Chapter IV. Developmental Trajectories of Children’s Aggressive Behaviors After the Birth of a Sibling

Dunn & Kendrick (1982) reported that some children were quite aggressive toward mothers and their infant siblings shortly after the birth. These early behaviors 3 weeks after the birth predicted the quality of sibling relations at 14 months after the birth, which, in turn, continued to predict early sibling relationship quality when the younger sibling was 6 years old (Stillwell & Dunn, 1985). Stewart (1990) also found that 36.5% of the 41 children in his study were involved in physical confrontations with the infant at 1 month, whereas 47. 5% were aggressive with mothers and 40% were aggressive with their fathers at 1 month after the birth. Although the number of children in physical confrontations with family members decreased substantially by 4 months, the number increased again by 8 months after the birth, with most physical confrontations focused on the sibling at 12 months. Thus, understanding early patterns of aggression after the birth of a sibling takes on added significance given the long-term continuity of aggression and sibling relationship quality from the earliest months after the birth and from early childhood (Hay et al., 2014; Song & Volling, 2015).

Development of Aggression in Early Childhood

Numerous studies have examined the development of aggression in early childhood, finding that aggression and other disruptive behaviors such as noncompliance and defiance are quite common in toddlerhood (i.e., terrible twos) and the preschool years (Campbell, 2002). In general, physical aggression increases from 2 to 3 years and begins to decline around 4 years of age (Alink et al., 2006; NICHD Early Child Care Research Network, 2004; Tremblay et al., 1999), a time during which many children experience the transition to siblinghood (Baydar et al., 1997). When aggressive behavior does not decline from toddlerhood into elementary school, children are considered at risk for numerous social-emotional problems and maladjustment (Campbell, 2002). Despite the decline in children’s aggression over early childhood, there are notable individual differences in young children’s aggression that are fairly stable over the preschool period (e.g., Alink et al., 2006). Several investigators have now examined individual differences in longitudinal trajectory patterns across toddlerhood, preschool and into elementary school in an effort to identify those children most at risk for developing persistent patterns of aggressive behavior.

Individual Differences in Young Children’s Aggression

There are now numerous studies using mixture modeling to examine group-based trajectories of aggressive behavioral patterns. We focus here on those studies that have examined the early toddler and preschool years as this is the period that coincides with the transition to siblinghood and thus, our findings need to be understood within this developmental framework. In a sample of 572 children followed longitudinally from 5 to 42 months, Tremblay and his colleagues (2004) identified three physical aggression trajectories across the 17, 30, and 42 month time points. The first was a group who exhibited little or no physical aggression over the course of time and accounted for 28% of the sample. The largest group of children (58%) followed a pattern in which children started relatively low but increased modestly in their physical aggression from 17 to 42 months. The final group consisting of 14% of children started out higher on aggression than the other two groups and continued to increase in their aggression from 30 to 42 months.

Other researchers examining trajectories of externalizing behavior problems, which includes aggression, during the toddler and preschool years, reveal similar longitudinal trajectory patterns; generally finding three or four groups of children who often differ with respect to their starting points (intercepts) and rates of change (i.e. linear slope). These groups often consist of a relatively large group of children with low externalizing problems over time, a smaller group of children with generally high and persistent externalizing problems, and groups who start at different points and show declines starting around 2 years of age (e.g., Hill, Degnan, Calkins, & Keane, 2006; NICHD Early Child Care Research Network, 2004). Tremblay et al (2004) documented that the largest risk factor for distinguishing the relatively small number of children in their high aggression trajectory group was the presence of young siblings in the family. Having a sibling as the target for physical aggression increased the odds of membership in the high aggression group by more than a factor of 4.

Based on these earlier findings, we hypothesized there would also be variability (i.e., individual differences) in children’s trajectories of aggressive behavior after the birth of a sibling and that we could identify distinct groups of children. We anticipated a large normative group of children with low aggression across the course of the year following the birth (i.e., low-stable). We also hypothesized that there may be a smaller, yet high-risk, group of children who would be persistently higher in their aggression over time and who may very well increase in aggression under the stress of the transition (e.g., high-stable or high-increasing). In her study of hard to manage preschoolers, Campbell (2002) noted that these children reacted strongly to the birth of a younger sibling. As noted earlier, we also tested for an adjustment and adaption response and a maladaptive response (i.e., sudden persistent change) in line with theories of family risk and resilience, and the emphasis on the transition to siblinghood as a psychosocial crisis for firstborns based on psychodynamic theorizing (Adler, 1928; Levy, 1934). Specifically, we tested whether increases in aggression from prenatal to 1 month after the birth increased and if so, how long-lived this increase was. Dunn & Kendrick (1982) claimed that emotional upset and disruptive behavior had diminished by the end of the first month after the birth and most children appeared to have adapted shortly thereafter. Thus, if aggression did increase, we expected to see an adjustment and adaptation response, in which increases in aggression from prenatal to 1 month would have returned to pre-birth levels by 4 months after the birth, indicating that children had adapted to changes in their families. The adjustment and adaptation response would provide evidence of potential disturbance in response to the stresses of change brought about by the birth, but this effect would subside by 4 months once the family had adapted to the arrival of the new family member and adjusted to the change in roles and expectations.

