Marital Disruption, Parent-Child Relationships, and Behavior Problems in Children

Marital Disruption, Parent-Child Relationships, and Behavior Problems in Children JAMES L. PETERSON NICHOLAS ZILL Child Trends, Inc. This study examines the effects of marital disruption on children's behavior, accounting for variations in postdisruption living arrangements and the effects of parent-child relationships and marital conflict. The study is based on a 1981 national sample of 1,400 children aged 12-16. Disruption was associated with a higher incidence of several behavior problems, negative effects being greatest with multiple marital transitions. The negative effects are lower if the child lives with the same-sex parent following divorce or maintains a good relationship with one or both parents. High, persistent conflict in intact families is al{o related to behavior problems. More than half of all children of elementaryschool age interviewed in the recent National Survey of Children said they feel afraid when their parents have arguments. Similarly, the proportion of children reported to have had a seriously upsetting experience is highest for those whose parents' marriages have been disrupted or whose parents' relationships are characterized by high conflict (ZiII, 1978). The same data also show that children's mental health is related to marital disruption and the level of conflict in the family (Zill and Peterson, 1983). To know why and to what extent marital disruption and parental conflict affect children is of This paper was presented at the annual meeting of the Society for Research in Child Development, April 1983. The authors acknowledge research support from Grants No. MH-34707 and No. MH-38664, National Institute of Mental Health, U.S. Department of Health and Human Services, and from a grimt from the Foundation for Child Development. Appreciation is also extended to Frank F. Furstenberg, Jr., a collaborator on the National Survey of Children, and to the anonymous referees who provided helpful and thoughtful comments on an earlier version of this article. Child Trends, Inc., 1990 M Street, N.W., Washington, DC 20036. practical importance to parents and to clinicians who seek to help parents in conflict; it is relevant for those who set policies that affect the functioning of families; and it relates to some central theoretical issues in the field of mental health. A major theoretical concern is the development of models that explain the role of family functioning in the development of behavior disorders in children. Research in this area has demonstrated the usefulness of examining interspousal and parent-child relations as factors influencing the development of behavior problems in children (Maccoby, 1980). More specifically, marital conflict and disruption have been found to be sources of behavior problems (Hetherington, Cox, and Cox, 1977; Wallerstein and Kelly, 1980; Zill and Peterson, 1983; Levinger and Moles, 1979; Levitin, 1979). For example, Hetherington, Cox, and Cox (1977) found that in the year following divorce, children became more dependent, disobedient, aggressive, and demanding, and less affectionate. The effects peaked at about one year and by two years had diminished for boys and largely disappeared for girls. Parents' childrearing behavior changed as well: parents became less consistent, less affectionate, and less able to control their children's behavior. Wallerstein and Kelly (1980) also found negative and more lasting effects of marital disruption Journal of Marriage and the Family 48 (May 1986): 295-307 295 - -- - - - - - -- - ---

296 JOURNAL OF MARRIAGE AND THE FAMILY on children. Five years after the divorce, one-third of the children in their sample were still seriously disturbed, and another third were having psychological difficulties. Even ten years later, many of the children who were preschoolers at the time of the divorce continued to struggle with emotional issues stemming from the divorce (Wallerstein, 1984). More favorable outcomes were found when the child had good relationships with both parents, and the parents were psychologically healthy. Rutter (1971), studying English children, found that marital discord accounted for many deleterious effects of separation and divorce, at least for boys. In particular, the longer the discord, the greater the antisocial behavior. The cessation of discord was followed by a decline in problem behavior. A good relationship with one parent mitigated but did not eliminate the effects of parental discord, a finding also supported by Hess and Camara (1979). Looking at parent-child interactions, Schwarz (1979) suggests that parent-child relationships in middle childhood are especially important for adjustment and personality development. Marital strife may seriously upset the way by which parent-child relationships normally lead to appropriate and well-adjusted behavior. Specifically, conflict between parents may place the child in a double approach-avoidance conflict in which closeness to one parent introduces the risk of rejection by the other. Then the child who cannot choose between the two parents is in danger of becoming disordered (Gassner and Murray, 1969; Schwarz and Getter, 1979). For the child who becomes stably allied with the parent of the same sex, Schwarz hypothesizes that the child becomes chauvinistic and alienated from the opposite sex. In the opposite-sex situation, the child experiences conflict and has problems with gender identity and lower self-esteem. Finally, the child who rejects both parents is likely to become antisocial because of weak socialization. Our present research is directed toward specifying in more detail how much and in what ways marital conflict and disruption lead to behavioral difficulties in children. Specifically, we are testing a series of hypotheses suggested by the research to date, as follows: 1. Child outcomes will be more adverse the higher the level of parental conflict, and the longer the duration of conflict. 2. Disruption and conflict will have stronger and more lasting effects on boys, leading especially to antisocial behavior. Girls, on the other hand, will be more likely to become depressed or withdrawn. 3. The effects of conflict and disruption will be worse when the conflict alienates the child from both parents. This situation is especially likely to lead to antisocial outcomes. 4. Intermediate effects will result when the child is able to maintain a strong, positive relationship with one parent. 5. The best outcomes will occur when the child, despite the disruption and conflict, is able to maintain a positive relationship with both parents. Sample METHODS The data on which the analyses are based are from the National Surveys of Children, which comprise two waves of a longitudinal study of children in the United States. The first wave, conducted in the fall and winter of 1976-77, was based on a national probability sample of households containing children aged 7-11. Altogether, information was gathered on 2,301 children in 1,747 households, this being a completion rate of 80070.Interviews were conducted with each eligible child and that parent most able to provide information about the child, usually the mother. Data were also collected by questionnaires sent to the children's teachers. The second wave of the survey was conducted in the spring and summer of 1981, when the children were 12 to 16 years old. Limitations in funding precluded reinterviewing the entire sample. In all, follow-up data were gathered on 1,423 children, or nearly 80% of those chosen for reinterview. Further information about the sample and the study design may be found in previous reports (Furstenberg, Nord, Peterson, and ZiII, 1983; Zill and Peterson, 1983). Since the focus of this paper is on the role of parent-child relationships in mediating the effects of marital disruption, the present analysis is limited to children both of whose parents were living at the time of the second survey, and who were living with at least one biological parent (or with two adoptive parents) and had had some contact with each parent within the last five years. These children comprise 76% of the sample. Within this study group, 20% were living with only one of their biological parents because of a separation or divorce. The mean time since the initial separation was eight years; and the mean age of the children at separation, six years. Yet much experience of disruption was fairly recent. Among children whose parents separated or divorced, one in four had this happen in the five-year interval between --. -----.

