You are on page 1of 26

Human Communication Research ISSN 0360-3989

ORIGINAL ARTICLE

Social Relations and Cyberbullying: The


Influence of Individual and Structural
Attributes on Victimization and Perpetration
via the Internet
Ruth Festl1 & Thorsten Quandt2
1 Institute of Communication Studies, University Hohenheim, 70599 Stuttgart, Germany
2 Institute of Communication Studies, University of Münster, 48149 Münster, Germany

Current research indicates that an alarming number of students are affected by cyberbullying.
However, most of the empirical research has focused on psychological explanations of the
phenomenon. In an explorative survey study based on the reconstruction of 2 complete
school networks (NP = 408), we expand the explanation strategies of cyberbullying to higher
levels of social abstraction. Using statistical and structural analysis, and visual inspection of
network environments, we compare explanations on individual and structural levels. In line
with previous research, the findings support traditional explanations via sociodemographic
and personality factors. However, the findings also reveal network positioning to be a
comparably strong predictor for cyberbullying. Therefore, we argue that without taking
structural factors into account, individual explanations will remain insufficient.

doi:10.1111/j.1468-2958.2012.01442.x

The rise of information and communication technologies (ICTs) has influenced


virtually every area of society. ICTs have opened a multitude of new possibilities
for communication, but as with all modern technologies, they have a double-edged
character: Positive, deliberative aspects go hand in hand with the possibility of abusing
new kinds of communicative freedom for other, less desirable purposes. In particular,
the chances and risks affect young people who have grown up with digital media and
intensively make use of ICTs as part of their everyday life (Lenhart, Purcell, Smith, &
Zickuhr, 2010).
Adolescents are increasingly confronted with content that is not suitable for their
age, as they gain (mostly) unrestricted access to networked information sources via
computers, mobile phones, and other networked devices (Livingstone & Haddon,
2008; Livingstone & Helsper, 2010). However, not all the damaging content is
produced by an adult ‘‘third’’ person. Recently, there have been many reports

Corresponding author: Ruth Festl; e-mail: ruth.festl@uni-hohenheim.de

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 101
Social Relations and Cyberbullying R. Festl & T. Quandt

of adolescents being affected by or taking part in cyberbullying—an ‘‘aggressive,


intentional act carried out by a group or individual, using electronic forms of contact,
repeatedly and over time against a victim who cannot easily defend him or herself,’’
as described by Smith, Mahdavi, Carvalho, Fisher, and Russell (2008, p. 376).
The phenomenon of cyberbullying is of increasing interest to academic study.
Although prevalence rates have yet to reach the levels of traditional forms of bullying
in schools or at the workplace (Raskauskas & Stoltz, 2007; Smith et al., 2008),
cyberbullying is a relevant problem in society. Ybarra and Mitchell (2004) found
that 19% of the 10–17 age group had been involved in online aggression, with 12%
being the aggressors, 3% being both aggressors and targets, and 4% being the targets.
Tokunaga (2010) summarized previous research on cyber victimization, reporting
that 20–40% of all adolescents had experienced cyberbullying at least once in their
lives.1
Despite the growing interest in the topic, research is following several closely
defined approaches. Most of the previous studies have focused on individual charac-
teristics found in the personality of adolescents such as extraversion and emotional
instability (Tani, Greenman, Schneider, & Fregoso, 2003) and demographics such
as gender or age differences (summarized by Tokunaga, 2010). In addition, some
studies analyzed shared features of traditional (offline) bullying and cyberbullying
(Kowalski, Limber, & Agatson, 2008; Raskauskas & Stoltz, 2007). Altogether, previ-
ous studies revealed that persons with specific personal characteristics are more or
less strongly prone to cyberbullying. Although personality variables and individual
psychological conditions have been proven to be relevant, this current perspective on
cyberbullying is somewhat reductionist, only taking individual-level explanations into
account.
Cyberbullying is not only an individual problem. It is clearly dependent on social
constellations, as it always includes perpetrators and victims who are embedded
in larger social structures and environments. This sociostructural character is most
relevant for the discussion of cyberbullying with reference to children and adolescents.
As they pass through various phases of personality development, the social context
is essential. Early social development research has stressed the need to analyze social
references in addition to individual aspects (Bronfenbrenner, 1943).
On the basis of this consideration, several theoretical approaches strive to
characterize the social milieu of young individuals. The social connections under
analysis vary from looser peer groups to more complex forms of friendship relations
and exclusive cliques (Cotterell, 2007). Some studies have focused on the interplay
of social and individual characteristics with regard to specific forms of adolescent
behavior. Early studies on aggressive behavior integrated the characteristics of a
person’s social network to explain the particular individual’s behavior (Cairns,
Cairns, Neckerman, Gest, & Gariepy, 1988). In a similar vein, research on traditional
face-to-face bullying expanded the individual-based focus by structural explanations
(see later). As these forms of deviant behavior are related, we can plausibly expect
that structural effects are also relevant to cyberbullying.

102 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Therefore, in addition to personal factors that influence cyberbullying behavior,


this study investigates the structural conditions that underlie the phenomenon.
On the basis of a network perspective, we apply a multilevel approach, starting
with micro level (personal) factors, and then extending the view to meso-level
structures (such as cliques and classes) and finally larger network structures (such as
complete school networks). The primary aim is to identify the personal and structural
principles behind cyberbullying, using statistical and network-analytical methods as
well as visual analysis of network patterns that can explain the phenomenon in a
larger, more ‘‘communicative and social’’ context than that achieved in previous
individualistic studies.

Cyberspace: A new quality of social relations and social harassment?


Growing up in the days of Web 2.0 is associated with increased use of social
Web activities. Membership in social network communities, such as Facebook,
has nearly become a condition sine qua non, especially for adolescents and young
adults (Schmidt, Paus-Hasebrink, & Hasebrink, 2009). The public and the scientific
community have engaged in controversial discussions on the communicative quality
of these online connections. On one hand, their volatile character can be seen as
helpful in enabling an open flow of information (Kavanaugh, Reese, Carool, &
Rosson, 2005); on the other hand, communication in social networks can be used in
a negative way. In particular, parents may worry about their children being contacted
or sexually harassed by strangers.
In contrast to these fears, empirical findings have shown a large congruence of
on- and offline networks of young persons (summarized by Livingstone, 2009). For
example, Ellison, Steinfield, and Lampe (2007) revealed that Facebook is mainly used
to maintain or strengthen existing offline relationships. Most of the relationships
in Facebook can be characterized as weak ties, but there is usually a shared offline
experience among individuals who friend one another, such as being in the same class
at school. This local component has also been confirmed in other studies (Boneva,
Quinn, Kraut, Kiesler, & Shklovski, 2006; Hampton & Wellman, 2003): Today’s
teenager seems to primarily communicate in an online environment with persons
they already know from local ‘‘real life’’ contexts, transferring their offline contacts
into the virtual world instead of searching for new relationships.
At least in this regard, one can assume that cyberbullying processes are similar
to normal communication in social networks. This assumption is plausible when
taking the ‘‘logic’’ of bullying into account: Perpetrators randomly harassing people
they do not know is uncommon,2 as their action is aimed at a social result, such
as strengthening one’s own social position or marginalizing opponents in a given
group (Sijtsema, Veenstra, Lindenberg, & Salmivalli, 2009). These prevalent bullying
motives refer to a specific social context and structure and, therefore, actions must
be directed toward identifiable and known members of the group (e.g., a school
class).