In contrast to these stable and adaptive patterns, we also tested for several other trajectory patterns that might manifest and reflect increases in problem behavior over time. One maladaptive response reflecting a sudden increase from prenatal to 1 month that persisted and remained high throughout the year could be tested with a quadratic polynomial or curvilinear pattern whereas a second pattern would show a gradual linear increase in aggression over time. Because no prior study has examined individual differences in growth trajectories of aggression across the transition to siblinghood, our modeling of trajectory patterns is based both on theories of family stress and resilience and prior research examining linear patterns of change in aggression over early childhood.

Risk and Protective Factors in the Prediction of Early Aggression

Potential risk factors that maintain stable aggressive tendencies in early childhood are generally grouped into child, parent, and family characteristics, which are also consistent with the developmental ecological systems perspective (Volling, 2005). With respect to child characteristics, temperamental difficulty or negative emotionality has been examined frequently and seems to be a developmental precursor to the emergence of externalizing behavior problems, although the exact process by which this occurs is not entirely clear (Bates, Maslin, & Frankel, 1985; Shaw, Keenan, & Vondra, 1994; Morris, Keane, Calkins, Shanahan, & O’Brien, 2014; Gartstein, Putnam, & Rothbart, 2012). High negative emotionality is posited as a central feature of psychopathology because difficult and fussy infants, and young children, often have difficulty regulating their anger and may be prone to the use of aggression in their social interactions (Hay et al., 2014).

In the parenting domain, hostile, power-assertive, and rejecting maternal behavior, or the lack of maternal acceptance and warmth, have been related consistently with the development of externalizing behaviors in young children (Dishion & Patterson, 2006; Miner & Clarke-Stewart, 2008; Choe, Olson, & Sameroff, 2013; Towe-Goodman & Teti, 2008), although the relations between maternal and child behaviors most likely reflect a transactional process of increasing coercion and punishment over time (e.g. parental reactivity hypothesis) rather than a simple unidirectional relation (Kochanska & Kim, 2012; Combs-Ronto, Olson, Lunkenheimer, & Sameroff, 2009; Collins, Maccoby, Steinberg, Hetherington, & Bornstein, 2000). Further, children with insecure avoidant and disorganized attachments to their mothers are more likely to have externalizing behavioral difficulties (Fearon et al., 2010; Fearon & Belsky, 2011), whereas positive father involvement and children with secure father-child attachments are less likely to have externalizing problem behaviors in early childhood (Boldt, Kochanska, Yoon, & Koenig Nordling, 2014; DeKlyen, Biernbaum, Speltz, & Greenberg, 1998; DeKlyen, Speltz, & Greenberg, 1998; Herbert, Harvey, Lugo-Candelas, & Breaux, 2013).

Maternal depressive symptoms have also been linked to the development of children’s disruptive behavior, whether this link is direct or mediated by its effect on ineffective parenting practices and inconsistent discipline (Tremblay et al., 2004; Owens & Shaw, 2003; Dix & Yan, 2014). Parents, both mothers and fathers, reporting lower parental self-efficacy also reported their children had more behavior problems (Gross & Tucker, 1994; Hill & Bush, 2001), although it is not clear if this association is a reflection of parent’s feelings of ineffectiveness in the parental role or an actual increase in problematic child behaviors. In an earlier report, we found that mothers’ and fathers’ parental self-efficacy at 1 month after the infant’s birth, specifically their inability to control and discipline their older child when naughty, predicted the older sibling’s antagonistic interactions with the younger sibling later in the year following the birth (Oh, Volling, & Gonzalez, 2015). When parents felt ineffective in disciplining the older sibling shortly after birth, their children were engaging in aggressive and antagonistic behavior with their 4-month-old infant siblings.