DIVORCE AND BEHAVIOR PROBLEMS 297 the two surveys, and one in two experienced some transition (first or subsequent) in this period. Measures Children were classified into three broad groups on the basis of their living arrangements (see Table I): those living with both biological parents (or two adoptive parents), those living with their biological mothers but not fathers (a step-, foster, or adoptive father might also be in the home), and those living with their biological fathers but not their mothers. The first group was further subdivided by the level of marital conflict, defined by combining items measuring marital happiness and marital arguments. Marriages were characterized as having low conflict if the parent respondent described the marriage as "very happy" and reported marital arguments in not more than two of nine areas that were covered. Marriages were characterized as having high conflict if the marriage was described as "not too happy" or if arguments were reported in five or more of the nine areas. All other situations were classified as having moderate conflict. Using the same conflict measure from Wave I, we divided children from highconflict situations in 1981 into those who experienced high conflict in 1976 as well (persistent high conflict) and those who did not (recent high conflict). Children living with their biological mothers were further ciassifed according to her marital status: single-parent mothers (separated, divorced, or never married); remarried; and redisrupted after a second or subsequent marriage. The number of children in the sample who lived with their biological father but not mother was too small to permit a similar subdivision. The parent-child relationship was measured by using the child's responses to a set of four items about each biological parent (whether present in the house or not). These items were: closeness; frequency of doing things together; amount of affection received; and aspiration to be like the kind of person the parent is. The items correlate well (the average correlation among pairs of variables is.50 for the fathers and.33 for the mothers), indicating that the items appear to tap the same underlying dimension. To form a summary variable of the relationship with both parents, we summed the items for each parent to form a scale that was then dichotomized so that at least half the sample fell in the upper part of each dichotomy. The final measure, a cross-classification of the dichotomized scales, had four categories: having a positive relationship with both parents; with the mother only; with the father only; or with neither parent. By this measure, half the children in the sample have a positive relationship with both parents, one in five has a positive relationship with the mother only; one in five, with neither; and about one in ten, with the father only (see Table I, first row). Five outcome variables were used. Three of these are subscales of a behavior problems index that was adapted from a longer checklist developed by Achenbach and Edelbrock (1981) to predict mental and behavior problems that may require therapy. We used three criteria to select items from the checklist: reliability; high loading on one of the subscales that emerged from their analyses, and adaptability to an interview situation. Factor analysis of the items in the National Survey of Children yielded four factors, similar to ones found in Achenbach's work. Three of the four factors appear to represent fairly serious behavior problems. These are depressed/withdrawn behavior, antisocial behavior, and impulsive / hyperactive behavior. Each of these subscales is composed of six items. (The fourth factor, measuring milder and more general behavior problems, is not used in this analysis.) While factor analysis was used to confirm which items belonged together in a subscale, a simple additive index was used to create each scale. This was done so that the scale score would have a straightforward meaning: namely, the number of items (from 0 to 6) that the parent reported as characterizing the child. All three measures are skewed, as might be expected. The means range from 1.0 (antisocial score for girls) to 2.3 (impulsive/hyperactive score for boys). The remaining two outcome variables are single items measuring behavior problems at school. These are whether, in the last four years, the child has had "any behavior or discipline problems at school resulting in your receiving a note or being asked to come in and talk to the teacher or principal," or has been "suspended, excluded, or expelled from school." RESULTS Previous analyses of the National Survey of Children have established that marital disruption and conflict are associated with psychological problems in children, some of which may persist for several years (ZiII, 1978; Zill and Peterson, 1983). The present analysis extends these results to other outcome measures. The aim is to learn whether these negative outcomes may be ameliorated by the child's maintaining a positive relationship with one or both parents and, given a positive relationship with one parent, to learn whether the sex of the parent or thesexof the -

298 JOURNAL OF MARRIAGE AND THE FAMIL Y child in relation to that of the parent is of any consequence. Marital Status and Parent-Child Relationships The disruption of a marriage by separation or divorce may undermine the relationship a child has with each parent. These weakened relationships, in turn, may contribute to some of the negative outcomes found among children from disrupted families. The drastic curtailment in the level of contact between the child and the noncustodial parent may weaken the child's relationship with that parent. Indeed, earlier analyses of data from this same survey found the level of contact to be strongly and positively associated with how close the child felt to the noncustodial parent. However, nearly half of the children who had a noncustodial parent had had virtually no contact with that parent in the year prior to the survey, and only one in six had seen him (in over 90070of cases the noncustodial parent was the father) as