مزاحمت سایبری: سمت پنهان دانشجویان

The purpose of this study was to investigate how university students perceive their involvement in the cyberbullying phenomenon, and its impact on their well-being. Thus, this study presents a preliminary approach of how college students’ perceived involvement in acts of cyberbullying can be measured. Firstly, Exploratory Factor Analysis (N = 349) revealed a unidimensional structure of the four scales included in the Cyberbullying Inventory for College Students. Then, Item Response Theory (N = 170) was used to analyze the unidimensionality of each scale and the interactions between participants and items. Results revealed good item reliability and Cronbach’s α for each scale. Results also showed the potential of the instrument and how college students underrated their involvement in acts of cyberbullying. Additionally, aggression types, coping strategies and sources of help to deal with cyberbullying were identified and discussed. Lastly, age, gender and course-related issues were considered in the analysis. Implications for researchers and practitioners are discussed.

School violence is a contemporary topic of discussion and one of the main causes of concern of students and professionals of the educational system. According to the literature, violence in educational settings has increased (Li, 2006), with aggravated consequences for the teaching and learning processes (Glover, Gough, Johnson, & Cartwright, 2000), as well as the socio-affective development of students (Clarke & Kiselica, 1997). Furthermore, school is the place where adolescents spend the majority of their time. Therefore, it is a critical arena of social support and academic development. Some of the literature has shown that students in schools with higher levels of bullying perform worse academically. (Strøm, Thoresen, Wentzel-Larsen, & Dyb, 2013). This type of violence affects many children and teenagers, at school and at home with the expansion and development of information and communication technologies (ICT). This insecurity is present at different grade levels, including university contexts and therefore, research involving the different forms of bullying is crucial in order to provide a better understanding of how it occurs, how students can deal with it and ultimately, how it can be prevented. As ICT have increasingly been incorporated into schools because they foster creative and autonomous ways of communicating and interacting, the risks and dangers associated with them also increase (Li, 2006). To specify, the rapid development of ICT (e.g. Internet and cell phones) has created more opportunities for bullies (Li, 2006 and Li, 2008) in the sense that the increased use or misuse of these electronic devices among teenagers (Slonje & Smith, 2008), has originated a new form of bullying (Beran & Li, 2007), that is, cyberbullying. Cyberbullying has a considerable impact on the lives of children and teenagers, considering it emerges at the elementary level and continues to higher education (Hinduja & Patchin, 2009) with increasing frequency and severity in and out of schools (Li, 2006). In light of these issues and because cyberbullying entails negative psychological and physical consequences that may affect interpersonal relationships (Anderson & Sturm, 2007), research should focus on the perceptions students have of their involvement in situations of cyberbullying, along with its associated dangers. Firstly, this study aims to understand how college students view and report their involvement in situations of cyberbullying. Hence, we present an inventory which could allow us to achieve this objective through the interpretation of its structure. We used Item Response Theory (IRT), which allowed us to calibrate our participants and items on a common scale (DeMars, 2010 and Embretson, 1996). This type of measurement presents an analysis of the interactions between people and items, enabling the interpretation of the variables in question. What’s more, the