Interparental or marital conflict has been consistently tied to young children’s disruptive behavior (Peterson & Zill, 1986; Lindsey, Caldera, & Tankersley, 2009; Amato & Cheadle, 2008) and externalizing behavior problems (Cummings, Goeke-Morey, & Papp, 2004; E. M. Cummings & Davies, 2011; Davies, Cicchetti, & Martin, 2012; Kaczynski, Lindahl, Malik, & Laurenceau, 2006). Coparenting between mothers and fathers in the form of either supportive or conflictual relations also predicts externalizing and disruptive behaviors in young children (Belsky, Woodworth, & Crnic, 1996; Schoppe-Sullivan et al., 2009). We have also found that home observations of coparenting (low support and undermining) between parents during triadic interaction with the firstborn before the birth predicted increases in externalizing behaviors 1 month after the birth, and that supportive and undermining coparenting interacted so that the greatest increases in externalizing behavior problems occurred when mothers and fathers were high on undermining coparenting and low on supportive coparenting (Kolak & Volling, 2013). Parenting stress (Shaw et al., 2001) and the social support parents receive from family and friends may play a role in the emergence of disruptive behaviors, most likely due to the fact that reliance on others for emotional, financial, and instrumental support can alleviate much of the caregiving stress, reduce role overload and enhance parental mental health (Lee, Anderson, Horowitz, & August, 2009; Guralnick, Hammond, Neville, & Connor, 2008; Hoagwood et al., 2010; DeGarmo, Patras, & Eap, 2008; Leahy-Warren, McCarthy, & Corcoran, 2012).

In the current study, we considered children’s temperamental characteristics (e.g., negative emotionality), parent characteristics (e.g., depression, parenting behaviors, feelings of efficacy, attitudes toward spanking), and family and social contextual factors (e.g., marital quality, social support, coparenting, parenting hassles) before the birth as predictors of our aggression trajectories as part of the variable selection algorithms. Variable selection procedures are preferable when the goal is to pinpoint which factors are most strongly related to outcomes and can provide the evidence base needed to target effective interventions. Should a high-risk group of children showing clinically-significant levels of aggression be identified, either before or after the birth of their infant sibling, this approach lets us identify those child, parent, and social-contextual factors most relevant for targeted interventions for these families.

Sibling Relationships and Early Aggression

According to Patterson (1986), coercive sibling interaction in childhood serves as a “training ground” for the development of aggressive behavior (i.e., the sibling training hypothesis). The strongest risk factor in predicting the trajectories of a highly aggressive group of toddler boys was the presence of a sibling in the household (Tremblay et al., 1999). The family is the first place that many young children encounter conflict and sibling conflict is commonplace in most households with two or more young children. Dunn & Munn (1986) reported that sibling conflict in early childhood occurred approximately 7 times an hour; similar to the 6.3 times per hour reported by Perlman & Ross (1997) with a sample of 2- and 4-year-old siblings. Although ubiquitous, sibling conflict can be both mild (e.g., argumentative) or severe (e.g., violent) and involve both destructive and constructive conflict resolution. Destructive conflict-resolution during sibling conflict, even for very young children, predicted aggressive behavior problems among preschool boys (Garcia et al., 2000). Early sibling conflict and aggression seems to play an influential role in the development of preschool children’s emerging emotion regulation and aggression. Further, sibling conflict is strongly associated with the development of children’s externalizing and internalizing difficulties (Dirks et al., 2015). Therefore, we examined whether the resulting aggression trajectories would predict sibling relationship quality, specifically more conflict and less positive involvement one year after the birth. We hypothesized that trajectories of aggression should predict the quality of the children’s sibling relationship at the end of the year, particularly levels of sibling conflict. Children expressing high levels of aggressive behavior over the course of the year would be engaged in more sibling conflict and less positive sibling relationships at 12 months.

In the current study, we used a group-based trajectory analysis (growth mixture modeling) to ascertain different patterns of children’s aggressive behavior across the transition, hypothesizing that children who evinced increased aggression or who had been highly aggressive before the infant’s birth and continued to remain high over the course of the year would be engaged in more sibling conflict 1 year after the birth compared to children who were low on aggression or experienced minimal change in aggression over the course of the transition.

Results

Individual Differences in Trajectories of Aggressive Behavior

Having determined in Chapter III that the unconditional Adjustment and Adaptation Response (AAR) model fit the aggression data better than both the unconditional latent linear and unconditional quadratic models, we used growth mixture modeling (GMM) to model the variability in the growth parameters and identify the patterns of aggression trajectories across the five timepoints of the study. We determined that the five-class Adjustment and Adaptation Response model (we will use AAR throughout the Results section for ease of presentation) solution with the fixed effects (i.e., intercept, linear slope, and AAR contrast) and the random effects (i.e., intercept, linear slope) of growth parameters was considered the best fitting model based on fit indices; the five-class model had a lower BIC and AIC, AIC = 5495.30, BIC = 5588.13, than both the four-, AIC = 5497.03, BIC = 5576.11, and six-class models, AIC = 5497.94, BIC = 5604.52, and higher entropy (.705) than the four-class model (.673). Table 4 presents estimates and standard errors for the fixed effects of each class, which provide interpretation of the changes observed within each class, in addition to the random effects. Figure 3 shows the different trajectory patterns for the five classes. In determining class labels, we used a naming convention in which we denoted the starting value (intercept) first (e.g., high, mid, low) and then the pattern of change most characteristic of this class (e.g. decreasing, AAR).