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 103
Social Relations and Cyberbullying R. Festl & T. Quandt

In this respect, the use of ICTs can be regarded as an expansion of the possibilities
of perpetration and victimization, by simplifying the options of reaching and
exposing intended targets known from other (mostly offline) contexts. In line with
this argument, empirical studies dealing with cyberbullying have identified a large
overlap of persons involved in face-to-face bullying and cyberbullying (Kowalski
et al., 2008; Raskauskas & Stoltz, 2007). ‘‘Real life’’ bullies are often cyberbullies, and
offline victims are frequently online victims.
Furthermore, traditional forms of bullying occur mainly within stable social
environments such as schools, as they rely on well-established social structures and
conflicts—there is strong evidence that cyberbullying becomes part of these school
bullying structures. For example, Jäger and Riebel (2009) reported that classmates are
the main initiators of bullying via the Internet. In addition, they reported a notable
overlap between students targeted by traditional bullying and by cyberbullying.
Hence, research into cyberbullying can be directly linked to research on antisocial
behavior in schools, and benefit from its findings.
Nevertheless, communicating in cyberspace features some peculiarities that must
be considered when analyzing cyberbullying. People assume that the anonymous
environment of the Web shifts or suspends the boundaries for socially accepted
behavior and that this could facilitate acts such as cyberbullying. This notion is
supported by Dehm and Storll’s (2010) study on the acceptance and use of taboos in
the media. The authors demonstrated that young people between 16 and 20 exchange
views about taboo topics considerably more often via the Internet than via other
media forms.
However, theories of disinhibition in computer-mediated communication often
are not that simple: Social identity models on deindividuation, for example, suggest
that an anonymous context reduces the visibility of the individual and heightens the
relevance of a shared social identity within a group. The latter will not only motivate
anonymous users to behave according to the in-group norm but also make them
prone to offend out-group members (e.g., Postmes, Spears, & Lea, 1998).
The anonymous quality of the net also leads to questions about the consequences
of cyberbullying. An unknown bully is supposed to evoke higher levels of emotional
stress (Kowalski & Limber, 2007), especially as victims cannot externally attribute the
reasons of their victimization to the specific personal characteristics of a perpetrator
(Huitsing, Veenstra, Sainio, & Salmivalli, 2010).
Heirman and Walrave (2008) name further specifics of online communication
that might contribute to more intensive consequences when compared with tradi-
tional forms of face-to-face bullying. First, online communication involves a high
degree of privacy, as the Internet is mostly accessed alone. Consequently, the act of
perpetration commonly remains unnoticed. Moreover, the lack of visible, nonverbal
reactions from the victim may disinhibit bullies or enable them to forget about the
true consequences of their online activities—an effect called ‘‘dissociative imagina-
tion’’ (Suler, 2004). Finally, harmful messages can be spread to an unlimited audience
and reach victims anywhere, at any time, given that local and time-based restrictions

104 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

disappear (Heirman & Walrave, 2008). Unlike unmediated bullying, the bullying
message can be forwarded (go ‘‘viral’’) and expanded by other online users who are
not part of the original group. This might lead to a perpetuation with a growing
number of bullying participants, and more severe consequences.
Following these findings and considerations, we assume that cyberbullying and
traditional bullying primarily differ with respect to the underlying processes (i.e.,
viral online distribution of the bullying message), as well as resulting consequences.
In contrast, both forms of behavior share many other characteristics (such as affected
persons, underlying perpetrator motivations, intended target groups, and social
context of the behavior). Therefore, this study considers the results of previous
research on traditional forms of school bullying, while taking into account the
above-mentioned specifics of online communication.

Social constellations and bullying in school


As noted earlier, structural research on cyberbullying is still in its infancy, despite the
obvious limitations of a purely individualistic approach. In contrast to this situation in
cyberbullying research, there is a notable interest in explaining bullying by integrating
a structural perspective within traditional bullying research. This interest in structure
can be linked to the very nature of the phenomenon: For example, indirect forms
of bullying, such as spreading rumors or social exclusion—often described as social
aggression (Cairns et al., 1988)—primarily aim to harm other persons by destroying
their social relations (Neal, 2009).
The structural perspective within traditional bullying research can be divided into
several approaches: (a) a positional approach that seeks to explain bullying behavior
by identifying the position a person occupies within a social network constellation,
(b) a structural attributes approach that considers the structural features of the
social environment (e.g., the density of a school class network), and (c) a mixed
structural/individual approach that strives to unite aspects of structural and behavioral
analysis (e.g., by analyzing how peer group influence might lead to individual activities
and vice versa). These approaches also offer valuable insights relevant to the analysis
of cyberbullying behavior.
Following the positional approach, Sijtsema et al. (2009) found that perpetrators
are often perceived as being popular in school, much in contrast to victims—an effect
that was especially notable for girls. Despite being described as popular, bullies scored
low on likability. Cillessen and Mayeux (2004) confirmed this social dilemma. They
identified a positive effect of social aggression on perceived popularity and a negative
effect on other persons’ preference for these popular students, even increasing over
time.
In addition to these findings, other studies showed a positive association between
network centrality and forms of social aggression (Ellis & Zarbatany, 2007; Xie, Swift,
Cairns, & Cairns, 2002). However, this relationship was found to be curvilinear
(Neal, 2009): A moderate position regarding centrality within class or school network
increases the likelihood to use social aggression and therefore become a perpetrator.

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 105
Social Relations and Cyberbullying R. Festl & T. Quandt

Individuals located in such a social constellation have the access that is necessary
to use this behavior. However, there is still room for an improvement of one’s own
social position.
Focusing on the social attributes on the school class level, Salmivalli, Huttunen,
and Lagerspetz (1997) found that cyberbullying occurs in specific peer configurations.
The more ‘‘cliquish’’ (i.e., divided into small groups) a class is, the more cases of
cyberbullying and, especially, the greater the number of targets that can be identified
in this class. In contrast, in more unified classes, there are fewer victims or they will
be positioned outside of the core of the group.
Connecting structural indices with bullying behavior and personal attributes,
Salmivalli et al. (1997) analyzed how friendship network structure relates to harass-
ment problems in Finnish schools. They assumed a certain level of uniformity within
networks, not only in relation to race and gender but also in relation to the attitudes
of the students. Accordingly, young people with similar opinions with regard to
bullying will likely form a network. As expected, the researchers found that bullies
typically build networks with other bullies. Moreover, students who support such a
behavior are overrepresented within these networks.
In contrast to the perpetrator side, victim networks more frequently included other
victims or persons who defended them. Other studies also noted a strong influence of
peers. Mouttapa, Valente, Gallaher, Rohrbach, and Unger (2004) included aggression
as a predictor of homophily in peer networks, and their findings support the assump-
tion that bullies and students who have already experienced both perpetration and
victimization more often nominate aggressive friends. An aggressive atmosphere and
conforming social norms therefore seem to increase the likelihood of actively partici-
pating in aggressive acts (a result confirmed by Rulison, Gest, Loken, & Welsh, 2010).
Research by Kärnä, Salmivalli, Poskiparta, and Voeten (2008) also supports the
notion that the perception of group structure and atmosphere is relevant. Their study
showed that the likelihood of victimization in a school class peaks when uninvolved
students support the behavior of bullies instead of defending victims. According to
Huitsing et al. (2010), the perceived structure is not only relevant to the bullying act
itself but also to the resulting consequences. Their research revealed that the correla-
tion between victimization and the psychological adjustment of victims is moderated
by the number of bullies and victims in the class. Classes with high proportions of
both are expected to contain a social climate of distrust. According to attribution
theory, victims in this context more easily can attribute their mistreatment externally
to the general conditions in class instead of considering the problem as a personal
trait (Kelley, 1973).
All the above studies on traditional bullying among adolescents consistently sup-
port the notion that structure matters. There is strong evidence that most bullying hap-
pens within stable groups, primarily within a school context, directed at individuals as
social beings embedded in that structure and context. In that sense, bullying is clearly
a structure-oriented behavior. As discussed earlier, there are convincing empirical
findings that support the transfer of these ideas to cyberbullying among adolescents.