regularly as once a week (Furstenberg et ai., 1983). As a consequence of such limited contact, the quality of the parent-child relationship suffered for most of these children. The relationship with the custodial parent may also suffer. In most cases the custodial parent has to take on an even greater share, if not all, of the burden of childrearing. She may take on or increase her level of employment out of economic necessity, thereby limiting the amount of time that can be spent in interaction with children. And the parent's ability to maintain a good relationship with her children may deteriorate as a result of the stress often associated with divorce. On balance, however, the relationship with the noncustodial parent may suffer more than the relationship with the custodial parent, because of the lack of contact. The data in Table I show that marital disruption is associated with poorer parent-child relationships, according to the child-based measure we used (p <.01, based on x'). Fifty-five percent of the children in intact families maintain a positive relationship with both parents, but only 25.JJo of those living with their mothers and 36% of those living with their fathers do so. The relationship with the noncustodial parent is especially likely to suffer. While 60% of those who live with a custodial mother report a positive relationship with her, only 36% of these same children report a positive relationship with the absent father. Similarly, 69% of those living with a custodial father have a positive relationship with him, com- TABLE I. RELATIONSHIPWITH PARENTS, BY CHILD'S living ARRANGMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Percenlage of Children Having Positive Relationship with: Both Mother Father Neither N Parents Only Only Parent (Unweighted) All childrena 49% 19070 12070 20070 1,080 Living arrangementsb Child lives with both biological (or adoptive) parents 55 15 12 18 813 62 13 12 13 345 51 12 23 320 50 20 II 19 96 20 29 28 52 Child lives with biological mother 25 35 II 29 232 Separated, divorced, or never married 26 32 12 30 132 32 39 9 20 74 Redisrupted 8 36 10 46 26 Child lives with biological father Sexc 36 22 33 9 28 Female 41 24 10 25 551 Male 57 13 16 529 athe data are based on children for whom the parent-child relationship variable is defined: namely, those having both biological (or adoptive) parents living, having had contact with each parent within the last five years, and having complete data on the items making up the scale. These children constitute 76070of all the children in the sample. Losses due to incomplete data are minor. bthe association between living arrangements and parent-child relationships is significant (x'. P <.01). cthe association between sex and parent-child relationships is significant (x'. P <.01).

f DIVORCE AND BEHAVIOR PROBLEMS 299 pared with 57070having a positive relationship with the absent mother. These last two sets of figures show that the relationship with the mother is more robust in the face of absence. Only a third of children maintain a positive relationship with absent fathers, compared with well over half with absent mothers. This difference may stem from the higher levels of contact that absent mothers maintain in comparison with absent fathers. Within intact marriages the relationship between the child and each parent, especially the father, suffers as the level of conflict rises. Especially harmful to good relationships is persistent high conflict (that is, high conflict at both survey dates). Thus 62070of the children in lowconflict families have a positive relationship with both parents, compared with only 29070of those in persistent high-conflict situations. The latter figure is actually comparable to the proportions for children in disrupted families. Since conflict usually attends marital disruption, much of the negative effect of separation or divorce may be a result of the associated conflict. Indeed, among children whose parents' marriages were disrupted between the two waves of the survey, those in a high-conflict situation at Wave I had nearly three times the rate of psychological distress between the two waves as those in a low- or moderate-conflict situation at Wave I (Zill and Peterson, 1983). After a disruption, the parent-child relationship appears to be strongly associated with the marital status of the custodial mother. In particular, the remarriage of the mother is associated with a modest improvement in the relationship with the mother and, to a lesser extent, the still-absent father. The disruption of a second marriage, on the other hand, is associated with markedly lower proportions having a positive relationship with either parent. Less than one child in ten from redisrupted families has a positive relationship with both parents, compared with about three in ten when the mother has either stayed remarried or not remarried at all. It is unclear from the present data whether these differences in proportions are changes over time induced by marital events (remarriage, redisruption), are due to the slightly younger age at initial disruption of children from redisrupted families, or are the result of selection factors (perhaps mothers more able to maintain good personal relationships manage to remarry and stay remarried, while those less competent remarry only to divorce or become separated again). Certainly, all these factors could contribute to the difference simultaneously. Sex and Parent-Child Relationships We might expect that, through the process of developing an appropriate sexual identity, girls will come to ha'{e a closer relationship with their mothers, and boys, with their fathers. This is clearly the case for girls, as can be seen in Table I. Thus 65% of girls have a positive relationship with their mothers, versus 51% with their fathers. The situation for boys is virtually even, with 70% and 71% having a positive relationship with their mothers and fathers, respectively. For children from disrupted families, however, the effect of the disruption appears to be more powerful than whether the parent is of the same sex (data not