MEASURES OF AGGRESSION AND VICTIMIZATION IN PORTUGUESE ADOLESCENTS

Victimization and aggression among adolescents in school settings is not a recent issue. It is however a growing problem not yet fully understood or properly evaluated. The goal of this thesis was to prepare and validate an instrument that addresses aggression, victimization and prosocial behavior in a quantitative, accessible and relatively fast format. The Revised Peer Experience Questionnaire (RPEQ) fulfills these needs since it accesses these tree constructs at once, for both aggressors and victims. For this purpose, a sample of 941 adolescent was recruited (49.1% males and 50.9% females). Participants’ age varied from 9 to 20 years old, and girls were significantly younger than boys. Regarding their school grade, 82.5% of the students attended middle school grades and 16.4% attended high school grades. Confirmatory factor analysis confirms the internal structure of the bully and victim versions of the instrument, as evaluating overt, relational and reputational aggression/ victimization and providing/ receiving prosocial behavior, respectively. This measurement model was invariant across genders, and presented adequate internal consistency and convergent validity indicators. Descriptive analysis furthermore shows that boys are more aggressive in overt and relational forms and victimized through overt aggression, whereas girls tend to be more aggressive and victimized relationally. More than any type of aggression or victimization, both boys and girls, revealed higher values for engaging and receiving prosocial behavior. Results from this research suggest that the RPEQ is a reliable and valid, structurally sound measure of aggression, victimization and prosocial behavior in this Portuguese school-based community sample. The use of this measure may assist researchers in gaining a better understanding of adolescent aggression, either performed or received, and may aid in the design of specific intervention strategies.


