Relevant Dimensions of Cyberbullying

Journal of Applied Developmental Psychology 34 (2013) 241–252

Contents lists available at SciVerse ScienceDirect

Journal of Applied Developmental Psychology

Relevant dimensions of cyberbullying — Results from two experimental studies Stephanie Pieschl ⁎, Torsten Porsch 1, Tobias Kahl 2, Rahel Klockenbusch 3 Institut für Psychologie, Westfälische Wilhelms-Universität Münster, Fliednerstr. 21, 48149 Münster, Germany

a r t i c l e

i n f o

Article history: Received 18 September 2012 Received in revised form 22 March 2013 Accepted 15 April 2013 Available online 30 May 2013 Keywords: Cyberbullying Experiment Power imbalance Media Vignettes

a b s t r a c t Cyberbullying is a prevalent problem of adolescents. However, several conceptual and measurement questions, regarding its defining characteristics and relevant dimensions in comparison to conventional bullying, remain unanswered. To this end we conducted two studies with experimental methods. Study I shows that power imbalance in terms of perceived popularity is relevant for the affective, cognitive, and behavioral experience of cyberbullying. Cyberbullying by a popular bully is more distressing than cyberbullying by an unpopular bully. Study II shows that factors unique to cyberbullying are also relevant for the experience of cyberbullying, namely the media and the type of cyberbullying. For example, different types of cyberbullying are related to different patterns of relevant coping strategies. Therefore, cyberbullying seems both a unique phenomenon and closely related to conventional bullying. © 2013 Elsevier Inc. All rights reserved.

Numerous studies have demonstrated that cyberbullying has become an important cross-national phenomenon with an estimated prevalence between 20 and 40% among adolescents (Tokunaga, 2010). In Germany, empirical studies have found cyber-victim prevalence to be between 3 and 36% and cyber-perpetrator prevalence between 5 and 42% (Katzer, Fetchenhauer, & Belschak, 2009a,b; Pieschl & Porsch, 2012; Riebel, Jäger, & Fischer, 2009; Schultze-Krumbholz & Scheithauer, 2009; Staude-Müller, Bliesener, & Nowak, 2009; Wachs & Wolf, 2011). A similar variance can be found internationally. Part of this variance between the findings can be attributed to methodology (Menesini & Nocentini, 2009): Self-reported prevalence is generally lower if adolescents have to report their experience regarding a short period of time (vs. their lifetime) and if it is measured with a single item directly referring to cyberbullying (vs. multiple behavioral items). However, we argue that this apparent methodological problem is confounded with and based on a deeper conceptual problem. The adequate conceptualization and definition of cyberbullying are still highly discussed because the exact composition of the construct of ⁎ Corresponding author. Tel.: +49 2518331386; fax: +49 2518339105. E-mail addresses: [email protected] (S. Pieschl), [email protected] (T. Porsch), [email protected] (T. Kahl), [email protected] (R. Klockenbusch). 1 Now at the LAFP (Landesamt für Ausbildung, Fortbildung und Personalangelegenheiten [Federal Bureau of Education and Human Resources]) of the Police in North-Rheine Westphalia, Germany. 2 Now a student at the Faculty of Psychology, Universität Bielefeld, Germany. 3 Now a student at Institut für Psychologie, Albert-Ludwigs-Universität Freiburg, Germany. 0193-3973/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.appdev.2013.04.002

cyberbullying is still an uncharted territory. One way to approach this problem is to explore which defining characteristics and additional dimensions are relevant for the experience of cyberbullying. In this paper we propose that experimental research could contribute valuable insights into these questions. We think that a definition of cyberbullying has to be based on empirical results as well as on a theoretical foundation. Therefore, we are reporting two exemplary studies – to our knowledge the first experimental studies to be published about cyberbullying – that explore the following conceptual issues: Are mandatory defining characteristics of conventional bullying also relevant for the experience of cyberbullying, specifically power imbalance in terms of perceived popularity (Study I)? Are further cyber-specific dimensions relevant for the experience of cyberbullying, specifically different types and media of cyberbullying (Study II)? Defining characteristics of bullying — The example of power imbalance Conventional bullying is defined as an intentional aggressive act carried out by a group or an individual repeatedly and over time against a victim who cannot easily defend him or herself (Olweus, 1996). Therefore, the three defining characteristics intention to harm, repetition, and power imbalance are central to conventional bullying because they set bullying apart from, for example, rough-and-tumble play, fights between friends, or singular acts of peer aggression. If this definition is transferred into cyberspace, we have to define cyberbullying as “bullying via electronic communication tools” (Li, 2006, p. 158) — including the defining characteristics of intention to harm, repetition, and power imbalance (cf. traditional definitions of cyberbullying, for example Smith et al., 2008).

242

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

One consideration in favor of this equalization is one of the most replicated findings in cyberbullying research, which suggests that the same adolescents are frequently involved in both conventional bullying and cyberbullying (Dempsey, Sulkowski, Dempsey, & Storch, 2011; Gradinger, Strohmeier, & Spiel, 2009; Hinduja & Patchin, 2008; Li, 2007; Riebel et al., 2009; Schultze-Krumbholz & Scheithauer, 2009; Wachs & Wolf, 2011). In Germany, Katzer and colleagues found a correlation of .55 between victims of conventional bullying and victims of cyberbullying (Katzer et al., 2009b) and a correlation of .59 between perpetrators of conventional bullying and perpetrators of cyberbullying (Katzer et al., 2009b), indicating a significant overlap between conventional and cyberbullying. Another consideration in favor of equating conventional bullying and cyberbullying is that some risk factors are the same for both kinds of bullying. Most of these risk factors are unspecific; an existent risk factor does not necessarily predict the specific behavior of cyberbullying. For example, cyber-victims report more personal problems, more peer relationship problems, more family-related problems, and more depressive and somatic symptoms than non-cyber-victims (Gradinger et al., 2009; Schultze-Krumbholz & Scheithauer, 2009; Sourander et al., 2010; Utsumi, 2010; Vandebosch & Van Cleemput, 2008). Furthermore, cyber-perpetrators show more aggression, a positive attitude towards aggression, less empathy, a less positive parent– child relationship, less perceived peer support, more delinquency, more smoking and drinking than non-cyber-perpetrators (Ang & Goh, 2010; Calvete, Orue, Estévez, Villardón, & Padilla, 2010; Katzer et al., 2009a; Schultze-Krumbholz & Scheithauer, 2009; Sourander et al., 2010; Utsumi, 2010). Due to these similarities between conventional bullying and cyberbullying it might be warranted to transfer defining characteristics. Still, this transfer of definition has been discussed controversially and therefore alternative definitions have been suggested in the literature (Dooley, Pyzalski, & Gross, 2009; Slonje & Smith, 2008; Tokunaga, 2010): First, the intention to harm cannot easily be applied to cyberbullying. (Computer-) mediated communication is impoverished in comparison with face-to-face communication (Kiesler, Siegel, & McGuire, 1984). The communication partners do not see body language, gestures or facial expressions and they do not hear prosody which distinguishes, for example, between irony, friendly teasing, and harassment. Therefore, cyber-victims might misinterpret messages intended as fun and cyber-perpetrators on the other hand may not be aware of the consequences of their actions because of this lack of physical and social cues about a target's reactions (Dehue, Bolman, & Völlink, 2008; Menesini & Nocentini, 2009; Raskauskas & Stoltz, 2007). Second, repetition is hard to define in cyberspace. For example, even the single act of uploading an embarrassing video might be considered cyberbullying if it is causing repeated humiliation (Dooley et al., 2009; Menesini & Nocentini, 2009). Third, power imbalance might be implied in the use of technology or take different shapes in cyberspace; it might be indicated by higher technological ability (Nocentini et al., 2010), by a higher rank in the virtual community (Menesini & Nocentini, 2009), or by anonymity (Dehue et al., 2008; Raskauskas & Stoltz, 2007; Vandebosch & Van Cleemput, 2008). It has also been argued that power imbalance might not be as important in cyberbullying as in conventional bullying. Cyber-victims presumably have more (technical) options of preventing and suppressing cyberbullying or of retaliating than victims have in conventional bullying; therefore they might feel less helpless (Nocentini et al., 2010). In the following, power imbalance will be discussed in further detail. Power imbalance has different facets in conventional bullying such as physical dominance, older age, higher social or verbal competence, higher intelligence, or higher social status of the bully (Scheithauer, Hayer, & Bull, 2007). Some of these aspects might also be relevant for cyberbullying, for example the social status of the bully. In conventional bullying, social status has been divided into two relatively independent constructs: Social preference and perceived popularity (Parkhurst &