shown). Both boys and girls who live with their mothers are more likely to have better relationships with their mothers than with their fathers. For those who live with their fathers, the results follow a same-sex pattern, boys having better relationships with their fathers and girls having slightly better relationships with their mothers. Overall, girls report poorer relationships with their parents than do boys. In both intact and disrupted families fewer girls than boys report positive relationships with both parents, and more report having a positive relationship with neither parent. While this may indicate a real sex difference in the quality of parent-child relationships, it could also be that boys and girls hold somewhat different standards for a positive relationship. Parent-Child Relationships and Child Outcomes Good parent-child relationships can ameliorate the negative effects of marital disruption only to the extent to which such relationships are associated with the outcomes in question. If the effects of positive relationships are at all additive, then having a positive relationship with both parents should produce the best results; with one parent, intermediate results: and with neither parent, the least desirable results. Given a positive relationship with only one parent, does it matter which parent that is? As Schwarz indicated, in the series of hypotheses reviewed earlier, it may depend on which outcomes one is looking at, and on the sex of the child in relation to that of the parent with whom the positive relationship is maintained. Whether or not this is the custodial parent may also be of importance. Table 2 presents the data for the five outcome variables we have examined. The data are presented separately for boys and girls because of sex differences in the outcome variables. In this regard, we note that girls score higher than boys on the depressed/withdrawn scale, but boys score

300 JOURNAL OF MARRIAGE AND THE FAMIL Y higher on all the other measures. These results are consistent with suggestions that behavior problems in girls are more likely to take the form of overcontrolled behavior, whereas in boys they take the form of undercontrolled behavior. The results may also prefigure the higher rates of depression of adult women in comparison with men. As expected, children who have a positive relationship with both parents are least likely to exhibit any of the problems measured. For boys, the worst outcomes usually occur when the child has a positive relationship with neither parent. An exception is the high rate of impulsive/hyperactive behavior for boys with a positive relationship just with the father. Given a positive relationship with only one parent, it makes little difference in three of the five outcomes which parent the relationship is with. In the other two cases-impulsive/hyperactive behavior, and being suspended or expelled -the outcome is somewhat worse if the relationship is with the father. The situation for girls is more complex. As with boys, the best outcomes are usually found for girls having a positive relationship with both parents. (For suspended/expelled, the "father only" category is just as low.) And in three of the five outcomes, the worst situation occurs when the girl has a positive relationshp with neither parent. But for antisocial behavior and impulsive/hyperactive behavior, having a positive relationship only with the father leads to the worst outcomes. It may be that since these two behaviors are especially associated with males (boys' average scores on these scales are much higher than the girls'), a father's influence, untempered by that of the mother, might increase these behaviors among girls. Yet the same argument could be made about the school misbehavior measures, and on these, girls who have a positive relationship just with the father score strikingly low. Another factor may come into play for these school variables. Parents, perhaps especially fathers, tend to have a double standard for girls' behavior, being much less tolerant of misbehavior in public than in private. For example, data from the National Survey show that girls exhibit problem behavior much less often than boys at school, but are only marginally better than boys on similar behaviors at home. If fathers do hold such a double standard, then girls who have a positive relationship only with the father may well exhibit more problem behavior of the sort that parents can observe, but less of the kind that is likely to come the attention of school personnel. Marital Status, Child Outcomes, and Parent-Child Relationship We consider next the associations between the marital situation of the child's family and the TABLE 2. SOCIALAND PSYCHOLOGICALOUTCOMES, BYSEX AND RELATIONSHIPWITH PARENTS, U.S CHILDREN AGED 12-16, 1981 Outcome Measures (parent reports) Females Depressed / withdrawn (mean)b Antisocial behavior (mean)b Impulsive / hyperactive (mean)b School note: misbehavior (OJo)C Suspended /expelled (OJo)C N (unweighted) Males Depressed/withdrawn (meaq)b Antisocial behavior (mean)b Impulsive/hyperactive (mean)b School note: misbehavior (07o)C Suspended /expelled (%)C N (unweighted) All Childrena 1.32 1.00 1.87 12 6 551 1.08 1.35 2.28 22 II 529 Children Having Positive Relationship with: Both Mother father Neither Parents Only Only Parent asee footnote a, Table I. bscores for the three scales range from 0 to 6, higher scores indicating more problem behavior. The standard errors of the means range from.06 to.20. The mean scores for "both parents" are all significantly less than those of the other three categories (p'::;'.05, using a one-tailed t test) except the impulsive/hyperactive score of males in the "mother only" category. cthe standard errors of the percentages range from 1.0 to 5.2. The percentages for "both parents" are significantly less than each of the other categories for males (p.::;..05, using a one-tailed t test). for females they are significantly less than the "neither parent" category for both variables, and the "mother only" category for the school note variable..88.75 1.42 5 3 220.87 1.02 2.09 13 4 274 1.31 1.00 1.99 15 6 6 1.17 1.72 2.24 29 18 82 1.76 1.40 2.50 8 3 53 1.31 1.63 2.89 29 22 78 1.86 1.26 2.24 23 II 132 1.56 1.97 2.45 44 21 95

