AGGRESSIVE BEHAVIOR Volume 41, pages 310–321 (2015)
Bullying in Preschool: The Associations Between Participant Roles, Social Competence, and Social Preference Marina Camodeca1*, Simona C.S. Caravita2, and Gabrielle Coppola1 Department of Neurosciences, Imaging, and Clinical Sciences, University “G. d’Annunzio”, Chieti‐Pescara, Italy Department of Psychology, Center for Research in Developmental and Educational Dynamics (C.R.I.d.e.e.), Catholic University of the Sacred Heart, Milan and Brescia, Italy
1 2
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. The different roles of bullying participation (bully, follower, victim, defender of the victim, and outsider) have not been investigated in preschool children. The aims of this study were to use a peer‐report measure to assess these roles and to investigate their associations with social competence among pre‐schoolers. We also explored whether status among peers, indicated by being socially preferred, mediates the relationship between social competence and bullying roles. Three hundred twenty 3‐ to 6‐year‐old children participated in the study. Bullying roles and social preference were assessed by means of peer reports, whereas social competence was investigated with a Q‐Sort methodology, based on observations in classrooms. Bullying was also assessed by means of teacher reports. The results showed quite a clear distinction among roles and a correspondence between peer and teacher assessments, except for the role of outsider. The role of defender was positively associated with social competence, whereas the other roles were negatively associated. In a subsample, social preference statistically predicted the role of bully and mediated between social competence and bullying. The findings are discussed in terms of the importance of assessing bullying and its correlates at a very young age, although roles may further develop when children grow up. Aggr. Behav. 41:310–321, 2015. © 2014 Wiley Periodicals, Inc.
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. Keywords: bullying; preschool; social competence; social preference; peer nominations
INTRODUCTION
Bullying at school is usually considered to be a widespread problem among children, preadolescents, and adolescents (Smith et al., 1999). It is usually defined as a form of aggression in which an individual (or a group of individuals), systematically and over time, hurts, harasses, humiliates, or excludes someone who is weaker or less powerful (Salmivalli, 2010). In particular, imbalance of power between bullies and victims and repetition over time are key factors in distinguishing bullying in comparison to general peer‐aggression (Ybarra, Espelage, & Mitchell, 2014). Although aggression and victimization are also daily experiences among young children, it may be questionable whether they show the specific pattern of bullying in early childhood. However, there are a number of recent claims that bullying is also found in preschool years (Alsaker & Gutzwiller‐Helfenfinger, 2010; Alsaker & Valkanover, 2001; Kochenderfer & Ladd, 1996; Monks, Ruiz, & Val, 2002; Monks, Palermiti, Ortega, & Costabile, 2011; Monks, Smith, & Swettenham, 2003, 2005; for a review Vlachou, Andreou, Botsoglou, & Didaskalou, 2011). The findings show that preschool bullying is more stable than victimization and characterized
by physical strength, even when verbal forms are also very common (Monks et al., 2003, 2011). Children bullied by aggressive peers have high scores on loneliness, school avoidance, and school dislike (Kochenderfer & Ladd, 1996); they lack leadership skills and have few friends (Alsaker & Gutzwiller‐Helfenfinger, 2010). Although the aforementioned works provide strong contributions to the study of bullying in the early years, they have mainly focused on the roles of bully and victim. Bullying cannot be fully understood if we do not consider the whole context of the classroom, where roles develop and contribute to bullying maintainability or limitation (Salmivalli, 2010). School children and adolescents can play specific roles in the bullying process: In addition to
Correspondence to: Marina Camodeca, Department of Neurosciences, Imaging, and Clinical Sciences, University “G. d’Annunzio”, Chieti‐ Pescara, via dei Vestini 31, Blocco A Psicologia, 66013 Chieti (CH), Italy. E‐mail: [email protected] Received 17 January 2014; Accepted 28 April 2014 DOI: 10.1002/ab.21541 Published online 28 May 2014 in Wiley Online Library (wileyonlinelibrary.com).
© 2014 Wiley Periodicals, Inc.
Bullying Roles and Social Competence in Pre‐Schoolers
the main roles of bully and victim, there are those of defender of the victim (who helps the victimized peers), follower of the bully (who supports the ringleader bully), and outsider (who pretends not to see or shies away). These roles have been found for children older than 7 and adolescents (Salmivalli, Lagerspetz, Björkqvist, Österman, & Kaukiainen, 1996; Sutton & Smith, 1999). Very little research has investigated bullying roles in the preschool, so that evidence that these roles occur at this age is scarce. Belacchi and Farina (2010) pointed to the existence, among pre‐schoolers, of the roles of pro‐bully, victim, defender, and outsider, and Monks and colleagues studied followers (Monks et al., 2002) and defenders (Monks et al., 2002, 2003, 2005, 2011), alongside bullies and victims. Belacchi and Farina (2010) employed teacher reports, which seemed to be a valid methodology for young children, since preschool teachers know and observe their pupils and the intra‐class dynamics better that those teaching older ages. However, it is possible that some episodes, as well as also relational forms of aggression, happen out of their sight or when they are not present (cf. Pellegrini & Bartini, 2000). In contrast, peer evaluations provide many opinions on each child, minimize social desirability, and have been found to be one of the best ways to assess bullying and roles, given that classmates are mostly informed about what happens among peers (Pellegrini & Bartini, 2000). Belacchi and Farina (2010) also suggested the value of testing a peer‐ nomination version of this instrument in order to verify whether peer nominations yield comparable results with the original roles model applied to older children (Salmivalli et al., 1996; Sutton & Smith, 1999). Peer reports have been found to reliably assess aggression and prosocial behavior even in kindergarten (Crick, Casas, & Mosher, 1997). Monks et al. (2002, 2003, 2005, 2011) assessed bullying among pre‐schoolers by means of peer reports, often in combination with teacher and self‐reports. They used a cartoon procedure to present children with a bullying episode and asked them about which classmates played different roles. The role of followers was not as identifiable as those of bullies, victims, and defenders (Monks et al., 2002). Moreover, the authors did not detect “outsiders,” that is, those who specifically remain indifferent or pretend not to be involved, and they only considered as “bystanders” those children scoring low on the other scales (Monks et al., 2002, 2011). Therefore, further research is needed on the applicability of the bullying roles model to the preschool years. As suggested by Vlachou et al. (2011), research on preschool bullying should focus on developing reliable and valid instruments, on providing their psychometric properties, and on considering the group context of
