Sibling Conflict and the Development of Conduct Problems
Submitted January 28, 1998
Introduction
Many researchers have shown the importance of sibling relationships in child development (see Dunn, 1983, for a review). Research suggests that siblings can be important contributors to each others' acquisition of pro-social skills and to the development of social understanding (Bryant & Crockenberg, 1980; Dunn, Brown, & Maguire, 1995; Youngblade & Dunn, 1995). Siblings are often each others' first playmates, and this early play may be the medium by which they influence each other in these areas. Theorists also believe that siblings play a role shaping the development of antisocial behavior from early to middle childhood (Patterson, 1984, 1986). For example, Patterson has hypothesized that coercive cycles may be a mechanism by which sibling interactions contribute to the development of externalizing behavior problems.
Much of the research on siblings and development focuses on sibling conflict, its causes, and its antecedents (Brody, Stoneman, & Burke, 1987; Stocker, Dunn, & Plomin, 1989; see Brody, 1998 for a review). Implicit in this research is the idea that extreme sibling conflict and aggression have implications for child adjustment; however, a connection between sibling conflict and conduct problems has yet to be demonstrated. It is probable that most sibling conflict is normal and harmless. However, prolonged aggressive and aversive conflict between siblings (i.e., destructive sibling conflict) may play a role in the development of conduct problems, or may be an indicator of coercive family processes associated with the development of conduct problems. Because antisocial behavior is relatively stable (Olweus, 1979), difficult to treat, and resistant to change (Kazdin, 1995), it is important to explore its early markers and signs. Aversive and prolonged (i.e., destructive) sibling conflict may prove to be one such marker that will increase our ability to identify and treat antisocial behavior.
The Influence of Siblings
Research has confirmed that siblings imitate each other, and that most often younger children imitate their older siblings rather than the reverse (Pepler, Abramovitch & Corter, 1981; Lamb, 1978). Because siblings often spend large amounts of time together, especially during early childhood, it makes sense that the behavior of an older sibling will affect the younger child's acquisition of certain behaviors, such as cooperating, sharing, and acting aggressively. For example, in the area of prosocial behavior, older siblings' prosocial behavior during sibling interaction is predictive of helping, sharing, and cooperating in younger siblings (Bryant & Crockenberg, 1980). Sibling interaction is also predictive of the development of later social and emotional understanding for (Dunn, et al., 1995; Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991; Youngblade & Dunn, 1995). Dunn and Munn (1986a) found that high levels of prosocial behavior by older siblings during sibling interaction were associated with more conciliating behavior and cooperating in the younger children 6 months later than in children whose older siblings did not behave as prosocially. They also found that teasing or justifying in a high proportion of sibling conflicts by older siblings was associated with the younger siblings justifying their own behavior and referring to rules at the 6 month follow-up (Dunn & Munn, 1986b). These findings show that sibling interaction is related to the development of prosocial behaviors and mature conflict resolution tactics in younger siblings.
A logical hypothesis that follows from this research is that if siblings can influence the development of prosocial behavior, they can also influence the course of disruptive and aggressive behavior. However, few researchers have examined whether sibling interaction is related to child conduct problems, and, if this relation exists, whether it is simply a marker for other processes (e.g., parenting) or whether it contributes uniquely to the prediction of conduct problems.
Coercion, Sibling Interaction, and Disruptive Behavior
Patterson's work, which links sibling interaction to the development of aggressive behavior, approaches the subject from a transactional perspective. In Patterson's model, parents unwittingly reinforce child oppositional and aggressive behavior through inconsistent and/or punitive discipline practices, which in turn, are exacerbated by the child's increasingly aversive responses. He dubbed these interactions coercive cycles; these cycles maintain and are reinforcing to the target child's problem behavior (Patterson, 1984, 1986, Patterson, Dishion, & Bank, 1984). In his work with families of clinically-referred boys, Patterson found that siblings were also aversive in their initiation of interactions, and that they were the family members most likely to become involved in coercive cycles with the target child. In his 1984 study, he analyzed these coercive family interactions to examine the role siblings played in the identified child's deviant behavior. Patterson's hypothesis was that in homes where family management practices had been disrupted, both children would exhibit problem behavior, and that the additional stress of a coercive older sibling would make the younger child even more coercive. The results of the study supported this hypothesis to some extent. Older siblings from abusive families and those with a clinically referred child were more coercive than siblings from normal families and were only slightly less coercive than the younger (referred) child. Siblings were involved in 80% of coercive exchanges with the target child that began as neutral events (e.g. fighting out of boredom). Patterson concluded from these data and previous studies that siblings can train younger children to be coercive by modeling and then reinforcing aversive behavior. A longitudinal design would improve the ability to predict later problem behavior; however, Patterson's data were cross-sectional and correlational, making it impossible to ascertain whether the older sibling was influencing or was influenced by the problem behavior of the younger child. It is also possible that these effects could be accounted for by other factors predictive of conduct problems such as parenting.