Hopmeyer, 1998). Social preference (also known as peer acceptance, likability, or sociometric status) describes how much a person is liked by others. Perceived popularity, on the other hand, describes if a person is considered popular in terms of prestige, visibility, or dominance (Caravita, Di Blasio, & Salmivalli, 2009; de Bruyn, Chillessen, & Wissink, 2010). Research indicates that conventional bullying is positively associated with perceived popularity (the “popular bully”) but negatively associated with social preference of the bully (Caravita et al., 2009; de Bruyn et al., 2010; Sijtsema, Veenstra, Lindenberg, & Salmivalli, 2009; Witvliet et al., 2010). It can also be assumed that the perceived popularity of the bully impacts victims' experience of the bullying episode. Power imbalance in terms of perceived popularity is only given if a high status (popular) bully bullies a lower status (less popular) victim. In this scenario victims have fewer means of preventing or adequately coping with bullying and thus they might feel helpless. On the other hand, if bullies and victims have similar power, for example similar perceived popularity, the intended victims could probably prevent bullying or could adequately cope with it and thus not feel as helpless. We empirically investigate these assumptions for cyberbullying in Study I. Unique aspects of cyberbullying — Media and type Besides the compelling empirical evidence that conventional bullying and cyberbullying overlap to a significant degree, there are numerous conceptual differences between conventional bullying and cyberbullying that go beyond the controversies concerning the defining characteristics. Most of these unique features are based on the electronic nature of communication: The cyber-perpetrator can remain anonymous (Slonje & Smith, 2008; Vandebosch & Van Cleemput, 2008) and not directly perceive the consequences of his/her actions (Slonje & Smith, 2008; Willard, 2007). These aspects can trigger disinhibition and might even facilitate cyber-perpetration. For the cyber-victim, escaping cyberbullying is almost impossible because of the omnipresence of electronic communication tools (Slonje & Smith, 2008). Additionally, cyber-bullying incidents can be spread to a large audience (Slonje & Smith, 2008) in a short amount of time and are hard to erase from the internet. Furthermore, there are some difficulties in detecting and reporting cyberbullying due to the lack of adult supervision. There is also some empirical evidence to support the notion that cyberbullying is not merely bullying in cyberspace but a unique phenomenon. The correlations between conventional bullying and cyberbullying are generally only of moderate effect size (see above). This indicates that there are a significant number of adolescents who are only involved in cyberbullying, but not in conventional bullying. Factor analyses demonstrate that acts of cyber aggression load on a unique factor compared with other acts of adolescent aggression (Dempsey et al., 2011). Additionally, there are cyber-specific risk factors for cyberbullying that are not relevant for conventional bullying. For example, both cyber-victimization and cyber-perpetration are related to more computer proficiency, more frequent internet use, more frequent use of electronic communication tools, and more frequent internet risk behavior (Erdur-Baker, 2010; Huang & Chou, 2010; Katzer et al., 2009a,b; Smith et al., 2008; Utsumi, 2010). Despite these findings, research into cyber-specific aspects of cyberbullying is still at its beginning. A unique aspect of cyberbullying that has been frequently discussed and researched is the involved media. For example, Smith et al. (2008) categorized cyberbullying according to seven different media or communication tools because “different media have different characteristics” (p. 377): Text messages, emails, phone calls, photo or video clips, instant massagers, websites, and chat rooms. Research showed that different media were indeed used with different frequencies for cyberbullying and had different effects on cyber-victims' experience of cyberbullying. However, the most frequently used tools for cyberbullying vary across studies. Whereas Smith et al. (2008)

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

found most cyber-bullies used phone calls; Slonje and Smith (2008) found most cyber-bullies used emails. Additionally, Smith and colleagues demonstrated that adolescents consider cyberbullying by photo or video clips to be more distressing than conventional bullying, cyberbullying by phone calls as distressing as conventional bullying, and cyberbullying by all other media less distressing than conventional bullying (Slonje & Smith, 2008; Smith et al., 2008). Recently it has been criticized that this kind of categorization might have become obsolete because the recent advent of the smart phones has caused different kinds of media to converge; media cannot be clearly distinguished anymore (Ortega, Mora-Merchán, & Jäger, 2007). We would like to point out that even though the hardware and technical applications converge other aspects of media do not. More specifically, we propose that one relevant dimension for explaining these effects (Slonje & Smith, 2008; Smith et al., 2008) is the representational code rather than the software applications: Verbal (written or spoken text) and visual codes (pictures and videos) are assumed to be processed differently (Paivio, 1986) and presumably have different effects on the experience of cyberbullying. Another suggested categorization, unique to cyberbullying, was derived from theoretical considerations. Conventional bullying is often subdivided into verbal, physical, and relational bullying (Scheithauer et al., 2007). For cyberbullying, Willard (2007) proposed eight types of “cyberbullying activities and other forms of online social cruelty” (p. 5). We consider five of these to be cyberbullying: harassment (insults or threats against the cyber-victim), denigration (spreading damaging rumors to harm the cyber-victim's reputation), impersonation (assuming a fake identity to impersonate the cyber-victim and behaving in an embarrassing or damaging way), outing and trickery (gaining and then violating the trust of the cyber-victim by publicly announcing private and embarrassing secrets, for example via photos or videos), and exclusion (systematically excluding the cyber-victim from online activities or online groups). We have excluded the other categories of Willard's (2007) taxonomy from our conceptualization of cyberbullying because they concern arguments between equally powerful peers (“flaming”), because we view them as being sub-categories of harassment (“cyberthreats”), or because we consider them more closely related to sexual harassment on the internet than to cyberbullying (“cyberstalking”). In Germany, a cyberbullying questionnaire (and adaptations) was developed based on this taxonomy (Pieschl & Porsch, 2012; Riebel et al., 2009; Wachs & Wolf, 2011). In these empirical studies, denigration and harassment were reported most frequently whereas impersonation, outing and exclusion were rarely reported by adolescents. However, the differential effects of these types of cyberbullying remain unexplored. We empirically investigate some of these uniquely cyber-specific aspects of cyberbullying in Study II. More specifically, we attempt to replicate the media effects that Smith and colleagues detected (Slonje & Smith, 2008; Smith et al., 2008) in a more experimental setting by comparing video-based and text-based incidents. Additionally, we explore the differential effects of different types of cyberbullying proposed by Willard (2007) by comparing the frequent type of harassment with the rare type of outing.

Methodological framework So far, cyberbullying research has used either qualitative methods such as focus groups and interviews of adolescents (Vandebosch & Van Cleemput, 2008) or quantitative survey methods (Menesini & Nocentini, 2009). These methods are well-suited to gather deep insights into adolescents' perspective on cyberbullying as well as representative data on the prevalence of cyberbullying and the relation between cyberbullying and potential risk and protective factors. However, these methods do not allow for conclusions about the

243

causality. Therefore, we chose a more experimental approach to directly address the conceptual issues of interest. We presented adolescents with hypothetical cyber incident vignettes. These scenarios were systematically manipulated to adhere to the experimental factors of interest. Vignettes in general are a valuable research tool because they allow collecting data on how people would act in situations that cannot be investigated with other methodologies because of their sensitive nature or their infrequent occurrence (Collett & Childs, 2011; LaFontana & Cillessen, 2010). However, the validity of this approach is a crucial issue: It has been shown that the emotional experience of real situations manipulated in the lab is more intense than reading hypothetical vignettes (Collett & Childs, 2011). These findings point to the general validity of this approach with regard to the type of effects but also to a potentially reduced size of effects. For cyberbullying, vignettes are a promising methodology to experimentally manipulate cyberbullying incidents, avoiding many of the methodological and ethical challenges of re-creating cyberbullying in the laboratory. Additionally, research on conventional bullying also successfully utilized vignettes for exploring conceptual questions. For example, vignettes have been used to explore children's views about their emotions (and their parents' reactions) as results of being a hypothetical bully (Ttofi & Farrington, 2008), parents' responses to hypothetical children's victimization events (Bonnet, Goossens, & Schuengel, 2011), and teachers' attitudes towards different kinds of bullying incidents (Bauman & Del Rio, 2006; Yoon, 2004). Our experimental approach can be interpreted within the framework of the General Aggression Model (GAM; Carnagey & Anderson, 2003) which is often used to explain effects of violent media consumption. The GAM assumes that the reaction to a specific incident depends on personal and situational characteristics as inputs. We systematically varied the situational characteristics via the vignette methodology. Additionally, we captured selected personal characteristics such as adolescents' age, sex, media use, and cyber(bullying) experiences. These inputs are assumed to be processed via three interrelated routes, namely cognition, affect, and physiological arousal which describe the current internal state of a person. Therefore, we selected our dependent variables accordingly: We captured affect via the Current Mood Scale (Dalbert, 1992); in Study I we also captured cognition via the Helpless Cognition Scale (Breitkopf, 1985). We did not measure physiological arousal. After appraisal this internal personal state is assumed to translate into action. To capture this behavioral facet, we asked adolescents about their intended coping strategies (Coping Strategies Question(naire)). Study I: Is power imbalance relevant to cyberbullying? Power imbalance has many facets; one of them is the social status of the perpetrator in terms of perceived popularity. We explored the difference between cyber incidents caused by a popular versus an unpopular bully in an online field experiment with a within-subject design. The following hypotheses and exploratory research questions were analyzed. Hypothesis 1: The Social Status of the cyber-bully has effects. Cyber-victims' affect (mood), cognitions (helplessness), and behavior (coping strategies) differ between being harassed by a popular cyber-bully and being harassed by an unpopular cyber-bully; being harassed by a popular cyber-bully is more distressing. Hypothesis 2: The difference between being harassed by a popular or unpopular cyber-bully is not independent from the Order of cyber incidents. By being repeatedly confronted with negative cyber experiences, cybervictims become desensitized. The first experience is more distressing than the second. Exploratory Research Question 3: Does the experience of cyber incidents vary according to personal characteristics, specifically the sex, age, perceived popularity, and cyberbullying experience of the cyber-victim, the degree of liking between the cyber-victim and the fictitious cyber-bully, and the sex of the fictitious cyber-bully?