Introduction
Victimization and aggression among adolescents in school settings is not a recent issue; it is, however, a growing problem not yet fully understood or properly evaluated (Costa, Pereira, Simões, & Farenzena, 2011).Most of these violent acts are frequently categorized as bullying, which consist of ill-intended and malicious behavior towards others.These acts occur typically with a certain level of repetition and are frequently perpetrated on those who have difficulties defending themselves (Olweus, 1983(Olweus, , 1986(Olweus, , 1991 cit by Olweus, 2011).Nevertheless, some of these acts do not qualify as bullying, since they do not fit into stables acts of interpersonal violence perpetuated through time via physical, verbal, social and even sexual acts.These random actions of violence tend to be purely physical aggression, consisting mostly of school yard fights (Carvalhosa, 2010).
Portuguese adolescents were placed 4th in a HBSC (Health Behavior in School-Aged Children) ranking, based on the reported frequency of aggressive acts.This ranking shows that about 30% of our adolescents are somehow involved in aggressive behavior (Currie, Hurrelmann, Settertobulte, Smith and Todd, 2000 cit by Carvalhosa, 2010).
Similar findings were obtained in other studies, revealing that the percentage of adolescents who bullied others grew from 9% to 25% in a between 2008 and 2010 (Carvalhosa, 2008;Costa et al., 2011).According to the same HBSC study, the percentage of Portuguese adolescents who were victims of bullying was approximately 29% (Carvalhosa, Moleiro, & Sales, 2009).Carvalhosa (2008) furthermore found a prevalence of more than 27% of Portuguese adolescents who are both bullies and victims.When assessing gender differentiation, boys, in comparison to girls, tend to engage more in aggressive behavior as perpetrators, victims or both (victim and aggressor simultaneously (Currie, Hurrelmann, Settertobulte, Smith and Todd, 2000 cit by Carvalhosa, 2010).
According to a longitudinal study by Olweus (2011), being a bully in early adolescence strongly predicted later criminality involvement.Rendering his finding, 55% of his bullying sample was convicted for one or more crimes in a 8-year period from ages 16 to 24 (Olweus, 2011).Regarding Takizawa and his colleagues (2014), being bullied at childhood puts well-being and mental health at risk up to four decades later.Childhood victims experience an extensive scope of poor social, health, and economic outcomes into middle adulthood (Takizawa, Maughan, & Arseneault, 2014).These finding were similar in terms of social outcomes to what had been found with Portuguese adolescents.This study indicated that boys and girls that were the most aggressive perceived a lower reassurance of personal value by others and decreased chances for providing support to others.Aggressiveness in girls was associated with decreased social inclusion, while for boys it was associated with diminished perception of a reliable bond and attachment to others (Neto, Grave, Caldeira, Morgado, & Vagos, 2013).
In order to better understand aggressive behavior, this concept was scrutinized regarding its form (i.e., method by which the harm is delivered) and function (i.e., purpose of the aggressive act).Regarding the forms of aggression, they have been are categorized as overt, relational and reputational (Prinstein & Cillessen, 2003).Direct or overt aggression refers to physically and verbally aggressive behavior (i.e., hitting, punching and threatening others) (De Los Reyes & Prinstein, 2004).Indirect forms of aggression such as relational do not include physical means to inflict harm; in contrast, they appeal to the relationship as a weapon used to harm others (i.e., by leaving others out of activities on purpose, knowing that they would like to participate) (Marsee et al., 2011;Putallaz et al., 2007).Reputational aggression or social aggression is very similar to relational aggression, but still presents slight differences.It is defined by its end-point, just like relational aggression, but in the case of reputational aggression the goal is to damage others' social standing by manipulating group acceptance (i.e., by saying mean things about another peer so that the group won't like him anymore) (Galen & Underwood, 1997 cit by Archer & Coyne, 2005).In sum, reputational aggression is a form of indirect aggression that attempts to damage peers' reputation, and it differs from relational aggression by using the other person's relationships instead of one's own relationships to inflict harm (Prinstein & Cillessen, 2003).Regarding the functions of aggression, they can be reactive and proactive.
The first one refers to the angry responses towards others in response to perceived threat or provocation, whereas proactive aggression refers to unprovoked free attack, generally with the attempt to gain something or dominate others (Marsee et al., 2011).
A study conducted with 493 adolescents over a two year period, revealed that higher levels of overt victimization in girls are associated with an increase in suicidal ideation (Heilbron & Prinstein, 2010).In general, boys tend to be more aggressive and victimized in overt forms than girls, and girls tend to be more victimized in relational forms than boys (Putallaz et al., 2007), even if there are studies reporting no gender differences or even higher rates of relational aggression among boys than girls (Underwood, 2003).
In order to properly address this issue and intervene in such matter, it is crucial that a good evaluation is conducted and the ideal method would be through direct observation.
However, in the public school system this may not be executable.Instead, a viable alternative is to resort to computer tasks and questionnaires (self and other informants).In this study, the authors favored the questionnaire option for its ability to reach a wider number of participants at once.(Giancola & Chermark, 1998;Anderson & Bushman, 1997 cit by Arriaga, Esteves, & Monteiro, 2004), in addition to the fact that psychometrically sound questionnaires for the evaluation of aggressive behavior in its perpetrator and victimized mode are still very scarce for the Portuguese adolescent population.Prinstein, 2004) will represent a good fit for the data of the bully and the victim version of the instrument, and will be invariant for boys and girls.As secondary goal of this work, and in an effort to make inferential analysis pertaining to construct validity of the Portuguese Version of the RPEQ (Appendix B), we considered gender differences on the levels of aggression, victimization and prosocial behavior.If results are in line with those previously found for the same constructs, we can infer that the RPEQ evaluates the same constructs.Following such previous findings, we expect that boys will report higher levels of overt aggression and victimization than girls, and girls will report higher levels of relational aggression and victimization than boys (Putallaz et al., 2007).

Participants
The participants for present study were 941 children and adolescents from 4 middle/high schools (461 males (49.1%) and 477 females (50.9%)

Instruments
The present study aimed at investigating the validity and reliability of the Revised Peer Experience Questionnaire (RPEQ; Prinstein et al., 2001).For construct validity purposes, Socioeconomic status was assigned according on parents' profession, and taking into account the Portuguese profession classification.
Examples of professions in the high socioeconomic status groups are judges, higher education professors, or M.D.s; in the medium socioeconomic status group are included nurses, psychologists, or school teachers; in the low socioeconomic group are included farmers, cleaning staff, or undifferentiated workers.
we used the Portuguese versions of the Peer Conflict Scale (PCS; Vagos, Rijo e Santos, 2014) and of the Revised Olweus Bully/Victim Questionnaire (Pereira, 2002).

