Adolescent Bullies, Victims, and Bully-victims: Comparing their Interpersonal Goals with Uninvolved Students
Amy Valerie Stauffer Student number: 12792934
Abstract
The differences between the social goals of uninvolved students and those of victims, bullies, and bully-victims are unclear. In this study, these differences were investigated and used to predict the belongingness to a bullying role while controlling for gender by answering the following research question: How do high school students who are victims, bullies, and bully-victims differ with regard to interpersonal goals compared to uninvolved students? The participants were high school students (Nboys = 66; Ngirls = 53) aged 12 to 16 (M = 13.94, SD = 0.94). A multinomial regression showed that the bullying role (assessed with a single item questionnaire) was not predicted by the differences in interpersonal goals (measured with the Interpersonal Goal Inventory) when controlling for gender differences in bullying behavior. These findings highlight the need to keep researching what makes students adopt specific bullying roles, to create effective interventions in order to reduce bullying behavior. Implications for alternative research include using different methods for the assessment of bullying roles and agentic goals. Suggestions for future research into the influence of for example social skills such as empathy and the SIP model on the bullying roles are proposed.
Adolescent Bullies, Victims, and Bully-victims: Comparing their Interpersonal Goals with Uninvolved Students
Almost half of the children who committed suicide in 2017 in the Netherlands encountered bullying behavior in school (Nederlandse Omroep Stichting, 2017). It is therefore unsurprising that the media is paying attention to this phenomenon. Examples are two published news articles about Marwan and Tharukshan, both just 15 years old, who committed suicide because they were bullied in high school (De Limburger, 2017;
EenVandaag, 2019). A worldwide study including data from 48 countries stated that students who are bullied are three times more likely to attempt suicide (Koyanagi et al., 2019).
However, not only being a victim, but also being a bully, a bully-victim, or even an uninvolved student can be associated with negative outcomes of bullying behavior.
Globally, one in three children encounters bullying behavior in school (UNESCO, 2019). In the Netherlands alone, 22.1% of children admit to being bullied (UNESCO, 2019), with almost a quarter being bullied during high school (Onderwijs in cijfers, 2018). Within adolescence, sex differences occur, with boys engaging more in bullying behavior than girls (Klomek, Marrocco, Kleinman, Schonfeld, & Gould, 2007; Reijntjes et al., 2013; Salmivalli & Peets, 2009). The involvement in this behavior puts adolescents at risk and research has mapped all kinds of negative outcomes. Examples of these consequences are that compared to uninvolved students, bullies, victims, and bully-victims have a higher risk for psychosomatic problems, such as headaches and stomach aches (Gini & Pozzoli, 2009), suicidal ideation, and suicidal behavior (Holt et al., 2015; Klomek et al., 2007). For victims, higher risks of sleeping problems (van Geel, Goemans, & Vedder, 2016) and internalizing problems such as somatic and psychotic problems have been identified (Moore et al., 2017). Even uninvolved students, who are mostly called bystanders or witnesses in literature, have a higher risk of anxiety and depressive symptoms (Midgett & Doumas, 2019). Back in the day, bullying was seen as a
harmless part of childhood (Arseneault, Bowes, & Shakoor, 2010), but in fact, bullying can harm children in many ways.
Worldwide research on school-based bullying interventions shows that they are overall effective in reducing bullying behavior (Gaffney, Farrington, & Ttofi, 2019). However, when zooming in on data coming from the Netherlands alone, the results of the interventions are nonsignificant. This is in line with the results of a large Dutch study investigating the
effectiveness of interventions for reducing bullying and victimization. This study showed that currently used interventions are not effective in reducing victimization overall (Orobio de Castro et al., 2018). Research into specific anti-bullying programs is done solely for primary school students and shows contradicting results. The intervention KiVa was effective in decreasing bullying behavior but the effect size was small (Huitsing et al., 2020) and the intervention Topper training was ineffective in reducing bullying behavior (Vliek, Overbeek, & Orobio de Castro, 2019). This study examines the role of interpersonal goals on bullying behavior because more insight is needed into the reasons why students engage in bullying behavior and the goals that motivate this behavior to increase the effectiveness of school-based bullying interventions.
But what exactly is bullying? Bullying is best defined as the intention to repetitively harm someone else, with a clear power imbalance between the bully and the victim. The victim is therefore unable to defend himself or herself (Olweus, 1994). There are different roles students can take on and one way to classify these roles is into the following four subgroups. First, the pure bullies, students who only bully others and from now on referred to as bullies. Second, the pure victims, those who are only victimized, referred to as victims from now on. Third, the bully-victims, those who are both bullies and victims. And fourth the
uninvolved students, who are neither bullies nor victims (Haynie et al., 2001; Nansel et al.,
All four bullying roles develop differently and for each of these roles a possible way of development is given. The development of the victim role can start with initial attacks by peers, caused by for example differences in appearance. This leads to the recurring process of double victimization involving external victimization, repeatedly being bullied by peers, and internal victimization, identifying with and behaving according to the victim role (Thornberg, Halldin, Bolmsjö, & Petersson, 2013). Becoming a bully can start with rejection by peers who witnessed the bullying behaviors. This activates the bully’s performance, increasing the display of more bullying behavior, and eventually leading them to identify as a bully (Lam & Liu, 2007). The development of a bully-victim starts with the single role phase where the student is either a victim or a bully (Sung, Chen, Yen, & Valcke, 2018), with most of them being victims first (Lereya, Copeland, Zammit, & Wolke, 2015). This phase is followed by the dual role phase, where the internal coping strategies of the victims are not sufficient anymore (Sung et al., 2018). To reduce the bullying and to satisfy their needs, they find more vulnerable students and mimic the bullying behavior of bullies. An initial victim is now a bully-victim. If students are a bully first, the bullying can become labeled as inappropriate by their peers, who will, therefore, reject the bully. Rejection is also a form of bullying, making the bully a bully-victim (Sung et al., 2018). The development of the roles differs for every group, however, most of the bully-victims were victims first, so they probably experienced the development of the victim role as well, possibly making these groups more similar. Interpersonal Goals and Bullying Behavior
From a social-cognitive perspective, behavior is motivated by goals and the goal is related to the desired outcome (Crick & Dodge, 1996). Social or interpersonal goals are therefore social outcomes that students prefer or want to avoid (Jarvinen & Nicholls, 1996). The interpersonal goals can be divided into two main categories. The agentic goals, aimed at achieving power, status, or influence (Locke, 2015). Behaviors associated with this goal are
more aggressive and less prosocial (Ojanen, Grönroos, & Salmivalli 2005). And the communal goals, aimed at achieving and maintaining positive relationships (Locke, 2015). Students with this goal are more prone to prosocial and withdrawal behaviors (Ojanen et al., 2005). Within these two groups, smaller categories exist adding the dimensions submission: avoiding arguments and conflict by going along with other’s expectations, and separation: concealing thoughts and feelings (Locke, 2010). These four dimensions create eight categories in total. An overview with examples is given in Figure 1.