interpretations of items in which participants have a higher or lower probability of dominating, have an important diagnostic convenience for our study, along with other group-related ratings, which we consider later. To complement our first analysis, the present study also explores the dynamics of cyberbullying in order to provide a better understanding of how college students view this phenomena from different perspectives (the roles of the victim, aggressor and observer of victims and/or aggressors). We also consider different aspects that are associated with cyberbullying, such as intimidation and image appropriation that may affect the lives of college students. Moreover, with the analyses presented in this study, we provide insights regarding the means through which cyberbullying occurs (i.e. type of ICT used), as well as the most common types of occurrence in Portuguese college settings.

4. Results 4.1. Cyberbullying Inventory for College Students – exploratory evidence We used IBM SPSS 22.0 and FACTOR 9.2 to interpret the internal structure of the four scales. Table 1 shows the correlations among all variables for each scale, as well as the descriptive statistics. Table 1. Item descriptive statistics, Exploratory Factor Analysis parameters, reliability and correlations of the CICS. Variables Structure coefficients Mean(SD) Correlations 1 2 3 4 5 6 7 8 Victims’ scale Item 1 .79 1.13(.35) Item 2 .74 1.07(.28) .53 Item 3 .91 1.19(.46) .72 .67 Item 4 .77 1.08(.28) .53 .59 .71 Item 5 .86 1.18(.45) .73 .65 .84 .54 Item 6 .91 1.25(.50) .80 .63 .81 .66 .89 Item 7 .84 1.13(.38) .70 .68 .73 .63 .72 .80 Item 8 .86 1.09(.33) .64 .67 .85 .69 .70 .71 .74 Item 9 .69 1.07(.26) .55 .54 .55 .79 .55 .57 .52 .59 Eigenvalues 6.15 % Explained variance 72% Cronbach’s alpha .96 Aggressors’ scale Item 2 .81 1.00(.05) Item 3 .95 1.01(.10) .76 Item 4 .84 1.03(.17) .82 .80 Item 5 .97 1.06(.24) .77 .94 .82 Item 6 .94 1.05(.24) .71 .93 .78 .96 Item 7 .88 1.02(.13) .71 .84 .70 .85 .83 Item 8 .93 1.01(.10) .73 .91 .72 .89 .88 .91 Item 9 .91 1.01(.12) .77 .84 .81 .87 .87 .81 .83 Eigenvalues 6.80 % Explained variance 85% Cronbach’s alpha .98 Observers of victims’ scale Item 1 .75 1.23(.51) Item 2 .73 1.21(.52) .45 Item 3 .92 1.48(.73) .66 .68 Item 4 .74 1.21(.54) .54 .56 .73 Item 5 .92 1.53(.77) .73 .62 .86 .61 Item 6 .88 1.49(.75) .77 .56 .84 .64 .94 Item 7 .90 1.38(.68) .74 .70 .80 .55 .82 .78 Item 8 .92 1.34(.66) .71 .73 .84 .66 .80 .74 .90 Item 9 .88 1.35(.68) .56 .72 .78 .83 .80 .70 .77 .83 Eigenvalues 6.83 % Explained variance 75% Cronbach’s alpha .97 Observers of aggressors’ scale Item 1 .83 1.09(.32) Item 2 .82 1.06(.29) .73 Item 3 .96 1.20(.52) .80 .76 Item 4 .85 1.09(.39) .73 .82 .80 Item 5 .97 1.23(.54) .82 .77 .94 .82 Item 6 .95 1.23(.53) .82 .71 .93 .78 .96 Item 7 .88 1.11(.39) .72 .71 .84 .70 .85 .83 Item 8 .93 1.14(.45) .79 .73 .91 .72 .89 .88 .91 Item 9 .89 1.12(.42) .60 .77 .84 .81 .87 .87 .81 .83 Eigenvalues 7.33 % Explained variance 83% Cronbach’s alpha .98 Table options We used polychoric correlations, as suggested in the literature, when univariate distributions of ordinal items are asymmetric for polytomous items (Brown, 2006, Muthén and Kaplan, 1985 and Muthén and Kaplan, 1992). We tested the data with the Kaiser–Meyer–Olkin (KMO) and the Bartlett’s Test of Sphericity to understand its underlying structure. The KMO measure of sampling adequacy was a good (i.e. .84, .87, .88, .83, respectively), while the Bartlett Sphericity was χ2(36) = 1385.3 (p < .001), χ2(28) = 2154.8 (p < .001), χ2(36) = 2080.6 (p < .001) and χ2(36) = 2375.9 (p < .001), demonstrating that the variables were suitable for factor analyses. Furthermore, we tested for multivariate normality. As Bollen and Long indicate (1993), if Mardia’s coefficient is lower than P(P + 2), where P is the number of observed variables, then there is multivariate normality. In this study, 9 observed variables were used in all scales except the aggressors’ scale (with 8 items) with a Mardia’s coefficient for