland: Clarendon.

Cherlin, Andrew. 1999. "Going to Extremes:Family Structure. Children's Well-Being, andSocial Science." Demoi-ruphy 36:421-28.

Cherlin. Andrew J.. Kathleen E. Kiernan. and P.Linsay Chase-Lansdale. 1995. "Parental Di-vorce in Childhood and Demographic Out-comes in Young Adulthood." Demof-raphv32:299-^318.

Cherlin, Andrew J., P. Lindsay Chase-Lansdale.and Christine McRae. 1998. "Eflects of Di-voree on Mental Health throughout the LifeCourse." American Sociological Review63:239-49.

Clogg, Clifford C . Eva Petkova. and AdamantiosHaritou. 1995. "Statistical Methods for Com-paring Regression Coefficients between Mod-els." American Journal of Sociologx 100:1261 -93.

Cohen, Jacoh. 1988. Statistical Power Analysisfor the Behavioral Sciences. Hillsdale, NJ;Erlbaum.

Conger. Rand D. and Glen H. Elder. 1994, Fami-lie.t in Troubled times: Adapting to Change inRural America. New York: Aldine de Gruyter.

Cooney, Teresa. M. 1994. "Young Adults' Rela-tions with Parents: The Influence of RecentParental Divorce." Journal of Marriage andtbe Fatnily 56:45-56.

Cooney. Teresa. M. and Jane Kurz. 1996. "Men-tal Health Outcomes Following Recent Paren-tal Divorce." Journal of Famih Issues 17:495-513.

Davies, Patrick T. and E. Mark Cummings. 1994."Marital Conflict and Child Adjustment: AnEmotional Security Hypothesis." Psychologi-cal Bulletin 116:387-411.

Elder. Glen. H., Jr. 1994. "Time, Agency, andSocial Change: Perspectives on the LifeCourse." Social Psychology Quarterly 57:5-15.

Emery, Robert E. 1999, Marriage, Divorce, andChildren's Adjustment. 2d ed. Beverly Hills,CA: Sage.

Emery. Rohert E.. K. M. Kiizniann, and M.Waldron. 1999. "Psychological Interventionsfor Separated and Divorced Families." Pp.323-44 in Coping with Divorce. SingleParenting, and Remarriage: A Ri.<ik and Resil-iency Perspective, edited hy E. M. Hether-ington. Mahwah. NJ: Lawrence Erlbaum.

Fauber. Roh L.. Rex Forehand, Amanda M. Tho-mas, and Michelle Wierson. 1990. "A Media-tional Model of the Impact of Marital Conflicton Adolescent Adjustmeni in Intact and Di-vorced Families: The Role of DisruptedParenting." Child Development 6\: 1 112-23.

Fergusson. David M.. L. John Horwood. andMichael Lynskey. 1994. "The Childhood of

Multiple Problem Adolescents: A 15-YearLongitudinal Study." Journal of Child Psy-chology and P.iychology 35:1123-40.

Furstenherg. Frank F. 2000. "The Sociology ofAdolescents and Youth in the 1990s: A Criti-cal Commentary." Journal of Marriage and theF«m/7.v 62:896-910.

Furstenberg. Frank F. and Kathleen E. Kiernan.2001. "Delayed Parental Divorce: How MuchDo Children Benefit?" Journal of Marriageand the Family 63:446-57.

Furstenberg. Frank F. and Julien O. Teitler. 1994."Reconsidering the Effecis of Marital Disrup-tion: What Happens to Children of Divorce inEarly Adulthood." Journal of Family Issues15:173-90.

Glenn, Norval D. 1991. "The Recent Trend InMarital Success in the United Slates." Journalof Marriage and the Family 53:261-70.

Glenn. Norval D. and Kathryn B. Kramer. 1985."The Psychological Well-Being of Adult Chil-dren of Divorce." Journal of Marriage and theFamily 41:905-\2.

. 1987. "The Marriages and Divorces ofthe Children t)f Divorce." Journal of Marriageand the Family 49:S]]~25.

Goldscheider. Frances K. and CalvinGoldscheider. 1994. "Leaving and ReturningHome in 20''' Century America." PopulationBulletin 4S',\-3>3.

Hanson, Thomas L. 1999. "Does Parental Con-flict Explain Why Divorce Is Negatively As-sociated wilh Child Welfare? Social Forces77:1283-1316.

Heckman, James J. 1979. "Sample Selection Biasas a Specification Error." Econometrics47:153-61..

Hetherington. E. Mavis. 1998. "Social Capitaland Ihe Development of Youth fromNondivorced, Divorced, and Remarried Fami-lies." Pp. 177-209 in Relationships as

Developmental Contexts: The Minnesota Sympo-sia on Child Psychology, edited by W. AndrewCollins. Mahwah. NJ: Lawrence Erlbaum.

Hetherington, E. Mavis and W. Glenn Clingem-peel. 1992. "Coping with Marital Transitions."Monographs of the Society for Research inChild Development, vol. 57, nos. 2-3. Chicago,IL: University of Chicago Press.