Figure 1. The placement of the interpersonal goal scales in the circumplex model, followed by an example item from each of
the eight scales. + A = Agentic; + A – C = Agentic and Separate; + A + C = Agentic and Communal; - C = Separate; + C = Communal; - A - C = Submissive and Separate; - A + C = Submissive and Communal; - A = Submissive. Adapted from “An interpersonal circumplex model of children's social goals: Links with peer-reported behavior and sociometric status,” by T. Ojanen, M. Grönroos, and C. Salmivalli, 2005, Developmental Psychology, 41, p. 701. Copyright 2005 by the American Psychological Association.
Adolescence is seen as an important social developmental period because popularity is prioritized (LaFontana & Cillessen, 2010) and interacting with peers becomes more important than with parents (Adler & Adler, 1998). Both agentic and communal goals increase during adolescence, due to an overall decrease in submission and separation goals (Trucco, Wright, & Colder, 2014), possibly caused by an increasing interest in (romantic) relationships (Pellegrini, 2002; Rose & Rudolph, 2006). Regarding this social aspect, findings on peer status show different outcomes related to the bullying roles. Research compared all bullying roles in three domains: social impact, the visibility of an individual in a group (Garandeau, Lee, & Salmivalli, 2013), social preference or sociometric popularity, how liked an individual is (Garandeau, Lee, & Salmivalli, 2014), and perceived popularity, the social dominance within the peer group (de Bruyn, Cillessen, & Wissink, 2010). The uninvolved students had the lowest social impact and bully-victims had more social impact than the victims (Guy, Lee, &Wolke, 2019). Regarding sociometric popularity, the victims and bully-victims scored lower than the bullies and the uninvolved students, making the latter the students who are most liked (Guy et al., 2019). Concerning perceived popularity, compared to the uninvolved students, the bullies scored significantly higher, and the victims as well as the bully-victims scored significantly lower (Guy et al., 2019). These findings indicate that every role has a different combination of peer status measures.
Research examining the relationship between bullying behavior and interpersonal goals is limited, general, and partly indirect. In general, early adolescents with higher agentic goals engage more in bullying behavior, when measured with the agentic questions of the Interpersonal Goal Inventory (IGI) (Sijtsema, Veenstra, Lindenberg, & Salmivalli, 2009). Another study investigated associations between interpersonal goals, bullying behavior, and popularity in early adolescence. Early adolescents with higher communal goals, scored higher on sociometric popularity and higher sociometric popularity was associated with less bullying
behavior (Caravita & Cillessen, 2012). On the other hand, students with high agentic goals scored higher on perceived popularity, which in turn was related to more bullying (Caravita & Cillessen, 2012). This shows an indirect relationship between bullying behavior and
interpersonal goals, via popularity.
This study
There is empirical evidence pointing towards, at least an indirect, relationship between interpersonal goals and bullying behavior in general. However, these studies did not
differentiate between the four bullying roles and are therefore limited (Caravita & Cillessen, 2012; Sijtsema et al., 2009). This study attempts to fill this gap by examining the direct links between bullying behavior and the differences in interpersonal goals between the uninvolved and three groups of involved students (i.e., bullies, victims, and bully-victims). The research question to be answered is: How do high school students who are victims, bullies, and bully-victims differ with regard to interpersonal goals compared to uninvolved students? This study examined the bullying roles and interpersonal goals of high school students, aged 12 to 16, taking into account gender differences (Klomek et al., 2007; Reijntjes et al., 2013; Salmivalli & Peets, 2009).
There is not enough research into these specific subgroups to formulate concrete hypotheses upfront, as previous research did not compare victims, bullies, and bully-victims with uninvolved students. However, the theory hints towards expected outcomes regarding the interpersonal goals of the bullying roles (Caravita & Cillessen, 2012; Sijtsema et al., 2009). Because the uninvolved students do not bully and scored high on sociometric
popularity, I expected them to score higher on communal goals. Bullies do engage in bullying behavior and scored higher on perceived popularity, both indicating higher agentic goals than uninvolved students. However, they scored similarly on sociometric popularity indicating
higher communal goals. Victims also scored lower than the uninvolved on both sociometric and perceived popularity, suggesting lower communal goals than the uninvolved students and lower agentic goals than the bullies. The bully-victims do engage in bullying behavior
therefore I expected higher agentic goals than the uninvolved, however, they scored lower than the uninvolved students on perceived and sociometric popularity suggesting lower agentic and communal goals. Based on this theoretical evidence, differences in interpersonal goals of all four groups can be inferred, but the theoretical evidence of bullies and bully-victims is contradicting. It is important to do research related to the possible differences in interpersonal goals for each of the four bullying roles because this information could shed light on the desired social outcomes of students in specific bullying roles. This research has not been performed yet and could potentially improve the production of tailored interventions to reduce and eventually stop bullying.
Method Participants
This study was conducted within a Dutch high school, including students following preparatory vocational education (VMBO in Dutch). Approximately 330 parents were contacted and the response rate was about 45%. Some students with consent from their parents did not participate due to illness, and three students did not consent themselves and therefore did not participate. Of the in total 120 participants, 55.8% were boys. The ages ranged from 12 to 16 years old, with a mean age of 13 years and 11 months (SD = 11 months). The sample was ethnically diverse consisting of Dutch (57.5%) and non-Dutch (42.5%) students. Students were classified as non-Dutch when at least one of their parents was born abroad (Centraal Bureau voor de Statistiek, n.d.). The four tracks of VMBO are given with the percentage of the sample that followed a particular track: “Basis” (35.0%),
“Kader” (32.5%), “Gemengd” (12.5%), and “Theoretisch” (20.0%). There was no missing data.
