Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2013 Aug 1.
Published in final edited form as: J Fam Psychol. 2012 Jul 9;26(4):523–531. doi: 10.1037/a0029083

Sibling Differences in Parent-Child Conflict and Risky Behavior: A 3-Wave Longitudinal Study

Chun Bun Lam 1, Anna R Solmeyer 1, Susan M McHale 1
PMCID: PMC3552291  NIHMSID: NIHMS422543  PMID: 22775198

Abstract

To better understand why siblings growing up in the same family are often as different as unrelated individuals, this study explored the role of differential experiences with parents in the development of sibling differences. Cross-lagged models tested directions of effect by examining whether differential parent-child conflict predicted sibling differences in risky behavior over time, or vice versa. Participants were mothers, fathers, and the two eldest adolescent siblings (mean ages at Time 1 = 15.12 and 12.58 years) from 355 European American, work- and middle-class families. On three occasions over a 2-year period, mothers and fathers reported on their conflict with each of the two siblings, and siblings reported on their own risky behavior. Results revealed that, controlling for sibling age differences and average levels of conflict and risky behavior at Time 1, youths who had more conflict with their mothers and fathers relative to their siblings subsequently engaged in relatively more risky behavior. Also, youths who engaged in more risky behavior relative to their siblings experienced relatively more conflict with their fathers, but not mothers, at later time points. Findings highlight the importance of examining both family dynamics and child characteristics in understanding sibling differentiation, and illuminate potential differences in parenting processes involving mothers versus fathers.

Keywords: adolescence, parent-child conflict, risky behavior, sibling differentiation, within-family influences

Sibling Differences in Parent-Child Conflict and Risky Behavior: A 3-Wave Longitudinal Study

Siblings raised in the same family are often as different from one another as they are from unrelated individuals (Turkheimer & Waldron, 2000). Questions of interest to family researchers and practitioners are: How do sibling differences unfold over time? And, what family dynamics are implicated in the development of sibling differences? Although genetic factors are part of the story, behavioral geneticists argue that the greatest environmental influence on behavioral development is the nonshared environment (Plomin & Daniels, 1987). An important distinction lies in the objective versus the effective environment: Objective environments refer to features that are observable from outside the family (such as by a researcher) and may be the same or different for siblings, regardless of whether they effectively contribute to sibling similarities or differences (Feinberg & Hetherington, 2001; Turkheimer & Waldron, 2000). In contrast, effective environments are defined by how they function to make siblings alike or different, regardless of whether they are objectively shared or nonshared by siblings. The focus of the current study is on sibling differences in parent-child conflict (a nonshared objective, and possibly effective, environmental feature), and their bidirectional links with sibling differences in risky behavior during adolescence. Specifically, we explored whether differences in parent-child conflict drive sibling differentiation (Festinger, 1954; Mullineaux, Deater-Deckard, Petrill, & Thompson, 2009), and whether effects of the reverse direction are operative, that is, whether sibling differences in behavior evoke differential experiences with parents (Conger & Conger, 1994; Scarr & McCartney, 1983).

Differential Parent-Child Conflict and Sibling Differences

Despite individualistic cultural norms that call for parents to treat their offspring equally, both parents and youths report that differential treatment is common (McHale & Crouter, 2003). Social comparison theory (Festinger, 1954) suggests that siblings use one another as sources for social comparison, and how parents treat oneself versus a sibling represents a ripe target for comparison. Feeling inferior or disfavored relative to a sibling can have negative implications for adjustment, an idea first described by Alfred Adler in his theory of individual psychology (Ansbacher & Ansbacher, 1956). Most existing research focuses on the implications of differential parent-child relationships for each child’s individual adjustment, and findings suggest that less favored siblings have higher levels of externalizing behavior (Feinberg & Hetherington, 2001; Tamrouti-Makkink, Dubas, Gerris, & van Aken, 2004). Fewer studies have conceptualized differential parent-child relationships as a force that drives sibling differentiation. This distinction is nuanced, but it is not trivial; the former analysis shows that disfavored children have relatively more adjustment problems compared to other youths, whereas the latter focuses on whether the disfavored child has more adjustment problems compared to her or his sibling. The current study takes the second approach, which is geared toward exploring within-family processes and how they work to create differences between siblings. This is an important direction for research and practice with families, which generally have assumed that information on the experiences of one child in a family is sufficient for understanding how families operate as social and socializing systems.

Behavioral geneticists have studied sibling differentiation using designs that account for the level of genetic similarity between siblings. A common approach is the monozygotic (MZ) twin design; because MZ twins share 100% of their genes, researchers can directly examine nonshared effective environmental processes by holding genetic variation constant. Findings from MZ twin studies provide some of the strongest evidence about the role of parent-child relationships in sibling differentiation. In general, the child who receives relatively more maternal negativity or less maternal positivity exhibits more problem behavior compared to her or his twin sibling, both concurrently (Asbury, Dunn, Pike, & Plomin, 2003) and longitudinally (Asbury, Dunn, & Plomin, 2006; Burt, McGue, Iacono, & Krueger, 2006; Mullineaux et al., 2009). Most MZ twin studies, however, focus on young children, with the exception ofBurt et al. (2006), who explored these associations in adolescence.

