Abstract
Adolescent bullying perpetration has previously been associated with dating benefits and costs in the short- and long-term, yet it is unclear how early temperament traits facilitate these associations. Therefore, the developmental pathways from temperament in early adolescence to bullying perpetration in middle adolescence and to dating outcomes in late adolescence were examined. Participants included 463 individuals who completed self-report measures on temperament traits at age 12, bullying perpetration at age 14, dating outcomes at age 19, and were in a romantic relationship at age 19. Findings from a path analysis revealed that an early adolescent temperament trait reflecting difficulty with self-regulation (i.e., lower inhibitory control) was associated with middle adolescent bullying perpetration and bullying perpetration was associated with late adolescent dating benefits (i.e., more dating partners) and costs (i.e., higher dating aggression perpetration). Lower inhibitory control also had significant indirect associations to the late adolescent dating outcomes through middle adolescent bullying perpetration. Findings suggest that although a temperament trait can facilitate adaptive dating outcomes through bullying, it can also come at a cost for romantic relationships. Results highlight the importance of early tailoring of bullying interventions to the self-regulatory difficulties of youth to prevent adverse long-term outcomes and to also recognize the challenges of developing interventions for behavior that can result in benefits.
Bullying in adolescent peer relationships may set the precedent for behavior in later romantic relationships. Given that up to 30% of youth report bullying within their peer relationships (e.g., Espelage & Holt, 2007; Hymel & Swearer, 2015; Vaillancourt, Brittain, et al., 2010), a significant portion of youth may be on a path to adverse romantic relationships. From an evolutionary framework, bullying is a form of aggression used within a power imbalance to adaptively harm peers and acquire resources (Hawley, 1999; Vaillancourt et al., 2008; Volk, Dane, & Marini, 2014). For example, bullying has been associated with a higher number of dating and/or sexual partners (e.g., Connolly, Pepler, Craig, & Taradash, 2000; Volk, Dane, Marini, & Vaillancourt, 2015). Despite these romantic benefits, bullying is not without its costs.
Bullying has been related to dating aggression, which can include direct (e.g., physical, verbal, and sexual attacks) and indirect (e.g., emotional manipulation) harm against a dating partner (Connolly et al., 2000; Pepler et al., 2006). There is also evidence for both bullying (e.g., Terranova, Morris, & Boxer, 2008) and dating aggression (e.g., Sullivan, Garthe, Goncy, Carlson, & Behrhorst, 2017) being rooted in similar temperament traits, such as difficulties with self-regulation. These findings support Pepler et al.’s (2006) developmental perspective of bullying (i.e., bullying is a relationship problem that generalizes to other forms of relationship aggression like dating violence). What is not clear is how temperament relates longitudinally to bullying perpetration and dating outcomes in one unified model. Accordingly, the potential associations between early adolescent temperament, middle adolescent bullying, and late adolescent dating outcomes were examined in this study.
Bullying From a Developmental Evolutionary Framework
Adolescence is a developmental period marked by biological changes including the onset of puberty and reproductive capability (Forbes & Dahl, 2010). These biological events can facilitate psychosocial changes like heightened sensation seeking (Steinberg, 2008) and social changes like the expansion of peer networks (Harris, 1995). From an evolutionary developmental framework (e.g., Bjorklund & Ellis, 2014), one reason that bullying may peak during adolescence is because this behavior can be used to obtain benefits (Volk, Camilleri, Dane, & Marini, 2012). As with any adaptive behavior, there is evidence from behavioral genetic research that bullying perpetration can be partially explained by genetic factors (e.g., Ball et al., 2008). Individual differences in temperament may be one of these factors with genetic foundations. Temperament includes tendencies for self-regulation and reactivity in emotions and behavior that are evident early in development and remain moderately stable throughout the life span (Rothbart & Bates, 2006). Temperament traits can influence how youth initiate or inhibit their behavior in social environments (A. H. Buss & Plomin, 1984; Rothbart, Ahadi, & Evans, 2000). Bullying perpetration has been associated with a higher activity level (e.g., Marini, Dane, Bosacki, & YLC-CURA, 2006) and difficulties with self-regulation (e.g., Terranova et al., 2008). Poorer self-regulation has also been associated with dating violence (e.g., Sullivan et al., 2017). From an evolutionary perspective, these traits may have been passed onto future generations because they can increase an individual’s likelihood of survival (D. M. Buss, 2011).
Temperament traits may fit naturally with the broader development of adolescent bullying as a means to compete for dating outcomes. An individual may consider their own temperament traits within a given situation or context to evaluate the costs and benefits of using bullying (Dane, Marini, Volk, & Vaillancourt, 2017). Alternatively, it is possible that individuals who possess particular temperamental traits may be more likely to utilize bullying as opposed to other strategies to obtain desired goals (Dane et al., 2017; Del Giudice & Belsky, 2011). Consequently, bullying may partly be a behavioral expression of temperament as a means to acquire adaptive resources.