Table 4.

Growth Mixture Model Results for Aggressive Behavior: Parameter Estimates and Standard Errors for Fixed Effects

Classes	Low-increasing	Low-AAR	Mid-stable	Mid-AAR-decreasing	High-decreasing
Parameters	n = 92	n = 68	n = 44	n = 18	n = 8
	(40%)	(29.6%)	(19.1%)	(7.8%)	(3.5%)
Intercept	5.222***	9.363***	12.565***	13.385***	19.376***
	(.483)	(.854)	(.786)	(.813)	(1.142)
Linear slope	.254**	−.143	−.109	−.431*	−.736**
	(.081)	(.139)	(.195)	(.177)	(.251)
AAR	.060	.710***	−.241	1.831***	−.044
	(.105)	(.215)	(.239)	(.224)	(.312)

Note. Standard errors in parentheses. AAR = Adjustment and Adaptation Response. The random effect (variance) of intercept est. = 2.949, SE = 1.386, p = .033; the random effect of linear slope est. = .037, SE = .048, p = ns; the random effect of the AAR contrast was constrained to be 0.

^*p < .05,

^**p < .01,

^***p ≤ .001.

Figure 3.

Trajectory classes for aggressive behavior from Growth Mixture Model (n=230).

The largest class was denoted as the low-increasing class and represented 40% of the sample (n = 92). As can be seen in Table 4 and Figure 3, children in this class showed the lowest levels of aggressive behavior at the prenatal time-point with a gradual linear increase over the first year after the birth of the sibling. A second class of children showed a significant increase in their aggressive behavior from prenatal to 1 month and then a decrease by 4 months (i.e., AAR effect). This low-AAR class, comprising 29.6% of the sample (n = 68), had low levels of aggressive behavior at the prenatal time point, followed by an increase from prenatal to 1 month, but a decrease from 1 to 4 months, and then remained stable for the rest of the year. A third class was labeled the mid-stable class (19.1%; n = 44) and showed moderate levels of aggressive behavior at the prenatal time point that remained stable throughout the first year after the birth of the sibling. The fourth, mid-AAR-decreasing class, accounted for only 7.8% of the sample (n = 18) and started out with mid-to-high levels of aggressive behavior at the prenatal time point, followed by a fairly dramatic increase from prenatal to 1 month in aggression, then a subsequent decrease from 1 to 4 months, with a further decrease in aggressive behavior for the rest of the year. The smallest class, high-decreasing, by contrast, accounted for 3.5% of the sample (n = 8) and showed the highest level of aggressive behavior at the prenatal timepoint with a significant linear decrease over time. Thus, results are consistent with theoretical patterns indicative of linear increase, stability and no change, as well as adjustment and adaptation as presented in Figure 1 and with none of the curvilinear patterns of change in Figure 2, including sudden, persistent change reflecting a psychosocial crisis model.

FIGURE 1.

Theorized trajectory patterns of change reflecting adjustment and adaptation, stability, and linear increase.

Figure 2.

Theorized curvilinear trajectory patterns reflecting sudden persistent change, delayed impact, and growth and maturity.

Spaghetti plots showing the individual trajectories for each child by class membership can be found in Figure 4. For each plot, the lower line represents the normative mean for the CBCL aggression subscale (1.5-5: Achenbach & Rescorla, 2000), the second line represents the 92.5^th percentile, which indicates the borderline clinical range cut-off, and the top line represents the 97^th percentile, which denotes the clinical range cut-off. What these figures clearly show is that the five classes fit precisely within these cut-off ranges, and are in line with what we would expect for a community-based low risk sample in which most children were low and fell well below the borderline or clinical cut-offs for aggressive behavior, and only a few children were identified in the clinical cut-off range. For instance, all children within the low-increasing class were well below the normative mean, whereas all the mid-stable children fit within the borderline clinical range. Only the 8 children in the high-decreasing class were near or above the clinical range cut-off. These figures demonstrate that the identified classes represent distinct and meaningful classes of children who fall within the normative, borderline and clinical ranges established for the CBCL. Also, the mid-AAR-decreasing class differs from the low-AAR class, with the low-AAR starting in the low range and reaching the borderline range after the birth, and the mid-AAR starting in the borderline range and hitting the clinical range cut-off. Thus, the GMM analyses identified two classes of children, both having an AAR effect, but one actually experienced clinical levels of problem behaviors in response to the sibling’s birth. This group of children is of particular interest because even though small in number there is evidence that they are experiencing behavioral changes severe enough to be clinically significant. Although several of the resulting classes have few children and could be considered statistical outliers or too few for class enumeration, these children are most deserving of our attention and should not be dismissed for purely statistical reasons.

Figure 4.