Although the MZ twin difference design has many strengths, a drawback is its lack of generalizability to the large majority of siblings. For example, MZ twins are the same age and the same gender. Research on siblings, however, suggests that the age and gender differences that characterize most sibling dyads may play a role in parents’ differential treatment and in sibling differentiation (McHale & Crouter, 2003; Tamrouti-Makkink et al., 2004). Another limitation of the available MZ twin studies is an almost exclusive focus on mothers, although a few studies on nontwin siblings have found that differential father-child relationships are also important for youth adjustment (Brody, Stoneman, & McCoy, 1992; Conger & Conger, 1994; Feinberg & Hetherington, 2001). We sought to replicate established MZ findings in a sample of nontwin siblings, and build on this research by studying child-specific experiences with both mothers and fathers.

Children from the same family can have quite different experiences with their parents in a wide variety of domains, such as privileges, chores, and discipline (Tucker, McHale, & Crouter, 2003). We focused on differential parent-child conflict, because it is a salient dynamic in adolescence as parents and youths navigate new issues around autonomy that can lead to disagreements (Laursen & Collins, 2009; Smetana, 1989). Using a genetically informed design, Anderson, Hetherington, Reiss, and Howe (1994) also demonstrated that parent-child conflict was linked to adolescent adjustment at the child-specific level, with little evidence to suggest that shared variance in parent-child conflict across siblings was associated with individual adjustment, making it a good candidate for addressing questions about differential environments. Further, differential parent-child conflict is one of the more commonly studied domains in the literature on sibling differentiation (e.g., Anderson et al., 1994; Asbury et al., 2003; Burt et al.,2006), and this focus allowed us to build on prior research.

Directions of Effect

Most findings on the links between differential parent-child relationships and youth adjustment have been interpreted as driven by family dynamics: Different experiences with parents cause siblings to become more different from one another over time. Few studies have tested the equally plausible hypothesis, that sibling differences drive differential parental treatment and parent-child relationships. A genotype → environment perspective (Scarr & McCartney, 1983) suggests that youths are not passive recipients of environmental forces; rather, they also affect their environments, such as when their personal characteristics elicit reactions from others. In the sibling context, these child-driven processes may also contribute to differential environments (Turkheimer & Waldron, 2000), as siblings evoke different responses from their parents when they behave in different ways. Indeed, parents perceive differences between their offspring in a variety of domains and report that these differences are a primary impetus for their differential treatment (McHale & Crouter, 2003). Thus, both differential parent-child relationships and child characteristics may play a role in sibling differentiation.

In this study, we focused on sibling differences in risky behaviors, such as truancy, delinquency, and substance use, because these types of behavior increase in adolescence (Farrington, 2009; Jessor, 1992) and theoretically may act as causes and/or consequences of differential parent-child conflict. For example, parents may react to adolescents’ risky behavior by becoming controlling or authoritarian, leading to an increase in conflict (Hipwell et al., 2008). This tendency may be magnified in the sibling context (Caspi, 2011): If parents have multiple children, they are able to readily compare the siblings’ behavior and may respond to the sibling who is relatively more delinquent by behaving more negatively toward that sibling. In turn, both siblings may behave in ways that conform to their distinct niches, with one sibling escalating her or his risky behavior and the other increasing in conformity to parents’ mandates, further perpetuating sibling differences. In this way, sibling differentiation may be both parent- and child-driven.

Although a body of research on how parent-child relationships and child characteristics jointly affect youth development exists outside the sibling context (Laursen & Collins, 2009), the issue of direction of effect in sibling differentiation has been addressed by only a handful of studies. In a study of MZ twin differences,Burt et al. (2006) showed that differential maternal conflict predicted later differences in externalizing behavior among adolescent twins who were highly discordant in risky behavior, but they found no evidence of the reverse direction of effect. A similar effect for mothers’ and fathers’ differential hostility on nontwin siblings’ delinquency was found by Conger and Conger (1994), who also reported a marginally significant, child-driven effect for mothers only: The sibling who reported relatively higher delinquency at Time 1 experienced relatively less maternal hostility 2 years later. Clearly, more research is needed on this topic, especially on the question of how fathers may react to sibling differences. Exploring these dynamics in the context of sibling differentiation can provide insights into the role of both mothers and fathers in the development of adolescent delinquency, an important issue for both parents and policy makers (Farrington, 2009).

In this study, we used data on both mother- and father-child conflict, collected on three occasions over a 2-year period, to fit a series of cross-lagged models and examine the associations between sibling differences in parent-child conflict and sibling differences in risky behavior. Based on prior theory and research, we expected that differential parent-child conflict would predict subsequent sibling differences in risky behavior, but we also expected sibling differences to predict differential parent-child conflict. Given the dearth of research on fathers’ roles in sibling differentiation, we proposed no specific hypotheses about how these processes would differ for mothers versus fathers.

Method

Participants

Participants were mothers, fathers, and the two eldest siblings in 355 families from two cohorts of a longitudinal study of family relationships. The sample included almost exclusively European American working- and middle-class families living in small cities, towns, and rural communities in a northeastern state. Recruitment letters were sent from schools to all families with youths in the target age group. Families were eligible if parents were married and employed and the two eldest siblings were not more than 4 years apart in age. Over 90% of families that returned postcards were eligible and eventually participated. We used three occasions of measurement (referred to as Times 1 through 3 hereafter) from each cohort. Attrition was low, and less than 5% of the data were missing. At Time 1, firstborns were in mid-adolescence (M = 15.12 years; SD = 0.74) and second-borns were in early adolescence (M = 12.58 years; SD = 1.06). Sibling dyads were divided evenly among the four possible gender constellations. At Time 1, the average family income was $67,592 (SD = 38,562). The sample came close to capturing the racial and economic backgrounds of two-parent families from the region where the study was conducted (US Census Bureau, 2000). Five fathers did not participate, and thus the models for mothers and fathers were based on 355 and 350 sibling dyads, respectively.