Researchers have proposed that bullying may be a form of adaptive behavior used to compete for resources such as social status, dominance, and power, as well as reproductively relevant resources such as dating and sexual partners (e.g., Volk et al., 2012, 2014). This elevated social status and reputation may help with intersexual selection (i.e., displaying qualities that would increase the chances of being chosen as a romantic partner by the opposite sex) through self-promotion (Archer, 2009; D. M. Buss, 1988, 2011; Vaillancourt & Krems, 2018; Volk et al., 2015). Demonstrating characteristics like social status, boldness, and the ability to provide resources to romantic partners may be attractive to prospective partners. Bullying may also help with intrasexual competition (i.e., “outcompeting” same-sex rivals as a preferred romantic partner) through derogation of rivals’ status and reputation (Benenson, 2009; Campbell, 2013; Vaillancourt, 2013; Vaillancourt & Krems, 2018; White, Gallup, & Gallup, 2010). This latter strategy has been used commonly by women via indirect forms of bullying and aggression. Taken together, adolescents may use bullying as an adaptive tool to gain benefits and the likelihood of using bullying may be associated with particular temperament traits.
The Dating Benefits and Costs of Bullying
Indeed, researchers have found that adolescents who bully their peers started dating earlier and had more advanced dyadic dating (i.e., dating a partner one on one as opposed to going on group dates; Connolly et al., 2000) and sexual experience (Volk et al., 2015). In two longitudinal studies, indirect aggression (e.g., rumor spreading, peer group exclusion) predicted dating status 1 year later (Arnocky & Vaillancourt, 2012; Lee, Brittain, & Vaillancourt, 2018). Researchers have attributed these dating benefits to the elevated status of individuals who use bullying. Often adolescents who bully are high in social power, dominance, and popularity (e.g., Reijntjes et al., 2013; Vaillancourt, Hymel, & McDougall, 2003; Veenstra, Lindenberg, Munniksma, & Dijkstra, 2010). The social status associated with bullying may facilitate intersexual selection by members of the opposite sex and outcompete same-sex rivals during intrasexual competition. Despite the dating benefits of bullying, there may also be some dating costs.
Adolescents who use bullying appear to be at a greater risk of dating aggression, which is affecting a large number of teens with 25% of girls and 13% of boys reporting physical dating violence perpetration and 10% of boys and 3% of girls reporting sexual dating violence perpetration (Wincentak, Connolly, & Card, 2017). In cross-sectional studies, bullying was associated with physical and indirect dating aggression (Connolly et al., 2000; Pepler et al., 2006). Additional concurrent (Niolon et al., 2015) and longitudinal (Cutbush, Williams, & Miller, 2016; Foshee et al., 2014; Miller et al., 2013; Walters & Espelage, 2018) links were found between bullying and dating violence. Both bullying and dating aggression may share a core need to control others, including peers and dating partners. As new forms of social relationships develop into late adolescence and early adulthood, previous experiences with power over peers may transform into attempts for power over dating partners (Foshee et al., 2014; Moffitt, 1993; Pepler et al., 2006). The need to control others in both bullying and dating may stem from similar temperamental risk factors.
Temperament and Bullying
Temperament traits can be viewed as basic individual differences with neurobiological foundations (e.g., Nigg, 2006; Whittle, Allen, Lubman, & Yücel, 2006). There is a general agreement among temperament researchers that there are three to four core temperament traits measured in children and adolescents. Two traits capture abilities to regulate behavior and emotions. For example, one trait includes voluntary abilities for inhibitory control, or stopping undesired behavior in order to engage in alternative forms of behavior (i.e., inhibitory control; Capaldi & Rothbart, 1992; L. K. Ellis & Rothbart, 2001). A second trait reflects abilities to regulate negative emotions such as frustration, irritability, or anger (i.e., frustration).
The other two temperament traits reflect positive and social tendencies. For example, one trait reflects sociability, low shyness, and approach for social and novel stimuli and activity (i.e., sociability; L. K. Ellis & Rothbart, 2001; Rothbart & Bates, 2006). There is evidence for the development of a fourth trait during adolescence that captures affiliativeness (e.g., Evans & Rothbart, 2007; Rothbart & Bates, 2006) but is not included in all temperament measures (e.g., see Shiner & De Young, 2013). Therefore, we chose to focus on the three traits most often captured in multiple temperament models: inhibitory control, frustration, and sociability.
Although several researchers have studied broader personality traits with bullying perpetration (e.g., Book, Volk, & Hosker, 2012), only a few researchers have examined temperament traits with bullying. Temperament traits may be important to examine as they may reveal more basic, biological activation and inhibitory processes that underlie personality traits and bullying. In the few existing studies, researchers have found support for the association between bullying and temperament traits that reflect higher frustration and difficulty with inhibitory control. For example, adolescents who engaged in bullying had higher concurrent anger-related coping biases in comparison to peers uninvolved in bullying (Marini et al., 2006). Difficulties regulating these negative emotions and behavior were also associated with bullying. Terranova, Morris, and Boxer (2008) found that lower effortful control (i.e., inhibitory control) in middle school predicted higher overt bullying 6 months later. In another study, lower inhibitory control was indirectly associated with higher bullying among middle school students (Bacchini, Affuso, & Trotta, 2008). It seems that youth who bully may be more prone to negative emotions and have difficulty regulating these emotions and behavior. In addition to these negative traits, bullying perpetration may be associated with some positive traits.