Spaghetti plots of individual trajectories for each of the five aggression classes showing the mean (lowest line), borderline-clinical (92.5%; middle line), and clinical (97.5%; highest line) levels.

Predicting Aggressive Trajectories

The combined recursive tree and random forest variable selection procedure identified four candidate predictors of the five classes of the aggression trajectories. These predictors were children’s negative emotionality, attachment security to mother, fathers’ parental-self efficacy, and observations of undermining coparenting during prenatal home visits. We then tested whether these four variables were statistically significant predictors of the five aggression classes by conducting a multinomial logistic regression with each predictor entered as a continuous, centered, additive predictor using the low-increasing class as the reference class. We computed a second multinomial logistic regression that also included all possible higher order interactions between the five predictors. The full model with all higher order interactions was not significantly better than the reduced model, LR Chisq (df = 44) =51.15, p = .21, so we report the results of the reduced model with additive main effects.

Table 5 presents the coefficients, the standard errors, the odds ratios, and z-tests for the multinomial regressions using the large, low-increasing class as the reference class. The z-tests show that children’s negative emotionality is a significant predictor of differences between each of the four classes and the low-increasing reference group (all z’s > 1.96). In all cases, there is a greater probability of being in the low-AAR, mid-stable, mid-AAR-decreasing, and high-decreasing class for children higher in negative emotionality compared to the low-increasing class. Children’s attachment security to mother significantly predicted the difference between the high-decreasing class and the low-increasing reference class such that lower attachment security to mother is associated with a higher probability of being in the high-decreasing class relative to the low-increasing reference class. Father’s parental efficacy was a significant predictor of the mid-stable and the high-decreasing class, with higher scores (i.e., greater feelings of ineffectiveness in disciplining) predicting a greater probability in both the mid-stable and high-decreasing classes relative to the low-increasing class.

Table 5.

Results of Multinomial Logistic Regression Analysis Examining Class Differences for Aggressive Behavior with the Low-Increasing Class as the Reference Class

Predictor	Low-increasing vs.	b	SE	z	P	OR
Child Negative Emotionality	Low-AAR	1.07	0.34	3.14	0.002	2.92
	Mid-stable	2.56	0.50	5.13	<0.001	12.97
	Mid-AAR	1.54	0.56	2.77	0.006	4.66
	High-decreasing	6.88	2.15	3.21	0.001	976.31
Undermining Coparenting	Low-AAR	0.11	0.10	1.05	n.s.	1.11
	Mid-stable	−0.19	0.15	−1.25	n.s.	0.83
	Mid-AAR	−0.32	0.19	−1.67	n.s.	0.72
	High-decreasing	1.14	0.71	1.60	n.s.	3.13
Attachment Security to Mother	Low-AAR	−1.38	0.99	−1.40	n.s.	0.25
	Mid-stable	−1.58	1.24	−1.27	n.s.	0.21
	Mid-AAR	−2.17	1.57	−1.39	n.s.	0.11
	High-decreasing	−23.24	10.2	−2.43	0.023	0.00
Fathers’ Parental Efficacya	Low-AAR	0.75	0.49	1.52	n.s.	2.12
	Mid-stable	2.49	0.70	3.59	<0.001	12.11
	Mid-AAR	0.65	0.79	0.83	n.s.	1.91
	High-decreasing	7.09	2.92	2.43	0.015	>1000

Note: n.s. = nonsignificant. OR= Odds Ratio.

^ahigher scores indicate lower sense of efficacy as a parent. AAR = Adjustment and Adaptation Response

Because of our interest in discerning differences between classes hitting different levels of clinical significance, we also conducted exploratory multinomial logistic regressions using several of the other classes as the reference class. Using the high-decreasing class as the reference class, in which children started near the clinical or above the clinical range on aggression before the birth, we found that children’s negative emotionality, attachment security to mother, undermining coparenting during triadic observations, and fathers’ parental self-efficacy significantly predicted class membership in the high-decreasing compared to the mid-AAR-decreasing class. Specifically, children in the high decreasing class were more negatively emotional, b = −5.34, z = −2.463, OR = .005, had parents engaged in more undermining coparenting, b = −1.47, z = −2.014, OR = .231, had fathers with low parental efficacy, b −6.44, z = −2.172, OR = 0.002, and had less secure attachments to their mothers, b = 21.06, z = 2.06, OR > 1000, than children in the mid-AAR-decreasing class, Actually, children in the high-decreasing class were higher on negative emotionality (all z’s less than −1.96) and had less secure attachments to their mothers than any of the four comparison classes (all z’s greater than 1.96). Further, fathers of children in the high-decreasing class felt less effective as a parent than fathers of children in the low-AAR class, b = −6.34, z = −2.178, OR = .002.