244

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

Method Procedure All students were confronted with two cyber scenarios in a two-by-two design. The Social Status of the cyber-bully in terms of perceived popularity was one within-subject factor. In one experimental condition a mean cyberbullying message was supposedly written by the most popular classmate (Bully+) in the other condition the entry was supposedly written by the most unpopular classmate (Bully−) of the students. The order of presentation was varied randomly in two between-subject conditions. Students in Group A (+/−) received the Bully + scenario first, followed by the Bully − scenario whereas students in Group B (−/+) received the Bully − scenario first, followed by the Bully + scenario. For each scenario students answered several questionnaires: the Current Mood Scale, the Helpless Cognition Scale, and the Coping Strategies Question. Afterwards they answered the Cyber Experience Questionnaire. On average students needed 18 min to complete these tasks. This study was conducted online; students could fill in the web-based questionnaires any time they wanted to. All students participated voluntarily and gave their informed consent. Among all participants six Amazon vouchers were raffled off as an incentive. At the end of the questionnaire it was explicitly pointed out that this was a fictitious experiment, we gave contact information for further questions, and all participants were informed of the results if they wished. Participants We recruited a voluntary online convenience sample of students between the ages of 12 to 19 through email and social network sites such as Facebook and schülerVZ (a German online social network). In total 465 participants started to answer the online questionnaires. However, 184 students only answered part of the questionnaires, 76 students were faster than 8 min which seemed unreasonable given the length of the questionnaires. Twelve students displayed no variance in their answers between the first and the second scenario even though the scales were reverse-coded which indicates that they did not seriously answer the questionnaires both times, and the open answers of 7 students indicated that they did not understand the instructions. Therefore, all the data of these students were excluded from further analyses and the final sample consists of n = 186 students. These 56 boys and 130 girls were between 12 and 19 years old (M = 15.88, SD = 2.06) and on average attended 10th grade (M = 9.88, SD = 2.16). Within the three-tier German school system, the majority (n = 128; 69 %) attended the highest track, 19 students (10 %) attended the middle track, and 3 students (2 %) attended the lowest track. Furthermore, 18 students (10 %) attended a comprehensive school that combines all tracks, 9 students (5 %) attended vocational school, and 9 students (5 %) indicated other types of education. On average students considered themselves fairly popular (M = 2.31, SD = .84; on a scale from 1 = almost nobody in my class considers me popular to 4 = almost everyone in class considers me popular).

visible to the public; it was posted by [insert name of the nominated student]. This entry read ‘You are nasty filth. At your birth the doctor had to puke after seeing how ugly you are.’”. Current mood scale. The ASTS (Aktuelle Stimmungsskala; Dalbert, 1992) is a scale with 19 adjectives; all adjectives had to be rated on 7-point scales regarding agreement (1 = low, 7 = high). Students had to answer all questions in reference to the fictitious cyber scenarios. The instrument consists of six subscales. The subscale Sorrow consists of three items (miserable, sad, and unhappy) and was reliable in all experimental conditions (Cronbach's α = .91–.93). The subscale Despair consists of three items (hopeless, desperate, and discouraged) and was reliable in all experimental conditions (Cronbach's α = .87–.94). The subscale Tiredness consists of four items (tired, weary, exhausted, and devitalized) and was reliable in all experimental conditions (Cronbach's α = .81–.90). The subscale Positive Mood consists of five items (comfortable, joyous, cheery, amused, and happy [the sixth item “frohgemuht” was excluded because it is an antiquated German term that is not generally understood by adolescents]) and was reliable in all experimental conditions (Cronbach's α = .84–.89). The subscale Anger consists of three items (angry, annoyed, furious) and was reliable in all experimental conditions (Cronbach's α = .88–.91). All scales were (re-)coded in a way that high values represent negative mood. These subscales were also combined into the overall ASTS scale Negative Mood which was also reliable in all experimental conditions (Cronbach's α = .91–.93). Helpless cognition scale. The HiS (Hilflosigkeitsskala; Breitkopf, 1985) is a scale with 20 items that have to be rated on 7-point scales regarding agreement (1 = low, 7 = high). For this study, an abridged version with 9 items was used (sample items: “I feel helpless”, “My situation is hopeless”, “No matter what I do nothing will change”). Students had to answer all questions in reference to the fictitious cyber scenarios. A confirmatory factor analysis confirmed the onedimensional structure of this scale. In the different experimental condition it explained between 52 and 70 percent of variance and was reliable (Cronbach's α = .86–.94). Coping Strategies Question. In an open question students were asked to elaborate what they would do if they were the victim of the fictitious cyber scenarios. These open answers were categorized as Social (seek support from adults or peers), Aggressive (online retaliation), Passive (ignore incident), Technical (notify the internet service provider, change account settings, or block contact), or Helpless coping (does not know what to do) and the additional categories of Confrontation (seek an active problem-solving contact with the perpetrator), Rationalization (seek an explanation, for example “it's my own fault” or “the perpetrator has problems”), and Depreciation (devalue the incident, for example “it's only a joke”). Two raters independently categorized the open answers of 40 students and agreed almost perfectly (Cohen's Kappa: Bully+ = .92; Bully− = .86). Therefore, the open answers of the remaining 146 students were categorized by only one of the two raters.

Material Cyber scenario vignettes. All students were asked to nominate the most popular student as well as the most unpopular student within their class by their first names. For example, one student could have nominated “Bob” as being the most popular student and “Lisa” as being the most unpopular student. Furthermore, students had to indicate the sex of these nominated classmates (male or female) as well as their personal liking (1 = I don't like [insert name of the nominated student] to 4 = I do like [insert name of the nominated student]). These names were automatically inserted into the following cyber scenario: “Imagine you logged into schülerVZ (a German online social network), looked at your bulletin board and saw a new entry that was

Cyber experience questionnaire. We adapted the cyberbullying questionnaire of Riebel et al. (2009) to include the following five of Willard's (2007) categories: harassment, denigration, impersonation, outing, and exclusion. Students were asked how often these incidents have happened to them via cell phone or via the internet (cyber-victim) and how often they had instigated such incidents themselves (cyber-perpetrator). All answers were given on 3-point scales with the categories never (= 0), once (= 1), and several times (= 2). Cyber Involvement was diagnosed if students gave at least once a different answer than never (cyber-victim and cyber-perpetrator). Note that we explicitly do not refer to these incidents as cyberbullying because not all defining criteria such as repetition are met.

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

100

Results

90

Results regarding the experimental conditions (Hypotheses 1 and 2) For each of the dependent variables – the current mood scales Sorrow, Despair, Tiredness, Positive Mood, Anger, and overall Negative Mood as well as the score of the Helpless Cognition Scale – we computed (M)ANOVAs with the within-subject repeated-measure factor Social Status of the cyber-bully (Bully+ vs. Bully−) and the between-subject factor Order of presentation (Group A vs. Group B). Additionally, students' open answers regarding the Coping Strategies Question were analyzed with Chi-Square and McNemar tests. For the corresponding descriptive results see Table 1 and Fig. 1. The ANOVA for overall Negative Mood indicates a significant main effect of the repeated-measure factor Social Status of the cyber-bully, F (1, 184) = 94.61, p b .001, ηp2 = .34, and a significant interaction between Social Status and the presentation Order of the cyber scenarios, F (1, 184) = 12.52, p b .001, ηp2 = .06; see Fig. 2, left. Table 1 Descriptive results for the Current Mood (Sub-)Scales. Current mood scales Sorrow Despair Tiredness Positive mood (reversed) Anger Negative mood (overall)

Bully+ M (SD) 4.04 2.97 2.44 6.25 4.50 4.20

(1.93) (1.86) (1.48) (1.12) (1.77) (1.22)

Bully− M (SD) 2.78 1.97 1.85 5.65 3.87 3.43

(1.75) (1.38) (1.09) (1.48) (1.88) (1.15)

Note. All answers were given on a scale from 1 = Positive Mood to 7 = Negative Mood.

80

Percent (%)

Descriptive results The random assignment of experimental conditions was successful. Approximately the same number of students was assigned to each condition (Group A: n = 91; Group B: n = 95) and these groups did not differ systematically in sex, χ2 (1) = 2.17, p = .153; age, t (184) = .730, p = .466; or perceived popularity, t (184) = −1.117, p = .266. Furthermore, students in both experimental groups (Group A vs. Group B) nominated popular and unpopular students who did not differ systematically in sex, Bully+: χ2 (1) = .01, p = .922; Bully−: χ2 (1) = .23, p = .630, or in being liked by the participating students, Bully+: t (184) = −1.123, p = .263; Bully−: t (184) = .916, p = .361. Therefore, all requirements for further analyses were met, all descriptive results will be reported for the overall sample, and the variables sex, age, and perceived popularity of the participating students as well as sex and being liked of the nominated popular and unpopular students will not be considered in the testing of Hypotheses 1 and 2. On average students' most pronounced mood after reading all cyber scenarios was Anger, followed by Sorrow, Despair, Tiredness, and Positive Mood (see Table 1). Overall, students reported low Helpless Cognitions (Bully+: M = 1.95, SD = 0.90; Bully−: M = 1.74, SD = 0.90) and only few coping strategies (Coping Strategies Question: Bully+: M = 2.03, SD = 0.94; Bully−: M = 2.15, SD = 0.95). Most often students indicated coping strategies that were categorized as Confrontation, followed by Social and Technical coping and Depreciation (see Fig. 1). Helpless, Passive, and Aggressive coping as well as Rationalization were indicated less frequently. Of the whole sample 132 students (71%) were categorized as cyber-victims and 94 students (51%) as cyber-perpetrators. Cyber-victimization and perpetration were significantly related (Phi coefficient: rφ = .22, p = .003). Seventy-six students (41%) were cyber-victims as well as cyber-perpetrators. Students most often nominated girls as popular (girls: n = 109, 59% vs. boys: n = 77, 41%) and unpopular classmates (girls: n = 105, 57% vs. boys: n = 81, 43%). The popular students were liked significantly more by the participating students (M = 3.23, SD = 0.79) than the unpopular students, M = 2.13, SD = 0.84; t (185) = 13.06, p b .001.