106 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Research questions
The present analysis chooses the school setting as the relevant context, in line with
many studies in traditional bullying. This choice follows the argument that even in
an online environment, adolescent bullying is primarily directed at the social near
field and relevant offline groups that are transferred to the online environment.
Furthermore, this choice allows for a mixed structural/individual approach as
described previously, so both structural and individual behavioral aspects can be
analyzed in a given, stable environment.
In a school environment, we can logically identify several factors or layers that
potentially influence cyberbullying processes:

1. Of course, there is the individual, psychological level, as also identified in most


previous studies on bullying and cyberbullying. Different features of an individual
induce different ways of thinking and acting. Demographic characteristics and
personality features of actors are supposed to directly or indirectly influence the
behavior (e.g., Ajzen, 1991).
Furthermore, we expect the following structural factors to affect the process:
2. In school, individuals are continuously placed in a social environment and inter-
act with other persons such as teachers or schoolmates. Everyone takes a social
position within the school or class network that is characterized by indicators
such as centrality or likeability (Neal, 2010). These aspects describe the structural
attributes of a person within a social system. As bullying behavior often seems to
be socially motivated (Sijtsema et al., 2009), the current social position of a person
should be considered as a relevant explanatory factor.
3. In addition to these structural aspects on the individual node level, other group-
based layers of social influence must be considered for structural analysis. The
most persistent group is the school class. Students spend many years of their
most important socialization periods in that (relatively) stable structure. Prior
studies already pointed to the relevance of the school class in shaping peer rela-
tions, mostly focusing on social homophily with respect to race and gender (e.g.,
Shrum, Cheek, & Hunter, 1988). In addition to these findings, we assume that
class-specific characteristics, such as the general climate or the teachers’ influence,
create a particular atmosphere that also must be considered as a relevant factor
in forming young persons’ behavior. This can be understood as an independent
meso-level effect, beyond a mere addition of individual friendship effects.
Class membership is something the students cannot decide on themselves, yet
they still form further informal connections such as peer groups and smaller
friendship networks inside the externally defined structure of the school class.
Although these groups might be much more volatile than the class structure itself,
they can be regarded as potentially influential because they are the most direct
social microenvironment of the students.
4. Beyond these proximate social environments, some wider factors of social influ-
ence might also play a role; for example, school atmosphere and general social

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 107
Social Relations and Cyberbullying R. Festl & T. Quandt

environment/milieu. Yet when it comes to the most plausible factors in the


given context of cyberbullying, as indicated by previous research, the friend-
ship and peer effects, as well as the class level, seem to be more directly
influential.
5. Finally, the online environment might be considered as an independent factor
beyond the direct real life of the students. However, previous research has shown
that online environments seem to be an extension of offline environments rather
than an independent virtual world, as they are often portrayed in public. That said,
a wider online environment is certainly relevant when it comes to cyberbullying
effects, as the unknown mass of users might work as a perceived or real amplifier
for the intended defamation effects.3

In the current study, we concentrate on the first three inner levels of influence,
as the phenomenon of cyberbullying itself is directly linked to these groups and
occurs in these environments. Although the exclusion of the outer layers is certainly
a restriction, it is necessary because of research-practical and theoretical consider-
ations: First, we suspect a combination and interplay of factors, so there is already
a notable complexity when looking at the chosen layers. Second, the outer layers
could be regarded as relevant contextual factors and amplifying agents, but they are
not crucial for cyberbullying to occur. In contrast, the phenomenon would not exist
without the inner layers. Therefore, we find that starting with the more direct effects
is logical, while reserving the outer layers for later follow-up studies.4
On the basis of these considerations, this study unifies two different perspectives:
individual and structural. The relevance of ego-based, psychological factors was
already confirmed in a range of studies (e.g., Ybarra & Mitchell, 2004). The structural
part is divided into two general approaches: node-level, focusing on structural
attributes in an ego-centered view (Neal, 2010), which can be regarded as being an
individualized structural approach (2, see above), and group/class level analysis (3).
The latter was primarily used to examine aspects of peer influence in substance use
of children and adolescents (Kirke, 2004; Sieving, Perry, & Williams, 2000). Related
to these conceptual approaches, we attempted to answer the following research
questions:
RQ1: How do individual factors contribute to the explanation of different cyberbullying
types?

RQ2: How does the social position of a person contribute to the explanation of different
cyberbullying types?

RQ3: Do these levels of influence (individual vs. structural) on the node level differ in
their predictive power regarding the explanation of cyberbullying types?

RQ4: How do structural factors (calculated with reference to individual-, group-, and
class-based variables) on the level of the school class contribute to the incidences of
cyberbullying in the respective class?

108 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Naturally, this research approach is explorative in the sense that it identifies factors
and patterns on various levels of social entities to discover whether individual reasons
or different kinds of meso-structures are important in explaining cyberbullying.
Therefore, although we work with both classic statistical and network analytical
methods, we will primarily use these to identify explanations for cyberbullying rather
than to test a given set of hypotheses.

Method

Study design
To explain cyberbullying on different levels of influence, we selected two schools for
carrying out complete school surveys and a full network/structure reconstruction.
On the basis of the literature and two focus groups, we developed a questionnaire
and pretested it for comprehensibility and clarity on 15 students aged 13–18. The
questionnaire included items about the respondents, as well as their social networks
in school. We also discussed various definitions of cyberbullying in the focus groups,
seeking the greatest consent and a common understanding of the term in our target
group. As a working definition for the survey, we finally decided to use the description
provided by Smith et al. (2008, p. 376, also quoted in the Introduction section). This
definition was readily understood by all students and seemed to be a common
denominator among them (which is crucial when using self-reports).
Research was carried out during 2 days in July 2010. The participating students
filled out a paper-and-pencil questionnaire during class hours at a given time so that
data from the whole school were collected in parallel. Therefore, data collection took
place in a natural environment, and through parallelization, we prevented students
from discussing the questionnaire before the field phase. We reached 82.6% of the
students, so there were minimal dropouts or missing samples (e.g., some students
were participating in an external event or taking part in exchange programs at
the time of the study).5 We achieved a high acceptance of the research through
an anonymization procedure, guaranteeing that the names of students as well as
the names of their social network connections (indicator explained later) remained
unknown to anybody involved in the analysis.6

Measures—Individual variables
Following the successive logic of our research questions, we developed or integrated
individual as well as structural variables.

Bullying
To measure cyberbullying involvement, we asked students to indicate whether they
had been involved in such bullying (using the working definition by Smith et al.,
2008, in a translated and easily understandable version). More specifically, we asked
whether they had taken part actively as a perpetrator, passively as a victim, or
both at the same time. The answers were dichotomized, generating three kinds

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 109
Social Relations and Cyberbullying R. Festl & T. Quandt

of cyberbullying roles based on the various combinations of values: perpetrators,


victims, and perpetrator/victims. Using the same procedure, we also measured
students’ experiences with traditional bullying in school.

ICT use
By definition, the use of ICT is a decisive facet of cyberbullying, differentiating it from
traditional forms of bullying. Although competence-related differences are likely
to exist within the group of so-called digital natives (Prensky, 2001), the technical
skills needed for the most common forms of cyberbullying are considerably low,
so they are not regarded to be a decisive factor. Accordingly, we concentrated on
frequency-based variables, expecting them to be more influential (as it is plausible
that a higher exposure might lead to a higher risk). We measured Internet use in
hours per day, as well as the number of online social network sites to which students
belong, as this is one of the most favored applications of online use and a likely risk
factor.

Sociodemographics and personality


In addition to classic sociodemographic variables such as age and gender, we
also analyzed the personality of the participants, as earlier research indicated that
personality traits can be decisive for bullying behavior (e.g., Caravita, Di Blasio, &
Salmivalli, 2009). In this study, we abstained from focusing on a certain personality
characteristic, but rather tried to apply an overall measure that covers the main
dimensions. Within personality research, it is assumed that individuals can be
differentiated according to five main dimensions of personality and that a different
constellation of these dimensions leads to a different kind of behavior. The so-
called big five—extraversion, conscientiousness, agreeableness, neuroticism, and
openness—have been proven to be a useful diagnostic instrument in a large array
of studies covering many directions of research. These considerations were also
transferred to bullying research, following an earlier study by Tani et al. (2003).
According to their results, different personality dimensions induce different forms
of bullying behavior. For example, bullies score high on extraversion and emotional
instability, whereas they seem to be less agreeable.
In case of these extraverted perpetrators, it can be assumed that bullying is used
as a method to additionally increase one’s own status. The reason why they do not try
to achieve these goals using prosocial behavior may be related to their emotionally
instable and less agreeable personality. This may also be valid for cybervictims,
although these are often expected to be more introverted and socially excluded.
Methodologically, we used a validated 10-item short scale by Rammstedt and John
(2007).7 Each of the five personality dimensions was measured using one positively,
one negatively coded item. The items were rated on a 5-point Likert scale ranging
from 1 (strongly disagree) to 5 (strongly agree).

110 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Measures—Structural variables
As described earlier, we endeavored to expand individual measures by structural
components, and therefore focus on social structures within which individuals are
embedded. We distinguish between two larger kinds of conceptual categories (see
research questions): node and group-/class-level factors.