Higher sociability or approach for novel stimuli is often associated with lower levels of shyness and fear (Rothbart & Bates, 2006). Previously, bullying has been correlated with higher activity level (Marini et al., 2006; Pellegrini, Bartini, & Brooks, 1999). The tendency to be active may motivate youth to be more socially dominant and bold via bullying to obtain resources like dating partners. If the attainment of these resources is threatened, it may be challenging for youth with self-regulation difficulties to inhibit frustration and the tendency to behave aggressively. This high approach motivation and potentially high social dominance may reduce the fear of the potential negative outcomes of bullying (Marini et al., 2006; Pellegrini et al., 1999; Terranova et al., 2008). Similar to bullying, dating outcomes may be associated with these temperament traits.
Temperament and Dating Outcomes
A fundamental feature of bullying is having power over others and as adolescents get older, they may engage in dating aggression as a means to control a romantic partner (Follingstad, Bradley, Helff, & Laughlin, 2002; Foshee et al., 2014). When this goal to control is interrupted, tendencies to be easily frustrated may facilitate dating aggression. Difficulty inhibiting this frustration may also facilitate dating aggression (Chapple & Hope, 2003; Walters & Espelage, 2018). In a longitudinal study with middle school students, higher anger dysregulation and lower inhibition were both associated with dating aggression (Sullivan et al., 2017). In another study focused on adult men, impulsivity in childhood and difficulty with concentration in adolescence were significant predictors of adult dating violence (Theobald & Farrington, 2012). However, similar to bullying, dating outcomes do not necessarily have to be preceded by negative traits.
Independent of bullying and aggressive behavior, sociable individuals who are low in negative emotions and high in self-regulation may be desirable dating partners. In one study, higher intensity pleasure (i.e., sociability) had a positive association with having had dating experience (Ivanova, Veenstra, & Mills, 2012). These findings are in line with the broader personality literature on mate preference, with adults preferring partners higher in extraversion and lower in negative affect (e.g., Botwin, Buss, & Shackelford, 1997). Youth who are less socially inhibited and more inclined to approach peers may be more likely to have friendships in middle adolescence and dating partners in late adolescence. However, difficulties with negative emotions like frustration may make individuals more prone to bullying and aggression. Therefore, we were interested in examining the potential differential developmental paths from temperament to bullying and dating outcomes.
Current Study
From a developmental framework, bullying may be a form of relationship aggression that is developmentally relevant in middle adolescence and may be a precursor to other forms of aggression that are more relevant in later adolescence, such as dating aggression. Both forms of relationship aggression may be rooted in particular temperament traits evident in early adolescence. Therefore, the developmental pathways from early adolescent temperament to middle adolescent bullying perpetration and late adolescent dating outcomes were examined. For temperament, we examined frustration and inhibitory control to capture emotional and behavioral regulation and examined sociability to capture social approach and motivation. Given that bullying may be associated with benefits and costs, for dating outcomes, we measured the number of dating partners as a dating benefit and dating aggression perpetration as a dating cost.
We also chose to assess temperament, bullying, and dating outcomes during developmental periods that may be salient to these constructs. We chose to examine temperament during early adolescence (i.e., Grade 6; age 12) as temperament is often stable across the life span and can have long-term social implications (e.g., Roberts & DelVecchio, 2000). We chose to examine bullying during middle adolescence (i.e., Grade 8; age 14) as this is the developmental period when bullying often peaks. In a population-based study of Canadian youth, rates of overall and multiple forms of bullying perpetration were highest in Grade 8 (Vaillancourt, Trinh, et al., 2010). Grade 8 is also the final elementary school year in most Canadian provinces including Ontario where our sample was drawn. We also chose Grade 8 bullying because we did not want to contaminate results with the transition to high school which has shown to be impact bullying rates (Wang, Brittain, McDougall, & Vaillancourt, 2015). Finally, we examined dating behavior in late adolescence (i.e., post high school; age 19) as researchers have found that approximately 80% of youth reported dating experience by age 18 (e.g., Carver, Joyner, & Udry, 2003).
Given previous findings with poorer self-regulation and higher social approach, we predicted higher frustration, lower inhibitory control, and higher sociability would be associated with bullying perpetration. Considering that aggression in peer relationships may transpire into additional forms of relationship aggression that are developmentally relevant, we also predicted bullying during middle adolescence would be associated with higher dating aggression perpetration in late adolescence. However, since bullying can be used adaptively for resources, we predicted bullying during middle adolescence would be associated with a higher number of dating partners in late adolescence. We also predicted that these three temperament traits would have indirect associations with the dating outcomes through bullying perpetration. Finally, we expected some direct associations between temperament and dating, specifically higher sociability for dating partners and higher frustration and lower inhibitory control for dating aggression.