DIVORCE AND BEHAVIOR PROBLEMS 301 TABLE 3. MEAN DEPRESSED/WITHDRAWN SCORES, BY CHILD'S LIVING ARRANGEMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Living Arrangements eta (column I); beta (columns 2, 3) Total variation explained N (Unweighted) 526 405 176 153 47 29 107 57 36 546 407 168 167 49 23 125 75 38 12 (I) Unadjusted Males 1.08 0.99 0.68 1.09 1.52 1.82 1.65 2.10 1.34 0.93 0.83 0.293 8.6070 Females 1.32 1.25 0.80 1.64 1.42 1.98 1.53 1.61 1.46 1.29 1.83 eta (column I); beta (columns 2, 3) 0.251 Total variation explained 6.3070 Mean Scores (2) Controlsa 1.00 0.67 1.10 1.59 1.87 1.57 2.02 1.23 0.89 0.82 0.292 12.6070 1.22 0.73 1.62 1.49 2.00 1.64 1.65 1.70 1.51 1.77 0.276 13.4070 acontrol variables are parent education, family income, race, and child's age. blnc1udes living with two adoptive parents. -These means are significantly different from that for the intact, low-conflict category (p < tests were calculated on the third column only. (3) Controls and Parent/Child Relationship 1.02 0.72 1.10-1.58-1.81-1.48-1.92-1.24-0.67 0.76 0.268.0070 0.246 17.3070.05). Significance various outcome measures, and the extent to which these associations are modified by taking the parent-child relationship into consideration. To examine these associations we made use of multiple classification analysis, a form of dummy multiple regression appropriate for use when the dependent variable is an interval scale (or a dichotomy) and the predictor variables are categorical. Each of the five outcome variables was used in turn as the dependent variable in a series of analyses. For each dependent variable three analyses were done. The first used the child's living arrangements (defined in terms of the parents' marital status and level of conflict) as the lone predictor. The second analysis added a set of background control variables-parent education, 1.24 0.79 1.60-1.54- I. 90-1.57-1.59-1.64-1.32 I. 77- family income, the child's age, and race-in addition to living arrangements. Finally, the third analysis added the parent-child relationship variable to the list of predictors to enable us to determine how much the addition of this variable explained the association between marital conflict or disruption and child behavior outcomes. Data from these analyses are presented in Tables 3-7, one table for each outcome variable. Column I in each table shows the outcome-variable means for each category of the child's living arrangements, unadjusted for any other predictors. Column 2 shows the means adjusted for the control variables; and column 3, adjusted for the controls and the parent-child relationship measure. Separate sets of means are presented for

302 JOURNAL OF MARRIAGE AND THE FAMILY TABLE 4. MEAN ANTISOCIALSCORES, BY CHILD'S LIVING ARRANGEMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Mean Scores (I) (2) (3) Controls and N Parent/Child Living Arrangements (Unweighted) Unadjusted Controlsa Relationship Males 526 1.35 405 1.23 1.26 1.29 176 0.92 0.92 1.01 153 1.26 1.32 1.33. 47 1.66 1.75 I. 75. 29 2.75 2.62 2.48 107 2.01 1.86 1.68. 57 2.30 2.09 I.90. 36 1.72 1.67 1.58 1.76 1.52 1.20 1.15 1.30 1.22 eta (column I); beta (columns 2, 3) 0.308 0.280 Total variation explained 9.5070.1070 547 408 169 167 49 23 125 75 38 12 Females 1.00 0.93 0.76 1.05 0.98 1.60 1.21 0.93 1.65 1.54 1.40 eta (column I); beta (columns 2, 3) 0.199 Total variation explained 4.0070 0.97 0.79 1.09 1.08 1.51 1.05 0.70 1.63 1.39 1.46 0.199 9.0070 acontrol variables are parent education, family income. race, and child's age. blncludes living with two adoptive parents..these means are significantly different from that for the intact, low-conflict category (p < tests were calculated on the third column only. 0.234 17.1070 0.98 0.81 1.08. 1.11 1.48. 1.02 0.68 1.60. 1.34 1.43 0.191 11.7070.05). Significance males and females because of the interactions between sex, parent-child relationships, and living arrangements. Depression/withdrawal. Children are least depressed and withdrawn when they live with both parents rather than the biological mother only (see Table 3, column I). When they live with their fathers, an interesting sex difference appears. In this case boys are at least as well off as those boys living with both parents, whereas girls are even more depressed and withdrawn than those living only with their mothers. However, the small sample necessitates caution in interpreting the means for this group. For children in intact families, the level of parental conflict strongly affects the level of depression and withdrawal. The depressed/withdrawn scores rise strongly with level of conflict, the scores for those in a situation of persistent conflict being even higher than for those who live with just one of their biological parents. Comparing girls and boys, it appears that girls may be a bit more sensitive to moderate levels of conflict. Among children living with their biological mothers but not fathers, the depressed/withdrawn scores are especially high for those living with a single mother (separated, divorced, or never married). This is especially so for boys, which is consistent with previous research findings on the greater vulnerability of boys to marital disruption. In comparison. living with a remarried mother is associated with lower depressed/withdrawn scores; the difference is especially large for boys. While these data are cross-sectional and do