Node level
On the node level, we measured the position a person takes within a particular net-
work, assuming that popular and isolated students differ with regard to cyberbullying
(Salmivalli, Karhunen, & Lagerspetz, 1996). To reconstruct the friendship networks,
and as a central network generator for our study, we asked students to nominate his
or her best friends in school (with no limit on the number of nominations). On the
basis of this value, we were not only able to directly determine a person’s outdegree
(i.e., the number of named friends) but also the number of perceived mentions for a
particular person (indegree) by others.
As the outdegree of a person can be considered as a self-reported indicator of
popularity, the indegree can be interpreted as a peer-related measure of a person’s
prestige, quantifying the rank an actor occupies within a set of actors (Wasserman
& Faust, 1994). Studies confirm a negative relationship between peers’ rating of a
person’s likeability and forms of social aggression (Crick, 1996; Zimmer-Gembeck,
Geiger, & Crick, 2005). The indegree might greatly differ from the outdegree, for
example, if a pupil overestimates her/his popularity.
In addition, we looked for another indicator of the structural importance of an
actor: the betweenness centrality (Freeman, Roeder, & Mulholland, 1980), which
indicates whether (and how strongly) actors are positioned between the geodesic
paths between other persons. Actors with a high betweenness centrality are assumed
to have a high amount of interaction control (Wasserman & Faust, 1994). We
assumed that this control of social relations enables actors to reach different parts of a
network and spread information for their own advantage. Occupying such a position
may be related to general status goals, as interaction control seems to be an intended
benefit.
Sijtsema et al. (2009) found that persons who bully others often pursue social
goals. Similar to forms of prosocial behavior, the use of bullying can thus be
considered as an alternative option to enhance one’s own position within peer
hierarchy. The social position of a person, therefore, can support bullying behavior,
which in turn can change the former. In general, previous studies revealed that
there is a positive relationship between the social embeddedness of a person and
the use of social aggression (e.g., Xie et al., 2002). In contrast, persons with a high
betweenness centrality may also be exposed to a higher degree of social pressure, as
they are positioned between different (groups of) persons. As a result, we expect that
these individuals also have an enhanced risk to become the target of cyberbullying.
Following this argument, the betweenness indicator may be especially relevant for
the group of perpetrator/victims.

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 111
Social Relations and Cyberbullying R. Festl & T. Quandt

Group/class level
Group relations between adolescents vary from very exclusive forms to looser
structures (Cotterell, 2007). For bullying behavior, close group concepts such as peer
cliques were found to be most relevant (Salmivalli et al., 1997). A clique is defined as
the largest possible group within which all members link directly without a mediator
(Wassermann & Faust, 1994, p. 254). Consequently, a clique can be characterized by a
high level of density and closeness. Although group structures can also be regarded on
separate levels of consideration, we analyzed these in relation to the larger structure
enclosing them—the school class. For this purpose, we generated new variables: The
frequency of cliques as well as bullying and victimization (traditional and cyber) were
measured by a simple standardization process. The respective incidence was divided
by the absolute number of students visiting the particular class.
In addition, we calculated demographic aspects of the classes, such as the average
age, average number of students, and gender composition, as well as the average in-
and outdegree. Finally, we applied two more variables to characterize the class as a
whole. Density was used as a measure of class cohesion. In a binary network, the
density describes the total number of ties divided by the total number of possible
ties (Shumate & Palazzolo, 2010). As each student simply named her or his friends
in school, we received dichotomous data, indicating if there is a link between nodes
(1) or not (0). Thus, the density measure is based on an asymmetric matrix, as the
nominations of two students are not necessarily reciprocal. A high density can be
regarded as an indicator of a positive class climate, as it describes an overall joint
structure in class. Every student is liked by a large part of his or her schoolmates.
As mentioned before, social likeability negatively correlates with the use of social
aggression. Moreover, a collective class climate is supposed to prevent a victimization
of singular persons, as thoughts of cohesion and common norms might contradict
forms of social exclusion. Nevertheless, we deem it possible that despite a high overall
class density, there are a few outsiders who then become the target of such a large
and joint group.
Following this line of argument, many unrelated groups within a class complicate
common attitudes and activities. According to Janis’ (1982) concept of ‘‘group-
think,’’ a cohesive group structure contains the risk that the self-related beliefs are
overestimated and foreign groups are stereotyped. We therefore controlled for the
classes’ cliquishness, which can be seen as an indicator of a more partitioned struc-
ture. Salmivalli et al. (1997) found that a cliquish structure is associated with higher
prevalence of bullying and victimization in class. Therefore, very dense classes are
assumed to be characterized by less cyberbullying incidents, whereas the occurrence
of many cliquish subgroups is anticipated to disturb solidarity and promote bullying
behavior.

Sample
Altogether, 408 students aged between 12 and 19 answered the questionnaire. For
ethical, legal, and practical research reasons, we excluded students below the seventh

112 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

grade from the study. The average age of the sample was 15.4 years (SD = 1.86). We
observed an overrepresentation of male participants in both types of schools (overall:
57.5%). Most of the students are very ICT literate, with an average daily Internet
use of approximately 2.5 hours. The diffusion of computers in the households of the
students is virtually complete, although roughly 30% of the students do not own
a PC themselves. Ninety-four percent of the students are members of one or more
social network sites, belonging to 2.5 sites on average. The average number of friends
named by the students was 3.56 (SD = 2.65). Almost two thirds of the participants
indicated that they like school much or very much (64.4%), 27.9% were indecisive,
and 7.7% do not like school much or at all. Overall, the level of student satisfaction
with their school life seems to be high, initially indicating against the widespread
diffusion of cyberbullying. Within the two schools, we identified 21 classes (7th to
12th grade).8 Mean class size was 24 persons. On average, we identified 4.7 cliques
per class with an average size of 3.6 members.

Results9

Bullying prevalence
As expected from the literature review, the incidence rates for traditional forms of
bullying in school are much higher than those for cyberbullying. More than half
of the students have been involved in traditional bullying incidents, whereas only
22% have been confronted with cyberbullying. Still, the data support the assumption
of a notable relevance of cyberbullying. Similar to the findings of Kowalski et al.
(2008), a plurality of those involved in cyberbullying are perpetrators (10.8%), 6.4%
are both offenders and victims, and 4.9% are solely victims. We found at least one
cyberbullying incident per class, with a mean value of over four cases per class.

Individual-level findings
Following our research approach, we analyzed the predictive power of ego-centered
variables on cyberbullying behavior. When we checked for sociodemographic factors,
we found a notable gender influence: Boys were more frequently perpetrators, whereas
girls were more frequently victims. These findings can be statistically confirmed, yet
the effect is weak (Cramer’s v = .18; p < .01). However, when separately testing the
condition of being a victim in a logistic regression model (instead of testing the
distribution in the 3 × 2 cross-tab), we found much stronger effects. The analysis
reveals a highly significant risk, almost six times greater, of girls becoming a victim
of cyberbullying, Exp(B) = 5.89; p < .001. Thus, gender provides a different role
allocation in the cyberbullying process.
Moreover, the literature has uniformly claimed that the age of a person is a
relevant demographic factor for harassment on the Internet. However, some studies
suggest that higher age predicts more bullying (Ybarra & Mitchell, 2004), whereas
other researchers report the opposite trend (Slonje & Smith, 2008). This study found
no significant age differences between the three groups of cyberbullying, although the