Method
Participants
Data were drawn from the McMaster Teen Study, an ongoing longitudinal study on bullying, mental health, and academic achievement. Following school board approval, in the spring of 2008, 51 primary schools that were randomly selected from a Southern Ontario school district participated in the study. At Time 1, a yearly household income estimated by parents had a median of $70,000–$80,000 (CAN), which was similar to the city of recruitment ($76,222) and province ($70,910; http://statscan.gc.ca). Participants were in Grade 5 at Time 1 and have been followed annually, with data collection ongoing (see Vaillancourt, et al, 2013 for a more detailed description of recruitment). For the longitudinal study, 875 individuals agreed to participate and 703 (80%) actually participated (mean age of 10.91 years at Time 1 [SD = 0.36]). To be included in this study, participants needed to be in a romantic relationship at the final time point (late adolescence) as the dating aggression items referred to perpetration against a dating partner in the last 12 months. Participants who indicated that the dating aggression items were not applicable as a result of not being in a relationship (n = 108) were excluded from the study. In addition, participants needed to have data in at least two of the following three time points examined in this study: Time 2 (Grade 6; age 12), Time 4 (Grade 8; age 14), or Time 9 (post high school; age 19). These three time points were selected for this study as they correspond with the developmental periods reflecting early adolescence, middle adolescence, and late adolescence in Ontario, Canada and data for these variables were available at these time points. The final analytic sample included 463 individuals (66% of participating cohort) with 208 boys and 255 girls. Participants in the analytic sample had a mean age of 12.19 in Grade 6 (SD = 0.42). Participants were also primarily White (Caucasian; 76.9%), had a median income of more than $80,000, and a median parental education level of a completed college diploma or trades certificate.
Procedure
Parents of participants were asked to provide consent every year and participants were also asked to provide assent. Starting at Time 6 (i.e., age 16), participants provided their own consent. Each year, measures were completed either online or on paper and pencil via prepaid mail (see Vaillancourt et al., 2013 for a complete description of the procedure). Participants received gift cards with increasing increments each year as compensation for participating. The study received ethics clearance from the school board and the associated university ethics boards.
Measures
Dating aggression
Late adolescent aggression 1 year post high school (i.e., Time 9) was measured using the 10-item Conflict in Adolescent Dating Relationships Inventory–Short Form (Fernández-González, Wekerle, & Goldstein, 2012), which has previously been used and validated in both high school and undergraduate samples. Participants responded to how often each event occurred with a dating partner in the last 12 months. Each item assessed different forms of dating aggression including verbal/emotional (e.g., “I spoke to my partner in a hostile or mean tone of voice”), physical (e.g., “I kicked, hit, or punched my partner”), threatening (e.g., “I threatened to hurt my partner”), relational (e.g., “I spread rumors about my partner”), and sexual (e.g., “I touched my partner sexually when they didn’t want me to”). Items were rated on a 4-point Likert-type scale (0 = never to 3 = often). Items were averaged to create a composite, with higher scores indicating higher dating aggression perpetration. The α was .82.
Dating partners
The number of dating partners 1 year post high school in late adolescence (i.e., Time 9) was assessed by asking an open-ended numerical question to participants, “How many different boyfriends/girlfriends have you had in your lifetime (including your current partner)?” Higher values indicated a higher number of dating partners.
Bullying
Participants were provided with the following definition of bullying: “There are lots of different ways to bully someone but a bully wants to hurt the other person (it’s not an accident), and does so repeatedly and unfairly (the bully has some advantage over the victim). Sometimes a group of students will bully a student. It is not bullying when two students of the same strength quarrel or fight.” Participants completed 5 items on bullying perpetration in Grade 6 (i.e., early adolescence; Time 2) and Grade 8 (i.e., middle adolescence; Time 4), with each item assessing a different form of bullying (Vaillancourt, Brittain, et al., 2010; Vaillancourt, Trinh et al., 2010). The first item assessed general bullying (i.e., “Since the start of the school year [September], how often have you taken part in bullying another student?”). The remaining 4 items each measured a different form of bullying including physical (e.g., hitting, shoving, kicking, spitting or beating up others), verbal (e.g., name-calling, mocking, hurtful teasing, verbally threatening), social (e.g., excluding others from groups, gossiping or spreading rumors about others), and cyber (e.g., using computer or e-mail messages or pictures to hurt someone’s feelings). Items were rated on a 5-point Likert-type scale (0 = not at all to 4 = many times a week). Items were averaged to create a composite, with higher scores indicating higher bullying perpetration. The αs were .72 in Grade 6 and .78 in Grade 8.
Temperament
Early adolescent temperament traits were assessed in Grade 6 (i.e., Time 2). Traits reflecting self-regulation were assessed using the frustration and inhibitory control subscales of the Early Adolescent Temperament Questionnaire-Revised (EATQ-R) self-report (Capaldi & Rothbart, 1992; L. K. Ellis & Rothbart, 2001). Seven items were used to assess frustration to reflect difficulty regulating negative emotions (e.g., “It bothers me when I try to make a phone call and the line is busy”), and 5 items were used to assess inhibitory control to reflect difficulty regulating behavior (e.g., “When someone tells me to stop doing something, it is easy for me to stop”). Items were rated on a 5-point Likert-type scale (0 = very false to 4 = very true). A trait reflecting approach for social and novel stimuli was assessed using an adapted version of the Revised Cheek and Buss Shyness and Sociability Scale (Cheek, 1983; Cheek & Buss, 1981). Five items were used to assess sociability (e.g., “I like to be with new people”) and were rated on a 5-point Likert-type scale (0 = not at all like me to 4 = extremely like me). Items for each trait were averaged to create a composite, with higher scores indicating higher frustration, inhibitory control, and sociability. The αs were .79 for frustration, .57 for inhibitory control, and .62 for sociability.