Procedure
Prior to data collection, parents were informed about the study and asked to give their consent for the participation of their child(ren) in this study. Meanwhile, two students
following a Master track from the University of Amsterdam were trained by the researcher to explain the questionnaire, stick to a script, and answer any questions regarding the research. To conduct the collection of the data, the researcher and at least one of the trained students went to the selected school. Data collection was spread out over two separate days within one week. The students with consent to participate from their parents were gathered and before handing out the questionnaire, we informed them about the data being confidential, the freedom to decline to answer any questions, and the freedom to decide not to participate or to quit at any time, without any further consequences. We explained that their parents had consented already, but that they also needed to consent themselves to participate. The first page of the questionnaire was a consent form that the children had to sign to participate. Besides the explanations on the questionnaire, we also explained everything out loud. After the consent form, the children filled out the questionnaire on paper. During the questionnaire, the researcher and trained students gave instructions to the group to make sure everybody was on the same pace and understood the upcoming sections. On average the questionnaire took 30 minutes in total.
Measures
Demographics. The following demographics were measured with the questionnaire. The age and gender of the child, the level of VMBO they followed (four options), the grade they were in (ranging from one to three), and where the students and their parents were born.
Bullying groups. To determine the bullying role, bullying behavior was measured with the questionnaire. Before answering the questions, we gave the students the definition of bullying both oral and in writing. The definition consisted of examples of bullying: saying and doing mean things, kicking and hitting, shutting people out, destroying and grabbing or stealing other people’s belongings, and cyberbullying. It is only considered bullying when it happens regularly and on purpose, and it has to be difficult for the victims to defend
themselves based on the bullying definition (Olweus, 1994) as used in previous empirical studies (Olweus, 2013).
The part measuring bullying behavior consisted of two questions; from the beginning of the school year (September), how often have you bullied others and how often have you been bullied? The answering options were 0 (Never), 1 (Once or twice), 2 (A few times a
week), and 3 (Multiple times a week). Students were classified as bullies when they scored 1
or higher on the perpetrator questions and 0 on the victimization question (N = 12). For victims, the score had to be a 0 on the perpetrator question and a 1 or more on the
victimization question (N = 25). Students were considered bully-victims when they scored a 1 or higher on both questions (N = 5) and the uninvolved category had to score a 0 on both questions (N = 78). This way of classifying is considered as using a conservative cut-off point, including both low and high levels of bullying (Goldbach, Sterzing, & Stuart, 2018).
Social goals. To measure the interpersonal goals of the students we used the
Interpersonal Goal Inventory (IGI) (Ojanen et al., 2005). This is a questionnaire consisting of 33 items all starting with “How important is it for you that…” Students would then read a statement and choose an answer from a four-point rating scale including the options, 0 (Not
important to me), 1 (Not really important to me), 2 (Quite important to me), and 3 (Very important to me). These 33 items measured the eight dimensions of social goals: agentic (e.g.,
things you want to say”), communal (e.g., “…you get along with others very well”),
submissive and communal (e.g., “…others like you”), submissive (e.g., “…others don’t get mad at you”), submissive and separate (e.g., “…you don’t do anything silly, weird, or ridiculous”), separate (e.g., “…you don’t show your feelings when others are present”), and agentic and separate (e.g., “…you decide what the group is going to do”). There was at least one of the 33 items missing from the questionnaires of 22 students. This added up to 32 missing answers, accounting for 0.01% of the total items of the IGI. When students did not answer one or more of the items of the IGI, the other answers were still used in the analyses. So, there were no students excluded from analyses based on missing data from the IGI.
Before running analyses, I prepared the data from the IGI. First, the reliability of the eight subscales from the IGI was calculated using Cronbach’s Alpha. The values ranged from .60 to .80, except for the dimension agentic and separate (.57), this is comparable to other studies (Ojanen et al., 2005). To standardize the data from the IGI scale, I ipsatized the scores using a procedure multiple-step (Fischer & Milfont, 2010). First, the mean and the standard deviation of all 33 items of the IGI were calculated per individual, followed by calculating the mean score per subscale per individual. To ipsatize the score, the mean of all 33 items is divided by the standard deviation of all 33 items, and the outcome of this is subtracted from the mean of one of the subscales. In total, this calculation was carried out eight times for all eight subscales of the IGI. This process helps to take out acquiescence responding, which can be explained as agreeing with all items regardless of their content (Fischer & Milfont, 2010). One participant answered all the items of the IGI with a zero. These scores could not be ipsatized because the calculation would include dividing by zero, which is mathematically impossible. This participant was therefore excluded, reducing the sample size to 119 participants. With the ipsatized scores, I calculated the vector scores for both agentic and communal goals, using two formulas (Ojanen et al., 2005):
Agenticvect = Agentic – Submissive + [.707 x (Agentic and Communal + Agentic and Separate – Submissive and Communal – Submissive and Separate)].
Communalvect = Communal – Separate + [.707 x (Agentic and Communal + Submissive and Communal – Agentic and Separate – Submissive and Separate)].
As expected because they measure different constructs, the agentic and communal vector scores were weakly correlated (r = -.09).
Data Analysis
For the main analysis, a multinomial regression was the best fitting test. Because the agentic and communal vector scores were the continuous independent variables and the bullying roles were the categorical dependent variable with four groups, using a multiple linear regression was not possible. Due to the small sample size, a simpler analysis would not provide enough power. With a multinomial regression, the goal is to try and predict the bullying role with the differences in agentic and communal vector scores between uninvolved and involved students, while controlling for gender because of the sex differences in bullying behavior. Before conducting the main analysis, the normality of the interpersonal goals and the assumptions for a multinomial logistic regression were investigated.