DIVORCE AND BEHAVIOR PROBLEMS 303 TABLE 5. MEAN IMPULSIVE/HYPERACTIVE SCORES, BY CHILD'S LIVING ARRANGEMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Living Arrangements eta (column I); beta (columns 2, 3) Total variation explained N (Unweighted) 526 405 176 153 47 29 107 57 36 547 408 169 167 49 23 125 75 38 12 (I) Unadjusted Males 2.29 2.17 1.66 2.49 2.67 3.26 2.89 2.86 2.90 2.91 2.38 0.285 8.1% Females 1.88 1.84 1.62 1.94 2.10 2.50 2.00 2.10 1.77 2.01 2.11 eta (column I); beta (columns 2, 3) 0.131 Total variation explained 1.7% Mean Scores (2) Controlsa 2.21 1.73 2.46 2.76 3.28 2.74 2.63 2.97 2.59 2.16 0.261 16.4070 1.85 1.59 1.94 2.22 2.64 1.96 2.01 1.88 1.89 2.10 0.5 11.9% acontrol variables are parent education, family income, race, and child's age. blncludes living with two adoptive parents. -These means are significantly different from that for the intact, low-conflict category (p < tests were calculated on the third column only. (3) Controls and Parent/Child Relationship 2.22 1.78 2.45-2.76-3.25-2.71-2.56-3.03-2.47 2.02 0.249 17.5% 1.88 1.67 1.94 2.24-2.55. 1.85 1.93 1.77 1.65 2.03 0.126 15.4%.05). Significance not directly test the causal effect of remarriage, they are consistent with the notion that gaining a stepfather improves outcomes for boys (the case of girls is addressed below). We expected redisruption to be associated with even higher levels of depression and withdrawal, an expectation not supported by the data. However, the numbers of cases in this category were quite small, so the means are subject to considerable sampling variability. Adding background variables to the model (Table 3, second column) does not appreciably change the overall pattern of results. For boys, the strength of living arrangements as a predictor is unchanged (both beta and eta are.29); for girls, it is actually strengthened slightly. The additional variables do, however, raise the proportion of variation explained in the depressed / withdrawn score from 9070to 13% for boys, and from 6% to 13% for girls. Introducing controls into the analysis produces one slight yet noteworthy change in the pattern of results. The adjusted depressed/withdrawn score for girls living with remarried mothers is slightly higher than for girls living with a single mother. For the unadjusted scores it is moderately lower. Santrock, Warshak, Lindbergh, and Meadows (1981) noted evidence for negative reactions of girls, but not boys, to the remarriage of their mothers. While our data do not indicate that girls with remarried mothers are substantially more depressed and withdrawn than girls with single mothers (the adjusted score is higher, but not significantly different), the data do indicate that a,...

304 JOURNAL OF MARRIAGE AND THE FAMIL Y TABLE 6. PERCENTAGE RECEIVING SCHOOL NOTE, BY CHILD'S LtVING ARRANGEMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Mean Scores (I) (2) (3) Controls and N Parent IChild Living Arrangements (Unweighted) Unadjusted Controlsa Relationship Males 525 22.2Clfo - - t' 404 19.0 19.7 20.9 Low connict 176 18.6 19.7 22.8 Moderate connict 152 18.0 18.0 18.3 High connict, recent 47 11.7.0 13.6 High connict, persistent 29 44.6 42.5 37.8 107 37.7 33.9 27.6 57 39.6 33.0 26.2 36 38.7 40.2 37.8. 29.2 22.6 9.6 33.4 34.5 32.2 ( eta (column I); beta (columns 2, 3) 0.221 0.189 0.158 Total variation explained 4.9070 10.1070 16.5070 Females 546 11.8070 - - 1\ 407 9.9 9.6 9.9 Low connict 169 8.1 7.7 9.3 Moderate connict 166 9.5 9.0 8.1 High connict, recent 49 10.8 12.8 13.2 High connict, persistent 23 29.3 25.9 23.1 125 20.3 21.6 20.0. 75 16.5 17.8 16.4 38 32.5 34.4 33.8. 12 9.9 9.5 4.9 3.8 2.6 4.3 C eta (column I); beta (columns 2, 3) 0.211 0.222 0.205 Total variation explained 4.4Clfo 10.9Clfo 15.1070 acontrol variables are parent education, family income, race, and child's age. blnc1udes living with two adoptive parents..these means are significantly different from that for the intact, low-connict category (p <.05). Significance tests were calculated on the third column only. [( remarriage is more troublesome for girls than for boys, which is consistent with Santrock's findings. We can test the ameliorative effects of a good parent-child relationship by adding this variable to the model and examining what happens to the predictive power of the living-arrangements variable. When this is done (compare columns 2 and 3 of Table 3), the explanatory power of living arrangements is reduced, but the magnitude of the reduction is modest at best. The betas drop from.29 to.27 for boys and from.28 to.25 for girls. The overall pattern of results is unchanged; however, the range of differences is somewhat reduced, consistent with the reduction in the betas. Significance tests, comparing the mean for each category with the mean for those in intact lowconflict families, show that most means are significantly higher. The exceptions are children living with mothers after a redisruption, and boys living with their fathers. Antisocial behavior. The results for antisocial behavior parallel those of depression/withdrawal in several respects (see Table 4). Specifically, antisocial behavior is higher for those living with their mothers than for those in intact families. Among those with their fathers, antisocial behavior is even higher for girls, whereas it is down to the levels of intact families for boys. In addition, conflict in intact families is strongly and positively related to antisocial behavior; and persistent high conflict is especially detrimental. The overall pattern of relationships is little changed by the addition of control variables and the parent-child relaa h r a h 1 - -------- -"-

DIVORCE AND BEHAVIOR PROBLEMS 305 TABLE 7. PERCENTAGE SUSPENDED OR EXPELLED, BY CHILD'S LIVING ARRANGEMENTSAND SEX, U.S. CHILDREN AGED 12-16, 1981 Living Arrangements eta (column I); beta (columns 2, 3) Total variation explained N (Unweighted) 526 405 176 153 47 29 107 57 36 acontrol variables are parent education, family income, race, and child's age. blncludes living with two adoptive parents..these means are significantly different from that for the intact, low-conflict category (p < tests were calculated on the third column only. (I) Unadjusted Males 11.1% 8.8 5.8 9.8 11.9 21.3 25.0 24.3 26.7 23.2 4.9 0.215 4.6% Females 546 5.4% 407 3.6 169 2.5 166 3.6 49 5.6 23 10.7 125 12.6 75 8.6 38 19.5 12 16.3 0.0 eta (column I); beta (columns 2, 3) 0.204 Total variation explained 4.2070 Mean Scores (2) Controlsa 9.9 6.2 12.1 13.7 17.4 18.3 13.9 27.1 12.4 9.9 0.16 15.1% 4.5 3.6 4.2 7.5 8.6 9.2 3.7 19.7 11.6 1.6 0.177 13.0% (3) Controls and Parent/Child Relationship 10.6 8.2 12.0 13.8.6.9 9.7 26.1 6.3 6.3 0.134 18.8% 4.5 4.0 3.7 7.7 7.7 9.2 3.6 20.1. 