Procedures and Measures

Trained interviewers conducted annual home interviews with mothers, fathers, and the two siblings. The family gave informed consent and received an honorarium at the beginning of the interview, and family members completed standardized questionnaires individually. Parent-child conflict was measured using a 10-item index adapted from Smetana (1988). At each occasion and at separate points in their interviews, mothers and fathers used a 6-point scale to rate “how often they had had conflict with” their firstborn and second-born children in 10 different domains (e.g., homework, social life, money) in the past year. Siblings’ risky behavior was measured using an 18-item index adapted from Eccles and Barber (1990). At each time point, firstborns and second-borns used a 4-point scale to rate “how often they had engaged in” 18 different risky activities (e.g., drinking, smoking, skipping school) in the past year. For both measures, items were summed, such that higher scores indicated higher levels of the construct. Table 1 shows the Cronbach’s alphas and concurrent correlations between firstborns’ and second-borns’ scores.

Table 1.

Descriptive Statistics for Firstborns’, Second-borns’, and Sibling Differences in Parent-Child Conflict and Risky Behavior, and Correlations between Siblings’ Scores (N = 336–355 sibling dyads)

Firstborns
 Second-borns
 Differencea
 Correlationb
Variables M SD α M SD α M SD r
Mother-child conflict T1 24.65 6.80 .82 25.03 7.18 .84 −0.38 6.94 .51
Mother-child conflict T2 24.04 6.67 .82 25.24 7.14 .83 −1.21 7.68 .39
Mother-child conflict T3 22.09 5.98 .81 23.52 6.66 .82 −1.41 6.97 .40
Father-child conflict T1 23.63 6.57 .84 24.63 6.45 .81 −1.00 6.85 .45
Father-child conflict T2 23.43 6.79 .84 24.32 6.37 .82 −0.90 6.81 .46
Father-child conflict T3 21.21 5.92 .84 22.88 5.56 .80 −1.69 6.12 .43
Risky behavior T1 25.24 6.41 .85 22.65 5.39 .86 2.59 7.06 .29
Risky behavior T2 26.73 7.91 .88 24.04 7.12 .90 2.68 9.22 .25
Risky behavior T3 27.78 8.22 .88 25.11 7.24 .88 2.71 9.20 .30
Open in a new tab
a

Computed by subtracting second-borns’ from firstborns’ scores.

b

All correlations were significant at p < .01.

To create scores that represented sibling differences in parent-child conflict and risky behavior, we subtracted second-borns’ from firstborns’ scores. Because our analysis compared two siblings from each family, the difference scores created by subtracting second-borns’ from firstborns’ scores were perfectly and negatively correlated with the difference scores created by subtracting firstborns’ from second-borns’ scores. Thus, the magnitude and interpretation of the associations between the difference scores would be the same, regardless of which sibling’s score was subtracted from the other’s. A positive correlation between differential parent-child conflict and sibling differences in risky behavior, for example, would indicate that whichever sibling experienced relatively more conflict also reported more risky behavior compared to her or his sibling.

Given prior research showing that parent-child conflict (Smetana, 1988) and risky behavior (Farrington, 2009) change across adolescence, we controlled for sibling differences in age. Moreover, to account for the overall levels of parent-child conflict and risky behavior and isolate the contribution of sibling differences (Kenny, Kashy, & Cook, 2006), we controlled for the average levels of parent-child conflict and risky behavior at Time 1.

Results

Analytic Plan

We tested a series of path analytic models using LISREL 8.80 (see Figure 1 for a conceptual model). A preliminary multigroup analysis comparing same- and mixed-sex sibling dyads revealed no differences in the path coefficients, and thus all models were tested on the full sample. The Full Information Maximum Likelihood (FIML) method in LISREL was used to account for missing data. To assess model fit, we used the chi-square statistic (X2), RMSEA, and CFI. Because the presence of missing data greatly reduces the ability of fit indices to detect mis-specified covariance matrices (Davey, Savla, & Lou, 2005), we used relative rather than absolute values of fit indices to choose the best among several competing models (Wu, West, & Taylor, 2009). By comparing the X2 of the nested models, we could identify the model that best reproduced the observed covariance matrices for mothers and for fathers via a series of significance tests. Because differences in RMSEA and CFI values do not follow a probability distribution, they were used as supplementary criteria for model selection. It is worth noting that there often are a large number of alternative models, whether theoretically plausible or not, that could fit the observed covariance matrices equally well or better (Tomarken & Waller, 2003). Thus, the chosen model only represented the best among the set of models that were of theoretical interest in our study.

Figure 1.

Figure 1

Open in a new tab

A conceptual model of differential parent-child conflict and sibling differences in risky behavior. Stability paths are indicated by c1c2, c2c3, b1b2, and b2b3. Time-lagged paths are indicated by c1b2, c2b3, b1c2, and b2c3. Double-headed arrows signify concurrent correlations, r1, re2, and re3. Sibling age differences and average levels of conflict and risky behavior at Time 1 were included as controls, but these paths are not shown for clarity.