Results Preliminary Results
To investigate the normality of the IGI items, I checked the skewness, kurtosis, normal Q-Q plots, and ran a Shapiro-Wilk test (de Vocht, 2014). To calculate the kurtosis and the skewness, the raw scores were divided by the standard error. For the agentic vector, this resulted in a skewness value of 0.63 and a kurtosis value of 3.95. For the communal vector, the skewness value was -0.48 and the kurtosis value was -1.48. To assure normality, all values should be ≤ |1.96| (de Vocht, 2014). Both skewness values were within a normality range. The
kurtosis value of the agentic vector was too big, indicating that there are outliers in the data. For both vectors, I conducted the Shapiro-Wilk test. For the communal vector score, the Shapiro-Wilk test was nonsignificant, W(119) = 0.99, p = .214, indicating a normally
distributed variable. For the agentic vector score, the Shapiro-Wilk was significant, W(119) = 0.97, p = .022. The boxplot of the agentic vector score showed outliers and I used the 3SD rule to determine problematic ones (Aggarwal, 2015). An outlier is considered problematic when it has a value of more than three standard deviations away from the mean. This rule allows for keeping 99% of the data included and this is valuable for smaller data sets. Two values exceeded the 3SD rule and were therefore labeled problematic outliers. These were changed into the highest or lowest score after that, called winsorisen (Lusk, Halperin, & Heilig, 2011). I chose this tactic over trimming to keep the sample size as big as possible and to prevent losing power because of a smaller sample size (Lusk et al., 2011). After adjusting the outliers, I ran the normality tests of the agentic vector again. The skewness value was 0.02 and the kurtosis value was 1.02, both within a normality range. The Shapiro-Wilk test was nonsignificant, W(199) = 0.98, p = .104, and the normal Q-Q plots were investigated, both indicating a normally distributed variable.
To ensure that the assumptions were met, I investigated the following. The
observations were independent of each other because we did not use repeated measurements or matched data. There was no multicollinearity, so the independent variables did not highly correlate with each other (Table 4). Normally a multinomial regression requires a large sample size, partly depending on the number of predictors. Even though the outliers were handled carefully so that they did not influence sample size, the actual sample size was still relatively small. However, with the small number of predictors and considering the variables and the data, it is still the best fitting analysis (Schreiber-Gregory, Jackson, & Bader, 2018; Statistics Solutions, n.d.). When adding the reliability and the normality of both interpersonal
goal scales, it can be stated that the assumptions are met. All the variables were entered into one model because using two models would reduce the power of the main analysis. This means that when looking at one of the predictors, the value is automatically controlled for all other inserted variables.
Descriptive Statistics
Table 1 shows the dispersion of the bullying roles for both sexes. Overall, most of the students were classified as uninvolved students. There were more victims than bullies and the bully-victims group was the smallest group in the sample. This pattern was consistent for both sexes. However, boys were overall more involved in bullying behavior than girls. They were more often bullies, victims, and bully-victims whereas almost three-quarters of the girls were uninvolved students. It was not possible to test whether the gender differences were
significant with a Chi-Square test, because the assumption that only 20% of the cells can have a participant number between 1 and 5 was violated (de Vocht, 2014). In this sample, 25% of the cells had a participant number between 1 and 5.
Table 1
Bullying Roles by Gender
Bully Victim Bully-victim Un-involved Total Gender Boy Count % within gender % of total 10 15.2 8.4 14 21.2 11.8 4 6.1 3.4 38 57.6 31.9 66 100.0 55.5 Girl Count % within gender % of total 2 3.8 1.7 11 20.8 9.2 1 1.9 0.8 39 73.6 32.8 53 100.0 44.5 Total Count % of total 12 10.1 25 21.0 5 4.2 77 64.7 119 100.0
Table 2 shows the descriptive statistics of the agentic and communal vector scores by gender and for the total sample. The average agentic vector score was negative, whereas the average communal vector score was positive. This shows that on average students were more communal than agentic. In addition, girls scored lower on agentic goals and higher on
communal goals. This indicates girls being more communal and boys being more agentic on average.
Table 2
Interpersonal Goals by Gender and in Total
Boys (N = 66) Girls (N = 53) Total (N = 119)
Mean SD Mean SD Mean SD
Agentic Goals -0.37 0.79 -0.84 1.10 -0.58 0.97
Communal Goals 1.55 1.13 1.94 1.22 1.72 1.18
The mean score and standard deviation of agentic and communal goals ordered by the bullying role can be found in Table 3. On average, bullies scored highest on the agentic goals, followed by bully-victims, victims, and lastly uninvolved students. On the communal scores, uninvolved students scored highest on average, followed by bullies, bully-victims, and lastly victims.
Table 3
Interpersonal Goals by Bullying Role
Bully Victim Bully-victim Uninvolved
Mean SD Mean SD Mean SD Mean SD
Agentic Goals -0.22 0.93 -0.62 1.21 -0.33 0.81 -0.64 0.90 Communal Goals 1.58 1.36 1.38 1.25 1.57 1.81 1.87 1.08
The correlation coefficient for agentic and communal goals was small and
nonsignificant (p = .315), indicating no relationship between these two variables (Table 4). For these two continuous variables, Pearson’s r was used (de Vocht, 2014). For the
correlation between gender and both interpersonal goals, Spearman’s r was used, because one of the variables was categorical (de Vocht, 2014). The correlation between gender and
communal goals was also nonsignificant (p = .071), however the correlation between gender and agentic goals was significant but small (p = .005).
Table 4
Pearson and Spearman Correlations
1. 2. 3.
1. Gender -
2. Agentic Goals -.257** -
3. Communal goals .166 -.093 -
Note. *p < .05. **p < .01. ***p < .001.
To test whether the differences in the interpersonal goals between boys and girls were significant, I conducted an independent-samples T-test. For the differences in agentic goals, the Levene’s test for equality of variances was significant, F = 6.02, p = .016, leading to a rejection of the null hypothesis of equal variances. I therefore decided to use Welch’s t-statistic instead of Student’s t t-statistic. Using α = .05 as the criterion for significance, I rejected the null hypothesis stating that there were no differences between the means of agentic goals for boys and girls, Welch’s t(91.454) = 2.61, p = .010, 95% CI [0.11, 0.83]. This significant difference was moderate with a calculated Cohen’s d of 0.49 (Social Science Statistics, n.d.). This means that boys had on average significantly higher agentic goals than girls. For the differences in communal goals, the Levene’s test for equality of variances was nonsignificant, F = 0.04, p = .835. Because the null hypothesis of equal variances could not
be rejected, it was appropriate to use the Student’s t-statistic. Using α = .05 as the criterion for significance, I did not reject the null hypothesis stating that there was no difference between the means of communal goals for boys and girls, Student’s t(117) = -1.77, p = .079, 95% CI [-0.81, 0.05]. This means that even though girls scored on average higher than boys, this difference was not significant. Because of the significant differences in the agentic vector score, it was important to control for the effect of gender in the main analysis.