311
bullying, especially examining the peripheral roles of defender, follower, and outsider. Our first aim was to investigate whether bullying roles were applicable to preschool children, using a peer report form of the instrument developed by Belacchi and Farina (2010). Peer evaluations were compared with teacher reports, and gender and age differences were analyzed. Children in various roles behave and relate to each other in different ways. Therefore, a salient dimension to discriminate roles may be social competence, which is the capacity to develop positive interactions with others and to achieve one’s own goals without hurting others. It embodies several cognitive, emotional, and behavioral skills, relevant for adequate child adjustment within social contexts (Vaughn et al., 2009; Waters & Sroufe, 1983). The entrance into preschool is for many children an important developmental step, because it allows them to interact with adults other than their parents and, especially, other peers, and to become members of stable peer‐networks. Therefore, since early childhood, social competence is an important protective factor for children’s adjustment, helping them to engage in shared activities, to adapt to different situations, to develop friendships, to respect turns, to regulate emotionality, to collaborate, to show prosocial behavior, to acquire a good reputation among peers, and to be accepted by peers (Denham et al., 2003; Eisenberg & Fabes, 1998; Rubin, Bukowski, & Parker, 1998). Although bullying and victimization may develop from a failure in the process of acquiring social skills, social competence in association with bullying has not been deeply investigated at this early age. To the best of our knowledge, only studies testing some behavioral correlates of social competence are available for pre‐ schoolers. For instance, Perren and Alsaker (2006) found that bullies obtained low scores on cooperativeness and prosociality, but high scores on leadership, whereas victims were not sociable, had the lowest scores on leadership, and the highest on submissiveness, isolation, and withdrawal. In slightly older samples, victims were also found to be low on prosociality, and high in internalizing problems and in emotional symptoms (Arseneault et al., 2006; Malti, Perren, & Buchmann, 2010). Another correlate of social competence, emotion comprehension (Denham et al., 2003), has been found to be positively associated with the defender role and negatively with the outsider and victim roles (Belacchi & Farina, 2010). Defenders also had high scores on theory of mind tasks (Monks et al., 2005). Findings from studies with school children and adolescents are more numerous and indicate that victims can be less competent than peers in interacting with others (cf. Camodeca & Goossens, 2005; Caravita, Di Blasio, & Salmivalli, 2010; Perren & Alsaker, 2006), whereas contrasting findings are available for bullies. Aggr. Behav.
312
Camodeca et al.
Although some studies suggest that bullies lack affective empathy (Caravita, Di Blasio, & Salmivalli, 2009) and the necessary social cognitive skills to positively interact with others (Camodeca & Goossens, 2005), other studies have found that bullies are able to correctly understand social situations and, thus, can be able to achieve goals by means of intelligent manipulation (Caravita et al., 2010; Gini, 2006; Sutton, Smith, & Swettenham, 1999). Followers have been found to show low social cognitive and mind‐reading abilities (Camodeca & Goossens, 2005; Sutton et al., 1999), whereas defenders have been found to be well‐liked among peers, prosocial, empathic, and morally competent (Caravita et al., 2009, 2010; Gini, Pozzoli, & Hauser, 2011; Menesini & Camodeca, 2008; Pozzoli & Gini, 2010; Salmivalli et al., 1996). Results concerning outsiders are less clear, with some studies showing that their social and moral skills are similar to those of defenders (Camodeca & Goossens, 2005; Gini, 2006; Olthof, 2012), and others suggesting low responsibility, morality, and self‐efficacy (Gini, Albiero, Benelli, & Altoè, 2008; Pozzoli & Gini, 2010), which may account for their silent support for bullies. A possible explanation of inconsistent results, especially for the roles of bullies and outsiders, may lay in the fact that the aforementioned studies used different definitions of social competence or focused on specific indicators of social competence. To avoid this problem, and to fill the gap of paucity of studies on social competence and bullying in young children, we explored social competence in relation to bullying roles by employing the theoretical model of social competence as a hierarchically organized construct (Bost, Vaughn, Washington, Cielinski, & Bradbard, 1998; Shin et al., 2011; Vaughn et al., 2009; Waters & Sroufe, 1983). According to this integrative approach, social competence is understood to be the organization of young children’s behavior, affect, and cognition at a higher level of psychological abstraction. In line with the literature embracing this view, we used two Q‐Sort instruments, which are culture‐specific, and allow the testing of a very large quantity of characteristics, taking into account the multidimensional nature of social competence (Vaughn et al., 2009). This measure has been found to be particularly useful for preschool children (Coppola & Camodeca, 2010). Social competence, which includes the ability to interact appropriately with others, to cooperate, and achieve goals in a positive way, shows high correlations with social status in the peer group (Coppola & Camodeca, 2010; Denham et al., 2003; Vaughn et al., 2009). Children with low social status (measured by social preference) have a greater risk of developing internalizing and (especially) externalizing problems (McDougall, Hymel, Vaillancourt, & Mercer, 2001; Aggr. Behav.