Dunn and Munn (1986b) examined a similar question with a longitudinal study of a nonclinical sample of siblings. Their study examined the relations between sibling conflict, physical aggression, and other conflict resolution strategies. They found that physical aggression by the sibling was negatively related to child conciliation and positively related to child physical aggression 6 months later. This was one of the few studies to associate aggressive behavior of a sibling with later aggressive behavior of the target child. Another interesting finding from this work is that a high level of maternal intervention in sibling conflicts was associated with physical aggression in the target child at follow-up. From the analyses performed it is difficult to tell if it was the sibling's behavior, mother's behavior, or both that accounted for increased levels of physical aggression in the target child. Dunn and Munn pointed to these results as suggestive evidence for Patterson's model showing siblings to be "shapers" of physical aggression (Patterson, 1984).
Longitudinal analyses that examine families whose children are at risk for conduct problems would permit more extensive tests of Patterson's sibling-training model. Beginning at school-age with a clinical sample, as Patterson did, makes it difficult to determine how the antisocial behavior developed initially. In addition, Patterson did not allow for alternate pathways in his model. A child who shows oppositional behavior in late infancy, possibly from genetic sources, may stress parenting resources, precipitating the breakdown of family management techniques. Bidirectional, transactional models have the potential to account for individual and interactional contributions of each family member. Patterson's data do suggest that coercive cycles may be a key component in the development of antisocial behavior, and that siblings are an important part of this process.
While the work of Patterson, and Dunn and Munn, suggest that sibling interactions may contribute to the development of antisocial behavior, important questions remain. There has been little work specifically linking qualitatively "destructive" sibling conflict with conduct problems. This link would be important not only for Patterson's model, but also for research examining gene-environment interactions in which sibling interaction has been hypothesized to be a factor in the development of antisocial behavior (Plomin & Daniels, 1987; Pike, McGuire, Hetherington, Reiss, & Plomin, 1996). Whereas sibling conflict may be statistically related to child conduct problems, it is also important to examine factors that may account for its association with these problems. Early child problem behavior and parenting are two variables that could potentially account for this relation. Early conduct problems frequently remain relatively stable throughout childhood, and this behavior often predicts later behavior problems (Campbell, 1990; Richman, Stevenson, & Graham, 1982). Rejecting parenting, which is characterized by hostility and a lack of warmth, is also associated with the development of conduct problems in children (Renken, Egeland, Marvinney, Mangelsdorf, & Sroufe, 1989; Zahn-Waxler, Iannotti, Cummings, & Denham, 1990). It is especially important to test whether rejecting parenting accounts for the relation between destructive sibling conflict and conduct problems, because in Patterson's model, disrupted parenting practices are one of the chief mechanisms by which coercive cycles begin and are spread to sibling interactions. It is important to measure parenting and child behavior at an early age (e.g., age 2) in order to account for these effects on later child behavior before siblings are playing together extensively and influencing each other both behaviorally and cognitively as well as receiving increasingly less adult supervision.
It is unclear from Patterson's model whether parenting, previous child behavior, and sibling interaction each contribute unique variance to the development of conduct problems, or whether variance contributed by sibling interaction is actually accounted for by one or both of these other variables. Therefore, it is important to determine whether sibling conflict continues to contribute unique variance in explaining conduct problems once the effects of earlier child externalizing behavior and parenting have been taken into account. Another possibility is that the interaction of destructive (aggressive and prolonged) sibling conflict and rejecting parenting contributes unique variance beyond that accounted for by the main effects of these three variables. In his model, Patterson posits that these two variables help to create and exacerbate coercive cycles of interaction in families and subsequent behavior problems in young children (Patterson, 1984, 1986). Another possibility is that sibling conflict takes on a different meaning in the presence of parental rejection. In this context, siblings may be aggressive towards each other because of rivalrous competition for scarce parental affection (Stocker, Dunn, & Plomin, 1989; Volling & Belsky, 1992).
In this study, two hypotheses involving the relation between sibling conflict and conduct problems were tested. First, we hypothesized that destructive sibling conflict would contribute unique variance to the prediction of conduct problems after controlling for early behavior problems and rejecting parenting. Based on the work of Patterson, and Dunn and Munn, it was expected that destructive sibling conflict would be directly related to aggressive and delinquent behavior at ages 5 and 6. These two types of antisocial behavior were examined separately to explore the specific types of behaviors that would be related to destructive sibling conflict. There is reason to believe that both aggressive behavior, or reactive aggression, and delinquent behavior, or proactive aggression, would be related with sibling conflict through modeling. Secondly, following Patterson's model, in which a child learns anger control and social (or antisocial) behaviors from his parents that are then reinforced through sibling interaction, it was hypothesized that the interaction between destructive sibling conflict and rejecting parenting would contribute unique variance to the prediction of conduct problems, over and above these factors' main effects. This is essentially a rival hypothesis, because a significant interaction would qualify any main effects found.