skewness of 136 > 9(9 + 2) = 99 and for kurtosis of 395 > 9(9 + 2) = 99 for the victims’ scale, a Mardia’s coefficient for skewness of 48 < 9(9 + 2) = 99 and for kurtosis of 232 > 9(9 + 2) = 99 for the observers of the victims’ scale and a Mardia’s coefficient for skewness of 212 > 9(9 + 2) = 99 and for kurtosis of 578 > 9(9 + 2) = 99 for the observers of the aggressors scale. In the aggressors’ scale we had 8 observed variables with a Mardia’s coefficient for skewness of 182 > 8(8 + 2) = 80 and for kurtosis of 483 > 8(8 + 2) = 80. Hence, considering our skewness and kurtosis values, we used Unweighted Least Squares (ULS) as the method for factor extraction, which is an estimation method that is not dependent on distributional assumptions ( Joreskog, 1977). In order to retain the appropriate number of factors, we applied various factor retention criteria, namely, Velicer’s MAP test and Horn Parallel analyses. According to the literature, these analyses perform optimally in determining the number of factors to extract ( Bandalos & Finney, 2010). By using different methods of extraction, we aimed to propose an approximation to a simple interpretable structure (see Table 2). We considered all items with structure coefficients values above .30 ( Bandalos and Finney, 2010 and Ford et al., 1986). In accordance with the different retention criteria, one factor was obtained for the victims’ scale (with 72% of explained variance), the observers of the victims’ scale (with 75% of explained variance) and for the observers of the aggressors’ scale (with 83% of explained variance). However, in the aggressors’ scale, item 1 presented loadings below .32 on three separate components. We removed item 1 and reran the analysis, hence, obtaining a unidimensional structure of the aggressors’ scale with 85% of explained variance. This preliminary study of the four scales included in the CICS suggested a unidimensional structure of all four scales (victims: α = .96, aggressors: α = .98, observers of victims: α = .97 and observers of aggressors: α = .98), with good reliability scores according to the psychometric literature ( Nunnally, 1978). What’s more, the values of goodness-of-fit (victims: GFI = .99, aggressors: GFI = 1.00, observers of victims: GFI = .99, observers of aggressors: GFI = 1.00), residuals statistics (RMSR = .06, .03, .06, .04, respectively) were also good in accordance with the literature ( McDonald, 1999, Nunnally, 1978 and Velicer, 1976). Table 2. Proposed unidimensional EFA model parameters of the CICS inventory. Proposed EFA modelsa Mardia’s coefficient Kaiser–Meyer–Olkin Bartlett Sphericity GFI RMSR S K Victims’ scale 136 > 9(9 + 2) = 99 395 > 9(9 + 2) = 99 .84 χ236 = 1385.3 (p < .001) .99 .06 Aggressors’ scale 182 > 8(8 + 2) = 80 483 > 8(8 + 2) = 80 .87 χ228 = 2154.8 (p < .001) 1.00 .03 Observers of victims’ scale 48 > 9(9 + 2) = 99 232 > 9(9 + 2) = 99 .88 χ236 = 2080.6 (p < .001) .99 .06 Observers of aggressors’ scale 212 > 9(9 + 2) = 99 578 > 9(9 + 2) = 99 .83 χ236 = 2375.9 (p < .001) 1.00 .04 a Velicer’s Minimum Partial Test used. Horn Parallel Analyses presented same values. Table options 4.2. Measuring the perceived level of involvement in situations of cyberbullying with the Item Response Theory approach We examined the reports of 170 college students’ involvement in situations of cyberbullying with the IRT approach in order to test the unidimensional structure of the scales included in the CICS and in order to understand whether participants underrated this involvement. In the victims scale and in the observers of victims’ scale none of the items showed an infit/outfit higher than 1.5, as well as z statistic higher than 2.00. In the aggressors’ scale, item 4 revealed, an infit/outfit higher than 1.5, and z statistic higher than 2.00. In the observers of the aggressors’ scale, items 2, 4 and 9 revealed an infit/outfit higher than 1.5, and z statistic higher than 2.00. Therefore, we removed these items and reran the analysis for these two scales. For all