245

Bully+ Bully-

70 60 50 40 30 20 10 0

Coping Strategy Categories Fig. 1. Percentage of students indicating coping strategy categories in the Coping Strategies Question in the two experimental conditions Bully+ and Bully−.

The MANOVA regarding all Current Mood Sub-Scales yields similar effects. We found a significant multivariate main effect of the repeated-measure factor Social Status of the cyber-bully, F (5, 180) = 20.09, p b .001; ηp2 = .36, and a multivariate interaction between Social Status and the presentation Order of the cyber scenarios, F (5, 180) = 3.94, p = .002; ηp2 = .10. The multivariate main effect of Social Status was replicated univariately on all subscales, Sorrow: F (1, 184) =88.34, p b .001, ηp2 = .32; Despair: F (1, 184) = 66.57, p b .001, ηp2 = .27; Tiredness: F (1, 184) = 45.56, p b .001, ηp2 = .20; Positive Mood: F (1, 184) = 36.80, p b .001, ηp2 = .17; Anger: F (1, 184) = 24.56, p b .001, ηp2 = .12. The multivariate interaction between Social Status and Order of presentation was univariately replicated on the scales Sorrow, F (1, 184) = 11.92, p = .001, ηp2 = .06; Despair, F (1, 184) = 10.43, p = .001, ηp2 = .05; and Anger, F (1, 184) = 13.15, p b .001, ηp2 = .07. In all cases the experimental condition Bully+ resulted in more negative mood than the experimental condition Bully−. In all cases of interactions the presentation Order in Group A (+/−) resulted in more pronounced differences between the experimental conditions Bully+ and Bully− whereas the presentation Order in Group B (−/+) resulted in less pronounced differences. The ANOVA for the Helpless Cognition Scale yielded a significant main effect of the repeated-measure factor Social Status of the cyber-bully, F (1, 184) = 14.78, p b .001, ηp2 = .07, and a significant interaction between Social Status and the presentation Order of the cyber scenarios, F (1, 184) = 74.89, p b .001, ηp2 = .29; see Fig. 2, right. On average the experimental condition Bully+ resulted in more helplessness than the experimental condition Bully −. The interaction indicates, however, that in both experimental groups the confrontation with the first cyber scenario resulted in more helplessness than the confrontation with the second cyber scenario. McNemar tests show that students indicate significantly more Helpless coping in the Bully+ condition than in the Bully− condition, χ2 (1) = 16.33, p b .001; Bully+: n = 31, 17 % vs. Bully−: n = 10, 5 %. On the other hand, students indicate significantly more Aggressive coping, χ2 (1) = 9.53, p = .003; Bully−: n = 36, 19 % vs. Bully+: n = 18, 10 %, and Depreciation, χ2 (1) = 15.75, p b .001; Bully−: n = 75, 40 % vs. Bully+: n = 44, 24 % in the Bully− condition than in the Bully+ condition (see Fig. 1). Chi-Square tests show no significant effects of the Order of presentation (Group A vs. Group B). Exploratory results (Research Question 3) In these exploratory analyses only two aggregated dependent variables were considered in both within-subject experimental conditions separately (Bully+ and Bully−) — the overall Negative Mood score from the Current Mood Scale and the score on the Helpless Cognition Scale. These dependent variables constitute criteria in four

246

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

Fig. 2. Negative Mood (left) and Helpless Cognition (right) by Social Status of the cyber-bully (Bully+ vs. Bully−) and presentation Order (Group A vs. Group B). For both scales ANOVAs yielded significant main effects of Social Status and significant interactions between Social Status and Order.

separate regressions. All potentially relevant descriptors of the participating students (sex, age, grade, perceived popularity, and experience as cyber-victim and as cyber-perpetrator), all potentially relevant descriptors of the fictitious Bully+ or Bully− (sex and the degree of being liked by the participating student), as well as the Order of presentation (Group A vs. Group B) were used as predictors in these regressions. Because of the exploratory (and not confirmatory) nature of this research question, aimed at understanding the most parsimonious explanation of these criteria, stepwise regressions were utilized (Thayer, 2002). This method only enters variables that significantly contribute to the regression model; all predictors that are not included in the final regression models do not significantly explain additional variance. The stepwise regression for Negative Mood in the Bully+ condition shows that (in the order of inclusion) if the Bully+ is a girl and if the participating student is unpopular, Negative Mood is more pronounced after reading the fictitious cyber scenario (see Table 2). The stepwise regression for Helpless Cognitions in the Bully + condition shows that (in the order of inclusion) if the participating student is unpopular, if the Bully+ scenario is presented first (vs. second), and if the student is young, Helpless Cognitions are more likely to occur after reading the fictitious cyber scenario (see Table 3). The stepwise regression for Negative Mood in the Bully − condition shows that (in the order of inclusion) if the participating student likes the Bully −, if the participating student is unpopular, if the Bully − scenario is presented first (vs. second), and if the participating student is a girl, Negative Mood is more pronounced after reading the fictitious cyber scenario (see Table 4). The stepwise regression for Helpless Cognitions in the Bully − condition shows that (in the order of inclusion) if the Bully − scenario is presented first (vs. second), and if the participating student is unpopular, young, and a girl, Helpless Cognitions are more likely to occur after reading the fictitious cyber scenario (see Table 5).

The interaction effects between the factors of Social Status (Bully+ vs. Bully−) and presentation Order (Group A vs. Group B) all indicate that the confrontation with the second cyber incident is not as distressing as the confrontation with the first. This effect was most pronounced for Helpless Cognitions which seem to be most susceptible to desensitization. Our Exploratory Research Question 3 also yielded interesting results. All indicators of distress were more pronounced if the participating student was unpopular (4/4 effects). Furthermore, being bullied by an unpopular cyber-bully (vs. a popular cyber-bully) was experienced to be more distressing for girls than for boys (2/2 effects) and younger students experienced more helpless cognitions (but not more negative mood; 2/2 effects). This shows that personal characteristics of the cyber-victim moderate the effects of cyber incident scenarios. In this context perceived popularity, sex, and age seem to be more important than the personal prior experience as a cybervictim or as a cyber-perpetrator. Additionally, personal characteristics of the hypothetical cyber-bully did not consistently elicit effects (1/4 effects of sex; 1/4 effects of being liked) and seem to play a secondary role. Despite these promising results this study has some specific limitations that need to be considered: We recruited the participants of this study via social network sites. Therefore, this study constitutes an uncontrolled field experiment. Even though we know that the majority of German adolescents uses the internet daily, most often to communicate via online communities (Medienpädagogischer Forschungsverbund Südwest (mpfs), 2011), we have no information about the representativeness of the final sample (Granello & Wheaton, 2004). Furthermore only the data of 40 % of the participants who started the online questionnaire could be used for analyses. Therefore, we do not know if selective drop-out might be another potential problem. Even though more girls than boys use social networking sites in Germany (mpfs, 2011), the uneven sex distribution in the final sample could be a sign of selective drop-out that might have influenced the results.

Discussion of Study I We confirmed our Hypothesis 1. Being harassed by a popular cyber-bully is indeed more distressing than being harassed by an unpopular cyber-bully. More specifically, this study indicates that in these scenarios cyberbullying by a popular cyber-bully elicited more Negative Mood (regarding all subscales), more Helpless Cognitions, and a different pattern of Coping Strategies, namely more Helpless coping, less Aggressive coping and less Depreciation. We confirmed our Hypothesis 2. By being repeatedly confronted with negative cyber incident vignettes cyber-victims become desensitized.

Table 2 Stepwise regression models for explaining Negative Mood in the Bully+ condition.

Step 1 Bully+ sexa Step 2 Bully+ sexa Student popularity

B

SE B

ß

T value

.735

.174

.297

4.221***

.691 −.245

.173 .102

.280 −.168

4.001*** −2.403*

*p b .05. **p b .01. ***p b .001.

a

1 = boy and 2 = girl.

Adj R2

F change

.083

17.813***

.106

5.775*

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252 Table 3 Stepwise regression models for explaining Helpless Cognitions in the Bully+ condition.