Based on prior statistical (Rovine & Liu, 2012) and empirical (Burt et al., 2006; Conger & Conger, 1994) research related to cross-lagged analyses, we tested four competing models separately for mothers and for fathers. First, we tested a stability model (Model 1), which served as a baseline for comparison and only estimated the temporal stability of differential parent-child conflict (i.e., c1c2, c2c3) and sibling differences in risky behavior (i.e., b1b2, b2b3) and their concurrent correlations (i.e., r1, re2, re3). Second, we tested two lagged regression models that examined whether sibling differences in parent-child conflict predicted later differences in risky behavior (Model 2, the conflict-driven model, included c1b2 and c2b3), and whether differences in siblings’ risky behavior predicted subsequent differences in parent-child conflict (Model 3, the behavior-driven model, included b1c2 and b2c3). Including the stability paths controlled for variance that was stable over time, and thus the time-lagged paths explained the residualized gains in the dependent variables (i.e., changes in sibling dyads’ rank order in the score distribution; Rovine & Liu, 2012). A significant X2 difference test comparing Model 1 and Model 2 or 3 would indicate that the time-lagged paths explained unique variance beyond the stability paths. Third, we tested a cross-lagged model that simultaneously estimated the four time-lagged paths included in Models 2 and 3 to examine whether each direction of effect uniquely predicted sibling differences (Model 4, the reciprocal model). A significant X2 difference between Model 4 and Model 2 or 3 would indicate that a bidirectional effect model fit the data better than either of the unidirectional effect models.

From Models 1–4, we selected the best-fitting model for mothers and for fathers. To further test whether the models could be made more parsimonious, we constrained the parallel stability paths (i.e., c1c2 = c2c3; b1b2 = b2b3) and time-lagged paths (i.e., c1b2 = c2b3; b1c2 = b2c3) between Times 1 and 2 and Times 2 and 3 to be equal. Constraints that did not result in a significant decrease in the X2 were retained in the final model (Model 5).

The last step of the analysis was exploratory and involved testing potential mother-father differences in the associations between differential parent-child conflict and sibling differences in risky behavior. We were not able to conduct a multigroup analysis here because, although the scores for differential parent-child conflict were different for mothers and fathers, the scores for sibling differences in risky behavior were identical for both parents; if a multigroup analysis with between-group constraints was conducted, the data for mothers and fathers would be analyzed simultaneously and the sibling difference scores would be counted twice in the model, a violation of the assumption of mutual exclusivity of groups (Jöreskog & Sörbom, 1999). Instead, after arriving at the final model for each parent, we assigned the values of the coefficients of interest of the mother model (Model 5a) to those of the father model (Model 5b). A significant X2 difference would indicate that the path coefficients for mothers and fathers were significantly different from each other.

Path Analytic Models

Table 1 shows the means and standard deviations for parent-child conflict and siblings’ risky behavior. The sign of the difference scores indicated that, on average, parents reported more conflict with second-borns than with firstborns and firstborns reported more risky behavior than second-borns. However, because our models were tested against the covariance matrix, the mean structure was of little importance to the interpretation of the results (Rovine & Liu, 2012). Table 2 shows the correlations between all variables included in the models.

Table 2.

Correlations between Study Variables (N = 332–355 sibling dyads)

Variables 1 2 3 4 5 6 7 8 9
1. Differential conflict T1 - .66** .57** .36** .34** .30** .08 .02 .10
2. Differential conflict T2 .71** - .70** .30** .31** .34** -.05 .07 .11
3. Differential conflict T3 .62** .70** - .24** .29** .33** -.01 -.03 .04
4. Differences in behavior T1 .31** .25** .18** - .68* .53** .19** .10 .23**
5. Differences in behavior T2 .26** .31** .25** .62** - .68** .20** .05 .11
6. Differences in behavior T3 .26** .35** .33** .47** .68** - .22** .07 .07
7. Age differences .00 -.01 -.03 .15** .21** .24** - -.00 -.17**
8. Average conflict T1 -.06 -.06 -.05 .07 .02 .05 .02 - .24**
9. Average risky behavior T1 .06 .10 .06 .21** .13* .09 -.15** .22** -
Open in a new tab

Note. Correlations for mothers are below the diagonal and correlations for fathers are above the diagonal.

*

p < .05.

**

p < .01.

The results for the mother models are shown in Table 3. The X2 difference between the stability (Model 1a) and the conflict-driven (Model 2a) models was significant, indicating that differences in mother-child conflict predicted sibling differences in risky behavior beyond the stability coefficients. The X2 difference between the stability and behavior-driven (Model 3a) models was not significant, suggesting that sibling differences in risky behavior did not predict subsequent differences in mother-child conflict. The reciprocal model (Model 4a) fit the data better than the behavior-driven model, but not the conflict-driven model, confirming that the behavior-driven effects were negligible. The conflict-driven model also had the lowest RMSEA and the highest CFI, and for these reasons it was chosen as the best-fitting model for mothers.

Table 3.