Prinstein, Rancourt, Guerry, & Brown, 2009). Studies on older age groups show that, although bullies may have a high status among peers because they are visible and powerful, they are usually low in social preference (Caravita et al., 2009, 2010; de Bruyn, Cillessen, & Wissink, 2010). Followers and victims are often rejected by their peers, whereas defenders are usually the most liked, and bystanders show a social status that is below average both in social acceptance and rejection, indicating that they are neglected by their peers (e.g., Caravita et al., 2009, 2010; de Bruyn et al., 2010; Goossens, Olthof, & Dekker, 2006; Salmivalli et al., 1996). Similar associations between social preference and bullying roles have been found in preschool children (Monks et al., 2002, 2003, 2011; Perren & Alsaker, 2006), although studies with young children are still scarce. The second aim of the present study is therefore to analyze the relationships among social competence, social preference, and roles in bullying among pre‐ schoolers. We expect social competence to be directly and positively associated with the role of defender, and negatively with the roles of bully, follower, and victim (Camodeca & Goossens, 2005; Caravita et al., 2009, 2010; de Bruyn et al., 2010; Pozzoli & Gini, 2010). Given the previous inconsistent results, we did not formulate any hypotheses for the role of outsiders. Within a sub‐sample, we further investigate whether social competence is indirectly related to bullying roles via social preference. We expect that being liked by peers is associated with social competence as assessed by Q‐methodology (Bost et al., 1998; Santos, Peceguina, João, Shin, & Vaughn, 2013). Based on findings with older children, we also expect that social preference is positively associated with defending behavior and negatively with bullying, victimization, and follower behavior. We also expect that it mediates the relationship between social competence and bullying roles. Social preference could be a way through which social competence influences roles, but, given that no studies are available on indirect influence in preschool children, we cannot formulate clearer hypotheses about the strength and directions of these associations. In sum, the present work aims to: (1) test a peer nomination instrument to detect bullying roles in pre‐ schoolers; (2) investigate the social competence of children involved in bullying with different roles and analyze its indirect effects on bullying roles through social preference. METHOD
Sample and Procedure Participants were 320 preschool children (151 boys and 169 girls), with an average age of 58.40 months
Bullying Roles and Social Competence in Pre‐Schoolers
313
(SD ¼ 11.74; range: 31–79 months), attending three kindergartens (13 classrooms) in central Italy and one kindergarten (6 classrooms) in northern Italy. This latter sub‐sample, consisting of 78 children (45 girls and 33 boys, age: 52–79 months, M ¼ 65.69; SD ¼ 7.25), was also assessed on social preference. The purpose and the methodology of the study were explained to the schools’ principals and teachers, who gave their consent. In one school, parents were informed about the study as one of the activities of an internal project, to which they had previously given their consent. In the other schools, parents received a letter to illustrate the research and were asked to sign if they allowed the participation of their children. About 12% of the contacted families did not allow participation of their children in the study. Graduates in Psychology, appropriately trained in interview and observation methods, administered the interview individually to assess bullying roles and social preference, and conducted the observations in the classroom in order to obtain the Q‐Sort scores. Teachers filled in the questionnaire for bullying for each pupil in their classroom.
TABLE I. Items Content of the PRQ‐P
Measures Participant roles questionnaire for teachers (PRQ‐T). This questionnaire was developed by Belacchi and Farina (2010) on the basis of the original Participant Roles Questionnaire (PRQ; Salmivalli et al., 1996). It aims to assess eight roles by means of teachers’ evaluations of children’s behavior, employing three items per role. As we were only interested in five roles (bully, follower of the bully, defender of the victim, outsider, and victim), we employed a shortened version of the original measure. Teachers were asked to rate the frequency with which each child in their classroom behaved on 15 items, using a 5‐point Likert scale from 0 to 4 (never, rarely, sometimes, often, and always). Psychometric characteristics of the questionnaire, together with a CFA to confirm the 5‐factor structure, are described in the Results Section. Participant roles questionnaire for pre‐schoolers (PRQ‐P). The PRQ based on peer nominations was developed for children older than 7 (Salmivalli et al., 1996; Sutton & Smith, 1999) and allows assessment of the roles of participation in bullying. Belacchi and Farina (2010) adapted some of the PRQ items for use by preschool teachers (PRQ‐T; see above). On the basis of this previous work, we developed a version to be used by children in the form of peer nominations, including 10 items, two for each role. The items are reported in Table I. Items were read individually to each child, paying particular attention that the children understood them well (e.g., words were explained if children had difficulties in making sense; questions were formulated
more than once; etc.). Children were asked to indicate or nominate the classmates whose behavior fitted with the one described in the items. Children were presented with the photos of all their classmates, which is a common and reliable procedure for young children, who are not able to keep in mind all the peers in the class (cf. Vaughn et al., 2009). In a very few classes, photos were not available, but great care was taken to obtain reliable responses (e.g., children were asked to think about all their classmates, even those not present that day; children were given more time to think; etc.). Item scores were the sum of the received nominations, which were standardized by class, in order to control for classroom size. Psychometric characteristics of the questionnaire are described in the Results Section. Q‐Sort descriptions for social competence. Two Q‐Sets assessing social competence were used: the California Child Q‐Set (CCQ; Block & Block, 1980; Waters, Noyes, Vaughn, & Ricks, 1985), a 100‐item set consisting of personality and behavior descriptive items of preschool children, and the Preschool Q‐Set (PQ; Waters et al., 1985), a 72‐item set consisting of items related to social competence in preschool age. Both instruments have a strong theoretical basis and have been successfully implemented to assess social competence in preschool age (e.g., Bost et al., 1998; Santos et al., 2013; Vaughn, 2001; Vaughn et al., 2009). A total of five trained observers separately conducted about 20 hr of observation in the classroom during 1 week, completing at least 1 hr and a half of observations per each child. They then described each child by sorting the two Q‐Sets into a
Scale Bullying
Follower of the bully
Defender of the victim
Outsider
Victim
Items 1. Makes fun of his/her classmates and makes them cry 5. Does nasty tricks and says bad things to other children 3. Offends someone because other schoolmates do it 7. Encourages those who tease and irritate other classmates 2. Defends the children who are being beaten or teased 10. Tries to console the children who cry because they have been beaten or isolated 4. Pretends not to see when someone bullies someone else 8. Remains indifferent if a child is being teased or left alone 6. The other children beat him/her and ruin his/her things 9. The other children isolate him/her—keep him/her apart—or tell him/her bad things
Note. The items number corresponds to the number in the questionnaire.
Aggr. Behav.