Method
Subjects
One hundred and eighty families participated in this study from a larger study of 310 low-income families focusing on the developmental precursors of antisocial behavior in boys. Boys were chosen because of their higher risk of developing conduct problems (Keenan & Shaw, 1997; Richman et al., 1992). The sample was recruited from the Women, Infant, and Children Nutritional Supplement Program (WIC) in the Pittsburgh metropolitan area, which provides nutritional resources to low-income families. At the time of recruitment, mothers were required to have at least two children at home, with the male target child approaching 1.5 years old. The average educational level of mothers and fathers at recruitment was 12.5 years, and the mean per capita income was $2,892 per year. The sample consisted of Caucasian (54%), African-American (40%), and mixed or Hispanic families (6%) (Shaw, Winslow, Owens, & Hood, 1998). Of this smaller sample, 57% are Caucasian, 37% African-American, and 6% Other. The majority of the mothers, 58%, identified themselves as married or living together, 27% stated they were single (never married), and 14% were separated, divorced, or widowed.
Observational and maternal report data were collected at the lab and/or home when the target child was 1.5, 2, 3.5, 5, and 6 years old. At each visit, families were reimbursed for their participation. Data for the current study are from the age 2, 5, and 6 year assessments. Attrition for mother-rated data for the entire sample from age 2 to 5 was low, with some data available on 94% of the sample at age 5. Of these families, 230 or 74% were able to participate in the sibling interaction paradigm. Data for teacher-reported outcomes at age 6 were available for 180 children (65% of the sample with age 5 data, 58% of the original sample). Because of strict criteria for inclusion in the study, sample sizes ranged from 180 to 115. One of the inclusion criteria was that the target child have a close-age sibling who was between one year younger and four years older, and who was able to participate in the study at the 5-year visit. Of the 180 siblings, 28 were one year younger, 48 one year older, 48 two years older, 39 three years older, and 17 four years older than the target child. This age range was chosen in order that the siblings would be close enough in age to be playmates, less likely to be in a caretaker role (as may happen with much older siblings), and similar in developmental range to the target children. This decision was supported empirically by using an ANOVA to test for age differences in the amount of destructive sibling conflict among children one year younger to four years older, which was nonsignificant, F (5, 176) = 1.81, ns (post hoc Tukey comparisons were also nonsignificant). Of this sample, 97 (54%) of the siblings were boys.
Procedure
The families were visited in their homes when the target child was 5 years old. The target child and his sibling were videotaped playing with sets of toys, while the examiner and mother completed questionnaires in the same room. The sibling dyad was videotaped for 1 hour during which the siblings played with up to three sets of toys brought by the examiner. The toys were (1) a castle with a shooting cannon and knights, (2) a Lion King set with moveable animal figures, (3) and Bendits bendable building materials. The children were allowed to change toys every 20 minutes, or they could continue playing with the same toy for the whole hour. However, they both had to agree to change before the assistant could introduce the next toy (based on Volling & Belsky, 1992). The mother was told that neither the examiner nor the assistant would intervene in the children's play, and that she should interact with them as she normally would. Sibling conflict was coded from these videotapes using the system described below.
Also as part of the age 5 assessment, mothers completed the CBCL. Follow-up reports of child behavior were also available from mothers (CBCL) and teachers (Teacher's Report Form; TRF) when the target children were age 6.
In addition, we used an observed parenting task and maternal report of the CBCL from the age 2 lab visit. The age 2 assessment involved a combined home-lab visit lasting approximately 3.5 hours. Parenting data were collected during the second half of the assessment in the lab (see description in Measures section).
Measures
Maternal and Teacher Report
Child Behavior Checklist (CBCL) and Teacher's Report Form (TRF) (Achenbach, 1992; Achenbach, 1991 respectively). These versions of the Achenbach behavior checklists are the most widely used reports of childhood behavior problems. In the present study, data from the age 2-3 version of the CBCL collected at the age 2 assessment and the age 4-16 parent version completed by mothers at ages 5 and 6 were used. Teachers also completed the TRF when the target children were 6 years old. This enabled the use of comparable scores across setting and time for each child. After age 3, scales are normalized by sex and age of child. Suggested clinical cut-off t-scores discriminated between children referred and not referred for mental health services (Achenbach, 1991, 1994).
For the present study, the narrow-band Aggression and Delinquency factors were used instead of the broad-band Externalizing scale in analyses because of our interest in how sibling conflict might lead to different forms of conduct problems. While the Aggression factor focuses primarily on symptoms related to reactive aggression and impulse control (e.g., "temper tantrums or hot temper" and "screams a lot"), the Delinquency scale is focused primarily on proactive antisocial behavior (e.g., "swearing or obscene language" and "stealing"). Alpha coefficients for the standardization sample for the CBCL Aggression and Delinquency scales are .92 and .74 respectively, and .96 and .70 for the same TRF scales (Achenbach, 1991, 1992).