Standardized Path Coefficients and Fit Statistics for Mother Models (N = 355 sibling dyads)

Model 1a Model 2a Model 3a Model 4a Model 5a
Stability Conflict Behavior Reciprocal Final
Path coefficients
Conflict T1 → Conflict T2 0.76** 0.77** 0.74** 0.76** 0.78**
Conflict T2 → Conflict T3 0.62** 0.64** 0.60** 0.62** 0.63**
Behavior T1 → Behavior T2 0.77** 0.74** 0.78** 0.75** 0.72**
Behavior T2 → Behavior T3 0.65** 0.60** 0.65** 0.60** 0.61**
Conflict T1 → Behavior T2 - 0.10† - 0.10† 0.16**
Conflict T2 → Behavior T3 - 0.19** - 0.19** 0.16**
Behavior T1 → Conflict T2 - - 0.05 0.04 -
Behavior T2 → Conflict T3 - - 0.04 0.03 -
Fit statistics
X2 (df) 46.41 (8) 27.19 (6) 43.90 (6) 25.31 (4) 28.63 (7)
Δ X2 compared to Model 1a - 19.22** 2.51 21.10** -
Δ X2 compared to Model 2a - - - 1.88 1.44
Δ X2 compared to Model 3a - - - 18.59** -
RMSEA .12 .10 .13 .12 .09
CFI .96 .98 .96 .98 .98
Open in a new tab
†

p < .10.

*

p < .05.

**

p < .01.

Comparisons testing for equivalence in the parallel paths indicated that the time-lagged paths from mother-child conflict to sibling differences in risky behavior between Times 1 and 2 and Times 2 and 3 were not significantly different from each other, and thus they were constrained to be equal in the final model (Model 5a). Focusing on the path coefficients, differential mother-child conflict was significantly and positively linked to sibling differences in risky behavior, meaning that youths who had more conflict with their mothers relative to their siblings subsequently engaged in more risky behavior relative to their siblings. The effect sizes were small (Kline, 2011), however, especially when compared to the large effect sizes of the stability paths. There were significant differences in the parallel stability paths for differential mother-child conflict, X2(1) = 6.09, p < .01, and sibling differences of risky behavior, X2(1) = 3.89, p < .05, indicating that the stability of these variables decreased over time.

Results for the father models are shown in Table 4. The X2 differences between the stability (Model 1b) and the conflict-driven (Model 2b) models and between the stability and the behavior-driven (Model 3b) models were significant, providing evidence of both directions of effect. Further, the reciprocal model (Model 4b) was a significant improvement over the conflict-driven and the behavior-driven models, suggesting that both the conflict-driven and behavior-driven paths explained unique variance in sibling differences. The reciprocal model also had the highest CFI, confirming that, among the four competing models, the reciprocal model best reproduced the observed covariance matrix for fathers.

Table 4.

Standardized Path Coefficients and Fit Statistics for Father Models (N = 350 sibling dyads)

Model 1b Model 2b Model 3b Model 4b Model 5b
Stability Conflict Behavior Reciprocal Final
Path coefficients
Conflict T1 → Conflict T2 0.65** 0.66** 0.62** 0.63** 0.63**
Conflict T2 → Conflict T3 0.64** 0.65** 0.61** 0.62** 0.63**
Behavior T1 → Behavior T2 0.88** 0.83** 0.89** 0.84** 0.83**
Behavior T2 → Behavior T3 0.64** 0.59** 0.65** 0.60** 0.61**
Conflict T1 → Behavior T2 - 0.14* - 0.14* 0.16**
Conflict T2 → Behavior T3 - 0.20* - 0.19** 0.16**
Behavior T1 → Conflict T2 - - 0.09* 0.09* 0.06*
Behavior T2 → Conflict T3 - - 0.06* 0.06* 0.06*
Fit statistics
X2 (df) 42.96 (8) 24.57 (6) 34.23 (6) 16.72 (4) 17.83 (7)
Δ X2 compared to Model 1b - 18.39** 8.73* 26.24** -
Δ X2 compared to Model 2b - - - 7.85* -
Δ X2 compared to Model 3b - - - 17.51** -
Δ X2 compared to Model 4b - - - - 1.11
RMSEA .11 .09 .12 .10 .07
CFI .96 .98 .97 .99 .99
Open in a new tab
*

p < .05.

**

p < .01.

Comparisons exploring the equivalence of the parallel paths indicated that the only parallel paths that showed significant differences between Times 1 and 2 and Times 2 and 3 were the stability paths for sibling differences in risky behavior, X2(1) = 11.43, p < .01, again suggesting that the temporal stability of these sibling differences decreased over time. All other parallel paths were constrained to be equal in the final model (Model 5b). Focusing on the path coefficients, differential father-child conflict was significantly and positively linked to sibling differences in risky behavior, meaning that youths who had more conflict with their fathers relative to their siblings subsequently engaged in more risky behavior relative to their siblings. Effects of the opposite direction were also significant, such that youths who engaged in more risky behavior relative to their siblings had more conflict with their fathers relative to their siblings at the next time point. The effect sizes were small (Kline, 2011), however, especially when compared to the large effect sizes of the stability paths.

The last step was to compare the final models for mothers and for fathers and test for potential differences by parent gender. The two final models overlapped in the time-lagged paths from differential parent-child conflict to sibling differences in risky behavior. The coefficient estimates were nevertheless the same (i.e., .16) for both parents, and thus the planned reassignment of values would not affect the father model’s ability to reproduce the observed covariance matrix. Because the final model for mothers did not include the behavior-driven paths, it was not possible to compare mothers and fathers on those associations. Taken together, our findings revealed that, although the conflict-driven effects were the same for the two parents, the behavior-driven effects were evident for fathers but not for mothers.