Demographic and control variables
Demographic variables were collected at Time 1 (i.e., Grade 5) including biological sex, race/ethnicity, household income, and parental education. Due to the small size of multiple racial backgrounds, race/ethnicity was recoded using child and parent reports as White (85%) and non-White (15%). Parents reported household income using an 8-point scale increasing in increments of $10,000 (e.g., 1 = <$19,999; 2 = $20,000–$29,999; 3 = $30,000–$39,999; 4 = $40,000–$49,999; 5 = $50,000–$59,999; 6 = $60,000–$69,999; 7 = $70,000–$79,999; 8 = >$80,000). Parents reported their highest education level using a 5-point scale (e.g., 1 = did not complete high school; 2 = high school; 3 = college diploma or trades certificate; 4 = university undergraduate degree; 5 = university graduate degree). These demographic variables were used as control variables in the primary analyses.
Plan of Data Analysis
Mplus Version 7.4 (Muthén & Muthén, 1998–2017) was used for all primary analyses. To allow for using all cases with at least one data point, full information maximum likelihood was used to handle missing data (Schafer & Graham, 2002). The primary interests of this analysis were the cross-lagged pathways from Grade 6 temperament to Grade 8 bullying perpetration and the pathways from Grade 8 bullying perpetration to post high school dating aggression and dating partners. Therefore, we estimated all cross-lagged paths between study variables at adjacent time points (e.g., Grade 6 sociability to Grade 8 bullying; Grade 8 bullying to post high school dating aggression). We also controlled for within-time covariance terms (e.g., Grade 6 sociability with Grade 6 inhibitory control) and for baseline Grade 6 bullying perpetration to Grade 8 bullying.
The four demographic variables were also controlled for by estimating the covariances among these variables with one another, with the three temperament traits, and with Grade 6 bullying perpetration. The paths from the control variables to Grade 8 bullying and the two post high school dating outcomes were also estimated. We controlled for sex differences as meta-analytic findings support that boys are higher in bullying perpetration than girls (e.g., Cook, Williams, Guerra, Kim, & Sadek, 2010). Also, mean levels of overall dating violence have been found to be similar between boys and girls in some studies (e.g., Connolly et al., 2000) and higher among girls in other studies (e.g., Walters & Espelage, 2018). Other researchers have found sex differences in particular forms of perpetration (e.g., Niolon et al., 2015; Walters & Espelage, 2018; Wincentak et al., 2017). Being a boy has also been associated with a higher number of dating partners (e.g., Volk et al., 2015). We controlled for indicators of socioeconomic status such as parental education and household income as parental education previously had negative associations with some forms of bullying (e.g., Foshee et al., 2014) and socioeconomic status had weak but significant negative associations with bullying perpetration (Tippett & Wolke, 2014). Ethnicity/race was also controlled for as there has been some evidence for race differences in some forms of dating aggression (e.g., Foshee, Reyes, & Ennett, 2010; Foshee et al., 2014).
Model fit was assessed using criteria of the comparative fit index (CFI) ≥ .95, root mean square error of approximation (RMSEA) ≤ .06, and standardized root mean square residual (SRMR) ≤ .08 (Browne & Cudeck, 1993; Hu & Bentler, 1999; Kline, 2011). For any significant cross-lagged pathways that included the three time periods (i.e., Grade 6, Grade 8, and post high school), tests of indirect effects were conducted. To estimate indirect effects, we used the MODEL INDIRECT command with percentile bootstrapping using 5,000 iterations to estimate 95% confidence intervals (CIs). Indirect effects were determined to be significant if the CIs did not cross over zero (Biesanz, Falk, & Savelei, 2010).
Results
Missing Data Analyses
To determine whether there were significant differences in the control and study variables between individuals in the analytic sample compared to individuals in the nonanalytic sample, a series of χ2 tests and t tests were conducted. In comparison to the nonanalytic sample, participants in the analytic sample were more likely to be White, were more likely to have a higher household income, and parents with higher levels of education (ps < .05). In comparison to the nonanalytic sample, participants in the analytic sample also had higher Grade 6 inhibitory control and a higher number of post high school dating partners (ps < .05). The χ2 tests and t tests were repeated to determine whether individuals in a romantic relationship at Time 9 significantly different from individuals who were excluded as a result of not being in a romantic relationship at Time 9. In comparison to individuals not in a relationship, individuals in a relationship were more likely to be White, higher in Grade 6 sociability, and have a higher number of post high school dating partners (ps < .05).
Descriptive Statistics and Bivariate Correlations
Descriptive statistics for all study variables are in Table 1 and bivariate correlations are presented in Table 2. All study variables demonstrated skewness and kurtosis values acceptable for path analyses (i.e., less than 10; Kline, 2011), with the exception of dating partners and dating aggression. These two variables along with bullying perpetration and inhibitory control had extreme univariate outliers. Winsorizing outliers preserved rank order and reduced the impact of these participants (Tabachnick & Fidell, 2013) and reduced skewness and kurtosis values of the dating variables to acceptable limits. Given the nonnormal distribution of bullying perpetration and dating outcomes at all time periods, maximum likelihood robust estimation was used for the path analysis. To determine whether study variables significantly differed by biological sex and race, a series of independent samples t tests were conducted. Girls were significantly higher than boys on post high school dating aggression. There were no other sex or race differences. Biological sex was also tested as a moderator, given the significant sex difference in t-test results. However, findings were nonsignificant and sex was kept as a control variable. Correlations were used to determine whether study variables were associated with household income and parental education. A lower household income was significantly correlated with higher levels of Grade 6 frustration, Grade 8 bullying, and post high school dating aggression (ps < .05). Lower parental education was significantly correlated with lower sociability, higher post high school number of dating partners, and higher post high school dating aggression perpetration (p < .05).