Step 1 Student popularity Step 2 Student popularity Presentation Ordera Step 3 Student popularity Presentation Ordera Student age

B

SE B

ß

T value

−.386

.075

−.339

−4.889***

−.347 −.349 −.319 −.371 −.079 a

*p b .05. **p b .01. ***p b .001.

.073 .123 .073 .121 .030

−.323 −.194 −.297 −.206 −.181

Adj R2

F change

.110

23.906***

.143

8.092**

.171

7.145**

−4.732*** −2.845** −4.377*** −3.062** −2.673**

1 = Group A (+/−) and 2 = Group B (−/+).

Additionally, this study used hypothetical scenarios. Therefore, these results are only valid if participants successfully put themselves in the position of the cyber-victim. We have no empirical confirmation of the validity of this approach, except for finding a reasonable pattern of results. We also do not know if these findings transfer to reality, namely if adolescents experience similar feelings and cognitions and display similar behaviors when confronted with real cyberbullying. For example, we do not know if sex effects can be generalized to real-world cyberbullying. One potential limitation is that girls might be better at putting themselves in these hypothetical scenarios than boys; another potential limitation is that girls might be better at expressing their emotions than boys whereas the experience of emotions might be similar. Another limitation concerns the very subtle manipulation of the cyber-bully's Social Status. The only differences between the experimental within-subject conditions (Bully+ vs. Bully−) were the inserted names of the fictitious cyber-bullies. We can only speculate if all of the participants adequately noticed this difference – if not our detected effects might underestimate the real effects – or if participants might have succumbed to a bias of social desirability – if yes our detected effects might overestimate the real effects. In a related issue, we do not know if the effects of Social Status transfer to other age groups given that the priority of peer reputation peaks during adolescence (LaFontana & Cillessen, 2010). Finally, we do not know if participants become desensitized to repeated cyberbullying, to being exposed to the same experimental treatment repeatedly, or become bored with answering the same questionnaire battery twice. Despite these limitations we can draw some tentative conclusions. This is to our knowledge the first study that could demonstrate that power imbalance is in fact relevant for the experience of cyber cruelty and that the cyber-bully's perceived popularity constitutes a relevant aspect of power imbalance in cyberspace as well as in conventional

Table 4 Stepwise regression models for explaining Negative Mood in the Bully− condition. B Step 1 Bully− being liked Step 2 Bully− being liked Student popularity Step 3 Bully− being liked Student popularity Presentation Ordera Step 4 Bully− being liked Student popularity Presentation Ordera Student sexb

SE B

ß

T value

.407

.096

.298

4.241***

.395 −.354

.093 .093

.290 −.259

4.261*** −3.806***

.416 −.380 .538

.090 .091 .152

.305 −.277 .235

4.614*** −4.196*** 3.548***

.397 −.356 .495 .383

*p b .05. **p b .01. ***p b .001. boy and 2 = girl.

a

.089 .090 .151 .165

.291 −.260 .216 .154

Adj R2

F change

.084

17.989***

.147

14.487***

.197

12.587***

.216

5.389*

4.435*** −3.955*** 3.275** 2.321*

1 = Group A (+/−) and 2 = Group B (−/+).

b

1=

247

Table 5 Stepwise regression models for explaining Helpless Cognitions in the Bully− condition. B Step 1 Presentation Ordera Step 2 Presentation Ordera Student popularity Step 3 Presentation Ordera Student popularity Student age Step 4 Presentation Ordera Student popularity Student age Student sexb

SE B

ß

T value

.592

.125

.331

4.753***

.630 −.277

.121 .072

.352 −.258

5.225*** −3.833***

.611 −.252 −.070

.119 .072 .029

.341 −.235 −.159

5.122*** −3.500** −2.377*

.583 −.235 −.069 .263

.119 .072 .029 .130

.326 −.220 −.159 .135

4.893*** −3.276** −2.385* 2.026*

*p b .05. **p b .01. ***p b .001. boy and 2 = girl.

a

Adj R2

F change

.105

22.594***

.167

14.693***

.187

5.652*

.201

4.107*

1 = Group A (+/−) and 2 = Group B (−/+).

b

1=

bullying. This is also, to our knowledge, the first study that experimentally investigated the effects of repeated exposure to online cruelty. Further research is needed to clarify if students in fact become desensitized to cyberbullying (see above). Desensitization could offer one explanation for the surprising finding that some adolescents do not seem to be affected by cyberbullying (Patchin & Hinduja, 2006); another explanation could be trait-like dismissal. Last but not least, our results underline that not all adolescents are equally affected by cyber incidents and that one's own perceived popularity as well as age and sex seem to be important moderators of these effects. However, further research is needed to determine if these effects hold true in more representative samples and if, for example, other types of power imbalance are equally relevant for cyberbullying such as the cyber-bully's technical internet skills or his or her anonymity. Study II: Are cyber-specific issues such as media and type relevant for the experience of cyber incidents? In this study two factors with pre-selected characteristics were investigated in a classroom field experiment with a two-by-two between-subject design: Medium (video vs. text) and Type of cyber incident (harassment vs. outing). The following hypotheses and exploratory research questions were addressed. Hypothesis 1: Different Media used for cyber incidents have different effects. Videos used for cyber incidents are more distressing for cyber-victims than texts. Consequently, they result in more negative affect (mood) and more planned behavior (coping strategies). Hypothesis 2: Different Types of cyber incidents have different effects. The affect (mood) and planned behavior (coping strategies) of cyber-victims differ between cyber harassment and cyber outing incidents. Exploratory Research Question 3: Does the cyber-victim's sex moderate the effects of Medium and Type of cyber incidents? Because of the strong effects of sex in study I this variable was included in all analyses of Study II. We did not include prior cyber(bullying) experience in this exploratory research question because we found no significant correlations between these variables and any dependent variable in this study. Method Procedure The study was conducted during students' regular classes. In each classroom two trained investigators collected the data in a standardized manner. One investigator read a standardized instruction and answered any (comprehension) questions. The other investigator dispersed the questionnaires. Students were randomly assigned to one of the four experimental conditions by handing out four different

248

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

versions of the questionnaire battery. Fictitious cyber scenario vignettes were developed for all experimental conditions of the two by two between-subject design (Media × Type of cyberbullying): HT (harassment via text), HV (harassment via video), OT (outing via text), and OV (outing via video). All students answered a questionnaire battery consisting of the Current Mood Scale, the Coping Strategies Questionnaire, the Cyber Experience Questionnaire, and the Internet Use Questionnaire. On average students needed approximately 25 min to answer all questions. All students participated voluntarily and gave their informed consent; furthermore, the school's headmaster gave his informed consent in loco parentis. All participants received sweets for their participation. At the end of the study the trained investigators disclosed the purpose of the study, pointed out that the cyber scenarios were fictitious, and answered all remaining questions. Participants We recruited a convenience sample of 138 students from three 6th grade classrooms and three 7th grade classrooms of a smalltown comprehensive school in Germany. Because of missing values or no variance in response, data from eleven students had to be excluded from further analyses. Therefore, the final sample consists of n = 127 students. These 65 boys and 62 girls were between 11 and 15 years old (M = 12.36, SD = 0.89). Within the three-tier German school system, 114 (90 %) of these students attended the middle track whereas 13 (10 %) attended the lowest track. Material Cyber scenario vignettes. Fictitious cyber scenario vignettes were developed and systematically manipulated according to the experimental conditions Type (harassment vs. outing) and Media (text vs. video) of cyberbullying. Four distinct experimental conditions resulted: HT (harassment via text), HV (harassment via video), OT (outing via text), and OV (outing via video). In the HT condition students were told that a written insult against them was posted on a public website and all students in their class received the link. The insult read “You are nasty filth. At your birth the doctor had to puke after seeing how ugly you are.” (See Study I.). In the HV condition the same insult was supposedly posted as a video message. In the OT condition students were told that their whole class received a message that disclosed an embarrassing secret about them. First, they had to imagine this fictitious secret: “You are in love with someone but nobody knows about that yet — not even the subject of your love. Someone overheard you finally telling somebody else on the telephone. The next day the message reads ‘Hey everyone! Guess what I overheard yesterday. [insert your name] is in love with [insert name of the subject of your secret love]. [insert name of the subject of your secret love] does not know it yet … but [insert your name] thinks of him/her constantly.’” In the OV condition the same secret was supposedly disclosed in a video about your phone conversation that was sent to everyone in class. Current mood scale. The ASTS (Aktuelle Stimmungsskala; Dalbert, 1992) is a scale with 19 adjectives; all adjectives had to be rated on 7-point scales regarding agreement (1 = low – 7 = high). Students had to answer all questions in reference to the fictitious cyber scenario. This instrument was adapted to the current study by adding the adjectives “troubled” and “scared”. Therefore, an exploratory factor analysis was computed and revealed three factors explaining 55.85% of variance: The scale Sad Mood (SM) consists of 10 adjectives (examples: desperate, miserable, scared, hopeless) and is reliable (Cronbach's α = .90). The scale Angry Mood (AM) consists of 3 adjectives (angry, annoyed, furious) and is reliable (Cronbach's α = .80). Finally, the scale Joyous Mood (JM) consists of 4 adjectives (comfortable, joyous, cheery,