of the scales, we also present possible models without participants with an infit/outfit higher than 1.5, as well as z statistic than 2.00 in order to see how the instruments would respond (see Table 3). Table 3. IRT parameters of the CICS. Model α Item separation reliability Person separation reliability Victims’ scale Model 1a .86 .91 .60 Model 2b .88 .90 .63 Aggressors’ scale Model 1a .82 .78 .55 Model 3b .72 .87 .55 Model 2c .79 .82 .48 Model 4d .72 .88 .53 Observers of victims’ scale Model 1a .91 .94 .75 Model 2b .93 .95 .79 Observers of aggressors’ scale Model 1a .92 .91 .76 Model 3b .92 .93 .81 Model 2c .94 .85 .83 Model 4d .95 .87 .85 a Model with all participants and all items. b Model without participants with high infit/outfit values. c Model without items with high infit/outfit values. d Model without participants and items with high infit/outfit values. Table options In the victims’ scale item 3 (“They spread rumors about my life.”) was the easiest item to report with a reported/difficulty level of −1.10 log, whereas the most difficult to report was item 9 (“They used my image without authorization.”) with a reported/difficulty level of 2.48 log. The distribution revealed a large range of difficulty (−1.10 < Di < 2.48). In the aggressors’ scale item 5 (“I made fun of someone.”) was the easiest item to report with a reported/difficulty level of −2.78 log, whereas the most difficult to report was item 9 (“I used someone’s image without authorization.”) with a reported/difficulty level of 2.15 log. The distribution revealed a large range of difficulty (−2.78 < Di < 2.15). In the observers of the victims’ scale item 5 (“Someone made fun of them.”) was the easiest item to report with a reported/difficulty level of −1.09 log, whereas the most difficult to report was item 4 (“Someone pretended to be them.”) with a reported/difficulty level of 1.28 log. The distribution revealed a moderate range of difficulty (−1.09 < Di < 1.28). In the observers of the aggressors’ scale item 5 (“They made fun of someone.”) was the easiest item to report with a reported/difficulty level of −1.68 log, whereas the most difficult to report were items 2 and 8 (“They harassed someone with sexual content.” and “They revealed data about someone’s private life.”) with a reported/difficulty level of 1.49 log. The distribution revealed a moderate range of difficulty (−1.68 < Di < 1.49). We also considered other reliability indicators from the Rasch measures for involvement in cyberbullying such as, Cronbach’s alpha, Person Separation Reliability and the Item Separation Reliability. The Person Separation Reliability shows the proportion of the sample variance which is not explained by the measure error, while the Item Separation Reliability indicates the percentage of item variance that is not explained by the measurement error (Smith, 2001). Table 3 shows the Cronbach’s α, the Person Separation Reliability and the Item Separation Reliability for all the scales. These scores indicate good internal consistency/reliability independently of removing participants and items with infit/outfit values higher than 1.5, and z statistic values higher than 2.00 ( Fox & Jones, 1998). The values for Person Separation Reliability in each scale, along with the difficulty indicators, reveal that these students may have had difficulty in responding to all of the items and underrated their involvement in situations of cyberbullying, thus confirming our first hypothesis. 4.3. Dynamics of cyberbullying in college students and their perceived form of involvement After testing the underlying structure of the four scales presented and having examined that college students underrated their level of involvement in situations of cyberbullying, we present a detailed analysis of the dynamics of cyberbullying (including parameters such as age, gender and year of course) in college students and their perceived form of involvement with the use of frequencies and non-parametric tests (e.g. the Mann–Whitney tests and the Kruskal Wallis test). 