Revised Peer Experience Questionnaire
The RPEQ (Prinstein et al., 2001) is a revised version of a measure originally developed by Vernberg and his colleagues (1999) to assess aggression and victimization among peers in school context.Although the original version assessed mostly overt forms of aggression and victimization, the RPEQ comported five items that were revised, created, or added from prior works (e.g., Lopez, 1998) to reflect developmentally appropriate forms of relational aggression and victimization in adolescents.At this point, the RPEQ can assess direct and indirect forms of aggression and victimization (i.e., relational, reputational) as well as the youths' engagement and receipt of prosocial behavior.Specifically, the RPEQ is composed of two counterbalanced sets of 18 items each.One of this sets of items assesses participants' peer victimization experiences.(i.e., victim version).In this version, participants are invited to rate how often an aggressive or prosocial behavior was directed towards them in the past year on a 5-poin scale that ranges from 1 (never) to 5 (a few times a week).The RPEQ assesses overt (i.e., "A kid threatened to hurt or beat me up"), relational (i.e., "A kid left me out of what they were doing") and reputational (i.e., "A teen tried to damage my social reputation by spreading rumors about me") forms of victimization, as well as receiving prosocial behavior (i.e., "Another teen helped me when I was having a problem").The other set of items assesses participants' practice of aggressive behavior towards peers (i.e., Bully version).Participants rate how often they engaged in an aggressive or prosocial behavior toward others in the past year on a 5-poin scale that ranges from 1 (never) to 5 (a few times a week).This instrument also assesses overt (i.e., "I threatened to hurt or beat up another teen"), relational (i.e., "I left another teen out of what I was doing") and reputational (i.e., "I tried to damage another teen's social reputation by spreading rumors about them") forms of aggression, as well as practicing prosocial behavior towards others (i.e., "I helped another teen when they were having a problem").The revised version of this questionnaire has demonstrated good test-retest reliability, internal consistency and convergent validity on both the victim and aggression versions of the measure (Prinstein et al., 2001).
The 36 original RPEQ items were translated and adapted into the Portuguese language following a translation and back-translation process (Hambleton, Merenda, & Spielberger, 2005).The 36 Portuguese items were back-translated into English by a Portuguese researcher, unrelated to this study, who is fluent in both Portuguese and English.The original and back-translated versions were considered equivalents by an English Portuguese teacher.The original and back-translated versions of one item from each proposed construct are shown in Table 2. Another teen gossiped about me so others would not like me.
Another teenager gossiped about me so that others wouldn't like me.

Note. a=aggression, v=victimization
For data analysis purpose, only 28 items were considered, similarly to what the original authors of the revised form of the instrument (used in the present work) have done, due to them not loading on any specific construct among those represented by the instrument.Items 4, 9, 13 and 17 were excluded from both the victim and aggressor versions.

Peer Conflict Scale
The PCS consists on a 40-item questionnaire that assesses four categories of aggressive behavior for perpetrators, consisting of reactive relational (i.e., "when someone makes me angry, I might badmouth that person"), reactive overt (i.e., "I have threatened somebody that has done something bad to me"), proactive relational (i.e., "I might tell somebody's secrets, it that's convenient for me") and proactive overt (i.e., "have hurt others to win a game or contest").Items are rated from 1 (it has very little to do with me) to 4 (it has everything to do with me) in its Portuguese version, with possible scores ranging from 10 to 40.The higher the score on each scale, the higher the level of selfreported aggression.
The PCS questionnaire may be applied to adolescents with ages up to 18 years old.The Portuguese version of the instrument presents good reliability (alphas ranging from .89 to .91) and construct validity in relation to other variables.The four-factor measurement model was validated to the Portuguese samples, and showed cross-gender invariance (Vagos et al., 2014).

Revised Olweus Bully/Victim Questionnaire
The ROBVQ (English version by Olweus, 1996, Portuguese version by Pereira, 2002) consists of 40 questions for the measurement of bully/victim problems such as, exposure to various physical, verbal, indirect, racial, or sexual forms of bullying/harassment, practice of various forms of bullying towards other students, probully and pro-victim attitudes, and the extent to which the social environment (teachers, peers, parents) is informed about and reacts to the bullying (Olweus, 2007).These questions are divided in 4 sections: socio-demographic information, victimization behavior, aggressive behavior and school environment perception.This instrument may be applied to adolescents aged 8 to 16 years old.To our knowledge, psychometric analysis regarding this instrument has not yet been presented or published either for its original or Portuguese version.For the present study, only 8 questions regarding 2 sections were used: victimization behavior and aggressive behavior (see appendix C), given that its purpose was solely to assess construct validity in relation to the construct evaluated by the RPEQ.