Descriptive Statistics for Study Variables.
*p < .05.
Bivariate Correlations Among Temperament, Bullying, and Dating Outcomes.
*p < .05. **p < .01.
Path Models of Temperament, Bullying, and Dating Outcomes
The model had adequate fit, χ2 = 7.861, df = 2, p =.020; CFI = 0.958; RMSEA = 0.080 (90% CI [.027, .141]); SRMR = 0.020 (see Figure 1 for model with significant standardized estimates). In Grade 6, higher frustration, lower inhibitory control, and lower sociability were all significantly correlated with one another. Lower inhibitory control and higher frustration were also significantly correlated with higher Grade 6 bullying perpetration. Grade 6 bullying perpetration had high stability with Grade 8 bullying perpetration. As expected, there were also several significant cross-lagged effects. Lower Grade 6 inhibitory control (b = −0.09, SE = .03, p = .003) significantly predicted higher Grade 8 bullying perpetration. Higher Grade 8 bullying perpetration significantly predicted higher post high school dating aggression perpetration (b = 0.09, SE = .04, p = .036) and number of dating partners (b = 1.21, SE = .47, p = .009). Higher grade 6 frustration significantly predicted higher post high school dating aggression (b = .05, SE = .02, p = .003).
Model of temperament, bullying perpetration, and dating outcomes for individuals in a romantic relationship in late adolescence. Values represent standardized coefficients or correlations. Only coefficients that are statistically significant at the p < .05 level are presented. Solid lines represent significant associations. Dashed lines represent significant indirect pathways including cross-lagged effects. Nonsignificant parameters and control variables are not displayed for ease of presentation.
Control variables
Post high school dating aggression was significantly predicted by being a girl, and post high school dating partners was significantly predicted by lower parental education. Grade 8 bullying was significantly predicted by lower household income. Lower household income was significantly correlated with higher Grade 6 frustration, and lower parental education was significantly correlated with lower sociability.
Indirect effects
There were two significant indirect effects that included cross-lagged pathways across the three time periods (i.e., Grade 6, Grade 8, and post high school; see Figure 1). Lower inhibitory control in Grade 6 significantly predicted higher bullying perpetration in Grade 8, which in turn predicted a higher number of post high school dating partners (b = −0.112, SE = .059, β = −.027, 95% CI [−.248, −.017]) and dating aggression perpetration (b = −0.008, SE = .005, β = −.026, 95% CI [−.020, −.001]). In summary, the same Grade 6 temperament trait had significant indirect effects to both dating benefits and costs through Grade 8 bullying perpetration.
Discussion
In the current study, the developmental pathways among early adolescent temperament, middle adolescent bullying perpetration, and late adolescent dating outcomes were examined among individuals in a romantic relationship in late adolescence. As predicted, bullying perpetration during middle adolescence was associated with both dating benefits and dating costs in late adolescence. Also as predicted, a temperament trait in early adolescence reflecting difficulty with self-regulation was associated with middle adolescent bullying and, in turn, was indirectly associated with late adolescent dating outcomes. Our findings indicate that bullying in middle adolescence can relate to additional forms of relationship violence in late adolescence, and both peer bullying and dating aggression may share a similar early temperamental indicator.
Pathways Between Temperament, Bullying, and Dating Outcomes
Consistent with previous associations between bullying and dating behavior, bullying perpetration during middle adolescence was associated with a higher number of dating partners 5 years later (e.g., Connolly et al., 2000; Pepler et al., 2006; Volk et al., 2015). In addition to these benefits, bullying was related to higher levels of dating aggression 5 years later (e.g., Connolly et al., 2000; Cutbush et al., 2016; Foshee et al., 2014; Niolon et al., 2015; Pepler et al., 2006). Our results support proposals of bullying as an adaptive behavior (e.g., Volk et al., 2012). Perpetrators of bullying may be more successful with both intersexual selection and intrasexual competition (Archer, 2009; Benenson, 2009; D. M. Buss, 1988; 2011; Vaillancourt & Krems, 2018; White et al., 2010). Adolescents who bully may self-promote features that may be attractive to opposite-sex dating partners such as an elevated dominance (Archer, 2009; D. M. Buss, 1988) and outcompete same-sex rivals (Benenson, 2009; Campbell, 2013; Vaillancourt, 2013). However, due to the coercive nature of bullying, it is possible that the desire to control and assert power over peers may carry over into later romantic relationships in the form of dating aggression in the long run (Foshee et al., 2014; Moffitt, 1993; Pepler et al., 2006).
Our findings support bullying as a relationship problem that can expand to other forms of developmentally relevant relationships (Pepler et al., 2006). Although bullying may be adaptive by successfully competing for and expanding the pool of dating partners, these dating relationships may be characterized by higher hostility and aggression and perhaps may be more characteristic of short-term dating and sexual opportunities rather than long-term romantic relationships. Although girls reported higher levels of dating aggression perpetration, we did not find that these associations were moderated by sex. Both girls and boys may be likely to use aggression to exert control and power over peers. Aggression may then generalize into dating relationships and used by both boys and girls to control and retain dating partners (Shackelford, Goetz, Buss, Euler, & Hoier, 2005). Researchers may want to examine sex differences in the forms and functions of dating aggression and the additional characteristics of these romantic relationships in future research. One source for the association between bullying and dating outcomes is reflected in our findings with temperament.