314
Camodeca et al.
predetermined rectangular distribution of nine categories (q‐profile), on the basis of the similarity of each item to the children’s behavior. The score assigned to each item is its placement in the distribution. By convention, the child’s social competence score is the Pearson correlation, ranging between 1 and þ1, between his/her q‐profile and a q‐criterion, which is a composite description obtained by averaging the scores assigned by experts to each item. In the present study, for each Q‐Set, we implemented an Italian Social Competence Criterion, describing the hypothetical most socially competent Italian pre‐schooler, generated by averaging the descriptions provided by Italian experts in children’s social development (Coppola & Camodeca, 2010). Although these two criteria are highly similar to the US ones reported by Waters et al. (1985), that is, r ¼ .89 for the CCQ and r ¼ .88 for the PQ, they are also sensitive to cultural specificities and have been shown to be valid and reliable in the assessment of social competence at preschool age (Coppola & Camodeca, 2010). One hundred sixty‐one children (50.3% of the total sample) were observed by two trained observers: Cross‐rater agreement on criterion scores was satisfactory (r ¼ .70, for the CCQ, and r ¼ .60, for the PQ) and in line with those reported elsewhere (Santos et al., 2013; Vaughn et al., 2009). Peer nominations for social preference. Social preference was assessed by asking the 78 children from the sub‐sample to nominate up to three classmates they liked most (preferences) and three that they liked least (refusals), a method derived from McCandless and Marshall (1957). They were also presented with pictures of all their classmates, to aid their recollection of members of the group. After standardizing the nominations for preferences and refusals by class, refusals were subtracted from preferences to obtain the total score for social preference. As recommended by Cillessen (2009), this combined score was further standardized by class (i.e., the reference group). Strategy of Analysis Data were analyzed by means of Structural Equation Modelling (Mplus 7.0; Muthén & Muthén, 1998–2012). The robust method of maximum likelihood estimation (MLR estimation) was employed, in order to control for the non‐normality of the variables’ distribution. The MLR estimation provides a scaled Chi‐square (Satorra & Bentler, 1994). For the models including the test of indirect effects the bias‐corrected bootstrap was used as recommended strategy (MacKinnon, 2008; MacKinnon, Lockwood, & Williams, 2004). The boostrapping procedure is also a strategy proposed for the test of models when the sample size is small (Adèr, Mellenbergh, & Hand, 2008). It estimates the model parameters, along Aggr. Behav.
with their standard errors, strictly from the study sample. In this procedure, the study sample is considered as a population from which several samples are drawn by replacement. Then, estimates of the expected values and the variability of the statistics are taken from this sampling distribution. According to the bootstrapping method, an effect is significant at the .05 and .01 levels, if the value of 0 is not part of, respectively, the 95% and 99% bootstrap confidence intervals around the effect. Mplus does allow the computation of bias‐corrected bootstrap effect confidence intervals at the .05, .01, and .10 (marginal significance) level, and we tested the models including indirect effects by drawing 2000 samples (a minimum number of 500 samples is recommended). In Mplus, when models are tested by bootstrapping, the use of the MLR estimation is not possible, and the Maximum Likelihood (ML) estimation is employed as the default option. The use of ML in place of the MLR estimation affects the estimates of standard errors but not the estimates of the parameters. Therefore, the estimator ML was employed for the models tested by bootstrapping. Goodness of fit of the model was evaluated using the following: Chi‐square statistics and the ratio x2/df, which is considered acceptable if lower than 3 (Bollen, 1989); the Comparative Fit Index (CFI), which provides acceptable fit when it is equal or higher than .90 and good when equal or higher than .95 (Hu & Bentler, 1999); and the Root Mean Square Error of Approximation (RMSEA), which is judged adequate when equal or lower than .08 and good when equal or lower than .05 (Hu & Bentler, 1999). Modification indices were also considered for the improvement of the models. Because Mplus does not provide modification indices for the models tested by bootstrapping, those models were re‐ run without using the bootstrapping method and using the MLR estimation in order to obtain the modification indices. Possible improvement of those models was tested when the suggestions from the modification indices were in accordance with the suggestions from derivatives. Lastly, a post‐hoc model trimming procedure was also used in order to further optimize the models. That is, the specified paths that resulted non‐significantly different from 0 (Wald test) at the first test of the model were fixed to 0 in the model specification. These modifications of the conceptual model were made when removing the paths was theoretically justifiable, and they were tested one by one, in order to ensure that each of these modifications resulted in a non‐significant increase of the fit of the model [Chi‐square difference test, which was computed with the corrected formula for the robust chi square test by Satorra (2000) when the MLR estimation was used]. Therefore, the most parsimonious models were our final models.
Bullying Roles and Social Competence in Pre‐Schoolers RESULTS
Adaptation of the PRQ‐P: CFA, Reliabilities, and Associations With PRQ‐T In order to test whether the distinction between different bullying roles could be applied to pre‐schoolers and whether peer reports are a valid instrument to assess roles, we conducted a Confirmatory Factor Analysis on the 10 items of the PRQ‐P, which were allowed to load on the expected five factors; other loadings were fixed to 0. Fit indices for the model were satisfactory (x2(25) ¼ 35.67, P ¼.08; x2/df ¼ 1.43; CFI ¼ .98; RMSEA ¼ .037). All items loaded on the expected factors and standardized factor loadings were higher than .42 (Ps < .01). All roles were inter‐correlated with each other, with the exception of the defender role that was significantly (positively) associated only with the roles of victim and outsider. Although reliabilities were good for the scales of bully and follower (a ¼ .85 and a ¼ .71, respectively), they were very low for the other scales (defender scale a ¼ .49, outsider scale a ¼ .37, victim scale a ¼ .36), especially for the outsider and victim roles. Based on these low values and on the suggestions from the modification indices, we tested a second model in which the roles of outsider and victim were each assessed by only one item, that is, the item with the highest loading on the respective factor (item 8 for the outsider role and item 9 for the victim role). The model had good fit indices (x2(13) ¼ 21.06, P ¼.07; x2/df ¼ 1.62; CFI ¼ .99; RMSEA ¼ .044), and we retained this model as the final one. Table II displays means, standard deviations, and items loadings of the five roles in the final model, together with correlations among the scale factors (below the diagonal). In this model, the associations between roles that were previously found were confirmed, with the exception of the positive associations of the defender role with the roles of victim and outsider, which became marginal, thus better distinguishing this role from the others. As second step of the analyses, we correlated the scales of the peer and teacher versions of the PRQ. Given that we used a shortened version of the instrument by Belacchi and Farina (2010), we run a CFA (MLR estimation method) to confirm the 5‐factor structure of the PRQ‐T. Fit indices were acceptable (x2(80) ¼ 175.40, P
Bullying in Preschool: The Associations Between Participant Roles, Social Competence, and Social Preference Marina Camodeca1*, Simona C.S. Caravita2, and Gabrielle Coppola1 Department of Neurosciences, Imaging, and Clinical Sciences, University “G. d’Annunzio”, Chieti‐Pescara, Italy Department of Psychology, Center for Research in Developmental and Educational Dynamics (C.R.I.d.e.e.), Catholic University of the Sacred Heart, Milan and Brescia, Italy
1 2
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. The different roles of bullying participation (bully, follower, victim, defender of the victim, and outsider) have not been investigated in preschool children. The aims of this study were to use a peer‐report measure to assess these roles and to investigate their associations with social competence among pre‐schoolers. We also explored whether status among peers, indicated by being socially preferred, mediates the relationship between social competence and bullying roles. Three hundred twenty 3‐ to 6‐year‐old children participated in the study. Bullying roles and social preference were assessed by means of peer reports, whereas social competence was investigated with a Q‐Sort methodology, based on observations in classrooms. Bullying was also assessed by means of teacher reports. The results showed quite a clear distinction among roles and a correspondence between peer and teacher assessments, except for the role of outsider. The role of defender was positively associated with social competence, whereas the other roles were negatively associated. In a subsample, social preference statistically predicted the role of bully and mediated between social competence and bullying. The findings are discussed in terms of the importance of assessing bullying and its correlates at a very young age, although roles may further develop when children grow up. Aggr. Behav. 41:310–321, 2015. © 2014 Wiley Periodicals, Inc.