Procedure
Prior to data collection, authorization for this work was sought and granted by the national evaluation committee on ethics and procedures to be followed by studies conducted in school settings (see appendix D).The study was conducted with adolescents from the 5 th to the 12 th grades in 4 middle/high schools.After receiving the authorization of the administration board of each school, parental consent forms were distributed through every head teacher of all targeted years (see appendix E).
Only students who received permission from their parents and who provided assent themselves took part in the investigation.The students were assessed in groups during their head teacher's class period at school.A standardized protocol in giving instructions was provided to the school boards in order to unify methods of data collection in every class.
All participants were informed that all of their answers were confidential and the instruments were posteriorly branded with an identification number for purely organization purposes.The instruments were presented alongside a socio-demographic questionnaire which included the following questions: "Age?"; "Gender?"; "School grade?"; "Number of years retained in the same school grade?"; "Have you ever had a disciplinary measure applied to you at school?"; "If yes, which measure did you receive?"; "Do you benefit from financial aid?", "What degree dos your father hold?"; "What degree does your mother hold?"; "What's your father's profession?" and "What's your mother's profession?".
Data analysis was conducted using Mplus (v6.2; (Muthén & Muthén, 1998-2010) and SPSS (v20.0).Mplus was used for single and multi-sample (i.e., by gender) confirmatory factor analysis.For evaluating model fit for confirmatory factor analysis, a 2index criterion was considered (Hu & Bentler, 1999), which combines a value of Standardized Root Mean Square Residual (SRMR) ≤ .08 with either a value of Comparative Fit Index (CFI) or Tucker-Lewis Index (TLI) ≥ .95,or a value of Root Mean Square Error of Approximation (RMSEA) ≤ .06.SPSS was used for descriptive, correlation and mean comparison analyses.

Confirmatory Factor Analysis
In order to define the measurement model of the RPEQ, two confirmatory factor analysis were performed following the procedure used by De los Reyes and Prinstein (2004).One included a four -factor model using the 14 items representing the bully version and the other included a four-factor model using the 14 items representing the victim version.
Analysis of responses indicated a normal distribution for the collected data for both version of the questionnaire (Bully version: multivariate kurtosis for 14 items=1.79;0 items with univariate kurtosis values ≥4 and 7 items with univariate skewnes values ≥2; victim version: multivariate kurtosis for 14 items=1.57;0 items with univariate kurtosis values ≥7 and 2 items with univariate skewness values ≥2) and consequently the Maximum Likelihood method was used for data analysis (Finney & diStefano, 2006).
The fit indexes were always reasonable and are presented in table 3 (i.e., CFI and TLI ≥ .90,RMSEA ≤ .07 and SRMR≤ .08)Item loadings on the four factor solution for each version of the questionnaire were always superior to .46 for the complete sample (Table 4).The internal consistency values for the four constructs proposed for each of the RPEQ's versions, and for the complete sample, were very good, according to the reference value of .07(Nunnally, 1978)   The measurement invariance of both the bully and the victim measurement model by gender was also tested (Table 3).For the bully version, the four-factor model seemed an adequate fit for the data of male and female participants separately.Additionally, full configural and metric invariance was found, given that the more constraint model (i.e., metric invariance) did not significantly worsen the fit of the less constraint model (i.e., configural model; ∆ 2 = 16.76,df = 10; p>0.05).Loadings for the items in the bully version were always superior to .54 for boys and .31for girls (Table 4).
For the victim version, the four-factor model again seemed an adequate fit for the data of male and female participants separately, but the constraint model represented a significant worsen of the fit, in comparison with the less constraint model (∆ 2 =31.35 df = 10 p<0.001).
Analysis of the modification indices for the metric invariance model suggested that items 3 and 5 should be allowed to correlate differently for girls and boys.Freeing this constraint did not question the metric invariance of the model, and was theoretically justified since both items are intended to evaluate being the recipient of prosocial behavior.Consequently, this modification was incorporated in the model and so full configural and metric invariance was achieved for the victim version of the instrument (∆ 2 =-8.25 df = 11 p>0.20).Loadings for the items in the victim version were always superior to .58 for boys and .52 for girls (Table 4)