Behavioral Observation
Rejecting Parenting. Maternal rejecting parenting was measured at age 2 using the Early Parenting Coding System (EPCS), which was designed to measure a range of parenting behaviors typically exhibited in interactions with young children (Shaw, Winslow, Owens, Vondra, Cohn, & Bell, 1998). The EPCS is an observational coding system consisting of 9 categories of parenting strategies coded molecularly, per behavior, as well as six global ratings (Winslow, 1995; Winslow, Shaw, Bruns, & Kiebler, 1995). Molecular and global ratings were made from videotaped mother-child interactions during a five minute structured clean-up task at the 24-month lab assessment. For the purposes of the present study, only molecular and global ratings relevant to hostile/rejecting parenting were employed. These included two molecular ratings--verbal/physical approval and critical statements--as well as three global ratings--hostility, warmth, and punitiveness. Verbal and physical approval included praising, verbal affirmations (e.g., good, yes), nodding the head affirmatively, smiling at child, and physical affection. Critical statements were verbal statements that prohibited the child from doing something or showed disapproval of his behavior or character. Hostility was defined as the mother's emotional expression of anger toward the child as indicated by the tone of voice and mannerisms. The warmth rating was an evaluation of the amount of positive affect expressed toward the child. Punitiveness was defined as the extent to which the mother was too strict, demanding, or harsh considering the child's behavior. Originally, global ratings were made on a 4-point scale; however, punitiveness and warmth were converted to 3-point scales (1 = none to a little, 2 = a little to some, 3 = some to a lot) to improve interrater agreement. Interrater agreement was calculated for each rating individually. For molecular ratings, Cohen's kappa coefficients were .87 for approval and .79 for critical statements. For global ratings, kappa coefficients were .94, .83, and .94 for hostility, warmth, and punitiveness, respectively. Five raters coded the molecular ratings, and three different raters coded the global ratings.
To create a single indicator corresponding to rejecting parenting, a principal components analysis was conducted with the five parenting variables. Only one factor emerged and loadings exceeded .40 in the expected directions for all five variables. The factor had an eigenvalue of 2.35 and accounted for 47% of the variance. To create the composite, each of the 5 variables was first standardized, and then the average of these variables was taken and used as the rejecting parenting factor. Alpha coefficient for this factor at 24 months was .70. This composite factor was found to predict later behavior problems in boys in an earlier study (Shaw, Winslow, Owens, Vondra, Cohn, & Bell, 1998).
Sibling Conflict Coding System (SCCS). Destructive sibling conflict was coded using the SCCS, an observational coding system for sibling interaction based on a live coding system developed by Volling and Belsky (1992). The system was designed to capture the amount and quality of sibling conflict during a one-hour videotaped sibling interaction. The procedure for the sibling interaction is described above in the Procedure section. Coders were instructed to rate each instance of conflict between the siblings in several areas: number of utterances from both siblings, amount of time (in seconds) spent in each conflict sequence, the intensity level of the affect, whether aggression was involved, and which sibling's action or utterance initiated the conflict sequence. The final "overall intensity of conflict" rating was made using a flow chart that combined number of utterances, intensity of affect (low, moderate, extreme), and aggression. There were five possible levels of intensity ranging from one unreciprocated utterance (level 1) to reciprocated physical aggression (level 5). Level 4 and 5 conflicts represent the most prolonged and aversive types of conflict.
Global ratings were added to the system to account for characteristics of the sibling interaction that might enhance or mitigate the effect of the conflict on child behavior. These 4-point scale codes included ratings for positive interaction, controlling behavior (examining one sibling's dominance over the other), constructive conflict (i.e., negotiation), and destructive conflict (the overall feel of the severity of the destructiveness of the sibling conflict sequences). Only the destructive conflict global rating was used for this study. The definition of destructive conflict encompasses the extremes of negative conflict sequences such as aggression, yelling in anger, swearing, and especially denigrating comments. The aim of this rating was to identify conflict that would be actually "destructive" to objects, people, and the sibling relationship. The 4-point scale included the following codes: "no destructive conflict" (1), "occasional destructiveness" (2), "some" (3), and "several instances/very destructive" (4). It was included to take into account siblings who had either many long conflict sequences but were not affectively charged (who would score low on the destructive conflict global), and those sibling pairs who had few very intense/angry and aggressive conflict sequences (who would receive high scores). The molecular codes were not complex enough take these qualitative distinctions into account, which necessitated the use of this global rating.
All coders attained and maintained an 80% minimum level of agreement for molecular ratings, as determined from 10% of tapes coded. Agreement between coders was reached for molecular codes when time of conflict, overall intensity rating, and conflict initiator were all in agreement. Consensus ratings were also used for questions on the molecular codes, in which all four coders watched the questionable piece of tape and decided together what the code should be. This was done for approximately 1/4 of all codes. Reliability was not achieved at the 80% level for global ratings; therefore, a consensus procedure was carried out in which the coding team had watched approximately 20 minutes of each tape. For each tape, the original coder was required to justify her global ratings, which were then verified by consensus. Both molecular and global ratings were coded by the same four coders. Maternal intervention was not included in any of the codes.
To strengthen the generalizability of the construct, the number of high intensity conflicts (levels 4 and 5), number of seconds spent in level 4 and 5 conflicts, and the global rating for destructive conflict were standardized and then summed to create a composite score of destructive sibling conflict for each sibling dyad, which was used in the analyses. The correlations among these variables ranged from .71-.88 (p < .001), and the alpha coefficient for the composite was .89. The composite takes into account how many times the siblings engaged in high intensity and aggressive conflict, how prolonged this conflict was, and the level of negative affect of the conflict throughout the play interaction.