amused) and is reliable (Cronbach's α = .67). All scales were (re-)coded in a way that high values represent negative mood. Coping Strategies Questionnaire. This questionnaire started with an open question where students were asked to elaborate what they would do if they were the victim of the fictitious cyber scenario. These open answers were categorized as Social (seek support from adults or peers), Cognitive (seek an active problem-solving contact with the perpetrator), Aggressive (online retaliation), Passive (ignore incident or helpless behavior), Technical (notify the internet service provider, change account settings, or block contact), and Legal coping (go to the police) and the additional categories of Identify bully (find out the identity of the perpetrator) and Defense (deny any problem or explain the situation to the audience). Because (partly) different cyber vignettes were used in Study I, the categorization of the students' answers was done for each study separately and thus not all categories were included in both studies. Two raters independently coded all open answers and for all categories inter rater agreement was almost perfect (Cohen's Kappa = .91–1.00). In the second part of this questionnaire, students had to rate the likelihood of employing 20 pre-defined coping strategies on 5-point scales (1 = very unlikely–5 = very likely). Items cover a range of coping strategies, for example “I go to the police”, “I show the incident to an adult”, “I try to delete the message” or “I try to solve the issue in class.” Exploratory factor analysis did not reveal a meaningful structure. Therefore, all items were aggregated into a mean value that denotes the overall extent of Active Coping. This aggregated scale proved to be reliable (Cronbach's α = .82). Cyber experience questionnaire. We adapted the cyberbullying questionnaire of Riebel et al. (2009) to include the following five of Willard's (2007) categories: harassment, denigration, impersonation, outing, and exclusion. Students were asked how often these incidents have happened to them via cell phone or via the internet in the last two months (cyber-victim) and how often they had instigated such incidents themselves (cyber-perpetrator). All answers were given on 5-point scales with the categories never (= 0), once or twice (= 1), several times a month (= 2), about once a week (= 3), and several times a week (= 4). In order to also derive the indicators of cyberbullying, including the criterion of repetition, we used this fine-grained answering scale. Two indicators of involvement in cyber incidents were derived from the students' answers: General Cyber Involvement was diagnosed if students gave at least once a different answer than never (cyber-victim and cyber-perpetrator). Cyberbullying was diagnosed including the criterion of repetition by computing a sum score across all categories. If this score was 3 or higher, students were assigned the appropriate category (cyberbullying-victim and cyberbullying-perpetrator). This was the case if students indicated that one type of incident happened at least once a week or, for example, if three different types of incidents happened once or twice. Internet Use Questionnaire. Students' Internet Use was explored by asking them how often they used nine internet applications that are popular in this age group (instant messengers, schülerVZ [a German social network], email, chat, video portals, blogs, Facebook, Google, and internet forums). Answers were recorded on 6-point scales (1 = never–6 = several times a day). Results Descriptive results The random assignment of experimental conditions was successful. Approximately the same number of students was assigned to each condition (HT: n = 29; HV: n = 33; OT: n = 32; OV: n = 33) and these groups did not differ systematically in sex, χ2 (3) = .27, p = .966, or age, F (3, 123) = .798, p = .503. Therefore, all necessary conditions

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

for further analyses are met and all descriptive results will be reported for the overall sample. On average students' most pronounced mood after reading the cyber scenarios was Angry Mood (AM: M = 5.29, SD = 1.41), followed by Sad Mood (SM: M = 3.51, SD = 1.36) and Joyous Mood (JM: M = 1.21, SD = 0.51). Furthermore, on average the students reported just about two coping strategies in their open answers (M = 1.70, SD = 0.91) and their average Active Coping score was M = 2.79 (SD = 0.62). More specifically, students' open answers indicate that, on average, they were most likely to employ Social coping and least likely to employ Legal coping (see Fig. 4). Seventy students (55%) were categorized as cyber-victims and 26 (21%) as cyberbullying-victims. Sixty-five students (51%) were categorized as cyber-perpetrators and 21 (17%) as cyberbullying-perpetrators. Cyber-victimization and perpetration were significantly related. Fiftytwo students (41%) were cyber-victims as well as cyber-perpetrators (Phi coefficient: rφ = .53, p b .001) and 8 students (6%) were cyberbullying-victims as well as cyberbullying-perpetrators (rφ = .21, p b .05). Students on average used Google (M = 4.60, SD = 1.38), schülerVZ (M = 4.43, SD = 1.83) and video portals (M = 4.28, SD = 1.60) several times a week; they used chat (M = 3.88, SD = 1.97), emails (M = 2.93, SD = 1.53), and instant messenger (M = 2.85, SD = 2.07) approximately once a week; and they rarely or never used Facebook (M = 2.29, SD = 1.99), internet forums (M = 1.77, SD = 1.29) or blogs (M = 1.36, SD = 0.95).

Results regarding hypotheses and exploratory research questions For each of the dependent variables – the current mood scales SM (Sad Mood), AM (Angry Mood), and JM (Joyous Mood) and Active Coping – we computed separate ANOVAs with the experimental between-subject factors Media (text vs. video) and Type (harassment vs. outing) of cyber incident. Additionally, we included students' sex as between-subject factor in each analysis (female vs. male). Students' open answers regarding the Coping Strategies Question were analyzed with Chi-Square tests. The ANOVA regarding SM showed a significant main effect of sex, F (119) = 8.09, p = .005, ηp2 = .06. Girls reported sadder mood than boys when confronted with cyber incidents. The ANOVA regarding AM showed a significant main effect of Medium of cyber incident, F (119) = 13.65, p b .001, ηp2 = .10, and a significant interaction between Medium and Type of cyber incident, F (119) = 3.99, p = .048, ηp2 = .03. The videos elicited angrier moods compared to the texts; this difference was more pronounced for harassment than for outing (see Fig. 3, left). The ANOVA regarding JM showed a significant main effect of sex, F (119) = 7.70, p = .006, ηp2 = .06. Girls reported less joyous mood than boys when confronted with cyber incidents. The ANOVA regarding the extent of Active Coping showed significant main effects of Media, F (119) = 6.53, p = .012, ηp2 = .05, and Type, F (119) = 19.31, p b .001, ηp2 = .14, of cyber incident as well as of sex, F (119) = 4.64, p = .033, ηp2 = .04. Videos elicited more coping than texts, harassment elicited more coping than outing, and girls reported more coping than boys (see Fig. 3, right). Chi-square tests computed separately for each of the factors (media, type, and sex) indicate the following effects regarding students' open coping strategies answers: We found no significant difference in the open answers between the Media of video and text. Regarding Types of cyber incidents, students confronted with harassment were more likely to employ Social, χ 2 (1) = 10.79, p = .001; Technical, χ 2 (1) = 5.52, p = .015; or Legal coping χ 2 (1) = 5.57, p = .018, and less likely to employ Passive coping, χ 2 (1) = 5.02, p = .022, than students confronted with outing (see Fig. 4). Regarding sex, girls were more likely to employ Social coping, χ 2 (1) = 13.29, p = .966, and less likely to employ Aggressive coping, χ 2 (1) = 8.93, p = .966, compared to boys.

249

Discussion of Study II We partly confirmed our Hypothesis 1. Videos used in the cyber incident scenarios are more distressing for cyber-victims than texts and result in more negative affect (Angry Mood) and more planned behavior (Active Coping). However, we found no effects of Media used in the cyber incident scenarios on Sad Mood, Joyous Mood, or students' open answers regarding their Coping Strategies. We partly confirmed our Hypothesis 2. Different Types of cyber incident vignettes have different effects and thus result in different affects and behaviors. More specifically, regarding Angry Mood, the effects of Media were more pronounced for harassment than for outing. Active Coping was more pronounced for harassment than for outing, and students' open answers indicate that harassment results in more Social, Technical, and Legal coping and in less Passive coping compared to outing. However, we found no effects of Type used in the cyber incident scenarios on Sad Mood or Joyous Mood. Our Exploratory Research Question 3 also yielded interesting results. Students' sex did not only moderate the effects of Media and Type of cyber incident, sex was the only variable significantly affecting students' Sad Mood and Joyous Mood. In both cases girls were more negatively affected by reading the cyber vignettes. Additionally, girls also reported more Active Coping and their open answers indicate that they would use more Social coping and less Aggressive coping than boys. Despite these promising results this study has some specific limitations that need to be considered (also see limitations of Study I, for example regarding the use of hypothetical vignettes and regarding the generalizability sex effects): In this study, we recruited whole classes and administered paper-and-pencil questionnaires. We therefore avoided some critical issues of online data collection, for example we avoided self-selection effects and numerous drop-outs (see Study I). However, the sample was fairly small and we have no information about its representativeness. We also do not know if the classroom context with the social dynamics of bullying being present, namely bullies, victims, and bystanders, had any effects on the results. However, the fact that we partly replicated the effects of Media detected by Smith and colleagues (Slonje & Smith, 2008; Smith et al., 2008) with a novel experimental approach underlines the validity of these effects and our methodology. Another important issue that needs to be addressed is the problem of confounded variables. In this study, it was impossible to create a fictitious cyber scenario that was only varied through the experimental dimensions of Medium (video vs. text) and Type (harassment vs. outing). Types of cyber incidents are not independent from Media. For example, it is hard to conceptualize a cyber scenario in which videos are used to exclude someone from an activity on Facebook. Besides these confounded variables, there might be other underlying dimensions relevant for media effects — these variables should be considered in future research. We were able to control these variables in this study by holding them constant across all experimental conditions. For example, synchronicity was controlled by choosing asynchronous forms of cyber incidents (vs. synchronous) and publicity was controlled by choosing a public website for all cyber incidents (vs. private vs. semi-public). Despite these limitations we can draw some tentative conclusions. For some affective variables (Sad Mood, Joyous Mood) sex seems to be more relevant than either Medium or Type of cyber incident. Furthermore, this is to our knowledge the first study that could demonstrate that the Types of cyberbullying proposed by Willard (2007) have different effects on cyber-victims. Therefore, this is the first attempt to test this theoretical framework empirically. However, further research is needed to determine if these effects hold true in more representative samples, and if the dimensions of interest truly caused these effects or if confounded variables are to blame.