Descriptive Analysis
Two mixed ANOVAs, with one within-subject factor for aggression/victimization and one between-subject factor as gender, were performed; school year, number of school holdbacks, and history of disciplinary measures were entered as covariates, given that boys and girls had significantly different mean ages and were not evenly distributed by school year, number of school holdbacks and history of disciplinary measures.
Regarding the bully version of the instrument, the Greenhouse-Geisser correction was used to account for violations to sphericity in the mixed ANOVA analysis ( = 0.387, p < .001).The within-subject effect was not significant (F < 1); the between-subject effect was significant for gender (F(1,650) = 20,11, p≤.001; p 2 = .030),and for the interaction between gender and aggression (F(1,81;3) = 13.08;p≤.001; p 2 = .020).The interaction effect between disciplinary measures and aggression was also significant (F(1,81; 1) = 3.65, p = .03;p 2 = .006).Pairwise comparisons for levels of aggression between boys and girls indicate that boys present significant higher values of the three types of bully behavior; girls presented higher levels of prosocial behavior towards others, even if this difference was not significant.Boys considered separately reported significantly higher levels of prosocial behavior towards others, followed by similar levels of both relational aggression and overt aggression, and finally, significantly lower levels of reputational aggression.In contrast, girls considered separately reported significantly higher levels of prosocial behavior towards others followed by relational aggression and finally, with similar levels of reputational aggression and overt aggression (Figure 1).
Note.Ccovariates appearing in the model are evaluated at the following values: age = 11.69;school year = 6.485;School holdbacks = .16,and disciplinary measure = 1.74 As for the victim version, the Greenhouse-Geisser correction was again used to account for violations to sphericity in the mixed ANOVA analysis ( = 0.577, p < .001).
Pairwise comparisons for levels of victimization between boys and girls indicate that they differ significantly only in overt victimization where boys revealed higher values.Boys present higher values than girls for overt and relational victimization, and girls present greater values than boys for reputational victimization and receiving prosocial behavior.
Considering only boys, they reported higher levels or receiving prosocial behavior, followed by being victim of relational aggression, then reputational aggression and then overt aggression; the difference was only not significant (p > .05)for the comparison between overt and reputational victimization.Considering only girls, they reported higher levels or receiving prosocial behavior, followed by being victim of relational aggression, then reputational aggression and then overt aggression; the difference was only not significant for the comparison between relational and reputational victimization (p > .05; Figure .2).
Note.Ccovariates appearing in the model are evaluated at the following values: age = 11.69;school year = 6.485;School holdbacks = .16,and disciplinary measure = 1.74

Convergent validity analysis
To establish convergent validity, comparisons between the RPEQ, PCS and ROBVQ were conducted.for the aggression scales of the RPEQ, PCS and ROBVQ, and table 7 outlines the Pearson rank-order correlation coefficients for the victimization scales of the RPEQ and ROBVQ.
Convergent validity analysis was not conducted for the engagement and receipt of prosocial behavior since the authors were not able to find a scale that would access the same constructs.
Moderate but significant correlations were observed between the RPEQ victim version and the ROBVQ's victim scale, whereas stronger correlations were found between all tree aggression scales from the RPEQ (i.e., overt aggression, relational aggression and reputational aggression) and the aggression measures taken from the ROBVQ and PCS.
The highest correlation for the aggression measures was found between the overt aggression measure of the RPEQ and the overt measures of the PCS.The lowest correlation value were obtained between the relational aggression measure of the RPEQ and the aggression measure taken from the ROBVQ.

…
The correlations between the victimization measures of the RPEQ and the ROBVQ were of similar magnitude, with the highest being for overt victimization and the lowest for reputational victimization.