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 113
Social Relations and Cyberbullying R. Festl & T. Quandt

average age of victims (M = 14.9) is lower than that of the other groups (perpetrators
M = 15.7; perpetrator/victims M = 15.3, and unaffected M = 15.4), possibly due to
the limited age range of the analyzed school classes.
In addition to sociodemographic factors, we analyzed specific personality traits
using the ‘‘big five’’ short scale. A principle axis factor analysis with oblique
rotation confirmed the expected five-factor solution, with only one of the two
intended items loading on the factor openness. For further analyses, we used the
five calculated factors (the value of each factor lies between −3 and +3; a high
value means a high agreement with each factor). Conscientiousness was a shared
quality among victims, although there is no significant distinction between victims
and unaffected students. By comparison, the perpetrators were significantly less
careful than the unaffected (M = −0.35; p < .05). The personalities of victims were
characterized by the highest degree of openness (M = 0.31). Openness—the pursuit
of new and creative experiences—might contribute to the risk of becoming an easy
target, especially in the environment of social networks. According to the big five
scale, victims were also reserved (M = 0.06). Moreover, victims show considerably
higher uncertainty than the other observed groups (however, these findings are
not statistically significant). Cyberbullies (M = 0.31; p < .01) and those who are
both perpetrators and victims (M = 0.43; p < .001) display significantly higher
extraversion than students who have not experienced cyberbullying (M = −0.08).
With regard to the social handling of others, perpetrators/victims are less tolerant and
have a tendency to criticize their classmates (agreeableness; M = −0.35; p < .01).
The same is true for perpetrators, although not as pronounced (M = −0.17; p < .01)
as for the perpetrators/victims and for the victims (not statistically significant). In
contrast, unaffected persons rank positively on the ‘‘agreeableness’’ dimension.
Another frequently cited assumption is that cyberbullying mainly constitutes a
new method for an old behavior (Li, 2005). Therefore, we analyzed the connections
between traditional forms of bullying and cyberbullying. With regard to the perpetra-
tors, the results were somewhat in sync with the cited assumption: 33.3% of bullying
offenders are also cyberbullying offenders. However, 57.6% of the bullying offenders
were not involved in any cyberbullying behavior, neither as offender nor as victim.
This finding might be explained by the much smaller number of students
affected by cyberbullying (when compared with traditional forms). When looking
at this connection from the cyberbullying perspective, we find that many of the
cyberbullying perpetrators are also perpetrators in traditional bullying (50.0%). The
findings are much less clear for the victims. Only 10.4% of traditional victims are also
harassed via the Internet, whereas 82.1% are completely unaffected by cyberbullying.
About one third of the cyberbullying victims are also victims of traditional bullying.
Results from cross-tabulation are summarized in Table 1.
Finally, we looked at the ICT use of the respondents. We assumed that with more
ICT use, students will experience more ICT-related risks, resulting in more cases
of bullying. To investigate the differences among perpetrator, perpetrator/victim,
and victim roles with regard to daily ICT use, we applied a one-way ANOVA,

114 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Table 1 Consistency of Traditional Bullying and Cyberbullying

Traditional Bullying
Perpetrator Perpetrator/Victim Victim

Cyberbullying Perpetrator 33.3 13.1 6.0


Perpetrator/victim 4.5 14.8 1.5
Victim 4.5 3.3 10.4
Uninvolved 57.6 68.9 82.1

Note: Results are from cross-tabulation analysis (column percentages); χ2 = 65.21; p < .001.

revealing that cyberbullies are more frequent Internet users (M = 3.71; p < .001)
than unaffected students (M = 2.28). The other bullying groups also spend more
hours per day on the Internet than the unaffected, but the difference was not
statistically significant. We examined the methods used to cyberbully. Harassment
via social networks constitutes the main channel (51.6% of the respondents who
admitted to cyberbullying used social networks to do so; the second most frequent
channel was Internet chats, at 19.7%). In addition, we asked the respondents to state
the number of social networks in which they participated, revealing that students who
have experienced cyberbullying were more frequently members of a diverse range of
social networks than are the unaffected control group. Both perpetrators (M = 3.16;
p < .001) and perpetrators/victims (M = 3.36; p < .001) were heavy users of the
new social possibilities of Web 2.0, whereas victims participated in fewer networks
than either perpetrator group.

Structural findings on the node level


As a first step in the structural analysis on the node level, we measured the
centrality of the students to demonstrate possible key and edge positions of the
interviewees in relation to their particular involvement in cyberbullying. Surprisingly,
when we compared the students’ outdegree, we found that victims indicated more
friends within class or school (M = 4.10; p < .05) than other groups involved
in cyberbullying (perpetrator: M = 3.00; perpetrator/victim: M = 3.08). Only the
difference between victims and unaffected persons (M = 3.64) is not statistically
significant, despite trending in the same direction of effect. Interestingly, fellow
students mentioned victims the least frequently (M = 2.45) as friends, meaning that
victims did not play a central role in class, which strongly contrasted with their
high outdegree (i.e., the persons they named as friends themselves). The difference
between the two values hints at a conflict for people who are open for new social
contacts and want to integrate but have not been accepted by their schoolmates.
Another surprising result was the high indegree of perpetrator/victims (M = 3.64;
p < .05), which differs significantly from unaffected students. This social parameter
reflected their extraverted personality, but it might also explain why they were active
as perpetrators and victims of cyberbullying; that is, they had a central communicative

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 115
Social Relations and Cyberbullying R. Festl & T. Quandt

role in their classes, which gives them the power to bully others and also exposes them
to the risk of being bullied. This assumption can be underlined by the betweenness
measure. The results reveal highly significant differences between perpetrators/victims
(M = 3.32; p < .001) and unaffected persons (M = 0.84). The betweenness value for
perpetrator/victims was also higher than for perpetrators (M = 1.24; p < .01) and
victims (M = 0.80; p < .01).10 Again, these findings support the notion that victims
were outsiders, whereas victims/perpetrators, and to a certain extent perpetrators,
played a much more central role in the networks. Still, the actual positioning needs
to be verified in the context of the full network.

Comparing individual and structural influences on the node level


Up to this point, we have presented the individual and structural results indepen-
dently, based on an ego-centered view. To develop one comprehensive model, we
applied a stepwise logistic regression (Table 2), integrating individual and struc-
tural factors. The criterion measures were the three cyberbullying roles: perpetrator,
perpetrator/victim, and victim.
Depending on the respective cyberbullying role, some influencing factors are
more important than others. Sociodemographic indicators are primarily relevant
for the victims. As mentioned earlier, young students and female students are more
likely to become the target of virtual attacks. There is a strong correlation between
being a traditional perpetrator in school and a cyberbully, underlining the above
argumentation that cyberbullying can be seen as an extension of traditional bullying
in an online environment.
For people involved in both sides of cyberbullying—as a perpetrator and a
victim—we received surprising results. Apart from personality and previous bullying
experiences in school being relevant (as is the case for the other groups), structural
factors strongly influence this kind of bullying behavior. As suggested in Table 2, the
students were socially preferred by their schoolmates (indegree) and occupied central
network positions in school, connecting other peers (betweenness). Therefore, an
explanation beyond individual aspects seems to play an important role for this
specific group of perpetrator/victims.

Structural findings on the class level


In the next step of the study, we shift the focus from an individual to an aggregated
perspective and analyzed classes with regard to cyberbullying figures and structural
indicators. For this purpose, we initially ran several linear regression analyses with
the cyberbullying roles in class as the criterion measure. Surprisingly, some basic
class-level factors were unimportant. The regression analyses yielded no significant
results for the cyberbullying prevalence of perpetrators, perpetrator/victims, or
victims regarding the number of students in a class. In a similar vein, gender balance
(i.e., the number of males vs. number of females in a class) as well as the average
age of the students showed no significant differences among the three observed
groups.

116 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Table 2 Comparing Individual and Structural Influences on Cyberbullying from a


Node-Level Perspective

Perpetrator Perpetrator/Victim Victim

Sociodemographics
Age 1.10 0.97 0.84
Gender 0.67 0.84 5.89∗∗
R2 .02 .01 .12∗∗
Personality
Conscientiousness 0.63∗ 1.07 1.28
Openness 0.72 0.87 1.74
Extraversion 1.61∗ 1.92∗ 1.09
Neuroticism 0.81 0.84 1.63
Agreeableness 0.67 0.46∗ 0.76
R2 .05 .10∗ .05
Traditional bullying
Perpetrator 7.40∗∗∗ 0.89 0.84
Perpetrator/victim 1.27 5.38∗∗∗ 0.57
Victim 0.46 0.25 2.71∗
R2 .16∗∗∗ .15∗∗ .01
ICT use
Online frequency (hours per day) 1.15∗∗ 1.08 1.04
Membership in online SNS 1.46∗∗ 1.55∗∗ 1.18
R2 .05∗∗ .01 .02
Structural indices of the person
Outdegree 0.91 0.92 1.08
Indegree 0.97 1.23∗ 0.91
Betweenness centrality 1.03 1.21∗∗∗ 0.95
R2 .00 .19∗∗∗ .02
Overall R2 .28 .46 .22

Note: N = 318; stepwise logistic regression was used. Effect size is indicated by odds ratios; an
increase in explained variance is specified by pseudo-R2 .
∗ p < .05. ∗∗ p < .01. ∗∗∗ p < .001.