We were also interested in the comparisons between the mid-stable and mid-AAR-decreasing classes because these two classes were similar at the prenatal time point, but evinced different patterns of change, with one showing an AAR and the other remaining stable over time. Using the mid-stable class as the reference class, father’s parental efficacy predicted the likelihood difference of being in the mid-AAR-decreasing class, b = −1.85, z = −2.110, OR =.158, indicating that fathers who felt more ineffective as a parent were more likely to have children in the mid-stable class. In addition, father’s parental efficacy, b = −1.74, z =−2.60, OR =.175, children’s negative emotionality, b = −1.49, z =−3.15, OR =.225, and observed undermining coparenting, b = .294, z = 2.05, OR = 1.34, predicted the low-AAR class relative to the mid-stable class, with greater feelings of ineffective fathering, more negatively emotional children, and less undermining coparenting associated with greater probability of membership in the mid-stable class compared to the low-AAR class. Finally, we were interested in the mid-AAR-decreasing and low-AAR classes. Both showed an adjustment and adaptation response, but they differed both with respect to their intercepts (i.e., prenatal scores) and the extent of the response with one showing a modest increase, and the other showing an increase hitting clinical levels. Only the observations of undermining coparenting predicted the difference between class membership, b = −.432, z = −2.22, OR = .649 (using low-AAR as reference), with parents in the low-AAR more likely to use undermining coparenting during home observations than parents in the mid-AAR-decreasing class.

Consequences of Children’s Aggression for Sibling Relationships at One Year

To determine whether children’s aggression trajectories predicted the quality of the sibling relationship at 12 months after the birth, we conducted a series of regression analyses across the three dimensions of sibling relationship quality (i.e., positive engagement, conflict, and avoidance) using the low-increasing class as a reference group because our predictor, the aggression trajectory classes, was categorical. Results revealed that at 12 months, parents of children in the mid-stable (est.= .243, SE= .108, p= .026) and mid-AAR-decreasing classes (est.= .382, SE= .143, p = .008) reported significantly more sibling conflict at 12 months than the low-increasing class. The mid-stable group was also significantly less likely to be positively engaged with their younger siblings at 12 months (est. = −.213, SE= .104, p= .041) than the low-increasing class.

Discussion of Children’s Aggression Trajectories

In this chapter, we examined the overall pattern of children’s aggression after the birth of a sibling, as well as subgroups of trajectories that captured individual differences within the overall sample. Additionally, we explored the predictors of these trajectories and tested whether distinct trajectory subtypes were associated with differences in sibling relationship quality when the younger sibling was one year of age.

Results examining the overall pattern of aggressive behavior across the sample indicated that the adjustment and adaptation response, with an initial increase in aggression from prenatal to 1 month and a subsequent decline in aggression from 1 to 4 months, best described children’s aggressive behavior after the birth of an infant sibling. Thus, firstborn children, in general, do appear to show an initial increase in aggression after their infant sibling’s birth, but this increase is short-lived and children return to pre-birth levels by 4 months. Testing the AAR contrast in our statistical models provides solid evidence for what we are referring to as an adjustment and adaptation response after a normative stressful life event (the birth of an infant sibling) that underscores the resilience of many children, and stands in contrast to a pattern of sudden and persistent change indicative of maladjustment that would have been captured by the quadratic effect. Thus, our results provide far more support for resilience in the face of stress and change for children undergoing the transition to siblinghood than any evidence of psychosocial trauma or a developmental crisis underscored in some of the earliest psychodynamic accounts of children’s reactions to the birth of a sibling (Levy, 1937; Winnicott, 1964) in which children supposedly would resort to any means necessary to recapture the lost attention and love of their parents.

Despite evidence of an adjustment and adaptation response to their sibling’s birth, heterogeneity in children’s aggression trajectories was further evident, and we were able to identify five trajectory patterns that described children in the present sample; several in line with current studies looking at trajectories of aggression and externalizing behavior in the toddler and preschool period. The largest class of children (40%) was low on aggression initially at the pre-birth time point and gradually increased (linear) in their aggression over the year following the sibling’s birth. This pattern may reflect a normative developmental trajectory in that children are known to increase over the toddler years in their aggressive behavior before aggression declines around 3 to 4 years of age. Children in the current study were 31 months of age, on average, so increases in aggressive behavior throughout the year following the birth due to normative age changes would not be completely consistent with extant findings.