Discussion/Conclusion
The present research intended to fill the existing gap in Portuguese psychological evaluation processes, when it comes to instruments available to the evaluation of aggression, victimization and prosocial behavior of adolescents in a quantitative, accessible and relatively fast format.The aim of this study was to psychometrically analyze an alternative to the two already available instruments that nonetheless neglect either the prosocial and victim experiences (i.e., the Peer Conflict Scale) or the quantitative analysis of such behavioral patterns (i.e., the Revised Olweus Bully-Victim Questionnaire).By assessing the behaviors of victims and aggressors and the practicing and receiving of prosocial behavior, the Revised Peer Experience Questionnaire may be an important alternative to the evaluation of aggressive acts in school settings that Portuguese adolescents are increasingly face (Costa, Pereira, Simões, & Farenzena, 2011).Not only it will facilitate the work of the research community, but will also provide a more holist perspective of the behaviors under study.
The Portuguese version of the Revised Peer Experience Questionnaire was developed and adapted through translations and back-translation procedures.The internal structure of the scale was evaluated based on a confirmatory factor analysis procedure.The four-factor model proposed by De Los Reyes and Prinstein (2004) for each version of the instrument (i.e., bully and victim) was a very good fit for the data, and included measures of overt aggression / victimization, relational aggression / victimization, reputational aggression / victimization, and engagement and receipt of prosocial behavior among peers.
This results make it possible to conduct multicultural studies based on this measure, since the same constructs seem to be evident in the measurement models for the RPEQ in north- The four factor model for both versions of the instrument was invariant across gender (i.e., boys vs girls) in this sample, suggesting that the four factor structure fit equally well for both boys and girls, and consequently that comparisons of these groups can be informative on whether or not the groups have equal amounts of the latent constructs (Hair, Black, Babin, & Anderson, 2009).Items 3 (i.e., Another teen helped me when I was having a problem) and 5 (i.e., Another teen was nice and friendly to me when I needed help) were differently correlated for boys and girls for the victim version, and indicated that girls in particular perceive these two behaviors as closely intertwined.
The possible differential practice of aggression, victimization and prosocial behavior by boys and girls studied within the current sample indicate that boys tend to be more aggressive and victimized in overt forms, similarly to what had been found (Prinstein et al., 2001;Putallaz et al., 2007).However, boys also tended to be more aggressive in relational and reputational forms, contrary to what was expected.Some studies point to no differences between boys and girls in relational aggression.In fact, these results were similar to other findings, namely the ones previously found using the north-American version of the RPEQ (Prinstein, et al., 2001).Particularly referring to aggression, previous works with a Portuguese sample found that boys practice more reactive relational and overt aggression then girls (Vagos et al., 2014); Additionally, boys tended to be more victimized in reputational and relational forms as well.Boys usually partner with other boys, more than with girl (Maccoby, 1998), and so probably suffer from the very forms of aggression they practice.It therefore seems that relational aggression is, in fact, no longer a female issue, but should instead be considered as a general and disturbing experience for adolescent boys and girls (Archer & Coyne, 2005).Girls in our sample revealed greater levels of prosocial behavior, either given or received, in comparison to boys.Adolescent girls tend to value closeness and intimacy in their social relationships, which may be fostered by the prosociality.Prosociality, in turn, may protect those who practice and receive it from becoming aggressors or victims (Kendrick, Jutengren & Stattin, 2012), in the last case possibly due to the aggressor's fear of retaliation or other negative consequence if targeting a prosocial peer (Hodges, Boivin, Vitaro, & Bukowski, 1999;Prinstein et al., 2001).Given this rational, it is consistent to find that girls are at the same less aggressive and victimized, but more prosocial.
Considering within-gender comparisons for the bully version, the results indicate similar patterns for both boys and girls.Both genders presented higher levels of relational aggression, followed by overt aggression and finally reputational aggression.Given that aggressive acts are increasingly common in schools nowadays (Archer & Coyne, 2005;Costa, Pereira, Simões, & Farenzena, 2011), it may be the case that adolescents are early on learning to strategically use this behavior, by practicing its relational form, which more hidden and less punished or recognized by external observers (Archer, 2001).As for the victim version of the instrument, results found with the current sample indicate that the patterns are different for boys and girls.For boys, the highest scores were for relational victimization, followed by overt victimization, and finally, reputational victimization, whereas girls scored higher for victimization though relational aggression, followed by reputational aggression and finally overt aggression.These results are partially concurrent to those found by Ptallaz and his colleagues (2007) particularly the ones referring to girls being more victimized in relational forms (Putallaz et al., 2007).However, contrary to what would be expected, boys are also more victimized in these form, which may again relate to boys affiliating more with boys (Maccoby, 1998) and therefore suffering from the very types of aggression they practice, particularly the relational and overt forms of aggressive acts.
The convergent validity in relation to other variables confers evidence on the construct validity of the Portuguese version of the Revised Peer Experience Questionnaire.
Its results were correlated to the results of other instruments that propose to measure the same constructs, and for that reason, all scales from both versions seem to be fit to evaluate its inherent constructs.For example, the overt aggression scale of the RPEQ has strongly correlated to the proactive overt and reactive overt scales of the PCS.Similarly, the relational aggression scale of the RPEQ reveals reasonable, yet significant, correlations to the proactive and reactive relational scales of the PCS.The victim scales revealed low, yet significant correlation with the ROVBQ's victim scales.This may be due to the fact that the RPEQ accesses the three forms of victimization whereas the ROBVQ only evaluates presence or absence of victimization per se.Convergent validity analysis was not conducted for the engagement and receipt of prosocial behavior since the authors were not able to find a scale that would access this very constructs.Nonetheless, previous studies have found that, similarly to our results, both boys and girls tend to reveal greater levels of acting and receiving prosocial behavior when compared to acting aggressively (Kelly, 2007), which may be indicative that the RPEQ is addressing its intended constructs, also when evaluating prosociality.This work is not without limitations, namely the fact that it is a cross-sectional study with a school-based community sample, using only self-report instruments.A longitudinal design is needed to determine whether aggression levels change over school grades, school holdbacks and disciplinary measures applied by the school, which were three significant covariates for aggression levels by gender found in the present work.This longitudinal investigations may contribute to a socio-cognitive and developmental perspective on adolescent aggression.Additionally, a wider and more diverse sample is required to create specific norms for aggressive behavior of Portuguese adolescents.It would be important as well to reconsider and test the reintegration of the items excluded from both version by the original authors and in this study.Similarly, the current research focused on replicating the measurement model proposed for north-American samples.
Even if this is relevant for the possibility of conducting multi-cultural studies, alternative measurement models may be verified, namely by exploratory factor analysis with Portuguese adolescents, by investigating confirmatory models that include simultaneously the 38 items originally derived for the RPEQ, or by testing diverse confirmatory models to the evaluation of aggression, victimization and the giving and receiving of prosocial behavior.Nevertheless, the goal of this study was fulfilled, namely adapting and presenting an instrument for experiences of aggression in Portuguese adolescents, capable of being administered to both aggressors and victims, and intending to also evaluate the giving and receiving of prosocial behavior.
Results from this research suggest that the RPEQ is a reliable and valid, structurally sound measure of aggression, victimization and prosocial behavior in this Portuguese school-based community sample.The use of this measure may assist researchers in gaining a better understanding of adolescent aggression, either performed or received, and may aid in the design of specific intervention strategies.This is particularly important since it may be used to evaluate the need for intervention, the intervention itself and the efficiency of this intervention.Aggressive acts in school settings have a widespread impact, to both its actors and the more general community.Consequently, preventing this acts and promoting healthy school environments where prosocial behaviors are the norm is paramount (World Health Organization, 1999 cit by Carvalhosa, 2010).The RPEQ may be an appropriate way to retrieve evidence-based information on the clinical and statistical change of adolescents subjected to such interventions, in addition to being usable for the screening of social behaviors in adolescents, and allowing for multi-cultural assessment and comparison of aggression, victimization and prosocial behavior.
Figure index American and Portuguese school-based samples.These four scales revealed a good reliability based on their internal consistency values.Nonetheless, the relational aggression and victimization measures, presented values under .70,which point to caution being needed when interpreting results as the source of clinical or diagnostic individual decisions.