Results Three hypotheses were tested in the present study. First, it was expected that destructive sibling conflict would be associated with conduct problems concurrently at age 5 and longitudinally at age 6. Second, it was expected that destructive sibling conflict would contribute unique variance to the prediction of conduct problems at age 5 and 6 after controlling for externalizing behavior problems and rejecting parenting at age 2. Third, it was expected that the interaction between destructive sibling conflict and rejecting parenting would account for unique variance in conduct problems after accounting for these factors' main effects.
Descriptive Statistics
Table 1 contains means and standard deviations of the hypothesized predictor and outcome variables for the sample used in the analyses. Compared to previous research using similar predictors, subjects' scores on certain measures deviated from those found in the general population. The mean SES on the Hollingshead Four Factor Index (Hollingshead, 1975) was 4, indicating that in most families, wage earners were of the occupational level of semi-skilled workers. At age 2, CBCL Externalizing t-scores matched those of the national average for non-referred children. At age 5 and 6 the narrow-band Aggression and Delinquency scores were greater than one standard deviation above national norms (Aggression normative mean t-score = 54.1, SD = 6.0; for Delinquency, normative mean t-score = 53.8, SD = 5.7).
Because there was a concern with specific variables that might influence relations between predictor variables and child conduct problems, these variables were examined prior to testing formal hypotheses. They included SES, sex of the sibling involved in the sibling conflict paradigm, and the age difference between the sibling and the target child. Only SES was significantly correlated with outcome variables and was added to the analyses as a covariate to all regression analyses. Sibling characteristics were not used as control variables in later analyses as they were not correlated with the outcome measures (see Table 2).
Correlational Analyses
Correlations among predictor and outcome variables are presented in Table 2. Early externalizing problems were related to maternal report of conduct problems at both age 5 and 6, but were not related to teacher report of conduct problems. Rejecting parenting was related to age 5 Aggression and Delinquency as reported by mother, and teacher report of Aggression. Destructive sibling conflict was associated with mother reported Aggression at age 5 and Delinquency at ages 5 and 6. Sibling conflict was not related to teacher reported conduct problems.
Intercorrelations among mother and teacher rated outcomes are also presented in Table 2. These correlations show that agreement between reporters was moderate for the Aggression factor, and low for the Delinquency factor.
Examining the Relation Between Sibling Conflict and Conduct Problems, Controlling for Early Predictors of Conduct Problems
The second set of analyses centered on the hypothesis that sibling conflict would continue to contribute unique variance to the prediction of conduct problems after controlling for SES, early externalizing problems, and rejecting parenting. To strengthen the constructs and to conserve space, age 5 and 6 outcomes for maternal report of conduct problems were combined by averaging the t-scores. Hierarchical multiple regression analyses were then performed using the combined age 5/6 scores for the Aggression and Delinquency factors. Externalizing problems (CBCL score) and rejecting parenting, both measured at age 2, were entered before sibling conflict in the regression equations. Socioeconomic status (SES)was entered first in all regression equations, in order to account for its influence.
Results for the multiple regression analyses using the age 5/6 CBCL scores as dependent variables are presented in Table 3. These results show destructive sibling conflict added unique variance to age 5/6 conduct problems, both Aggression and Delinquency, after the entry of age 2 Externalizing behavior and rejecting parenting.
Regression analyses predicting to age 6 teacher report of conduct problems (TRF) confirmed the results of the correlation analyses; destructive sibling conflict failed to add significant variance to TRF Aggression and Delinquency after controlling for the other predictors. These results are presented in Table 4.
Testing for Interactive Effects of the Predictor Variables
The third hypothesis entailed examining the interaction between destructive sibling conflict and rejecting parenting in the prediction of age 5/6 child conduct problems. Because interactive effects had not been previously tested in this area, all possible 2- and 3-way interactions were explored, but it was hypothesized that only the interaction between destructive sibling conflict and rejecting parenting would account for unique variance once main effects had been taken into account. Hierarchical regressions were computed to examine these interactive effects, the results of which are presented in Table 3. In all analyses, predictor variables previously used in regression analyses were entered before two- and three-way interaction terms. For the maternal report of age 5/6 CBCL Aggression, the hypothesized interaction between sibling conflict and rejecting parenting did account for unique variance, as did the interaction between early externalizing behavior and sibling conflict, which had not been predicted. These interactions call into question the validity of the main effect of sibling conflict on CBCL Aggression. The interaction between destructive sibling conflict and rejecting parenting was then examined through an analysis of simple slopes using the procedure recommended by Aiken and West (1991). Socio-economic status and the age 2 Externalizing score were entered in these analyses as covariates as they were in the original regressions. Figure 1 shows the regression lines for sibling conflict and rejecting parenting, both plotted at 1 standard deviation below and 1 standard deviation above the means for each variable. This figure illustrates that a rise in aggression scores is only experienced by those children who have both high levels of destructive conflict with their siblings and experience a rejecting parent, t = 2.74, p < .01. The size of the difference in CBCL Aggression scores between high and low sibling conflict at high levels of rejecting parenting was about 1 standard deviation. There were no significant interactions for age 5/6 CBCL Delinquency.