In parallel with the individual-level analysis, we also compared traditional forms of


bullying with ICT counterparts on the basis of class indices. For perpetrators/victims,
the bullying behavior in an offline context was found to be highly predictive for the
corresponding cyberbullying behavior (β = .74; p < .001). This correlation can be
explained by a constant ‘‘bullying atmosphere’’ within a class. Obviously, if there are
many students who bully and are bullied, the class has many cases of cyberbullying.
As mentioned earlier, we examined emerging social structures not only from an
ego-centered perspective but also on the class level. To control for the correlation of
the independent variables, we ran multiple linear regression analyses. As indicated
by individual-level findings, the social likeability of a person seems to be a relevant

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 117
Social Relations and Cyberbullying R. Festl & T. Quandt

Table 3 Cyberbullying and Social Indices of the Class

Average Average Perpe- Average


Perpetrator trators/Victim Victims Indices
Indices (β) Indices (β) (β)

Outdegree −.57 .11 .17


Indegree −.35 .47 −.83∗
Cliquishness parameter −.34 .18 .12
Number of cliques .79∗ .27 .30
Density .09 −.76 .34

Note: N = 20; results are from multiple linear regression analyses; the standardized β coefficient
is indicated.

predictor of being involved in cyberbullying—at least for the mixed group of


perpetrator/victims. As Table 3 suggests, this holds not to be true on the class level, as
the average indegree of students in class does not significantly influence the average
number of perpetrator/victims. However, this structural value negatively predicts
the average number of victims in class, so that a high number of socially preferred
adolescents seem to reduce the likelihood of victimization (β = −.83; p < .05).
The literature assumes that the more cliquish the individual’s environment, the
higher the chance that a student will attempt to demonstrate the power and unity
of the group via bullying or to be harassed by other surrounding cliques (Cotterell,
2007; Schuster, 1999). Following this line of argumentation, we found that pure
perpetrators tended to locate in more cliquish areas, as the number of cliques in
class significantly predicts their average occurrence (β = .79; p < .05). In such a case,
harassing others without being punished or without becoming victimized seemed
easier, as perpetrators receive social support and protection from their own dense
clique. However, due to the small sample size, further analyses must be performed to
support this assumption.
As a last structural measure, we tested the density of the classes as an indicator of
interdependency within networks. In contrast to the clique concept, a class with a high
density should exhibit less cyberbullying. Again, no significant findings supported
this hypothesis.

Overall network: The school graph


As a helpful tool to evaluate the findings on the ego and structural levels, and as
an overall frame of reference, we also applied visual network data analysis.11 This
approach enables us to illustrate cyberbullying indices on the macrolevel of the
social environment—the whole network of a school system. We regard this visual
inspection as an extension of the previous structural analysis, which might provide
additional insights for the interpretation of data. The network graph of one school is
illustrated in Figure 1.

118 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Figure 1 Exemplary network of one school (perpetrator = black triangle up; perpetra-
tor/victim = dark grey triangle down; victim = light grey square, unaffected = transparent
circle).

Visual analysis reveals an accumulation of offenders and victims in certain


parts of the network. Apparently, the cyberbullying phenomenon is not evenly
distributed throughout schools. Perpetrators were commonly located at the core
of (sub)networks (much more so than one would suspect purely on the level of
their indegrees), and perpetrators/victims were often located at ‘‘bridging’’ positions.
The latter were likely using the ‘‘strength of weak ties’’ (Granovetter, 1983): They
transferred communication between various subgroups, which gave them a strategic
position and also made them more vulnerable to bullying by being exposed to several
group influences. In contrast to the ‘‘active’’ participants in cyberbullying, victims
appeared to be located more often at the edge of the networks.
Finally, network inspection showed interesting micropatterns: There were many
instances where perpetrators and victims seem to be closely connected. Nevertheless,
we registered only a few single cases of direct friendship nominations. Of an
overall number of 1,366 direct links, only one victim unidirectional named another
perpetrator as a friend. Moreover, there were only three pairs of perpetrators and
victims, who nominate each other.
In contrast to these hotspots of cyberbullying, many areas of the network were
completely unaffected. We suspect that these hotspots indicate specific incidents of
cyberbullying, probably with initial dyadic offender–victim constellations, followed
by other incidents and retaliatory bullying. This interpretation seems plausible;
however, due to ethical concerns, we did not ask the participants about specific
cyberbullying incidents and the persons involved in these incidents. In addition, such a
dynamic diffusion of cyberbullying needs to be analyzed in a future longitudinal study.

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 119
Social Relations and Cyberbullying R. Festl & T. Quandt

Discussion

Summary and interpretation of results


The current study has presented findings on the causes and conditions for cyber-
bullying on both the individual and structural level. On the individual, node-level,
gender clearly contributes to cyberbullying risk, as female adolescents are more likely
to become targets of aggressive behavior. The data further imply that bullying is a
stable behavior, as a considerable percentage of active perpetrators not only show this
behavior with regard to traditional bullying but also via the Internet. Cyberbullies
spend more time on the Internet than their fellow students, particularly in social
networks—which are the number one tool for aggressors. In addition, the various
cyberbullying types can be differentiated according to personality attributes. On aver-
age, victims are more diligent, and also uncertain and reserved, while aggressors are
more extraverted, but less conscientious and agreeable. Perpetrators/victims seem to
be a special group within the cyberbullying process. Their personality characteristics
are more similar to those of perpetrators and also differ from the pure victims:
They are very extroverted persons, who tend to criticize other people quickly. Inter-
estingly, unaffected persons score very neutral on the five personality dimensions,
whereas involved groups—whether these are perpetrators, perpetrator/victims or
victims—show more extreme characteristics on each of the factors. That said, per-
petrators and victims do not seem to be complete opposites, but rather intertwined,
with personality features deviating from the characteristics of individuals who are
not involved in cyberbullying.
With regard to their social relations, members of the perpetrator/victim group
are characterized by a high and central social standing within the observed school
networks. They are socially preferred by their schoolmates. In contrast, perpetrators
and, even more so, victims seem not to be as popular as those individuals who already
experienced both—perpetration and victimization via the Internet. Especially the
latter were found to be nominated least often as a friend.
Although these structural findings on the node level might sound surprising at
first, they are in line with the research literature on bullying and cyberbullying. Earlier
research (Salmivalli et al., 1996) found support for the notion that aggressors and
especially targets can be seen as unpopular, whereas people who have experienced
both sides have more central positions. We suspect that this reflects the skills of the vic-
tim/perpetrator group in dealing with both situations and their possession of a certain
mental dexterity, which makes it easier to maintain social contact with various groups.
The current study did not stop with node-level data, but also focused on larger
social constellations (e.g., school classes) as the unit of analysis, using complete school
network data—something lacking from most previous research on (cyber)bullying.
As mentioned in the node-level analysis, bullying attributes seem to be consistent,
independent of the methods and channels conducting the behavior—the same
findings find support on the class level. In general, we found that in classes with many
‘‘traditional’’ bullies, there are also many cases of cyberbullying. Data furthermore

120 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

revealed that characteristics based on social constellations within the classes such
as density, average outdegree, and the measure of inherent cliquishness had no
statistically relevant meaning in the cyberbullying process.
In contrast to cliquishness as an aggregated value, the number of cliques in class
had a significant influence: On average, we identified almost five cliques per class
through network analysis. The more cliques in class the higher the average value of
perpetration. This is in line with previous research by Salmivalli et al. (1997) that
considers cliquish structures as a kind of competitive climate, which promotes acts
such as bullying to enhance one’s own group status.
In general, it can be assumed that the lack of significant findings may be related
to the fact that the location-independent context of cyberbullying is not affected by
physically anchored social structures such as these analyzed in this study. However,
according to previous findings in the literature, we could confirm a strong conformity
between traditional bullying and cyberbullying. Owing to this overlap, we expect
social group and class indices to be equally important for both forms of behavior.
The absence of significant findings on this level may be explained by the nature and
size of the data set, as well as the limited ‘‘reach’’ of small-range social groups.
Finally, a visual ‘‘environmental’’ inspection of the complete school networks
supported some of the findings, primarily on the lower node-level analyses. It
underlined the importance of taking structural factors into account, as the strategic
positioning of perpetrator/victims as well as victims’ positions outside the core
network became clearly visible. Furthermore, we identified clusters of perpetrators
and victims at certain parts of the networks. These structures hint at specific
cyberbullying incidents, where a given bully and his/her supporters harassed victims
in their network neighborhoods. Our statistical analyses could not confirm an
influence of social clusters in class on cyberbullying prevalence. Maybe school class
boundaries are too restrictive, as they limit group-based effects on cyberbullying
behavior to this predefined normative structure. Structural effects likely refer to other
groups outside the own class. However, that could not be confirmed by the design of
the current study.