In contrast, these gradual increases in aggressive behavior over the months after the birth may coincide with normative developmental changes in the infant during the first year after birth. Maturation in infant mobility with advances in exploratory play and social interest in others as the year progresses means greater social contact between older and younger siblings, yet infants are also limited in their social abilities. These sorts of encounters of invading on the older siblings’ play space, while grabbing toys and destroying carefully planned fantasy worlds may often be upsetting to older siblings who themselves are still relatively unsophisticated in their own abilities to resolve conflict. Dunn & Munn (1986) provided detailed descriptions of the conflict between young siblings of this age and the frequent number of conflicts in which parents must intervene over the course of a day (approximately 7 per hour). This interpretation of increasing aggression coinciding with developmental changes in the infant is substantiated by our findings showing that the different aggression trajectories predicted the quality of children’s relationships with their younger sibling one year after the birth. Children in the mid-stable class exhibited more conflict and were less positively engaged with their infant siblings than children in the low-increasing class. The largest class of children showing low, but linearly increasing aggression over the year following the birth of their younger siblings, appears to describe most children in this sample. Thus, a modest linear increase in children’s aggressive behavior may be expected during the year following the birth of a second child.

Because of our interest in identifying the effects of the transition on children’s behavioral adjustment and whether these effects were short- or long-lived, we tested different trajectory patterns indicative of no change and linear change (increases and decreases), as well as more specific patterns we hypothesized to reflect different patterns of adjustment and adaptation, one in which children would exhibit initial behavioral disruption in response to the stress of the transition, but adapt over time to the changing family circumstances and return to pre-birth levels of behavioral functioning, and a second, more indicative of behavioral maladjustment or difficulties with the transition which would reflect sudden increases in problematic behavior in response to the transition that persisted over time (quadratic model) and did not return to pre-birth levels. The quadratic model would also allow testing of other possible developmental trajectories that might coincide with changes in the family and the infant’s developmental milestones over time such as the gradual impact and decline model wherein children increased but returned gradually back to pre-birth levels over the year, or the delayed impact model wherein children show little to no change immediately after the birth, but showed significantly greater increases by 8 and 12 months when infant mobility and social intrusions become more frequent (Stewart, 1990). The unconditional latent growth curve model indicated that the model testing the adjustment and adaptation response was a better fit to our data than either the linear or quadratic models suggesting that, on average, there was evidence of initial increases in children’s aggression shortly after the birth that subsided by 4 months of age. This model was a better fit to the longitudinal patterns than a maladaptation or developmental crisis model, the delayed impact model, or the gradual impact and decline model. These findings are consistent with Dunn & Kendrick’s (1982) work reporting that changes in children’s problematic behavior after the sibling’s birth were short-lived and most children appeared to adjust to the birth by 8 months of age, although our data suggest that most children have adjusted already by 4 months after the birth.

Regardless of this general pattern, though, there was evidence of heterogeneity in children’s aggression trajectories. There was a very small group of children (N = 8) who evinced very high levels (clinical and borderline-clinical cut-off range) of aggression even before the birth and actually showed significant declines in their aggression over time. This pattern may reflect the normative developmental decreases in aggression often reported in the literature, but based on our prediction analysis examining the most discriminating prenatal predictors, we offer another possible interpretation focusing on family processes interacting with children’s temperament. These 8 children differed from the low-increasing class in that they were higher in negative emotionality, had lower attachment security scores to their mothers, had fathers who felt more ineffective in their abilities to parent and discipline the older sibling, and parents engaged in more undermining coparenting during prenatal home observations. Once the infant was born and mothers were busy with the care of a newborn infant, these insecurely-attached children’s aggression actually decreased over time, suggesting the possibility that children’s behavioral adjustment may have far more to do with family dynamics and changes in the family, brought about by the birth of the infant, than the actual infant per se. One possible explanation is that children fare better and engage in less aggressive behavior once their mothers’ attention is redirected to another child, and the older child is no longer the recipient of the insensitive or intrusive parenting that gave rise to the insecure attachment to their mothers. It is also possible that these children expressed their aggressive tendencies and emotional distress through other means, perhaps becoming more emotionally reactive once the infant was born. Of these 8 aggressive children, 5 were in the mid-increasing group of emotionally reactive children to be described later, suggesting that decreases in aggressive behavior were accompanied by increases in emotional distress over the year. We do not know if these children may have turned to their fathers during this time for security and comfort and whether a secure father-child attachment may account in some ways for this decrease in aggressive behavior. It should be noted, however, that fathers of these children reported feeling less efficacious in their parenting skills so even if children did turn to their fathers for support, it is not clear that their fathers were able to provide the emotional comfort and security needed to ameliorate their distress.

A similar picture emerged for some of the other trajectory patterns wherein children’s negative emotionality and family relationship functioning before the sibling’s birth played a large part in predicting class membership. For instance, most children in the three classes that fell within the borderline clinical and clinical cut-offs of the aggression scale, the high-decreasing, mid-stable, and mid-AAR-decreasing classes, were higher on negative emotionality than children in the low-increasing class. Children in the mid-stable class also had fathers who felt less effective in their discipline than those in the low-increasing class, yet had higher attachment security scores with their mothers than children in the riskier, high-decreasing class.