As seen in Figure 1, a trait indicating poorer behavioral regulation in Grade 6 (i.e., lower inhibitory control) predicted higher bullying perpetration 2 years later in Grade 8. The result is consistent with previous associations between bullying perpetration with lower inhibitory and effortful control (e.g., Bacchini et al., 2008; Terranova et al., 2008). This finding supports evolutionary proposals that traits may have been selected for if they increase chances of survival (D. M. Buss, 2011). Individuals with particular temperament traits may be more likely to use bullying as a strategy (Dane et al., 2017; Del Giudice & Belsky, 2011). Youth who demonstrate early difficulty with behavioral self-regulation may become adolescents who are less likely and able to reduce impulses to bully, and/or less likely and able to find alternative, nonaggressive means of securing resources. We did not find a significant association between Grade 6 frustration and Grade 8 bullying perpetration despite previous associations between bullying, anger, and negative emotions (e.g., Marini et al., 2006). However, we found concurrent associations between frustration and inhibitory control as well as bullying (see Figure 1). These concurrent associations suggest that difficulty regulating negative emotions may be relevant for bullying but may not be the strongest self-regulatory temperamental predictor of bullying over time. Frustration was also the only temperament trait to have an additional significant direct effect on dating aggression perpetration, suggesting that dating aggression and peer bullying may have some temperamental differences.
We also did not find the expected positive association between Grade 6 sociability and Grade 8 bullying despite previous associations between higher activity level and bullying (e.g., Marini et al., 2006; Pellegrini et al., 1999). The lack of an association in the present study may have been the result of the effect from inhibitory control to bullying. Researchers who have previously examined temperament (e.g., Marini et al., 2006; Pellegrini et al., 1999) and personality (e.g., Connolly & O’Moore, 2003) have explained that activity level or extraversion may be associated with aggressive behavior in individuals who may also have difficulty with self-control and/or lack fear of consequences. Indeed, researchers who have previously found associations between bullying perpetration and temperament or personality traits reflecting social approach found additional associations between bullying and traits related to emotion regulation and antisocial tendencies (e.g., Connolly & O’Moore, 2003; Mynard & Joseph, 1997; Pellegrini et al., 1999; Volk, Schiralli, Xia, Zhao, & Dane, 2018). Therefore, tendencies for social approach may be related to behavior in peer relationships, but in comparison to other traits, sociability may not be the strongest predictor of bullying over time.
Finally, household income and parental education had significant associations with some primary study variables. These findings are consistent with previous evidence linking adverse socioeconomic conditions with bullying (Tippett & Wolke, 2014) and emotional or behavioral problems (e.g., Shetgiri, Lin, Avila, & Flores, 2012). According to developmental evolutionary frameworks such as the life history theory, individuals experiencing harsh or unpredictable environments (i.e., lower economic resources and status) may utilize strategies that can help increase opportunities for survival such as higher frequencies of aggression, risk-taking, and mating behavior (B. J. Ellis et al., 2012; Figueredo et al., 2006). Within conditions of economic scarcity or inequality, there may be more adaptive benefits to gain than lose through the use of aggressive forms of behavior (Daly, 2016).
Indirect Effects From Temperament to Dating Outcomes
In addition to the associations from Grade 6 inhibitory control to Grade 8 bullying, this trait had indirect associations through Grade 8 bullying perpetration to both forms of post high school dating outcomes as evident in Figure 1. Our findings support previous researchers’ (e.g., Foshee et al., 2014) hypothesis that dating outcomes and bullying may share roots in traits related to self-regulation. Adolescents who engage in bullying may initially be appealing friends and dating partners (Foshee et al., 2014). The lack of fear when engaging in dominant or bold behavior like bullying may at first appear attractive to peers. Indeed, researchers have previously found that socially uninhibited adolescents have had higher success with peer and dating relationships (e.g., Ivanova et al., 2012; Sanson, Hemphill, & Smart, 2004). Eventually as adolescents get older, the greater access and control over peers may translate into the control and power over dating partners that may become more relevant in late adolescence (Faris & Felmlee, 2011).
Individuals may manipulate and exploit weaknesses in their dating partners in late adolescence, in a way similar to their previous middle adolescent bullying (Faris & Felmlee, 2011; Pepler et al., 2006). Higher levels of impulsivity and difficulty with behavioral regulation may additionally facilitate coercive behavior like bullying and dating aggression when the attempts to control dating partners are interrupted or delayed. Our results are consistent with previous links between difficulties with self-regulation and lower inhibitory control with dating aggression (e.g., Sullivan et al., 2017; Theobald & Farrington, 2012; Walters & Espelage, 2018). We extended these previous findings by demonstrating an indirect association through peer bullying perpetration in middle adolescence. Together, our findings suggest that adolescent bullying can result in dating benefits and costs, and both bullying and dating aggression may be rooted in early adolescent temperament.