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. Keywords: bullying; preschool; social competence; social preference; peer nominations
INTRODUCTION
Bullying at school is usually considered to be a widespread problem among children, preadolescents, and adolescents (Smith et al., 1999). It is usually defined as a form of aggression in which an individual (or a group of individuals), systematically and over time, hurts, harasses, humiliates, or excludes someone who is weaker or less powerful (Salmivalli, 2010). In particular, imbalance of power between bullies and victims and repetition over time are key factors in distinguishing bullying in comparison to general peer‐aggression (Ybarra, Espelage, & Mitchell, 2014). Although aggression and victimization are also daily experiences among young children, it may be questionable whether they show the specific pattern of bullying in early childhood. However, there are a number of recent claims that bullying is also found in preschool years (Alsaker & Gutzwiller‐Helfenfinger, 2010; Alsaker & Valkanover, 2001; Kochenderfer & Ladd, 1996; Monks, Ruiz, & Val, 2002; Monks, Palermiti, Ortega, & Costabile, 2011; Monks, Smith, & Swettenham, 2003, 2005; for a review Vlachou, Andreou, Botsoglou, & Didaskalou, 2011). The findings show that preschool bullying is more stable than victimization and characterized
by physical strength, even when verbal forms are also very common (Monks et al., 2003, 2011). Children bullied by aggressive peers have high scores on loneliness, school avoidance, and school dislike (Kochenderfer & Ladd, 1996); they lack leadership skills and have few friends (Alsaker & Gutzwiller‐Helfenfinger, 2010). Although the aforementioned works provide strong contributions to the study of bullying in the early years, they have mainly focused on the roles of bully and victim. Bullying cannot be fully understood if we do not consider the whole context of the classroom, where roles develop and contribute to bullying maintainability or limitation (Salmivalli, 2010). School children and adolescents can play specific roles in the bullying process: In addition to
Correspondence to: Marina Camodeca, Department of Neurosciences, Imaging, and Clinical Sciences, University “G. d’Annunzio”, Chieti‐ Pescara, via dei Vestini 31, Blocco A Psicologia, 66013 Chieti (CH), Italy. E‐mail: [email protected] Received 17 January 2014; Accepted 28 April 2014 DOI: 10.1002/ab.21541 Published online 28 May 2014 in Wiley Online Library (wileyonlinelibrary.com).
© 2014 Wiley Periodicals, Inc.
Bullying Roles and Social Competence in Pre‐Schoolers
the main roles of bully and victim, there are those of defender of the victim (who helps the victimized peers), follower of the bully (who supports the ringleader bully), and outsider (who pretends not to see or shies away). These roles have been found for children older than 7 and adolescents (Salmivalli, Lagerspetz, Björkqvist, Österman, & Kaukiainen, 1996; Sutton & Smith, 1999). Very little research has investigated bullying roles in the preschool, so that evidence that these roles occur at this age is scarce. Belacchi and Farina (2010) pointed to the existence, among pre‐schoolers, of the roles of pro‐bully, victim, defender, and outsider, and Monks and colleagues studied followers (Monks et al., 2002) and defenders (Monks et al., 2002, 2003, 2005, 2011), alongside bullies and victims. Belacchi and Farina (2010) employed teacher reports, which seemed to be a valid methodology for young children, since preschool teachers know and observe their pupils and the intra‐class dynamics better that those teaching older ages. However, it is possible that some episodes, as well as also relational forms of aggression, happen out of their sight or when they are not present (cf. Pellegrini & Bartini, 2000). In contrast, peer evaluations provide many opinions on each child, minimize social desirability, and have been found to be one of the best ways to assess bullying and roles, given that classmates are mostly informed about what happens among peers (Pellegrini & Bartini, 2000). Belacchi and Farina (2010) also suggested the value of testing a peer‐ nomination version of this instrument in order to verify whether peer nominations yield comparable results with the original roles model applied to older children (Salmivalli et al., 1996; Sutton & Smith, 1999). Peer reports have been found to reliably assess aggression and prosocial behavior even in kindergarten (Crick, Casas, & Mosher, 1997). Monks et al. (2002, 2003, 2005, 2011) assessed bullying among pre‐schoolers by means of peer reports, often in combination with teacher and self‐reports. They used a cartoon procedure to present children with a bullying episode and asked them about which classmates played different roles. The role of followers was not as identifiable as those of bullies, victims, and defenders (Monks et al., 2002). Moreover, the authors did not detect “outsiders,” that is, those who specifically remain indifferent or pretend not to be involved, and they only considered as “bystanders” those children scoring low on the other scales (Monks et al., 2002, 2011). Therefore, further research is needed on the applicability of the bullying roles model to the preschool years. As suggested by Vlachou et al. (2011), research on preschool bullying should focus on developing reliable and valid instruments, on providing their psychometric properties, and on considering the group context of
311