Discussion

Consistent with theory (Ansbacher & Ansbacher, 1956; Festinger, 1954) and prior work with MZ twins (Asbury et al., 2003, 2006; Burt et al., 2006), our findings indicated that parents played a role in the differentiation of nontwin siblings: Youths who experienced relatively more conflict with their parents subsequently reported higher levels of risky behavior compared to their siblings. A number of studies have documented stronger differentiation effects among highly discordant MZ twins (Burt et al., 2006) and in high-risk environments (Asbury et al., 2003, 2006), lending support to a long-maintained behavioral genetics notion that it requires extremely unfavorable experiences to exert sustainable environmental influences on youth development (Scarr & McCartney, 1983; Turkheimer & Waldron, 2000). Our results offered another perspective by showing that objectively nonshared family environments within an average, expectable range also contributed to sibling differentiation: Even within a relatively low-risk, community sample of nontwin siblings, differential mother- and father-child conflict had a consistent, longitudinal influence on sibling differences in risky behavior. Perhaps more importantly, these time-lagged associations, though modest in magnitude, emerged in the context of remarkably stable differential conflict and risky behavior scores. The stability of differential mother-child conflict and sibling differences in risky behavior declined over time, however, reflecting the observation that adolescence is a period when parents and youths undergo intense exchanges and realignment of expectations (Laursen & Collins, 2009) and when youths begin to explore new roles and experiment with a wide range of behaviors (Jessor, 1992). Given that adolescence is a time of change and that mother- and father-child conflict play similar roles in sibling differentiation in risky behavior, promoting conflict resolution strategies may constitute an important focus of interventions directed at preventing adolescent delinquency. Moreover, considering that siblings often use one another as sources of comparison (Ansbacher & Ansbacher, 1956; Festinger, 1954), working with parents to address potential feelings of jealousy and rivalry that may arise when youths have differential family experiences relative to their siblings, such as by helping them to understand the reasons for differential treatment, may be a promising approach for practitioners to help families with deviant youths (Caspi, 2011).

A number of theorists have pointed to the roles of youths in shaping their own environments (McHale & Crouter, 2003; Scarr & McCartney, 1983), and yet there is a tendency in research and practice to focus on unidirectional effects from parents to children. While we did find evidence of parental influences, our results also revealed that fathers’ differential conflict with their children was predicted by differences in their children’s characteristics. Our study contributes to the small body of work on child-driven effects (Laursen & Collins, 2009) in its focus on the sibling subsystem: Youths’ behavior and personal qualities elicit reactions from their environments, and this evocation process may be particularly salient in a within-family context as there are frequent opportunities to compare siblings side-by-side. Because youths become more able to actively construct their environments as they get older, such child-driven effects may increase in magnitude over time (Scarr & McCartney, 1993). The magnitude of the child-driven effects we identified was consistently modest, however. A longer time frame covering adolescence through adulthood, when youths begin to leave home and become more self-sufficient, may be needed to capture this hypothesized change in child-driven effects.

In general, our exploration of mother-father differences revealed similar effects for both parents. The single gender difference that emerged indicated that differential paternal, but not maternal, conflict was predicted by sibling differences, suggesting that fathers may be more sensitive to siblings’ unique behavior or personal qualities. Although this effect was small, it is consistent with the larger literature on fathering, which posits that mothers’ roles are more scripted than fathers’ in two-parent families (Parke & Buriel, 1998). The traditional paternal role does not require highly engaged parenting, meaning that fathers’ parenting behavior may be more voluntary and therefore more conditioned by offspring’s characteristics. In contrast, the maternal role of caregiver prescribes active parental efforts, and mothers’ parenting behaviors may be less dependent upon what their offspring are like. Indeed, there is evidence that child characteristics, such as gender and temperament, are more strongly associated with fathers’ than with mothers’ emotional warmth (McBride, Schoppe, & Rane, 2002) and general involvement (Parke & Buriel, 1998). Because little research has directly examined fathers’ roles in sibling differentiation, our results should be treated as hypothesis generating. However, to the extent that fathers’ parenting activities are less scripted than mothers’, fathers may be an important lever for family interventions. Parents, especially fathers, should be educated about how sibling differences may evoke differential responses from them and how such child-specific experiences may further reinforce existing sibling differences. Programs designed to teach parents to refrain from overt comparisons between siblings (e.g., “Why can’t you be more like your brother?”) and instead to attend to the unique strengths and needs of each child (Caspi, 2011) have the potential to reduce confirmatory biases (on the part of parents) and self-fulfilling prophesies (on the part of youths), both of which may contribute to unhealthful sibling differentiation. More generally, family-focused prevention and intervention programs very rarely incorporate sibling dynamics, but our findings suggest that focusing on only one child in a family may limit understanding of how families operate as social and socializing systems.

It is noteworthy that the child-driven effects in our model, which showed that fathers had more conflict with the more delinquent sibling, were in the opposite direction from those found by Conger and Conger (1994), which showed that mothers exhibited less hostility toward the more deviant sibling. One possible reason for the discrepancy is that Conger and Conger (1994) measured parent-child relationships through observation, which captured specific hostile behaviors during the observation period, whereas we measured parent-child relationships through self-reports, which tapped into a general style of parent-child conflict across a variety of domains. Another possibility is that, although mothers may be less likely to be influenced by child characteristics compared to fathers, when it does happen, mothers and fathers may react in different ways. A study by Sheeber, Hops, Andrews, Alpert, and Davis (1998), for example, demonstrated that, in an experimental setting, mothers exhibited increased facilitative behavior when their child was more depressed, whereas fathers responded by becoming less aggressive. The authors interpreted the findings as evidence that mothers may become more engaged with, but fathers, more distant from, youths with psychosocial problems. An important direction for future studies is to examine the overlapping as well as distinctive features of mothers’ versus fathers’ child-specific parenting and how these features are linked to youth adjustment. From a prevention standpoint, findings indicating differential reactions of mothers versus fathers to sibling differences may underscore the importance of tailored programing for each parent.