Table index Table 1 :
Sample description regarding, retentions, disciplinary measures,

Table 2 :
Adaptation and back-translation of the Peer Experience

Table 4 :
Item loading on the four factor model, for the two versions………….…….

Table 1
.e., 9 th , 10 th , 11 th and 12 th grade; 16.4%).The majority of the students had never been retained in the same school year (n= 608; 64.6%) and had never been subjected to a disciplinary measure due to inappropriate behavior (n = 498; 52.9%).Those who had been subjected to such measures were expelled from the class (n = 47, 10.2%).As for socioeconomic status 1 , most of the students belonged to a low socioeconomic status (n= 509, 54.1%), but only 22.6% (n = 213) of the students received financial aid from the school.Sample description regarding, retentions, disciplinary measures, specification of the disciplinary measure, financial aid, and socioeconomic level.

Table 3
Confirmatory factor analysis on the four-factor measurement model for the internal structure of the Portuguese version of the Revised Peer Experience Questionnaire Note.RMSEA = root-mean-square error of approximation; WRMR = standardized root mean square residual; CI for RMSEA = confidence interval for RMSEA; CFI = comparative fit index; TLI = Tucker-Lewis Index.aadjustment values presented refer to the modified metric variance model All chi-square values were significant at p < .001 Table 6 outlines the Pearson rank-order correlation coefficients

Table 6
Pearson rank-order correlations between the Revised Peer Experience Questionnaire (RPEQ), the Peer Conflict Scale (PCS) and the Revised Olweus Bully/victim Questionnaire (ROBVQ)