bullying, especially examining the peripheral roles of defender, follower, and outsider. Our first aim was to investigate whether bullying roles were applicable to preschool children, using a peer report form of the instrument developed by Belacchi and Farina (2010). Peer evaluations were compared with teacher reports, and gender and age differences were analyzed. Children in various roles behave and relate to each other in different ways. Therefore, a salient dimension to discriminate roles may be social competence, which is the capacity to develop positive interactions with others and to achieve one’s own goals without hurting others. It embodies several cognitive, emotional, and behavioral skills, relevant for adequate child adjustment within social contexts (Vaughn et al., 2009; Waters & Sroufe, 1983). The entrance into preschool is for many children an important developmental step, because it allows them to interact with adults other than their parents and, especially, other peers, and to become members of stable peer‐networks. Therefore, since early childhood, social competence is an important protective factor for children’s adjustment, helping them to engage in shared activities, to adapt to different situations, to develop friendships, to respect turns, to regulate emotionality, to collaborate, to show prosocial behavior, to acquire a good reputation among peers, and to be accepted by peers (Denham et al., 2003; Eisenberg & Fabes, 1998; Rubin, Bukowski, & Parker, 1998). Although bullying and victimization may develop from a failure in the process of acquiring social skills, social competence in association with bullying has not been deeply investigated at this early age. To the best of our knowledge, only studies testing some behavioral correlates of social competence are available for pre‐ schoolers. For instance, Perren and Alsaker (2006) found that bullies obtained low scores on cooperativeness and prosociality, but high scores on leadership, whereas victims were not sociable, had the lowest scores on leadership, and the highest on submissiveness, isolation, and withdrawal. In slightly older samples, victims were also found to be low on prosociality, and high in internalizing problems and in emotional symptoms (Arseneault et al., 2006; Malti, Perren, & Buchmann, 2010). Another correlate of social competence, emotion comprehension (Denham et al., 2003), has been found to be positively associated with the defender role and negatively with the outsider and victim roles (Belacchi & Farina, 2010). Defenders also had high scores on theory of mind tasks (Monks et al., 2005). Findings from studies with school children and adolescents are more numerous and indicate that victims can be less competent than peers in interacting with others (cf. Camodeca & Goossens, 2005; Caravita, Di Blasio, & Salmivalli, 2010; Perren & Alsaker, 2006), whereas contrasting findings are available for bullies. Aggr. Behav.
312
Camodeca et al.
Although some studies suggest that bullies lack affective empathy (Caravita, Di Blasio, & Salmivalli, 2009) and the necessary social cognitive skills to positively interact with others (Camodeca & Goossens, 2005), other studies have found that bullies are able to correctly understand social situations and, thus, can be able to achieve goals by means of intelligent manipulation (Caravita et al., 2010; Gini, 2006; Sutton, Smith, & Swettenham, 1999). Followers have been found to show low social cognitive and mind‐reading abilities (Camodeca & Goossens, 2005; Sutton et al., 1999), whereas defenders have been found to be well‐liked among peers, prosocial, empathic, and morally competent (Caravita et al., 2009, 2010; Gini, Pozzoli, & Hauser, 2011; Menesini & Camodeca, 2008; Pozzoli & Gini, 2010; Salmivalli et al., 1996). Results concerning outsiders are less clear, with some studies showing that their social and moral skills are similar to those of defenders (Camodeca & Goossens, 2005; Gini, 2006; Olthof, 2012), and others suggesting low responsibility, morality, and self‐efficacy (Gini, Albiero, Benelli, & Altoè, 2008; Pozzoli & Gini, 2010), which may account for their silent support for bullies. A possible explanation of inconsistent results, especially for the roles of bullies and outsiders, may lay in the fact that the aforementioned studies used different definitions of social competence or focused on specific indicators of social competence. To avoid this problem, and to fill the gap of paucity of studies on social competence and bullying in young children, we explored social competence in relation to bullying roles by employing the theoretical model of social competence as a hierarchically organized construct (Bost, Vaughn, Washington, Cielinski, & Bradbard, 1998; Shin et al., 2011; Vaughn et al., 2009; Waters & Sroufe, 1983). According to this integrative approach, social competence is understood to be the organization of young children’s behavior, affect, and cognition at a higher level of psychological abstraction. In line with the literature embracing this view, we used two Q‐Sort instruments, which are culture‐specific, and allow the testing of a very large quantity of characteristics, taking into account the multidimensional nature of social competence (Vaughn et al., 2009). This measure has been found to be particularly useful for preschool children (Coppola & Camodeca, 2010). Social competence, which includes the ability to interact appropriately with others, to cooperate, and achieve goals in a positive way, shows high correlations with social status in the peer group (Coppola & Camodeca, 2010; Denham et al., 2003; Vaughn et al., 2009). Children with low social status (measured by social preference) have a greater risk of developing internalizing and (especially) externalizing problems (McDougall, Hymel, Vaillancourt, & Mercer, 2001; Aggr. Behav.