Our study is not without limitations. First, genetic differences are confounded with both environmental and behavioral differences between siblings (Turkheimer & Waldron, 2000). Because nontwin siblings only share an average of 50% of their genes, their behavioral differences may provide for more variance to explain and may result in a greater number of significant effects in our models. Our full sibling sample, however, did not allow us to control for genetic variation as a confounding factor. A longitudinal design that includes sibling pairs of different age spacing and gender constellations and with varying degrees of genetic similarity (e.g., half-siblings, adopted siblings) is needed to provide definitive evidence supporting the effective role of differential parent-child conflict in creating differences between siblings. Second, although the inclusion of multiple informants helped reduce potential bias due to common method variance, our measures of parent-child conflict and risky behavior were based on self-reports and might be affected by respondents’ self-presentation tendencies. The incorporation of less subjective measures, such as behavioral observation (e.g., Conger & Conger, 1994), would be useful for validating our findings. Finally, our sample was ethnically homogeneous and not representative of the diversity of families in the US. Given that youths react less negatively to parental differential treatment when they regard it as fair (Caspi, 2011; McHale & Crouter, 2003), our findings need to be replicated in more diverse samples, such as families with collectivistic cultural backgrounds.

Despite these limitations, our focus on directions of effect and both mothers and fathers contributes to the literature on sibling differentiation by illustrating how sibling differences unfold over time and how mothers and fathers may play different roles in the underlying processes. On an applied level, our findings alert practitioners who work directly with youths and families about the importance of understanding families as systems: Our use of between-sibling comparisons directs attention beyond youths as individuals to their relative positions in a larger sibling subsystem; our emphasis on within-family processes also illuminates how youths simultaneously drive and react to their unique experiences with parents.

Acknowledgments

The authors wish to express gratitude to Michael J. Rovine and Siwei Liu for their insights into our data analytic approach; to our undergraduate and graduate assistants, staff, and faculty collaborators for their help in conducting this study; and to the participating families for their time and cooperation. This work was funded by grants from the National Institute of Child Health and Human Development (R01-HD32336 and R01-HD29409) to Ann C. Crouter and Susan M. McHale, Co-Principal Investigators.

Footnotes

Portions of this article were presented at the Society for Research on Adolescence Biennial Meeting, Philadelphia, PA, March 2010.