250

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

Fig. 3. Students' Angry Mood (left; significant main effect of Medium and significant interaction Medium × Type) and Active Coping (right; significant main effects of Medium, Type, and Sex) by sex, medium (T = text vs. V = video) and type (H = harassment vs. O = outing) of cyberbullying.

Overall conclusions, implications, and significance The superordinate question posed at the beginning of this paper was which dimensions are relevant for cyberbullying. The results of these two exemplary studies show that not only defining characteristics transferred from conventional bullying but also uniquely cyber-specific aspects are relevant to cyberbullying. Even though these studies only focus on few selected dimensions that are potentially relevant for cyberbullying, first conclusions can be drawn for the conceptualization and measurement of cyberbullying. On the conceptual side, defining characteristics of bullying serve to distinguish bullying from other acts of aggression. Presumably, the underlying rationale is that bullying according to these criteria is a distinct and distressing experience. For cyberbullying, defining characteristics are not yet carved in stone. One viable option would be to define the boundaries of cyberbullying based on the cybervictim's level of distress. Against this background, Study I indicates that at least some mandatory defining criteria of conventional bullying can be transferred to cyberbullying. We demonstrated that one specific facet of power imbalance, namely perceived popularity, impacts the experience of cyberbullying. But it is also feasible that other aspects of power imbalance as well as other defining criteria, specifically the intention to harm and repetition, could have similar effects. These findings underline the similarity between conventional bullying and cyberbullying. The findings of Study II on the other hand underline conceptual differences between conventional bullying and 100

harassment

90

Percent (%)

80

outing

70 60 50 40 30 20 10 0

Coping Strategy Categories Fig. 4. Coping in response to cyber incidents as indicated by students' open answers by Type of cyber incident (harassment v. outing).

cyberbullying. The dimensions of media and type of cyberbullying also have significant effects on the experience of cyberbullying, indicating that these uniquely cyber-specific aspects of online cruelty might be important for the definition of cyberbullying as well. These findings might also be relevant for the measurement of cyberbullying. Valid measurement instruments of cyberbullying have to capture the full breadth and width of the construct according to the corresponding definition. For example, if a single question is asked respondents have to have a correct understanding of the technical term “cyberbullying” to answer adequately (Menesini & Nocentini, 2009). Depending on the underlying definition, this understanding should be sensitive to defining characteristics such as power imbalance and should include cyberbullying with, for example, different media and of different types. However, not all adolescents might be aware that outing or exclusion in cyberspace constitutes cyberbullying. It is a problem, when respondents have a biased understanding of “cyberbullying” or if questionnaires only cover a fraction of the construct. Detected prevalence rates might underestimate (or overestimate) the real occurrence of cyberbullying. Besides these straightforward implications for the conceptualization and measurement of cyberbullying, we would like to outline some implications on a methodological level. Both studies in this paper show that experimental research can significantly contribute to clarifying conceptual issues in cyberbullying research. Even though these two studies should only be seen as the starting point of an emerging research program, they contribute something unique to the field of cyberbullying research, mostly because of the novel methodological approach. On top of this, experimental cyberbullying research could also contribute to answering questions of causality, for example between presumed risk and protective factors, cyberbullying, and presumed consequences or effects. So far we do not know if depressive symptoms are caused by cyberbullying or vice versa. Therefore, this paper should also be read as a strong argument for the broadening of the methodological repertoire of cyberbullying research to also include experiments — outside and inside the laboratory. Last but not least, these results also have applied practical implications because they reveal specific vulnerabilities that can be addressed in preventions or interventions. In Study I female sex, low perceived popularity, and young age were risk factors for distress associated with negative cyber incidents. Especially these adolescents need to be taught numerous and adequate tools for coping with potential negative cyber incidents (Pieschl & Porsch, 2012). Study II shows that cyber incidents' media and type are also relevant risk factors for distress. For example, cyberbullying via video caused more distress than cyberbullying via text. Adolescents can be taught

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

adequate (technical) coping strategies that confine such incidents, such as contacting the internet service providers or filing charges against unauthorized use of one's own picture/video (Pieschl & Porsch, 2012). Please note, that these vulnerabilities are not risk factors for cyberbullying per se. For example, (perceived) popularity does not significantly predict cyber-victimization; rather cybervictimization seems to foster (perceived) popularity, at least for girls (Gradinger, Strohmeier, Schiller, Stefanek, & Spiel, 2012). Equally, a reverse u-shaped relation between age and cyberbullying is assumed with a peak during adolescence (Tokunaga, 2010). The results of these studies partly confirm this assumption: In the younger sample of Study II 55% of students report cyber-victimization whereas the corresponding percentage is 71 % for the older sample of study I. Therefore, all suggested applications of prevention or intervention programs do not necessarily reduce the risk of cyber-victimization; rather they could help buffer its negative consequences. Acknowledgment The data for Study I was collected as part of the fourth author's Bachelor thesis requirements, and the data for Study II was collected as part of the third author's Bachelor thesis requirements. References Ang, R. P., & Goh, D. H. (2010). Cyberbullying among adolescents: The role of affective and cognitive empathy, and gender. Child Psychiatry and Human Development, 41(4), 387–397, http://dx.doi.org/10.1007/s10578-010-0176-3. Bauman, S., & Del Rio, A. (2006). Preservice teachers' responses to bullying scenarios: Comparing physical, verbal, and relational bullying. Journal of Educational Psychology, 98(1), 219–231, http://dx.doi.org/10.1037/0022-0663.98.1.219. Bonnet, M., Goossens, F. A., & Schuengel, C. (2011). Parental strategies and trajectories of peer victimization in 4 to 5 year olds. Journal of School Psychology, 49(4), 385–398. Breitkopf, L. (1985). Hilflosigkeitsskala [The helplessness scale]. Diagnostica, 31(3), 221–233. Calvete, E., Orue, I., Estévez, A., Villardón, L., & Padilla, P. (2010). Cyberbullying in adolescents: Modalities and aggressors' profile. Computers in Human Behavior, 26, 1128–1135, http://dx.doi.org/10.1016/j.chb.2010.03.017. Caravita, S. S., Di Blasio, P., & Salmivalli, C. (2009). Unique and interactive effects of empathy and social status on involvement in bullying. Social Development, 18(1), 140–163, http://dx.doi.org/10.1111/j.1467-9507.2008.00465.x. Carnagey, N. L., & Anderson, C. A. (2003). Theory in the study of media violence: The general aggression model. In D. A. Gentile (Ed.), Media violence and children: A complete guide for parents and professionals (pp. 87–105). Westport, CT: Praeger Publishers/Greenwood Publishing Group. Collett, J. L., & Childs, E. (2011). Minding the gap: Meaning, affect, and the potential shortcomings of vignettes. Social Science Research, 40(2), 513–522, http://dx.doi.org/ 10.1016/j.ssresearch.2010.08.008. Dalbert, C. (1992). Subjektives Wohlbefinden junger Erwachsener: Theoretische und empirische Analysen der Struktur und Stabilität [Subjective well-being of young adults: Theoretical and empirical analysis of structure and stability]. Zeitschrift für Differentielle und Diagnostische Psychologie, 13(4), 207–220. de Bruyn, E. H., Chillessen, A. N., & Wissink, I. B. (2010). Associations of peer acceptance and perceived popularity with bullying and victimization in early adolescence. The Journal of Early Adolescence, 30(4), 543–566, http://dx.doi.org/10.1177/ 0272431609340517. Dehue, F., Bolman, C., & Völlink, T. (2008). Cyberbullying: Youngsters' experiences and parental perception. Cyberpsychology & Behavior, 11(2), 217–223. Dempsey, A. G., Sulkowski, M. L., Dempsey, J., & Storch, E. A. (2011). Has cyber technology produced a new group of peer aggressors? Cyberpsychology, Behavior and Social Networking, 14(5), 297–302, http://dx.doi.org/10.1089/cyber.2010.0108. Dooley, J. J., Pyzalski, J., & Gross, D. (2009). Cyberbullying versus face-to-face bullying. Zeitschrift für Psychologie/Journal of Psychology, 217(4), 182–188. Erdur-Baker, Ö. (2010). Cyberbullying and its correlation to traditional bullying, gender and frequent and risky usage of internet-mediated communication tools. New Media & Society, 12(1), 109–125. Gradinger, P., Strohmeier, D., Schiller, E., Stefanek, E., & Spiel, C. (2012). Cyber-victimization and popularity in early adolescence: Stability and predictive associations. European Journal of Developmental Psychology, 9(2), 228–243, http://dx.doi.org/ 10.1080/17405629.2011.643171. Gradinger, P., Strohmeier, D., & Spiel, C. (2009). Traditional bullying and cyberbullying identification of risk groups for adjustment problems. Zeitschrift für Psychologie/ Journal of Psychology, 217(4), 205–213. Granello, D., & Wheaton, J. E. (2004). Online data collection: Strategies for research. Journal of Counseling and Development, 82(4), 387–393, http://dx.doi.org/10.1002/ j.1556-6678.2004.tb00325.x. Hinduja, S., & Patchin, J. W. (2008). Cyberbullying: An exploratory analysis of factors related to offending and victimization. Deviant Behavior, 29, 129–156.