Limitations
There were some limitations to this study. First, all measures were self-reports and our findings may have been inflated because of shared method variance. Also, one of the major limitations was the low reliabilities for the measure of inhibitory control and sociability. Researchers who have used the same inhibitory control measure on early adolescent samples (as opposed to middle or late adolescent samples) have also previously noted lower reliabilities for this subscale (e.g., Gardner, Dishion, & Connell, 2008; Muris, Meesters, & Blijlevens, 2007). Also, the sociability measure has primarily been validated in adult samples (e.g., Cheek & Buss, 1981; Crozier, 2005) but is still used in child and youth samples (e.g., Tang, Santesso, Segalowitz, & Schmidt, 2016). The younger age of our early adolescent sample (i.e., Grade 6) may have been one reason why the reliabilities of these two subscales were low. Another reason may be that inhibitory control and sociability are more narrow components of broader temperament traits. Inhibitory control is one aspect of a broader effortful control subscale that is also comprised of activation control and attention, whereas sociability is one component of a broader surgency that is also comprised of high-intensity pleasure, low fear, and low shyness (L. K. Ellis & Rothbart, 2001). Moreover, our sociability measure was from a separate measure than the EATQ-R. Researchers may want to incorporate some of these additional items from the same temperament questionnaire in future research to see whether this increases reliability. To further counteract these biases and low reliabilities, researchers may want to use peer reports. We also made sure to use measures and subscales that have been commonly used and well known by researchers in the field. Second, we did not have measures of temperament in Grade 8 and post high school or measures of dating variables in Grade 8. This may have inflated our findings, as we were unable to run more robust models that control for prior associations. However, we were able to control for Grade 6 bullying perpetration and our small effect sizes were consistent with those generally found in longitudinal research (e.g., Adachi & Willoughby, 2015). Third, our sample included participants who were in a romantic relationship during late adolescence. These participants had some significant differences in demographics, temperament, and the number of dating partners, indicating that our findings may be less generalizable to individuals who were not in a dating relationship at that specific time point. Continued longitudinal studies that capture relationship status each year may help indicate whether these findings are generalizable across all adolescent years. Fourth, although the study was longitudinal, it does not allow us to draw true causal conclusions. Researchers may want to examine additional mechanisms between individual differences, bullying, and dating outcomes. Fifth, we examined a composite of bullying and a composite of dating aggression. Previously, researchers have found that girls engaged in higher physical and verbal dating violence whereas boys engaged in higher sexual dating violence (e.g., Pepler et al., 2006; Walters & Espelage, 2018; Wincentak et al., 2017). Researchers may want to extend our findings by determining whether different temperament traits are associated with different forms of bullying and dating aggression. Given that our sample would have been underpowered to detect sex differences in these different forms of dating violence, our goal was to establish an overall indirect link among early adolescent temperament, middle adolescent bullying, and late adolescent dating outcomes.
Implications and Conclusions
Our results suggest that a temperament trait evident in early adolescence can have a lasting impact on bullying in middle adolescence and, in turn, dating outcomes in late adolescence. From a theoretical perspective, our findings support evolutionary proposals of bullying as an adaptive behavior (e.g., Volk et al., 2012). Bullying may be partially explained by individual differences in temperament. Difficulty with behavioral self-regulation may be associated with adaptive uses of bullying to compete for dating partners. However, in line with a developmental framework, bullying may be one form of relationship conflict that generalizes to other forms of relationships including dating aggression (Pepler et al., 2006).
From a practical perspective, our findings highlight that intervention efforts should incorporate an integrated approach that recognizes individual and contextual risk factors that are developmentally relevant and to recognize the challenges of developing interventions for a behavior that can also result in benefits (i.e., opportunities for dating). Early identification and intervention of difficulty regulating behavior may be important for preventing both bullying and dating aggression (Foshee et al., 2014). Given that using power and aggression with peers may generalize to other relationships, healthy relationship and conflict resolution skills may be important to prevent multiple forms of aggression (Pepler et al., 2006). Also, the absence of a sex moderation indicates that efforts to reduce overall levels of bullying and dating aggression should be equally applied to adolescent girls and boys (Connolly et al., 2000; O’Leary & Slep, 2012).
Regarding temperamental differences, parents and teachers may encourage adolescents to discuss negative behavior and emotions (e.g., Gray & Steinberg, 1999; Marini et al., 2006) to counteract difficulties with behavioral self-regulation. Teaching conflict management skills along with broader interpersonal communication skills in early adolescence may also be beneficial for youth with difficulties inhibiting impulses during peer and dating conflict (e.g., Williams, Connolly, Pepler, Craig, & Laporte, 2008). The additional associations between indicators of socioeconomic status suggest that adolescents experiencing socioeconomic adversity may be further at risk of self-regulation difficulties and aggressive behavior. Efforts should be made to provide resources for youth who may otherwise have difficulties accessing resources as a result of economic or social limitations. Tailoring methods to improve social cognitions and habits to the youth involved may then hopefully have lasting positive effects on all future peer and dating relationships.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Social Sciences and Humanities Research Council of Canada (grant numbers 833-2004-1019, 435-2016-1251), Ontario Mental Health Foundation (grant number PA-13-303), and Canadian Institutes of Health Research (grant numbers 201009MOP-232632-CHI-CECA-136591, 201603PJT-365626-PJT-CECA-136591).