Jekielek. Susan. 1998. "Parental Conflict. Mari-tal Disruption, and Children's Emotional Weli-Being." Social Forces 76:905-36.

Kendler. Kenneth S.. Michael C. Neale, RonaldC. Kessler, Andrew C. Heath, and Lindon J.Eaves. 1992. "Childhood Parental Loss andAdult Psychopathology in Women." .4rr/(/rf.vof General P.sychiatry 49:109-1 16.

Kessler. Ronald C. 1982. "A Disaggregation ofthe Relationship between Socioeconomic Sta-lus and Psychological Distress." y4m(?nVa/i So-

DIVORCE, MARITAL DISCORD, AND CHILDREN'S WELL-BEING 9 2 1

ciuhgical Review 47:752-64.Kessler. Ronald C, and William J. Magee. 1993.

"Childhood Adversities and Aduli Depression:Basic Patterns of Association in a US NationalSiimple." Psychological Medicine 23:679-90.

Kurdek, Lawrence, A. 1990. "Divorce Historyand Self-Reporled Psychological Distress inHusbands and Wives." Journal of Marriageand the Family 52:701-708.

Langner, Thomas, S. 1962. "A Twenty-Two ItemScreening Score of Psychatric Symptoms Indi-cating Impairment." Journal of Health andHuman Behavior 3:269-76.

Lc Bourdais. Celine and Nicole Marcil-Gratlon.1998. "The Impact of Family Disruplion inChildhood on Demographic Outcomes inYoung Adulthood." Pp. 91-205 in LabourMarkets. Social Institutions, and the Future ofCamuhi's Children, edited by M. Corak. Ot-tawa, Canada: Statistics Canada and HumanResources Devetopmeni Canada.

McLanahan. Sara and Larry Bumpass. 1988. "In-tergenerational Consequences of Family Dis-ruption." American Journal of Sociology 94:130-52.

McLanahan, Sara and Gary Sandefur. 1995.Growing Up with a Single Parent. Cambridge,MA: Harvard University Press.

McLeod, Jane D. 1991, "Childhood ParentalLoss and Adult Depression." Journal of HealthandSocial Behavior 32:205-20.

Nock. Steven L. 1998. Marriage in Men's Lives.New York: Oxford University Press.

Overall. John E., B. W. Henry, and Arthur Wood-ward. 1974. "Dependence of Marital Problemson Parental Family History." Journal of Ab-normal Psychology 83:446-50.

Pryor, Jan and Bryan Rodgers. 2001. Children inChanging Families: Life after Parental Sepa-ration. Oxford, England: Blackweli,

Rodgers, Bryan. 1994. "Pathways between Pa-rental Divorce and Adult Depression." Journalof Child Psychology and Psychiatry 35:1289-1994.

Rosenberg, M. 1965. Society and AdolescentSelf-image. Princeton NJ; Princeton University

Ross, Catherine E. 1995. "ReconceptualizingMarital Status as a Continuum of Social At-tachment." Journal of Marriage and the Fam-ily 57:129-40.

Ross. Catherine E. and John Mirowsky. 1999."Parental Divorce. Life-Course Disruption,and Adull Depression." Journal of Marriageand the Family (}]:\0M-45.

Ross, Catherine E. and Chia-ling Wu. 1995. "TheLinks between Education and Health." Amer/-can Sociological Review 60:719-45.

Rossi, Alice and Peter Rossi. 1990. Of HumanBonding: Parent Child Relations across theLife Course. New York: Aldine de Gruyter.

Silverstein, Merril and Vern L. Bengtson. 1997."intergenerational Solidarity and the Structureof Adult Child-Parent Relationships in Ameri-can FamlWes." American Journal of Sociology103:429-60.

Simons, Ronald L. and Associates. 1996. Under-standing Differences between Divorced and In-tact Families. Thousand Oaks, CA: Sage.

Snarey, John R. 1993. How Fathers Care for theNext Generation: A Four-Decade Study. Cam-bridge, MA: Harvard University Press.

Sun, Youngmin. 2001. "Family Environment andAdolescents' Well-Being Before and AfterParents' Marital Disruption: A LongitudinalAnalysis." Journal of Marriage and the Fam-ily 6^: 691-1 ]3.

Umberson, Debra. 1992. "Relationships betweenAduli Children and Their Parents: Psychologi-cal Consequences for Both Generations." Jour-nal of Marriage and the Family 54:664-74.

Waite, Linda. 1995. "Does Marriage Matter?"Demography 32:483-507.

Wu. Lawrence L. and Brian C. Martinson. 1993."Family Structure and the Risk of a PremaritalBirth." American Sociological Review 58:210-32.

Zill, Nicholas, Donna R. Morrison, and Mary J.Coiro. 1993. "Long-Term Effects of ParentalDivorce on Parent-Child Relationships, Ad-justment, and Achievement in Young Adult-hood." Journal of Family Psychology 7:91-103.