We were particularly interested in comparisons between groups where children were comparable at the prenatal visit, but showed a different pattern of change after the birth, specifically comparisons of the mid-stable group with the mid-AAR-decreasing group, where children had a large increase (some hitting clinical levels) in their aggression after the birth that subsequently declined by 4 months, and comparison of the low-increasing with the low-AAR group, where there was a more modest increase in aggression from prenatal to 1 month that eventually declined by 4 months. Children in the mid-AAR-decreasing group were distinguished from the mid-stable class in that their fathers reported feeling more efficacious in their parenting abilities than those in the mid-stable class. Perhaps the more efficacious fathers of the mid-AAR-decreasing class were able to effectively respond and discipline their children in response to aggressive behavior that allowed children to return to pre-birth levels of functioning during a particularly stressful developmental transition.

Most children in the low-increasing and low-AAR classes were low on negative emotionality, yet children in the low-AAR class did have higher scores on negative emotionality than children in the low-increasing class, suggesting that these children may have been more temperamentally reactive to changes in the family after the sibling’s birth. It is interesting that children in the low-AAR families had parents who engaged in more undermining coparenting during prenatal home observations than parents in both the mid-stable and mid-AAR-decreasing classes underscoring that mothers and fathers in these families had difficulties working together as coparents and this may have played some role in why the low-AAR children had the significant increase in aggressive behaviors that they did. The low-AAR class represented 30% of the sample and was the second largest class after the larger, low-increasing class. Some have argued that the father’s involvement during this transition is critical for children’s adjustment (Kreppner, Paulsen, & Schuetze, 1982; Stewart, 1990), and the current findings suggest that children may experience little to no disruption after the birth when their fathers are actively involved in the family either because of their active role as a coparent or their feelings of competence as an effective father. No study before ours has examined fathers’ feelings of parental efficacy or coparenting in predicting children’s adjustment after the birth of a sibling, but our findings are consistent with earlier research underscoring the strong connections between marital quality, father-child relationships, and children’s behavioral adjustment (Cummings, Merrilees, & George, 2010; Jouriles & Farris, 1992; Stevenson et al., 2014; Davies, Sturge-Apple, Woitach, & Cummings, 2009; Booth & Amato, 1994; Schacht, Cummings, & Davies, 2009; Goeke-Morey & Cummings, 2007).

Two of the five classes revealed a period of adjustment (i.e., the initial response to a potential stressor) and adaptation (i.e., a period in which the child reorganizes and resumes family life much as before the stressful life event). We believe our research program is the first to explicitly test for an adjustment and adaptation response after a normative life event using latent growth curve and growth mixture modeling. The timing of our assessments was critical in capturing this response, given that children had already returned to pre-birth levels by 4 months after birth. Had our first post-birth assessment not occurred at 1 month so soon after the birth, we would have missed the adjustment and adaptation response for children’s aggression and most likely concluded that children’s aggression remained stable, when in fact the results clearly suggested a perturbation in the family system that eventually resolved itself. Because our analysis strategy was guided by our desire to identify prenatal predictors that would be useful targets for pre-birth intervention, we were unable to delineate what family processes after the birth may be responsible for these different change patterns. Future reports will be able to provide more fine-grained analyses addressing the family processes and developmental cascades between child, parent, and family context over time.

Overall, the results examining children’s aggression after the birth of a second child indicate that for most children, parents can expect a gradual linear increase in the firstborn children’s aggression that more than likely coincides with developmental changes in the infant’s growing capacity to socially engage others, both negatively and positively. There is also evidence, however, that there are significant individual differences in how children react to the birth of their infant sibling and that family processes occurring before the birth forecast these individual differences in aggression trajectories over the first year. Temperamentally difficult children with insecure attachments to their mothers in families where fathers felt ineffective in disciplining the firstborn, with more undermining coparenting observed during home observations were at greatest risk for being aggressive before and staying aggressive after the birth of their infant sibling, or evincing a period of initial disruption that eventually subsided by 4 months. Most children appeared to be resilient and returned to pre-birth levels by 4 months after the birth. Intervention efforts targeting the mother-child relationship, fathers’ effectiveness in managing the older child’s behaviors, and the coparenting relationship before the birth may assist these children and their families with the transition. These results are consistent with prior research finding that preschool children with externalizing or aggressive behavior problems often have difficult temperaments, have insecure attachments to their mothers and fathers, and have parents engaged in more coparenting conflict (Kochanska & Kim, 2012; Fearon et al., 2010; Gartstein et al., 2012; Fearon & Belsky, 2011; Rothbart & Bates, 1998). In sum, these results suggest that children’s aggressive behaviors before and after the birth of a sibling may be exacerbated by the stresses and changes in family life surrounding the birth, but many of these changes are short-lived and children are resilient.