Main Analysis
For the main analysis, I conducted a multinomial regression. I assessed the model fit first. Using α = .05 as the criterion for significance, the null hypothesis stating that the model with the predictors was a better fit than the model without predictors (intercept only model) could not be rejected, χ2 (9) = 10.97, p = .278. The model with the predictors explained only 4.7% of the variance (McFadden R2 = 0.047). This means that the model without the
predictors was a better fit.
All entered predictors were nonsignificant in the analysis, indicating that the bullying roles were not significantly predicted by the communal vector score, χ2 (3) = 3.07, p = .382, by the agentic vector score, χ2 (3) = 1.23, p = .747, or gender, χ2 (3) = 5.13, p = .163. Even though the predictors as a whole were not significant, the odds ratios are displayed in Table 5. The values in table 5 are all relative to the uninvolved students. Because all the predictors were entered into the model simultaneously, the odds ratio for one variable is controlled for all other predictors. Noteworthy is the wide range between the confidence intervals, a smaller range indicates a more precise prediction. This wide range showed a poor model fit again (Petrucci, 2009).
Table 5
The Odds Ratios of Belonging to a Specific Bully Role
OR [95% CI] Sig. B
Bully roles Victim Boys vs. girls 1.18 [0.46, 3.07] .732 0.17 Agentic 0.96 [0.59, 1.57] .880 -0.04 Communal 0.70 [0.47, 1.05] .085 -0.35 Bully Boys vs. girls 4.38 [0.88, 21.82] .071 1.48
Agentic 1.44 [0.70, 2.98] .320 0.37 Communal 0.91 [0.52, 1.56] .721 -0.10 Bully-victim Boys vs. girls 3.59 [0.37, 34.93] .271 1.28
Agentic 1.26 [0.44, 3.61] .669 0.23 Communal 0.88 [0.39, 1.96] .750 -0.13
Note. The uninvolved student group is used as a reference category.
Even though table 3 shows differences in the average scores of both agentic and communal goals between all four groups, the main analysis indicates that it was not possible to predict the belongingness to a specific bullying role by using gender, agentic goals, or communal goals as predictors.
Discussion
The main aim of this study was to gain more insight into the role of interpersonal goals in the belongingness to bullying roles. By conducting a multinomial regression, I tested whether the different bullying roles could be predicted, at least partly, by differences in interpersonal goals, measured as the agentic and communal vector scores. This study did not find differences between the interpersonal goals of the uninvolved students and the bullies, victims, or bully-victims. This indicates that in this study the agentic and communal vector scores did not play a role in explaining why students belong to a certain bullying role.
Although it was not possible to formulate hypotheses based on empirical research upfront, differences between the uninvolved students and the three groups of involved
differences in the interpersonal goals of the uninvolved and any of the involved bullying roles were identified. These findings could be explained by the use of different measures. The research showing that students with higher agentic goals engage more in bullying behavior in general used peer nominations for assessing the bullying roles. In addition, the agentic goals were measured with only the agentic subscale of the IGI (Sijtsema et al., 2009). The study showing the indirect association of interpersonal goals and bullying behavior via popularity also used peer nominations to measure bullying behavior (Caravita & Cillessen, 2012). The interpersonal goals were measured using the vector scores for agentic and communal goals (Caravita & Cillessen, 2012). Although these studies did not include the same bullying roles, the results could be different from the findings of this study due to the use of peer
nominations for bullying roles and the use of only the agentic subscale for agentic goals, instead of self-reports and the agentic vector scores respectively. It would be interesting to include all measuring methods for these constructs in one study, to compare them and decide what method is the best to use in which circumstances.
Another explanation for our results could be due to the choice of the bullying roles used. The division of the sample in only four subgroups may not cover all the roles properly. An alternative way to split the sample would be to use the categories of Salmivalli (1999). To add to the bullies, victims, bully-victims, and uninvolved students, with the groups: assistants, reinforcers, outsiders, and defenders. Using these categories maybe makes the separate groups more homogeneous with regard to interpersonal goals, causing the differences between the groups to become more apparent. Another change to the used bullying roles could be to split the bully-victims group into two groups. When looking at the development of this role (Sung et al., 2018), part of this group was a bully first, and the other part was a victim first so maybe it is not justified to classify those students as one group. Bullies and victims differ in their outcomes and development. It could be possible that the part of the bully-victims who were
victims first differ from those who were bullies first and that they should be viewed as two distinct categories.
A final explanation could be that it is not due to our measurement methods or the way we operationalized our constructs, but that the social goals might not have the expected influence on the bullying roles. Perhaps, the bullying roles are determined by other factors, such as social deficits. Is has long been thought that bullies are socially impaired (Crick & Dodge, 1996, 1999). This statement is tested with research on the social information
processing model (SIP). This model states that social information is processed mentally and step-wise resulting in a behavioral response (Crick & Dodge,1994). It consists of five steps: encoding social cues, interpretation of social cues, clarification of goals, response
construction, and response decision followed by the enactment of the behavioral response. To react appropriately to social stimuli, accurate assessment in all steps has to take place.
Research regarding the role of the SIP model is contradicting. It identified distinct SIP processes for victims, bullies, and uninvolved students and found the bully-victims to be similar to the bullies (Ziv, Leibovich, & Shechtman, 2013). However, others did not find deficiencies in the early steps of the SIP model of bullies but did find biases in the SIP models of victims and bully-victims (Guy, Lee, & Wolke, 2017). Besides the SIP model, certain social skills are found to be related to bullying behavior. Examples are that students with more empathy skills engage less in bullying behavior (Elliott, Hwang, & Wang, 2019; Mitsopoulou & Giovazolias, 2015). In addition, students with higher assertion skills engage less in bullying behavior (Elliott et al., 2019). Finally, lower overall social skills lead to more maladjustment, making a student less popular, and more involved in bullying behavior (Postigo, González, Mateu, & Montoya, 2012). This evidence all indicates that there is a possibility that the social capacities have a bigger influence on the bullying roles than social
goals do because we identified no differences between uninvolved students and all three other roles.
Strengths and Limitations
The first strength is the sample of this study. The sample consisted of only VMBO-students, making the study generalizable for this particular track in the Dutch system. In addition, the sample included similar amounts of boys and girls, and was ethnically diverse being almost half Dutch and half non-Dutch. Second, the use of self-reports made it possible to provide the definition of bullying, with an emphasis on the need for a power imbalance. It also provided a source of data that did not rely on other participants, as peer-nominations would (Felix, Sharkey, Green, Furlong, & Tanigawa, 2011).