Despite the absence of direct relations between destructive sibling conflict and TRF (teacher-reported) Aggression and Delinquency, the hypothesized interaction between sibling conflict and rejecting parenting was significant in predicting TRF Aggression at age 6. This result is displayed in Table 4. This interaction was also examined with the analysis of simple slopes described above. Again, increased aggression scores are only experienced by children who have both high levels of destructive sibling conflict and a rejecting parent, t = 2.53, p < .05. The difference in TRF Aggression scores between high and low sibling conflict at high levels of rejecting parenting was about 2/3 of a standard deviation.
Discussion The present study yielded several interesting findings. First, destructive sibling conflict was directly related to maternal report of conduct problems, but not teacher-reported outcomes. Second, sibling conflict accounted for unique variance in the prediction of delinquent behavior once SES, early child behavior, and parenting were accounted for. Again this was found only for maternal report of delinquent behavior. Third, and most importantly, the interaction between destructive sibling conflict and rejecting parenting accounted for unique variance in the prediction of aggressive behavior as reported by both mothers and teachers, after accounting for these predictors main effects and SES.
These findings suggest that the pathways towards the development of aggressive and delinquent behavior are somewhat different. The results were strongest for the Aggression factor; children who experienced high levels of aversive sibling conflict and high levels of parental rejection were rated as having higher levels of aggressive behavior by both mothers and teachers than children who only experienced one of these predictors (see Figure 1). It is possible that these results were only consistent for the Aggression factor because of the types of behaviors measured in that scale. It consists of items involving reactive dysregulation, including temper tantrums, destruction of objects, fighting, and screaming. These aversive behaviors are also likely to elicit parents' use of aversive control strategies, as described in Patterson's coercion theory of dyadic family interaction. It is also probable that teachers would observe these kinds of self-regulatory problems in classroom situations. This finding challenges the validity of direct relations between sibling conflict and aggressive behavior problems, as it is only in conjunction with rejecting parenting that sibling conflict appears to influence child aggressive behavior.
This result is also consistent with an additive risk model, suggesting that children who experience conflictual relationships with both adults and children in the home are more likely to demonstrate conduct problems at home and in school. The effect sizes for these results were consistent across informant, and this interaction was the only significant result to emerge from regressions using the teacher report data. This leads one to believe that this interactive effect is valid and is a powerful indicator as to siblings' role in the development of aggressive behavior problems. Sibling conflict appears to be not solely a marker for aggressive problem behavior, but in conjunction with rejecting parenting, may be implicated in the development of this behavior.
Alternatively, these results do not rule out the possibility that the child is the driving force behind the interactions both with his sibling and with a rejecting parent. For instance, a disruptive child might elicit more rejecting parenting and be involved in higher rates of destructive sibling conflict. In addition, the fact that the interaction between rejecting parenting and sibling conflict contributed unique variance after accounting for early externalizing problems is consistent with models in which both members of the parent-child and sibling dyads contribute to coercive interactions.
Prediction of the Delinquency factor was not as clear. Although there was a main effect for destructive sibling conflict predicting to maternal report of delinquent behavior after controlling for other predictors, this relation was not found for teacher report. The nature of the Delinquency factor items may be informative as to both why there was no effect for teacher report, and why there was a main effect instead of an interactive effect. The Delinquency factor items include more proactive, discrete acts such as stealing, lying, and swearing which are more likely to be modeled by an antisocial sibling (a main effects model), and which teachers may not be witnessing as much as aggressive and dysregulated behavior. The relations with delinquent behavior also support Patterson's sibling training hypothesis for children with close-age siblings. Our results are consistent with the notion that aversive and aggressive exchanges with a close-age sibling may promote more serious types of problem behaviors in the target child, as spending time with a deviant peer group has been shown to do. Quality of conflict is important because it is probable that a certain amount of conflict between siblings is helpful in teaching children conflict-resolution skills and other socially important strategies and behaviors (Brody, 1998; Dunn et al., 1991). For example, Dunn and Munn (1986b) showed that certain types of conflict with an older sibling were associated with more mature behaviors in the younger sibling.
To test the strength of the relation between sibling conflict and conduct problems over time, in additional analyses, we tested whether sibling conflict would still predict maternal report of age 6 Delinquency after partialing out the variance attributed to age 5 Delinquency. Sibling conflict continued to add unique variance to age 6 Delinquency scores after controlling for age 5 scores (r2 change = .036, p = .01). It is possible that destructive sibling conflict makes an independent contribution to the development of delinquent behavior as children's relationships with siblings play an increasingly important role in their lives. It is also possible that this study was more able to accurately assess the development of aggressive behavior, which develops at an earlier age, than delinquent behavior, which continues to develop into adolescence. If the effect of sibling interaction on delinquent behavior is increasing with time, then it may be that we will be better to assess its effects at a later date.