4.3.1. Victims’ scale In accordance with some of the theoretical issues presented previously, we decided to provide a detailed analysis considering two main aspects of cyberbullying in the victims’ scale, namely acts of intimidation and image appropriation. As seen in the theoretical section, intimidation involves acts that are quite similar to those practiced in the context of bullying, such as threatening, harassing, making fun, insulting and spreading rumors (among others). Image appropriation is more specific of cyberbullying, because it involves the identity and/or image of the victim. Considering the items covered these issues, which were presented in the definitions and characterizations of cyberbullying provided by the literature (Belsey, 2005, Hinduja and Patchin, 2009, Li, 2006 and Willard, 2005), we were interested in analyzing the items individually, and these two aspects separately. In this study, we found 145 (27.94%) participant victims of cyberbullying (143 victims of intimidation and 48 victims of image appropriation), 72.2% of which were female victims. The most frequent aggressions (see Table 4) reported were: “They insulted me” (73.7%), followed by “They spread rumors about my life” (59.3%), “They made fun of me” (55.8%) and “They threatened me” (46.9%). From the victims of Intimidation, 28.9% reported they did not know who the aggressor was, although this type of harassment was more frequently perpetrated by a mixed group (29.3%), followed by single boys (26.0%), and by single girls (17.1%). It is interesting to see the difference between the groups, in comparison to individual situations, verifying that girls in groups (10.6%) were more likely to harm others than boys in groups (1.6%). Image Appropriation was more frequent in boys individually and the mixed group (both 25.6%), followed by girls individually (18.6%), and group of girls (7%). Nonetheless, a great percentage (33.3%) of students did not know who the aggressor was. This last percentage is similar to one of the cyberbullying specificities – anonymity, and in this case, was more frequent in image appropriation (33.3%) than in intimidation (28.9%). The same tendency regarding the behavior of groups was found for image appropriation, with a higher percentage in the group of girls (7%), and no aggressions perpetrated by the group of boys. The percentage of aggressors as classmates of victims was similar in acts of intimidation and image appropriation (53.2% and 55%). Table 4. Types of aggression experienced by victims, committed by aggressors and observed by observers. Items Victims % Aggressors % Observers of victims % Observers of aggressors % Acts of intimidation Acts of image appropriation Threatening someonea 46.9 143 (27.55) – 43.4 44.4 Harassing with sexual contenta 28.9 9.5 35.5 23.6 Spreading rumors about one’s lifea 59.3 16.7 74.2 66 Making fun of someonea 55.8 35.7 30.8 31.2 Insulting mea 73.7 71.4 77.2 82.1 Demonstrating to have information about one’s life that may affect one’s psychological well-beinga 43.4 59.5 73.0 82 Revealing data about one’s private lifea 33.1 19.0 57.4 47.2 Pretending to be someoneb 26.2 – 48 (9.25) 19.0 57.8 50 Using someone’s image without authorizationb 17.9 7.2 47.3 33 Note: All items are presented in the gerund to represent all perspectives of involvement in cyberbullying. a Item of intimidation. b Item of Image appropriation. Table options If we consider the unidimensional structure of the victims’ scale, the last aggression occurred mainly in secondary education (48.3%), then in higher education (34.5%), and lastly in primary education (26.2%). If we examine acts of intimidation and image appropriation separately, the last act of intimidation occurred mainly in secondary education (47.6%), then in higher education (33.6%), and lastly in primary education (25.9%). Regarding acts of image appropriation, the aggressions had a different distribution: 52.1% in secondary, 31.3% in primary, and 22.9% in higher