10.9 2.7 0.179.6%.05). Significance tionship variable. As before, the addition of the latter does somewhat weaken the predictive power of living arrangements, at least for boys, indicating that a good relationship does have a modest ameliorative effect in the face of conflict and disruption. One notable difference between the results for antisocial behavior and depressed/withdrawn behavior is that girls living with single mothers are no worse off than those in intact, low-conflict families. Perhaps antisocial behavior, more often a male than a female characteristic, is less likely to occur in the absence of a father figure in the household. In contrast, girls with remarried mothers exhibit the very highest levels of antisocial behavior. These results are again consistent with Santrock's notion that the remarriage is troublesome for girls and may also reflect the addition of a male figure to the household. Boys, it can be seen, actually fare somewhat better with remarried mothers than with single ones. Impulsive/hyperactive behavior. A similar pattern of results emerges once again from the analysis of impulsive Ihyperactive behavior (see Table 5). As before, the detrimental effects of conflict, especially persistent conflict, can be seen. Among children in disrupted families there are differential effects according to the sex of the child in relation to the sex of the custodial parent, with boys doing relatively well if they live with their fathers, and girls, if they live with their mothers. The addition of the parent-child relationship variable to the model again reduces the -

306 JOURNAL OF MARRIAGE AND THE FAMILY effect of living arrangements, much.. but not by very In contrast to the previous two variables, impulsive/hyperactive behavior is more common among boys living with remarried mothers than among those with single mothers, whereas the reverse is true for girls (although the difference is small). Significance tests for the full model indicate that for girls the effect of living arrangements on impulsive/hyperactive behavior is fairly weak. Only the two categories for high conflict in intact families have mean scores significantly higher than that for girls in low-conflict intact families. Schoo/ behavior problems. The results for the two variables measuring school behavior problems are less clear-cut (see Tables 6 and 7). While higher levels of conflict in intact families generally result in higher proportions with behavior problems, the results are not always as linear or as strong as with the first three variables (see especially the boys). Living in an intact family is generally best, except that girls living with their fathers have especially low rates of behavior problems. This is in sharp contrast to the earlier pattern of results. Finally, the rates of behavior problems are especially high for both boys and girls living with a remarried mother. In most cases, the addition of the parent-child relationship variable does reduce the strength of living arrangements as a predictor ance. and also adds to the explained vari- The time frame for the two school questions is broad. Each refers to problems anytime in the period between the two waves of the survey-a span of four years. This vagueness in the time frame may have muddied the results, as some of the marital transitions (though a minority) occurred after the school behavior problems. Also, the two school measures, though based on parent reports, are objective measures of actions taken by school personnel. As such they are indicators of fairly extreme public behavior. This kind of behavior may be less predictable from such variables as conflict, marital disruption, and parent-child relationships. DISCUSSION In line with our hypotheses, the data show that marital disruption is associated with a range of negative outcomes for children. Both overcontrolled and undercontrolled behavior are more prevalent among children who had experienced some form of marital disruption. Equally important, the data demonstrate that marital disruption is not a simple dichotomy. The postdisruption living arrangements and marital history make a great deal of difference for children's behavior. Marital conflict in intact homes, especiaily if persistent, appears to be as harmful as disruption itself. Furthermore, poor parent-child relationships lead to more negative child behavior, yet maintaining good relationships with parents can go some way in reducing the affects of conflict and disruption. Beyond these findings, which are generally consistent with our original hypotheses, the results raise several issues and new research questions that merit discussion. Remarriage of the custodial mother. For girls, living with a remarried mother. rather than one who remains single, is associated with more behavior problems on all measures except impulsive/hyperactive behavior. Why is this so, especiaily when a remarriage brings the advantages of a second parent, usually a higher income, and the promise of some stability in living arrangements? And why shouldn't boys experience the same degree of difficulty? Remarriage is itself a transition to which the child must adapt. Though eventuaily beneficial, the transition may cause shortterm disturbances. Indeed, other analyses of this data support the notion that in time the behavior of a child living with a remarried parent is likely to improve. That girls have more difficulty with this transition may stem from having lived with the same-sex parent. These girls may come to identify more strongly with their mothers, forming a bond that may be threatened by the arrival of a new father. Sex of the custodial parent. The results suggest that a child living with a parent of the opposite sex is especially prone to problem behavior. This finding must be viewed with caution because the numbers in the sample who are living with their fathers are quite small. But if the results can be corroborated, the implications for such issues as child custody are quite important. RecaIl that the children in this survey were adolescents, a time when the establishment of an appropriate sexual identity is an especially important aspect of development. Perhaps at these ages living with a parent of the opposite sex is detrimental to progress with this developmental stage. Age may playa minor role in another way. On the average, boys living with their fathers tend to be slightly older than girls or than boys living with their mothers. If the behavior problems we measured decline with age, this may account for some of the low rates of behavior problems among boys living with fathers. Such an age-related decline is clearly evident for impulsive/hyperactive behavior but not so much for the other variables. Sex differences. We expected boys to be more vulnerable to the effects of disruption and con-