References

  1. Anderson ER, Hetherington EM, Reiss D, Howe G. Parents' nonshared treatment of siblings and the development of social competence during adolescence. Journal of Family Psychology. 1994;8:303–320. [Google Scholar]
  2. Ansbacher HL, Ansbacher RR. The individual psychology of Alfred Adler. New York, NY: Basic Books; 1956. [Google Scholar]
  3. Asbury K, Dunn JF, Pike A, Plomin R. Nonshared environmental influences on individual differences in early behavioral development: A monozygotic twin differences study. Child Development. 2003;74:933–943. doi: 10.1111/1467-8624.00577. [DOI] [PubMed] [Google Scholar]
  4. Asbury K, Dunn JF, Plomin R. Birthweight-discordance and differences in early parenting relate to monozygotic twin differences in behavior problems and academic achievement at age 7. Developmental Science. 2006;9:22–31. doi: 10.1111/j.1467-7687.2006.00469.x. [DOI] [PubMed] [Google Scholar]
  5. Brody GH, Stoneman Z, McCoy JK. Parental differential treatment of siblings and sibling differences in negative emotionality. Journal of Marriage and the Family. 1992;54:643–651. [Google Scholar]
  6. Burt SA, McGue M, Iacono WG, Krueger RF. Differential parent-child relationships and adolescent externalizing symptoms: Cross-lagged analyses within a monozygotic twin differences design. Developmental Psychology. 2006;42:1289–1298. doi: 10.1037/0012-1649.42.6.1289. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Caspi J. Sibling aggression: Assessment and treatment. New York, NY: Springer; 2011. [Google Scholar]
  8. Conger KJ, Conger RD. Differential parenting and change in sibling differences in delinquency. Journal of Family Psychology. 1994;8:287–302. [Google Scholar]
  9. Davey A, Savla J, Luo Z. Issues in assessing model fit with missing data. Structural Equation Modeling. 2005;12:578–597. [Google Scholar]
  10. Eccles J, Barber B. The risky behavior scale. University of Michigan; 1990. Unpublished manuscript. [Google Scholar]
  11. Farrington DP. Conduct disorder, aggression, and delinquency. In: Lerner RM, Steinberg L, editors. Handbook of adolescent psychology: Vol. 1. Individual bases of adolescent development. 3rd ed. Hoboken, NJ: Wiley; 2009. pp. 683–722. [Google Scholar]
  12. Feinberg ME, Hetherington EM. Differential parenting as a within-family variable. Journal of Family Psychology. 2001;15:22–37. doi: 10.1037//0893-3200.15.1.22. [DOI] [PubMed] [Google Scholar]
  13. Festinger L. A theory of social comparison processes. Human Relations. 1954;7:117–140. [Google Scholar]
  14. Hipwell AE, Keenan K, Kasza K, Bean T, Loeber R, Stouthamer-Loeber M. Reciprocal influences between girls’ behavioral and emotional problems and parental punishment and warmth: A six-year prospective analysis. Journal of Abnormal Child Psychology. 2008;36:663–677. doi: 10.1007/s10802-007-9206-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jessor R. Risk behavior in adolescence: A psychosocial framework for understanding and action. Developmental Review. 1992;12:374–390. doi: 10.1016/1054-139x(91)90007-k. [DOI] [PubMed] [Google Scholar]
  16. Jöreskog KG, Sörbom D. LISREL 8 user's reference guide. 2nd ed. Lincolnwood, IL: Scientific Software International; 1999. [Google Scholar]
  17. Kenny DA, Kashy DA, Cook WL. Dyadic data analysis. New York, NY: Guilford; 2006. [Google Scholar]
  18. Kline RB. Principles and practice of structural equation modeling. 3rd ed. New York, NY: Guilford; 2010. [Google Scholar]
  19. Laursen B, Collins WA. Parent-child relationships during adolescence. In: Lerner RM, Steinberg L, editors. Handbook of adolescent psychology: Vol. 2. Contextual influences on adolescent development. 3rd ed. New York, NY: Wiley; 2009. pp. 3–42. [Google Scholar]
  20. McBride BA, Schoppe SJ, Rane TR. Child characteristics, parenting stress, and parental involvement: Fathers versus mothers. Journal of Marriage and Family. 2002;64:998–1011. [Google Scholar]
  21. McHale SM, Crouter AC. How do children exert an impact on family life? In: Crouter AC, Booth A, editors. Children's influence on family dynamics: The neglected side of family relationships. Mahwah, NJ: Erlbaum; 2003. pp. 207–220. [Google Scholar]
  22. Mullineaux PY, Deater-Deckard K, Petrill SA, Thompson LA. Parenting and child behaviour problems: A longitudinal analysis of non-shared environment. Infant and Child Development. 2009;18:133–148. doi: 10.1002/ICD.593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Parke RD, Buriel R. Socialization in the family: Ethnic and ecological perspectives. In: Eisenberg N, Damon W, editors. Handbook of child psychology: Vol. 3. Social, emotional, and personality development. 5th ed. New York, NY: Wiley; 1998. pp. 463–552. [Google Scholar]
  24. Plomin R, Daniels D. Why are children in the same family so different from one another? Behavioral and Brain Sciences. 1987;10:1–16. [Google Scholar]
  25. Rovine M, Liu S. Structural equation modeling approaches to longitudinal data. In: Newsom JT, Jones RN, Hofer SM, editors. Longitudinal data analysis: A practical guide for researchers in aging, health, and social sciences. New York, NY: Routledge; 2012. pp. 243–270. [Google Scholar]
  26. Scarr S, McCartney K. How people make their own environments: A theory of Genotype → environment effects. Child Development. 1983;54:424–435. doi: 10.1111/j.1467-8624.1983.tb03884.x. [DOI] [PubMed] [Google Scholar]
  27. Sheeber L, Hops H, Andrew J, Alpert T, Davis B. Interactional processes in families with depressed and nondepressed adolescents: Reinforcement of depressive behavior. Behavioral Research and Therapy. 1998;36:417–427. doi: 10.1016/s0005-7967(97)10030-4. [DOI] [PubMed] [Google Scholar]
  28. Smetana JG. Concepts of self and social convention: Adolescents’ and parents’ reasoning about hypothetical and actual family conflicts. In: Gunnar MR, Collins WA, editors. Minnesota Symposium on Child Psychology: Vol. 21. Development during the transition to adolescence. Hillsdale, NJ: Erlbaum; 1988. pp. 79–122. [Google Scholar]
  29. Tamrouti-Makkink ID, Dubas JS, Gerris JRM, van Aken MAG. The relation between the absolute level of parenting and differential parental treatment with adolescent siblings’ adjustment. Journal of Child Psychology and Psychiatry. 2004;45:1397–1406. doi: 10.1111/j.1469-7610.2004.00846.x. [DOI] [PubMed] [Google Scholar]
  30. Tomarken AJ, Waller NG. Potential problems with ‘‘well fitting” models. Journal of Abnormal Psychology. 2003;112:578–598. doi: 10.1037/0021-843X.112.4.578. [DOI] [PubMed] [Google Scholar]
  31. Tucker CJ, McHale SM, Crouter AC. Dimensions of mothers’ and fathers’ differential treatment of siblings: Links with adolescents’ sex-typed personal qualities. Family Relations. 2003;52:82–88. [Google Scholar]
  32. Turkheimer E, Waldron M. Nonshared environment: A theoretical, methodological, and quantitative review. Psychological Bulletin. 2000;126:78–108. doi: 10.1037/0033-2909.126.1.78. [DOI] [PubMed] [Google Scholar]
  33. US Census Bureau. Census 2000 Gateway. 2000 Retrieved from http://www.census.gov/
  34. Wu W, West SG, Taylor AB. Evaluating model fit for growth curve models: Integration of fit indices from SEM and MLM frameworks. Psychological Methods. 2009;14:183–201. doi: 10.1037/a0015858. [DOI] [PubMed] [Google Scholar]

RESOURCES