education. Thus, the distribution varies according to the two different aspects. Regarding the technologies used, acts of intimidation and image appropriation had similar frequencies of computer use (68.5% and 72.9%), and use of mobile phones (42% and 37.5%). For acts of intimidation, the preferred methods reported were SMS/MMS (35%), Facebook (30.1%), Messenger (25.9%), Hi5 (22.4%), while Messenger (37.5%), Hi5 (35.4%), SMS/MMS (29.2%), and Facebook (22.9%) were mainly used for image appropriation. Table 5 shows the different types of technology through which victims were offended, considering the level of education the aggression took place. Table 5. Technologies used in victimization/aggression acts. Basic education Secondary education Higher education Victims Aggressors Victims Aggressors Victims Aggressors Computer 71.1 75 70 75 72 90.9 Cell phone 42.1 37.5 44.3 37.5 42 54.5 Blog 7.9 0 8.6 12.5 20 0 Chat 13.2 25 14.3 12.5 14 9.1 Email 10.5 12.5 17.1 12.5 14 9.1 Facebook 7.9 0 25.7 29.2 58 81.8 Hi5 39.5 25 25.7 16.7 10 0 Messenger 36.8 37.5 32.9 25 16 27.3 Myspace 0 0 0 4.2 0 0 Secondlife 0 0 0 0 0 0 SMS/MMS 42.1 25 40 20.8 34 54.5 Youtube 5.3 12.5 4.3 12.5 8 9.1 Table options Victims from primary education refer that the main technologies used were SMS/MMS, Hi5 and Messenger. In the victims from secondary education, the tendency is the same regarding technologies, but the percentages change: SMS/MMS, Messenger, Hi5 and Facebook. In higher education, the cyberbullying escalades through Facebook, continues occurring through SMS/MMS, and increases in Blogs. It is important to refer that the primary education in Portugal is comprised of three cycles, encompassing ages between 6–10, 11–12, and 13–15 years of age. The percentage of aggressions through Facebook and Youtube coincide with the beginning of these technologies in Portugal, therefore, these respondents were most likely in the last cycle of primary education. The most common feelings related with acts of intimidation and image appropriation were: insecurity (48.3%; 64.6%), anger (48.3%; 58.3%), concern (40.6%; 52.1%), sadness (36.4%; 50%), embarrassment (31.5%; 43.8%), and pride (30.1%, 35.4%). Furthermore, 74.6% of victims of acts of intimidation and 75% of the victims of acts of image appropriation tried to prevent the continuation of the situation by “confronting the aggressor”; “avoiding contact with the aggressor”; “excluding the aggressor from the social network” and “stopping answering anonymous calls” (see Table 6). Table 6. Coping strategies used by victims. Coping strategies Acts of intimidation (N = 143) (%) Acts of image appropriation (N = 48) (%) I changed my profile privacy 19.6 27.1 I deleted my Facebook page 4.2 8.3 I confronted the aggressor 35.7 37.5 I contacted the site manager 2.8 4.2 I contacted the police 2.8 4.2 I stopped answering anonymous calls 21.7 22.9 I closed my email account 6.3 6.3 I avoided contact with the aggressor 22.4 20.8 I excluded the aggressor from my social network 21.0 22.9 I ignored the aggression 16.1 10.4 I changed my mobile number 10.5 14.6 I sought help from someone trustworthy 6.3 6.3 I sought professional support 3.5 6.3 I tried to find out who the aggressor was 7.7 14.6 Paid more attention to my computer’s webcam 2.8 4.2 Other 3.5 3.0 Table options Overall, victims of acts of intimidation and image appropriation included as sources of help their friends (72.7% and 81.3%), parents (62.9% and 70.8%), police (44.8% and 41.7%), teachers (28.7% and 35.4%), and classmates (21.7% and 31.3%). Despite these high percentages, the fact is that it occurs little in reality, since only 6.3% of victims reported using the following coping strategy: “sought help of someone trustworthy”. 