Desiderata and future research


There is no doubt that cyberbullying is a central phenomenon of ICTs, with a
negative impact on adolescents and children. With social networking, mobile media,
and open distribution platforms, offenders have a wide array of bullying weapons
at their disposal. Our research found support for the relevance of the phenomenon,
with a worrying number of students having been affected by cyberbullying.
We identified reasons for this prevalence on the individual and structural
levels. However, small-scale social group indicators did not prove to be as helpful
in predicting micro- or higher-level factors for cyberbullying, which seems to
be contradictory to previous findings that this behavior occurs in small groups.
However, the lack of significant findings may also be due to the limitations of the
explorative sample. Although the current study is not meant to be representative but

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 121
Social Relations and Cyberbullying R. Festl & T. Quandt

rather reconstructive (i.e., directed at the identification and reconstruction of social


logics), the small number of classes and groups poses some restrictions for the data
analysis. Several tests did not achieve statistical significance due to these limitations.
Another critical aspect of our study is the collection of data via self-reports,
especially with respect to a sensitive topic such as cyberbullying. Salmivalli et al. (1996)
reported that people tend to underestimate their own involvement in perpetration
when compared with peers’ ratings. Likewise, when asked directly, victims more often
remain secretive about their experiences. On the basis of these considerations, we
would propose to include peer ratings as well as other methods (such as vignette-based
interviews) in future studies to limit the effect of underreporting.
Research on the behavior of adolescents is a developmental-based topic, strongly
influenced by age-related differences. In this study, we used a large age span of
participants, ranging from 12 to 19 years (with most respondents being in the range
of 14–18 years). Although we are aware of the problems involved, the aim of our
study was to reconstruct complete school networks. Limiting the age span to a smaller
range would prevent that kind of analysis. Furthermore, a larger age span allows
for the comparison of various age groups in one single study. Our findings already
indicated that victims of cyberbullying are rather young students located in classes
with lower average age, so we statistically controlled for the age of the participants
in subsequent analyses. However, a more detailed analysis of age-related differences
might be a logical next step for further research.
Clearly, there are further areas for improvement that should be addressed in
follow-up studies with larger sample sets, which would allow structural analyses
to be performed with a more robust number of cases. This would also offer more
possibilities in terms of integrating multiple levels of analysis in a unified model
and do multilevel modeling, rather than comparing the explanatory value of these
levels in three separate steps. In addition, more detailed comparisons of specific
cyberbullying incidents, case-based media use, particular aspects of social capital,
and finer differentiation between the various bullying roles/groups in off- and online
context could be included in follow-up studies.
Finally, temporal patterns of cyberbullying are a promising research route. The
structural patterns detected in the network analysis can be interpreted on the basis
of previous incidents and, probably, chain reactions that result in the diffusion of
cyberbullying. However, such an interpretation needs more support from longitu-
dinal empirical data. As it is, our research is a first step in exploring the structural
and environmental aspects of cyberbullying—not as a replacement for ‘‘traditional’’
ego/variable-oriented approaches, but as a useful extension of that approach.

Notes
1 Previous studies have reported variable prevalence rates, depending on country, point of
time, and definition of cyberbullying (e.g., Kowalski et al., 2008; Slonje & Smith, 2008).
2 One could argue that there is a random element in any forum, ‘‘trolling,’’ but this form
of behavior is usually not meant to hurt specific individuals, but to deliberately sabotage

122 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

online discussions with the sole aim of causing trouble. Although there might be victims
of trolling, they are usually not intentionally targeted for personal reasons beyond their
participation in the forum.
3 A popular depiction of cyberbullying is a locally initiated online defamation that gets
out of control ‘‘in the wild’’ of the Internet. Although such cases are potentially the most
harmful, they are not the usual case according to previous studies (as described in the
literature review): Most cyberbullying attacks still happen in a localized online context.
4 A follow-up study is under preparation, taking the ‘‘outer’’ layers into account.
5 Missing data were completely excluded from ego-based and structural analyses. In social
network analysis, it is strongly criticized that ignoring missing data can have negative
effects on the structural properties of the networks. However, Huisman (2010) found
that these limitations cannot be reduced by simple imputation processes. Therefore, we
decided to exclude missing values from our analyses following Huisman’s line of
argument.
6 The anonymization included (a) anonymization of the students’ names and
(b) anonymization of friend’s names independently from each other and the
questionnaire. This was achieved through a questionnaire that was separated into three
parts (i.e., name, friendship information, and other personal data). The data handling
and anonymization of the three parts were performed independently by different
persons. Optical character recognition was used to automatically code the survey
responses. After the scanning procedure, the questionnaires were destroyed. In a last
step, the anonymous data were reconnected through a separate key table, only available
to the researchers.
7 We decided against the use of longer scales due to the expected attention span of
younger students and other practical restrictions of a school survey.
8 One class was excluded from the statistical analysis because of an exceptionally low
response rate. We suspected some interference through group interaction or a
nonsupportive teacher.
9 Data were analyzed using SPSS for the ego-centered approach and UCINET for
structural aspects.
10 For this purpose, an ANOVA and a subsequent Scheffe post hoc test were performed to
distinguish the specific cyberbullying groups.
11 Visual inspection of network data can supersede statistical analysis, and computer-based
pattern analysis often fails to yield satisfactory results due to the complexity of the
calculations and the handling of the process (Klösgen & Zytkow, 2002; Ling, Gerth, &
Hanrahan, 2006). In other words, while there are advanced pattern-detection
algorithms, their actual use is not crucial, and in many cases, the selection of proper
solutions is more complicated and less useful than the visual inspection. In many ways,
pattern detection can be compared with factor analysis or structural equation modeling,
as there is no singular, correct solution. Mathematically ideal solutions might not be
useful at all; consequently, the selection of the solution with the highest explanatory
substance depends on meaningful decisions made by the researcher her/himself.

References
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human
Decision Processes, 50, 179–211.

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 123
Social Relations and Cyberbullying R. Festl & T. Quandt

Boneva, B., Quinn, A., Kraut, R., Kiesler, S., & Shklovski, I. (2006). Teenage communication
in the instant messaging era. In R. Kraut, M. Brynin, & S. Kiesler (Eds.), Computers,
phones, and the Internet: Domestication information technology (pp. 201–218). Oxford,
NY: Oxford University Press.
Bronfenbrenner, U. (1943). A constant frame of reference for sociometric research.
Sociometry, 6, 363–397.
Cairns, R., Cairns, B., Neckerman, H., Gest, S., & Gariepy, J.-L. (1988). Social networks and
aggressive behavior: Peer support or peer rejection? Developmental Psychology, 24,
815–823.
Caravita, S., Di Blasio, P., & Salmivalli, C. (2009). Unique and interactive effects of empathy
and social status on involvement in bullying. Social Development, 18, 140–163.
Cillessen, A. H. N., & Mayeux, L. (2004). From censure to reinforcement: Developmental
changes in the association between aggression and social status. Child Development, 75,
147–163.
Cotterell, J. (2007). Social networks in youth and adolescence. London and New York:
Routledge.
Crick, N. R. (1996). The role of overt aggression, relational aggression, and prosocial
behavior in the prediction of children’s future social adjustment. Child Development, 67,
2317–2327.
Dehm, U., & Storll, D. (2010). Medien und Tabus [Media and taboos]. Media Perspektiven, 9,
410–431.
Ellis, W. E., & Zarbatany, L. (2007). Peer group status as a moderator of group influence on
children’s deviant, aggressive, and prosocial behavior. Child Development, 78, 1240–1254.
Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook ‘‘Friends’’: Social
capital and college students’ use of online social network sites. Journal of
Computer-Mediated Communication, 12, 1143–1168.
Freeman, L. C., Roeder D., & Mulholland, R. R. (1980). Centrality in social networks II:
Experimental results. Social Networks, 2, 119–141.
Granovetter, M. (1983). The strength of weak ties: A network theory revisited. Sociological
Theory, 1, 201–233.
Hampton, K., & Wellman, B. (2003). Neighboring in netville: How the internet supports
community and social capital in a wired suburb. City and Community, 2, 277–311.
Heirman, W., & Walrave, M. (2008). Assessing concerns and issues about the mediation of
technology in cyberbullying. Cyberpsychology: Journal of Psychosocial Research on
Cyberpspace, 2, http://cyberpsychology.eu/view.php?cisloclanku=2008111401&article=1
Huisman, M. (2010). Imputation of missing network data: Some simple procedures. Journal
of Social Structure, 10, 1–29.
Huitsing, G., Veenstra, R., Sainio, M., & Salmivalli, C. (2010). ‘‘It must be me’’ or ‘‘It could
be them?’’ The impact of the social network position of bullies and victims on victims’
adjustment. Social Networks. doi:10.1016/j.socnet.2010.07.002.
Jäger, R., & Riebel, J. (2009). Mobbing bei Schülern in der Bundesrepublik Deutschland—Eine
empirische Untersuchung auf der Grundlage einer Online-Befragung im Jahre 2009.
[Bullying among students in the federal Republic of Germany—An empirical study based
on an online survey 2009]. Retrieved October 4, 2010, from www.zepf.uni-lindau.de.
Janis, I. L. (1982). Groupthink: Psychological studies of policy decisions and fiascoes (2nd ed.).
Boston: Houghton Mifflin.