Even though self-reports are a worldly accepted measure, it comes with the risk of social desirability bias, causing participants to present themselves more favorably (Olsen & George, 2004). This becomes even more probable with research showing that single item questionnaires including only one question about involvement in bullying behavior lead to lower prevalence rates than multiple item assessment based on behaviors related to bullying (Green, Felix, Sharkey, Furlong, & Kras, 2013; Jetelina et al., 2019). Possible causes for this are the fear of sanctions from adults and peers when admitting to bullying or the belief that their actions are just harmless jokes (Guy et al., 2019). Our single-item assessment of bullying behavior may have led to underreporting it and a relatively big group of uninvolved students.
Related to this, is the limitation regarding the sample size. Partly caused by the lower prevalence rates, but also due to needing active consent from the parents and students the overall sample size was quite small. This was especially apparent in the bully-victim group, including only five participants. The low consent rate poses a risk for the nonresponse bias because it was not possible to make sure the participating group did not differ significantly
from the not participating group (Navarro & Foxcroft, 2018). The data needs to be interpreted with caution the bully-victims are represented by only five participants in this study.
Implications for Research and Practice
First, future research should focus on measuring interpersonal goals with both self-reports and peer nominations because these methods could lead to different outcomes. Likewise, using only the agentic subscale of the IGI and the agentic vector should be compared. Conducting longitudinal research with an extended target group including both younger and older adolescents would make the changing patterns of interpersonal goals in this important developmental time visible. These patterns can then be used to select the best fitting methods to assess bullying behavior and interpersonal goals at different ages within
adolescence.
Second, the bullying roles of Salmivalli (1999) should be used in future research. This could make the uninvolved students as a reference group smaller and more centered in values, possibly exposing differences between the interpersonal goals of the bullying groups. Another valuable alteration to our study design would be adding a multiple-item part assessing
different behaviors considered bullying to the single item questions. This could have an impact on the prevalence rates and prevent or reduce underreporting, possibly making the assessment of the influence of interpersonal goals on the involvement in specific bullying roles more valid.
Finally, the option that these findings showed us that interpersonal goals might not have an impact on the bullying roles should be considered. This would lead us to look beyond the influence of interpersonal goals and investigate the relation between for example the social skills and bullying roles further. Future studies could, for instance, look into the social information processing models, empathy, and assertion of the different involved and
interventions could focus on improving for instance social skills and reducing biases in the SIP model.
Conclusion
The findings suggest that interpersonal goals might not be one of the determinants for belongingness to a bullying role – or, that we were not able to detect them with our used methods. Reassessing these findings in future research with the extended bullying roles and more sophisticated measures to address bullying behavior from different perspectives might produce more insight and more valid conclusions. It is also recommended to look beyond the interpersonal goals as an explanation and expand the research on, for example, the social skills of students in relation to bullying behaviors and roles. This more in-depth knowledge will enable us to – eventually – have a better understanding of the reasoning behind the enrollment of students in specific bullying roles. When that knowledge is gained, effective interventions could be developed targeting for instance the nature of interpersonal goals, teaching alternative behavior for reaching interpersonal goals, or increasing important social skills. Interventions should explicitly be designed and modified to the different bullying roles to create effective anti-bullying interventions for adolescents.
References
Adler, P. A., & Adler, P. (1998). Peer power: Preadolescent culture and identity. New Brunswick, NJ: Rutgers University Press.
Aggarwal C. C. (2015). Data mining. doi:10.1007/978-3-319-14142-8_8
Arseneault, L., Bowes, L., & Shakoor, S. (2010). Bullying victimization in youths and mental health problems: ‘Much ado about nothing’? Psychological Medicine, 40(5), 717-729. doi:10.1017/S0033291709991383
Caravita, S. C. S., & Cillessen, A. H. N. (2012). Agentic or communal? Associations between interpersonal goals, popularity, and bullying in middle childhood and early
adolescence. Social Development, 21(2), 376-395. doi:10.1111/j.1467- 9507.2011.00632.x Centraal Bureau voor de Statistiek (n.d.). Wat verstaat het CBS onder een ‘allochtoon’?
Retrieved from https://www.cbs.nl/nl-nl/faq/specifiek/wat-verstaat-het-cbs-onder-een-allochtoon-
Crick, N. R. & Dodge, K. A. (1994). A review and reformulation of social information-processing mechanisms in children's social adjustment. Psychological Bulletin, 115(1), 74-101. doi:10.1037/0033-2909.115.1.74
Crick, N. R., & Dodge, K. A. (1996). Social information-processing mechanisms in reactive and proactive aggression. Child Development, 67(3), 993-1002. Retrieved from
https://srcd.onlinelibrary.wiley.com/journal/14678624
Crick, N. R., & Dodge, K. A. (1999). ‘Superiority’ is in the eye of the beholder: A comment on Sutton, Smith, and Swettenham. Social Development, 8(1), 128-131. doi:10.1111/1467-9507.00084
de Bruyn, E. H., Cillessen, A. H. N., & Wissink, I. B. (2010). Associations of peer acceptance and perceived popularity with bullying and victimization in early adolescence. The Journal
De Limburger (2017). Hoe ‘een lieve jongen’ als Tharukshan (15) voor de dood koos. Retrieved from https://www.limburger.nl/cnt/dmf20170624_00042346/hoe-een-lieve-jongen-als-tharukshan-voor-de-dood-koos
de Vocht, A. (2014). Basishandboek SPSS 21: IBM SPSS Statistics 21. Utrecht, Utrecht: Bijleveld Press.