4.3.2. Aggressors’ scale The sample in this analysis consisted of 8% of respondents (42 students) who were aggressors, 59.5% of which were female aggressors. Their targets were mainly boys (29.3%) and girls (19.5%) individually and mixed groups (19.5%). What’s more, these individuals reported that 45% of the victims were their schoolmates. Hence, there is the same inverted U relationship, as noted earlier in the victims, regarding the level of education the aggressor attended when the cyberbullying was committed. That is, 19% of the aggressions were committed in primary education, while 57.1% were carried out in secondary education and 26.2% in higher education. In terms of the most common aggressions (see Table 4), the following examples illustrate what respondents reported: “I made fun of someone” (71.4%), “I insulted someone” (59.5%) and “I pretended to be someone else” (35.7%). As for the technology used to perpetrate the aggression, respondents mentioned the computer as the main technology used (71.4%), then the mobile phone (35.7%). Facebook (33.3%), Messenger (26.2%) and SMS/MMS (26.2%) prevailed as the preferred digital tools in order to carry out the aggression. As we can see in Table 5, the use of both computer and cell phones increases from primary to higher education and Facebook and SMS/MMS are the most common technologies used in aggressions in higher education. The respondents of the aggressor’s sub-scale reported that anger (63.6%), concern (54.4%) embarrassment (45.5%), insecurity (36.4%), fear and sadness (27.3%) were some of the feelings they believe to have inflicted on their victims. We also analyzed the motives behind the aggressions (see Table 7) and respondents reported mainly that it was: “As revenge regarding past episodes” (54.5%), “Just for fun” and “Because I didn’t like the person’s attitudes” (both 36.4%). Table 7. Motives aggressors mentioned for cyberbullying others. Motives % For the group to accept me 0.0 Just for fun 36.4 For not being able to be personally affirmative to the person 9.1 For revenge regarding past episodes 54.5 Because I wanted to assert myself 18.2 Because the person fits those stereotypes usually mocked 9.1 Because the person has a strange personality 9.1 Because I don’t like the person’s attitudes 36.4 Because there’s no problem acting this way 9.1 Because he belongs to a rival group 27.3 Because if someone abuses me, I can also abuse 27.3 Other 27.3 Table options 4.3.3. Observers of victims’ scale and observers of aggressors’ scale Many respondents (45.7%) reported observing victimization incidents with the majority of acts involving girls individually (47.4%), followed by mixed groups (19.4%) and boys individually (12.1%). The number of the aggressors’ observers was lower (20.4%) because cyberbullying is a phenomenon that allows aggressors to remain anonymous (Slonje & Smith, 2008). The aggressors were mostly observed in the act of cyberbullying in mixed groups (26%), followed by a single girl situation (25%), a single boy situation (20%) and in a group of girls (12%). The victims’ observers reported that the last aggression they observed occurred mainly when the victims were attending secondary education (53.6%), then, higher education (33.2%), and lastly, primary education (16.2%). In terms of the aggressors’ observers, the trend reverses slightly, in the sense that the highest percentage was for secondary education (56.2%), followed by primary education (24.8%), and lastly, higher education (21.9%). According to the acts mentioned by the victims’ observers and the aggressors’ observers, the most common were: “They made fun of someone” (77.2% and 82.1%); “They spread rumors about someone’s life” (74.2% and 66%) and “They insulted someone” (73% and 82%). Table 4 shows the types of aggressions observed by both types of observers. Many of the victims’ observers (54.6%) tried to prevent the continuation of aggressions and the main coping strategies reported were “I tried to support the victim” (44.5%), followed by “I tried to understand the gravity of the situation” (24.2%) and “I advised the victim to tell someone trustworthy” (21.6%). A smaller percentage of the aggressors’ observers (49%) tried to prevent the situation, and the preferred coping strategies were “Prevent the victim” (20.8%), “Dissuade the aggressor” (19.8%), “Denounce the aggressor” (15.1%). 