Two null findings are worth note. First the rejecting parenting factor was directly related to the outcomes in the correlational analysis, but not additively in the regressions once early externalizing behavior had been accounted for. This was unexpected, but understandable given the time lag between measurement periods (3 to 4 years), and the length and context of the assessment of parenting (i.e., 5 minutes during a laboratory based clean-up task). However, the same parenting measure has been shown to be predictive of age 3½ conduct problems with the same sample in an earlier study (Shaw, et.al., 1998). Second, the presence of only one significant finding with teacher outcomes is worth consideration. These results are in line with previous work showing that parents and teachers disagree more often than not when rating problem behaviors in children (Offord et al., 1996). We can also see from the differences in the mean t-scores from maternal to teacher reports that the teachers in general rated the children as having fewer problem behaviors (Table 1). These results highlight the importance of analyzing data from multiple informants separately to account for the various contexts of the child's environment.
This study was limited by the lack of earlier data about the sibling involved in the interaction. Because data were not available on siblings' conduct problems, questions could not be answered about which child was driving the interaction. Patterson's sibling training hypothesis would postulate that it is the older sibling who is shaping the younger child's behavior, and other researchers also assume an older-to-younger sibling influence (Patterson, 1984; Dunn, et al., 1995). However, it is also possible that close-age siblings have mutual influence on each other.
Another limitation of the study was that only the target child and his closest-age sibling were observed. This means that for families with multiple children, the interactions with other siblings were not taken into account. It is possible that the sibling used in the play interaction was not the sibling the target child interacts with most. Having data only on one sibling, albeit the one closest in age, may have provided an inaccurate representation of the quality of the target child's interactions with other siblings in the family.
A third limitation was the measure of sibling conflict. The children were videotaped playing with their mother and two research assistants in the room with them. Although they were in the natural surroundings of their own homes, adult supervision most likely had an effect on their behavior. Maternal intervention into sibling conflict was not coded, which limits the conclusions to be drawn from this measure as it undoubtedly had some effect. It is also probable that the children were less argumentative and aggressive with each other than they would have been if they had been playing together alone. Accordingly, it is also possible that the reduced variability in sibling conflict may have attenuated the size of the relations between destructive sibling conflict and conduct problems.
Finally, the target children in this study were all low-income boys. This limits the generalizability of these findings to brother-brother and brother-sister pairs of siblings and to the predictive validity of sibling conflict for conduct problems in boys. Further studies should address these relations for girls and for sister-sister sibling pairs. These children were also at higher risk for conduct problems due to their low-income status as evidenced by mean scores on the CBCL and TRF which are above national norms. This issue may also limit generalizability.
Despite these caveats, the results offer support to the idea that sibling relations play an important role in children's social development, specifically in the development of early school-age conduct problems. They show that just as there are associations among sibling relationships and prosocial behavior and social/emotional understanding, there are also similar connections between destructive sibling conflict and conduct problems. The results also suggest areas for future research, for instance, linkages between sibling and peer relationships. Since the interaction between rejecting parenting and sibling conflict has been shown to be related to teacher report of conduct problems, it is possible that sibling conflict is related to peer interaction in school. If children are learning aggressive conflict resolution strategies through interactions with their siblings, then they also may be adopting these strategies with peers. Further areas for research include taking into account child genetic and biological vulnerabilities in this model. Another is to include a whole family approach to this question in which the effects of marital conflict are taken into account both in the development of sibling conflict and conduct problems.
These results may have implications for the assessment and treatment of conduct problems. The effect sizes found for the relations between destructive sibling conflict and conduct problems are comparable to those seen for relations between marital conflict and behavior problems (Fincham, 1994). Just as knowledge of the parental relationship is important in determining family functioning for the assessment and treatment of conduct problems, sibling conflict may also play a role in family functioning. Knowledge of the child's behavior in the context of the sibling relationship may be an important indicator from which to ascertain the pervasiveness of coercive relationships in the family. The results also indicate that sibling interaction is especially important to consider in conjunction with early rejecting parenting, because the interaction was associated with an increased risk of conduct problems over time and across settings. The results argue for a family approach to the assessment and possibly treatment of conduct problems in boys.
References
Achenbach, T. M. (1991). Manual for the Teacher's Report Form and 1991 profile. Burlington, VT: University of Vermont, Department of Psychiatry.
Achenbach, T. M. (1992). Manual for the Child Behavior Checklist and Revised Child Behavior profile. Burlington, VT: University of Vermont, Department of Psychiatry.
Achenbach, T. M. (1994). Child behavior checklist and related instruments. In M. E. Maruish (Ed.), The use of psychological testing for treatments planning and outcome assessment (pp.517-549). Hillsdale, NJ: Lawrence Erlbaum Associates.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage Publications.
Brody, G. H. (1998). Sibling relationship quality: Its causes and consequences. Annual Review of Psychology, 49, 1-24.
Brody, G. H., Stoneman, Z., & Burke, M. (1987). Child temperaments, maternal differential behavior, and sibling relationships. Developmental Psychology, 23, 354-362.
Bryant, B. K., & Crockenberg, S. B. (1980). Correlates and dimensions of prosocial behavior: A study of female siblings with their mothers. Child Development, 51, 529-544.
Campbell, S. B. (1990). Behavior problems in preschool children: Clinical and developmental issues. New York: Guilford Press.