Prinstein, Rancourt, Guerry, & Brown, 2009). Studies on older age groups show that, although bullies may have a high status among peers because they are visible and powerful, they are usually low in social preference (Caravita et al., 2009, 2010; de Bruyn, Cillessen, & Wissink, 2010). Followers and victims are often rejected by their peers, whereas defenders are usually the most liked, and bystanders show a social status that is below average both in social acceptance and rejection, indicating that they are neglected by their peers (e.g., Caravita et al., 2009, 2010; de Bruyn et al., 2010; Goossens, Olthof, & Dekker, 2006; Salmivalli et al., 1996). Similar associations between social preference and bullying roles have been found in preschool children (Monks et al., 2002, 2003, 2011; Perren & Alsaker, 2006), although studies with young children are still scarce. The second aim of the present study is therefore to analyze the relationships among social competence, social preference, and roles in bullying among pre‐ schoolers. We expect social competence to be directly and positively associated with the role of defender, and negatively with the roles of bully, follower, and victim (Camodeca & Goossens, 2005; Caravita et al., 2009, 2010; de Bruyn et al., 2010; Pozzoli & Gini, 2010). Given the previous inconsistent results, we did not formulate any hypotheses for the role of outsiders. Within a sub‐sample, we further investigate whether social competence is indirectly related to bullying roles via social preference. We expect that being liked by peers is associated with social competence as assessed by Q‐methodology (Bost et al., 1998; Santos, Peceguina, João, Shin, & Vaughn, 2013). Based on findings with older children, we also expect that social preference is positively associated with defending behavior and negatively with bullying, victimization, and follower behavior. We also expect that it mediates the relationship between social competence and bullying roles. Social preference could be a way through which social competence influences roles, but, given that no studies are available on indirect influence in preschool children, we cannot formulate clearer hypotheses about the strength and directions of these associations. In sum, the present work aims to: (1) test a peer nomination instrument to detect bullying roles in pre‐ schoolers; (2) investigate the social competence of children involved in bullying with different roles and analyze its indirect effects on bullying roles through social preference. METHOD
Sample and Procedure Participants were 320 preschool children (151 boys and 169 girls), with an average age of 58.40 months
Bullying Roles and Social Competence in Pre‐Schoolers
313
(SD ¼ 11.74; range: 31–79 months), attending three kindergartens (13 classrooms) in central Italy and one kindergarten (6 classrooms) in northern Italy. This latter sub‐sample, consisting of 78 children (45 girls and 33 boys, age: 52–79 months, M ¼ 65.69; SD ¼ 7.25), was also assessed on social preference. The purpose and the methodology of the study were explained to the schools’ principals and teachers, who gave their consent. In one school, parents were informed about the study as one of the activities of an internal project, to which they had previously given their consent. In the other schools, parents received a letter to illustrate the research and were asked to sign if they allowed the participation of their children. About 12% of the contacted families did not allow participation of their children in the study. Graduates in Psychology, appropriately trained in interview and observation methods, administered the interview individually to assess bullying roles and social preference, and conducted the observations in the classroom in order to obtain the Q‐Sort scores. Teachers filled in the questionnaire for bullying for each pupil in their classroom.
TABLE I. Items Content of the PRQ‐P
Measures Participant roles questionnaire for teachers (PRQ‐T). This questionnaire was developed by Belacchi and Farina (2010) on the basis of the original Participant Roles Questionnaire (PRQ; Salmivalli et al., 1996). It aims to assess eight roles by means of teachers’ evaluations of children’s behavior, employing three items per role. As we were only interested in five roles (bully, follower of the bully, defender of the victim, outsider, and victim), we employed a shortened version of the original measure. Teachers were asked to rate the frequency with which each child in their classroom behaved on 15 items, using a 5‐point Likert scale from 0 to 4 (never, rarely, sometimes, often, and always). Psychometric characteristics of the questionnaire, together with a CFA to confirm the 5‐factor structure, are described in the Results Section. Participant roles questionnaire for pre‐schoolers (PRQ‐P). The PRQ based on peer nominations was developed for children older than 7 (Salmivalli et al., 1996; Sutton & Smith, 1999) and allows assessment of the roles of participation in bullying. Belacchi and Farina (2010) adapted some of the PRQ items for use by preschool teachers (PRQ‐T; see above). On the basis of this previous work, we developed a version to be used by children in the form of peer nominations, including 10 items, two for each role. The items are reported in Table I. Items were read individually to each child, paying particular attention that the children understood them well (e.g., words were explained if children had difficulties in making sense; questions were formulated
more than once; etc.). Children were asked to indicate or nominate the classmates whose behavior fitted with the one described in the items. Children were presented with the photos of all their classmates, which is a common and reliable procedure for young children, who are not able to keep in mind all the peers in the class (cf. Vaughn et al., 2009). In a very few classes, photos were not available, but great care was taken to obtain reliable responses (e.g., children were asked to think about all their classmates, even those not present that day; children were given more time to think; etc.). Item scores were the sum of the received nominations, which were standardized by class, in order to control for classroom size. Psychometric characteristics of the questionnaire are described in the Results Section. Q‐Sort descriptions for social competence. Two Q‐Sets assessing social competence were used: the California Child Q‐Set (CCQ; Block & Block, 1980; Waters, Noyes, Vaughn, & Ricks, 1985), a 100‐item set consisting of personality and behavior descriptive items of preschool children, and the Preschool Q‐Set (PQ; Waters et al., 1985), a 72‐item set consisting of items related to social competence in preschool age. Both instruments have a strong theoretical basis and have been successfully implemented to assess social competence in preschool age (e.g., Bost et al., 1998; Santos et al., 2013; Vaughn, 2001; Vaughn et al., 2009). A total of five trained observers separately conducted about 20 hr of observation in the classroom during 1 week, completing at least 1 hr and a half of observations per each child. They then described each child by sorting the two Q‐Sets into a
Scale Bullying
Follower of the bully
Defender of the victim
Outsider
Victim
Items 1. Makes fun of his/her classmates and makes them cry 5. Does nasty tricks and says bad things to other children 3. Offends someone because other schoolmates do it 7. Encourages those who tease and irritate other classmates 2. Defends the children who are being beaten or teased 10. Tries to console the children who cry because they have been beaten or isolated 4. Pretends not to see when someone bullies someone else 8. Remains indifferent if a child is being teased or left alone 6. The other children beat him/her and ruin his/her things 9. The other children isolate him/her—keep him/her apart—or tell him/her bad things
Note. The items number corresponds to the number in the questionnaire.
Aggr. Behav.