251

Huang, Y. -Y., & Chou, C. (2010). An analysis of multiple factors of cyberbullying among junior high school students in Taiwan. Computers in Human Behavior, 26, 1581–1590. Katzer, C., Fetchenhauer, D., & Belschak, F. (2009a). Cyberbullying in Internet-Chatrooms — Wer sind die Täter? Ein Vergleich von Bullying in Internet-Chatrooms mit Bullying in der Schule aus der Täterperspektive [Cyberbullying in internet-chatrooms — Who are the bullies? A comparison between bullying in internet-chatrooms and bullying in school from the perspective of the bully]. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 41(1), 33–44, http://dx.doi.org/10.1027/1864-1105.21.1.25. Katzer, C., Fetchenhauer, D., & Belschak, F. (2009b). Cyberbullying: Who are the victims? A comparison of victimization in internet chatrooms and victimization in school. Journal of Media Psychology, 21(1), 25–36. Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Social psychological aspects of computer-mediated communication. American Psychologist, 39(10), 1123–1134, http://dx.doi.org/10.1037/0003-066X.39.10.1123. LaFontana, K. M., & Cillessen, A. N. (2010). Developmental changes in the priority of perceived status in childhood and adolescence. Social Development, 19(1), 130–147, http://dx.doi.org/10.1111/j.1467-9507.2008.00522.x. Li, Q. (2006). Cyberbullying in schools: A research of gender differences. School Psychology International, 27(2), 157–170, http://dx.doi.org/10.1177/0143034306064547. Li, Q. (2007). New bottle but old wine: A research of cyberbullying in schools. Computers in Human Behavior, 23(4), 1777–1791, http://dx.doi.org/10.1016/ j.chb.2005.10.005. Medienpädagogischer Forschungsverbund Südwest (mpfs) (2011). JIM-studie 2011, jugend, information, (multi-)media. Basisstudie zum Medienumgang 12-bis 19-Jähriger in Deutschland [JIM-study 2011. Youth, information, (multi)media]. Representative study about media use of 12–19 year olds in Germany. Stuttgart: MPFS. Menesini, E., & Nocentini, A. (2009). Cyberbullying definition and measurement: Some critical considerations. Zeitschrift Für Psychologie/Journal of Psychology, 217(4), 230–232, http://dx.doi.org/10.1027/0044-3409.217.4.230. Nocentini, A., Calmaestra, J., Schultze-Krumbholz, A., Scheithauer, H., Ortega, R., & Menesini, E. (2010). Cyberbullying: Labels, behaviours and definition in three European countries. Australian Journal of Guidance & Counselling, 20(2), 129–142, http://dx.doi.org/10.1375/ajgc.20.2.129. Olweus, D. (1996). Gewalt in der Schule: Was Lehrer und Eltern wissen sollten — und tun können [Violence in school: What teachers and parents should know — And can do]. Bern: Huber. Ortega, R., Mora-Merchán, J. A., & Jäger, T. (Eds.). (2007). Acting against school bullying and violence. The role of media, local authorities and the Internet [e-book]. Landau: Verlag Empirische Pädagogik, http://dx.doi.org/10.1027/0044-3409.217.4.197. Paivio, A. (1986). Mental representations: A dual coding approach. New York, NY: Oxford University Press. Parkhurst, J. T., & Hopmeyer, A. (1998). Sociometric popularity and peer-perceived popularity: Two distinct dimensions of peer status. The Journal of Early Adolescence, 18(2), 125–144, http://dx.doi.org/10.1177/0272431698018002001. Patchin, J. W., & Hinduja, S. (2006). Bullies move beyond the schoolyard: A preliminary look at cyberbullying. Youth Violence and Juvenile Justice, 4(2), 148–169, http://dx.doi.org/ 10.1177/1541204006286288. Pieschl, S., & Porsch, T. (2012). Schluss mit Cybermobbing! Das Trainings- und Präventionsprogramm “Surf-Fair” [Stop cyberbullying! The training and prevention program “Surf-Fair”]. Weinheim: Beltz. Raskauskas, J., & Stoltz, A. D. (2007). Involvement in traditional and electronic bullying among adolescents. Developmental Psychology, 43(3), 564–575, http://dx.doi.org/ 10.1037/0012-1649.43.3.564. Riebel, J., Jäger, R., & Fischer, U. C. (2009). Cyberbullying in Germany — An exploration of prevalence, overlapping with real life bullying and coping strategies. Psychology Science Quarterly, 51(3), 298–314. Scheithauer, H., Hayer, T., & Bull, H. D. (2007). Gewalt an Schulen am Beispiel von Bullying. Aktuelle Aspekte eines populären Themas [Violence at schools regarding the example of bullying. New aspects of a popular topic]. Zeitschrift für Sozialpsychologie, 38(3), 141–152, http://dx.doi.org/10.1024/0044-3514.38.3.141. Schultze-Krumbholz, A., & Scheithauer, H. (2009). Social–behavioral correlates of cyberbullying in a German student sample. Zeitschrift für Psychologie/Journal of Psychology, 217(4), 224–226. Sijtsema, J. J., Veenstra, R., Lindenberg, S., & Salmivalli, C. (2009). Empirical test of bullies' status goals: Assessing direct goals, aggression, and prestige. Aggressive Behavior, 35(1), 57–67, http://dx.doi.org/10.1002/ab.20282. Slonje, R., & Smith, P. K. (2008). Cyberbullying: Another main type of bullying? Scandinavian Journal of Psychology, 49(2), 147–154, http://dx.doi.org/10.1111/ j.1467-9450.2007.00611.x. Smith, P. K., Mahdavi, J., Carvalho, M., Fischer, S., Rusell, S., & Tippett, N. (2008). Cyberbullying: Its nature and impact in secondary school pupils. Journal of Child Psychology and Psychiatry, 49(4), 376–385, http://dx.doi.org/10.1111/j.1469-7610. 2007.01846.x. Sourander, A., Klomek, A., Ikonen, M., Lindroos, J., Luntamo, T., Koskelainen, M., et al. (2010). Psychosocial risk factors associated with cyberbullying among adolescents: A population-based study. Archives of General Psychiatry, 67(7), 720–728, http://dx.doi.org/10.1001/archgenpsychiatry.2010.79. Staude-Müller, F., Bliesener, T., & Nowak, N. (2009). Cyberbullying und Opfererfahrungen von Kindern und Jugendlichen im Web 2.0 [Cyberbullying and victim experience of children and adolescents in the Web 2.0]. Kinder- und Jugendschutz in Wissenschaft und Praxis, 2, 42–47. Thayer, J. D. (2002). Stepwise regression as an exploratory data analysis procedure. Paper presented at the Annual Meeting of the American Educational Research Association (AERA), New Orleans, LA, USA.

252

S. Pieschl et al. / Journal of Applied Developmental Psychology 34 (2013) 241–252

Tokunaga, R. S. (2010). Following you home from school: A critical review and synthesis of research on cyberbullying victimization. Computers in Human Behavior, 26, 277–287. Ttofi, M. M., & Farrington, D. P. (2008). Reintegrative shaming theory, moral emotions and bullying. Aggressive Behavior, 34(4), 352–368, http://dx.doi.org/10.1002/ ab.20257. Utsumi, S. (2010). Cyberbullying among middle school students: Association with children's perception of parental control and relational aggression. Japanese Journal of Educational Psychology, 58(1), 12–22. Vandebosch, H., & Van Cleemput, K. (2008). Defining cyberbullying: A qualitative research into the perceptions of youngsters. Cyberpsychology & Behavior, 11(4), 499–503, http://dx.doi.org/10.1089/cpb.2007.0042.

Wachs, S., & Wolf, K. D. (2011). Zusammenhänge zwischen Cyberbullying und Bullying — erste Ergebnisse aus einer Selbstberichtsstudie [Correlates of cyberbullying and bullying — First results of a self-report study]. Praxis der Kinderpsychologie und Kinderpsychiatrie, 60, 735–744. Willard, N. E. (2007). Cyberbullying and cyberthreats. Responding to the challenge of online social aggression, threats, and distress. Champaign, Illinois: Research Press. Witvliet, M., Olthof, T., Hoeksma, J. B., Goossens, F. A., Smits, M. I., & Koot, H. M. (2010). Peer group affiliation of children: The role of perceived popularity, likeability, and behavioral similarity in bullying. Social Development, 19(2), 285–303, http://dx.doi.org/10.1111/ j.1467-9507.2009.00544.x. Yoon, J. S. (2004). Predicting teacher interventions in bullying situations. Education and Treatment of Children, 27(1), 37–45.