Dunn, J. (1983). Sibling relationships in early childhood. Child Development, 54, 787-811.
Dunn, J., Brown, J. R., & Maguire, M. (1995). The development of children's moral sensibility: Individual differences and emotion understanding. Developmental Psychology, 31, 649-659.
Dunn, J., Brown, J., Slomkowski, C., Telsa, C., & Youngblade, L. M. (1991). Young children's understanding of other people's feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 1352-1366.
Dunn, J. & Munn, P. (1986a). Siblings and the development of prosocial behaviour. International Journal of Behavioral Development, 9, 265-284.
Dunn, J. & Munn, P. (1986b). Sibling quarrels and maternal intervention: Individual differences in understanding and aggression. Journal of Child Psychology and Psychiatry, 27, 583-595.
Fincham, F. D. (1994). Understanding the association between marital conflict and child adjustment: Overview. Journal of Family Psychology, 8, 123-127.
Hetherington, E. M., Reiss, D., & Plomin, R. (Eds.). (1994). Separate social worlds of siblings: Impact of the nonshared environment on development. Hillsdale, NJ: Erlbaum.
Hollingshead, A. B. (1975). Four Factor Index of Social Status. Unpublished manuscript, Yale University, Department of Sociology, New Haven, CT.
Kazdin, A. E. (1995). Conduct Disorders in Childhood and Adolescence (2nd ed.), Thousand Oaks, CA: Sage Publications.
Keenan, K., & Shaw, D. (1997). Developmental and social influences on young girls' early problem behavior. Psychological Bulletin, 121, 95-113.
Lamb, M. E. (1978). The development of sibling relationships in infancy: A short-term longitudinal study. Child Development, 49, 1189-1196.
Offord, D. R., Boyle, M. H., Racine, Y., Szatmari, P., Fleming, J. E., Sanford, M., and Lipman, E. L. (1996). Integrating assessment data from multiple informants. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 1078-1085.
Olweus, D. (1979). Stability of aggressive reaction patterns in males: A review. Psychological Bulletin, 86, 852-875.
Patterson, G. R. (1984). Siblings: Fellow travelers in coercive family processes. In R.J. Blanchard (Ed.), Advances in the study of aggression (pp. 174-213). New York: Academic Press.
Patterson, G. R. (1986). The contribution of siblings to training for fighting: A microsocial analysis. In D. Olweus, J. Block, & M. Radke-Yarrow (Eds.), Development of antisocial and prosocial behavior: Research theories and issues (pp. 235-261). New York: Academic Press.
Patterson, G. R., Dishion, T. J., & Bank, L. (1984). Family interaction: A process model of deviancy training. Aggressive Behavior, 10, 253-267.
Pepler, D. J., Abramovitch, R., Corter, C. (1981). Sibling interaction in the home: A longitudinal study. Child Development, 52, 1344-1347.
Pike, A., McGuire, S., Hetherington, E. M., Reiss, D., & Plomin, R. (1996). Family environment and adolescent depressive symptoms and antisocial behavior: A multivariate genetic analysis. Developmental Psychology, 32, 590-603.
Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from one another? Behavioral and Brain Sciences, 10, 1-16.
Renken, B., Egeland, B., Marvinney, D., Mangelsdorf, S., & Sroufe, A. (1989). Early childhood antecedents of aggression and passive-withdrawal in early elementary school. Journal of Personality, 57, 257-281.
Richman, M., Stevenson, J., & Graham, P. J. (1982). Preschool to school: A behavioral study. London: Academic Press.
Shaw, D. S., Winslow, E. B., Owens, E. B., & Hood, N. (1998). Young children's adjustment to chronic family adversity: A longitudinal study of low-income families. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 545-553.
Shaw, D. S., Winslow, E. B., Owens, E. B., Vondra, J. I., Cohn, J. F., & Bell, R. Q. (1998). The development of early externalizing problems among children from low-income families: A transformational perspective. Journal of Abnormal Child Psychology, 26, 95-107.
Stocker, C., Dunn, J., & Plomin, R. (1989). Sibling relationships: Links with child temperament, maternal behavior, and family structure. Child Development, 60, 715-727.
Volling, B. L., & Belsky, J. (1992). The contribution of mother-child and father-child relationships to the quality of sibling interaction: A longitudinal study. Child Development, 63, 1209-1222.
Winslow, E. (1995). Relations among contextual factors and early parenting styles. Unpublished master's thesis. University of Pittsburgh.
Winslow, E. B., Shaw, D. S., Bruns, H., & Kiebler, K. (1995). Parenting as a mediator of child behavior problems and maternal stress, support, and adjustment. Paper presented at the biennial meeting of the Society for the Research in Child Development, Indianapolis, IN.
Youngblade, L. M., & Dunn, J. (1995). Individual differences in young children's pretend play with mother and sibling: Links to relationships and understanding of other people's feelings and beliefs. Child Development, 66, 1472-1492.
Zahn-Waxler, C., Iannotti, R. J., Cummings, E. M., & Denham, S. (1990). Antecedents of problem behaviors in children of depressed mothers. Development and Psychopathology, 2, 271-291.