EenVandaag (2019). Zelfdoding Hilversumse Marwan (15): Ouders wisten niks van jarenlange pesterijen. Retrieved from https://eenvandaag.avrotros.nl/item/zelfdoding-hilversumse-marwan-15-ouders-wisten-niks-van-jarenlange-pesterijen/
Elliott, S. N., Hwang, Y.-S., & Wang, J. (2019). Teachers' ratings of social skills and problem behaviors as concurrent predictors of students' bullying behavior. Journal of Applied
Developmental Psychology, 60, 119-126. doi:10.1016/j.appdev.2018.12.005
Felix, E. D., Sharkey, J. D., Green, J. G., Furlong, M. J., & Tanigawa, D. (2011). Getting precise and pragmatic about the assessment of bullying: The development of the California Bullying Victimization Scale. Aggressive Behavior, 37(3), 234-247. doi:10.1002/ab.20389 Fischer, R., & Milfont, T. L. (2010). Standardization in psychological research. International
Journal of Psychological Research, 3(1), 88-96. doi:10.21500/20112084.852
Gaffney, H., Farrington, D. P., & Ttofi, M. M. (2019). Examining the effectiveness of school-bullying intervention programs globally: A meta-analysis. International Journal of
Bullying Prevention, 1(1), 14-31. doi:10.1007/s42380-019-0007-4
Garandeau, C. F., Lee, I. A., & Salmivalli, C. (2013). Inequality matters: Classroom status hierarchy and adolescents’ bullying. Journal of Youth and Adolescence, 43(7), 1123- 1133. doi:10.1007/s10964-013-0040-4
Garandeau, C. F., Lee, I. A., & Salmivalli, C. (2014). Differential effects of the KiVa anti-bullying program on popular and unpopular bullies. Journal of Applied Developmental
Gini, G., & Pozzoli, T. (2009). Association between bullying and psychosomatic problems: A meta-analysis. Pediatrics, 123(3), 1059-1065. doi:10.1542/peds.2008-1215
Goldbach, J. T., Sterzing, P. R., & Stuart, M. J. (2018). Challenging conventions of bullying thresholds: exploring differences between low and high levels of bully-only, victim-only, and bully-victim roles. Journal of Youth and Adolescence, 47(3), 586-600.
doi:10.1007/s10964-017-0775-4
Green, J. G., Felix, E. D., Sharkey, J. D., Furlong, M. J., & Kras, J. E. (2013). Identifying bully victims: Definitional versus behavioral approaches. Psychological Assessment, 25(2), 651. doi:10.1037/a0031248
Guy, A., Lee, K., & Wolke, D. (2017). Differences in the early stages of social information processing for adolescents involved in bullying. Aggressive Behavior, 43(6), 578-587. doi:10.1002/ab.21716
Guy, A., Lee, K., & Wolke, D. (2019). Comparisons between adolescent bullies, victims, and bully-victims on perceived popularity, social impact, and social preference. Frontiers in
Psychiatry, 10, 1-10. doi:10.3389/fpsyt.2019.00868
Haynie, D. L., Nansel, T., Eitel, P., Crump, A. D., Saylor, K., Yu, K., & Simons-Morton, B. (2001). Bullies, victims, and bully/victims: Distinct groups of at-risk youth. The Journal of
Early Adolescence, 21(1), 29-49. doi:10.1177/0272431601021001002
Holt, M. K., Vivolo-Kantor, A. M., Polanin, J. R., Holland, K. M., DeGue, S., Matjasko, J. L., ... Reid, G. (2015). Bullying and suicidal ideation and behaviors: A
meta-analysis. Pediatrics, 135(2), e496-e509. doi:10.1542/peds.2014-1864
Huitsing, G., Lodder, G. M. A., Browne, W. J., Oldenburg, B., van der Ploeg, R., & Veenstra, R. (2020). A large-scale replication of the effectiveness of the KiVa Antibullying Program: A randomized controlled trial in the Netherlands. Prevention Science, 1-12.
Jarvinen, D. W., & Nicholls, J. G. (1996). Adolescents' social goals, beliefs about the causes of social success, and satisfaction in peer relations. Developmental Psychology, 32(3), 435-441. doi:10.1037/0012-1649.32.3.435
Jetelina, K. K., Gonzalez, J. M. R., Cuccaro, P. M., Peskin, M. F., Pompeii, L., Atem, F., ... Schuster, M. A. (2019). Self-reporting discrepancies of bullying victimization and perpetration measures. Annals of Epidemiology, 32, 58-63.
doi:10.1016/j.annepidem.2019.01.008
Klomek, A. B., Marrocco, F., Kleinman, M., Schonfeld, I. S., & Gould, M. S. (2007). Bullying, depression, and suicidality in adolescents. Journal of the American Academy of
Child & Adolescent Psychiatry, 46(1), 40-49. doi:10.1097/01.chi.0000242237.84925.18
Koyanagi, A., Oh, H., Carvalho, A. F., Smith, L., Haro, J. M., Vancampfort, D., … DeVylder, J. E. (2019). Bullying victimization and suicide attempt among adolescents aged 12–15 years from 48 countries. Journal of the American Academy of Child & Adolescent
Psychiatry, 58(9), 907-918. doi:10.1016/j.jaac.2018.10.018
LaFontana, K. M., & Cillessen, A. H. N. (2010). Developmental changes in the priority of perceived status in childhood and adolescence. Social Development, 19(1), 130-147. doi:10.1111/j.1467-9507.2008.00522.x
Lam, D. O. B., & Liu, A. W. H. (2007). The path through bullying—A process model from the inside story of bullies in Hong Kong secondary schools. Child and Adolescent Social
Work Journal, 24(1), 53-75. doi:10.1007/s10560-006-0058-5
Lereya, S. T., Copeland, W. E., Zammit, S., & Wolke, D. (2015). Bully/victims: A longitudinal, population-based cohort study of their mental health. European Child &
Locke, K. D. (2010). Circumplex scales of interpersonal values: Reliability, validity, and applicability to interpersonal problems and personality disorders. Journal of Personality
Assessment, 75(2), 249 –267. doi:10.1207/S15327752JPA7502_6
Locke, K. D. (2015). Agentic and communal social motives. Social and Personality
Psychology Compass, 9(10), 525-538. doi:10.1111/spc3.12201
Lusk, E. J., Halperin, M., & Heilig, F. (2011). A note on power differentials in data
preparation between trimming and winsorizing. Business Management Dynamics, 1(2), 23-31. Retrieved from https://www.bmdynamics.com/
Midgett, A., & Doumas, D. M. (2019). Witnessing bullying at school: The association between being a bystander and anxiety and depressive symptoms. School Mental
Health, 11(3), 454-463. doi:10.1007/s12310-019-09312-6
Mitsopoulou, E., & Giovazolias, T. (2015). Personality traits, empathy and bullying behavior: A meta-analytic approach. Aggression and Violent Behavior, 21, 61-72.
doi:10.1016/j.avb.2015.01.007
Moore, S. E., Norman, R. E., Suetani, S., Thomas, H. J., Sly, P. D., & Scott, J. G. (2017). Consequences of bullying victimization in childhood and adolescence: A systematic review and meta-analysis. World Journal of Psychiatry, 7(1), 60-76.