314
Camodeca et al.
predetermined rectangular distribution of nine categories (q‐profile), on the basis of the similarity of each item to the children’s behavior. The score assigned to each item is its placement in the distribution. By convention, the child’s social competence score is the Pearson correlation, ranging between 1 and þ1, between his/her q‐profile and a q‐criterion, which is a composite description obtained by averaging the scores assigned by experts to each item. In the present study, for each Q‐Set, we implemented an Italian Social Competence Criterion, describing the hypothetical most socially competent Italian pre‐schooler, generated by averaging the descriptions provided by Italian experts in children’s social development (Coppola & Camodeca, 2010). Although these two criteria are highly similar to the US ones reported by Waters et al. (1985), that is, r ¼ .89 for the CCQ and r ¼ .88 for the PQ, they are also sensitive to cultural specificities and have been shown to be valid and reliable in the assessment of social competence at preschool age (Coppola & Camodeca, 2010). One hundred sixty‐one children (50.3% of the total sample) were observed by two trained observers: Cross‐rater agreement on criterion scores was satisfactory (r ¼ .70, for the CCQ, and r ¼ .60, for the PQ) and in line with those reported elsewhere (Santos et al., 2013; Vaughn et al., 2009). Peer nominations for social preference. Social preference was assessed by asking the 78 children from the sub‐sample to nominate up to three classmates they liked most (preferences) and three that they liked least (refusals), a method derived from McCandless and Marshall (1957). They were also presented with pictures of all their classmates, to aid their recollection of members of the group. After standardizing the nominations for preferences and refusals by class, refusals were subtracted from preferences to obtain the total score for social preference. As recommended by Cillessen (2009), this combined score was further standardized by class (i.e., the reference group). Strategy of Analysis Data were analyzed by means of Structural Equation Modelling (Mplus 7.0; Muthén & Muthén, 1998–2012). The robust method of maximum likelihood estimation (MLR estimation) was employed, in order to control for the non‐normality of the variables’ distribution. The MLR estimation provides a scaled Chi‐square (Satorra & Bentler, 1994). For the models including the test of indirect effects the bias‐corrected bootstrap was used as recommended strategy (MacKinnon, 2008; MacKinnon, Lockwood, & Williams, 2004). The boostrapping procedure is also a strategy proposed for the test of models when the sample size is small (Adèr, Mellenbergh, & Hand, 2008). It estimates the model parameters, along Aggr. Behav.
with their standard errors, strictly from the study sample. In this procedure, the study sample is considered as a population from which several samples are drawn by replacement. Then, estimates of the expected values and the variability of the statistics are taken from this sampling distribution. According to the bootstrapping method, an effect is significant at the .05 and .01 levels, if the value of 0 is not part of, respectively, the 95% and 99% bootstrap confidence intervals around the effect. Mplus does allow the computation of bias‐corrected bootstrap effect confidence intervals at the .05, .01, and .10 (marginal significance) level, and we tested the models including indirect effects by drawing 2000 samples (a minimum number of 500 samples is recommended). In Mplus, when models are tested by bootstrapping, the use of the MLR estimation is not possible, and the Maximum Likelihood (ML) estimation is employed as the default option. The use of ML in place of the MLR estimation affects the estimates of standard errors but not the estimates of the parameters. Therefore, the estimator ML was employed for the models tested by bootstrapping. Goodness of fit of the model was evaluated using the following: Chi‐square statistics and the ratio x2/df, which is considered acceptable if lower than 3 (Bollen, 1989); the Comparative Fit Index (CFI), which provides acceptable fit when it is equal or higher than .90 and good when equal or higher than .95 (Hu & Bentler, 1999); and the Root Mean Square Error of Approximation (RMSEA), which is judged adequate when equal or lower than .08 and good when equal or lower than .05 (Hu & Bentler, 1999). Modification indices were also considered for the improvement of the models. Because Mplus does not provide modification indices for the models tested by bootstrapping, those models were re‐ run without using the bootstrapping method and using the MLR estimation in order to obtain the modification indices. Possible improvement of those models was tested when the suggestions from the modification indices were in accordance with the suggestions from derivatives. Lastly, a post‐hoc model trimming procedure was also used in order to further optimize the models. That is, the specified paths that resulted non‐significantly different from 0 (Wald test) at the first test of the model were fixed to 0 in the model specification. These modifications of the conceptual model were made when removing the paths was theoretically justifiable, and they were tested one by one, in order to ensure that each of these modifications resulted in a non‐significant increase of the fit of the model [Chi‐square difference test, which was computed with the corrected formula for the robust chi square test by Satorra (2000) when the MLR estimation was used]. Therefore, the most parsimonious models were our final models.
Bullying Roles and Social Competence in Pre‐Schoolers RESULTS
Adaptation of the PRQ‐P: CFA, Reliabilities, and Associations With PRQ‐T In order to test whether the distinction between different bullying roles could be applied to pre‐schoolers and whether peer reports are a valid instrument to assess roles, we conducted a Confirmatory Factor Analysis on the 10 items of the PRQ‐P, which were allowed to load on the expected five factors; other loadings were fixed to 0. Fit indices for the model were satisfactory (x2(25) ¼ 35.67, P ¼.08; x2/df ¼ 1.43; CFI ¼ .98; RMSEA ¼ .037). All items loaded on the expected factors and standardized factor loadings were higher than .42 (Ps < .01). All roles were inter‐correlated with each other, with the exception of the defender role that was significantly (positively) associated only with the roles of victim and outsider. Although reliabilities were good for the scales of bully and follower (a ¼ .85 and a ¼ .71, respectively), they were very low for the other scales (defender scale a ¼ .49, outsider scale a ¼ .37, victim scale a ¼ .36), especially for the outsider and victim roles. Based on these low values and on the suggestions from the modification indices, we tested a second model in which the roles of outsider and victim were each assessed by only one item, that is, the item with the highest loading on the respective factor (item 8 for the outsider role and item 9 for the victim role). The model had good fit indices (x2(13) ¼ 21.06, P ¼.07; x2/df ¼ 1.62; CFI ¼ .99; RMSEA ¼ .044), and we retained this model as the final one. Table II displays means, standard deviations, and items loadings of the five roles in the final model, together with correlations among the scale factors (below the diagonal). In this model, the associations between roles that were previously found were confirmed, with the exception of the positive associations of the defender role with the roles of victim and outsider, which became marginal, thus better distinguishing this role from the others. As second step of the analyses, we correlated the scales of the peer and teacher versions of the PRQ. Given that we used a shortened version of the instrument by Belacchi and Farina (2010), we run a CFA (MLR estimation method) to confirm the 5‐factor structure of the PRQ‐T. Fit indices were acceptable (x2(80) ¼ 175.40, P