124 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association
R. Festl & T. Quandt Social Relations and Cyberbullying

Kärnä, A., Salmivalli, C., Poskiparta, E., & Voeten, M. J. M. (2008). Do bystanders influence
the frequency of bullying in a classroom? Paper presented at the XIth EARA conference,
Turin, Italy.
Kavanaugh, A. L., Reese, D., Carool, J. M., & Rosson, M. B. (2005). Weaked ties in
networked communities. The Information Society, 21, 119–131.
Kelley, H. H. (1973). The process of causal attribution. American Psychologist, 28, 107–128.
Kirke, D. M. (2004). Chain reactions in adolescents’ cigarette, alcohol and drug use:
Similarity through peer influence or the patterning of ties in peer networks? Social
Networks, 26, 3–28.
Klösgen, W., & Zytkow, J. (2002). Handbook of data mining and knowledge discovery. New
York: Oxford University Press.
Kowalski, R. M., & Limber, S. P. (2007). Electronic bullying among middle school students.
Journal of Adolescent Health, 41, 22–30.
Kowalski, R. M., Limber, S. P., & Agatston, P. W. (2008). Cyberbullying. Malden, MA:
Blackwell.
Lenhart, A., Purcell, K., Smith, A., & Zickuhr, K. (2010). Social media and mobile internet use
among teens and young adults. Pew Internet and American Life Project. Retrieved October
5, 2011, from http://67.192.40.213/∼/media/Files/Reports/2010/PIP_Social_Media_
and_Young_Adults_Report_Final_with_toplines.pdf.
Li, Q. (2005). New bottle but old wine: A research of cyberbullying in schools. Computers in
Human Behavior, 23, 1777–1791.
Livingstone, S. (2009). Children and the internet: Great expectations, challenging realities.
Cambridge: Polity Press.
Livingstone, S., & Haddon, L. (2008). Risky experiences for children online: Charting
European research on children and the internet. Children & Society, 22, 314–323.
Livingstone, S., & Helsper, E. (2010). Balancing opportunities and risks in teenagers’ use of
the internet: The role of online skills and internet self-efficacy. New Media and Society, 12,
309–329.
Mouttapa, M., Valente, T., Gallaher, P., Rohrbach, L. A., & Unger, J. B. (2004). Social
network predictors of bullying and victimization. Adolescence, 39, 315–336.
Neal, J. W. (2009). Network ties and mean lies: A relational approach to relational
aggression. Journal of Community Psychology, 37, 737–753.
Neal, J. W. (2010). Social aggression and social position in middle childhood and early
adolescence: Burning bridges or building them? The Journal of Early Adolescence, 30,
122–137.
Postmes, T., Spears, R., & Lea, M. (1998). Breaching or building social boundaries?
SIDE-effects of computer-mediated communication. Communication Research, 25,
689–715.
Prensky, M. (2001). Digital natives, digital immigrants part I. On the Horizon, 9, 1–6.
Rammstedt, B., & John, O. P. (2007). Measuring personality in one minute or less: A 10-item
short version of the Big Five Inventory in English and German. Journal of Research in
Personality, 41, 203–212.
Raskauskas, J., & Stoltz, A. D. (2007). Involvement in traditional and electronic bullying
among adolescents. Developmental Psychology, 43, 564–575.
Rulison, K. L., Gest, S. D., Loken, E., & Welsh, J. A. (2010). Rejection, feeling bad und being
hurt: Using multilevel modeling to clarify the link between peer group aggression and
adjustment. Journal of Adolescence, 33, 787–800.

Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association 125
Social Relations and Cyberbullying R. Festl & T. Quandt

Salmivalli, C., Huttunen, A., & Lagerspetz, K. M. J. (1997). Peer networks and bullying in
schools. Scandinavian Journal of Psychology, 38, 305–312.
Salmivalli, C., Karhunen, J., & Lagerspetz, K. M. J. (1996). How do the victims respond to
bullying? Aggressive Behavior, 22, 99–109.
Schmidt, J.-H., Paus-Hasebrink, I., & Hasebrink, P. (Eds.). (2009). Heranwachsen mit dem
Social Web. Zur Rolle von Web 2.0-Angeboten im Alltag von Jugendlichen und jungen
Erwachsenen [Growing up with the social web. The importance of web 2.0-offers in the
everyday life of adolescents and young adults]. Berlin: Vistas Verlag.
Schuster, B. (1999). Outsiders at school: The prevalence of bullying and its relation with
social status. Group Processes and Intergroup Relations, 2, 175–190.
Sieving, R. E., Perry, C. L., & Williams, C. L. (2000). Change friendships? Examining paths of
influence in young adolescents’ alcohol use. Journal of Adolescence Health, 26, 27–35.
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,
57–67.
Shrum, W., Cheek, N. H., & Hunter, S. (1988). Friendship in school: Gender and racial
homophily. Sociology of Education, 61, 227–239.
Shumate, M., & Palazzolo, E. T. (2010). Exponential random graph (p!) models as a method
for social network analysis in communication research. Communication Methods and
Measures, 4, 341–371.
Slonje, R., & Smith, P. K. (2008). Cyberbullying—Another main type of bullying?
Scandinavian Journal of Psychology, 49, 147–154.
Smith, P. K., Mahdavi, J., Carvalho, M., Fisher, S., & Russell, S. (2008). Cyberbullying—Its
nature and impact in secondary school students. Journal of Child Psychology and
Psychiatry, 49, 376–385.
Suler, J. (2004). The online disinhibition effect. CyberPsychology & Behavior, 7, 321–326.
Tani, F., Greenman, P. S., Schneider, B. H., & Fregoso, M. (2003). Bullying and the big five. A
study of childhood personality and participant roles in bullying incidents. School
Psychology International, 24, 131–146.
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.
Wassermann, S., & Faust, K. (1994). Social network analysis. Cambridge: University Press.
Xiao, L., Gerth, J., & Hanrahan, P. (2006). Enhancing visual analysis of network traffic using
a knowledge representation. In P. C. Wong & D. Keim (Eds.), IEEE Symposium onVisual
Analytics Science and Technology: VAST’06 (pp. 107–114). Baltimore, MD: IEEE
Computer Society Press.
Xie, H., Swift, D. J., Cairns, B., & Cairns, R. B. (2002). Aggressive behaviors in social
interaction and developmental adaptation: A narrative analysis of interpersonal conflicts
during early adolescence. Social Development, 11, 205–224.
Ybarra, M. L., & Mitchell, K. J. (2004). Online aggressors, victims, and aggressor/victims: A
comparison of associated youth characteristics. Journal of Child Psychology and
Psychiatry, 45, 1308–1316.
Zimmer-Gembeck, M. J., Geiger, T. C., & Crick, N. R. (2005). Relational and physical
aggression, prosocial behaviors, and peer relations: Gender moderation and bidirectional
associations. Journal of Early Adolescence, 25, 421–452.

126 Human Communication Research 39 (2013) 101–126 © 2013 International Communication Association

You might also like