doi:10.5498/wjp.v7.i1.60
Nansel, T. R., Overpeck, M., Pilla, R. S., Ruan, W. J., Simons-Morton, B., & Scheidt, P. (2001). Bullying behaviors among US youth: Prevalence and association with psychosocial adjustment. JAMA, 285(16), 2094-2100. doi:10.1001/jama.285.16.2094
Navarro, D., & Foxcroft, D. (2018). Learning statistics with jamovi: A tutorial for psychology
Nederlandse Omroep Stichting (2017). Jongeren die in 2017 zelfmoord pleegden vaker verstrikt in web van zorg. Retrieved from https://nos.nl/artikel/2318851-jongeren-die-in-2017-zelfmoord-pleegden-vaak-verstrikt-in-web-van-zorg.html
Ojanen, T., Grönroos, M., & Salmivalli, C. (2005). An interpersonal circumplex model of children's social goals: Links with peer-reported behavior and sociometric
status. Developmental Psychology, 41(5), 699-710. doi:10.1037/0012-1649.41.5.699 Olsen, C., & George, D. M. M. (2004). Cross-sectional study design and data analysis.
Retrieved from
https://www.physioblasts.org/f/public/1355667773_1_FT0_4297_module_05.pdf
Olweus, D. (1994). Bullying at school: Basic facts and effects of a school based intervention program. Journal of Child Psychology and Psychiatry, 35(7), 1171-1190.
doi:10.1111/j.1469-7610.1994.tb01229.x
Olweus, D. (2013). School bullying: Development and some important challenges. Annual
Review of Clinical Psychology, 9, 751-780. doi:10.1146/annurev-clinpsy-050212-185516
Onderwijs in cijfers (2018). Pesten op school. Retrieved from https://www.onderwijsincijfers.nl/themas/pesten-op-school
Orobio de Castro, B., Mulder, S., van der Ploeg, R., Onrust, S., van den Berg, Y., Stoltz, S., ... Scholte, R. (2018). Wat werkt tegen pesten? Effectiviteit van kansrijke programma’s tegen pesten in de Nederlandse onderwijspraktijk. Retrieved from
https://www.leraar24.nl/app/uploads/Watwerkt_Pesten.pdf
Pellegrini, A. D. (2002). Affiliative and aggressive dimensions of dominance and possible functions during early adolescence. Aggression and Violent Behavior, 7(1), 21-31. doi:10.1016/S1359-1789(00)00033-1
Petrucci, C. J. (2009). A primer for social worker researchers on how to conduct a multinomial logistic regression. Journal of social service research, 35(2), 193-205. doi:10.1080/01488370802678983
Postigo, S., González, R., Mateu, C., & Montoya, I. (2012). Predicting bullying:
maladjustment, social skills and popularity. Educational Psychology, 32(5), 627-639. doi:10.1080/01443410.2012.680881
Reijntjes, A., Vermande, M., Goossens, F. A., Olthof, T., van de Schoot, R., Aleva, L., & van der Meulen, M. (2013). Developmental trajectories of bullying and social dominance in youth. Child Abuse & Neglect, 37(4), 224-234. doi:10.1016/j.chiabu.2012.12.004 Rose, A. J., & Rudolph, K. D. (2006). A review of sex differences in peer relationship
processes: Potential trade-offs for the emotional and behavioral development of girls and boys. Psychological Bulletin, 132(1), 98-131. doi:10.1037/0033-2909.132.1.98
Salmivalli, C. (1999). Participant role approach to school bullying: Implications for interventions. Journal of Adolescence, 22(4), 453-459. doi:10.1006/jado.1999.0239 Salmivalli, C., & Peets, K. (2009). Bullies, victims, and bully-victim relationships in middle
childhood and early adolescence. In K. H. Rubin, W. M. Bukowski, & B. Laursen (Eds.), Social, emotional, and personality development in context. Handbook of peer interactions, relationships, and groups (p. 322–340). The Guilford Press.
Schreiber-Gregory, D. N., Jackson, H. M., & Bader, K. S. (2018). Logistic and Linear Regression Assumptions: Violation Recognition and Control. Retrieved from
https://www.semanticscholar.org/paper/Logistic-and-Linear-Regression-Assumptions-
Sijtsema, J. J., Veenstra, R., Lindenberg, S., & Salmivalli, C. (2009). Empirical test of bullies' status goals: Assessing direct goals, aggression, and prestige. Aggressive Behavior, 35(1), 57-67. doi:10.1002/ab.20282
Social Science Statistics (n.d.). Effect size calculator for T-test. Retrieved from https://www.socscistatistics.com/effectsize/default3.aspx
Statistics Solutions (n.d.). Assumptions of logistic regression
https://www.statisticssolutions.com/assumptions-of-logistic-regression/ Sung, Y.-H., Chen, L.-M., Yen, C.-F., & Valcke, M. (2018). Double trouble: The
developmental process of school bully-victims. Children and Youth Services Review, 91, 279-288. doi:10.1016/j.childyouth.2018.06.025
Thornberg, R., Halldin, K., Bolmsjö, N., & Petersson, A. (2013). Victimising of school bullying: A grounded theory. Research Papers in Education, 28(3), 309-329.
doi:10.1080/02671522.2011.641999
Trucco, E. M., Wright, A. G. C., & Colder, C. R. (2014). Stability and change of social goals in adolescence. Journal of Personality, 82(5), 379-389. doi:10.1111/jopy.12069
UNESCO (2019). Behind the numbers: Ending school violence and bullying. Retrieved from https://unesdoc.unesco.org/ark:/48223/pf0000366483
van Geel, M., Goemans, A., & Vedder, P. H. (2016). The relation between peer victimization and sleeping problems: A meta-analysis. Sleep Medicine Reviews, 27, 89-95.
doi:10.1016/j.smrv.2015.05.004
Vliek, L., Overbeek, G., & Orobio de Castro, B. (2019). Effects of Topper Training on psychosocial problems, self-esteem, and peer victimisation in Dutch children: A randomised trial. PLoS One, 14(11), 1-22. doi:10.1371/journal.pone.0225504
Ziv, Y., Leibovich, I., & Shechtman, Z. (2013). Bullying and victimization in early adolescence: Relations to social information processing patterns. Aggressive