DIVORCE AND BEHAVIOR PROBLEMS 307 flict, especially in areas where boys are particularly prone to behavior problems (such as antisocial behavior). Similarly we expected girls to be most susceptible to depression and withdrawal. The data are not entirely consistent with these hypotheses. The unadjusted effects of the conflict / disruption measure (living arrangements) are stronger for boys in all five areas, including depression / withdrawal. However, the effects for the two school measures are quite close for boys and for girls. Furthermore, once controls, including the parent-child relationship, are introduced, both school measures show stronger relationships to disruption/conflict for girls than for boys. In part, the stronger effects on boys may be due to differences introduced by the sex of the custodial parent. That is, because most boys live in a relatively high-risk living arrangement after disruption, compared with girls, the effect of the living arrangement is bound to be stronger for boys. The stronger adjusted effect of disruption and conflict on girls' school behavior is more puzzling. Usually, girls' public behavior (as at school) is more controlled than their private behavior, whereas for boys this difference is minimal. Perhaps under the stress of conflict or disruption, the greater control girls maintain in public breaks down. Or perhaps teachers hold different standards for girls and that an equivalent rise in misbehavior by girls and boys is more likely to lead to discipline for girls. Parent-child relationships. The data support our hypothesis that positive relationships with parents can ameliorate the negative effects of marital conflict or disruption. But the moderating influence is modest. These moderating effects are above and beyond any associated with the previously introduced controls for parent background. In fact, the impact of the relationship variable, in terms of the reduction in betas, is generally as large or larger than that of the entire group of background variables. We examined the interrelationships among these variables more closely by separate analyses within each category of the parent-child relationship variable. Though small samples cloud the results, these analyses show reductions in negative effects associated with better parent-child relations for children in high-conflict or disrupted family situations. The analysis and discussion has focused on the effects of conflict and disruption and on sex differences in effects. However, we must not lose sight of the fact that among all the statuses considered, children are clearly better off living in intact families with low to moderate levels of conflict, and that the majority of children enjoy just such a situation. While the incidence of behavioral disturbances is higher in cases of conflict, disruption, and poor relationships, in each case only a minority exhibited such behavior. REFERENCES Achenbach, Thomas and C. Edelbrock. 1981. "Behavioral problems and competencies reported by parents of normal and disturbed children aged four through sixteen." Monographs of the Society for Research in Child Development, Serial No. 188. Furstenberg, Frank F., Jr., C. Nord, J. Peterson, and N. Zili. 1983. "The life course of children of divorce: Marital disruption and parental contact." American Sociological Review 48 (October): 656-668. Gassner, Suzanne, and E. Murray. 1969. "Dominance and connict in the interactions bel ween parents of normal and neurotic children." Journal of Abnormal and Social Psychology 74: 33-41. Hess, Robert, and K. Camara. 1979. "Post-divorce family relationships as mediating factors in the consequences of divorce for children." Journal of Social Issues 35: 79-96. Hetherington, E. Mavis, M. Cox, and R. Cox. 1977. "The aftermath of divorce." In J. H. Stevens, Jr., and M. Matthews (eds.), Mother-Child, Father-Child Relations. Washington, DC: National Association for the Education of Young Children. Levinger, George, and O. Moles. (eds.). 1979. Divorce and Separation: Context, Causes, and Consequences. New York: Basic Books. Levitin, Teresa. 1979. "Children of divorce." Journal of Social Issues 35 (Fall): entire issue. Maccoby, Eleanor. 1980. Social Development: Psychological Growth and the Parent-Child Relationship. New York: Harcourt Brace Jovanovich. Rutter, Michael. 1971. "Parent-child separation: Psychological effects on the children." Journal of Child Psychology and Psychiatry 12: 233-260. Santrock, John W., P. Warshak, C. Lindbergh, and L. Meadows. 1981. "Children's and parents" observed social behavior in stepfather families." Child Development 53: 472-4110. Schwarz, J. Conrad. 1979. "Childhood origins of psychopathology." American Psychologist 34: 879-883. Schwarz, J. Conrad, and H. Getter. 1979. "Parental factors in neuroticism: A triple interaction model." Unpublished manuscript. Wallerstein, Judith. 1984. "Children of divorce: Preliminary report of a ten year follow-up of young children." American Journal of Orthopsychiatry 54: 444-458. Wallerstein, Judith, and J. Kelly. 1980. Surviving the Break-up: How Children Actually Cope with Divorce. New York: Basic Books. ZiII, Nicholas. 1978. "Divorce, marital happiness, and the mental health of children." Paper presented at the NIMH workshop on divorce and children, Bethesda, MD. ZiII, Nicholas, and J. Peterson. 1983. "Marital disruption and the child's need for psychological help." Report prepared for NIMH.