4.3.4. Gender, course, school year and age differences We found that men revealed a greater tendency towards being victims (Mean Rank = 278.44, U = 21,119, p < .05, r = −0.09) and aggressors (Mean Rank = 278.92, U = 21,063, p < .01, r = .10). Then, we found significant differences concerning age. Subjects that were 20 years of age or less were more prone to being observers of victims (Mean Rank = 277.16), χ2(3, N = 519) = 15.22, p = .01, hp2 = .03 and observers of aggressors (Mean Rank = 270.54), χ2(3, N = 519) = 9.97, p = .01, hp2 = .04, than other age groups. We found no significant differences in terms of college year (i.e. years 1, 2 and 3 of the different courses) regarding reported victims and aggressors. Nonetheless, first year students reported a significantly higher tendency to be the observers of victims (Mean Rank = 274.31), χ2(2, N = 519) = 4.97, p = .1, hp2 = .01 and of aggressors (Mean Rank = 276.54), χ2(2, N = 519) = 9.29, p = .01, hp2 = .02. What’s more, we analyzed courses separately because of the different course cultures. We found significant differences between courses. Specifically, the Social Service students were more prone to being victims (Mean Rank = 326.25), χ2(6, N = 519) = 17.72, p = .01, hp2 = .03, whereas students from Social Animation were more prone to being aggressors (Mean Rank = 306.88), χ2(6, N = 519) = 14.11, p = .05, hp2 = .03. In regards to the strategies used by victims, female students (Mean Rank = 260.86, U = 22,771, p = .01, r = .02) and students between the ages of 21 and 23, Mean Rank = 261.50, χ2(3, N = 518) 21.65, p < .001, hp2 = .04, were more prone to contacting the site’s administration, whereas male students were more inclined towards avoiding the aggressors (Mean Rank = 268.31, U = 22777, p = .05, r = .05). Also, students enrolled in courses in Education, Journalism and Social Service (Mean Rank = 264.50) tended to deactivate their email, χ2(6, N = 519) 22.63, p < .001, hp2 = .04, whereas only students in Journalism were more prone to exclude the aggressor from their social network, Mean Rank = 275.50, χ2(6, N = 519) 14.14, p < .05, hp2 = .03. As for the strategies of the observers of victims and aggressors, we found no significant differences in regards to gender, age and course, with the exception of the Science Education students who tried to understand the gravity of the situation as the observers of victims, Mean Rank = 274.00, χ2(6, N = 518) 11.61, p < .10, hp2 = .02. Nonetheless, we found a small effect size for this difference. Generally, the main coping strategies mentioned by all of the students were: “Block contacts” (64.2%), “Inform authorities” (63.5%), “Ask someone you trust for help” (59.1%), and “Change email accounts” (54.6%)”. They also refer to “Contact site managers” (27.1%), “Ignore” (25.1%) and “Create or appeal to a support group” (20.6%). As for finding sources of help, the male students were more inclined towards asking parents (Mean Rank = 276.64, U = 19,465, p = .01, r = .11) and teachers (Mean Rank = 275.93, U = 19,543, p = .05, r = .11) for help than female students. Also, students between the ages of 21 and 23, Mean Rank = 260.20, χ2(3, N = 507) 6.54, p < .10, hp2 = .10, were more prone to turn to colleagues for help, whereas students aged 24–26 were more inclined to ask parents, Mean Rank = 298.23, χ2(3, N = 507) 8.27, p < .05, hp2 = .02, and teachers, Mean Rank = 284.34, χ2(3, N = 507) 9.83, p < .05, hp2 = .02. Also, students enrolled in courses in Social Service were more prone to turn to parents (Mean Rank = 359.35) χ2(6, N = 508) 31.33, p < .001, hp2 = .06, and teachers Mean Rank = 356.23, χ2(6, N = 508) 41.95, p < .001, hp2 = .08. In general, all respondents were asked about who could help resolve cyberbullying situations and the following were mentioned: friends (73.4%), parents (72%), police (59.1%), teachers (43.9%), responsible by the institution (32.7%), colleagues (28.1%) and (6.5%) other person.